Repository navigation
Fix macOS test fixtures for paths and disk rollback - #512
Conversation
|
Merge plan for this four-PR stack: Stack order: #512 → #513 → #514 → #516. Each PR should ultimately merge into After a predecessor merges, retarget the next PR to This PR is the first step and is already ready for review. It changes test fixtures, not guest behavior. macOS workspace tests and the deliberate rollback regression check passed; Linux CI and CodeQL are green. Obtain the usual maintainer approval and confirm required checks remain green, then merge #512 into The next action is to retarget #513 from #515 is an independent follow-up for existing |
0bd6b9f to
6a13b64
Compare
hbrodin
left a comment
There was a problem hiding this comment.
Reviewed the full rebased diff at 6a13b64 using current main guidance. No actionable findings remain. Linux workspace tests, workspace Clippy, formatting, dependency checks, and applicable pre-commit hooks pass. Deliberately removing the spawn-error rollback makes the repaired test fail; restoring it passes. macOS tests were reported before the rebase and were not repeated afterward. Production behavior is unchanged; fresh CI must pass before merge.
The macOS unit suite now avoids creating an invalid UTF-8 filename on APFS and deterministically tests Lima's YAML rollback when
truncatecannot start. The rollback fixture runs in an isolated child with a controlled executable search path, so it neither finds a hosttruncatenor changes sibling tests' environment. Linux retains the filesystem-backed invalid-path test.Closes #508. Closes #509.
Validation: before rebasing, macOS workspace tests, workspace Clippy with
-D warnings, formatting, and all pre-commit checks passed. After rebasing onto currentmain, Linux workspace tests, workspace Clippy with-D warnings, formatting, TOML formatting, dependency checks, and all applicable pre-commit hooks pass. Removing the spawn-error rollback in the rebased tree deliberately makes the repaired test fail on the changed YAML; restoring it makes the test pass. The full rebased diff passed closeout review. Production behavior is unchanged. macOS tests were not repeated after the rebase; fresh Linux CI and all CodeQL analyses passed on the rebased head.