In jquery.signalR-1.1.3.js the longpolling error function passes back the response text instead of the entire jqXHR object.
Is there a good way to determine in the hub error handler what the error actually is?
$.connection.hub.error(function (error) {
system.log('SignalR error: ' + error);
});
I've used the SignalR [Authorize] attribute to return a HTTP 403 Forbidden when a connection should be refused, but in my client all I get is the IIS error page response page in the error object because SignalR doesn't pass back the entire response object see below: $(instance).triggerHandler(events.onError, [data.responseText]);. This happens in a tight loop because SignalR just keeps trying to reconnect a few hundred/thousand/whatever times until the reconnect timeout passes.
I'd like to be able to tell that it's a 403, or whatever custom status code I might return, and then let my app respond accordingly...in this case, quit trying to reconnect and log out.
Here's the longpolling transport's error handler:
error: function (data, textStatus) {
// Stop trying to trigger reconnect, connection is in an error state
// If we're not in the reconnect state this will noop
window.clearTimeout(reconnectTimeoutId);
reconnectTimeoutId = null;
if (textStatus === "abort") {
connection.log("Aborted xhr requst.");
return;
}
// Increment our reconnect errors, we assume all errors to be reconnect errors
// In the case that it's our first error this will cause Reconnect to be fired
// after 1 second due to reconnectErrors being = 1.
reconnectErrors++;
if (connection.state !== signalR.connectionState.reconnecting) {
connection.log("An error occurred using longPolling. Status = " + textStatus + ". " + data.responseText);
$(instance).triggerHandler(events.onError, [data.responseText]);
}
// Transition into the reconnecting state
transportLogic.ensureReconnectingState(instance);
// If we've errored out we need to verify that the server is still there, so re-start initialization process
// This will ping the server until it successfully gets a response.
that.init(instance, function () {
// Call poll with the raiseReconnect flag as true
poll(instance, true);
});
}
And the Server side hub authorization code
public override bool AuthorizeHubConnection(Microsoft.AspNet.SignalR.Hubs.HubDescriptor hubDescriptor, Microsoft.AspNet.SignalR.IRequest request)
{
// If we want them to connect, return true, else false
// Asp.NET will return a 403 if we return false here
return false;
}
A couple of questions arise...
- Why does signalR reconnect in a tight loop like that?
- Is there a better way to control the signalR connection lifecycle based upon the Authorization of the hub connection?