SIGSEGV signal - Mesh quality criteria only

[Guillaume THIRIET]

Good evening everyone,

I recently tried to run a case using the Mesh quality criteria only execution mode in order to therefore be able to use the preprocessor log to define my boundary zones. I did not have so far implemented any parameters nor models. This is just a pre-calculation with mesh only. I did not get any error while creating the mesh with SALOME.

Unfortunately, I got the following error SIGSEGV signal (access to forbidden memory area) intercepted !. It seems that several persons already got that kind of problem before looking at the forum history but I did not find so far a solution to my problem. I am also quiet troubled as I already run some cases in the past few weeks without having this mistake whereas the cases I was dealing with where more complicated in term of meshing.

My case is rather simple and composed only of uniform squared cells (picture below).

Here is the error I get in the preprocessor.log file

Code: Select all

Warning
=======
The MED mesh contains 7803 elements of type seg2
which are ignored by the Preprocessor.

  Number of faces                             :    65920
                                   Family 0   :    43280
                                   Family 29  :      800
                                   Family 30  :      100
                                   Family 31  :    21740
  Number of cells                             :   180800
                                   Family 0   :   180800
SIGSEGV signal (access to forbidden memory area)  intercepted !

Call stack 
   1: 0x56077d55b2de <+0x82de>                        (cs_preprocess)
   2: 0x56077d592e97 <ecs_table_att__cree_att_fam+0x727> (cs_preprocess)
   3: 0x56077d59f93c <ecs_maillage__cree_attributs+0x9c> (cs_preprocess)
   4: 0x56077d5811b0 <ecs_pre_med__lit_maillage+0x1ef0> (cs_preprocess)
   5: 0x56077d56355a <ecs_pre__lit_maillage+0x16a>    (cs_preprocess)
   6: 0x56077d55b390 <main+0xa0>                      (cs_preprocess)
   7: 0x7fea2bb43083 <__libc_start_main+0xf3>         (libc.so.6)
   8: 0x56077d55bbae <_start+0x2e>                    (cs_preprocess)
End of stack

Thank you very much for your help.

Yours faithfully

Guillaume THIRIET

[Yvan Fournier]

Hello,

Could you also provide the mesh (posting it if it is not too large, otherwise sending it through some file transfer or sharing system).

Best regards,

Yvan

[Guillaume THIRIET]

Good evening/morning Yvon,

Please find there the Mesh file (MED format)

Yours faithfully

[Yvan Fournier]

Hello,

Which tool and version did you use to generate the MED file ? Using a MED dump utility, you can check that there are elements of family 30 (faces according to the code_saturne preprocessor), but in the family descriptions, family 30 does not exist (we go directly from 29 to 31).

So the file is not consistent, which causes the crash of the code_saturne import. I could probably adapt the import so as to emit a warning and ignore family (group) information when the file is incomplete, but I suspect a bug in the tool that generated the file.

Best regards,

Yvan

[Guillaume THIRIET]

Good morning Yvan,

I apologise for the late reply. I was on Chinese new year holiday so I did not check the forum this week.

I created the mesh simply using SALOME V9.9.0. I did not get any errors while creating the mesh. As you can see, the later remains pretty simple as it is only composed of squared cells. I therefore used the "Export" --> "MED file" format and that's all.

Yours faithfully.
Again, please accept my apologies for the late reply.

Yours faithfully

Guillaume THIRIET

[Yvan Fournier]

Hello,

I'll see what happens if I import and export it out of Salome 9.10 (which is not on the salome.org site yet, at least not in a "built" form), and if you ha a script or .hdf file allowing to reproduce the mesh generation we could file a bug for Salome.

On the code_saturne site, I can try to make the reader more robust, so that it ignores missing définitions instead of crashes. I'll try that in.the next 2 or 3 weeks.

Best regards,

Yvan

[Guillaume THIRIET]

Dear Yvan,

I maybe found this morning the solution related with the crash.

In this simulation, I want to implement a squared fuel inlet on the "ground" of my room as shown on figure 1 (View from below). In this way, I created a plane using the "create a plane" tool and I stuck it on the low face of the "air" domain (the big blue rectangle).

Then, I again created multiple planes using the "create a plane" tool in order to partition the whole domain using the "partition" tool which leads to what we can see on figure 2. (I did something similar to what it is done on the "Full domain simplified nuclear vessel" tutorial).

The thing is that maybe I first needed to partition the "air domain" with the "fire inlet" squared rectangle and then partition the newly get domain in order to get enough "line" to ease the mesh creation. That's what I did this morning and it seems it removed the previously quoted error.

In any case, the mesh does not change but this additional step allows to remove the issue while processing with "mesh quality criteria only".

If things are not clear in my explanation, I am available to discuss it online on your request.

Yours faithfully

Guillaume THIRIET