codespell doesn't ignore files in .gitignore

[mr-cal]

Here's a reproducer:

> python3 -m venv venv
> tox

lint-codespell: commands[0]> codespell --toml /home/developer/dev/starbase/pyproject.toml
./venv/lib/python3.10/site-packages/tox/tox_env/python/api.py:53: interpeter ==> interpreter
./venv/lib/python3.10/site-packages/pip/_internal/vcs/bazaar.py:64: readd ==> re-add, read
./venv/lib/python3.10/site-packages/pip/_internal/vcs/subversion.py:90: readd ==> re-add, read
./venv/lib/python3.10/site-packages/pip/_internal/utils/misc.py:340: becouse ==> because
...

Assuming it will be an easy fix, but I've hit my mental quota for linting issue today.

[lengau]

It looks like this is blocked by codespell-project/codespell#1887

I tried a workaround of feeding it the output of git ls-files, but the result in the snapcraft repo was that files that had been skipped previously were no longer skipped since they'd been explicitly added on the command line. Another option would be to filter the output of git ls-files, but that would also be reduplicating the skip configuration, and at minimum we'd probably want a comment in pyproject.toml about not using skip. I think for now it's best to just add anything extra we want to the skip parameter.

Leaving open for further discussion and marking as blocked since the "full"/straightforward fix is blocked from upstream. If we decide to do a full workaround with filtering the output of git ls-files we should probably put it as an answer in the upstream bug too.

[mr-cal]

I think for now it's best to just add anything extra we want to the skip parameter.

Agreed, that seems like the best workaround for now!

[setharnold]

A friend maintains a GitHub action to check spelling; I understand it's built around git ls-files as a starting point, so it might already have the feature you're after: https://github.com/marketplace/actions/check-spelling

The MIR team has been trialing check-spelling ( canonical/ubuntu-mir#19 ).

Edit: Here's an example of check-spelling in the wild, microsoft/terminal#15107 (comment)