Temporary workaround for MOOSE apps (mis)using LibmeshPetscCall

[nmnobre]

No description provided.

[moosebuild]

Job Coverage, step Generate coverage on b5e4298 wanted to post the following:

Coverage

Coverage did not change

Full coverage report

This comment will be updated on new commits.

[roystgnr]

Copied to https://github.com/libMesh/libmesh/tree/farscape_workaround_test so we can test this in the MOOSE PRs. Please remind me to delete that when we're done with it.

[lindsayad]

Is this what just passed in idaholab/moose#28969?

[nmnobre]

Is this what just passed in idaholab/moose#28969?

Affirm.

• 1) Delete https://github.com/libMesh/libmesh/tree/farscape_workaround_test
• 2) Merge this PR
• 3) Update MOOSE's libMesh submodule
• 4) Rebase and merge Switch to using LibmeshPetscCall wrapper macros idaholab/moose#28929
• 5) Revert this PR and fb0abb5
• 6) Kindly ask devs to update their apps
• 7) Reuse Phantom PR for libMesh testing purposes only 👻 🎃 idaholab/moose#28969 to police devs 😛
• 8) Update MOOSE's libMesh submodule

[lindsayad]

Such a simple process 😆

[jwpeterson]

@nmnobre I think this PR caused a failure in the "Test No Exceptions" testing that we do for the devel -> master merge, see e.g. here. There are many error error messages from the compiler, but some representative ones are:

In file included from ./include/libmesh/petsc_shell_matrix.h:33,
                 from ./include/libmesh/petsc_matrix_shell_matrix.h:26,
                 from ./include/libmesh/static_condensation.h:26,
                 from ../src/base/dof_map.C:42:
./include/libmesh/petsc_vector.h: In constructor 'libMesh::PetscVector<T>::PetscVector(Vec, const libMesh::Parallel::Communicator&)':
./include/libmesh/petsc_vector.h:598:3: error: there are no arguments to 'LibmeshPetscCall' that depend on a template parameter, so a declaration of 'LibmeshPetscCall' must be available [-fpermissive]
   LibmeshPetscCall(VecGetLocalSize(_vec, &petsc_local_size));
   ^~~~~~~~~~~~~~~~
./include/libmesh/petsc_vector.h:598:3: note: (if you use '-fpermissive', G++ will accept your code, but allowing the use of an undeclared name is deprecated)

And many others along the same lines:

./include/libmesh/petsc_vector.h:602:3: error: there are no arguments to 'LibmeshPetscCall' that depend on a template parameter, so a declaration of 'LibmeshPetscCall' must be available [-fpermissive]
   LibmeshPetscCall(VecGetType(_vec, &ptype));
   ^~~~~~~~~~~~~~~~
./include/libmesh/petsc_vector.h:637:7: error: there are no arguments to 'LibmeshPetscCall' that depend on a template parameter, so a declaration of 'LibmeshPetscCall' must be available [-fpermissive]
       LibmeshPetscCall(VecGetLocalToGlobalMapping(_vec, &mapping));
       ^~~~~~~~~~~~~~~~

I see from your previous comment that reverting this PR is part of the 8 step plan so I guess the issue will eventually be resolved, but we won't have any devel -> master merges in the meantime.

[nmnobre]

The errors make sense, the temporary definition is only on the with-exceptions code path. I'm at SC24, I'll try to push a patch this afternoon (or feel free to go ahead with a patch).

[jwpeterson]

It's not urgent for me (and I don't know exactly what the fix is!) so I'll let you take care of it when you have some time. Thanks!