fix(test): drain stdin in scheduler-wake test fixture to close a SIGPIPE flake

[seonghobae]

The bug

tests/test_opencode_required_verdict_regression.py::test_scheduler_wake_reuses_trusted_receipt_predicate intermittently fails with AssertionError: assert 141 == 0 (SIGPIPE) — first observed on #1519 (an unrelated Dependabot version bump) via the Hourly cadence, immutable source, NIM credential, and conflict scope required check, then reproduced locally at roughly 1/20 runs.

Root cause: the test's fake gh script's repos/ContextualWisdomLab/.github/dispatches branch never reads stdin before exiting:

elif [[ "$*" == *"repos/ContextualWisdomLab/.github/dispatches"* ]]; then
  printf 'dispatch\n' >>"$DISPATCH_CALLS"
fi

The real production pipe under test is jq -cn ... | GH_TOKEN="$app_token" gh api -X POST repos/ContextualWisdomLab/.github/dispatches --input -. When the fake gh process (the pipe's reader) exits before jq -cn (the pipe's writer) finishes writing — a timing race, not deterministic — the writer receives SIGPIPE, and under this script's set -euo pipefail, that propagates as the observed exit code 141.

Confirmed test-harness-only, not a production bug: the real gh api -X POST ... --input - does read and use its stdin, so it wouldn't exit early and trigger this race outside the test.

The fix

Drain stdin (cat >/dev/null) in the fake gh's dispatches branch before it proceeds, so the pipe's reader always fully consumes the writer's output regardless of scheduling timing:

elif [[ "$*" == *"repos/ContextualWisdomLab/.github/dispatches"* ]]; then
  cat >/dev/null
  printf 'dispatch\n' >>"$DISPATCH_CALLS"
fi

Validation

• Reproduced the flake locally before the fix: ~1/20 runs of test_scheduler_wake_reuses_trusted_receipt_predicate failed with assert 141 == 0.
• After the fix: 60/60 clean runs of the same test.
• PYTHONPATH=. python3 -m pytest tests -q: 2246 passed, 1 skipped, 21 subtests.
• ruff check .: no new findings (one pre-existing, unrelated F401 in tests/test_opencode_model_pool_runner.py, confirmed present before this change via git stash).
• interrogate -c pyproject.toml .: 100.0%.
• git diff --check: clean.

Scope

Test-only change (one added line, cat >/dev/null), no production code touched.

---

Generated by Claude Code

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[seonghobae]

The source fix is current-main, mergeable, test-only, and locally verified, but this PR is still Draft. The connected Ready-for-review mutation is currently broken by a connector GraphQL schema error (Repository.fullDatabaseId does not exist), not by repository state. To avoid leaving a valid one-line flake fix indefinitely unreachable by the normal reviewer pipeline, I am replacing this draft conversation with a fresh non-draft PR from the identical unchanged branch/head. No checks/reviews from this PR transfer; the successor must regenerate exact-head evidence.