One more correction to mpirun exit codes - cleanup the application proc's exit codes in the orted so that non-zero exit codes generated by mpirun itself don't get "munged".

Modify the multi_abort function so they all return different exit codes - allows us to tell which one was being reported.

This commit was SVN r17895.

orte/mca/odls/base/odls_base_default_fns.c

@@ -1436,9 +1436,6 @@ GOTCHILD:
         goto MOVEON;
     }
 
-    /* save the exit code */
-    child->exit_code = status;
-
     /* If this child was the (vpid==0), we hooked it up to orterun's
        STDIN SOURCE earlier (do not change this without also changing
        odsl_default_fork_local_proc()).  So we have to tell the SOURCE
@@ -1481,6 +1478,9 @@ GOTCHILD:
 
     /* determine the state of this process */
     if(WIFEXITED(status)) {
+        /* set the exit status appropriately */
+        child->exit_code = WEXITSTATUS(status);
+        
         /* even though the process exited "normally", it is quite
          * possible that this happened via an orte_abort call - in
          * which case, we need to indicate this was an "abnormal"
@@ -1556,6 +1556,13 @@ GOTCHILD:
          * abnormal, so indicate that condition
          */
         child->state = ORTE_PROC_STATE_ABORTED_BY_SIG;
+        /* If a process was killed by a signal, then make the
+         * exit code of orterun be "signo + 128" so that "prog"
+         * and "orterun prog" will both set the same status
+         * value for the shell
+         */
+        child->exit_code = WTERMSIG(status) + 128;
+                
         OPAL_OUTPUT_VERBOSE((5, orte_odls_globals.output,
                              "%s odls:wait_local_proc child process %s terminated with signal",
                              ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),

orte/test/mpi/multi_abort.c

@@ -18,7 +18,7 @@ int main(int argc, char* argv[])
 
     printf("Hello, World, I am %d of %d\n", rank, size);
 
-    if (0 != rank) MPI_Abort(MPI_COMM_WORLD, 2);
+    if (0 != rank) MPI_Abort(MPI_COMM_WORLD, rank);
 
     MPI_Finalize();
     return 0;

orte/tools/orterun/orterun.c

@@ -580,19 +580,6 @@ static void job_completed(int trigpipe, short event, void *arg)
                    num_killed, ((num_killed > 1) ? "es" : ""), orterun_basename);
         }
     }
-    /* Make sure we propagate the exit code */
-    if (WIFEXITED(orte_exit_status)) {
-        orte_exit_status = WEXITSTATUS(orte_exit_status);
-    }  else if (ORTE_JOB_STATE_FAILED_TO_START == exit_state ||
-                ORTE_JOB_STATE_ABORTED_WO_SYNC == exit_state) {
-        /* ensure we don't treat this like a signal */
-    } else {
-        /* If a process was killed by a signal, then make the
-         * exit code of orterun be "signo + 128" so that "prog"
-         * and "orterun prog" will both set the same status
-         * value for the shell */
-        orte_exit_status = WTERMSIG(orte_exit_status) + 128;
-    }
     
     /* the job is complete - now setup an event that will
      * trigger when the orteds are gone and tell the orteds that it is