Jan Nikl

Also as @pazner mentioned in #4091 , we could unify the way how ownership of data is handled 😉 . The same pattern of raw pointers, flags and eventually some...

I totally agree, the same pattern for handling ownership is repeated at many places in the code, which is very prone to a mistake like noninitialized flag in one of...

Yes, I agree it should be discussed for MFEM 5.0 in a separate issue perhaps 😉 This PR, however, is rather meant for MFEM 4.7 (and actually was in my...

Yes, you should use `VectorFE` variants, because Nedelecs are vector finite elements, i.e., certain components are stored at DOFs. This should not be confused with `Vector` variants, which assume scalar...

@v-dobrev , as suspected, the build fails 😏 :smile:

Yeah, that Ubuntu is really weird :face_with_spiral_eyes: , it looks like `pow(float, float)` gives `double` there. That should not be according to C++98 already. However, my theory is that it...

Yes, seems to be working 😉

I added [najlkin/add-github-ci](https://github.com/mfem/github-actions/tree/najlkin/add-single-ci) also to github-actions for a cleaner configuration 👍 .

It seems to be all working 👍 , just this GitHub web app shows it is waiting for some tests, but they are not present in the setup anymore 😏

Based on the discussion, this is what needs to be done (in order): 1. mfem/github-actions#11 must be merged 2. the new head of `mfem/github-actions:master` should be tagged (`v2.5` probably) 3....