Jan Nikl

Ok, @cjvogl , I have put together [example 4](https://github.com/mfem/mfem/blob/najlkin/mixed-face-forms-supp/examples/ex4.cpp) (in branch `najlkin/mixed-face-forms-supp`) with the Dirichlet BC imposed weakly through the constraint 👍 . Run it like normal, but `-bc` option...

I just changed a little bit how empty matrices are built in `Compute*FaceMatrix()` methods following the discussion in #4123 .

I just removed `virtual` as @cjvogl wanted anyway and there is now consensus on it (see #4399 ) :wink: .

Also look here #4128 , internal boundaries are sometimes generated when loading `ParMesh` from files 😮 . By coincidence, it also happened originally with a `ParSubMesh` in #4083 , which...

The problem with the template is that the specialization cannot be reused in another specialization, so everything would have to go to header files. However, that might not be a...

Ok, the code is much cleaner now without the terrible (and potentially dangerous) C-style and static casts 😉 The const variant I am not enforcing, which would make a lot...

A point for discussion, do we want a typedef for `TBlockOperator`? Something like `BlockHyperParMatrix`? It would make sense, because it appears very often in the code, but it might be...

Based on the results of [the poll](https://github.com/orgs/mfem/discussions/4223), I am closing this PR.

Or also my #1951 (which has been approved but not yet merged 🙄 )

Besides that, do you have a distributed mesh, i.e. load it as `ParMesh`? If you do, maybe look here #4128 , there is a known bug 🪲 and its fix...