Jan Nikl

As we have discussed at the meeting, I have done something like this already here #1951 :wink:.

That `bcast` is needed because of the debugging check, where `values_counter` is compared with the essential vdofs, which are synchronized, if I am not wrong 🤔

> > That `bcast` is needed because of the debugging check, where `values_counter` is compared with the essential vdofs, which are synchronized, if I am not wrong 🤔 > >...

Ok, to add some concurrency here 😁 , I added this check (removed `pfes->GetLocalTDofNumber(i) == -1`) to #1951 .

@thartland , see above ☝️ , I have updated the description. That change from `CalcShape()` to `CalcPhysShape()` has a slight downside that the `ElementTransformation` must be retrieved even when it...

Ok, I have (indirectly) reverted the change of the interface of `BlockOperator` in #4136 . So everything should work as before, but still it is perhaps a good practice to...

@dylan-copeland put it in #4235

I am not sure about `bool generate_bdr = false;`. It is an intended behavior perhaps 🤔

To move things forward a little bit, @jandrej , @sohailreddy , what are your opinions on `generate_bdr = false`? Note that `ParMesh::Finalize()` is a public method and `Mesh::Finalize()` generates boundaries...

> I'm still not fully convinced by this solution, and wonder if it's even necessary to support generating boundary elements for parallel mesh objects (why wouldn't you just do it...