I missed the fact that the app doesn't force it. But if this is indeed the case then it is extremely weird that you are seing someone else releasing your proc.
Regarding the destruction of the proc, the OPAL layer only does in a single place, when the local proc is set (opal_proc_local_set). Moreover, it does call OBJ_RETAIN when it does this, so the proc should not vanish without you having control over it. I looked at the code and noticed that it only crash in apps, the place where the ORTE proc is not provided to the OPAL layer. George. On Fri, Aug 1, 2014 at 12:12 PM, Ralph Castain <r...@open-mpi.org> wrote: > > On Aug 1, 2014, at 8:27 AM, George Bosilca <bosi...@icl.utk.edu> wrote: > > This commit brings two things. One if the renaming suggested by Gilles. > The second one is forcing the ORTE process down on the OPAL. This doesn't > fit the current design of the BTL move. The current design assumes that the > local OPAL process is part of the local OMPI process. > > > Your statement isn't accurate - this commit sets the opal_proc_t for all > *non-MPI* processes. As the comment in ess_base_std_app.c notes, and the > commit message states, ORTE sets and controls the opal_proc_local structure > for all ORTE tools and non-MPI procs as (shockingly) they don't call > MPI_Init, and hence don't go thru ompi_proc_init, and were therefore > leaving the opal_proc_local structure set to the default "nothing" state. > This caused all the OPAL layer functions that access it to think nothing > had been setup yet. > > My destruct issue is caused by the OPAL layer destructing the object, > which seems odd to me but <shrug> > > > George. > > PS: If it doesn't break lose everywhere is because the OMPI layer reset > it's own process after the RTE (which explain why Ralph notice that his > proc has been OBJ_DESTRUCT). > > > > On Fri, Aug 1, 2014 at 10:44 AM, <svn-commit-mai...@open-mpi.org> wrote: > >> Author: rhc (Ralph Castain) >> Date: 2014-08-01 10:44:11 EDT (Fri, 01 Aug 2014) >> New Revision: 32398 >> URL: https://svn.open-mpi.org/trac/ompi/changeset/32398 >> >> Log: >> Some more cleanups. Remove direct references to ORTE by changing >> OMPI_CAST_ORTE_NAME -> OMPI_CAST_RTE_NAME. Ensure that ORTE tools (mpirun, >> orted, tools) set the OPAL proc structure fields so OPAL knows what is >> going on and uses the correct print functions (still need to fix the >> problem for non-MPI apps). Properly return uint32_t from the opal utilities >> instead of int32_t as that is what the ORTE process name fields contain. >> >> Thanks to Gilles for pointing out some of the discrepancies. >> >> Text files modified: >> trunk/ompi/mca/bcol/basesmuma/bcol_basesmuma_smcm.c | 2 >> trunk/ompi/mca/coll/hierarch/coll_hierarch.c | 2 >> trunk/ompi/mca/coll/sm/coll_sm_module.c | 6 ++-- >> trunk/ompi/mca/dpm/orte/dpm_orte.c | 10 ++++---- >> trunk/ompi/mca/pml/bfo/pml_bfo_failover.c | 6 ++-- >> trunk/ompi/mca/rte/orte/rte_orte.h | 2 >> trunk/ompi/proc/proc.c | 14 >> ++++++------ >> trunk/ompi/runtime/ompi_mpi_abort.c | 4 +- >> trunk/ompi/runtime/ompi_mpi_init.c | 4 +- >> trunk/opal/util/proc.c | 9 +++---- >> trunk/opal/util/proc.h | 4 +- >> trunk/orte/mca/ess/base/ess_base_std_orted.c | 9 ++++++++ >> trunk/orte/mca/ess/base/ess_base_std_tool.c | 9 ++++++++ >> trunk/orte/mca/ess/hnp/ess_hnp_module.c | 8 +++++++ >> trunk/orte/runtime/orte_init.c | 42 >> ++++++++++++++++++++++++++++++++++++++++ >> trunk/orte/util/proc_info.c | 6 +++++ >> trunk/orte/util/proc_info.h | 4 ++ >> 17 files changed, 108 insertions(+), 33 deletions(-) >> >> >> Diff not shown due to size (21547 bytes). >> To see the diff, run the following command: >> >> svn diff -r 32397:32398 --no-diff-deleted >> >> _______________________________________________ >> svn mailing list >> s...@open-mpi.org >> http://www.open-mpi.org/mailman/listinfo.cgi/svn >> > > _______________________________________________ > devel mailing list > de...@open-mpi.org > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel > Link to this post: > http://www.open-mpi.org/community/lists/devel/2014/08/15456.php > > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > Subscription: http://www.open-mpi.org/mailman/listinfo.cgi/devel > Link to this post: > http://www.open-mpi.org/community/lists/devel/2014/08/15457.php >
