[v0.11] Vendor the Version package, use it instead of distutils

[Stannislav]

Fixes #2724

Cherry-picked a0f7bf8 / #2659 as discussed in #2724

[mwaskom]

Thanks @Stannislav! Looks like CI is not set up to run on PRs against the v0.11 branch; would you mind adding that as part of this PR?

[codecov]

Codecov Report

Merging #2766 (65d91cb) into v0.11 (e8a83c8) will increase coverage by 0.04%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##            v0.11    #2766      +/-   ##
==========================================
+ Coverage   96.98%   97.03%   +0.04%     
==========================================
  Files          17       17              
  Lines        6233     6233              
==========================================
+ Hits         6045     6048       +3     
+ Misses        188      185       -3     

Impacted Files	Coverage Δ
seaborn/_core.py	98.13% <100.00%> (ø)
seaborn/_statistics.py	99.02% <100.00%> (ø)
seaborn/axisgrid.py	95.99% <100.00%> (ø)
seaborn/categorical.py	98.00% <100.00%> (ø)
seaborn/rcmod.py	100.00% <100.00%> (ø)
seaborn/_testing.py	98.11% <0.00%> (+5.66%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update e8a83c8...65d91cb. Read the comment docs.

[Stannislav]

@mwaskom has the v0.11 branch been tested recently? Not sure some failing tests have to do with the changes in this PR. I'll try and look into the details tonight.

[mwaskom]

@mwaskom has the v0.11 branch been tested recently? Not sure some failing tests have to do with the changes in this PR

no, I wasn’t expecting to do an 0.11.3 so it’s been dormant for a while. I don’t think the test failures are caused by this pr; matplotlib broke some tests with their 3.5 release (by changing an API that only the tests hit); they’ve also added some deprecation warnings that get a lot of false alarms and add noise.

I’m less sure what’s happening with the failing setup in the doc build, but also doubt this PR caused it.

[Stannislav]

@mwaskom the docs build is failling because of sphinx_bootstrap_theme==0.7.1 in docs/requirements.txt. The problem is that this package is using setup(use_2to3=True, ...) in setup.py, which was removed in setuptools>=58. To reproduce

pip install 'setuptools>=58'
pip install 'sphinx_bootstrap_theme==0.7.1'

The next higher version is sphinx_bootstrap_theme==0.8.0 and it fixes the error. Any reason not to bump? I see it was updated on master in #2704.

[mwaskom]

The next higher version is sphinx_bootstrap_theme==0.8.0 and it fixes the error. Any reason not to bump? I see it was updated on master in #2704.

Ah yeah, that sounds right.

[Stannislav]

@mwaskom I bumped sphinx-bootstrap-theme and cherry-picked #2690 to try and fix the tests after matplotlib 3.5 upgrade. Is it possible to re-trigger the CI to try and test?

[mwaskom]

Thanks @Stannislav sorry this is turning into such a headache for you. I'm not sure why new contributors need every PR push to be approved for CI (rather than just the first one).

[Stannislav]

Hey @mwaskom no worries. Maybe the approval is needed because it's a forked branch.

Anyway, a small change adopted from master (from #2413) fixed the test locally, so we can try to retest.

[mwaskom]

Maybe the approval is needed because it's a forked branch.

Yeah — I think it's to stop people from opening a PR that mines crypto in CI, so it makes sense to require approval for the first run, but I wish there were a checkbox to say "run CI on subsequent pushes for this PR".

In any case, everything looks good here now — thanks!

[EwoutH]

Could a new release on the v0.11.x branch be tagged, which includes this change?