Is this some known issue (two jobs fail)?

  Cloning https://github.com/larsoner/python-pytest-harvest (to revision doctest) to /tmp/pip-req-build-b8_nwlwr
  WARNING: Did not find branch or tag 'doctest', assuming revision or ref.
ERROR: Command errored out with exit status 1: git checkout -q doctest Check the logs for full command output.

Is this some known issue (two jobs fail)?

Fixed in master, will restart jobs

Azure jobs fail because they can't create the temp file. Is there a recommended mechanism that works?

@cbrnr try using our _TempDir() function?

Thanks, this should fix the problem. FWIW, the problem with tempfile.NamedTemporaryFile was that I opened it for writing and in the same block I wanted to read that file. This works on Unix, but not on Windows (and I think it is ugly anyway).

can you just add an entry in latest.inc ?

Sure. Have we already converged on a new format? I think we agreed to include the GitHub PR number, so my entry would look like this:

- :func:`mne.io.read_raw_edf` now supports EDF files with invalid recording dates (by `Clemens Brunner`_, [#8283](https://github.com/mne-tools/mne-python/pull/8283))

If you agree I could change the only other entry by @larsoner accordingly.

remember we use rst and not md. I am fine adding url to the PR

Right, so this would be:

- :func:`mne.io.read_raw_edf` now supports EDF files with invalid recording dates (by `Clemens Brunner`_, `#8283 https://github.com/mne-tools/mne-python/pull/8283`_)

Or leaving the contributor outside parens like before:

- :func:`mne.io.read_raw_edf` now supports EDF files with invalid recording dates by `Clemens Brunner`_ (`#8283 https://github.com/mne-tools/mne-python/pull/8283`_)

And if we really want the # in front, you or I could update the substitution code easily to use that instead of the gh- lead-in.

I think this would be a good idea. Where do I change this?

Found it, let's see if this works.

And if we really want the # in front, you or I could update the substitution code easily to use that instead of the gh- lead-in.

I think this would be a good idea. Where do I change this?

Can we have "GH-XXXX"? Seems more generic and, should we ever move to another platform, will help us distinguish

I'd prefer #XXX because it looks more familiar and other platforms (e.g. GitLab) also use this notation. Actually, the platform is not relevant for users, so showing GH-XXX might raise questions why GH is important, and should we switch to GitLab using GL-XXX would be similar. Instead, if we use #XXX we don't include the specific platform at all.

@larsoner could it be that the current substitution mechanism only works for issues?

See comment a few lines down:

 # issue/PR mode (issues/PR-num will redirect to pull/PR-num)

and try it (have to use code-mode so that GitHub doesn't substitute; copy-paste enter and it should redirect here):

https://github.com/mne-tools/mne-python/issues/8283

Sorry I don't get it. I'd like to link to a PR, how could the substitution distinguish between issues and PRs if both use identical (but separate) ranges starting from 1? We already have some issue links in our roadmap, so we can't just change everything to point to PRs.

Sorry I don't get it. I'd like to link to a PR, how could the substitution distinguish between issues and PRs if both use identical (but separate) ranges starting from 1

They are not separate really. They don't overlap. Try this link and see where it goes:

https://github.com/mne-tools/mne-python/issues/8283

Or directly try the output links

https://22434-1301584-gh.circle-artifacts.com/0/dev/whats_new.html

They are not separate really. They don't overlap.

🤯 TIL!

All green 🚀 !

@cbrnr try using our _TempDir() function?

IMO _TempDir use in tests should be discouraged, the pytest tmpdir fixture is more standard and cleaner. We can fix this later though

Thanks @cbrnr

@hoechenberger I'll backport this at the same time as #8287