Don't assume things about the contact info string in the general case. There

is no need for the IP address in most cases (filem being one dubious
exception), so just publish and hand around the supposedly opaque contact
info strings

This commit was SVN r15638.

orte/mca/filem/base/filem_base_fns.c

@@ -30,11 +30,13 @@
 #include "opal/mca/base/base.h"
 
 #include "opal/util/output.h"
+#include "opal/util/argv.h"
 #include "opal/mca/base/mca_base_param.h"
 #include "opal/util/os_dirpath.h"
 
 #include "orte/mca/gpr/gpr.h"
 #include "orte/mca/rml/rml.h"
+#include "orte/mca/rml/base/rml_contact.h"
 
 #include "orte/mca/filem/filem.h"
 #include "orte/mca/filem/base/base.h"
@@ -157,7 +159,7 @@ int orte_filem_base_get_proc_node_name(orte_process_name_t *proc, char **machine
     /* if it is the root then we need a different key :/ */
     if(proc->jobid  == 0 &&
        proc->vpid   == 0) {
-        keys[0] = ORTE_PROC_RML_IP_ADDRESS_KEY;
+        keys[0] = ORTE_PROC_RML_CONTACT_KEY;
     }
     else {
         keys[0] = ORTE_NODE_NAME_KEY;
@@ -221,6 +223,33 @@ int orte_filem_base_get_proc_node_name(orte_process_name_t *proc, char **machine
         }
     }
 
+    if (proc->jobid == 0 && proc->vpid == 0) {
+        /* we have contact info -- need an IP address.  This assumes
+           that we're using the OOB RML, but that's probably a safe
+           enough assumption in here. */
+
+        char *contact_info = *machine_name;
+        orte_process_name_t peer;
+        char **uris;
+        char *ip, *port;
+
+        *machine_name = NULL;
+        ret = orte_rml_base_parse_uris(contact_info, &peer, &uris);
+        free(contact_info);
+        if (ORTE_SUCCESS == ret) {
+            exit_status = ret;
+            goto cleanup;
+        }
+
+        ip = strrchr(uris[0], '/') + 1;
+        port = strrchr(uris[0], ':');
+        port[0] = '\0';
+
+        *machine_name = strdup(ip);
+
+        opal_argv_free(uris);
+    }
+    
     if (NULL == *machine_name ){
         exit_status = ORTE_ERROR;
         goto cleanup;

orte/mca/rmaps/base/rmaps_base_registry_fns.c

@@ -84,7 +84,7 @@ int orte_rmaps_base_get_job_map(orte_job_map_t **map, orte_jobid_t jobid)
     char* dkeys[] = {
         ORTE_PROC_NAME_KEY,
         ORTE_NODE_NAME_KEY,
-        ORTE_PROC_RML_IP_ADDRESS_KEY,
+        ORTE_PROC_RML_CONTACT_KEY,
         NULL
     };
     bool daemon_exists;
@@ -346,7 +346,7 @@ int orte_rmaps_base_get_job_map(orte_job_map_t **map, orte_jobid_t jobid)
                 }
                 continue;
             }
-            if (strcmp(keyval->key, ORTE_PROC_RML_IP_ADDRESS_KEY) == 0) {
+            if (strcmp(keyval->key, ORTE_PROC_RML_CONTACT_KEY) == 0) {
                 /* we don't care about the value here - the existence of the key is
                  * enough to indicate that this daemon must already exist, so flag it
                  */

orte/mca/rmaps/base/rmaps_base_support_fns.c

@@ -630,7 +630,7 @@ int orte_rmaps_base_define_daemons(orte_job_map_t *map)
     char* dkeys[] = {
         ORTE_PROC_NAME_KEY,
         ORTE_NODE_NAME_KEY,
-        ORTE_PROC_RML_IP_ADDRESS_KEY,
+        ORTE_PROC_RML_CONTACT_KEY,
         NULL
     };
     orte_gpr_value_t **dvalues=NULL, *value;
@@ -681,7 +681,7 @@ int orte_rmaps_base_define_daemons(orte_job_map_t *map)
                 }
                 continue;
             }
-            if (strcmp(keyval->key, ORTE_PROC_RML_IP_ADDRESS_KEY) == 0) {
+            if (strcmp(keyval->key, ORTE_PROC_RML_CONTACT_KEY) == 0) {
                 /* we don't care about the value here - the existence of the key is
                  * enough to indicate that this daemon must already exist, so flag it
                  */

orte/mca/rml/base/rml_base_contact.c

@@ -14,8 +14,6 @@
 #include "orte/mca/gpr/gpr.h"
 #include "orte/mca/oob/oob_types.h"
 
-#define ORTE_RML_BASE_CONTACT_KEY "rml-contact"
-
 extern opal_list_t       orte_rml_base_subscriptions;
 
 struct orte_rml_base_subscription_t {
@@ -31,8 +29,7 @@ static int get_contact_info(orte_jobid_t job, char **tokens, orte_gpr_notify_dat
 {
     char *segment;
     char *keys[] = {
-        ORTE_RML_BASE_CONTACT_KEY,
-        ORTE_PROC_RML_IP_ADDRESS_KEY,
+        ORTE_PROC_RML_CONTACT_KEY,
         NULL
     };
     orte_gpr_value_t **values;
@@ -151,7 +148,7 @@ orte_rml_base_register_subscription(orte_jobid_t jobid, char *trigger)
                                          ORTE_GPR_KEYS_OR | ORTE_GPR_TOKENS_OR | ORTE_GPR_STRIPPED,
                                          segment,
                                          NULL,  /* look at all containers on this segment */
-                                         ORTE_RML_BASE_CONTACT_KEY,
+                                         ORTE_PROC_RML_CONTACT_KEY,
                                          orte_rml_base_contact_info_notify, NULL))) {
         ORTE_ERROR_LOG(rc);
         free(sub_name);
@@ -177,10 +174,10 @@ int
 orte_rml_base_register_contact_info(void)
 {
     orte_std_cntr_t i, num_tokens;
-    orte_data_value_t *values[2];
-    char *tmp, *tmp2, *tmp3;
+    orte_data_value_t *values[1];
+    char *tmp;
     char *segment, **tokens;
-    char *keys[] = { ORTE_RML_BASE_CONTACT_KEY, ORTE_PROC_RML_IP_ADDRESS_KEY};
+    char *keys[] = { ORTE_PROC_RML_CONTACT_KEY };
     int rc;
     
     /* setup to put our contact info on registry */
@@ -194,28 +191,6 @@ orte_rml_base_register_contact_info(void)
     values[0]->data = strdup(tmp);
     free(tmp);
 
-    /* setup the IP address for storage */
-    tmp = orte_rml.get_contact_info();
-    tmp2 = strrchr(tmp, '/') + 1;
-    tmp3 = strrchr(tmp, ':');
-    if(NULL == tmp2 || NULL == tmp3) {
-        opal_output(0, "%s orte_rml_base_init: invalid address \'%s\' "
-                    "returned for selected oob interfaces.\n",
-                    ORTE_NAME_PRINT(orte_process_info.my_name), tmp);
-        ORTE_ERROR_LOG(ORTE_ERROR);
-        free(tmp);
-        return ORTE_ERROR;
-    }
-    *tmp3 = '\0';
-    values[1] = OBJ_NEW(orte_data_value_t);
-    if (NULL == values[1]) {
-        ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
-        return ORTE_ERR_OUT_OF_RESOURCE;
-    }
-    values[1]->type = ORTE_STRING;
-    values[1]->data = strdup(tmp2);
-    free(tmp);
-
     /* define the segment */
     if (ORTE_SUCCESS != (rc = orte_schema.get_job_segment_name(&segment, ORTE_PROC_MY_NAME->jobid))) {
         ORTE_ERROR_LOG(rc);
@@ -236,7 +211,7 @@ orte_rml_base_register_contact_info(void)
 
     /* put our contact info in registry */
     if (ORTE_SUCCESS != (rc = orte_gpr.put_N(ORTE_GPR_OVERWRITE | ORTE_GPR_TOKENS_XAND,
-                                        segment, tokens, 2, keys, values))) {
+                                        segment, tokens, 1, keys, values))) {
         ORTE_ERROR_LOG(rc);
     }
 
@@ -247,7 +222,6 @@ orte_rml_base_register_contact_info(void)
     }
     if (NULL != tokens) free(tokens);
     OBJ_RELEASE(values[0]);
-    OBJ_RELEASE(values[1]);
 
     return rc;
 }
@@ -273,7 +247,7 @@ orte_rml_base_contact_info_notify(orte_gpr_notify_data_t* data,
 
                 /* check to make sure this is the requested key */
                 keyval = value->keyvals[j];
-                if(strcmp(keyval->key, ORTE_RML_BASE_CONTACT_KEY) != 0)
+                if(strcmp(keyval->key, ORTE_PROC_RML_CONTACT_KEY) != 0)
                     continue;
                 orte_dss.get((void**)&(contact_info), keyval->value, ORTE_STRING);
                 orte_rml.set_contact_info(contact_info);

orte/mca/schema/schema_types.h

@@ -115,7 +115,7 @@
 #define ORTE_PROC_NUM_AT_STG3                   "orte-proc-num-stg3"
 #define ORTE_PROC_NUM_FINALIZED                 "orte-proc-num-finalized"
 #define ORTE_PROC_NUM_TERMINATED                "orte-proc-num-terminated"
-#define ORTE_PROC_RML_IP_ADDRESS_KEY            "orte-proc-rml-ip-addr"
+#define ORTE_PROC_RML_CONTACT_KEY               "orte-proc-rml-contact"
 #define ORTE_PROC_CPU_LIST_KEY                  "orte-proc-cpu-list"
 
 #define ORTE_JOB_CKPT_STATE_KEY                "orte-job-ckpt-state"