Repository navigation
Make pr_review.py and Merge Gate Item 2 Agree on an Earlier-Head Copilot Refusal #3571
Description
Summary
pr_review.py and pr-review-conduct "Merge Gate" item 2 disagree on a pull request into develop whose Copilot error refusal landed on an earlier head.
- The rule says
refusal=QUOTAorrefusal=ERRORis "the one nothing the loop does clears". That holds when the newest Copilot review is such a refusal, "on this head or an earlier one". - The tool reads an attested local pass on the current head as
coverage=localandreview_on_head=localeven then.waitexits 0, and the digest still printsrefusal=ERROR.
Failure
Take a constructed feature pull request into develop. Copilot posts an error refusal on head A. The driver pushes head B and attests a local pass on it. status prints coverage=local refusal=ERROR, wait exits 0, and the driver merges. By the rule's text the item was unmet, so the merge needed the maintainer.
This happened on #3569. The maintainer accepted that merge on 2026-10-28 and asked for this issue.
Expected
The tool and the rule state one behavior. Either the tool holds item 2 unmet under an earlier-head refusal, or the rule says what clears it. That would be an attested local pass on a later head, for a pull request into a non-default branch.
🤖 Generated with Claude Code
Activity
ptr727 commentedon Nov 1, 2026
Decision needed (decision triage, 2026-11-01)
Question: When a pull request into a non-default branch has a Copilot error or quota refusal on an earlier head and an attested local pass on its current head, should
pr_review.pyhold Merge Gate item 2 unmet, or should pr-review-conduct's Merge Gate text say the later-head attested local pass clears it?Options, recommended first:
local_coveralso refuses when the newest Copilot review on any head is a QUOTA or ERROR refusal. Makes the text literally true, but under quota exhaustion every feature PR goes to the maintainer and pr_merge.py stops auto-merging them. No work in flight.Evidence: pr-review-conduct SKILL.md:42 "every head can carry an attested local pass instead,
coverage=local"; lines 69, 74-79 "A refusal is not that coverage ... can sit on this head or an earlier one". scripts/pr_review_rounds.py:638-656local_coverrequiresnot refusing_review(pr), which checks only the newest review on the current head (line 242, docstring line 649). scripts/pr_merge.py:35 includes "local" in MET_COVERAGE and ignoresrefusal=, so since #3704 this disagreement also decides automated merges into develop. Overlaps #2180.Blocks: none
ptr727 commentedon Nov 1, 2026
Maintainer answer (decision triage, 2026-11-01): Option 1: change the pr-review-conduct Merge Gate text to say a later-head attested local pass clears item 2 on a pull request into a non-default branch, even after an earlier-head Copilot quota or error refusal. A promotion into main still always needs a Copilot round on its head. This is now work for the next burndown round.
ptr727 commentedon Nov 1, 2026
Released (2026-11-01): worktree and branch removed, so this claim holds no files. Its fix merged to develop in PR #3775. The issue closes when that is promoted.
Backlog burndown claim (round 1, group G1). Branch:
feature/burn-3758. Issues: #3758, #3571, #2180, #3732, #3713, #3726.Predicted file set:
scripts/pr_review.py,scripts/pr_review_rounds.py,scripts/pr_review_copilot.py,scripts/reviewers/copilot.jsontests/test_pr_review.py,tests/test_pr_review_copilot.py.agents/skills/pr-review-conduct/SKILL.mdand its generated trees (.github/skills/pr-review-conduct/,.claude-plugin/fleet-skills/skills/pr-review-conduct/,.claude-plugin/fleet-skills/.source-digests/pr-review-conduct)drive-prgenerated trees (included section),GOVERNANCE.md"PR Review Etiquette",.github/copilot-instructions.mdRead Copilot's Excluded-Files Refusal as Its Own Kind and Align Merge…