[Mikola Lysenko (mikolalysenko)]

LLM Description written by Claude Code:claude-fable-5-2

Fixes #2080

Summary

A test-only change. hot_patch::tests::a_start_is_a_project_pass_then_a_global_pass
fails on Windows, which has kept the scheduled main run's
test (windows-latest, 1) leg, and so ci-ok on main, red since #2068.
The two manifest-path assertions now compare paths instead of strings.

Root cause

passes() builds each pass's manifest path with Path::join, which writes
the platform's separator. The test compared the result to a /-joined
string literal:

left:  "/state\\project\\manifest.json"
right: "/state/project/manifest.json"

The product code is right: the path is handed to the scan as a path, and
either separator works on Windows. Only the test's expectation was wrong.
Path equality compares components, and Windows parses both / and \
as separators, so comparing as Paths keeps the readable literal and holds
on every platform. Both assertions (project pass and global pass) are
changed; they are the only /state/... string comparisons in the file, and
the failing job's log shows no other failed test.

Evidence

• Failing run on main (63f791d): https://github.com/SocketDev/socket-patch/actions/runs/43562815783, job test (windows-latest, 1); that test is its only failure.
• Green on Windows at this head (5fffc73): a PR run skips the Windows test legs, so I dispatched CI with os=windows on the branch. https://github.com/SocketDev/socket-patch/actions/runs/43600463417: test (windows-latest, 1) (the leg that failed on main), test (windows-latest, 2) and ci-ok (windows only) all pass.
• cargo test -p socket-patch-cli --lib: 943 passed, 0 failed (Linux).
• cargo clippy --locked --workspace --all-features -- -D warnings (CI's command): clean.
• rustfmt --check on the touched file: clean.
• /code-review high on the diff: no findings. Cursor Bugbot: no issues.

Note: cargo clippy -p socket-patch-cli --all-targets --all-features -- -D warnings
fails locally on unmodified origin/main too (duplicate_mod and
format_in_format_args in test targets). CI doesn't lint test targets, so
it isn't visible there, and it isn't part of this change.

https://claude.ai/code/session_011f9aePVwfWQikf61EL6Ec7

---

Generated by Claude Code

[Mikola Lysenko (mikolalysenko)]

bugbot run

---

Generated by Claude Code

[cursor]

PR Summary

Low Risk
Test-only assertion change with no impact on production hot-patch behavior.

Overview
Makes a_start_is_a_project_pass_then_a_global_pass assert project and global manifest locations with Path equality instead of raw string equality, so Windows CI passes when passes() stores paths built via Path::join (backslashes) against forward-slash literals.

No runtime behavior changes—only the test expectation is corrected, with a short comment explaining why Path comparison is used.

^{Reviewed by Cursor Bugbot for commit 5fffc73. Configure here.}

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 5fffc73. Configure here.}

[Mikola Lysenko (mikolalysenko)]

[agent] PR shepherd: review brief for the approver

What it does: This only changes a test. hot_patch::tests::a_start_is_a_project_pass_then_a_global_pass now compares the two manifest paths as Paths, not as strings, so it passes on Windows, where Path::join writes \\. That should turn the nightly main run's test (windows-latest, 1) leg (and ci-ok on main) green again (#2080).

Risk: Low. Test-only, two assertions in one file, and no product code changes.

Look here

• socket-patch/crates/socket-patch-cli/src/hot_patch.rs

  Lines 622 to 627 in 5fffc73

  	// Compared as paths: the separator `Path::join` writes is the
  	// platform's, and a `/` in a literal is one on every platform.
  	assert_eq!(
  	Path::new(&project.common.manifest_path),
  	Path::new("/state/project/manifest.json")
  	);

  : the project-pass assertion

• socket-patch/crates/socket-patch-cli/src/hot_patch.rs

  Lines 634 to 637 in 5fffc73

  	assert_eq!(
  	Path::new(&global.common.manifest_path),
  	Path::new("/state/global/manifest.json")
  	);

  : the global-pass assertion

• socket-patch/crates/socket-patch-cli/src/hot_patch.rs

  Lines 225 to 230 in 5fffc73

  	fn passes(args: &ScanArgs, state: &Path) -> Vec<(&'static str, ScanArgs)> {
  	let with_manifest = |mut pass: ScanArgs, scope: &'static str| {
  	pass.common.manifest_path = state
  	.join(scope)
  	.join("manifest.json")
  	.to_string_lossy()

  : passes(), which builds the path with Path::join (unchanged)

Verified

• Read the diff against Windows: hot_patch test compares a Path::join result against a /-joined string (main CI red since #2068) #2080's failing log. Only these two /state/... string comparisons exist, and Path equality compares components, so / and \\ match on Windows. The assertions still check the full path on every platform.
• The workflow_dispatch run on this head, https://github.com/SocketDev/socket-patch/actions/runs/43600463417, passed test (windows-latest, 1), test (windows-latest, 2) and ci-ok (windows only).
• cargo test -p socket-patch-cli --lib hot_patch passes locally on Linux (14 tests).
• The PR checks on the head are green (ci-ok, clippy). The branch is up to date with main. There are no open review threads, Bugbot found no issues, and CHANGELOG.md is untouched.

Changes I made: none.

Open questions: none.

---

Generated by Claude Code