Skip to content

Decide How the auto-1947 Build-Profile Fix Proceeds Past Its Second Review Budget #2027

Description

@ptr727

Question

The fix for #1947 on branch feature/auto-1947 (handoff #1981) spent this push's two-round edit budget, and the local strict review pass on the result still raises findings on text the branch wrote or should have written. How should the lane proceed?

Where it stands

The branch now carries the #1982 answer (the local development loop scoped to the uv form, the pip-form type-check sentence qualified to declared directories), develop merged in, and a second round fixing the first pass's findings: the pip-form checker pointed at its environment, the VS Code task mirror, the CI-gate bullet, references/testing.md's default invocation, and the devcontainer check all scoped to the uv form. Head b1b4e37, pushed. Two pre-existing findings were filed as #2025 and #2026.

The second pass raised these, classed as introduced:

  1. The pip form's local environment is never defined. The text says to run pytest "from the environment its requirements*.txt installed", but names no command, and CI does three things (uv venv, one install over every requirements*.txt together, then uv pip install -e . where [project] exists). Installing only the requirements leaves a src-layout package unimportable, so local pytest fails where CI passes.
  2. SKILL.md's unchanged "CI runs the same clean-compile commands" now contradicts the pip-form text for the undeclared root, which CI type-checks with a bare uvx <checker>@latest while the local text points the checker at the environment.
  3. references/profiles.md says CI runs uv sync --frozen, while CI runs uv sync --all-groups --frozen, which installs dependency groups but not the [project.optional-dependencies] extras the same bullet allows for dev tools.
  4. references/profiles.md still says the profiles differ by third-party runtime dependencies showing up in pyproject.toml, which disagrees with the restated trait and with a pip form whose dependencies live in requirements*.txt.
  5. uv-only text the scoping missed: the Layout block's committed uv.lock, the Type checking paragraph's uv run pyright as the CI command, and the dependency-declaration axis naming only uv mechanisms.
  6. The skill tells a pip-form repo to adapt the vscode-tasks-python.json snippet, whose header names no pip-form adaptation and whose lint-only description also fits a pip-form directory, which would drop the pytest and coverage tasks the pip form owes.

Counts for this push: first pass 5 findings (4 introduced, 1 pre-existing, filed as #2025), second pass 7 findings (6 introduced, 1 pre-existing, filed as #2026).

Options

  1. Recommended: open the pull request with the branch as it stands and file the six findings as one follow-up issue. Reason: python-codestyle Profiles Reference Calls the Build Profile PEP 621, Then Admits a pyproject.toml With No [project] Table #1947's ask is met and every edit so far is a strict improvement on develop, while each round's findings are moving outward onto adjacent pre-existing text rather than converging, so another round is likely to produce the next set. The parked work opens its pull request into develop with Closes on promotion: #1947 and drives it, and the follow-up issue carries the pip-form local environment and the remaining uv-only surfaces.
  2. Authorize one more edit round fixing the six findings. Reason: they are all in the same skill, and fixing them leaves the pip form fully described in one pull request. The parked work takes a third round on this branch, a fresh local strict review pass, then opens the pull request if that pass raises nothing introduced.
  3. Narrow the branch to the references/profiles.md bullet alone and file the uv-command scoping as its own issue. Reason: keeps this pull request to exactly what python-codestyle Profiles Reference Calls the Build Profile PEP 621, Then Admits a pyproject.toml With No [project] Table #1947 asked. The parked work drops the scoping commits from this branch.

Belongs to

Handoff #1981 (track auto-1947), issue #1947. Branch feature/auto-1947, pushed. No pull request is open yet, so none is blocked.

Activity

  1. added
    decisionA question waiting on the maintainer, alone or beside whatever else the issue carries
    on Sep 29, 2026
  2. ptr727 commented on Sep 29, 2026

    @ptr727
    OwnerAuthor

    Answered by the maintainer: option 1, open the pull request with the branch as it stands and file the six findings as one follow-up issue.

  3. removed
    decisionA question waiting on the maintainer, alone or beside whatever else the issue carries
    on Sep 29, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions