Skip to content

Hive-authored PR superseded by another merged PR parks at needs-decision instead of closing itself; adopted originals keep stale 085 #15010

Closed

Description

@clubanderson

Observed (hive's own hosted spoke, 2026-10-28)

Case A — superseded hive PR parked at needs-human. #14965 (scanner, Fixes #14952) was opened concurrently with #14957, which also fixed #14952. #14957 merged → GitHub closed #14952 → #14965 became conflicting. The review swarm then returned requires_human ("should #14965 be closed in favour of #14957?") and the PR sat at stage/085-needs-decision for >1h until the owner clicked close-superseded. Owner: "if this is from a hive agent it should go straight through — why is this stuck in needs human?"

Case B — adopted original still wears 085. #14975 (thatguywob, docs for #14972) was adopted at 15:54Z (hive:adopted … mode=supersede superseded_by=14980) and #14980 (App-authored, same head de64cb1) carries it through CI/Review. 38 min later #14975 still carried stage/085-needs-decision and the Activity Monitor showed "Needs human decision" while the adoption comment claimed stage/074-adopted-by-hive. Same pattern earlier that day: #14947 → #14977 — and the duplicate-fix sweep then closed the adoptee #14977 as a duplicate of #14947.

Owner's stage rule (12:03 ET): "If it needs adoption then it should go back to the CI/Review column as adopted. Anything in merge gate that needs a human, and then we decide and the PR needs to be reworked, goes back to a different column — it LEAVES merge gate."

Root causes

  1. No rule anywhere decided "hive PR whose issues were all closed by another merged PR" — the supersession sweep only reported the refused rebase once and left the PR for a person; the self-authored lane dispatched a review that could only ask the human question.
  2. stage/085-needs-decision never yielded to a live adoption: the 085 label was re-derived from the stale requires_human verdict each pass, and the SPA had no 074 → CI/Review word.
  3. The duplicate-fix sweep did not recognise an adoption pair (Supersedes #N in the App PR body) and treated the adoptee as a duplicate.
  4. The open-time dup guard (findDuplicatePR) exists in the relay hive-open-pr path; why it did not refuse 🐛 fix(dashboard): name the blocking PR on tracker rows and keep webhook unblocks from being overwritten #14965 could not be reconstructed (logs rotated). Tracked here, not fixed.

Expected

  • A hive-authored PR whose closing targets are all CLOSED, one of them by a different merged PR, and that is conflicting (or whose files that PR already covers) is closed by the hive itself as hive/superseded with a comment naming the superseding PR — before any review cycle, in the supersession sweep, and immediately in close-on-merge of the superseding PR. Human-authored PRs get the note and keep the decision. Knob governor.auto_close_superseded, default on.
  • Adopt-supersede clears 085 on the original in the same pass; the tracker shows "Adopted → #N" in the CI/Review column, never "Needs attention"; the original is closed as superseded when the adoptee merges.
  • Rework decisions leave the merge gate; close-superseded/wontfix → closed column.

Activity

  1. added
    help wantedDenotes an issue that needs help from a contributor. Must meet "help wanted" guidelines.
    on Oct 28, 2026
  2. github-actions commented on Oct 28, 2026

    @github-actions
    Contributor

    @clubanderson this issue was closed by @hivecommons-hive[bot] as completed.

    If this was closed by a merged PR, the fix only takes effect once it ships in a tagged release and your deployment updates to it — check your running version against the fix before concluding the fix failed.

    If you are still seeing the problem, comment /reopen and it will be reopened for you — you do not need write access. Please add what you are still observing (running version, logs, steps) so it is not closed the same way again.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    help wantedDenotes an issue that needs help from a contributor. Must meet "help wanted" guidelines.

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions