source-control: exec-bit-check tests lose the REPO_SEQ increment in a subshell, so every fixture shares repo1

[kyle-sexton]

Context

Found by the repo-wide /code-tidying:batch-simplify sweep (62 groups / 986 files, branch claude/repo-code-tidying-batch-twhddw). The sweep's doctrine files latent defects instead of fixing them mid-sweep, because the fix changes observable behavior; this issue is that filing.

plugins/source-control/skills/pull-request/scripts/tests/exec-bit-check.test.sh:27-46,602-610 — mkrepo increments REPO_SEQ inside $( ); the increment never escapes the subshell, so every fixture shares $TEST_TMPDIR/repo1. The case-group-19b comment records this already biting once.

Proposed work

• Return the path without mutating the counter in a subshell (mutate first, then capture), or derive per-case names.

Acceptance criteria

• Two consecutive mkrepo calls produce distinct repo directories.

References

• plugins/source-control/skills/pull-request/scripts/tests/exec-bit-check.test.sh:27-46

Filed by the /code-tidying:batch-simplify closeout (Phase 8).

[kyle-sexton]

This was generated by AI during work-loop.

Already fixed on main. mkrepo in plugins/source-control/skills/commit/scripts/exec-bit-check.test.sh increments REPO_SEQ in the function body before the fixture subshell, then prints the path. Consecutive mkrepo calls therefore get repo1, repo2, … The leftover comment at the dis- fixtures describes the historical bug, not current behavior.

Closing as done.