Fix: sessiond: assert on empty payload when handling client out event
[lttng-tools.git] / src / bin / lttng-sessiond / action-executor.cpp
index f43716c63986673377890c8a635f25c149a4f7d7..0b910d474ca3c01c0ee01c245d0c4c25c52b8a85 100644 (file)
@@ -233,6 +233,15 @@ static int client_handle_transmission_status(
        case CLIENT_TRANSMISSION_STATUS_COMPLETE:
                DBG("Successfully sent full notification to client, client_id = %" PRIu64,
                                client->id);
+               /*
+                * There is no need to wake the (e)poll thread. If it was waiting for
+                * "out" events on the client's socket, it will see that no payload
+                * in queued and will unsubscribe from that event.
+                *
+                * In the other cases, we have to wake the the (e)poll thread to either
+                * handle the error on the client or to get it to monitor the client "out"
+                * events.
+                */
                update_communication = false;
                break;
        case CLIENT_TRANSMISSION_STATUS_QUEUED:
@@ -821,7 +830,7 @@ static void clean_up_action_executor_thread(void *_data)
 struct action_executor *action_executor_create(
                struct notification_thread_handle *handle)
 {
-       struct action_executor *executor = (action_executor *) zmalloc(sizeof(*executor));
+       struct action_executor *executor = zmalloc<action_executor>();
 
        if (!executor) {
                goto end;
@@ -893,7 +902,7 @@ enum action_executor_status action_executor_enqueue_trigger(
                goto error_unlock;
        }
 
-       work_item = (action_work_item *) zmalloc(sizeof(*work_item));
+       work_item = zmalloc<action_work_item>();
        if (!work_item) {
                PERROR("Failed to allocate action executor work item: trigger name = `%s`",
                                get_trigger_name(trigger));
This page took 0.025032 seconds and 4 git commands to generate.