Crash in AWSIoTMQTTClient.m line 635 -[AWSIoTMQTTClient openStreams:]

gadget-man commented 1 year ago

Describe the bug I'm seeing a lot of crash reports in Crashlytics which all point to the same line: AWSIoTMQTTClient.m line 635 -[AWSIoTMQTTClient openStreams:].

We are currently using AWS v2.27.13

Steps to reproduce the behavior: Very difficult to reproduce - only seen intermittently on device when resuming from background.

Environment SDK Version: [2.27.13] Dependency Manager: Cocoapods

Device Information

Device: iPhone 14 Pro Max, iPhone 12 Pro
iOS Version: iOS 16.1.1

Stack Trace Crashed: Thread #1 EXC_BAD_ACCESS KERN_INVALID_ADDRESS 0x000000000000012e

Crashed: Thread 0 libobjc.A.dylib 0x6cf0 objc_opt_respondsToSelector + 52 1 CoreFoundation 0x13b14c _inputStreamCallbackFunc + 48 2 CoreFoundation 0x105fb4 _signalEventSync + 216 3 CoreFoundation 0x117e84 _cfstream_solo_signalEventSync + 224 4 CoreFoundation 0x10e42c _CFStreamSignalEvent + 304 5 CFNetwork 0x147eb8 _CFNetworkErrorGetLocalizedDescription + 180236 6 CFNetwork 0x1533e4 _CFNetworkErrorGetLocalizedDescription + 226616 7 CoreFoundation 0x11a41c CFSocketPerformV0 + 648 8 CoreFoundation 0xd5f54 CFRUNLOOP_IS_CALLING_OUT_TO_A_SOURCE0_PERFORM_FUNCTION + 28 9 CoreFoundation 0xe232c CFRunLoopDoSource0 + 176 10 CoreFoundation 0x66210 CFRunLoopDoSources0 + 244 11 CoreFoundation 0x7bba8 CFRunLoopRun + 836 12 CoreFoundation 0x80ed4 CFRunLoopRunSpecific + 612 13 Foundation 0x42334 -[NSRunLoop(NSRunLoop) runMode:beforeDate:] + 212 14 AWSIoT 0x5a728 -[AWSIoTMQTTClient openStreams:] + 759 (AWSIoTMQTTClient.m:759) 15 Foundation 0x5b808 NSThreadstart__ + 716 16 libsystem_pthread.dylib 0x16cc _pthread_start + 148 17 libsystem_pthread.dylib 0xba4 thread_start + 8

I see a very similar issue was previously report at https://github.com/aws-amplify/aws-sdk-ios/issues/1209 however this was reported resolved in a prior build - we are still seeing it in 2.27.13.

lostyoung commented 1 year ago

We have also met this issue in SDK with version 2.27.7 and 2.28.5.

Banner2404 commented 1 year ago

The same issue in SDK 2.29.1

Banner2404 commented 1 year ago

I was able to reliably reproduce this issue (SDK 2.30.1) with the following code:

    func connect(_ iteration: Int) {
        print(iteration, "Connect")
        dataManager.connect(withClientId: UUID().uuidString, cleanSession: true, certificateId: "certificate") { status in
            print(iteration, "Connection callback", status.string)
        }
        DispatchQueue.main.asyncAfter(deadline: .now() + 0.2) {
            print(iteration, "Disconnect")
            self.dataManager.disconnect()
            print(iteration, "Reconnect")
            self.connect(iteration + 1)
        }
    }

Basically here I start a connection to MQTT broker, and after a small delay I terminate the connection and start another connection. In theory this can happen when user doesn't wait for connection to be established and switches to another IoT device.

The crash always occurs in background thread, however the stack trace may differ. Seems like it crashes in different parts of the AWSIoTMQTTClient code. The crash can also occur not immediately but only after 1, 2, 3... reconnection loops.