Ensure we don't incorrectly return a non-zero exit code when normally terminating a slurm job.

Slurm, of course, must always be different...

This commit was SVN r23371.

orte/mca/errmgr/hnp/errmgr_hnp.c

@@ -399,7 +399,11 @@ static int update_state(orte_jobid_t job,
             /* if we have ordered orteds to terminate, see if this one failed to tell
              * us it had terminated
              */
-            if (orte_orteds_term_ordered) {
+            if (orte_orteds_openend_term_ordered) {
+                record_dead_daemon(jdata, proc->vpid, state, 0);
+                check_job_complete(jdata);
+                break;
+            } else if (orte_orteds_term_ordered) {
                 record_dead_daemon(jdata, proc->vpid, state, exit_code);
                 check_job_complete(jdata);
                 break;

orte/mca/plm/slurm/plm_slurm_module.c

@@ -505,6 +505,7 @@ static int plm_slurm_terminate_orteds(void)
     /* tell them to die without sending a reply - we will rely on the
      * waitpid to tell us when they have exited!
      */
+    orte_orteds_openend_term_ordered = true;
     if (ORTE_SUCCESS != (rc = orte_plm_base_orted_exit(ORTE_DAEMON_EXIT_CMD))) {
         ORTE_ERROR_LOG(rc);
     }

orte/runtime/orte_globals.c

@@ -96,6 +96,7 @@ bool orte_abnormal_term_ordered = false;
 bool orte_routing_is_enabled = false;
 bool orte_job_term_ordered = false;
 bool orte_orteds_term_ordered = false;
+bool orte_orteds_openend_term_ordered = false;
 
 int orte_startup_timeout;
 

orte/runtime/orte_globals.h

@@ -604,6 +604,7 @@ ORTE_DECLSPEC extern bool orte_abnormal_term_ordered;
 ORTE_DECLSPEC extern bool orte_routing_is_enabled;
 ORTE_DECLSPEC extern bool orte_job_term_ordered;
 ORTE_DECLSPEC extern bool orte_orteds_term_ordered;
+ORTE_DECLSPEC extern bool orte_orteds_openend_term_ordered;
 
 ORTE_DECLSPEC extern int orte_startup_timeout;