Repository navigation
test(work): isolate lock lifetime checks from parallel forks - #1279
Conversation
|
@codex review |
|
@codex security review |
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
|
Codex Review: Didn't find any major issues. Hooray! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
🛡️ Codex Security ReviewSecurity review completed. No security issues were found in this pull request. Reviewed commit: Only the user who started this review can view the report in Codex. ℹ️ About Codex security reviews in GitHubThis is an experimental Codex feature. Security reviews are triggered when:
Once complete, Codex will leave suggestions, or a comment if no findings are found. |
Fixes the lock-lifetime test flakes reported while qualifying #1265. T1 (test-only).
Linux flock locks belong to an open file description. A parallel libtest test can fork while another test holds a lock; CLOEXEC closes the inherited descriptor at exec. Dropping every intended owner therefore does not guarantee an immediate reopen can acquire the lock. This affects the exact durable native-completion assertion and both shadow admission replay/reopen tests. Widening the fork-to-exec window also exposed the adjacent native queue settlement test.
Run these four existing tests individually in a fresh, environment-cleared test process. Their locks are created after exec without unrelated test peers; the helper verifies exactly one passing test. All original assertions and real lock operations are preserved byte for byte. The test inventory remains exactly 106 library tests. The 49-line change is entirely test infrastructure; production behavior is unchanged.
Validation at exact head
6ac4025caa3d7c6f4ac54b2a47cca53786ff4c71:b72f5fc189a407d918a2426375f4b5ebcc39d40c. A controlled probe retaining fork-inherited descriptors reproduces both shadowalready in useerrors and native completionWouldBlock. A 500ms fork-to-exec delay in the full parallel library suite exposes the adjacent native queue settlement assertion (105 passed, 1 failed). Eight uninstrumented full library runs passed; the reported natural 2-of-4 failure rate was not reproduced.oxa-work,oxa-cli,oxa-parity. Lane 44225113910 SUCCESS on x86_64 and ARM64: 1,560/1,560 passed, zero skipped on each architecture, including all four repaired tests.git diff --checkand scopedrustfmt --checkpass. CI was started by ordinary branch pushes/PR creation; no manual dispatch or restart.