[CI Failure Doctor] CI Failure Investigation - Run #37460

[github-actions]

Summary

test job fails because TestWasmGolden_CompileFixtures/smoke-copilot now generates a playwright job step with --browser-arg --no-sandbox, but the golden file still expects the previous argument list.

Failure Details

• Run: 22291205041
• Commit: 6740e50
• Trigger: push

Root Cause Analysis

The new Playwright change to "disable Chromium sandbox in Playwright MCP to allow localhost access" (commit 6740e50e5e21c749d84e8bce670cea08d99f80c8) adds "--browser-arg", "--no-sandbox" to the entrypointArgs for the smoke workflows. TestWasmGolden_CompileFixtures/smoke-copilot compares the compiled workflow to pkg/workflow/testdata/wasm_golden/TestWasmGolden_CompileFixtures/smoke-copilot.golden, which still lists only "--output-dir", "/tmp/gh-aw/mcp-logs/playwright", so the golden diff fails as soon as this test runs.

Failed Jobs and Errors

• test: pkg/workflow - TestWasmGolden_CompileFixtures/smoke-copilot (golden diff showing missing --browser-arg --no-sandbox in entrypointArgs)

Investigation Findings

• Golden diff excerpt:

  --- Expected
  +++ Actual
  @@ -497,3 +497,3 @@
                  "args": ["--init", "--network", "host", "--security-opt", "seccomp=unconfined", "--ipc=host"],
  -                "entrypointArgs": ["--output-dir", "/tmp/gh-aw/mcp-logs/playwright"],
  +                "entrypointArgs": ["--output-dir", "/tmp/gh-aw/mcp-logs/playwright", "--browser-arg", "--no-sandbox"],
                  "mounts": ["/tmp/gh-aw/mcp-logs:/tmp/gh-aw/mcp-logs:rw"]
  

• This failure is deterministic once the new Playwright call is compiled, so the test runs fail every CI run until the golden file is updated.
• Reproduction:
  1. go test ./pkg/workflow -run TestWasmGolden_CompileFixtures/smoke-copilot -count=1
  2. Observe the diff against pkg/workflow/testdata/wasm_golden/TestWasmGolden_CompileFixtures/smoke-copilot.golden.
  3. Update the golden to match the compiled workflow (or adjust the compilation logic if another change is expected).

Recommended Actions

• Update pkg/workflow/testdata/wasm_golden/TestWasmGolden_CompileFixtures/smoke-copilot.golden to include the --browser-arg --no-sandbox entries in entrypointArgs.
• Re-run go test ./pkg/workflow -run TestWasmGolden_CompileFixtures -count=1 (or make test-unit if broader coverage is desired) to ensure no other golden mismatches surfaced.
• Document the sandbox flag change in the changelog or release notes since it affects workflow manifests that are validated by golden tests.

Prevention Strategies

• Whenever Playwright/MCP invocation arguments change, rerun TestWasmGolden_CompileFixtures to refresh the golden fixtures before pushing, ensuring the compiled workflow assets stay in sync with test expectations.
• Consider adding a script that regenerates the wasm golden files automatically whenever smoke workflow templates are modified, so CI failures highlight unexpected test regressions instead of expected updates.

AI Team Self-Improvement

When you modify Playwright entrypoint arguments (e.g., to disable the sandbox), immediately rerun the wasm golden fixture tests and update pkg/workflow/testdata/wasm_golden/TestWasmGolden_CompileFixtures/*.golden files before creating CI-affecting commits.

Historical Context

There are no open `[CI Failure Doctor]` issues describing this golden mismatch; the closest search hit was issue #17486, but it is closed and unrelated.

🩺 Diagnosis provided by CI Failure Doctor

To install this workflow, run gh aw add githubnext/agentics/workflows/ci-doctor.md@ea350161ad5dcc9624cf510f134c6a9e39a6f94d. View source at https://github.com/githubnext/agentics/tree/ea350161ad5dcc9624cf510f134c6a9e39a6f94d/workflows/ci-doctor.md.

• expires on Feb 24, 2026, 2:59 AM UTC

[github-actions]

This issue is being closed as outdated. A newer issue has been created: #17821

View newer issue

---

This action was performed automatically by the CI Failure Doctor workflow.