Hard-to-understand error messages for simplex triangulation

[nfehn]

Creating a simplex triangulation by calling

dealii::TriangulationDescription::Utilities::
      create_description_from_triangulation_in_groups<dim, dim>(..., mesh_smoothing, triangulation_description_setting);

with the last two parameters set to

      mesh_smoothing = dealii::Triangulation<dim>::limit_level_difference_at_vertices;
      triangulation_description_setting =
        dealii::TriangulationDescription::construct_multigrid_hierarchy;

produces hard to understand error messages:

An error occurred in line <1283> of file <dealii/include/deal.II/grid/tria_accessor.templates.h> in function
    unsigned int dealii::TriaAccessor<<anonymous>, <anonymous>, <anonymous> >::vertex_index(unsigned int) const [with int structdim = 3; int dim = 3; int spacedim = 3]
The violated condition was: 
    ::dealii::deal_II_exceptions::internals::compare_less_than(corner, this->n_vertices())
Additional information: 
    Index 4 is not in the half-open range [0,4).

It seems as if construct_multigrid_hierarchy and/or (?) limit_level_difference_at_vertices is not supported for simplex. It would be helpful if certain functionality not supported for simplex triangulations were asserted with understandable error messages.

The documentation https://dealii.org/developer/doxygen/deal.II/namespaceTriangulationDescription_1_1Utilities.html#af155557fb9b46c44224a6aeeb2493cc8 seems to be incomplete as well. According to the note in the documentation, one parameter internally manipulates the other one, differing from the value set by the user? If there are proper asserts in the code, I do not see a need to extend the documentation. In the current setting with the present error message, however, I consider it very hard to use such functionality.

[bangerth]

@nfehn Can you create a small but complete program that shows the error? I'll be happy to fix the error then.

[bangerth]

@nfehn Ping?

[nfehn]

Sorry, I did not find time to construct a minimal program this week. Here you can see how I resolved it in ExaDG for the time being, https://github.com/exadg/exadg/blob/1a98c62467c3bdee68a46b6fb90f743f16beeb60/include/exadg/grid/grid_utilities.h#L197-L215.

According to my current understanding, such a check would ideally be done inside dealii, i.e. if construct_multigrid_hierarchy or limit_level_difference_at_vertices is set by the user, assert that tria.all_reference_cells_are_hypercube() (or similar).

What happens if https://github.com/dealii/dealii/blob/master/tests/fullydistributed_grids/copy_distributed_tria_01.cc is done with a simplex grid? The problem I reported might, however, not be visible here since the triangulation constructed first is distributed, which might already contain a human-readable assert regarding simplex, I don't know. I could try to adapt this test, or are there already tests in dealii that are closer to my problem, @peterrum ?