lldplay appears to be doing calls to the message callback method *after* lldplay_destroy has returned

[jackjansen]

Specifically, I'm getting Output: metadata not equal but comparable by value from someone/something that is apparently still alive.

I am only seeing this when I have both an lldplay and an lldpkg at the same time, and the lldpkg is destroyed after the lldplay`. Unsure if this is a coincidence or not.

But anyway: it's dangerous.Not only may I discard the Python/C# callback methods after lldplay_destroy() has returned, but actually I think the handle may be freed, so the pointer to the callback routine may have become invalid.

I think lldplay_destroy() should not return until it is sure there are no references to the callback still alive in some thread or other.

[rbouqueau]

The code seems to guarantee that objects are deallocated. The Signals "Pipeline" (master object) and the Logger from the llplay wrapper are both destroyed UNLESS the destroy method failed (which is unlikely + in which case a message is emitted but that's all since function returns void).

My favorite hypothesis is that we keep pushing packets to a deallocated object and it works by chance.