Hi @oscargus,

I have a working patch for boldsymbol in my local repository. Since #25359 is a dependency, I am unsure about opening a new PR.

Also, for generating the tests for boldsymbol, does it need all tests for the baseline_images, or similar to #22173 adding a lightweight test is fine?

Great!

I think a lightweight test is enough.

Hopefully, #25359 is merged soon, so maybe one can wait for a bit before adding it? It is possible to open a new PR as well and just state that it depends on #25359.

As mentioned in the linked PR, these changes currently do not build without PR #25359.

Xref #1366

Discussion there seems to have centered around "do we wish to include amsmath things in the mathtext implementation or keep it 'pure latex', for simplicity and scoping".

That said, the barrier for amsmath is breached by \genfrac and \binom already, so that could be an argument for including \boldsymbol

However, the behavior of \boldsymbol is not quite the same as bfit as noted by @devRD here.

My personal prior would be that if \boldsymbol is included, it should behave as close to the latex behavior as possible, and thus is not as simple as aliasing bfit, though the implementation may mirror bfit.

There are more things that are from ams and other packages (especially symbols). I'd say add what is useful.