fix: complete path arguments before later options

[tomatotomata]

What changed

When Bash is completing a custom path argument and the next token already starts with -, the generated completer now keeps the path completer active. This covers editing a command such as ... -d -w and prevents the later option from being inserted as the path value.

Verification

• python -m compileall -q shtab
• git diff --check
• Added a regression test covering a directory argument followed by another option.

I could not run the full pytest suite on this Windows checkout because the repository's POSIX pty test helper requires termios, and the local environment also has incompatible pytest recording plugins. Let me know what you think.

[casperdcl]

Unsure what your intention is. Is it related to #116?

In any case the test is failing.

[tomatotomata]

I traced this back to #116. The generated Bash function was keeping the path completer active, but it was still passing the already-typed option token (-w) into that completer, so it could not list the directory entries.

Commit 5df353b keeps the path completer selected and passes an empty word when the current token starts with -. That makes the middle-of-command case complete the pending -d path argument while leaving later options in place. The existing regression now returns folder for COMP_WORDS=(test -d -w); COMP_CWORD=2.

Validation on this checkout:

• focused Bash completion scenario: passed
• python -m compileall -q shtab: passed
• git diff --check: passed

The full pytest suite still cannot start on this Windows checkout because its POSIX pty helper imports termios. Let me know what you think of this narrower fix.

[casperdcl]

this currently breaks examples/pathcomplete.py:

pathcomplete -<TAB>   # doesn't list --options
pathcomplete . -<TAB> # correctly lists --options

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 85.13%. Comparing base (22cc55d) to head (65ca5cc).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #248   +/-   ##
=======================================
  Coverage   85.13%   85.13%           
=======================================
  Files           3        3           
  Lines         471      471           
  Branches       92       92           
=======================================
  Hits          401      401           
  Misses         39       39           
  Partials       31       31           

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[tomatotomata]

I reproduced the regression in the generated Bash state machine. The path completer should still win when an option token is actually the value for a pending option argument, but after a positional path is complete the user should be able to start a new - option. I adjusted that branch to distinguish those two states and added a regression test for COMP_WORDS=(test . -), while keeping the existing middle-of-command path test. The generated Bash script passes bash -n, Python compilation and git diff --check pass, and the fix is on d35ce54. Let me know what you think.

[casperdcl]

Thanks! Fixes #116