Crash in WebSocketTransport due to data races

[tahirmt]

Summary

In WebSocketTransport the subscribers dictionary is not Atomic and is modified using the processingQueue however it is read in processMessage(text:) function which is called on a different thread by the WebSocketClient. This can lead to a data race if the updates are coming in on the web socket very fast and at the same time new subscriptions are started or stopped.

By adding @Atomic to the subscribers variable, this crash is completely resolved.

While I didn't encounter a crash on subscriptions variable, I suspect the same can happen there.

Version

1.16.1

Steps to reproduce the behavior

This is tough because for us it only happened during a really high load scenario where we were receiving tens of messages per second on the web socket and a lot of those messages also resulted in our code stopping and starting subscriptions.

We have seen this crash happen on production as well. The crash shows up as EXC_BAD_ACCESS KERN_INVALID_ADDRESS

Logs

Anything else?

No response

[github-actions]

Do you have any feedback for the maintainers? Please tell us by taking a one-minute survey. Your responses will help us understand Apollo iOS usage and allow us to serve you better.