Veselin Dobrev

To be more specific, in the above setting where I see failures, I also enabled debug more in MFEM with `MFEM_DEBUG=YES`. Some of the failures may be due to checks...

As I suspected the above failures are not specific to HIP builds -- the same issues can be seen with CUDA (HYPRE + CUDA + UVM and MFEM + MPI...

I just pushed a fix for one of the issues -- this came up recently in another context too, so thought it's good to address it.

The way to enable debug mode in the CMake build system is to set the build type to `Debug` instead of the default `Resease`, e.g. by adding the flag `-DCMAKE_BUILD_TYPE=Debug`...

Another potential issue on Lassen could be that you need to tell CMake how to properly run parallel jobs, e.g. using `-DMFEM_MPIEXEC="lrun" -DMFEM_MPIEXEC_NP="-n"`.

> This fails because `miniapps/meshing/pmesh-fitting.cpp` terminates if `HYPRE_USING_GPU` is defined. This is an oversight -- examples that return `MFEM_SKIP_RETURN_VALUE` (i.e. `242`) have to manually be skipped in the CMake build...

> I commented out these lines and the test passes, so it might be ok to remove this now? I'm not sure `pmesh-fitting.cpp` actually works correctly with `HYPRE_USING_GPU`, even if...

@helloworld922, the issue you see is related to this issue: #4351. I'm testing a fix for that and I'll push it when ready.

@helloworld922, I was not able to reproduce the issues you mentioned with HYPRE+HIP+UVM builds.

This PR is now [under review](https://github.com/mfem/mfem/blob/master/CONTRIBUTING.md#pull-requests) (see the table in the PR description). To help with the review process, please **do not force push** to the branch.