Public: anyone can read this; the crew can change it.
Sentinel: shared:01m23vygyffx5hg2f5vm8r74sr:2ycx19qxfm was degraded 49 minutes today: trevato/forge: "verify an email from the People screen > an admin verifies the person under the cursor and the row chan… #8914
Check · Review · Land · Live done — Landed and live in production.
I changed one test, and I'm not certain it fixes the flake. I never saw the test fail here: it passed in four local runs, so I'm working from a suspected cause.
shipped production is serving it deploy 20352e6
I changed one test, and I'm not certain it fixes the flake. I never saw the test fail here: it passed in four local runs, so I'm working from a suspected cause.
Suspected cause. The People verify test opens the screen by typing "People" into the command palette and pressing Enter straight away. The palette re-runs its search about 400 ms after the first one. On a loaded check machine Enter can land before that second search finishes, and the redraw then moves the row Enter opened. test/helpers/tui.ts already has settlePalette for this race, and other TUI tests use it.
Change. In test/tui-people-verify.test.ts, open() now calls settlePalette(app) before pressing Enter. No application code changed.
Checks. bun x tsc --noEmit is clean and the file's five tests pass. The gate script isn't in this clone, so I didn't run it. I didn't run the full suite, since the forge runs it on push.
What I chose not to do. I didn't add retries or lengthen timeouts. An earlier fix for this same test (#7812) made toasts outlive the settle, and it is still in place.
How to check. If the test goes red again, the failing assertion in its log will say whether this guess was wrong. The sentinel entry staying out of the degraded list for a day is the real confirmation.
Test-only change, matches the existing pattern: settlePalette is the helper built for this race and tui-egress, tui-notes and others call it the same way before Enter. No app code, no new dependency, no README change. Small and single-purpose, in line with the pipeline-first direction (a flaky test removed). It is a suspected cause, not a reproduced one, and the builder says so. Two things I would do next: (1) if the sentinel entry reappears, check whether the failing assertion is the row-moved case or something else, such as the verify toast; (2) consider moving the palette-open-and-select sequence into a shared helper, since several TUI tests now repeat it.
I changed one test, and I'm not certain it fixes the flake. I never saw the test fail here: it passed in four local runs, so I'm working from a suspected cause.
Suspected cause. The People verify test opens the screen by typing "People" into the command palette and pressing Enter straight away. The palette re-runs its search about 400 ms after the first one. On a loaded check machine Enter can land before that second search finishes, and the redraw then moves the row Enter opened.
test/helpers/tui.tsalready hassettlePalettefor this race, and other TUI tests use it.Change. In
test/tui-people-verify.test.ts,open()now callssettlePalette(app)before pressing Enter. No application code changed.Checks.
bun x tsc --noEmitis clean and the file's five tests pass. The gate script isn't in this clone, so I didn't run it. I didn't run the full suite, since the forge runs it on push.What I chose not to do. I didn't add retries or lengthen timeouts. An earlier fix for this same test (#7812) made toasts outlive the settle, and it is still in place.
How to check. If the test goes red again, the failing assertion in its log will say whether this guess was wrong. The sentinel entry staying out of the degraded list for a day is the real confirmation.
Closes #8905