Rene, The code reads
+ /* If ORTE was not setup yet, don't use orte_show_help */ + if (orte_setup) { + orte_show_help("help-mpi-runtime", + "mpi_init:startup:internal-failure", true, + "MPI_INIT", "MPI_INIT", error, err_msg, ret); + } else { + opal_show_help("help-mpi-runtime", + "mpi_init:startup:internal-failure", true, + "MPI_INIT", "MPI_INIT", error, err_msg, ret); + } The difference is indeed very subtle: in one case we call ORTE_show_help while on the other we call OPAL_show_help. Exactly as the log message indicates. George. On Jan 31, 2014, at 15:49 , René Oertel <rene.oer...@cs.tu-chemnitz.de> wrote: > Dear OpenMPI developers, > > I think the source code modification titled > "Ensure that we don't try to call orte_show_help() before orte_init() > succeeds." is buggy (see [1] at lines 755-763). > There is no difference between the if and the else path of the code, > i.e. checking orte_setup/rte_setup makes no sense. > You will find the erroneous code below the 'error' label in > ompi/runtime/ompi_mpi_init.c in SVN trunk, too. > > Kind regards, > > René > > [1] > https://github.com/open-mpi/ompi-svn-mirror/commit/b200ea0e1554d40ed4948f9823fc6dd7b0cc3ba6#diff-3673b21a7f42dc0665ea4470b3171df1 > > > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/devel