lost connection inconstencies?

[nathanday]

Trying to handle a loss of network connection in our app I have found inconsistencies when trying to reconnect. The two ways I test loss of network are, using Apples "Network Link Conditioner" with 100% loss and pulling out the Ethernet cable. When my error delegate method (websocketDidDisconnect:error:) is called I just wait a small amount of time and call connect again on the same JFRWebSocket instance, I also call writePing: periodically when I have a connection to test its still connected. If the connection is broken by Network Link Conditioner, my error delegate method is called every time I try to reconnect until the network is restored and I can connect. But if I pull out the Ethernet cable, my error delegate method is called when the connection is broken, but when I try reconnecting again my error delegate is never called so there is no way for me to know the re-connection attempt failed and I should wait a bit and try again. I guess the second behaviour based on the method name websocketDidDisconnect:error: make senses, but then there isn't a way to an attempt to connect failed. Is reusing the same JFRWebSocket after network loss OK or should I create a new instance all together? Maybe I should be calling disconnect on JFRWebSocket before I try reconnect? Maybe I have discovered a bug? Thanks

[nathanday]

I have had a chance to look at this myself, the problem is a race condition with the instances variable didDisconnect, if you try to connect and it fails immediately, the method -[JFRWebSocket doDisconnect:] is called before the block to set self.didDisconnect = NO in -[JFRWebSocket connect] is called. Is setting self.didDisconnect in a queue for key value observing? If so the testing of self.didDisconnecting should occur in the queue as well. I would provide the fix for this myself, but where I work I'm behind a very strict fire wall.