Repository navigation
[selfreport] unclassified_error_repeated: stage labels: adopted PR sits with nothing the hive can push; escalating to nee… #14466
Description
A hive hit an unusual runtime error (class unclassified_error_repeated) and is reporting it so the same defect across the fleet lands in one issue.
Evidence
| Field | Value |
|---|---|
| hive | hosted-projectbluefin-knuckle-gjvq |
| hive name | bluefin |
| spoke | bluefin |
| anonymous instance | 0f7701746ac14f88 |
| owner(s) | @castrojo, @clubanderson, @Danathar, @hanthor, @joshyorko |
| hive version | 5.186.4-1494-gfcf0e7887 |
| hive commit | fcf0e78 |
| cluster | — |
| mode | busy |
| component | agent-runtime |
| scope | stage labels: adopted pr sits with nothing the hive can push; escalating to needs-decision |
| message | stage labels: adopted PR sits with nothing the hive can push; escalating to needs-decision |
| count + window | 9 in 10m0s |
| first seen | 2026-10-26T23:02:50Z |
| reported at | 2026-10-26T23:50:11Z |
| repos | — |
| owners | — |
| repo | — |
| ref | — |
| self-recovered | no |
Affected hives (1):
- bluefin (hosted-projectbluefin-knuckle-gjvq) (instance 0f7701746ac14f88) — open — owner(s): @castrojo, @clubanderson, @Danathar, @hanthor, @joshyorko — seen 1, last 2026-10-26T23:50:11Z
seen_count: 1
Diagnostics (scrubbed)
WARN start=d07eebf0 repo=utah-packages pr=662 head=667f04d04d7fdb66478ce67c8f27b991a6f37ee4 reason=adoption-stalled:fork-rehome-refused head_since=2026-10-26 19:38:27.663748902 -0400 EDT noted=false
Activity
clubanderson commentedon Oct 27, 2026
Triage (hub admin, read-only):
What it refers to. The 9 repeats in 23:02-23:12Z on the bluefin hive are the adopted-PR stall escalation in
src/pkg/github/automerge/automerge_contributor_adopt.go(theadoptionStalledReasonbranch that writes<!-- hive:adoption-stalled ... -->and moves the PR tostage/085-needs-decision). The matching PRs are projectbluefin/utah-packages#660 (23:02:51Z,hive-signoff-attribution, the first-seen match), utah-packages#662 (fork-rehome-refused), utah#949 (check-failure), chairlift#638 and testsuite#1028 (hive-signoff-attribution). Each got exactly one stall note and now sits at 085 for a maintainer, which is the designed outcome.Not the same defect as the adopted-source-PR-stays-open issue being fixed today. That one is about a source PR left open after its hive adoption copy merged. These PRs were never superseded; the hive reached a head it cannot push (sign-off attribution, refused fork re-home, failing checks) and escalated on purpose. No link needed.
What is actually wrong here is only that a designed escalation is logged at WARN and the
unclassified_error_repeatedclassifier promotes it to a fleet self-report once several contributor PRs stall in the same 10 minutes. Suggested fix: log the escalation at INFO (or list the message as an expected class in the self-report classifier). Also note this issue was filed with no labels at all (nohive/selfreport,class:*,severity:*), so the org's add-help-wanted workflow never fired on it; that looks like a label write that failed at filing time.hivecommons-hive commentedon Oct 27, 2026
🔒 Claimed by agent
scanner(agent on hive hosted-available-oke-11-placeholder-r05x). The claim lasts 2h0m0s without activity and renews while the work continues. Reply/unclaimto release, or/claimto take it over if you outrank the holder.Danathar commentedon Oct 27, 2026
@clubanderson, two follow-ups to your triage above.
1. Your suggestion, on v6: #14481. It logs the stall escalation at INFO, the same line as #14480 (which targets v5), with a test that runs the real
slogfeedhandler. It also corrects the notes' "theSign-off survives the squashgate stays red": none of the three repos here runsdco-squash-attribution.yml, so the notes now name that check only when the head's required or reported checks include it.2. A proposal, not a PR yet: the sign-off park stops the hive working on its own head. projectbluefin/testsuite#1028 shows it (times UTC, 2026-10-26/27):
stage/074-adopted-by-hive, grace elapsed with the swarm's ask open).021bb6f8, signed off as the App.stage/085asadoption-stalled:hive-signoff-attribution. The note says the head "has not moved for 1m".021bb6f8(the hive's own head) and asked for changes (verdictchanges_requested): delete a stale row indocs/skills/test-authoring/suite-map/SKILL.md:234, and mention the extra app downloads in the PR description.review fix dispatch excluded … reason=not-hive-authored-squash-signoff, thenreview fix withheld: PR not hive-authored; review published, nothing pushed.Why, at 392df48:
adoptionStalledReasonreturns the sign-off reason before the 074 dwell and staleness check (automerge_contributor_adopt.go:1122). So the PR parks as soon as the hive's own commit is seen, 17 minutes before the swarm had reviewed it.stage/074-adopted-by-hive(fixPushAllowed,review/dispatch.go:1034). The park removed that label.So the swarm wants changes on the hive's own commit, the hive has stopped working on it, and the 085 note asks a maintainer to approve that same head.
Proposed rule: park on
hive-signoff-attributiononly once nothing else is open on the head. That means the swarm's verdict on this head is in and does not block (approveorrequires_human), and no required check is red or pending. Until then:Unchanged: such a head is never re-homed (the superseder would carry the same head), the 085 note is the same, and approving the head still selects the merge commit.
Under this rule utah-packages#660 (swarm
approve) and chairlift#638 (requires_human) would park exactly as they did; only heads shaped like #1028 change. If you agree, I'll open the PR against v6.hivecommons-hive commentedon Oct 27, 2026
🔓 Claim by agent
scannerreleased (holding agent relaunched).🐛 fix(automerge): sign-off stall notes name the squash gate only wher…