dispatcher_connection_new_requested_channel: fix a use-after-free crash

If an error occurs when requesting the channel, we call
dispatcher_request_failed which as the side effect of destroying the
DispatcherRequestData. When calling dispatcher_flush_outstanding_operations
we used to deference this pointer.
Fixes this crash by using the "self" pointer instead and reffing it to be sure
it stays valid.

diff --git a/libempathy/empathy-dispatcher.c b/libempathy/empathy-dispatcher.c
index 201e48409c1f03857cb92313f323fd057d84258e..70f97f97453f0f9a4486ffc9b7cf76e89f669663 100644 (file)
--- a/libempathy/empathy-dispatcher.c
+++ b/libempathy/empathy-dispatcher.c
@@ -1316,6 +1316,12 @@ dispatcher_connection_new_requested_channel (EmpathyDispatcher *self,
   EmpathyDispatchOperation *operation = NULL;
   ConnectionData *conn_data;
 
+  /* The DispatcherRequestData owns a ref on the self object. As the request
+   * data could be destroyed (when calling dispatcher_request_failed for
+   * example) we keep a ref on self to be sure it stays alive while we are
+   * executing this function. */
+  g_object_ref (self);
+
   conn_data = g_hash_table_lookup (priv->connections,
     request_data->connection);
 
@@ -1398,8 +1404,8 @@ dispatcher_connection_new_requested_channel (EmpathyDispatcher *self,
   g_object_unref (operation);
 
 out:
-  dispatcher_flush_outstanding_operations (request_data->dispatcher,
-    conn_data);
+  dispatcher_flush_outstanding_operations (self, conn_data);
+  g_object_unref (self);
 }
 
 static void