@@ -231,6 +231,30 @@ tf_call_channel_dispose (GObject *object)
...
+      while (g_hash_table_iter_next (&iter, &key, &value))
+        {
+          TfCallContent *content = TF_CALL_CONTENT (value);
+          if (content)
+            tf_call_content_on_call_channel_gone (content);

Why do you have a if (content) here? In which case could the value be NULL ?

The other patches are committed

Reflex safety I guess, I do not have a particular reason to believe it could be NULL.

Branch rebased, amended to remove that check, and repushed.

Now that I think more about it, maybe it shoudl just call g_object_run_dispose(content); and make sure every entry point and async callback in the CallContent object checks if the object is still valid.

Branch updated with a version that forces dispose on the content, and adds checks to avoid the last ref holder crashing.
I don't like this much tbh.

Note that for added fun, calling g_simple_async_result_complete in got_content_media_properties will cause that dispose to be called, so the content becomes invalid in the middle of that function. This is guarded for in process_codec_offer now.

*** Bug 34816 has been marked as a duplicate of this bug. ***

merged