Hi, I am trying to get the latest snapshot of qutecom uploaded to Debian experimental. However I am facing the issue that the qutecom process never terminates.
I tracked the problem to its cause, and decided to write to the ML as it related to multiple trac tickets / patches. The problem is that qutecom schedules the invocation of purple_core_quit() on the default glib loop [1]. However, the fix for [2] merges the default glib with the Qt loop. However the GUI is destroyed very early during the exit process; along with that, the main glib loop is terminated before libpurple has the chance to be deinitialized. As a consequence of that, the thread that is waiting fro libpurple deadlocks. The best solution IMHO would be running the model on a glib loop (seperate from Qt!) and having libpurple use that same loop. That would also help for other reasons I mentioned in [1]. To that ticket I attached two pacthes to implement this, but they still have some issues I was not able to debug. Alternative solutions? Making sure the Qt loop waits for the model class that is waiting for libpurple? By the way, do you have more information about [2]? Do you have a pointer to the changeset that fixed that bug? What was a separate glib loop causing that crash? [1] http://trac.qutecom.org/ticket/175 [2] http://trac.qutecom.org/ticket/173 Thanks! Ludovico _______________________________________________ QuteCom-dev mailing list QuteCom-dev@lists.qutecom.org http://lists.qutecom.org/mailman/listinfo/qutecom-dev
