SignalR crashes when app is moved to background and hub connection is closed

[nysander]

I have such reconnect policy:

class RTReconnectPolicy: ReconnectPolicy {
    private var connect: () -> Void
    private var retryTimeout = 1

    public init(connect: @escaping () -> Void) {
        self.connect = connect
    }

    func nextAttemptInterval(retryContext: RetryContext) -> DispatchTimeInterval {
        if let error = retryContext.error as? SignalRError, error.statusCode == 401 {
            Logger.shared.log(error.localizedDescription, logLevel: .error)
            connect()
            return DispatchTimeInterval.seconds(30)
        }

        Logger.shared.log(retryContext.error.localizedDescription, logLevel: .error)

        return DispatchTimeInterval.seconds(calculateTimeout(failedAttemptsCount: retryContext.failedAttemptsCount))
    }

    func calculateTimeout(failedAttemptsCount: Int) -> Int {
        if retryTimeout < 32 {
            retryTimeout = failedAttemptsCount * 2
            return retryTimeout
        } else {
            retryTimeout = 32
            return retryTimeout
        }
    }
}

and my connection method

        refreshTokenUseCase.execute()
            .sink(receiveValue: { accessToken in
                if !accessToken.isEmpty {
                    self.connecting = true

                    self.connection = HubConnectionBuilder(url: url)
                        .withLogging(minLogLevel: rtLogLevel, logger: Logger.shared)
                        .withJSONHubProtocol()
                        .withAutoReconnect(reconnectPolicy: RTReconnectPolicy(connect: { self.connect() }))
                        .withPermittedTransportTypes(.webSockets)
                        .withHttpConnectionOptions(configureHttpOptions: { options in
                            // to authorize RealtimeBroker connections with Access Token `skipNegotiation` parameter has to be set to `false`
                            options.accessTokenProvider = { accessToken }
                            options.skipNegotiation = false
                        })
                        .build()
                    self.connection?.delegate = self

                    self.connection?.start()
                }
            })
            .store(in: &cancellables)

disconnect method

    public func disconnect() {
        Logger.shared.log("Disconnecting from Realtime server with address: \(url?.absoluteString ?? "--")", logLevel: .info, file: "\(#file)", function: "\(#function)")
        connection?.stop()
    }

App uses access token for securing hub connection (before I used access token this error did not happened (and then I used default .autoReconnect() method but when using it with token it caused to many reconnections failing.

Scenario:

App is in foreground it has AT (access token) to open web socket, then I push it to background or lock a device. Connection is then closed with .disconnect() method. When I unlock device I have report that app had crashed. and in crash log I see what's on image below.

I log also SignalR Error - 8

[moozzyk]

Thanks for reporting. It seems it comes from recently added keepAlive. I merged a change yesterday that should allow disabling keepAlive by setting the interval to nil. You may want to try this until I figure out what is happening.

[moozzyk]

Any chance to include logs?

[nysander]

(edited)

I will check if disabling it will help and I will try to prepare some anonimized logs for you.

[moozzyk]

this crash is from version 0.8.0

This does not seem right. Support for keep alive was added less than 3 weeks ago in 5ddfab6 and was not released yet. It seems like the changes are from master.

[nysander]

yes you are right, I thougt I use tagged version not the master. edited above

[nysander]

thank you for quick fix, I will write you back next week if this resolved a crash as it is not something I can trigger on demand. I have also set keepAliveInterval to nil

[moozzyk]

If it doesn’t fix the crash you can take a commit before the one I pointed to.

…

On Thu, Oct 21, 2021 at 7:29 AM Paweł Madej ***@***.***> wrote: thank you for quick fix, I will write you back next week if this resolved a crash as it is not something I can trigger on demand. I have also set keepAliveInterval to nil — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#208 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAK7JJARPQJNSYFT3XD7PU3UIAPWHANCNFSM5GNNSG5Q> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.

[nysander]

My first tries and it looks that app is not killing itself in backround, so that makes hope ;)

I am now releasing new testflight version so I will have more users to try it out.

[nysander]

I think the issue is fixed now. I didn't get any new crash report from Signal R until today. Thanks again for quick fix

[moozzyk]

Thanks for confirming!

[moozzyk]

I am reopening because the root cause has not been addressed.

[KateXe]

We have the same issue, and my colleague found the bug, there is a double semaphore.wait, the one in the closure of resetKeepAlive, and the one in cleanupKeepAlive. So it comes to a freeze/crash. I try to fix it.

[cosmer-work]

This also frequently happens when HttpConnection.stop(stopError:) is called.

[moozzyk]

@cosmer-work seems like a different issue. Can you create a new issue and provide more details including logs at the debug level?

[moozzyk]

This should now be fixed with 48942fb and 4f3a9b2