The gap-suite output gets truncated in CI, which has previously hidden live regressions. Emit full output plus a pass/fail table as a job annotation.
Robustness criteria (from @proggeramlug's field data): "Stop truncating" is necessary but incomplete. test_parity_timers_promises (root cause #712, now closed) emitted 5.7M identical lines and the runner's O(n²) bash normalize_output burned ~3 hours before being killed. Acceptance criteria for this sub-issue must also include:
- Per-test output cap — hard limit on stdout bytes/lines (e.g. 50k lines) per test, with a clear "TRUNCATED at N" marker so the cap is visible and not silent.
- Linear-time
normalize_output — replace the O(n²) bash implementation, or pathological output DOSes the whole CI job.
Without these, the job is one bug away from another 3-hour timeout. Same robustness constraint applies to #812's CI job.
Part of #793.
The gap-suite output gets truncated in CI, which has previously hidden live regressions. Emit full output plus a pass/fail table as a job annotation.
Robustness criteria (from @proggeramlug's field data): "Stop truncating" is necessary but incomplete.
test_parity_timers_promises(root cause #712, now closed) emitted 5.7M identical lines and the runner's O(n²) bashnormalize_outputburned ~3 hours before being killed. Acceptance criteria for this sub-issue must also include:normalize_output— replace the O(n²) bash implementation, or pathological output DOSes the whole CI job.Without these, the job is one bug away from another 3-hour timeout. Same robustness constraint applies to #812's CI job.
Part of #793.