pflotran@develop build failures due to fortran interface changes in petsc@main

[balay]

https://gitlab.com/xsdk-project/spack-xsdk/-/jobs/7100286934

     51    petsc_utility.F90:42:36:
     52    
     53       42 |   call MatSetValuesBlockedLocal(A,1,irow-1,1,icol-1,ndof_mat
           ,ADD_VALUES, &
     54          |                                    1
  >> 55    Error: Rank mismatch in argument 'c' at (1) (rank-1 and scalar)

[balay]

There are multiple changes on PETSc fortran interface. The first one:

• VecLockPop - which is deprecated (since 3.11) for a while appears to be removed now (only for fortran? C still exists).

i.e the following change gets this part of the issue resolved [and this change should work with older petsc versions aswell:

-  call VecLockPop(xx,ierr);CHKERRQ(ierr)
+  call VecLockReadPop(xx,ierr);CHKERRQ(ierr)

etc..

[balay]

Next issue: triggered by https://gitlab.com/petsc/petsc/-/merge_requests/7559

[0]PETSC ERROR: Cannot create PETSC_NULL_XXX object
[0]PETSC ERROR: See https://petsc.org/release/faq/ for trouble shooting.
[0]PETSC ERROR: Petsc Development GIT revision: v3.21.1-241-g0764c050596  GIT Date: 2024-05-29 11:11:27 -0500
[0]PETSC ERROR: /home/balay/pflotran/src/pflotran/pflotran on a arch-linux-c-debug named pj01 by balay Fri Jun 14 15:01:33 2024
[0]PETSC ERROR: Configure options --with-hdf5-fortran-bindings --with-hdf5-dir=/home/balay/pinstall-hdf5
[0]PETSC ERROR: #1 snescreate_() at /home/balay/petsc/arch-linux-c-debug/src/snes/interface/ftn-auto/snesf.c:725
[0]PETSC ERROR: #2 solver.F90:439

The following usage is in pfortran - which is no longer correct? So I'm not sure what changes is required here [and stay backward compatible with current release]

  solver%snes = PETSC_NULL_SNES
  <snip>
  call SNESCreate(comm,solver%snes,ierr);CHKERRQ(ierr)
  <snip>
  if (solver%snes /= PETSC_NULL_SNES) then
    call SNESDestroy(solver%snes,ierr);CHKERRQ(ierr)

[balay]

also cc: @rtmills

[ghammond86]

Sorry. I completely missed this.

Are there any Fortran examples for SNESCreate/Destroy()?

[balay]

I think we'll need @BarrySmith 's help here.

I'll have to recheck issues with latest petsc/main

[ghammond86]

@BarrySmith @balay @rtmills

No need. I figured it out (helps to think before asking a question). Fortran is now using classes (extended derived types) instead of derived types. These changes are relatively comprehensive. We have a release coming up in a few weeks that will be based on 3.21. After that release, we will refactor.

[balay]

Sure thanks! This work is more for the next xsdk release due nov (i.e with the next petsc version 3.22 due sept etc..)..