[Dune] gmsh-documentation

[Dan Popovic]

Hi Oliver!

> Thanks for writing documentation about the gmsh import.  However,
> I am not really convinced that you put them in the proper place.
> The external_libraries directory has so far been only for installation
> instructions, and not more.  I think we should keep it that way and
> put usage instructions elsewhere.  Could you maybe merge your
> usage instructions into the grid-howto?  Thank you!

You are right, I will seperate the usage instructions and send a patch
for the dune-grid-howto.

> I gather from your notes that the problem with orientation inconsistencies
> can at least be circumvented somehow within gmsh.  Can you give
> me an update on the situation, please?

It seems as if the grid orientation inconsistencies only appeared if one
used the 'refine' or 'refine by splitting' button during the mesh
creation in Gmsh-2.3.1 or Gmsh 2.4.2, respectively, to create fine
grids. Using other settings for the mesh generator (e.g., max/min
element size) to get fine meshes didn't result in inconsistent grids,
interestingly. So the work-around is not to use refinement in Gmsh.

I have tested all the mesh files in ./dune-grid/doc wether they are
consistently oriented, and luckily they are still usable with the new
UGGridFactory with the non-quadratic algorithm. 

That means that the insertion/creation of linear boundary segments by
default has nothing to do with the consistency errors we discussed last
week. You could revert your last patch where you disabled the creation
of linear boundary segments for meshes with linear boundary segments.
I discussed the topic with Peter as well. According to him, he had to
create the linear boundary segments explicitly because otherwise the
GridFactory would number them in an intrinsic manner, which wouldn't be
consistent to any indexing done by Gmsh any more. Thus, data attached
'in Gmsh' to such linear boundary segments wouldn't be assigned to the
appropriate boundary.

Best wishes,
Dan

> best wishes,
> Oliver