Repository navigation
π fix: preserve supersession decisions across stage reconciliation - #14821
thatguywob wants to merge 3 commits into
Conversation
Keep a supersession cause label separate from CI and author repair state, retain escalation evidence in tracker audit replay, and expose the closer and keep-open reason even when a stale record still says 071/072. Allow trusted maintainer decisions to close superseded contributor PRs without a branch push. Detect held duplicates without writes, allow ancillary file differences only with implementation overlap, and skip closed contributor queue snapshots. Hive-Run: hivecommons#14814 Hive-Plan: issue-14814-supersession-sweep-and-stage-tracker Hive-Spec: issue-14814#fleet-generic-fix Co-authored-by: clubanderson <407614+clubanderson@users.noreply.github.com> Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
|
@thatguywob this PR is back in your court ( Failing checks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Your relay session received this as a |
Persist supersession park labels in the dashboard sweep fixture and assert the second pass makes no additional writes. Preserve slash-containing label names in the GitHub fixture so reopened-issue cleanup is observed. Signed-off-by: thatguywob <thatguywob@users.noreply.github.com>
|
@thatguywob this PR is back in your court ( Failing checks:
Reply here or push a new head to this branch (same sign-off); CI and the review swarm re-run automatically and the PR returns to the normal path. Queued in your relay fix queue (0 waiting): fixes are handed out before new work as soon as you have capacity. |
There was a problem hiding this comment.
CI red is this diff's own test fixture, and the fix is one case line.
test (dashboard 1/4) and test (dashboard shuffle 3/3) fail in TestReviewQueue_SweepDetectionWithoutAutoClose (src/pkg/dashboard/api_review_queue_superseded_test.go:414,430):
unexpected GitHub request: GET /repos/myorg/repo1/issues/2
sweep: β¦ err=parking superseded PR repo1#2 for a decision: GET β¦/repos/myorg/repo1/issues/2: 404
Mechanism, read at 30a71c9: the test now expects the sweep to park the contributor PR (labelWrites == 1, needs-human,hive/supersession-pending), so trySweepSupersededPR β parkSupersededPR (pr_supersession_resolve.go:145) β AddLabels (served by the fake) β sink.Parked(...) β stage_facts_sink.go:111 Issues.Get(repo1, 2), which the fake server's default: branch 404s. The coverage job only fails downstream of those shards ("sliced=failure"). Adding a case "GET /repos/myorg/repo1/issues/2": that returns {"number":2,"labels":[β¦current labelsβ¦]} should turn both shards green; the test (dashboard β¦) jobs on v5 do not run this path because v5 never parked in this scenario.
Also read, and looks correct to me: derive_pr_classify.go:329 derives 085/supersession-decision-pending from needs-human + hive/supersession-pending unless hive/keep-open; both ApplyStageLabels gates in automerge_stage_labels.go:810,827 and the stageengine DCO/route hops skip that reason, so CI, rebase and DCO reconciliation no longer evict the park β matching #14814. releaseStaleSupersessionPark removes both labels when detection stops. No textual conflict with #14822 (both touch pr_unpark_decision.go; git merge-tree is clean).
I have not approved, merged, or changed anything.
β hive: agent=reviewer backend=copilot model=claude-fable-5.1 copilot=1.0.88
There was a problem hiding this comment.
Read the diff against v5 plus pr_supersession_sweep.go, pr_supersession_resolve.go, pr_unpark_decision.go:36-80, derive_pr_classify.go and the dashboard tracker. The design holds together and matches #14814 Β§1β3 and Β§5; two things a maintainer should look at before this lands.
1. test (dashboard 1/4) and test (dashboard shuffle 3/3) are red at head 30a71c9. I could not read the job logs from here, so I cannot say whether the cause is in this diff, but both shards cover the package this PR changes most (stage_tracker.go, api_review_queue_superseded.go, worksource_identity.go), so it is worth triaging before anyone reads further.
2. A non-superseded /hive decision reject on a hive-authored PR is now swallowed (src/pkg/github/pr_unpark_decision.go:70-80, pr_supersession_decision.go:19-22). The close-decision branch runs before the selfAuthored switch, and when detectSupersession finds no live closer it replies "No merged PR currently supersedes this PR's closing claims; the PR stays open" and returns true β the comment is marked answered and never reaches the existing changes-requested path. The doc comment on supersessionCloseDecision says other text "continues to the existing changes-requested path", but a bare reject does not. Before this PR, reject on a hive PR at 085 kicked the author with that word; after it, the owner gets a "stays open" reply and no fix kick. Low severity, but the reply should at least tell the owner what will act (close it themselves, or give an instruction).
What I verified and have no concern about:
- The close path is gated by
opts.TrustedMerger(pr_unpark_decision.go:50-52) and by a freshdetectSupersessionwithcandidate.Closing(pr_supersession_decision.go:38-47); it deliberately skips the file-overlap and size guards, which is right for an explicit human decision. releaseStaleSupersessionParkis only reached from the!detectedbranch and underhive/keep-open(pr_supersession_sweep.go:235-237), so it cannot flap againstescalateSupersededPR's park on alternating passes.- The held path only adds reads:
detectSupersessionruns, then returns before any comment/label/close (pr_supersession_sweep.go:185-192). ReasonSupersessionPendingshort-circuits both the human-route and DCO/fixing downgrades inderive_pr.go:249,262andautomerge_stage_labels.go:810,827, which is the sticky-085 behaviour Β§2 asked for.
Scope note: the worksource_identity.go closed-snapshot skip is not something #14814 asks for. It is small and plausible, but it rides on a Fixes #14814 PR with no issue of its own; fine if the maintainer is happy with that.
β hive: agent=reviewer-2 backend=copilot model=claude-fable-5.1 copilot=1.0.88
Open PR overlap checkBase branch: Conflicts are checked with every PR merged onto the current tip of
|
|
Supersession sweep found that this PR's claimed issue is already closed by another merged PR.
Decision needed (@hivecommons): #14823 already merged and closed the claimed issue, but this PR touches files the merged PR did not. Recommended: keep open until the extra changes are reviewed, or close as superseded if they are not wanted. Another option: merge anyway. Files the merged PR did not touch: This comment is edited in place by the supersession sweep; it is not duplicated. |
|
PR coverage cross-reference: this PR already resolves #14868 through the closed-snapshot skip in A maintainer should add β hive: backend=codex π Hive Agent: |
|
Two review-backlog findings filed against the hive's adopted copy (#14862, closed unmerged) apply to this branch's diff and not to
π Hive Agent: β hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88 |
|
/hive decision supersede now: adopt #14821 in supersede mode, preserving the author's commits and sign-off; do not wait for the grace deadline Rationale: the contributor is offline; adoption in supersede mode keeps the work moving (#13975). Decided by the hive at @clubanderson's request as the owner's delegate under the owner-delegate policy (rule |
|
@thatguywob the hive has adopted this PR ( |
|
Closing: #14979 (the hive's adoption of this PR, with @thatguywob's commits) has merged. Thank you! |
β¦rge queue A PR parked at 085 because a person has to act on the PR itself (rerun CI, approve workflow runs, resolve a GitHub permission, answer review findings or red checks still open on its head) is not queued, merging or merge-ready. Its chevron now lights PR in progress β "Waiting for @who Β· rerun shared ci on GitHub", person icon kept β and the Merge queue column stays empty. Only rows actually queued / merging / merged, or a merge-ready PR (review approved, CI green) parked on a merge decision, sit in Merge queue. A bare 085 with no track evidence keeps the gate (#14821). - stageTrackerItemColumn (Go) / stageTrackerAnchor (JS): new stageTrackerPersonActsOnPR predicate, mirrored on both sides. - Unauthorized word/caption: "Waiting for @author Β· <step> on GitHub" (a hive / bot / clanker author yields "a maintainer"). - Nested PR rows carry unauthorized_step so the word matches. - Tests: Go table (#16524 shape and its twins) + rows fixture. Fixes #16592 Signed-off-by: Andrew Anderson <andy@clubanderson.com>
β¦rge queue (#16593) * π Activity Monitor: person-owed PR steps light PR in progress, not Merge queue A PR parked at 085 because a person has to act on the PR itself (rerun CI, approve workflow runs, resolve a GitHub permission, answer review findings or red checks still open on its head) is not queued, merging or merge-ready. Its chevron now lights PR in progress β "Waiting for @who Β· rerun shared ci on GitHub", person icon kept β and the Merge queue column stays empty. Only rows actually queued / merging / merged, or a merge-ready PR (review approved, CI green) parked on a merge decision, sit in Merge queue. A bare 085 with no track evidence keeps the gate (#14821). - stageTrackerItemColumn (Go) / stageTrackerAnchor (JS): new stageTrackerPersonActsOnPR predicate, mirrored on both sides. - Unauthorized word/caption: "Waiting for @author Β· <step> on GitHub" (a hive / bot / clanker author yields "a maintainer"). - Nested PR rows carry unauthorized_step so the word matches. - Tests: Go table (#16524 shape and its twins) + rows fixture. Fixes #16592 Signed-off-by: Andrew Anderson <andy@clubanderson.com> * π trackerrow: sync shared tracker-row-anchor.js with the spoke copy Signed-off-by: Andrew Anderson <andy@clubanderson.com> --------- Signed-off-by: Andrew Anderson <andy@clubanderson.com> Co-authored-by: hivecommons-hive[bot] <280983584+hivecommons-hive[bot]@users.noreply.github.com>
Fixes #14814
Supersession notices could leave closed-issue/open-PR rows looking like ordinary author repair work. Keep a
hive/supersession-pendingcause label at 085 across CI, rebase and DCO stage reconciliation, including migration of existing sweep parks. Clear it when the cause disappears while preserving unrelated human parks./hive decision rejector/hive decision close superseded by #N, with fresh closing-claim/merged-closer validation and a comment, without pushing to contributor branches.Validation: added regression tests for the stage engine and automerge adapter, tracker rendering/actions and audit replay, trusted close commands and unauthorized/held cases, supplemental-file closure and size protection, park cleanup, and closed queue snapshots.
gofmt -landgit diff --checkare clean. Tests were not run locally, as required by AGENTS.md; CI supplies the test verdict.β hive: backend=codex
π Hive Agent:
contributor| SHA:0d69dd4ed