[KSchlobohm]

Description

Use uv run pytest -n auto instead of -n 2 in the existing six-job OS/Python matrix. Let pytest-xdist select the worker count for each runner. Keep the full suite, matrix, dependencies, fixture behavior, and separate Ruff job unchanged.

Builds on #5051, which already added pytest-xdist and faster test-server shutdown.

In a three-attempt fork experiment, auto selected four workers on Ubuntu and Windows and five on macOS. Ubuntu test steps were approximately 23% faster than two workers. Windows differences were within runner noise; macOS coverage was insufficient for a speed claim. This is observed evidence, not a guaranteed speedup.

Testing

• Tested locally with uv run specify --help
• Ran existing tests with uv sync && uv run pytest
• Tested with a sample project (if applicable)

Hosted experiment at source commit 380663f9474cd1257a440813dc9d63229cea9c57:

• uv run pytest -n auto: Ubuntu and Windows passed on Python 3.13 and 3.14 in all three attempts. macOS/Python 3.14 passed once; the other five macOS auto jobs were cancelled without running tests because hosted runners could not be acquired.
• Across all worker configurations (2, 4, auto), 42 test jobs succeeded and 12 were cancelled. No completed test job failed. The overall workflow was not green because of the cancellations.
• These are experiment results, not a full-matrix validation of this rebased one-line patch. The PR's normal CI run must provide that evidence.

Local Windows/Python 3.14.8 checks on this patch:

uv sync --extra test
uv run specify --help
.venv\Scripts\python -m pytest tests\test_security_workflow.py tests\specify_cli\integrations\test_installed_adapters.py -k 'test_security_workflow or test_round9_optional_metadata' -n auto --tb=short -q
git diff --check

Results: dependency setup, CLI help, and whitespace checks passed; 80 focused tests passed with eight auto-selected workers, including valid and invalid adapter metadata cases.

The broader command .venv\Scripts\python -m pytest tests\test_security_workflow.py tests\specify_cli\integrations\test_installed_adapters.py -n auto -q was interrupted after failures and stalled progress. A repeat with -x --tb=short ended with 302 passed, 2 skipped, and 3 failed. All three captured failures were WinError 1314 (missing symlink privilege). Repeating those three cases with upstream's -n 2 reproduced all three failures:

.venv\Scripts\python -m pytest tests\specify_cli\integrations\test_installed_adapters.py::test_same_adapter_key_in_different_projects_uses_its_own_paths tests\specify_cli\integrations\test_installed_adapters.py::test_symlinked_storage_never_follows_target 'tests\specify_cli\integrations\test_installed_adapters.py::test_output_paths_are_validated_against_selected_project[False]' -n 2 --tb=short -q

No local full-suite pass is claimed. Sample-project testing is not applicable to this CI-only change.

AI Disclosure

• I did not use AI assistance for this contribution
• I did use AI assistance (fill in the disclosure below)

AI disclosure: Prepared with GitHub Copilot in Auto mode; the underlying model and reasoning effort were not exposed. The agent autonomously prepared the one-line workflow change, ran local checks, inspected hosted results, and drafted this description within a human-approved scope. Earlier experimental work was AI-assisted as disclosed in the linked fork experiment.

[Copilot]

🔵 Needs a closer look

Full six-job validation of the rebased patch remains unverified, particularly macOS/Python 3.13 with automatic workers.

No open issues 🎉

What changed in this PR

Updates Spec Kit’s existing parallel test setup from #5051 so pytest-xdist selects the worker count automatically for each CI runner.

Changes:

• Replace pytest -n 2 with pytest -n auto.
• Keep the full suite, six-job OS/Python matrix, dependencies, fixtures, and separate Ruff job unchanged.

File	Description
.github/​workflows/​test.yml	Enable automatic pytest worker selection.

_{Review effort: Max}