Document the OpenID Connect tokens that Gitea Actions jobs can request since https://github.com/go-gitea/gitea/pull/39052, so they can sign in to cloud providers without long-lived secrets.
- Add an OpenID Connect page covering how to trust Gitea, request a token and match its claims
- Document the `id-token` permission scope and its maximum
- Translate both for zh-cn and zh-tw, which still listed `id-token` as unsupported
Co-authored-by: silverwind <[email protected]>
Co-authored-by: silverwind <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/526
Reviewed-by: silverwind <[email protected]>
Co-authored-by: Max P. <[email protected]>
Follow-up to #588.
- rename `DocusaurusSidebarItem` -> `SidebarSourceItem`, `fromDocusaurus` -> `fromSidebarSource`, `docusaurusSidebars` -> `sidebarSources`
- reword comments and READMEs that described behaviour by comparison with docusaurus; fix the stale note in `sites/docs/src/config/api.ts`
- drop `.docusaurus/` from `.gitignore`
No functional change. `pnpm check` has 0 errors and `pnpm build` still builds 6690 pages.
<!-- cloudflare-preview --> Preview: https://pr-589.docs-gitea-com.pages.dev
Reviewed-on: https://gitea.com/gitea/docs/pulls/589
Reviewed-by: silverwind <[email protected]>
The v28 release (#564) added the `prepare-awesome-v%` Makefile rule, but the checks and publish workflows still only prepare awesome.md up to 1.27, so version-28/awesome.md is not generated in CI. Both workflows now call `prepare-awesome-v28`.
The preview step is skipped when Cloudflare secrets are unavailable, which is the case for pull requests from forks (#557 and #565 have no preview line, while same-repo PRs like #533 and #563 do). The skip is now a visible notice. Previewing fork PRs would need a separate trusted workflow and is not part of this change.
Reviewed-on: https://gitea.com/gitea/docs/pulls/579
Reviewed-by: bircni <[email protected]>
Co-authored-by: Lunny Xiao <[email protected]>
Bump the v28 docs to the 28.1.0 release, following the 1.27.3 bump (#532):
- point the `28` version variables (`version`, `sourceVersion`, `dockerVersion`, `displayVersion`) at 28.1.0
- update the v28 entry in `update_api_docs.sh` to `28.1.0`
- regenerate `static/swagger-28.json` and `static/openapi3-28.json` from the `v28.1.0` tag
---------
Co-authored-by: Lunny Xiao <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/578
Reviewed-by: TheFox0x7 <[email protected]>
Co-authored-by: bircni <[email protected]>
This PR contains the following updates:
| Package | Change | [Age](https://docs.renovatebot.com/merge-confidence/) | [Confidence](https://docs.renovatebot.com/merge-confidence/) |
|---|---|---|---|
| [pnpm](https://github.com/pnpm/pnpm/tree/main/pnpm) ([source](https://github.com/pnpm/pnpm/tree/HEAD/pnpm/npm/pnpm)) | [`12.6.0` → `12.8.2`](https://renovatebot.com/diffs/npm/pnpm/12.6.0/12.8.2) |  |  |
| [sharp](https://sharp.pixelplumbing.com) ([source](https://github.com/lovell/sharp), [changelog](https://github.com/lovell/sharp/blob/main/docs/src/content/docs/changelog.md)) | [`0.35.4` → `0.35.5`](https://renovatebot.com/diffs/npm/sharp/0.35.4/0.35.5) |  |  |
| [starlight-openapi](https://github.com/HiDeoo/starlight-openapi) ([source](https://github.com/HiDeoo/starlight-openapi/tree/HEAD/packages/starlight-openapi)) | [`0.26.2` → `0.26.3`](https://renovatebot.com/diffs/npm/starlight-openapi/0.26.2/0.26.3) |  |  |
---
### Release Notes
<details>
<summary>pnpm/pnpm (pnpm)</summary>
### [`v12.8.2`](https://github.com/pnpm/pnpm/releases/tag/v12.8.2): pnpm 12.8.2
[Compare Source](https://github.com/pnpm/pnpm/compare/v12.8.1...v12.8.2)
pnpm 12.8.2 fixes a startup crash on Linux ppc64le and `UnknownIssuer` errors on systems without CA certificates. `pnpm run` no longer installs before every script on CI when `autoDedupe` is enabled, and resolution and hoisted installs on macOS are faster.
##### Patch Changes
##### Platforms and environments
- Fixed pnpm crashing on startup on Linux ppc64le [#​16380](https://github.com/pnpm/pnpm/issues/16380).
- Fixed installs failing with `UnknownIssuer` on Linux systems without CA certificates, such as `node:24-slim`, when `NODE_EXTRA_CA_CERTS` is set. The extra certificates now extend the bundled CA roots [#​16365](https://github.com/pnpm/pnpm/issues/16365).
- pnpm now creates its store operation locks and other per-user lock files in `$XDG_RUNTIME_DIR` when it points to a directory only the user can write to. Otherwise, pnpm still uses `/tmp` on Linux and macOS. Sandboxes that block writes to `/tmp` can point `XDG_RUNTIME_DIR` at a writable directory [#​16390](https://github.com/pnpm/pnpm/issues/16390).
- POSIX bin shims and the `pnpm`, `pn`, `pnpx`, and `pnx` launchers now run inside a Nix build, where the system default path holds none of the utilities they call. Installing again replaces the shims already in `node_modules` [#​16377](https://github.com/pnpm/pnpm/issues/16377).
- In a project that pins another pnpm version, pnpm now passes a command with an option it does not know to the pinned version. Before, pnpm rejected the option before switching, so `pnpm install --auto-dedupe` failed with "Unknown option" even though the pinned pnpm supports it [#​16353](https://github.com/pnpm/pnpm/issues/16353).
##### Installing and resolving dependencies
- `pnpm install --frozen-lockfile` now fails when `Cargo.lock` does not satisfy a dependency requirement in `Cargo.toml`. The error names the crate and the version the lockfile holds [#​16355](https://github.com/pnpm/pnpm/issues/16355).
- `pnpm install` returns "Already up to date" again in a workspace with injected workspace dependencies and a shared lockfile. Since 12.7.0 every repeat install in such a workspace ran the full install and copied the injected projects again.
- With `injectWorkspacePackages: true`, a fresh `pnpm install` now records a workspace dependency as `link:` when its injected copy differs from the project only by an optional peer that peer-dependent dedupe merges. It was recorded as a peer-suffixed `file:` copy [#​16354](https://github.com/pnpm/pnpm/issues/16354).
- `pnpm dedupe --check` now passes right after `pnpm dedupe` when deduplication merges variants of a package that differ only in their peers. A lockfile key whose peer suffix named a merged variant now names the variant that replaced it [#​16356](https://github.com/pnpm/pnpm/issues/16356).
- When `minimumReleaseAge` hides the version that `latest` points to, pnpm now falls back to a prerelease of the same major before a stable version of an older major. A stable version of the same major is still preferred. Before, while a new `1.0.0` was too new, `latest` fell back to an old `0.0.1` even though `1.0.0-beta.4` had been `latest` until then [#​16388](https://github.com/pnpm/pnpm/issues/16388).
- Git-hosted dependencies now respect `pmOnFail`. If it is set to anything other than `download`, a git-hosted dependency that pins a pnpm version is prepared by the running pnpm, and pnpm does not download the pinned version [#​16376](https://github.com/pnpm/pnpm/issues/16376).
- pnpmfile hooks such as `readPackage` now run once for a dependency that several packages request at the same time. They could run twice for it before.
- `childConcurrency` now defaults to 5, the documented value. It used to be capped at 4 and to follow the host's CPU count.
##### Running scripts
- `pnpm run` and `pnpm exec` no longer install dependencies before every script on CI when `autoDedupe` is enabled. `pnpm install --frozen-lockfile` now keeps the deduplication record left by an earlier install [#​16374](https://github.com/pnpm/pnpm/issues/16374).
- On macOS and Linux, lifecycle scripts and `pnpm run` now always get `PATH` from the `PATH` variable. When the environment also held a `Path` variable, a script sometimes got `Path`'s value, and failed with `node: not found` [#​16308](https://github.com/pnpm/pnpm/issues/16308).
- `pnpm run` now exits after a `SIGTERM` in a container where pnpm is PID 1 and the script runs pnpm again, as `"start": "pnpm serve"` does. Since 12.6.0 it kept waiting after the script had shut down, until the container runtime killed it.
##### Other commands and settings
- `pnpm config get globalShims`, `pnpm shim list`, and global installs no longer read `globalShims` from a project's `pnpm-workspace.yaml`. Only the global config file, the pnpm home's own `pnpm-workspace.yaml`, and `PNPM_CONFIG_GLOBAL_SHIMS` set it, so a repository cannot choose which globally installed packages get project-aware shims.
- `pnpm config set --location=project` refuses a machine-level setting such as `stateDir` or `scope` with `ERR_PNPM_CONFIG_SET_NOT_A_PROJECT_SETTING`, which names where the setting belongs. `pnpm config delete` still clears such a key from a project's `pnpm-workspace.yaml`.
- `pnpm deploy` no longer fails with `ERR_PNPM_DEPLOY_AMBIGUOUS_PEER` in a workspace with `injectWorkspacePackages: true` when a workspace package lists its peer dependency as a dev dependency too [#​16375](https://github.com/pnpm/pnpm/issues/16375).
- `pnpm deploy` no longer copies the workspace root's `packageManager` and `devEngines.packageManager` fields into the deployed `package.json` [#​16403](https://github.com/pnpm/pnpm/issues/16403).
- `pnpm publish` now includes bare `README` files and README files with Markdown extensions such as `readme.markdown` in registry metadata [#​12704](https://github.com/pnpm/pnpm/issues/12704).
- `pnpm store prune` now removes the packages that only expired `pnpm dlx` cache entries used. They were left in the store until the next `pnpm store prune` [#​16383](https://github.com/pnpm/pnpm/discussions/16383).
##### Performance
- Sped up dependency resolution in large workspaces, and when many dependencies request different ranges of the same package. Resolution also uses less memory.
- Sped up `pnpm install` with `nodeLinker: hoisted` on macOS when the lockfile is re-resolved, such as with `autoDedupe` enabled [#​16397](https://github.com/pnpm/pnpm/issues/16397).
- Sped up extracting package tarballs.
- `pnpm install` without `--frozen-lockfile` is faster on some machines in projects with a `pnpm-workspace.yaml`. Those installs linked with one worker thread per core, half of what a frozen install uses.
- On Windows, warm `pnpm install --frozen-lockfile` runs are 4-5% faster on 4- and 8-core machines. pnpm now links with one worker thread per core on Windows, between 4 and 16. This changes frozen installs and installs in projects without a `pnpm-workspace.yaml` on machines with 3 to 15 cores.
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://notion.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/notion.svg" width="80" alt="Notion"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://latitude.so/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/latitude.svg" width="160" alt="Latitude"></a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
### [`v12.8.1`](https://github.com/pnpm/pnpm/releases/tag/v12.8.1): pnpm 12.8.1
[Compare Source](https://github.com/pnpm/pnpm/compare/[email protected])
pnpm 12.8.1 fixes `pnpm install --frozen-lockfile` rejecting lockfiles with injected workspace packages that have peers, restores the executable bit on files of local directory dependencies, makes `pnpm dedupe` converge, and uses less CPU on many-core machines.
##### Patch Changes
- `pnpm install --frozen-lockfile` no longer rejects a freshly generated lockfile when an injected workspace package has peer dependencies [#​16332](https://github.com/pnpm/pnpm/issues/16332).
- Executable files in a `file:` directory dependency or an injected workspace package keep their executable bit again. Since 12.8.0, pnpm installed these files without the permissions they have in their project.
- `pnpm dedupe` now reaches a stable lockfile when a package's peer suffix is long enough to be hashed. Before, each run could switch that package's key between the hashed and the spelled-out suffix, so `pnpm dedupe --check` always failed [#​16331](https://github.com/pnpm/pnpm/issues/16331).
- `pnpm install --frozen-lockfile`, the default in CI, now uses less CPU on machines with more than 8 cores. Warm installs on many-core Windows machines got up to 10% faster. Frozen installs now link with at most 16 worker threads.
- `verifyDepsBeforeRun` no longer reports dependencies as outdated after a filtered install just because `pnpm-lock.yaml` has a newer modification time. It checks the lockfile against the packages that install put in place. Before, `pnpm run` reinstalled the whole workspace with lifecycle scripts on, for example after a Docker `COPY` brought in a lockfile with a newer mtime [#​16322](https://github.com/pnpm/pnpm/issues/16322).
After a filtered install, `verifyDepsBeforeRun` now also checks that the install put the selected projects' dependencies in place. A `node_modules` directory alone no longer counts as proof.
- `pnpm run` and `pnpm exec` no longer install a project that has never been installed and has nothing to install. Such a project declares no dependencies, no peer dependencies that `autoInstallPeers` would fetch, and no install lifecycle scripts. The command now runs without writing `node_modules` or `pnpm-lock.yaml` [#​16313](https://github.com/pnpm/pnpm/issues/16313).
- `pnpm update -g --latest` now upgrades globally installed packages beyond their saved version ranges [#​16320](https://github.com/pnpm/pnpm/issues/16320).
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://notion.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/notion.svg" width="80" alt="Notion"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://latitude.so/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/latitude.svg" width="160" alt="Latitude"></a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
### [`v12.8.0`](https://github.com/pnpm/pnpm/releases/tag/v12.8.0): pnpm 12.8
[Compare Source](https://github.com/pnpm/pnpm/compare/[email protected])
pnpm 12.8.0 warns when `pnpm pack` or `pnpm publish` would ship a `.env` file that `files` does not list, installs `sharedWorkspaceLockfile: false` workspaces concurrently, applies every setting passed as `--config.<name>=<value>`, and no longer leaves the Windows terminal stuck after Ctrl+C in a script.
##### Minor Changes
- `pnpm pack` and `pnpm publish` now warn when the tarball includes a `.env` or `.env.*` file that the `files` field of `package.json` does not list. Templates such as `.env.example` are not reported. List the file in `files` to publish it on purpose, or exclude it in `.npmignore` or `.gitignore` [#​7826](https://github.com/pnpm/pnpm/issues/7826).
- `pnpm pack` now honors `--silent`, `--reporter=silent`, and `--loglevel=silent` to hide the tarball contents and summary. With `--json`, lifecycle script output and the final JSON output remain visible [#​10297](https://github.com/pnpm/pnpm/issues/10297).
##### Patch Changes
##### Installing packages
- Installing through a `pnpr` server now records the pnpmfile checksum in the lockfile, so a later `pnpm install --frozen-lockfile` accepts that lockfile [#​14460](https://github.com/pnpm/pnpm/issues/14460). A frozen install through the pnpr server now fails if the pnpmfile changed. If the pnpmfile defines a `readPackage`, `afterAllResolved` or `preResolution` hook or custom resolvers, pnpm resolves dependencies locally and prints a warning that the pnpr server was not used.
Installing through a `pnpr` server also links a workspace project at the directory its `publishConfig.directory` names. A server that does not forward the setting makes the install fail with `ERR_PNPM_PNPR_PUBLISH_DIRECTORY_MISMATCH`, so pnpm never writes a lockfile that points at the wrong directory. The server rejects a `publishConfig.directory` that points outside its project.
- Installing a git-hosted dependency that has to be built no longer fails when that dependency's own dependencies have build scripts nobody approved. pnpm skips those builds while preparing the dependency, as it does without `strictDepBuilds` [#​9764](https://github.com/pnpm/pnpm/issues/9764).
- A git-hosted dependency that is a pnpm workspace with no committed lockfile is now detected as a pnpm project [#​14011](https://github.com/pnpm/pnpm/issues/14011).
- `pnpm install --dev` and `pnpm fetch --dev` now install the optional dependencies of devDependencies, such as the platform binaries of Biome and oxlint. The project's own `optionalDependencies` are still skipped [#​9678](https://github.com/pnpm/pnpm/issues/9678).
- `pnpm install --offline` and `pnpm add --offline` now resolve a version range to the newest matching version whose tarball is already in the store. They used to pick the newest version in the cached metadata and fail with `ERR_PNPM_NO_OFFLINE_TARBALL` when its tarball was missing [#​10715](https://github.com/pnpm/pnpm/issues/10715).
- If an offline install fails because the registry metadata cache uses the layout from before pnpm 11.27 and 12.4, the error now names the older mirror on disk and explains that one online install repopulates the cache. The error also carries the `ERR_PNPM_NO_OFFLINE_META` code. `pnpm cache prune --help` now says that pnpm 11.26 and earlier, and pnpm 12.3 and earlier, depend on the directories it removes [#​15656](https://github.com/pnpm/pnpm/issues/15656).
- Running `pnpm install` now refreshes dependencies when a package declared with a local `file:` directory changes its dependencies [#​4623](https://github.com/pnpm/pnpm/issues/4623).
- A repeat `pnpm install` now keeps its fast up-to-date check when an override replaces a declared local `file:` dependency [#​12892](https://github.com/pnpm/pnpm/issues/12892).
- `pnpm install` now removes an optional dependency from `node_modules` if its install script fails. Code that checks whether the package is installed no longer finds a package that cannot load [#​8756](https://github.com/pnpm/pnpm/issues/8756).
- With `nodeLinker: hoisted`, `pnpm install` now restores a workspace project's `node_modules` after it was deleted. Before, the install printed "Already up to date" and left the project without the dependencies nested under it. On Windows, the install also no longer fails with "Access is denied" when another project's copy of a shared dependency links to the deleted directory.
- Under `nodeLinker: hoisted`, `pnpm install` now clears orphaned package directories that an interrupted or failed install leaves in a project's `node_modules`. A directory recorded by the previous install is removed, while an unrecorded directory is moved to `node_modules/.ignored`. A copy already in `.ignored` is never overwritten [#​13676](https://github.com/pnpm/pnpm/issues/13676).
- Concurrent installs no longer fail when they replace the same stale hoisted dependency link. Virtual store cleanup now keeps the temporary lockfiles that concurrent installs write.
##### Resolving and linking dependencies
- `pnpm install` no longer aborts on a failed allocation of many gigabytes when peer dependency ranges combine overlapping `||` alternatives [#​15867](https://github.com/pnpm/pnpm/issues/15867).
- `pnpm install` no longer fails when a package from the registry declares a `file:` dependency on a directory inside itself, such as `"@​types/css-tree": "file:./typings/css-tree"`. pnpm links that dependency to the directory inside the package, as npm and Yarn do. The lockfile records it as `link:<root>/typings/css-tree` [#​9141](https://github.com/pnpm/pnpm/issues/9141).
- An `npm:` alias written by `overrides` now stays in place when a change elsewhere makes pnpm re-resolve the aliased dependency. Before, pnpm could look up the alias name at the aliased version, which failed with `ERR_PNPM_NO_MATCHING_VERSION` or locked an unrelated package [#​16309](https://github.com/pnpm/pnpm/issues/16309).
- A peer dependency no longer resolves to two different versions for one package. This happened when the package peer-depends on another package and on one of that package's peers, and it is installed deeper than a direct dependency of the package that provides them [#​12098](https://github.com/pnpm/pnpm/issues/12098).
- An optional peer dependency is no longer resolved from another workspace project's package when the project provides one of that package's own peers at a version it rejects. This avoids bogus unmet peer errors [#​13989](https://github.com/pnpm/pnpm/issues/13989).
- `pnpm dedupe` no longer changes the lockfile on every run when a nested peer dependency is provided through an npm alias [#​15709](https://github.com/pnpm/pnpm/issues/15709).
- With `resolutionMode: time-based` and `minimumReleaseAge` both set, `pnpm install` no longer reports a subdependency as too new when only the time-based cutoff excludes it. Such subdependencies used to fail a strict install with `ERR_PNPM_NO_MATURE_MATCHING_VERSION`, or were added to `minimumReleaseAgeExclude` [#​13569](https://github.com/pnpm/pnpm/issues/13569). A transitive dependency that has no matching version published before the time-based cutoff now resolves to the lowest matching version allowed by `minimumReleaseAge`. pnpm picks a version younger than `minimumReleaseAge` only if no older version matches [#​16298](https://github.com/pnpm/pnpm/issues/16298).
- `pnpm install` retries registry metadata fetches that fail with a timeout, a dropped connection, or an interrupted response body before it applies `trustPolicy` or `minimumReleaseAge`. A transient fetch failure is not reported as `TRUST_DOWNGRADE` or `MINIMUM_RELEASE_AGE_VIOLATION` [#​12031](https://github.com/pnpm/pnpm/issues/12031).
- pnpm's built-in package compatibility database no longer applies to a project's own manifest. A project named like a published package, such as `vue-loader`, no longer gains dependencies on `pnpm install` or `pnpm update`. User-configured `packageExtensions` still apply to project manifests [#​11700](https://github.com/pnpm/pnpm/issues/11700).
- Packages in an external `virtualStoreDir` can resolve the project's direct dependencies selected by `hoistPattern`. Run `pnpm install --force` to repair an existing installation [#​5652](https://github.com/pnpm/pnpm/issues/5652).
- `pnpm install` now links the executables of auto-installed peer dependencies into the workspace root's `node_modules/.bin`, including after a frozen-lockfile reinstall [#​8511](https://github.com/pnpm/pnpm/issues/8511).
##### Lockfiles and frozen installs
- `pnpm install --frozen-lockfile` now works on a detached HEAD when `gitBranchLockfile` is enabled. The install reads the lockfiles of the local and remote-tracking branches that contain the checked-out commit. It still writes the shared `pnpm-lock.yaml` [#​7672](https://github.com/pnpm/pnpm/issues/7672).
- `pnpm install --frozen-lockfile` now accepts a lockfile that has no importer entry for a workspace package without dependencies. Such a package added after the lockfile was written made the install fail with `ERR_PNPM_PACKAGE_MANAGER_NO_IMPORTER` [#​15875](https://github.com/pnpm/pnpm/issues/15875).
- `pnpm install` now fails with `ERR_PNPM_LOCKFILE_MISSING_DEPENDENCY` when an importer references a dependency version that has no snapshot entry. Before, the install succeeded and left a `node_modules` symlink pointing at a missing virtual-store directory [#​14764](https://github.com/pnpm/pnpm/issues/14764).
- `pnpm install` on CI now fails on an outdated lockfile when `preferFrozenLockfile` is explicitly set to `true`. Setting it to `true` used to let CI update the lockfile [#​9072](https://github.com/pnpm/pnpm/pull/9072).
- With `gitBranchLockfile` enabled, each emoji or other character outside the Basic Multilingual Plane in a branch name now becomes `!!` in the lockfile name. Before, each such character became one `!`.
##### Workspaces and filtering
- `pnpm install` in a workspace with `sharedWorkspaceLockfile: false` now installs projects concurrently, up to `workspaceConcurrency` at a time [#​14480](https://github.com/pnpm/pnpm/issues/14480). A project is resolved, fetched, and written to its virtual store without waiting for the workspace projects it depends on. It waits for them only before it links its dependencies and runs its lifecycle scripts, so its scripts still run after theirs. A project with a `preinstall` or `pnpm:devPreinstall` script, or with an injected or `file:` workspace dependency, waits for its workspace dependencies before it starts.
The installs of the projects also share their package metadata, lockfile verification, and store caches, so they use less CPU and memory when several projects depend on the same packages. An install with a pnpmfile no longer starts an extra Node.js process when the pnpmfile has no `preResolution` hook.
- With `enableGlobalVirtualStore` and `sharedWorkspaceLockfile: false`, each project now keeps its current lockfile and its hidden hoisted dependencies in its own `node_modules/.pnpm`. Before, every project wrote them to the workspace root's `node_modules/.pnpm`, so each repeat install treated the other projects' packages as its own and relinked them [#​14480](https://github.com/pnpm/pnpm/issues/14480).
- `pnpm rebuild`, `pnpm approve-builds`, and `pnpm ignored-builds` now work on the current project's `node_modules` when they run inside a project of a workspace with `sharedWorkspaceLockfile: false`. They used to read the workspace root's `node_modules`, so `pnpm rebuild` did not rebuild the project's dependencies and created a second virtual store at the workspace root [#​9402](https://github.com/pnpm/pnpm/issues/9402).
- `pnpm install` no longer creates a `node_modules` symlink inside the `publishConfig.directory` of a workspace package linked with `linkDirectory`. A build tool that cleaned its output directory through that symlink deleted the files of the package's dependencies. `pnpm install` also removes a symlink that an earlier install left there [#​16226](https://github.com/pnpm/pnpm/issues/16226). It also no longer fails with `ERR_PNPM_CMD_SHIM_RESOLVE_PATH` when such a package has a `bin` field and its `publishConfig.directory` does not exist yet.
- `pnpm install` no longer fails for an injected workspace dependency whose package publishes from a `publishConfig.directory` that its own `prepare` script builds. The injected copy now picks up that directory once `prepare` finishes building it. `pnpm install --frozen-lockfile` no longer reports the dependency as outdated while the directory has not been built yet [#​7811](https://github.com/pnpm/pnpm/issues/7811).
- An in-place edit to the source of an injected workspace package now shows up in its injected copy, unless a build writes to that package or `packageImportMethod` is set. pnpm hardlinks such packages under the default import method [#​4410](https://github.com/pnpm/pnpm/issues/4410). Scripts listed in `syncInjectedDepsAfterScripts` now update injected dependencies while they run, so a watcher on the injected package, such as a dev server, sees each change before the script exits.
- With `sharedWorkspaceLockfile: false`, an injected workspace package that has lifecycle scripts is now hard linked into the projects that depend on it. Before, pnpm left a plain copy, so later edits to the package did not reach those projects [#​9828](https://github.com/pnpm/pnpm/issues/9828).
- `injectWorkspacePackages` now hard links a workspace dependency declared with a relative path, such as `workspace:../foo`, the same way it already does for `workspace:*` [#​10446](https://github.com/pnpm/pnpm/issues/10446).
- Workspace discovery prunes dot-prefixed directories, so a `packages` pattern such as `**` no longer matches projects inside `.cache` and other hidden directories [#​16250](https://github.com/pnpm/pnpm/issues/16250).
- `pnpm import` in a workspace now keeps the versions pinned by a `yarn.lock` inside a workspace project [#​4385](https://github.com/pnpm/pnpm/issues/4385).
##### Store and caches
- Files imported from the store now follow the umask of the install that writes them. Installing with a umask of `077` no longer leaves imported files readable by the group and others [#​3807](https://github.com/pnpm/pnpm/issues/3807).
- `pnpm install` keeps the owner, group, and mode of files already in a shared store, including `index.db`. New store files and directories inherit the store directory's group-write bit. When that directory is setgid, new files inherit its group. pnpm does not change a file's owner or group [#​12765](https://github.com/pnpm/pnpm/issues/12765).
- When `pnpm install` repairs a store file that was modified through a hard link in `node_modules`, the repair now keeps the file's inode on Linux and macOS, so hard-linked copies in other projects are healed at the same time. On Windows the repair still replaces the file, so other projects are healed on their next install [#​3445](https://github.com/pnpm/pnpm/issues/3445).
- `pnpm install` now reports a full store at once when writing package files fails. It no longer retries the tarball [#​8581](https://github.com/pnpm/pnpm/issues/8581).
- pnpm now warns when it cannot hard link packages from an existing store in the pnpm home directory and falls back to a store on the project's filesystem. This can happen when the project is on another filesystem, such as a bind-mounted workspace in a container. The warning names both stores and suggests setting `storeDir` [#​14505](https://github.com/pnpm/pnpm/issues/14505).
- The side-effects cache now restores the symlinks that a build script creates inside a package. A warm install used to replace each of them with a copy of its target [#​12859](https://github.com/pnpm/pnpm/issues/12859).
After upgrading, every package with a build script is built once more.
- The global virtual store and the side-effects cache now key built packages by the Node.js version that the root project's `devEngines.runtime` or `engines.runtime` pins. That is the Node.js their build scripts run with. A dependency that declares its own `engines.runtime` no longer changes the key for every other package.
- With `enableGlobalVirtualStore`, an install into a fresh `node_modules` no longer runs the build scripts of a dependency whose global virtual store slot an earlier install already built. `pnpm rebuild` still runs them [#​14480](https://github.com/pnpm/pnpm/issues/14480).
- Concurrent installs that share a global virtual store now run a package's build in its shared slot one at a time. A failed build leaves the slot in place and marks it for the next install to rebuild [#​15568](https://github.com/pnpm/pnpm/issues/15568).
- A warm `pnpm install` reuses on-disk package metadata for five minutes when the registry does not send an ETag. Registries that send an ETag, including the public npm registry, still revalidate with a conditional request. `pnpm update` still fetches current metadata [#​13976](https://github.com/pnpm/pnpm/issues/13976).
- pnpm no longer revalidates cached registry metadata when the registry sends `Cache-Control: max-age=0`, `no-cache`, or `no-store`. It downloads the metadata again, so a version newly published to such a registry is visible on the next install [#​13487](https://github.com/pnpm/pnpm/issues/13487).
- `pnpm install` honors `Cache-Control` for dependencies named with an `http:` or `https:` tarball URL. A fresh response is taken from the store with no request, and a stale one is revalidated with `If-None-Match` [#​15648](https://github.com/pnpm/pnpm/issues/15648).
##### Patched dependencies
- `pnpm install` now repairs a `pnpm-lock.yaml` whose `(patch_hash=<hash>)` dependency paths disagree with its `patchedDependencies` map, including paths that lack the hash their patch calls for. Before, pnpm accepted such a lockfile as up to date and kept the old patched files. `pnpm install --frozen-lockfile` now fails on such a lockfile with `ERR_PNPM_INCONSISTENT_PATCH_HASH`. It fails with `ERR_PNPM_UNCHECKABLE_PATCH_HASH` when a patch hash in the lockfile is malformed, or when the lockfile lacks the package version or patch entry that the check needs [#​15336](https://github.com/pnpm/pnpm/pull/15336).
- `pnpm install` with `nodeLinker: hoisted` now applies a patch once to each copy of a patched dependency in a workspace. Before, a copy that several workspace projects shared could receive the patch twice and end up with the patched content duplicated [#​7565](https://github.com/pnpm/pnpm/issues/7565).
- `pnpm install` and `pnpm fetch` now fail with `ERR_PNPM_PATCH_NOT_FOUND` when a patch file listed in `patchedDependencies` does not exist [#​5268](https://github.com/pnpm/pnpm/issues/5268).
- `engineStrict` now checks the patched `package.json` when a `patchedDependencies` entry changes `engines`. A patch that relaxes `engines.node` no longer fails the install against the published range [#​9603](https://github.com/pnpm/pnpm/issues/9603).
- `pnpm patch` now applies the existing patch file to the edit directory of a git-hosted dependency, as it already does for packages from the registry [#​9699](https://github.com/pnpm/pnpm/issues/9699).
##### Adding, updating, and removing dependencies
- `pnpm add <dir>` now warns when the added directory declares peer dependencies, as `pnpm link` does. The directory is saved as a `link:` dependency, and its peers are not resolved from the project that adds it. Use the `file:` protocol to have them resolved [#​5523](https://github.com/pnpm/pnpm/issues/5523).
- `pnpm add --save-types` no longer adds a `@types/*` package whose resolved version is deprecated. DefinitelyTyped publishes such stubs for packages that ship their own types, such as `@types/typescript` for `typescript` [#​15636](https://github.com/pnpm/pnpm/issues/15636).
- `pnpm version`, `pnpm add`, and `pnpm pkg set` keep JSON5 style when they update `package.json5`. ASCII identifier keys stay unquoted, strings keep JSON5 quotes, and indented files keep trailing commas [#​15717](https://github.com/pnpm/pnpm/issues/15717).
##### Running scripts and commands
- `pnpm run` and `pnpm exec` no longer install dependencies automatically when the root `package.json` still keeps `overrides`, `packageExtensions`, `patchedDependencies`, or `ignoredOptionalDependencies` in its `pnpm` field. pnpm no longer reads that field, so the install rewrote the lockfile without those settings. The command now fails and asks to move the settings to `pnpm-workspace.yaml` [#​16278](https://github.com/pnpm/pnpm/issues/16278).
- When `verifyDepsBeforeRun` triggers an install before a filtered `pnpm run` or `pnpm exec`, pnpm now installs only the selected projects and their dependencies. A later filtered command also installs a selected project that an earlier filtered install skipped [#​11865](https://github.com/pnpm/pnpm/issues/11865).
- `pnpm -r run /regexp/` now honors the `tasks` `dependsOn` declared for each script the selector matches, like running the script by name does. Matched scripts that depend on each other run in order. Each matched script runs once [#​15596](https://github.com/pnpm/pnpm/issues/15596).
- `pnpm run` exits with the code of a script that handles Ctrl+C and shuts down. A script that finished cleanly is not reported as a lifecycle failure. The commands after it in the same script still run [#​9945](https://github.com/pnpm/pnpm/issues/9945).
- pnpm no longer hangs after a lifecycle script exits while a process it started in the background keeps the script's output open. pnpm stops reading that output one second after the script exits [#​5730](https://github.com/pnpm/pnpm/issues/5730).
- `pnpm run` and lifecycle scripts use the configured `scriptShell`, including Git Bash on Windows, when `shellEmulator` is also enabled. `shellEmulator` still runs scripts when `scriptShell` is not set. Extra arguments passed to `pnpm run` are quoted for the shell that runs the script, so a Windows path stays intact [#​14719](https://github.com/pnpm/pnpm/issues/14719).
- With `enableGlobalVirtualStore`, dependency build scripts now see the workspace root's `node_modules/.bin`, as they do with a local virtual store. A `postinstall` script that runs `node` finds the Node.js installed by `devEngines.runtime` and no longer fails with "command not found" on machines without a system Node.js [#​15652](https://github.com/pnpm/pnpm/issues/15652). Dependency build scripts also see the bins of privately hoisted dependencies.
- Dependency install scripts now find the node-gyp bundled with pnpm when pnpm runs through a symlink, such as `node_modules/.bin/pnpm` or the `pnpm` that `npm install -g pnpm` links. They used to fail with `node-gyp: command not found` on macOS [#​15694](https://github.com/pnpm/pnpm/issues/15694).
- `pnpm run` and lifecycle scripts now set `npm_config_node_gyp` to the bundled `node-gyp` entry point. Tools that read the variable resolve the same `node-gyp` pnpm builds with. An `npm_config_node_gyp` value the environment already sets is kept as is [#​16270](https://github.com/pnpm/pnpm/issues/16270).
- Scripts now see the `npm_command` environment variable that npm sets. It holds `run-script` when the command runs a script, and the command's own name otherwise [#​16265](https://github.com/pnpm/pnpm/issues/16265).
- Commands run from a POSIX shell through a dependency's own `node_modules/.bin`, such as `node_modules/vite/node_modules/.bin/esbuild`, no longer fail with `MODULE_NOT_FOUND` [#​10189](https://github.com/pnpm/pnpm/issues/10189).
- `pnpx --version` and `pnpm dlx --version` now print the pnpm version. Other unknown options before the command are reported as errors. Before, pnpm tried to download a package named after the option [#​16259](https://github.com/pnpm/pnpm/issues/16259).
- `pnpm dlx` now keeps the virtual store of its cached installs in `node_modules/.pnpm`, like every other install [#​13955](https://github.com/pnpm/pnpm/issues/13955). `pnpm pack-app` now names the manifest of its runtime install directory `pnpm-pack-app-<target>`.
##### Publishing, packing, and deploying
- `pnpm pack` and `pnpm publish` now ship a file that the `files` field names even when another entry excludes the directory holding it. For example, `["**", "!dist", "dist/index.d.ts"]` ships `dist/index.d.ts` [#​16213](https://github.com/pnpm/pnpm/issues/16213).
- `pnpm pack` prunes a directory that a `files` field exclusion names, such as `!**/test`, excluding the directory and its contents from the packed package [#​15738](https://github.com/pnpm/pnpm/issues/15738).
- `pnpm publish` now waits at least 5 minutes for the registry to answer a publish request, like npm. This fixes "409 Conflict - Failed to save packument" errors when the registry is slow to answer [#​11454](https://github.com/pnpm/pnpm/issues/11454).
- `pnpm deploy --prod` no longer fails with `ERR_PNPM_OUTDATED_LOCKFILE` when the deployed project declares a `devEngines.runtime` with `onFail: download`. The runtime stays out of the deployed `node_modules` with the rest of the dev dependencies [#​15703](https://github.com/pnpm/pnpm/issues/15703).
- `pnpm deploy` with a shared lockfile now copies workspace dependencies into the deploy directory, even when `packageImportMethod` is set to `hardlink`. Before, their files were hard-linked to the workspace sources, so editing a source file also changed the deployed copy [#​12176](https://github.com/pnpm/pnpm/issues/12176).
- `pnpm deploy --legacy` no longer leaves broken links to nested local dependencies of workspace packages [#​9575](https://github.com/pnpm/pnpm/issues/9575).
##### Configuration and pnpmfile hooks
- Every setting pnpm supports can now be set with `--config.<name>=<value>` on the command line, not only the ones whose command also carries a matching flag. Before, `pnpm install --config.frozen-lockfile=true` dropped the setting and rewrote `pnpm-lock.yaml` as though the install had not been frozen [#​16276](https://github.com/pnpm/pnpm/issues/16276).
- Settings given on the command line, such as `--registry` and `--store-dir`, now take precedence over the values a pnpmfile `updateConfig` hook sets [#​14063](https://github.com/pnpm/pnpm/issues/14063).
- `pnpm config set --location=project` and `pnpm config delete --location=project`, run from a package inside a workspace, now write settings that belong in `pnpm-workspace.yaml` to the workspace root's `pnpm-workspace.yaml`. Before, they created a new `pnpm-workspace.yaml` in the current package, which made that package the workspace root. Settings stored in `.npmrc` are still written to the current directory [#​13757](https://github.com/pnpm/pnpm/issues/13757).
- pnpm now reads the workspace directory override from `PNPM_CONFIG_WORKSPACE_DIR`, like other settings. `NPM_CONFIG_WORKSPACE_DIR` still works as a fallback [#​16275](https://github.com/pnpm/pnpm/issues/16275).
- pnpm now fails with `ERR_PNPM_AUTH_INVALID_BASE64` when a registry's `_password` in `.npmrc` is not valid base64. Before, it sent the value as the raw password. A `username` or `_password` left empty, for example by an unset environment variable, now supplies no credential [#​16273](https://github.com/pnpm/pnpm/issues/16273).
- `proxy=false` now turns proxying off even when `HTTP_PROXY`, `HTTPS_PROXY`, or `ALL_PROXY` is set. pnpm no longer sends requests through a proxy named only in `ALL_PROXY`.
- `pnpm install` now runs the install hooks of a config dependency plugin's pnpmfile, including `readPackage`, `afterAllResolved`, and custom resolvers. Its pnpmfile is also counted in `pnpmfileChecksum`. Before, only the plugin's `updateConfig` hook ran, so a plugin could not change the resolved dependencies.
- A pnpmfile `fetchers` hook now runs once per package on a fresh install when it handles a resolution with a custom `type` or delegates a git-hosted one to the same subdirectory [#​15584](https://github.com/pnpm/pnpm/issues/15584). These packages were fetched a second time for installation, so the installed files could come from a different archive than the one their dependencies were read from. The hook also no longer runs twice when a `resolvers` hook returns a tarball resolution without a manifest [#​15025](https://github.com/pnpm/pnpm/issues/15025).
- `pnpm install` now re-fetches a package from a custom resolver when the `integrity` of its resolution changes, with or without `enableGlobalVirtualStore`. It used to update the lockfile but keep the old files in `node_modules` [#​15670](https://github.com/pnpm/pnpm/issues/15670).
- `pnpm install` now rejects invalid results from a `readPackage` hook. A hook that returns a non-object value fails with `ERR_PNPM_BAD_READ_PACKAGE_HOOK_RESULT` [#​15730](https://github.com/pnpm/pnpm/issues/15730). A hook that sets a dependency range to a value other than a string, such as `undefined`, fails with an error that names the dependency, the package and the pnpmfile. Delete the property to remove a dependency [#​15705](https://github.com/pnpm/pnpm/issues/15705).
##### Global packages, pnpm versions, and runtimes
- `pnpm update --global` now reinstalls the global packages that pnpm 10 installed into the previous global directory, `<global-dir>/5`, so their commands are linked into the pnpm home `bin` directory again and `pnpm list --global` lists them. Once every package is migrated, pnpm deletes the previous directory and the commands pnpm 10 linked into the pnpm home [#​11528](https://github.com/pnpm/pnpm/issues/11528).
- A signal sent to pnpm, such as `SIGTERM`, now reaches the pnpm that pnpm switches to because of `packageManager` or `devEngines.packageManager`, and the one that `pnpm with` runs. The signal used to be dropped, so scripts running under that pnpm never got to shut down [#​9948](https://github.com/pnpm/pnpm/issues/9948).
- On arm64 musl Linux, such as Alpine on ARM, switching to a pinned pnpm older than 12 now runs the JavaScript `pnpm` package. The standalone executable of those versions crashed at startup on that platform [#​10443](https://github.com/pnpm/pnpm/issues/10443).
- Global shims such as `node` now work when pnpm runs through a relative symlink, as with a Homebrew install. They were copies of that symlink and did not resolve from the global bin directory [#​15691](https://github.com/pnpm/pnpm/issues/15691).
- `pnpm env remove --global` deletes Node.js versions that pnpm installed into its own store, including when another tool installed pnpm [#​8357](https://github.com/pnpm/pnpm/issues/8357).
- `pnpm self-update` no longer suggests a downgrade when `minimumReleaseAge` holds back the registry's `latest` release. It now says that release is still within the cutoff [#​12006](https://github.com/pnpm/pnpm/issues/12006).
##### Windows
- Interrupting a script with Ctrl+C on Windows no longer leaves the terminal stuck [#​14860](https://github.com/pnpm/pnpm/issues/14860). A script that runs through a batch shim, as `vite dev` does through `vite.CMD`, made cmd.exe wait forever on its "Terminate batch job (Y/N)?" answer, and every following keystroke went to that prompt. pnpm now ends a cmd.exe script shell once it has sat for a second after the interrupt with nothing running under it. A script that takes longer to shut down is still waited for. A second Ctrl+C ends the script's shell at once.
- On Windows, `pnpm run` now passes the arguments after the script name to the script as typed. Before, `cmd` expanded `%VAR%` in them and backslashes arrived doubled. Line breaks still arrive as the two characters `\n`, because `cmd` cannot pass them. The command line pnpm prints for the script quotes the arguments the same way on every platform [#​16257](https://github.com/pnpm/pnpm/issues/16257).
- The Windows `pnpm.exe` runs on a clean Windows install that does not have the Visual C++ Redistributable. It used to exit immediately on startup because that runtime was missing [#​15723](https://github.com/pnpm/pnpm/issues/15723).
- On Windows, the `.cmd` command shims in `node_modules/.bin` now keep a `%` in the project path. Before, cmd.exe expanded it as a variable reference, so the command received a mangled `NODE_PATH` [#​15716](https://github.com/pnpm/pnpm/issues/15716). Command shims also run tools whose paths contain non-ASCII characters [#​6999](https://github.com/pnpm/pnpm/issues/6999), including the PowerShell shims in Windows PowerShell 5.1 [#​16217](https://github.com/pnpm/pnpm/issues/16217).
- Bin shims in `node_modules/.bin` run from Cygwin on Windows again. The shims passed a `/cygdrive/c/...` path to the Windows `node` found on `PATH`, so Node.js failed with `Cannot find module 'C:\cygdrive\c\...'` [#​12845](https://github.com/pnpm/pnpm/issues/12845).
- On Windows, installing pnpm with npm inside a project now writes `node_modules/.bin` shims that run `pnpm.exe`. A global install with `npm install --location=global` now gets the same shims as `npm install -g` [#​15688](https://github.com/pnpm/pnpm/issues/15688).
- `pnpm install` no longer fails with `ERR_PNPM_WORKSPACE_INVALID_GLOB` on Windows for a wildcard pattern such as `plugins/*/*` in `pnpm-workspace.yaml` when the workspace is on a different drive than the pnpm cache or state directory [#​16239](https://github.com/pnpm/pnpm/issues/16239).
- On Windows, `pnpm install` no longer skips a dependency's build script on a later install when the package ships an executable file and the script changes nothing inside the package directory [#​15667](https://github.com/pnpm/pnpm/issues/15667).
- `pnpm setup` no longer writes the `pn.ps1`, `pnpx.ps1`, and `pnx.ps1` PowerShell wrappers. It also removes the ones an earlier setup wrote. PowerShell now runs `pn`, `pnpx`, and `pnx` through their `.cmd` wrappers, like `pnpm` itself. Before, these aliases failed with a "not digitally signed" error wherever the execution policy blocks unsigned scripts [#​8444](https://github.com/pnpm/pnpm/issues/8444).
- `pnpm setup` on Windows no longer panics when an unrelated environment variable has a name containing a non-ASCII character. It skips that variable [#​15684](https://github.com/pnpm/pnpm/issues/15684).
- On Windows, `pnpm setup` repairs the `PNPM_HOME` registry type left by older pnpm versions, even when the configured directory has not changed.
- On Windows, the `ERR_PNPM_BAD_ENV_FOUND` error of `pnpm setup` now shows the value `PNPM_HOME` is currently set to. Before, it showed the directory pnpm wanted to set.
- On Windows, pnpm expands nested `%VAR%` references in `PNPM_HOME` and the other directory environment variables it uses for its home, store, cache, state, and config directories. pnpm fails with an error when a `%VAR%` reference remains after expansion [#​13236](https://github.com/pnpm/pnpm/issues/13236).
- On Windows, if the global bin directory is not in `PATH` and a `PATH` entry still contains an unexpanded variable such as `%PNPM_HOME%`, the error now names that entry. A variable referenced from the user `Path` must be set to a full path and stored as a plain string (`REG_SZ`) for the entry to expand [#​5283](https://github.com/pnpm/pnpm/issues/5283).
##### Inspecting dependencies
- `pnpm audit` and `pnpm audit signatures` now fail with an error when the lockfile contains unresolvable dependency references [#​13638](https://github.com/pnpm/pnpm/issues/13638).
- `pnpm licenses list` now reports the actual on-disk package locations when using `nodeLinker: hoisted` or `shamefully-hoist: true` [#​8589](https://github.com/pnpm/pnpm/issues/8589). With `--json`, its `paths` array now includes every installed copy of a package, including hoisted copies and isolated installations with different peer dependencies.
- `pnpm root` now prints the configured `modulesDir`. It used to print `node_modules` regardless of the setting. A project's own `modulesDir` from `packageConfigs` is printed too [#​9113](https://github.com/pnpm/pnpm/issues/9113).
##### Output and messages
- With the default and append-only reporters, installs with `--loglevel warn` or `--loglevel error` now print the full output of a failed install script. The output of successful scripts, including the root project's own install hooks, stays hidden. With `--loglevel warn`, pnpm also prints ignored build script warnings.
- When a dependency fails to resolve, the error now shows the cause. For example, a Node.js runtime download behind a proxy that re-signs TLS now reports `invalid peer certificate: UnknownIssuer` [#​9556](https://github.com/pnpm/pnpm/issues/9556).
- When installing a git dependency over SSH fails with `Permission denied (publickey)`, pnpm suggests checking the loaded keys with `ssh-add -l`. Resolving an SSH URL that refuses the key also shows a local HTTPS rewrite that leaves the recorded URL alone [#​13743](https://github.com/pnpm/pnpm/issues/13743).
- Lockfile verification now fails with `ERR_PNPM_TARBALL_URL_MISMATCH`, `ERR_PNPM_TARBALL_REVISION_MISMATCH`, or `ERR_PNPM_MISSING_NAMED_REGISTRY` when every rejected entry failed that check. These failures were reported as the generic `ERR_PNPM_LOCKFILE_RESOLUTION_VERIFICATION`.
- The lockfile verification error now suggests relaxing the policy that flagged an entry only if a fresh resolution still fails and you trust the affected packages. Errors from checks that no policy controls, such as a missing tarball integrity, no longer suggest relaxing a policy [#​14411](https://github.com/pnpm/pnpm/issues/14411).
- `pnpm install` no longer prints an extra `Progress:` line after the progress line is marked `done` [#​16184](https://github.com/pnpm/pnpm/issues/16184).
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://notion.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/notion.svg" width="80" alt="Notion"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://latitude.so/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/latitude.svg" width="160" alt="Latitude"></a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
### [`v12.7.0`](https://github.com/pnpm/pnpm/releases/tag/v12.7.0): pnpm 12.7
[Compare Source](https://github.com/pnpm/pnpm/compare/v12.6.0...v12.7.0)
pnpm 12.7.0 ships with `.nvmrc` and `.node-version` support in the global `node` shim, `pnpm install --allow-build`, `pnpm publish --publish-wait-timeout`, and `pnpm-workspace.yaml` created from the `workspaces` field. `pnpm install --force` no longer installs optional dependencies built for other platforms. This release also carries security fixes for bin shims on Nix, for lifecycle scripts of packages in a `storeDir` inside the workspace, and for `userAgent` placeholders in `pnpm-workspace.yaml`.
##### Minor Changes
- `pnpm install --force` now keeps skipping optional dependencies whose `os`, `cpu` or `libc` do not match the host. It still refetches every package and lifts `engineStrict`. The new `forceIgnoresPlatform` setting restores the previous behaviour, installing optional dependencies of every platform under `--force` [#​6133](https://github.com/pnpm/pnpm/issues/6133).
- The global `node` shim created by pnpm now uses the Node.js version from the nearest `.nvmrc` or `.node-version` file when the project does not declare a Node.js runtime in `devEngines.runtime` or `engines.runtime` [#​4471](https://github.com/pnpm/pnpm/issues/4471). The nearest directory with a Node.js runtime declaration decides the version. Within one directory, `package.json` takes precedence over `.node-version`, which takes precedence over `.nvmrc`. An `.nvmrc` value that only nvm can act on, such as `system` or a custom alias, is ignored.
- `pnpm install` now supports the `--allow-build` option to selectively allow or deny package lifecycle scripts and record them in `pnpm-workspace.yaml` [#​15388](https://github.com/pnpm/pnpm/issues/15388).
- Added `pnpm publish --publish-wait-timeout <milliseconds>` to wait for published versions and their tarballs to become available from the registry. Set `publishWaitTimeout` in `pnpm-workspace.yaml` to configure a default. A value of `0` disables the check.
Recursive publishing confirms availability before publishing dependent packages. If confirmation times out, the command fails.
When `pnpm publish -r --report-summary` fails after some uploads were accepted, the summary file now lists those packages.
- `pnpm install` now creates `pnpm-workspace.yaml` from the `workspaces` field of the root `package.json` when the repository has no `pnpm-workspace.yaml`. The projects the field lists are linked on that same install. An existing `pnpm-workspace.yaml` is never changed. With `--ignore-workspace`, no file is created. If the `workspaces` field later differs from `packages` in `pnpm-workspace.yaml`, pnpm prints a warning [#​2255](https://github.com/pnpm/pnpm/issues/2255).
- When a project pins a pnpm version or a runtime that another pnpm process is installing at that moment, pnpm now waits a few seconds and then installs and runs a private copy of its own. It used to wait up to five minutes and then use the shared install directory without the lock. The private copy is removed once the command has run. `pnpm store prune` removes any private copy that a killed process left behind [#​15413](https://github.com/pnpm/pnpm/issues/15413).
- pnpm now keeps the blank lines between entries of `package.json` when it updates the file, for example on `pnpm add` [#​5602](https://github.com/pnpm/pnpm/issues/5602).
##### Patch Changes
##### Security
- pnpm no longer expands environment variables in a `userAgent` set in a project's `pnpm-workspace.yaml`. A `userAgent` with a placeholder in that file is now ignored. Before this fix, pnpm sent the variable's value to the configured registry [#​15415](https://github.com/pnpm/pnpm/issues/15415).
- On Nix, a dependency's bin named like a system utility such as `sed` can no longer redirect a POSIX bin shim or the `pnpm`, `pn`, `pnpx`, and `pnx` launchers. The shims and launchers now ignore `node_modules` and relative `PATH` entries while they locate their own files. Installing again replaces the shims already in `node_modules` [#​14883](https://github.com/pnpm/pnpm/issues/14883).
- pnpm no longer treats manifests inside its store, cache, state, or modules directories as workspace projects. Before, a `storeDir` inside the workspace could let lifecycle scripts of packages in the store run without `allowBuilds` approval [#​15033](https://github.com/pnpm/pnpm/issues/15033).
- Packages that run a lifecycle script are no longer hard-linked into the virtual store, so a build script can no longer rewrite the workspace source of an injected package or the store copy it was imported from [#​15483](https://github.com/pnpm/pnpm/issues/15483).
##### Installing packages
- Fixed `pnpm install`, `pnpm add`, `pnpm remove`, and `pnpm peers check` running out of memory when many packages share a missing peer dependency. This mostly affected projects with `autoInstallPeers: false` [#​15362](https://github.com/pnpm/pnpm/issues/15362).
- pnpm no longer hangs for up to 5 minutes after a pnpm process was killed while setting up the pnpm version pinned in `packageManager` or `devEngines` [#​15360](https://github.com/pnpm/pnpm/issues/15360), [#​15393](https://github.com/pnpm/pnpm/issues/15393). The killed process left behind a lock that every later pnpm command in the project waited on. pnpm now detects that the process holding a lock is gone and takes the lock over at once. The same applies to the locks pnpm takes while installing a managed runtime or writing the global bin directory. Two pnpm processes that are both still running keep waiting for each other as before.
- Requests to a registry or tarball server whose TLS certificate fails verification now fail at once. Such requests were retried for more than a minute without any output [#​9134](https://github.com/pnpm/pnpm/issues/9134).
- On macOS, pnpm now falls back to its bundled CA roots when system trust evaluation is unavailable, such as in a sandbox or when macOS cannot create an SSL policy for a registry connection. Installs failed or crashed on the first registry request in that case. Custom `ca` certificates are now honored directly [#​15329](https://github.com/pnpm/pnpm/issues/15329), [#​14461](https://github.com/pnpm/pnpm/issues/14461).
- `pnpm install` now caps concurrent connections to a proxy at 50 sockets by default [#​15280](https://github.com/pnpm/pnpm/issues/15280). It also immediately retries transient connection resets when downloading package archives.
- `pnpm install` now reuses a package already present in the store when an existing lockfile entry satisfies the dependency, avoiding registry requests that fail without authorization [#​2522](https://github.com/pnpm/pnpm/issues/2522).
- Installing or adding dependencies no longer fails when a previously installed local tarball file was deleted from disk [#​8367](https://github.com/pnpm/pnpm/issues/8367).
- `pnpm install` now installs the new version of a local tarball dependency whose file was replaced at the same path [#​2437](https://github.com/pnpm/pnpm/issues/2437). `pnpm install --frozen-lockfile` rejects such a changed tarball, even when the previous archive contents are in the store [#​1889](https://github.com/pnpm/pnpm/issues/1889).
- `pnpm install` now fetches committed submodules of git dependencies [#​1470](https://github.com/pnpm/pnpm/issues/1470).
- `pnpm install` now applies patches produced by `pnpm patch-commit` when an edit removes the trailing lines of a file along with its newline. The install no longer fails with `ERR_PNPM_INVALID_PATCH` ("expected end of hunk") [#​12451](https://github.com/pnpm/pnpm/issues/12451).
- `pnpm install` now preserves existing `node_modules` directories when a cross-device move reports `EXDEV` [#​14504](https://github.com/pnpm/pnpm/issues/14504).
- `pnpm install` no longer fails when writing the workspace state file encounters an error. Failures to update the state file now emit a warning instead of aborting the install [#​14550](https://github.com/pnpm/pnpm/issues/14550).
- Interrupting `pnpm install` with Ctrl+C or SIGTERM no longer leaves a temporary lockfile (`.pnpm-lock.yaml.*.tmp`) behind in the project [#​1418](https://github.com/pnpm/pnpm/issues/1418).
- `pnpm install` now relinks a direct dependency whose link in `node_modules` points to a missing target. Before, it reported "Already up to date" and left the broken link [#​9758](https://github.com/pnpm/pnpm/issues/9758).
- `pnpm install` uses less CPU when it links packages from a warm store. On Windows, a warm install could take several times longer than with pnpm 11 [#​15439](https://github.com/pnpm/pnpm/issues/15439).
- `pnpm install` now runs `node --version` once per run. A workspace whose projects keep their own lockfiles (`sharedWorkspaceLockfile: false`) previously ran the probe once or twice for every project, and on macOS the concurrent launches waited on each other, so a project could wait several seconds before its linking started.
- A repeat `pnpm install --frozen-lockfile` with `nodeLinker: hoisted` in a workspace no longer re-links `node_modules` when nothing changed.
- Custom fetcher hooks no longer run a second time during installation when an archive was already fetched during dependency resolution [#​15025](https://github.com/pnpm/pnpm/issues/15025).
- Fixed a package resolved by a `resolvers` pnpmfile hook installing without its own dependencies. This happened when the hook returned no `manifest` and a `fetchers` hook handled the resolution [#​15552](https://github.com/pnpm/pnpm/issues/15552).
- `pnpm install --prod` and other installs that skip `devDependencies` no longer run the `pnpm:devPreinstall` script [#​7065](https://github.com/pnpm/pnpm/issues/7065). They skip `prepare` lifecycle scripts too, as do installs given package arguments.
- `pnpm prune --prod` and production installs now remove devDependencies when `lockfile: false` is configured [#​2677](https://github.com/pnpm/pnpm/issues/2677).
- `pnpm install --prod`, `pnpm fetch --prod` and `pnpm deploy --prod` no longer install a devDependency that is only there to satisfy an optional peer dependency of a production dependency. `pnpm list`, `pnpm why`, `pnpm licenses`, `pnpm sbom` and `pnpm audit` leave it out of `--prod` results too. The same applies to `--dev`. A peer that is not optional is still installed and audited [#​15344](https://github.com/pnpm/pnpm/issues/15344).
- `pnpm install` no longer skips optional dependencies that the Node.js version locked for a `devEngines.runtime` range supports, when the range uses `onFail: download`. An explicitly set `nodeVersion` still takes priority [#​14628](https://github.com/pnpm/pnpm/issues/14628).
- `pnpm fetch` now also installs the pnpm version that `pnpm-lock.yaml` pins, when it differs from the running pnpm. A later `pnpm install --offline` that switches to the pinned version no longer fails because that version is missing from the store [#​11808](https://github.com/pnpm/pnpm/issues/11808).
- A dependency that ships a `binding.gyp` and sets `gypfile: false` no longer gets the `node-gyp rebuild` install script pnpm synthesizes for it. Such a dependency needs no `allowBuilds` entry and is no longer listed under "Ignored build scripts".
- `pnpm install` no longer adds `allowBuilds` placeholder entries to `pnpm-workspace.yaml` when it runs in CI or without a terminal. Interactive installs still add them [#​11574](https://github.com/pnpm/pnpm/issues/11574).
- pnpm now detects the same CI environments as pnpm 11, including AWS CodeBuild, which does not set `CI`. On these services `pnpm install` uses a frozen lockfile by default and fails with `ERR_PNPM_OUTDATED_LOCKFILE` when the lockfile is outdated.
##### Resolving and linking dependencies
- Installing through a pnpr server now installs a project's peer dependencies when `autoInstallPeers` is enabled. A project that declared only peer dependencies failed with `ERR_PNPM_OUTDATED_LOCKFILE` or skipped its peers [#​14833](https://github.com/pnpm/pnpm/issues/14833).
- pnpm now installs a dependency that a package also declares as an optional peer dependency, for example `lightningcss` in some vite builds. The dependency was missing from `node_modules`, so the package failed to import it [#​8912](https://github.com/pnpm/pnpm/issues/8912).
- Removal overrides such as `"parent>peer": "-"` now prevent optional peers from being installed from another workspace package [#​15008](https://github.com/pnpm/pnpm/issues/15008).
- Removing an entry from `overrides` now re-resolves the packages it targeted. A version the override had locked is no longer kept just because the declared range still accepts it [#​4587](https://github.com/pnpm/pnpm/issues/4587).
- `packageExtensions` and `overrides` entries with a ranged selector (such as `@<X` or `@*`) no longer match a dependency that has no `package.json`, such as a local directory dependency [#​15007](https://github.com/pnpm/pnpm/issues/15007).
- Trim leading and trailing whitespace from dependency override selectors in `pnpm.overrides` [#​6356](https://github.com/pnpm/pnpm/issues/6356).
- With `trustPolicy: no-downgrade`, pnpm now resolves the newest matching version that is not a trust downgrade. Previously a dependency failed with `ERR_PNPM_TRUST_DOWNGRADE` even when an older version satisfied its range. `pnpm self-update` picks its target version the same way. A request for an exact version still fails [#​14176](https://github.com/pnpm/pnpm/issues/14176).
- `pnpm install` now re-resolves a dependency when its manifest range is updated from a prerelease to a stable version. The lockfile previously retained the prerelease version and caused `--frozen-lockfile` to fail [#​15528](https://github.com/pnpm/pnpm/issues/15528).
- `pnpm install --ignore-pnpmfile` no longer removes `pnpmfileChecksum` from an up-to-date `pnpm-lock.yaml`. `pnpm install --frozen-lockfile --ignore-pnpmfile` no longer fails with `ERR_PNPM_LOCKFILE_CONFIG_MISMATCH` when the lockfile records a `pnpmfileChecksum`. A command that resolves dependencies with the pnpmfile ignored still writes the lockfile without it [#​10944](https://github.com/pnpm/pnpm/issues/10944).
- `pnpm install` and `pnpm peers check` now use local tarball packages' actual versions when checking peer dependencies. Compatible packages no longer fail with `strictPeerDependencies` enabled.
- `pnpm peers check` and the install-time peer dependency check now resolve peer dependencies from the workspace root when `resolvePeersFromWorkspaceRoot` is enabled [#​14982](https://github.com/pnpm/pnpm/issues/14982).
- `autoDedupe` and `pnpm dedupe` now move transitive dependencies to the version a `catalog:` dependency pins, as they already did for versions written directly in `package.json`. Previously they could move those dependencies to a higher version and keep both versions in the lockfile.
- `pnpm dedupe` now produces a stable lockfile when a dependency's range matches both a direct dependency and an `npm:` alias of the same package. The dependency resolves to the version of the direct dependency. Repeated runs previously alternated between two lockfiles [#​15588](https://github.com/pnpm/pnpm/issues/15588).
- Merging lockfiles now preserves recorded configuration fields such as `overrides`, `neverBuiltDependencies`, `patchedDependencies`, `packageExtensionsChecksum`, `settings`, and `catalogs` [#​8366](https://github.com/pnpm/pnpm/issues/8366).
- A lockfile entry whose resolution is unchanged now keeps its recorded `deprecated` message [#​5772](https://github.com/pnpm/pnpm/issues/5772).
- pnpm no longer writes a package's legacy array-form `engines`, such as `["node >= 0.8"]`, to the lockfile. It was recorded as an object keyed by index, such as `{'0': node >= 0.8}` [#​4518](https://github.com/pnpm/pnpm/issues/4518).
- Tarball URLs recorded in the lockfile now strip default HTTP and HTTPS ports (`:80` and `:443`) [#​15539](https://github.com/pnpm/pnpm/issues/15539).
- `node_modules/.package-map.json` no longer contains entries that point at directories that do not exist. Such entries appeared for packages installed only with peer dependencies, most visibly with `enableGlobalVirtualStore` [#​14938](https://github.com/pnpm/pnpm/issues/14938).
- With `nodeLinker: hoisted`, `hoistWorkspacePackages` now links each workspace project that `hoistPattern` or `publicHoistPattern` selects into the root `node_modules`, unless a hoisted package or a root dependency already uses its name. The project's bins are linked into the root `node_modules/.bin` [#​7553](https://github.com/pnpm/pnpm/issues/7553).
- With `nodeLinker: hoisted`, `pnpm install` now removes the commands of the packages it removes from `node_modules/.bin`, such as a nested copy deduped into the root `node_modules` [#​7568](https://github.com/pnpm/pnpm/issues/7568).
- `pnpm install` no longer puts a dependency's bin on `PATH` for that dependency's own lifecycle scripts before the bin's file exists. pnpm links such a bin after the dependency's build has run. It also removes such a bin left by an earlier install. This fixes installing the `node` package on Windows [#​15501](https://github.com/pnpm/pnpm/issues/15501).
- Dependencies and executable binaries are now correctly linked and accessible for workspace packages using `publishConfig.directory` and `publishConfig.linkDirectory` [#​8338](https://github.com/pnpm/pnpm/issues/8338).
- Bin linking leaves workspace and linked dependency files outside node\_modules unchanged. Already executable bin files no longer receive redundant permission changes.
##### Workspaces and filtering
- `pnpm install` now finds workspace projects reached through a symlink, such as a `packages` directory that links to a folder outside the workspace. It installs their dependencies, and the links in their `node_modules` resolve [#​1044](https://github.com/pnpm/pnpm/issues/1044).
- A dependency declared with `catalog:` now counts as a workspace dependency when its catalog entry points at a workspace project, for example `workspace:*` [#​15587](https://github.com/pnpm/pnpm/issues/15587). With `linkWorkspacePackages` enabled, so does an `npm:` alias of a workspace project, such as `"math-alias": "npm:math@^1.0.0"`. `pnpm -r run` runs that project first. `--filter <pkg>...` selects it.
- A `workspace:` dependency now resolves to a workspace project whose version is not valid semver, such as `1` or `1.0`. `workspace:*`, `workspace:^`, and `workspace:~` match it. A range identical to the version also matches it [#​4567](https://github.com/pnpm/pnpm/issues/4567).
- A `workspace:` dependency with an exact version now resolves to a workspace project whose version carries SemVer build metadata. For example, `workspace:0.5.6-next.3` matches a project at `0.5.6-next.3+f60facc` [#​6483](https://github.com/pnpm/pnpm/issues/6483).
- Secondary dependencies now prefer the version resolved by the local project's direct dependencies over versions from sibling workspace projects [#​7191](https://github.com/pnpm/pnpm/issues/7191).
- `pnpm install` now re-resolves a workspace project's auto-installed peer dependency when another workspace project changes its specifier for that package to one that excludes the locked version but still overlaps the peer range. The peer then resolves to the version a fresh install would pick [#​11800](https://github.com/pnpm/pnpm/issues/11800).
- `pnpm install --frozen-lockfile` now fails with `ERR_PNPM_OUTDATED_LOCKFILE` when `pnpm-lock.yaml` lists a workspace project whose directory or manifest file is missing. The install used to report success without installing that project's dependencies [#​7667](https://github.com/pnpm/pnpm/issues/7667).
- `pnpm install -r` now installs every workspace project when `recursiveInstall` is set to `false` in `pnpm-workspace.yaml` [#​7504](https://github.com/pnpm/pnpm/issues/7504).
- `pnpm install` with `--filter` now installs only the dependencies of the selected projects when using `nodeLinker: hoisted` [#​8882](https://github.com/pnpm/pnpm/issues/8882).
- `pnpm install` now updates an injected workspace dependency after that package's own dependencies change, when `shared-workspace-lockfile` is `false` [#​7209](https://github.com/pnpm/pnpm/issues/7209).
- `pnpm install` now copies the output of a workspace package's own `prepare`, `install`, or `postinstall` script into the injected copies of that package. Before, the injected copies kept only the files that existed before the script ran. `syncInjectedDepsAfterScripts` now also works when `modulesDir` is set [#​9464](https://github.com/pnpm/pnpm/issues/9464).
- `syncInjectedDepsAfterScripts` now copies files into injected dependencies when `node_modules` is on another filesystem than the package sources. The sync previously failed with a cross-device link error and made the script run exit with an error [#​14703](https://github.com/pnpm/pnpm/issues/14703).
- A `modulesDir` with several path segments, such as `www/modules`, now puts each workspace project's dependencies in `<project>/www/modules` on both fresh and frozen installs, and `pnpm bin` prints `<project>/www/modules/.bin` [#​15484](https://github.com/pnpm/pnpm/issues/15484).
- With `nodeLinker: hoisted`, pnpm now installs the root project's dependencies into a custom `modulesDir` instead of `node_modules`. With a custom `modulesDir`, the virtual store and its `lock.yaml` now default to `<modulesDir>/.pnpm`.
- A repeat `pnpm install` in a workspace with a custom `modulesDir` now takes the up-to-date fast path. Before, pnpm looked for each workspace project's dependencies in `node_modules` and ran a full install every time.
- pnpm now warns when a workspace install covers a project that has its own `pnpm-workspace.yaml`. The nested file's settings, such as `patchedDependencies`, do not apply when the outer workspace installs that project. pnpm reads settings only from the `pnpm-workspace.yaml` at the workspace root [#​11724](https://github.com/pnpm/pnpm/issues/11724).
- The `[<since>]` filter selector now compares against the commit where the current branch forked from `<since>`. Projects changed only by newer commits on `<since>` are no longer selected. Uncommitted changes are still included. In a shallow clone without that commit, pnpm compares against `<since>` directly, as before [#​9907](https://github.com/pnpm/pnpm/issues/9907).
- `--filter "[<since>]"` now selects workspace packages when dependency versions change in a catalog in `pnpm-workspace.yaml` [#​8718](https://github.com/pnpm/pnpm/issues/8718). It also selects projects that files were moved out of when git detects the move as a rename [#​15481](https://github.com/pnpm/pnpm/issues/15481).
- `--filter` now evaluates selectors in order, so later inclusion filters can re-include packages that an earlier exclusion filter excluded [#​9354](https://github.com/pnpm/pnpm/issues/9354).
##### Adding, updating, and removing dependencies
- `pnpm add` now saves changes to `package.json` before running lifecycle scripts, so a postinstall script failure leaves the added dependency in `package.json` [#​8627](https://github.com/pnpm/pnpm/issues/8627).
- `pnpm add` now saves the requested exact version when adding a dependency, even when the manifest already contains a version range [#​6040](https://github.com/pnpm/pnpm/issues/6040).
- `pnpm add <pkg>@​<version>` and `pnpm update <pkg>@​<version>` now move the catalog entry onto the named version when the entry's range already covers it. For example, `^7.22.17` becomes `^7.29.6`, the same way `pnpm update <pkg>` moves an entry to the version it resolves [#​13715](https://github.com/pnpm/pnpm/issues/13715).
- `pnpm add` and `pnpm install` keep an empty `peerDependencies`, `dependencies`, `devDependencies`, or `optionalDependencies` field that was already in `package.json`. pnpm still drops such a field when it removes the last entry itself, as `pnpm remove` does [#​5096](https://github.com/pnpm/pnpm/issues/5096).
- `pnpm update` now keeps a version range whose shape has no save prefix, such as `<= 3.0.0` or `>=1.0.0 <2.0.0`, when the updated version still satisfies it. Before, `<= 3.0.0` became `^3.0.0` [#​6714](https://github.com/pnpm/pnpm/issues/6714).
- `pnpm update <pkg>` now moves a package off a locked version the registry no longer serves, such as an unpublished release. The lockfile check for supply-chain policies such as `minimumReleaseAge` used to reject that version before the update could replace it [#​9953](https://github.com/pnpm/pnpm/issues/9953).
- `pnpm update --prod` no longer installs devDependencies when run in a project installed with `--prod` [#​8038](https://github.com/pnpm/pnpm/issues/8038).
- `pnpm update --interactive --workspace` now allows external dependencies to be updated.
- `pnpm outdated` and `pnpm update` now apply `minimumReleaseAge` to GitHub Actions. `minimumReleaseAgeExclude` entries match action names such as `actions/checkout` [#​13923](https://github.com/pnpm/pnpm/issues/13923).
- `pnpm remove` now accepts `--trust-lockfile` and `--no-trust-lockfile` to control supply-chain policy checks while removing a package [#​14406](https://github.com/pnpm/pnpm/issues/14406).
- `pnpm unlink` now removes the `link:` dependency that `pnpm link <dir>` added to `package.json`. The linked package is removed from `node_modules` and the lockfile. A `link:` dependency to another directory is kept [#​4219](https://github.com/pnpm/pnpm/issues/4219).
- `pnpm install` now prunes unreferenced catalog entries from `pnpm-workspace.yaml` when `catalogPrune: true` is configured [#​15273](https://github.com/pnpm/pnpm/issues/15273).
- `minimumReleaseAgeExcludePrune` and `trustPolicyExcludePrune` now work in workspaces with `shared-workspace-lockfile=false`. Once every project has been installed, pnpm drops an entry only if no project lockfile records it. Undecided `allowBuilds` entries are pruned the same way [#​14612](https://github.com/pnpm/pnpm/issues/14612).
- Exclude entries that pnpm writes to `pnpm-workspace.yaml` now match the file's list indentation and dominant quote style [#​15571](https://github.com/pnpm/pnpm/issues/15571), [#​15079](https://github.com/pnpm/pnpm/issues/15079).
- `pnpm import` now converts dependencies that use Yarn's `patch:` protocol. The dependency keeps the version it patches, and the patch file is added to `patchedDependencies` in `pnpm-workspace.yaml`. If the patch file is missing, pnpm prints a warning and imports the dependency without the patch [#​10278](https://github.com/pnpm/pnpm/issues/10278).
- `pnpm import` in a workspace now keeps the versions pinned by the root `yarn.lock`, `package-lock.json`, or `npm-shrinkwrap.json` when another workspace project's range allows a newer version. Before, the root project got the newest version in its range [#​4385](https://github.com/pnpm/pnpm/issues/4385).
- `pnpm patch`, `pnpm patch-commit`, and `pnpm patch-remove` now work in a project of a workspace with `sharedWorkspaceLockfile: false`. `pnpm patch` failed there with `ERR_PNPM_PATCH_NO_LOCKFILE` after a successful install. The reinstall after committing or removing a patch left the project's own `node_modules` unchanged [#​9926](https://github.com/pnpm/pnpm/issues/9926).
- `pnpm patch-commit` now resolves default patch directory locations when passed a package name or package specifier (such as `pnpm patch-commit <pkg>` or `pnpm patch-commit <pkg>@​<version>`).
- `pnpm patch-commit` now updates the lockfile snapshot and prunes removed dependencies when the patch modifies `package.json` [#​6866](https://github.com/pnpm/pnpm/issues/6866).
- `pnpm patch-commit` now falls back to copying package files when hard linking fails.
##### Running scripts and commands
- A script that pnpm runs without a terminal now ends when pnpm itself is killed. Killing pnpm's process group, as Playwright's `webServer` does to stop the command it started, used to leave the script running and holding the caller's output pipes open [#​15555](https://github.com/pnpm/pnpm/issues/15555).
- `pnpm --filter <project> <command>` and `pnpm -r <command>` now run a command installed in the selected projects' dependencies when none of them has a script by that name. This matches `pnpm <command>` in a single project. `pnpm run` with `--filter` or `-r` still reports the missing script [#​10151](https://github.com/pnpm/pnpm/issues/10151).
- `pnpm exec` and `pnpm dlx` now set `npm_execpath`, `INIT_CWD`, `npm_node_execpath`, and `NODE` in child environments when Node.js is available. Stale inherited `NODE` and `npm_node_execpath` variables are cleared when Node.js cannot be found on PATH [#​7037](https://github.com/pnpm/pnpm/issues/7037). Scripts that `pnpx` and `pnx` run now get pnpm itself as `npm_execpath`. A script that ran `$npm_execpath install` there ran `pnpm dlx install`.
- `pnpm exec` now sets the `PWD` environment variable to the directory the command runs in. Shells and tools that read `PWD` now report the logical path of a workspace package reached through a symlink [#​1550](https://github.com/pnpm/pnpm/issues/1550).
- A script that runs `pnpm run` no longer adds duplicate `node_modules/.bin` and `node-gyp-bin` entries to `PATH` [#​5352](https://github.com/pnpm/pnpm/issues/5352).
- Concurrent `pnpm run` and `pnpm exec` commands now serialize their dependency installs [#​14551](https://github.com/pnpm/pnpm/issues/14551).
- `pnpm run` and `pnpm exec` with `verifyDepsBeforeRun` now accept a moved project whose store is on the project's volume. Before, the check reported that the workspace structure had changed whenever the default store was not on the home volume.
- `verifyDepsBeforeRun` checks now account for project-specific `packageConfigs` overrides in workspaces with `sharedWorkspaceLockfile: false` [#​15545](https://github.com/pnpm/pnpm/issues/15545).
- `pnpm restart` now runs the "stop" and "start" scripts when the package has no "restart" script. Previously it ran "stop" and then failed with "Missing script: restart" [#​4750](https://github.com/pnpm/pnpm/issues/4750).
- `pnpm dlx` now keeps a separate cache entry for each Node.js major version. A package built under one Node.js major version, such as a native addon, is no longer reused under another [#​8611](https://github.com/pnpm/pnpm/issues/8611).
- `pnpm pipeline` no longer fails when run in a project outside a Git work tree or on a system without `git`. Tasks in those projects run without caching, and pnpm prints a warning explaining why [#​15601](https://github.com/pnpm/pnpm/issues/15601).
- A `runtime:` version range that contains `||` or a space, such as a `devEngines.runtime` version of `^22.18.0 || ^24.0.0`, now installs the requested runtime. pnpm used to install the npm package with the same name, such as `node` [#​14817](https://github.com/pnpm/pnpm/issues/14817).
- When the configured `scriptShell` does not exist, running a script now fails with an error that names the shell. Previously pnpm printed only an exit code or the package directory [#​7562](https://github.com/pnpm/pnpm/issues/7562).
- A script killed by a signal now fails with an error that names the signal, such as `Command failed with signal SIGKILL.` [#​9821](https://github.com/pnpm/pnpm/issues/9821).
##### Publishing, packing, and deploying
- `pnpm publish` now resolves `workspace:` dependencies from workspace manifests when `node_modules` is not installed. Previously, publishing without `node_modules` failed with `ERR_PNPM_CANNOT_RESOLVE_WORKSPACE_PROTOCOL` [#​6567](https://github.com/pnpm/pnpm/issues/6567).
- `pnpm publish` now honors `publishConfig["@​scope:registry"]` for a package in that scope. It takes precedence over the registry set for the same scope in `.npmrc` and over `publishConfig.registry` [#​12071](https://github.com/pnpm/pnpm/issues/12071).
- `pnpm pack` and `pnpm publish` now include bundled dependencies when using the isolated linker. This covers workspace packages and the dependencies of each bundled package. Bundled dependencies are also included when `publishConfig.directory` selects a build directory [#​1643](https://github.com/pnpm/pnpm/issues/1643).
- `pnpm pack`, `pnpm deploy`, and installs of local directory dependencies now keep symlinks that point to files or directories included in the package. `pnpm pack` leaves out symlinks that point outside the package [#​8208](https://github.com/pnpm/pnpm/issues/8208).
- `pnpm pack` now preserves file executable permissions in the packed tarball when source files are executable on disk.
- `pnpm publish` and `pnpm pack` now report a missing `version` or `name` field on a workspace dependency. Previously, pnpm reported that the dependency was not installed [#​4164](https://github.com/pnpm/pnpm/issues/4164).
- `pnpm publish` and `pnpm pack` now report an error when a bin script has a shebang line ending with CRLF [#​7311](https://github.com/pnpm/pnpm/issues/7311).
- `pnpm deploy` now copies the `packageManager` and `devEngines.packageManager` fields of the workspace root `package.json` into the deployed `package.json`, unless the deployed project pins a package manager itself [#​9079](https://github.com/pnpm/pnpm/issues/9079).
- `pnpm deploy` now puts the virtual store at `virtualStoreDir`, resolved against the deploy directory. A shared-lockfile deploy records `virtualStoreDir` in the deployed `pnpm-workspace.yaml`. With the global virtual store enabled or an absolute `virtualStoreDir`, the deploy still uses `node_modules/.pnpm` [#​8787](https://github.com/pnpm/pnpm/issues/8787).
- `pnpm deploy` now respects `--package-import-method` passed on the command line and reports the package import method correctly [#​7593](https://github.com/pnpm/pnpm/issues/7593).
- `pnpm deploy` does not run the `prepare` scripts of the deployed project [#​7282](https://github.com/pnpm/pnpm/issues/7282).
- `pnpm deploy --legacy` no longer rewrites the source workspace's `node_modules/.pnpm-workspace-state-v1.json` to describe only the deployed project [#​15352](https://github.com/pnpm/pnpm/issues/15352).
##### Manifests and configuration files
- pnpm now reads and updates `package.json5` project manifests. Manifest updates retain comments, and workspace discovery prefers `package.json`, then `package.json5`, then `package.yaml` [#​15129](https://github.com/pnpm/pnpm/issues/15129). `pnpm pack` includes exactly one `package.json` in the archive when the project uses an alternative manifest format, even when `.npmignore` or `files` excludes the source file.
- Git-hosted dependencies that use a `package.yaml` or `package.json5` manifest now honor its `files` field [#​7906](https://github.com/pnpm/pnpm/issues/7906).
- Fixed `pnpm version` failing on projects using a `package.yaml` manifest.
Fixed `pnpm init` creating an extra `package.json` when `package.yaml` is already present.
- `pnpm version` now applies pending bumps to private workspace packages. A private package's changelog is written to its committed `CHANGELOG.md`, also when `versioning.changelog.storage` is `registry` [#​13736](https://github.com/pnpm/pnpm/issues/13736), [#​13519](https://github.com/pnpm/pnpm/issues/13519).
- `pnpm init` now supports the `--bare` option. It creates a `package.json` file with only the required fields [#​15538](https://github.com/pnpm/pnpm/issues/15538).
- The `reporter` setting is now honored when it is configured in `pnpm-workspace.yaml`, the global configuration, or the `PNPM_CONFIG_REPORTER` environment variable. Configured `reporter: silent` makes silent output the default. An explicit `--reporter` takes precedence [#​4879](https://github.com/pnpm/pnpm/issues/4879).
- `.npmrc` and `pnpm-workspace.yaml` files now support npm's `${VAR?}` placeholder. It expands to the value of `VAR`, or to an empty string without a warning when `VAR` is unset [#​14404](https://github.com/pnpm/pnpm/issues/14404).
- pnpm now expands environment variables in `_auth.authToken` values loaded from global `config.yaml` and `pnpm_config__auth`.
- pnpm now keeps the configured default registry when `_auth` holds credentials for several registries and some of those registries serve package scopes.
Lockfile verification checks a tarball hosted on a scoped registry against that registry's metadata, unless the package's own scope has a registry assigned [#​15530](https://github.com/pnpm/pnpm/issues/15530).
- pnpm now parses the first setting in a `.npmrc` that starts with a UTF-8 byte order mark. Previously, the leading byte order mark caused the first line's key to be ignored [#​15353](https://github.com/pnpm/pnpm/issues/15353).
- pnpm now prints a warning when a `.npmrc`, `auth.ini`, or the file set by `npmrcAuthFile` exists but cannot be read. The settings in such a file were ignored without any message. A `.npmrc` that contains invalid UTF-8 is now read [#​5065](https://github.com/pnpm/pnpm/issues/5065).
- `pnpm config set` and `pnpm config delete` now preserve comments and repeated keys such as `ca=` in `.npmrc` [#​14851](https://github.com/pnpm/pnpm/issues/14851).
- `pnpm login` now logs back in to an existing user on registries without web login, such as verdaccio. The classic login request sends the username and password as basic auth, as `npm login` does [#​12055](https://github.com/pnpm/pnpm/issues/12055).
- Commands that do not use the store no longer create a temporary file in the project directory when they load their settings. These include `pnpm view`, `pnpm config`, `pnpm root`, `pnpm bin`, `pnpm exec`, `pnpm run`, the script shortcuts such as `pnpm test`, and the registry commands such as `pnpm whoami`, `pnpm dist-tag`, and `pnpm search`. `pnpm exec`, `pnpm run`, and the script shortcuts still create one in projects that declare `configDependencies`.
##### Global packages, pnpm versions, and runtimes
- Global commands such as `pnpm add --global`, `pnpm list --global`, and `pnpm bin --global` now run with the pnpm you invoked, even in a project that pins another pnpm version. Previously, a pin with `onFail: "download"` switched them to the pinned pnpm, and a pinned pnpm 10 or older failed because its global bin directory was not in `PATH` [#​14531](https://github.com/pnpm/pnpm/issues/14531).
- `pnpm add -g`, `pnpm update -g`, and `pnpm remove -g` no longer fail with `ERR_PNPM_PACKAGE_MANIFEST_IO_ERROR` when another global package's link into the store dangles, for example after the store was pruned. The pnpm install script failed the same way on such a machine.
- `pnpm update --global` now skips a global package installed from a `file:` path that no longer exists, prints a warning, and updates the remaining global packages. Previously the whole update failed with `ERR_PNPM_LINKED_PKG_DIR_NOT_FOUND` [#​12533](https://github.com/pnpm/pnpm/issues/12533).
- `pnpm self-update` run in a project that pins pnpm through `packageManager` or `devEngines.packageManager` now also updates the global pnpm, as it does outside a project [#​14747](https://github.com/pnpm/pnpm/issues/14747).
- `pnpm self-update` no longer leaves the previous pnpm in the global packages when it was installed as `@pnpm/exe`. `pnpm ls -g` now lists a single pnpm [#​14709](https://github.com/pnpm/pnpm/issues/14709). `pnpm setup` now installs pnpm under the package name `pnpm` too, so both commands leave the same shims in the global bin directory. After a self-update on Windows, PowerShell ran pnpm through `pnpm.cmd` and asked "Terminate batch job (Y/N)?" on Ctrl+C [#​15567](https://github.com/pnpm/pnpm/issues/15567).
- `pnpm setup` failed with `Text file busy (os error 26)` when `$PNPM_HOME/bin` already held `pn`, `pnpx`, or `pnx` as links to the running pnpm executable. It now replaces those files and completes [#​15494](https://github.com/pnpm/pnpm/issues/15494).
- `pnpm setup` no longer deletes aliases and other lines that sit between a `# pnpm` comment and the pnpm block in a shell startup file [#​7067](https://github.com/pnpm/pnpm/issues/7067).
- When pnpm switches to the version a project pins, the `minimumReleaseAge` approvals for that version are now added to `minimumReleaseAgeExclude` in the project's `pnpm-workspace.yaml`. A project without that file gets one. Global commands leave the project's settings unchanged [#​15396](https://github.com/pnpm/pnpm/issues/15396).
- `pnpm env remove` now cleans up dangling Node.js executables and shims. Surviving global commands remain intact.
- Node.js runtime resolution now supports Windows ARM64. Node.js 20 and newer resolve native `win-arm64` builds, and older versions fall back to `win-x64` under emulation [#​7123](https://github.com/pnpm/pnpm/issues/7123).
##### Windows and WSL
- On Windows, `pnpm clean` and installs no longer fail immediately when another process uses a package in `node_modules`. pnpm waits up to a minute for an open file. It waits up to 5 seconds for a running program [#​15081](https://github.com/pnpm/pnpm/issues/15081).
- `pnpm install` in WSL now waits out Windows file locks on a Windows drive such as `/mnt/c`, as it already does on Windows. Before, an antivirus or indexer scan holding a file open could fail the install with `EACCES` [#​6155](https://github.com/pnpm/pnpm/issues/6155).
- On Windows, pnpm now retries saving `pnpm-lock.yaml` for up to a minute while another process holds the file open. The save used to fail at once with `EPERM`, `EBUSY`, or "Access is denied" [#​9461](https://github.com/pnpm/pnpm/issues/9461).
- pnpm now escapes trailing dots and spaces in `node_modules/.pnpm` directory names. Windows strips these characters, so a dependency such as `"parent-pkg": "file:../"` created a directory that could not be deleted or failed to install [#​8101](https://github.com/pnpm/pnpm/issues/8101).
- On Windows, bin shims run from Git Bash, MSYS2, or Cygwin now pass `NODE_PATH` to Node.js as Windows paths. A project installed from cmd or PowerShell gave its bins a `NODE_PATH` under the Git install directory when they ran from Git Bash. Installing again replaces the shims already in `node_modules` [#​3360](https://github.com/pnpm/pnpm/issues/3360).
- Fixed scripts failing with errors such as `'an-compile' is not recognized` when `scriptShell` is set to `cmd.exe` on Windows [#​7181](https://github.com/pnpm/pnpm/issues/7181).
- On Windows, the error for a `node_modules` directory that pnpm cannot move out of the way now names the directory and says that a file in it is probably in use by another process [#​7505](https://github.com/pnpm/pnpm/issues/7505).
##### Inspecting dependencies
- `pnpm audit` and `pnpm audit signatures` now check only the dependencies of the projects selected by `--filter`, `--filter-prod`, or `--workspace-root`. The filter used to be ignored, so a filtered audit reported the whole workspace [#​10982](https://github.com/pnpm/pnpm/issues/10982).
- `pnpm audit` now lists at least one dependency path from every workspace project that depends on a vulnerable package. Before, a project whose dependency was reached through more than 100 paths filled the path list, and other projects that depend on the same package were left out [#​12200](https://github.com/pnpm/pnpm/issues/12200).
- `pnpm audit --fix` now prunes redundant overrides when one vulnerable range is a subset of another for the same package [#​8577](https://github.com/pnpm/pnpm/issues/8577).
- Running `pnpm list` inside a workspace package without `--recursive` or a filter now lists only the current package [#​14494](https://github.com/pnpm/pnpm/issues/14494). `pnpm licenses list` does the same. Use `--recursive` or `--filter` to list the licenses of other workspace projects [#​5689](https://github.com/pnpm/pnpm/issues/5689).
- `pnpm list --only-projects` now prints every project selected with `--filter` or `--recursive`, including a project that has no workspace dependencies [#​9770](https://github.com/pnpm/pnpm/issues/9770). It also lists the workspace projects when `sharedWorkspaceLockfile` is `false` [#​7151](https://github.com/pnpm/pnpm/issues/7151), and a project that sets `publishConfig.directory` [#​10635](https://github.com/pnpm/pnpm/issues/10635). It no longer reports packages in `node_modules` that are missing from the lockfile [#​9528](https://github.com/pnpm/pnpm/issues/9528).
- `pnpm licenses list` failed or reported nothing in a workspace with `sharedWorkspaceLockfile: false`. It now reads the lockfile of each selected project [#​10140](https://github.com/pnpm/pnpm/issues/10140).
- With `nodeLinker: hoisted`, `pnpm licenses list` reported every license as `Unknown` and listed paths under `node_modules/.pnpm` that do not exist. It now reads each package from the directory where the hoisted linker placed it [#​8589](https://github.com/pnpm/pnpm/issues/8589).
- `pnpm outdated` and `pnpm -r outdated` now fail with `ERR_PNPM_NO_PACKAGE_IN_DEPENDENCIES` when a requested package selector does not match any dependency in the inspected projects [#​2319](https://github.com/pnpm/pnpm/issues/2319).
- `pnpm -r outdated --json` now includes every outdated workspace dependency when multiple projects depend on different versions or dependency types of the same package. Such a package is keyed by its current version and dependency type, for example `[email protected] (dev)` [#​7693](https://github.com/pnpm/pnpm/issues/7693).
- `pnpm sbom` filtered to a single workspace project now takes the `author`, `description`, `license`, `repository`, and `bugs` fields from the workspace root `package.json` when the project does not declare them. A field the project declares is never taken from the root, even when it is blank or `null` [#​14882](https://github.com/pnpm/pnpm/issues/14882).
- `pnpm store status` no longer reports a package as modified when it has build or postinstall scripts, peer dependencies, or skipped optional dependencies [#​15383](https://github.com/pnpm/pnpm/issues/15383). When packages were mutated, it now lists only those packages and no longer suggests running `pnpm install --force` [#​919](https://github.com/pnpm/pnpm/issues/919).
- `pnpm peers check` and the `ERR_PNPM_PEER_DEP_ISSUES` error now group peer dependency issues under the workspace project they were found in [#​15351](https://github.com/pnpm/pnpm/issues/15351).
##### Output and messages
- The error for an incompatible pnpm-lock.yaml now reports the lockfileVersion the file was generated with and the lockfileVersion the current pnpm supports [#​848](https://github.com/pnpm/pnpm/issues/848).
- When the registry stops sending data for longer than `fetchTimeout`, pnpm now reports that the metadata or tarball request timed out. Previously the error did not mention the timeout [#​3646](https://github.com/pnpm/pnpm/issues/3646).
- Fatal peer dependency errors and their hints are now written to stderr [#​5419](https://github.com/pnpm/pnpm/issues/5419).
- `pnpm run` with `--loglevel` set to `warn`, `error`, or `silent` (or the same `loglevel` setting) no longer prints the `$ <command>` line before a script, nor the summary of the install that `verifyDepsBeforeRun` runs first. Both are info-level output [#​8944](https://github.com/pnpm/pnpm/issues/8944).
- `pnpm run` and `pnpm exec` now print `No projects matched the filters in "<workspace>"` when `--filter` selects no project [#​8408](https://github.com/pnpm/pnpm/issues/8408).
- `pnpm dedupe` now counts each package reused from the store once in its progress output [#​15303](https://github.com/pnpm/pnpm/issues/15303).
- `pnpm add` now warns when replacing an existing dependency with a specifier pointing to a different source [#​14869](https://github.com/pnpm/pnpm/issues/14869).
- `pnpm link` now warns when linking a package that declares one or more peer dependencies, explaining that the linked dependency will not resolve peer dependencies from the target `node_modules` and suggesting the `file:` protocol instead.
- `pnpm import` now warns when package.json lists projects in a "workspaces" array and there is no "pnpm-workspace.yaml". Without that file, the import writes a lockfile for the root project only [#​5240](https://github.com/pnpm/pnpm/issues/5240).
- Bash completion now completes script names that contain a colon, such as `pnpm run test:u` to `pnpm run test:unit` [#​5482](https://github.com/pnpm/pnpm/issues/5482).
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://notion.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/notion.svg" width="80" alt="Notion"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://latitude.so/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/latitude.svg" width="160" alt="Latitude"></a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
</details>
<details>
<summary>lovell/sharp (sharp)</summary>
### [`v0.35.5`](https://github.com/lovell/sharp/releases/tag/v0.35.5)
[Compare Source](https://github.com/lovell/sharp/compare/v0.35.4...v0.35.5)
<https://github.com/lovell/sharp-libvips/releases/tag/v1.3.4>
- Add upper bounds check on length of `linear` and GIF `delay` arrays.
- Improve error handing when WebAssembly fallback also fails.
[#​4593](https://github.com/lovell/sharp/pull/4593)
[@​lazerg](https://github.com/lazerg)
- TypeScript: Allow multi-frame options for JXL output.
[#​4602](https://github.com/lovell/sharp/pull/4602)
[@​ramin-010](https://github.com/ramin-010)
- TypeScript: Remove non-existent named export.
[#​4604](https://github.com/lovell/sharp/issues/4604)
- Increase accepted dimensions when extending an image.
[#​4605](https://github.com/lovell/sharp/issues/4605)
- Improve gain map support for `extract` and `rotate` operations.
[#​4606](https://github.com/lovell/sharp/issues/4606)
- Tests: Ensure composite tests pass on big endian platforms.
[#​4609](https://github.com/lovell/sharp/issues/4609)
</details>
<details>
<summary>HiDeoo/starlight-openapi (starlight-openapi)</summary>
### [`v0.26.3`](https://github.com/HiDeoo/starlight-openapi/blob/HEAD/packages/starlight-openapi/CHANGELOG.md#0263)
[Compare Source](https://github.com/HiDeoo/starlight-openapi/compare/[email protected]@0.26.3)
##### Patch Changes
- [#​181](https://github.com/HiDeoo/starlight-openapi/pull/181) [`e81882f`](https://github.com/HiDeoo/starlight-openapi/commit/e81882ffd4a85b1c113fd477422ee6b524715d58) Thanks [@​HiDeoo](https://github.com/HiDeoo)! - Fixes a potential spacing issue with heading anchor links.
</details>
---
### Configuration
📅 **Schedule**: (UTC)
- Branch creation
- Between 12:00 AM and 03:59 AM, only on Monday (`* 0-3 * * 1`)
- Automerge
- At any time (no schedule defined)
🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied.
♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
👻 **Immortal**: This PR will be recreated if closed unmerged. Get [config help](https://github.com/renovatebot/renovate/discussions) if that's undesired.
---
- [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check this box
---
This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).
<!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0My4xOTEuMiIsInVwZGF0ZWRJblZlciI6IjQzLjE5MS4yIiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6W119-->
<!-- cloudflare-preview --> Preview: https://pr-575.docs-gitea-com.pages.dev
Reviewed-on: https://gitea.com/gitea/docs/pulls/575
Reviewed-by: silverwind <[email protected]>
Co-authored-by: Renovate Bot <[email protected]>
The Docker Compose example currently mounts the PostgreSQL data directory to `/var/lib/postgresql/data`.
With PostgreSQL 18, the official PostgreSQL Docker image expects the volume to be mounted at `/var/lib/postgresql` instead. Mounting `/var/lib/postgresql/data` causes the container to fail during startup because the image detects PostgreSQL data at an unused mount point.
This change updates the example to use the recommended `/var/lib/postgresql` mount path for PostgreSQL 18 and newer.
---------
Co-authored-by: silverwind <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/570
Reviewed-by: silverwind <[email protected]>
Co-authored-by: JWheelMaker <[email protected]>
The cheat sheet lists `/img/repo_default.png` as the `REPOSITORY_AVATAR_FALLBACK_IMAGE` default, but public files moved under `/assets` in Gitea 1.15 (https://github.com/go-gitea/gitea/pull/15219), so that path 404s. Like `app.example.ini` in https://github.com/go-gitea/gitea/pull/39514, document the default as empty, falling back to Gitea's builtin repository avatar, and note that a custom value is used as-is without the sub-path. Applies to all documented versions and translations.
_Written by Claude Code (Opus 5.5)._
Reviewed-on: https://gitea.com/gitea/docs/pulls/569
Reviewed-by: wxiaoguang <[email protected]>
Co-authored-by: silverwind <[email protected]>
Documents the settings changed by https://github.com/go-gitea/gitea/pull/39426: `EGRESS_MODE`, the `[migrations]` `ALLOWED_HOST_LIST`/`BLOCKED_HOST_LIST` replacing `ALLOWED_DOMAINS`, `BLOCKED_DOMAINS` and `ALLOW_LOCALNETWORKS`, and the host list syntax with ports, CGNAT and reserved addresses. Also fixes the `socks://` examples, Go only speaks `socks5://`.
_Written by Claude._
---------
Co-authored-by: Lunny Xiao <[email protected]>
Co-authored-by: TheFox0x7 <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/557
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: silverwind <[email protected]>
Automated update of the generated runner reference pages from gitea/runner (main for the develop docs, the newest stable tag for every documented series), opened by the "update runner reference" scheduled workflow.
---------
Co-authored-by: Lunny Xiao <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/555
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Gitea Bot <[email protected]>
This PR contains the following updates:
| Package | Change | [Age](https://docs.renovatebot.com/merge-confidence/) | [Confidence](https://docs.renovatebot.com/merge-confidence/) |
|---|---|---|---|
| [@astrojs/starlight](https://starlight.astro.build) ([source](https://github.com/withastro/starlight/tree/HEAD/packages/starlight)) | [`0.42.0` → `0.42.1`](https://renovatebot.com/diffs/npm/@astrojs%2fstarlight/0.42.0/0.42.1) |  |  |
| [pnpm](https://github.com/pnpm/pnpm/tree/main/pnpm) ([source](https://github.com/pnpm/pnpm/tree/HEAD/pnpm/npm/pnpm)) | [`12.4.0` → `12.4.2`](https://renovatebot.com/diffs/npm/pnpm/12.4.0/12.4.2) |  |  |
| [yaml](https://eemeli.org/yaml/) ([source](https://github.com/eemeli/yaml)) | [`2.9.0` → `2.9.1`](https://renovatebot.com/diffs/npm/yaml/2.9.0/2.9.1) |  |  |
---
### Release Notes
<details>
<summary>withastro/starlight (@​astrojs/starlight)</summary>
### [`v0.42.1`](https://github.com/withastro/starlight/blob/HEAD/packages/starlight/CHANGELOG.md#0421)
[Compare Source](https://github.com/withastro/starlight/compare/@astrojs/[email protected]...@astrojs/[email protected])
##### Patch Changes
- [#​4198](https://github.com/withastro/starlight/pull/4198) [`96a44d4`](https://github.com/withastro/starlight/commit/96a44d4f296b512febc6e631a587103ec5297872) Thanks [@​mayank99](https://github.com/mayank99)! - Removes the `popover` attribute from the sidebar pane on desktop viewports. This ensures the desktop sidebar doesn't stay in a "hidden" popover visibility state.
</details>
<details>
<summary>pnpm/pnpm (pnpm)</summary>
### [`v12.4.2`](https://github.com/pnpm/pnpm/releases/tag/v12.4.2): pnpm 12.4.2
[Compare Source](https://github.com/pnpm/pnpm/compare/v12.4.1...v12.4.2)
pnpm 12.4.2 includes security fixes for executable shims and GitHub Actions links, more reliable installs, faster peer dependency checks in workspaces, and Python lockfiles that work across compatible targets.
#### Patch Changes
##### Security
- Dependency executables can no longer take over another package's POSIX bin shim through its shell helpers. Reinstall dependencies to replace existing shims [#​14837](https://github.com/pnpm/pnpm/issues/14837).
On Cygwin, MSYS2, and WSL, shims still use `PATH` for Windows path conversion, so dependency executables can still redirect them there.
- GitHub Actions homepage links no longer expose server credentials. GitHub server URLs now require HTTPS, with HTTP allowed only for loopback hosts.
##### Installing packages
- pnpm no longer crashes at startup on FreeBSD and other Unix-like platforms. Platforms other than Windows and macOS use `~/.local/share/pnpm/store` by default [#​14859](https://github.com/pnpm/pnpm/issues/14859).
- `pnpm install` on Windows no longer fails with `ERR_PNPM_PACKAGE_MANAGER_REMOVE_MODULES_DIR` when clearing `node_modules` containing linked dependencies, such as when changing `nodeLinker` [#​14790](https://github.com/pnpm/pnpm/issues/14790).
- `pnpm install <pkg>` now accepts `--prod` and `--dev`, including `--prod=false` [#​14868](https://github.com/pnpm/pnpm/issues/14868).
- `pnpm install` and `pnpm update` now honor `--ignore-workspace` in nested projects excluded from the surrounding workspace. The flag also skips that workspace's settings during the `packageManager` check [#​14809](https://github.com/pnpm/pnpm/issues/14809).
- `pnpm install` on macOS no longer reuses stale files for `file:` tarball or git-hosted tarball dependencies.
- `pnpm install` in a single-project directory now detects `package.json` edits made while the previous install was finishing [#​14890](https://github.com/pnpm/pnpm/issues/14890).
- `pnpm install --frozen-lockfile` now removes packages no longer reachable from any project in `pnpm-lock.yaml`. This also prevents repeated lifecycle script execution and unnecessary installs before `pnpm run` and `pnpm exec` with `verifyDepsBeforeRun` [#​14891](https://github.com/pnpm/pnpm/issues/14891).
##### Resolving and updating dependencies
- Node.js runtime resolution now reports network failures from unofficial-builds.nodejs.org. These failures previously omitted musl builds from `pnpm-lock.yaml`, making its contents depend on network access [#​14813](https://github.com/pnpm/pnpm/issues/14813).
- `pnpm install` now rejects invalid `peerDependencies` specifiers with `ERR_PNPM_INVALID_PEER_DEPENDENCY_SPECIFICATION`. A value such as `"foo": "[email protected]"` previously created a broken directory link [#​14791](https://github.com/pnpm/pnpm/issues/14791).
- `pnpm deploy` now writes plain registry versions in the deployed `package.json`, without peer dependency suffixes. The lockfile retains peer bindings, and npm aliases retain their target package names [#​14873](https://github.com/pnpm/pnpm/issues/14873).
- `pnpm add <git repository>` now names repositories without a `package.json` as `@owner/repo`, allowing dependencies on equally named repositories from different owners [#​14870](https://github.com/pnpm/pnpm/issues/14870).
- Peer dependency resolution now deduplicates packages whose child dependency resolves an optional peer in only some workspace projects, such as `next` with `styled-jsx`'s optional `babel-plugin-macros` peer [#​14800](https://github.com/pnpm/pnpm/issues/14800).
- `pnpm update` now settles the lockfile in one run when an upgrade removes the package providing an optional peer dependency [#​14895](https://github.com/pnpm/pnpm/issues/14895).
- `pnpm update --no-save` now preserves override-applied specifiers for dependencies it is not updating, preventing subsequent frozen installs from failing with `ERR_PNPM_OUTDATED_LOCKFILE` [#​14836](https://github.com/pnpm/pnpm/issues/14836).
- `pnpm update --no-save` now succeeds under `minimumReleaseAgeStrict` when every resolved version is old enough [#​14835](https://github.com/pnpm/pnpm/issues/14835).
##### Performance
- Workspace installs and `pnpm peers check` are faster when projects depend on each other, fixing a slowdown introduced in 12.3.0. Unmet peer dependencies of workspace packages are now reported only under projects that link them directly [#​14906](https://github.com/pnpm/pnpm/issues/14906).
- Hoisted installs use less memory when packages are cached. Frozen-lockfile hoisted installs on macOS are also faster when reusable package directories are cached.
##### Python projects
- `pnpm install --frozen-lockfile` now reuses `pylock.toml` across compatible Python targets, including after kernel updates. Reuse requires unchanged requirements, index, and `requires-python`, compatible wheels, and a locked dependency graph matching the target's markers [#​14843](https://github.com/pnpm/pnpm/issues/14843).
The lockfile's `environments` marker now includes only the interpreter version and marker variables used by the dependency graph. Without `--frozen-lockfile`, pnpm warns and resolves again when the locked graph no longer matches the target.
- Python resolution no longer fails on malformed `Requires-Python` values, such as the trailing comma in `openpyxl` 3.0.x. pnpm treats these releases as declaring no interpreter range [#​14910](https://github.com/pnpm/pnpm/issues/14910).
- `pnpm add pypi:...` now rejects unsupported `--save-prefix` values before editing the manifest or resolving dependencies.
##### Workspaces and scripts
- Scripts listed in `syncInjectedDepsAfterScripts` no longer fail with `ERR_PNPM_INJECTED_DEPS_SYNC_READ_DIR` when the lockfile contains an injected package copy that no project depends on.
- `shellEmulator` now expands `${VAR}`, `${VAR:-default}`, and `${VAR:+alternative}` in scripts [#​14814](https://github.com/pnpm/pnpm/issues/14814).
- Cargo and Python project discovery now honors `!` exclusions in `pnpm-workspace.yaml` `packages`, skipping both parsing and generated source configuration for excluded projects [#​14844](https://github.com/pnpm/pnpm/issues/14844).
- `pnpm --filter "./packages/{app,lib}"` now selects either alternative. Brace alternatives can nest, span path separators, and combine with other wildcards.
- GitHub Actions updates now stop if an action reference changes during version resolution, and preserve unrelated workflow edits.
##### CLI and output
- `pn`, `pnpx`, and `pnx` now run the pnpm installed alongside them, even when that directory is absent from `PATH` or another pnpm comes first [#​14803](https://github.com/pnpm/pnpm/issues/14803).
- `pnpm --version` now reports failures to install or record a project's pinned pnpm, then prints the running CLI's version. It also honors `--store-dir` and `--store` [#​14831](https://github.com/pnpm/pnpm/issues/14831).
- `pnpm self-update` no longer reinstalls the active version when it was installed by the standalone installation script [#​14823](https://github.com/pnpm/pnpm/issues/14823).
- `pnpm t` and `pnpm tst` work again as aliases for `pnpm test`.
- `pnpm sbom` now emits valid repository URLs in CycloneDX `externalReferences[].url` and SPDX `homepage`. Shorthands such as `vercel/ms` become `git+https` URLs, embedded credentials are removed, and invalid repository values are omitted [#​14773](https://github.com/pnpm/pnpm/issues/14773).
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://notion.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/notion.svg" width="80" alt="Notion"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://latitude.so/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/latitude.svg" width="160" alt="Latitude"></a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
### [`v12.4.1`](https://github.com/pnpm/pnpm/releases/tag/v12.4.1): pnpm 12.4.1
[Compare Source](https://github.com/pnpm/pnpm/compare/v12.4.0...v12.4.1)
pnpm 12.4.1 fixes installs that failed on filesystems refusing hard links or clones, on Android, and under `nodeLinker: hoisted`. Repeat installs are faster.
#### Patch Changes
##### Installing packages
- `pnpm install` no longer fails with `Operation not permitted` when the filesystem refuses a hard link or a copy-on-write clone [#​14722](https://github.com/pnpm/pnpm/issues/14722). Under `packageImportMethod: auto` and `clone-or-copy`, pnpm copies the file instead. EdenFS checkouts, which have no hard links, and rootless containers, which refuse the clone syscall, both hit this. An explicit `packageImportMethod: hardlink` or `clone` still reports the error.
pnpm also copies a package file whose store entry has reached the filesystem's limit on names for one file, 1024 on NTFS and 65000 on ext4. Such a file failed the install under `packageImportMethod: hardlink`, and under `auto` it stopped pnpm hard linking for the rest of the install.
- `pnpm install` no longer writes a package file through a symlink left at the path it is importing to. Copying such a file overwrote whatever the link pointed at, and created that file when the link pointed nowhere. An executable package file also made the link's target executable.
- Fixed `pnpm install` and `pnpm dlx` on Android. Registry requests crashed because pnpm found no system CA certificates, so pnpm uses bundled ones there [#​14777](https://github.com/pnpm/pnpm/issues/14777). Imports also failed with "Permission denied" on filesystems that deny hard links and reflinks, and now fall back to copying [#​14780](https://github.com/pnpm/pnpm/issues/14780).
- `pnpm install` no longer fails with "Invalid cross-device link" while preserving a package's nested `node_modules` directory during a Docker build [#​14758](https://github.com/pnpm/pnpm/issues/14758).
- `pnpm install` no longer fails on a package tarball that carries a file at the archive root, such as the `._*` entries macOS `tar` adds [#​14701](https://github.com/pnpm/pnpm/issues/14701). The file is installed at the root of the package.
A `file:` tarball packed without the usual `package/` directory is now recorded under the name and version from its own `package.json`. It was recorded under the alias the dependency was given, at version 0.0.0.
- Under `nodeLinker: hoisted`, `pnpm install` no longer re-imports packages that are already in place. A repeat install replaced the whole `node_modules` tree and reported `Packages: +N`. A package is still imported when its directory is missing, when its `package.json` no longer carries the installed version, when it is a `file:` dependency, and when it is patched. Lifecycle scripts no longer run again for a package left in place, and `pnpm rebuild` and a change to `allowBuilds` still reach it.
- `pnpm install` now runs a dependency's build scripts again when its side-effects cache entry has no files to restore [#​14717](https://github.com/pnpm/pnpm/issues/14717). Such builds were skipped and nothing was put in their place, so a script whose whole effect lands outside its own package directory, such as a git hook installer, never took effect. pnpm no longer publishes empty artifacts to the shared side-effects cache either.
##### Resolving and linking dependencies
- `pnpm install`, `pnpm add`, and `pnpm dedupe` now apply `ignoredOptionalDependencies` [#​14729](https://github.com/pnpm/pnpm/issues/14729). Matching optional dependencies are left out of the lockfile and are not installed. pnpm 12 installed them whenever it resolved dependencies from scratch.
- `pnpm install` no longer links a transitive dependency to a workspace package when `linkWorkspacePackages` is `true` and the dependency is declared with a plain version range [#​14781](https://github.com/pnpm/pnpm/issues/14781). Enabling `preferWorkspacePackages` does not change this. Set `linkWorkspacePackages: deep` to link them.
- `pnpm install` no longer leaves dangling dependency links in workspace packages located above the workspace root [#​14726](https://github.com/pnpm/pnpm/issues/14726).
- `pnpm install` and `pnpm add` no longer leave a dangling symlink in `node_modules` when a project starts depending directly on a package that the lockfile holds only as a transitive dependency with resolved peer dependencies [#​14714](https://github.com/pnpm/pnpm/issues/14714).
- `pnpm dedupe` now keeps a compatible auto-installed peer when another workspace project depends on a newer major [#​14697](https://github.com/pnpm/pnpm/issues/14697). Repeated runs alternated between compatible and incompatible peer versions.
- `pnpm peers check` no longer reports a peer dependency declared as `workspace:^`, `workspace:~`, or a bare `workspace:` as unmet [#​14770](https://github.com/pnpm/pnpm/issues/14770). pnpm reported these as unmet whatever version the linked workspace project supplied.
##### Performance
- Sped up repeat installs [#​14540](https://github.com/pnpm/pnpm/issues/14540). pnpm checks the store's files only for the packages it links into `node_modules`, instead of every package in the lockfile. Creating the command shims in `node_modules/.bin` makes about 1,500 fewer filesystem calls in a 76 project workspace. Installs that use the global virtual store read their slot paths from the cache directory instead of deriving them every time. Verifying a large lockfile also allocates less memory.
- Sped up `pnpm install` in Cargo workspaces with many member crates. Repeated installs reuse verified Cargo checksum metadata.
- Installing several packages from the same Git repository and commit now downloads the source once per install [#​14725](https://github.com/pnpm/pnpm/issues/14725). Each package still runs its prepare scripts in its own copy of the checkout.
##### Running scripts and tasks
- pnpm now passes Ctrl+C on to the script or command it started and waits for it to shut down [#​14723](https://github.com/pnpm/pnpm/issues/14723). pnpm exited first, so a script that was still writing landed on the shell prompt.
- `pnpm run "/pattern/" --no-bail` now lets every matched script finish after one of them fails [#​14718](https://github.com/pnpm/pnpm/issues/14718). The command exits with `ERR_PNPM_RUN_FAILED`, and its message lists the scripts that failed in the order they were selected.
- `pnpm pipeline` no longer fails on a project that tracks a symlink, such as a `CLAUDE.md` pointing at `AGENTS.md` [#​14692](https://github.com/pnpm/pnpm/issues/14692). Changing a symlinked input's target invalidates that task's cache, and `pnpm pipeline --no-cache` no longer hashes task inputs.
##### Commands
- `pnpm add -g`, `pnpm update -g`, and `pnpm remove -g` no longer change global bins or install directories after reading only part of an installed package group [#​13796](https://github.com/pnpm/pnpm/issues/13796). If any declared package manifest is missing, malformed, or unreadable, pnpm now fails before it activates or removes anything and leaves the existing global installation intact.
- `pnpm dedupe` now processes every workspace project by default, including workspaces that keep a separate lockfile per project [#​14732](https://github.com/pnpm/pnpm/issues/14732). Workspace filters select which projects it processes, and `--fail-if-no-match` exits with an error when no project matches.
- `pnpm update <name>@​<version>` now keeps the range operator the manifest declares [#​14745](https://github.com/pnpm/pnpm/issues/14745). Running `pnpm update [email protected]` on `"react": "^19.2.8"` writes `"react": "^19.3.0"`. A `jsr:` entry keeps its `jsr:` prefix, and a plain `pnpm update` now moves a `jsr:` range the way it moves an npm range.
- `pnpm --filter` directory selectors now support `?` wildcards and character classes such as `[ab]`. A `*` or `?` wildcard no longer selects a directory whose name starts with a dot, as on pnpm 11.
- `pnpm deploy --legacy` now prefers the dependency versions pinned in the source workspace lockfile when they still satisfy the deployed project's range [#​13857](https://github.com/pnpm/pnpm/issues/13857).
- `pnpm sbom` now leaves out a package's author field when the manifest author name is empty or contains only whitespace [#​14685](https://github.com/pnpm/pnpm/issues/14685). In a filtered or split workspace run, only a project with no `author` field inherits the workspace root's author.
`pnpm sbom --sbom-format spdx` now writes `creationInfo.created` with whole seconds, such as `2026-09-08T10:38:21Z` [#​14684](https://github.com/pnpm/pnpm/issues/14684). The fractional seconds it carried were rejected by strict SPDX consumers.
##### Configuration
- The `updateConfig` pnpmfile hook now receives the resolved configuration, including settings that came from `.npmrc`, the command line, or a default [#​14676](https://github.com/pnpm/pnpm/issues/14676). Scoped registries are reported under `registriesByScope`, and a hook may rewrite that map to change where packages are fetched from. Registry credentials are reported under `configByUri`, as pnpm 11 reports them. An unset setting is left out rather than reported as `null`.
- `pnpm audit --fix` and the `minimumReleaseAgeStrict` approval prompt now keep the comments in `minimumReleaseAgeExclude` when they append an entry to it in `pnpm-workspace.yaml`. The rest of the list is left as written, and the `trustPolicyExcludePrune` and `minimumReleaseAgeExcludePrune` cleanups keep the comments of the entries they retain.
`pnpm install` and `pnpm dedupe` now run those cleanups too [#​14759](https://github.com/pnpm/pnpm/issues/14759). Only `pnpm add`, `pnpm update`, and `pnpm remove` pruned the entries that the freshly written lockfile no longer resolves.
- `pnpm config set --global node-download-mirrors` no longer rejects the key [#​13611](https://github.com/pnpm/pnpm/issues/13611). The global config file already accepted `nodeDownloadMirrors`, but the command refused to write it.
- `NO_PROXY` entries that start with a dot, such as `.npmjs.org`, now bypass the proxy for the domain and its subdomains [#​14686](https://github.com/pnpm/pnpm/issues/14686).
- pnpm no longer creates a project `pnpm-lock.yaml` when `devEngines.packageManager.onFail` is `download` and lockfile writing is off through `lockfile: false` or `--no-lockfile` [#​14728](https://github.com/pnpm/pnpm/issues/14728). pnpm still switches to the pinned version.
- pnpm now writes `node_modules/.package-map.json` only when `nodeExperimentalPackageMap` is enabled. Nothing reads the file without that setting, and an install that stops writing the map removes the one a previous install left.
##### Windows
- `pnpm pipeline` no longer fails with intermittent access denied errors when concurrent tasks save their cache entries on Windows.
- Windows filesystem operations now retry permission errors for up to one second [#​14682](https://github.com/pnpm/pnpm/issues/14682). A permanent permission error delayed the failure by a minute. Sharing and lock violations keep their one minute retry budget.
##### Messages and output
- pnpm now warns when the root `package.json` declares a non-empty `workspaces` array and the project has no `pnpm-workspace.yaml` [#​2255](https://github.com/pnpm/pnpm/issues/2255). Such an install linked no project and said nothing about why.
- `ERR_PNPM_PACKAGE_MANAGER_REMOVE_MODULES_DIR` now names the file or directory in `node_modules` that pnpm could not clean up. It reported only the underlying OS error, such as "Access is denied (os error 5)".
- `pnpm --help` no longer describes pnpm as experimental.
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://notion.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/notion.svg" width="80" alt="Notion"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://latitude.so/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/latitude.svg" width="160" alt="Latitude"></a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
</details>
<details>
<summary>eemeli/yaml (yaml)</summary>
### [`v2.9.1`](https://github.com/eemeli/yaml/releases/tag/v2.9.1)
[Compare Source](https://github.com/eemeli/yaml/compare/v2.9.0...v2.9.1)
- Limit recursive merge aliases ([#​685](https://github.com/eemeli/yaml/issues/685), [#​713](https://github.com/eemeli/yaml/issues/713))
- Simplify line unfolding during quoted string parsing ([#​714](https://github.com/eemeli/yaml/issues/714))
</details>
---
### Configuration
📅 **Schedule**: (UTC)
- Branch creation
- Between 12:00 AM and 03:59 AM, only on Monday (`* 0-3 * * 1`)
- Automerge
- At any time (no schedule defined)
🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied.
♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
👻 **Immortal**: This PR will be recreated if closed unmerged. Get [config help](https://github.com/renovatebot/renovate/discussions) if that's undesired.
---
- [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check this box
---
This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).
<!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0My4xOTEuMiIsInVwZGF0ZWRJblZlciI6IjQzLjE5MS4yIiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6W119-->
<!-- cloudflare-preview --> Preview: https://pr-554.docs-gitea-com.pages.dev
Reviewed-on: https://gitea.com/gitea/docs/pulls/554
Reviewed-by: silverwind <[email protected]>
Co-authored-by: Renovate Bot <[email protected]>
Fixes https://github.com/go-gitea/gitea/issues/39310
The docker install pages tell readers to use `:nightly` / `:nightless-rootless` for the latest development version, but those tags are no longer published — the registry only carries `main-nightly` and `main-nightly-rootless` (plus the per-release-branch `1.x-nightly` tags):
```
$ curl -s https://docker.gitea.com/v2/gitea/tags/list?n=1000 | tr ',' '\n' | grep -E '"(main-)?nightly(-rootless)?"'
"main-nightly"
"main-nightly-rootless"
```
Anyone who followed the rootless page has been pinned to an image that has not been rebuilt since 2025-08-23.
Changes:
- point the docker and rootless docker install pages at `:main-nightly` / `:main-nightly-rootless`, in English and in the zh-cn/zh-tw translations
- use `1.27-nightly` instead of the long gone `1.16-nightly` as the release branch example (the oldest one still published is `1.19-nightly`)
- fix the zh-cn/zh-tw claim that `:latest` installs a development version, and mention the nightly tags there the way the English page does
- render `@dockerVersion@` as `main-nightly` for the next version, so the compose snippets on the unreleased docs no longer reference `gitea:nightly`
Not touched: the 1.22–1.26 versioned docs carry the same stale tag, but those versions are already out of maintenance — happy to include them if preferred.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
<!-- cloudflare-preview --> Preview: https://pr-550.docs-gitea-com.pages.dev
---------
Co-authored-by: bircni <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/550
Reviewed-by: silverwind <[email protected]>
Co-authored-by: Lunny Xiao <[email protected]>
Several cron tasks in the cheat sheet listed `ENABLE_SUCCESS_NOTICE` or `NO_SUCCESS_NOTICE`. Gitea reads neither, only `NOTICE_ON_SUCCESS`. `NO_SUCCESS_NOTICE` was removed in https://github.com/go-gitea/gitea/pull/19221.
1. Replace both dead keys with `NOTICE_ON_SUCCESS` (default `false`).
2. Fix "Notice if not success" descriptions, which inverted the key's meaning.
3. List the key for `cleanup_hook_task_table`, `cleanup_actions`, `cleanup_action_runs` and `gc_lfs`, which honor it but omitted it.
Matches https://github.com/go-gitea/gitea/pull/39352.
Written by Claude.
Reviewed-on: https://gitea.com/gitea/docs/pulls/552
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: silverwind <[email protected]>
Two unrelated but small fixes, closes#258 and closes#32.
**The reference does not name its types (#258)**
Since the api reference is rendered from the openapi 3.0 document, a field of type `ReviewStateType` is rendered as the bare string enum the `$ref` points at: starlight-openapi dereferences everything and the schema has no `title`, so the name of the type is nowhere on the page. Searching the current reference for `ReviewStateType` returns nothing, which is what the issue reports (the plugin also has no route for `components/schemas`, so there is no page per type either).
`normalizeSchema()` now titles every named schema of an openapi 3.0 document with its own name, which the plugin renders next to the type of a field:
- before: `state` · `string` · Allowed values: APPROVED …
- after: `state` · `ReviewStateType` · `string` · Allowed values: APPROVED …
The go doc comments start with the name of the type, so the description would repeat it ("ChangeFilesOptions — ChangeFilesOptions options for creating …"); a leading `<name> ` is stripped from the description.
For the record, how the three generations of documents describe that type:
| Versions | Type name | Allowed values |
| --- | --- | --- |
| 1.22 – 1.25 (swagger, `ReviewStateType` definition without `enum`) | shown | missing |
| 1.26 – 1.27 (swagger, enum inlined upstream) | missing | shown |
| 1.27, next (openapi 3.0, with this change) | shown | shown |
The released swagger documents are frozen, so 1.22 – 1.26 keep what they have; from 1.27 on the reference answers the question the issue asks.
**Gmail needs an app password (#32)**
The configuration in the guide is still correct after Google turned off password-only access, because it already asks for an app password. What was missing is why there is no alternative: Gitea's mailer authenticates with `CRAM-MD5`, `PLAIN` or `LOGIN` and has no XOAUTH2, so "Sign in with Google" is not an option. Added that, together with the two things people run into — an app password needs 2-step verification and a Workspace administrator can disable app passwords for the domain — and the `587` + `smtp+starttls` variant.
Verified by building the api and the next docs: the type names show up on the operation pages, the descriptions are not duplicated, and 1.26 and older are unchanged. `pnpm check` passes.
<!-- cloudflare-preview --> Preview: https://pr-541.docs-gitea-com.pages.dev
---------
Co-authored-by: bircni <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/541
Reviewed-by: bircni <[email protected]>
Co-authored-by: Lunny Xiao <[email protected]>
Closes#243.
The `sha` of the contents API is compared against the ID of the blob the commit has at that path, and for a file tracked by Git LFS that blob is the pointer file. Sending the `lfs_oid` or a checksum of the file content fails with `sha does not match`, and nothing in the documentation said so.
Adds a section to the LFS page: use the `sha` from `GET /repos/{owner}/{repo}/contents/{filepath}`, not `lfs_oid`, and note that the API does turn the content into an LFS object by itself when `.gitattributes` tracks the path (`modifyFile()` stores the object and commits the pointer), so nothing else has to be done differently.
The swagger comments upstream are being fixed in https://github.com/go-gitea/gitea/pull/38915, which will reach `/api/next/` through the scheduled job and the released reference with 1.28. The documents of 1.22 – 1.27 are frozen, so the guide is the only place where readers of the current release can find it.
<!-- cloudflare-preview --> Preview: https://pr-542.docs-gitea-com.pages.dev
Reviewed-on: https://gitea.com/gitea/docs/pulls/542
Reviewed-by: bircni <[email protected]>
Co-authored-by: Lunny Xiao <[email protected]>
Automated update of static/swagger-latest.json and static/openapi3-latest.json from the gitea main branch, opened by the "update api spec files" scheduled workflow.
Reviewed-on: https://gitea.com/gitea/docs/pulls/531
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Gitea Bot <[email protected]>
## Summary
Adds Sealos to the cloud-provider installation page for the current and next
Gitea documentation.
## Why
Sealos provides a public, community-maintained one-click Gitea template. The
entry fits alongside the existing Cloudron, Linode, alwaysdata, and Exoscale
options.
## Proposed files
- `docs/installation/on-cloud-provider.md`
- `versioned_docs/version-1.27/installation/on-cloud-provider.md`
## Validation
- Sealos App Store: https://sealos.io/products/app-store/gitea
- Template source: https://github.com/labring-actions/templates/tree/kb-0.9/template/gitea
- Gitea image: `docker.gitea.com/gitea:1.27.0-rootless`
- Live deployment validated on 2026-07-24
- Verified administrator initialization, login, repository creation, commits,
issues, release uploads, persistence, and S3 upload, read, and delete flows
## Maintenance
I maintain the Sealos template and can update the documentation when its
deployment path changes. Deployment-specific support is handled in the
[Sealos templates repository](https://github.com/labring-actions/templates/issues).
## AI assistance
This proposal was prepared with OpenAI Codex assistance. I reviewed the
deployment evidence and proposed content and will personally handle review
feedback.
---------
Co-authored-by: Lunny Xiao <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/525
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: yangchuansheng <[email protected]>
Automated update of the generated runner reference pages from gitea/runner (main for the develop docs, the newest stable tag for every documented series), opened by the "update runner reference" scheduled workflow.
Reviewed-on: https://gitea.com/gitea/docs/pulls/528
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Gitea Bot <[email protected]>
Automated update of static/swagger-latest.json and static/openapi3-latest.json from the gitea main branch, opened by the "update api spec files" scheduled workflow.
Reviewed-on: https://gitea.com/gitea/docs/pulls/522
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Gitea Bot <[email protected]>
The docs site is served by Cloudflare Pages now, so the AWS deployment path is redundant.
- Drop the AWS CLI installation step.
- Drop the `configure-aws-credentials` step and the `aws s3 sync` + CloudFront invalidation step.
- Keep only the Cloudflare Pages deploy, and simplify the `_headers` / `_redirects` comment since it no longer needs to explain S3 ordering.
The AWS_* secrets are no longer used by this workflow and can be removed from the repository settings.
Reviewed-on: https://gitea.com/gitea/docs/pulls/518
Reviewed-by: silverwind <[email protected]>
Automated update of static/swagger-latest.json and static/openapi3-latest.json from the gitea main branch, opened by the "update api spec files" scheduled workflow.
Reviewed-on: https://gitea.com/gitea/docs/pulls/519
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Gitea Bot <[email protected]>
The `checks` workflow deploys a Cloudflare Pages preview to the `pr-<number>` branch of the `docs-gitea-com` project for every pull request, but nothing ever removed it. Closed and merged pull requests therefore keep an alias and a growing list of deployments around.
This adds a `cleanup preview` workflow that runs when a pull request is closed (merged or not), lists the deployments of the project through the Cloudflare API and deletes the ones whose trigger branch is `pr-<number>`. `force=true` is used so the deployment the branch alias points at is removed as well.
Pull requests from forks do not receive the secrets, so the cleanup is skipped there, exactly like the preview deployment itself.
---------
Co-authored-by: bircni <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/517
Reviewed-by: bircni <[email protected]>
Co-authored-by: Lunny Xiao <[email protected]>
Five documentation fixes that do not depend on anything upstream.
**PostgreSQL version in the docker guide** (closes#224)
The compose example pinned `postgres:14`, the oldest version Gitea still tests against, and PostgreSQL does not upgrade its data directory on its own, so every new deployment that copies the example is one major upgrade behind from day one. The example now uses `postgres:18`, with a note about why the major version is pinned and a pointer to the supported range.
**SELinux** (closes#211)
`setcap 'cap_net_bind_service=+ep'` is what the docs suggest for binding to a port below 1024, but it is denied under SELinux depending on the label of the binary, and it is lost on every upgrade because the capability lives in an extended attribute of the file. A new page describes granting the capability through the systemd unit instead (`AmbientCapabilities`, plus `PrivateUsers=false` when the sandboxing gets in the way), the `semanage fcontext` / `restorecon` labels for the binary and the data directory, and how to read the denials with `ausearch` and `audit2allow`. It is linked from the binary and source installation pages and from the `gitea web` notes.
**OAuth2 and OpenID Connect** (closes#106)
The authentication page documented LDAP, PAM, SMTP, FreeIPA, SPNEGO and the reverse proxy, but not that Gitea can log users in against an external OAuth2 or OpenID Connect provider, which made it look like the only OAuth2 support is Gitea acting as the provider. The new section covers the fields of the authentication source, the callback URL to register at the provider, the group and claim mappings, and the `[oauth2_client]` settings, and separates it from the older OpenID 2.0 sign-in.
**IIS reverse proxy** (closes#39)
Application Request Routing rewrites the host of the `Location` header back to the internal one while "Reverse rewrite host in response headers" is on, which sends the browser to `127.0.0.1:3000` after a sign-in. The IIS section now says to turn it off and where the setting lives.
**ACME in containers** (closes#292)
`ACME_DIRECTORY` defaults to the relative path `https`, which is resolved against the working directory of the process. In the official image that is `/app/gitea`, outside the `/data` volume, so a recreated container asks the CA for a new certificate and runs into the Let's Encrypt limit of 5 certificates per week for the same domains. The HTTPS page now warns about it and shows an absolute path inside the volume; the cheat sheet entry points at the warning. The default itself is being changed in https://github.com/go-gitea/gitea/pull/35851, the note stays useful for every released version.
Only `docs/` is touched, the released versions keep their content.
<!-- cloudflare-preview --> Preview: https://pr-513.docs-gitea-com.pages.dev
Reviewed-on: https://gitea.com/gitea/docs/pulls/513
Reviewed-by: bircni <[email protected]>
Co-authored-by: Lunny Xiao <[email protected]>
Closes#37.
Gitea generates an OpenAPI 3.0 document since 1.27 (`templates/swagger/v1_openapi3_json.tmpl`, renamed to `templates/swagger/v1-openapi3.generated.json` on main, served by an instance at `/openapi3.v1.json`), but this site never picked it up and still published Swagger 2.0 only. That is what the issue is about: code generators such as `openapi-python-client` reject the document with "You may be trying to use a Swagger document; this is not supported by this project".
**What this does**
- `update_api_docs.sh` downloads the OpenAPI 3.0 document next to the Swagger 2.0 one, rewrites the same placeholders (they sit in `servers` instead of `basePath`) and writes `static/openapi3-latest.json` and `static/openapi3-<minor>.json`. Versions that do not have one (1.26 and older) are skipped with a message and keep their Swagger 2.0 document.
- The OpenAPI 3.0 document is rendered where it exists (`/api/` and `/api/next/`), the older versions keep rendering Swagger 2.0. Both documents describe the same api and carry the same operation ids, so no url of an operation page changes; the `basePath` normalization is now only applied to the Swagger 2.0 documents, since a full url in `servers` is valid.
- The overview page of each version links both documents for download, which is what the issue actually needs: `https://docs.gitea.com/openapi3-27.json` for the current release, `https://docs.gitea.com/openapi3-latest.json` for gitea main.
- The scheduled workflow now refreshes and proposes both documents of gitea main.
**Verification**
- `pnpm dlx @redocly/cli lint static/openapi3-latest.json`: valid, only content warnings that come from upstream (missing 4xx responses, ambiguous paths).
- `openapi-python-client generate --path static/openapi3-latest.json`, the tool from the issue, generates a client. The only warning left is `POST /markdown/raw`, whose request body is `text/plain`.
- Operation ids of `static/swagger-27.json` and `static/openapi3-27.json` are identical (481 on both sides), so the operation pages keep their urls.
- Built the api product and compared a rendered operation page against the Swagger 2.0 one: same sections, the request body is now marked required where the document says so.
The released Swagger 2.0 documents are kept: they are the only description 1.26 and older ever produced, and some tooling still expects them.
<!-- cloudflare-preview --> Preview: https://pr-514.docs-gitea-com.pages.dev
Reviewed-on: https://gitea.com/gitea/docs/pulls/514
Reviewed-by: bircni <[email protected]>
Closes#146.
The published site runs pagefind, not docsearch: `PUBLIC_DOCSEARCH_APP_ID` and `PUBLIC_DOCSEARCH_API_KEY` are not set, so `astro.config.mjs` falls back to `pagefind: true`, and a page of docs.gitea.com ships starlight's `#starlight__search` with the `/pagefind/` bundle. The version scoping we have is only implemented for docsearch (`docsearch:product` / `docsearch:version` meta tags plus `facetFilters` in `src/config/docsearch.ts`), so nothing consumes it today.
Pagefind builds one index for the whole site, partitioned only by the html language, and nothing tags the pages with a version, so a search started in the 1.23 docs answers with every version of the same page and the latest one usually wins. That is what the issue reports; the original cause (the docusaurus search plugin) is gone, this one replaced it.
**What this does**
- `src/components/MarkdownContent.astro` wraps the content in two `data-pagefind-filter` elements, `product` and `version`, so the filters land in the index. One filter per element on purpose: pagefind reads the whole attribute as a single `name:value` pair, a comma separated value ends up as one filter named `product` with the value `docs,version:1.26`.
- The same scope is written to a `gitea:search-filters` meta tag next to the docsearch ones (`src/lib/search.ts`).
- Starlight builds the search ui from a build time configuration and has no option for per page filters, so its `@pagefind/default-ui` import is redirected to a subclass in `src/lib/pagefind-ui.ts` by the `gitea-pagefind-filters` plugin in `astro.config.mjs`, the same approach the `gitea-openapi-overview` plugin already uses. The subclass selects the filters of the page the modal was opened on.
- A "Search all versions" checkbox below the search input drops the filters again and labels each result with the version (and the product for the api and the runner) it comes from, so the duplicates are distinguishable. Translated into 简体中文 and 繁體中文.
- Pagefind's own filter panel is hidden, it is redundant next to the checkbox and starlight does not style it. The checkbox gets its native rendering back, pagefind resets everything it renders with `all: unset`.
- A `gitea-pagefind-filters-check` integration fails the build if the redirect did not run, so a starlight upgrade that moves the import cannot silently bring back an unscoped search.
`@pagefind/default-ui` is added to `sites/docs/package.json`; it was only reachable as a transitive dependency of starlight, and the version is the one already in the lockfile.
**Verification**
Built the whole site and drove the search with a headless browser against `astro preview` (the search does not run in `pnpm dev`):
| Page | Query | Results |
| --- | --- | --- |
| `/1.23/usage/actions/comparison/` | email | only `/1.23/...` |
| `/usage/actions/overview/` | email | only the version served at the root |
| `/runner/registration/` | label | only `/runner/...` |
| `/api/operations/tags/issue/` | label | only `/api/...` |
| `/zh-cn/1.26/usage/actions/overview/` | email | only `/zh-cn/1.26/...` |
| `/1.23/...` with the checkbox ticked | email | every version, results labelled `... · API 1.25`, `... · next` |
`pnpm check` passes. The docsearch path is unchanged: when the credentials are set the plugin takes over, the redirect does not run and the check integration skips.
<!-- cloudflare-preview --> Preview: https://pr-512.docs-gitea-com.pages.dev
Reviewed-on: https://gitea.com/gitea/docs/pulls/512
## Problem
`https://docs.gitea.com/help/seek-help` returns 404. It is still linked as the
*Support* item of the blog.gitea.com navbar, and it is not the only url in that
state: `/usage/webhooks`, `/usage/agit`, `/usage/protected-tags` and
`/usage/repo-mirror` are linked from gitea itself and 404 as well.
## Root cause
Not a rendering problem of the new site: those pages were moved years ago.
`docs/help/seek-help.md` became `docs/help/support.md` in 2023 (`slug: support`),
and the flat `/usage/<page>`, `/advanced/<page>`, `/developers/<page>` layout of
1.19 and 1.20 was reorganized into `usage/repository/`, `administration/` and
`development/`.
The old urls kept working because the site was published to S3 with
`aws s3 sync` **without** `--delete`: every page built before a move stayed in
the bucket and CloudFront kept serving it. The archived copy of
`/help/seek-help` still carries `last-modified: Wed, 12 Jul 2023` - the file was
six years stale but reachable.
The Astro deployment publishes to Cloudflare Pages, which only serves what the
current build emits, so all of those leftovers disappeared at once.
I crawled the urls the Wayback Machine has for `docs.gitea.com` and checked them
against the live site: 335 non-removed-version urls 404 today, and they are all
of this kind.
## Fix
301s in `cloudflare/_redirects`, from every legacy url that has a current
equivalent to that page, generated for the root, the `zh-cn` and `zh-tw`
prefixes and the `next` (develop) version. Also the runner docs that were
published under a full version number (`/runner/0.2.11/`, `/runner/1.0.8/`) and
`/runner/next/`, which is `/runner/develop/` now.
Checked while generating them, against a full local build:
- no rule shadows a page that exists today - the zh translations of `usage/agit`
and `usage/profile-readme` were never moved, so they are still served at the
old path and are left alone
- every target resolves to a page that exists, in the language and version the
rule redirects within
- no duplicated source, 417 static and 10 dynamic rules, well inside the
Cloudflare Pages limits (2100 / 100)
## Out of scope
- `/contributing/*`, `/developers/guidelines-*` and `/developers/hacking-on-gitea`
404 too, but those pages moved to the gitea repository, there is nothing on
docs.gitea.com to point them at.
- The `aliases:` frontmatter of the sources (`/en-us/seek-help`, `/webhooks`,
...) is a leftover of the Hugo site. Docusaurus ignored it and so does the
Astro build; those urls have been 404 for years and are not resurrected here.
- The blog.gitea.com navbar should link to `/help/support` directly, that needs a
pull request against gitea/blog.
Reviewed-on: https://gitea.com/gitea/docs/pulls/507
Reviewed-by: bircni <[email protected]>
Co-authored-by: Lunny Xiao <[email protected]>
Replace [#307](https://gitea.com/gitea/docs/issues/307)
Closes [#237](https://gitea.com/gitea/docs/issues/237)
## Summary
Rebuilds docs.gitea.com with [Astro](https://astro.build) and [Starlight](https://starlight.astro.build), replacing Docusaurus. Every
published url keeps working and no content file was moved: the site is a new rendering layer over the existing `docs/`, `versioned_docs/`, `i18n/`, `runner-docs/` and `static/swagger-*.json` trees.
For the first time the API reference is part of the site rather than a Redoc
bundle: all seven swagger documents are rendered into real pages, one per
operation, so they are linkable, crawlable and searchable.
## Why
- the Docusaurus build needs an 8 GB heap and about 2 minutes for 2478 pages;
the Astro build produces 5751 pages, API reference included, in about 100
seconds
- the API reference was a single client rendered Redoc page per version, absent
from the site search and from search engines
- sidebars, version lists and language lists were configured in three different
places and drifted apart
## What is in here
- **`packages/content-loader`** — the product × version × language matrix
(`products.ts`) and an Astro content loader that reads the existing markdown
trees directly. The matrix is the single source of truth: content loading,
sidebars, the version and language pickers, the version banner, the search
facets and the API schemas are all derived from it.
- **Content compatibility** — frontmatter `slug`, `sidebar_position` and
`sidebar_label` keep working, `_category_.json` still drives sidebar labels,
order and the generated category pages, `@version@` style release variables are
still substituted, `:::note` admonitions become Starlight asides and the 2778
relative `*.md` links are rewritten to urls while the site is built.
- **API reference** — `starlight-openapi` renders the seven swagger documents.
Operation urls are hyphenated (`/api/operations/list-admin-workflow-jobs/`),
the sidebar shows the operations of the version being read with a coloured
HTTP method badge, and the overview page links to one page per tag instead of
repeating all 484 operations.
- **Navigation** — a version picker that follows the reader to the same page in
another version, a language picker that only offers the languages the current
product is published in, and product links for Docs, API, Runner and
Enterprise.
- **Theme** — the light and dark palettes of about.gitea.com, mapped onto the
Starlight variables.
- **Search** — Pagefind by default; Algolia DocSearch takes over when
`PUBLIC_DOCSEARCH_APP_ID` and `PUBLIC_DOCSEARCH_API_KEY` are set. Every page
carries `docsearch:product`, `docsearch:version` and `docsearch:language` meta
tags, so a search stays inside what is being read.
`cloudflare/docsearch-crawler.json` holds the crawler configuration.
## Bugs fixed along the way
- `GET /user/applications/oauth2` and `GET /user/applications/oauth2/{id}` have
operation ids that only differ in case, so they collapsed onto the same url
and one of the two pages was silently dropped. They are now
`user-get-oauth2-application` and `user-get-oauth2-application-by-id`.
- The API overview and every tag page were all titled "Overview".
## Url compatibility
The routes of both builds were compared page by page during the migration with
`scripts/url-diff.mjs`:
```
identical: 2368, missing: 0, accepted: 110, added: 3292
```
The 110 accepted ones are all redirected in `cloudflare/_redirects`: the
localized copies of the English-only API and Runner docs, the Docusaurus search
page, and `/1.27/`, `/runner/3/`, `/api/1.27/` which are aliases of the versions
served at the product root. The added ones are the API operation and tag pages
plus the routes the Starlight language fallback serves in English when a
translation is missing — Docusaurus answered those with a 404.
## Workflows
- `checks` builds the site and type checks it on every pull request
- `Build and Publish Docs site` publishes `sites/docs/dist` to S3/CloudFront and
to Cloudflare Pages, and copies `cloudflare/_headers` and
`cloudflare/_redirects` into the deployment
- `update swagger files` and `update runner reference` are unchanged, they only
touch content
- `make cut-version PRODUCT=docs VERSION=1.28` replaces
`docusaurus docs:version`
## Removed
`docusaurus.config.js`, the swizzled theme under `src/`, and the Docusaurus UI
translations (`i18n/*/code.json`, `i18n/*/docusaurus-theme-classic/`). The
strings those carried — the product names, the footer column titles and the
outdated translation notice — were ported to
`sites/docs/src/config/strings.ts`; Starlight ships the rest of its interface in
both Chinese locales. The documentation content itself is untouched.
## Follow ups
- nine relative links are broken in the sources and reported by every build, the
same ones Docusaurus warned about; `GITEA_DOCS_STRICT_LINKS=true` turns them
into an error once they are fixed
- `cloudflare/worker.js` has to be deployed for `/enterprise/` to keep resolving
- the Algolia index has to be created and crawled before the search credentials
are set
## Testing
```shell
make install
make serve-fast # english, the version served at the root
make serve # the whole matrix
make build # 5751 pages
make check # 0 errors
make serve-built # build and serve, the only way to try the search locally
```
## Screenshots
<img width="1371" alt="image.png" src="attachments/69acdd77-cc89-4635-a8bd-1163a34afa86">
<img width="1810" alt="image.png" src="attachments/76212416-4825-4b4d-bf59-3e549900c96f">
<img width="1789" alt="image.png" src="attachments/7be84cda-5abe-48bd-8839-7ae1ee7806e7">
---------
Co-authored-by: bircni <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/496
Reviewed-by: bircni <[email protected]>
This PR contains the following updates:
| Package | Change | [Age](https://docs.renovatebot.com/merge-confidence/) | [Confidence](https://docs.renovatebot.com/merge-confidence/) | Type | Update | Pending |
|---|---|---|---|---|---|---|
| [@easyops-cn/docusaurus-search-local](https://github.com/easyops-cn/docusaurus-search-local) ([source](https://github.com/easyops-cn/docusaurus-search-local/tree/HEAD/packages/docusaurus-search-local)) | [`0.55.2` → `0.55.3`](https://renovatebot.com/diffs/npm/@easyops-cn%2fdocusaurus-search-local/0.55.2/0.55.3) |  |  | dependencies | patch | |
| [pnpm](https://pnpm.io) ([source](https://github.com/pnpm/pnpm/tree/HEAD/pnpm11/pnpm)) | [`11.17.0` → `11.20.0`](https://renovatebot.com/diffs/npm/pnpm/11.17.0/11.20.0) |  |  | packageManager | minor | `11.21.0` |
| [pnpm/action-setup](https://github.com/pnpm/action-setup) ([changelog](https://github.com/pnpm/action-setup/compare/0ebf47130e4866e96fce0953f49152a61190b271..0977fd99725f1db4007ccb2928dbb4e90d06cc86)) | `0ebf471` → `0977fd9` | | | action | digest | |
---
### Release Notes
<details>
<summary>easyops-cn/docusaurus-search-local (@​easyops-cn/docusaurus-search-local)</summary>
### [`v0.55.3`](https://github.com/easyops-cn/docusaurus-search-local/releases/tag/v0.55.3)
[Compare Source](https://github.com/easyops-cn/docusaurus-search-local/compare/v0.55.2...v0.55.3)
##### Bug Fixes
- trigger a release for refining search icon for [#​577](https://github.com/easyops-cn/docusaurus-search-local/issues/577) ([9a7eb91](https://github.com/easyops-cn/docusaurus-search-local/commit/9a7eb91393f389838cc31f45e4f4737dc4bb325e))
</details>
<details>
<summary>pnpm/pnpm (pnpm)</summary>
### [`v11.20.0`](https://github.com/pnpm/pnpm/releases/tag/v11.20.0): pnpm 11.20
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.19.0...v11.20.0)
#### Minor Changes
- **Security fix.** Affects projects using `namedRegistries` on pnpm 11.1.0–11.19.x. It is **semi-breaking** for those projects — see "If you use named registries" below.
The lockfile recorded no marker for which registry a package came from. Packages were keyed by `name@version` alone, and entry lookup went through `refToRelative(ref, name)`, so a dependency you declared against one registry could be satisfied by an entry that was actually resolved from another. When two registries served the same name and version, both collapsed onto a single `packages:` entry and whichever resolved first decided the tarball every consumer got.
That is a package-substitution risk: a package you expect from your private registry could be installed from a different registry that publishes the same name and version, and the lockfile recorded nothing that would let you tell.
Packages resolved from a named registry are now recorded under registry-qualified keys (`<name>@​<registryName>:<version>`, e.g. `foo@work:1.0.0`), so each registry gets its own entry and the lockfile pins which one a dependency came from.
The lockfile format version is unchanged. Registry-qualified keys appear only for packages resolved from a named registry, so a project that does not use `namedRegistries` sees no difference, and older pnpm versions keep reading the file.
##### If you use named registries
Your next non-frozen install re-keys those entries, which shows up as a lockfile diff. Commit it — that diff is the fix being applied. Review it: an entry that moves to a registry you did not expect is worth investigating.
Everyone working on the project should be on this version or newer before you do. An older pnpm reads the re-keyed lockfile fine — frozen installs are unaffected — but it does not produce registry-qualified keys itself, so any install that updates the lockfile writes those entries back to the old shape, and the next install on a current pnpm re-qualifies them. The result is a lockfile that flips back and forth, and while it is in the old shape the project is exposed again. Because the lockfile format version is deliberately unchanged, pnpm cannot detect this and warn you about it.
There is no setting to keep the old behavior: the old shape is the vulnerability.
Tarball URLs that follow the standard registry layout are no longer written to the lockfile for named-registry packages; they are recomputed from the `namedRegistries` setting on demand.
To use named registries, map your aliases in `pnpm-workspace.yaml`:
```yaml
namedRegistries:
work: https://npm.enterprise.example.com/
```
##### New built-in `npmjs:` alias
`npmjs:` now resolves to `https://registry.npmjs.org/` with no configuration, alongside the existing `gh:` alias for GitHub Packages. It pins a dependency to the public registry even when `registry` points elsewhere, such as an internal proxy:
```json
{ "dependencies": { "left-pad": "npmjs:^1.3.0" } }
```
`npm:` cannot do this — it is the alias protocol (`npm:<name>@​<range>`) and resolves through whatever `registry` points at.
**If you mirror or proxy npmjs, point the alias at your mirror:**
```yaml
namedRegistries:
npmjs: https://npm.internal.example.com/
```
Built-in registry URLs are also the prefixes a lockfile's recorded tarball URL is matched against when pnpm verifies a package. Without the override, an entry whose tarball URL is on `registry.npmjs.org` is verified against the public registry rather than your mirror. This only affects lockfiles that record such URLs — a canonical URL for your configured registry is omitted from the lockfile and unaffected — and only when a tarball-URL, `minimumReleaseAge`, or `trustPolicy` check runs. Overriding the alias is the same escape hatch GHES users already have for `gh`.
Every alias the lockfile references must stay in `namedRegistries`: reading an entry whose alias is gone fails with `ERR_PNPM_MISSING_NAMED_REGISTRY` rather than silently falling back to the default registry, since that would fetch a different package. Renaming an alias re-resolves the packages that used it.
Named registry aliases that shadow a reserved dependency specifier prefix (`file`, `link`, `workspace`, `runtime`, `npm`, `jsr`, ...) are now rejected with `ERR_PNPM_RESERVED_NAMED_REGISTRY_NAME` instead of being silently shadowed by the corresponding resolver.
`pnpm licenses` and `pnpm sbom` now keep the two artifacts apart as well: license records carry the registry alias, and SBOM components carry the purl `repository_url` qualifier.
#### Patch Changes
- An empty `http-proxy`, `https-proxy`, `proxy`, or `no-proxy` value — from the `.npmrc`, `pnpm-workspace.yaml`, the CLI, or the `HTTP_PROXY` / `HTTPS_PROXY` / `PROXY` / `NO_PROXY` environment variables — no longer fails the install with `ERR_PNPM_INVALID_PROXY`. Empty settings read as unset, so a shell exporting `HTTP_PROXY=` disables the proxy, and an empty `proxy=` in the `.npmrc` no longer suppresses `HTTPS_PROXY` [#​13533](https://github.com/pnpm/pnpm/issues/13533).
`proxy=false` in the `.npmrc` or `proxy: false` in `pnpm-workspace.yaml` now turns proxying off instead of being read as a proxy host named `false`. `false` and `null` on `https-proxy` / `http-proxy` / `no-proxy` read as unset, and on the command line they are ordinary host names, since a flag carries its value verbatim.
- The env lockfile no longer pins `@pnpm/exe` alongside `pnpm` when the wanted pnpm version is 12 or newer. From v12 the unscoped `pnpm` package is itself the native executable, so `@pnpm/exe` is not published for it and resolving it would fail. The engine identity check now verifies the native binary through whichever package ships it.
- `lexCompare` and `nerfDart` are now published as `@pnpm/text.ordinal-comparator` and `@pnpm/config.registry-auth-key`. Use these instead of `@pnpm/util.lex-comparator` and `@pnpm/config.nerf-dart`.
- Fixed the order in which pnpm matches a lockfile's recorded tarball URL against known registry URLs. Two registry URLs of equal length were previously ordered arbitrarily, so which one a tarball URL matched could differ between runs.
- Dependency resolution is faster: package metadata is now filtered once per packument instead of once per dependency edge when `minimumReleaseAge` is active, and parsed semver versions and ranges are reused instead of re-parsed on every comparison.
- Security: `pnpm rebuild` now refuses a lockfile whose `packages` key carries a path traversal in the package name (e.g. `../../../[email protected]`), instead of running that package's lifecycle scripts and linking its bins in a directory outside the virtual store. Such a name is rejected with `ERR_PNPM_INVALID_DEPENDENCY_NAME`.
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
### [`v11.19.0`](https://github.com/pnpm/pnpm/releases/tag/v11.19.0): pnpm 11.19
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.18.0...v11.19.0)
#### Minor Changes
- `pnpm login` no longer requires an interactive terminal when the registry supports web-based login: without a TTY it prints the authentication URL (skipping the QR code and the "Press ENTER to open the URL in your browser" prompt) and polls the registry until the browser approval completes. Only the classic username/password login still fails with `ERR_PNPM_LOGIN_NON_INTERACTIVE` in a non-interactive terminal.
- The `save-prefix` setting now accepts `=`: newly added dependencies are saved with an explicit `=` operator (`=1.2.3`) instead of the setting being silently treated as the default `^`.
#### Patch Changes
- `allowBuilds` entries can now approve git-hosted packages that pnpm downloads as a tarball, such as `github:` dependencies (which are fetched from `codeload.github.com` rather than cloned), by their repository URL without the resolved commit hash. This matches the hashless `git+` matching already supported for cloned git dependencies. For example:
```yaml
allowBuilds:
"foo@git+https://github.com/org/foo.git": true
```
This approves the package whether pnpm clones it or downloads a tarball, so the entry no longer has to be updated every time the pinned commit changes. GitLab and Bitbucket tarball downloads are matched the same way. Approving or denying a specific resolved commit by its full tarball dep path continues to work.
- `pnpm outdated --include-github-actions` no longer blocks on an interactive git credential prompt when a workflow uses a private action repo.
- Prevented `minimumReleaseAge` from replacing `latest` with a SemVer-greater version than the registry tag target [#​13034](https://github.com/pnpm/pnpm/issues/13034).
- Fixed empty `bundledDependencies` and `bundleDependencies` arrays causing nondeterministic lockfile changes. See [#​13123](https://github.com/pnpm/pnpm/issues/13123).
- The install summary no longer prints `(X is available)` when the registry's `dist-tags.latest` is still held back by the active `minimumReleaseAge` policy. The hint only ever names the actual latest tag, so an immature latest suppresses the hint instead of advertising the version pnpm just refused to install [#​11698](https://github.com/pnpm/pnpm/issues/11698).
- `pnpm update` keeps the explicit `=` operator of an exact version pin: a dependency saved as `=3.5.1` now updates to `=3.5.2` instead of the bare `3.5.2`. See [#​13168](https://github.com/pnpm/pnpm/issues/13168).
- Preserve a workspace dependency's `link:` entry when a run does not target it — e.g. `pnpm update <other-pkg>` (with or without `--recursive`), or a plain install after a root/catalog dependency change — with `injectWorkspacePackages`, instead of spuriously rewriting it to a peer-suffixed `file:` protocol. See [#​10433](https://github.com/pnpm/pnpm/issues/10433).
- Workspace dependencies declared with a relative path (e.g. `"foo": "workspace:../foo"`) are no longer silently dropped from the workspace projects graph, so `--filter` selection and the topological order of recursive commands take them into account.
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
### [`v11.18.0`](https://github.com/pnpm/pnpm/releases/tag/v11.18.0): pnpm 11.18
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.17.0...v11.18.0)
#### Minor Changes
- Fixed an installed optional dependency being left without one of its own required dependencies. When a package reached through `optionalDependencies` is installable on the current system but one of its regular `dependencies` is not, a lockfile-based install skipped that dependency and installed the parent anyway, so importing the parent failed with `MODULE_NOT_FOUND`. The dependency is now installed, and an install-check warning reports the incompatibility. A dependency is still only skipped when every path to it is optional, or when the package that pulls it in was itself skipped [#​13286](https://github.com/pnpm/pnpm/issues/13286).
- `pnpm setup` now appends `PNPM_HOME` and the global bin directory to the GitHub Actions environment files (`GITHUB_ENV` and `GITHUB_PATH`), so later steps in the same job can run `pnpm add --global` and other global commands [#​9191](https://github.com/pnpm/pnpm/issues/9191).
- Added support for `publishConfig.name`, which publishes a package under a different name than the one its manifest carries in the workspace. It is for a project whose published name is already taken by a sibling project, which otherwise has to be renamed by a build step just before publishing. Only the published artifact is renamed — dependents, `pnpm-lock.yaml`, and release tooling keep addressing the project by its manifest name — and the new name reaches the packed manifest, the tarball filename, and everything that addresses the package at the registry: the already-published check of `pnpm publish -r`, its registry selection, and the release-planning probes of `pnpm change status` and `pnpm version -r` [#​13345](https://github.com/pnpm/pnpm/issues/13345).
- `pnpm self-update` no longer takes any instruction from the project it is run in:
- pnpm is fetched through the same trusted registry and auth configuration used when switching pnpm versions, so a project `.npmrc` or `pnpm-workspace.yaml` can no longer redirect the download or attach credentials to it, and the project's default `.pnpmfile.(c|m)js` is no longer loaded. Pnpmfiles from trusted sources (the `pnpmfile` setting, the global pnpmfile, config dependencies) still apply.
- The `minimumReleaseAge` settings in `pnpm-workspace.yaml` no longer affect `self-update`. They still govern the project's own dependencies; for `self-update` the cooldown now comes from the built-in default, your global config, a `PNPM_CONFIG_*` environment variable, or a command-line flag. This fixes `self-update` failing inside a workspace that raises the cutoff while succeeding everywhere else, and stops a repository from either waiving the cooldown or keeping you on an outdated pnpm by raising it.
- The same applies to the `trustPolicy` settings and to `ci`: a project can no longer weaken the trust check that guards the pnpm download, nor re-enable the confirmation prompt that a CI run suppresses.
When `self-update` refuses a version that is younger than the cutoff, an interactive run now offers to update anyway; non-interactive runs still fail. CI never prompts, even on a runner that attaches a TTY.
#### Patch Changes
- Fixed `pnpm licenses list` to report every version when the same package is installed under multiple aliases [pnpm/pnpm#13438](https://github.com/pnpm/pnpm/issues/13438).
- Sort `pnpm dedupe --check` snapshot changes for stable output across pnpm implementations.
- Strip Unicode formatting characters from registry- and manifest-derived terminal output.
- Speed up installs after compatible catalog or direct dependency range changes by retaining the locked version without resolving the dependency graph again.
- Speed up installs after safe override changes by reusing unambiguous compatible dependency resolutions, pruning obsolete dependencies, applying independent replacements and removals together, and handling parent-scoped `"-"` overrides without full lockfile resolution.
- Installing a local `file:` directory dependency with the global virtual store enabled no longer fails with `TypeError: Cannot read properties of undefined (reading 'split')` [#​13335](https://github.com/pnpm/pnpm/issues/13335).
Local directory dependencies — `file:` directories and injected workspace packages — now get a global-virtual-store slot of their own per project. They used to share one slot across every project that depended on a directory of the same name, so a project could end up linked to another project's copy of the dependency.
- The `Workspace` column of `pnpm update --interactive` now falls back to the project's path when its `name` is only whitespace, as it already did for a missing or empty one — all three render an equally blank label otherwise.
- Checking GitHub Actions dependencies for updates is now opt-in for every command. Neither `pnpm outdated` nor `pnpm update` reads the workflow files unless `--include-github-actions` is passed or `update.githubActions` is set to `true` in `pnpm-workspace.yaml`. Reading them runs `git ls-remote` against every referenced repository, which fails in environments where GitHub is not reachable the way pnpm assumes (a GitHub Enterprise Server, a custom certificate authority, or an offline network) [#​13254](https://github.com/pnpm/pnpm/issues/13254).
`pnpm outdated` accepts the `--include-github-actions` option too.
- `pnpm update --interactive` now measures its table in terminal columns rather than in characters. A package name, workspace name, or version containing wide characters (CJK, most emoji) no longer knocks its row's columns out of line with the rest of the group, and a wide character in a version no longer aborts the command with `Subject parameter value width cannot be greater than the container width` [#​13357](https://github.com/pnpm/pnpm/issues/13357).
- The `Workspace` column of `pnpm update --interactive` is more informative in two cases. A dependency outdated at the same version in several workspace projects is offered as one choice, since selecting it updates every project — that choice now names all of them instead of only the first. And a workspace project without a `name` is now labelled with its path rather than left blank, so several unnamed projects can be told apart.
- An auto-installed *optional* peer is no longer hoisted at a version the workspace root's own dependency on that package excludes. `resolvePeersFromWorkspaceRoot` already made the workspace root's specifier decide which version a missing *required* peer is installed at; the optional-peer picker ignored it and always took the highest version present anywhere in the graph. In a workspace whose root pins `postcss: 8.5.10`, an importer that depends on `webpack` and declares no `postcss` of its own got `[email protected]` hoisted for `terser-webpack-plugin`'s optional `postcss` peer, leaving two `[email protected]` instances in the graph [#​13320](https://github.com/pnpm/pnpm/issues/13320).
- `overrides` now also govern peers that pnpm auto-installs. Previously an override only rewrote dependencies declared in a manifest, so a peer nobody declares — installed because `autoInstallPeers` is on — resolved against its declared peer range and could bring in a second copy of the very package the override pinned. For example, with `overrides: { react: npm:[email protected] }` and a lone `lucide-react` dependency, pnpm installed `[email protected]`; it now installs the pinned `[email protected]` [#​13320](https://github.com/pnpm/pnpm/issues/13320).
- Under `resolvePeersFromWorkspaceRoot`, a workspace root dependency declared with `link:` or `file:` (or the path form of `workspace:`, such as `workspace:../pkg`) now satisfies another project's missing peer dependency at the linked package's own version, instead of being hoisted as a path. Those specifiers are relative to the project that declares them, so the same specifier reached a different directory — or none — from the project the peer was hoisted into, leaving a broken link. The root now has the same authority over the peer as it has when it declares the package with a version range [#​13373](https://github.com/pnpm/pnpm/issues/13373).
- Installs through a pnpr server now apply the project's whole verification policy. `minimumReleaseAgeExclude`, `minimumReleaseAgeIgnoreMissingTime`, `trustPolicy`, `trustPolicyExclude`, `trustPolicyIgnoreAfter`, and `trustLockfile` were ignored, so excluded packages were still held back and a lockfile containing them could be rejected.
`trustPolicy: no-downgrade` no longer fails with `TRUST_POLICY_INCOMPATIBLE_WITH_PNPR` when a pnpr server is configured.
`--frozen-lockfile` and `--no-prefer-frozen-lockfile` are now honored on the pnpr path, instead of resolving and rewriting the lockfile anyway. Since `frozenLockfile` defaults to `true` on CI, a CI install through a pnpr server now fails on an out-of-date lockfile rather than updating it.
- Workspace installs through a pnpr server no longer crash with `Cannot read properties of undefined (reading 'filter')` after linking, when `minimumReleaseAge` is active [#​13275](https://github.com/pnpm/pnpm/issues/13275).
- Fixed `pnpm dedupe` updating valid catalog resolutions when another matching version exists in the lockfile.
- `pnpm -r run "/pattern/" --no-bail` no longer exits zero when one of a project's matched scripts fails and a later one passes. The run summary carries a single status per project, and the passing script overwrote the recorded failure.
- Restored the store block a first install prints, naming how packages were materialized and where the stores live [#​13315](https://github.com/pnpm/pnpm/issues/13315):
```text
Packages are hard linked from the content-addressable store to the virtual store.
Content-addressable store is at: ~/.local/share/pnpm/store/v11
Virtual store is at: node_modules/.pnpm
```
- The root project's `pnpm:devPreinstall` script now runs before resolution and linking, as it does in pnpm 11. It is skipped under `--ignore-scripts`, `--lockfile-only` and `--dry-run`, by `pnpm fetch` and `pnpm rebuild`, and by a repeat install that is already up to date. Workspaces that use the hook to prepare state the install depends on — such as [next.js](https://github.com/vercel/next.js), which generates a placeholder `next` bin with it — were left with dependents linked against files that were never created [#​13313](https://github.com/pnpm/pnpm/issues/13313).
- Prevented `pnpm dedupe --check` from removing an incompatible `node_modules` directory.
- `pnpm update --workspace` no longer links dependencies the user never named:
- Running it with `updateConfig.ignoreDependencies` configured no longer fails with `ERR_PNPM_WORKSPACE_PACKAGE_NOT_FOUND` for a dependency that is only published to the registry. Such dependencies keep their specifiers, as they already did when no dependencies were ignored.
- Passing package selectors that match no direct dependency no longer falls back to linking every workspace dependency.
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
</details>
---
### Configuration
📅 **Schedule**: (UTC)
- Branch creation
- Between 12:00 AM and 03:59 AM, only on Monday (`* 0-3 * * 1`)
- Automerge
- At any time (no schedule defined)
🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied.
♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
👻 **Immortal**: This PR will be recreated if closed unmerged. Get [config help](https://github.com/renovatebot/renovate/discussions) if that's undesired.
---
- [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check this box
---
This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).
<!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0My4xOTEuMiIsInVwZGF0ZWRJblZlciI6IjQzLjE5MS4yIiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6W119-->
<!-- cloudflare-preview --> Preview: https://pr-500.docs-gitea-com.pages.dev
---------
Co-authored-by: Lunny Xiao <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/500
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Renovate Bot <[email protected]>
Pull requests now get a Cloudflare Pages preview deployment after the site build succeeds, and the preview URL is written back as the last line of the pull request description.
- `concurrency` cancels the previous preview build when a pull request is pushed again.
- After `make build`, the workflow copies `cloudflare/_headers` into `build/` (same as the production deployment, so the file is never published to S3/CloudFront) and runs `wrangler pages deploy build --project-name docs-gitea-com --branch pr-<number>`, then extracts the `*.pages.dev` URL from the wrangler output.
Notes:
- Pull requests from forks do not receive the secrets, in that case the preview steps are skipped with a log message and the build check still runs.
- Requires `CLOUDFLARE_ACCOUNT_ID` / `CLOUDFLARE_API_TOKEN` (Pages edit permission) and `DEPLOY_TOKEN` (pull request write permission).
<!-- cloudflare-preview --> Preview: https://pr-498.docs-gitea-com.pages.dev
Reviewed-on: https://gitea.com/gitea/docs/pulls/498
Reviewed-by: Zettat123 <[email protected]>
The scheduled `update swagger files` job has failed on **every** run since it was added (I checked the run list on gitea.com, ~250 consecutive failures). Two reasons:
1. Upstream moved the file: `go-gitea/gitea` main no longer has `templates/swagger/v1_json.tmpl`, it is now `templates/swagger/v1-swagger.generated.json`, and the placeholders changed from `{{.SwaggerAppVer}}` / `{{.SwaggerAppSubUrl}}` to `0.0.0+GITEA-API-APP-VERSION` / `/GITEA-API-APP-SUBURL/api/v1`. Because the script used `curl --silent` without `--fail`, the 404 body was written into `static/swagger-latest.json`.
2. `main` is protected (`required_approvals=1`), so the job's `git push` to main could never succeed.
Changes:
- `update_api_docs.sh`: `set -euo pipefail`, download with `--fail` (a broken download now aborts instead of committing garbage), try the new generated json first and fall back to the old template for released tags, handle the 1.28+/1.24+/<1.24 placeholder variants, and add a `--latest-only` flag.
- `Makefile`: new `update-api-docs-latest` target for the cron job; `clean` no longer deletes the committed `static/swagger-*.json` files (deleting them breaks `make build`).
- `.gitea/workflows/update_swagger.yaml`: every 12h (plus `workflow_dispatch`) it regenerates only `static/swagger-latest.json`, verifies no placeholder is left and the file is valid JSON, force pushes to `bot/update-swagger-latest` and opens a pull request via the API (HTTP 409 = a PR is already open, it just now points at the new commit). It skips entirely when the bot branch already carries the same file, so no PR churn.
- `README.md`: document the two make targets and the automation.
`secrets.DEPLOY_TOKEN` must belong to a real user with `write:repository` so the PR checks are triggered; the job fails with a clear message if it is empty.
Tested locally in a scratch clone: `make update-api-docs-latest` produces a valid `dev` spec with no placeholders left, `make update-api-docs` still regenerates all released versions, and the workflow shell logic was simulated against a local bare repo for all three paths (open PR / bot branch already up to date / main already up to date), including a shallow clone to confirm `fetch-depth: 1` force pushes work.
Fixes#454
Reviewed-on: https://gitea.com/gitea/docs/pulls/481
Reviewed-by: techknowlogick <[email protected]>
The `usage/actions/runner` has duplicated runner documentations as `/runner`. This PR remove the duplicated content and leave obtain registry token in `usage/actions/runner` and runner install/configurations in `/runner`.
Reviewed-on: https://gitea.com/gitea/docs/pulls/479
Reviewed-by: silverwind <[email protected]>
Fixes https://gitea.com/gitea/docs/issues/386
`gitea.event` was documented as `{...}` only. Adds an "Event payload" section listing the commonly used fields per event, including `gitea.event.schedule`, and notes where Gitea's payloads deviate from GitHub's.
Also corrects the trigger event table:
1. `pull_request` takes `synchronize`, not `synchronized`
1. `workflow_run` also emits `in_progress`
1. `schedule`, `workflow_call` and `pull_request_target` were missing
Reviewed-on: https://gitea.com/gitea/docs/pulls/478
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: silverwind <[email protected]>
Documents the WebSocket endpoint `/-/ws` added in https://github.com/go-gitea/gitea/pull/36965 and fixes the examples that could not serve it.
1. Nginx: adds the missing `proxy_http_version 1.1`, without it Nginx talks HTTP/1.0 upstream and Gitea rejects the handshake.
1. Apache HTTPD: `upgrade=websocket` on the existing `ProxyPass` rule, handled by `mod_proxy_http` since 2.4.48.
1. Nginx Proxy Manager: use the `Websockets Support` toggle, a duplicate `proxy_http_version` stops Nginx from starting.
1. IIS: the `WebSocket Protocol` Windows feature is required.
---------
Co-authored-by: silverwind <[email protected]>
Co-authored-by: silverwind <[email protected]>
Co-authored-by: mohammad rahimi <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/463
Reviewed-by: Lunny Xiao <[email protected]>
Reviewed-by: silverwind <[email protected]>
Co-authored-by: mohammad rahimi <[email protected]>
This PR contains the following updates:
| Package | Change | [Age](https://docs.renovatebot.com/merge-confidence/) | [Confidence](https://docs.renovatebot.com/merge-confidence/) |
|---|---|---|---|
| [pnpm](https://pnpm.io) ([source](https://github.com/pnpm/pnpm/tree/HEAD/pnpm11/pnpm)) | [`11.11.0` → `11.15.0`](https://renovatebot.com/diffs/npm/pnpm/11.11.0/11.15.0) |  |  |
---
### Release Notes
<details>
<summary>pnpm/pnpm (pnpm)</summary>
### [`v11.15.0`](https://github.com/pnpm/pnpm/releases/tag/v11.15.0): pnpm 11.15
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.14.0...v11.15.0)
##### Minor Changes
- Optional peer dependencies declared only via `peerDependenciesMeta` (for example `debug`'s `supports-color` peer) are now resolved from a satisfying version already present in the dependency graph, the same way explicitly declared optional peer dependencies are. Previously such peers were only resolved this way when the package's metadata was read back from the lockfile, so an unrelated dependency change could rewrite peer resolutions across the whole lockfile.
##### Patch Changes
- Updated `adm-zip` to prevent crafted ZIP archives from causing excessive memory allocation.
- `pnpm version -r` no longer writes a versioning-ledger entry with no consumed intents as a bare `intents:` key, which the next run failed to read with `ERR_PNPM_INVALID_VERSIONING_LEDGER`. Empty intent lists are now written as `intents: []`, and the ledger reader accepts the bare form left by earlier releases.
- Fixed pnpr workspace resolution to preserve project names and versions for `workspace:` dependencies.
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
### [`v11.14.0`](https://github.com/pnpm/pnpm/releases/tag/v11.14.0): pnpm 11.14
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.13.1...v11.14.0)
#### Minor Changes
- `peerDependencies` now accept dependency specifiers that carry a scheme — a named-registry spec (`<registry>:<version>`), an `npm:` alias, or a `file:`/git/URL spec — instead of rejecting them with `ERR_PNPM_INVALID_PEER_DEPENDENCY_SPECIFICATION` [#​13095](https://github.com/pnpm/pnpm/issues/13095). Such a peer is matched against the semver range carried by the specifier (`work:5.x.x` is checked as `5.x.x`, `npm:bar@^5` as `^5`), or against `*` when it carries no version, while the original specifier still selects the package to auto-install. Bare `name@version` values, which are almost always a mistake, are still rejected.
- Added `pnpm doctor`, which diagnoses the pnpm installation and the environment it runs in: the versions and install method, whether the global bin directory is on `PATH`, whether the store and cache are writable, which link strategies (reflink, hardlink, symlink) the store's filesystem supports, registry connectivity, and an offline `file:` install that exercises the resolve/store/link path end to end. Each check reports how to fix what it finds, and the command exits non-zero when any check fails.
Use `--offline` to skip the checks that need network access, `--json` for machine-readable output, and `--benchmark` to time the filesystem and install checks.
- Added support for executing multiple scripts matching a RegExp passed to `pnpm run` (e.g., `pnpm run "/^build:.*/"`), running matched scripts in deterministic lexicographical order. Restored the `--sequential` (`-s`) CLI option for `pnpm run`, which forces `workspaceConcurrency` to 1 so that matched scripts run sequentially one by one across and within packages.
#### Patch Changes
- Fixed `pnpm install` failing with `ERR_PNPM_LOCKFILE_IS_SYMLINK` when `pnpm-lock.yaml` is a symlink, as build sandboxes such as Bazel and Nix stage it [#​13073](https://github.com/pnpm/pnpm/issues/13073). Reading a lockfile through a symlink is allowed again, and an install that leaves the lockfile unchanged no longer rewrites it, so `--frozen-lockfile` no longer needs to write at all. Writing a *changed* lockfile through a symlink is still refused, as that would redirect the write onto the symlink's target.
- Fixed frozen installs incorrectly treating equivalent Git dependency specifiers as a stale lockfile. See [#​13039](https://github.com/pnpm/pnpm/issues/13039).
- `pnpm owner ls` now reports authentication and authorization failures (401/403) as dedicated errors that include the registry's response body, matching `pnpm owner add`/`rm`, instead of a generic `Failed to fetch owners` message.
- Recover from a metadata cache entry that disappears (concurrent cache cleanup, antivirus) after the registry has already answered the conditional request with `304 Not Modified`. The metadata is re-requested once without cache validators instead of failing the install with `ERR_PNPM_CACHE_MISSING_AFTER_304`.
- A project pinned to a broken pnpm release via `packageManager` or `devEngines.packageManager` now reports which release is broken and what to do about it, instead of failing inside the installer. `pnpm self-update` already refused these releases; the version switch does too.
- Prevent broken-lockfile errors from including snippets of the lockfile's contents.
- `pnpm self-update` now checks that the version it installed can run before making it the active pnpm. A release that installs but cannot execute is discarded with an error instead of replacing a working installation.
- Fixed an out-of-memory regression when workspace projects concurrently resolve a package with large registry metadata [pnpm/pnpm#13077](https://github.com/pnpm/pnpm/issues/13077).
- Fixed `pnpm update` rewriting exact version pins that use the `=` operator (for example `=3.5.1`) to a caret range (`^3.5.1`). Exact pins are now preserved and written back as the bare version. See [#​12745](https://github.com/pnpm/pnpm/issues/12745).
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
### [`v11.13.1`](https://github.com/pnpm/pnpm/releases/tag/v11.13.1): pnpm 11.13.1
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.13.0...v11.13.1)
#### Patch Changes
- Fixed `pnpm pack` applying workspace-root ignore rules when a workspace package has its own `.npmignore` file.
- Keep the interactive `minimumReleaseAge` approval prompt visible during `pnpm install`. The progress reporter now pauses its redraws while a prompt is waiting for input instead of overwriting it, so the install no longer hangs on a question the user cannot see [#​13019](https://github.com/pnpm/pnpm/issues/13019).
- Fixed `pnpm self-update` failing to link native platform binaries stored in sibling global virtual store slots.
### [`v11.13.0`](https://github.com/pnpm/pnpm/releases/tag/v11.13.0): pnpm 11.13
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.12.0...v11.13.0)
#### Minor Changes
- Added `versioning.epics` to `pnpm-workspace.yaml`. An epic ties a group of member packages to a lead package, constraining every member's major version to a band derived from the lead's major: while the lead is on major `M`, members live in `M*100 … M*100+99`. Members move independently inside the band (patch, minor, and a `major` intent that stays in-band); a bump that would carry a member past the band ceiling is rejected until the lead advances its own major. When a release plan takes the lead to a new stable major, every member re-bases to the band floor in the same plan. Membership is matched with pnpm's package selectors — name globs, `./`-prefixed directory globs, and `!`-prefixed negations.
- Added the `team` command for managing organization teams and team memberships on the registry, with create, destroy, add, rm, and ls subcommands and support for --otp, --parseable, and --json flags.
- Added native workspace release management [#​12952](https://github.com/pnpm/pnpm/issues/12952): the new `pnpm change` command records change intents as changesets-compatible `.changeset/*.md` files (`pnpm change status` shows the pending release plan), and the bare `pnpm version -r` consumes them — bumping versions across the workspace with dependent propagation through `workspace:` ranges, fixed groups, a `maxBump` cap, `--filter` narrowing, and `--dry-run` — writing changelogs, and recording consumed intents in a committed ledger that keeps cherry-picks and merge-backs between release branches safe. Packages can be moved onto per-package release lanes with the new `pnpm lane <name> --filter <pkg>` command and back with `pnpm lane main --filter <pkg>` (`pnpm lane` shows the membership), releasing `X.Y.Z-lane.N` prereleases from the same runs that release stable versions of the packages on the main lane. Configuration lives under the new `versioning` key of `pnpm-workspace.yaml` (`fixed`, `ignore`, `maxBump`, `lanes`, `changelog`). When two workspace projects publish the same name, intent files, `versioning.lanes`, and `versioning.fixed`/`ignore` may reference a project by its workspace-relative directory path (e.g. `"./pnpm/npm/pnpm"`) — the one additive extension to the changesets format, applied automatically by `pnpm change`.
Release changelogs default to `registry` storage (`versioning.changelog.storage`): no `CHANGELOG.md` is committed. Each release's section is composed at publish time and packed into the published tarball on top of the previously published version's changelog, and the consumed change intents are garbage-collected by a later `pnpm version -r` only once the registry confirms the version is published with its section. Set `versioning.changelog.storage: repository` to keep committed `CHANGELOG.md` files instead.
- Added a new override selector form with an empty range — `"pkg@": "<version>"` — called a convergence override. It rewrites a dependency edge only when its exact version satisfies the edge's declared range, so compatible consumers converge on one version while incompatible consumers keep their own resolution — now and for any dependent added in the future [#​12794](https://github.com/pnpm/pnpm/issues/12794).
```yaml
overrides:
"form-data@": 4.0.6
```
The value must be an exact version. When a full resolution detects that every declared range also admits a newer version, pnpm warns that the override is stale and names the version to converge on. Previously an empty range in an override selector was undocumented and behaved like a bare (unscoped) override.
#### Patch Changes
- A `tokenHelper` set in the global pnpm `auth.ini` is no longer rejected as project-level configuration. The guard that blocks `tokenHelper` from a project `.npmrc` only treated `~/.npmrc` as a trusted source, so a helper written to `auth.ini` (for example by `pnpm config set`) failed on every command and could not even be removed with `pnpm config delete`. A `tokenHelper` in a workspace or project `.npmrc` is still rejected.
- `pnpm cache delete` now removes a package's metadata from every metadata cache directory (`metadata`, `metadata-full`, and `metadata-full-filtered`), instead of only the one the current resolution mode reads. Previously a package cached under a different mode (e.g. `metadata-full-filtered`) was left behind. Closes [#​12753](https://github.com/pnpm/pnpm/issues/12753).
- Fixed an injected workspace dependency (`injectWorkspacePackages: true`) incorrectly staying as `file:` instead of deduping back to `link:` when an unrelated, ordinary shared dependency resolved to a peer-suffixed variant for the target project's own copy but not for the injected occurrence. See [#​10433](https://github.com/pnpm/pnpm/issues/10433).
- `pnpm deploy` now supports workspaces that use catalogs.
- Fixed `pnpm deploy` with a shared lockfile so local `file:` tarball dependencies keep their package name in the generated deploy lockfile. This prevents warm-store deploys from failing with `ERR_PNPM_UNEXPECTED_PKG_CONTENT_IN_STORE` when the tarball filename includes the version.
- Options that follow `create`, `exec`, or `test` appearing as a subcommand of another command are now parsed instead of being silently treated as positional parameters. For example, `pnpm team create @​org:team --registry <url>` previously ignored the `--registry` option and sent the request to the default registry.
- `pnpm add -g`, `pnpm update -g`, `pnpm setup`, and the self-updater no longer fail with `ERR_PNPM_MISSING_TIME` when `trustPolicy: no-downgrade` or `resolutionMode: time-based` is set in the global config [#​12883](https://github.com/pnpm/pnpm/issues/12883). The decision to fetch full registry metadata now lives in one place, and the `no-downgrade` trust policy always requests full metadata (matching the self-updater), since the trust evidence it checks is missing from abbreviated metadata even on registries that include the `time` field.
- `pnpm list` and `pnpm why` no longer crash with `EMFILE: too many open files` when a project has a large number of unsaved dependencies (packages present in `node_modules` but not in the lockfile). The reads of those packages are now concurrency-limited.
- The published `pnpm` package no longer declares `dependencies` or `devDependencies`. Because the CLI bundles its runtime dependencies into `dist/node_modules`, those fields are dropped when packing, so `npm install` of the tarball no longer tries to resolve internal-only packages such as `@pnpm/test-ipc-server`. Closes [#​12955](https://github.com/pnpm/pnpm/issues/12955).
- Fixed `pnpm publish --otp` and `pnpm publish --batch --otp` to send the configured OTP to the registry.
- `pnpm publish` again sends the package's README to the registry as metadata, so registries can render it on the package page. The readme is always included in the published metadata (matching the npm CLI), while the `embed-readme` setting continues to control only whether the readme is written into the `package.json` inside the tarball. This restores the behavior that was lost when publishing became fully native. Closes [#​12966](https://github.com/pnpm/pnpm/issues/12966).
- Fixed the dependency status check wrongly reporting "up to date" when a `package.json`, `.pnpmfile.cjs`, or patch file was edited in the same second as the previous install, on filesystems that record mtimes at whole-second resolution (for example ext4 with 128-byte inodes). The optimistic repeat-install fast path and `verify-deps-before-run` compared mtimes strictly, so a same-second edit whose mtime rounded down looked unchanged and re-resolution was skipped. Such a file's whole second is now treated as possibly-modified, falling through to the content check; behavior on sub-second filesystems is unchanged.
- Retry package metadata requests when a registry or proxy returns `304 Not Modified` to an unconditional request, preventing false `ERR_PNPM_CACHE_MISSING_AFTER_304` failures [pnpm/pnpm#12882](https://github.com/pnpm/pnpm/issues/12882).
If the retry also returns `304`, report `ERR_PNPM_META_NOT_MODIFIED_WITHOUT_CACHE` instead.
- Fixed `pnpm update` removing transitive lockfile entries when `dedupePeerDependents` is disabled and the selected package is absent [pnpm/pnpm#12456](https://github.com/pnpm/pnpm/issues/12456).
- Limit modern deploy lockfiles and localized virtual stores to dependencies reachable from the selected dependency groups.
- A `tokenHelper` command is now given a 60-second time limit. A helper that hangs (deadlock, stuck I/O) is killed and reported as an error instead of leaving the command waiting forever.
- Fixed orphaned child processes on Windows when pnpm exits on an error while commands spawned by `pnpm exec` or `pnpm dlx` are still running (for example, when one project's command fails during `pnpm --recursive exec`). The PIDs of these commands are now recorded when they are spawned and their whole process trees are terminated with `taskkill` on an error exit. Previously the cleanup relied on enumerating the system process list, which is so slow on Windows that the enumeration hit its timeout and the cleanup was silently skipped [#​12406](https://github.com/pnpm/pnpm/issues/12406).
- `pnpm pack` now respects workspace-root `.npmignore` and `.gitignore` files when packing workspace packages.
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
### [`v11.12.0`](https://github.com/pnpm/pnpm/releases/tag/v11.12.0): pnpm 11.12
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.11.0...v11.12.0)
#### Minor Changes
- [`a897ef7`](https://github.com/pnpm/pnpm/commit/a897ef7): Custom fetchers exported from a pnpmfile can now delegate by returning a `{ delegate: <resolution> }` envelope: pnpm rewrites the package's resolution to the delegated shape and runs the built-in fetcher on it. This is the portable delegation form that also works in pacquet, where `cafs` and `fetchers` cannot be passed to the hook. Related to [pnpm/pnpm#11685](https://github.com/pnpm/pnpm/issues/11685).
#### Patch Changes
- [`2b02764`](https://github.com/pnpm/pnpm/commit/2b02764): The changed-packages filter (`--filter "...[<since>]"`) no longer allows an option-like `<since>` value (such as `--output=<path>`) to be interpreted as a git option — git now rejects it as a bad revision. The repository root is also resolved to the nearest `.git` entry, so the filter works in a git worktree checked out inside another repository's tree.
- [`43711ce`](https://github.com/pnpm/pnpm/commit/43711ce): `pnpm outdated` no longer checks the registry for dependencies that are resolved from local `link:`, `file:`, or `workspace:` references in the lockfile [#​12827](https://github.com/pnpm/pnpm/issues/12827).
- [`3c6718b`](https://github.com/pnpm/pnpm/commit/3c6718b): Fixed a deadlock in peer dependency resolution: `pnpm install` hung forever when a peer dependency cycle spanned a project's own dependencies and auto-installed peer providers, for example when installing `[email protected].3` [#​12921](https://github.com/pnpm/pnpm/issues/12921).
- [`252f15e`](https://github.com/pnpm/pnpm/commit/252f15e): Fixed peer dependency auto-install picking a version the peer range rejects. In a workspace with several projects, a package declaring a peer dependency with a semver range (for example `^1.0.0`) could get the highest version found anywhere in the workspace (for example a `2.0.0` resolved for another project) instead of a version that satisfies the range. Peers are now deduplicated onto the highest preferred version that satisfies the declared range, and when none does, the range is resolved from the registry.
Also fixed re-resolving with an existing lockfile hoisting a different peer version than a fresh install of the same manifest: root dependencies reused from the lockfile were invisible to peer hoisting, so a peer that a root dependency provides could be bound to another version.
- [`a38adda`](https://github.com/pnpm/pnpm/commit/a38adda): `pnpm self-update <version>` now installs the requested pnpm version when it matches the currently running version but is missing from the global self-update directory.
- [`6a85968`](https://github.com/pnpm/pnpm/commit/6a85968): `pnpm stage list` now stops paginating after a fail-safe cap of 1000 pages, so a misbehaving registry cannot keep the command looping forever.
- [`eee7c9a`](https://github.com/pnpm/pnpm/commit/eee7c9a): `verify-deps-before-run` no longer spawns a `pnpm install` when pnpm is executed in a directory that has no `package.json`. A mistyped command run outside a project (for example `pnpm witch 10 login`) used to crash with a confusing error from the spawned install; now it fails with the regular "no package.json found" error.
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
</details>
---
### Configuration
📅 **Schedule**: (UTC)
- Branch creation
- At any time (no schedule defined)
- Automerge
- At any time (no schedule defined)
🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied.
♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
🔕 **Ignore**: Close this PR and you won't be reminded about this update again.
---
- [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check this box
---
This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).
<!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0My4xOTEuMiIsInVwZGF0ZWRJblZlciI6IjQzLjE5MS4yIiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6W119-->
---------
Co-authored-by: Lunny Xiao <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/469
Reviewed-by: silverwind <[email protected]>
Co-authored-by: Renovate Bot <[email protected]>
This PR contains the following updates:
| Package | Change | [Age](https://docs.renovatebot.com/merge-confidence/) | [Confidence](https://docs.renovatebot.com/merge-confidence/) |
|---|---|---|---|
| [pnpm](https://pnpm.io) ([source](https://github.com/pnpm/pnpm/tree/HEAD/pnpm11/pnpm)) | [`11.10.0` → `11.11.0`](https://renovatebot.com/diffs/npm/pnpm/11.10.0/11.11.0) |  |  |
---
### Release Notes
<details>
<summary>pnpm/pnpm (pnpm)</summary>
### [`v11.11.0`](https://github.com/pnpm/pnpm/blob/HEAD/pnpm11/pnpm/CHANGELOG.md#11110)
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.10.0...v11.11.0)
##### Minor Changes
- [`508b8c2`](https://github.com/pnpm/pnpm/commit/508b8c2): Added the `pnpm access` command for managing package access and visibility on the registry, supporting listing packages and collaborators, getting and setting package status and MFA requirements, and granting or revoking team access.
##### Patch Changes
- [`c70e33e`](https://github.com/pnpm/pnpm/commit/c70e33e): Allow `allowBuilds` entries for git-hosted packages to match by repository URL without pinning the resolved commit hash. This lets trusted git repositories keep running their build scripts after branch updates without approving each new commit, while package-name-only rules still do not approve git-hosted artifacts.
- [`3067e4f`](https://github.com/pnpm/pnpm/commit/3067e4f): Reduced peak memory usage during cold-cache dependency resolution. The metadata fetch is memoized for the whole resolution phase, and it was retaining each package's raw registry response body (used only to mirror the response to disk) for that entire time. The memoized cache now holds a body-less copy, so the raw body only lives as long as the call that writes the disk mirror. On large graphs that fetch full metadata (e.g. with `minimumReleaseAge` or `trustPolicy` enabled) this cuts peak RSS by roughly 30%, back in line with pnpm 10. The resolved lockfile is unchanged.
- [`51300fd`](https://github.com/pnpm/pnpm/commit/51300fd): Prevent a crafted `pnpm-lock.yaml` from writing package content outside the virtual store. A dependency path key whose name reconstructs to a path-traversal sequence (e.g. `../../../tmp/[email protected]`) is now rejected by the isolated (virtual-store) linker and the Plug'n'Play resolver map, matching the containment already applied to the hoisted linker. Under the global virtual store, a traversal in the version-derived path segment (e.g. a snapshot `version: "../../x"`) is now rejected at `formatGlobalVirtualStorePath`, the single point every global-virtual-store slot path funnels through — closing the same escape in the isolated linker, the resolver's dependency-graph builder, and the config-dependency installer.
- [`f8058eb`](https://github.com/pnpm/pnpm/commit/f8058eb): Reject symlinked `pnpm-lock.yaml` files when reading or writing the env lockfile document.
- [`9318a11`](https://github.com/pnpm/pnpm/commit/9318a11): Allow `registries` and `namedRegistries` to be configured in the global `config.yaml` file.
- [`51300fd`](https://github.com/pnpm/pnpm/commit/51300fd): Fixed a path traversal vulnerability where a dependency whose manifest `name` was a scoped path traversal (e.g. `@x/../../../<path>`) could be written outside `node_modules` to an attacker-controlled location during `pnpm install`, even with `--ignore-scripts`. The isolated linker now validates the package name before using it as a directory name, matching the existing protection in the hoisted linker.
- [`14332f0`](https://github.com/pnpm/pnpm/commit/14332f0): Fail instead of silently removing an optional dependency's locked entries from `pnpm-lock.yaml` when the registry cannot resolve it. Previously, when registry metadata lacked a version that the lockfile already pinned (for example, a mirror that had not synced a recent release yet), `pnpm install` and `pnpm dedupe` silently dropped the optional dependency's entries — emptying maps such as the platform binaries of `@napi-rs/canvas` — so the lockfile differed between machines and frozen installs on other hosts had nothing to link [#​12853](https://github.com/pnpm/pnpm/issues/12853).
- [`fecfe83`](https://github.com/pnpm/pnpm/commit/fecfe83): Fixed peer dependency resolution with `autoInstallPeers` when a workspace package depends on a version of a package that a transitive dependency's self-contained closure also provides for itself. The peer providers that are attached to the root project for reuse are no longer peer-resolved a second time in the root context, so packages inside such a closure no longer get their peers bound to the root project's incompatible version [#​4993](https://github.com/pnpm/pnpm/issues/4993).
- [`5a4daec`](https://github.com/pnpm/pnpm/commit/5a4daec): `${...}` environment-variable placeholders in the `httpProxy`, `httpsProxy`, `noProxy`, `proxy`, and `noproxy` settings are no longer expanded when these settings come from a project's `pnpm-workspace.yaml`. They now receive the same protection already applied to `registry`, `namedRegistries`, and `pnprServer`.
- [`d1da02e`](https://github.com/pnpm/pnpm/commit/d1da02e): `pnpm publish` no longer prints credentials when the target registry is configured with inline `user:pass@` credentials (e.g. `registry=https://user:[email protected]/`). They are now redacted both from the "publishing to registry" line and from the OIDC (trusted publishing) failure messages.
- [`dcfc611`](https://github.com/pnpm/pnpm/commit/dcfc611): `pnpm self-update` now honors `trustPolicy=no-downgrade`. It resolves the target pnpm version against full registry metadata, so it refuses to switch to a version whose supply-chain trust evidence is weaker than an earlier-published one, the same way a regular install does.
- [`a8ad82d`](https://github.com/pnpm/pnpm/commit/a8ad82d): Register the `pn` alias in generated shell completion scripts.
- [`25bd5c3`](https://github.com/pnpm/pnpm/commit/25bd5c3): Fixed standalone installer downgrades from pnpm v12 to v11.
- [`23996e9`](https://github.com/pnpm/pnpm/commit/23996e9): `pnpm runtime set <name> <version>` now validates its arguments: the name must be `node`, `deno`, or `bun`, and the version must not contain a comma. Previously these were interpolated straight into a `pnpm add` selector, where an unsupported name or a comma (e.g. `node 22,is-positive`) could be misread as a list of packages or a local directory and install unintended packages or bins.
</details>
---
This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).
<!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0My4xOTEuMiIsInVwZGF0ZWRJblZlciI6IjQzLjE5MS4yIiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6W119-->Reviewed-on: https://gitea.com/gitea/docs/pulls/465
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Renovate Bot <[email protected]>
This PR contains the following updates:
| Package | Change | [Age](https://docs.renovatebot.com/merge-confidence/) | [Confidence](https://docs.renovatebot.com/merge-confidence/) |
|---|---|---|---|
| [pnpm](https://pnpm.io) ([source](https://github.com/pnpm/pnpm/tree/HEAD/pnpm11/pnpm)) | [`11.9.0` → `11.10.0`](https://renovatebot.com/diffs/npm/pnpm/11.9.0/11.10.0) |  |  |
---
### Release Notes
<details>
<summary>pnpm/pnpm (pnpm)</summary>
### [`v11.10.0`](https://github.com/pnpm/pnpm/blob/HEAD/pnpm11/pnpm/CHANGELOG.md#11100)
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.9.0...v11.10.0)
##### Minor Changes
- [`e2e3c81`](https://github.com/pnpm/pnpm/commit/e2e3c81): Added the `issues` command as an alias of `bugs`, so `pnpm issues` opens the package's bug tracker URL in the browser.
- [`8491f8e`](https://github.com/pnpm/pnpm/commit/8491f8e): Added the `prefix` command which prints the current package prefix directory (or global prefix directory if `-g` / `--global` is used).
- [`3425e80`](https://github.com/pnpm/pnpm/commit/3425e80): Added an `_auth` setting for configuring registry authentication as a single structured (URL-keyed) value. It can be set in the **global** pnpm config (`config.yaml`) or, for CI, via the `pnpm_config__auth` environment variable. The env form sidesteps the GitHub Actions / bash / zsh limitation that broke the existing `pnpm_config_//host/:_authToken=…` form (env var names containing `/`, `:`, or `.` are silently dropped). Closes [#​12314](https://github.com/pnpm/pnpm/issues/12314).
The value is keyed by registry URL so each secret is explicitly bound to the host that may receive it. Registry URL keys must use `http` or `https` and must not include credentials, query strings, or fragments:
```sh
export pnpm_config__auth='{"https://registry.npmjs.org":{"@​":{"authToken":"npm-token"},"@​org":{"authToken":"org-token"}}}'
```
The equivalent in the global `config.yaml`:
```yaml
_auth:
https://registry.npmjs.org:
"@​":
authToken: npm-token
"@​org":
authToken: org-token
```
Within each registry URL, `@` means registry-wide/default credentials and package scopes like `@org` bind credentials to that scope on the same host. The only supported credential field is `authToken` (maps to `_authToken` / bearer auth); the deprecated `basicAuth` / `username` + `password` forms are intentionally not accepted here.
Each entry also infers a trusted registry route: `@` routes the default registry (and `pnpm add <pkg>` resolves there), and `@org` routes that scope. Because the credential and destination host arrive in one trusted value, repo-controlled `pnpm-workspace.yaml` or project `.npmrc` cannot redirect the token to a different host. `_auth` is honored **only** from the env var and the global config — it is ignored in a project `pnpm-workspace.yaml` / `.npmrc`, so repo-controlled config can never supply registry auth. Precedence: CLI flags (`--registry`, `--@​scope:registry`) > `pnpm_config__auth` > global `config.yaml` `_auth` > `pnpm-workspace.yaml`.
Both `pnpm_config__auth` (lowercase, documented form) and `PNPM_CONFIG__AUTH` (all-caps, the shell convention some CI runners apply) are honored. If both are set, lowercase wins unless it is empty, in which case uppercase is used. The env var wins over the global `config.yaml` `_auth` on a conflicting key. `tokenHelper` is not supported in `_auth`. Parsing is strict: a malformed value (bad JSON, wrong shape, invalid registry URL or scope, an unsupported credential field) fails fast with an error rather than being silently dropped.
**Pacquet parity note:** the pacquet (Rust) port supports the same single credential field as the TS CLI: `authToken`.
- [`a33eeec`](https://github.com/pnpm/pnpm/commit/a33eeec): `pnpm self-update` and `packageManager` version-switching can now install and link pnpm v12 (the Rust port), published with equal content under both the `pnpm` and `@pnpm/exe` names on the `next-12` dist-tag. Its native binaries ship as `@pnpm/exe.<platform>-<arch>` packages, which pnpm's built-in installer links directly — no Node.js launcher, so the command pays no Node startup cost. v12 is initialized exactly like `@pnpm/exe`, including per-platform global-virtual-store hashing. From v12 onward the install converges on the unscoped `pnpm` package (the Rust exe) — even when updating from the SEA `@pnpm/exe` build.
- [`1dd12bd`](https://github.com/pnpm/pnpm/commit/1dd12bd): When resolving through a pnpr install-accelerator server, pnpm no longer forwards its own upstream registry credentials in the resolve request. Only the `Authorization` header identifying the caller to pnpr is sent. The pnpr server now selects upstream credentials from its own route policy (operator-configured upstream credential aliases), so private dependencies resolve through a pnpr-managed alias the caller is authorized to use, rather than by sending the client's registry tokens to the server.
- [`1e81761`](https://github.com/pnpm/pnpm/commit/1e81761): Expose web authentication `authUrl` and `doneUrl` in JSON error output when OTP is required in a non-interactive terminal [#​12724](https://github.com/pnpm/pnpm/issues/12724).
##### Patch Changes
- [`2f389d6`](https://github.com/pnpm/pnpm/commit/2f389d6): Added the Node.js release team's new signing key (Stewart X Addison, `655F3B5C1FB3FA8D1A0CA6BDE4A7D232B936D2FD`) to the embedded Node.js release keys, so runtimes whose `SHASUMS256.txt` is signed by the new releaser verify successfully.
- [`acbdb94`](https://github.com/pnpm/pnpm/commit/acbdb94): Fixed shell tab completion not suggesting workspaces after the `-F` alias for `--filter` option.
- [`dcabb78`](https://github.com/pnpm/pnpm/commit/dcabb78): Fixed `pnpm up -r <pkg>` bumping unrelated packages that have open semver ranges. Previously, any update mutation nullified the lockfile-derived `preferredVersions` globally, so packages with `^x.y.z` ranges could re-resolve to newer compatible versions even though the user only asked to update a specific package. The install layer now always seeds `preferredVersions` from the lockfile, and caller-supplied preferred versions (such as the vulnerability penalties of `pnpm audit --fix`) layer on top of the seed instead of replacing it. The targeted package still bumps: the per-resolve `updateRequested` flag makes the resolver ignore the target's own lockfile pins.
Closes [#​10662](https://github.com/pnpm/pnpm/issues/10662).
- [`d539172`](https://github.com/pnpm/pnpm/commit/d539172): Fixed pnpm pack and pnpm publish failing when prepack generates files that are included in the package and postpack cleans them up.
- [`be6505a`](https://github.com/pnpm/pnpm/commit/be6505a): Hardened global package management:
- On Windows, removing or updating a global package now also cleans up the `node.exe` flavor of a bin, so a stale `node.exe` no longer survives on `PATH` after uninstall, and a new global install no longer silently overwrites an existing `node.exe`.
- `pnpm add -g pnpm@<version>` (and `@pnpm/exe@<version>`) is now rejected like the bare `pnpm` form, pointing to `pnpm self-update`.
- Dependency aliases read from a global package's manifest are validated before being joined onto `node_modules` paths, preventing a tampered manifest from escaping the install directory.
- Each global install group is created in its own freshly-made directory (no longer reusing a colliding or pre-existing path).
- Removing or updating a global package no longer unlinks a bin that belongs to a different globally installed package.
- [`25c7388`](https://github.com/pnpm/pnpm/commit/25c7388): pnpm now rejects `jsr:` specifiers whose package name is not a valid npm package name — an empty scope or name (e.g. `jsr:@​scope/`), path separators inside the name, or any other shape `validate-npm-package-name` rejects — with `ERR_PNPM_INVALID_JSR_PACKAGE_NAME` instead of silently converting them into a malformed `@jsr/...` npm package name.
- [`25c7388`](https://github.com/pnpm/pnpm/commit/25c7388): pnpm now rejects named-registry specifiers (e.g. `gh:`) whose package name is not a valid npm package name — an empty scope (e.g. `gh:@​/bar`), path separators inside the name (e.g. `gh:@​scope/../name`), or any other shape `validate-npm-package-name` rejects — with `ERR_PNPM_INVALID_NAMED_REGISTRY_PACKAGE_NAME` instead of passing the name through to registry URLs and metadata cache file paths.
- [`96da7c5`](https://github.com/pnpm/pnpm/commit/96da7c5): node-gyp's `gyp_main.py` and `gyp` entrypoints are now packed with the executable bit in the `pnpm` and `@pnpm/exe` tarballs. Without it, building native addons from source could fail with a permission error.
- [`99982b9`](https://github.com/pnpm/pnpm/commit/99982b9): Sped up resolution and reduced memory use against registries that ignore npm's abbreviated metadata format and always return the full package document (for example, Azure DevOps Artifacts). pnpm now strips such documents down to the abbreviated field set before caching them. Resolution output is unchanged, and registries that honor the abbreviated format (such as the npm registry) pay no extra cost.
- [`11a7fdd`](https://github.com/pnpm/pnpm/commit/11a7fdd): Sped up offline and `--prefer-offline` resolution on large workspaces (e.g. `pnpm dedupe --offline`, `pnpm install --offline`). Package metadata loaded from the local cache is now kept in memory, so each package's metadata is parsed once per command instead of once per dependent that references it.
- [`2c7369d`](https://github.com/pnpm/pnpm/commit/2c7369d): `pnpm pack-app` now rejects `--entry` / `pnpm.app.entry` and `--output-dir` / `pnpm.app.outputDir` values that are absolute paths or escape the project directory via `..` (or a symlink that resolves outside it), and refuses to write the produced executable when its target path already exists as a symlink (or other non-regular file). This prevents a repository-controlled `package.json` from embedding host files (such as an SSH key) into the produced executable, writing build artifacts outside the project, or overwriting an arbitrary file through a committed symlink. The new error codes are `ERR_PNPM_PACK_APP_ENTRY_OUTSIDE_PROJECT`, `ERR_PNPM_PACK_APP_OUTPUT_DIR_OUTSIDE_PROJECT`, and `ERR_PNPM_PACK_APP_OUTPUT_FILE_NOT_REGULAR`.
When ad-hoc signing macOS targets, `pnpm pack-app` now runs the system `codesign` by absolute path and resolves `ldid` to a location outside the project, so a repository-controlled `node_modules/.bin` on `PATH` cannot hijack the signer.
- [`ce5d5a5`](https://github.com/pnpm/pnpm/commit/ce5d5a5): Relative paths in `patchedDependencies` are now resolved against the lockfile directory when computing patch file hashes, so running `pnpm install` from a subdirectory no longer fails with `ENOENT` looking for the patch file in the wrong location [#​12762](https://github.com/pnpm/pnpm/pull/12762).
- [`ebb4096`](https://github.com/pnpm/pnpm/commit/ebb4096): `pnpm peers` no longer reports a conflict for a missing peer dependency that is ignored via `pnpm.peerDependencyRules.ignoreMissing`.
- [`dcabb78`](https://github.com/pnpm/pnpm/commit/dcabb78): Fixed a prototype-pollution hazard when seeding preferred versions: a dependency named `__proto__` in a manifest or in `pnpm-lock.yaml` could write through `Object.prototype` (or crash the install) while the preferred-versions map was being built. The maps are now null-prototype objects, so crafted package names land as plain keys.
- [`f38e696`](https://github.com/pnpm/pnpm/commit/f38e696): Hardened `pnpm deploy --force` so it refuses unsafe deploy targets such as workspace roots, parent directories, out-of-workspace paths, and symlinked target parents.
- [`806c3ec`](https://github.com/pnpm/pnpm/commit/806c3ec): pnpm no longer warns about ignored project-level auth settings when `PNPM_CONFIG_NPMRC_AUTH_FILE` points at the project `.npmrc` — setting it to that file is an explicit opt-in to trusting it, so auth env variables in it are expanded [pnpm/pnpm#12480](https://github.com/pnpm/pnpm/issues/12480).
- [`991405e`](https://github.com/pnpm/pnpm/commit/991405e): Restore differential rendering (`ansi-diff`) to fix duplicated output lines introduced by [#​12351](https://github.com/pnpm/pnpm/issues/12351).
- [`c121235`](https://github.com/pnpm/pnpm/commit/c121235): Fixed the topological order of `--filter`ed commands (`pnpm run`, `pnpm exec`, `pnpm publish`, `pnpm pack`, `pnpm rebuild`) when the selected projects depend on each other only transitively through projects that were not selected. Previously such selected projects could run concurrently or in the wrong order; now a project always runs after the selected projects it transitively depends on, while projects without a real dependency relationship still run concurrently. This now also holds for prod-only filters (`--filter-prod`), which resolve order through the production dependency graph so transitive production dependencies are respected without pulling back the dev dependencies the filter drops, and for selections that mix `--filter` with `--filter-prod` [#​8335](https://github.com/pnpm/pnpm/issues/8335).
- [`d539172`](https://github.com/pnpm/pnpm/commit/d539172): `pnpm pack` and `pnpm publish` no longer follow a symlinked workspace `LICENSE` file when injecting it into a package that has no license of its own. Following the symlink could pack bytes from outside the workspace into the published tarball.
- [`dcabb78`](https://github.com/pnpm/pnpm/commit/dcabb78): Fixed `pnpm up <pkg>` producing a different result than a fresh install of the same manifests would. The resolver now distinguishes `updateRequested` (true only for packages that match the user's update target) from the broader `update` flag, and for the targeted package ignores only its own lockfile-derived preferred-version pins — so the target re-resolves exactly as if its lockfile entries were deleted and `pnpm install` ran. Preferred versions a fresh install applies (manifest pins, versions propagated down the dependency chain, and the vulnerability-avoidance penalties of `pnpm audit --fix`) stay in effect, so an update never installs duplicate versions that a reinstall from scratch would not reproduce. When a preferred version holds the update target below the newest version its range admits, pnpm now prints a warning explaining that reaching the newer version everywhere requires an override.
- [`dcabb78`](https://github.com/pnpm/pnpm/commit/dcabb78): `pnpm update <dep>@​<version>` now prints a warning when `<dep>` is only present as a transitive dependency: the requested version cannot be applied there (updates resolve the target the way a fresh install would), and the warning recommends adding the version to `pnpm.overrides` instead, which is the mechanism that does pin transitive dependencies. Closes [#​12744](https://github.com/pnpm/pnpm/issues/12744).
- [`a6c4d5f`](https://github.com/pnpm/pnpm/commit/a6c4d5f): When a dependency cannot be found in the registry (404) or the registry has no matching version, and a workspace project with the same name exists only at non-matching versions, the error now reports the available workspace versions (`ERR_PNPM_NO_MATCHING_VERSION_INSIDE_WORKSPACE`) instead of the raw registry failure [pnpm/pnpm#1379](https://github.com/pnpm/pnpm/issues/1379). Other registry failures (authorization, network, server errors) still propagate unchanged. The pacquet (Rust) resolver applies the same behavior.
</details>
---------
Co-authored-by: Lunny Xiao <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/460
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Renovate Bot <[email protected]>
This PR contains the following updates:
| Package | Change | [Age](https://docs.renovatebot.com/merge-confidence/) | [Confidence](https://docs.renovatebot.com/merge-confidence/) |
|---|---|---|---|
| [pnpm](https://pnpm.io) ([source](https://github.com/pnpm/pnpm/tree/HEAD/pnpm)) | [`11.7.0` → `11.9.0`](https://renovatebot.com/diffs/npm/pnpm/11.7.0/11.9.0) |  |  |
---
### Release Notes
<details>
<summary>pnpm/pnpm (pnpm)</summary>
### [`v11.9.0`](https://github.com/pnpm/pnpm/releases/tag/v11.9.0): pnpm 11.9
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.8.0...v11.9.0)
#### Minor Changes
- [`bae694f`](https://github.com/pnpm/pnpm/commit/bae694f): Some registries generate tarballs on-demand and cannot provide an integrity checksum in their package metadata. In that case pnpm now computes the integrity from the downloaded tarball and stores it in the lockfile, so the entry is verifiable on subsequent installs instead of being written without an integrity (which would fail the next install). This also applies to `--lockfile-only`: the tarball is downloaded so its integrity can be computed. A lockfile entry that is still missing its integrity is rejected as a `ERR_PNPM_MISSING_TARBALL_INTEGRITY` lockfile verification violation (the install fails closed) rather than being silently re-fetched.
- [`6c35a43`](https://github.com/pnpm/pnpm/commit/6c35a43): Added `--exclude-peers` to `pnpm sbom`. With `auto-install-peers` (the default), peer dependencies resolve into the lockfile and are otherwise indistinguishable from the package's own dependencies. The flag drops peer dependencies (and any transitive subtree reachable only through them) from the SBOM. CycloneDX 1.7 has no scope or relationship that expresses "consumer-provided peer", so omission is the only spec-clean handling. The flag name matches `pnpm list --exclude-peers`; note the SBOM flag prunes a peer's exclusive subtree, which is stricter than `pnpm list` (which only hides leaf peers).
#### Patch Changes
- [`25a829e`](https://github.com/pnpm/pnpm/commit/25a829e): `pnpm audit --fix` now writes a single combined `minimumReleaseAgeExclude` entry per package (e.g. `[email protected].1 || 0.21.1`) instead of one entry per version, matching the format documented for the setting. Existing per-version entries in `pnpm-workspace.yaml` are merged into the combined form rather than left as duplicates. Installs that auto-collect immature versions into `minimumReleaseAgeExclude` now report the same combined entries, so the "Added N entries" message matches what is written to the manifest [#​12534](https://github.com/pnpm/pnpm/issues/12534).
- [`1cbb5f2`](https://github.com/pnpm/pnpm/commit/1cbb5f2): Fixed non-deterministic peer resolution that could add or remove an optional transitive peer — for example `@babel/core`, reached through `styled-jsx` — from a package's peer-dependency suffix across otherwise identical installs, churning the lockfile and causing intermittent `pnpm dedupe --check` failures in CI. When a package's children are resolved by one occurrence (the "owner") and reused by a deeper consumer, whether that consumer inherited the owner's missing peers depended on whether the owner's resolution had finished yet — a race under concurrent resolution. The decision is now a function of the dependency graph's structure rather than resolution-completion order.
- [`d577eea`](https://github.com/pnpm/pnpm/commit/d577eea): Fixed a Windows flakiness in `pnpm dlx` where a failed install could surface a spurious `EBUSY: resource busy or locked` error. The cleanup of a partially-populated dlx cache is now best-effort with retries and no longer masks the original error.
- [`ec7cf70`](https://github.com/pnpm/pnpm/commit/ec7cf70): Shortened the `pnpm dlx` cache path so deep dependency trees no longer overflow Windows' `MAX_PATH`, which could make a dependency's lifecycle script fail with `spawn cmd.exe ENOENT`.
- [`05b95ab`](https://github.com/pnpm/pnpm/commit/05b95ab): Fixed `pnpm` hanging (and crashing with an unhandled promise rejection) when a non-retryable network error such as `SELF_SIGNED_CERT_IN_CHAIN` occurs while fetching from a registry. The error is now rejected through the returned promise instead of being thrown inside the detached retry callback.
- [`d3f68e2`](https://github.com/pnpm/pnpm/commit/d3f68e2): Fix a `pnpm audit` performance regression on lockfiles that contain dependency cycles. The reachable-vulnerability pruning added in pnpm 11.5.1 only memoized acyclic subtrees, so any node whose subtree touched a cycle — together with all of its ancestors — was recomputed on every query, making the path walk quadratic. Reachability is now computed once per node using Tarjan's strongly-connected-components algorithm, so cyclic graphs are handled in linear time [#​12212](https://github.com/pnpm/pnpm/issues/12212).
The audit path walk also no longer recurses, so a deeply nested dependency graph can no longer overflow the call stack, and the install path to each finding is tracked without per-node copying, keeping memory linear in the graph depth.
- [`322f88f`](https://github.com/pnpm/pnpm/commit/322f88f): Fix failed optional dependency updates so they don't rewrite unrelated dependency specs [#​11267](https://github.com/pnpm/pnpm/issues/11267).
- [`1488db1`](https://github.com/pnpm/pnpm/commit/1488db1): When `enableGlobalVirtualStore` is toggled on for a project that was previously installed without it, stale hoisted symlinks under `node_modules/.pnpm/node_modules` are now replaced instead of being left pointing at the old per-project virtual store location [#​9739](https://github.com/pnpm/pnpm/issues/9739).
- [`6545793`](https://github.com/pnpm/pnpm/commit/6545793): Fixed `pnpm install --ignore-workspace` overwriting the `allowBuilds` map in `pnpm-workspace.yaml`. The ignored builds of a package with a build script were auto-populated into `allowBuilds` even though `--ignore-workspace` was passed, clobbering committed `true`/`false` values with the `set this to true or false` placeholder [#​12469](https://github.com/pnpm/pnpm/issues/12469).
- [`fbdc0eb`](https://github.com/pnpm/pnpm/commit/fbdc0eb): Fixed `minimumReleaseAgeExclude` and `trustPolicyExclude` so multiple exact-version entries for the same package behave the same as a single `||` disjunction entry. Previously only the first matching rule's versions were honored, so a config like `[[email protected], [email protected]]` could still flag `[email protected]` as violating `minimumReleaseAge`, while `[[email protected] || 2.5.6]` worked as expected [#​12463](https://github.com/pnpm/pnpm/issues/12463).
- [`fa7004b`](https://github.com/pnpm/pnpm/commit/fa7004b): The in-memory package metadata cache is now populated on the exact-version disk fast path, so repeated resolutions of the same package within one install no longer re-read and re-parse the on-disk metadata. In large monorepos this brings the time for adding a new package down from minutes to seconds. The in-memory cache key now also includes the registry, so a package of the same name served by two different registries in a single install can no longer share a cache slot and resolve the wrong tarball.
- [`0a154b1`](https://github.com/pnpm/pnpm/commit/0a154b1): Fixed `pnpm patch` dropping the package name (and leaking internal option fields) when the patched dependency resolves to a single git-hosted version.
- [`4d3fe4b`](https://github.com/pnpm/pnpm/commit/4d3fe4b): The pnpr resolver endpoints moved under the reserved `/-/pnpr` namespace: `POST /v1/resolve` is now `POST /-/pnpr/v0/resolve` and `POST /v1/verify-lockfile` is now `POST /-/pnpr/v0/verify-lockfile`. The capability handshake at `GET /-/pnpr` advertises protocol version `0` to match. This keeps every pnpr-proprietary route in npm's reserved namespace, so it can never collide with a package path.
- [`0ec878d`](https://github.com/pnpm/pnpm/commit/0ec878d): Removing a runtime dependency now removes the matching `devEngines.runtime` or `engines.runtime` entry that was materialized from it. Blank runtime selectors are normalized to `latest`.
- [`17e7f2c`](https://github.com/pnpm/pnpm/commit/17e7f2c): `pnpm sbom` now emits a CycloneDX `issue-tracker` external reference for components (and the root) whose `package.json` declares a `bugs` URL. Email-only `bugs` entries are skipped, since the reference requires a URL.
- [`a84d2a1`](https://github.com/pnpm/pnpm/commit/a84d2a1): Add `@pnpm/resolving.tarball-url`, which builds and recognizes the canonical npm tarball URL of a package. It vendors `getNpmTarballUrl` (previously the external `get-npm-tarball-url` package) and adds `isCanonicalRegistryTarballUrl`, the predicate the lockfile writer uses to decide whether a tarball URL is derivable from name+version+registry (and can therefore be omitted from `pnpm-lock.yaml`).
Exposing `isCanonicalRegistryTarballUrl` lets a custom resolver (pnpmfile `resolvers`) fronting a proxy that serves tarballs on a non-canonical path (e.g. an ephemeral `localhost:<port>`) rewrite the resolved tarball to the canonical form, so nothing host-specific is persisted to the lockfile. Previously this logic was private to `@pnpm/lockfile.utils`.
Two correctness fixes are included while consolidating the logic: the scoped-package unescape now handles uppercase `%2F` as well as `%2f` (percent-encoding is case-insensitive), and protocol-insensitive comparison strips only a leading `http(s)://` scheme instead of splitting on the first `://` (which could truncate URLs containing a later `://`).
- [`852d537`](https://github.com/pnpm/pnpm/commit/852d537): Lockfile verification no longer reports a registry metadata fetch failure (for example a `403`/`401` on a private registry, or a network error) as `ERR_PNPM_TARBALL_URL_MISMATCH`. When the registry can't be reached to verify an entry, the install now aborts with the registry's own fetch error (such as `ERR_PNPM_FETCH_403`, which already explains the authentication situation) instead of mislabeling a transport failure as lockfile tampering. Registry fetch errors no longer leak basic-auth credentials embedded in the registry URL (`https://user:pass@host/`) into their message.
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
### [`v11.8.0`](https://github.com/pnpm/pnpm/releases/tag/v11.8.0): pnpm 11.8
[Compare Source](https://github.com/pnpm/pnpm/compare/v11.7.0...v11.8.0)
#### Minor Changes
- [`c112b61`](https://github.com/pnpm/pnpm/commit/c112b61): Added a `--dry-run` option to `pnpm install`. It runs a full dependency resolution and reports what an install would change, but writes nothing to disk (no lockfile, no `node_modules`) and always exits with code 0. This mirrors the preview semantics of `npm install --dry-run` [#​7340](https://github.com/pnpm/pnpm/issues/7340).
- [`179ebc4`](https://github.com/pnpm/pnpm/commit/179ebc4): `pnpm run --no-bail` now exits with a non-zero exit code when any of the executed scripts fail, while still running every matched script to completion. This makes the exit-code behavior of `--no-bail` consistent between recursive and non-recursive runs (recursive runs already failed at the end). Previously, a non-recursive `pnpm run --no-bail` always exited with code 0, even when a script failed [#​8013](https://github.com/pnpm/pnpm/issues/8013).
- [`0474a9c`](https://github.com/pnpm/pnpm/commit/0474a9c): Added support for generating Node.js package maps at `node_modules/.package-map.json` during isolated and hoisted installs. Added the `node-experimental-package-map` setting to inject the generated map into pnpm-managed Node.js script environments, and the `node-package-map-type` setting to choose between `standard` and `loose` package maps.
- [`dcededc`](https://github.com/pnpm/pnpm/commit/dcededc): `pnpm sbom` now marks components reachable only through `devDependencies` with CycloneDX `scope: "excluded"` and the `cdx:npm:package:development` property. The `excluded` scope documents "component usage for test and other non-runtime purposes", which matches the semantics of a devDependency; the property is the CycloneDX npm-taxonomy marker emitted by `@cyclonedx/cyclonedx-npm`, so both modern (scope) and existing (property) consumers are covered. Components reachable at runtime (including installed `optionalDependencies`) omit `scope` and default to `required`.
- [`1495cb0`](https://github.com/pnpm/pnpm/commit/1495cb0): Added per-package SBOM generation with `--out` and `--split` flags. Use `--out out/%s.cdx.json` to write one SBOM per workspace package to individual files, or `--split` for NDJSON output to stdout. When `--filter` selects a single package, the SBOM root component now uses that package's metadata. Workspace inter-dependencies (`workspace:` protocol) and their transitive dependencies are included. Author, repository, and license fall back to the root manifest when the package doesn't define them.
- [`293921a`](https://github.com/pnpm/pnpm/commit/293921a): feat(view): support searching project manifest upward when package name is omitted
When running `pnpm view` without a package name, the command now searches
upward for the nearest project manifest (`package.json`, `package.yaml`, or `package.json5`) and uses its `name` field.
If the manifest exists but lacks a `name` field, an error is thrown.
This change also replaces the `find-up` dependency with `empathic` for
improved performance and consistency across workspace tools.
#### Patch Changes
- [`29ab905`](https://github.com/pnpm/pnpm/commit/29ab905): Fixed `pnpm update` overriding the version range policy of a named catalog whose name parses as a version (e.g. `catalog:express4-21`). The `catalog:` reference carries no pinning of its own, so the prefix from the catalog entry (such as `~`) is now preserved instead of being widened to `^` [#​10321](https://github.com/pnpm/pnpm/issues/10321).
- [`bee4bf4`](https://github.com/pnpm/pnpm/commit/bee4bf4): Security: validate config dependency names and versions from the env lockfile (`pnpm-lock.yaml`) before using them to build filesystem paths. A committed lockfile with a traversal-shaped `configDependencies` name (such as `../../PWNED`) or version (such as `../../../PWNED`) could previously cause `pnpm install` to create symlinks or write package files outside `node_modules/.pnpm-config` and the store. Names must now be valid npm package names and versions must be exact semver versions; the same validation is applied to optional subdependencies of config dependencies, and to the legacy workspace-manifest format before any lockfile is written. See [GHSA-qrv3-253h-g69c](https://github.com/pnpm/pnpm/security/advisories/GHSA-qrv3-253h-g69c).
- [`96bdd57`](https://github.com/pnpm/pnpm/commit/96bdd57): Fix `link:` workspace protocol switching to `file:` after `pnpm rm` is run from inside a workspace package whose target workspace dependency has its own dependencies, when `injectWorkspacePackages: true` is set. Follow-up to [#​10575](https://github.com/pnpm/pnpm/pull/10575), which fixed the same symptom for workspace packages without dependencies.
- [`302a2f7`](https://github.com/pnpm/pnpm/commit/302a2f7): No longer warn about using both `packageManager` and `devEngines.packageManager` when the two fields pin the same package manager at the same version with the same integrity hash (e.g. both `[email protected]+sha512.…`). Previously the hash was stripped from the legacy `packageManager` field but not from `devEngines.packageManager`, so even identical specifications looked like a mismatch [#​12028](https://github.com/pnpm/pnpm/issues/12028).
The warning still fires on any genuine divergence, and several cases now state the specific reason instead of a single generic message: a different package manager, a different version, or contradictory integrity hashes for the same version.
- [`3f0fb21`](https://github.com/pnpm/pnpm/commit/3f0fb21): Fixed the progress line showing leftover characters from external processes that write to the terminal between progress updates (e.g. an SSH passphrase prompt would leave a fragment like `added 0sa':`). The interactive reporter now redraws each frame in place, erasing to the end of the display before reprinting, so any such remnants are cleared [#​12350](https://github.com/pnpm/pnpm/issues/12350).
- [`564619f`](https://github.com/pnpm/pnpm/commit/564619f): Fixed `pnpm approve-builds` reporting "no packages awaiting approval" when a build-script dependency whose approval was revoked (e.g. after `git stash` drops the `allowBuilds` from `pnpm-workspace.yaml`) is re-added. The revoked packages are now correctly recorded in `.modules.yaml` so `approve-builds` can find them. [#​12221](https://github.com/pnpm/pnpm/issues/12221)
- [`3d1fd20`](https://github.com/pnpm/pnpm/commit/3d1fd20): Skip the redundant "target bin directory already contains an exe called node" warning on Windows when the existing `node.exe` already matches the target (same hard link or identical content) [pnpm/pnpm#12203](https://github.com/pnpm/pnpm/issues/12203).
- [`1b02b47`](https://github.com/pnpm/pnpm/commit/1b02b47): Fix macOS Gatekeeper blocking native binaries (`.node`, `.dylib`, `.so`) by removing the `com.apple.quarantine` extended attribute after importing them from the store.
When pnpm imports files from its content-addressable store into `node_modules`, macOS preserves extended attributes, including `com.apple.quarantine`. If this xattr is present on a store blob (e.g. it was first written under a Gatekeeper-enabled app such as a Git client), it propagates to `node_modules`, and Gatekeeper blocks the native binary from loading even though pnpm already verified the file's integrity against the lockfile.
After importing a package, pnpm now strips `com.apple.quarantine` from its native binaries, matching Homebrew's behaviour of dropping quarantine from verified downloads. The cleanup is macOS-only, runs in a single batched `xattr` call per package, is restricted to native binaries (other files are untouched), and is non-fatal (it logs a warning on unexpected errors).
Fixes [#​11056](https://github.com/pnpm/pnpm/issues/11056)
- [`61969fb`](https://github.com/pnpm/pnpm/commit/61969fb): Fix `pnpm install` with `optimisticRepeatInstall` incorrectly reporting `Already up to date` when `pnpm-lock.yaml` changed but project manifests did not. This affected workflows such as checking out or restoring only the lockfile [#​12100](https://github.com/pnpm/pnpm/issues/12100).
Also fixes `checkDepsStatus` to use the correct lockfile path when `useGitBranchLockfile` is enabled, so the optimistic fast-path and lockfile modification detection work with `pnpm-lock.<branch>.yaml` files instead of always stat'ing `pnpm-lock.yaml`. Merge-conflict detection now reads the resolved lockfile name as well, and with `mergeGitBranchLockfiles` enabled every `pnpm-lock.*.yaml` is scanned for modifications and conflicts. The git branch is now resolved by reading `.git/HEAD` directly (no process spawn) and uses the workspace directory rather than `process.cwd()`.
- [`5c12968`](https://github.com/pnpm/pnpm/commit/5c12968): Fix recursive updates of transitive dependencies when the update command mixes transitive dependency patterns with direct dependency selectors. For example, `pnpm up -r "@​babel/core" uuid` now updates matching transitive `@babel/core` dependencies even when `uuid` is a direct dependency selector [#​12103](https://github.com/pnpm/pnpm/issues/12103).
- [`9d79ba1`](https://github.com/pnpm/pnpm/commit/9d79ba1): Register the `pnpm update --no-save` flag in the CLI help and option parser.
- [`0474a9c`](https://github.com/pnpm/pnpm/commit/0474a9c): Fixed `pnpm import` for Yarn v2 lockfiles when `js-yaml` v4 is installed.
- [`9e0c375`](https://github.com/pnpm/pnpm/commit/9e0c375): Fixed `pnpm install` repeatedly prompting to remove and reinstall `node_modules` in a workspace package when `enableGlobalVirtualStore` is enabled. The post-install build step recorded a per-project `node_modules/.pnpm` virtual store directory in `node_modules/.modules.yaml`, overwriting the global `<storeDir>/links` value the install step had written. The next install then detected a virtual-store mismatch (`ERR_PNPM_UNEXPECTED_VIRTUAL_STORE`). The build step now derives the same global virtual store directory as the install step [#​12307](https://github.com/pnpm/pnpm/issues/12307).
- [`223d060`](https://github.com/pnpm/pnpm/commit/223d060): Document the `--cpu`, `--os` and `--libc` flags in the output of `pnpm install --help`. These flags were already supported but were only documented on the website [#​12359](https://github.com/pnpm/pnpm/issues/12359).
- [`e85aea2`](https://github.com/pnpm/pnpm/commit/e85aea2): Avoid reading `README.md` from disk when publishing if the publish manifest already provides a `readme` field. The README is now only read lazily, inside `createExportableManifest`, when it is actually needed.
- [`3188ae7`](https://github.com/pnpm/pnpm/commit/3188ae7): Fixed `pnpm peers check` to accept loose peer dependency ranges such as `>=3.16.0 || >=4.0.0-` when the installed peer version satisfies the range [#​12149](https://github.com/pnpm/pnpm/issues/12149).
- [`531f2a3`](https://github.com/pnpm/pnpm/commit/531f2a3): Fixed `pnpm update` rewriting a `workspace:` dependency that points at a local path (e.g. `workspace:../packages/foo/dist`) into a normalized `link:` or version-range specifier. Such specifiers are now preserved verbatim when the workspace protocol is preserved [#​3902](https://github.com/pnpm/pnpm/issues/3902).
- [`fe66535`](https://github.com/pnpm/pnpm/commit/fe66535): Fixed a lockfile non-convergence bug where an incremental install kept a duplicate transitive dependency that a fresh install would not produce. When a package is reused from the lockfile, its child edges are taken verbatim and bypass the preferred-versions walk, so a transitive dependency could stay pinned to an older version even after a direct dependency resolved to a higher version that satisfies the same range. The resolver now refreshes such a stale pin to the higher direct-dependency version during resolution — so the older version is never resolved or fetched, and the incremental result converges to the fresh one.
- [`6d35338`](https://github.com/pnpm/pnpm/commit/6d35338): `pnpm install` detects changes inside local file dependencies again. The optimistic repeat-install fast path only tracks manifest and lockfile modification times, so edits inside a local dependency's directory (or a repacked local tarball) were reported as "Already up to date". Projects with local file dependencies (`file:` and bare local path or tarball specifiers, declared directly or through `pnpm.overrides`) now always run a full install, which refetches those dependencies, matching pnpm v10 behavior [#​11795](https://github.com/pnpm/pnpm/issues/11795).
- [`4ca9247`](https://github.com/pnpm/pnpm/commit/4ca9247): Preserve the existing Node.js runtime version prefix when resolving `node@runtime:<range>` to a concrete version.
- [`30c7590`](https://github.com/pnpm/pnpm/commit/30c7590): Create shorter CAFS temporary package directories to leave room for lifecycle scripts that create IPC socket paths under TMPDIR.
- [`13815ad`](https://github.com/pnpm/pnpm/commit/13815ad): Reporter output (warnings, progress) for `pnpm store` and `pnpm config` subcommands now goes to stderr instead of stdout. This fixes scripts that capture their stdout (e.g. `PNPM_STORE=$(pnpm store path)`, `pnpm config list --json | jq`) from getting warnings mixed into the result.
- [`1c05876`](https://github.com/pnpm/pnpm/commit/1c05876): Avoid relinking unchanged child dependencies and remove stale child links during warm installs.
- [`817f99d`](https://github.com/pnpm/pnpm/commit/817f99d): Fixed lockfile churn where a package's `transitivePeerDependencies` could be dropped (and shift between packages) when the package participates in a dependency cycle. A cycle re-entry resolves against truncated children, so it must not be cached as "pure"; otherwise sibling occurrences of the same package short-circuit and lose transitive peers depending on traversal order [#​5108](https://github.com/pnpm/pnpm/issues/5108).
- [`eba03e0`](https://github.com/pnpm/pnpm/commit/eba03e0): Fix `pnpm install` reporting "Already up to date" after a catalog entry in `pnpm-workspace.yaml` was reverted to a previous version. After an update modified a catalog, the workspace state cache stored the pre-update catalog versions, so reverting the entry back to its original version was not detected as an outdated state [#​12418](https://github.com/pnpm/pnpm/issues/12418).
- [`3b54d79`](https://github.com/pnpm/pnpm/commit/3b54d79): `pnpm update` now keeps lockfile `overrides` that resolve through a catalog in sync with the catalog. Previously, when an override referenced a catalog (e.g. `overrides: { foo: 'catalog:' }`) and `pnpm update` bumped that catalog entry, the lockfile's `catalogs` advanced while the resolved `overrides` kept the old version. The resulting lockfile was internally inconsistent, so a later `pnpm install --frozen-lockfile` failed with `ERR_PNPM_LOCKFILE_CONFIG_MISMATCH`.
- [`9d0a300`](https://github.com/pnpm/pnpm/commit/9d0a300): Fixed `pnpm version --recursive` so it honors the workspace selection. In recursive mode the version bump now applies to the packages resolved from the workspace filter (`selectedProjectsGraph`), matching the behavior of `pnpm publish --recursive`, instead of always bumping every workspace package [#​11348](https://github.com/pnpm/pnpm/issues/11348).
<!-- sponsors -->
#### Platinum Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://bit.cloud/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/bit.svg" width="80" alt="Bit"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://openai.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/openai_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/openai_light.svg" />
<img src="https://pnpm.io/img/users/openai_dark.svg" width="160" alt="OpenAI" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
#### Gold Sponsors
<table>
<tbody>
<tr>
<td align="center" valign="middle">
<a href="https://sanity.io/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/sanity.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/sanity_light.svg" />
<img src="https://pnpm.io/img/users/sanity.svg" width="120" alt="Sanity" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://discord.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/discord.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/discord_light.svg" />
<img src="https://pnpm.io/img/users/discord.svg" width="220" alt="Discord" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://vite.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer"><img src="https://pnpm.io/img/users/vitejs.svg" width="42" alt="Vite"></a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://serpapi.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/serpapi_dark.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/serpapi_light.svg" />
<img src="https://pnpm.io/img/users/serpapi_dark.svg" width="160" alt="SerpApi" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://coderabbit.ai/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/coderabbit.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/coderabbit_light.svg" />
<img src="https://pnpm.io/img/users/coderabbit.svg" width="220" alt="CodeRabbit" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://stackblitz.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/stackblitz.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/stackblitz_light.svg" />
<img src="https://pnpm.io/img/users/stackblitz.svg" width="190" alt="Stackblitz" />
</picture>
</a>
</td>
</tr>
<tr>
<td align="center" valign="middle">
<a href="https://workleap.com/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/workleap.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/workleap_light.svg" />
<img src="https://pnpm.io/img/users/workleap.svg" width="190" alt="Workleap" />
</picture>
</a>
</td>
<td align="center" valign="middle">
<a href="https://nx.dev/?utm_source=pnpm&utm_medium=release_notes" target="_blank" rel="noopener noreferrer">
<picture>
<source media="(prefers-color-scheme: light)" srcset="https://pnpm.io/img/users/nx.svg" />
<source media="(prefers-color-scheme: dark)" srcset="https://pnpm.io/img/users/nx_light.svg" />
<img src="https://pnpm.io/img/users/nx.svg" width="50" alt="Nx" />
</picture>
</a>
</td>
</tr>
</tbody>
</table>
<!-- sponsors end -->
</details>
---
### Configuration
📅 **Schedule**: (UTC)
- Branch creation
- At any time (no schedule defined)
- Automerge
- At any time (no schedule defined)
🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied.
♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
🔕 **Ignore**: Close this PR and you won't be reminded about this update again.
---
- [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check this box
---
This PR has been generated by [Mend Renovate](https://github.com/renovatebot/renovate).
<!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0My4xOTEuMiIsInVwZGF0ZWRJblZlciI6IjQzLjE5MS4yIiwidGFyZ2V0QnJhbmNoIjoibWFpbiIsImxhYmVscyI6W119-->
Reviewed-on: https://gitea.com/gitea/docs/pulls/446
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Renovate Bot <[email protected]>
Co-committed-by: Renovate Bot <[email protected]>
Adds a versioned docs snapshot for Gitea Runner v1.0.8 (latest stable),
updates lastVersion and navbar to point to the new version while keeping
0.2.11 in the dropdown for historical reference.
---------
Co-authored-by: Lunny Xiao <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/435
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Nicolas <[email protected]>
Co-committed-by: Nicolas <[email protected]>
Adds documentation for the `DEFAULT_TITLE_SOURCE` setting.
- Added `DEFAULT_TITLE_SOURCE` to the config cheat sheet under `[repository.pull-request]`
- Added a "Default pull request title" section to the pull request usage docs explaining both modes (`first-commit` and `auto`)
---------
Co-authored-by: Nicolas <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/391
Reviewed-by: Nicolas <[email protected]>
Co-authored-by: 0xGREG <[email protected]>
Co-committed-by: 0xGREG <[email protected]>
Sync the zh-cn and zh-tw `config-cheat-sheet.md` translations with the English source for current and 1.26 versions.
Changes:
- Add 17 missing sections (`qos`, `mailer.override_header`, `markup.external-render-name`, `lfs_client`, `storage_minio`, `storage_azureblob`, `storage_override`, and 10 cron task subsections including the `Actions` cron tasks group)
- Add ~30 missing individual settings across existing sections
- Restructure storage sections to match English (separate Minio/Azure/Override subsections)
- Remove stale settings (`TEST_CONFLICTING_PATCHES_WITH_GIT_APPLY`, old inline MINIO duplicates, removed git timeout settings)
- Fix `MERMAID_MAX_SOURCE_CHARACTERS` default value (5000 -> 50000)
- Replace full-width Unicode punctuation (U+FF1A, U+FF08, U+FF09) with ASCII equivalents
All 4 i18n files now have 745 settings matching the English source.
---
This PR was written with the help of Claude Opus 4.6
Reviewed-on: https://gitea.com/gitea/docs/pulls/381
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: silverwind <[email protected]>
Co-committed-by: silverwind <[email protected]>
Add documentation for the new `REVERSE_PROXY_LOGOUT_REDIRECT` setting in the `[security]` section, which allows redirecting users to an external URL or relative path after logout when using reverse proxy authentication.
For Gitea 1.26, ref https://github.com/go-gitea/gitea/pull/36085.
---
This PR was written with the help of Claude Opus 4.6
Reviewed-on: https://gitea.com/gitea/docs/pulls/380
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: silverwind <[email protected]>
Co-committed-by: silverwind <[email protected]>
This PR contains the following updates:
| Package | Type | Update | Change |
|---|---|---|---|
| [pnpm/action-setup](https://github.com/pnpm/action-setup) | action | major | `v4` → `v5` |
---
### Release Notes
<details>
<summary>pnpm/action-setup (pnpm/action-setup)</summary>
### [`v5`](https://github.com/pnpm/action-setup/compare/v4...v5)
[Compare Source](https://github.com/pnpm/action-setup/compare/v4...v5)
</details>
---
### Configuration
📅 **Schedule**: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined).
🚦 **Automerge**: Disabled by config. Please merge this manually once you are satisfied.
♻ **Rebasing**: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.
🔕 **Ignore**: Close this PR and you won't be reminded about this update again.
---
- [ ] <!-- rebase-check -->If you want to rebase/retry this PR, check this box
---
This PR has been generated by [Renovate Bot](https://github.com/renovatebot/renovate).
<!--renovate-debug:eyJjcmVhdGVkSW5WZXIiOiI0My43Ni41IiwidXBkYXRlZEluVmVyIjoiNDMuNzYuNSIsInRhcmdldEJyYW5jaCI6Im1haW4iLCJsYWJlbHMiOltdfQ==-->
---------
Co-authored-by: silverwind <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/365
Reviewed-by: silverwind <[email protected]>
Reviewed-by: Nicolas <[email protected]>
Co-authored-by: Renovate Bot <[email protected]>
Co-committed-by: Renovate Bot <[email protected]>
The standalone `environment-to-ini` tool was removed in https://github.com/go-gitea/gitea/pull/35735 and its functionality is now built into the Gitea binary. This PR:
- Removes outdated links to the removed `contrib/environment-to-ini` directory
- Updates the "Use environment variables to setup Gitea" section in the config cheat sheet
- Adds examples showing how env vars map to `app.ini` settings and the `__FILE` suffix for secrets
*This PR was authored by Claude.*
Reviewed-on: https://gitea.com/gitea/docs/pulls/355
Reviewed-by: wxiaoguang <[email protected]>
Reviewed-by: TheFox0x7 <[email protected]>
Fix act runner doc typos and errors and simplify some grammar.
Co-authored-by: Peter Ryszkiewicz <[email protected]>
Co-committed-by: Peter Ryszkiewicz <[email protected]>
Original Gitea's repository no longer contains "Environment to INI" document referenced in "Managing Deployments With Environment Variables" section.
This commit fixes the reference by pointing the link to the last commit where needed doc was still present.
Reviewed-on: https://gitea.com/gitea/docs/pulls/315
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: andriibratanin <[email protected]>
Co-committed-by: andriibratanin <[email protected]>
This adds explanations, that were missing regarding Token scopes and the packages feature requiring a token with said scope. Also added a missing explanation, that tokens can be created in the user interface, not just the API. The reason this PR is created was due to misunderstandings around the package feature, that arose during the discussion in https://github.com/go-gitea/gitea/issues/32048 .
This PR aims to clarify the misunderstood points around tokens, scopes and the package feature.
During the creation of this PR bug https://github.com/go-gitea/gitea/issues/32078 was found, the scopes are always reported as `null` during the response of the initial `POST` to create the token.
Co-authored-by: Wladislav ヴラド Artsimovich <[email protected]>
Co-authored-by: Lunny Xiao <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/72
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: FrostKiwi <[email protected]>
Co-committed-by: FrostKiwi <[email protected]>
The SMTPS protocol uses port 465, not 587. Port 587 is instead used for STARTTLS.
Helps getting administrators stuck in cryptic errors such as:
Failed to send a testing email to "[email protected]": could not initiate SMTP session: tls: first record does not look like a TLS handshake
Reviewed-on: https://gitea.com/gitea/docs/pulls/295
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: kairosys-dev <[email protected]>
Co-committed-by: kairosys-dev <[email protected]>
Fix JSON syntax error in token creation request
The previous JSON was malformed:
{"name":test_token","scopes":[...]}
The value of "name" was missing opening quotes, causing the error:
invalid character 't' looking for beginning of value
Fixed by enclosing the string in double quotes:
{"name":"test_token","scopes":[...]}
Reviewed-on: https://gitea.com/gitea/docs/pulls/269
Reviewed-by: techknowlogick <[email protected]>
Co-authored-by: desysoft <[email protected]>
Co-committed-by: desysoft <[email protected]>
Added instructions on how to download the signature file using wget, and fixed a typo in the wget commands that would have caused the command to fail when copied/pasted.
Reviewed-on: https://gitea.com/gitea/docs/pulls/262
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: kmanwar89 <[email protected]>
Co-committed-by: kmanwar89 <[email protected]>
Documentation for my changes to ssh commit signing in <https://github.com/go-gitea/gitea/pull/34341>.
* Renames "GPG Commit Signatures" to "GPG/SSH Commit Signatures", since the url is called just signing and creating a new page would duplicate a lot
* _"The default option and repository specific signing keys are not supported for ssh keys"_
* the original draft implementation did support this for ssh keys
* this warning should avoid people expecting this gitconfig style things to work for ssh keys, since verifying is disabled in this case (e.g. pre 1.25 undocumented behavior preserved).
Co-authored-by: Lunny Xiao <[email protected]>
Reviewed-on: https://gitea.com/gitea/docs/pulls/241
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: Christopher Homberger <[email protected]>
Co-committed-by: Christopher Homberger <[email protected]>
## Description
Fix step 1 of the [Database Preparation](https://docs.gitea.com/installation/database-prep) docs.
## Changes
When a user modifies the */etc/mysql/my.cnf* config file as instructed in step 1 of the database preparation guide, they get an error message from MySQL when trying to log in as the root user after saving changes.
```shell
mysql: [ERROR] Found option without preceding group in config file /etc/mysql/my.cnf at line 37.
mysql: [ERROR] Fatal error in defaults handling. Program aborted!
```
The solution is to include precede `bind-address` with a group like this.
```ini
[mysqld]
bind-address = 203.0.113.3
```
Reviewed-on: https://gitea.com/gitea/docs/pulls/173
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: b-nagaj <[email protected]>
Co-committed-by: b-nagaj <[email protected]>
I can not understand the old logic:
- download the web page (not the raw data) from GitHub
- replace `AppVer`
- remove the file???
It does nothing.
The correct logic:
- download the raw file from GitHub
- replace `AppVer` and `AppSubUrl`
- move the file to the correct place
As the script is not correct for a long time, added the support for all version since 1.19.
Fix#142
Reviewed-on: https://gitea.com/gitea/docs/pulls/150
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: yp05327 <[email protected]>
Co-committed-by: yp05327 <[email protected]>
"body": "Automated update of the generated runner reference pages from gitea/runner (main for the develop docs, the newest stable tag for every documented series), opened by the \"update runner reference\" scheduled workflow."
"body": "Automated update of static/swagger-latest.json and static/openapi3-latest.json from the gitea main branch, opened by the \"update api spec files\" scheduled workflow."
"_comment":"Algolia DocSearch crawler configuration. The three facets come from the docsearch:* meta tags every page carries; only the versions people read are indexed, the older ones stay reachable through the version picker.",
@@ -12,10 +12,10 @@ Some jurisdictions (such as EU), requires certain legal pages (e.g. Privacy Poli
## Getting Pages
Gitea source code ships with sample pages, available in `contrib/legal` directory. Copy them to `custom/public/assets/`. For example, to add Privacy Policy:
Gitea source code ships with sample pages, available in `contrib/service/sample-page` directory. Copy them to `custom/public/assets/`. For example, to add Privacy Policy:
Now you need to edit the page to meet your requirements. In particular you must change the email addresses, web addresses and references to "Your Gitea Instance" to match your situation.
Disabling `synchronous_commit` makes PostgreSQL less resilient to crashes, but makes the import a lot faster. Since we already have a backup of the original database and we can check that the import completes successfully, it should be a good trade-off.
After the import is completed, set up Gitea to use PostgreSQL and start the Gitea server again. Good luck!
After the import is completed, set up Gitea to use PostgreSQL and start the Gitea server again. Good luck!
@@ -34,13 +34,44 @@ A full restart is required for Gitea configuration changes to take effect.
## Use environment variables to setup Gitea
There is [environment-to-ini](https://github.com/go-gitea/gitea/tree/main/contrib/environment-to-ini) to help to
generate Gitea's `app.ini` from environment variables.
Gitea supports setting `app.ini` values via environment variables of the form `GITEA__<section>__<KEY>`. These require running `gitea config edit-ini --in-place --apply-env` before starting Gitea to write them into `app.ini`. The official Docker images do this automatically on container startup.
:::info
If `<section>` or `<KEY>` contains a `.`, replace that with `_0X2E_`; if it contains a `-`, replace it with `_0X2C_`
:::
For example, to configure the database connection via environment variables:
```
GITEA__database__DB_TYPE=postgres
GITEA__database__HOST=127.0.0.1:5432
GITEA__database__NAME=gitea
GITEA__database__USER=gitea
GITEA__database__PASSWD=secret
```
This is equivalent to the following `app.ini` configuration:
```ini
[database]
DB_TYPE = postgres
HOST = 127.0.0.1:5432
NAME = gitea
USER = gitea
PASSWD = secret
```
You can also load values from files by appending `__FILE` to the variable name. This is useful for Docker secrets:
These values are environment-dependent but form the basis of a lot of values. They will be
reported as part of the default configuration when running `gitea help` or on start-up.
reported as part of the default configuration when running `gitea help` or on start-up.
The order they are emitted there is slightly different but we will list them here in the order they are set-up.
- _`AppPath`_: This is the absolute path of the running gitea binary.
@@ -80,26 +111,32 @@ In addition, there is _`StaticRootPath`_ which can be set as a built-in at build
A relative path is interpreted as **`{AppWorkPath}/{ROOT}`**.
- `SCRIPT_TYPE`: **bash**: The script type this server supports. Usually this is `bash`,
but some users report that only `sh` is available.
- `DEFAULT_OBJECT_FORMAT`: **sha1**: The default object format for new repositories. Valid values are `sha1` and `sha256`.
- `DETECTED_CHARSETS_ORDER`: **UTF-8, UTF-16BE, UTF-16LE, UTF-32BE, UTF-32LE, ISO-8859, windows-1252, ISO-8859, windows-1250, ISO-8859, ISO-8859, ISO-8859, windows-1253, ISO-8859, windows-1255, ISO-8859, windows-1251, windows-1256, KOI8-R, ISO-8859, windows-1254, Shift_JIS, GB18030, EUC-JP, EUC-KR, Big5, ISO-2022, ISO-2022, ISO-2022, IBM424_rtl, IBM424_ltr, IBM420_rtl, IBM420_ltr**: Tie-break order of detected charsets - if the detected charsets have equal confidence, charsets earlier in the list will be chosen in preference to those later. Adding `defaults` will place the unnamed charsets at that point.
- `ANSI_CHARSET`: **_empty_**: Default ANSI charset to override non-UTF-8 charsets to.
- `FORCE_PRIVATE`: **false**: Force every new repository to be private.
- `DEFAULT_PRIVATE`: **last**: Default private when creating a new repository: `last`, `private`, `public`
- `DEFAULT_PUSH_CREATE_PRIVATE`: **true**: Default private when creating a new repository with push-to-create.
- `MAX_CREATION_LIMIT`: **-1**: Global maximum creation limit of repositories per user,
`-1` means no limit.
- `MAX_CREATION_LIMIT`: **-1**: Global maximum creation limit of repositories per user/org,
`-1` means no limit. Acts as a shortcut: sets the default value for both
`USER_MAX_CREATION_LIMIT` and `ORG_MAX_CREATION_LIMIT` when those are not configured.
- `USER_MAX_CREATION_LIMIT`: **-1**: Global maximum creation limit of repositories per user,
applied at creation time. `-1` means no limit. Takes precedence over `MAX_CREATION_LIMIT` when set.
- `ORG_MAX_CREATION_LIMIT`: **-1**: Global maximum creation limit of repositories per organization,
applied at creation time. `-1` means no limit. Takes precedence over `MAX_CREATION_LIMIT` when set.
- `PREFERRED_LICENSES`: **Apache License 2.0,MIT License**: Preferred Licenses to place at
the top of the list. Name must match file name in options/license or custom/options/license.
- `DISABLE_HTTP_GIT`: **false**: Disable the ability to interact with repositories over the
HTTP protocol.
- `USE_COMPAT_SSH_URI`: **false**: Force ssh:// clone url instead of scp-style uri when
default SSH port is used.
- `GO_GET_CLONE_URL_PROTOCOL`: **https**: Value for the "go get" request returns the repository url as https or ssh
default is https.
- `GO_GET_CLONE_URL_PROTOCOL`: **_empty_**: Scheme of the returned URL for the "go get" response.
Default is "https" if DISABLE_HTTP_GIT=false or SSH is disabled, otherwise "ssh".
- `ACCESS_CONTROL_ALLOW_ORIGIN`: **_empty_**: Value for Access-Control-Allow-Origin header,
default is not to present.
:::warning
This maybe harmful to you website if you do not give it a right value.
This maybe harmful to your website if you do not give it a right value.
:::
- `DEFAULT_CLOSE_ISSUES_VIA_COMMITS_IN_ANY_BRANCH`: **false**: Close an issue if a commit on a non default branch marks it as closed.
@@ -118,7 +155,7 @@ In addition, there is _`StaticRootPath`_ which can be set as a built-in at build
- `ALLOW_DELETION_OF_UNADOPTED_REPOSITORIES`: **false**: Allow non-admin users to delete unadopted repositories
- `ALLOW_FORK_WITHOUT_MAXIMUM_LIMIT`: **true**: Allow to fork repositories without maximum number limit
- `ALLOW_FORK_INTO_SAME_OWNER`: **false**: Allow to fork repositories into the same owner (user or organization). This feature is experimental, not fully tested, and may be changed in the future
- `ALLOW_FORK_INTO_SAME_OWNER`: **false**: Allow to fork repositories into the same owner (user or organization).
### Repository - Editor (`repository.editor`)
@@ -141,8 +178,12 @@ In addition, there is _`StaticRootPath`_ which can be set as a built-in at build
- `DEFAULT_MERGE_MESSAGE_OFFICIAL_APPROVERS_ONLY`: **true**: In default merge messages only include approvers who are officially allowed to review.
- `POPULATE_SQUASH_COMMENT_WITH_COMMIT_MESSAGES`: **false**: In default squash-merge messages include the commit message of all commits comprising the pull request.
- `ADD_CO_COMMITTER_TRAILERS`: **true**: Add co-authored-by and co-committed-by trailers to merge commit messages if committer does not match author.
- `TEST_CONFLICTING_PATCHES_WITH_GIT_APPLY`: **false**: PR patches are tested using a three-way merge method to discover if there are conflicts. If this setting is set to **true**, conflicting patches will be retested using `git apply` - This was the previous behaviour in 1.18 (and earlier) but is somewhat inefficient. Please report if you find that this setting is required.
- `RETARGET_CHILDREN_ON_MERGE`: **true**: Retarget child pull requests to the parent pull request branch target on merge of parent pull request. It only works on merged PRs where the head and base branch target the same repo.
- `DEFAULT_DELETE_BRANCH_AFTER_MERGE`: **false**: Set the default value for "Delete pull request branch after merge by default" for new repositories.
- `DEFAULT_TITLE_SOURCE`: **auto**: Default source for the pull request title when opening a new PR. Valid options:
- `first-commit`: Uses the oldest commit's summary as the title. If there are multiple commits, still uses the first commit's message.
- `auto`: Uses the commit's summary when the PR contains a single commit; when there are multiple commits, converts the branch name into a human-readable title by normalizing separators and casing.
- `branch-name`: Always uses the PR's branch name.
### Repository - Issue (`repository.issue`)
@@ -159,9 +200,11 @@ In addition, there is _`StaticRootPath`_ which can be set as a built-in at build
### Repository - Release (`repository.release`)
- `ALLOWED_TYPES`: **_empty_**: Comma-separated list of allowed file extensions (`.zip`), mime types (`text/plain`) or wildcard type (`image/*`, `audio/*`, `video/*`). Empty value or `*/*` allows all types.
- `ALLOWED_TYPES`: **_empty_**: Comma-separated list of allowed release attachment file extensions (`.zip`), mime types (`text/plain`) or wildcard type (`image/*`, `audio/*`, `video/*`). Empty value or `*/*` allows all types.
- `DEFAULT_PAGING_NUM`: **10**: The default paging number of releases user interface
- For settings related to file attachments on releases, see the `attachment` section.
- `FILE_MAX_SIZE`: **2048**: Max filesize limit for release attachments (MB)
- `MAX_FILES`: **5**: Maximum number of release attachments that can be uploaded at once for a release.
- For other settings related to file attachments on releases, see the `attachment` section.
### Repository - Signing (`repository.signing`)
@@ -209,7 +252,6 @@ The following configuration set `Content-Type: application/vnd.android.package-a
- `MAX_AGE`: **10m**: max time to cache response
- `ALLOW_CREDENTIALS`: **false**: allow request with credentials
- `HEADERS`: **Content-Type,User-Agent**: additional headers that are permitted in requests
- `X_FRAME_OPTIONS`: **SAMEORIGIN**: Set the `X-Frame-Options` header value.
## UI (`ui`)
@@ -224,6 +266,8 @@ The following configuration set `Content-Type: application/vnd.android.package-a
- `DEFAULT_THEME`: **gitea-auto**: Set the default theme for the Gitea installation, custom themes could be provided by `{CustomPath}/public/assets/css/theme-*.css`.
- `SHOW_USER_EMAIL`: **true**: Whether the email of the user should be shown in the Explore Users page.
- `THEMES`: **_empty_**: All available themes by `{CustomPath}/public/assets/css/theme-*.css`. Allow users select personalized themes.
- `FILE_ICON_THEME`: **material**: The icon theme for files (basic/material).
- `FOLDER_ICON_THEME`: **basic**: The icon theme for folders (basic/material).
- `MAX_DISPLAY_FILE_SIZE`: **8388608**: Max size of files to be displayed (default is 8MiB)
- `AMBIGUOUS_UNICODE_DETECTION`: **true**: Detect ambiguous unicode characters in file contents and show warnings on the UI
- `REACTIONS`: All available reactions users can choose on issues/prs and comments
@@ -233,6 +277,7 @@ The following configuration set `Content-Type: application/vnd.android.package-a
- `CUSTOM_EMOJIS`: **gitea, codeberg, gitlab, git, github, gogs**: Additional Emojis not defined in the utf8 standard.
By default, we support Gitea (:gitea:), to add more copy them to public/assets/img/emoji/emoji_name.png and
add it to this config.
- `ENABLED_EMOJIS`: **_empty_**: Comma separated list of enabled emojis, for example: "smile, thumbsup, thumbsdown". Leave it empty to enable all emojis.
- `DEFAULT_SHOW_FULL_NAME`: **false**: Whether the full name of the users should be shown where possible. If the full name isn't set, the username will be used.
- `SEARCH_REPO_DESCRIPTION`: **true**: Whether to search within description at repository search on explore page.
- `ONLY_SHOW_RELEVANT_REPOS`: **false**: Whether to only show relevant repos on the explore page when no keyword is specified and default sorting is used.
@@ -276,17 +321,23 @@ The following configuration set `Content-Type: application/vnd.android.package-a
## Markdown (`markdown`)
- `ENABLE_HARD_LINE_BREAK_IN_COMMENTS`: **true**: Render soft line breaks as hard line breaks in comments, which
means a single newline character between paragraphs will cause a line break and adding
trailing whitespace to paragraphs is not necessary to force a line break.
- `ENABLE_HARD_LINE_BREAK_IN_DOCUMENTS`: **false**: Render soft line breaks as hard line breaks in documents, which
means a single newline character between paragraphs will cause a line break and adding
trailing whitespace to paragraphs is not necessary to force a line break.
- `RENDER_OPTIONS_COMMENT`: **short-issue-pattern, new-line-hard-break**: Customize render options for different contexts.
Set to "none" to disable the defaults, or use comma separated list:
- short-issue-pattern: recognized "#123" issue reference and render it as a link to the issue
- new-line-hard-break: render soft line breaks as hard line breaks, which means a single newline character between
paragraphs will cause a line break and adding trailing whitespace to paragraphs is not
necessary to force a line break.
- `RENDER_OPTIONS_WIKI`: **short-issue-pattern**: see also RENDER_OPTIONS_COMMENT
- `RENDER_OPTIONS_REPO_FILE`: **_empty_**: see also RENDER_OPTIONS_COMMENT
- `CUSTOM_URL_SCHEMES`: Use a comma separated list (ftp,git,svn) to indicate additional
URL hyperlinks to be rendered in Markdown. URLs beginning in http and https are
always displayed. If this entry is empty, all URL schemes are allowed
- `FILE_EXTENSIONS`: **.md,.markdown,.mdown,.mkd,.livemd**: List of file extensions that should be rendered/edited as Markdown. Separate the extensions with a comma. To render files without any extension as markdown, just put a comma.
- `FILE_EXTENSIONS`: **.md,.markdown,.mdown,.mkd,.livemd**: List of file extensions that should be rendered/edited as Markdown.
Separate the extensions with a comma. To render files without any extension as markdown, just put a comma.
- `ENABLE_MATH`: **true**: Enables detection of `$...$`, `$$...$$`, ``` $`...`$$ ``` blocks as math blocks.
- `MATH_CODE_BLOCK_DETECTION`: **inline-dollar,block-dollar**: Enable delimiters for math code block detection.
Set to "none" to disable all, or use comma separated list: inline-dollar, inline-parentheses, block-dollar, block-square-brackets.
Default value follows GitHub.
## Server (`server`)
@@ -297,10 +348,14 @@ The following configuration set `Content-Type: application/vnd.android.package-a
- `PROXY_PROTOCOL_TLS_BRIDGING`: **false**: When protocol is https, expect PROXY protocol headers after TLS negotiation.
- `PROXY_PROTOCOL_HEADER_TIMEOUT`: **5s**: Timeout to wait for PROXY protocol header (set to 0 to have no timeout)
- `PROXY_PROTOCOL_ACCEPT_UNKNOWN`: **false**: Accept PROXY protocol headers with Unknown type.
- `DOMAIN`: **localhost**: Domain name of this server.
This is useful if the internal and the external URL don't match (e.g. in Docker).
This is useful if the internal and the external URL don't match (e.g. behind a reverse proxy).
- `PUBLIC_URL_DETECTION`: **`auto`**: Controls how to generate the public URL. Most instances should use the "auto" behavior,
especially when the Gitea instance needs to be accessed in a container network.
- "legacy": generate the public URL by "Host" header if "X-Forwarded-Proto" header exists, otherwise use "ROOT_URL".
- "auto": always use "Host" header, and also use "X-Forwarded-Proto" header if it exists. If no "Host" header, use "ROOT_URL".
- "never": always use "ROOT_URL", never detect from request headers.
- `STATIC_URL_PREFIX`: **_empty_**:
Overwrite this option to request static resources from a different URL.
This includes CSS files, images, JS files and web fonts.
@@ -348,23 +403,23 @@ The following configuration set `Content-Type: application/vnd.android.package-a
If it is set to `(DOER_USERNAME)`, it will use current signed-in user's username.
This option is only for some advanced users who have configured their SSH reverse-proxy and need to use different usernames for git SSH clone.
Most users should just leave it blank and/or modify the `BUILTIN_SSH_SERVER_USER`.
- `SSH_DOMAIN`: **`{DOMAIN}`**: Domain name of this server, used for displayed clone URL.
- `SSH_PORT`: **22**: SSH port displayed in clone URL.
- `SSH_DOMAIN`: **_empty_**: Domain name to be exposed in clone URL, defaults to the domain part of ROOT_URL
- `SSH_PORT`: **22**: SSH port displayed in clone URL. If you need a different "SSH clone port" from the real "SSH listen port", set the SSH_LISTEN_PORT separately.
- `SSH_LISTEN_HOST`: **0.0.0.0**: Listen address for the built-in SSH server.
- `SSH_LISTEN_PORT`: **`{SSH_PORT}`**: Port for the built-in SSH server.
- `SSH_ROOT_PATH`: **~/.ssh**: Root path of SSH directory.
- `SSH_CREATE_AUTHORIZED_KEYS_FILE`: **true**: Gitea will create a authorized_keys file by default when it is not using the internal ssh server. If you intend to use the AuthorizedKeysCommand functionality then you should turn this off.
- `SSH_ROOT_PATH`: **_empty_**: Root path of SSH user directory for the system's standalone SSH server if Gitea is not using its builtin SSH server. Default is the '.ssh' directory in the run user's home directory.
- `SSH_CREATE_AUTHORIZED_KEYS_FILE`: **true**: Gitea will create an authorized_keys file by default when it is not using the builtin SSH server. If you intend to use the AuthorizedKeysCommand functionality then you should turn this off.
- `SSH_AUTHORIZED_KEYS_BACKUP`: **false**: Enable SSH Authorized Key Backup when rewriting all keys, default is false.
- `SSH_TRUSTED_USER_CA_KEYS`: **_empty_**: Specifies the public keys of certificate authorities that are trusted to sign user certificates for authentication. Multiple keys should be comma separated. E.g.`ssh-<algorithm> <key>` or `ssh-<algorithm> <key1>, ssh-<algorithm> <key2>`. For more information see `TrustedUserCAKeys` in the sshd config man pages. When empty no file will be created and `SSH_AUTHORIZED_PRINCIPALS_ALLOW` will default to `off`.
- `SSH_TRUSTED_USER_CA_KEYS_FILENAME`: **`RUN_USER`/.ssh/gitea-trusted-user-ca-keys.pem**: Absolute path of the `TrustedUserCaKeys` file Gitea will manage. If you're running your own ssh server and you want to use the Gitea managed file you'll also need to modify your sshd_config to point to this file. The official docker image will automatically work without further configuration.
- `SSH_AUTHORIZED_PRINCIPALS_ALLOW`: **off** or **username, email**: \[off, username, email, anything\]: Specify the principals values that users are allowed to use as principal. When set to `anything` no checks are done on the principal string. When set to `off` authorized principal are not allowed to be set.
- `SSH_CREATE_AUTHORIZED_PRINCIPALS_FILE`: **false/true**: Gitea will create a authorized_principals file by default when it is not using the internal ssh server and `SSH_AUTHORIZED_PRINCIPALS_ALLOW` is not `off`.
- `SSH_CREATE_AUTHORIZED_PRINCIPALS_FILE`: **false/true**: Gitea will create an authorized_principals file by default when it is not using the builtin SSH server and `SSH_AUTHORIZED_PRINCIPALS_ALLOW` is not `off`.
- `SSH_AUTHORIZED_PRINCIPALS_BACKUP`: **false/true**: Enable SSH Authorized Principals Backup when rewriting all keys, default is true if `SSH_AUTHORIZED_PRINCIPALS_ALLOW` is not `off`.
- `SSH_AUTHORIZED_KEYS_COMMAND_TEMPLATE`: **`{{.AppPath}} --config={{.CustomConf}} serv key-{{.Key.ID}}`**: Set the template for the command to passed on authorized keys. Possible keys are: AppPath, AppWorkPath, CustomConf, CustomPath, Key - where Key is a `models/asymkey.PublicKey` and the others are strings which are shellquoted.
- `SSH_SERVER_CIPHERS`: **`[email protected]`, `aes128-ctr`, `aes192-ctr`, `aes256-ctr`, `[email protected]`, `[email protected]`**: For the built-in SSH server, choose the ciphers to support for SSH connections, for system SSH this setting has no effect.
- `SSH_SERVER_KEY_EXCHANGES`: **curve25519-sha256, ecdh-sha2-nistp256, ecdh-sha2-nistp384, ecdh-sha2-nistp521, diffie-hellman-group14-sha256, diffie-hellman-group14-sha1**: For the built-in SSH server, choose the key exchange algorithms to support for SSH connections, for system SSH this setting has no effect.
- `SSH_SERVER_MACS`: **`[email protected]`, `hmac-sha2-256`, `hmac-sha1`**: For the built-in SSH server, choose the MACs to support for SSH connections, for system SSH this setting has no effect
- `SSH_SERVER_HOST_KEYS`: **ssh/gitea.rsa, ssh/gogs.rsa**: For the built-in SSH server, choose the keypairs to offer as the host key. The private key should be at `SSH_SERVER_HOST_KEY` and the public `SSH_SERVER_HOST_KEY.pub`. Relative paths are made absolute relative to the `APP_DATA_PATH`. If no key exists a 4096 bit RSA key will be created for you.
- `SSH_SERVER_HOST_KEYS`: **ssh/gitea.rsa, ssh/gitea.ed25519, ssh/gitea.ecdsa, ssh/gogs.rsa**: For the built-in SSH server, choose the keypairs to offer as the host key. The private key should be at `SSH_SERVER_HOST_KEY` and the public `SSH_SERVER_HOST_KEY.pub`. Relative paths are made absolute relative to the `APP_DATA_PATH`. If none of the keys exist, a 4096 bit RSA, a 256 bit ECDSA and an ED25519 key will be created for you.
- `SSH_KEY_TEST_PATH`: **/tmp**: Directory to create temporary files in when testing public keys using ssh-keygen, default is the system temporary directory.
- `SSH_KEYGEN_PATH`: **_empty_**: Use `ssh-keygen` to parse public SSH keys. The value is passed to the shell. By default, Gitea does the parsing itself.
- `SSH_EXPOSE_ANONYMOUS`: **false**: Enable exposure of SSH clone URL to anonymous visitors, default is false.
@@ -396,46 +451,14 @@ The following configuration set `Content-Type: application/vnd.android.package-a
- `REDIRECT_OTHER_PORT`: **false**: If true and `PROTOCOL` is https, allows redirecting http requests on `PORT_TO_REDIRECT` to the https port Gitea listens on.
- `REDIRECTOR_USE_PROXY_PROTOCOL`: **`{USE_PROXY_PROTOCOL}`**: expect PROXY protocol header on connections to https redirector.
- `PORT_TO_REDIRECT`: **80**: Port for the http redirection service to listen on. Used when `REDIRECT_OTHER_PORT` is true.
- `SSL_MIN_VERSION`: **TLSv1.2**: Set the minimum version of ssl support.
- `SSL_MAX_VERSION`: **_empty_**: Set the maximum version of ssl support.
- `SSL_CURVE_PREFERENCES`: **X25519,P256**: Set the preferred curves,
- `SSL_CIPHER_SUITES`: **ecdhe_ecdsa_with_aes_256_gcm_sha384,ecdhe_rsa_with_aes_256_gcm_sha384,ecdhe_ecdsa_with_aes_128_gcm_sha256,ecdhe_rsa_with_aes_128_gcm_sha256,ecdhe_ecdsa_with_chacha20_poly1305,ecdhe_rsa_with_chacha20_poly1305**: Set the preferred cipher suites.
- If there is no hardware support for AES suites, by default the ChaCha suites will be preferred over the AES suites.
- supported suites as of Go 1.18 are:
- TLS 1.0 - 1.2 cipher suites
- "rsa_with_rc4_128_sha"
- "rsa_with_3des_ede_cbc_sha"
- "rsa_with_aes_128_cbc_sha"
- "rsa_with_aes_256_cbc_sha"
- "rsa_with_aes_128_cbc_sha256"
- "rsa_with_aes_128_gcm_sha256"
- "rsa_with_aes_256_gcm_sha384"
- "ecdhe_ecdsa_with_rc4_128_sha"
- "ecdhe_ecdsa_with_aes_128_cbc_sha"
- "ecdhe_ecdsa_with_aes_256_cbc_sha"
- "ecdhe_rsa_with_rc4_128_sha"
- "ecdhe_rsa_with_3des_ede_cbc_sha"
- "ecdhe_rsa_with_aes_128_cbc_sha"
- "ecdhe_rsa_with_aes_256_cbc_sha"
- "ecdhe_ecdsa_with_aes_128_cbc_sha256"
- "ecdhe_rsa_with_aes_128_cbc_sha256"
- "ecdhe_rsa_with_aes_128_gcm_sha256"
- "ecdhe_ecdsa_with_aes_128_gcm_sha256"
- "ecdhe_rsa_with_aes_256_gcm_sha384"
- "ecdhe_ecdsa_with_aes_256_gcm_sha384"
- "ecdhe_rsa_with_chacha20_poly1305_sha256"
- "ecdhe_ecdsa_with_chacha20_poly1305_sha256"
- TLS 1.3 cipher suites
- "aes_128_gcm_sha256"
- "aes_256_gcm_sha384"
- "chacha20_poly1305_sha256"
- Aliased names
- "ecdhe_rsa_with_chacha20_poly1305" is an alias for "ecdhe_rsa_with_chacha20_poly1305_sha256"
- "ecdhe_ecdsa_with_chacha20_poly1305" is alias for "ecdhe_ecdsa_with_chacha20_poly1305_sha256"
- `SSL_MIN_VERSION`: **_empty_**: Minimum supported TLS version, one of TLSv1.0, TLSv1.1, TLSv1.2, TLSv1.3. Unset uses Go's default.
- `SSL_MAX_VERSION`: **_empty_**: Maximum supported TLS version, one of TLSv1.0, TLSv1.1, TLSv1.2, TLSv1.3. Unset uses Go's default.
- `SSL_CURVE_PREFERENCES`: **_empty_**: Comma-separated list of SSL curve preferences. Unset uses Go's default.
- `SSL_CIPHER_SUITES`: **_empty_**: Comma-separated list of SSL cipher suites. Unset uses Go's default.
- `ENABLE_ACME`: **false**: Flag to enable automatic certificate management via an ACME capable Certificate Authority (CA) server (default: Lets Encrypt). If enabled, `CERT_FILE` and `KEY_FILE` are ignored, and the CA must resolve `DOMAIN` to this gitea server. Ensure that DNS records are set and either port `80` or port `443` are accessible by the CA server (the public internet by default), and redirected to the appropriate ports `PORT_TO_REDIRECT` or `HTTP_PORT` respectively.
- `ACME_URL`: **_empty_**: The CA's ACME directory URL, e.g. for a self-hosted [smallstep CA server](https://github.com/smallstep/certificates), it can look like `https://ca.example.com/acme/acme/directory`. If left empty, it defaults to using Let's Encerypt's production CA (check `LETSENCRYPT_ACCEPTTOS` as well).
- `ACME_ACCEPTTOS`: **false**: This is an explicit check that you accept the terms of service of the ACME provider. The default is Lets Encrypt [terms of service](https://letsencrypt.org/documents/LE-SA-v1.2-November-15-2017.pdf).
- `ACME_DIRECTORY`: **https**: Directory that the certificate manager will use to cache information such as certs and private keys.
- `ACME_DIRECTORY`: **https**: Directory that the certificate manager will use to cache information such as certs and private keys. A relative path is resolved against the working directory of the Gitea process, which is not necessarily a directory that survives a restart, see the note in [HTTPS setup](https-support.md#using-acme-default-lets-encrypt).
- `ACME_EMAIL`: **_empty_**: Email used for the ACME registration. Usually it is to notify about problems with issued certificates.
- `ACME_CA_ROOT`: **_empty_**: The CA's root certificate. If left empty, it defaults to using the system's trust chain.
- `ALLOW_GRACEFUL_RESTARTS`: **true**: Perform a graceful restart on SIGHUP
@@ -465,7 +488,7 @@ The following configuration set `Content-Type: application/vnd.android.package-a
- `require`: Enable TLS without any verifications.
- `verify-ca`: Enable TLS with verification of the database server certificate against its root certificate.
- `verify-full`: Enable TLS and verify the database server name matches the given certificate in either the `Common Name` or `Subject Alternative Name` fields.
- `SQLITE_TIMEOUT`: **500**: Query timeout for SQLite3 only.
- `SQLITE_TIMEOUT`: **20000**: Query timeout in milliseconds, for SQLite3 only.
- `SQLITE_JOURNAL_MODE`: **""**: Change journal mode for SQlite3. Can be used to enable [WAL mode](https://www.sqlite.org/wal.html) when high load causes write congestion. See [SQlite3 docs](https://www.sqlite.org/pragma.html#pragma_journal_mode) for possible values. Defaults to the default for the database file, often DELETE.
- `ITERATE_BUFFER_SIZE`: **50**: Internal buffer size for iterating.
- `PATH`: **data/gitea.db**: For SQLite3 only, the database file path.
@@ -485,14 +508,14 @@ relation to port exhaustion.
## Indexer (`indexer`)
- `ISSUE_INDEXER_TYPE`: **bleve**: Issue indexer type, currently supported: `bleve`, `db`, `elasticsearch` or `meilisearch`.
- `ISSUE_INDEXER_TYPE`: **bleve**: Issue indexer type, currently supported: `bleve`, `db`, `elasticsearch` (also compatible with OpenSearch) or `meilisearch`.
- `ISSUE_INDEXER_CONN_STR`: ****: Issue indexer connection string, available when ISSUE_INDEXER_TYPE is elasticsearch (e.g. `http://elastic:password@localhost:9200`) or meilisearch (e.g. `http://:apikey@localhost:7700`)
- `ISSUE_INDEXER_NAME`: **gitea_issues**: Issue indexer name, available when ISSUE_INDEXER_TYPE is elasticsearch or meilisearch.
- `ISSUE_INDEXER_PATH`: **indexers/issues.bleve**: Index file used for issue search; available when ISSUE_INDEXER_TYPE is bleve and elasticsearch. Relative paths will be made absolute against _`AppWorkPath`_.
- `REPO_INDEXER_ENABLED`: **false**: Enables code search (uses a lot of disk space, about 6 times more than the repository size).
- `REPO_INDEXER_REPO_TYPES`: **sources,forks,mirrors,templates**: Repo indexer units. The items to index could be `sources`, `forks`, `mirrors`, `templates` or any combination of them separated by a comma. If empty then it defaults to `sources` only, as if you'd like to disable fully please see `REPO_INDEXER_ENABLED`.
- `REPO_INDEXER_TYPE`: **bleve**: Code search engine type, could be `bleve` or `elasticsearch`.
- `REPO_INDEXER_TYPE`: **bleve**: Code search engine type, could be `bleve` or `elasticsearch` (also compatible with OpenSearch).
- `REPO_INDEXER_PATH`: **indexers/repos.bleve**: Index file used for code search.
- `REPO_INDEXER_CONN_STR`: ****: Code indexer connection string, available when `REPO_INDEXER_TYPE` is elasticsearch. i.e. `http://elastic:password@localhost:9200`
- `REPO_INDEXER_NAME`: **gitea_codes**: Code indexer name, available when `REPO_INDEXER_TYPE` is elasticsearch
@@ -503,6 +526,10 @@ relation to port exhaustion.
- `MAX_FILE_SIZE`: **1048576**: Maximum size in bytes of files to be indexed.
- `STARTUP_TIMEOUT`: **30s**: If the indexer takes longer than this timeout to start - fail. (This timeout will be added to the hammer time above for child processes - as bleve will not start until the previous parent is shutdown.) Set to -1 to never timeout.
## Redis (`redis`)
- `CONN_STR`: **_empty_**: Shared default Redis connection string. When a redis-backed subsystem (`[cache]`, `[session]`, `[queue]`, `[global_lock]`) is configured to use Redis but its own connection string is empty, it falls back to this value. A subsystem's own connection string always takes precedence, so different subsystems can still use different Redis instances.
## Queue (`queue` and `queue.*`)
Configuration at `[queue]` will set defaults for queues with overrides for individual queues at `[queue.*]`. (However see below.)
@@ -511,7 +538,7 @@ Configuration at `[queue]` will set defaults for queues with overrides for indiv
- `DATADIR`: **queues/common**: Base DataDir for storing level queues. `DATADIR` for individual queues can be set in `queue.name` sections. Relative paths will be made absolute against `{APP_DATA_PATH}`.
- `LENGTH`: **100000**: Maximal queue size before channel queues block
- `BATCH_LENGTH`: **20**: Batch data before passing to the handler
- `CONN_STR`: **redis://127.0.0.1:6379/0**: Connection string for the redis queue type. Several redis connections schemes are supported. To see all available `uri.Scheme` types, see [here](https://github.com/go-gitea/gitea/blob/main/modules/nosql/manager_redis.go#L98-L123). For example, if you're running a Redis cluster, use `redis+cluster://127.0.0.1:6379/0`. Options can be set using query params. Similarly, LevelDB options can also be set using: **leveldb://relative/path?option=value** or **leveldb:///absolute/path?option=value**, and will override `DATADIR`
- `CONN_STR`: **redis://127.0.0.1:6379/0**: Connection string for the redis queue type. If left empty while `TYPE` is `redis`, it falls back to [redis/CONN_STR](#redis-redis) when that is set. Several redis connections schemes are supported. To see all available `uri.Scheme` types, see [here](https://github.com/go-gitea/gitea/blob/main/modules/nosql/manager_redis.go#L98-L123). For example, if you're running a Redis cluster, use `redis+cluster://127.0.0.1:6379/0`. Options can be set using query params. Similarly, LevelDB options can also be set using: **leveldb://relative/path?option=value** or **leveldb:///absolute/path?option=value**, and will override `DATADIR`
- `QUEUE_NAME`: **_queue**: The suffix for default redis and disk queue name. Individual queues will default to **`name`**`QUEUE_NAME` but can be overridden in the specific `queue.name` section.
- `SET_NAME`: **_unique**: The suffix that will be added to the default redis and disk queue `set` name for unique queues. Individual queues will default to **`name`**`QUEUE_NAME`_`SET_NAME`_ but can be overridden in the specific `queue.name` section.
- `MAX_WORKERS`: **(dynamic)**: Maximum number of worker go-routines for the queue. Default value is "CpuNum/2" clipped to between 1 and 10.
@@ -559,8 +586,8 @@ And the following unique queues:
- `SECRET_KEY`: **\<random at every install\>**: Global secret key. This key is VERY IMPORTANT, if you lost it, the data encrypted by it (like 2FA secret) can't be decrypted anymore.
- `SECRET_KEY_URI`: **_empty_**: Instead of defining SECRET_KEY, this option can be used to use the key stored in a file (example value: `file:/etc/gitea/secret_key`). It shouldn't be lost like SECRET_KEY.
- `LOGIN_REMEMBER_DAYS`: **31**: How long to remember that a user is logged in before requiring relogin (in days).
- `COOKIE_REMEMBER_NAME`: **gitea\_incredible**: Name of cookie used to store authentication
information.
- `COOKIE_REMEMBER_NAME`: **gitea\_remember**: Name of cookie used to store authentication information.
- `REVERSE_PROXY_LOGOUT_REDIRECT`: **_empty_**: URL or relative path to redirect users to after logout when authentication is handled by a reverse proxy or SSO. For example: `/my-sso/logout?return=/my-sso/home`.
- `REVERSE_PROXY_AUTHENTICATION_USER`: **X-WEBAUTH-USER**: Header name for reverse proxy
authentication.
- `REVERSE_PROXY_AUTHENTICATION_EMAIL`: **X-WEBAUTH-EMAIL**: Header name for reverse proxy
@@ -570,6 +597,9 @@ And the following unique queues:
- `REVERSE_PROXY_LIMIT`: **1**: Interpret X-Forwarded-For header or the X-Real-IP header and set this as the remote IP for the request.
Number of trusted proxy count. Set to zero to not use these headers.
- `REVERSE_PROXY_TRUSTED_PROXIES`: **127.0.0.0/8,::1/128**: List of IP addresses and networks separated by comma of trusted proxy servers. Use `*` to trust all.
- `X_FRAME_OPTIONS`: **SAMEORIGIN**: Set the `X-Frame-Options` header value for all HTTP responses. Set to `unset` to not send the header. Previously located in `[cors]`.
- `X_CONTENT_TYPE_OPTIONS`: **nosniff**: Set the `X-Content-Type-Options` header value for all HTTP responses. Set to `unset` to not send the header.
- `CONTENT_SECURITY_POLICY_GENERAL`: **_empty_**: The value of the general Content-Security-Policy for most web pages. Leave it empty to apply the default policy, or set it to "unset" to disable Content-Security-Polic.
- `DISABLE_GIT_HOOKS`: **true**: Set to `false` to enable users with Git Hook privilege to create custom Git Hooks.
:::warning
@@ -614,6 +644,31 @@ And the following unique queues:
- `PASSWORD_CHECK_PWN`: **false**: Check [HaveIBeenPwned](https://haveibeenpwned.com/Passwords) to see if a password has been exposed.
- `SUCCESSFUL_TOKENS_CACHE_SIZE`: **20**: Cache successful token hashes. API tokens are stored in the DB as pbkdf2 hashes however, this means that there is a potentially significant hashing load when there are multiple API operations. This cache will store the successfully hashed tokens in a LRU cache as a balance between performance and security.
- `DISABLE_QUERY_AUTH_TOKEN`: **false**: Reject API tokens sent in URL query string (Accept Header-based API tokens only). This setting will default to `true` in Gitea 1.23 and be deprecated in Gitea 1.24.
- `TWO_FACTOR_AUTH`: **_empty_**: set to enforced to enforce two factor authentication. Only available in Gitea 1.24 and later.
- `EGRESS_MODE`: **lax**: Egress mode toggles between strictness of outgoing requests:
- `lax` non-public targets (private, loopback, link-local, CGNAT and special-use ranges) to be allowed, it allows all public ones
- `strict` requires an explicit allow of all addresses
- `ALLOWED_HOST_LIST`: **_empty_**: Webhook and oauth2 clients can only call allowed hosts for security reasons. Comma separated list, eg: `192.168.1.0/24:3000`, `[2001:db8::/32]:9090`, `*.mydomain.com:[80|443]`
- Built-in networks:
- `loopback`: 127.0.0.0/8 for IPv4 and ::1/128 for IPv6, localhost is included.
- `private`: RFC 1918 (10.0.0.0/8, 172.16.0.0/12, 192.168.0.0/16), RFC 4193 (FC00::/7) and RFC 6598 CGNAT (100.64.0.0/10). Also called LAN/Intranet.
- CIDR list: `1.2.3.0/8`, `2001:db8::/32`, and with a port `[2001:db8::/32]:9090` (IPv6 addresses and CIDRs need brackets when a port is given)
- Host matching: `example.com` matches the host and any subdomain, dot-anchored so `notexample.com` never matches. `*.example.com` and `.example.com` match only subdomains, the apex itself is excluded. IDN names must be given as punycode.
- All addresses can have ports specified. Accepted port specs:
- a single port: `192.168.1.0/24:3000`
- a range, both bounds inclusive: `*.mydomain.com:3000-3010`
- a bracketed set of ports and ranges, `|` separated: `*.mydomain.com:[80|443|3000-3010]`
- all ports: `*.mydomain.com:*`
- A portless entry covers all ports in `lax` mode, only 80 and 443 in `strict` mode
- Port specs apply only where the list is consulted: in `lax` mode that is non-public targets alone, public targets are allowed on every port whatever the list says. In `strict` mode every target is checked, so ports restrict public hosts too.
- Non-public targets need an IP or built-in entry, a host name entry alone never covers them.
- Reserved addresses (the IPv4-embedding NAT64, Teredo and 6to4 ranges, this-network, multicast and broadcast) are denied whatever the list says. To reach them configure an HTTP proxy
- This list is enforced on direct connections only. When an HTTP proxy is configured, restricting the proxied target is the proxy server's responsibility.
## Audit (`audit`)
- `RECORD_OUTPUT`: **disabled**: Where security-relevant events are recorded. `disabled` records nothing. `database` writes events to the `audit_event` table and shows them in the admin, organization, repository and user settings. Invalid values fall back to `disabled`. See [Audit Logging](audit-logging.md).
- `RETENTION_DAYS`: **30**: Days to keep recorded events. `0` keeps them forever. Pruning is done by `cron.delete_old_audit_events`. That task is only registered while `RECORD_OUTPUT` records events and `RETENTION_DAYS` is greater than `0`.
## Camo (`camo`)
@@ -624,8 +679,8 @@ And the following unique queues:
## OpenID (`openid`)
- `ENABLE_OPENID_SIGNIN`: **true**: Allow authentication in via OpenID.
- `ENABLE_OPENID_SIGNUP`: **! DISABLE\_REGISTRATION**: Allow registering via OpenID.
- `ENABLE_OPENID_SIGNIN`: **false**: Allow authentication in via OpenID.
- `ENABLE_OPENID_SIGNUP`: **false**: Allow registering via OpenID.
- `WHITELISTED_URIS`: **_empty_**: If non-empty, list of POSIX regex patterns matching
OpenID URI's to permit.
- `BLACKLISTED_URIS`: **_empty_**: If non-empty, list of POSIX regex patterns matching
@@ -660,7 +715,7 @@ And the following unique queues:
Requires `Mailer` to be enabled.
- `REGISTER_MANUAL_CONFIRM`: **false**: Enable this to manually confirm new registrations.
Requires `REGISTER_EMAIL_CONFIRM` to be disabled.
- `DISABLE_REGISTRATION`: **false**: Disable registration, after which only admin can create
- `DISABLE_REGISTRATION`: **true**: Disable registration, after which only admin can create
accounts for users.
- `REQUIRE_EXTERNAL_REGISTRATION_PASSWORD`: **false**: Enable this to force externally created
accounts (via GitHub, OpenID Connect, etc) to create a password.
@@ -670,15 +725,20 @@ And the following unique queues:
:::
- `REQUIRE_SIGNIN_VIEW`: **false**: Enable this to force users to log in to view any page or to use API.
It could be set to "expensive" to block anonymous users accessing some pages which consume a lot of resources,
for example: block anonymous AI crawlers from accessing repo code pages.
The "expensive" mode is experimental and subject to change.
- `ENABLE_NOTIFY_MAIL`: **false**: Enable this to send e-mail to watchers of a repository when
something happens, like creating issues. Requires `Mailer` to be enabled.
- `ENABLE_BASIC_AUTHENTICATION`: **true**: Disable this to disallow authentication using HTTP
BASIC and the user's password. Please note if you disable this you will not be able to access the
tokens API endpoints using a password. Further, this only disables BASIC authentication using the
password - not tokens or OAuth Basic.
- `ENABLE_PASSWORD_SIGNIN_FORM`: **true**: Show the password login form (for password-based login).
- `ENABLE_PASSWORD_SIGNIN_FORM`: **true**: Show the password login form (for password-based login),
otherwise, only show OAuth2 or passkey login methods if they are enabled.
If you set it to false, maybe it also needs to set `ENABLE_BASIC_AUTHENTICATION` to false to
completely disable password-based authentication.
- `ENABLE_PASSKEY_AUTHENTICATION`: **true**: Allow users to sign-in with a passkey
- `ENABLE_REVERSE_PROXY_AUTHENTICATION`: **false**: Enable this to allow reverse proxy authentication for web requests
- `ENABLE_REVERSE_PROXY_AUTHENTICATION_API`: **false**: Enable this to allow reverse proxy authentication for API requests, the reverse proxy is responsible for ensuring that no CSRF is possible.
- `ENABLE_REVERSE_PROXY_AUTO_REGISTRATION`: **false**: Enable this to allow auto-registration
@@ -714,7 +774,7 @@ And the following unique queues:
- `DEFAULT_ALLOW_ONLY_CONTRIBUTORS_TO_TRACK_TIME`: **true**: Only allow users with write permissions to track time.
- `EMAIL_DOMAIN_ALLOWLIST`: **_empty_**: If non-empty, comma separated list of domain names that can only be used to register on this instance, wildcard is supported.
- `EMAIL_DOMAIN_BLOCKLIST`: **_empty_**: If non-empty, comma separated list of domain names that cannot be used to register on this instance, wildcard is supported.
- `SHOW_REGISTRATION_BUTTON`: **! DISABLE\_REGISTRATION**: Show Registration Button
- `SHOW_REGISTRATION_BUTTON`: **false**: Show Registration button, defaults to true only if both DISABLE_REGISTRATION and ALLOW_ONLY_EXTERNAL_REGISTRATION are false
- `SHOW_MILESTONES_DASHBOARD_PAGE`: **true** Enable this to show the milestones dashboard page - a view of all the user's milestones
- `AUTO_WATCH_NEW_REPOS`: **true**: Enable this to let all organisation users watch new repos when they are created
- `AUTO_WATCH_ON_CHANGES`: **false**: Enable this to make users watch a repository after their first commit to it
@@ -724,8 +784,7 @@ And the following unique queues:
- `DEFAULT_ORG_MEMBER_VISIBLE`: **false** True will make the membership of the users visible when added to the organisation.
- `ALLOW_ONLY_INTERNAL_REGISTRATION`: **false** Set to true to force registration only via Gitea.
- `ALLOW_ONLY_EXTERNAL_REGISTRATION`: **false** Set to true to force registration only using third-party services.
- `NO_REPLY_ADDRESS`: **noreply.DOMAIN** Value for the domain part of the user's email address in the Git log if user has set KeepEmailPrivate to true. DOMAIN resolves to the value in server.DOMAIN.
The user's email will be replaced with a concatenation of the user name in lower case, "@" and NO_REPLY_ADDRESS.
- `NO_REPLY_ADDRESS`: **_empty_**: Value for the domain part of the user's email address in the git log if user has set KeepEmailPrivate to true. The user's email will be replaced with a concatenation of the user name in lower case, "@" and NO_REPLY_ADDRESS. Default value is "noreply." + the domain part of ROOT_URL
- `USER_DELETE_WITH_COMMENTS_MAX_TIME`: **0** Minimum amount of time a user must exist before comments are kept when the user is deleted.
- `VALID_SITE_URL_SCHEMES`: **http, https**: Valid site url schemes for user profiles
@@ -736,6 +795,13 @@ And the following unique queues:
- `DISABLE_ORGANIZATIONS_PAGE`: **false**: Disable the organizations explore page.
- `DISABLE_CODE_PAGE`: **false**: Disable the code explore page.
### Request Quality of Service (`qos`)
- `ENABLED`: **false**: Enable request quality of service and overload protection.
- `MAX_INFLIGHT`: **(dynamic)**: The maximum number of concurrent requests that the server will process before enqueueing new requests. Default is "CpuNum * 4".
- `MAX_WAITING`: **100**: The maximum number of requests that can be enqueued before new requests will be dropped.
- `TARGET_WAIT_TIME`: **250ms**: Target maximum wait time a request may be enqueued for. Requests that are enqueued for less than this amount of time will not be dropped. When wait times exceed this amount, a portion of requests will be dropped until wait times have decreased below this amount.
- `PAGING_NUM`: **10**: Number of webhook history events that are shown in one page.
- `PROXY_URL`: **_empty_**: Proxy server URL, support http://, https//, socks://, blank will follow environment http_proxy/https_proxy. If not given, will use global proxy setting.
- `PROXY_URL`: **_empty_**: Proxy server URL, support http://, https://, socks5://, blank will follow environment http_proxy/https_proxy. If not given, will use global proxy setting.
- `PROXY_HOSTS`: **_empty_`**: Comma separated list of host names requiring proxy. Glob patterns (*) are accepted; use ** to match all hosts. If not given, will use global proxy setting.
When a proxy is configured, Gitea does not enforce `[security]``ALLOWED_HOST_LIST` on the proxied target, the proxy server is expected to restrict it.
## Mailer (`mailer`)
:::warning
@@ -793,7 +853,7 @@ and
- Please note: authentication is only supported when the SMTP server communication is encrypted with TLS (this can be via `STARTTLS`) or SMTP host is localhost. See [Email Setup](email-setup.md) for more information.
- `ENABLE_HELO`: **true**: Enable HELO operation.
- `HELO_HOSTNAME`: **(retrieved from system)**: HELO hostname.
- `FROM`: **_empty_**: Mail from address, RFC 5322. This can be just an email address, or the "Name" `\<[email protected]\>` format.
- `FROM`: **_empty_**: Mail from address, RFC 5322. This can be just an email address, or the `"Name" <[email protected]>` format.
- `ENVELOPE_FROM`: **_empty_**: Address set as the From address on the SMTP mail envelope. Set to `<>` to send an empty address.
- `FROM_DISPLAY_NAME_FORMAT`: **`{{ .DisplayName }}`**: If gitea sends mails on behave of users, it will just use the name also displayed in the WebUI. If you want e.g. `Mister X (by CodeIt) <[email protected]>`, set it to `{{ .DisplayName }} (by {{ .AppName }})`. Available Variables: `.DisplayName`, `.AppName` and `.Domain`.
- `SUBJECT_PREFIX`: **_empty_**: Prefix to be placed before e-mail subject lines.
@@ -803,6 +863,7 @@ and
- `SENDMAIL_CONVERT_CRLF`: **true**: Most versions of sendmail prefer LF line endings rather than CRLF line endings. Set this to false if your version of sendmail requires CRLF line endings.
- `SEND_BUFFER_LEN`: **100**: Buffer length of mailing queue. **DEPRECATED** use `LENGTH` in `[queue.mailer]`
- `SEND_AS_PLAIN_TEXT`: **false**: Send mails only in plain text, without HTML alternative.
- `EMBED_ATTACHMENT_IMAGES`: **false**: Embed attached images in base64 format in HTML emails. (for clients that do not load external images or disconnected vpn users still receiving emails; ATTENTION: online webclients like gmail will not show base64 embedded images)
- `ADAPTER`: **memory**: Cache engine adapter, either `memory`, `redis`, `twoqueue` or `memcache`. (`twoqueue` represents a size limited LRU cache.)
- `INTERVAL`: **60**: Garbage Collection interval (sec), for memory and twoqueue cache only.
- `HOST`: **_empty_**: Connection string for `redis` and `memcache`. For `twoqueue` sets configuration for the queue.
- `ADAPTER`: **memory**: Either `memory`, `redis`, `memcache`, or `twoqueue`.
- `INTERVAL`: **60**: For `memory` and `twoqueue`, GC interval in seconds.
- `HOST`: **_empty_**: For `redis` and `memcache`, connection host address. For `twoqueue` sets configuration for the queue. If left empty while `ADAPTER` is `redis`, it falls back to [redis/CONN_STR](#redis-redis) when that is set.
- Redis: `redis://:[email protected]:6379/0?pool_size=100&idle_timeout=180s` (See [queue/CONN_STR](#queue-queue-and-queue) for information on additional supported connection types)
- For a Redis cluster: `redis+cluster://:[email protected]:6379/0?pool_size=100&idle_timeout=180s`
- For a Redis sentinel: `redis+sentinel://:macaron@sentinel0:26379,sentinel1:26379,sentinel2:26379/0?pool_size=100&idle_timeout=180s&master_name=mymaster`
- Memcache: `127.0.0.1:9090;127.0.0.1:9091`
- Memcache: `127.0.0.1:11211`
- TwoQueue LRU cache: `{"size":50000,"recent_ratio":0.25,"ghost_ratio":0.5}` or `50000` representing the maximum number of objects stored in the cache.
- `ITEM_TTL`: **16h**: Time to keep items in cache if not used, Setting it to -1 disables caching.
- `ITEM_TTL`: **16h**: Time to keep items in cache if not used. Setting it to -1 disables caching.
- `ITEM_TTL`: **8760h**: Time to keep items in cache if not used, Setting it to -1 disables caching.
- `ITEM_TTL`: **8760h**: Time to keep items in cache if not used. Setting it to -1 disables caching.
- `COMMITS_COUNT`: **1000**: Only enable the cache when repository's commits count great than.
## Session (`session`)
- `PROVIDER`: **memory**: Session engine provider \[memory, file, redis, db, mysql, couchbase, memcache, postgres\]. Setting `db` will reuse the configuration in `[database]`
- `PROVIDER_CONFIG`: **data/sessions**: For file, the root path; for db, empty (database config will be used); for others, the connection string. Relative paths will be made absolute against _`AppWorkPath`_.
- `PROVIDER`: **file**: Either "memory", "file", "redis" or "db", default is "file". "db" will reuse the configuration in `[database]`
- `PROVIDER_CONFIG`: **data/sessions**: For file, the root path; for db, empty (database config will be used); for others, the connection string. Relative paths will be made absolute against _`AppWorkPath`_. For the `redis` provider, if left empty it falls back to [redis/CONN_STR](#redis-redis) when that is set.
- `COOKIE_SECURE`:**_empty_**: `true` or `false`. Enable this to force using HTTPS for all session access. If not set, it defaults to `true` if the ROOT_URL is an HTTPS URL.
- `COOKIE_NAME`: **i\_like\_gitea**: The name of the cookie used for the session ID.
- `COOKIE_NAME`: **gitea\_session**: Session cookie name
- `GC_INTERVAL_TIME`: **86400**: GC interval in seconds.
- `SESSION_LIFE_TIME`: **86400**: Session life time in seconds, default is 86400 (1 day)
- `DOMAIN`: **_empty_**: Sets the cookie Domain
@@ -863,8 +924,7 @@ In-Reply-To =
## Picture (`picture`)
- `GRAVATAR_SOURCE`: **gravatar**: Can be `gravatar`, `duoshuo` or anything like
`http://cn.gravatar.com/avatar/`.
- `GRAVATAR_SOURCE`: **gravatar**: Can be `gravatar`, `libravatar` or any other Gravatar-compatible source, like `https://cn.gravatar.com/avatar/`.
- `DISABLE_GRAVATAR`: **false**: Enable this to use local avatars only. **DEPRECATED [v1.18+]** moved to database. Use admin panel to configure.
- `ENABLE_FEDERATED_AVATAR`: **false**: Enable support for federated avatars (see
[http://www.libravatar.org](http://www.libravatar.org)). **DEPRECATED [v1.18+]** moved to database. Use admin panel to configure.
@@ -883,7 +943,7 @@ In-Reply-To =
- none = no avatar will be displayed
- random = random avatar will be generated
- image = default image will be used (which is set in `REPOSITORY_AVATAR_FALLBACK_IMAGE`)
- `REPOSITORY_AVATAR_FALLBACK_IMAGE`: **/img/repo_default.png**: Image used as default repository avatar (if `REPOSITORY_AVATAR_FALLBACK` is set to image and none was uploaded)
- `REPOSITORY_AVATAR_FALLBACK_IMAGE`: **_empty_**: Image used as default repository avatar (if `REPOSITORY_AVATAR_FALLBACK` is set to image and none was uploaded). The value is used as-is, without prepending the sub-path of `ROOT_URL`. Empty uses Gitea's builtin repository avatar.
## Project (`project`)
@@ -892,12 +952,15 @@ Default templates for project board view:
- `PROJECT_BOARD_BASIC_KANBAN_TYPE`: **To Do, In Progress, Done**
- `PROJECT_BOARD_BUG_TRIAGE_TYPE`: **Needs Triage, High Priority, Low Priority, Closed**
## Issue and pullrequest attachments (`attachment`)
## Issue, pull-request and release attachments (`attachment`)
- `ENABLED`: **true**: Whether issue and pullrequest attachments are enabled.
- `ALLOWED_TYPES`: **.avif,.cpuprofile,.csv,.dmp,.docx,.fodg,.fodp,.fods,.fodt,.gif,.gz,.jpeg,.jpg,.json,.jsonc,.log,.md,.mov,.mp4,.odf,.odg,.odp,.ods,.odt,.patch,.pdf,.png,.pptx,.svg,.tgz,.txt,.webm,.webp,.xls,.xlsx,.zip**: Comma-separated list of allowed file extensions (`.zip`), mime types (`text/plain`) or wildcard type (`image/*`, `audio/*`, `video/*`). Empty value or `*/*` allows all types.
- `MAX_SIZE`: **2048**: Maximum size (MB).
- `MAX_FILES`: **5**: Maximum number of attachments that can be uploaded at once.
ALLOWED_TYPES/MAX_SIZE/MAX_FILES in this section only affect issue and pull-request attachments, not release attachments.
Release attachment has its own config options in `[repository.release]` section.
- `ENABLED`: **true**: Whether issue, pull-request and release attachments are enabled.
- `ALLOWED_TYPES`: **.avif,.cpuprofile,.csv,.dmp,.docx,.fodg,.fodp,.fods,.fodt,.gif,.gz,.jpeg,.jpg,.json,.jsonc,.log,.md,.mov,.mp4,.odf,.odg,.odp,.ods,.odt,.patch,.pdf,.png,.pptx,.svg,.tgz,.txt,.webm,.webp,.xls,.xlsx,.zip**: Comma-separated list of allowed issue/pull-request attachment file extensions (`.zip`), mime types (`text/plain`) or wildcard type (`image/*`, `audio/*`, `video/*`). Empty value or `*/*` allows all types.
- `MAX_SIZE`: **100**: Max size of each issue/pull-request attachment file in MB.
- `MAX_FILES`: **5**: Maximum number of issue/pull-request attachments that can be uploaded at once.
- `STORAGE_TYPE`: **local**: Storage type for attachments, it could be `<blank>`, `local`, `minio`, `azureblob` or `xxx` which defined in another section with `[storage.xxx]`.
For `STORAGE_TYPE = ` or there is no this configuration item, all storages will be derived from `[storage]` if configured or defult values.
@@ -906,7 +969,7 @@ Default templates for project board view:
- `PATH`: **attachments**: Path to store attachments only available when STORAGE_TYPE is `local`, relative paths will be resolved to `{AppDataPath}/{attachment.PATH}`.
For `STORAGE_TYPE = minio`, the configurations can be found at [Storage Minio](#storage_minio), you can override some configurations like below.
For `STORAGE_TYPE = minio`, the configurations can be found at [Storage Minio](#minio-storage-configuration-storage_minio), you can override some configurations like below.
- `MINIO_BASE_PATH`: **attachments/**: Minio base path on the bucket only available when STORAGE_TYPE is `minio`
@@ -979,7 +1042,6 @@ Default templates for project board view:
- `ENABLED`: **false**: Enable to run all cron tasks periodically with default settings.
- `RUN_AT_START`: **false**: Run cron tasks at application start-up.
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE` accept formats
- Full crontab specs, e.g. `* * * * * ?`
- Descriptors, e.g. `@midnight`, `@every 1h30m` ...
@@ -991,30 +1053,40 @@ Default templates for project board view:
- `ENABLED`: **true**: Enable service.
- `RUN_AT_START`: **true**: Run tasks at start up time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Whether to emit notice on successful execution too
- `SCHEDULE`: **@midnight**: Cron syntax for scheduling repository archive cleanup, e.g. `@every 1h`.
- `OLDER_THAN`: **24h**: Archives created more than `OLDER_THAN` ago are subject to deletion, e.g. `12h`.
- `SCHEDULE`: **@every 10m**: Cron syntax for scheduling update mirrors, e.g. `@every 3h`.
- `RUN_AT_START`: **false**: Run Update mirrors task when Gitea starts.
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `PULL_LIMIT`: **50**: Limit the number of mirrors added to the queue to this number (negative values mean no limit, 0 will result in no mirrors being queued effectively disabling pull mirror updating).
- `PUSH_LIMIT`: **50**: Limit the number of mirrors added to the queue to this number (negative values mean no limit, 0 will result in no mirrors being queued effectively disabling push mirror updating).
#### Cron - Repository Health Check (`cron.repo_health_check`)
- `ENABLED`: **true**: Enable running repository health check task periodically.
- `SCHEDULE`: **@midnight**: Cron syntax for scheduling repository health check.
- `RUN_AT_START`: **false**: Run repository health check task when Gitea starts.
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `TIMEOUT`: **60s**: Time duration syntax for health check execution timeout.
- `ARGS`: **_empty_**: Arguments for command `git fsck`, e.g. `--unreachable --tags`. See more on http://git-scm.com/docs/git-fsck
- `RUN_AT_START`: **false**: Run cleanup hook_task at start time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@midnight**: Cron syntax for cleaning hook_task table.
- `CLEANUP_TYPE`**OlderThan** OlderThan or PerWebhook Method to cleanup hook_task, either by age (i.e. how long ago hook_task record was delivered) or by the number to keep per webhook (i.e. keep most recent x deliveries per webhook).
- `OLDER_THAN`: **168h**: If CLEANUP_TYPE is set to OlderThan, then any delivered hook_task records older than this expression will be deleted.
@@ -1030,10 +1102,20 @@ Default templates for project board view:
#### Cron - Update Migration Poster ID (`cron.update_migration_poster_id`)
Update migrated repositories' issues and comments' posterid, it will always attempt synchronization when the instance starts.
- `ENABLED`: **true**: Enable update migration poster id job.
- `RUN_AT_START`: **true**: Update migrated repositories' issues and comments' posterid when starting server
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@midnight** : Interval as a duration between each synchronization, it will always attempt synchronization when the instance starts.
Synchronize external user data (only LDAP user synchronization is supported)
- `ENABLED`: **true**: Enable synchronize external user data job
- `RUN_AT_START`: **false**: Synchronize external user data when starting server
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@midnight** : Interval as a duration between each synchronization, it will always attempt synchronization when the instance starts.
- `UPDATE_EXISTING`: **true**: Create new users, update existing user data and disable users that are not in external source anymore (default) or only create new users if UPDATE_EXISTING is set to false.
@@ -1041,8 +1123,16 @@ Default templates for project board view:
- `RUN_AT_START`: **false**: Run job at start time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to log a success message.
- `SCHEDULE`: **@annually*: Cron syntax for scheduling synchronization of repository licenses
### Extended cron tasks
#### Cron - Delete all repository archives (`cron.delete_repo_archives`)
@@ -1074,7 +1171,14 @@ Default templates for project board view:
- `ENABLED`: **false**: Enable service.
- `RUN_AT_START`: **false**: Run tasks at start up time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@every 72h**: Cron syntax for scheduling repository archive cleanup, e.g. `@every 1h`.
- `SCHEDULE`: **@every 72h**: Cron syntax for scheduling synchronization of gitea ssh keys, e.g. `@every 1h`.
#### Cron - Update the '.ssh/authorized_principals' file (`cron.resync_all_sshprincipals`)
- `ENABLED`: **false**: Enable service.
- `RUN_AT_START`: **false**: Run tasks at start up time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@every 72h**: Cron syntax for scheduling update of authorized_principals, e.g. `@every 1h`.
#### Cron - Resynchronize pre-receive, update and post-receive hooks of all repositories (`cron.resync_all_hooks`)
@@ -1112,11 +1216,21 @@ Default templates for project board view:
- `SCHEDULE`: **@every 168h**: Cron syntax to set how often to check.
- `OLDER_THAN`: **8760h**: any action older than this expression will be deleted from database, suggest using `8760h` (1 year) because that's the max length of heatmap.
#### Cron - Delete old audit events (`cron.delete_old_audit_events`)
Deletes audit events older than the retention period. The task is only registered when `[audit].RECORD_OUTPUT` records events and `[audit].RETENTION_DAYS` is greater than `0`. See [Audit Logging](audit-logging.md).
- `ENABLED`: **true**: Enable service.
- `RUN_AT_START`: **false**: Run tasks at start up time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@every 24h**: Cron syntax to set how often to check.
- `OLDER_THAN`: **720h**: any audit event older than this expression will be deleted from the database. Defaults to `[audit].RETENTION_DAYS`.
#### Cron - Check for new Gitea versions (`cron.update_checker`)
- `ENABLED`: **true**: Enable service.
- `RUN_AT_START`: **false**: Run tasks at start up time (if ENABLED).
- `ENABLE_SUCCESS_NOTICE`: **true**: Set to false to switch off success notices.
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@every 168h**: Cron syntax for scheduling a work, e.g. `@every 168h`.
- `HTTP_ENDPOINT`: **https://dl.gitea.com/gitea/version.json**: the endpoint that Gitea will check for newer versions
@@ -1124,7 +1238,7 @@ Default templates for project board view:
- `ENABLED`: **false**: Enable service.
- `RUN_AT_START`: **false**: Run tasks at start up time (if ENABLED).
- `NO_SUCCESS_NOTICE`: **false**: Set to true to switch off success notices.
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@every 168h**: Cron syntax to set how often to check.
- `OLDER_THAN`: **8760h**: any system notice older than this expression will be deleted from database.
@@ -1132,12 +1246,50 @@ Default templates for project board view:
- `ENABLED`: **false**: Enable service.
- `RUN_AT_START`: **false**: Run tasks at start up time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@every 24h**: Cron syntax to set how often to check.
- `OLDER_THAN`: **168h**: Only attempt to garbage collect LFSMetaObjects older than this (default 7 days)
- `LAST_UPDATED_MORE_THAN_AGO`: **72h**: Only attempt to garbage collect LFSMetaObjects that have not been attempted to be garbage collected for this long (default 3 days)
- `NUMBER_TO_CHECK_PER_REPO`: **100**: Minimum number of stale LFSMetaObjects to check per repo. Set to `0` to always check all.
- `PROPORTION_TO_CHECK_PER_REPO`: **0.6**: Check at least this proportion of LFSMetaObjects per repo. (This may cause all stale LFSMetaObjects to be checked.)
### Actions cron tasks
#### Cron - Rebuild issue index (`cron.rebuild_issue_indexer`)
- `ENABLED`: **true**: Enable service.
- `RUN_AT_START`: **true**: Run tasks at start up time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@annually**: Cron syntax to set how often to rebuild
#### Cron - Stop running tasks which haven't been updated for a long time (`cron.stop_zombie_tasks`)
- `ENABLED`: **true**: Enable service.
- `RUN_AT_START`: **true**: Run tasks at start up time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@every 5m**: Cron syntax to set how often tasks to run the check
#### Cron - Stop running tasks which have running status and continuous updates but don't end for a long time (`cron.stop_endless_tasks`)
- `ENABLED`: **true**: Enable service.
- `RUN_AT_START`: **true**: Run tasks at start up time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@every 30m**: Cron syntax to set how often to run the check.
#### Cron - Cancel jobs which haven't been picked up for a long time (`cron.cancel_abandoned_jobs`)
- `ENABLED`: **true**: Enable service.
- `RUN_AT_START`: **false**: Run tasks at start up time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@every 6h**: Cron syntax to set how often to run the check.
#### Cron - Start cron based actions (`cron.start_schedule_tasks`)
- `ENABLED`: **true**: Enable service.
- `RUN_AT_START`: **false**: Run tasks at start up time (if ENABLED).
- `NOTICE_ON_SUCCESS`: **false**: Set to true to switch on success notices.
- `SCHEDULE`: **@every 1m**: Cron syntax to set how often to schedule tasks
## Git (`git`)
- `PATH`: **""**: The path of Git executable. If empty, Gitea searches through the PATH environment.
@@ -1158,24 +1310,23 @@ Default templates for project board view:
- `LARGE_OBJECT_THRESHOLD`: **1048576**: (Go-Git only), don't cache objects greater than this in memory. (Set to 0 to disable.)
- `DISABLE_CORE_PROTECT_NTFS`: **false** Set to true to forcibly set `core.protectNTFS` to false.
- `DISABLE_PARTIAL_CLONE`: **false** Disable the usage of using partial clones for git.
- `DIFF_RENAME_SIMILARITY_THRESHOLD`: **50%** Set the similarity threshold passed to git commands via `--find-renames=<threshold>`. Default is 50%, the same as git. Must be a integer percentage between 0% and 100%.
The key/value pairs in this section will be used as git config.
This section only does "set" config, a removed config key from this section won't be removed from git config automatically. The format is `some.configKey = value`.
The key/value pairs in this section will be used as git config. The format is `some.configKey = value`.
These options are written into the gitconfig file under `[git] HOME_PATH` when Gitea starts, and are applied **after** Gitea's built-in defaults — so values set here take precedence over the defaults listed below.
**Note:** This section only does "set" config; a removed config key from this section won't be removed from git config automatically. Some config options might affect the behavior of git and cause Gitea's git operations to fail — make sure you know what you are doing before making changes.
- `diff.algorithm`: **histogram**
- `core.logAllRefUpdates`: **true**
- `gc.reflogExpire`: **90**
## Metrics (`metrics`)
@@ -1203,7 +1354,10 @@ This section only does "set" config, a removed config key from this section won'
- `JWT_SECRET_URI`: **_empty_**: Instead of defining JWT_SECRET in the configuration, this configuration option can be used to give Gitea a path to a file that contains the secret (example value: `file:/etc/gitea/oauth2_jwt_secret`)
- `JWT_SIGNING_PRIVATE_KEY_FILE`: **jwt/private.pem**: Private key file path used to sign OAuth2 tokens. The path is relative to `APP_DATA_PATH`. This setting is only needed if `JWT_SIGNING_ALGORITHM` is set to `RS256`, `RS384`, `RS512`, `ES256`, `ES384` or `ES512`. The file must contain a RSA or ECDSA private key in the PKCS8 format. If no key exists a 4096 bit key will be created for you.
- `MAX_TOKEN_LENGTH`: **32767**: Maximum length of token/cookie to accept from OAuth2 provider
- `DEFAULT_APPLICATIONS`: **git-credential-oauth, git-credential-manager, tea**: Pre-register OAuth applications for some services on startup. See the [OAuth2 documentation](/development/oauth2-provider.md) for the list of available options.
- `DEFAULT_APPLICATIONS`: **git-credential-oauth, git-credential-manager, tea, gitea-app**: Pre-register OAuth applications for some services on startup. See the [OAuth2 documentation](/development/oauth2-provider.md) for the list of available options.
- `CUSTOM_SCHEMES`: **_empty_**: By default, OAuth2 applications can only use "http" and "https" as their redirect URI schemes.
If you need to use other schemes (e.g. for desktop applications), you can specify them here as a comma-separated list.
For example: set "my-scheme, com.example.app" to support "my-scheme://..." and "com.example.app://..." redirect URIs.
## i18n (`i18n`)
@@ -1213,7 +1367,9 @@ This section only does "set" config, a removed config key from this section won'
## Markup (`markup`)
- `MERMAID_MAX_SOURCE_CHARACTERS`: **5000**: Set the maximum size of a Mermaid source. (Set to -1 to disable)
- `MERMAID_MAX_SOURCE_CHARACTERS`: **50000**: Set the maximum size of a Mermaid source. (Set to -1 to disable)
Gitea can support Markup using external tools. The example below will add a markup named `asciidoc`.
@@ -1227,7 +1383,6 @@ IS_INPUT_FILE = false
```
- ENABLED: **false** Enable markup support; set to **true** to enable this renderer.
- NEED\_POSTPROCESS: **true** set to **true** to replace links / sha1 and etc.
- FILE\_EXTENSIONS: **_empty_** List of file extensions that should be rendered by an external
command. Multiple extensions needs a comma as splitter.
- RENDER\_COMMAND: External command to render all matching extensions.
@@ -1236,6 +1391,10 @@ IS_INPUT_FILE = false
- sanitized: Sanitize the content and render it inside current page, default to only allow a few HTML tags and attributes. Customized sanitizer rules can be defined in `[markup.sanitizer.*]`.
- no-sanitizer: Disable the sanitizer and render the content inside current page. It's **insecure** and may lead to XSS attack if the content contains malicious code.
- iframe: Render the content in a separate standalone page and embed it into current page by iframe. The iframe is in sandbox mode with same-origin disabled, and the JS code are safely isolated from parent page.
- RENDER_CONTENT_SANDBOX: **_empty_** The sandbox applied to the iframe and Content-Security-Policy header when RENDER_CONTENT_MODE is `iframe`. It defaults to a safe set of "allow-*" restrictions (space separated). You can also set it by your requirements or use "disabled" to disable the sandbox completely. When set it, make sure there is no security risk:
- PDF-only content: generally safe to use "disabled", and it needs to be "disabled" because PDF only renders with no sandbox.
- HTML content with JS: if the "RENDER_COMMAND" can guarantee there is no XSS, then it is safe, otherwise, you need to fine tune the "allow-*" restrictions.
- NEED_POST_PROCESS: **false** Whether post-process the rendered HTML content, including: resolve relative links and image sources, recognizing issue/commit references, escaping invisible characters, mentioning users, rendering permlink code blocks, replacing emoji shorthands, etc. By default, this is true when RENDER_CONTENT_MODE is `sanitized`, otherwise false.
Two special environment variables are passed to the render command:
@@ -1279,11 +1438,13 @@ in this mapping or the filetype using heuristics.
- `MAX_ATTEMPTS`: **3**: Max attempts per http/https request on migrations.
- `RETRY_BACKOFF`: **3**: Backoff time per http/https request retry (seconds)
- `ALLOWED_DOMAINS`: **_empty_**: Domains allowlist for migrating repositories, default is blank. It means external hosts will be allowed. Multiple domains could be separated by commas. Wildcard is supported: `github.com, *.github.com`.
- `BLOCKED_DOMAINS`: **_empty_**: Domains blocklist for migrating repositories, default is blank. Multiple domains could be separated by commas. When `ALLOWED_DOMAINS` is not blank, this option has a higher priority to deny domains. Wildcard is supported.
- `ALLOW_LOCALNETWORKS`: **false**: Allow private addresses defined by RFC 1918, RFC 1122, RFC 4632 and RFC 4291. If a domain is allowed by `ALLOWED_DOMAINS`, this option will be ignored.
- `EGRESS_MODE`: **lax**: Mode toggles between strictness of scanning outgoing requests, same format as `EGRESS_MODE` in [`security`](#security-security).
- `ALLOWED_HOST_LIST`: **_empty_**: Hosts migrations and mirrors may call, same format as `ALLOWED_HOST_LIST` in [`security`](#security-security). Private, loopback and link-local addresses need a builtin or CIDR entry. When unset, the deprecated `ALLOWED_DOMAINS` entries apply on every port with `strict` as the default mode, and the deprecated `ALLOW_LOCALNETWORKS` adds `private, loopback`.
- `BLOCKED_HOST_LIST`: **_empty_**: Hosts migrations and mirrors may never call, portless entries cover all ports.
These lists are enforced on direct connections only. When a proxy is configured (`[proxy]`, `[git.config]``http.proxy` or the environment), restricting the proxied target is the proxy server's responsibility.
@@ -1326,6 +1487,7 @@ in this mapping or the filetype using heuristics.
- `LIMIT_SIZE_RUBYGEMS`: **-1**: Maximum size of a RubyGems upload (`-1` means no limits, format `1000`, `1 MB`, `1 GiB`)
- `LIMIT_SIZE_SWIFT`: **-1**: Maximum size of a Swift upload (`-1` means no limits, format `1000`, `1 MB`, `1 GiB`)
- `LIMIT_SIZE_VAGRANT`: **-1**: Maximum size of a Vagrant upload (`-1` means no limits, format `1000`, `1 MB`, `1 GiB`)
- `LIMIT_SIZE_TERRAFORM_STATE``: **-1**: Maximum size of a Terraform state upload (`-1` means no limits, format `1000`, `1 MB`, `1 GiB`)
## Mirror (`mirror`)
@@ -1349,7 +1511,7 @@ is `data/lfs` and the default of `MINIO_BASE_PATH` is `lfs/`.
- `PATH`: **./data/lfs**: Where to store LFS files, only available when `STORAGE_TYPE` is `local`. If not set it fall back to deprecated LFS_CONTENT_PATH value in [server] section.
For `STORAGE_TYPE = minio`, the configurations can be found at [Storage Minio](#storage_minio), you can also define configurations like below to override derived or default values.
For `STORAGE_TYPE = minio`, the configurations can be found at [Storage Minio](#minio-storage-configuration-storage_minio), you can also define configurations like below to override derived or default values.
- `MINIO_BASE_PATH`: **attachments/**: Minio base path on the bucket only available when STORAGE_TYPE is `minio`
@@ -1500,7 +1662,7 @@ is `data/repo-archive` and the default of `MINIO_BASE_PATH` is `repo-archive/`.
- `PATH`: **./data/repo-archive**: Where to store archive files, only available when `STORAGE_TYPE` is `local`.
For `STORAGE_TYPE = minio`, the configurations can be found at [Storage Minio](#storage_minio), you can override some configurations like below.
For `STORAGE_TYPE = minio`, the configurations can be found at [Storage Minio](#minio-storage-configuration-storage_minio), you can override some configurations like below.
- `MINIO_BASE_PATH`: **repo-archive/**: Minio base path on the bucket only available when `STORAGE_TYPE` is `minio`
@@ -1517,14 +1679,16 @@ is `data/repo-archive` and the default of `MINIO_BASE_PATH` is `repo-archive/`.
## Proxy (`proxy`)
- `PROXY_ENABLED`: **false**: Enable the proxy if true, all requests to external via HTTP will be affected, if false, no proxy will be used even environment http_proxy/https_proxy
- `PROXY_URL`: **_empty_**: Proxy server URL, support http://, https//, socks://, blank will follow environment http_proxy/https_proxy
- `PROXY_URL`: **_empty_**: Proxy server URL, support http://, https://, socks5://, blank will follow environment http_proxy/https_proxy
- `PROXY_HOSTS`: **_empty_**: Comma separated list of host names requiring proxy. Glob patterns (*) are accepted; use ** to match all hosts.
When a proxy is configured, Gitea does not enforce the egress lists (`[security]`, `[migrations]`) on the proxied target, the proxy server is expected to restrict it.
- `DEFAULT_ACTIONS_URL`: **github**: Default platform to get action plugins, `github` for `https://github.com`, `self` for the current Gitea instance.
- `STORAGE_TYPE`: **local**: Storage type for actions logs, `local` for local disk or `minio` for s3 compatible object storage service, default is `local` or other name defined with `[storage.xxx]`
- `MINIO_BASE_PATH`: **actions_log/**: Minio base path on the bucket only available when STORAGE_TYPE is `minio`
- `LOG_RETENTION_DAYS`: **365**: Logs retention time in days. Old logs will be deleted after this period.
- `LOG_RETENTION_DAYS`: **365**: Days to keep logs. Old logs will be deleted after this period. 0 means keep forever.
- `LOG_COMPRESSION`: **zstd**: Log compression type, `none` for no compression, `zstd` for zstd compression.
Other compression types like `gzip` are NOT supported, since seekable stream is required for log view.
It's always recommended to use compression when using local disk as log storage if CPU or memory is not a bottleneck.
And for object storage services like S3, which is billed for requests, it would cause extra 2 times of get requests for each log view.
But it will save storage space and network bandwidth, so it's still recommended to use compression.
- `ARTIFACT_RETENTION_DAYS`: **90**: Default number of days to keep artifacts. Artifacts could have their own retention periods by setting the `retention-days` option in `actions/upload-artifact` step.
- `ARTIFACT_RETENTION_DAYS`: **90**: Days to keep artifacts. Old artifacts will be deleted after this period. 0 means keep forever.
Changes only apply to newly uploaded artifacts, existing ones keep the expiry stored when they were uploaded.
Artifacts could have their own retention periods by setting the `retention-days` option in `actions/upload-artifact` step.
- `ARTIFACT_PREVIEW_MAX_SIZE`: **10485760**: Maximum total artifact size in bytes that can be browsed or previewed in the web UI. Set to `0` to disable artifact previews or `-1` for no limit. Individual files are also limited by [`MAX_DISPLAY_FILE_SIZE`](#ui-ui).
- `RUN_RETENTION_DAYS`: **400**: Days to keep completed runs. Old runs and everything under them will be deleted after this period. 0 means keep forever.
- `ZOMBIE_TASK_TIMEOUT`: **10m**: Timeout to stop the task which have running status, but haven't been updated for a long time
- `ENDLESS_TASK_TIMEOUT`: **3h**: Timeout to stop the tasks which have running status and continuous updates, but don't end for a long time
- `ABANDONED_JOB_TIMEOUT`: **24h**: Timeout to cancel the jobs which have waiting status, but haven't been picked by a runner for a long time
- `SKIP_WORKFLOW_STRINGS`: **[skip ci],[ci skip],[no ci],[skip actions],[actions skip]**: Strings committers can place inside a commit message or PR title to skip executing the corresponding actions workflow
- `WORKFLOW_DIRS`: **.gitea/workflows,.github/workflows**: Comma-separated list of workflow directories, the first one to exist in a repo is used to find Actions workflow files.
- `SCOPED_WORKFLOW_DIRS`: **.gitea/scoped_workflows**: Comma-separated list of directories holding [scoped workflows](usage/actions/scoped-workflows.md) (workflows defined in a central source repository that run on other repositories). Must not overlap with `WORKFLOW_DIRS`. Leave empty so no directory is scanned for scoped workflows; none are found or run.
- `MAX_RERUN_ATTEMPTS`: **50**: Maximum number of attempts a single workflow run can have (initial run + reruns). Defaults value is 50. Set any positive value that fits your workflow needs.
`DEFAULT_ACTIONS_URL`indicates where the Gitea Actions runners should find the actions with relative path.
For example, `uses: actions/checkout@v4`means`https://github.com/actions/checkout@v4` since the value of `DEFAULT_ACTIONS_URL` is `github`.
And it can be changed to `self` to make it `root_url_of_your_gitea/actions/checkout@v4`.
Please note that using `self` is not recommended for most cases, as it could make names globally ambiguous.
Additionally, it requires you to mirror all the actions you need to your Gitea instance, which may not be worth it.
Therefore, please use `self` only if you understand what you are doing.
In earlier versions (`<= 1.19`), `DEFAULT_ACTIONS_URL` could be set to any custom URLs like `https://gitea.com` or `http://your-git-server,https://gitea.com`, and the default value was `https://gitea.com`.
However, later updates removed those options, and now the only options are `github` and `self`, with the default value being `github`.
However, if you want to use actions from other git server, you can use a complete URL in `uses` field, it's supported by Gitea (but not GitHub).
Like `uses: https://gitea.com/actions/checkout@v4` or `uses: http://your-git-server/actions/checkout@v4`.
`DEFAULT_ACTIONS_URL`sets where runners download actions without a host.
With `github`, `uses: actions/checkout@v4`resolves to`https://github.com/actions/checkout@v4`, with `self` to `<ROOT_URL>/actions/checkout@v4`.
`self` requires mirroring every action you use to your instance, so prefer naming the host per action instead: `uses: self:owner/repo@ref` for your instance, or a full URL like `uses: https://gitea.com/actions/checkout@v4` for any other host.
Customizing Gitea is typically done using the `CustomPath` folder - by default this is
the `custom` folder from the working directory (WorkPath), but may be different if your build has
the `custom` folder from the working directory (WorkPath), but may be different if your [installation](../installation/from-binary.md) has
set this differently. This is the central place to override configuration settings,
templates, etc. You can check the `CustomPath` using `gitea help`. You can also find
the path on the _Configuration_ tab in the _Site Administration_ page. You can override
@@ -23,7 +23,7 @@ the Linux Filesystem Standard. Gitea will attempt to create required folders, in
`custom/`. Distributions may provide a symlink for `custom` using `/etc/gitea/`.
Application settings can be found in file `CustomConf` which is by default,
`$GITEA_CUSTOM/conf/app.ini` but may be different if your build has set this differently.
`$GITEA_CUSTOM/conf/app.ini` but may be different if your [installation](../installation/from-binary.md) has set this differently.
Again `gitea help` will allow you review this variable and you can override it using the
`--config` option on the `gitea` binary.
@@ -85,6 +85,8 @@ To obtain any embedded file (including templates), the [`gitea embedded` tool](.
Be aware that any statement contained inside `{{` and `}}` are Gitea's template syntax and
shouldn't be touched without fully understanding these components.
Gitea pages are protected by Content-Security-Policy, use `<script nonce="{{ctx.CspScriptNonce}}">...</script>` for script tags.
### Customizing startpage / homepage
Copy [`home.tmpl`](https://github.com/go-gitea/gitea/blob/main/templates/home.tmpl) for your version of Gitea from `templates` to `$GITEA_CUSTOM/templates`.
@@ -136,10 +138,10 @@ The data is encoded and sent to the PlantUML server which generates the picture.
demo server at http://www.plantuml.com/plantuml, but if you (or your users) have sensitive data you
can set up your own [PlantUML server](https://plantuml.com/server) instead. To set up PlantUML rendering,
copy JavaScript files from https://gitea.com/davidsvantesson/plantuml-code-highlight and put them in your
`$GITEA_CUSTOM/public/assets/` folder. Then add the following to `custom/footer.tmpl`:
`$GITEA_CUSTOM/public/assets/` folder. Then add the following to `$GITEA_CUSTOM/templates/custom/footer.tmpl`:
```html
<script>
<script nonce="{{ctx.CspScriptNonce}}">
$(async () => {
if (!$('.language-plantuml').length) return;
await Promise.all([
@@ -165,52 +167,186 @@ Alice <-- Bob: Another authentication Response
The script will detect tags with `class="language-plantuml"`, but you can change this by providing a second argument to `parsePlantumlCodeBlocks`.
### Example: STL Preview
### Example: CAD Files Preview using Online 3D Viewer
You can display STL file directly in Gitea by adding:
You can implement CAD file preview inside your Gitea instance. This implemenation uses [`Online 3D Viewer`](https://github.com/kovacsv/Online3DViewer).
console.log('New Raw file link detected after delay:', fileName);
onPageChange();
}
};
const observer = new MutationObserver(() => {
// Delay execution by 300ms after last mutation
clearTimeout(timeoutId);
timeoutId = setTimeout(checkAndRun, 300);
});
observer.observe(document.body, {
childList: true,
subtree: true,
});
}
</script>
```
to the file `templates/custom/footer.tmpl`
#### Part 2: Add public files
You also need to download the content of the library [Madeleine.js](https://github.com/beige90/Madeleine.js) and place it under `$GITEA_CUSTOM/public/assets/` folder.
Now we need to download latest version of O3DV. Go to "$GITEA_CUSTOM/public/assets/".
Create folder using (and cd into it):
You should end-up with a folder structure similar to:
```
mkdir o3dv
cd o3dv
```
Copy latest release zip link from [`GitHub`](https://github.com/kovacsv/Online3DViewer/releases) (v0.18.0 as of now).
Now the last thing. Change permissions on the public folder:
```
chown -R git:git $GITEA_CUSTOM/public
```
Now we have everything ready! Restart your gitea instance to apply these changes and test it in your browser.
Sanity check. You should end-up with a folder structure similar to this:
```
$GITEA_CUSTOM/templates
@@ -218,49 +354,14 @@ $GITEA_CUSTOM/templates
`-- footer.tmpl
$GITEA_CUSTOM/public/assets/
-- Madeleine.js
|-- LICENSE
|-- README.md
|-- css
| |-- pygment_trac.css
| `-- stylesheet.css
|-- examples
| |-- ajax.html
| |-- index.html
| `-- upload.html
|-- images
| |-- bg_hr.png
| |-- blacktocat.png
| |-- icon_download.png
| `-- sprite_download.png
|-- models
| |-- dino2.stl
| |-- ducati.stl
| |-- gallardo.stl
| |-- lamp.stl
| |-- octocat.stl
| |-- skull.stl
| `-- treefrog.stl
`-- src
|-- Madeleine.js
|-- css
| `-- Madeleine.css
|-- icons
| |-- logo.png
| |-- madeleine.eot
| |-- madeleine.svg
| |-- madeleine.ttf
| `-- madeleine.woff
`-- lib
|-- MadeleineConverter.js
|-- MadeleineLoader.js
|-- detector.js
|-- stats.js
`-- three.min.js
-- o3dv
|-- o3dv.zip
|-- o3dv.license.md
|-- o3dv.min.js
|-- envmaps
\...
```
Then restart Gitea and open a STL file on your Gitea instance.
## Customizing Gitea mails
The `$GITEA_CUSTOM/templates/mail` folder allows changing the body of every mail of Gitea.
@@ -295,15 +396,16 @@ Starting with Gitea 1.20, you can customize the git configuration via the `git.c
### Enabling signed git pushes
To enable signed git pushes, set these two options:
[Signed pushes](https://git-scm.com/docs/git-push#Documentation/git-push.txt---signedtruefalseif-asked) allow clients to cryptographically sign the push operation itself (not just individual commits). To enable signed pushes, add the following to `app.ini`:
```ini
[git.config]
receive.advertisePushOptions = true
receive.certNonceSeed = <randomstring>
```
`certNonceSeed` should be set to a random string and be kept secret.
`certNonceSeed` should be set to a random string and be kept secret. It is used to generate anti-replay nonces. Gitea already sets `receive.advertisePushOptions = true` by default, so no additional configuration is needed. Note that Gitea does not read `/etc/gitconfig`, so this option must be set via `app.ini` as shown above.
On the client side, pushes can be signed via `git push --signed` or enabled permanently using `git config --global push.gpgSign if-asked`.
### Labels
@@ -321,7 +423,7 @@ The [legacy file format](https://github.com/go-gitea/gitea/blob/main/options/lab
`#hex-color label name ; label description`
For more information, see the [labels documentation](usage/labels.md).
For more information, see the [labels documentation](../usage/issues-prs/labels.md).
### Licenses
@@ -380,13 +482,27 @@ To make a custom theme available to all users:
The value of `$GITEA_CUSTOM` of your instance can be queried by calling `gitea help` and looking up the value of "CustomPath".
2. Add `<theme-name>` to the comma-separated list of setting `THEMES` in `app.ini`, or leave `THEMES` empty to allow all themes.
A custom theme file named `theme-my-theme.css` will be displayed as `my-theme` on the user's theme selection page.
It could add theme meta information into the custom theme CSS file to provide more information about the theme.
If a custom theme is a dark theme, please set the global css variable `--is-dark-theme: true` in the `:root` block.
This allows Gitea to adjust the Monaco code editor's theme accordingly.
An "auto" theme could be implemented by using "theme-gitea-auto.css" as a reference.
```css
gitea-theme-meta-info {
--theme-display-name: "My Awesome Theme"; /* this theme will be display as "My Awesome Theme" on the UI */
}
:root {
--is-dark-theme: true; /* if it is a dark theme */
--color-primary: #112233;
/* more custom theme variables ... */
}
```
Community themes are listed in [gitea/awesome-gitea#themes](https://gitea.com/gitea/awesome-gitea#themes).
The default theme sources can be found [here](https://github.com/go-gitea/gitea/blob/main/web_src/css/themes).
If your custom theme is considered a dark theme, set the global css variable `--is-dark-theme` to `true`.
This allows Gitea to adjust the Monaco code editor's theme accordingly.
Gitea has mailer functionality for sending transactional emails (such as registration confirmation). It can be configured to either use Sendmail (or compatible MTAs like Postfix and msmtp) or directly use SMTP server.
:::note
Be sure to set ENABLE_NOTIFY_MAIL=true to allow Gitea to send email notifications. Check the [Config Cheat Sheet](../administration/config-cheat-sheet.md#service-service) for details.
:::
## Using Sendmail
Use `sendmail` command as mailer.
@@ -41,7 +45,7 @@ ENABLED = true
FROM = gitea@mydomain.com
PROTOCOL = smtps
SMTP_ADDR = mail.mydomain.com
SMTP_PORT = 587
SMTP_PORT = 465
USER = gitea@mydomain.com
PASSWD = `password`
```
@@ -56,7 +60,7 @@ For the full list of options check the [Config Cheat Sheet](../administration/co
Authentication is only supported when the SMTP server communication is encrypted with TLS or `HOST=localhost`. TLS encryption can be through:
:::
- STARTTLS (also known as Opportunistic TLS) via port 587. Initial connection is done over cleartext, but then be upgraded over TLS if the server supports it.
- STARTTLS (also known as Opportunistic TLS) via port 587 with `PROTOCOL=smtp+starttls`. Initial connection is done over cleartext, but then be upgraded over TLS if the server supports it.
- SMTPS connection (SMTP over TLS) via the default port 465. Connection to the server use TLS from the beginning.
- Forced SMTPS connection with `PROTOCOL=smtps`. (These are both known as Implicit TLS.)
This is due to protections imposed by the Go internal libraries against STRIPTLS attacks.
@@ -82,6 +86,21 @@ PROTOCOL = smtps
Note that you'll need to create and use an [App password](https://support.google.com/accounts/answer/185833?hl=en) by enabling 2FA on your Google
account. You won't be able to use your Google account password directly.
An app password is the only option: Google
[turned off access for apps that sign in with a password alone](https://workspaceupdates.googleblog.com/2023/09/winding-down-google-sync-and-less-secure-apps-support.html)
in 2024 and asks applications to use OAuth instead, which Gitea's mailer does
not implement — it authenticates with `CRAM-MD5`, `PLAIN` or `LOGIN`.
Two things to keep in mind:
- app passwords require 2-step verification on the account, and a Google
Workspace administrator can switch them off for the whole domain. If they are
- port `587` works as well, with `SMTP_PORT = 587` and
`PROTOCOL = smtp+starttls`.
### ProtonMail
This feature is currently only available for select Proton for Business customers and those with Visionary and Family plans with custom domain addresses. See [ProtonMail's SMTP documentation](https://proton.me/support/smtp-submission) for more information. This limitation can be circumvented by using the ProtonMail Bridge application.
@@ -16,6 +16,8 @@ The logging configuration of Gitea mainly consists of 3 types of components:
There is a fully functional log output by default, so it is not necessary to define one.
Security-relevant events (repository transfers, token changes, impersonation, and similar) are recorded separately. See [Audit Logging](audit-logging.md).
## Collecting Logs for Help
To collect logs for help and issue report, see [Support Options](help/support.md).
1. Set `[server] ROOT_URL = https://git.example.com/` in your `app.ini` file.
2. Make the reverse-proxy pass `https://git.example.com/foo` to `http://gitea:3000/foo`.
3. Make sure the reverse-proxy does not decode the URI. The request `https://git.example.com/a%2Fb` should be passed as `http://gitea:3000/a%2Fb`.
4. Make sure `Host` and `X-Fowarded-Proto` headers are correctly passed to Gitea to make Gitea see the real URL being visited.
4. Make sure `Host` and `X-Forwarded-Proto` headers are correctly passed to Gitea to make Gitea see the real URL being visited.
5. Make sure your webserver has a certificate, including all intermediate and RootCA certificates, for `git clone` and `git pull` to work.
6. Gitea uses a WebSocket connection at `/-/ws` for real-time updates like the notification counter. Make sure the reverse-proxy talks HTTP/1.1 and forwards the `Upgrade` and `Connection` headers, see the examples below. Without it, Gitea falls back to periodic polling.
### Use a sub-path
@@ -40,6 +42,7 @@ server {
location / {
client_max_body_size 512M;
proxy_pass http://localhost:3000;
proxy_http_version 1.1;
proxy_set_header Connection $http_connection;
proxy_set_header Upgrade $http_upgrade;
proxy_set_header Host $host;
@@ -66,6 +69,7 @@ server {
rewrite ^ $request_uri;
rewrite ^/(gitea($|/))?(.*) /$3 break;
proxy_pass http://127.0.0.1:3000$uri;
proxy_http_version 1.1;
# other common HTTP headers, see the "Nginx" config section above
proxy_set_header Connection $http_connection;
@@ -112,6 +116,7 @@ server {
}
location / {
# other directives, see the "Nginx" config section above
proxy_pass http://localhost:3000;
}
}
@@ -128,6 +133,7 @@ server {
server_name git.example.com;
location / {
# other directives, see the "Nginx" config section above
proxy_pass http://localhost:3000;
}
}
@@ -149,6 +155,20 @@ server {
}
```
## Nginx Proxy Manager
If you are using Nginx Proxy Manager to serve your Gitea instance it differs slighly from the raw Nginx.
It is [adding some directives](https://github.com/NginxProxyManager/nginx-proxy-manager/blob/master/docker/rootfs/etc/nginx/conf.d/include/proxy.conf) for a custom location by default, so you have to skip these from the above mentioned Nginx config. Otherwise Nginx will produce `400 bad request` error due to duplicated directives (`proxy_set_header Host $host` is the particularly problematic one).
So while creating the `/` Custom location, just add the following line to it's configuration:
```nginx
client_max_body_size 512M;
```
Enable `Websockets Support` on the proxy host, Nginx Proxy Manager then adds the WebSocket directives itself. Adding `proxy_http_version` on top of that makes Nginx refuse to start.
## Apache HTTPD
If you want Apache HTTPD to serve your Gitea instance, you can add the following to your Apache HTTPD configuration (usually located at `/etc/apache2/httpd.conf` in Ubuntu):
@@ -159,13 +179,15 @@ If you want Apache HTTPD to serve your Gitea instance, you can add the following
RequestHeader set "X-Forwarded-Proto" expr=%{REQUEST_SCHEME}
</VirtualHost>
```
:::note
The following Apache HTTPD mods must be enabled: `proxy`, `proxy_http`.
`upgrade=websocket` requires Apache HTTPD 2.4.48 or later. On older versions you **MUST** remove it, otherwise Apache HTTPD will not start.
:::
If you wish to use Let's Encrypt with webroot validation, add the line `ProxyPass /.well-known !` before `ProxyPass` to disable proxying these requests to Gitea.
@@ -183,7 +205,7 @@ In case you already have a site, and you want Gitea to share the domain name, yo
</Proxy>
AllowEncodedSlashes NoDecode
# Note: no trailing slash after either /git or port
RequestHeader set "X-Forwarded-Proto" expr=%{REQUEST_SCHEME}
</VirtualHost>
@@ -192,7 +214,55 @@ In case you already have a site, and you want Gitea to share the domain name, yo
Then you **MUST** set something like `[server] ROOT_URL = http://git.example.com/git/` correctly in your configuration.
:::note
The following Apache HTTPD mods must be enabled: `proxy`, `proxy_http`.
The following Apache HTTPD mods must be enabled: `proxy`, `proxy_http`.`upgrade=websocket` requires Apache HTTPD 2.4.48 or later, remove it on older versions.
:::
## Apache HTTPD and serve static resources directly
We can tune the performance in splitting requests into categories static and dynamic.
CSS files, JavaScript files, images and web fonts are static content.
The front page, a repository view or issue list is dynamic content.
Apache HTTPD can serve static resources directly and proxy only the dynamic requests to Gitea.
Download a snapshot of the Gitea source repository to `/path/to/gitea/`.
After this, run `make frontend` in the repository directory to generate the static resources. We are only interested in the `public/` directory for this task, so you can delete the rest.
(You will need to have [Node with npm](https://nodejs.org/en/download/) and `make` installed to generate the static resources)
Depending on the scale of your user base, you might want to split the traffic to two distinct servers,
or use a cdn for the static files.
### Single node and single domain
Set `[server] STATIC_URL_PREFIX = /_/static` in your configuration.
```apacheconf
<VirtualHost *:80>
...
<Proxy *>
Order allow,deny
Allow from all
</Proxy>
ProxyPass /_/static/ !
Alias /_/static/ /path/to/gitea/public/
<Directory /path/to/gitea/public/>
Options FollowSymlinks
AllowOverride None
Require all granted
</Directory>
AllowEncodedSlashes NoDecode
# Note: no trailing slash after either /git or port
RequestHeader set "X-Forwarded-Proto" expr=%{REQUEST_SCHEME}
</VirtualHost>
```
:::note
The following Apache HTTPD mods must be enabled: `proxy`, `proxy_http`. `upgrade=websocket` requires Apache HTTPD 2.4.48 or later, remove it on older versions.
:::
## Caddy
@@ -229,11 +299,13 @@ If you wish to run Gitea with IIS. You will need to setup IIS with URL Rewrite a
- Install Application Request Routing (ARR for short) either by using the Microsoft Web Platform Installer 5.1 (WebPI) or downloading the extension from [IIS.net](https://www.iis.net/downloads/microsoft/application-request-routing)
- Once the module is installed in IIS, you will see a new Icon in the IIS Administration Console called URL Rewrite.
- Install the `WebSocket Protocol` Windows feature, otherwise ARR cannot proxy Gitea's WebSocket endpoint.
- Open the IIS Manager Console and click on the `Gitea Proxy` Website from the tree view on the left. Select and double click the URL Rewrite Icon from the middle pane to load the URL Rewrite interface.
- Choose the `Add Rule` action from the right pane of the management console and select the `Reverse Proxy Rule` from the `Inbound and Outbound Rules` category.
- In the Inbound Rules section, set the server name to be the host that Gitea is running on with its port. e.g. if you are running Gitea on the localhost with port 3000, the following should work: `127.0.0.1:3000`
- Enable SSL Offloading
- In the Outbound Rules, ensure `Rewrite the domain names of the links in HTTP response` is set and set the `From:` field as above and the `To:` to your external hostname, say: `git.example.com`
- Turn `Reverse rewrite host in response headers` off, in the `Server Proxy Settings` of Application Request Routing (select the server node in the IIS Manager, open `Application Request Routing Cache`, then `Server Proxy Settings` in the right pane). While it is on, ARR rewrites the host of the `Location` header back to the internal one, so the redirects Gitea sends after a sign-in or a form submission send the browser to `127.0.0.1:3000` instead of `git.example.com`.
- Now edit the `web.config` for your website to match the following: (changing `127.0.0.1:3000` and `git.example.com` as appropriate)
Gitea will verify GPG commit signatures in the provided tree by
Gitea will verify gpg/ssh commit signatures in the provided tree by
checking if the commits are signed by a key within the Gitea database,
or if the commit matches the default key for Git.
Additionally Gitea will verify commits signed by ssh keys, which public keys are part of [`TRUSTED_SSH_KEYS`](#general-configuration).
Keys are not checked to determine if they have expired or revoked.
Keys are also not checked with keyservers.
@@ -46,6 +48,11 @@ for GPG - in particular it is probably advisable to only install a
signing secret subkey without the master signing and certifying secret
key.
## Installing and generating a SSH key for Gitea
You can run `ssh-keygen -t ed25519 -f gitea-signing-key` to generate the private/public keypair for commit signing without any password. Usually you would store the key next to the gitea configuration, then point `SIGNING_KEY` to the generated public key `/path/to/gitea-signing-key.pub`. Gitea generates all its commits using the server `git` command at present - and therefore the server `ssh-keygen` will be used for
signing (if configured.)
## General Configuration
Gitea's configuration for signing can be found with the
For SSH commit signing, you need to specify the `SIGNING_FORMAT` to `ssh` instead of the default `openpgp`. `SIGNING_NAME` and `SIGNING_EMAIL` are required for verifing the signatures.
This looks like this:
```ini
...
[repository.signing]
SIGNING_KEY = /path/to/gitea-signing-key.pub
SIGNING_NAME =
SIGNING_EMAIL =
SIGNING_FORMAT = ssh
INITIAL_COMMIT = always
CRUD_ACTIONS = pubkey, twofa, parentsigned
WIKI = never
MERGES = pubkey, twofa, basesigned, commitssigned
...
```
- `/path/to/gitea-signing-key` is expected to be the private key for `/path/to/gitea-signing-key.pub` [see here how to generate a new ssh keypair](#installing-and-generating-a-ssh-key-for-gitea).
- `TRUSTED_SSH_KEYS = ssh-<algorithm> <key>` or `TRUSTED_SSH_KEYS = ssh-<algorithm> <key1>, ssh-<algorithm> <key2>` can be used for rotating the global ssh signing key to continue verifying commits signed by the previous keys.
### `SIGNING_KEY`
The first option to discuss is the `SIGNING_KEY`. There are three main
options:
- `none` - this prevents Gitea from signing any commits
- `default` - Gitea will default to the key configured within `git config`
- `default` - Gitea will default to the gpg key configured within `git config`
- `KEYID` - Gitea will sign commits with the gpg key with the ID
`KEYID`. In this case you should provide a `SIGNING_NAME` and
`SIGNING_EMAIL` to be displayed for this key.
- `/path/to/gitea-signing-key.pub` - Gitea will sign commits with the ssh key without the `.pub` suffix `/path/to/gitea-signing-key`. In this case you should provide a `SIGNING_NAME` and
`SIGNING_EMAIL` to be displayed for this key and set `SIGNING_FORMAT` to `ssh`.
The `default` option will interrogate `git config` for
`commit.gpgsign` option - if this is set, then it will use the results
@@ -97,6 +129,10 @@ Related home files for git command (like `.gnupg`) should also be put in Gitea's
If you like to keep the `.gnupg` directory outside of `{[git].HOME_PATH}/`, consider setting the `$GNUPGHOME` environment variable to your preferred location, otherwise Gitea will use the gpg keys only under `{[git].HOME_PATH}/.gnupg`.
:::
:::warning
The default option and repository specific signing keys are not supported for ssh keys
:::
### `INITIAL_COMMIT`
This option determines whether Gitea should sign the initial commit
@@ -168,3 +204,9 @@ In cases where there is a repository specific key this can be obtained from:
```sh
/api/v1/repos/:username/:reponame/signing-key.gpg
```
For ssh commit signing the default ssh public key can be obtained via the API at:
specification (not RFC 9421). The signature is sent in the `Signature` header,
and SSH certificates additionally include an `x-ssh-certificate` header. The
official [go-sdk](https://gitea.com/gitea/go-sdk) implements this flow if you
need a reference implementation.
## Generating and listing API tokens
API tokens can be created either in the user interface or via the API. Tokens have by default limited permissions and it is important to create tokens with the correct permissions for your task.
### User Interface
Tokens can be created via the `Manage Access Tokens` dialog, accessed via `User Settings` / `Applications` or via the link `gitea-domain.example/user/settings/applications`. The interface allows you to create tokens and manage their permissions via the `Select permissions` sub-menu.
Once created, the token is displayed in a toast message above the `Manage Access Tokens` dialog. Please note, that you can view this toast only once and it is not possible to redisplay the token for security reasons.
### Token API
A new token can be generated with a `POST` request to
`/users/:name/tokens`.
@@ -35,8 +54,10 @@ Note that `/users/:name/tokens` is a special endpoint and requires you
to authenticate using `BasicAuth` and a password, as follows:
By default, this creates a token with very limited permissions. To complete your tasks, your token may require extra permissions. These permissions are created via the json variable `scopes`, which takes an array of permissions as strings. Eg.: `"scopes":["all"]`. Possible permissions, as reflected in the user interface are:
- `activitypub`
- `admin`
- `issue`
- `misc`
- `notification`
- `organization`
- `package`
- `repository`
- `user`
Each permission may be set to `read` or `write`. `write` implies both Read & Write. To set permissions you write the permissions string as `<read or write>:<permission name>`, eg.: `write:package` or `read:notification`. A properly formatted API call may look like:
@@ -33,7 +33,7 @@ then the config file (app.ini) should exists in the "custom/conf" directory of y
Some package vendors might use "/etc/gitea" to store the config file, while some others don't.
You could manually find the config file (app.ini) by checking Gitea's startup logs
or reading the Gitea Web's Site Administrator -> Confugiraton Summary.
or reading the Gitea Web's Site Administrator -> Configuration Summary.
If you are using some isolated enviroments like container (docker),
the path you see usually is not what it is in the host's filesystem.
@@ -330,9 +330,9 @@ It is highly recommended to back-up your database before running these commands.
- Add your (bare) repositories to the correct spot for your configuration (`repository.ROOT`), ensuring they are in the correct layout `<REPO_ROOT>/[user]/[repo].git`.
- **Note:** the directory names must be lowercase.
- You can also check `<ROOT_URL>/admin/config` for the repository root path.
- You can also check `<ROOT_URL>/-/admin/config` for the repository root path.
- Ensure that the user/org exists that you want to adopt repositories for.
- As an admin, go to `<ROOT_URL>/admin/repos/unadopted` and search.
- As an admin, go to `<ROOT_URL>/-/admin/repos/unadopted` and search.
- Users can also be given similar permissions via config [`ALLOW_ADOPTION_OF_UNADOPTED_REPOSITORIES`](../administration/config-cheat-sheet.md#repository-repository).
- If the above steps are done correctly, you should be able to select repositories to adopt.
- If no repositories are found, enable [debug logging](../administration/config-cheat-sheet.md#repository-repository) to check for any specific errors.
@@ -13,7 +13,7 @@ Gitea was originally forked from [Gogs](https://gogs.io) and almost all the code
:::warning
Gitea does not sent or cherry-picked commits from upstream, so there is no guarantee it will work if you upgrade from Gogs to Gitea. The recommended method is to migrate repositories from Gogs to Gitea.
Gitea does not send commits to upstream or cherry-pick commits from it, so there is no guarantee it will work if you upgrade from Gogs to Gitea. The recommended method is to migrate repositories from Gogs to Gitea.
:::
@@ -31,11 +31,11 @@ You can try it out using [the online demo](https://demo.gitea.com).
- **Code Hosting**
Gitea supports creating and managing repositories, browsing commit history and code files, reviewing and merging code submissions, managing collaborators, handling branches, and more. It also supports many common Git features such as tags, Cherry-pick, hooks, integrated collaboration tools, and more.
Gitea supports creating and managing repositories, browsing commit history and code files, reviewing and merging code submissions, managing collaborators, handling branches, and more. It also supports many common Git features such as tags, cherry-picking, hooks, integrated collaboration tools, and more.
- **Lightweight and Fast**
One of Gitea's design goals is to be lightweight and fast in response. Unlike some large code hosting platforms, it remains lean, performs well in terms of speed, and is suitable for resource-limited server environments. Due to its lightweight design, Gitea has relatively low resource consumption and performs well in resource-constrained environments.
One of Gitea's design goals is to be lightweight and fast in response. Unlike some large code hosting platforms, it remains lean, performs well in terms of speed, and is suitable for resource-limited server environments.
- **Easy Deployment and Maintenance**
@@ -43,7 +43,7 @@ You can try it out using [the online demo](https://demo.gitea.com).
- **Security**
Gitea places a strong emphasis on security, offering features such as user permission management, access control lists, and more to ensure the security of code and data.
Gitea places a strong emphasis on security, offering features such as user permission management, access control lists, [audit logging](administration/audit-logging.md), and more to ensure the security of code and data.
- **Code Review**
@@ -105,6 +105,6 @@ For more detailed information, please refer to: https://docs.gitea.com/installat
@@ -24,9 +24,10 @@ All steps below requires that the database engine of your choice is installed on
## MySQL/MariaDB
1. For remote database setup, you will need to make MySQL listen to your IP address. Edit`bind-address` option on `/etc/mysql/my.cnf` on database instance to:
1. For remote database setup, you will need to make MySQL listen to your IP address. Configure the`bind-address` option in `/etc/mysql/my.cnf` on database instance by adding the following lines:
Note that the above command will download Gitea @version@ for 64-bit Linux.
## Verify GPG signature
## Verify signature
### Sigstore
Starting with v1.27 Gitea signs all binaries using [sigstore](https://sigstore.dev/) to prevent against unwanted modification of binaries.
To validate the binary, download the bundle file which ends in `sigstore.json` for the binary you downloaded and use the [cosign](https://docs.sigstore.dev/cosign/system_config/installation/) command line tool.
If the command outputs `Verified OK`, binary was not modified.
:::note
The above command will match any release workflow. You may choose to restrict the identity further by requiring a specific branch or workflow to be matched. For example this url will match only nightly builds made on main branch: `https://github.com/go-gitea/gitea/.github/workflows/release-nightly.yml@refs/heads/main`
:::
### GPG
Gitea signs all binaries with a [GPG key](https://keys.openpgp.org/search?q=teabot%40gitea.io) to prevent against unwanted modification of binaries.
To validate the binary, download the signature file which ends in `.asc` for the binary you downloaded and use the GPG command line tool.
A script to enable bash-completion can be found at [`contrib/autocompletion/bash_autocomplete`](https://raw.githubusercontent.com/go-gitea/gitea/main/contrib/autocompletion/bash_autocomplete). This can be copied to `/usr/share/bash-completion/completions/gitea`
or sourced within your `.bashrc`.
### Adding shell autocompletion (from 1.25)
Similarly a script for zsh-completion can be found at [`contrib/autocompletion/zsh_autocomplete`](https://raw.githubusercontent.com/go-gitea/gitea/main/contrib/autocompletion/zsh_autocomplete). This can be copied to `/usr/share/zsh/_gitea` or sourced within your
`.zshrc`.
Shell completion can be generated directly from binary with:
```sh
gitea completion <shell>
```
YMMV and these scripts may need further improvement.
Supported values for `<shell>` are `bash`, `fish`, `pwsh` and `zsh`. Details on how to load the completion for your shell can be found in the completion command help.
## Running Gitea
After you complete the above steps, you can run Gitea two ways:
:::note
On a distribution with SELinux in enforcing mode (Fedora, RHEL, CentOS Stream,
Rocky Linux, AlmaLinux and others), the binary and the data directory need
labels the default policy does not give them, and binding to a port below 1024
takes an extra step. See [running Gitea with SELinux](selinux.md).
:::
### 1. Creating a service file to start Gitea automatically (recommended)
See how to create [Linux service](installation/run-as-service-in-ubuntu.md)
@@ -214,9 +239,7 @@ As of v1.8, there is a problem with the arm7 version of Gitea, and it doesn't ru
It is recommended to switch to the arm6 version, which has been tested and shown to work on Raspberry Pis and similar devices.
<!---
please remove after fixing the arm7 bug
--->
{/* please remove after fixing the arm7 bug */}
### Git error after updating to a new version of Gitea
If during the update, the binary file name has been changed to a new version of Gitea,
This can be combined with `CC`, `GOOS`, and `GOARCH` as above.
### Adding bash/zsh autocompletion (from 1.19)
### Adding shell autocompletion (from 1.25)
A script to enable bash-completion can be found at [`contrib/autocompletion/bash_autocomplete`](https://raw.githubusercontent.com/go-gitea/gitea/main/contrib/autocompletion/bash_autocomplete). This should be altered as appropriate and can be `source` in your `.bashrc`
or copied as `/usr/share/bash-completion/completions/gitea`.
Shell completion can be generated directly from binary with:
```sh
gitea completion <shell>
```
Similarly, a script for zsh-completion can be found at [`contrib/autocompletion/zsh_autocomplete`](https://raw.githubusercontent.com/go-gitea/gitea/main/contrib/autocompletion/zsh_autocomplete). This can be copied to `/usr/share/zsh/_gitea` or sourced within your
`.zshrc`.
YMMV and these scripts may need further improvement.
Supported values for `<shell>` are `bash`, `fish`, `pwsh` and `zsh`. Details on how to load the completion for your shell can be found in the completion command help.
@@ -12,7 +12,7 @@ You can run Gitea as a Linux service, using either systemd or supervisor. The st
## Using systemd
Copy the sample [gitea.service](https://github.com/go-gitea/gitea/blob/main/contrib/systemd/gitea.service) to `/etc/systemd/system/gitea.service`, then edit the file with your favorite editor.
Copy the sample [gitea.service](https://github.com/go-gitea/gitea/blob/main/contrib/service/systemd/gitea.service) to `/etc/systemd/system/gitea.service`, then edit the file with your favorite editor.
Uncomment any service that needs to be enabled on this host, such as MySQL.
@@ -56,115 +58,15 @@ Note that the volume should be owned by the user/group with the UID/GID specifie
sudo chown 1000:1000 config/ data/
```
> If you don't give the volume correct permissions, the container may not start.
> If you don't give the volume correct permissions, the container may present the following errors in the logs:
For a stable release you could use `:latest-rootless`, `:1-rootless` or specify a certain release like `:@dockerVersion@-rootless`, but if you'd like to use the latest development version then `:nightly-rootless` would be an appropriate tag. If you'd like to run the latest commit from a release branch you can use the `:1.x-nightly-rootless` tag, where x is the minor version of Gitea. (e.g. `:1.16-nightly-rootless`)
## Custom port
To bind the integrated ssh and the webserver on a different port, adjust
the port section. It's common to just change the host port and keep the ports within
For a stable release you could use `:latest-rootless`, `:1-rootless` or specify a certain release like `:@dockerVersion@-rootless`, but if you'd like to use the latest development version then `:main-nightly-rootless` would be an appropriate tag. If you'd like to run the latest commit from a release branch you can use the `:1.x-nightly-rootless` tag, where x is the minor version of Gitea. (e.g. `:1.27-nightly-rootless`)
## Named volumes
@@ -174,8 +76,6 @@ create the required volume. You don't need to worry about permissions with
named volumes; Docker will deal with that automatically.
- GITEA__mailer__PASSWD="""${GITEA__mailer__PASSWD:?GITEA__mailer__PASSWD not set}"""
```
To set required TOKEN and SECRET values, consider using Gitea's built-in [generate utility functions](../administration/command-line.md#generate).
# SSH Container Passthrough
Since SSH is running inside the container, SSH needs to be passed through from the host to the container if SSH support is desired. One option would be to run the container SSH on a non-standard port (or moving the host port to a non-standard port). Another option which might be more straightforward is to forward SSH commands from the host to the container. This setup is explained in the following.
This guide assumes that you have created a user on the host called `git` with permission to run `docker exec`, and that the Gitea container is called `gitea`. You will need to modify that user's shell to forward the commands to the `sh` executable inside the container, using `docker exec`.
First, create the file `/usr/local/bin/gitea-shell` on the host, with the following contents:
```bash
#!/bin/sh
/usr/bin/docker exec -i --env SSH_ORIGINAL_COMMAND="$SSH_ORIGINAL_COMMAND" gitea sh "$@"
```
Note that `gitea` in the docker command above is the name of the container. If you named yours differently, don't forget to change that.
You should also make sure that you’ve set the permissions of the shell wrapper correctly:
```bash
sudo chmod +x /usr/local/bin/gitea-shell
```
Once the wrapper is in place, you can make it the shell for the `git` user:
```bash
sudo usermod -s /usr/local/bin/gitea-shell git
```
Now that all the SSH commands are forwarded to the container, you need to set up the SSH authentication on the host. This is done by leveraging the [SSH AuthorizedKeysCommand](../administration/command-line.md#keys) to match the keys against those accepted by Gitea. Add the following block to `/etc/ssh/sshd_config`, on the host:
@@ -12,29 +12,34 @@ Gitea provides automatically updated Docker images within its Docker Hub organiz
possible to always use the latest stable tag or to use another service that handles updating
Docker images.
This reference setup guides users through the setup based on `docker-compose`, but the installation
of `docker-compose` is out of scope of this documentation. To install `docker-compose` itself, follow
the official [install instructions](https://docs.docker.com/compose/install/).
This reference setup guides users through the setup based on `dockercompose`.
Dockercompose has been included in the Docker Engine since the shift to Compose v2.
If, for some reason, compose is not available on your system, follow the official [install instructions](https://docs.docker.com/compose/install/).
This document targets for the default rootful image. For the rootless image,
see "[Installation with Docker (rootless)](./with-docker-rootless.md)"
ATTENTION: the rootful/rootless images are not compatible with the other.
If you have chosen one, you should always use the same one,
don't switch to the other one by changing the compose file's `image` value.
## Basics
The most simple setup just creates a volume and a network and starts the `docker.io/gitea/gitea:latest`
The most simple setup just creates a volume and a network and starts the `docker.gitea.com/gitea:latest`
image as a service. Since there is no database available, one can be initialized using SQLite3.
Create a directory like `gitea` and paste the following content into a file named `docker-compose.yml`.
Note that the volume should be owned by the user/group with the UID/GID specified in the config file.
If you don't give the volume correct permissions, the container may not start.
For a stable release you can use `:latest`, `:1` or specify a certain release like `:@dockerVersion@`, but if you'd like to use the latest development version of Gitea then you could use the `:nightly` tag. If you'd like to run the latest commit from a release branch you can use the `:1.x-nightly` tag, where x is the minor version of Gitea. (e.g. `:1.16-nightly`)
For a stable release you can use `:latest`, `:1` or specify a certain release like `:@dockerVersion@`, but if you'd like to use the latest development version of Gitea then you could use the `:main-nightly` tag. If you'd like to run the latest commit from a release branch you can use the `:1.x-nightly` tag, where x is the minor version of Gitea. (e.g. `:1.27-nightly`)
```yaml
version: "3"
networks:
gitea:
external: false
services:
server:
image: docker.io/gitea/gitea:@dockerVersion@
image: docker.gitea.com/gitea:@dockerVersion@
container_name: gitea
environment:
- USER_UID=1000
@@ -58,15 +63,13 @@ the port section. It's common to just change the host port and keep the ports wi
the container like they are.
```diff
version: "3"
networks:
gitea:
external: false
services:
server:
image: docker.io/gitea/gitea:@dockerVersion@
image: docker.gitea.com/gitea:@dockerVersion@
container_name: gitea
environment:
- USER_UID=1000
@@ -93,15 +96,13 @@ To start Gitea in combination with a MySQL database, apply these changes to the
`docker-compose.yml` file created above.
```diff
version: "3"
networks:
gitea:
external: false
services:
server:
image: docker.io/gitea/gitea:@dockerVersion@
image: docker.gitea.com/gitea:@dockerVersion@
container_name: gitea
environment:
- USER_UID=1000
@@ -143,16 +144,21 @@ services:
To start Gitea in combination with a PostgreSQL database, apply these changes to
the `docker-compose.yml` file created above.
```diff
version: "3"
:::note
The example pins a major version on purpose: PostgreSQL does not upgrade its data
directory on its own, moving to a newer major version later requires `pg_upgrade`
or a dump and restore. Start a new instance on the newest version Gitea supports,
see [database preparation](database-preparation.md) for the supported versions.
:::
```diff
networks:
gitea:
external: false
services:
server:
image: docker.io/gitea/gitea:@dockerVersion@
image: docker.gitea.com/gitea:@dockerVersion@
container_name: gitea
environment:
- USER_UID=1000
@@ -176,7 +182,7 @@ services:
+ - db
+
+ db:
+ image: docker.io/library/postgres:14
+ image: docker.io/library/postgres:18
+ restart: always
+ environment:
+ - POSTGRES_USER=gitea
@@ -185,7 +191,7 @@ services:
+ networks:
+ - gitea
+ volumes:
+ - ./postgres:/var/lib/postgresql/data
+ - ./postgres:/var/lib/postgresql
```
## Named volumes
@@ -196,8 +202,6 @@ create the required volume. You don't need to worry about permissions with
named volumes; Docker will deal with that automatically.
```diff
version: "3"
networks:
gitea:
external: false
@@ -208,7 +212,7 @@ networks:
+
services:
server:
image: docker.io/gitea/gitea:@dockerVersion@
image: docker.gitea.com/gitea:@dockerVersion@
container_name: gitea
restart: always
networks:
@@ -227,17 +231,11 @@ MySQL or PostgreSQL containers will need to be created separately.
## Startup
:::note
From July 2023 Compose V1 stopped receiving updates. It's also no longer available in new releases of Docker Desktop.
To start this setup based on `docker compose`, execute `docker compose up -d`,
to launch Gitea in the background. Using `docker compose ps` will show if Gitea
started properly. Logs can be viewed with `docker compose logs`.
Compose V2 is included with all currently supported versions of Docker Desktop. Please use V2 to do below operations.
:::
To start this setup based on `docker-compose`, execute `docker-compose up -d`,
to launch Gitea in the background. Using `docker-compose ps` will show if Gitea
started properly. Logs can be viewed with `docker-compose logs`.
To shut down the setup, execute `docker-compose down`. This will stop
To shut down the setup, execute `docker compose down`. This will stop
and kill the containers. The volumes will still exist.
:::note
@@ -247,9 +245,9 @@ If using a non-3000 port on http, change app.ini to match
## Installation
After starting the Docker setup via `docker-compose`, Gitea should be available using a
After starting the Docker setup via `dockercompose`, Gitea should be available using a
favorite browser to finalize the installation. Visit http://server-ip:3000 and follow the
installation wizard. If the database was started with the `docker-compose` setup as
installation wizard. If the database was started with the `dockercompose` setup as
documented above, please note that `db` must be used as the database hostname.
## Configure the user inside Gitea using environment variables
@@ -266,6 +264,8 @@ files; for named volumes, this is done through another container or by direct ac
`/var/lib/docker/volumes/gitea_gitea/_data`. The configuration file will be saved at
`/data/gitea/conf/app.ini` after the installation.
Example: Analogous to the non-docker-installation customization linked above, you can create a `/public` folder within `/data/gitea` and place your custom `robots.txt` there which will then be served normally.
## Upgrading
:::warning
@@ -277,17 +277,18 @@ To upgrade your installation to the latest release:
```bash
# Edit `docker-compose.yml` to update the version, if you have one specified
# Pull new images
docker-compose pull
dockercompose pull
# Start a new container, automatically removes old one
docker-compose up -d
dockercompose up -d
```
## Managing Deployments With Environment Variables
In addition to the environment variables above, any settings in `app.ini` can be set
or overridden with an environment variable of the form: `GITEA__SECTION_NAME__KEY_NAME`.
These settings are applied each time the docker container starts, and won't be passed into Gitea's sub-processes.
Full information [here](https://github.com/go-gitea/gitea/tree/master/contrib/environment-to-ini).
or overridden with an environment variable of the form: `GITEA__section_name__KEY_NAME=value`.
These settings are applied each time the docker container starts by `environment-to-ini` command
(a warpper of `gitea config edit-ini`), and won't be passed into Gitea's sub-processes.
See `./gitea config edit-ini --help` for more details.
These environment variables can be passed to the docker container in `docker-compose.yml`.
The following example will enable an smtp mail server if the required env variables
@@ -316,8 +317,8 @@ Gitea will generate new secrets/tokens for every new installation automatically
The following commands will output a new `SECRET_KEY` and `INTERNAL_TOKEN` to `stdout`, which you can then place in your environment variables.
```bash
docker run -it --rm docker.io/gitea/gitea:1 gitea generate secret SECRET_KEY
docker run -it --rm docker.io/gitea/gitea:1 gitea generate secret INTERNAL_TOKEN
docker run -it --rm docker.gitea.com/gitea:1 gitea generate secret SECRET_KEY
docker run -it --rm docker.gitea.com/gitea:1 gitea generate secret INTERNAL_TOKEN
```
```yaml
@@ -329,316 +330,7 @@ services:
- GITEA__security__INTERNAL_TOKEN=[value returned by generate secret INTERNAL_TOKEN]
```
## SSH Container Passthrough
### SSH with multiple IP addresses
Since SSH is running inside the container, SSH needs to be passed through from the host to the container if SSH support is desired. One option would be to run the container SSH on a non-standard port (or moving the host port to a non-standard port). Another option which might be more straightforward is for Gitea users to ssh to a Gitea user on the host which will then relay those connections to the docker.
### Understanding SSH access to Gitea (without passthrough)
To understand what needs to happen, you first need to understand what happens without passthrough. So we will try to explain this:
1. The client adds their SSH public key to Gitea using the webpage.
2. Gitea will add an entry for this key to the `.ssh/authorized_keys` file of its running user, `git`.
3. This entry has the public key, but also has a `command=` option. It is this command that Gitea uses to match this key to the client user and manages authentication.
4. The client then makes an SSH request to the SSH server using the `git` user, e.g. `git clone git@domain:user/repo.git`.
5. The client will attempt to authenticate with the server, passing one or more public keys one at a time to the server.
6. For each key the client provides, the SSH server will first check its configuration for an `AuthorizedKeysCommand` to see if the public key matches, and then the `git` user's `authorized_keys` file.
7. The first entry that matches will be selected, and assuming this is a Gitea entry, the `command=` will now be executed.
8. The SSH server creates a user session for the `git` user, and using the shell for the `git` user runs the `command=`
9. This runs `gitea serv` which takes over control of the rest of the SSH session and manages gitea authentication & authorization of the git commands.
Now, for the SSH passthrough to work, we need the host SSH to match the public keys and then run the `gitea serv` on the docker. There are multiple ways of doing this. However, all of these require some information about the docker being passed to the host.
### SSHing Shim (with authorized_keys)
In this option, the idea is that the host simply uses the `authorized_keys` that gitea creates but at step 9 the `gitea` command that the host runs is a shim that actually runs ssh to go into the docker and then run the real docker `gitea` itself.
- To make the forwarding work, the SSH port of the container (22) needs to be mapped to the host port 2222 in `docker-compose.yml` . Since this port does not need to be exposed to the outside world, it can be mapped to the `localhost` of the host machine:
```yaml
ports:
# [...]
- "127.0.0.1:2222:22"
```
- Next on the host create the `git` user which shares the same `UID`/ `GID` as the container values `USER_UID`/ `USER_GID`. These values can be set as environment variables in the `docker-compose.yml`:
```yaml
environment:
- USER_UID=1000
- USER_GID=1000
```
- Mount `/home/git/.ssh` of the host into the container. This ensures that the `authorized_keys` file is shared between the host `git` user and the container `git` user otherwise the SSH authentication cannot work inside the container.
```yaml
volumes:
- /home/git/.ssh/:/data/git/.ssh
```
- Now a SSH key pair needs to be created on the host. This key pair will be used to authenticate the `git` user on the host to the container. As an administrative user on the host run: (by administrative user we mean a user that can sudo to root)
- Please note depending on the local version of ssh you may want to consider using `-t ecdsa` here.
- `/home/git/.ssh/authorized_keys` on the host now needs to be modified. It needs to act in the same way as `authorized_keys` within the Gitea container. Therefore add the public key of the key you created above ("Gitea Host Key") to `~/git/.ssh/authorized_keys`. As an administrative user on the host run:
```bash
sudo -u git cat /home/git/.ssh/id_rsa.pub | sudo -u git tee -a /home/git/.ssh/authorized_keys
Important: The pubkey from the `git` user needs to be added "as is" while all other pubkeys added via the Gitea web interface will be prefixed with `command="/usr [...]`.
`/home/git/.ssh/authorized_keys` should then look somewhat like
- The next step is to create the fake host `gitea` command that will forward commands from the host to the container. The name of this file depends on your version of Gitea:
- For Gitea v1.16.0+. As an administrative user on the host run:
Here is a detailed explanation what is happening when a SSH request is made:
1. The client adds their SSH public key to Gitea using the webpage.
2. Gitea in the container will add an entry for this key to the `.ssh/authorized_keys` file of its running user, `git`.
- However, because `/home/git/.ssh/` on the host is mounted as `/data/git/.ssh` this means that the key has been added to the host `git` user's `authorized_keys` file too.
3. This entry has the public key, but also has a `command=` option.
- This command matches the location of the Gitea binary on the container, but also the location of the shim on the host.
4. The client then makes an SSH request to the host SSH server using the `git` user, e.g. `git clone git@domain:user/repo.git`.
5. The client will attempt to authenticate with the server, passing one or more public keys in turn to the host.
6. For each key the client provides, the host SSH server will first check its configuration for an `AuthorizedKeysCommand` to see if the public key matches, and then the host `git` user's `authorized_keys` file.
- Because `/home/git/.ssh/` on the host is mounted as `/data/git/.ssh` this means that the key they added to the Gitea web will be found
7. The first entry that matches will be selected, and assuming this is a Gitea entry, the `command=` will now be executed.
8. The host SSH server creates a user session for the `git` user, and using the shell for the host `git` user runs the `command=`
9. This means that the host runs the host `/usr/local/bin/gitea` shim that opens an SSH from the host to container passing the rest of the command arguments directly to `/usr/local/bin/gitea` on the container.
10. Meaning that the container `gitea serv` is run, taking over control of the rest of the SSH session and managing gitea authentication & authorization of the git commands.
**Notes**
SSH container passthrough using `authorized_keys` will work only if
- `opensshd` is used in the container
- if `AuthorizedKeysCommand` is _not used_ in combination with `SSH_CREATE_AUTHORIZED_KEYS_FILE=false` to disable authorized files key generation
- `LOCAL_ROOT_URL` is not changed (depending on the changes)
If you try to run `gitea` on the host, you will attempt to ssh to the container and thence run the `gitea` command there.
Never add the `Gitea Host Key` as a SSH key to a user on the Gitea interface.
### SSHing Shell (with authorized_keys)
In this option, the idea is that the host simply uses the `authorized_keys` that gitea creates but at step 8 above we change the shell that the host runs to ssh directly into the docker and then run the shell there. This means that the `gitea` that is then run is the real docker `gitea`.
- In this case we setup as per SSHing Shim except instead of creating `/usr/local/bin/gitea`
we create a new shell for the git user. As an administrative user on the host run:
Be careful here - if you try to login as the git user in future you will ssh directly to the docker.
Here is a detailed explanation what is happening when a SSH request is made:
1. The client adds their SSH public key to Gitea using the webpage.
2. Gitea in the container will add an entry for this key to the `.ssh/authorized_keys` file of its running user, `git`.
- However, because `/home/git/.ssh/` on the host is mounted as `/data/git/.ssh` this means that the key has been added to the host `git` user's `authorized_keys` file too.
3. This entry has the public key, but also has a `command=` option.
- This command matches the location of the Gitea binary on the container.
4. The client then makes an SSH request to the host SSH server using the `git` user, e.g. `git clone git@domain:user/repo.git`.
5. The client will attempt to authenticate with the server, passing one or more public keys in turn to the host.
6. For each key the client provides, the host SSH server will first check its configuration for an `AuthorizedKeysCommand` to see if the public key matches, and then the host `git` user's `authorized_keys` file.
- Because `/home/git/.ssh/` on the host is mounted as `/data/git/.ssh` this means that the key they added to the Gitea web will be found
7. The first entry that matches will be selected, and assuming this is a Gitea entry, the `command=` will now be executed.
8. The host SSH server creates a user session for the `git` user, and using the shell for the host `git` user runs the `command=`
9. The shell of the host `git` user is now our `ssh-shell` which opens an SSH connection from the host to container, (which opens a shell on the container for the container `git`).
10. The container shell now runs the `command=` option meaning that the container `gitea serv` is run, taking over control of the rest of the SSH session and managing gitea authentication & authorization of the git commands.
**Notes**
SSH container passthrough using `authorized_keys` will work only if
- `opensshd` is used in the container
- if `AuthorizedKeysCommand` is _not used_ in combination with `SSH_CREATE_AUTHORIZED_KEYS_FILE=false` to disable authorized files key generation
- `LOCAL_ROOT_URL` is not changed (depending on the changes)
If you try to login as the `git` user on the host in future you will ssh directly to the docker.
Never add the `Gitea Host Key` as a SSH key to a user on the Gitea interface.
### Docker Shell (with authorized_keys)
Similar to the above ssh shell technique we can use a shell which simply uses `docker exec`. As an administrative user on the host run:
```bash
cat <<"EOF" | sudo tee /home/git/docker-shell
#!/bin/sh
/usr/bin/docker exec -i -u git --env SSH_ORIGINAL_COMMAND="$SSH_ORIGINAL_COMMAND" gitea sh "$@"
EOF
sudo chmod +x /home/git/docker-shell
sudo usermod -s /home/git/docker-shell git
```
Here is a detailed explanation what is happening when a SSH request is made:
1. The client adds their SSH public key to Gitea using the webpage.
2. Gitea in the container will add an entry for this key to the `.ssh/authorized_keys` file of its running user, `git`.
- However, because `/home/git/.ssh/` on the host is mounted as `/data/git/.ssh` this means that the key has been added to the host `git` user's `authorized_keys` file too.
3. This entry has the public key, but also has a `command=` option.
- This command matches the location of the Gitea binary on the container.
4. The client then makes an SSH request to the host SSH server using the `git` user, e.g. `git clone git@domain:user/repo.git`.
5. The client will attempt to authenticate with the server, passing one or more public keys in turn to the host.
6. For each key the client provides, the host SSH server will first check its configuration for an `AuthorizedKeysCommand` to see if the public key matches, and then the host `git` user's `authorized_keys` file.
- Because `/home/git/.ssh/` on the host is mounted as `/data/git/.ssh` this means that the key they added to the Gitea web will be found
7. The first entry that matches will be selected, and assuming this is a Gitea entry, the `command=` will now be executed.
8. The host SSH server creates a user session for the `git` user, and using the shell for the host `git` user runs the `command=`
9. The shell of the host `git` user is now our `docker-shell` which uses `docker exec` to open a shell for the `git` user on the container.
10. The container shell now runs the `command=` option meaning that the container `gitea serv` is run, taking over control of the rest of the SSH session and managing gitea authentication & authorization of the git commands.
Note that `gitea` in the docker command above is the name of the container. If you named yours differently, don't forget to change that. The host `git` user also has to have
permission to run `docker exec`.
**Notes**
Docker shell passthrough using `authorized_keys` will work only if
- `opensshd` is used in the container
- if `AuthorizedKeysCommand` is _not used_ in combination with `SSH_CREATE_AUTHORIZED_KEYS_FILE=false` to disable authorized files key generation
- `LOCAL_ROOT_URL` is not changed (depending on the changes)
If you try to login as the `git` user on the host in future you will `docker exec` directly to the docker.
A Docker execing shim could be created similarly to above.
### Docker Shell with AuthorizedKeysCommand
The AuthorizedKeysCommand route provides another option that does not require many changes to the compose file or the `authorized_keys` - but does require changes to the host `/etc/sshd_config`.
In this option, the idea is that the host SSH uses an `AuthorizedKeysCommand` instead of relying on sharing the `authorized_keys` file that gitea creates. We continue to use a special shell at step 8 above to exec into the docker and then run the shell there. This means that the `gitea` that is then run is the real docker `gitea`.
- On the host create a `git` user with permission to run `docker exec`.
- We will again assume that the Gitea container is called `gitea`.
- Modify the `git` user's shell to forward commands to the `sh` executable inside the container using `docker exec`. As an administrative user on the host run:
```bash
cat <<"EOF" | sudo tee /home/git/docker-shell
#!/bin/sh
/usr/bin/docker exec -i -u git --env SSH_ORIGINAL_COMMAND="$SSH_ORIGINAL_COMMAND" gitea sh "$@"
EOF
sudo chmod +x /home/git/docker-shell
sudo usermod -s /home/git/docker-shell git
```
Now all attempts to login as the `git` user on the host will be forwarded to the docker - including the `SSH_ORIGINAL_COMMAND`. We now need to set-up SSH authentication on the host.
We will do this by leveraging the [SSH AuthorizedKeysCommand](../administration/command-line.md#keys) to match the keys against those accepted by Gitea.
Add the following block to `/etc/ssh/sshd_config`, on the host:
(From 1.16.0 you will not need to set the `-c /data/gitea/conf/app.ini` option.)
Finally restart the SSH server. As an administrative user on the host run:
```bash
sudo systemctl restart sshd
```
Here is a detailed explanation what is happening when a SSH request is made:
1. The client adds their SSH public key to Gitea using the webpage.
2. Gitea in the container will add an entry for this key to its database.
3. The client then makes an SSH request to the host SSH server using the `git` user, e.g. `git clone git@domain:user/repo.git`.
4. The client will attempt to authenticate with the server, passing one or more public keys in turn to the host.
5. For each key the client provides, the host SSH server will checks its configuration for an `AuthorizedKeysCommand`.
6. The host runs the above `AuthorizedKeysCommand`, which execs in to the docker and then runs the `gitea keys` command.
7. Gitea on the docker will look in it's database to see if the public key matches and will return an entry like that of an `authorized_keys` command.
8. This entry has the public key, but also has a `command=` option which matches the location of the Gitea binary on the container.
9. The host SSH server creates a user session for the `git` user, and using the shell for the host `git` user runs the `command=`.
10. The shell of the host `git` user is now our `docker-shell` which uses `docker exec` to open a shell for the `git` user on the container.
11. The container shell now runs the `command=` option meaning that the container `gitea serv` is run, taking over control of the rest of the SSH session and managing gitea authentication & authorization of the git commands.
**Notes**
Docker shell passthrough using `AuthorizedKeysCommand` will work only if
- The host `git` user is allowed to run the `docker exec` command.
If you try to login as the `git` user on the host in future you will `docker exec` directly to the docker.
A Docker execing shim could be created similarly to above.
### SSH Shell with AuthorizedKeysCommand
Create a key for the host `git` user as above, add it to the docker `/data/git/.ssh/authorized_keys` then finally create and set the `ssh-shell` as above.
Add the following block to `/etc/ssh/sshd_config`, on the host:
(From 1.16.0 you will not need to set the `-c /data/gitea/conf/app.ini` option.)
Finally restart the SSH server. As an administrative user on the host run:
```bash
sudo systemctl restart sshd
```
Here is a detailed explanation what is happening when a SSH request is made:
1. The client adds their SSH public key to Gitea using the webpage.
2. Gitea in the container will add an entry for this key to its database.
3. The client then makes an SSH request to the host SSH server using the `git` user, e.g. `git clone git@domain:user/repo.git`.
4. The client will attempt to authenticate with the server, passing one or more public keys in turn to the host.
5. For each key the client provides, the host SSH server will checks its configuration for an `AuthorizedKeysCommand`.
6. The host runs the above `AuthorizedKeysCommand`, which will SSH in to the docker and then run the `gitea keys` command.
7. Gitea on the docker will look in it's database to see if the public key matches and will return an entry like that of an `authorized_keys` command.
8. This entry has the public key, but also has a `command=` option which matches the location of the Gitea binary on the container.
9. The host SSH server creates a user session for the `git` user, and using the shell for the host `git` user runs the `command=`.
10. The shell of the host `git` user is now our `git-shell` which uses SSH to open a shell for the `git` user on the container.
11. The container shell now runs the `command=` option meaning that the container `gitea serv` is run, taking over control of the rest of the SSH session and managing gitea authentication & authorization of the git commands.
**Notes**
SSH container passthrough using `AuthorizedKeysCommand` will work only if
- `opensshd` is running on the container
If you try to login as the `git` user on the host in future you will `ssh` directly to the docker.
Never add the `Gitea Host Key` as a SSH key to a user on the Gitea interface.
SSHing shims could be created similarly to above.
This assumes that the host machine has more than one reachable IP address: `192.168.1.1` (host) `192.168.1.2` (gitea)
On the host machine, configure SSHD in `/etc/ssh/sshd_config` to listen on one IP address `ListenAddress 192.168.1.1`. In the compose file the SSH port forwarding then needs to be changed to `"192.168.1.2:22:22"`. The port forwarding needs to be adjusted similarily for all other forwarded ports to avoid problems with DNS.
And there are some differences for permissions between individual repositories and organization repositories.
## Individual Repository
For individual repositories, the creators are the only owners of repositories and have no limit to change anything of this
repository or delete it. Repositories owners could add collaborators to help maintain the repositories. Collaborators could have `Read`, `Write` and `Admin` permissions.
For a private repository, the experience is similar to visiting an anonymous public repository. You have access to all the available content within the repository, including the ability to clone the code, create issues, respond to issue comments, submit pull requests, and more. If you have 'Write' permission, you can push code to specific branches of the repository, provided it's permitted by the branch protection rules. Additionally, you can make changes to the wiki pages. With 'Admin' permission, you have the ability to modify the repository's settings.
But you cannot delete or transfer this repository if you are not that repository's owner.
## Organization Repository
For individual repositories, the owner is the user who created it. For organization repositories, the owners are the members of the owner team on this organization. All the permissions depends on the team permission settings.
### Owner Team
The owner team will be created when the organization is created, and the creator will become the first member of the owner team. The owner team cannot be deleted and there is at least one member.
### Admin Team
When creating teams, there are two types of teams. One is the admin team, another is the general team. An admin team can be created to manage some of the repositories, whose members can do anything with these repositories. Only members of the owner or admin team can create a new team.
### General Team
A general team in an organization has unit permissions settings. It can have members and repositories scope.
- A team could access all the repositories in this organization or special repositories.
- A team could also be allowed to create new repositories or not.
The General team can be created to do the operations allowed by their permissions. One member could join multiple teams.
Protected branches prevent unwanted changes by enforcing push and merge policies on selected branches. The rules are enforced for every Git protocol (HTTP(S), SSH), the web editor, the API, and background jobs such as auto-merge. Only repository owners and administrators can manage the rules, and the Branches page is read-only while a repository is archived.
## Creating or editing a rule
1. Open the repository and select **Settings → Branches** (repository admin permission required).
2. Select **Add new rule** or **Edit** next to an existing rule.
3. Fill in the **Protected branch name pattern** and optional file patterns, then configure the push, merge, and review options described below.
4. Select **Save rule**.
The rule immediately applies to all matching branches, even if the branches are created in the future.
## Rule matching and priorities
- The **Protected branch name pattern** accepts [glob](https://github.com/gobwas/glob) expressions and matches the entire branch name. Patterns are case-sensitive. Using a simple name such as `main` without glob special characters always matches that specific branch (case-insensitive).
- If multiple rules match the same branch, only the first rule is used. Reorder the list on the **Branches** page by dragging the grab handle. The first entry (priority 1) has the highest priority, so place patterns such as `main` or `release/*` before generic fallbacks such as `*`.
| `hotfix/**` | Nested branches such as `hotfix/security/CVE` |
| `*` | Every branch (use as a fallback) |
### File-level pattern controls
- **Protected file patterns** block changes to sensitive files (for example `.drone.yml` or `/docs/**/*.txt`). Patterns are case-insensitive and separated by semicolons. Commits and merge attempts that touch one of the files are rejected.
- **Unprotected file patterns** do the opposite: if pushes are blocked, users with write access can still push commits that modify only the listed files. This is useful for letting contributors update documentation while still requiring pull requests for code.
Both fields use the same `glob` syntax and match paths relative to the repository root.
## Controlling direct pushes
The **Push** section controls direct pushes (including the web editor and API).
- **Disable push** makes the branch read-only. Any attempt to push directly fails, and changes must be merged through pull requests.
- **Enable push** allows anyone with [write access](./permissions.md) to push (force pushes are still blocked unless explicitly allowed).
- **Allowlist restricted push** requires being on the allowlist. Choose users and, for organization-owned repositories, teams. Deploy keys that already have write access can also be allowlisted.
When a push is blocked, the server-side hook rejects the update with an explanation.
### Force pushes
Force pushes have their own set of options:
- **Disable force push** completely forbids rewriting history on the branch.
- **Enable force push** allows anyone who can push to also force push.
- **Allowlist restricted force push** limits force pushes to a separate allowlist (users, teams, and optionally deploy keys) **and** requires the person to already have regular push access.
## Pull request merges and approvals
- **Merge allowlist**: keep the default to let anyone with write access merge pull requests, or enable the allowlist to restrict merges to selected users/teams.
- **Required approvals**: specify how many approvals are needed before a merge is allowed. Reviews from users with write access count, unless the **Restrict approvals to allowlisted users or teams** option is enabled.
- **Dismiss stale approvals** removes existing approvals whenever new commits that change the pull request content are pushed.
- **Ignore stale approvals** keeps approvals but does not count reviews that were made on older commits. This option is disabled while dismissal of stale approvals is enabled.
- **Block merge on rejected reviews** prevents merging while any official reviewer has requested changes.
- **Block merge on official review requests** blocks merges while there are outstanding review requests (for example when CODEOWNERS requires a review).
- **Block merge if the pull request is outdated** makes sure the head branch is up to date with the base branch before it can be merged.
- **Administrators must follow branch protection rules** removes the ability for repository administrators to bypass the rules with the "Force merge" button.
Protected file patterns apply to pull requests as well. When a pull request changes one of the protected files, the pull request banner shows the affected paths and merging stays disabled.
## Status checks
Enable status checks to require one or more CI jobs to succeed before merging:
1. Check **Enable status check**.
2. Enter one pattern per line in **Status check patterns**. Each pattern is a `glob` expression that matches the context name reported by Actions, Drone, Woodpecker, or another Check API client (for example `actions/test-*`).
3. Pick contexts from the table of jobs that have reported results in the last week to verify their names.
When the option is active, Gitea requires at least one context that matches each pattern to report success on the pull request head commit. An empty list is not allowed; use `*` to require the latest commit to be successful regardless of the context name.
## Signed commits and other safeguards
- **Require signed commits** rejects pushes that contain unsigned or unverifiable commits. The check runs inside the server hook before the push is accepted.
- **Protected file patterns** (see above) prevent both pushes and merges that modify sensitive files.
- **Unprotected file patterns** allow limited pushes while keeping the branch protected.
Protected tags allow control over who has permission to create or update Git tags. Each rule allows you to match either an individual tag name, or use an appropriate pattern to control multiple tags at once.
## Setting up protected tags
To protect a tag, you need to follow these steps:
1. Go to the repository’s **Settings** > **Tags** page.
1. Type a pattern to match a name. You can use a single name, a [glob pattern](https://pkg.go.dev/github.com/gobwas/glob#Compile) or a regular expression.
1. Choose the allowed users and/or teams. If you leave these fields empty no one is allowed to create or modify this tag.
1. Select **Save** to save the configuration.
## Pattern protected tags
The pattern uses [glob](https://pkg.go.dev/github.com/gobwas/glob#Compile) or regular expressions to match a tag name. For regular expressions you need to enclose the pattern in slashes.
Examples:
| Type | Pattern Protected Tag | Possible Matching Tags |
This page will introduce the [act runner](https://gitea.com/gitea/act_runner) in detail, which is the runner of Gitea Actions.
## Requirements
Currently the runner supports run in two modes. One is running in docker container, another is running in host. It is recommended to run jobs in a [docker](https://docker.com) container, if you chose this mode, you need to [install docker](https://docs.docker.com/engine/install/) first and make sure that the docker daemon is running.
Other OCI container engines which are compatible with Docker's API should also work, but are untested.
However, if you are sure that you want to run jobs directly on the host only, then docker is not required.
There are multiple ways to install the act runner.
## Installation with binary
### Download the binary
You can download the binary from the [release page](https://gitea.com/gitea/act_runner/releases).
However, if you want to use the latest nightly build, you can download it from the [download page](https://dl.gitea.com/act_runner/).
When you download the binary, please make sure that you have downloaded the correct one for your platform.
You can check it by running the following command if you are in a Unix-Style OS.
```bash
chmod +x act_runner
./act_runner --version
```
If you see the version information, it means that you have downloaded the correct binary.
### Obtain a registration token
You can register a runner in different levels, it can be:
- Instance level: The runner will run jobs for all repositories in the instance.
- Organization level: The runner will run jobs for all repositories in the organization.
- Repository level: The runner will run jobs for the repository it belongs to.
Note that the repository may still use instance-level or organization-level runners even if it has its own repository-level runners. A future release may provide an option to allow more control over this.
Before register the runner and run it, you need a registration token. The level of the runner determines where to obtain the registration token.
- Instance level: The admin settings page, like `<your_gitea.com>/admin/actions/runners`.
- Organization level: The organization settings page, like `<your_gitea.com>/<org>/settings/actions/runners`.
- Repository level: The repository settings page, like `<your_gitea.com>/<owner>/<repo>/settings/actions/runners`.
If you cannot see the settings page, please make sure that you have the right permissions and that Actions have been enabled.
The format of the registration token is a random string `D0gvfu2iHfUjNqCYVljVyRV14fISpJxxxxxxxxxx`.
A registration token can also be obtained from the gitea [command-line interface](../../administration/command-line.md#actions-generate-runner-token):
You can also use `GITEA_RUNNER_REGISTRATION_TOKEN`/`GITEA_RUNNER_REGISTRATION_TOKEN_FILE` environment variable to set a global runner registration token when Gitea starts, for example:
The token from environment is valid until you reset the token (re-create a new one) via web UI or API.
Tokens are valid for registering multiple runners, until they are revoked and replaced by a new token using the token reset link in the web interface.
### Configuration
Configuration is done via a configuration file. It is optional, and the default configuration will be used when no configuration file is specified. You can generate a configuration file by running the following command:
```bash
./act_runner generate-config
```
The default configuration is safe to use without any modification, so you can just use it directly.
```bash
./act_runner generate-config > config.yaml
./act_runner --config config.yaml [command]
```
### Register the runner
Registration is required before running the act runner, because the runner needs to know where to get jobs from. And it is also important to Gitea instance to identify the runner.
If this has been installed using the binary package, the act runner can be registered by running the following command.
```bash
./act_runner register
```
Alternatively, you can use the `--config` option to specify the configuration file mentioned in the previous section.
```bash
./act_runner --config config.yaml register
```
You will be asked to input the registration information step by step. Includes:
- The Gitea instance URL, like `https://gitea.com/` or `http://192.168.8.8:3000/`.
- The registration token.
- The runner name, which is optional. If you leave it blank, the hostname will be used.
- The runner labels, which is optional. If you leave it blank, the default labels will be used.
You may be confused about the runner labels, which will be explained later.
If you want to register the runner in a non-interactive way, you can use arguments to do it.
When you have registered the runner, you can find a new file named `.runner` in the current directory.
This file stores the registration information.
Please do not edit it manually.
If this file is missing or corrupted, you can simply remove it and register again.
If you want to store the registration information in another place, you can specify it in the configuration file,
and don't forget to specify the `--config` option.
### Start the runner in command line
After you have registered the runner, you can run it by running the following command:
```shell
./act_runner daemon
```
or
```bash
./act_runner daemon --config config.yaml
```
The runner will fetch jobs from the Gitea instance and run them automatically.
### Start the runner with Systemd
It is also possible to run act-runner as a [systemd](https://en.wikipedia.org/wiki/Systemd) service. Create an unprivileged `act_runner` user on your system, and the following file in `/etc/systemd/system/act_runner.service`. The paths in `ExecStart` and `WorkingDirectory` may need to be adjusted depending on where you installed the `act_runner` binary, its configuration file, and the home directory of the `act_runner` user.
If using Docker, the `act_runner` user should also be added to the `docker` group before starting the service. Keep in mind that this effectively gives `act_runner` root access to the system [[1]](https://docs.docker.com/engine/security/#docker-daemon-attack-surface).
### Start the runner with LaunchDaemon(macOS)
Mac uses `launchd` in place of systemd for registering daemon processes. By default daemons run as the root user, so if desired an unprivileged `_act_runner` user can be created via the `dscl` tool. The following file should then be created at the directory `/Library/LaunchDaemon/com.gitea.act_runner.plist`. The paths for `WorkingDirectory`, `ProgramArguments`, `StandardOutPath`, `StandardErrPath`, and the `HOME` environment variable may need to be updated to reflect your installation. Also note that any executables outside of the example `PATH` shown will need to be explicitly included and will not be inherited from existing configurations.
```xml
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
You can also set up a Linux service or Windows service to let the runner run automatically.
## Install with the docker image
### Pull the image
You can use the docker image from the [docker hub](https://hub.docker.com/r/gitea/act_runner/tags).
Just like the binary, you can use the latest nightly build by using the `nightly` tag, while the `latest` tag is the latest stable release.
```bash
docker pull docker.io/gitea/act_runner:latest # for the latest stable release
```
If you want to test newly features, you could also use nightly image
```bash
docker pull docker.io/gitea/act_runner:nightly # for the latest nightly build
```
### Configuration
Configuration is optional, but you could also generate config file with docker:
```bash
docker run --entrypoint="" --rm -it docker.io/gitea/act_runner:latest act_runner generate-config > config.yaml
```
When you are using the docker image, you can specify the configuration file by using the `CONFIG_FILE` environment variable. Make sure that the file is mounted into the container as a volume:
```bash
docker run -v $PWD/config.yaml:/config.yaml -e CONFIG_FILE=/config.yaml ...
```
You may notice the commands above are both incomplete, because it is not the time to run the act runner yet.
Before running the act runner, we need to register it to your Gitea instance first.
### Start the runner with docker
If you are using the docker image, behaviour will be slightly different. Registration and running are combined into one step in this case, so you need to specify the registration information when running the act runner.
A quick start with docker run like below. You need to get `<registration_token>` from the above step, and give
When using docker, there is no requirement to enter the container and manually run `./act_runner daemon` command as shown below. Once the container has been started successfully, it will show up as an active runner in your Gitea instance.
## Advantage Configurations
### Configuring cache when starting a Runner using docker image
If you do not intend to use `actions/cache` in workflow, you can ignore this section.
If you use `actions/cache` without any additional configuration, it will return the following error:
> Failed to restore: getCacheEntry failed: connect ETIMEDOUT IP:PORT
The error occurs because the runner container and job container are on different networks, so the job container cannot access the runner container.
Therefore, it is essential to configure the cache action to ensure its proper functioning. Follow these steps:
- 1.Obtain the LAN IP address of the host machine where the runner container is running.
- 2.Find an available port number on the host machine where the runner container is running.
- 3.Configure the following settings in the configuration file:
```yaml
cache:
enabled: true
dir: ""
# Use the LAN IP obtained in step 1
host: "192.168.8.17"
# Use the port number obtained in step 2
port: 8088
```
- 4.When starting the container, map the cache port to the host machine:
```bash
docker run \
--name gitea-docker-runner \
-p 8088:8088 \
-d docker.io/gitea/act_runner:nightly
```
### Labels
The labels of a runner are used to determine which jobs the runner can run, and how to run them.
The default labels are `ubuntu-latest:docker://node:16-bullseye,ubuntu-22.04:docker://node:16-bullseye,ubuntu-20.04:docker://node:16-bullseye,ubuntu-18.04:docker://node:16-buster`.
It is a comma-separated list, and each item is a label.
Let's take `ubuntu-22.04:docker://node:16-bullseye` as an example.
It means that the runner can run jobs with `runs-on: ubuntu-22.04`, and the job will be run in a docker container with the image `node:16-bullseye`.
If the default image is insufficient for your needs, and you have enough disk space to use a better and bigger one, you can change it to `ubuntu-22.04:docker://<the image you like>`.
You can find more useful images on [act images](https://github.com/nektos/act/blob/master/IMAGES.md).
If you want to run jobs in the host directly, you can change it to `ubuntu-22.04:host` or just `ubuntu-22.04`, the `:host` is optional.
However, we suggest you to use a special name like `linux_amd64:host` or `windows:host` to avoid misusing it.
Starting with Gitea 1.21, you can change labels by modifying `runners.labels` in the runner configuration file (if you don't have a configuration file, please refer to [configuration tutorials](#configuration)).
The runner will use these new labels as soon as you restart it, i.e., by calling `./act_runner daemon --config config.yaml`.
Workflow runs can store files such as test results, coverage reports, and build output as artifacts. Artifacts are uploaded by workflows, for example with [`actions/upload-artifact`](https://github.com/actions/upload-artifact), and appear on the run page after the upload completes.
## Previewing artifacts
Select an artifact on the run page to open its file browser and preview in the same tab. The preview identifies the originating run and attempt, and provides links back to that run and to download the complete artifact. Selecting files in the browser updates the preview in the same tab.
Artifact previews:
- require sign-in and access to the Actions run to open the preview page
- render text-based files, images (including SVG), and PDFs
- run HTML and JavaScript content in a sandboxed frame
Preview content is loaded through signed URLs that expire after one hour and do not require a session cookie. Reopen the preview page to obtain a fresh URL after it expires.
The file browser lists up to 2,000 files per artifact. Larger artifacts show a notice when the list is truncated.
Unsupported or oversized files can still be downloaded with the complete artifact.
## Instance configuration
Site administrators can limit the total artifact size that can be browsed or previewed with `ARTIFACT_PREVIEW_MAX_SIZE` in the `[actions]` section of `app.ini`. The default is 10 MiB. Set it to `0` to disable previews or `-1` to allow artifacts of any size.
The size of an individual previewed file is also limited by [`MAX_DISPLAY_FILE_SIZE`](../../administration/config-cheat-sheet.md#ui-ui) in the `[ui]` section.
Even though Gitea Actions is designed to be compatible with GitHub Actions, there are some differences between them.
Gitea Actions is designed to be compatible with GitHub Actions, with the differences below.
## Additional features
### Absolute action URLs
Gitea Actions supports defining actions via absolute URL, which means that you can use actions from any git repository.
Like `uses: https://github.com/actions/checkout@v4` or `uses: http://your_gitea.com/owner/repo@branch`.
`uses:` accepts an absolute URL to any git host, e.g. `uses: https://gitea.com/actions/checkout@v4`.
### Action prefixes
`uses:` accepts these prefixes:
- `self:` references your own Gitea instance, e.g. `uses: self:owner/repo@v1` or `uses: self:owner/repo/.gitea/workflows/build.yml@v1`
- `builtin:` runs an action built into Gitea Runner, which needs no download and no Node in the job image, e.g. `uses: builtin:checkout`
`builtin:checkout` accepts the `repository`, `ref`, `token`, `path` and `fetch-depth` inputs of `actions/checkout` with the same defaults, and fails on any other input.
### Workflow-relative actions
`$/` references the repository and commit of the workflow, or of the enclosing composite action, e.g. `uses: $/.gitea/actions/build`.
Unlike `./`, it needs no checkout first.
### Actions written in Go
Gitea Actions supports writing actions in Go.
See [Creating Go Actions](https://blog.gitea.com/creating-go-actions/).
### Support the non-standard syntax @yearly, @monthly, @weekly, @daily, @hourly on schedule
Github Actions doesn't support that. https://docs.github.com/en/actions/using-workflows/events-that-trigger-workflows#schedule
`schedule` also accepts `@yearly`, `@monthly`, `@weekly`, `@daily` and `@hourly`, which [GitHub does not](https://docs.github.com/en/actions/using-workflows/events-that-trigger-workflows#schedule).
## Unsupported workflows syntax
### `concurrency`
It's used to run a single job at a time.
See [Using concurrency](https://docs.github.com/en/actions/using-jobs/using-concurrency).
It's ignored by Gitea Actions now.
### `run-name`
The name for workflow runs generated from the workflow.
See [Workflow syntax for GitHub Actions](https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#run-name).
It's ignored by Gitea Actions now.
### `permissions` and `jobs.<job_id>.permissions`
See [Workflow syntax for GitHub Actions](https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#permissions).
It's ignored by Gitea Actions now.
### `jobs.<job_id>.timeout-minutes`
See [Workflow syntax for GitHub Actions](https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#jobsjob_idtimeout-minutes).
It's ignored by Gitea Actions now.
### `jobs.<job_id>.continue-on-error`
See [Workflow syntax for GitHub Actions](https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#jobsjob_idcontinue-on-error).
It's ignored by Gitea Actions now.
### `jobs.<job_id>.environment`
See [Workflow syntax for GitHub Actions](https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#jobsjob_idenvironment).
It's ignored by Gitea Actions now.
[`jobs.<job_id>.environment`](https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#jobsjob_idenvironment) is ignored.
### Complex `runs-on`
See [Workflow syntax for GitHub Actions](https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#jobsjob_idruns-on).
The `GITEA_TOKEN`for a job running within a repository should be able to publish to the associated package repository (i.e. to upload OCI images). See the "packages" scope for the "default access" in [Automatic token authentication](https://docs.github.com/en/actions/security-for-github-actions/security-guides/automatic-token-authentication#permissions-for-the-github_token).
This is not implemented in Gitea Actions now. A workaround for Gitea Actions is to use a Personal Access Token (PAT). See this [github issue and comment](https://github.com/go-gitea/gitea/issues/23642#issuecomment-2119876692) for tracking this feature.
`GITEA_TOKEN`cannot publish to the package registry of its repository, e.g. to push OCI images.
Use a personal access token instead, see [this issue](https://github.com/go-gitea/gitea/issues/23642#issuecomment-2119876692).
### Problem Matchers
Problem Matchers are a way to scan the output of actions for a specified regex pattern and surface that information prominently in the UI.
See [Problem matchers](https://github.com/actions/toolkit/blob/main/docs/problem-matchers.md).
It's ignored by Gitea Actions now.
[Problem matchers](https://github.com/actions/toolkit/blob/main/docs/problem-matchers.md) are ignored.
### Create an error annotation
See [Creating an annotation for an error](https://docs.github.com/en/actions/using-workflows/workflow-commands-for-github-actions#example-creating-an-annotation-for-an-error)
It's ignored by Gitea Actions now.
[Error annotations](https://docs.github.com/en/actions/using-workflows/workflow-commands-for-github-actions#example-creating-an-annotation-for-an-error) are ignored.
### Expressions
For [expressions](https://docs.github.com/en/actions/learn-github-actions/expressions), only [`always()`](https://docs.github.com/en/actions/learn-github-actions/expressions#always) is supported.
[Expressions](https://docs.github.com/en/actions/learn-github-actions/expressions) support the standard GitHub functions and contexts.
## Missing UI features
### Pre and Post steps
Pre and Post steps don't have their own section in the job log user interface.
Pre and post steps have no section of their own in the job log.
### Services steps
Services steps don't have their own section in the job log user interface.
Service steps have no section of their own in the job log.
## Different behavior
### Job token permissions (`permissions`)
`permissions` and `jobs.<job_id>.permissions` control the `GITEA_TOKEN` permissions, clamped by the repository and owner settings and further restricted for fork pull requests and cross-repository access.
GitHub-only scopes like `statuses`, `checks`, `deployments`, `security-events` and `pages` are not supported, while Gitea adds `code`, `releases`, `wiki` and `projects`. `id-token` works like on GitHub, see [OpenID Connect](openid-connect.md).
See [Actions job token permissions](token-permissions.md).
### Downloading actions
Previously (Pre 1.21.0),`[actions].DEFAULT_ACTIONS_URL`defaulted to `https://gitea.com`.
We have since restricted this option to only allow two values (`github` and `self`).
When set to `github`, the new default, Gitea will download non-fully-qualified actions from `https://github.com`.
For example, if you use `uses: actions/checkout@v4`, it will download the checkout repository from `https://github.com/actions/checkout.git`.
If you want to download an action from another git hoster, you can use an absolute URL, e.g. `uses: https://gitea.com/actions/checkout@v4`.
If your Gitea instance is in an intranet or a restricted area, you can set the URL to `self` to only download actions from your own instance by default.
Of course, you can still use absolute URLs in workflows.
More details about the `[actions].DEFAULT_ACTIONS_URL` configuration can be found in the [Configuration Cheat Sheet](../../administration/config-cheat-sheet.md#actions-actions)。
Actions without a host, like `uses: actions/checkout@v4`, are downloaded from `https://github.com`, or from your own instance when`[actions].DEFAULT_ACTIONS_URL`is `self`.
See the [Configuration Cheat Sheet](../../administration/config-cheat-sheet.md#actions-actions).
For other sources, see [Absolute action URLs](#absolute-action-urls), [Action prefixes](#action-prefixes) and [Workflow-relative actions](#workflow-relative-actions).
### Context availability
Context availability is not checked, so you can use the env context on more places.
See [Context availability](https://docs.github.com/en/actions/learn-github-actions/contexts#context-availability).
[Context availability](https://docs.github.com/en/actions/learn-github-actions/contexts#context-availability) is not checked, so contexts like `env` work in more places.
Gitea Actions has multiple components. This document describes them individually.
## Act
## Gitea Runner
The [nektos/act](https://github.com/nektos/act) project is an excellent tool that allows you to run your GitHub Actions locally.
We were inspired by this and wondered if it would be possible to run actions for Gitea.
However, while [nektos/act](https://github.com/nektos/act) is designed as a command line tool, what we actually needed was a Go library with modifications specifically for Gitea.
So we forked it as [gitea/act](https://gitea.com/gitea/act).
This is a soft fork that will periodically follow the upstream.
Although some custom commits have been added, we will try our best to avoid changing too much of the original code.
The forked act is just a shim or adapter for Gitea's specific usage.
There are some additional commits that have been made, such as:
- Outputting execution logs to logger hook so they can be reported to Gitea
- Disabling the GraphQL URL, since Gitea doesn't support it
- Starting a new container for every job instead of reusing to ensure isolation.
These modifications have no reason to be merged into the upstream.
They don't make sense if the user just wants to run trusted actions locally.
However, there may be overlaps in the future, such as a required bug fix or new feature needed by both projects.
In these cases, we will contribute the changes back to the upstream repository.
## Act runner
Gitea's runner is called act runner because it's based on act.
Gitea Runner is based on a hard fork of [nektos/act](https://github.com/nektos/act).
Like other CI runners, we designed it as an external part of Gitea, which means it should run on a different server than Gitea.
@@ -43,11 +19,11 @@ Additionally, the runner will introduce itself to Gitea and declare what kind of
Earlier, we mentioned that `runs-on: ubuntu-latest` in a workflow file means that the job will be run on a runner with the `ubuntu-latest` label.
But how does the runner know to run `ubuntu-latest`? The answer lies in mapping the label to an environment.
That's why when you add custom labels during registration, you will need to input some complex content like `my_custom_label:docker://centos:7`.
That's why each label of a runner names an environment, like `my_custom_label:docker://centos:7`.
This means that the runner can take the job which needs to run on `my_custom_label`, and it will run it via a docker container with the image `centos:7`.
Docker isn't the only option, though.
The act also supports running jobs directly on the host.
The runner also supports running jobs directly on the host.
This is achieved through labels like `linux_arm:host`.
This label indicates that the runner can take a job that needs to run on `linux_arm` and run it directly on the host.
@@ -58,16 +34,15 @@ So,
- `my_custom_label:docker://node:18`: Run jobs labeled with `my_custom_label` using the `node:18` Docker image.
- `my_custom_label:host`: Run jobs labeled with `my_custom_label` directly on the host.
- `my_custom_label`: Same as `my_custom_label:host`.
- `my_custom_label:vm:ubuntu-latest`: (Example only, not implemented) Run jobs labeled with `my_custom_label` using a virtual machine with the `ubuntu-latest` ISO.
## Communication protocol
As act runner is an independent part of Gitea, we needed a protocol for runners to communicate with the Gitea instance.
As the runner is an independent part of Gitea, we needed a protocol for runners to communicate with the Gitea instance.
However, we did not think it was a good idea to have Gitea listen on a new port.
Instead, we wanted to reuse the HTTP port, which means we needed a protocol that is compatible with HTTP.
We chose to use gRPC over HTTP.
We use [actions-proto-def](https://gitea.com/gitea/actions-proto-def) and [actions-proto-go](https://gitea.com/gitea/actions-proto-go) to wire them up.
The protocol is defined in [actions-proto-def](https://gitea.com/gitea/actions-proto-def), its Go code lives in [actionslib](https://gitea.com/gitea/actionslib).
More information about gRPC can be found on [its website](https://grpc.io/).
## Network architecture
@@ -79,9 +54,9 @@ This will help you troubleshoot some problems and explain why it's a bad idea to
There are four network connections marked in the picture, and the direction of the arrows indicates the direction of establishing the connections.
### Connection 1, act runner to Gitea instance
### Connection 1, runner to Gitea instance
The act runner must be able to connect to Gitea to receive tasks and send back the execution results.
The runner must be able to connect to Gitea to receive tasks and send back the execution results.
### Connection 2, job containers to Gitea instance
@@ -92,10 +67,11 @@ Fetching code is not always necessary to run some jobs, but it is required in mo
If you use a loopback address to register a runner, the runner can connect to Gitea when it is on the same machine.
However, if a job container tries to fetch code from localhost, it will fail because Gitea is not in the same container.
### Connection 3, act runner to internet
### Connection 3, runner to internet
When you use some actions like `actions/checkout@v4`, the act runner downloads the scripts, not the job containers.
By default, it downloads from [github.com](http://github.com/), so it requires access to the internet. If you configure the `DEFAULT_ACTIONS_URL` to `self`, then it will download from your Gitea instance by default. Then it will not connect to internet when downloading the action itself.
When you use some actions like `actions/checkout@v4`, the runner downloads the scripts, not the job containers.
By default, it downloads them from [github.com](https://github.com/), which requires internet access.
With `DEFAULT_ACTIONS_URL` set to `self`, or with `uses: self:owner/repo@ref`, it downloads them from your Gitea instance instead.
It also downloads some docker images from Docker Hub by default, which also requires internet access.
However, internet access is not strictly necessary.
@@ -27,33 +27,24 @@ Still, this is completely optional since both options have the same effect at th
## Where will the runner download scripts when using actions such as `actions/checkout@v4`?
There are tens of thousands of [actions scripts](https://github.com/marketplace?type=actions) in GitHub, and when you write `uses: actions/checkout@v4`, it downloads the scripts from [github.com/actions/checkout](http://github.com/actions/checkout) by default.
But what if you want to use actions from other places such as gitea.com instead of GitHub?
From `https://github.com` by default, or from your own instance when `[actions].DEFAULT_ACTIONS_URL` is `self`, see the [Configuration Cheat Sheet](../../administration/config-cheat-sheet.md#actions-actions).
The good news is that you can specify the URL prefix to use actions from anywhere.
This is an extra syntax in Gitea Actions.
For example:
To use an action from elsewhere, name its host:
- `uses: https://gitea.com/xxx/xxx@xxx`
- `uses: https://github.com/xxx/xxx@xxx`
- `uses: http://your_gitea_instance.com/xxx@xxx`
Be careful, the `https://` or `http://` prefix is necessary!
This is one of the differences from GitHub Actions which supports actions scripts only from GitHub.
But it should allow users much more flexibility in how they run Actions.
Alternatively, if you want your runners to download actions from your own Gitea instance by default, you can configure it by setting `[actions].DEFAULT_ACTIONS_URL`.
See [Configuration Cheat Sheet](../../administration/config-cheat-sheet.md#actions-actions).
- `uses: self:owner/repo@ref` for your own Giteainstance
## How to limit the permission of the runners?
Runners have no more permissions than simply connecting to your Gitea instance.
When any runner receives a job to run, it will temporarily gain limited permission to the repository associated with the job.
If you want to give more permissions to the runner, allowing it to access more private repositories or external systems, you can pass [secrets](usage/actions/secrets.md) to it.
Runners only connect to your Gitea instance.
For each job, a runner gets a `GITEA_TOKEN` limited to the job's repository, see [Actions job token permissions](token-permissions.md).
To give a job access to more private repositories or external systems, pass it [secrets](usage/actions/secrets.md).
Refined permission control to Actions is a complicated job.
In the future, we will add more options to Gitea to make it more configurable, such as allowing more write access to repositories or read access to all repositories in the same organization.
## Which operating systems are supported by Gitea Runner?
Official binaries are released for Linux, macOS and Windows.
Other systems supported by Go and Docker work in theory.
## How to avoid being hacked?
@@ -72,22 +63,6 @@ Here's how we do it on [gitea.com](http://gitea.com/):
- To run actions for fork pull requests, approval is required. See [#22803](https://github.com/go-gitea/gitea/pull/22803).
- If someone registers their own runner for their repository or organization on [gitea.com](http://gitea.com/), we have no objections and will just not use it in our org. However, they should take care to ensure that the runner is not used by other users they do not know.
## Which operating systems are supported by act runner?
It works well on Linux, macOS, and Windows.
While other operating systems are theoretically supported, they require further testing.
One thing to note is that if you choose to run jobs directly on the host instead of in job containers, the environmental differences between operating systems may cause unexpected failures.
For example, bash is not available on Windows in most cases, while act tries to use bash to run scripts by default.
Therefore, you need to specify `powershell` as the default shell in your workflow file, see [defaults.run](https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#defaultsrun).
```yaml
defaults:
run:
shell: powershell
```
## Why choose GitHub Actions? Why not something compatible with GitLab CI/CD?
[@lunny](https://gitea.com/lunny) has explained this in the [issue to implement actions](https://github.com/go-gitea/gitea/issues/13539).
@@ -98,69 +73,55 @@ It is exciting to be able to reuse them.
## What if it runs on multiple labels, such as `runs-on: [label_a, label_b]`?
This is valid syntax.
It means that it should run on runners that have both the `label_a`**and**`label_b` labels, see [Workflow syntax for GitHub Actions](https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#jobsjob_idruns-on).
Unfortunately, act runner does not work this way.
As mentioned, we map labels to environments:
- `ubuntu` → `ubuntu:22.04`
- `centos` → `centos:8`
But we need to map label groups to environments instead, like so:
- `[ubuntu]` → `ubuntu:22.04`
- `[with-gpu]` → `linux:with-gpu`
- `[ubuntu, with-gpu]` → `ubuntu:22.04_with-gpu`
We also need to re-design how tasks are assigned to runners.
A runner with `ubuntu`, `centos`, or `with-gpu` does not necessarily indicate that it can accept jobs with `[centos, with-gpu]`.
Therefore, the runner should inform the Gitea instance that it can only accept jobs with `[ubuntu]`, `[centos]`, `[with-gpu]`, and `[ubuntu, with-gpu]`.
This is not a technical problem, it was just overlooked in the early design.
See [runtime.go#L65](https://gitea.com/gitea/act_runner/src/commit/90b8cc6a7a48f45cc28b5ef9660ebf4061fcb336/runtime/runtime.go#L65).
Currently, the act runner attempts to match everyone in the labels and uses the first match it finds.
The job runs on a runner that has all of the labels, as on [GitHub](https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#jobsjob_idruns-on).
The runner uses the environment of the first label it has.
## What is the difference between agent labels and custom labels for a runner?

Agent labels are reported to the Gitea instance by the runner during registration.
Custom labels, on the other hand, are added manually by a Gitea administrator or owners of the organization or repository (depending on the level of the runner).
However, the design here needs improvement, as it currently has some rough edges.
You can add a custom label such as `centos` to a registered runner, which means the runner will receive jobs with `runs-on: centos`.
However, the runner may not know which environment to use for this label, resulting in it using a default image or leading to a logical dead end.
This default may not match user expectations.
See [runtime.go#L71](https://gitea.com/gitea/act_runner/src/commit/90b8cc6a7a48f45cc28b5ef9660ebf4061fcb336/runtime/runtime.go#L71).
In the meantime, we suggest that you re-register your runner if you want to change its labels.
Gitea no longer has custom labels.
A runner declares its labels every time it starts, from its `runner.labels` configuration or the labels given at registration, so change them there and restart the runner.
## Will there be more implementations for Gitea Actions runner?
Although we would like to provide more options, our limited manpower means that act runner will be the only officially supported runner.
However, both Gitea and act runner are completely open source, so anyone can create a new/better implementation.
We support your choice, no matter how you decide.
In case you fork act runner to create your own version: Please contribute the changes back if you can and if you think your changes will help others as well.
Although we would like to provide more options, our limited manpower means that Gitea Runner will be the only officially supported runner at the moment.
However, both Gitea and Gitea Runner are completely open source under MIT License, so anyone can modify the code to satisfy their requirements.
In case you fork Gitea Runner to create your own version: Please contribute the changes back if you can and if you think your changes will help others as well.
## What workflow trigger events does Gitea support?
All events listed in this table are supported events and are compatible with GitHub.
For events supported only by GitHub, see GitHub's [documentation](https://docs.github.com/en/actions/using-workflows/events-that-trigger-workflows).
> Without an explicit `types:` filter, `pull_request` and `pull_request_target` only run on `opened`, `reopened` and `synchronize`, matching GitHub. All other events run on every activity type they support.
> For `pull_request` events, in [GitHub Actions](https://docs.github.com/en/actions/using-workflows/events-that-trigger-workflows#pull_request), the `ref` is `refs/pull/:prNumber/merge`, which is a reference to the merge commit preview. However, Gitea has no such reference.
> Therefore, the `ref` in Gitea Actions is `refs/pull/:prNumber/head`, which points to the head of pull request rather than the preview of the merge commit.
## How to share actions and reusable workflows from private repositories?
Go to the repository's **Settings** > **Actions** > **General** page and add collaborative owners.
The private repositories of collaborative owners are allowed to access the actions and workflows in the current repository.
To share within the same user or organization, add the private repositories under its **Settings** > **Actions** > **General** > **Cross-Repository Access** instead.
See [Cross-repository access](token-permissions.md#cross-repository-access).
Jobs can request short-lived OpenID Connect (OIDC) tokens to sign in to cloud providers and other services without storing long-lived secrets. The tokens use the GitHub Actions format, so actions like `aws-actions/configure-aws-credentials`, `google-github-actions/auth` and `azure/login` work unchanged.
## Trusting Gitea
Add Gitea as an OIDC identity provider in the service:
- Issuer: `<ROOT_URL>/api/actions/oidc`
- Audience: `<ROOT_URL>/<owner>`, unless the workflow requests another one
Then limit which workflows can sign in with a condition on the [`sub` claim](#claims).
Gitea signs the tokens with a key it creates at `<APP_DATA_PATH>/jwt/actions_oidc.pem`. Instances behind a load balancer must share this file.
## Requesting a token
Grant the `id-token` permission to the workflow or job:
```yaml
permissions:
id-token: write
```
Default permissions never include it, and jobs from fork pull requests never get it. Owners and repositories can disable it with their [maximum token permissions](token-permissions.md#maximum-token-permissions-clamping).
Actions like the ones above request the token themselves. In a script, use `ACTIONS_ID_TOKEN_REQUEST_URL` and `ACTIONS_ID_TOKEN_REQUEST_TOKEN`:
The `pull_request` event uses the second form, all other events use the first, including `pull_request_target`.
The other claims match GitHub's, like `repository`, `ref`, `sha`, `workflow_ref` and `job_workflow_ref`. The discovery document at `<ROOT_URL>/api/actions/oidc/.well-known/openid-configuration` lists them all.
Differences from GitHub Actions:
- There is no `environment` claim.
- Pull request runs use `refs/pull/<index>/head` and its commit, not the merge ref and merge commit.
- `run_number` counts runs per repository, not per workflow.
@@ -25,12 +25,12 @@ If you want to learn more or encounter any problems while configuring it, please
### Set up runner
Gitea Actions requires [act runner](https://gitea.com/gitea/act_runner) to run the jobs.
Gitea Actions requires [Gitea Runner](https://gitea.com/gitea/runner) to run the jobs.
In order to avoid consuming too many resources and affecting the Gitea instance, it is recommended to start runners on separate machines from the Gitea instance.
You can use the [pre-built binaries](http://dl.gitea.com/act_runner) or the [docker images](https://hub.docker.com/r/gitea/act_runner/tags) to set up the runner.
You can use the [pre-built binaries](http://dl.gitea.com/gitea-runner) or the [docker images](https://hub.docker.com/r/gitea/runner/tags) to set up the runner.
Before proceeding any further, we suggest running it as a command line with pre-built binaries to ensure that it works with your environment, especially if you are running a runner on your localhost.
Before proceeding any further, we suggest running it as a command line with pre-built binaries to ensure that it works with your environment, especially if you are running a runner on your localhost.
And it could be easier to debug if something goes wrong.
The runner can run the jobs in isolated Docker containers, so you need to make sure that the Docker has been installed and Docker daemon is running.
@@ -40,7 +40,7 @@ However, it is recommended to use Docker to run the jobs, because it is more sec
Before running a runner, you should first register it to your Gitea instance using the following command:
You can find more information by visiting [Act runner](usage/actions/act-runner.md).
You can find more information by visiting [Gitea Runner](runner.mdx).
### Use Actions
@@ -115,7 +115,7 @@ jobs:
:::warning
Certain actions may not function correctly within SHA256 repositories. This includes [actions/checkout](https://github.com/actions/checkout/issues/1843).
Certain actions may not function correctly within SHA256 repositories or when Gitea runs on subpath. This includes [actions/checkout](https://github.com/actions/checkout/issues/1843).
This page will introduce the [Gitea Runner](https://gitea.com/gitea/runner), which is the runner for Gitea Actions.
:::tip
For detailed installation, configuration, and usage instructions, see the [Gitea Runner documentation](/runner/).
:::
## Obtain a registration token
You can register a runner at different levels. It can be:
- Instance level: The runner will run jobs for all repositories in the instance.
- Organization level: The runner will run jobs for all repositories in the organization.
- Repository level: The runner will run jobs for the repository it belongs to.
Note that the repository may still use instance-level or organization-level runners even if it has its own repository-level runners. A future release may provide an option to allow more control over this.
Before registering the runner and running it, you need a registration token. The level of the runner determines where to obtain the registration token.
- Instance level: The admin settings page, like `<your_gitea.com>/-/admin/actions/runners`.
- Organization level: The organization settings page, like `<your_gitea.com>/<org>/settings/actions/runners`.
- Repository level: The repository settings page, like `<your_gitea.com>/<owner>/<repo>/settings/actions/runners`.
If you cannot see the settings page, please make sure that you have the right permissions and that Actions have been enabled.
The format of the registration token is a random string `D0gvfu2iHfUjNqCYVljVyRV14fISpJxxxxxxxxxx`.
A registration token can also be obtained from the Gitea [command-line interface](../../administration/command-line.md#actions-generate-runner-token):
You can also use `GITEA_RUNNER_REGISTRATION_TOKEN`/`GITEA_RUNNER_REGISTRATION_TOKEN_FILE` environment variables to set a global runner registration token when Gitea starts, for example:
Scoped workflows let you maintain Actions workflows in one central **source** repository and have them run automatically on many other repositories, without copying the workflow files into each one.
A workflow file committed under a source repository's scoped-workflow directory runs on every repository that the source applies to (a "consuming" repository). Each run executes **in the consuming repository's own context**: its runners, secrets, `GITEA_TOKEN`, and branch, while the workflow **content is read from the source repository**. This is useful for mandating shared CI across an organization or instance: linting, security scans, compliance or policy checks, and similar.
## Levels
A source repository can be registered at two levels:
- **Owner level**: registered by an organization or user. The source's workflows run on every repository owned by that organization or user.
- **Instance level**: registered by a site administrator. The source's workflows run on **every** repository on the instance.
The same repository may be registered at both levels; it is still evaluated once per matching event.
:::note
Instance-level sources apply to **every** repository on the instance, and a required one cannot be opted out. Detection runs on every repository's events. Register them deliberately.
:::
## Configuration
Scoped workflows live in a directory that is separate from regular workflows. It is controlled by `SCOPED_WORKFLOW_DIRS` in the `[actions]` section of `app.ini`:
```ini
[actions]
SCOPED_WORKFLOW_DIRS = .gitea/scoped_workflows
```
- The default is `.gitea/scoped_workflows`.
- It may list multiple directories (comma-separated).
- It **must not overlap** with `WORKFLOW_DIRS` (the regular workflow directories, by default `.gitea/workflows` and `.github/workflows`).
- Leaving it empty means no directory holds scoped workflows, so none are found or run. The settings page still appears and sources can still be registered, but no scoped workflow runs.
Keeping scoped workflows in their own directory means a source repository's *own* Actions are unaffected: only files under `SCOPED_WORKFLOW_DIRS` are treated as scoped workflows.
## Providing scoped workflows from a repository
1. In the would-be source repository, commit your workflow files under the scoped-workflow directory on its **default branch**, for example `.gitea/scoped_workflows/lint.yaml`:
Search for and add the repository. From then on, its scoped workflows apply to the consuming repositories at that level.
The source repository is within its own scope, so it runs these workflows on itself like any other consumer.
## How scoped workflows run
- The run belongs to the **consuming** repository and uses its runners, secrets, `GITEA_TOKEN`, and ref. It behaves like a normal run there, including rerun, logs, and commit statuses.
- The workflow **content is read from the source** repository, pinned to the source's default-branch commit at the time of the event.
- Detection uses the consuming repository's event, so the workflow's `on:` triggers (such as `push` and `pull_request`) must match that event.
## Opting out
A consuming repository can disable a scoped workflow it does not want, from its **Actions** page (the workflow appears under its source). A workflow that has been marked **required** (see below) can never be disabled or opted out by a consuming repository.
## Required workflows and the merge gate
On the Scoped Workflows settings page you can mark individual workflows as **required** and give each one or more **status-check patterns** (one glob per line). A required scoped workflow:
- Cannot be disabled by consuming repositories.
- Gates pull-request merges: a consuming pull request can only be merged once a commit status matching **every** pattern has **passed** (must-present-and-pass). A required check that never posts a status **blocks** the merge rather than being silently skipped, so disabling Actions on the consumer cannot bypass it.
The status-check context a scoped run posts has the form:
```
<source repo full name>: <workflow display name> / <job> (<event>)
```
The settings page previews these "Expected status checks" for each workflow and marks the ones your patterns match, so you can confirm a pattern is correct. A common pattern wildcards the job and event, for example `my-org/ci-repo: Lint / *`.
Enforcement scope:
- A required check is enforced on any target branch that has a **branch protection rule**, even one whose own status checks are disabled.
- A target branch with **no** protection rule is not gated.
:::warning
Only a workflow that runs on an event that posts a commit status (`push`, `pull_request`, `pull_request_target`, `release`) can be required. A workflow that runs only on events like `workflow_dispatch`, `schedule` or `workflow_call` posts no status, so marking it required would block every consuming pull request forever. The settings page warns you in this case.
:::
## Reusable workflows (`uses:`)
Scoped workflows can call reusable workflows:
- A **local** reference (`uses: ./...` or `uses: $/...`) resolves against the **source** repository: the repository the calling workflow's content came from, not the consuming repository.
- A **cross-repository** reference (`uses: owner/repo/...@ref` or `uses: self:owner/repo/...@ref`) is resolved with the **consuming repository's** read permission. If it points to a private repository, make sure every consuming repository can read it; otherwise the workflow will fail there.
A reusable workflow can live under `SCOPED_WORKFLOW_DIRS` (or `WORKFLOW_DIRS`) of the source repository.
## Security considerations
A source repository's workflow content is executed in every repository it applies to, and its step scripts and their output are written to that repository's Actions logs, readable by anyone who can view the consuming repository's Actions.
- Registering a **private** repository as a source therefore discloses its workflow logic through those logs. Only register repositories whose workflow content may be shared with every consuming repository.
## Limitations
- `on: schedule` and `on: workflow_run` are currently not supported as scoped workflow triggers.
- Scoped workflow content is read from the source's default branch; other branches of the source are not used.
Every Actions job receives a built-in token (`GITEA_TOKEN`) which can be used to access Gitea (Git over HTTP(S), API requests, etc.).
This page documents how Gitea decides what the token is allowed to do.
In workflows, it is available as `${{ secrets.GITEA_TOKEN }}`. These settings and `permissions:` only affect `GITEA_TOKEN` (not other secrets like personal access tokens). For API calls, see [API authentication](../../development/api-usage.md#authentication).
## Where permissions come from
Gitea determines the job token permissions in this order:
- For repositories in an organization, the repository can follow the owner-level configuration, or opt out with **Override owner-level configuration**.
Note: Repository-level token permission settings are only shown once **Repository Actions** are enabled.
## Fork pull requests
For security reasons, workflows triggered by pull requests from forks are always restricted to read-only permissions for repository contents, regardless of workflow `permissions:` or settings. They also get no OpenID Connect tokens.
## Cross-repository access
By default, `GITEA_TOKEN` only has access to:
- The job's repository (according to the computed permissions)
- Public repositories (read-only)
Access to other private repositories is denied by default.
A user or organization can allow read-only access to selected private repositories via **Settings → Actions → General → Cross-Repository Access**.
Private repositories can also allow selected collaborative owners via **Settings → Actions → General**.
This allows the private repositories of those owners to read the current private repository for Actions use, such as private actions and reusable workflows.
Fork pull request workflows never gain cross-repository access to private repositories, even if those repositories are listed.
These environment variables are set automatically in every workflow run and can be accessed directly (e.g. `$CI` in shell scripts).
| Name | Description | Example |
|---|---|---|
| `CI` | Always set to `true`. | `true` |
| `GITEA_ACTIONS` | Always set to `true`. Useful to distinguish Gitea Actions from other CI systems. | `true` |
| `GITEA_ACTIONS_RUNNER_VERSION` | The version of the runner executing the workflow. | `1.0.8` |
| `GITEA_DOCKER_WORKSPACE` | The workspace path as the Docker daemon sees it, for bind mounts into containers a job starts, e.g. `${GITEA_DOCKER_WORKSPACE:-.}/data:/app/data` in a compose file. Only set for jobs running in a container. | `/var/lib/docker/volumes/GITEA-ACTIONS-TASK-42_***/_data` |
| `GITEA_ENV`<br/>`GITHUB_ENV` | Path to the file that sets environment variables for subsequent steps. | `/home/runner/work/_temp/_runner_file_commands/set_env_***` |
| `GITEA_OUTPUT`<br/>`GITHUB_OUTPUT` | Path to the file that sets step output parameters. | `/home/runner/work/_temp/_runner_file_commands/set_output_***` |
| `GITEA_PATH`<br/>`GITHUB_PATH` | Path to the file that adds system `PATH` entries for subsequent steps. | `/home/runner/work/_temp/_runner_file_commands/add_path_***` |
| `GITEA_STATE`<br/>`GITHUB_STATE` | Path to the file that sets step state variables. | `/home/runner/work/_temp/_runner_file_commands/save_state_***` |
| `GITEA_STEP_SUMMARY`<br/>`GITHUB_STEP_SUMMARY` | Path to the file for writing job summaries. | `/home/runner/work/_temp/_runner_file_commands/step_summary_***` |
| `GITHUB_ACTIONS` | Always set to `true`. | `true` |
| `GITHUB_ACTION` | The name of the action currently running, or the step `id`. | `__run` |
| `GITHUB_ACTION_PATH` | The path where the action is located. | `/home/runner/work/_actions/actions/checkout/v4` |
| `GITHUB_ACTION_REF` | The ref of the action being executed. | `v4` |
| `GITHUB_ACTION_REPOSITORY` | The owner and repository of the action. | `actions/checkout` |
| `GITHUB_ACTOR` | The username of the user that triggered the workflow. | `silverwind` |
| `GITHUB_API_URL` | The URL of the REST API. | `https://gitea.com/api/v1` |
| `GITHUB_BASE_REF` | The target branch of a pull request. | `main` |
| `GITHUB_EVENT_NAME` | The name of the event that triggered the workflow. | `push` |
| `GITHUB_EVENT_PATH` | Path to the file containing the event webhook payload. | `/home/runner/work/_temp/_github_workflow/event.json` |
| `GITHUB_GRAPHQL_URL` | Empty in Gitea (GraphQL is not supported). | _(empty)_ |
| `GITHUB_HEAD_REF` | The source branch of a pull request. | `feature-branch` |
| `GITHUB_JOB` | The `job_id` of the current job. | `build` |
| `GITHUB_REF` | The fully-formed ref that triggered the workflow. | `refs/heads/main` |
| `GITHUB_REF_NAME` | The short ref name. | `main` |
| `GITHUB_REF_TYPE` | The type of ref: `branch` or `tag`. | `branch` |
| `GITHUB_REPOSITORY` | The owner and repository name. | `gitea/docs` |
| `GITHUB_REPOSITORY_OWNER` | The repository owner's username. | `gitea` |
| `GITHUB_RETENTION_DAYS` | The number of days that workflow run logs and artifacts are kept. | `90` |
| `GITHUB_RUN_ATTEMPT` | The attempt number of the workflow run. | `1` |
| `GITHUB_RUN_ID` | A unique number for each workflow run. | `1234` |
| `GITHUB_RUN_NUMBER` | A unique number for each run of a particular workflow. | `42` |
| `GITHUB_SERVER_URL` | The URL of the Gitea instance. | `https://gitea.com` |
| `GITHUB_SHA` | The commit SHA that triggered the workflow. | `a1b2c3d4e5f6a1b2c3d4e5f6a1b2c3d4e5f6a1b2` |
| `GITHUB_WORKFLOW` | The name of the workflow. | `CI` |
| `GITHUB_WORKSPACE` | The default working directory on the runner. | `/workspace/gitea/docs` |
### Runner environment variables
| Name | Description | Example |
|---|---|---|
| `RUNNER_ARCH` | The architecture of the runner. | `X64` |
| `RUNNER_OS` | The operating system of the runner. | `Linux` |
| `RUNNER_TEMP` | Path to a temporary directory on the runner. | `/tmp` |
| `RUNNER_TOOL_CACHE` | Path to the tool cache directory on the runner. | `/opt/hostedtoolcache` |
### Internal environment variables
These are used internally by the runner and actions. They are typically not needed in workflows directly.
| Name | Description | Example |
|---|---|---|
| `ACTIONS_CACHE_URL` | URL for the actions cache service. | `http://192.168.1.10:8088/` |
| `ACTIONS_ID_TOKEN_REQUEST_TOKEN` | Token for OIDC requests. Only set when `id-token: write` is granted. | `***` |
| `ACTIONS_ID_TOKEN_REQUEST_URL` | URL to request OIDC tokens. Only set when `id-token: write` is granted. | `https://gitea.com/api/actions/oidc/token?` |
| `ACTIONS_RESULTS_URL` | URL for storing artifacts. | `https://gitea.com` |
| `ACTIONS_RUNTIME_TOKEN` | Authentication token for the Actions pipeline API. | `***` |
| `ACTIONS_RUNTIME_URL` | URL for the Gitea Actions pipeline API. | `https://gitea.com/api/actions_pipeline/` |
## Context variables
These variables are available in workflow expressions via `${{ gitea.<name> }}`. For compatibility, `${{ github.<name> }}` works as an alias.
| Name | Description | Example |
|---|---|---|
| `gitea.action`<br/>`github.action` | The name of the action currently running, or the `id` of a step. | `__run` |
| `gitea.action_path`<br/>`github.action_path` | The path where an action is located. Only supported in composite actions. | `/home/runner/work/_actions/actions/checkout/v4` |
| `gitea.action_ref`<br/>`github.action_ref` | The ref of the action being executed. | `v4` |
| `gitea.action_repository`<br/>`github.action_repository` | The owner and repository name of the action. | `actions/checkout` |
| `gitea.action_status`<br/>`github.action_status` | The current result of a composite action. | `success` |
| `gitea.actor`<br/>`github.actor` | The username of the user that triggered the initial workflow run. | `silverwind` |
| `gitea.api_url`<br/>`github.api_url` | The URL of the REST API. | `https://gitea.com/api/v1` |
| `gitea.base_ref`<br/>`github.base_ref` | The target branch of a pull request. Only set for `pull_request` and `pull_request_target` events. | `main` |
| `gitea.env`<br/>`github.env` | Path on the runner to the file that sets environment variables from workflow commands. Unique to each step. | `/home/runner/work/_temp/_runner_file_commands/set_env_***` |
| `gitea.event`<br/>`github.event` | The full event webhook payload as an object. See [Event payload](#event-payload). | `{...}` |
| `gitea.event_name`<br/>`github.event_name` | The name of the event that triggered the workflow run. | `push` |
| `gitea.event_path`<br/>`github.event_path` | Path on the runner to the file containing the full event webhook payload. | `/home/runner/work/_temp/_github_workflow/event.json` |
| `gitea.head_ref`<br/>`github.head_ref` | The source branch of a pull request. Only set for `pull_request` and `pull_request_target` events. | `feature-branch` |
| `gitea.job`<br/>`github.job` | The `job_id` of the current job. | `build` |
| `gitea.ref`<br/>`github.ref` | The fully-formed ref that triggered the workflow. | `refs/heads/main` |
| `gitea.ref_name`<br/>`github.ref_name` | The short ref name. | `main` |
| `gitea.ref_protected`<br/>`github.ref_protected` | `true` if branch protections are configured for the ref that triggered the workflow run. | `true` |
| `gitea.ref_type`<br/>`github.ref_type` | The type of ref: `branch` or `tag`. | `branch` |
| `gitea.path`<br/>`github.path` | Path on the runner to the file that sets system `PATH` variables from workflow commands. Unique to each step. | `/home/runner/work/_temp/_runner_file_commands/add_path_***` |
| `gitea.repository`<br/>`github.repository` | The owner and repository name. | `gitea/docs` |
| `gitea.repository_owner`<br/>`github.repository_owner` | The repository owner's username. | `gitea` |
| `gitea.repositoryUrl`<br/>`github.repositoryUrl` | The HTML URL to the repository. | `https://gitea.com/gitea/docs` |
| `gitea.retention_days`<br/>`github.retention_days` | The number of days that workflow run logs and artifacts are kept. | `90` |
| `gitea.run_id`<br/>`github.run_id` | A unique number for each workflow run within a repository. Does not change on re-run. | `1234` |
| `gitea.run_number`<br/>`github.run_number` | A unique number for each run of a particular workflow. Starts at 1 and increments with each new run. | `42` |
| `gitea.run_attempt`<br/>`github.run_attempt` | A unique number for each re-run attempt. Starts at 1 and increments with each re-run. | `1` |
| `gitea.secret_source`<br/>`github.secret_source` | The source of a secret used in a workflow. Always `Actions` in Gitea. | `Actions` |
| `gitea.server_url`<br/>`github.server_url` | The URL of the Gitea instance. | `https://gitea.com` |
| `gitea.sha`<br/>`github.sha` | The commit SHA that triggered the workflow. | `a1b2c3d4e5f6a1b2c3d4e5f6a1b2c3d4e5f6a1b2` |
| `gitea.token`<br/>`github.token` | A token to authenticate on behalf of the Gitea App installed on the repository. See [Token permissions](token-permissions.md). | `ghs_***` |
| `gitea.triggering_actor`<br/>`github.triggering_actor` | The username of the user that initiated the workflow run. May differ from `actor` on re-runs. | `silverwind` |
| `gitea.workflow`<br/>`github.workflow` | The name of the workflow. If unnamed, the full path of the workflow file. | `CI` |
| `gitea.workspace`<br/>`github.workspace` | The default working directory on the runner and the default location of your repository when using the `checkout` action. | `/workspace/gitea/docs` |
| `gitea.gitea_default_actions_url` | The default URL for downloading actions. Gitea-specific. | `https://github.com` |
### Event payload
`gitea.event` is the webhook payload of the event that triggered the run, so which fields exist depends on `gitea.event_name`.
The same payload is written as JSON to the file at `gitea.event_path`.
Payloads use Gitea's webhook format, which is close to GitHub's but not identical. Differences are noted below.
Reading a field that the payload does not have returns an empty value instead of failing, so `${{ gitea.event.pull_request.number }}` is empty on a `push` run.
These fields exist in almost every payload:
| Name | Description | Example |
|---|---|---|
| `gitea.event.action` | The activity type. Only set for events that have activity types, see [supported events](faq.md#what-workflow-trigger-events-does-gitea-support). | `opened` |
| `gitea.event.repository.full_name` | The owner and name of the repository. | `gitea/docs` |
| `gitea.event.repository.default_branch` | The default branch of the repository. | `main` |
| `gitea.event.repository.private` | Whether the repository is private. | `false` |
| `gitea.event.repository.html_url` | The HTML URL of the repository. | `https://gitea.com/gitea/docs` |
| `gitea.event.sender.login` | The username of the user that triggered the event. Also available as `sender.username`. | `silverwind` |
The remaining fields are specific to one or a few events:
| Name | Events | Description | Example |
|---|---|---|---|
| `gitea.event.schedule` | `schedule` | The cron expression that triggered the run. | `*/5 * * * *` |
| `gitea.event.inputs.<name>` | `workflow_dispatch`, `workflow_call` | A raw input value. Prefer `${{ inputs.<name> }}`, which also applies the input's `default`. | `true` |
| `gitea.event.workflow` | `workflow_dispatch` | The file name of the dispatched workflow. | `ci.yaml` |
| `gitea.event.ref` | `push`, `create`, `delete`, `workflow_dispatch` | The ref of the event. Unlike GitHub, `create` and `delete` use the full ref instead of the short name. | `refs/heads/main` |
| `gitea.event.ref_type` | `create`, `delete` | The type of ref: `branch` or `tag`. | `tag` |
| `gitea.event.before`<br/>`gitea.event.after` | `push` | The commit SHA before and after the push. | `a1b2c3d4e5f6a1b2c3d4e5f6a1b2c3d4e5f6a1b2` |
| `gitea.event.commits` | `push` | The commits contained in the push, each with `id`, `message`, `url`, `author`, `added`, `removed` and `modified`. | `[{...}]` |
| `gitea.event.head_commit.message` | `push` | The message of the most recent commit of the push. | `fix: broken link` |
| `gitea.event.total_commits` | `push` | The number of commits in the push. Gitea-specific. | `3` |
| `gitea.event.compare_url` | `push` | The URL comparing `before` with `after`. GitHub names this field `compare`. | `https://gitea.com/gitea/docs/compare/a1b2c3d...e5f6a1b` |
| `gitea.event.pusher.login` | `push` | The username of the user that pushed. | `silverwind` |
| `gitea.event.forkee.full_name` | `fork` | The owner and name of the new fork. | `silverwind/docs` |
| `gitea.event.page` | `gollum` | The name of the wiki page. | `Home` |
| `gitea.event.number` | `issues`, `pull_request`, `pull_request_target`, `pull_request_review`, `pull_request_review_comment` | The index of the issue or pull request. Not set for `issue_comment`, use `gitea.event.issue.number` there. | `386` |
| `gitea.event.issue.number`<br/>`gitea.event.issue.title`<br/>`gitea.event.issue.body`<br/>`gitea.event.issue.state`<br/>`gitea.event.issue.labels`<br/>`gitea.event.issue.user.login` | `issues`, `issue_comment` | The issue the event happened on. | `Expand gitea.event in actions docs` |
| `gitea.event.comment.body` | `issue_comment` | The body of the comment. | `LGTM` |
| `gitea.event.is_pull` | `issue_comment` | Whether the comment was made on a pull request. Gitea-specific, GitHub sets `issue.pull_request` on the issue object instead. | `true` |
| `gitea.event.pull_request.number`<br/>`gitea.event.pull_request.title`<br/>`gitea.event.pull_request.body`<br/>`gitea.event.pull_request.draft`<br/>`gitea.event.pull_request.merged`<br/>`gitea.event.pull_request.state`<br/>`gitea.event.pull_request.labels`<br/>`gitea.event.pull_request.user.login` | `pull_request`, `pull_request_target`, `pull_request_review`, `pull_request_review_comment` | The pull request the event happened on. | `Add caching to CI` |
| `gitea.event.pull_request.head.ref`<br/>`gitea.event.pull_request.head.sha`<br/>`gitea.event.pull_request.head.repo.full_name` | `pull_request`, `pull_request_target`, `pull_request_review`, `pull_request_review_comment` | The source branch of the pull request. | `feature-branch` |
| `gitea.event.pull_request.base.ref`<br/>`gitea.event.pull_request.base.sha` | `pull_request`, `pull_request_target`, `pull_request_review`, `pull_request_review_comment` | The target branch of the pull request. | `main` |
| `gitea.event.review.type` | `pull_request_review`, `pull_request_review_comment` | The kind of review: `pull_request_review_approved`, `pull_request_review_rejected` or `pull_request_review_comment`. GitHub uses `review.state` with the values `approved`, `changes_requested` and `commented`. | `pull_request_review_approved` |
| `gitea.event.review.content` | `pull_request_review`, `pull_request_review_comment` | The body of the review. GitHub names this field `review.body`. | `Ship it` |
| `gitea.event.changes.title.from`<br/>`gitea.event.changes.body.from` | `issues`, `issue_comment`, `pull_request` | The previous value. Only set when `action` is `edited`. | `Old title` |
| `gitea.event.release.tag_name`<br/>`gitea.event.release.name`<br/>`gitea.event.release.body`<br/>`gitea.event.release.draft`<br/>`gitea.event.release.prerelease` | `release` | The release the event happened on. | `v1.2.3` |
| `gitea.event.package.name`<br/>`gitea.event.package.type`<br/>`gitea.event.package.version` | `registry_package` | The package the event happened on. | `gitea` |
| `gitea.event.workflow_run.id`<br/>`gitea.event.workflow_run.event`<br/>`gitea.event.workflow_run.status`<br/>`gitea.event.workflow_run.conclusion`<br/>`gitea.event.workflow_run.head_branch`<br/>`gitea.event.workflow_run.head_sha` | `workflow_run` | The workflow run that the event reports on. | `success` |
`schedule` runs have no webhook of their own. Their payload is the payload of the push that registered the schedule, with `schedule`, `repository`, `sender` and, for organization-owned repositories, `organization` filled in.
Only those fields are meaningful, and `sender.login` is always `gitea-actions`.
## User-defined variables
You can create configuration variables on the user, organization and repository level.
The level of the variable depends on where you created it. When creating a variable, the
key will be converted to uppercase. You need use uppercase on the yaml file.
## Naming conventions
### Naming conventions
The following rules apply to variable names:
@@ -21,12 +181,12 @@ The following rules apply to variable names:
- Variable names must be unique at the level they are created at.
- Variable names must not start with `CI`.
## Using variable
### Using variables
After creating configuration variables, they will be automatically filled in the `vars` context.
They can be accessed through expressions like `${{ vars.VARIABLE_NAME }}` in the workflow.
## Precedence
### Precedence
If a variable with the same name exists at multiple levels, the variable at the lowest level takes precedence:
A repository variable will always be chosen over an organization/user variable.
In Gitea `1.13`, support for [AGit](https://git-repo.info/en/2020/03/agit-flow-and-git-repo/) was added. AGit enables users to create pull requests directly, even without write permissions of the repository, eliminating the need to fork it. This helps reduce the number of duplicated repositories and minimizes unnecessary disk usage.
:::note
Git version 2.29 or higher is required on the server side for this to work.
:::
## Creating PRs with AGit
AGit allows to create PRs while pushing code to the remote repo.
This can be done by pushing to a special refspec (a location identifier known to git) instead of to a branch.
- `<remote>`: The remote pointing at the target repository (not a fork!), `origin` in the example above
- `<local-rev>`: The local branch or commit containing the changes you are proposing, `HEAD` in the example above
- `refs/for/<branch>`: The target PR type and configuration
- `for`: Create a normal PR with `<branch>` as the target branch. This is the only prefix supported at the moment, `refs/draft/...` and `refs/for-review/...` are reserved for future use and do not create a PR.
- `<branch>`: The branch you want your changes to be merged into
- `<topic>`: Optional, an alternative to `-o topic=<topic>`. The push is rejected with `topic-branch is not set` if the topic is given in neither place.
- `-o <option>[=<value>]`: Options for the PR
- `topic`: The topic of this change. It will become the name of the branch holding the changes waiting for review (prefixed with your user name). This is REQUIRED to trigger a pull request, unless it is already part of the refspec.
- `title`: The PR title (optional but recommended), only used for topics not already having an associated PR.
- `description`: The PR description (optional but recommended), only used for topics not already having an associated PR.
- `force-push=true`: Specifies whether to force-update the target branch.
- Note: omitting the value and using just `-o force-push` will also work.
Here's another example for creating a new PR targeting `main` with `topic`, `title`, and `description`:
```shell
git push origin HEAD:refs/for/main -o topic="topic_of_my_PR" -o title="Title of the PR" -o description="# The PR Description\nThis can be **any** markdown content.\n- [x] Ok"
Some projects have a standard list of questions that users need to answer
when creating an issue or pull request. Gitea supports adding templates to the
**default branch of the repository** so that they can autopopulate the form when users are
creating issues and pull requests. This will cut down on the initial back and forth
of getting some clarifying details.
It is currently not possible to provide generic issue/pull-request templates globally.
Additionally, the New Issue page URL can be suffixed with `?title=Issue+Title&body=Issue+Text` and the form will be populated with those strings. Those strings will be used instead of the template if there is one. See [Pre-filling Issues and Pull Requests](prefill-issues-pull-requests.md) for all parameters.
## File names
Possible file names for issue templates:
- `ISSUE_TEMPLATE.md`
- `ISSUE_TEMPLATE.yaml`
- `ISSUE_TEMPLATE.yml`
- `issue_template.md`
- `issue_template.yaml`
- `issue_template.yml`
- `.gitea/ISSUE_TEMPLATE.md`
- `.gitea/ISSUE_TEMPLATE.yaml`
- `.gitea/ISSUE_TEMPLATE.yml`
- `.gitea/issue_template.md`
- `.gitea/issue_template.yaml`
- `.gitea/issue_template.yml`
- `.github/ISSUE_TEMPLATE.md`
- `.github/ISSUE_TEMPLATE.yaml`
- `.github/ISSUE_TEMPLATE.yml`
- `.github/issue_template.md`
- `.github/issue_template.yaml`
- `.github/issue_template.yml`
Possible file names for issue config:
- `.gitea/ISSUE_TEMPLATE/config.yaml`
- `.gitea/ISSUE_TEMPLATE/config.yml`
- `.gitea/issue_template/config.yaml`
- `.gitea/issue_template/config.yml`
- `.github/ISSUE_TEMPLATE/config.yaml`
- `.github/ISSUE_TEMPLATE/config.yml`
- `.github/issue_template/config.yaml`
- `.github/issue_template/config.yml`
Possible file names for PR templates:
- `PULL_REQUEST_TEMPLATE.md`
- `PULL_REQUEST_TEMPLATE.yaml`
- `PULL_REQUEST_TEMPLATE.yml`
- `pull_request_template.md`
- `pull_request_template.yaml`
- `pull_request_template.yml`
- `.gitea/PULL_REQUEST_TEMPLATE.md`
- `.gitea/PULL_REQUEST_TEMPLATE.yaml`
- `.gitea/PULL_REQUEST_TEMPLATE.yml`
- `.gitea/pull_request_template.md`
- `.gitea/pull_request_template.yaml`
- `.gitea/pull_request_template.yml`
- `.github/PULL_REQUEST_TEMPLATE.md`
- `.github/PULL_REQUEST_TEMPLATE.yaml`
- `.github/PULL_REQUEST_TEMPLATE.yml`
- `.github/pull_request_template.md`
- `.github/pull_request_template.yaml`
- `.github/pull_request_template.yml`
## Directory names
Alternatively, users can create multiple issue templates inside a special directory and allow users to choose one that more specifically
addresses their problem.
Possible directory names for issue templates:
- `ISSUE_TEMPLATE`
- `issue_template`
- `.gitea/ISSUE_TEMPLATE`
- `.gitea/issue_template`
- `.github/ISSUE_TEMPLATE`
- `.github/issue_template`
- `.gitlab/ISSUE_TEMPLATE`
- `.gitlab/issue_template`
Inside the directory can be multiple markdown (`.md`) or yaml (`.yaml`/`.yml`) issue templates of the form.
## Syntax for markdown template
```md
---
name: "Template Name"
about: "This template is for testing!"
title: "[TEST] "
ref: "main"
assignees: ["user1"]
projects:
- Example Project
labels:
- bug
- "help needed"
---
This is the template!
```
In the above example, when a user is presented with the list of issues they can submit, this would show as `Template Name` with the description
`This template is for testing!`. When submitting an issue with the above example, the issue title would be pre-populated with
`[TEST] ` while the issue body would be pre-populated with `This is the template!`.
The issue would be assigned to `user1`.
The issue would be assigned to the project `Example Project`.
The issue would also be assigned two labels,
`bug` and `help needed`, and the issue will have a reference to `main`.
## Syntax for yaml template
This example YAML configuration file defines an issue form using several inputs to report a bug.
```yaml
name: Bug Report
about: File a bug report
title: "[Bug]: "
body:
- type: markdown
attributes:
value: |
Thanks for taking the time to fill out this bug report!
# some markdown that will only be visible once the issue has been created
- type: markdown
attributes:
value: |
This issue was created by an issue **template** :)
visible: [content]
- type: input
id: contact
attributes:
label: Contact Details
description: How can we get in touch with you if we need more info?
placeholder: ex. email@example.com
validations:
required: false
- type: textarea
id: what-happened
attributes:
label: What happened?
description: Also tell us, what did you expect to happen?
placeholder: Tell us what you see!
value: "A bug happened!"
validations:
required: true
- type: dropdown
id: version
attributes:
label: Version
description: What version of our software are you running?
options:
- 1.0.2 (Default)
- 1.0.3 (Edge)
validations:
required: true
- type: dropdown
id: browsers
attributes:
label: What browsers are you seeing the problem on?
multiple: true
options:
- Firefox
- Chrome
- Safari
- Microsoft Edge
- type: textarea
id: logs
attributes:
label: Relevant log output
description: Please copy and paste any relevant log output. This will be automatically formatted into code, so no need for backticks.
render: shell
- type: checkboxes
id: terms
attributes:
label: Code of Conduct
hide_label: true
description: By submitting this issue, you agree to follow our [Code of Conduct](https://example.com)
options:
- label: I agree to follow this project's Code of Conduct
required: true
- label: I have also read the CONTRIBUTION.MD
required: true
visible: [form]
- label: This is a TODO only visible after issue creation
visible: [content]
```
### Markdown
You can use a `markdown` element to display Markdown in your form that provides extra context to the user, but is not submitted by default.
| label | A brief description of the expected user input, which is also displayed in the form. | Required | String | - | - |
| hide_label | If true, the label normally used as a headline is not visible. | Optional | Boolean | false | - |
| description | A description of the text area to provide context or guidance, which is displayed in the form. | Optional | String | Empty String | - |
| placeholder | A semi-opaque placeholder that renders in the text area when empty. | Optional | String | Empty String | - |
| value | Text that is pre-filled in the text area. | Optional | String | - | - |
| render | If a value is provided, submitted text will be formatted into a codeblock. When this key is provided, the text area will not expand for file attachments or Markdown editing. | Optional | String | - | Languages known to Gitea. |
| required | Prevents form submission until element is completed. | Optional | Boolean | false | - |
| is_number | Prevents form submission until element is filled with a number. | Optional | Boolean | false | - |
| regex | Prevents form submission until element is filled with a value that match the regular expression. | Optional | String | - | a [regular expression](https://en.wikipedia.org/wiki/Regular_expression) |
visible: Default is **[form, content]**
### Dropdown
You can use a `dropdown` element to add a dropdown menu in your form.
| label | A brief description of the expected user input, which is displayed in the form. | Required | String | - | - |
| hide_label | If true, the label normally used as a headline is not visible. | Optional | Boolean | false | - |
| description | A description of the dropdown to provide extra context or guidance, which is displayed in the form. | Optional | String | Empty String | - |
| multiple | Determines if the user can select more than one option. | Optional | Boolean | false | - |
| list | If true, display as a list. If false, print items on one line with commas. | Optional | Boolean | false | - |
| options | An array of options the user can choose from. Cannot be empty and all choices must be distinct. | Required | String array | - | - |
| label | A brief description of the expected user input, which is displayed in the form. | Required | String | - | - |
| hide_label | If true, the label normally used as a headline is not visible. | Optional | Boolean | false | - |
| description | A description of the set of checkboxes, which is displayed in the form. Supports Markdown formatting. | Optional | String | Empty String | - |
| options | An array of checkboxes that the user can select. For syntax, see below. | Required | Array | - | - |
For each value in the options array, you can set the following keys.
| label | The identifier for the option, which is displayed in the form. Markdown is supported for bold or italic text formatting, and hyperlinks. | Required | String | - | - |
| required | Prevents form submission until element is completed. | Optional | Boolean | false | - |
| visible | Whether a specific checkbox appears in the form only, in the created issue only, or both. Valid options are "form" and "content". | Optional | String array | false | - |
You can use labels to classify issues and pull requests and to improve your overview over them.
## Creating Labels
For repositories, labels can be created by going to `Issues` and clicking on `Labels`.
For organizations, you can define organization-wide labels that are shared with all organization repositories, including both already-existing repositories as well as newly created ones. Organization-wide labels can be created in the organization `Settings`.
Labels have a mandatory name, a mandatory color, an optional description, and must either be exclusive or not (see `Scoped Labels` below).
When you create a repository, you can ensure certain labels exist by using the `Issue Labels` option. This option lists a number of available label sets that are [configured globally on your instance](../../administration/customizing-gitea.md#labels). Its contained labels will all be created as well while creating the repository.
## Scoped Labels
Scoped labels are used to ensure at most a single label with the same scope is assigned to an issue or pull request. For example, if labels `kind/bug` and `kind/enhancement` have the Exclusive option set, an issue can only be classified as a bug or an enhancement.
A scoped label must contain `/` in its name (not at either end of the name). The scope of a label is determined based on the **last**`/`, so for example the scope of label `scope/subscope/item` is `scope/subscope`.
## Filtering by Label
Issue and pull request lists can be filtered by label. Selecting multiple labels shows issues and pull requests that have all selected labels assigned.
By holding alt to click the label, issues and pull requests with the chosen label are excluded from the list.
- PullRequestPosterName: Pull request's poster name
- PullRequestIndex: Pull request's index number
- PullRequestReference: Pull request's reference char with index number. i.e. #1, !2
- ClosingIssues: return a string contains all issues which will be closed by this pull request i.e. `close #1, close #2`
- ReviewedOn: Which pull request this commit belongs to. For example `Reviewed-on: https://gitea.com/foo/bar/pulls/1`
- ReviewedBy: Who approved the pull request before the merge. For example `Reviewed-by: Jane Doe <[email protected]>`
## Rebase
When rebasing without a merge commit, `REBASE_TEMPLATE.md` modifies the message of the last commit. The following additional variables are available in this template:
A Pull Request (PR) is a way to propose changes to a repository.
It is a request to merge one branch into another, accompanied by a description of the changes that were made.
Pull Requests are commonly used as a way for contributors to propose changes and for maintainers to review and merge those changes.
## Creating a pull request
To create a PR, you'll need to follow these steps:
1. **Fork the repository** - If you don't have permission to make changes to the repository directly, you'll need to fork the repository to your own account.
This creates a copy of the repository that you can make changes to.
2. **Create a branch (optional)** - Create a new branch on your forked repository that contains the changes you want to propose.
Give the branch a descriptive name that indicates what the changes are for.
3. **Make your changes** - Make the changes you want, commit, and push them to your forked repository.
4. **Create the PR** - Go to the original repository and go to the "Pull Requests" tab. Click the "New Pull Request" button and select your new branch as the source branch.
Enter a descriptive title and description for your Pull Request and click "Create Pull Request".
## Reviewing a pull request
When a PR is created, it triggers a review process. The maintainers of the repository are notified of the PR and can review the changes that were made.
They can leave comments, request changes, or approve the changes.
If the maintainers request changes, you'll need to make those changes in your branch and push the changes to your forked repository.
The PR will be updated automatically with the new changes.
If the maintainers approve the changes, they can merge the PR into the repository.
## Closing a pull request
If you decide that you no longer want to merge a PR, you can close it.
To close a PR, go to the open PR and click the "Close Pull Request" button. This will close the PR without merging it.
## "Work In Progress" pull requests
Marking a pull request as being a work in progress will prevent that pull request from being accidentally merged.
To mark a pull request as being a work in progress, you must prefix its title by `WIP:` or `[WIP]` (case insensitive).
Those values are configurable in your `app.ini` file:
```ini
[repository.pull-request]
WORK_IN_PROGRESS_PREFIXES=WIP:,[WIP]
```
The first value of the list will be used in helpers.
## Default pull request title
When opening a new pull request, Gitea pre-fills the title field. The source of that title is controlled by the `DEFAULT_TITLE_SOURCE` setting in `app.ini`:
```ini
[repository.pull-request]
DEFAULT_TITLE_SOURCE = first-commit
```
Two modes are available:
- **`first-commit`** (default): The title is taken from the summary line of the oldest commit in the branch. This applies regardless of how many commits are included in the PR.
- **`auto`**: When the PR contains a single commit, its summary line is used as the title (same as `first-commit` for one commit). When the PR contains multiple commits, Gitea converts the branch name into a human-readable sentence: dashes, underscores, and `camelCase` word boundaries are replaced with spaces, and the first letter is capitalized.
Example: branch name `fix-user-login-flow` with multiple commits produces the title `Fix user login flow` under `auto`, but would use the oldest commit's message under `first-commit`.
## Pull Request Templates
You can find more information about pull request templates at the page [Issue and Pull Request templates](issue-pull-request-templates.md).
Now add the registry configuration to `/etc/pacman.conf`.
```conf
[{owner}.gitea.example.com]
[{repository}]
SigLevel = Required
Server = https://gitea.example.com/api/packages/{owner}/arch/{repository}/{architecture}
Server = https://gitea.example.com/api/packages/{owner}/arch/$repo/$arch
```
| Placeholder | Description |
| -------------- | ----------- |
| `owner` | The owner of the packages. |
| `repository` | The repository to use. |
| `architecture` | The architecture to use. |
Consult the owners package overview to see what `repository` and `architecture` is available.
If the registry is private, provide credentials in the url. You can use a password or a [personal access token](development/api-usage.md#authentication):
```
Server = https://{username}:{your_password_or_token}@gitea.example.com/api/packages/{owner}/arch/{repository}/{architecture}
[{repository}]
SigLevel = Required
Server = https://{username}:{your_password_or_token}@gitea.example.com/api/packages/{owner}/arch/$repo/$arch
**The following paragraphs only apply if Packages are not globally disabled!**
@@ -56,6 +57,8 @@ and shows a link to the repository on the package site (as well as a link to the
| **read** access | public, if user is public too; otherwise for this user only | public, if org is public, otherwise for org members only |
| **write** access | owner only | org members with admin or write access to the org |
Please note, that your respective package manager needs to have a [personal access token](development/api-usage.md#authentication) with the permissions `package` set to either Read or Read & Write to download and publish packages respectively.
N.B.: These access restrictions are [subject to change](https://github.com/go-gitea/gitea/issues/19270), where more finegrained control will be added via a dedicated organization team permission.
| `owner` | The owner of the state (user or organization). |
| `name` | The name of the state. |
| `username` | Your Gitea username. |
| `password` | Your Gitea password or [personal access token](development/api-usage.md#authentication). |
If you are using 2FA or OAuth use a [personal access token](development/api-usage.md#authentication) instead of the password.
## Initialize the backend
After configuring the backend, you can initialize it with:
```shell
terraform init
```
Terraform will prompt you to migrate your state if you already have one locally.
## State Locking
Gitea supports state locking to prevent concurrent modifications. The `lock_address` and `unlock_address` should point to the `/lock` sub-route of your state URL.
- **Locking**: Performed via a `POST` request to `{address}/lock`.
- **Unlocking**: Performed via a `DELETE` request to `{address}/lock`.
Terraform handles these requests automatically when configured as shown above.
## Encrypted state
The state registry supports [encrypted state](https://opentofu.org/docs/language/state/encryption/).
## State Versions and Management
Gitea keeps track of your Terraform state versions. You can use the API to retrieve or delete specific versions.
Repository mirroring allows for the mirroring of repositories to and from external sources. You can use it to mirror branches, tags, and commits between repositories.
## Use cases
The following are some possible use cases for repository mirroring:
- You migrated to Gitea but still need to keep your project in another source. In that case, you can simply set it up to mirror to Gitea (pull) and all the essential history of commits, tags, and branches are available in your Gitea instance.
- You have old projects in another source that you don’t use actively anymore, but don’t want to remove for archiving purposes. In that case, you can create a push mirror so that your active Gitea repository can push its changes to the old location.
## Pulling from a remote repository
For an existing remote repository, you can set up pull mirroring as follows:
1. Select **New Migration** in the **Create...** menu on the top right.
2. Select the remote repository service.
3. Enter a repository URL.
4. If the repository needs authentication fill in your authentication information.
5. Check the box **This repository will be a mirror**.
6. Select **Migrate repository** to save the configuration.
The repository now gets mirrored periodically from the remote repository. You can force a sync by selecting **Synchronize Now** in the repository settings.
:::warning
:exclamation::exclamation: You can only set up pull mirroring for repos that don't exist yet on your instance. Once the repo is created, you can't convert it into a pull mirror anymore. :exclamation::exclamation:
:::
## Pushing to a remote repository
For an existing repository, you can set up push mirroring as follows:
1. In your repository, go to **Settings** > **Repository**, and then the **Mirror Settings** section.
2. Enter a repository URL.
3. If the repository needs authentication expand the **Authorization** section and fill in your authentication information. Note that the requested **password** can also be your access token.
4. Select **Add Push Mirror** to save the configuration.
The repository now gets mirrored periodically to the remote repository. You can force a sync by selecting **Synchronize Now**. In case of an error a message displayed to help you resolve it.
:::warning
:exclamation::exclamation: This will force push to the remote repository. This will overwrite any changes in the remote repository! :exclamation::exclamation:
:::
### Setting up a push mirror from Gitea to GitHub
To set up a mirror from Gitea to GitHub, you need to follow these steps:
1. Create a [GitHub personal access token](https://docs.github.com/en/github/authenticating-to-github/creating-a-personal-access-token) with the *public_repo* box checked. Also check the **workflow** checkbox in case your repo uses GitHub Actions for continuous integration.
2. Create a repository with that name on GitHub. Unlike Gitea, GitHub does not support creating repositories by pushing to the remote. You can also use an existing remote repo if it has the same commit history as your Gitea repo.
3. In the settings of your Gitea repo, fill in the **Git Remote Repository URL**: `https://github.com/<your_github_group>/<your_github_project>.git`.
4. Fill in the **Authorization** fields with your GitHub username and the personal access token as **Password**.
5. (Optional, available on Gitea 1.18+) Select `Sync when new commits are pushed` so that the mirror will be updated as well as soon as there are changes. You can also disable the periodic sync if you like.
6. Select **Add Push Mirror** to save the configuration.
The repository pushes shortly thereafter. To force a push, select the **Synchronize Now** button.
### Setting up a push mirror from Gitea to GitLab
To set up a mirror from Gitea to GitLab, you need to follow these steps:
1. Create a [GitLab personal access token](https://docs.gitlab.com/ee/user/profile/personal_access_tokens.html) with *write_repository* scope.
2. Fill in the **Git Remote Repository URL**: `https://<destination host>/<your_gitlab_group_or_name>/<your_gitlab_project>.git`.
3. Fill in the **Authorization** fields with `oauth2` as **Username** and your GitLab personal access token as **Password**.
4. Select **Add Push Mirror** to save the configuration.
The repository pushes shortly thereafter. To force a push, select the **Synchronize Now** button.
### Setting up a push mirror from Gitea to Bitbucket
To set up a mirror from Gitea to Bitbucket, you need to follow these steps:
1. Create a [Bitbucket app password](https://support.atlassian.com/bitbucket-cloud/docs/app-passwords/) with the *Repository Write* box checked.
2. Fill in the **Git Remote Repository URL**: `https://bitbucket.org/<your_bitbucket_group_or_name>/<your_bitbucket_project>.git`.
3. Fill in the **Authorization** fields with your Bitbucket username and the app password as **Password**.
4. Select **Add Push Mirror** to save the configuration.
The repository pushes shortly thereafter. To force a push, select the **Synchronize Now** button.
### Mirror an existing ssh repository
Currently Gitea supports no ssh push mirrors. You can work around this by adding a `post-receive` hook to your Gitea repository that pushes manually.
1. Make sure the user running Gitea has access to the git repo you are trying to mirror to from shell.
2. On the web interface at the repository settings > git hooks add a post-receive hook for the mirror. I.e.
Gitea supports viewing the line-by-line revision history for a file also known as blame view.
You can also use [`git blame`](https://git-scm.com/docs/git-blame) on the command line to view the revision history of lines within a file.
1. Navigate to and open the file whose line history you want to view.
1. Click the `Blame` button in the file header bar.
1. The new view shows the line-by-line revision history for a file with author and commit information on the left side.
1. To navigate to an older commit, click the  icon.
## Ignore commits in the blame view
All revisions specified in the `.git-blame-ignore-revs` file are hidden from the blame view.
This is especially useful to hide reformatting changes and keep the benefits of `git blame`.
Lines that were changed or added by an ignored commit will be blamed on the previous commit that changed that line or nearby lines.
The `.git-blame-ignore-revs` file must be located in the root directory of the repository.
For more information like the file format, see [the `git blame --ignore-revs-file` documentation](https://git-scm.com/docs/git-blame#Documentation/git-blame.txt---ignore-revs-fileltfilegt).
### Bypassing `.git-blame-ignore-revs` in the blame view
If the blame view for a file shows a message about ignored revisions, you can see the normal blame view by appending the url parameter `?bypass-blame-ignore=true`.
Gitea supports the execution of several actions through incoming mails. This page describes how to set this up.
## Requirements
Handling incoming email messages requires an IMAP-enabled email account.
The recommended strategy is to use [email sub-addressing](https://en.wikipedia.org/wiki/Email_address#Sub-addressing) but a catch-all mailbox does work too.
The receiving email address contains a user/action specific token which tells Gitea which action should be performed.
This token is expected in the `To` and `Delivered-To` header fields.
Gitea tries to detect automatic responses to skip and the email server should be configured to reduce the incoming noise too (spam, newsletter).
## Configuration
To activate the handling of incoming email messages you have to configure the `email.incoming` section in the configuration file.
The `REPLY_TO_ADDRESS` contains the address an email client will respond to.
This address needs to contain the `%{token}` placeholder which will be replaced with a token describing the user/action.
This placeholder must only appear once in the address and must be in the user part of the address (before the `@`).
An example using email sub-addressing may look like this: `incoming+%{token}@example.com`
If a catch-all mailbox is used, the placeholder may be used anywhere in the user part of the address: `incoming+%{token}@example.com`, `incoming_%{token}@example.com`, `%{token}@example.com`
## Security
Be careful when choosing the domain used for receiving incoming email.
It's recommended receiving incoming email on a subdomain, such as `incoming.example.com` to prevent potential security problems with other services running on `example.com`.
Gitea uses MarkDown structured text in many places, you can recognise it by the `.md` file extension.
Markdown is plain text format for writing structured documents, allowing one to write
"plain" text files that nonetheless have _"fancy"_**formatting**, while still keeping the plain-text version readable.
Unfortunately, due to historical implementation differences, many subtle dialects exist.
To avoid adding to the confusion, Gitea tries to follow "[GitHub Flavoured Markdown (GFM)](https://help.github.com/articles/github-flavored-markdown/)" as close as possible.
"GFM" is an extension on top of the rigorously-specified [CommonMark](https://commonmark.org/) standard.
CommonMark flavours are used by most major web platforms (e.g. Reddit, Stack Overflow, Discourse)
and git forges (GitHub, GitLab and our very own Gitea), thus you shouldn't run into any surprises with Gitea.
For a quick introduction to the syntax and features of GitHub Flavoured Markdown, see the GitHub documentation:
Gitea supports GitHub-like math expression formatting.
### Inline expression
- ``` $\alpha$ ```: quoted by single-dollar `$`
- ``` $$\alpha$$ ```: quoted by double-dollar `$$`
- ``` $`\alpha`$ ```: quoted by dollar with backquotes, the backquotes could repeat like normal code blocks.
### Block expression
Using `$$`:
````
$$
\alpha
$$
````
Using code-block with language:
````
```math
\alpha
```
````
## Frontmatter
Gitea supports frontmatter and Table of Contents (TOC) rendering. By default, frontmatter rendering is enabled, and TOC rendering is disabled.
### Enabling TOC rendering
To enable TOC rendering for a markdown file, add `include_toc: true` to its frontmatter section.
### Disabling frontmatter
To disable frontmatter rendering for a markdown file, add `gitea: none` to its frontmatter section.
### Example
```yaml
---
include_toc: true
gitea: none
---
```
Then, when you preview this markdown file in Gitea, the frontmatter section will not be rendered, and a Table of Contents will be generated at the top of the content.
To migrate from GitLab to Gitea, you can use this non-affiliated tool:
https://github.com/loganinak/MigrateGitlabToGogs
## How to migrate from AWS CodeCommit to Gitea
- To use the AWS CodeCommit API, Gitea requires an access key ID and a secret access key. For security reasons, we recommend creating a new user with the minimum necessary permissions and generating an access key ID and secret access key for the migration. The minimum permissions required for this user are as follows:
- If you do not need to migrate pull requests, you can remove the `ListPullRequests`, `GetPullRequest`, and `GetCommentsForPullRequest` actions.
- For instructions on how to create an IAM user and assign permissions, you can refer to this [AWS documentation](https://docs.aws.amazon.com/IAM/latest/UserGuide/id_users_create.html).
- To clone this repository, Gitea requires HTTPS Git credentials. You can create HTTPS Git credentials according to this [AWS documentation](https://docs.aws.amazon.com/codecommit/latest/userguide/setting-up-gc.html).
To display a Markdown file in your Gitea user or organization profile page, create a repository named `.profile` and add a new file named `README.md` to it.
Gitea will automatically display the contents of the file on your profile, in a new "Overview" above your repositories.
Making the `.profile` repository private will hide the Profile README.
There are some additional features when pushing commits to Gitea server.
## Open PR through Push
When you push commits to a non-default branch for the first time,
you will receive a link you can click on to visit the compare page of your branch compared to your main branch.
From there, it's easy to create a pull request, even if you want to target another branch.

## Push Options
In Gitea `1.13`, support for some [push options](https://git-scm.com/docs/git-push#Documentation/git-push.txt--oltoptiongt)
were added.
### Supported Options
- `repo.private` (true|false) - Change the repository's visibility.
This is particularly useful when combined with push-to-create.
- `repo.template` (true|false) - Change whether the repository is a template.
Example of changing a repository's visibility to public:
```shell
git push -o repo.private=false -u origin main
```
## Push To Create
Push to create is a feature that allows you to push to a repository that does not exist yet in Gitea. This is useful for automation and for allowing users to create repositories without having to go through the web interface. This feature is disabled by default.
### Enabling Push To Create
In the `app.ini` file, set `ENABLE_PUSH_CREATE_USER` to `true` and `ENABLE_PUSH_CREATE_ORG` to `true` if you want to allow users to create repositories in their own user account and in organizations they are a member of respectively. Restart Gitea for the changes to take effect. You can read more about these two options in the [Configuration Cheat Sheet](../../administration/config-cheat-sheet.md#repository-repository).
### Using Push To Create
Assuming you have a git repository in the current directory, you can push to a repository that does not exist yet in Gitea by running the following command:
```shell
# Add the remote you want to push to
git remote add origin git@{domain}:{username}/{repo name that does not exist yet}.git
# push to the remote
git push -u origin main
```
This assumes you are using an SSH remote, but you can also use HTTPS remotes as well.
Push-to-create will default to the visibility defined by `DEFAULT_PUSH_CREATE_PRIVATE` in `app.ini`.
Repository mirroring allows for the mirroring of repositories to and from external sources. You can use it to mirror branches, tags, and commits between repositories.
## Use cases
The following are some possible use cases for repository mirroring:
- You migrated to Gitea but still need to keep your project in another source. In that case, you can simply set it up to mirror to Gitea (pull) and all the essential history of commits, tags, and branches are available in your Gitea instance.
- You have old projects in another source that you don’t use actively anymore, but don’t want to remove for archiving purposes. In that case, you can create a push mirror so that your active Gitea repository can push its changes to the old location.
## Pulling from a remote repository
For an existing remote repository, you can set up pull mirroring as follows:
1. Select **New Migration** in the **Create...** menu on the top right.
2. Select the remote repository service.
3. Enter a repository URL.
4. If the repository needs authentication fill in your authentication information.
5. Check the box **This repository will be a mirror**.
6. Select **Migrate repository** to save the configuration.
The repository now gets mirrored periodically from the remote repository. You can force a sync by selecting **Synchronize Now** in the repository settings.
:::warning
:exclamation::exclamation: You can only set up pull mirroring for repos that don't exist yet on your instance. Once the repo is created, you can't convert it into a pull mirror anymore. :exclamation::exclamation:
:::
## Pushing to a remote repository
For an existing repository, you can set up push mirroring as follows:
1. In your repository, go to **Settings** > **Repository**, and then the **Mirror Settings** section.
2. Enter a repository URL.
3. If the repository needs authentication expand the **Authorization** section and fill in your authentication information. Note that the requested **password** can also be your access token.
4. Select **Add Push Mirror** to save the configuration.
The repository now gets mirrored periodically to the remote repository. You can force a sync by selecting **Synchronize Now**. In case of an error a message displayed to help you resolve it.
:::warning
:exclamation::exclamation: This will force push to the remote repository. This will overwrite any changes in the remote repository! :exclamation::exclamation:
:::
### Setting up a push mirror from Gitea to GitHub
To set up a mirror from Gitea to GitHub, you need to follow these steps:
1. Create a [GitHub personal access token](https://docs.github.com/en/github/authenticating-to-github/creating-a-personal-access-token) with the *public_repo* box checked. Also check the **workflow** checkbox in case your repo uses GitHub Actions for continuous integration.
2. Create a repository with that name on GitHub. Unlike Gitea, GitHub does not support creating repositories by pushing to the remote. You can also use an existing remote repo if it has the same commit history as your Gitea repo.
3. In the settings of your Gitea repo, fill in the **Git Remote Repository URL**: `https://github.com/<your_github_group>/<your_github_project>.git`.
4. Fill in the **Authorization** fields with your GitHub username and the personal access token as **Password**.
5. (Optional, available on Gitea 1.18+) Select `Sync when new commits are pushed` so that the mirror will be updated as well as soon as there are changes. You can also disable the periodic sync if you like.
6. Select **Add Push Mirror** to save the configuration.
The repository pushes shortly thereafter. To force a push, select the **Synchronize Now** button.
### Setting up a push mirror from Gitea to GitLab
To set up a mirror from Gitea to GitLab, you need to follow these steps:
1. Create a [GitLab personal access token](https://docs.gitlab.com/ee/user/profile/personal_access_tokens.html) with *write_repository* scope.
2. Fill in the **Git Remote Repository URL**: `https://<destination host>/<your_gitlab_group_or_name>/<your_gitlab_project>.git`.
3. Fill in the **Authorization** fields with `oauth2` as **Username** and your GitLab personal access token as **Password**.
4. Select **Add Push Mirror** to save the configuration.
The repository pushes shortly thereafter. To force a push, select the **Synchronize Now** button.
### Setting up a push mirror from Gitea to Bitbucket
To set up a mirror from Gitea to Bitbucket, you need to follow these steps:
1. Create a [Bitbucket app password](https://support.atlassian.com/bitbucket-cloud/docs/app-passwords/) with the *Repository Write* box checked.
2. Fill in the **Git Remote Repository URL**: `https://bitbucket.org/<your_bitbucket_group_or_name>/<your_bitbucket_project>.git`.
3. Fill in the **Authorization** fields with your Bitbucket username and the app password as **Password**.
4. Select **Add Push Mirror** to save the configuration.
The repository pushes shortly thereafter. To force a push, select the **Synchronize Now** button.
### Mirror an existing ssh repository
Currently Gitea supports no ssh push mirrors. You can work around this by adding a `post-receive` hook to your Gitea repository that pushes manually.
1. Make sure the user running Gitea has access to the git repo you are trying to mirror to from shell.
2. On the web interface at the repository settings > git hooks add a post-receive hook for the mirror. I.e.
Since Gitea `1.11`, you can create template repositories.
When creating a repo based on a template, you can copy most of its content, and even auto-inject variables into it.
By default, variables will not be expanded in any file.
Only files contained in a pattern inside the `.gitea/template` file will be checked for if they contain variables.
When creating the template, all files are included except for this `.gitea/template` file.
Gitea uses [gobwas/glob](https://github.com/gobwas/glob) for its glob syntax.
It closely resembles a traditional `.gitignore`, however there may be slight differences.
## Example `.gitea/template` file
All paths are relative to the base of the repository
```gitignore
# Expand all .go files, anywhere in the repository
**.go
# All text files in the text directory
text/*.txt
# A specific file
a/b/c/d.json
# Batch files in both upper or lower case can be matched
**.[bB][aA][tT]
```
Gitea also supports to use sections to exclude some files:
```
[expand]
*.md
[exclude]
*.log
```
## Variable Expansion
In any file matched by the above globs, the variables below will be expanded.
Matching filenames and paths can also be expanded, and are conservatively sanitized to support cross-platform filesystems.
You can use variables by prefixing them with `$` or surrounding them with `${}`, so both `$VAR` and `${VAR}` insert the value of `VAR` at this location.
To escape an expansion, use `$$`, such as `$$VAR` or `$${VAR}`.
These are the variables Gitea currently recognizes:
| YEAR | The year of generating the repository (i.e. `2024`) | ✘ |
| MONTH | The month of generating the repository (i.e. `03`) | ✘ |
| MONTH_ENGLISH | The month but in English (i.e. `March`) | ✓ |
| DAY | The day of generating the repository (i.e. `02`) | ✘ |
| REPO_NAME | The name of the generated repository | ✓ |
| TEMPLATE_NAME | The name of the template repository | ✓ |
| REPO_DESCRIPTION | The description of the generated repository | ✘ |
| TEMPLATE_DESCRIPTION | The description of the template repository | ✘ |
| REPO_OWNER | The owner of the generated repository | ✓ |
| TEMPLATE_OWNER | The owner of the template repository | ✓ |
| REPO_LINK | The URL to the generated repository | ✘ |
| TEMPLATE_LINK | The URL to the template repository | ✘ |
| REPO_HTTPS_URL | The HTTP(S) clone link for the generated repository | ✘ |
| TEMPLATE_HTTPS_URL | The HTTP(S) clone link for the template repository | ✘ |
| REPO_SSH_URL | The SSH clone link for the generated repository | ✘ |
| TEMPLATE_SSH_URL | The SSH clone link for the template repository | ✘ |
## Transformers :robot:
Since Gitea `1.12.0`, variables marked as transformable in the table above also have alternative versions where the given transformer has been applied.
Transformed variables can be called by appending the transformer name to the variable name.
For example, to get `REPO_NAME` in `PASCAL`-case, you should use the variable `${REPO_NAME_PASCAL}`.
The following transformers are available (assuming `go-sdk` is the input):
with each webhook delivery. This is independent from the webhook secret:
- Use the secret to verify integrity with HMAC.
- Use the `Authorization` header when the receiving endpoint requires
application-level authentication.
## Delivery
This section describes how Gitea sends webhook deliveries and how receivers can
identify and verify them.
### Delivery behavior
- Webhooks are delivered asynchronously over HTTP.
- Generic `Gitea` and `Gogs` webhooks support `POST` and `GET`; `POST` is the
normal choice.
- For `POST` requests, the payload can be sent either as JSON
(`application/json`) or as a form field named `payload`
(`application/x-www-form-urlencoded`).
- Provider-specific integrations may use the HTTP method and body format
required by that provider.
### Delivery headers
Every delivery includes a unique delivery ID and event headers. For
GitHub-compatible integrations, Gitea also sends the corresponding GitHub and
Gogs header names.
| Header | Description |
| --- | --- |
| `X-Gitea-Delivery` | Unique delivery UUID for this attempt. |
| `X-Gitea-Event` | Normalized event name, such as `push`, `issues`, or `pull_request`. |
| `X-Gitea-Event-Type` | More specific event type, such as `issue_assign` or `pull_request_review_comment`. |
| `X-Gitea-Signature` | Hex-encoded HMAC-SHA256 of the raw request body, without a prefix. |
| `X-Gitea-Hook-Installation-Target-Type` | Where the webhook is defined: typically `repository`, `organization`, `user`, or `system`. Default webhooks are copied into repositories before delivery, so they are typically delivered as `repository`. |
| `X-Gogs-Delivery`, `X-Gogs-Event`, `X-Gogs-Event-Type`, `X-Gogs-Signature` | Compatibility headers with the same values as the Gitea variants. |
Multi-factor Authentication (also referred to as MFA or 2FA) enhances security by requiring a time-sensitive set of credentials in addition to a password.
If a password were later to be compromised, logging into Gitea will not be possible without the additional credentials and the account would remain secure.
Gitea supports both TOTP (Time-based One-Time Password) tokens and FIDO-based hardware keys using the Webauthn API.
MFA can be configured within the "Security" tab of the user settings page.
## MFA Considerations
Enabling MFA on a user does affect how the Git HTTP protocol can be used with the Git CLI.
This interface does not support MFA, and trying to use a password normally will no longer be possible whilst MFA is enabled.
If SSH is not an option for Git operations, an access token can be generated within the "Applications" tab of the user settings page.
This access token can be used as if it were a password in order to allow the Git CLI to function over HTTP.
:::warning
By its very nature, an access token sidesteps the security benefits of MFA.
It must be kept secure and should only be used as a last resort.
:::
The Gitea API supports providing the relevant TOTP password in the `X-Gitea-OTP` header, as described in [API Usage](development/api-usage.md).
This should be used instead of an access token where possible.
content:`<a href="https://about.gitea.com/products/cloud">Try Gitea Cloud ☁️ for 30 days <span aria-hidden="true">→</span> Accelerate your Development & Deploys!</a>`,
Loaded 100 of 3068 files, more files were not shown because too many files have changed in this diff.
Show more
Reference in new issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.