Fix for PetscCtxRt and PetscCtx argument types

[stevendargaville]

@BarrySmith I've brought the changes from your fork into this branch that will be merged after your PETSc MR. The commit hash you will need for PFLARE.py in PETSc is ad0dba9.

[BarrySmith]

Thanks. The PETSc CI is getting an error I don't understand with PFLARE; it doesn't seem to be related to this change but suddenly it appeared, maybe you understand it.

/home/svcpetsc/soft/mpich-4.2.2/bin/mpif90 -o src/PMISR_DDC.o -c -fPIC -Wall -ffree-line-length-none -ffree-line-length-0 -Wno-lto-type-mismatch -Wno-unused-dummy-argument -g -O   -Werror -Werror -I/scratch/svcpetsc/glci-builds/HXRAfDpzy/2/petsc/petsc/arch-ci-freebsd-pkgs-opt/externalpackages/git.pflare -Iinclude -I/scratch/svcpetsc/glci-builds/HXRAfDpzy/2/petsc/petsc/include -I/scratch/svcpetsc/glci-builds/HXRAfDpzy/2/petsc/petsc/arch-ci-freebsd-pkgs-opt/include -I/home/svcpetsc/petsc-hash-pkgs/4e9453/include -I/home/svcpetsc/soft/mpich-4.2.2/include    -I/scratch/svcpetsc/glci-builds/HXRAfDpzy/2/petsc/petsc/arch-ci-freebsd-pkgs-opt/externalpackages/git.pflare -Iinclude src/PMISR_DDC.F90
src/PMISR_DDC.F90:294:51:

  294 |       if (comm_size/=1) assigned_nonlocal = .FALSE.
      |                                                   ^
Error: 'assigned_nonlocal.dim[0].ubound' may be used uninitialized [-Werror=maybe-uninitialized]
src/PMISR_DDC.F90:157:93:

  157 |       PFLARE_PETSCBOOL_C_TYPE, dimension(:), allocatable, target :: assigned_local, assigned_nonlocal
      |                                                                                             ^
note: 'assigned_nonlocal.dim[0].ubound' was declared here
src/PMISR_DDC.F90:423:90:

  423 |             call boolscatter_mat_end_c(A_array, assigned_local_ptr, assigned_nonlocal_ptr)
      |                                                                                          ^
Error: 'assigned_local_ptr' may be used uninitialized [-Werror=maybe-uninitialized]
src/PMISR_DDC.F90:158:61:

  158 |       type(c_ptr) :: measure_nonlocal_ptr, assigned_local_ptr, assigned_nonlocal_ptr
      |                                                             ^
note: 'assigned_local_ptr' was declared here
src/PMISR_DDC.F90:423:90:

  423 |             call boolscatter_mat_end_c(A_array, assigned_local_ptr, assigned_nonlocal_ptr)
      |                                                                                          ^
Error: 'assigned_nonlocal_ptr' may be used uninitialized [-Werror=maybe-uninitialized]
src/PMISR_DDC.F90:158:84:

  158 |       type(c_ptr) :: measure_nonlocal_ptr, assigned_local_ptr, assigned_nonlocal_ptr
      |                                                                                    ^
note: 'assigned_nonlocal_ptr' was declared here
f951: all warnings being treated as errors
gmake[2]: *** [/scratch/svcpetsc/glci-builds/HXRAfDpzy/2/petsc/petsc/lib/petsc/conf/rules:171: src/PMISR_DDC.o] Error 1

https://gitlab.com/petsc/petsc/-/jobs/12774297747

https://gitlab.com/petsc/petsc/-/jobs/12774297747

[stevendargaville]

Thanks. The PETSc CI is getting an error I don't understand with PFLARE; it doesn't seem to be related to this change but suddenly it appeared, maybe you understand it.

Ah apologies, that was due to some recent changes in main and not this PR, but I thought I had the PFLARE CI setup to ensure all warnings were on and set as errors like the PETSc CI so that this wouldn't happen. I'll try and fix this now.

[stevendargaville]

@BarrySmith Ok this should be fixed now. Missed warnings were caused by old gcc version in the PFLARE CI and a missed -Werror flag for Fortran files, which have now been fixed.

The commit you need is 4ca6c2d

[BarrySmith]

I had to change the Interface_XXX function to exclude the final end interface line and require the user to provide that so you will need to make another change to the branch.

[stevendargaville]

@BarrySmith Done

[BarrySmith]

@stevendargaville It's gotten better. It now does not need the use of Interface_xxx so could you please remove those? Thanks

[stevendargaville]

@BarrySmith no worries, done

[BarrySmith]

@stevendargaville One of the pflare tests in the PETSc CI is failing. Could you please take a look at it https://gitlab.com/petsc/petsc/-/jobs/12854064147

[stevendargaville]

@stevendargaville One of the pflare tests in the PETSc CI is failing. Could you please take a look at it https://gitlab.com/petsc/petsc/-/jobs/12854064147

Apologies, I've managed to replicate this and am working through now.

[stevendargaville]

@BarrySmith Ok that should be fixed (and the PFLARE CI has been updated with -fp_trap on to catch this in the future).

As it looks like your PETSc MR is close to merging, any chance you could use the PFLARE commit hash f4ad40b rather than this branch name, as I tag any commit that makes it into PETSc main so github doesn't garbage collect away a branch.

[BarrySmith]

Satish likes to enforce hashes before merge so we will use a hash when merged. Until then I find it is simpler to use the branch name otherwise I have to constantly change commits and will make mistakes.

[stevendargaville]

Sure no worries 😄