Mesh joining

[eduardp]

Hello to everyone,

I encountered a problem in joining meshes and I would like to know if it exists a limit for how many meshes can be joined. I have something like 60 .med meshes and I tried to use Code_Saturne 2.0.2, but I got this error:

/opt/saturne-2.0.2/installer/ecs-2.0.2/src/pre-post/ecs_pre.c:378: Fatal error.

Mesh file "/home/eduard/CFD/MESH/5Mesh_3" does not have
a known extension. A keyword is required to determine
the format in this case.

Call stack
1: 0x7f7e8e8af152 (libbft.so.1)
2: 0x40f154 (cs_preprocess)
3: 0x40915e (cs_preprocess)
4: 0x409ee6 (cs_preprocess)
5: 0x7f7e8d8c1c4d (libc.so.6)
6: 0x404209 (cs_preprocess)
End of stack

But it is not a extension problem. 5Mesh_3 is the ~ 30th mesh. If I try with around 10-15 meshes everything is ok.

Should I use a newer version or is no working around for this problem.

Thanks for any hint,
Eduard

[Yvan Fournier]

Hello,

In theory, there should be no hard limitation to the number of meshes read.

Are you sure this is not an extension problem ? If the mesh has no ".med" extension, you need to provide the format info using the appropriate "--format" option. On 30+ meshes, it is easy to forget the option on 1 of the meshes.

This could also be a bug, but this would seem surprising.

In any case, if you are still using 2.0, upgrading to at least 2.0.4 is recommended (though none of the bugs fixed between 2.0.2 and the current 2.0.6 would seem related to this issue if I rememeber correctly, it is always safe to do).

Best regards,

Yvan

[eduardp]

Hi Yvan,

I think I must explain a little what I am doing so we can rule out the file names problem.
So, I created a geometry in Salome and from this I obtained 7 solids (Solid_1, Solid_2... Solid_7). These solids are then meshed and Solid_2-Solid_6 are translated (creating copies) a number of times (n) and Solid_7 is translated once at the end of my setup. All the generated meshes are then saved to a specific folder (no room for error here - it is done automatically). I also defined faces that are shared for every solid, they are named Shared.
In Code_Saturne I made a new case and specified the meshes, specified that the Group for joining should be Shared (all using the GUI interface).
And now comes the interesting part: when I have n=1,2,3 everything works, the Check mesh has no complains and I can get the boundaries from the processor list and the case runs, but when n>4 the error appears. Sometimes complains of something like "/home/eduard/CFD/MESH/5Mesh_3", but sometimes is just "/home/eduard/CFD/M" or similar.
I'll try to upgrade to 2.0.6, but if it is no known bug, I do not think it will help much.

Thanks again for any hint.

With best wishes,
Eduard

[Yvan Fournier]

Hello,

This would seem like a mesh input list resizing/owerwrite bug in the Preprocessor.

Running a quick test with a similar copy, I don't reproduce this. Could you run:

code_saturne check mesh -m *.med -j Shared

manually from the directory in which you have your meshes ? This would help determining whether what looks like a command-line truncation is due to the GUI or the Preprocessor itself.

Note also that with version 2.2 and above, you could simply generate one mesh, and read it multiple times/translate it / join it from the main Code_Saturne kernel.

Best regards,

Yvan

[eduardp]

Hi Yvan,

I tried as you've suggested and it is working perfectly. To be sure I increased the number of med files to 280. In the end it seems it is a GUI bug.

Thanks for the help.

With best wishes,
Eduard

[Yvan Fournier]

Hello Eduard,

OK, great. I'll check the GUI to see if the bug is easy to find/fix, but there is no fixed date for the next maintenace/bugfix release version 2.0.7 (depends on the number and criticality of bugs fixed or the necessity of new machine ports).

Best regards,

Yvan