Wow, that's bizarre, matplotlib 3.4.2 doesn't xfail, but the development version does.

Hmm, I see they merged matplotlib/matplotlib#18971 but we still see their dev version as Matplotlib: 3.4.2.post689+gb46e5039d.

But the more relevant question, do you know which matplotlib PR fixed this dpi issue and are you sure they forward-ported that PR back to master?

p.s. To avoid the bot from wrongly removing the label you assigned, please wait for the Triage job to complete first before you manually apply the labels defined in the bot rules. This is a known issue that has not been fixed by GitHub. Thanks!

Sorry, no idea about what the matplotlib PR is that fixed things...

In that case, maybe we should just revisit again in the near future until the forward-port happens. I don't think we can merge this until dev job is green again with this patch. What do you think?

Yes, makes sense. Though what do we do with tests currently failing with matplotlib 3.4.2?
Eg., https://github.com/astropy/astropy/pull/11720/checks?check_run_id=2547206829

Just skip the test for now?

tests currently failing with matplotlib 3.4.2?

Hmm... How about we say something like "if matplotlib >= 3.4.1 but !=3.4.2" -- that should work for both jobs? Then we can open a follow-up issue to revisit this logic in the future.

Hmm, looks like I didn't even write the test correctly. It may be simpler than we thought. Pushed to test that...

RTD failure is unrelated (All Sesame queries failed.). Dev jobs passed! 👏

OK, so still a bit strange, but I guess in matplotlib dev they are simply still thinking about how to do it... Anyway, will merge this so that the tests elsewhere start passing again!

Thanks!