EXC_BAD_ACCESS crash in PersistentStorage.swift

[mt00chikin]

EXC_BAD_ACCESS crash in PersistentStorage.swift

[crleona]

Thanks @mt00chikin - I believe this is fixed by #159 which should release later this week.

[crleona]

Hi @mt00chikin, The stack trace points to a crash calling dispatch.sync on one of our internal queues, which probably indicates a deadlock. Could you share more information about how you are calling ApplicationAnalyticsService.identifyIfNeeded (in terms of dispatch), as shown in the background thread in the stack trace?

[mt00chikin]

Sure thing. That service manages multiple different analytics destinations, Amplitude being one of several, and on any analytics event publish(_:) call it first checks several user properties to ensure that the user's context hasn't changed since the last event was fired. If it has, it then calls identify on each downstream service. In our case, it calls Amplitude's public func identify(identify: Identify, options: EventOptions? = nil) -> Amplitude function. Once that's done, it proceeds to publish the event using amplitude.track().

This is all happening on an internal serial DispatchQueue.

// ApplicationAnalyticsService.swift 

func publish(_ event: SubwayEvent) {
    eventQueue.async { [weak self] in
        guard let self else { return }
        ...
        publish(event)
    }
} 

private func publish(_ event: AnalyticEvent) {
    identifyIfNeeded()
    // Publish on each downstream service
    services.forEach { $0.publish(event) }
}

private func identifyIfNeeded() {
    guard let userIdentity = contextCache?.currentUserIdentity() else { return }

    // If identity did not change there's no need to re-identify the user.
    guard userIdentity != currentUserIdentity else { return }
    services.forEach { $0.identify(userIdentity) }
    currentUserIdentity = userIdentity
}

[crleona]

This should be fixed in 1.5.0.