On 04/04/2012 07:20 AM, Tanu Kaskinen wrote: > On Tue, 2012-04-03 at 19:03 +0200, David Henningsson wrote: >> Make sure we can't be called into by remaining references to >> sink-inputs and source-outputs after we have teared down, as >> that will likely lead to segfaults or assertion failures. >> >> Signed-off-by: David Henningsson<david.henningsson at canonical.com> >> --- >> >> This is a continuation of the "loopback: Fix crash if max request change at >> unload" thread. Not sure if this makes sense, it probably does, but I have not >> verified how the core will react when all of these are NULL, so won't push without >> additional comments/acks. >> >> Frederic, it would be nice to know if this improves your situation as well. > > pa_sink_input_unlink() already resets the callbacks. Ok, that's good. However, looking at it a bit closer, the most important one is not reset in the reset_callbacks: u->sink_input->parent.process_msg = NULL; or, perhaps even better: u->sink_input->parent.process_msg = pa_sink_input_process_msg; Perhaps it would make sense to put the above into teardown()? > Did my suggestion > of a one-line fix get any consideration? > http://lists.freedesktop.org/archives/pulseaudio-discuss/2012-March/013091.html Hmm, I don't understand how your suggestion could make a difference - at least now with the patch I pushed, adjust_time is set to zero at the same time as the sink is unlinked. (Sorry for misspelling your email, btw) -- David Henningsson, Canonical Ltd. http://launchpad.net/~diwic
