TST: lift xfail from tests fixed upstream in matplotlib - #11315

Please note that this is a draft and includes a temporary change to tox.ini to try building against @tacaswell's branch.

Thanks! I converted this PR to draft status for now, as indicated.

@pllim, did I revert the correct changes from #11292? Looks like that PR might have been addressing two unrelated Matplotlib issues...

Re: Not sure what happened but seems like you reverted too much? All you need is to delete the lines added in the final diff of https://github.com/astropy/astropy/pull/11292/files

The tests were failing, and I noticed that @tacaswell's branch was well behind master, so I thought I would try reverting more. I'll go back to just removing those two xfail markers, but then I'll need to rebase his branch in order to test it.

It's failing with a whole bunch of 404 errors when it tries to download baseline images...
https://app.circleci.com/pipelines/github/astropy/astropy/4017/workflows/9d0697fe-5cf6-410e-b392-aa62b10ce629/jobs/84521

Is some piece of infrastructure down at the moment?

Does matplotlib use setuptools_scm now? If so, maybe because your fork doesn't have the tags necessary to build the correct version number?

Interesting thought. I pushed the tags to my fork, then try again.

mpldev is green! 👏

Woot! OK, I'll report back to Matplotlib that this fix will work for us.

Does matplotlib use setuptools_scm now? If so, maybe because your fork doesn't have the tags necessary to build the correct version number?

No, but we are working on switching: matplotlib/matplotlib#18971 currently we use versioneer.

I re-worked my mpl PR, is there an easy way to re-trigger this CI work?

I re-worked my mpl PR, is there an easy way to re-trigger this CI work?

You mean astropy's? Please rebase against master and push all of matplotlib's version tags to your fork. Then I can trigger it.

Done. I have to pushed tags to my fork in a very long time (it pushed 1.5.2 on 🤣 ) .

I triggered a build a little while ago, before you had pushed the latest tag, by pushing your branch to my fork. It's still passing :-)

matplotlib/matplotlib#19438 is in now, so this is ready for review.

Thanks! Remote failure unrelated and you can ignore coverage complain.

Before I get this in, can you please remove MATPLOTLIB_DEV altogether? I don't think it is used anymore.