test(windows): stabilize root marker race evidence - #3872
Conversation
|
@Astro-Han Exact-head est and windows_recovery are green. The hosted Windows job directly ran the new root initialization replacement race step as 1 test / 1 pass / 0 skip. Local focused evidence is 20 consecutive passes. Please review when available. |
eed5e6d to
78e4c95
Compare
Generated-by: Codex
Generated-by: Codex
78e4c95 to
e354489
Compare
M4n5ter
left a comment
There was a problem hiding this comment.
I reviewed exact head e354489615c9940c56d6a99600cb47586d83043d and found no P0–P3 issues.
The fixture now keys the interception on the unique marker basename instead of one particular full-path spelling. That is appropriately bounded: the child performs one root initialization, and production still constructs the temporary marker under that root with the same marker filename plus a process ID and UUID. This removes the Windows long/short, namespaced, and case-spelling dependency without changing production authority or state.
The dedicated Windows workflow step also proves that exactly one matching test ran and passed with none skipped. The workflow contract test protects that evidence from being silently removed or broadened.
Validation on this head:
- exact-head hosted
testandwindows_recoverychecks completed successfully; - Storage and Core builds passed;
- the focused race test passed exactly 1/1 with 0 skipped;
- the complete root-authority suite passed 28 tests with 3 platform-specific tests skipped and no failures;
- all 39 CI workflow contract tests passed;
- Biome and
git diff --checkpassed; - the current-main merge tree is clean and preserves all three reviewed files byte-for-byte.
Posted by an automated review agent operated by @M4n5ter. This is not an
independent human review and does not satisfy the committer review required by
CONTRIBUTING.md. A human is accountable for this comment — please push back if
anything here is wrong.
简体中文
本条评论由 @M4n5ter 运行的自动化审查程序发出。它不构成 CONTRIBUTING.md
所要求的独立人类审查,也不能替代人类审查。有人类对本条评论负责,如有错误请直接指出。
Summary
Close the remaining deterministic Windows root-initialization race failure found while validating #2624.
realpathstring spelling;windows_recoverywith a strict1 test / 1 pass / 0 skipassertion;Why
The first post-#3789 full Windows baseline on byte-identical
main@32a1db0ebreproduced one deterministic failure in both the focused Storage gate and the complete Storage suite:initialization race fixture reported resolved while waiting for marker_open_pending.The child fixture compared the opened temporary path against a
realpath-derived parent prefix. Windows can expose equivalent parents with different long/short, namespaced, case, or canonical spellings. This child performs exactly one root initialization, so the marker filename uniquely identifies the intended open cut without relying on parent spelling.All other baseline groups passed: install/build, CLI and Electron smoke, Runtime PTY 3/3, PowerShell UTF-8 2/2, the remaining Storage path/lock gates, and residual-process audit.
Verification
Local Windows, Node 24:
git diff --check: passed.Hosted exact-head
windows_recoverypassed and directly executed the new root race step. Hostedtestalso passed.Three consecutive fork full-baseline runs on this exact head are clean: run 1, run 2, and run 3. Each reports focused Storage 88/0/12, full Storage 940/0/38, empty residual-process evidence, and successful PTY/UTF-8/smoke steps.
AI use
Tool and scope: Codex inspected the full Windows baseline artifact, isolated the path-spelling-sensitive fixture cut, implemented the fixture and workflow contract changes, and ran the listed local gates.
Refs #2624. Refs #2142.
Checklist
testandwindows_recoverypass on exact head6dae318b9Does this PR entail a change in behavior?