Doc: Continuous integration information

[tupui]

In the effort to improve contributing documentation (#12633), this adds information about CI.

Also fix a missing command in a quickstart guide.

Feel free to add more information and complete this 😃

[tupui]

I can also see that skipping multiple CI works properly 😅 I will just add to put this on a new line to make it clean (as this last commit).

[andyfaff]

I'm not sure that we want to write a table of what we check in CI, this changes regularly.
Perhaps the biggest aspect that might help is to ask users to run the relevant test suite locally before making a commit on a PR.

[tupui]

I'm not sure that we want to write a table of what we check in CI, this changes regularly.
Perhaps the biggest aspect that might help is to ask users to run the relevant test suite locally before making a commit on a PR.

That's why I tried to not list all of them and wrote matrix for some.

For local testing, I can put a line and refer to the existing doc and the checklist (EDIT: done)

[rgommers]

Thanks @tupui. This is a good start. Some comments to keep it more maintainable.

You are still missing some of the key we want for this section I think: how to run things locally if CI fails. In particular for the custom steps, like refguide check and linting. Can you add the exact invocations to run those, or better - add them in http://scipy.github.io/devdocs/dev/contributor/runtests.html#runtests and link to those?

[tupui]

Interesting, CI did not skip this time. The only difference is the blank line. I will force push just removing the blank line to check.

[tupui]

Confirmed. The blank line is incompatible and prevent the check to work. Also [skip github] is not working.

So I propose to remove the blank line (making it work with the blank line might be too involved for this PR) and use again [skip actions].

[rgommers]

Confirmed. The blank line is incompatible and prevent the check to work. Also [skip github] is not working.

@larsoner is this expected?

[larsoner]

@larsoner is this expected?

At least for [skip azp] the string comes from git log --max-count=1 --skip=1 --pretty=format:"%s", which just gets the commit "title" (first line) and not the "description" (anything that follows). This can be useful at least when doing a squash+merge so that intermediate commits with [skip azp] don't cause the main build to skip building. Personally I also don't mind putting the skips in the title, but I'm okay with expanding it to look in the description, too.

[rgommers]

Thanks for explaining @larsoner!

Personally I also don't mind putting the skips in the title, but I'm okay with expanding it to look in the description, too.

That would be great. I think skips naturally go at the end of the description.

[tupui]

FYI, I just used [skip ci] in gh-14022 and it skipped everything BUT azure.

[rgommers]

FYI, I just used [skip ci] in gh-14022 and it skipped everything BUT azure.

Yes, that has always been the case. [skip ci] is kind of the universal command (except Azure devs are really dense and refuse to honor it for PRs from forks). The per-provider ones are the pain point.

[tupui]

Shall I add anything else? Shall I write that the global skip is not working with azure so if you want to skip all you need to also add the instruction for azure?

[rgommers]

Shall I write that the global skip is not working with azure so if you want to skip all you need to also add the instruction for azure?

That doesn't work either, Azure just refuses to be skipped.

I think it looks good as is.

[rgommers]

Merged, thanks @tupui, all.