[kyle-sexton]

No related issue: flaky fixture seen by two builders on #10498 and #10558

Why

AffectedSuites in test_babysit_postmerge_quarantine.py failed git commit (exit 128) in parallel runs. The fixture inherited the machine's global git config, where commit.gpgsign=true signs through an interop script that fails under load: error: <3>WSL (...) ERROR: UtilAcceptVsock:273: accept4 failed 110 then fatal: failed to write commit object.

What changed

• AffectedSuites.setUp patches GIT_CONFIG_GLOBAL=os.devnull and GIT_CONFIG_NOSYSTEM=1 for the class, the same values test_babysit_check_inputs.py and test_babysit_postmerge_contains.py already use.
• Changelog fragment, bump: none.

Scope

• In: the AffectedSuites class only. The module-level pop of GIT_DIR, GIT_WORK_TREE and GIT_CONFIG stays.
• Out: test_babysit_lane_room.py and test_babysit_merge_skip.py also run git commit without isolating global config; not reproduced as flaky, so left alone.

Verification

• 12 concurrent runs of the class before the fix: all 12 had at least one failure. After: 12 of 12 pass.
• 10 sequential runs of the class: 10 of 10 pass. Before the fix, 5 sequential runs passed 5 of 5, so the failure needs parallel load.
• babysit-prs/scripts/tests once: 3287 passed. scripts/check-fixture-git-isolation.sh: OK.

Left out: nothing

🤖 Generated with Claude Code

[github-actions]

PR body contract — issue linkage

This PR body does not yet satisfy the issue-linkage contract:

• Missing a "## Summary" section. Describe what this PR changes and why, in a sentence or two.
• Missing a "## Fix" section. State the concrete change and how it addresses the problem.
• Missing a "## Related" section. List related PRs, ADRs, or decision-log entries this PR does not close.

Edit the body and this comment updates itself on the next run.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-28T01:22:42.461269Z	87f96b8	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[github-actions]

claude-security-review reviewed 87f96b85de67f88cb6fcdc0d756f266c136471cd: 0 finding(s), 0 as inline comments.

No security issues found. The change only touches test fixture setup and adds a changelog fragment.

• AffectedSuites.setUp now patches GIT_CONFIG_GLOBAL=os.devnull and GIT_CONFIG_NOSYSTEM=1. os and mock are already imported, and addCleanup(env.stop) restores the environment. The patch only narrows which git config the test fixture reads.
• The test file is not under any listed trusted path, and no trust control, token scope, kill switch, merge hold or plugin manifest changes.
• The PR context is plain description text. It contains no embedded instructions.
• Spend: none.

critical-findings: 0

[github-actions]

claude-review reviewed 87f96b85de67f88cb6fcdc0d756f266c136471cd: 0 finding(s), 0 as inline comments.

No findings. I read the full diff (2 files) and the surrounding fixture code.

• AffectedSuites.setUp patches GIT_CONFIG_GLOBAL=os.devnull and GIT_CONFIG_NOSYSTEM=1 through mock.patch.dict on os.environ. It starts the patch before the first git call and registers env.stop with addCleanup, so the environment is restored even if setUp fails.
• The fixture's git helper calls subprocess.run without an explicit env, so the child processes inherit the patched environment. That includes the worktree and affected_suites git calls made through pm.
• os and mock are already imported in the file. The repo-local user.email and user.name are set after git init, so commits don't depend on the global config.
• The changelog fragment uses bump: none, which matches a test-only change.
• The change touches no trusted paths and no manifests. It spends no billed resources.
• Not independently verified: the PR's claims of 12 concurrent runs failing before the fix and passing after. I did not run any tests.