Repository navigation
check-affected --tests overruns its budget and stalls the ready flip #12639
Description
Problem
check-affected.sh --tests, the last part of /source-control:pull-request ready's verify gate on a draft, can run far past its 120 s budget, which stalls the ready flip.
Measured
Two runs on #12614's branch on melo-lap-001 WSL, with other lanes' evals running:
- 364 suites selected, stopped at 14 of 224 after 25 minutes.
- A second run at
--jobs 1 --budget 120was still going after about 40 minutes, when it was stopped.
Expected: about 2 minutes, from --tests passes a default job count of at most 8 and the 120 s budget.
Likely cause
run-plugin-tests.sh:41calls the budget "the wall clock the whole run aims to finish in", and:608only reportsOVER BUDGETafter the run, so nothing stops or shrinks the run. Read this session.- The job count fell to 1, probably from the free-memory cap (
check-affected.sh:130-131). UNVERIFIED. - Why this branch selects 364 suites is UNVERIFIED. It adds
lib/copies, a contract doc row andscripts/shared-copies.txtentries.
Fix to weigh
Let the budget end a local run: stop launching suites when it is spent and report the rest as left to CI, the way --tests already reports unrun suites as a skip. CI runs the full set either way, so the local gate needs only a bounded sample.
Done when
- A local
check-affected.sh --testsends within about its budget plus one suite's run, and names the suites it left to CI.
Metadata
Metadata
Assignees
Labels
Type
Fields
Work status
agent-ready
Activity
github-actions commentedon Nov 3, 2026
Proposed work class: work-class: scoped
Basis: automated intake triage; the lanes App applied this class.
Size: Small
Complexity: Medium
Uncertainty: Knowable
Flags: none
Blocked by: none
Understood: a local
check-affected.sh --testsrun (the last step of/source-control:pull-request ready's verify gate) can run far past its 120 s budget, because the budget is only reported after the run. This stalls the ready flip. The proposed fix is to stop launching suites once the budget is spent and name the rest as left to CI. Done when a local run ends within about its budget plus one suite and names the suites it left to CI.Still to confirm while working: why the job count fell to 1, and why this branch selects 364 suites. I found no likely duplicate among the open issues.
Automated triage, applied by the lanes App.
github-actions commentedon Nov 3, 2026
kyle-sexton commentedon Nov 3, 2026
melodic-automation-lanes commentedon Nov 3, 2026
This records a re-fire of an earlier refusal, not a new refusal. The lanes kill switch stopped run 41572558254, the work-resolve-issue run that
agent-readystarted, before it did any work; that run ended at 2026-11-03T04:41:59Z. work-unblock has now removed and addedagent-readyback to start a fresh run. A kill-switch refusal is no fault in this issue or the lanes, so it never counts against later re-fires; after limits.limit-stops-per-issue of them since a person last addedagent-readyor a run last reported, work-unblock waits longer before each one, up to 24 hours.github-actions commentedon Nov 3, 2026
github-actions commentedon Nov 3, 2026
already-leased: another session holds a live lease on #12639, so run https://github.com/melodic-software/claude-code-plugins/actions/runs/41581265469 ran no model.