Mythbackend resched segfault with UPnP code

[otto at kolsi dot fi]

About 40-50% of the "mythbackend --resched" executions end up with segfault in UPnP code. I've attached a backtrace. This is with FC5 dual-core machine.

[Ash Bostock <abostock@…>]

Would this cause the same thing that I'm seeing with mythwelcome, mythfrontend and also mythtv-setup? They segfault 40-50% of the time on first load, last terminal output is about deleting upnp. They'll either segfault straight away or if not will run perfectly fine afterwards, only occurs on initial loading.

[Nigel]

Yes, and the same as #4346. The more GDB traces we get, the more likely we will be able to find the problem.

[otto at kolsi dot fi]

Just tested with SVN rev 15284 to see if [15283] helped for this problem, but it does not. Attached is a new file: bactrace-v2.txt just in case more info is needed.

[otto at kolsi dot fi]

I've attached one more backtrace, this time it is from running simply "mythtv" without any arguments. It looks like a bit different this time..

[Anduin Withers]

(In [15286]) References #4383

Make m_addressList local to SendNotifyMsg?, this is the last stop before true avoidance with a switch to QWaitCondition and letting the task queue send the disable (would also prevent the double that can/will happen now).

[otto at kolsi dot fi]

This seems to be fixed in [15286] or some other lately change.