Peter Munch

> suggestion for renaming: is_mg -> matrix_free_operates_on_coarser_tria_level Sounds reasonable. There are other places where the term `is_mg` is used. I think similar to https://github.com/exadg/exadg/pull/640, we could rename these.

I am wondering whether it makes sense to extract a boundary mesh from a mesh with manifold. While refining the boundary mesh the new points will be positioned according to...

See also my comment https://github.com/dealii/dealii/pull/11566#issuecomment-761808020: > The function GridGenerator::extract_boundary_mesh() does not attach any manifolds (as documented in the documentation). However, since the function might refine the mesh, manifolds have to...

We already attach the flat manifolds: https://github.com/dealii/dealii/blob/732766ac2383f48e6fc987aea25249e70514affd/source/grid/grid_generator.cc#L8540-L8549 Everything else seems to be wrong and very error-prone as long as we don't have submanifolds.

> Please also add a changelog entry. For that, let me do some additional changes. At the moment, I am including `numerics/matrix_creator.templates.h` in my code, since the functions are not...

> Currently, the indent check fails, but the PR is too old and the logs are no longer available. @bangerth Done!

You should be able to remove that.

@kronbichler I think we can close this issue. Using `MatrixFree::intialize_dof_vector()` is in my opinion the right way if someone wants to use `MatrixFree`; together with @mschreter, we have improved the...

I have updated the PR. We are still long before release so that we could attempt to merge this now.

@Rombur Do you have a small test?