#386 - ci: main is red - `bootstrap (linux-x64)` fails on push and nightly at dc07765 - torbscript/language

orion commented

2026-10-19 22:03:12 +00:00

Collaborator

Main CI is red since the push of dc077652a0 (test(conformance): Option equality over ArrayList, RingList and Buffer): job bootstrap (linux-x64) failed after ~7 min on the push run (https://git.torb.dev/torbscript/language/actions/runs/1243) and again on the scheduled nightly run (https://git.torb.dev/torbscript/language/actions/runs/1244); pack the seed and gates fail as followers. The same commit was green as a pull request (run 1241).

Acceptance: find the cause from the job log (flake, runner, or a real regression between the PR run and the push run), fix it on main through a normal loop PR, main CI green on the landed commit.

Tracking: orion/tasks#97

Main CI is red since the push of dc077652a0 (`test(conformance): Option equality over ArrayList, RingList and Buffer`): job `bootstrap (linux-x64)` failed after ~7 min on the push run (https://git.torb.dev/torbscript/language/actions/runs/1243) and again on the scheduled nightly run (https://git.torb.dev/torbscript/language/actions/runs/1244); `pack the seed` and `gates` fail as followers. The same commit was green as a pull request (run 1241). Acceptance: find the cause from the job log (flake, runner, or a real regression between the PR run and the push run), fix it on `main` through a normal loop PR, `main` CI green on the landed commit. Tracking: orion/tasks#97

orion added the

loop

ready

labels

2026-10-19 22:03:12 +00:00

orion commented

2026-10-19 22:03:16 +00:00

Author

Collaborator

[loop:torbscript] claim implement by session:loop-torbscript-20261020 until 2026-10-20T04:00Z

orion commented

2026-10-19 22:04:26 +00:00

Author

Collaborator

[loop:torbscript] implement: diagnosed, not a code regression - a stale lock in the CI runner's action cache. Next: waiting (forge administrator must clear the lock on the runner, then re-run run 1243 or push to main again).

Evidence (job logs of run 1243 bootstrap (linux-x64) and nightly run 1244):

The compiler build itself passes in both runs: step 1, step 2 and step 3 all complete, "the fixpoint holds: steps 2 and 3 emitted the same C", program.c hash printed. The "emitted different C" line before it is the expected seed-older-than-codegen notice (same text in green run 1241).
The job then fails on the next step, Keep the release binary (a remote action, upload-artifact): [runner]: could not fetch remote 'origin': fatal: Unable to create '/data/.cache/act/09/afeb67125058141a431bed995e8fb4425f52500e08803ce95700e36a06ec2f/shallow.lock': File exists. Another git process seems to be running in this repository, or the lock file may be stale. The step is skipped, so no bootstrap artifact is uploaded.
pack the seed (and with it gates) fail only as followers: Unable to download artifact(s): Failed to ListArtifacts ... Not Found, because the upload never happened.
The failure repeats identically in the nightly run about 3 hours later (same lock path), so the lock is stale, not a transient race. Push runs 1236-1238 and PR run 1241 on the same commit were green, so the sources are fine.

Fix: remove /data/.cache/act/09/afeb67125058141a431bed995e8fb4425f52500e08803ce95700e36a06ec2f/shallow.lock (or the whole cache dir .../09/afeb6712...) on the runner that hosts it; act re-clones the action on the next run. This is runner state, outside the repo and outside what the loop may touch. A repo-side workaround (re-pinning the action to force a new cache dir) would only hide the stale lock, so I did not open a PR. After the lock is gone, a re-run of run 1243 should turn main green; if it stays red, the next failure is a different cause and this issue stays open.

Token usage: about 60k tokens (log analysis only, no worktree, no code changes).

[loop:torbscript] implement: diagnosed, not a code regression - a stale lock in the CI runner's action cache. Next: waiting (forge administrator must clear the lock on the runner, then re-run run 1243 or push to main again). Evidence (job logs of run 1243 `bootstrap (linux-x64)` and nightly run 1244): - The compiler build itself passes in both runs: step 1, step 2 and step 3 all complete, "the fixpoint holds: steps 2 and 3 emitted the same C", program.c hash printed. The "emitted different C" line before it is the expected seed-older-than-codegen notice (same text in green run 1241). - The job then fails on the next step, `Keep the release binary` (a remote action, `upload-artifact`): `[runner]: could not fetch remote 'origin': fatal: Unable to create '/data/.cache/act/09/afeb67125058141a431bed995e8fb4425f52500e08803ce95700e36a06ec2f/shallow.lock': File exists. Another git process seems to be running in this repository, or the lock file may be stale`. The step is skipped, so no `bootstrap` artifact is uploaded. - `pack the seed` (and with it `gates`) fail only as followers: `Unable to download artifact(s): Failed to ListArtifacts ... Not Found`, because the upload never happened. - The failure repeats identically in the nightly run about 3 hours later (same lock path), so the lock is stale, not a transient race. Push runs 1236-1238 and PR run 1241 on the same commit were green, so the sources are fine. Fix: remove `/data/.cache/act/09/afeb67125058141a431bed995e8fb4425f52500e08803ce95700e36a06ec2f/shallow.lock` (or the whole cache dir `.../09/afeb6712...`) on the runner that hosts it; act re-clones the action on the next run. This is runner state, outside the repo and outside what the loop may touch. A repo-side workaround (re-pinning the action to force a new cache dir) would only hide the stale lock, so I did not open a PR. After the lock is gone, a re-run of run 1243 should turn main green; if it stays red, the next failure is a different cause and this issue stays open. Token usage: about 60k tokens (log analysis only, no worktree, no code changes).

orion referenced this issue

2026-10-24 02:14:42 +00:00

checker: infer a type parameter through a trait bound (WEB F3) #101

orion commented

2026-10-24 07:09:11 +00:00

Author

Collaborator

[loop:torbscript] close: bootstrap (linux-x64) is green again on main, on push and on the nightly schedule: success at a5bdcf3e, fe8c4b8b, 7d4fe624, bc504c24, c4a84435 (push and nightly) and 93280923 (2026-10-23 16:40Z to 2026-10-24 03:38Z). The stale runner cache lock is gone; the latest run at e594f248 was cancelled (superseded), not failed. Closing; a new red bootstrap gets a new issue. next: none

[loop:torbscript] close: `bootstrap (linux-x64)` is green again on main, on push and on the nightly schedule: success at a5bdcf3e, fe8c4b8b, 7d4fe624, bc504c24, c4a84435 (push and nightly) and 93280923 (2026-10-23 16:40Z to 2026-10-24 03:38Z). The stale runner cache lock is gone; the latest run at e594f248 was cancelled (superseded), not failed. Closing; a new red bootstrap gets a new issue. next: none

orion closed this issue

2026-10-24 07:09:11 +00:00

Rows
Columns

ci: main is red - bootstrap (linux-x64) fails on push and nightly at dc07765 #386

ci: main is red - `bootstrap (linux-x64)` fails on push and nightly at `dc07765` #386