[ManSio]

Three defects found while measuring why the suite looked like an 18-minute hang. All three produced real orphaned processes or unbounded waits on this machine.

1. The hook killed only the direct child (.githooks/pre-commit)

On TimeoutExpired it called proc.kill(). The gate is a Python script that spawns its own children: verify_diary -> pytest tests/. The direct child died, the grandchild kept running for 10+ minutes and competed with the next run. Observed twice (PIDs 7732 -> 14156 -> 16264).

Now gates are process-group leaders (CREATE_NEW_PROCESS_GROUP / start_new_session) and the timeout path kills the tree (taskkill /PID <pid> /T /F or killpg), with proc.kill() as a fallback.

2. The launch lock was unbounded (src/core/embedder/onnx_client.py)

ensure_server_running() did with _client_lock: with no timeout. A leaked thread holding it turns the next call into an indefinite wait: in the full run one test was reported at 654s while the same file takes 2.7s alone and that test < 0.01s. It now acquires with _LOCK_TIMEOUT = 150s and returns False (fail closed) instead of hanging.

3. Two suites could run on one checkout (tests/conftest.py)

A session file lock (.pytest_cache/suite.lock) makes the second run exit 3 with an explicit message. A nested pytest spawned by a test inherits MSCB_SUITE_LOCK_INHERITED=1 and skips the check, so test_bootstrap_pipeline still works. The OS releases the lock when a process dies, so a crash cannot leave it stale.

Evidence

tests: 20 passed (test_suite_lock, test_precommit_hook_contract, test_idle_reload)
ruff:  All checks passed

Real processes, not fakes:

parent=14720 child=5544 child_alive_before=True
child_alive_after=False parent_poll=1          # _kill_tree removed the grandchild

second_pytest_exit=3
! Another pytest run already holds .pytest_cache/suite.lock. ...

The new tests are the negative controls: test_timeout_kills_the_process_tree fails if the /T flag or fallback disappears, and test_second_acquire_fails_and_release_frees fails if the lock stops excluding.

Limits

• Full pytest tests was not run to completion for this commit (1092s under a concurrent benchmark); the commit used the documented MSCB_PRECOMMIT_FAST=1 mode, so gate-zero runs in CI.
• A benchmark process does not take the suite lock yet; it still competes for CPU. Wiring experiments/bench_2026_11/run.py to the same lock is a reasonable follow-up.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: a3a76168-538e-46f5-8f6d-f739ea7560fe

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• Trigger review

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn