recommended Ubuntu version and mesh file format

[Jean-Baptiste Leydet]

Hey all,

Two questions:

1. Linux: I need to install Linux on a machine to test Code_Saturne 2.0rc2. Which version of Ubuntu would you recommend? Also, Is there a clearly better distro than Ubuntu out there, for intended use with EDF open source solutions (Saturne, Aster, ...)?
2-a. Mesh generation: So far we used the gmsh format, which did the job for fairly simple geometries. Now I'm looking into CastNet for a more "industrial" solution (prism layers, abstract modeling). CastNet cannot export Saturne-friendly files yet, but the support is proposing to implement of a patch. They sent me a .neu file to test with Saturne (test_1.2.1.neu in attached zip), but when I perform a mesh check from 2.0rc1, I receive an error message (see error.txt in attached zip). I found on the web another .neu file that I can import without a problem (test_2.4.6.neu in attached zip). Can you tell me what the problem with the 1st file is? (too old version? element types? ...)
2-b. Mesh generation again, but a more general question: Is there a way to avoid using color numbers to identify BCs and volumes in Code_Saturne? Ideally I'd like to name my fluid zones and BC's in a meshing software, and Saturne would detect these at import and make them appear in the GUI. For example, using the attached test_2.4.6.neu file, I could identify named BC's in Paraview. Do I have to type these same names in Saturne? That could be rather painful for complex models.

Thanks!

Jean-Baptiste

[Yvan Fournier]

Hello,

Any recent Ubuntu version should do fine (I use 10.04). At EDF, the reference platform uses Debian Etch, which is quite older, but we install newer gcc versions on those platforms (which have 4.1.2 by default, on which some difficult to reproduce issues have been reported). Any recent distribution should be fine. Note that the next Debian and Ubuntu releases will include Code_Saturne as a package, but it might be built without MPI, and the version may lag slighty behind our releases.

I just debugged the mesh that causes the crash, and it simply amount to this:
- in a "correct" Gambit neutral file, element group section lines start with: "       ELEMENT GROUP" (with 7 initial white spaces), while in the file that causes a crash, the line start with "      ELEMENT GROUP) (with only 6 initial white spaces)...

If the CastNet support adds one whitespace in their output format for element groups, things should work better (you may test the difference by adding a whitespace at the beginning of line 14363 with a text editor). We could also test for both cases with Code_Saturne, but if you strictly follow the Gambit neutral documentation, adding a whitespace is the correct solution.

Finally, to identify BC's and volumes in Code_Saturne, the use of color numbers or group names depend on the mesh format. The Gmsh format uses physical group numbers (so you have colors), while the Gambit format uses group names (so you may use those under Code_Saturne). You don't ned to type the names in Code_Saturne, as the GUI allows you to run the Preprocessor on a file so as to identify group names, and parses the preprocessor log to extract those names and suggest them. If you use the SALOME CFD_STUDY extension, you may even highlight groups uner the SMESH view to check your boundary conditions.

Best regards,

  Yvan

[Jean-Baptiste Leydet]

Hello Yvan,
Thanks for the input. I tried what you suggested (additional white space): it made the warning message disappear, but i still receive a SIGSEGV error when checking the mesh.
Any clue?
Jean-Baptiste

[Yvan Fournier]

Hello,
I just reproduced the crash on 2.0-rc1, but the bug that caused this crash seems to be one we corrected early June, as 2.0-rc2 seems OK.
Best regards,
  Yvan

[Jean-Baptiste Leydet]

Alright. One more reason for me to install 2.0rc2, then ;-)
Thanks for your help,
Jean-Baptiste