[Gmsh] Extruded Mesh Again

[Christophe Geuzaine]

Nicolas Tardieu wrote:

>   Dear Christophe,
> 
> Here I come again with my extruded mesh....
> Gmsh tells me "The mesh will be incorect" : in what
> sense? And why a mesh can be correct with
> pyramids and hexas and wrong with tets? Is it
> a problem of negative Jacobians?

No: it's a more general problem due to the fact that we must respect the
surface mesh when we divide the interior elements. The way we swap the
edges is a little bit primitive at the moment.

Can you replace 'Mesh/3D_Extrude.cpp' with the file attached to this
message? I just introduced a "random" swap when the "deterministic" one
fails. (Please check that the resulting mesh is correct--I didn't have
much time to test.)

> 
> Moreover, if I try not to extrude the mesh but to generate
> it in regular Delaunay manner, I obtain some weird
> mesh.....

That's because your "bottom" surface clearly doesn't project cleanly
into a plane (using our simple projection technique--see 5.2 in the
FAQ). There is a solution to this problem (see our TODO list), but
someone has to implement it...

Christophe

-- 
Christophe A. Geuzaine
Applied and Computational Mathematics, Caltech
geuzaine at acm.caltech.edu - http://geuz.org