Repository navigation
chore(pr-pipeline): remove the lane-record annotation and its fallback after 2026-10-25 #7775
Description
What
Remove the lane-record annotation that report-check-run still posts, and the annotation fallback in .github/scripts/pr-unblock/limit-stop.mjs. Since #7774, the record lives on the first line of the posted check run's output.summary, and the dispatchers read it from there. The annotation exists only so that limit stops posted before #7774 still restart.
When
On or after 2026-10-25. A limit stop's reset is clamped to at most 7 days after its run ended (limit-stop.mjs), and #7774 merged on 2026-10-17. After that date, no check run posted before the change can still be waiting to restart.
Done when
report-check-runposts noannotationsfor the lane record. DeletelaneRecordAnnotationand its tests.limitResetsAtandanyHandlesread only the summary record, andlistAnnotationsis gone fromdispatch.mjsandfeedback-dispatch.mjs.- The
app_idandcheck_namefilters on the dispatchers' check-run list stay. The security review of feat(pr-pipeline): carry the lane record in the posted check run's output #7774 noted that dropping the fallback also closes the path where an::error::annotation inside the job is read as a lane record. docs/conventions/pr-pipeline/pr-run-activity.mdno longer mentions the annotation.
🤖 Generated with Claude Code
Activity
github-actions commentedon Oct 17, 2026
Understood: this is a scheduled follow-up to #7774 — remove the lane-record annotation fallback from
report-check-runandlimit-stop.mjs(and the annotation-reading paths indispatch.mjs/feedback-dispatch.mjs), once no pre-#7774 limit stop can still be waiting to restart (on or after 2026-10-25, per the 7-day reset clamp). The done-list (deletelaneRecordAnnotationand its tests, droplistAnnotations, keep theapp_id/check_namefilters, updatepr-run-activity.md) is specific enough to act on directly once that date passes. No duplicates found; #7772 and #7765 are the already-closed predecessors this refers to. Nothing further needed from the reporter — this looks ready to pick up after the date gate.Automated triage. A maintainer reviews it.
kyle-sexton commentedon Oct 17, 2026
The wait is unnecessary. The live proof in pr-pipeline-sandbox (2026-10-17, run 39056191404, check run 117464603970) showed that GitHub sets every check-run annotation title to
<path>#L<line>when it is posted without one.limit-stop.mjskeeps only annotations with no title, so the fallback has never matched an annotationreport-check-runposted, and no limit stop waits on it. Removing the fallback and the annotation now changes no behavior.The same proof confirmed the summary path end to end:
report-lane-outcomeandreport-check-runpostedlane=pr-unblock activity=resolve-conflicts class=rate-limit resets=… turns=1 max-turns=30.dispatchConflictingskipped the PR beforecompleted_at + 1h(also at 1h−1s) and dispatched exactly once at 1h+1s.