Closed maxbrunsfeld closed 1 month ago
Note that the stack trace on the livekit.multicast
thread is similar to the one in this other open issue: https://github.com/livekit/client-sdk-swift/issues/257. I decided to open a separate issue because the Zombie profiler (and multiple class implementations) is not relevant to my case.
Might be related to statistics timer, maybe some kind of retain cycle occurring... Will investigate, thanks for the report.
Hi @maxbrunsfeld! We are big fans of your counterfeiter lib.
My guess is that it's related to DispatchQueueTimer
which i patched already in v1,
or some issue when the TrackStatistics get swapped with the new value.
Do you read the Track.statistics
property at all ?
I've released v1.1.5, can you try it out ?
Do you read the Track.statistics property at all ?
We don't read that property.
We can try out v1.1.5 in a couple of weeks. For the moment, we have downgraded to 1.0.12, as that's the version we were using before upgrading to 1.1.4.
Hello, v2.0.8 has different lock mechanism now, I will close this. Please re-open if issue persists .
Describe the bug
When leaving a Room with several participants, our application is occasionally crashing. It looks like process is being killed with
SIGILL
. Whenever this happens thelivekit.multicast
thread always has a very deep call stack that looks like this:Does anyone know if there is ever some linked-list-like structure that would be deallocated on the
livekit.multicast
thread, such that many deeply nested destructors would be called?SDK Version
1.1.4
iOS/macOS Version Various macOS versions. One example occurred on macOS 14.1.2. Another occurred on macOS 13.x
Steps to Reproduce Unfortunately, I don't know how to reliably reproduce this. It happens when leaving a Room. I think it happens more often when leaving a Room with several participants, so there are several audio tracks.
Expected behavior No crash.
Screenshots N/A
Logs N/A