On Wed, Aug 17, 2022 at 07:31:30PM +0200, Drew Parsons wrote:
> Source: fenics-dolfinx
> Followup-For: Bug #1017284
>
> The cmake error logs can be too effusive, they tend to obscure the
> error. In this case the real problem seems to be that it never found
> scotch, even though it found it. A mystery indeed. Schrödinger's
> SCOTCH.
My reading of cpp/cmake/modules/FindSCOTCH.cmake is that it fails in
if (SCOTCH_dgraphInit(&dgrafdat, MPI_COMM_WORLD) != 0) {
if (MPI_THREAD_MULTIPLE > provided) {
std::cout << \"MPI implementation is not thread-safe:\" << std::endl;
std::cout << \"SCOTCH should be compiled without SCOTCH_PTHREAD\" <<
std::endl;
exit(1);
glibc 2.34 merged libpthread into libc, some kind of threading
detection no longer working would not be a complete mystery.
cu
Adrian
