[Gmsh] gmsh problem with 2d mesh

Christophe Geuzaine c.geuzaine at ulg.ac.be
Wed Sep 3 08:39:15 CEST 2003


krazz at ceoc.com wrote:

> in this example, for lc <= 2 gmsh will produce a 2d mesh as it should,
> but with lc=4 for example, it will give me an error message saying "No
> triangles in surface mesh" or produce new lines from the 1d mesh points
> of the line loop(16) to infinity in some direction
> 
> The problem seems to occur as soon as the triangles that I export are
> small compared to lc. How do these values depend on eachother, or to put
> it another way, how can I compute the minimum value for the
> characteristic length of a triangle, so that gmsh will use it correctly?

Yep, it's a known bug in the old 2D algorithm (that we should definitely 
fix...). Meanwhile, you might want to use the other isotropic 2D 
algorithm (based on Triangle, which I think is actually also used in VTK):

on the command line: gmsh -algo triangle
in input files: Mesh.Algorithm = 3;
in the GUI: Options->Mesh->2D->Triangle

> 
> Another question, is there a real windows port of gmsh available, that
> will compile with Visual C++ for example? I have spent quite some time
> trying to get the library to work without cygwin, but no success..
> 

It used to work with mingw (pure win32 gcc) a couple of years ago, but I 
didn't test it recently. You'll probably have to add some #ifdefs around 
the Unix socket code (Fltk/GmshServer.*) and other Unix commands 
("unlink()", etc.), but this shouldn't be too hard.

If you send a patch, I'll incorporate it.

Christophe

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