So far, the PR supports therm, bf, it, bfit, tt, sf font classes. From what I have found so far, we also have aLatin Modern Math that covers all the glyphs for the math font. However,LM Math does not containbold, or italics, so we end up with something like this, where the lowercase Greek mappings are fetched from the latinmodermath font because the regular lm fonts do not contain those glyphs.:

In LaTeX, I believe that there is a small set of bold glyphs supported so Fake Bold is being utilized(?).https://tex.stackexchange.com/questions/639337/latin-modern-math-doesnt-have-a-bold-font-but-does-have-glyphs-for-unicode-blo

https://ctan.math.washington.edu/tex-archive/fonts/ps-type1/lm/doc/fonts/lm/lm-info.pdf

Also, theLatin Modern Math doc metions something aboutcal and frak font names to be excerpted by other packages.https://www.gust.org.pl/projects/e-foundry/lm-math/download/readme-latin-modern-math.txt/view

Not sure how to proceed further, so opening the PR up for discussion.

Sorry, I have missed this completely. I cannot really provide any relevant feedback though, except that it is probably better to support what can be supported and possibly extend it later. (Just state any current limitations in the what's new note.)

Wouldn't it make sense to add an otf-directory rather than putting the otf-files in the ttf-directory? Not sure it anything else must be adjusted though. I'll enable the wheel-building tests as well so one can download one of those wheels and see that everything is indeed included as expected.

Any idea why some of the tests are failing?

The last change was updated quite some time ago. Not sure if there was a CI issue at the time. I'll just close and reopen to check.

Looks like it is fetching Latin Modern font from a different directory than the font directory in matplotlib...