restart:247 and lanes:311 carry the same BSD-incompatible grep -v 'pattern' -- file shape PR #53 fixes at skills/restart/SKILL.md:58, and the hygiene checker's own text-split can't see it #54

Lane: openrepotools-3 (openRepoTools-3)

PR #53 (issue #34, Copilot review 5202775204) fixed skills/restart/SKILL.md:58's lread fence — grep -v 'not fetching; reading origin/' -- "$ln_err_file" puts the PATTERN before --, so a non-permuting getopt (BSD's, every macOS shell) reads the pattern as the first OPERAND and then reads -- itself as a second one, a FILENAME. Verified here, no macOS runner needed, by setting POSIXLY_CORRECT=1, which turns off GNU grep's own permutation and reproduces the exact BSD shape:

$ POSIXLY_CORRECT=1 grep -v 'not fetching; reading origin/' -- /tmp/x
grep: --: No such file or directory
/tmp/x:<the file's own content, now prefixed with its name>

That PR's own description names this as "the third site of the shape lanes and restart took at 29d3417" — and the other two sites are still live, untouched by that PR because its diff only ever reached SKILL.md's copy:

• restart:247 — grep -v 'not fetching; reading origin/' -- "$lc_err_file" >&2 || :
• lanes:311 — grep -v 'not fetching; reading origin/' -- "$local_err_file" >&2 || :

Both are byte-identical to the line #53 just fixed, and both are shipped commands a person runs directly (not only a skill's embedded fence), so this is live on every macOS workstation today: any OTHER line on that stderr stream — "this workstation's session records could not be read", say — reaches the person prefixed with a temp filename and preceded by a spurious grep: --: No such file or directory, instead of the clean forwarding the fence exists to do.

Why test_no_shipped_bash_ends_its_options_after_an_operand did not catch it, even though both files are in ALL_BASH: _end_of_options_after_an_operand (tests/test_repo_hygiene.py:2193) splits each line on a bare re.split(r"[|;&()]|\$\(|", line)before tokenizing — not quote-aware. The pattern text itself contains a; ('not fetching; reading origin/'), so the split cuts the command in two before the tokenizer ever sees the --, and neither half is left holding both an operand and the --that follows it. Confirmed:python3 -m pytest tests/test_repo_hygiene.py -k test_no_shipped_bash_ends_its_options_after_an_operand -qis 11 passed on unmodifiedmain, restartandlanes` included. A checker that calls a real defect clean is worse than no checker, because it is the one thing a reviewer trusts instead of reading the line.

Ask:

1. restart:247 and lanes:311: the same one-word fix /restart's lread fence keeps a read's stderr and drops only the known no-fetch notice; two module comments now point at the lists instead of restating a stale file count (#34) #53 applied — grep -v -e 'not fetching; reading origin/' -- "$file".
2. _end_of_options_after_an_operand: make the per-line split quote-aware (tokenize the whole line first, e.g. with something like the existing _shell_tokens, then split THAT token stream on the separator tokens, rather than regex-splitting the raw text) so a pattern argument containing ;, &, |, ( or ) can no longer hide this shape from the checker. A fixture built from this exact line, asserted red before the split fix and green after, belongs beside it.

🤖 Generated with Claude Code