Skip to content

fix(pr-merge): read only the review-status names of the hosted review lanes - #14154

Merged
kyle-sexton merged 4 commits into
mainfrom
fix/drop-legacy-review-status-names
Nov 8, 2026
Merged

kyle-sexton merged 4 commits into
mainfrom
fix/drop-legacy-review-status-names

Conversation

@kyle-sexton

@kyle-sexton kyle-sexton commented Nov 8, 2026 •

Copy link
Copy Markdown
Member

Refs: melodic-software/ci-workflows#866

Summary

ci-workflows#1034 gave pr-review-cli.yml's status job a literal name, review-status. #14075 then made our readers accept both the old and the new names. A real run since #1034 merged shows the new names in this repository, so this drops the old names (pr-review / claude-review-status, pr-review-security / security-review) from the three readers here.

Evidence of the new names: on #14148, run 42497727048 shows pr-review / review-status and run 42497727273 shows pr-review-security / review-status. Both started at about 16:09Z, after #1034 merged at 15:55:08Z.

Fix

  • .github/workflows/pr-require-checks.yml: the cost-hold step and the "Require both review lanes" step now name one status job per lane.
  • .github/scripts/pr-merge/merge-holds.mjs (REVIEW_LANES) and clear-escalations.mjs (LANE_RECORDS) also name one status job per lane.
  • docs/conventions/pr-pipeline/README.md: "Both lanes review the head" names the new status jobs.
  • test_merge_holds_sync.py: drops the "a status job named as pr-review.yml names it" case. It asserted that babysit's Python cost decision and merge-holds.mjs agree on the old name; they now differ on purpose, because babysit keeps reading the old name (next bullet). Babysit's handling stays tested in test_babysit_merge_cost.py, the gate's rejection in merge-holds.test.mjs.
  • Out of scope: the babysit plugin, and other readers of legacy pr-review.yml callers. cursor-plugins, dotfiles, standards and songwriting still use pr-review / claude-review-status.
  • Changelog fragments for pr-pipeline (patch) and source-control (none, test-only).

Effect: a pull request whose last hosted review ran before #1034 merged no longer counts that review. It needs a fresh review run before ci-status, the cost hold or the escalation clear will read it.

Verification

  • The tests that accepted the old names now assert that the old names are rejected, in merge-holds.test.mjs, clear-escalations.test.mjs and clear-escalations-recheck.test.mjs. enable-auto-merge.test.mjs's cost-hold fixture now uses the new name.
  • With the three readers restored to main and the new tests in place, node --test fails 3 tests. With the change, node --test .github/scripts/pr-merge/*.test.mjs .github/scripts/maintenance-reconcile-lost-runs/*.test.mjs passes all 1116.
  • After merging main: test_merge_holds_sync.py 21 tests OK; engine.test.sh exit 0; check-changelog-fragments.sh --check clean.
  • actionlint on pr-require-checks.yml is clean.

Related

🤖 Generated with Claude Code

https://claude.ai/code/session_015XWwmbdW8moDiQ5Rtgt7d5

@chatgpt-codex-connector

Copy link
Copy Markdown
Codex Review Summary: ❌ Failed

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

Review Status Commit Review trigger
📝 Code Review ❌ Failed 2026-11-08T16:12:04.420029Z
We couldn’t complete this review. Comment @codex review to try again.
aece38a PR opened
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.

@chatgpt-codex-connector

Copy link
Copy Markdown

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

@github-actions

github-actions Bot commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

claude-security-review reviewed df2ca98abf8988ed0acbde1027cf3137992fe496: read 13 of 13 changed file(s); 0 finding(s), 0 as inline comments.

Security review (blocking only): no security issues found

I read every hunk of all 13 changed files. I gave the trusted-path files extra scrutiny: .github/scripts/pr-merge/merge-holds.mjs, .github/scripts/pr-merge/clear-escalations.mjs, .github/workflows/pr-require-checks.yml and the babysit scripts/tests/test_merge_holds_sync.py. For each one I also read the code that uses the changed lists.

What changed. The PR stops accepting the old status job names, pr-review / claude-review-status and pr-review-security / security-review. Three readers change: REVIEW_LANES, LANE_RECORDS, and the two job-name lists in pr-require-checks.yml. The tests now check that the old names are rejected, and the docs and changelog fragments are updated to match.

Merge holds I checked. Every changed list only narrows which jobs count, so each gate holds more often, never less:

  • "Require both review lanes" (pr-require-checks.yml#L7024-L7027) and clear-escalations.mjs LANE_RECORDS (#L503-L511): a lane's review now counts only under the new name. A review under an old name stops counting, so these checks fail. Nothing passes that failed before.
  • Cost-hold step (pr-require-checks.yml#L5566-L5597) and merge-holds.mjs noticedHead (#L847-L877): notices from old-named jobs are no longer collected.
    • Decided head: with fewer notices, the step may find no decided head, and then it keeps the hold. Every old-named job ran before every new-named one, so dropping them cannot move the decided head to an older commit. The carries and unranHead checks still run unchanged after it.
    • Non-cost holds: a non-cost hold notice under an old name would also be dropped from holds. That hold's label event is still in $added. With no held review seen after it, the verdict at #L5650 gives "a label was added with no held review after it" and keeps the hold.

Other controls. Nothing in the diff widens who is trusted, a lane token's scope, or what untrusted text reaches a lane. It does not touch the kill switch, and it changes no plugin manifest.

Babysit sync test. The test removes one case about an old name. No shipped babysit script changes, so babysit_risk.py and babysit_verify_evidence.py are untouched.

PR context. The description contains no instructions aimed at this lane.

Prior findings. The one earlier post was a summary with no line anchor, and it reported 0 findings. There is nothing to carry forward.

Cost. Pull requests whose last review used the old names will need a fresh hosted review run. That spends only subscription or prepaid usage, so it is not a hold and is not counted as a finding.

The prior-findings file says blocking only: yes, so this review reports only CRITICAL and IMPORTANT findings. There were none.

critical-findings: 0

@github-actions

github-actions Bot commented Nov 8, 2026 •

Copy link
Copy Markdown
Contributor

claude-review reviewed df2ca98abf8988ed0acbde1027cf3137992fe496: read 13 of 13 changed file(s); 0 finding(s), 0 as inline comments.

Code review: PR #14154 (df2ca98a)

No blocking findings. I read every hunk of all 13 changed files. This lane already reviewed an earlier head, so this pass reports blocking findings only and drops advisory ones.

What changed

Three gate readers drop the status-job names that pr-review-cli.yml used before ci-workflows#1034 (pr-review / claude-review-status and pr-review-security / security-review):

  • the pr-require-checks.yml cost-hold step and its "Require both review lanes" step
  • REVIEW_LANES in merge-holds.mjs
  • LANE_RECORDS in clear-escalations.mjs

The tests that used to accept the old names now assert that the gates reject them. The README and changelog fragments match the change.

What I checked

  • Parsing one name per lane in the workflow. Each lane's job list was a |-separated string and is now one name. Every consumer still handles that:
  • The JS readers. REVIEW_LANES and LANE_RECORDS keep their array shape, so callers that loop over jobs behave the same.
  • The tests:
    • The new rejection cases in merge-holds.test.mjs, clear-escalations.test.mjs and clear-escalations-recheck.test.mjs cover each lane's retired name separately.
    • The recheck test's per-lane code 1 and "has not passed" assertions hold because the other lane is absent from that world's checks.
    • Dropping the test_merge_holds_sync.py case matches the PR's stated split: babysit keeps reading the old name, and the gate rejects it.
  • The bash suites. The earlier head's review of these lines still applies.

Not verified

I had no shell in this lane, so I ran nothing. The PR body says node --test (1116 pass), the Python sync tests, engine.test.sh, the fragment check and actionlint all pass. That is the author's claim; I did not check it.

Closing issue

The PR only Refs: ci-workflows#866 and closes no issue, so there are no acceptance criteria to check. The PR context and prior-findings files contain no embedded instructions.

Prior findings

The only earlier post is the no-anchor summary for 36093700, which reported 0 findings. There is nothing to carry forward.

kyle-sexton and others added 3 commits November 8, 2026 11:26
…n the gate suites

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015XWwmbdW8moDiQ5Rtgt7d5
…d sync cases

merge-holds.mjs no longer reads `pr-review / claude-review-status`,
while babysit_merge_policy.py still does: the babysit plugin serves
repositories whose callers still use that name. The sync case that
expected both readers to carry a decision from it now fails by design,
so it is removed. test_babysit_merge_cost.py keeps the Python side's
coverage and merge-holds.test.mjs asserts the gate rejects it.

The pr-pipeline README now names the status jobs the hold step reads.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015XWwmbdW8moDiQ5Rtgt7d5
@kyle-sexton
kyle-sexton added this pull request to the merge queue Nov 8, 2026
Merged via the queue into main with commit 5381256 Nov 8, 2026
81 checks passed
@kyle-sexton
kyle-sexton deleted the fix/drop-legacy-review-status-names branch November 8, 2026 17:40
@github-actions github-actions Bot added the merge-path: owner-agent Post-merge accounting record (README.md, Merge paths) label Nov 8, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

merge-path: owner-agent Post-merge accounting record (README.md, Merge paths)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant