Similar to bug 670687. See https://bugzilla.mozilla.org/show_bug.cgi?id=664179#c61

Why is that getting the origin from the URI and not from the principal?

Comment on attachment 547568 [details] [diff] [review] v1 r=me if you address bz' comment.

Comment on attachment 547568 [details] [diff] [review] v1 Hmm.. mUTF16Origin is a somewhat misleading name. GetASCIIOrigin returns the puny-code encoded origin. I would have expected mUTF16Origin to contain a unicode encoded origin based on the explicit 'utf16' in the name. I'd also have guessed that the event should contain the unicode variant. Is there a reason you explicitly ask for the punycode coded variant?

Sicking, did you look at bug 67068?

Comment on attachment 547568 [details] [diff] [review] v1 Actually, no, this is not right Message events don't need ascii representation of the origin.

"the origin attribute must be set to the Unicode serialization of the origin of the event stream's URL"

But we may need to do something to the origin, see https://bugzilla.mozilla.org/show_bug.cgi?id=670687#c5

nsContentUtils::GetUTFOrigin should do the right thing. (GetXXXOrigin really should not be in nsContentUtils)

> Why is that getting the origin from the URI and not from the principal? Because "the origin attribute must be set to the Unicode serialization of the origin of the event stream's URL". I don't know if I have misunderstood that, but I think per spec it is correct. WebSockets uses the origin from the principal, which is get from the entry script. It seems strange to me, but the specs define these behaviors... > nsContentUtils::GetUTFOrigin should do the right thing. ok

> mUTF16Origin is a somewhat misleading name. So I will let mOrigin

Comment on attachment 549658 [details] [diff] [review] v2 Could we get some tests for this.

> Could we get some tests for this. In bug 664179 we have already some tests.

updated for bug 664179

Comment on attachment 570446 [details] [diff] [review] v3 You don't actually need to re-request review when just updating to tip.

> You don't actually need to re-request review when just updating to tip. Ok, sorry. I didn't know that.

Checked in to inbound: https://hg.mozilla.org/integration/mozilla-inbound/rev/3ab7b95bfcb5 Thanks for the quick turnaround and for the patch in general!! Please do poke someone if you see bugs lingering like this, that should never happen.

Backed out in https://hg.mozilla.org/mozilla-central/rev/abdbf0646a21 - see bug 664179 comment 85

Fixed on bug 664179. See https://bugzilla.mozilla.org/show_bug.cgi?id=664179#c97