Remove StartTestNoRangeServer and its tests since the real bug is about
servers that don't return Content-Length at all (like GitHub's zip downloads),
not servers with Content-Length but no Range support.
The chunked tests (TestFetcher_DownloadChunkedContinue and
TestFetcher_DownloadChunkedNotExceedSize) correctly simulate the real-world
bug using StartTestCustomServer which serves files without Content-Length.
Co-authored-by: monkeyWie <[email protected]>
When HTTP server doesn't return Content-Length or doesn't support Range
requests, the download can accumulate extra bytes on pause/resume because:
1. Each retry/resume re-downloads from the beginning but writes at the
accumulated offset
2. No size limit enforcement for non-range downloads
3. onDownloadComplete incorrectly uses chunk.remain() for non-range
downloads, which gives negative values for chunk(0,0)
Fixes:
- Reset chunk state in downloadChunkOnce and runConnectionFallback for
non-range downloads since server will send from beginning
- Add size limit checks in download loops when Size is known
- Fix onDownloadComplete to use conn.Completed for non-range downloads
instead of unreliable chunk.remain()
- Add test cases for non-range download, pause/resume, and size limit
Co-authored-by: monkeyWie <[email protected]>
copilot-swe-agent[bot]andmonkeyWie
Levi
Matteo Pietro Dazzi