Repository navigation
tests-macos: test_a_path_that_is_not_utf8_is_a_row_not_a_traceback cannot build its fixture on APFS (git worktree add refuses the non-UTF-8 path) #177
Description
Activity
CLAIMED — lane openRepoTools-3, session 83619980-06bb-4469-ae73-2a8c5e61981a@Eagle, 2026-10-06T18:04:03Z, for #177
Logged in
lanes/log/openRepoTools-3.mdat2f11530072e1eaf49f22fa37752f8dc9749a9b99.The three reads (lane-collision-protocol Rule 1):
1. existing `CLAIMED —` comments on opensoft/openRepoTools#177: none 2. `gh pr list --repo opensoft/openRepoTools --state all --search "#177"`, kept where the row or its body names `#177`: none 3. `git ls-remote --heads git@github.com:opensoft/openRepoTools` branches naming `177` as a whole token: none
On the
readygate run of #168 (head df0d691, run 37492625698, job 112368980621) tests-macos ended 1 failed, 1160 passed in 1:55:25. The one failure is #168's own testtests/test_lane_worktrees.py::test_a_path_that_is_not_utf8_is_a_row_not_a_traceback: its fixture runsgit worktree addwith a path containing a non-UTF-8 byte and git on macOS (APFS, which only stores valid UTF-8 names) refuses it withsubprocess.CalledProcessErrorbefore the assertion under test is reached.The tool under test is fine; the fixture cannot exist on that filesystem. Fix: build the non-UTF-8 directory with the raw bytes name first and
pytest.skip("filesystem refuses non-UTF-8 names")when thatOSErrors or when git refuses the worktree add; keep the test live on Linux (where it proves B2 of the Opus review).Filed as the landing-gate follow-up of #168 under Brett Heap's ruling 1 of 2026-09-16 pending his word on where the fix lands (#168's head, or the #169 fold-in that follows it).
Lane: openRepoTools-3