refactor: drop logrus and other direct dependencies (#1319)

After https://gitea.com/gitea/runner/pulls/1304, logrus stayed only to forward logs from docker/cli and moby/go-archive. The runner used a small part of those: volume parsing, the build context tar, the `ssh://` connection and registry credentials. Own code that matches them replaces them, so logrus leaves the module graph. docker/cli stays for its logrus-free `opts` and `connhelper/ssh` packages.

Other dependencies that the standard library or existing ones cover go too: mergo, httprouter (replaced by `http.ServeMux`), retry-go, go-isatty, go-selinux, container-device-interface, gotest.tools and go-cmp (replaced by testify). The binary shrinks by 5%.

Behaviour differences:

1. An unreadable file in a build context is skipped with a warning, where go-archive left a truncated tar.
2. Build contexts no longer carry `security.capability` xattrs.
3. The cache and artifact servers answer HEAD on GET routes, and no longer redirect trailing slashes or answer OPTIONS.

*Written by Claude.*

---------

Co-authored-by: bircni <[email protected]>
Reviewed-on: https://gitea.com/gitea/runner/pulls/1319
Reviewed-by: bircni <[email protected]>
Reviewed-by: Lunny Xiao <[email protected]>
Co-authored-by: silverwind <[email protected]>

internal/app/cmd/daemon.go

@@ -33,9 +33,8 @@ import (
 
 	"connectrpc.com/connect"
 	runnerv1 "gitea.dev/actionslib/runner/v1"
-	"github.com/avast/retry-go/v5"
-	"github.com/mattn/go-isatty"
 	"github.com/spf13/cobra"
+	"golang.org/x/term"
 )
 
 func runDaemon(ctx context.Context, daemArgs *daemonArgs, configFile *string) func(cmd *cobra.Command, args []string) error {
@@ -189,22 +188,22 @@ func runDaemon(ctx context.Context, daemArgs *daemonArgs, configFile *string) fu
 		}()
 
 		// declare the labels of the runner before fetching tasks
-		resp, err := retry.NewWithData[*connect.Response[runnerv1.DeclareResponse]](
-			retry.Context(ctx),
-			retry.UntilSucceeded(),
-			retry.Delay(declareRetryDelay),
-			retry.DelayType(retry.FixedDelay),
-			retry.RetryIf(shouldRetryDeclare),
-			retry.OnRetry(func(attempt uint, err error) {
-				if attempt == 0 {
-					logging.WithError(err).Warn("fail to invoke Declare, retrying until Gitea is reachable")
-				} else {
-					logging.WithError(err).Debug("fail to invoke Declare, retrying")
-				}
-			}),
-		).Do(func() (*connect.Response[runnerv1.DeclareResponse], error) {
-			return runner.Declare(ctx, ls.Names())
-		})
+		var resp *connect.Response[runnerv1.DeclareResponse]
+		for attempt := 0; ; attempt++ {
+			resp, err = runner.Declare(ctx, ls.Names())
+			if err == nil || !shouldRetryDeclare(err) || ctx.Err() != nil {
+				break
+			}
+			if attempt == 0 {
+				logging.WithError(err).Warn("fail to invoke Declare, retrying until Gitea is reachable")
+			} else {
+				logging.WithError(err).Debug("fail to invoke Declare, retrying")
+			}
+			select {
+			case <-time.After(declareRetryDelay):
+			case <-ctx.Done():
+			}
+		}
 		switch {
 		case err != nil && ctx.Err() != nil && !client.IsRegistrationRejected(err):
 			return nil
@@ -340,7 +339,7 @@ func initLogging(cfg *config.Config) {
 		return funcname, filename
 	}
 
-	isTerm := isatty.IsTerminal(os.Stdout.Fd())
+	isTerm := term.IsTerminal(int(os.Stdout.Fd()))
 	logging.SetTextFormat(logging.TextOptions{
 		DisableColors:    !isTerm,
 		FullTimestamp:    true,

internal/app/cmd/daemon_test.go

@@ -188,10 +188,3 @@ func TestShouldRetryDeclare(t *testing.T) {
 		require.Equal(t, tc.retry, shouldRetryDeclare(tc.err), tc.err.Error())
 	}
 }
-
-// A dependency that reconfigures logrus's global logger when it is loaded makes its own log lines
-// show up twice or not at all, so this fails when upgrading one starts doing that. It does not
-// cover a dependency that does so later, while running.
-func TestLogrusBridgeSurvivesDependencies(t *testing.T) {
-	require.True(t, logging.LogrusBridgeIntact())
-}

internal/app/cmd/register.go

@@ -24,15 +24,15 @@ import (
 	"connectrpc.com/connect"
 	pingv1 "gitea.dev/actionslib/ping/v1"
 	runnerv1 "gitea.dev/actionslib/runner/v1"
-	"github.com/mattn/go-isatty"
 	"github.com/spf13/cobra"
+	"golang.org/x/term"
 )
 
 // runRegister registers a runner to the server
 func runRegister(ctx context.Context, regArgs *registerArgs, configFile *string) func(*cobra.Command, []string) error {
 	return func(cmd *cobra.Command, args []string) error {
 		logging.SetReportCaller(false)
-		isTerm := isatty.IsTerminal(os.Stdout.Fd())
+		isTerm := term.IsTerminal(int(os.Stdout.Fd()))
 		logging.SetTextFormat(logging.TextOptions{
 			DisableColors:    !isTerm,
 			DisableTimestamp: true,

internal/app/cmd/register_test.go

@@ -14,7 +14,6 @@ import (
 	"gitea.com/gitea/runner/internal/pkg/config"
 
 	"github.com/stretchr/testify/require"
-	"gotest.tools/v3/assert"
 )
 
 func TestRegisterNonInteractiveReturnsLabelValidationError(t *testing.T) {
@@ -23,7 +22,7 @@ func TestRegisterNonInteractiveReturnsLabelValidationError(t *testing.T) {
 		Token:        "token",
 		InstanceAddr: "http://localhost:3000",
 	})
-	assert.ErrorContains(t, err, "empty label")
+	require.ErrorContains(t, err, "empty label")
 }
 
 func TestRegisterInputsValidate(t *testing.T) {
@@ -262,6 +261,6 @@ func TestDoRegisterReturnsPingErrorWithoutRetrying(t *testing.T) {
 	cfg := &config.Config{}
 	cfg.Runner.File = filepath.Join(t.TempDir(), ".runner")
 	err := doRegister(t.Context(), cfg, &registerInputs{InstanceAddr: server.URL, RunnerName: "runner"})
-	assert.ErrorContains(t, err, "cannot ping the Gitea instance server")
-	assert.Equal(t, int32(1), requests.Load())
+	require.ErrorContains(t, err, "cannot ping the Gitea instance server")
+	require.Equal(t, int32(1), requests.Load())
 }

internal/app/run/workflow_test.go

@@ -10,7 +10,6 @@ import (
 	runnerv1 "gitea.dev/actionslib/runner/v1"
 	"github.com/stretchr/testify/require"
 	"go.yaml.in/yaml/v4"
-	"gotest.tools/v3/assert"
 )
 
 func Test_generateWorkflow(t *testing.T) {
@@ -58,7 +57,7 @@ jobs:
 				},
 			},
 			assert: func(t *testing.T, wf *model.Workflow) {
-				assert.DeepEqual(t, wf.GetJob("job9").Needs(), []string{"job1", "job2"})
+				require.Equal(t, []string{"job1", "job2"}, wf.GetJob("job9").Needs())
 			},
 			want1:   "job9",
 			wantErr: false,
@@ -83,8 +82,8 @@ jobs:
 			},
 			assert: func(t *testing.T, wf *model.Workflow) {
 				job := wf.GetJob("test")
-				assert.DeepEqual(t, job.Needs(), []string{})
-				assert.Equal(t, len(job.Steps), 2)
+				require.Equal(t, []string{}, job.Needs())
+				require.Len(t, job.Steps, 2)
 			},
 			want1:   "test",
 			wantErr: false,
@@ -125,9 +124,9 @@ jobs:
 			assert: func(t *testing.T, wf *model.Workflow) {
 				job := wf.GetJob("deploy")
 				needs := job.Needs()
-				assert.DeepEqual(t, needs, []string{"build", "lint", "test"})
-				assert.Equal(t, wf.Jobs["test"].Outputs["coverage"], "80%")
-				assert.Equal(t, wf.Jobs["lint"].Result, "failure")
+				require.Equal(t, []string{"build", "lint", "test"}, needs)
+				require.Equal(t, "80%", wf.Jobs["test"].Outputs["coverage"])
+				require.Equal(t, "failure", wf.Jobs["lint"].Result)
 			},
 			want1:   "deploy",
 			wantErr: false,
@@ -165,11 +164,11 @@ jobs:
 				},
 			},
 			assert: func(t *testing.T, wf *model.Workflow) {
-				assert.Equal(t, wf.Name, "Complex workflow")
-				assert.Equal(t, wf.Env["NODE_ENV"], "production")
-				assert.Equal(t, wf.Env["CI"], "true")
+				require.Equal(t, "Complex workflow", wf.Name)
+				require.Equal(t, "production", wf.Env["NODE_ENV"])
+				require.Equal(t, "true", wf.Env["CI"])
 				job := wf.GetJob("build")
-				assert.Equal(t, len(job.Steps), 4)
+				require.Len(t, job.Steps, 4)
 			},
 			want1:   "build",
 			wantErr: false,
@@ -200,8 +199,8 @@ jobs:
 			assert: func(t *testing.T, wf *model.Workflow) {
 				job := wf.GetJob("integration")
 				container := job.Container()
-				assert.Equal(t, container.Image, "node:18")
-				assert.Equal(t, job.Services["postgres"].Image, "postgres:15")
+				require.Equal(t, "node:18", container.Image)
+				require.Equal(t, "postgres:15", job.Services["postgres"].Image)
 			},
 			want1:   "integration",
 			wantErr: false,
@@ -231,7 +230,7 @@ jobs:
 				job := wf.GetJob("test")
 				matrixes, err := job.GetMatrixes()
 				require.NoError(t, err)
-				assert.Equal(t, len(matrixes), 2)
+				require.Len(t, matrixes, 2)
 			},
 			want1:   "test",
 			wantErr: false,
@@ -245,9 +244,9 @@ jobs:
 				},
 			},
 			assert: func(t *testing.T, wf *model.Workflow) {
-				assert.Equal(t, wf.Name, "Special: characters & test")
+				require.Equal(t, "Special: characters & test", wf.Name)
 				job := wf.GetJob("test")
-				assert.Equal(t, len(job.Steps), 3)
+				require.Len(t, job.Steps, 3)
 			},
 			want1:   "test",
 			wantErr: false,
@@ -283,7 +282,7 @@ jobs:
 			}
 			require.NoError(t, err)
 			tt.assert(t, got)
-			assert.Equal(t, got1, tt.want1)
+			require.Equal(t, tt.want1, got1)
 		})
 	}
 }
@@ -301,7 +300,7 @@ func Test_yamlV4NodeRoundTrip(t *testing.T) {
 
 		out, err := yaml.Marshal(&node)
 		require.NoError(t, err)
-		assert.Equal(t, string(out), "- a\n- b\n- c\n")
+		require.Equal(t, "- a\n- b\n- c\n", string(out))
 	})
 
 	t.Run("unmarshal and re-marshal workflow", func(t *testing.T) {
@@ -310,7 +309,7 @@ func Test_yamlV4NodeRoundTrip(t *testing.T) {
 		var wf map[string]any
 		err := yaml.Unmarshal(input, &wf)
 		require.NoError(t, err)
-		assert.Equal(t, wf["name"], "test")
+		require.Equal(t, "test", wf["name"])
 
 		out, err := yaml.Marshal(wf)
 		require.NoError(t, err)
@@ -318,7 +317,7 @@ func Test_yamlV4NodeRoundTrip(t *testing.T) {
 		var wf2 map[string]any
 		err = yaml.Unmarshal(out, &wf2)
 		require.NoError(t, err)
-		assert.Equal(t, wf2["name"], "test")
+		require.Equal(t, "test", wf2["name"])
 	})
 
 	t.Run("node kind constants", func(t *testing.T) {

internal/pkg/labels/labels_test.go

@@ -7,7 +7,6 @@ import (
 	"testing"
 
 	"github.com/stretchr/testify/require"
-	"gotest.tools/v3/assert"
 )
 
 func TestParse(t *testing.T) {
@@ -77,7 +76,7 @@ func TestParse(t *testing.T) {
 				return
 			}
 			require.NoError(t, err)
-			assert.DeepEqual(t, got, tt.want)
+			require.Equal(t, tt.want, got)
 		})
 	}
 }

internal/pkg/report/reporter.go

@@ -8,6 +8,7 @@ import (
 	"context"
 	"errors"
 	"fmt"
+	"math/rand/v2"
 	"regexp"
 	"slices"
 	"strings"
@@ -24,7 +25,6 @@ import (
 	"connectrpc.com/connect"
 	"gitea.dev/actionslib/pkg/model"
 	runnerv1 "gitea.dev/actionslib/runner/v1"
-	"github.com/avast/retry-go/v5"
 	"google.golang.org/protobuf/proto"
 	"google.golang.org/protobuf/types/known/timestamppb"
 )
@@ -777,7 +777,23 @@ func (r *Reporter) Close(lastWords string) error {
 func (r *Reporter) flushFinal(fn func() error) error {
 	ctx, cancel := context.WithTimeout(context.Background(), 3*r.effectiveCloseTimeout())
 	defer cancel()
-	return retry.New(retry.Context(ctx)).Do(fn)
+	var errs []error
+	for attempt := range 10 {
+		err := fn()
+		if err == nil {
+			return nil
+		}
+		errs = append(errs, err)
+		if attempt == 9 {
+			break
+		}
+		select {
+		case <-time.After(100*time.Millisecond<<attempt + rand.N(100*time.Millisecond)):
+		case <-ctx.Done():
+			return errors.Join(append(errs, ctx.Err())...)
+		}
+	}
+	return errors.Join(errs...)
 }
 
 // effectiveCloseTimeout returns closeTimeout, or 10s when unset, so a zero

internal/pkg/report/reporter_test.go

@@ -15,6 +15,7 @@ import (
 	"sync"
 	"sync/atomic"
 	"testing"
+	"testing/synctest"
 	"time"
 
 	"gitea.com/gitea/runner/act/common/logging"
@@ -1161,6 +1162,19 @@ func TestReporter_EffectiveCloseTimeout(t *testing.T) {
 	assert.Equal(t, 5*time.Second, (&Reporter{closeTimeout: 5 * time.Second}).effectiveCloseTimeout())
 }
 
+func TestReporter_FlushFinalRetriesUntilSuccess(t *testing.T) {
+	synctest.Test(t, func(t *testing.T) {
+		calls := 0
+		require.NoError(t, (&Reporter{}).flushFinal(func() error {
+			if calls++; calls < 3 {
+				return errors.New("unavailable")
+			}
+			return nil
+		}))
+		assert.Equal(t, 3, calls)
+	})
+}
+
 func TestReporter_ParseResult(t *testing.T) {
 	r := &Reporter{}
 

internal/pkg/telemetry/telemetry.go

@@ -12,6 +12,7 @@ import (
 	"fmt"
 	"io"
 	"math"
+	"math/rand/v2"
 	"net/http"
 	"net/url"
 	"os"
@@ -26,7 +27,6 @@ import (
 
 	"gitea.dev/actionslib/pkg/model"
 	runnerv1 "gitea.dev/actionslib/runner/v1"
-	"github.com/avast/retry-go/v5"
 	"github.com/docker/go-connections/tlsconfig"
 	"github.com/prometheus/client_golang/prometheus"
 	"go.opentelemetry.io/otel/attribute"
@@ -226,10 +226,10 @@ func (c *client) upload(ctx context.Context, data proto.Message) error {
 	}
 	ctx, cancel := context.WithTimeout(ctx, c.timeout)
 	defer cancel()
-	return retry.New(retry.Context(ctx), retry.Attempts(5), retry.LastErrorOnly(true), retry.DelayType(retryDelay)).Do(func() error {
+	for attempt := 1; ; attempt++ {
 		req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.endpoint, bytes.NewReader(body))
 		if err != nil {
-			return retry.Unrecoverable(err)
+			return err
 		}
 		req.Header = c.headers.Clone()
 		req.Header.Set("Content-Type", "application/x-protobuf")
@@ -237,33 +237,39 @@ func (c *client) upload(ctx context.Context, data proto.Message) error {
 			req.Header.Set("Content-Encoding", "gzip")
 		}
 		resp, err := c.httpClient.Do(req)
-		if err != nil {
+		if err == nil {
+			_, _ = io.Copy(io.Discard, resp.Body)
+			_ = resp.Body.Close()
+			if resp.StatusCode < http.StatusMultipleChoices {
+				return nil
+			}
+			err = &statusError{resp.Status, resp.Header.Get("Retry-After")}
+			if !slices.Contains([]int{http.StatusTooManyRequests, http.StatusBadGateway, http.StatusServiceUnavailable, http.StatusGatewayTimeout}, resp.StatusCode) {
+				return err
+			}
+		}
+		if attempt == 5 {
 			return err
 		}
-		_, _ = io.Copy(io.Discard, resp.Body)
-		_ = resp.Body.Close()
-		if resp.StatusCode < http.StatusMultipleChoices {
-			return nil
+		select {
+		case <-time.After(retryDelay(attempt, err)):
+		case <-ctx.Done():
+			return ctx.Err()
 		}
-		err = &statusError{resp.Status, resp.Header.Get("Retry-After")}
-		if slices.Contains([]int{http.StatusTooManyRequests, http.StatusBadGateway, http.StatusServiceUnavailable, http.StatusGatewayTimeout}, resp.StatusCode) {
-			return err
-		}
-		return retry.Unrecoverable(err)
-	})
+	}
 }
 
 type statusError struct{ status, retryAfter string }
 
 func (e *statusError) Error() string { return "OTLP export failed: " + e.status }
 
-func retryDelay(attempt uint, err error, config retry.DelayContext) time.Duration {
+func retryDelay(attempt int, err error) time.Duration {
 	if status := (*statusError)(nil); errors.As(err, &status) {
 		if seconds, err := strconv.ParseUint(status.retryAfter, 10, 31); err == nil {
 			return time.Duration(seconds) * time.Second
 		}
 	}
-	return retry.CombineDelay(retry.BackOffDelay, retry.RandomDelay)(attempt, err, config)
+	return 100*time.Millisecond<<(attempt-1) + rand.N(100*time.Millisecond)
 }
 
 func StartJob(ctx context.Context, task *runnerv1.Task) (context.Context, func(runnerv1.Result)) {

internal/pkg/telemetry/telemetry_test.go

@@ -152,7 +152,7 @@ func TestExportRetriesAndMatchesOTLP(t *testing.T) {
 	<-traceBodies
 	var tracesData tracepb.TracesData
 	require.NoError(t, proto.Unmarshal(gunzip(<-traceBodies), &tracesData))
-	assert.Equal(t, 2*time.Second, retryDelay(1, &statusError{"429 Too Many Requests", "2"}, nil))
+	assert.Equal(t, 2*time.Second, retryDelay(1, &statusError{"429 Too Many Requests", "2"}))
 	assert.Equal(t, "op", tracesData.GetResourceSpans()[0].GetScopeSpans()[0].GetSpans()[0].GetName())
 
 	resourceMetrics := metricsData.GetResourceMetrics()[0]