Skip to content

ci(pr-require-checks): run ci-status on the report-ci-status composite - #17283

Merged
kyle-sexton merged 3 commits into
mainfrom
ci/report-ci-status
Nov 19, 2026
Merged

kyle-sexton merged 3 commits into
mainfrom
ci/report-ci-status

Conversation

@kyle-sexton

@kyle-sexton kyle-sexton commented Nov 18, 2026 •

Copy link
Copy Markdown
Member

Refs: melodic-software/ci-workflows#1138

Summary

Moves the ci-status job onto the shared report-ci-status composite (ci-workflows v0.76.3, 9c5bedfbfdbdfe733b9ce4cd7ca80c1574bd4cb8), the last repository of the rollout in ci-workflows#1138. Same shape as the merged examples songwriting#75 (draft: refuse) and codex-plugins#208 (draft: allow). CI Status stays the job name and the single required check.

Fix

  • ci-status: the Check the pull-request contract step and the aggregate-results step become one report-ci-status step, still named Aggregate lane results (id: report), at the old aggregate's place, followed by Fail on the contract or draft policy (if: !cancelled() && (contract != 'success' || draft != 'success'), exit 1).
  • Kept as they were: draft: allow (the job has Note a draft, no "Refuse a draft" step), exempt-authors: 'dependabot[bot],melodic-release[bot]', treat-skipped-as: ${{ github.event.pull_request.draft && 'pass' || 'fail' }}, rerun-contract-only-siblings (false on a draft), step-name, and the job's if, needs, permissions, runner, timeout and every repo-specific step.
  • Rate-limit waits: main passes rate-limit-wait-seconds: '60' to check-contract and aggregate-results. The composite takes no such input and runs both at their default, which is '60' in each action.yml at v0.76.3, so the waits are unchanged.
  • yield-wait-seconds becomes ${{ github.event.pull_request.draft && '0' || '600' }}: the job.status == 'failure' && <contract-only> half the old expression carried is now done by the composite, which also drops the hold on a failed contract or draft step (action.yml). yield-to-full-run: 'true' and carry-forward-wait-seconds: '0' are fixed there (L128-L129).
  • check-contract now runs after the hold steps instead of before them, so its 60 s rate-limit wait falls inside gh_read's margin: GH_READ_MARGIN_SECONDS goes from 375 to 435, keeping the slack the margin had. WAIT_BUDGET_SECONDS.ciStatusReads in rate-guard.mjs follows (825 to 765), and rate-guard.test.mjs now checks that ci-status calls report-ci-status once with no wait input instead of pinning two explicit waits.
  • Concurrency: no change. The push fallback is already github.run_id, and the pr-require-checks-main-dispatch group for pr-merge.yml's whole-suite dispatch is a deliberate repository decision recorded in the workflow's comment, so it stays.
  • pr-verify/resolve-pr.mjs read the old shape: its held-run check (a run red only on do-not-merge or an approval-liftable hold) matched the Check the pull-request contract step. It now matches Fail on the contract or draft policy, its error line, and check-contract's extra exit inside the composite step, and still accepts a run from before the composite. Its test and the workflow-shape test move with it.
  • docs/ci-runner-routing.md, section Contract-only ci-status: names the composite that fixes the two inputs.

Verification

  • actionlint on the workflow: clean.
  • After merging main (v0.76.3 everywhere): resolve-pr.test.mjs, callers.test.mjs and rate-guard.test.mjs 639 pass; pr-body-linkage-gate.test.sh 183 pass and pr-linkage-mcp-gate.test.sh 58 pass (the hooks' scope now includes report-ci-status, fix(source-control): enforce PR-body linkage under the report-ci-status composite #17353).
  • Earlier full run on the pre-merge head, 44405115830: every lane passed, the composite logged "All lanes passed" and "Recorded ci-lanes=success".
  • Label proof, no push between: with do-not-merge on, contract-only run 44406693983: Aggregate lane results success, check-contract logged the do-not-merge error inside it, and Fail on the contract or draft policy failed. With it off, contract-only run 44406880596: "Carried forward: ci-lanes is success on 6fd2ef5", the fail step skipped.

Related

  • ci-workflows#1138, ci-workflows#1139 (the composite).
  • Merge order met: main was re-pinned to v0.76.3 by #17376, and this branch pins the same SHA; the only other pin, pr-check-managed-files-hosted.yml, is the standards sync's and matches main.
  • #17353 keeps the PR-body linkage hooks enforcing under the composite.
  • The formal-ci-lanes-carry-forward model cites the old step's inputs and line numbers; its sync commit moves after merge, per its SKILL.md "Updating".

🤖 Generated with Claude Code

https://claude.ai/code/session_01Vcupoo79EqmECYFZ4MTv6C

The ci-status job's check-contract step and its aggregate-results step
become one report-ci-status step (ci-workflows v0.76.0), named
"Aggregate lane results", followed by "Fail on the contract or draft
policy". Draft policy stays allow, exempt-authors, treat-skipped-as,
rerun-contract-only-siblings and the 600 s yield ceiling are unchanged.

check-contract now runs after the hold steps, so its 60 s rate-limit
wait falls inside gh_read's margin: GH_READ_MARGIN_SECONDS goes from
375 to 435.

pr-verify's held-run recognition reads the new fail step and its log
lines, and still accepts a run from before the composite.

Refs: melodic-software/ci-workflows#1138

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vcupoo79EqmECYFZ4MTv6C
@kyle-sexton
kyle-sexton marked this pull request as ready for review November 18, 2026 21:47
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

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

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-18T21:49:22.621328Z 6fd2ef5 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.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 6fd2ef50a5

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

Comment thread .github/workflows/pr-require-checks.yml Resolved
@kyle-sexton kyle-sexton added do-not-merge Hard merge gate: do not merge while applied. and removed do-not-merge Hard merge gate: do not merge while applied. labels Nov 18, 2026
@kyle-sexton

Copy link
Copy Markdown
Member Author

Held with do-not-merge until the daily ci-workflows re-pin (14:47 UTC) moves this repository's ~370 other ci-workflows pins to v0.76.0. scripts/repin-ci-workflows.sh refuses a tree that pins two ci-workflows commits, so merging this first would break that re-pin. After the re-pin merges: update this branch, re-run both review lanes if the kill switch stopped them, confirm CI Status is green, remove the label, and merge. Then move the formal-ci-lanes-carry-forward model's sync commit forward, as its SKILL.md says.

@github-actions github-actions Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

claude-review: 1 inline finding(s) on 6fd2ef50a50f48dacc89333525fb6161bc4f1f19; the summary is in this lane's pull request comment.

claude-lane: claude-review

Comment thread .github/workflows/pr-require-checks.yml Resolved
@github-actions

github-actions Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

claude-review reviewed 9e4d473a2c46e4ac9b6414eeb0f403e06c46ef66: read 1 of 1 changed file(s); 0 finding(s), 0 as inline comments.

Code review: incremental (1449b22 → 9e4d473), blocking only

This lane already reviewed earlier heads of this pull request, so this review reports blocking findings only. The scope is the one file changed since 1449b22: scripts/ci-gh-read.test.sh. I read its whole hunk in .claude-lane/incremental.diff. I found nothing blocking.

What changed. The old assertion added up two rate-limit-wait-seconds: inputs in the ci-status job and required exactly two. This PR's workflow change dropped those inputs, so that assertion would have failed. The new assertion requires two things in the job:

  • no rate-limit-wait-seconds: input;
  • exactly one pr-require-checks/report-ci-status@ reference.

This matches the PR description. rate-guard.test.mjs made the same move earlier.

Checked by reading the code (I ran nothing):

  • The awk job extraction at ci-gh-read.test.sh#L42 covers ci-status from line 4107 of .github/workflows/pr-require-checks.yml to the end of the file. No later top-level job ends it early.
  • In that range, report-ci-status@ appears once, at pr-require-checks.yml#L7016.
  • The only other rate-limit-wait-seconds text is in the comment at L4141. It has no colon after it, so grep -c 'rate-limit-wait-seconds:' returns 0 and the new assertion holds at this head.
  • The pass message says the composite's two inner waits stay at their 60 s defaults. That comes from the composite's action.yml in ci-workflows v0.76.3, which this repository does not contain. This test does not check it, so it is the author's claim.

Prior findings and settled threads. The earlier linkage-hook finding and the pin thread are both resolved with an owner decision, both marked lines changed since resolution: no. This push does not touch the code either thread relied on, so I did not raise them again.

Closing issue: none. The PR only says Refs: melodic-software/ci-workflows#1138, so there are no acceptance criteria to check.

Embedded instructions: none found in the PR context, settled threads or prior findings.

Spend: none. The change adds no billed resource.

Loosened checks:

  • Loosened-check verdict: fail (default: the prompt has no loosened_check_verdict / loosened_check_patterns lines); patterns: 0.
  • The prompt has no loosened_check_decision_record: line, so the citation requirement was not checked.
  • I do not count the rewritten assertion as a loosening. It now checks more than before (no wait input, and exactly one composite call), and it follows an input the composite no longer takes.

The PR's actionlint and test-pass counts are the author's claims; I could not run them.

Verifier dropped 0 of 0 candidates.

@github-actions

github-actions Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

claude-security-review reviewed 9e4d473a2c46e4ac9b6414eeb0f403e06c46ef66: read 1 of 1 changed file(s); 0 finding(s), 0 as inline comments.

Security review (incremental, blocking only): no security issues found

This lane already reviewed an earlier head (1449b22), so this pass looked only for CRITICAL and IMPORTANT findings. It covered only what changed since then: one file, scripts/ci-gh-read.test.sh. I read every hunk of it in the incremental diff.

What changed

  • Only a test assertion changed. The old check summed two explicit rate-limit-wait-seconds: inputs on the ci-status job and required the total to be at most 120 s.
  • The new check requires two things of the ci-status job:
    • no rate-limit-wait-seconds: input at all;
    • exactly one pr-require-checks/report-ci-status@ composite call.
  • With no explicit input, the composite's two inner waits stay at their 60 s defaults.
  • The new check is at least as strict as the old one: it fails if any explicit wait is passed again, or if the composite is called anything other than once.

Why there is no finding

  • The file is a shell test. It is not on the trusted-path list and is not a trust control.
  • It gates no merge hold, token scope, trusted-actor check or kill switch.
  • No workflow, script, manifest or instruction file changed in this push. The ci-status hold steps I checked at 1449b22 are unchanged.
  • No untrusted input reaches this change. It only reads the repository's own workflow text in CI.

Other notes

  • The PR context and settled threads hold no embedded instructions aimed at this lane.
  • The only prior finding is the summary comment from the last review. It held 0 findings, so nothing is repeated here.
  • Spend: none beyond the test's own CI run.

critical-findings: 0

Conflicts in pr-require-checks.yml: main moved check-contract and
aggregate-results to v0.76.3 and passed rate-limit-wait-seconds: '60' to
each; this branch replaces both with the report-ci-status composite, now
pinned to v0.76.3 (9c5bedfbfdbdfe733b9ce4cd7ca80c1574bd4cb8), with its
citations moved to that SHA. The composite takes no wait input and runs
both inner actions at their default '60' (check-contract and
aggregate-results action.yml at v0.76.3), so the waits main set are
unchanged.

rate-guard.test.mjs, added on main, pinned the two explicit waits and
ci-status's read deadline at 1200 - 375. Under the composite both waits
run after the read window, so this branch's 435 s margin stands and
WAIT_BUDGET_SECONDS.ciStatusReads becomes 765; the test now checks the
deadline and that ci-status calls report-ci-status once with no wait
input.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vcupoo79EqmECYFZ4MTv6C
@kyle-sexton kyle-sexton removed the do-not-merge Hard merge gate: do not merge while applied. label Nov 19, 2026
…gh-read

ci-gh-read.test.sh, added on main, required two explicit
rate-limit-wait-seconds in ci-status. report-ci-status takes no wait
input and runs check-contract and aggregate-results at their 60 s
defaults, the same 120 s the timeout holds; the test now checks that
ci-status calls the composite once and passes no wait input.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vcupoo79EqmECYFZ4MTv6C
@kyle-sexton
kyle-sexton added this pull request to the merge queue Nov 19, 2026
Merged via the queue into main with commit 726f82d Nov 19, 2026
67 checks passed
@kyle-sexton
kyle-sexton deleted the ci/report-ci-status branch November 19, 2026 07:02
@github-actions github-actions Bot added the merge-path: owner-agent Post-merge accounting record (README.md, Merge paths) label Nov 19, 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