One-off doctest failures in a fresh install because of matplotlib

[defeo]

When running Sage doctests for the very first time (with an empty DOT_SAGE directory):

sage -t --long src/sage/geometry/fan.py
**********************************************************************
File "src/sage/geometry/fan.py", line 2887, in sage.geometry.fan.?.plot
Failed example:
    fan.plot()
Expected:
    Graphics object consisting of 31 graphics primitives
Got:
    doctest:273: UserWarning: Matplotlib is building the font cache using fc-list. This may take a moment.
    Graphics object consisting of 31 graphics primitives
**********************************************************************
1 item had failures:
   1 of   3 in sage.geometry.fan.?.plot
    [492 tests, 1 failure, 9.60 s]
sage -t --long src/sage/categories/coxeter_groups.py
**********************************************************************
File "src/sage/categories/coxeter_groups.py", line 1681, in sage.categories.coxeter_groups.CoxeterGroups.ElementMethods.bruhat_lower_cov
ers
Failed example:
    P.show()
Expected nothing
Got:
    doctest:273: UserWarning: Matplotlib is building the font cache using fc-list. This may take a moment.
**********************************************************************
1 item had failures:
   1 of  15 in sage.categories.coxeter_groups.CoxeterGroups.ElementMethods.bruhat_lower_covers
    [461 tests, 1 failure, 18.31 s]

Of course, the failures go away when the tests are rerun. Any way we can avoid these failures?

[jdemeyer]

Running tests inside a fresh docker container could be a good way to check for problems like this.

[embray]

Still don't have automated builds of the sagemath-develop container running yet, but hope to have that sooner or later.

[jdemeyer]

Can't we simply ignore these warnings?

[defeo]

Replying to jdemeyer:

Can't we simply ignore these warnings?

What's the standard way to ignore warnings in doctests, again?

[embray]

I wonder if the initial sage startup shouldn't also take care of this.

[jdemeyer]

Replying to embray:

I wonder if the initial sage startup shouldn't also take care of this.

That's not really possible because the relevant matplotlib files are user files (inside $DOT_SAGE, which is typically in $HOME), not system files.

[embray]

I think I see what you're saying--I'm so used at this point to running my own copy of sage that it's warped my viewpoint in ugly ways :)

But isn't there some initial creation of user files in $DOT_SAGE the first time a user runs Sage? That's what I meant.

[mkoeppe]

Upstream issue: ​https://github.com/matplotlib/matplotlib/issues/5836

[jdemeyer]

New commits:

​e2083fc

Build Matplotlib font cache before running doctests

[embray]

Huh, I thought this was already fixed. I haven't run into it in a while.

[vbraun]

sage -t --long --warn-long 81.6 src/sage/tests/startup.py
**********************************************************************
File "src/sage/tests/startup.py", line 6, in sage.tests.startup
Failed example:
    'numpy' in sys.modules
Expected:
    False
Got:
    True
**********************************************************************
File "src/sage/tests/startup.py", line 8, in sage.tests.startup
Failed example:
    'pyparsing' in sys.modules
Expected:
    False
Got:
    True
**********************************************************************
1 item had failures:
   2 of   7 in sage.tests.startup
    [6 tests, 2 failures, 1.47 s]

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​dd1dfe4

Build Matplotlib font cache before running doctests

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​59e671e

Build Matplotlib font cache before running doctests

[jdemeyer]

New version which imports matplotlib in a forked child process.

[git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

​7169df8

Build Matplotlib font cache before running doctests

[jhpalmieri]

I'm seeing these failures again in 8.3.beta1.

[jhpalmieri]

See #25375 for a followup ticket.