Remove an abstraction barrier I thought was gone long-ago. The OOB subscription really shouldn't be defined as an OMPI subscription.

I know it's just a technicality, but it is time to address such things rather than just letting them continue to propagate. :-)

This commit was SVN r12954.

orte/mca/oob/oob_types.h

@@ -26,4 +26,6 @@
 #include "orte_config.h"
 #include "orte/orte_constants.h"
 
+#define ORTE_OOB_SUBSCRIPTION   "orte-oob-sub"
+
 #endif  /* MCA_OOB_TYPES_H */

orte/mca/oob/tcp/oob_tcp.c

@@ -46,7 +46,6 @@
 #include "orte/mca/errmgr/errmgr.h"
 #include "orte/mca/ns/ns.h"
 #include "orte/mca/gpr/gpr.h"
-#include "ompi/constants.h"
 
 /*
  * Data structure for accepting connections.
@@ -961,7 +960,7 @@ int mca_oob_tcp_resolve(mca_oob_tcp_peer_t* peer)
     }
 
     if (ORTE_SUCCESS != (rc = orte_schema.get_std_subscription_name(&sub_name,
-                                OMPI_OOB_SUBSCRIPTION, peer->peer_name.jobid))) {
+                                ORTE_OOB_SUBSCRIPTION, peer->peer_name.jobid))) {
         ORTE_ERROR_LOG(rc);
         return rc;
     }
@@ -1109,7 +1108,7 @@ int mca_oob_tcp_init(void)
     }
 
     if (ORTE_SUCCESS != (rc = orte_schema.get_std_subscription_name(&sub_name,
-                                OMPI_OOB_SUBSCRIPTION, jobid))) {
+                                ORTE_OOB_SUBSCRIPTION, jobid))) {
         ORTE_ERROR_LOG(rc);
         return rc;
     }