I have added this code: if(sgid != -1 ) { SETERRQ(PETSC_COMM_SELF,PETSC_ERR_PLIB,"Have un-symmetric graph (apparently). Use '-pc_gamg_sym_graph true' to symetrize the graph or '-pc_gamg_threshold 0.0' if the matrix is structurally symmetric."); } I'm trying to build now. LuLu: I've changed the defaults so it looks like you want to add '-pc_gamg_sym_graph true' to get back to the old semantics. Mark On Jul 13, 2012, at 6:20 PM, Jed Brown wrote:

Mark, in a nonsymmetric (but structurally symmetric) bs=2 problem from Lulu, we are tripping this assertion.

else{ /* I'm stealing this local, owned by a ghost */ assert(sgid==-1);

What does this mean? Can't we at least always use SETERRQ for errors?

Also from that section of code, I notice you are using PETSC_ERR_LIB for internal PETSc logic/corruption errors. They look like they should be PETSC_ERR_PLIB.