updated documentation for slurm_receive_msg and fixes where that function wasn't used correctly

src/api/slurm_pmi.c

@@ -110,6 +110,7 @@ int  slurm_get_kvs_comm_set(struct kvs_comm_set **kvs_set_ptr,
 	uint16_t port;
 	kvs_get_msg_t data;
 	char *env_pmi_ifhn;
+	List ret_list = NULL;
 
 	if (kvs_set_ptr == NULL)
 		return EINVAL;
@@ -183,12 +184,20 @@ int  slurm_get_kvs_comm_set(struct kvs_comm_set **kvs_set_ptr,
 		error("slurm_accept_msg_conn: %m");
 		return errno;
 	}
-	while (slurm_receive_msg(srun_fd, &msg_rcv, 0) < 0) {
+	while ((ret_list = slurm_receive_msg(srun_fd, &msg_rcv, 0)) == NULL) {
 		if (errno == EINTR)
 			continue;
 		error("slurm_receive_msg: %m");
 		return errno;
 	}
+	if(ret_list) {
+		if(list_count(ret_list)>0) {
+			error("We didn't do things correctly "
+			      "got %d responses didn't expect any",
+			      list_count(ret_list));
+		}
+		list_destroy(ret_list);
+	}
 	msg_rcv.conn_fd = srun_fd;
 	if (msg_rcv.msg_type != PMI_KVS_GET_RESP) {
 		error("slurm_get_kvs_comm_set msg_type=%d", msg_rcv.msg_type);

src/common/slurm_protocol_api.c

@@ -697,11 +697,13 @@ int slurm_close_accepted_conn(slurm_fd open_fd)
 \**********************************************************************/
 
 /*
- * NOTE: memory is allocated for the returned msg and must be freed at 
- *	some point using the slurm_free_functions
+ * NOTE: memory is allocated for the returned msg and the returned list
+ *       both must be freed at some point using the slurm_free_functions 
+ *       and list_destroy function.
  * IN open_fd	- file descriptor to receive msg on
  * OUT msg	- a slurm_msg struct to be filled in by the function
- * RET int	- size of msg received in bytes before being unpacked
+ * RET List	- List containing the responses of the childern (if any) we 
+ *                forwarded the message to.
  */
 List slurm_receive_msg(slurm_fd fd, slurm_msg_t *msg, int timeout)
 {
@@ -1316,6 +1318,8 @@ int slurm_send_rc_msg(slurm_msg_t *msg, int rc)
 
 /*
  * Send and recv a slurm request and response on the open slurm descriptor
+ * with a list containing the responses of the childern (if any) we 
+ * forwarded the message to.
  */
 static List 
 _send_and_recv_msg(slurm_fd fd, slurm_msg_t *req, 
@@ -1408,7 +1412,7 @@ int slurm_send_recv_controller_msg(slurm_msg_t *req, slurm_msg_t *resp)
 		if(ret_list) {
 			if(list_count(ret_list)>0) {
 				error("We didn't do things correctly "
-				      "missed %d responses",
+				      "got %d responses didn't expect any",
 				      list_count(ret_list));
 			}
 			list_destroy(ret_list);

src/common/slurm_protocol_api.h

@@ -277,8 +277,9 @@ int inline slurm_shutdown_msg_engine(slurm_fd open_fd);
  *    (msg->data) is allocated from within this function, and must be
  *    freed at some point using one of the slurm_free* functions.
  *
- *  Returns SLURM_SUCCESS if an entire message is successfully 
- *    received. Otherwise SLURM_ERROR is returned.
+ *  Returns List containing the responses of the childern (if any) we 
+ *  forwarded the message to if an entire message is successfully 
+ *  received. Otherwise NULL is returned.
  */
 List slurm_receive_msg(slurm_fd fd, slurm_msg_t *resp, int timeout);
 

testsuite/slurm_unit/common/slurm_protocol_message_client.c

@@ -12,6 +12,7 @@ int main ( int argc , char * argv[] )
 	slurm_msg_t resp;
 	int16_t port = 0;
 	update_node_msg_t *in_msg, out_msg;
+	List ret_list = NULL;
 	
 	/* init address sturctures */
 	if (argc > 1)
@@ -32,9 +33,16 @@ int main ( int argc , char * argv[] )
 	slurm_send_node_msg( worker_socket , &msg ) ;
 
 	printf("Sending message=%s\n", out_msg.node_names);
-	if (slurm_receive_msg (worker_socket, &resp, 0) < 0) {
+	if ((ret_list = slurm_receive_msg(worker_socket, &resp, 0)) == NULL) {
 		printf("Error reading slurm_receive_msg %m\n");
 		exit(1);
+	} else {
+		if(list_count(ret_list)>0) {
+			error("We didn't do things correctly "
+			      "got %d responses didn't expect any",
+			      list_count(ret_list));
+		}
+		list_destroy(ret_list);
 	}
 	if (resp.msg_type != REQUEST_UPDATE_NODE) {
 		printf("Got wrong message type: %u\n", resp.msg_type);

testsuite/slurm_unit/common/slurm_protocol_message_server.c

@@ -13,6 +13,7 @@ int main ( int argc , char * argv[] )
 	slurm_msg_t resp;
 	int16_t port = 0;
 	update_node_msg_t *in_msg, out_msg;
+	List ret_list = NULL;
 		
 	/* init address sturctures */
 	if (argc > 1)
@@ -29,9 +30,17 @@ int main ( int argc , char * argv[] )
 	printf ( "worker socket %i\n", worker_socket ) ;
 
 	while (1) {
-		if (slurm_receive_msg (worker_socket, &msg, 0) == SLURM_SOCKET_ERROR ) {
-			printf ("slurm_receive_msg error\n");
+		if ((ret_list = slurm_receive_msg(worker_socket, &resp, 0)) 
+		    == NULL) {
+			printf ("Error reading slurm_receive_msg %m\n");
 			break;
+		} else {
+			if(list_count(ret_list)>0) {
+				error("We didn't do things correctly "
+				      "got %d responses didn't expect any",
+				      list_count(ret_list));
+			}
+			list_destroy(ret_list);
 		}
 
 		if (msg.msg_type == REQUEST_SHUTDOWN_IMMEDIATE) {