fix(tests): repair changed-scope drift and stale noema cancel-step test

[seonghobae]

Summary

Fixes 2 pre-existing tests/ failures that independently block agent-review-runtime-quality-ci.yml's "Verify scheduler and contextual-orchestrator review-repair contracts" step — it runs pytest -q --cov=... --cov-fail-under=100 with no positional test-file arguments, so it implicitly collects and runs pytest's default full-tests/-directory discovery. Any failing test anywhere in tests/ fails this required check. Confirmed unrelated to .github#1875 (which fixes 3 separate stale hourly-cron tests) — these two test files are untouched by that PR's diff.

Bug 1 — test_gate_job_is_byte_identical_across_the_five_workflows_apart_from_if

strix.yml's changed-scope job had textually drifted from its byte-identical siblings in security-scan.yml/sast-semgrep.yml. Root cause: PR #1869 ("retire review scans when PRs return to draft") folded strix.yml's if: condition onto a multi-line >- block scalar to add converted_to_draft handling:

if: >-
  github.event_name != 'pull_request_target' ||
  (github.event.action != 'closed' && github.event.action != 'converted_to_draft')

The test's normalizer only strips lines that literally start with if: (by design — each workflow is allowed its own single-line if:), so the two continuation lines survived and broke the intended byte-for-byte parity of the rest of the job. Fix: collapsed it to one physical if: line, same expression, no semantic change:

if: github.event_name != 'pull_request_target' || (github.event.action != 'closed' && github.event.action != 'converted_to_draft')

Bug 2 — test_noema_close_cleanup_selects_only_the_closed_pr_across_shared_display_titles

This was a stale test, not a production regression. The same PR #1869 generalized noema-review.yml's cancel-closed-pr-runs job to also cover converted_to_draft (not just closed):

• step renamed "...for the closed pull request" → "...for the inactive pull request"
• env vars renamed CLOSED_PR_NUMBER/CLOSED_PR_HEAD_SHA → INACTIVE_PR_NUMBER/INACTIVE_PR_HEAD_SHA, plus new PR_ACTION
• added a live_target_matches live-PR re-verification (mirroring strix.yml's identical job) before every cancellation pass

tests/test_noema_review_gate.py's equivalent tests were updated for this at the time; this one test in test_noema_orchestrator_workflow_contract.py was missed, so it still looked up the old step name and passed the old env var, and the script's un-mocked live_target_matches gh lookup would have failed regardless once the step name matched.

Verified the protected invariant is intact, not broken: the PR #1507 "sibling Noema runs evade cancellation" fix (matching via GitHub's own pull_requests[] array, not just display_title) is unchanged in the current script — confirmed by reading the full cancel_runs() jq selector. No production regression; only the test needed updating. Updated it to the new step name/env vars and taught its fake gh to answer the new pulls/<number> live-state lookup.

Validation

1. PYTHONPATH=. python3 -m pytest tests/test_docs_only_pr_runner_admission.py tests/test_noema_orchestrator_workflow_contract.py tests/test_required_workflow_queue_contract.py -v — 91 passed.
2. Exact unscoped command from the CI step (--cov=scripts.ci.pr_review_conflict_scope --cov=scripts.ci.pr_review_autofix_context --cov=scripts.ci.zdr_policy --cov=scripts.ci.contextual_orchestrator_review_policy --cov-branch --cov-fail-under=100) — 100% coverage reached; only the pre-existing stale-cron failures (owned by .github#1875) and the known local-sandbox gh-missing limitation (test_strix_rerun_job_selection.py) remain (20 failed, down from 22 on a clean baseline; both targeted tests now pass).
3. Full suite (coverage run -m pytest tests --ignore=tests/test_queue_cancellation_open_pr_revalidation.py --ignore=tests/test_queue_cancellation_revalidation.py -q): baseline on clean origin/main was 22 failed, 2830 passed; after this fix, 20 failed, 2832 passed — exactly the 2 targeted bugs fixed, zero new regressions.
4. coverage report --include='scripts/ci/*' --show-missing and interrogate scripts/ci: both identical to the clean-baseline result (99% / 98.3%, from the two revalidation test files excluded above, which the harness doesn't mock gh/git for) — confirmed unaffected by this test-only/workflow-YAML-only change via a stash-and-compare.
5. git diff --check clean; strix.yml re-parses as valid YAML; the edited test file re-parses as valid Python.
6. Added a CHANGELOG.md entry under the first ## [Unreleased] heading.

---

🤖 Generated with Claude Code

https://claude.ai/code/session_01KPmJErfkcHer4UVEgrQxUX

---

Generated by Claude Code

[coderabbitai]

Warning

Review limit reached

Next included review available in 17 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Team

Run ID: 2467e4c8-bab1-48ef-9d49-98238f687c7a

📥 Commits

Reviewing files that changed from the base of the PR and between 1b65dbc and f374f5b.

📒 Files selected for processing (7)

• .github/workflows/strix.yml
• CHANGELOG.md
• scripts/ci/test_strix_quick_gate.sh
• tests/test_github_hourly_conflict_repair.py
• tests/test_hourly_review_repair_callers.py
• tests/test_noema_orchestrator_workflow_contract.py
• tests/test_strix_rerun_job_selection.py

---

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]

Exact-head dependency RCA at 183f9703991fff868559cf4847265ead1da7ff03: the Noema/Strix focused slice reaches 23 passing tests plus 17 subtests, but tests/test_strix_rerun_job_selection.py::test_dispatch_strix_reruns_scan_job_not_sibling_publisher still fails (stale_head vs rerun). That missing fixture correction is owned by unmerged #1871, so this PR is not independently GREEN against current main. I returned it to Draft rather than treating two repaired oracles as a complete current-main fix. Re-evaluate after #1871 merges or reconstruct this head on that protected-main generation; no failing test was excluded.

[seonghobae]

Re-evaluated per the diagnosis above (stale_head vs rerun at 183f9703, attributed to the fixture correction in unmerged #1871).

Rather than pulling #1871 wholesale (~1000 unreviewed lines across 22 files, actively churning), I independently traced the exact call path: dispatch_strix_evidence → live_dispatch_head_matches → fetch_pr. The test mocked rerun_actions_job but not fetch_pr, so with a real gh on PATH the live-head re-read hits the actual GitHub API for a synthetic PR that doesn't exist there, returning a live/head mismatch ("stale_head") instead of "rerun" — confirmed this is the only remaining gh call in that code path (matching_actions_job_id, review_dispatch_admitted, and validate_github_repository are all pure/in-memory).

Cherry-picked only the one confirmed-correct fixture line — monkeypatch.setattr(sched, "fetch_pr", lambda *_args: [pr]) — added alongside the existing rerun_actions_job mock, nothing else from #1871's branch.

Verified:

• tests/test_strix_rerun_job_selection.py now passes cleanly (no gh FileNotFoundError, no stale_head/rerun mismatch).
• Full suite (coverage run -m pytest tests --ignore=tests/test_queue_cancellation_open_pr_revalidation.py --ignore=tests/test_queue_cancellation_revalidation.py -q): 19 failed / 2833 passed — one fewer failure than this branch's prior 20/2832, all 19 remaining failures pre-existing and unrelated (hourly-cron schedule drift, owned by .github#1875), confirmed via stash-and-compare.
• coverage report --include='scripts/ci/*' and interrogate scripts/ci: identical to this branch's pre-existing baseline (99% / 98.3%) — test-only change, no production code touched.
• git diff --check clean.
• Appended a third-fix addendum to this PR's existing CHANGELOG.md bullet (not a new bullet).

One caveat for the reviewer: GitHub currently reports this PR's mergeable_state as dirty — main advanced independently via #1878, which added an unrelated ## [Unreleased] entry to the same CHANGELOG.md section, producing a textual conflict unrelated to this fix. Pre-existing, out of scope for this targeted change; flagging rather than silently resolving it myself.

Marking ready for review again — this is exactly the re-evaluation your comment called for.

---

Generated by Claude Code

[seonghobae]

Merged origin/main into this branch to resolve the mergeable_state: "dirty" GitHub reported after #1878 ("remove organization queue sweep") landed on main and edited the same ## [Unreleased] CHANGELOG.md section this PR touches — a textual conflict, not a logic one (two unrelated bullets added near the same spot).

Resolution: kept the first ## [Unreleased] heading as the active section (per convention, this file has multiple historical ## [Unreleased] headings) and kept both bullets — this PR's own "Fix two pre-existing tests/ failures..." entry and #1878's three bullets (queue-sweep removal, Noema concurrency-group move, coalescer admission scoping) — merged cleanly with no content dropped.

Verified no other files conflicted (confirmed strix.yml/noema-review.yml/this PR's test files were untouched by #1878's diff, as expected).

Validation after the merge commit (13a71d25):

• git diff --check: clean.
• Full suite (coverage run -m pytest tests, excluding the two live-revalidation files per the standing exclusion): 19 failed / 2814 passed — the identical 19 pre-existing, environment/date-driven test_hourly_review_repair_callers.py / test_github_hourly_conflict_repair.py failures present on both the pre-merge PR head and clean origin/main (confirmed via a side-by-side worktree comparison); no new failures introduced by the merge or by #1878.
• This PR's own targeted files (test_docs_only_pr_runner_admission.py, test_noema_orchestrator_workflow_contract.py, test_strix_rerun_job_selection.py): all 19 tests pass.

Pushed as a normal (non-force) merge commit to fix/changed-scope-drift-and-noema-cancel-step.

---

Generated by Claude Code