Add some further debug to the dstore framework. When doing comm_spawn, we have to exchange any provided cpu bitmaps to ensure both sides compute the same locality, else various mpi frameworks can go bonkers.

This commit was SVN r31572.

ompi/proc/proc.c

@@ -194,9 +194,15 @@ int ompi_proc_set_locality(ompi_proc_t *proc)
 
             /* retrieve the binding for the other proc */
             OBJ_CONSTRUCT(&myvals, opal_list_t);
-            if (OMPI_SUCCESS != opal_dstore.fetch(opal_dstore_internal,
-                                                  (opal_identifier_t*)&proc->proc_name,
-                                                  OPAL_DSTORE_CPUSET, &myvals)) {
+            if (OMPI_SUCCESS != (ret = opal_dstore.fetch(opal_dstore_internal,
+                                                         (opal_identifier_t*)&proc->proc_name,
+                                                         OPAL_DSTORE_CPUSET, &myvals))) {
+                /* check the nonpeer data in case of comm_spawn */
+                ret = opal_dstore.fetch(opal_dstore_nonpeer,
+                                        (opal_identifier_t*)&proc->proc_name,
+                                        OPAL_DSTORE_CPUSET, &myvals);
+            }
+            if (OMPI_SUCCESS != ret) {
                 /* we don't know their cpuset, so nothing more we can say */
                 locality = OPAL_PROC_ON_NODE;
             } else {

opal/mca/dstore/base/dstore_base_stubs.c

@@ -88,6 +88,9 @@ int opal_dstore_base_store(int dstorehandle,
         return OPAL_ERR_NOT_FOUND;
     }
 
+    opal_output_verbose(1, opal_dstore_base_framework.framework_output,
+                        "storing data in %s dstore", (NULL == hdl->name) ? "NULL" : hdl->name);
+
     return hdl->module->store((struct opal_dstore_base_module_t*)hdl->module, id, kv);
 }
 
@@ -102,6 +105,8 @@ void opal_dstore_base_commit(int dstorehandle,
     }
 
     if (NULL != hdl->module->commit) {
+        opal_output_verbose(1, opal_dstore_base_framework.framework_output,
+                            "committing data in %s dstore", (NULL == hdl->name) ? "NULL" : hdl->name);
         hdl->module->commit((struct opal_dstore_base_module_t*)hdl->module, id);
     }
 }
@@ -119,6 +124,9 @@ int opal_dstore_base_fetch(int dstorehandle,
         return OPAL_ERR_NOT_FOUND;
     }
 
+    opal_output_verbose(1, opal_dstore_base_framework.framework_output,
+                        "fetching data from %s dstore", (NULL == hdl->name) ? "NULL" : hdl->name);
+
     return hdl->module->fetch((struct opal_dstore_base_module_t*)hdl->module, id, key, kvs);
 }
 
@@ -133,6 +141,9 @@ int opal_dstore_base_remove_data(int dstorehandle,
         return OPAL_ERR_NOT_FOUND;
     }
 
+    opal_output_verbose(1, opal_dstore_base_framework.framework_output,
+                        "removing data from %s dstore", (NULL == hdl->name) ? "NULL" : hdl->name);
+
     return hdl->module->remove((struct opal_dstore_base_module_t*)hdl->module, id, key);
 }
 

opal/mca/dstore/hash/dstore_hash.c

@@ -179,6 +179,9 @@ static int fetch(struct opal_dstore_base_module_t *imod,
                 OPAL_ERROR_LOG(rc);
                 return rc;
             }
+            OPAL_OUTPUT_VERBOSE((5, opal_dstore_base_framework.framework_output,
+                                 "dstore:hash:fetch: adding data for key %s on proc %" PRIu64 "",
+                                 (NULL == kv->key) ? "NULL" : kv->key, id));
             /* add it to the output list */
             opal_list_append(kvs, &knew->super);
         }

orte/util/nidmap.c

@@ -1151,6 +1151,13 @@ int orte_util_decode_pidmap(opal_byte_object_t *bo)
                     OBJ_DESTRUCT(&kv);
                     goto cleanup;
                 }
+                /* also need a copy in nonpeer to support dynamic spawns */
+                if (ORTE_SUCCESS != (rc = opal_dstore.store(opal_dstore_nonpeer,
+                                                            (opal_identifier_t*)&proc, &kv))) {
+                    ORTE_ERROR_LOG(rc);
+                    OBJ_DESTRUCT(&kv);
+                    goto cleanup;
+                }
                 OBJ_DESTRUCT(&kv);
                 free(cpu_bitmap);
             }