Hong, you are doing well with the formal workflow, but in this last round with 'hzhang/mat-cleanup', it looks like you didn't check the nightly logs before merging (they all failed).
http://ftp.mcs.anl.gov/pub/petsc/nightlylogs/make_next_arch-linux-gcc-ifc_crank.log Matt fixed the problem below, unbreaking 'master', but this should never have been merged when (a) the nightlies failed, (b) the branch cannot be compiled on any machine, and (c) the result of the merge cannot be compiled on any machine. Remember that when integrating (e.g., to 'next', 'master', etc), you should _always_ build the result of the merge before pushing. commit 275476c694b7affb79a129990509c16cc25cadec Author: Matthew G. Knepley <[email protected]> Date: Wed Jul 3 17:23:45 2013 -0500 Mat: Need a scope when declaring variables in a switch block diff --git a/src/mat/impls/aij/mpi/mpimatmatmult.c b/src/mat/impls/aij/mpi/mpimatmatmult.c index bf6f620..f924000 100644 --- a/src/mat/impls/aij/mpi/mpimatmatmult.c +++ b/src/mat/impls/aij/mpi/mpimatmatmult.c @@ -823,6 +823,7 @@ PetscErrorCode MatTransposeMatMult_MPIAIJ_MPIAIJ(Mat P,Mat A,MatReuse scall,Pets ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ_nonscalable(P,A,fill,C);CHKERRQ(ierr); break; case 2: + { Mat Pt; Mat_PtAPMPI *ptap; Mat_MPIAIJ *c; @@ -833,6 +834,7 @@ PetscErrorCode MatTransposeMatMult_MPIAIJ_MPIAIJ(Mat P,Mat A,MatReuse scall,Pets ptap->Pt = Pt; (*C)->ops->mattransposemultnumeric = MatTransposeMatMultNumeric_MPIAIJ_MPIAIJ_matmatmult; PetscFunctionReturn(0); + } break; default: ierr = MatTransposeMatMultSymbolic_MPIAIJ_MPIAIJ(P,A,fill,C);CHKERRQ(ierr);
pgpWfSaH18siE.pgp
Description: PGP signature
