Thanks for testing for looking at the rc!

It looks like this was intentially changed in fc73b77 / #12317 . We pinged @felixonmars for Arch but got no response.

Sorry for being inactive on that issue. Since Arch don't split packages I tend to provide mostly the default set. I didn't notice the tests installation change.

@tacaswell What I’m saying is that matplotlib install now defaults to tests = True, while still saying the opposite. It might be a side-effect of the linked changes, but likely not the intended behaviour. I’ll look at those files and see if I find something weird.

OK, indeed: https://github.com/matplotlib/matplotlib/pull/12317/files#diff-c28358ce9623af21f7b6e068f69085b7R662

I don’t think this was intended, am I right?

I think that was an intended change, but we failed to update the documentation / template. Is that going to be a problem for Arch?

@anntzer can you confirm the intent?

Also, thank you to both @ArchangeGabriel and @felixonmars for your work on Arch packaging 🎉 (it is what I use on my personal machines).

No, it won’t be an issue at all, we build two copies anyway (one with the tests to check everything but fonts work, one without the tests for distribution). So currently the two copies end up being identical, but if it gets confirmed that this was intended, then we will switch from enabling tests in one copy to disabling them in the other. ;)

Always installing tests was not an intended change of that PR (in fact as @tacaswell may remember I have a PR in the works for splitting the test install even more), looks like it got in accidentally... making a PR to revert that.
Edit: #14008.

Reverted in #14008

in fact as @tacaswell may remember I have a PR in the works for splitting the test install even more

No, my memory is a bit sieve like....

Thanks for following up on thin @anntzer !