minrk
commented
10 months ago lldb gives this stack:
* frame #0: 0x000000010c8f2b79 libpetsc.3.20.0.dylib`PetscOptionsFindPair + 153
frame #1: 0x000000010c8f7595 libpetsc.3.20.0.dylib`PetscOptionsDeprecated_Private + 117
frame #2: 0x000000010c8c3ee0 libpetsc.3.20.0.dylib`PetscInfoProcessClass + 80
frame #3: 0x000000010cec8e5c libpetsc.3.20.0.dylib`MatMFFDInitializePackage + 156
frame #4: 0x000000010ced3a2d libpetsc.3.20.0.dylib`MatInitializePackage + 61
frame #5: 0x000000010cf1b5fd libpetsc.3.20.0.dylib`MatCreate + 45
frame #6: 0x000000010a830ba5 libdolfin.2019.1.0.dylib`dolfin::PETScDMCollection::create_transfer_matrix(dolfin::FunctionSpace const&, dolfin::FunctionSpace const&) + 13621tracing it to MatCreate called here, which ultimately calls MatMFFDInitializePackage -> PetscInfoProcessClass
and finally the segfault occurs here:
PetscCall(PetscOptionsDeprecated_Private(NULL, "-info_exclude", NULL, "3.13", "Use ~ with -info to indicate classes to exclude"));which calls:
PetscCall(PetscOptionsFindPair(options, prefix, oldname, &value, &found));with *prefix=NULL, options=NULL, oldname="-info_exclude"
So the segfault is somewhere in PetscOptionsFindPair. I can reproduce this with the simple C program:
#include "petsc.h"
int main() {
PetscBool found;
const char* value;
char *prefix = NULL;
PetscOptions options = NULL;
const char oldname[] = "-info_exclude";
PetscCall(PetscOptionsFindPair(options, prefix, oldname, &value, &found));
return 0;
}
sblauth
keiyamamo
Solution to issue cannot be found in the documentation.
Issue
Hi, I am having an issue using
fenics.PETScDMCollectionor rather the result of its methodcreate_transfer_matrix. I have a MWE here:Using this, I get a segmentation fault or PETSc error 98 (general MPI error, even though I am not using MPI to run the code).
Is this expected? What workaround is possible? I am using this functionality inside my package to create a reusable interpolation matrix.
Many thanks in advance.
Installed packages
Environment info