Release PR metadata rejected when it should be exempt, and @auto-review ignores user comments directing approval #287
Labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
Slipstream/auto_review#287
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Summary
Two recurring, related defects in @auto-review behavior, both reproduced again on Slipstream/eventcore#396 (see Slipstream/eventcore#396 (comment)). It should not have taken multiple tries to get an approval. Auto-review should have accepted the PR on the first pass; and even if it didn't, it should have approved after the first user comment saying the metadata is acceptable.
Problem A: Release PR title/description rejected when it should not be
Release PRs (e.g. title
chore(release): v0.8.1, with a body that mirrors the title) are being failed under the PR metadata quality pre-merge check. On PR #396 the review reported:This is wrong for release PRs.
chore(release): vX.Y.Zis a valid, conventional title for an automated release, and a body that restates the release version is acceptable. The metadata check needs to recognize release PRs and exempt them from the "needs a descriptive summary / explain why" requirements, rather than treating them like normal feature/fix PRs.This has been raised before; it is recurring and needs a durable fix (not another one-off).
Problem B: @auto-review does not respond naturally to user comments directed at it
When a user replies to @auto-review correcting it — e.g. on PR #396 the user wrote:
— auto-review should treat that as an authoritative human instruction and re-evaluate / approve accordingly. Instead it took multiple rounds to get an approval, and the agent does not appear to be incorporating the user's prior comments ("as I told you" implies it had already been told once). Directed comments should:
Acceptance criteria
chore(release): vX.Y.Zand a release-style body passes the PR metadata quality check on the first review pass (no human intervention required).References