Clock skew not being handled by didCompleteWithError method

[ghost]

Describe the bug When making a request on a device with wrong time, the response comes with the following error Signature not yet current: 20130909T170846Z is still later than 20130909T170823Z (20130909T170323Z + 5 min.).

To Reproduce Steps to reproduce the behavior:

1. Put your phone clock back 20 minutes
2. Try to make a request to an AWSLambda instance

Which AWS service(s) are affected? • AWSLambda

Expected behavior When making a request on a device with wrong time, the SDK should attempt a new request with the clock skew fixed.

Screenshots Not applicable.

Environment(please complete the following information):

• SDK Version: 2.6.32
• Dependency Manager: Cocoapods
• Swift Version : Not applicable (Objective-C)

Device Information (please complete the following information):

• Device: All devices
• iOS Version: iOS 12.0

Additional context When an error related to clock skew returns from the request it should call the method - (void)URLSession:(NSURLSession *)session task:(NSURLSessionTask *)sessionTask didCompleteWithError:(NSError *)error in order to retry the task with clock skew fixed (AWSNetworkingRetryTypeShouldCorrectClockSkewAndRetry).

[mutablealligator]

@ghearly Sorry for the inconvenience. Can you post a code snippet that reproduces the issue?

[ghost]

@kvasukib This is the AWSAPIGatewayRequest we are using: AWSAPIGatewayRequest *request = [[AWSAPIGatewayRequest alloc] initWithHTTPMethod:@"POST" URLString:@"/users" queryParameters:queryParameters headerParameters:_headersPOST HTTPBody:HTTPBody];

We call it like this [[[APIClient defaultClient] usersPostWithHTTPBody:HTTPBody] continueWithBlock:^id _Nullable(AWSTask * _Nonnull task) { BLOCK GOES HERE }];.

The APIClient class is generated by AWS, it works fine when the client has the correct hour and date set.

[palpatim]

The APIGatewayClient and its API-specific subclasses use custom network handling rather than the core stack, which means that even defining a retry handler and injecting it into the API client's configuration wouldn't be respected.

The request returned from the API Gateway includes enough information that a caller could potentially inspect and retry manually, but that's not really a good solution--the SDK should do this on the client's behalf.

Because the APIGateway network stack is custom, the behavior isn't technically a bug so much as a feature gap. Therefore, I'm characterizing this as a feature request to this capability to bring this SDK into parity with other SDKs. At implementation time, we'll need to decide whether to expand the client with retry-specific custom code in both invoke methods of the API Gateway client, or migrate the APIGatewayClient networking code to use AWSCore networking.

[svsharsha]

We are also facing this issue. In IOS we implemented direct lambda invocation, when user changes iPads time manually got below exception "{"message":"Signature not yet current: 20190604T152329Z is still later than 20190604T090443Z (20190604T085943Z + 5 min.)"}" AND "Error Domain=com.amazonaws.AWSLambdaErrorDomain Code=0 "InvalidSignatureException" UserInfo={StatusCode=403, responseStatusCode=403, message=Signature not yet }"

And there is no proper solution for Clock skew issue in IOS. Please let us know the solution for this.

[atierian]

Thanks for opening this feature request. We’re closing this due to the lack of community engagement on the request.