Open termi-official opened 11 months ago
L2 spaces have no dofs associated with faces (i.e. dofs that are shared with face-neighbor elements) and consequently they have no FEs associated with faces and no order associated with faces. Thus, producing an error if a user asks for the face order in a DG space is the correct behavior.
When you say "crashes", do you mean the code segfaults? If that is the case, we need to fix this and call one of the error macros, MFEM_ABORT
, MFEM_VERIFY
, or similar.
The provided MWE unfortunately just segfaults when compiled with MFEM_USE_LIBUNWIND=YES
and MFEM_DEBUG=YES
. See:
> ./mwe_l2_faceorder_crash -o 0
Options used:
--mesh ../data/star-hilbert.mesh
--order 0
zsh: segmentation fault (core dumped) ./mwe_l2_faceorder_crash -o 0
Trying to use the Kelly error indicator on L2 spaces crashes. While the Kelly error indicator is in not necessarily the recommended method for DG problems it should at least not crash. This can be tracked down to a crash in
GetFaceOrder
. Discovered in while testing https://github.com/mfem/mfem/pull/3693 , but resolving this is a separate issue.Minimal reproducer:
We might want to add back these tests after the issue has been fixed to capture a regression in the error indicator, in addition to test coverage for
GetFaceOrder
on L2 spaces.