Problem with libple.so.2

[jgd23]

Hello,

Sorry I moved my post to the right section.

I have an error when I try to compute Mesh processing or solver with CS 7.2, I use Ubuntu 22.04 via a Virtual machine
The Mesh is done with Salome 9.9 export in Med format.

I have the following message in the console:
/home/julien/code_saturne/7.2.0/code_saturne-7.2.0/arch/Linux_x86_64/libexec/code_saturne/cs_solver: error while loading shared libraries: libple.so.2: cannot open shared object file: No such file or directory
solver script exited with status 127.

Thank you for your help

Best regards

Julien

[Yvan Fournier]

Hello,

The recommended info also includes the installation logs (at least config.log), and a short description of your system if possible.

Also, was this an initial install in a clean (empty) build and install directory, or a re-install ?

Best regards,

Yvan

[jgd23]

Hello,

Thank you for your answer.
It is a fresh installation. I use VMware machine on a Windows 10 OS. Host is Ubuntu 22.04.
Code Saturne is 7.2 and Salome version is 9.9.

Below is my config.log file from this code saturne 7.2 build folder

Julien

[niuniuaba]

Hi there,

I have the same problem. CS-7.2.0 in Ubuntu 22.04 native, hdf5/med/medcoupling from SALOME installation.
Besides libple, CS claims medcoupling library not found either. Now I am using it by

Code: Select all

LD_LIBRARY_PATH=..... code_saturne ...

Best regards,
niuniuaba

[Yvan Fournier]

Hello,

I'm not sure why there is this behavior on Ubuntu. I have noticed over the years that Ubuntu seems to have specific low-level linker settings, which make everything more difficult (i.e. more users have issues on Ubuntu than on any other Linux distribution, as you can see if you search this Forum's posts for Ubuntu).

One related factor may be that PLE was switched from using libtool to simpler Makefile rules in code_saturne 7.1, and this was done for the main part of code_saturne itself only after v7.2, and some "implicit" expectations may not work on Ubuntu.
Depending on the library used, some "rpath" settings were or were not given. This should be more systematic in code_saturne v7.3, and can probably further be improved/automated.

I'll run tests at least for code_saturne v7.3 on Ubuntu in the future.
In the meantime, the suggested workaround is a good solution.

Best regards,

Yvan

[jgd23]

Hello,

Could you be more precise niuniuaba?
I have to modify the bashrc file in adding a path towards what?
Sorry I am not an experienced user of Linux.
Could you write a clear example of what you mean? Thank you

Best regards

Julien

[niuniuaba]

Hello,

In your $HOME directory (/home/julien) there is a shell script named '.bashrc'. Add a line

Code: Select all

alias code_saturne='LD_LIBRARY_PATH=/home/julien/code_saturne/7.2.0/code_saturne-7.2.0/arch/Linux_x86_64/lib:$LD_LIBRARY_PATH /home/julien/code_saturne/7.2.0/code_saturne-7.2.0/arch/Linux_x86_64/bin/code_saturne'

to this file, logout and login again, then you can use code_saturne command as usual.

Good luck.

[jgd23]

Hello,

Thank you, it works now for single processor.

If I try with several processor I have the following message

"
Starting calculation
--------------------

./cs_solver was built without MPI support,
so option "--mpi" may not be used.
solver script exited with status 1.

Error running the calculation.

Check run_solver.log and error* files for details.

Domain None (code_saturne):
run_solver.log, error*.

Post-calculation operations
---------------------------

Run failed in calculation stage.
"
Here is the run solver log file.
How can I resolve that?
I don't have install Scotch and Parametis because I have some errors

Best regards

Julien

[jgd23]

Hello,

I see in a previous post of Yvan that Scotch and Parametis are not really important for a basic use.

viewtopic.php?f=3&t=3005

I will try to switch to Debian OS

Best regards

Julien

[Yvan Fournier]

Hello,

MPI should work on Ubuntu also. You need to make sure some MPI library is installed, then help the "configure" script either with the '--with-mpi' option (to provide a path), or use "CC=mpixx CXX=mpicxx" (no need to change to use an MPI wrapper for Fortran, which does not directly call MPI).

On a relatively recent Debian system with OpenMPI packages installed, I need to add '--with-mpi=/usr/lib/x86_64-linux-gnu/openmpi' for MPI to be detected. The path is probably similar on Ubuntu.

Regards,

Yvan