Comment 3 for bug 1406124

The backtrace seems to indicate that when the VU meter CO was set by the engine, the CDP then rebroadcasted the valueChanged signal to a COSlave that was concurrently being deleted by the main thread. That's consistent with changing the skin since the VU meter CO slave was just deleted. I suspect if Daniel had posted a backtrace of all threads we would see that the main thread was in the middle of ~WVuMeter.

(I'm not sure what smart pointers would tell you here? In this particular case, the control is held by ControlWidgetConnection and it's a ControlObjectSlave wrapped in a QScopedPointer)

So, shoot. That's a legit race condition.

Using an AutoConnection for CDP <-> COSlave signals would solve the issue for COSlaves that are deleted from the main thread. But if a CO slave's life cycle is not controlled by the main thread and that thread does not have a Qt event loop then we're kind of stuck.

One option is to use QObject::deleteLater for COSlaves. But if a COSlave is not created in the main thread then it won't be deleted until that thread ends. We could moveToThread(mainThread) and then call deleteLater? Thoughts?