Ali jawwad a4972da717 fix(scripts): count subdirectory-only dirs as non-empty in PowerShell (parity with bash) (#3137) ...

* fix(scripts): count subdirectory-only dirs as non-empty in PowerShell

Test-DirHasFiles (the documented PowerShell twin of bash check_dir) tested
non-emptiness with `Get-ChildItem | Where-Object { -not $_.PSIsContainer }`,
counting only top-level FILES and ignoring subdirectories. Bash check_dir
(`-n $(ls -A ...)`) and the PowerShell JSON-path contracts checks
(check-prerequisites.ps1 / setup-tasks.ps1, no PSIsContainer filter) both
count ANY entry. So a contracts/ directory whose only contents are
subdirectories (e.g. contracts/v1/openapi.yaml) was reported present by
bash, by bash JSON, and by PowerShell JSON, but [FAIL]/absent by PowerShell
text mode — the lone outlier.

Drop the PSIsContainer filter so Test-DirHasFiles counts any entry, matching
the other three code paths.

Add bash + PowerShell parity tests asserting a subdir-only contracts/ dir is
reported non-empty in both shells.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* review: accurate non-empty comment + drop doubled test prefix

Address review feedback on Test-DirHasFiles parity fix:

- Reword the common.ps1 comment so it no longer claims exact `ls -A` parity (Get-ChildItem omits hidden entries without -Force); it now points at the in-repo PowerShell JSON contracts checks as the matching reference and keeps the subdir-only-is-non-empty rationale.

- Rename test_test_dir_has_files_ps_... -> test_dir_has_files_ps_... to drop the doubled 'test_' prefix.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* test: assert dir-non-emptiness via stdout marker, not exit code

Address Copilot review: check_dir always exits 0 (it echoes the marker rather than setting an exit code) and Test-DirHasFiles returns a boolean (pwsh still exits 0 when it returns $false), so 'result.returncode == 0' validated nothing. Drop the misleading assertion and rely on the [OK]/checkmark marker in stdout, which is the actual behavioral signal; document why inline.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix: keep common.ps1 ASCII-only (PowerShell 5.1 compatibility)

My reworded Test-DirHasFiles comment introduced an em dash (U+2014), which tripped tests/test_ps1_encoding.py::test_ps1_file_is_ascii_only -- .ps1 files must stay ASCII for Windows PowerShell 5.1. Replace it with '--', matching the existing comment style in this file (e.g. the Resolve-SpecifyInitDir docstring).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* test: decode dir-parity subprocess output as UTF-8 explicitly

Address Copilot review: check_dir echoes the non-ASCII markers ✓/✗, and subprocess.run with text=True but no encoding decodes via the platform locale (cp1252 on Windows), which can raise UnicodeDecodeError or mangle stdout. Pin encoding='utf-8' on both the bash and PowerShell dir-parity helpers so decoding is deterministic across CI runners.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-29 14:38:39 -05:00

32 KiB

Raw Blame History

View Raw