Transition to newer version of Code Saturne SIGSEGV error

[Boone11]

Hi all,

I tried to use SRC files of version 5.1.6 with the version 5.3.4 of Code Saturne but I am experiencing a SIGSEGV error. I isolated the issue. It is coming from my file typecl.f90 (see attached document) that is used to impose a pressure difference between 2 channels that are linked with a pipe.

With the version 5.1.6 every is fine but when I used the version 5.3.4 I faced successive issues:

Firstly I had the following error:
"error : The attributes of this name conflict with those made accessible by a USE statement. [IFRSLB] integer ifrslb, itbslb." during the compilation. So I replaced the variable ifrslb and itbslb involved in the typecl.f90 file by ifrslb2 and itbslb2 which allowed the code to compile.

Now I am facing a SIGSEGV error at the begin of the file (before using ifrslb2 and itbslb2):

Code: Select all

WRITE(*,*) "BEGINNING OF DO LOOP"
do ifac = 1, nfabor
  ityp = itypfb(ifac)
  if(ityp.le.0.or.ityp.gt.ntypmx) then
    itypfb(ifac) = 0
    iok = iok + 1
  endif
enddo
WRITE(*,*) "END OF DO LOOP"

The SIGSEGV error occurs inside the do loop since I see the "BEGINNING OF DO LOOP" message but not the "END OF DO LOOP" one. I do not see the reason for this error. May you help me? Note that without the typecl.f90 file the calculation is successfully run even with the 5.3.4 version.

Thanks a lot!

[Yvan Fournier]

Hello,

Note that version 5.3 was retired 5 years ago (https://www.code-saturne.org/cms/web/do ... n_timeline).

Also, since you are using a modified non-user-defined subroutine, this amounts to using your own development branch. In this case, using development tools (Git to help update/merge files, and a debugger to analyze errors) is strongly recommended. In any case, at least use "meld" (or your favorite text comparison tool) to compare your modified typecl.f90 with the one from the code, and check if arguments match...

Since in a previous post you mentioned moving your turbulence model to a new version I strongly recommend starting with this when you get the chance, at least if you plan on using the model for a significant amount of time (since turbulence code has mode to C, this is a significant task if there is a lot of non-user-defined code)...

Also, did you choose to use versions 5.1 and 5.3, or did someone force that choice on you ? If that is the case, they should do all the hard work...

Best regards,

Yvan

[Boone11]

Thank you very much Yvan for your answer. It helped me updating the typecl.f90 file to version 5.3.4.

The turbulence model is in version 5.1.6 and 5.3.4, but this particular modified non-user-defined subroutine, which is used for very specific purpose was not moved to version 5.3.4, which I just did.

To be more precise, the turbulence model I am using has been moved to a newer version (I think it is version 7.0) but there are issues with the model on the newer version that are still not resolved. I do not have the skills and the search for the reason of the issues is out of my project. By the time the issue is solved, I work with version 5.3.4.

Thanks again,
Have a nice day,

[Yvan Fournier]

Hello,

Do you have more info on the issues with that turbulence model ? Is it something we need to look at ? If yes, do you have a test case ?

Note that intermediate versions only have minimal testing, while v7, v8, .... have undergone complete validation. This does not totally prevent issues but makes them less probable. If there are remaining issues, bug reports are welcome/useful.

Best regards,

Yvan