Multiple sigchld reports can occur within a single event callback, so have to reap them until none remain. Also, need to ensure the daemon is flagged as alive prior to calling wait_cb

Refs trac:4717

This commit was SVN r32020.

The following Trac tickets were found above:
  Ticket 4717 --> https://svn.open-mpi.org/trac/ompi/ticket/4717

orte/mca/plm/rsh/plm_rsh_module.c

@@ -921,6 +921,7 @@ static void process_launch_list(int fd, short args, void *cbdata)
         }
         caddy = (orte_plm_rsh_caddy_t*)item;
         /* register the sigchild callback */
+        ORTE_FLAG_SET(caddy->daemon, ORTE_PROC_FLAG_ALIVE);
         orte_wait_cb(caddy->daemon, rsh_wait_daemon, (void*)caddy);
 
         /* fork a child to exec the rsh/ssh session */

orte/runtime/orte_wait.c

@@ -248,12 +248,14 @@ static void wait_signal_callback(int fd, short event, void *arg)
         return;
     }
 
-    /* retrieve the pid */
- retry:
+    /* we can have multiple children leave but only get one
+     * sigchild callback, so reap all the waitpids until we
+     * don't get anything valid back */
+    while (1) {
         pid = waitpid(-1, &status, WNOHANG);
         if (-1 == pid && EINTR == errno) {
             /* try it again */
-        goto retry;
+            continue;
         }
         /* if we got garbage, then nothing we can do */
         if (pid <= 0) {
@@ -270,11 +272,8 @@ static void wait_signal_callback(int fd, short event, void *arg)
                 }
                 opal_list_remove_item(&pending_cbs, &t2->super);
                 OBJ_RELEASE(t2);
-            return;
+                break;
+            }
         }
     }
-
-    /* if we get here, then this sigchild occurred prior to someone
-     * registering it, or after someone mistakenly removed it. Either
-     * way, there really isn't anything we can do with it */
 }