[hgunduzoglu]

Part of #2497 (1/3).

Problem

rtk git status (compact path) reported a clean working tree with exit 0 when git status actually failed for any reason other than "not a git repository" — e.g. a corrupt index, a lock conflict, or unreadable refs. The non-compact path already guards on result.success() (git.rs:824); the compact path only special-cased the not-a-repo string and let every other failure fall through to the formatted summary + Ok(0).

# repo with a corrupted .git/index
$ git status            # native
fatal: .git/index: index file smaller than expected     # exit 128
$ rtk git status        # before this PR
Clean working tree                                       # exit 0  ✗

Fix

Replace the narrow "not a git repository" check with a general if !result.success() guard (mirroring the non-compact path). The friendly Not a git repository message is preserved for that common case; any other failure surfaces git's stderr and returns its real exit code.

$ rtk git status        # after
fatal: .git/index: index file smaller than expected
                        # exit 128  ✓ (matches native git)

Testing

• New test test_run_status_compact_propagates_non_repo_failure: inits a temp repo, corrupts .git/index, asserts run_status returns a non-zero exit (uses git -C via global args; assertion is != 0 so it's exit-code-agnostic and cross-platform).
• cargo fmt --all --check && cargo clippy --all-targets && cargo test — clean, 2200 passed.
• Manual verification vs native git:
  • corrupt index → exit 128 (was 0)
  • clean repo → * main / clean — nothing to commit, exit 0 (no regression)
  • dirty repo → ?? f2, exit 0 (no regression)
  • non-git dir → Not a git repository, exit 128 (friendly message preserved)

Same masking-failure class as #1581 / #1535 / #2446 / #2494. Sibling fixes for run_worktree and run_stash will follow as separate PRs under #2497.

[aeppling]

Hey @hgunduzoglu thanks clean fix, could you remove comment in all your conditions since this is pretty explicit by code ? This would just avoid accumulating noise in code.

I'll merge the three parts then.
Thanks for this contribution, looks good to me by code

[hgunduzoglu]

Done — dropped the explanatory comments on the guards across all three (#2498, #2499, #2500); the code speaks for itself and the rationale lives in the commit messages / #2497. Tests still green. Thanks for the review!

[aeppling]

ci checks are manually triggered :)

[hgunduzoglu]

ci checks are manually triggered :)

windows tests returns error, I wanted to check whether it was temporary or related to the code