Elliott Sales de Andrade

No, as I [mentioned earlier](https://github.com/matplotlib/matplotlib/pull/29816#issuecomment-2762857817), that barely changed anything. I'm trying to see if that's true or whether I didn't set it up right.

Ah, I had set the `_classic_test` style to `hinting_factor=1`, but I missed this testing setup: https://github.com/matplotlib/matplotlib/blob/7d5d0278fb5296a368a1ef910e67a8fa9a1f2fb8/lib/matplotlib/testing/__init__.py#L20-L23 With that removed/ignored, I get 1018 test failures, which seems more expected. Many of...

I've put together a script to show a few results without having to go through all the test images. Test script ```python import string import matplotlib.pyplot as plt from matplotlib.testing...

And one more, if we change `text.hinting` from `none` to `default`, we get:  Note that our _actual_ defaults are `text.kerning_factor=0`, `text.hinting_factor=8`, `text.hinting='force_autohint'`, while testing uses 6, 8, `none`, respectively.

As mentioned in the dev call, the above images show the changes as they are _for tests_ to the new versions, but this does not correctly show how things would...

> `text.hinting_factor` was 8 in both `classic` and `mpl20` and with FreeType 2.6.1, I would not have suggested changing it. But with FreeType 2.13.3, this seems more reasonable. Oh, as...

While looking at #29936, I came across a bug in the kerning calculation for mathtext; I've added the fix here in https://github.com/matplotlib/matplotlib/pull/29816/commits/6c7b9d02b98be7f4291fb1d16db39e3de0dee3aa It only affects 27 tests, but I can...

For the minimum version builds, it looks like we're hitting https://github.com/mesonbuild/meson-python/pull/424, which was fixed in meson-python 0.13.2, so I guess we'll have to bump that requirement (but it's already 0.13.1,...

Hmm, something seems a bit off with the kerning. For example, https://github.com/matplotlib/matplotlib/pull/29816/files#diff-a98ee9405471b4a49ef6fa8d8083e4a485936d4d8f8568782b8e1d70bc311f3e spreads the `Te` apart, but the opposite happens once libraqm is involved, so there may be some unnecessary...

After discussing it a bit, I think the kerning is fine (or as good as it'll be.) I was misled by only testing "Text" and not "2D Text" as in...