Repository navigation
Flaky gap test: test_gap_turnloop_p3_timer_heap intermittently output-mismatches #11148
Description
test_gap_turnloop_p3_timer_heap fails intermittently in the PR-tier gap suite with an output mismatch, then passes when the same SHA is re-run.
Seen twice on 2026-09-23:
- train 270 (Merge train 270: 7 PRs (v0.5.1653) — folds train 267 #11132): failed, passed on re-run of the same SHA.
- fix(gc): rearm closure young log after slot rewrites #11118, run 35860774280,
gap-suite (3):test_gap_turnloop_p3_timer_heap: pass -> parity_fail, the only regression in the shard. Re-run requested.
Probably timing-sensitive output (timer ordering or elapsed-time-dependent lines) under a loaded runner. Until it's fixed it reddens pr-gate on unrelated PRs. Either make the test's output independent of scheduling jitter, or triage it as known-flaky in test-parity/known_failures.json with this issue linked.
Activity
proggeramlug commentedon Sep 24, 2026
More intermittent gap failures from 2026-09-24, same pattern (pass on re-run, unrelated diffs):
test_gap_turnloop_p3_timer_heapagain, on perf(runtime): replace Map side tables with an owned MapStore #11171 (gap-suite 6).test_gap_cron_cronjobon fix(stdlib): enable base64 for turnloop-http-client so fetch programs get a stripped runtime #11116 (a one-line Cargo feature change; passed on re-run) and on fix(hir): new/instanceof, async bodies and inlined factories use a class declaration's self-binding (#11142; stacked on #11188) #11190 (gap-suite 2).test_gap_turnloop_p9_worker_agent_neton perf: share boxed closure capture layouts by function #11177 (gap-suite 6).All timer or network timing-sensitive.