fix: job container, daemon startup, action cache and exec fixes (#1294)

- Keep `GITHUB_ENV`, `GITHUB_OUTPUT`, `GITHUB_STATE` and `GITHUB_PATH` when a job times out or is cancelled, so post steps still get their state
- Retry the startup Declare while Gitea is unreachable instead of exiting, still failing fast on rejected registrations, auth and misconfiguration
- Reuse a local image when a platform variant such as `linux/arm/v7` is requested instead of pulling again
- Read `PATH` from the created job container so cross-architecture jobs get tool paths for the right architecture
- Extract action copies relative to the act volume so they never cross an image's `/var/run` symlink, which rootless Podman rejects
- Keep the job container's entrypoint when container options set `--entrypoint`, warning instead of appending it to the keep-alive command
- Lock action cache entries across runner processes that share a cache directory
- `exec`: omit `GITHUB_SERVER_URL` and `GITHUB_API_URL` without `--gitea-instance` instead of exporting a bare `https://`, and read owner and repository from any origin remote

Reviewed-on: https://gitea.com/gitea/runner/pulls/1294
Reviewed-by: bircni <bircni@icloud.com>
Co-authored-by: silverwind <me@silverwind.io>

act/common/git/git.go

@@ -10,8 +10,10 @@ import (
 	"encoding/base64"
 	"errors"
 	"fmt"
+	"net/url"
 	"os"
 	"os/exec"
+	"path"
 	"path/filepath"
 	"regexp"
 	"runtime"
@@ -29,8 +31,6 @@ import (
 var (
 	codeCommitHTTPRegex = regexp.MustCompile(`^https?://git-codecommit\.(.+)\.amazonaws.com/v1/repos/(.+)$`)
 	codeCommitSSHRegex  = regexp.MustCompile(`ssh://git-codecommit\.(.+)\.amazonaws.com/v1/repos/(.+)$`)
-	githubHTTPRegex     = regexp.MustCompile(`^https?://.*github.com.*/(.+)/(.+?)(?:.git)?$`)
-	githubSSHRegex      = regexp.MustCompile(`github.com[:/](.+)/(.+?)(?:.git)?$`)
 	hexRefRegex         = regexp.MustCompile(`^[0-9a-fA-F]+$`)
 
 	LocalEnvVars = []string{ // `git rev-parse --local-env-vars`, inherited ones point git at another repository
@@ -44,12 +44,29 @@ var (
 	ErrShortRef = errors.New("short SHA references are not supported")
 )
 
-// AcquireCloneLock returns an unlock function after locking the per-directory mutex for dir.
-// Only concurrent operations targeting the same directory are serialized; clones into different directories run in parallel.
-// Callers reading files inside dir (e.g. tarring a checked-out action into a job container) must hold this lock too,
-// otherwise a concurrent NewGitCloneExecutor on the same dir can mutate the worktree mid-read.
-func AcquireCloneLock(dir string) func() {
-	return cloneLocks.Lock(dir)
+// AcquireCloneLock serializes access to dir across runner processes, or within this one where no lock file can be created. Readers of dir must hold it too.
+func AcquireCloneLock(ctx context.Context, dir string) (func(), error) {
+	dir = filepath.Clean(dir)
+	if runtime.GOOS == "plan9" {
+		return cloneLocks.Lock(dir), nil
+	}
+	if err := os.MkdirAll(filepath.Dir(dir), 0o755); err != nil {
+		return nil, err
+	}
+	for {
+		release, err := lock.TryLock(dir)
+		if err == nil {
+			return func() { _ = release() }, nil
+		}
+		if !errors.Is(err, lock.ErrLocked) {
+			return cloneLocks.Lock(dir), nil
+		}
+		select {
+		case <-ctx.Done():
+			return nil, ctx.Err()
+		case <-time.After(50 * time.Millisecond):
+		}
+	}
 }
 
 type Error struct {
@@ -109,13 +126,12 @@ func FindGitRef(ctx context.Context, file string) (string, error) {
 }
 
 // FindGithubRepo get the repo
-func FindGithubRepo(ctx context.Context, file, githubInstance string) (string, error) {
+func FindGithubRepo(ctx context.Context, file string) (string, error) {
 	url, err := findGitRemoteURL(ctx, file)
 	if err != nil {
 		return "", err
 	}
-	_, slug := findGitSlug(url, githubInstance)
-	return slug, nil
+	return findGitSlug(url)
 }
 
 func findGitRemoteURL(ctx context.Context, file string) (string, error) {
@@ -130,25 +146,25 @@ func findGitRemoteURL(ctx context.Context, file string) (string, error) {
 	return url, nil
 }
 
-func findGitSlug(url, githubInstance string) (string, string) {
-	if matches := codeCommitHTTPRegex.FindStringSubmatch(url); matches != nil {
-		return "CodeCommit", matches[2]
-	} else if matches := codeCommitSSHRegex.FindStringSubmatch(url); matches != nil {
-		return "CodeCommit", matches[2]
-	} else if matches := githubHTTPRegex.FindStringSubmatch(url); matches != nil {
-		return "GitHub", fmt.Sprintf("%s/%s", matches[1], matches[2])
-	} else if matches := githubSSHRegex.FindStringSubmatch(url); matches != nil {
-		return "GitHub", fmt.Sprintf("%s/%s", matches[1], matches[2])
-	} else if githubInstance != "github.com" {
-		gheHTTPRegex := regexp.MustCompile(fmt.Sprintf(`^https?://%s/(.+)/(.+?)(?:.git)?$`, githubInstance))
-		gheSSHRegex := regexp.MustCompile(githubInstance + "[:/](.+)/(.+?)(?:.git)?$")
-		if matches := gheHTTPRegex.FindStringSubmatch(url); matches != nil {
-			return "GitHubEnterprise", fmt.Sprintf("%s/%s", matches[1], matches[2])
-		} else if matches := gheSSHRegex.FindStringSubmatch(url); matches != nil {
-			return "GitHubEnterprise", fmt.Sprintf("%s/%s", matches[1], matches[2])
-		}
+func findGitSlug(remoteURL string) (string, error) {
+	if matches := codeCommitHTTPRegex.FindStringSubmatch(remoteURL); matches != nil {
+		return matches[2], nil
+	} else if matches := codeCommitSSHRegex.FindStringSubmatch(remoteURL); matches != nil {
+		return matches[2], nil
 	}
-	return "", url
+	if host, repoPath, ok := strings.Cut(remoteURL, ":"); ok && filepath.VolumeName(remoteURL) == "" && !strings.Contains(host, "/") && !strings.HasPrefix(repoPath, "//") {
+		remoteURL = "ssh://" + host + "/" + repoPath // scp-like [user@]host:path
+	}
+	parsed, err := url.Parse(remoteURL)
+	if err != nil {
+		return "", errors.New("cannot parse the origin remote URL")
+	}
+	owner, repo := path.Split(strings.TrimSuffix(strings.Trim(parsed.Path, "/"), ".git"))
+	owner = path.Base(owner)
+	if parsed.Host == "" || repo == "" || owner == "." || owner == "/" {
+		return "", errors.New("cannot determine owner and repository from the origin remote URL")
+	}
+	return owner + "/" + repo, nil
 }
 
 // NewGitCloneExecutorInput the input for the NewGitCloneExecutor
@@ -182,6 +198,9 @@ func CloneIfRequired(ctx context.Context, input NewGitCloneExecutorInput, logger
 	if err == nil && origin == input.URL {
 		return true, nil
 	}
+	if ctxErr := ctx.Err(); ctxErr != nil { // a cancelled origin lookup must not discard a valid clone
+		return false, ctxErr
+	}
 	if err == nil {
 		logger.Debugf("Removing cached clone at %s because origin URL changed from %s to %s", input.Dir, origin, input.URL)
 	} else if _, statErr := os.Stat(input.Dir); statErr == nil {
@@ -209,7 +228,11 @@ func NewGitCloneExecutor(input NewGitCloneExecutorInput) common.Executor {
 			logger.Infof("git clone '%s' # ref=%s", input.URL, input.Ref)
 		}
 		logger.Debugf("  cloning %s to %s", input.URL, input.Dir)
-		defer AcquireCloneLock(input.Dir)()
+		unlock, err := AcquireCloneLock(ctx, input.Dir)
+		if err != nil {
+			return err
+		}
+		defer unlock()
 
 		reused, err := CloneIfRequired(ctx, input, logger)
 		if err != nil {

act/common/git/git_test.go

@@ -18,9 +18,11 @@ import (
 	"sync"
 	"syscall"
 	"testing"
+	"testing/synctest"
 	"time"
 
 	"gitea.com/gitea/runner/act/common"
+	"gitea.com/gitea/runner/internal/pkg/lock"
 
 	log "github.com/sirupsen/logrus"
 	logrustest "github.com/sirupsen/logrus/hooks/test"
@@ -29,29 +31,33 @@ import (
 )
 
 func TestFindGitSlug(t *testing.T) {
-	assert := assert.New(t)
-
-	slugTests := []struct {
-		url      string // input
-		provider string // expected result
-		slug     string // expected result
+	for _, testcase := range []struct {
+		remoteURL string
+		slug      string
 	}{
-		{"https://git-codecommit.us-east-1.amazonaws.com/v1/repos/my-repo-name", "CodeCommit", "my-repo-name"},
-		{"ssh://git-codecommit.us-west-2.amazonaws.com/v1/repos/my-repo", "CodeCommit", "my-repo"},
-		{"git@github.com:nektos/act.git", "GitHub", "nektos/act"},
-		{"git@github.com:nektos/act", "GitHub", "nektos/act"},
-		{"https://github.com/nektos/act.git", "GitHub", "nektos/act"},
-		{"http://github.com/nektos/act.git", "GitHub", "nektos/act"},
-		{"https://github.com/nektos/act", "GitHub", "nektos/act"},
-		{"http://github.com/nektos/act", "GitHub", "nektos/act"},
-		{"git+ssh://git@github.com/owner/repo.git", "GitHub", "owner/repo"},
-		{"http://myotherrepo.com/act.git", "", "http://myotherrepo.com/act.git"},
-	}
-
-	for _, tt := range slugTests {
-		provider, slug := findGitSlug(tt.url, "github.com")
-		assert.Equal(tt.provider, provider)
-		assert.Equal(tt.slug, slug)
+		{"https://git-codecommit.us-east-1.amazonaws.com/v1/repos/my-repo-name", "my-repo-name"},
+		{"ssh://git-codecommit.us-west-2.amazonaws.com/v1/repos/my-repo", "my-repo"},
+		{"git@github.com:nektos/act.git", "nektos/act"},
+		{"git@github.com:nektos/act", "nektos/act"},
+		{"https://github.com/nektos/act.git", "nektos/act"},
+		{"http://github.com/nektos/act.git", "nektos/act"},
+		{"https://github.com/nektos/act", "nektos/act"},
+		{"http://github.com/nektos/act", "nektos/act"},
+		{"git+ssh://git@github.com/owner/repo.git", "owner/repo"},
+		{"https://example.com:3000/gitea/owner/repo.git/", "owner/repo"},
+		{"ssh://git@example.com:2222/gitea/owner/repo.git", "owner/repo"},
+		{"example.com:owner/repo.git", "owner/repo"},
+		{"http://myotherrepo.com/act.git", ""},
+		{"https://user:secret@example.com/act.git", ""},
+		{"/tmp/owner/repo.git", ""},
+		{"./owner/repo:branch", ""},
+	} {
+		t.Run(testcase.remoteURL, func(t *testing.T) {
+			slug, err := findGitSlug(testcase.remoteURL)
+			assert.Equal(t, testcase.slug == "", err != nil)
+			assert.NotContains(t, fmt.Sprint(err), "secret")
+			assert.Equal(t, testcase.slug, slug)
+		})
 	}
 }
 
@@ -107,7 +113,7 @@ func TestFindGitMetadataOfSHA256RepositoryUsesOrigin(t *testing.T) {
 	require.NoError(t, err)
 	require.Equal(t, remoteURL, url)
 
-	slug, err := FindGithubRepo(context.Background(), basedir, "github.com")
+	slug, err := FindGithubRepo(context.Background(), basedir)
 	require.NoError(t, err)
 	require.Equal(t, "owner/repo", slug)
 }
@@ -310,6 +316,12 @@ func TestGitCloneExecutorReclonesWhenOriginURLChanges(t *testing.T) {
 	markerPath := filepath.Join(cacheDir, "stale-marker")
 	require.NoError(t, os.WriteFile(markerPath, []byte("stale"), 0o644))
 
+	cancelled, cancel := context.WithCancel(t.Context())
+	cancel()
+	_, err := CloneIfRequired(cancelled, NewGitCloneExecutorInput{URL: oldRemoteDir, Ref: "main", Dir: cacheDir}, log.New())
+	require.ErrorIs(t, err, context.Canceled)
+	require.FileExists(t, markerPath)
+
 	require.NoError(t, NewGitCloneExecutor(NewGitCloneExecutorInput{
 		URL: newRemoteDir,
 		Ref: "main",
@@ -631,14 +643,42 @@ func gitCmd(args ...string) error {
 }
 
 func TestAcquireCloneLock(t *testing.T) {
+	t.Run("blocks on a lock held by another process despite a trailing separator", func(t *testing.T) {
+		dir := filepath.Join(t.TempDir(), "cache")
+		release, err := lock.TryLock(dir)
+		require.NoError(t, err)
+		defer func() { require.NoError(t, release()) }()
+		synctest.Test(t, func(t *testing.T) {
+			ctx, cancel := context.WithTimeout(t.Context(), time.Second)
+			defer cancel()
+			_, err := AcquireCloneLock(ctx, dir+string(filepath.Separator))
+			require.ErrorIs(t, err, context.DeadlineExceeded)
+		})
+	})
+
+	t.Run("read-only parent falls back to an in-process lock", func(t *testing.T) {
+		parent := t.TempDir()
+		dir := filepath.Join(parent, "cache")
+		require.NoError(t, os.Mkdir(dir, 0o755))
+		require.NoError(t, os.Chmod(parent, 0o555))
+		defer func() { require.NoError(t, os.Chmod(parent, 0o755)) }()
+		unlock, err := AcquireCloneLock(t.Context(), dir)
+		require.NoError(t, err)
+		unlock()
+	})
+
 	t.Run("same directory serializes", func(t *testing.T) {
 		dir := t.TempDir()
 
-		unlock1 := AcquireCloneLock(dir)
+		unlock1, err := AcquireCloneLock(t.Context(), dir)
+		require.NoError(t, err)
 
 		secondAcquired := make(chan struct{})
 		go func() {
-			unlock := AcquireCloneLock(dir)
+			unlock, err := AcquireCloneLock(t.Context(), dir)
+			if !assert.NoError(t, err) {
+				return
+			}
 			close(secondAcquired)
 			unlock()
 		}()
@@ -662,12 +702,16 @@ func TestAcquireCloneLock(t *testing.T) {
 		dirA := t.TempDir()
 		dirB := t.TempDir()
 
-		unlockA := AcquireCloneLock(dirA)
+		unlockA, err := AcquireCloneLock(t.Context(), dirA)
+		require.NoError(t, err)
 		defer unlockA()
 
 		done := make(chan struct{})
 		go func() {
-			unlock := AcquireCloneLock(dirB)
+			unlock, err := AcquireCloneLock(t.Context(), dirB)
+			if !assert.NoError(t, err) {
+				return
+			}
 			unlock()
 			close(done)
 		}()

act/container/docker_images.go

@@ -8,10 +8,10 @@ package container
 
 import (
 	"context"
-	"fmt"
 
 	cerrdefs "github.com/containerd/errdefs"
 	"github.com/moby/moby/client"
+	specs "github.com/opencontainers/image-spec/specs-go/v1"
 )
 
 // ImageExistsLocally returns a boolean indicating if an image with the
@@ -30,11 +30,23 @@ func ImageExistsLocally(ctx context.Context, imageName, platform string) (bool,
 		return false, err
 	}
 
-	if platform == "" || platform == "any" || fmt.Sprintf("%s/%s", inspectImage.Os, inspectImage.Architecture) == platform {
+	if platform == "" || platform == "any" {
 		return true, nil
 	}
 
-	return false, nil
+	requested, err := parsePlatform(platform)
+	if err != nil {
+		return false, err
+	}
+	return platformMatches(requested, &specs.Platform{OS: inspectImage.Os, Architecture: inspectImage.Architecture, Variant: inspectImage.Variant}), nil
+}
+
+func platformMatches(requested, image *specs.Platform) bool {
+	variant := image.Variant
+	if variant == "" {
+		variant = map[string]string{"amd64": "v1", "arm": "v7", "arm64": "v8"}[image.Architecture]
+	}
+	return image.OS == requested.OS && image.Architecture == requested.Architecture && (requested.Variant == "" || requested.Variant == variant)
 }
 
 // RemoveImage removes image from local store, the function is used to run different

act/container/docker_platform_test.go

@@ -61,3 +61,26 @@ func TestParsePlatform(t *testing.T) {
 		})
 	}
 }
+
+func TestPlatformMatches(t *testing.T) {
+	for _, testcase := range []struct {
+		requested, image string
+		want             bool
+	}{
+		{"linux/amd64", "linux/amd64", true},
+		{"linux/amd64/v1", "linux/amd64", true},
+		{"linux/arm64", "linux/amd64", false},
+		{"linux/arm64/v8", "linux/arm64", true},
+		{"linux/arm/v7", "linux/arm/v7", true},
+		{"linux/arm", "linux/arm/v7", true},
+		{"linux/arm/v6", "linux/arm/v7", false},
+		{"linux/arm", "linux/arm/v6", true},
+		{"windows/arm/v7", "linux/arm/v7", false},
+	} {
+		requested, err := parsePlatform(testcase.requested)
+		require.NoError(t, err)
+		image, err := parsePlatform(testcase.image)
+		require.NoError(t, err)
+		assert.Equal(t, testcase.want, platformMatches(requested, image), "%s on %s", testcase.requested, testcase.image)
+	}
+}

act/container/docker_run.go

@@ -31,7 +31,7 @@ import (
 	cerrdefs "github.com/containerd/errdefs"
 	"github.com/docker/cli/cli/compose/loader"
 	"github.com/docker/cli/cli/connhelper"
-	"github.com/joho/godotenv"
+	"github.com/docker/cli/opts"
 	"github.com/moby/moby/api/pkg/stdcopy"
 	"github.com/moby/moby/api/types/container"
 	"github.com/moby/moby/api/types/mount"
@@ -95,8 +95,8 @@ func (cr *containerReference) Start(attach bool) common.Executor {
 				cr.attach().IfBool(attach),
 				cr.start(),
 				cr.wait().IfBool(attach),
-				cr.tryReadUID(),
-				cr.tryReadGID(),
+				cr.tryReadID("-u", func(id int) { cr.UID = id }),
+				cr.tryReadID("-g", func(id int) { cr.GID = id }),
 				func(ctx context.Context) error {
 					// If this fails, then folders have wrong permissions on non root container
 					if cr.UID != 0 || cr.GID != 0 {
@@ -305,7 +305,25 @@ func (cr *containerReference) UpdateFromEnv(srcPath string, env *map[string]stri
 }
 
 func (cr *containerReference) UpdateFromImageEnv(env *map[string]string) common.Executor {
-	return cr.extractFromImageEnv(env).IfNot(common.Dryrun)
+	envMap := *env
+	return common.Executor(func(ctx context.Context) error {
+		inspect, err := cr.cli.ContainerInspect(ctx, cr.id, client.ContainerInspectOptions{})
+		if err != nil {
+			common.Logger(ctx).Error(err)
+			return fmt.Errorf("inspect container: %w", err)
+		}
+		if inspect.Container.Config == nil {
+			return nil
+		}
+		for key, value := range opts.ConvertKVStringsToMap(inspect.Container.Config.Env) {
+			if envMap[key] == "" {
+				envMap[key] = value
+			} else if key == "PATH" {
+				envMap[key] += ":" + value
+			}
+		}
+		return nil
+	}).IfNot(common.Dryrun)
 }
 
 func (cr *containerReference) Exec(command []string, env map[string]string, user, workdir string) common.Executor {
@@ -567,20 +585,6 @@ func (cr *containerReference) mergeContainerConfigs(ctx context.Context, config
 		return nil, nil, fmt.Errorf("cannot process container options: '%s': '%w'", options, err)
 	}
 
-	// FIXME: If everything is fine after gitea/act v0.260.0, remove the following comment.
-	// In the old fork version, the code is
-	// if len(copts.netMode.Value()) == 0 {
-	// 	if err = copts.netMode.Set("host"); err != nil {
-	// 		return nil, nil, fmt.Errorf("cannot parse networkmode=host. This is an internal error and should not happen: '%w'", err)
-	// 	}
-	// }
-	// And it has been commented with:
-	//   If a service container's network is set to `host`, the container will not be able to
-	//   connect to the specified network created for the job container and the service containers.
-	//   So comment out the following code.
-	// Not the if it's necessary to comment it in the new version,
-	// since it's cr.input.NetworkMode now.
-
 	if len(copts.netMode.Value()) == 0 {
 		if err = copts.netMode.Set(cr.input.NetworkMode); err != nil {
 			return nil, nil, fmt.Errorf("cannot parse networkmode=%s. This is an internal error and should not happen: '%w'", cr.input.NetworkMode, err)
@@ -644,6 +648,9 @@ func (cr *containerReference) mergeContainerConfigs(ctx context.Context, config
 	if flags.Changed("network") || flags.Changed("net") {
 		logger.Warn("--network and --net in the options will be ignored.")
 	}
+	if len(cr.input.Entrypoint) != 0 && flags.Changed("entrypoint") {
+		logger.Warn("--entrypoint in the options will be ignored.")
+	}
 	hostConfig.NetworkMode = networkMode
 	logger.Debugf("Merged container.HostConfig ==> %+v", hostConfig)
 
@@ -673,17 +680,11 @@ func (cr *containerReference) create(capAdd, capDrop []string) common.Executor {
 			ExposedPorts: exposedPorts,
 			Tty:          input.AllocatePTY,
 		}
-		// For Gitea, reduce log noise
-		// logger.Debugf("Common container.Config ==> %+v", config)
 
 		if len(input.Cmd) != 0 {
 			config.Cmd = input.Cmd
 		}
 
-		if len(input.Entrypoint) != 0 {
-			config.Entrypoint = input.Entrypoint
-		}
-
 		mounts := make([]mount.Mount, 0)
 		for mountSource, mountTarget := range input.Mounts {
 			mounts = append(mounts, mount.Mount{
@@ -718,20 +719,20 @@ func (cr *containerReference) create(capAdd, capDrop []string) common.Executor {
 			PortBindings: portBindings,
 			AutoRemove:   input.AutoRemove,
 		}
-		// For Gitea, reduce log noise
-		// logger.Debugf("Common container.HostConfig ==> %+v", hostConfig)
 
 		config, hostConfig, err = cr.mergeContainerConfigs(ctx, config, hostConfig)
 		if err != nil {
 			return err
 		}
 
+		if len(input.Entrypoint) != 0 {
+			config.Entrypoint = input.Entrypoint
+		}
+
 		// For Gitea
 		config, hostConfig = cr.sanitizeConfig(ctx, config, hostConfig)
 
 		var networkingConfig *network.NetworkingConfig
-		// For Gitea, reduce log noise
-		// logger.Debugf("input.NetworkAliases ==> %v", input.NetworkAliases)
 		n := hostConfig.NetworkMode
 		// IsUserDefined and IsHost are broken on windows
 		if n.IsUserDefined() && n != "host" && len(input.NetworkAliases) > 0 {
@@ -764,44 +765,6 @@ func (cr *containerReference) create(capAdd, capDrop []string) common.Executor {
 	}
 }
 
-func (cr *containerReference) extractFromImageEnv(env *map[string]string) common.Executor {
-	envMap := *env
-	return func(ctx context.Context) error {
-		logger := common.Logger(ctx)
-
-		inspect, err := cr.cli.ImageInspect(ctx, cr.input.Image)
-		if err != nil {
-			logger.Error(err)
-			return fmt.Errorf("inspect image: %w", err)
-		}
-
-		if inspect.Config == nil {
-			return nil
-		}
-
-		imageEnv, err := godotenv.Unmarshal(strings.Join(inspect.Config.Env, "\n"))
-		if err != nil {
-			logger.Error(err)
-			return fmt.Errorf("unmarshal image env: %w", err)
-		}
-
-		for k, v := range imageEnv {
-			if k == "PATH" {
-				if envMap[k] == "" {
-					envMap[k] = v
-				} else {
-					envMap[k] += `:` + v
-				}
-			} else if envMap[k] == "" {
-				envMap[k] = v
-			}
-		}
-
-		env = &envMap
-		return nil
-	}
-}
-
 func (cr *containerReference) exec(cmd []string, env map[string]string, user, workdir string) common.Executor {
 	return func(ctx context.Context) error {
 		if cr.id == "" {
@@ -819,11 +782,6 @@ func (cr *containerReference) exec(cmd []string, env map[string]string, user, wo
 
 		logger.Debugf("Exec command '%s'", cmd)
 		isTerminal := cr.input.AllocatePTY
-		envList := make([]string, 0)
-		for k, v := range env {
-			envList = append(envList, fmt.Sprintf("%s=%s", k, v))
-		}
-
 		var wd string
 		if workdir != "" {
 			if strings.HasPrefix(workdir, "/") {
@@ -840,7 +798,7 @@ func (cr *containerReference) exec(cmd []string, env map[string]string, user, wo
 			User:         user,
 			Cmd:          cmd,
 			WorkingDir:   wd,
-			Env:          envList,
+			Env:          getEnvListFromMap(env),
 			TTY:          isTerminal,
 			AttachStderr: true,
 			AttachStdout: true,
@@ -857,7 +815,7 @@ func (cr *containerReference) exec(cmd []string, env map[string]string, user, wo
 		}
 		defer resp.Close()
 
-		err = cr.waitForCommand(ctx, resp.HijackedResponse, idResp, user, workdir)
+		err = cr.waitForCommand(ctx, resp.HijackedResponse)
 		if err != nil {
 			return err
 		}
@@ -907,15 +865,7 @@ func (cr *containerReference) tryReadID(opt string, cbk func(id int)) common.Exe
 	}
 }
 
-func (cr *containerReference) tryReadUID() common.Executor {
-	return cr.tryReadID("-u", func(id int) { cr.UID = id })
-}
-
-func (cr *containerReference) tryReadGID() common.Executor {
-	return cr.tryReadID("-g", func(id int) { cr.GID = id })
-}
-
-func (cr *containerReference) waitForCommand(ctx context.Context, resp client.HijackedResponse, _ client.ExecCreateResult, _, _ string) error {
+func (cr *containerReference) waitForCommand(ctx context.Context, resp client.HijackedResponse) error {
 	logger := common.Logger(ctx)
 
 	// Buffered so the copy goroutine never blocks on send if the grace-period
@@ -957,6 +907,7 @@ func (cr *containerReference) waitForCommand(ctx context.Context, resp client.Hi
 }
 
 func (cr *containerReference) copyDir(dstPath, srcPath string, useGitIgnore, skipGitDir bool) common.Executor {
+	dstPath = path.Clean(dstPath)
 	return func(ctx context.Context) error {
 		if cr.id == "" {
 			return cr.missingContainerError("copy directory to %s", dstPath)
@@ -978,7 +929,11 @@ func (cr *containerReference) copyDir(dstPath, srcPath string, useGitIgnore, ski
 				logger.Error(err)
 			}
 		}(tarFile)
-		if err := writeDirTar(ctx, tarFile, dstPath, srcPath, useGitIgnore, skipGitDir, cr.UID, cr.GID); err != nil {
+		extractionRoot := "/"
+		if actPath := cr.GetActPath(); strings.HasPrefix(dstPath+"/", actPath+"/") {
+			extractionRoot = actPath // keeps tar entries from crossing the image's /var/run symlink
+		}
+		if err := writeDirTar(ctx, tarFile, path.Join("/", strings.TrimPrefix(dstPath, extractionRoot)), srcPath, useGitIgnore, skipGitDir, cr.UID, cr.GID); err != nil {
 			return err
 		}
 
@@ -987,14 +942,7 @@ func (cr *containerReference) copyDir(dstPath, srcPath string, useGitIgnore, ski
 		if err != nil {
 			return fmt.Errorf("failed to seek tar archive: %w", err)
 		}
-		_, err = cr.cli.CopyToContainer(ctx, cr.id, client.CopyToContainerOptions{
-			DestinationPath: "/",
-			Content:         tarFile,
-		})
-		if err != nil {
-			return fmt.Errorf("failed to copy content to container: %w", err)
-		}
-		return nil
+		return cr.upload(ctx, extractionRoot, tarFile)
 	}
 }
 
@@ -1009,17 +957,17 @@ func (cr *containerReference) copyContent(dstPath string, files ...*FileEntry) c
 		}
 
 		common.Logger(ctx).Debugf("Extracting content to '%s'", dstPath)
-		_, err := cr.cli.CopyToContainer(ctx, cr.id, client.CopyToContainerOptions{
-			DestinationPath: dstPath,
-			Content:         &buf,
-		})
-		if err != nil {
-			return fmt.Errorf("failed to copy content to container: %w", err)
-		}
-		return nil
+		return cr.upload(ctx, dstPath, &buf)
 	}
 }
 
+func (cr *containerReference) upload(ctx context.Context, dstPath string, content io.Reader) error {
+	if _, err := cr.cli.CopyToContainer(ctx, cr.id, client.CopyToContainerOptions{DestinationPath: dstPath, Content: content}); err != nil {
+		return fmt.Errorf("failed to copy content to container: %w", err)
+	}
+	return nil
+}
+
 func (cr *containerReference) attach() common.Executor {
 	return func(ctx context.Context) error {
 		out, err := cr.cli.ContainerAttach(ctx, cr.id, client.ContainerAttachOptions{

act/container/docker_run_test.go

@@ -5,6 +5,7 @@
 package container
 
 import (
+	"archive/tar"
 	"bufio"
 	"bytes"
 	"context"
@@ -31,44 +32,28 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
-func TestDocker(t *testing.T) {
-	requireDocker(t)
-	ctx := context.Background()
-	client, err := GetDockerClient(ctx)
-	require.NoError(t, err)
-	defer client.Close()
-
-	dockerBuild := NewDockerBuildExecutor(NewDockerBuildExecutorInput{
-		ContextDir: "testdata",
-		ImageTag:   "envmergetest",
-	})
-
-	err = dockerBuild(ctx)
-	assert.NoError(t, err) //nolint:testifylint // pre-existing issue from nektos/act
-
-	cr := &containerReference{
-		cli: client,
-		input: &NewContainerInput{
-			Image: "envmergetest",
-		},
-	}
+func TestDockerEnvironmentUsesContainerPath(t *testing.T) {
+	client := &mockDockerClient{}
+	client.On("ContainerInspect", mock.Anything, "amd64-job", mobyclient.ContainerInspectOptions{}).
+		Return(mobyclient.ContainerInspectResult{Container: container.InspectResponse{Config: &container.Config{Env: []string{
+			"PATH=/opt/node/x64/bin:/usr/bin:/bin", "CONFLICT_VAR=container-value", "SOME_RANDOM_VAR=",
+			"ANOTHER_ONE=old-value", "ANOTHER_ONE=BUT_I_HAVE_VALUE", "LITERAL='quoted'=$HOME\nsecond line",
+		}}}}, nil)
 	env := map[string]string{
-		"PATH":         "/usr/local/bin:/usr/bin:/usr/sbin:/bin:/sbin",
+		"PATH":         "/custom/bin",
 		"RANDOM_VAR":   "WITH_VALUE",
 		"ANOTHER_VAR":  "",
-		"CONFLICT_VAR": "I_EXIST_IN_MULTIPLE_PLACES",
+		"CONFLICT_VAR": "step-value",
 	}
-
-	envExecutor := cr.extractFromImageEnv(&env)
-	err = envExecutor(ctx)
-	assert.NoError(t, err) //nolint:testifylint // pre-existing issue from nektos/act
+	require.NoError(t, (&containerReference{cli: client, id: "amd64-job"}).UpdateFromImageEnv(&env)(t.Context()))
 	assert.Equal(t, map[string]string{
-		"PATH":            "/usr/local/bin:/usr/bin:/usr/sbin:/bin:/sbin:/this/path/does/not/exists/anywhere:/this/either",
+		"PATH":            "/custom/bin:/opt/node/x64/bin:/usr/bin:/bin",
 		"RANDOM_VAR":      "WITH_VALUE",
 		"ANOTHER_VAR":     "",
 		"SOME_RANDOM_VAR": "",
 		"ANOTHER_ONE":     "BUT_I_HAVE_VALUE",
-		"CONFLICT_VAR":    "I_EXIST_IN_MULTIPLE_PLACES",
+		"CONFLICT_VAR":    "step-value",
+		"LITERAL":         "'quoted'=$HOME\nsecond line",
 	}, env)
 }
 
@@ -498,25 +483,45 @@ func TestRejectsMissingContainer(t *testing.T) {
 	check("Inspect after removal", err)
 }
 
-// End-to-end: a stale cr.id is cleared, repopulated from name lookup,
-// and the Copy completes against the fresh id.
 func TestPublicCopyPipelineHandlesStaleID(t *testing.T) {
-	ctx := context.Background()
-	client := &mockDockerClient{}
-	client.On("ContainerInspect", ctx, "stale", mobyclient.ContainerInspectOptions{}).
-		Return(mobyclient.ContainerInspectResult{}, cerrdefs.ErrNotFound.WithMessage("gone"))
-	client.On("ContainerList", ctx, mobyclient.ContainerListOptions{All: true}).
-		Return(mobyclient.ContainerListResult{Items: []container.Summary{
-			{ID: "fresh", Names: []string{"/job-1"}},
-		}}, nil)
-	client.On("CopyToContainer", ctx, "fresh", mock.MatchedBy(func(opts mobyclient.CopyToContainerOptions) bool {
-		return opts.DestinationPath == "/var/run/act"
-	})).Return(mobyclient.CopyToContainerResult{}, nil)
+	for _, testcase := range []struct {
+		name, destination, extractionRoot, entry string
+		directory                                bool
+	}{
+		{name: "files", destination: "/var/run/act", extractionRoot: "/var/run/act", entry: "x"},
+		{name: "action below act volume", directory: true, destination: "/var/run/act/actions/checkout/", extractionRoot: "/var/run/act", entry: "actions/checkout/x"},
+		{name: "sibling prefix", directory: true, destination: "/var/run/actor/action", extractionRoot: "/", entry: "var/run/actor/action/x"},
+	} {
+		t.Run(testcase.name, func(t *testing.T) {
+			ctx := t.Context()
+			client := &mockDockerClient{}
+			client.On("ContainerInspect", ctx, "stale", mobyclient.ContainerInspectOptions{}).
+				Return(mobyclient.ContainerInspectResult{}, cerrdefs.ErrNotFound.WithMessage("gone"))
+			client.On("ContainerList", ctx, mobyclient.ContainerListOptions{All: true}).
+				Return(mobyclient.ContainerListResult{Items: []container.Summary{
+					{ID: "fresh", Names: []string{"/job-1"}},
+				}}, nil)
+			client.On("CopyToContainer", ctx, "fresh", mock.Anything).
+				Run(func(args mock.Arguments) {
+					opts := args.Get(2).(mobyclient.CopyToContainerOptions)
+					assert.Equal(t, testcase.extractionRoot, opts.DestinationPath)
+					header, err := tar.NewReader(opts.Content).Next()
+					require.NoError(t, err)
+					assert.Equal(t, testcase.entry, header.Name)
+				}).Return(mobyclient.CopyToContainerResult{}, nil).Once()
 
-	cr := &containerReference{id: "stale", cli: client, input: &NewContainerInput{Name: "job-1"}}
-	require.NoError(t, cr.Copy("/var/run/act", &FileEntry{Name: "x", Mode: 0o644})(ctx))
-	assert.Equal(t, "fresh", cr.id)
-	client.AssertExpectations(t)
+			cr := &containerReference{id: "stale", cli: client, input: &NewContainerInput{Name: "job-1"}}
+			if testcase.directory {
+				source := t.TempDir()
+				require.NoError(t, os.WriteFile(filepath.Join(source, "x"), nil, 0o644))
+				require.NoError(t, cr.CopyDir(testcase.destination, source+string(filepath.Separator), false, true)(ctx))
+			} else {
+				require.NoError(t, cr.Copy(testcase.destination, &FileEntry{Name: "x", Mode: 0o644})(ctx))
+			}
+			assert.Equal(t, "fresh", cr.id)
+			client.AssertExpectations(t)
+		})
+	}
 }
 
 // Type assert containerReference implements ExecutionsEnvironment
@@ -923,6 +928,35 @@ func TestMergeContainerConfigsWarnsOnlyAboutOptionsThatWereGiven(t *testing.T) {
 	assert.Equal(t, 1, warnings("", "--privileged"))
 }
 
+func TestDockerCreateKeepsInputEntrypoint(t *testing.T) {
+	for _, entrypoint := range [][]string{nil, {"sleep", "3600"}} {
+		for _, optionEntrypoint := range []string{"/sbin/dumb-init", ""} {
+			logger, hook := test.NewNullLogger()
+			cr := &containerReference{
+				input: &NewContainerInput{
+					NetworkMode:     "bridge",
+					Entrypoint:      entrypoint,
+					RunnerOptions:   "--entrypoint=runner-entrypoint",
+					WorkflowOptions: "--entrypoint=" + optionEntrypoint + " --init",
+				},
+				cli: &probeClient{create: func(opts mobyclient.ContainerCreateOptions) (mobyclient.ContainerCreateResult, error) {
+					if len(entrypoint) > 0 {
+						assert.Equal(t, entrypoint, opts.Config.Entrypoint)
+					} else {
+						assert.Equal(t, []string{optionEntrypoint}, opts.Config.Entrypoint)
+					}
+					require.NotNil(t, opts.HostConfig.Init)
+					assert.True(t, *opts.HostConfig.Init)
+					return mobyclient.ContainerCreateResult{ID: "created"}, nil
+				}},
+			}
+			require.NoError(t, cr.create(nil, nil)(common.WithLogger(t.Context(), logger)))
+			assert.Equal(t, "created", cr.id)
+			assert.Equal(t, len(entrypoint) > 0, len(hook.AllEntries()) == 1)
+		}
+	}
+}
+
 func TestMergeContainerConfigsKeepsNetworkAliasesFromOptions(t *testing.T) {
 	logger, _ := test.NewNullLogger()
 	cr := &containerReference{input: &NewContainerInput{

act/container/testdata/Dockerfile

@@ -1,5 +0,0 @@
-FROM scratch
-ENV PATH="/this/path/does/not/exists/anywhere:/this/either"
-ENV SOME_RANDOM_VAR=""
-ENV ANOTHER_ONE="BUT_I_HAVE_VALUE"
-ENV CONFLICT_VAR="I_EXIST_ONLY_HERE"

act/ghcontext/github_context.go

@@ -130,11 +130,11 @@ func SetSha(ctx context.Context, ghc *model.GithubContext, repoPath string) {
 
 // SetRepositoryAndOwner resolves the repository of the context from the git
 // remote in repoPath when it is not set yet, and derives its owner.
-func SetRepositoryAndOwner(ctx context.Context, ghc *model.GithubContext, githubInstance, repoPath string) {
+func SetRepositoryAndOwner(ctx context.Context, ghc *model.GithubContext, repoPath string) {
 	if ghc.Repository == "" {
-		repo, err := findGithubRepo(ctx, repoPath, githubInstance)
+		repo, err := findGithubRepo(ctx, repoPath)
 		if err != nil {
-			common.Logger(ctx).Warningf("unable to get git repo (githubInstance: %v, repoPath: %v): %v", githubInstance, repoPath, err)
+			common.Logger(ctx).Warningf("unable to get git repo (repoPath: %v): %v", repoPath, err)
 			return
 		}
 		ghc.Repository = repo

act/runner/action.go

@@ -145,7 +145,11 @@ func maybeCopyToActionDir(ctx context.Context, step actionStep, actionDir, actio
 		containerActionDirCopy += `/`
 	}
 
-	defer git.AcquireCloneLock(actionDir)()
+	unlock, err := git.AcquireCloneLock(ctx, actionDir)
+	if err != nil {
+		return err
+	}
+	defer unlock()
 
 	if !rc.Config.NoActionPatch {
 		// A concurrent job's prepare resets this directory, so patch under the copy's lock.
@@ -366,7 +370,11 @@ func execAsDocker(ctx context.Context, step actionStep, actionName, actionDir, b
 				// Held across the whole build: the daemon drains contextDir lazily.
 				inner := prepImage
 				prepImage = func(ctx context.Context) error {
-					defer git.AcquireCloneLock(actionDir)()
+					unlock, err := git.AcquireCloneLock(ctx, actionDir)
+					if err != nil {
+						return err
+					}
+					defer unlock()
 					return inner(ctx)
 				}
 			}

act/runner/action_test.go

@@ -143,6 +143,7 @@ runs:
 }
 
 func TestActionRunner(t *testing.T) {
+	t.Chdir(t.TempDir())
 	table := []struct {
 		name        string
 		step        actionStep
@@ -366,7 +367,10 @@ func TestMaybeCopyToActionDirHoldsCloneLock(t *testing.T) {
 
 	peerAcquired := make(chan struct{})
 	go func() {
-		unlock := git.AcquireCloneLock(actionDir)
+		unlock, err := git.AcquireCloneLock(t.Context(), actionDir)
+		if !assert.NoError(t, err) {
+			return
+		}
 		close(peerAcquired)
 		unlock()
 	}()
@@ -400,7 +404,9 @@ func TestMaybeCopyToActionDirHoldsCloneLock(t *testing.T) {
 func TestExecAsDockerHoldsCloneLockForRemoteUncached(t *testing.T) {
 	actionDir := t.TempDir()
 
-	unlockOnce := sync.OnceFunc(git.AcquireCloneLock(actionDir))
+	unlock, err := git.AcquireCloneLock(t.Context(), actionDir)
+	require.NoError(t, err)
+	unlockOnce := sync.OnceFunc(unlock)
 	defer unlockOnce()
 
 	innerEntered := make(chan struct{})

act/runner/job_executor.go

@@ -29,7 +29,10 @@ import (
 	"gitea.dev/actionslib/pkg/model"
 )
 
-const maxJobSummaryBytes = 1024 * 1024
+const (
+	jobCleanupTimeout  = time.Minute
+	maxJobSummaryBytes = 1024 * 1024
+)
 
 // jobSummaryTruncationMarker is appended to a summary that exceeded the size limit
 // so the rendered output makes the truncation visible instead of silently cutting off.
@@ -249,7 +252,7 @@ func newJobExecutor(info jobInfo, sf stepFactory, rc *RunContext) common.Executo
 		jobError := common.JobError(ctx)
 		var err error
 		// always allow 1 min for stopping and removing the runner, even if we were cancelled
-		ctx, cancel := context.WithTimeout(common.WithLogger(context.Background(), common.Logger(ctx)), time.Minute)
+		ctx, cancel := context.WithTimeout(common.WithLogger(context.Background(), common.Logger(ctx)), jobCleanupTimeout)
 		defer cancel()
 
 		logger := common.Logger(ctx)

act/runner/reusable_workflow.go

@@ -10,7 +10,6 @@ import (
 	"net/url"
 	"path"
 	"regexp"
-	"strings"
 
 	"gitea.com/gitea/runner/act/common"
 	"gitea.com/gitea/runner/act/common/git"
@@ -27,7 +26,7 @@ func newLocalReusableWorkflowExecutor(rc *RunContext) common.Executor {
 	if !rc.Config.NoSkipCheckout {
 		// resolve the local workflow against the workspace root, not the process
 		// working directory, so it is found regardless of where the runner is invoked
-		return newReusableWorkflowExecutor(rc, rc.Config.Workdir, localWorkflow.Path)
+		return newReusableWorkflowExecutor(rc, rc.Config.Workdir, localWorkflow.Path, false)
 	}
 
 	uses := fmt.Sprintf("%s/%s@%s", rc.Config.PresetGitHubContext.Repository, localWorkflow.Path, rc.Config.PresetGitHubContext.Sha)
@@ -41,7 +40,7 @@ func newLocalReusableWorkflowExecutor(rc *RunContext) common.Executor {
 
 	return common.NewPipelineExecutor(
 		cloneRemoteReusableWorkflow(rc, remoteReusableWorkflow.CloneURL(), remoteReusableWorkflow.Ref, workflowDir),
-		newReusableWorkflowExecutor(rc, workflowDir, remoteReusableWorkflow.Path),
+		newReusableWorkflowExecutor(rc, workflowDir, remoteReusableWorkflow.Path, true),
 	)
 }
 
@@ -59,7 +58,7 @@ func newRemoteReusableWorkflowExecutor(rc *RunContext) common.Executor {
 
 	return common.NewPipelineExecutor(
 		cloneRemoteReusableWorkflow(rc, remoteReusableWorkflow.CloneURL(), remoteReusableWorkflow.Ref, workflowDir),
-		newReusableWorkflowExecutor(rc, workflowDir, remoteReusableWorkflow.Path),
+		newReusableWorkflowExecutor(rc, workflowDir, remoteReusableWorkflow.Path, true),
 	)
 }
 
@@ -87,11 +86,16 @@ func cloneRemoteReusableWorkflow(rc *RunContext, cloneURL, ref, targetDirectory
 	}
 }
 
-func newReusableWorkflowExecutor(rc *RunContext, directory, workflow string) common.Executor {
+func newReusableWorkflowExecutor(rc *RunContext, directory, workflow string, cached bool) common.Executor {
 	return func(ctx context.Context) error {
-		// Serialize workflow reads with cache updates.
 		planner, err := func() (model.WorkflowPlanner, error) {
-			defer git.AcquireCloneLock(directory)()
+			if cached {
+				unlock, err := git.AcquireCloneLock(ctx, directory)
+				if err != nil {
+					return nil, err
+				}
+				defer unlock()
+			}
 			return model.NewWorkflowPlanner(path.Join(directory, workflow), true)
 		}()
 		if err != nil {
@@ -135,11 +139,7 @@ type remoteReusableWorkflow struct {
 }
 
 func (r *remoteReusableWorkflow) CloneURL() string {
-	// In Gitea, r.URL always has the protocol prefix, we don't need to add extra prefix in this case.
-	if strings.HasPrefix(r.URL, "http://") || strings.HasPrefix(r.URL, "https://") {
-		return fmt.Sprintf("%s/%s/%s", r.URL, r.Owner, r.Repo)
-	}
-	return fmt.Sprintf("https://%s/%s/%s", r.URL, r.Owner, r.Repo)
+	return fmt.Sprintf("%s/%s/%s", withHTTPS(r.URL), r.Owner, r.Repo)
 }
 
 var absoluteReusableWorkflowURLRegex = regexp.MustCompile(`^(https?://.*)/([^/]+/[^/]+/\.[^/]+/workflows/[^@]+@.*)$`)
@@ -233,10 +233,7 @@ func onInstanceHost(host string, instanceURLs ...string) bool {
 		if candidate == "" {
 			continue
 		}
-		if !strings.HasPrefix(candidate, "http://") && !strings.HasPrefix(candidate, "https://") {
-			candidate = "https://" + candidate
-		}
-		if u, err := url.Parse(candidate); err == nil && u.Host == host {
+		if u, err := url.Parse(withHTTPS(candidate)); err == nil && u.Host == host {
 			return true
 		}
 	}

act/runner/reusable_workflow_test.go

@@ -8,7 +8,6 @@ import (
 	"crypto/tls"
 	"crypto/x509"
 	"encoding/pem"
-	"io/fs"
 	"net/http"
 	"net/http/httptest"
 	"os"
@@ -84,14 +83,16 @@ func TestNewReusableWorkflowExecutorHoldsCloneLock(t *testing.T) {
 	workflowDir := t.TempDir()
 	require.NoError(t, os.WriteFile(filepath.Join(workflowDir, "reusable.yml"), []byte(":"), 0o644))
 
-	unlockOnce := sync.OnceFunc(git.AcquireCloneLock(workflowDir))
+	unlock, err := git.AcquireCloneLock(t.Context(), workflowDir)
+	require.NoError(t, err)
+	unlockOnce := sync.OnceFunc(unlock)
 	defer unlockOnce()
 
 	rc := &RunContext{
 		Config: &Config{},
 		Run:    &model.Run{Workflow: &model.Workflow{Jobs: map[string]*model.Job{}}},
 	}
-	exec := newReusableWorkflowExecutor(rc, workflowDir, "reusable.yml")
+	exec := newReusableWorkflowExecutor(rc, workflowDir, "reusable.yml", true)
 
 	done := make(chan error, 1)
 	go func() { done <- exec(context.Background()) }()
@@ -113,8 +114,9 @@ func TestNewReusableWorkflowExecutorHoldsCloneLock(t *testing.T) {
 }
 
 func TestNewLocalReusableWorkflowExecutorFindsSameRepositoryPaths(t *testing.T) {
-	workdir := t.TempDir()
+	workdir := filepath.Join(t.TempDir(), "workspace")
 	require.NoError(t, os.MkdirAll(filepath.Join(workdir, ".gitea", "workflows"), 0o755))
+	require.NoError(t, os.Mkdir(workdir+".lock", 0o755))
 	require.NoError(t, os.WriteFile(filepath.Join(workdir, ".gitea", "workflows", "reusable.yml"), []byte(":"), 0o644))
 
 	for _, uses := range []string{"./.gitea/workflows/reusable.yml", "$/.gitea/workflows/reusable.yml"} {
@@ -123,8 +125,7 @@ func TestNewLocalReusableWorkflowExecutorFindsSameRepositoryPaths(t *testing.T)
 			Run:    &model.Run{JobID: "job", Workflow: &model.Workflow{Jobs: map[string]*model.Job{"job": {Uses: uses}}}},
 		}
 		err := newLocalReusableWorkflowExecutor(rc)(t.Context())
-		require.Error(t, err)
-		require.NotErrorIs(t, err, fs.ErrNotExist)
+		require.ErrorContains(t, err, "workflow is not valid")
 	}
 }
 

act/runner/run_context.go

@@ -1541,14 +1541,8 @@ func (rc *RunContext) getGithubContext(ctx context.Context) *model.GithubContext
 			merged.RunnerPerflog, merged.RunnerTrackingID = ghc.RunnerPerflog, ghc.RunnerTrackingID
 			*ghc = merged
 
-			instance := rc.Config.GitHubInstance
-			if !strings.HasPrefix(instance, "http://") &&
-				!strings.HasPrefix(instance, "https://") {
-				instance = "https://" + instance
-			}
-			ghc.ServerURL = instance
-			ghc.APIURL = instance + "/api/v1" // the version of Gitea is v1
-			ghc.GraphQLURL = ""               // Gitea doesn't support graphql
+			rc.setInstanceURLs(ghc)
+			ghc.GraphQLURL = "" // Gitea doesn't support graphql
 			return ghc
 		}
 	}
@@ -1562,7 +1556,7 @@ func (rc *RunContext) getGithubContext(ctx context.Context) *model.GithubContext
 
 	ghc.SetBaseAndHeadRef()
 	repoPath := rc.Config.Workdir
-	ghcontext.SetRepositoryAndOwner(ctx, ghc, rc.Config.GitHubInstance, repoPath)
+	ghcontext.SetRepositoryAndOwner(ctx, ghc, repoPath)
 	if ghc.Ref == "" {
 		ghcontext.SetRef(ctx, ghc, repoPath)
 	}
@@ -1572,27 +1566,7 @@ func (rc *RunContext) getGithubContext(ctx context.Context) *model.GithubContext
 
 	ghc.SetRefTypeAndName()
 
-	// defaults
-	ghc.ServerURL = "https://github.com"
-	ghc.APIURL = "https://api.github.com"
-	ghc.GraphQLURL = "https://api.github.com/graphql"
-	// per GHES
-	if rc.Config.GitHubInstance != "github.com" {
-		ghc.ServerURL = "https://" + rc.Config.GitHubInstance
-		ghc.APIURL = fmt.Sprintf("https://%s/api/v3", rc.Config.GitHubInstance)
-		ghc.GraphQLURL = fmt.Sprintf("https://%s/api/graphql", rc.Config.GitHubInstance)
-	}
-
-	{ // Adapt to Gitea
-		instance := rc.Config.GitHubInstance
-		if !strings.HasPrefix(instance, "http://") &&
-			!strings.HasPrefix(instance, "https://") {
-			instance = "https://" + instance
-		}
-		ghc.ServerURL = instance
-		ghc.APIURL = instance + "/api/v1" // the version of Gitea is v1
-		ghc.GraphQLURL = ""               // Gitea doesn't support graphql
-	}
+	rc.setInstanceURLs(ghc)
 
 	// allow to be overridden by user
 	if rc.Config.Env["GITHUB_SERVER_URL"] != "" {
@@ -1608,6 +1582,20 @@ func (rc *RunContext) getGithubContext(ctx context.Context) *model.GithubContext
 	return ghc
 }
 
+func (rc *RunContext) setInstanceURLs(ghc *model.GithubContext) {
+	if instance := strings.TrimRight(rc.Config.GitHubInstance, "/"); instance != "" {
+		ghc.ServerURL = withHTTPS(instance)
+		ghc.APIURL = ghc.ServerURL + "/api/v1"
+	}
+}
+
+func withHTTPS(u string) string {
+	if strings.HasPrefix(u, "http://") || strings.HasPrefix(u, "https://") {
+		return u
+	}
+	return "https://" + u
+}
+
 func isLocalCheckout(ghc *model.GithubContext, step *model.Step) bool {
 	if step.Type() == model.StepTypeInvalid {
 		// This will be errored out by the executor later, we need this here to avoid a null panic though
@@ -1657,20 +1645,16 @@ func (rc *RunContext) withGithubEnv(ctx context.Context, github *model.GithubCon
 	env["GITHUB_HEAD_REF"] = github.HeadRef
 	env["GITHUB_SERVER_URL"] = github.ServerURL
 	env["GITHUB_API_URL"] = github.APIURL
-	env["GITHUB_GRAPHQL_URL"] = github.GraphQLURL
 
-	{ // Adapt to Gitea
-		instance := rc.Config.GitHubInstance
-		if !strings.HasPrefix(instance, "http://") &&
-			!strings.HasPrefix(instance, "https://") {
-			instance = "https://" + instance
+	for _, name := range []string{"GITHUB_SERVER_URL", "GITHUB_API_URL"} {
+		if env[name] == "" {
+			delete(env, name)
 		}
-		env["GITHUB_SERVER_URL"] = instance
-		env["GITHUB_API_URL"] = instance + "/api/v1" // the version of Gitea is v1
-		env["GITHUB_GRAPHQL_URL"] = ""               // Gitea doesn't support graphql
-		env["GITHUB_RUN_ATTEMPT"] = github.RunAttempt
 	}
 
+	env["GITHUB_GRAPHQL_URL"] = ""
+	env["GITHUB_RUN_ATTEMPT"] = github.RunAttempt
+
 	env["RUNNER_NAME"] = rc.Config.RunnerName
 	env["RUNNER_ENVIRONMENT"] = "self-hosted"
 	if workspace := parentDir(github.Workspace); workspace != "" {

act/runner/run_context_test.go

@@ -915,6 +915,7 @@ func TestGetGitHubContext(t *testing.T) {
 		Run: &model.Run{
 			Workflow: &model.Workflow{
 				Name: "GitHubContextTest",
+				Jobs: map[string]*model.Job{"job1": {}},
 			},
 		},
 		Name:        "GitHubContextTest",
@@ -960,6 +961,36 @@ func TestGetGitHubContext(t *testing.T) {
 	assert.Equal(t, "preset-actor", ghc.Actor)
 	assert.Equal(t, "triggerer", ghc.TriggeringActor)
 	assert.Equal(t, "job1", ghc.Job)
+
+	preset := &model.GithubContext{ServerURL: "https://preset.test", APIURL: "https://preset.test/api/v1"}
+	for _, testcase := range []struct {
+		name, instance, serverURL, apiURL string
+		preset                            *model.GithubContext
+		env                               map[string]string
+	}{
+		{name: "unknown instance"},
+		{name: "bare host with trailing slash", instance: "gitea.test/", serverURL: "https://gitea.test", apiURL: "https://gitea.test/api/v1"},
+		{name: "scheme port and base path", instance: "http://gitea.test:3000/gitea/", serverURL: "http://gitea.test:3000/gitea", apiURL: "http://gitea.test:3000/gitea/api/v1"},
+		{name: "daemon instance replaces preset endpoints", instance: "gitea.test", preset: preset, serverURL: "https://gitea.test", apiURL: "https://gitea.test/api/v1"},
+		{name: "daemon keeps preset endpoints without instance", preset: preset, serverURL: "https://preset.test", apiURL: "https://preset.test/api/v1"},
+		{name: "environment overrides", env: map[string]string{"GITHUB_SERVER_URL": "https://override.test", "GITHUB_API_URL": "https://override.test/api"}, serverURL: "https://override.test", apiURL: "https://override.test/api"},
+	} {
+		t.Run(testcase.name, func(t *testing.T) {
+			rc.Config.GitHubInstance = testcase.instance
+			rc.Config.PresetGitHubContext = testcase.preset
+			rc.Config.Env = map[string]string{"GITHUB_REPOSITORY": "owner/repo", "GITHUB_REF": "refs/heads/main", "SHA_REF": "1234567890"}
+			maps.Copy(rc.Config.Env, testcase.env)
+			github := rc.getGithubContext(t.Context())
+			assert.Equal(t, testcase.serverURL, github.ServerURL)
+			assert.Equal(t, testcase.apiURL, github.APIURL)
+			env := map[string]string{}
+			rc.withGithubEnv(t.Context(), github, env)
+			_, exported := env["GITHUB_SERVER_URL"]
+			assert.Equal(t, testcase.serverURL != "", exported)
+			assert.Equal(t, testcase.serverURL, env["GITHUB_SERVER_URL"])
+			assert.Equal(t, testcase.apiURL, env["GITHUB_API_URL"])
+		})
+	}
 }
 
 func TestGetGithubContextRef(t *testing.T) {

act/runner/runner.go

@@ -5,6 +5,7 @@
 package runner
 
 import (
+	"cmp"
 	"context"
 	"fmt"
 	"net/http"
@@ -49,7 +50,7 @@ type Config struct {
 	ContainerDaemonSocket         string                                        // Path to Docker daemon socket
 	ContainerOptions              string                                        // Options for the job container
 	UseGitIgnore                  bool                                          // controls if paths in .gitignore should not be copied into container, default true
-	GitHubInstance                string                                        // GitHub instance to use, default "github.com"
+	GitHubInstance                string                                        // instance hostname or URL, empty when unknown
 	ContainerCapAdd               []string                                      // list of kernel capabilities to add to the containers
 	ContainerCapDrop              []string                                      // list of kernel capabilities to remove from the containers
 	ArtifactServerPath            string                                        // the path where the artifact server stores uploads
@@ -111,13 +112,7 @@ func (c Config) GetToken() string {
 
 // DefaultActionURL returns the host used for implicit remote actions.
 func (c Config) DefaultActionURL() string {
-	if c.DefaultActionInstance != "" {
-		return c.DefaultActionInstance
-	}
-	if c.GitHubInstance != "" {
-		return c.GitHubInstance
-	}
-	return "github.com"
+	return cmp.Or(c.DefaultActionInstance, c.GitHubInstance, "github.com")
 }
 
 type caller struct {

act/runner/step.go

@@ -186,19 +186,25 @@ func runStepExecutor(step step, stage stepStage, executor common.Executor) commo
 		if err == nil {
 			err = insecureErr
 		}
-		if fileErr := processRunnerEnvFileCommand(ctx, envFileCommand, rc, rc.setEnvFile); fileErr != nil && err == nil {
+		fileCommandCtx := ctx
+		if ctx.Err() != nil { // an interrupted step still records its file commands, a live deadline stays in force
+			var cancelFileCommands context.CancelFunc
+			fileCommandCtx, cancelFileCommands = context.WithTimeout(context.WithoutCancel(ctx), jobCleanupTimeout)
+			defer cancelFileCommands()
+		}
+		if fileErr := processRunnerEnvFileCommand(fileCommandCtx, envFileCommand, rc, rc.setEnvFile); fileErr != nil && err == nil {
 			err = fileErr
 		}
-		if fileErr := processRunnerEnvFileCommand(ctx, stateFileCommand, rc, rc.saveState); fileErr != nil && err == nil {
+		if fileErr := processRunnerEnvFileCommand(fileCommandCtx, stateFileCommand, rc, rc.saveState); fileErr != nil && err == nil {
 			err = fileErr
 		}
-		if fileErr := processRunnerEnvFileCommand(ctx, outputFileCommand, rc, rc.setOutput); fileErr != nil && err == nil {
+		if fileErr := processRunnerEnvFileCommand(fileCommandCtx, outputFileCommand, rc, rc.setOutput); fileErr != nil && err == nil {
 			err = fileErr
 		}
-		if fileErr := rc.UpdateExtraPath(ctx, path.Join(actPath, pathFileCommand)); fileErr != nil && err == nil {
+		if fileErr := rc.UpdateExtraPath(fileCommandCtx, path.Join(actPath, pathFileCommand)); fileErr != nil && err == nil {
 			err = fileErr
 		}
-		_ = rc.JobContainer.Copy(actPath, files...)(ctx)
+		_ = rc.JobContainer.Copy(actPath, files...)(fileCommandCtx)
 
 		if err == nil {
 			logger.WithField("stepResult", stepResult.Conclusion).Infof("Success - %s %s", stage, stepString)

act/runner/step_action_remote.go

@@ -88,6 +88,12 @@ func (sar *stepActionRemote) prepareActionExecutor() common.Executor {
 			}
 		}
 
+		unlock, err := git.AcquireCloneLock(ctx, actionDir)
+		if err != nil {
+			return err
+		}
+		defer unlock()
+
 		// Best effort: the download report falls back to the ref alone when the commit is unknown.
 		if _, sha, err := git.FindGitRevision(ctx, actionDir); err != nil {
 			common.Logger(ctx).Debugf("unable to resolve the commit of %s: %v", sar.remoteAction.Reference(), err)
@@ -100,7 +106,6 @@ func (sar *stepActionRemote) prepareActionExecutor() common.Executor {
 			return f, f, err
 		}
 
-		defer git.AcquireCloneLock(actionDir)()
 		actionModel, err := sar.readAction(ctx, sar.Step, actionDir, sar.remoteAction.Path, remoteReader, os.WriteFile)
 		sar.action = actionModel
 		return err
@@ -251,8 +256,8 @@ func (ra *remoteAction) CloneURL(u string) string {
 	if ra.URL == "" {
 		// keep an absolute local path as-is (used by tests to resolve actions from a local
 		// repo); only bare host names get the https:// scheme prepended
-		if !strings.HasPrefix(u, "http://") && !strings.HasPrefix(u, "https://") && !filepath.IsAbs(u) {
-			u = "https://" + u
+		if !filepath.IsAbs(u) {
+			u = withHTTPS(u)
 		}
 	} else {
 		u = ra.URL
@@ -301,6 +306,9 @@ func newRemoteAction(action string, github *model.GithubContext, instanceURL str
 		}
 		ra.Org, ra.Repo, _ = strings.Cut(repo, "/")
 		ra.URL, ra.Ref = github.ServerURL, ref
+		if ra.URL == "" {
+			return nil, fmt.Errorf("unable to resolve %q without a Gitea instance", action)
+		}
 		if ra.Org == "" || ra.Repo == "" || ra.Ref == "" {
 			return nil, fmt.Errorf("unable to resolve %q without a repository and commit", action)
 		}

act/runner/step_action_remote_test.go

@@ -193,7 +193,7 @@ func TestStepActionRemote(t *testing.T) {
 				RunContext: &RunContext{
 					Config: &Config{
 						GitHubInstance: "github.com",
-						ActionCacheDir: "/tmp/test-cache",
+						ActionCacheDir: t.TempDir(),
 					},
 					Run: &model.Run{
 						JobID: "1",
@@ -447,7 +447,7 @@ func TestStepActionRemotePost(t *testing.T) {
 				RunContext: &RunContext{
 					Config: &Config{
 						GitHubInstance: "https://github.com",
-						ActionCacheDir: "/tmp/test-cache",
+						ActionCacheDir: t.TempDir(),
 					},
 					JobContainer: cm,
 					Run: &model.Run{
@@ -688,6 +688,7 @@ func Test_newRemoteActionSelfRepo(t *testing.T) {
 				Ref:  "v1",
 			},
 		},
+		{name: "unknown instance", action: "$/.gitea/actions/build", github: &model.GithubContext{Repository: "owner/workflow-repo", Sha: "abc123"}},
 		{name: "empty path", action: "$/", github: workflow},
 		{name: "ref suffix is not allowed", action: "$/.gitea/actions/build@v1", github: workflow},
 		{name: "path traversal", action: "$/../escape", github: workflow},
@@ -867,7 +868,7 @@ func TestStepActionRemoteCloneTokenSurvivesNilSecrets(t *testing.T) {
 					Config: &Config{
 						GitHubInstance:        tt.gitHubInstance,
 						DefaultActionInstance: tt.defaultActionInstance,
-						ActionCacheDir:        "/tmp/test-cache",
+						ActionCacheDir:        t.TempDir(),
 						// Mirrors the state of a composite RunContext: job secrets are
 						// stripped, but the job token is still reachable via Config.Token.
 						Secrets: nil,

act/runner/step_test.go

@@ -7,9 +7,11 @@ package runner
 import (
 	"context"
 	"errors"
+	"io"
 	"os"
 	"strings"
 	"testing"
+	"testing/synctest"
 
 	"gitea.com/gitea/runner/act/common"
 	"gitea.com/gitea/runner/act/container"
@@ -189,7 +191,6 @@ func TestSetupEnv(t *testing.T) {
 		"GITHUB_ACTION_PATH":       "",
 		"GITHUB_ACTION_REF":        "",
 		"GITHUB_ACTION_REPOSITORY": "",
-		"GITHUB_API_URL":           "https:///api/v1", // Gitea uses api/v1 (upstream GitHub: api/v3)
 		"GITHUB_BASE_REF":          "",
 		"GITHUB_EVENT_NAME":        "",
 		"GITHUB_EVENT_PATH":        "/var/run/act/workflow/event.json",
@@ -200,7 +201,6 @@ func TestSetupEnv(t *testing.T) {
 		"GITHUB_RUN_ATTEMPT":       "",
 		"GITHUB_RUN_ID":            "runId",
 		"GITHUB_RUN_NUMBER":        "1",
-		"GITHUB_SERVER_URL":        "https://",
 		"GITHUB_WORKFLOW":          "",
 		"INPUT_ID":                 "1234567",
 		"INPUT_STEP_WITH":          "with-value",
@@ -475,6 +475,26 @@ func TestRunStepExecutorDoesNotLeakRefusalToNextStep(t *testing.T) {
 	require.NoError(t, errB)
 }
 
+type fileCommandTestContainer struct {
+	*container.HostEnvironment
+}
+
+func (c *fileCommandTestContainer) UpdateFromEnv(srcPath string, env *map[string]string) common.Executor {
+	return func(ctx context.Context) error {
+		if err := ctx.Err(); err != nil {
+			return err
+		}
+		return c.HostEnvironment.UpdateFromEnv(srcPath, env)(ctx)
+	}
+}
+
+func (c *fileCommandTestContainer) GetContainerArchive(ctx context.Context, srcPath string) (io.ReadCloser, error) {
+	if err := ctx.Err(); err != nil {
+		return nil, err
+	}
+	return c.HostEnvironment.GetContainerArchive(ctx, srcPath)
+}
+
 func TestRunStepExecutorParity(t *testing.T) {
 	newStep := func(t *testing.T, stepModel *model.Step) *stepRun {
 		rc := createRunContext(t)
@@ -527,6 +547,30 @@ func TestRunStepExecutorParity(t *testing.T) {
 		}
 	})
 
+	t.Run("file commands survive a job timeout", func(t *testing.T) {
+		synctest.Test(t, func(t *testing.T) {
+			step := newStep(t, &model.Step{ID: "interrupted"})
+			ctx, cancel := applyJobTimeout(t.Context(), step.RunContext, &model.Job{TimeoutMinutes: "1"})
+			defer cancel()
+			step.RunContext.JobContainer = &fileCommandTestContainer{step.RunContext.JobContainer.(*container.HostEnvironment)}
+			err := runStepExecutor(step, stepStageMain, func(execCtx context.Context) error {
+				for name, content := range map[string]string{
+					"GITHUB_ENV": "KEPT_ENV=value\n", "GITHUB_STATE": "saved=value\n",
+					"GITHUB_OUTPUT": "kept=value\n", "GITHUB_PATH": "/kept/bin\n",
+				} {
+					require.NoError(t, os.WriteFile(step.env[name], []byte(content), 0o600))
+				}
+				<-execCtx.Done()
+				return execCtx.Err()
+			})(ctx)
+			require.ErrorIs(t, err, context.DeadlineExceeded)
+			assert.Equal(t, "value", step.RunContext.Env["KEPT_ENV"])
+			assert.Equal(t, "value", step.RunContext.IntraActionState[step.Step.ID]["saved"])
+			assert.Equal(t, "value", step.RunContext.StepResults[step.Step.ID].Outputs["kept"])
+			assert.Contains(t, step.RunContext.ExtraPath, "/kept/bin")
+		})
+	})
+
 	t.Run("composite child files do not leak", func(t *testing.T) {
 		outer := newStep(t, &model.Step{ID: "outer"})
 		childRC := &RunContext{

internal/app/cmd/daemon.go

@@ -5,8 +5,10 @@ package cmd
 
 import (
 	"context"
+	"crypto/tls"
 	"errors"
 	"fmt"
+	"net/http"
 	"os"
 	"path"
 	"path/filepath"
@@ -29,6 +31,8 @@ import (
 	"gitea.com/gitea/runner/internal/pkg/ver"
 
 	"connectrpc.com/connect"
+	runnerv1 "gitea.dev/actionslib/runner/v1"
+	"github.com/avast/retry-go/v5"
 	"github.com/mattn/go-isatty"
 	log "github.com/sirupsen/logrus"
 	"github.com/spf13/cobra"
@@ -185,8 +189,25 @@ func runDaemon(ctx context.Context, daemArgs *daemonArgs, configFile *string) fu
 		}()
 
 		// declare the labels of the runner before fetching tasks
-		resp, err := runner.Declare(ctx, ls.Names())
+		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 {
+					log.WithError(err).Warn("fail to invoke Declare, retrying until Gitea is reachable")
+				} else {
+					log.WithError(err).Debug("fail to invoke Declare, retrying")
+				}
+			}),
+		).Do(func() (*connect.Response[runnerv1.DeclareResponse], error) {
+			return runner.Declare(ctx, ls.Names())
+		})
 		switch {
+		case err != nil && ctx.Err() != nil && !client.IsRegistrationRejected(err):
+			return nil
 		case err != nil && connect.CodeOf(err) == connect.CodeUnimplemented:
 			log.Errorf("Your Gitea version is too old to support runner declare, please upgrade to v1.21 or later")
 			return err
@@ -352,6 +373,20 @@ func initLogging(cfg *config.Config) {
 	}
 }
 
+var declareRetryDelay = 30 * time.Second
+
+func shouldRetryDeclare(err error) bool {
+	if errors.As(err, new(*tls.CertificateVerificationError)) || errors.As(err, new(tls.AlertError)) || errors.Is(err, http.ErrSchemeMismatch) { // misconfigurations never heal
+		return false
+	}
+	switch connect.CodeOf(err) {
+	case connect.CodeUnauthenticated, connect.CodeUnimplemented, connect.CodePermissionDenied, connect.CodeInvalidArgument:
+		return false
+	default:
+		return !client.IsRegistrationRejected(err)
+	}
+}
+
 var commonSocketPaths = []string{
 	"/var/run/docker.sock",
 	"/run/podman/podman.sock",

internal/app/cmd/daemon_test.go

@@ -5,11 +5,13 @@ package cmd
 
 import (
 	"context"
+	"crypto/tls"
 	"errors"
 	"net/http"
 	"net/http/httptest"
 	"os"
 	"path/filepath"
+	"sync/atomic"
 	"testing"
 	"time"
 
@@ -87,12 +89,19 @@ var errUnregisteredRunner = errors.New("rpc error: code = Unauthenticated desc =
 type daemonTestService struct {
 	runnerv1connect.UnimplementedRunnerServiceHandler
 	declareErr, fetchErr error
+	declareCalls         atomic.Int32
 	beforeTask           bool
 	cancel               context.CancelFunc
 }
 
 func (s *daemonTestService) Declare(context.Context, *connect.Request[runnerv1.DeclareRequest]) (*connect.Response[runnerv1.DeclareResponse], error) {
+	if s.declareCalls.Add(1) == 1 {
+		return nil, connect.NewError(connect.CodeUnavailable, errors.New("bad gateway"))
+	}
 	if s.declareErr != nil {
+		if s.beforeTask {
+			s.cancel()
+		}
 		return nil, s.declareErr
 	}
 	return connect.NewResponse(&runnerv1.DeclareResponse{Runner: &runnerv1.Runner{Name: "test"}}), nil
@@ -117,7 +126,9 @@ func (*daemonTestService) UpdateLog(_ context.Context, req *connect.Request[runn
 	return connect.NewResponse(&runnerv1.UpdateLogResponse{AckIndex: req.Msg.Index + int64(len(req.Msg.Rows))}), nil
 }
 
-func TestDaemonRemovesOnlyConsumedEphemeralRegistration(t *testing.T) {
+func TestDaemonRetriesDeclareOutagesAndRemovesOnlyConsumedEphemeralRegistration(t *testing.T) {
+	defer func(delay time.Duration) { declareRetryDelay = delay }(declareRetryDelay)
+	declareRetryDelay = 0
 	for _, tc := range []struct {
 		name                  string
 		ephemeral, beforeTask bool
@@ -132,11 +143,13 @@ func TestDaemonRemovesOnlyConsumedEphemeralRegistration(t *testing.T) {
 		{"ephemeral proxy 401 on fetch", true, false, nil, connect.NewError(connect.CodeUnauthenticated, errors.New("HTTP status 401 Unauthorized")), false},
 		{"ephemeral rejected on declare", true, false, errUnregisteredRunner, nil, true},
 		{"persistent rejected on declare", false, false, errUnregisteredRunner, nil, false},
+		{"ephemeral stopped during declare outage", true, true, connect.NewError(connect.CodeUnavailable, errors.New("bad gateway")), nil, false},
 	} {
 		t.Run(tc.name, func(t *testing.T) {
 			ctx, cancel := context.WithTimeout(t.Context(), 5*time.Second)
 			defer cancel()
-			_, handler := runnerv1connect.NewRunnerServiceHandler(&daemonTestService{declareErr: tc.declareErr, fetchErr: tc.fetchErr, beforeTask: tc.beforeTask, cancel: cancel})
+			service := &daemonTestService{declareErr: tc.declareErr, fetchErr: tc.fetchErr, beforeTask: tc.beforeTask, cancel: cancel}
+			_, handler := runnerv1connect.NewRunnerServiceHandler(service)
 			server := httptest.NewServer(http.StripPrefix("/api/actions", handler))
 			defer server.Close()
 			dir := t.TempDir()
@@ -145,11 +158,12 @@ func TestDaemonRemovesOnlyConsumedEphemeralRegistration(t *testing.T) {
 			require.NoError(t, os.WriteFile(configFile, []byte("runner:\n  file: "+regFile+"\n  idle_cleanup_interval: 0s\ncache:\n  enabled: false\n"), 0o600))
 			require.NoError(t, config.SaveRegistration(regFile, &config.Registration{Address: server.URL, UUID: "test", Name: "test", Token: "test", Labels: []string{"host:host"}, Ephemeral: tc.ephemeral}))
 			err := runDaemon(ctx, &daemonArgs{Once: true}, &configFile)(nil, nil)
-			if tc.declareErr == nil && tc.fetchErr == nil {
+			if tc.declareErr == nil && tc.fetchErr == nil || tc.beforeTask {
 				require.NoError(t, err)
 			} else {
 				require.Error(t, err)
 			}
+			require.Equal(t, int32(2), service.declareCalls.Load())
 			if tc.removed {
 				require.NoFileExists(t, regFile)
 			} else {
@@ -158,3 +172,20 @@ func TestDaemonRemovesOnlyConsumedEphemeralRegistration(t *testing.T) {
 		})
 	}
 }
+
+func TestShouldRetryDeclare(t *testing.T) {
+	for _, tc := range []struct {
+		err   error
+		retry bool
+	}{
+		{connect.NewError(connect.CodeUnavailable, errors.New("bad gateway")), true},
+		{connect.NewWireError(connect.CodeInternal, errors.New("update runner")), true},
+		{connect.NewError(connect.CodeUnavailable, &tls.CertificateVerificationError{Err: errors.New("unknown authority")}), false},
+		{connect.NewError(connect.CodeUnavailable, http.ErrSchemeMismatch), false},
+		{connect.NewError(connect.CodeUnavailable, tls.AlertError(42)), false},
+		{connect.NewError(connect.CodePermissionDenied, errors.New("forbidden")), false},
+		{connect.NewWireError(connect.CodeUnknown, errors.New("unregistered runner")), false},
+	} {
+		require.Equal(t, tc.retry, shouldRetryDeclare(tc.err), tc.err.Error())
+	}
+}

scripts/test-dind.sh

@@ -9,7 +9,7 @@
 #
 # Usage: scripts/test-dind.sh [target] [-- go-test-args...]
 #   target:        dind (default), dind-rootless, or podman to run PODMAN_TEST_IMAGE's API service instead
-#   go-test-args:  passed verbatim to `go test`. Defaults cover image env extraction,
+#   go-test-args:  passed verbatim to `go test`. Defaults cover image platform detection,
 #                  symlink copying, a mounted Docker job using cached images and the
 #                  image's s6 supervision, or the Docker proxy probe for podman.
 #
@@ -55,7 +55,7 @@ else
   daemon_args=(-e DOCKER_TLS_CERTDIR= --entrypoint dockerd-entrypoint.sh "$image" --host=tcp://0.0.0.0:2375)
   if [ $# -eq 0 ]; then
     default_tests=true
-    set -- -count=1 -race -run '^TestDocker$' ./act/container/
+    set -- -count=1 -race -run '^TestImageExistsLocally$' ./act/container/
   fi
 fi