Fix bugs in the oob base - ensure we get the components in high-to-low priority, and that we correctly track reachability via all components. Adjust the priority of the tcp component to leave headroom for others

Refs trac:267

This commit was SVN r30740.

The following Trac tickets were found above:
  Ticket 267 --> https://svn.open-mpi.org/trac/ompi/ticket/267

orte/mca/oob/base/oob_base_select.c

@@ -93,7 +93,7 @@ int orte_oob_base_select(void)
         added = false;
         OPAL_LIST_FOREACH(cmp, &orte_oob_base.actives, mca_base_component_list_item_t) {
             c3 = (mca_oob_base_component_t *) cmp->cli_component;
-            if (c3->priority < component->priority) {
+            if (c3->priority > component->priority) {
                 continue;
             }
             opal_output_verbose(5, orte_oob_base_framework.framework_output,

orte/mca/oob/base/oob_base_stubs.c

@@ -76,6 +76,7 @@ void orte_oob_base_send_nb(int fd, short args, void *cbdata)
              * it can reach it
              */
             reachable = false;
+            pr = NULL;
             OPAL_LIST_FOREACH(cli, &orte_oob_base.actives, mca_base_component_list_item_t) {
                 component = (mca_oob_base_component_t*)cli->cli_component;
                 if (NULL != component->is_reachable) {
@@ -83,13 +84,15 @@ void orte_oob_base_send_nb(int fd, short args, void *cbdata)
                         /* there is a way to reach this peer - record it
                          * so we don't waste this time again
                          */
-                        pr = OBJ_NEW(orte_oob_base_peer_t);
+                        if (NULL == pr) {
-                        if (OPAL_SUCCESS != (rc = opal_hash_table_set_value_uint64(&orte_oob_base.peers, ui64, (void*)pr))) {
+                            pr = OBJ_NEW(orte_oob_base_peer_t);
-                            ORTE_ERROR_LOG(rc);
+                            if (OPAL_SUCCESS != (rc = opal_hash_table_set_value_uint64(&orte_oob_base.peers, ui64, (void*)pr))) {
-                            msg->status = ORTE_ERR_ADDRESSEE_UNKNOWN;
+                                ORTE_ERROR_LOG(rc);
-                            ORTE_RML_SEND_COMPLETE(msg);
+                                msg->status = ORTE_ERR_ADDRESSEE_UNKNOWN;
-                            OBJ_RELEASE(cd);
+                                ORTE_RML_SEND_COMPLETE(msg);
-                            return;
+                                OBJ_RELEASE(cd);
+                                return;
+                            }
                         }
                         /* mark that this component can reach the peer */
                         opal_bitmap_set_bit(&pr->addressable, component->idx);

orte/mca/oob/tcp/oob_tcp_component.c

@@ -111,7 +111,7 @@ mca_oob_tcp_component_t mca_oob_tcp_component = {
             MCA_BASE_METADATA_PARAM_CHECKPOINT
         },
         0,   // reserve space for an assigned index
-        100, // default priority of this transport
+        30, // default priority of this transport
         component_available,
         component_startup,
         component_shutdown,