Cannot build Code_Saturne with PETSc

[paspro]

I have been unable to build Code_Saturne with PETSc for any combination of configure petsc-related options because the relative m4 file is simply wrong, at least for a regular system-wide installation of the petsc package using a typical Linux repository. I modified some files attached here to make it work.

On my system with these modifications and the command line options:

--with-petsc=/usr --with-petsc-lib=/usr/lib64/openmpi/lib

the compilation works correctly.

You may want to consider updating them.

cs_petsc.zip

[YvanFournier]

The patch breaks the build using user/project account installs (installed using --prefix). I will try to merge the patch so as to work with both, so I may need to get back to you soon to let you test the patch on your configuration.

[paspro]

Happy to test any patch you have.

[YvanFournier]

Hello,

Trying to get back to this.

On a Debian sid system using a packaged install of PETSc, the detection configuration works OK without the patch using --with-petsc=/usr/lib/petsc (which is actuallly a symbolic link to etc/alternatives/petsc, itself including other symbolic links). I do have a link issue if I do not force "reload_config=True" in build_aux/cs_compile_build.py in the package constructor, but this is probably a separate Python issue.

Which type of system are you running on ? What is the structure under /usr/lib64/openmpi/lib ? Did you try with the unmodified system using --with-petsc-include=/usr/include --with-petsc=/usr/lib64/openmpi/lib ?

Best regards,

[paspro]

Contributor

I am on Fedora which has somewhat different installation paths for petsc compared to ubuntu/Debian.

[YvanFournier]

Could you specify how the install paths for PETSc are organized on Fedora ?

[paspro]

On Red Hat / Fedora / CentOS the installation filelist can be seen in: https://koji.fedoraproject.org/koji/rpminfo?rpmID=31788325