Backend deadlock in normal operation

[stuartm]

Found the backend had hung during the night. Backtrace attached. Only shows two threads which is odd and may in fact make the BT useless. I was unable to step through the first thread, it was just stuck at that point.

[stuartm]

Backtrace obtained by attaching to pid of hung process

[anonymous]

Have you tried using pstack on the backend when it is hung?

[danielk]

The first thread in the backtrace is the one that appears hung, the second thread is waiting on the first to finish.

I'm doubtful there enough info to fix it for someone familiar with the UPnP code, so I'm changing the milestone to unknown. dblain or greg can close it if this is not fixable with this info.

[Shane Shrybman]

That trace looks similar, if not identical, to a trace I get from the 2nd frontend started on a single system.

Steps to reproduce: start one frontend and let it idle, start another frontend and it often hangs on startup, then attach gdb to the second pid.

[beirdo]

If this reoccurs, please open a new ticket. With all the changes in the past 13 months, I would doubt we can even recreate the precise symptoms.