No vtkGhostLevels when dgify_fields is true with continuous model

[markgoffin]

In subroutine vtk_write_fields in VTK_interfaces.F90, if the model mesh is continuous but the state contains discontinuous fields to be output, then no vtkGhostLevels are output. This is because the element_halos are not copied to the discontinuous model_mesh.

It seems to me that the element_halos could be copied across from model to model_mesh and thus the vtkGhostLevels would be present in the pvtu. But maybe this is not correct or always the case.

You could argue that the output mesh should be chosen to be a discontinuous mesh in the options file in the first place but it could at least give a warning. Or if possible include the element halos in the dgify process.

[stephankramer]

All sounds correct to me. Alternatively, it could directly use model%element_halos ? Afair these should be the same for all meshes anyway.

You could argue that the output mesh should be chosen to be a discontinuous mesh in the options file in the first place but it could at least give a warning. Or if possible include the element halos in the dgify process.

I would argue that indeed :-) This automatic dgify-ing is pretty annoying as you have to hand remove all dg-fields from output to avoid getting humongous vtus. Virtual beers for the person that removes it. I'd prefer it to do a masslumped galerkin projection for the vtu fields.