Welcome to the forums. Please post in English or French.

You are not logged in.

#1 2022-01-20 00:10:39

jlucas
Member
Registered: 2021-12-15
Posts: 4

compilation errors of code-aster on debian10(bullseye).

Hello,

I've got some errors when trying to compile code-aster under debian10. Tried both versions 14 and 15 (same problems).

Errors on tfel-3.2.1 and mumps-5.1.2 libraries. The other libraries compiled without any problem.

The problem with mumps is related with gfortran v10. as it seams it requires an additional flag -fallow-argument-mismatch to compile without errors.

By editing the Makefile.inc under the /tmp/install_mumps.*/mumps-5.1.2 folder to add the mentioned flag to OPTF, ie:

#Begin Optimized options
OPTF    = -O -O2 -fPIC -O -DPORD_INTSIZE64 -fopenmp -fallow-argument-mismatc

and recompiling with 'make all', compilation is successful.

I could not get the fix by using the waf system. Tried to add the flag to F90FLAGS on the setup.cfg but did not work. so installed the files manually.

The tfel error could not yet get it fixed. By cloning the repository and compiling with cmake. ver. 3.2.1 shows the same error, but ver 4.0.0 not (and compiles successfully). Not sure it helps. I attach the setup.log file in case anyone knows the solution and is willing to help.

I guess the errors I am experiencing are compiler related and will be quite common on new systems as gcc10 will be the standard. So it would be nice if the developers were to get the system up-to-date.

Thanks,
Best wishes,
Jorge

Offline

#2 2022-01-26 18:40:53

mathieu.courtois
Administrator
From: France
Registered: 2007-11-21
Posts: 1,178

Re: compilation errors of code-aster on debian10(bullseye).

Hello,

Debian bulleye is Debian 11.
Debian 10 still uses GCC 8.3.

aster-full packages will probably die because a lack of maintainer(s). Compiling some the prerequisites on "all" platform may be a pain!

That's why our effort is to maintain Singularity images that embed all the needed prerequisites to build the stable and development versions.
Singularity images are available from the Download page of salome_meca.
Current containers are based on Debian 9 (which is the reference OS for salome_meca). The 2022 version will be based on Debian 10.
I already know the patch for MUMPS. Feel free to share your patch for Debian 11.

If there are some volunteers to continue aster-full setup scripts. I can try to start a repository of its files.

MC


Code_Aster release : last unstable on Ubuntu 16.04 64 bits - GNU Compilers

Please do not forget to tag your first post as *SOLVED* when it is!

Offline

#3 2022-01-28 01:22:27

jlucas
Member
Registered: 2021-12-15
Posts: 4

Re: compilation errors of code-aster on debian10(bullseye).

Dear Mathieu Courtois,

Thank you for your quick reply (and corrections). I totally understand the difficulty and complexity of building such a complex software as code-aster and fit it into the so many existent OS platforms and flavours...

I am a new user of code-aster and so far with little experience. I must say: it is not the easiest FEM software to get into and to start with. In comparison CalculiX was much easier to install and get through with the documentation and examples. But as it lacks functionality, I decided to move into the more powerful and painful domain of code-aster... (The book from Jean-Pierre Aubry is 5/5 star, but not enough to get the job done...)

Please forgive me for criticising, but I find it hard to work with a non standard build system as the waf. Other than keeping the default options, I couldn't barely do any adjustment/fine tuning. For example, I would like to link the PETSc library and to try the iterative parallel solvers... but judging from the comments in the forum, the process seams to be other than straight forward and probably very painful.

The singularity container you mentioned is 5GB! To be honest, code-aster solver + gmsh suffices my present needs. It seems awkward to me to mount the 5GB container, specially if I have a working alternative. But really what putted me off the salome meca 2021 container was the non matching sha1/md5 checksum. First, what is what (they are not the same) and then the key on the web simply don't match to the software I downloaded (3 times!). Again, it seams a long going issue as it can be found on the forum, and sorry to say it shows a bit of floppiness.. Cannot understand how difficult it must be to replace that number to a correct one, or simply to delete this key away from the website...

To finalise, in an "ideal world" (at least on mine) it would be just super nice to issue an apt install code-aster command and get all the pain over the back (even if it would mean with some functionality restrictions)... 
...but off course, the world rests in rough reality...

Best wishes,
Jorge

Offline