We seem to not consistently have the getInterface on the global this. This causes issues with some test262, that compare the result of two consecutive for ... in over the global object. Testcase: http://jsfiddle.net/sKYa6/1/ For some reason this doesn't always reproduce for me. My own build shows the problem, but the nightly I am using, does not.

I say we just add nsIInterfaceRequestor to window classinfo for now...

What causes it to appear sometimes? The Firefox UI QIing it? Does that propagate through xrays?

> What causes it to appear sometimes? The Firefox UI QIing it? Afaik, yes. > Does that propagate through xrays? Again, afaik yes.

I'll be glad when that goes away for DOM objects ;-)

I assume this is going to be fixed by webidl DOMWindow? Is that going to happen soon?

WebIDL window will get us "never", yes, for web content. But it's not going to happen until a month or more from now at best; no one is actively workin on it. For this bug, we could just add nsIInterfaceRequestor to the classinfo for window, I guess...

This seems to be breaking actual web content (see bug 965790). I think we should just add this to classinfo for the moment.

Comment on attachment 8368214 [details] [diff] [review] Just always define getInterface on the window for now, so we stop breaking people. It'll go away once we ship WebIDL window. [Approval Request Comment] Bug caused by (feature/regressing bug #): Been with us for a while, but has recently started breaking sites. User impact if declined: Some websites (e.g. Udacity) are broken in Firefox. Testing completed (on m-c, etc.): Passes our tests. Risk to taking this patch (and alternatives if risky): Risk should be low. All this patch does is make the property always visible to the web page, instead of being either visible or not depending on what UI code happens to have run. String or IDL/UUID changes made by this patch: None.

https://hg.mozilla.org/releases/mozilla-aurora/rev/be285c8df31a https://hg.mozilla.org/releases/mozilla-beta/rev/0450f3309a58

Cannot document this if we don't know when it started to happen.

I don't think this needs developer documentation. It "started" to happen probably over a decade ago.

Got it.

I re-enabled the test262 that this bug or WebIDL window should have fixed: https://hg.mozilla.org/integration/mozilla-inbound/rev/a67d187584ed