Breakpoint exception when executing request AF5

Question

So I'm trying to check if the bearer is valid before before each request, but when I run the refresh token request Alamofire gets an exception on dispatch precondition: What am I doing wrong? is this even possible?

extension Session: SessionStateProvider {
    func request(for task: URLSessionTask) -> Request? {
        dispatchPrecondition(condition: .onQueue(rootQueue))

        return requestTaskMap[task]
    }

Main caller of the request where first the token is checked

private static func createRequest(urlConvertible:  URLRequestConvertible)-> Alamofire.DataRequest?{
        if let manager = sessionManager{
            return manager.request(urlConvertible)
///custom manager that checks the tokens
        }else if let manager = authenticatedManager{
            manager.verifyRequest(url: urlConvertible) { (result) in
                return result
            }

        }
        return nil
    }

    ///  This method make a request using the JSONAPI's Format - Only to the 'data' structure
    ///
    /// - Parameter urlConvertible: Codabel Onject to parse teh response
    /// - Returns: The Codable object if success or not
    static func requestJSONAPI<T: Codable> (_ urlConvertible: URLRequestConvertible) -> Observable<T> {
        initManager(urlConvertible: urlConvertible)

        //Create an RxSwift observable, which will be the one to call the request when subscribed to
        return Observable<T>.create { observer in
            let request = createRequest(urlConvertible: urlConvertible)?
                .debugLog()
                .validate()
                .responseCodableJSONAPI(keyPath: "data", completionHandler: { (response: AFDataResponse<T>) in
                    switch response.result {
                    case .success(let value):

                        let jsonDataString = String(data: response.data ?? Data(), encoding: String.Encoding.utf8)!
                        print("Post Request Params : \(jsonDataString)")
                        observer.onNext(value)

Here is my code where auth.refreshtokens gets the new tokens with an AF.request

    private func shouldRefreshToken() -> Bool {

        do {
            let jwt = try decode(jwt: accessToken)
            return jwt.expired
        } catch {

        }
        return false
    }

    func verifyRequest(url: URLRequestConvertible, completion: @escaping (_ result: DataRequest) -> Void){

        if  shouldRefreshToken() {
            let auth = interceptor
///makes an AF.request to get the new tokens - where it crashes
            auth.refreshTokens { [weak self] succeeded, accessToken, refreshToken in
                self?.accessToken = accessToken!
                completion((self?.sessionManager?.request(url))!)
            }
        }else{
            completion((self.sessionManager?.request(url))!)
        }

    }
´´´

Answer 1

I know this is a couple years old, but on the latest versions of Alamofire, I found this is entirely reproducible for my own particular use case. I am restarting the session manager (user signed out, so any tasks in flight are no longer valid or desired, since since there's no token, we could get numerous "unauthorized" from those tasks).

My original code:

public func restartSessionManager() {
    self.manager.session.invalidateAndCancel()
    self.startSessionManager()
}

This always results in a crash with a large stack trace that is caused by a precondition in Session.swift in cancelRequestsForSessionInvalidation(with:):

dispatchPrecondition(condition: .onQueue(rootQueue))

I've tried working around this a number of ways, such as manually invalidating all requests and closing. Same crash, same place.

I could modify Session.swift to remove the precondition, and the world doesn't end. However, the next pod update with a new release can overwrite this.

Only work-around (within my own code) is to replace my original code with:

public func restartSessionManager() {
    manager.rootQueue.async {
        self.manager.session.invalidateAndCancel()
        DispatchQueue.main.async {
            self.startSessionManager()
        }
    }
}

The requeue on .main is just a safety-belt issue. Probably not necessary, but it makes me feel better to do it that way.

Answer 2

This is probably an Alamofire bug. Could you create a new issue on our GitHub with the full stack trace from the crash and the snippets of code you posted here or a sample project?

Also, much of your async handling won't work correctly, like your attempt to return result inside verifyRequest, but I'm not sure if that has anything to do with this issue.

Additionally, I would suggest refactoring your token handling into a RequestInterceptor with an adapter that adds the token and a retrier that will refresh the token when necessary to retry the request. There's no need to do the checking manually, Alamofire supports this sort of thing automatically.