noqa functions chaotically or not at all

[d-w-moore]

In ruff-linter workflow, noqa is not processed as expected.

1. # noqa at head of a file is flagged as an unused directive.
2. # noqa: <code>... directives often do not produce the desired effect. In particular the addition of such lines often leads to their being flagged as unused. This and the first point run counter to our stated team goal of trying to attain a green condition in the ruff-checks section of the linter workflow.
3. occasionally when a source line is quoted along with line number and lint check error under "Annotations" above the workflow output, I've found that the line does not actually exist in the current state of the PR. (Perhaps owing to an artifact left over in a .ruff_cache/ directory?)

All of the above results in the lack of progress toward merging a body of code changes, once all reasonable linter-requested corrections have been made to the PR.

[alanking]

Can you provide links / screenshots / more specific anecdotes? It's a little hard to diagnose the problem without something to look at. Thanks

This may be more appropriate in https://github.com/irods/irods_reusable_github_workflows as a git_ruff issue, but we can hold on that til we have a better understanding of what is happening.

Also, for everyone's reference: https://docs.astral.sh/ruff/linter/#disabling-fixes

[alanking]

# noqa at head of a file is flagged as an unused directive.

According to the Ruff docs, to disable all the warnings/fixes for a file, you need # ruff: noqa (not just # noqa).

# noqa: <code>... directives often do not produce the desired effect. In particular the addition of such lines often leads to their being flagged as unused. This and the first point run counter to our stated team goal of trying to attain a green condition in the ruff-checks section of the linter workflow.

This could be a Ruff-specific issue. Perhaps the docs can shed some light on this situation.

occasionally when a source line is quoted along with line number and lint check error under "Annotations" above the workflow output, I've found that the line does not actually exist in the current state of the PR. (Perhaps owing to an artifact left over in a .ruff_cache/ directory?)

This could be a git_ruff issue, but would like to see an example to be sure.

[d-w-moore]

@alanking @SwooshyCueb ....

I tried to read and trace through the github workflow to see how I might find the git_ruff action/script locally either
- in a Bash shell
- within a Vim or IDE window.
- as a subcommand of git (this would be ideal, since - if it exists - it could limit the ruff reports to the diff itself.)

Ran into a block in that I can't seem to decode what the workflow is doing.

[SwooshyCueb]

The script can be run in a bash shell. if you pass --help it will tell you how to use it. It lives in the reusable workflows repository.

[d-w-moore]

@alanking
Here's an example - see https://github.com/irods/python-irodsclient/actions/runs/24040558975/job/70110952905?pr=808

It's being recommended by ruff check that I remove the noqa: D400 comment, but as soon as I do, I am fairly confident D400 will once again be flagged as an error. (The preceding doc string does not end in a period - nor should it, in accordance with my own understanding of English grammar.)

Catch 22. You leave it as it is, you have a red indication. You take it away, and get a different red indicaation

Neither leaves you with a green status on ruff-check.

Let me know if you have a suggestion, @SwooshyCueb

[d-w-moore]

The script can be run in a bash shell. if you pass --help it will tell you how to use it. It lives in the reusable workflows repository.

Right. I'm just not sure where the git_ruff script lives.

[SwooshyCueb]

@alanking Here's an example - see https://github.com/irods/python-irodsclient/actions/runs/24040558975/job/70110952905?pr=808

It's being recommended by ruff check that I remove the noqa: D400 comment, but as soon as I do, I am fairly confident D400 will once again be flagged as an error. (The preceding doc string does not end in a period - nor should it, in accordance with my own understanding of English grammar.)

Catch 22. You leave it as it is, you have a red indication. You take it away, and get a different red indicaation

Neither leaves you with a green status on ruff-check.

Let me know if you have a suggestion, @SwooshyCueb

I think it's a bug with ruff. sometimes it works as expected, sometimes it doesn't. in cases where it doesn't, just remove the noqa and ignore the check.

[SwooshyCueb]

The script can be run in a bash shell. if you pass --help it will tell you how to use it. It lives in the reusable workflows repository.

Right. I'm just not sure where the git_ruff script lives.

https://github.com/irods/irods_reusable_github_workflows/blob/main/git_ruff.py