code_saturne User's Forum

Hi all,

I have recently reinstalled Code_Saturne 4.0.8 with PT_Scotch support(6.0.4). When I ran a simulation with a mesh density of 22 million the results produced by the installation could not be read. Trying to restart the calculation i get the following error.

SIGSEGV signal (forbidden memory area access) intercepted!
Call stack:
1: 0x2af5c123917b <fvm_selector_create+0xc4b> (libsaturne.so.0)
2: 0x2af5c13330d7 <cs_mesh_init_selectors+0xb7> (libsaturne.so.0)
3: 0x2af5c10f1c05 <cs_preprocess_mesh+0x465> (libsaturne.so.0)
4: 0x2af5c1063c47 <cs_run+0x127> (libsaturne.so.0)
5: 0x2af5c1063a01 <main+0x111> (libsaturne.so.0)
6: 0x2af5c4148b35 <__libc_start_main+0xf5> (libc.so.6)
7: 0x401439 <> (cs_solver)
End of stack

Conversely with a smaller mesh density there are no issues.

I have attached the error outputs and listing files.

Any help is much appreciated

Hello,

Did the same case run "restarted" under a previous version ?

It is strange, as the error occurs during the preprocessing step, prior to restart, so If the error only occurs upon restart, I suspect a prior memory override bug occuring earlier.

Could you test a debug build (a build configured with "--enable-debug") ? This may catch and report errors earlier, making debugging easier (but is 2 to 3 times slower, so good for testing and seting up a simulation, then switching to a "standard" production build to run iterations).

Also, could you test if you have the same error not using any user subroutines (especially cs_user_parameters.f90; you other subroutines are called later, unless I missed something).

Best regards,

Yvan

Hi Yvan,

Thank you for replying.

The issue appears to have been linked to the method I used to install code_saturne.

Initially I was installing code_saturne with the automatic installer. The installation appeared
to progress correctly. But results generated by that installation, could not be read in paraview, as I would get a segmentation error and trying to restart the simulation would crash code saturne.

Installing code_saturne manually seems to have resolved some of the issues, I can now read the result in paraview and restart the case. I will try running for longer iterations and restarting the case to confirm everything still works.

I do not understand what caused the earlier issues but it seems to be working now.

Kind regards,

Kenneth