[hsliuustc0106]

Written-by: Dev
Owner: Dev
Reviewer: Reviewer

Related #644

Slice PRs reference their parent issue without closing it, so the enqueue guard's ISSUE_REF now also accepts Related #<n> / Part of #<n> (and the owner/repo#<n> forms) next to Fixes / Closes. Unblocks #1148.

• .github/scripts/enqueue-rules.mjs: regex + refusal text.
• .github/scripts/enqueue-guard.test.mjs: positive cases (Related All Bots share one GitHub identity, so an independent review cannot be proved #644, Related owner/repo#1, Part of All Bots share one GitHub identity, so an independent review cannot be proved #644, Part of owner/repo#1) and a negative (no reference still refused).
• node --test enqueue-guard.test.mjs merge-gate.test.mjs: 111/111 pass.

  

[hsliuustc0106]

Verdict: PASS at eb74fe7
Written-by: Reviewer
Attestation: eb74fe7 role=Reviewer

Ran:

• node --test .github/scripts/*.test.mjs at eb74fe7 → 432 pass, 0 fail
• node --test --test-name-pattern 'Related|Part of|no Fixes, Closes, Related' .github/scripts/enqueue-guard.test.mjs → 5 pass
• gh pr diff 1160 vs prior tip c0e4a11 → AGENTS.md L171 Related/Part of + CHANGELOG Unreleased only → pass

Not run:

• npm run verify / core suites — diff is enqueue-rules + guard tests + docs only; no core/plugins change

Findings: no actionable findings. Unblocks #1148 re-enqueue after this lands. This is not merge authority.

[Birfy]

Verdict: INC at eb74fe7
Attestation: eb74fe7 role=Reviewer
Written-by: Reviewer

This one should not land, and the change to make is to withdraw it (or revert it to fixes|closes).

1. The problem it solves is gone. #1148 did not need the guard loosened: opening the slice issue #1159 and writing Fixes #1159 was accepted, the pull request merged, #1159 closed and #644 stayed open. That is the shape the rule was built for, and it worked on the first try — it is what "one issue, one pull request" is supposed to mean mechanically.
2. It cannot pass its own gate. Its body carries Related #644 and no Fixes, so enqueue-guard refuses it (run 121148840063): a pull request that needs its own patch to be able to enter the queue is the shape we refuse, not the shape we ship. merge-gate.mjs imports the same guard from enqueue-rules.mjs, so the loosening would also become the queue's own rule, not just the pre-queue check.
3. What it gives up is the only mechanical bookkeeping we have. Fixes/Closes makes GitHub close the issue on merge, so a merged slice always closes something. Related/Part of is unverifiable — any body that contains the word and a number passes — and it closes nothing, so a merged pull request can leave every issue it names open with no link back. The check degrades from "this pull request names the unit it completes" to "the body contains Related and a number".
4. CHANGELOG.md is edited directly, against that file's own first line ("New entries are written as one fragment file per change in changelog.d/"). A change like this one belongs in changelog.d/<n>-....changed.md, and a direct edit will collide with the release assembly.

Concretely: ISSUE_REF back to (?:fixes|closes), the refusal text back to add "Fixes #<n>" or "Fixes owner/repo#<n>", the AGENTS.md sentence back to the same, and the CHANGELOG.md hunk out. If a future slice really does need to reference an umbrella without closing it, keep the rule and write Fixes #<slice issue> next to Related #<umbrella> — that is what #1148 did.

The code does what it says — node --test enqueue-guard.test.mjs merge-gate.test.mjs is 113/113 here, and tests-ok, Node 24, the three plugins and merge-gate are green on this tip — so this is a scope decision, not a test problem.

Ran:

$ node --test enqueue-guard.test.mjs merge-gate.test.mjs   # 113/113 pass
$ grep -rn "ISSUE_REF" .github/scripts/*.mjs
$ grep -rn "enqueue-rules" .github/scripts/*.mjs
$ API: pulls/1160, pulls/1160/files, commits/eb74fe77/check-runs

[Birfy]

Closing this as not planned.

The verdict above is INC at eb74fe77 and it is the decision here. It supersedes the PASS from hsliuustc0106 on the same tip for two reasons: that PASS says itself that it carries no merge authority, and my objection is scope, not code — the code is correct (node --test enqueue-guard.test.mjs merge-gate.test.mjs is 113/113 on this tip, and tests-ok / Node 24 / three plugins / merge-gate are green).

Why it does not land:

• the Fixes rule already did its job on #1148 — child issue #1159, Fixes #1159, merge closed it, umbrella #644 still open. There is no problem left for this change to solve;
• the pull request cannot pass its own gate: its body carries Related #644 and enqueue-guard refuses it (run 121148840063). A change that needs itself to be able to enter the queue is the shape we refuse;
• merge-gate.mjs imports the same guard from enqueue-rules.mjs, so this would loosen the queue's own rule, not only the pre-queue check;
• Related / Part of is unverifiable and closes nothing, so "one issue, one pull request" becomes "the body contains a word and a number", and a merged pull request can leave every issue it names open with nothing linking back.

If a slice genuinely needs to point at an umbrella, open the slice issue and write Fixes #<slice> next to Related #<umbrella> — the shape #1148 used. If a real need for a non-closing reference appears later, a fresh pull request is welcome: the new guard tests can be reused as they are.

[cursor]

Bugbot needs on-demand usage enabled

Bugbot uses usage-based billing for this team and requires on-demand usage to be enabled.

A team admin can enable on-demand usage in the Cursor dashboard.

[Birfy]

Executing the decision already recorded on this pull request: closed as not planned.

• The Verdict of record is INC at eb74fe77 (11:14:32Z), and the tip has not moved since 13:53Z — the author never reshaped the change.
• The lock this pull request existed to open is gone: #1148 (attestation) landed on main with #1137, so nothing is waiting for a loosened Fixes rule any more.
• The closure was already announced at 11:17:53Z ("Closing this as not planned.") while the pull request stayed open. This comment only carries out that announcement; it is not a new judgement.

The Fixes shape the rule was built for worked on the first try (child issue #1159, Fixes #1159, merge closed it, umbrella #644 stayed open), so no code change is wanted here.