strange behavior with TRIA and code_aster 16.5

AsterO'dactyle

Entre vieux machins, on peut discuter.
Je n'ai vu personne se plaindre récemment de ce problème. Les tests tournent sans problèmes.
Le seul truc que j'ai personnellement modifié cette année est la détection des mailles non-planes mais c'est pour les QUAD4 (DKQ donc) et pas en 16.5.

Ca ressemble donc à un problème dans ta version ou une variable non initialisée depuis des années que tu révèles dans ton cas. Tu as compilé comment ? Peut-être un problème avec les BLAS

jeanpierreaubry

AsterO'dactyle
je suspectais aussi une possiblité de ce genre
c'est la version conda de notre ami Luca et consort
dans le conteneur 2023 je n'arrive à lancer ni as_run ni run_aster ni astk
quant à salome-meca tu connais mon avis

ceci étant dit c'est juste pour aider sur ce forum mes besoins personnels sont satisfaits avec 14.6 et 15.2

AsterO'dactyle

Ah, il faut éviter les installations autour de conda et rester en Python Standard. Vu ton plantage, c'est probablement lié à ça puisque MECA_STATIQUE est full-Python maintenant.

jeanpierreaubry

AsterO'dactyle
merci pour l'info
mais comment fait on pour installer une version 16.5 ou 17
ailleurs que dans le conteneur 2023
qui souffre, au moins chez moi, de dysfonctionnements rédhibitoires
rien ne se lance ni salome-meca, ni astk, ni run_aster, ni as_run
avec à chaque fois des messages assez incompréhensibles

k_zurawski

AsterO'dactyle This is the same situation in my case as well. It might be, that you are working at EDF on different version of Salome_Meca 2023 than is provided on your site.

AsterO'dactyle

Hello,

Two "official" versions of salome_meca: the salome_meca for LGPL (on the website) and the salome_meca for EDF in-house uses.
The only difference between the two is the presence of more tools in EDF's version, not released under the LGPL. The structure and basis are the same, including code_aster.
It works on a Singularity container. Internally validated on Debian10 and Debian11. Normally, the container ensures compatibility on many platforms (including WIndows). There are occasional problems, but mostly on the graphics side (graphics card support). On the code_aster part, it should work.
The salome_meca2024 LGPL version will be released in a few days.

k_zurawski

AsterO'dactyle so there's something else wrong with the 2023 container available to us. I tried to run it on Debian 11 and Debian 12. 2022 runs, 2023 unfortunately not. There's no point in looking where the error was since you'll be issuing 2024. Could you ask someone to test before posting it?

Which version of Singularity/Apptainer should we use for 2024?

jeanpierreaubry

k_zurawski
same remarks and same question from me

mf

k_zurawski
Hello,

I never got SM 2023 to work properly on Ubuntu 24.04, see thread

https://forum.code-aster.org/public/d/27539-new-version-sm-salome-meca-lgpl-202310-4-20240327-questions/22

Hoping this goes better next time,

Mario.

jeanpierreaubry

hello

my original topic was a strange behavior with TRIA3 elements
it drifted away a bit

back to it
with an update to
Version 17.2.2 modifiée le 04/01/2025
on
https://github.com/conda-forge/code-aster-feedstock/
the trouble does not seems to exist any more
i have tested only 2 problems yet,
one very simple with only 1 TRIA3 and 3 QUAD4 which helped to identify the trouble
and one more representative with over 4000 element
both runs well

AsterO'dactyle

Hello,

I asked my colleagues to test on non-EDF machines before releasing salome_meca 2024.
For the graphics part of salome_meca, we're dependent on graphics card support in Singularity, so we'll still have problems, unfortunately.
On the other hand, for the 'classic' code_aster part, it should work.

k_zurawski

AsterO'dactyle Thank you.

Please tell us which version of Singularity/Apptainer we should be using with Salome_Meca 2024 and which system and version should be the host for Singularity/Apptainer.

Thank you in advance.