Yes, I agree - the handle is invalid in the session referenced in the argument. Therefore MPI_T_ERR_INVALID_SESSION seems reasonable to me. Martin On Jul 25, 2013, at 2:03 PM, Kathryn Mohror <[email protected]> wrote:

Hi everybody,

I got asked a question from Nathan:

What should (MPI_T_pvar_read, _write, _reset, etc) return if bad handle is given?

He gave this example of erroneous code: { MPI_T_pvar_session session1, session2; MPI_T_pvar_handle handle; char buffer[256]; int count;

MPI_T_pvar_session_create (&session1); MPI_T_pvar_session_create (&session2);

MPI_T_pvar_handle_alloc (session1, 0, NULL, &handle, &count);

MPI_T_pvar_start (session1, handle);

/* erroneus statement. should return MPI_T_ERR_INVALID_SESSION or MPI_T_ERR_INVALID_HANDLE */ MPI_T_pvar_read (session2, handle, buffer); }

I argue it should return MPI_T_ERR_INVALID_HANDLE since the session is good, it's just that the handle doesn�t match.

Do you all agree? If so, we should specify it since it's ambiguous. I believe this is an errata item.

Kathryn ______________________________________________________________ Kathryn Mohror, [email protected], http://people.llnl.gov/mohror1 CASC @ Lawrence Livermore National Laboratory, Livermore, CA, USA

_______________________________________________ Mpi3-tools mailing list [email protected] http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi3-tools

________________________________________________________________________ Martin Schulz, [email protected], http://people.llnl.gov/schulzm CASC @ Lawrence Livermore National Laboratory, Livermore, USA