Josh Matthews [:jdm] Reporter
	Description • 14 years ago

This is extremely frustrating, as the callbacks fail silently.  I had |do_check_eq(this._variable, NONEXISTANT_CONSTANT)| which gave me no output.

	Josh Matthews [:jdm] Reporter
	Comment 1 • 14 years ago

Actually, I suppose the real problem is that ReferenceErrors are not detected or something, because when I changed NONEXISTANT_CONSTANT to an actual constant, the test fails just fine.

	Josh Matthews [:jdm] Reporter
	Comment 2 • 14 years ago

So there's a reason that the default ChannelListener callbacks are all wrapped in try/catch blocks. I propose we implement a ChannelForwarder that takes a listener and forwards all calls to it safely, so that we can write custom listeners without the try/catch annoyances.

	Josh Matthews [:jdm] Reporter
	Comment 4 • 14 years ago

Comment on attachment 477553 [details] [diff] [review]
Add ChannelForwarder to make use of custom channel listeners easier in network xpcshell tests.

Here's what I had in mind.

	Jason Duell
	Comment 5 • 14 years ago

Comment on attachment 477553 [details] [diff] [review]
Add ChannelForwarder to make use of custom channel listeners easier in network xpcshell tests.

"ChannelForwarder" is not a very descriptive name.  How about "ListenerExceptionReporter"?  Either way, couldn't hurt to add a comment like this:

// Exceptions from OnStart/Data/Stop are silently eaten by default: this class
// converts them to xpcshell errors.
 
>+function ChannelForwarder(listener) {
>+  this._listener = listener;
>+}

	Patrick McManus [:mcmanus]
	Comment 6 • 9 years ago

there is a r+ patch here.. should it be checked in? should the bug be otherwise closed?

	Josh Matthews [:jdm] Reporter
	Comment 7 • 9 years ago

The fact that this is an opt-in tool means that nobody will use it, in all likelihood. I don't think it's worth trying to pursue this solution.

	Jason Duell
	Comment 8 • 9 years ago

OK, then.