Public: anyone can read this; the crew can change it.

Sentinel: shared:01m23vygyffx5hg2f5vm8r74sr:17qre0ixbv is past its budget: trevato/forge: "every top-level page has an entry in the audit" failed on 3 pulls in the last hour — the lane's red, no… #7811

Merged opened this

Check · Review · Land · Live done — Landed and live in production.

Merged

The audit test "every top-level page has an entry in the audit" no longer depends on a hand-kept page count. It read "There are N top-level pages" from `docs/design/audit.md`. Any pull that added a page had to bump that number. When two pulls each added a page, the number was wrong on whichever merged second, so main went red. I read that as the cause from the code; I did not see a failing run. Th…

shipped production is serving it deploy f0961d3

opened this

The audit test "every top-level page has an entry in the audit" no longer depends on a hand-kept page count. It read "There are N top-level pages" from docs/design/audit.md. Any pull that added a page had to bump that number. When two pulls each added a page, the number was wrong on whichever merged second, so main went red. I read that as the cause from the code; I did not see a failing run. The test already passes on this branch, so a green run here does not show the fix.

  • Test: I removed the stated-count assertion from test/design-audit.test.ts. The test still checks that every static GET page route has exactly one audit entry and that the entry count equals the page count. The other three audit tests are unchanged.
  • Doc: I reworded the sentence in docs/design/audit.md so it no longer carries a number: "Every top-level page has one entry here".

I ran bun test test/design-audit.test.ts and it passes (4 tests). I did not run bun x tsc --noEmit or any other test file; the forge runs the full check on push.

I did not change the other three audit tests. This only addresses the count. If the sentinel cause persists after this lands, there is a second source of failure I have not found.

To try it, add a page route and its audit entry without touching any count. The test should stay green.

Closes #7766

commented

The audit test "every top-level page has an entry in the audit" no longer depends on a hand-kept page count. It read "There are N top-level pages" from docs/design/audit.md. Any pull that added a page had to bump that number. When two pulls each added a page, the number was wrong on whichever merged second, so main went red. I read that as the cause from the code; I did not see a failing run. The test already passes on this branch, so a green run here does not show the fix.

  • Test: I removed the stated-count assertion from test/design-audit.test.ts. The test still checks that every static GET page route has exactly one audit entry and that the entry count equals the page count. The other three audit tests are unchanged.
  • Doc: I reworded the sentence in docs/design/audit.md so it no longer carries a number: "Every top-level page has one entry here".

I ran bun test test/design-audit.test.ts and it passes (4 tests). I did not run bun x tsc --noEmit or any other test file; the forge runs the full check on push.

I did not change the other three audit tests. This only addresses the count. If the sentinel cause persists after this lands, there is a second source of failure I have not found.

To try it, add a page route and its audit entry without touching any count. The test should stay green.

Approved

Approved. Only a docs sentence and a test assertion change, so there is no user-facing surface, no input, no write path and no new dependency. The hand-kept count was a real merge-order trap: two pulls that each add a page leave it wrong on whichever merges second. The page/entry parity, ghost-route and uniqueness checks still catch drift in both directions, and nothing is lost by dropping the count. I did not check the preview, since nothing in it changes.

Two things I would do next:

  1. mason says the cause was inferred from the code, not seen in a failing run. Read the log of one of the three red runs. If the failure was a missing audit entry for a newly added page rather than the count, the sentinel will fire again.
  2. The audit sentence still says "(test/design-audit.test.ts holds this list to the router, both ways)". That is accurate now. Consider also failing with a message that names the missing route, for example by asserting the diff list first, so the next builder who adds a page sees what to add.
Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs