You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Part of the test strategy (captain, 2026-10-05: scenarios that test the changes together; cleaner, faster suites; TESTING.md). Epic #305.
Scope.nightly.yml: INPUTLAYER_ORACLE_SEEDS=300, INPUTLAYER_SOAK=1 soak (#298: 10 min, oracle-checked histories with 200 subscribers and slow consumers), cargo fuzz parser and WS frames for 10 min each (#301), the HNSW recall test with the level-scale fix (#302) and brute-force recall over 3 sizes, perf-gate-remote.sh on the bench host under the shared lock (PR 273), coverage once a week; results posted as a job summary; a nightly failure opens or updates one tracking issue.
Acceptance. First nightly green or with named XFAILs; the perf gate report lands in the summary; #[ignore] count stays zero.
Done: the nightly workflow (T-11) is merged, started by hand while hosted CI is paused (the schedule line is present but commented out). An unregistered bench runner shows as a named XFAIL; fuzzing replays the checked-in seeds and regressions. The first nightly run is still pending a manual start.
Activity
jsam commentedon Oct 6, 2026
Done: the nightly workflow (T-11) is merged, started by hand while hosted CI is paused (the schedule line is present but commented out). An unregistered bench runner shows as a named XFAIL; fuzzing replays the checked-in seeds and regressions. The first nightly run is still pending a manual start.