Michael, > I have managed to install deal.II-6.3.pre (revision 19078) with PETSC, > Trilinos, Metis and threading on two separate Ubuntu systems. Once > installed I tested some of the examples to verify that the libraries > were working. I found that step-1 and steps which use MPI ran (I > tested step-17 and step-31) but steps which did not use MPI failed > (specifically step-2 and step-22). The specific order of the stack > trace is different in both cases, but I have noticed that both > originate in the Subscriptor deconstructor and that both involve that > ConstraintMatrix class.
My suspicion for the cause of the problem would be one of these: - you built Trilinos or PETSc with MPI enabled - deal.II was built without MPI enabled or - you built Trilinos or PETSc with MPI enabled - deal.II was built with MPI enabled - as a consequence when any of the programs that don't use MPI explicitly (such as step-2 and step-22) are run, the ConstraintMatrix does something with MPI anyway; since MPI_Initialize wasn't called in main(), we fail Does any of these scenarios sound right? In particular, if the second scenario were right, what happens if you add the MPI_Initialize statement just like in step-17? Best Wolfgang ------------------------------------------------------------------------- Wolfgang Bangerth email: [email protected] www: http://www.math.tamu.edu/~bangerth/ _______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
