diff --git a/.github/AGY-REVIEWER.md b/.github/AGY-REVIEWER.md index 3e768779..b554dc52 100644 --- a/.github/AGY-REVIEWER.md +++ b/.github/AGY-REVIEWER.md @@ -272,6 +272,7 @@ runner, so disable fork workflows there or gate them the same way. | `AGY_TIMEOUT_SECONDS_PER_MIB` | `900` | extra seconds per MiB of diff. A diff handed over as a file costs agy 6-8 minutes almost whatever its size, so this is deliberately steep (240 left a 426 KB diff at 397 s, too little) | | `AGY_PRINT_TIMEOUT_MAX_SECONDS` | `1800` (workflow: `900`) | ceiling on one attempt after the raise | | `AGY_RETRIES` | `3` (workflow: `2`) | attempts; a retry is for a transient backend error, so more of them at the same budget only repeat a timeout | +| `AGY_LOCK_WAIT` | `600` (workflow: `1860`) | seconds to wait for the host-wide agy lock before giving up. Every repository's review on the host shares it (agy is one singleton per user), and a review holds it across all its attempts and the pauses between them, so the wait must outlast another review's whole hold: 1860 s covers one review at two 900 s attempts plus the 15 s pause. It counts toward the job's worst case, so raising it means raising `timeout-minutes` (the selftest checks the sum) | | `AGY_DIFF_MODE` | `auto` | how the diff reaches agy: `inline` (in the `--print` prompt), `file` (written to a file agy reads with its tools), or `auto` (inline if it fits the arg-size budget, else file — so **large PRs are never truncated**) | | `MAX_DIFF_BYTES` | `5000000` | sanity cap on a pathological diff (5 MB). Not the arg-size limit — a large diff goes to agy as a file, not an argv value | | `MAX_PROMPT_BYTES` | `120000` | the inline/file threshold in `auto` mode, and a hard backstop on the **inlined** prompt. agy takes an inlined prompt as a `--print` argv value, and a single `execve` argument cannot exceed `MAX_ARG_STRLEN` (128 KiB); over it `execve` fails with E2BIG before agy starts. Clamped below 128 KiB | @@ -281,6 +282,14 @@ runner, so disable fork workflows there or gate them the same way. ## Troubleshooting +- **"agy lock timed out after Ns" and a red job with no review:** another + review on the same host held the agy lock longer than `AGY_LOCK_WAIT`. The + lock is shared by every repository's review there, so two PRs opened at once + in different repositories queue behind each other; with a wait shorter than + one review (the old 300 s), the second gives up. Re-run the job once the other + review finishes, and keep `AGY_LOCK_WAIT` above another review's worst-case + hold (`AGY_RETRIES` x the per-attempt cap, plus the pauses). The lock fails + closed: a timeout exits rather than running two agy processes at once. - **Review comment never appears / job succeeds but empty:** the classic Antigravity **issue #76** — `agy -p` drops stdout when it isn't attached to a TTY. This template runs `agy` under a PTY (`unbuffer`, or `scripts/_agy_print.sh` diff --git a/.github/workflows/antigravity-review.yml b/.github/workflows/antigravity-review.yml index c4d94d7a..d1c4eef0 100644 --- a/.github/workflows/antigravity-review.yml +++ b/.github/workflows/antigravity-review.yml @@ -69,10 +69,11 @@ jobs: # for a large one handed over as a file. The cap must still exceed the # script's own worst case -- lock wait + every retry at its maximum timeout + # backoff -- or the job is killed mid-retry and posts nothing. With the - # settings below that is 300 + 2 x 900 + 15 = 2115 s, under 45 minutes with + # settings below that is 1860 + 2 x 900 + 15 = 3675 s, under 80 minutes with # room for checkout and the diff fetch; the selftest checks this sum against - # this value. - timeout-minutes: 45 + # this value. (Until 2026-11-15: 300 + 2 x 900 + 15 = 2115 s under 45 + # minutes; see AGY_LOCK_WAIT for why the wait grew.) + timeout-minutes: 80 steps: # Check out the DEFAULT BRANCH, never the PR head. This job runs the checked-out # `scripts/agy-review.sh` on a self-hosted runner with a token in the environment, @@ -125,7 +126,17 @@ jobs: # Two attempts, not three: a retry is for a transient backend error; a third # attempt at the same budget only repeats a timeout. AGY_RETRIES: "2" - AGY_LOCK_WAIT: "300" # wait for another review's lock (default 600) + # The lock is shared by EVERY repository's review on this host, not only this + # one's (agy is one singleton per user). 300 s was shorter than one review + # elsewhere: on 2026-11-15 RustyNES_MiSTer's PR #79 and RustyNES's PR #664 + # opened together, #664's review of a large diff held the lock, and #79's job + # gave up after 300 s ("agy lock timed out") and went red with no review. + # 1860 s outlasts another review's whole hold: the lock is held across both of its + # attempts at the 15-minute cap AND the 15 s pause between them (2 x 900 + 15 = + # 1815 s), plus a margin for process start-up; this is what moved + # timeout-minutes to 80. The runner is self-hosted, so the longer wait + # bills nothing; the wait is a queue, not a fail-open. + AGY_LOCK_WAIT: "1860" # wait for another review's lock (default 600) # AGY_DIFF_MODE: auto # auto|inline|file — how the diff reaches agy. `auto` # # (default) inlines a small diff and hands a large one to # # agy as a file, so PR size never truncates the review.