juanfont/headscale · error

getting working directory: %w

Error message

getting working directory: %w

What it means

Returned by `createGoTestContainer` when `os.Getwd()` fails while resolving the repo path to bind-mount into the test container. This is an OS-level error (not Docker's): the process's current working directory was deleted or is otherwise unreadable. The cwd is needed to compute `findProjectRoot(pwd)` for the source mount.

Source

Thrown at cmd/hi/docker.go:227

	if config.TestPattern != "" {
		cmd = append(cmd, "-run", config.TestPattern)
	}

	if config.FailFast {
		cmd = append(cmd, "-failfast")
	}

	cmd = append(cmd, "-timeout", config.Timeout.String())
	cmd = append(cmd, "-v")

	return cmd
}

// createGoTestContainer creates a Docker container configured for running integration tests.
func createGoTestContainer(ctx context.Context, cli *client.Client, config *RunConfig, containerName, logsDir string, goTestCmd []string) (container.CreateResponse, error) {
	pwd, err := os.Getwd()
	if err != nil {
		return container.CreateResponse{}, fmt.Errorf("getting working directory: %w", err)
	}

	projectRoot := findProjectRoot(pwd)

	runID := dockertestutil.ExtractRunIDFromContainerName(containerName)

	env := []string{
		fmt.Sprintf("HEADSCALE_INTEGRATION_POSTGRES=%d", boolToInt(config.UsePostgres)),
		"HEADSCALE_INTEGRATION_RUN_ID=" + runID,
	}

	// Pass through CI environment variable for CI detection
	if ci := os.Getenv("CI"); ci != "" {
		env = append(env, "CI="+ci)
	}

	// Pass through all HEADSCALE_INTEGRATION_* environment variables
	for _, e := range os.Environ() {

View on GitHub (pinned to 565fd254d0)

Solutions

1. Re-run from a directory that exists and is stable for the whole test (repo root).
2. In CI, ensure workspace cleanup steps cannot race test execution.
3. Check `pwd` works in the same shell before invoking.

// verify cwd is readable before launching the runner
if wd, err := os.Getwd(); err != nil || wd == "" {
    return fmt.Errorf("invalid working directory: %v", err)
}

if err := runDockerTest(ctx, config); err != nil {
    if strings.Contains(err.Error(), "getting working directory") {
        // restart the process from the repo root; the old cwd was deleted
    }
}

When it happens

Trigger: Starting `hi run` from a directory that is then deleted before this call; running the binary with a cwd on an unmounted/network filesystem that returns an error; extremely unusual containerized CI where cwd is invalid.

Common situations: Running `hi` from a repo clone that gets removed/replaced mid-session (e.g. CI workspace cleanup racing the job); deleted scratch directories; FUSE mounts returning EIO.

Related errors

• getting absolute path for logs directory: %w
• creating directory failed with permission error
• creating temp dir: %w
• writing config: %w
• dumping config: %w

AI-assisted analysis of juanfont/headscale@565fd254d0 (2026-08-15). Data as JSON: /api/errors/1e30208e2fad57c6. Report an issue: GitHub.