Thanks! On Tue, Mar 26, 2019 at 9:21 AM John Peterson <jwpeter...@gmail.com> wrote:
> > > On Tue, Mar 26, 2019 at 7:59 AM Charlie Talbot <chazti...@gmail.com> > wrote: > >> Hi, I recently configured petsc after pulling from the master branch, >> configuring using the >> >> arch-linux-pkgs-64idx.py >> >> script provided, and specified this when trying to install libmesh-1.4.0, >> which failed with error message: >> ... >> CXX src/numerics/libmesh_dbg_la-sparse_matrix.lo >> CXX src/numerics/libmesh_dbg_la-sparse_shell_matrix.lo >> CXX src/numerics/libmesh_dbg_la-sum_shell_matrix.lo >> CXX src/numerics/libmesh_dbg_la-tensor_shell_matrix.lo >> In file included from ./include/libmesh/petsc_vector.h:31, >> from ../src/numerics/petsc_vector.C:24: >> ../src/numerics/petsc_vector.C: In member function ‘virtual void >> libMesh::PetscVector<T>::localize(libMesh::NumericVector<T>&) const’: >> ./include/libmesh/petsc_macro.h:94:58: error: there are no arguments to >> ‘VecScatterCreateWithData’ that depend on a template parameter, so a >> declaration of ‘VecScatterCreateWithData’ must be available [-fpermissive] >> # define LibMeshVecScatterCreate(xin,ix,yin,iy,newctx) >> VecScatterCreateWithData(xin,ix,yin,iy,newctx) >> > > > The code in question is this: > > // Once PETSc-3.11.0 is released, "&& PETSC_VERSION_RELEASE" should be > removed > #if PETSC_VERSION_LESS_THAN(3,11,0) && PETSC_VERSION_RELEASE > # define LibMeshVecScatterCreate(xin,ix,yin,iy,newctx) > VecScatterCreate(xin,ix,yin,iy,newctx) > #else > # define LibMeshVecScatterCreate(xin,ix,yin,iy,newctx) > VecScatterCreateWithData(xin,ix,yin,iy,newctx) > #endif > > Since you are using PETSc master, you are getting into the #else case > since master == !PETSC_VERSION_RELEASE. But somehow your version of master > doesn't have VecScatterCreateWithData, so it may have been in master for a > while and then they took it back out... > > > > Any idea how to work around this? Should I stick to petsc-3.9._? >> > > Yes. At this point we probably don't have a version of libmesh that works > with PETSc master, it is a bit of a moving target as you can see.. > > -- > John > _______________________________________________ Libmesh-users mailing list Libmesh-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-users
