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

The forge alarms when main is green and ahead of production for more than 90… #11592

Parked opened this
Closed without merging

The release-heartbeat and main-ahead alarm change from #11507 is now pushed with both check failures addressed. I ran only the two failing test files (they pass) and bun x tsc --noEmit (clean). I did not run bun scripts/gate.ts or the full suite, which the forge's check runs on this push.

Notes

stopped parked by the crew

opened this

The release-heartbeat and main-ahead alarm change from #11507 is now pushed with both check failures addressed. I ran only the two failing test files (they pass) and bun x tsc --noEmit (clean). I did not run bun scripts/gate.ts or the full suite, which the forge's check runs on this push.

  • header-width: the new comment in src/system/home_truth.ts ran past 160 columns. I wrapped it onto two lines.
  • stall-in-flight: this test was already tolerant of a loaded machine on main. My branch still had the strict toContainEqual version, so a real stall charged to the same cause could fail it. I merged the latest origin/main, which brings in the tolerant version. It passes now, but I haven't shown it was flaky, so it could still fail on a loaded machine.

Neither test was changed on this branch.

Continues #11592

2 earlier comments
commented

The release-heartbeat and main-ahead alarm change from #11507 is now pushed with both check failures addressed. I ran only the two failing test files (they pass) and bun x tsc --noEmit (clean). I did not run bun scripts/gate.ts or the full suite, which the forge's check runs on this push.

  • header-width: the new comment in src/system/home_truth.ts ran past 160 columns. I wrapped it onto two lines.
  • stall-in-flight: this test was already tolerant of a loaded machine on main. My branch still had the strict toContainEqual version, so a real stall charged to the same cause could fail it. I merged the latest origin/main, which brings in the tolerant version. It passes now, but I haven't shown it was flaky, so it could still fail on a loaded machine.

Neither test was changed on this branch.

commented

I made no code change for this failure. No test failed. The check machine's CPU governor stopped the run for using 16.9 cores for 31s against a limit of 16.

  • Not my change: the busiest process was a sh that had used 438s of CPU. The "egress refused" lines for Chrome update hosts and DuckDuckGo come from browser-based tests, and none of them touch the code in this pull.
  • No new commits on main: nothing new landed on main since the last merge, so I changed nothing.
  • Re-run: I pushed an empty commit (0de1cd32a) so the forge runs the check again.

If the governor stops the check again at the same point, the cause is the load on the check machine. It would need a person to look at the shard scheduling, since a change to this pull wouldn't fix it.

commented

Parked by the crew: the check stayed red through 2 automatic tries

Sign in to join this conversation.

↑↓ navigate ↵ open ⌘↵ its app esc close

Keyboard shortcuts

Docs