Repository navigation
π docs(v6): behind fork leaves the lane front at once; serialized lane readiness row from the dd406b3 re-run - #14939
Conversation
β¦e readiness row from the dd406b3 re-run operator-reference.md said a behind fork pull request leaves the front once the front timeout runs out; forkUpdateOrLeave leaves at once and posts one comment, which the live exercise on hivecommons#7563 observed. The readiness row now records the full re-run at dd406b3: steps 1-3 and 5-7 pass, step 4 (hold at the front) failed and is tracked in hivecommons#14934. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_0113KwWU4TkJ8wfy3XjusyGo Signed-off-by: Danathar <Danathar@users.noreply.github.com>
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
|
@Danathar 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 |
|
Hive filed review findings that do not block this PR as backlog items:
Cap: 10 backlog item(s) per PR per day. |
There was a problem hiding this comment.
Intent alignment β read head cd420aa. The fork paragraph in src/docs/operator-reference.md:771-779 matches forkUpdateOrLeave (src/pkg/mergelane/lane.go:729-752: update only when viewerCanUpdateBranch && baseSync, else ensureForkWaitComment + leave at once, reason strings at :736-746) and the readiness row is a faithful status update. One added sentence overstates what the code does:
src/docs/operator-reference.md:780-782 β "A queued pull request whose new head its queue approval no longer covers (for example, the update also changed files) must be queued again." The parenthetical implies a tip-only merge keeps the approval. In the code the queue approval covers exactly one commit: latestHiveQueueApproval records review.GetCommitID() (src/pkg/github/automerge/automerge_sweep.go:2820), and both queueApprovalStillHolds (:2041-2043) and the queued sweep path (:2657-2661, "approval head changed since approval β re-queue required") compare SHAs, with no content/tree exception. I found no code that re-approves a tip-only merge. If sandbox#13 merged without a re-queue it was on a path that is not the /hive queue approval, which the doc should not generalise. Suggested wording:
rejoins the lane and merges at a later turn. A pull request merged through a
`/hive queue` approval must be queued again after any new head, including a
merge of the tip: the approval covers exactly the approved commit.
(If there is a tip-only exception I missed, please point at it and I'll withdraw this.)
β hive: agent=adjudicator-3 backend=copilot model=claude-fable-5.1 copilot=1.0.88
|
hive/check-report (head π Hive Agent: β hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88 |
There was a problem hiding this comment.
Read against src/pkg/mergelane/lane.go on v6; the new wording matches the code.
- "leaves the front at once" β
forkUpdateOrLeave(lane.go:729-750) callsr.leave(...)directly whenViewerCanUpdateBranchis false orContributorBaseSyncAllowedis off; there is no deadline wait on that path. - The quoted reason string and the three
viewerCanUpdateBranch/base_syncsuffixes are the literal strings atlane.go:736-745. - "posts one comment" β
ensureForkWaitComment(lane.go:753) checksIssueCommentsContain(forkWaitCommentMarker)before creating one, so it is idempotent per PR. - "head its queue approval no longer covers must be queued again" β consistent with the authorization re-check per head (
gate_test.go:152"lgtm approval covers another head" β waiting).
The v6-readiness.md row edit is evidence bookkeeping and links the issues it names. Looks correct to me.
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 No conflicting open PRs or same-file overlaps were found. |
|
Deferred to shared-CI incident #14976: the red check here failed on a π Hive Agent: β hive: agent=scanner backend=copilot model=claude-fable-5.1 copilot=1.0.88 |
A /hive queue approval records the reviewed commit SHA (latestHiveQueueApproval) and both queueApprovalStillHolds and the queued sweep compare it by exact SHA, so any new head, including a plain merge of the tip, needs a re-queue. Drop the parenthetical that implied a tip-only merge keeps the approval. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01BQYnPGFJx81c8JXR3o1sg1 Signed-off-by: Danathar <Danathar@users.noreply.github.com>
|
Re the adjudicator-3 intent-alignment finding ( |
|
stage 060-reviewed β 050-review-pending: check-pending |
Summary: Two doc updates from the v6 serialized-lane live exercise on #7563 (evidence).
operator-reference.mdsaid a behind fork pull request leaves the lane front "once the front timeout runs out". The code (forkUpdateOrLeave,src/pkg/mergelane/lane.go:729) leaves at once and posts one comment. The live run saw exactly that: Case 7: fork PR G (behind, base_sync off)Β Danathar/hive-lane-sandbox#13 left 2.6s after reaching the front.v6-readiness.md: the serialized merge lane row now reflects the full re-run atdd406b3. Steps 1β3 and 5β7 have evidence. Step 4 (hold at the front) failed and is filed as π [v6] serialized lane: aholdlabel on the front PR never releases the front, so the lane is blocked until the front timeoutΒ #14934. The row stays π¨ Partial and unchecked.Changes
src/docs/operator-reference.md("Fork pull requests"):viewerCanUpdateBranchis true andreview.contributor_prs.base_syncis on. Otherwise it leaves the front immediately with the recorded reason, and Hive posts one comment.src/docs/v6-readiness.md: the serialized lane row's status, with links to the evidence comment, π [v6] serialized lane: aholdlabel on the front PR never releases the front, so the lane is blocked until the front timeoutΒ #14934, π [v6] contributor-lane auto-merge bypasses the serialized lane on hive-serialized reposΒ #14935 and π [v6] serialized lane: a 403/405/422 from the lane's merge call is recorded as an uncertain effect, so the front is stuck until the front timeoutΒ #14936. The earlierf55b5fe66history is kept.Evidence for the operator-reference change
front-enter; 12:29:42Zfront-exit"fork pull request is behind the target branch and Hive is not allowed to update it: viewerCanUpdateBranch is false and review.contributor_prs.base_sync is disabled". This was on a 60m front timeout.hive:lane-fork-waitcomment.Docs only; no code or tests change.
π€ Generated with Claude Code
https://claude.ai/code/session_0113KwWU4TkJ8wfy3XjusyGo