[EvanProgramming]

Problem and root cause

A clean-candidate full-suite run failed while removing the bug-fix workflow fixture's temporary SQLite directory; state.sqlite3-shm remained. Instrumentation found the fixture began cleanup with a submitted post-turn learning task still pending in 3 of 50 repetitions. This proves the fixture has a cleanup race; the original intermittent failure's exact timing was not captured, so its cause is not asserted as certain.

Change

Disable post-turn learning within this bug-fix workflow fixture. The fixture exercises reproduction, repair, verification, and feedback; post-turn learning is covered independently and does not belong to these assertions. This isolates the fixture without changing production behavior or blocking user-facing turns. A regression test fails if the fixture starts real background learning.

Stacked on #323 (Evan/v3-history-database-protection). No milestone issue is claimed closed by this test-only follow-up.

Validation

• Regression test failed before the fixture change and passes after it.
• Instrumented 50-run comparison: 3/50 cleanups had background learning pending without isolation; 0/50 with isolation. No cleanup failures occurred during either sample.
• Full local suite: 1,186 tests, 2 platform skips; make check, make lint, and make docs-check passed.
• Exact-candidate CI and six platform soaks are pending.

Risks and follow-up

This test-only change does not establish the root cause of the prior Windows reviewer/cancellation failures or prove application shutdown waits for background tasks. R260 remains blocked; the benchmark is incomplete, and the prior candidate's records and waivers do not apply to this commit. No release manifest or publication is authorized by this PR alone.

Update: The full suite was rerun on the clean signed commit and passed: 1,186 tests, two platform skips. Exact-candidate CI and platform soaks remain in progress. The exact-candidate matched benchmark stopped on its first V2 request with AI_RELAY_STREAM_INCOMPLETE and unknown usage; it is inconclusive, and the reservation is held. R260 remains unresolved.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

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

Tip

🎯 Customize your risk assessments

Guide how CodeRabbit assesses merge risk and security architecture risk. Highlight sensitive changes and explain your security boundaries using Merge Risk Instructions and Security Risk Instructions in your review settings. Security architecture risk assessments require the Advanced plan or above.

✅ Automatic approvals, with your rules

Let CodeRabbit approve pull requests once required threads are resolved, pre-merge checks pass, and your risk limits are met. Enable Automatic approvals in review settings, with overrides for teams or file paths.

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

CodeRabbit Verify

• Run CodeRabbit Verify to test the changes in this pull request.