✅ Done! Reviewed and approved #5619 ✅ It's a clean, test-only fix and the right approach: it stops the tests from recreating the exact macOS socket-length failure that the production code in #5559 was designed to avoid. Someone had already independently filed #5620 for the same regression, and it carries

ready-for-dev

, so the linked-issue gate on the PR now passes as well. What I checked: • Applied the diff to a local clone and ran the file: 47 passed (the tmux subset: 9 passed). • The new

counts_canonical_socket_bytes[macos]

/

[linux]

cases pass, so the 103/107 boundary is exercised from Linux CI too. •

ruff check

+

ruff format --check

are clean, and no production code is touched. Left two minor non-blocking notes in the review: the new test hardcodes the 103/107 socket limits (worth referencing

_MAX_TMUX_SOCKET_PATH

), and one path-selection test dropped its posix

skipif

— still safe because

_fits_tmux_socket

returns

True

off-posix, but the marker keeps intent explicit. Neither blocks merge. CI is green except the already-resolved PR-description run.