gate: accept MEMBER/selfReview/admin PASS when GitHub stamps CONTRIBUTOR #936

Activity

1. Birfy commented on Oct 24, 2026

   

   Birfy

   on Oct 24, 2026

   Collaborator

   More actions

   Triage: the need is real, but one of the three qualifications must not ship. Line numbers on main 7be4e9b7.

   The need, confirmed

   author_association is GitHub's stamp about the commenter's relationship to the repository, and it says CONTRIBUTOR for a login that plainly holds admin on this repository (seen on #925: the tip PASS was discarded, and selfReview did not clear it). Accepting OWNER/MEMBER/COLLABORATOR is right, MEMBER is the org bar, and a permission-API check for a repository admin is the only way to recover the truth GitHub did not put in the stamp. Keep that.

   What must not ship: selfReview qualifying other people's pull requests

   selfReview is documented in .github/agent/review-policy.json as the own-author exception — "A pull request's own author does not count unless the author is listed in selfReview". #928/#931 shipped it as exactly one narrow carve-out for that case (review-gate.mjs:166-168 on 7be4e9b7: selfOnly = selfReview.has(login(r)) && login(r) === authorLogin), with a pinned negative case. Making the same list qualify any pull request keeps the name but changes the meaning: it becomes a second reviewers list, keyed on a list the policy file defines as something else, and it makes the #931 carve-out untestable (the negative case disappears).

   Note it is also not needed for this issue: hsliuustc0106 is in reviewers, and the admin-permission path is what recovers his PASS on #925 — not selfReview.

   Requirements

   1. Association qualifies on OWNER/MEMBER/COLLABORATOR (review-gate.mjs:36).
   2. The own-author selfReview carve-out stays exactly as merged (:166-168), negative case included.
   3. A repository admin qualifies a CONTRIBUTOR, asked once per such conclusion, only for a login already in reviewers, only for a Tip+Verdict report, fail closed on 403/404/HTML — and never called when the tier needs no PASS.
   4. Nothing else moves: the Written-by check, the own-author rule (:183), the reviewers list (:184).

   Acceptance

   1. Replay of Fix mock admin @all and in-place compaction in smoke (#917) #925: a tip PASS by a login in reviewers whose association is CONTRIBUTOR and whose permission is admin → counts: true.
   2. The fix(gate): count a listed self-reviewer's PASS on its own pull request #931 negative case stays green unchanged: a selfReview login, CONTRIBUTOR, not an admin, judging someone else's pull request → counts: false, same reason text as today. The test is not rewritten to expect the new behaviour.
   3. An admin who is the pull request's author and is not in selfReview → own PASS still false.
   4. An admin who is not in reviewers → still false (the platform-admin case must not open the reviewers list).
   5. permission: admin and permissions.admin: true both count; 403, 404, a non-JSON body and a write/triage/maintain permission do not.
   6. adminsFor is not called when the result is exempt (T0/T1), nor for a report that names no tip or carries no verdict.
   7. The three scripts share decide: the same behaviour is asserted in review-gate.test.mjs, merge-gate.test.mjs and enqueue-guard.test.mjs.

   #935 currently implements requirement 2 the other way round (it deletes the own-PR restriction at review-gate.mjs:166-168 and rewrites the #931 test to expect true). It must drop that branch before it is judged; I noted the exact change on the pull request.

   Written-by: Triage
2. Birfy

   added

   agent-prPull request opened by an agentPull request opened by an agent

   area:automationScheduled tasks and automationScheduled tasks and automation

   enhancementNew feature or requestNew feature or request

   on Oct 24, 2026
3. Birfy

   added 2 commits that reference this issue on Oct 24, 2026

   fix(gate): keep the selfReview carve-out own-author only, accept a re…

   ...

   This issue will close once commit 031d1e8 is merged into the 'main' branch.

   031d1e8

   fix(gate): ask the permission API for a reviewer who is not the pull …

   ...

   This issue will close once commit f8d7ddc is merged into the 'main' branch.

   f8d7ddc
4. cursor

   added a commit that references this issue on Oct 24, 2026

   docs(gate): the permission API is not the own-author carve-out

   ...

   This issue will close once commit 381b94c is merged into the 'main' branch.

   381b94c
5. Birfy

   added a commit that references this issue on Oct 24, 2026

   docs(gate): say that selfReview and admin do not replace each other

   ...

   This issue will close once commit 2288844 is merged into the 'main' branch.

   2288844
6. Birfy

   closed this as completedin #935on Oct 24, 2026
7. Birfy

   removed

   agent-prPull request opened by an agentPull request opened by an agent

   on Oct 24, 2026
8. HugoXK

   added a commit that references this issue on Oct 24, 2026

   fix(gate): accept MEMBER/selfReview/admin PASS under CONTRIBUTOR stamp (

   ...

   This issue will close once commit c812e5d is merged into the 'main' branch.

   c812e5d