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
{{ message }}
Repository navigation
Site rig follow-ups after stable: judges at margin 0 (#1914) #1939
Follow-ups from the final review of #1914 (head dbc4332) on 28 Oct 2026. None of them makes the PR worse than master or breaks the owner's fuse rule, so they wait until after stable.
test(site): pin the width of s18's skip after the house step
test(site): replay a rig record through the judges without the rig
control: s18 sim with several house steps at margin 0 (keep review's closed-loop case)
test(site): pin the width of s18's skip after the house step
TestS18OwnCommandsSkipTheHouseStep checks overages at 50 s and 150 s, but nothing between 10 s and 50 s after small_step. Review mutation on dbc4332: change the second window from minor (small+10) to d.window(small+45, none). The test still passes, so the judge could skip 45 s of Core's own commands. Test idea: add a 4 s overage that starts at small+12 (inside the minor window) and must fail; add one at small+6 that must pass.
test(site): replay a rig record through the judges without the rig
For each judge change in this stack (s18 on dbc4332), the PR body quotes numbers from a rig record, such as "own commands peak at 16.008 A". Nobody can check them without the runner, because no command re-judges a saved results directory. Test idea: a go test ./test/site -run TestRejudge -args -dir <results> hook that loads ftw.jsonl and marks, then prints metrics and criteria. Then check that the fdfbfec s18 record gives max_i_own_commands <= 16.01.
control: s18 sim with several house steps at margin 0 (keep review's closed-loop case)
Review ran s18 at margin 0 with house steps +0.1 A, -0.7 A, +0.5 A on one phase only, -0.5 A and +0.3 A, one minute apart. Core's own commands peaked at 16.005 A outside the 10 s after each upward step. The longest stretch over 16 A was 3 s, and the physical peak was 16.50 A. It passes on dbc4332. Today only TestBatteryBesideMinimumCarHoldsTheFuseLine covers a single +0.1 A step. Test idea: add this sequence as a sitesim test asserting own peak <= 16.01 A and each overage < 10 s, so that a later raise that overshoots after a house drop gets caught.
Activity