571981 - "ASSERTION: Got the wrong answer from the custom method call" with frameset root

Status: RESOLVED FIXED

(Core :: XPConnect, defect)

Description

[Jesse Ruderman]

Attachment: testcase

###!!! ASSERTION: Got the wrong answer from the custom method call!: 'NS_SUCCEEDED(debug_rv) == NS_SUCCEEDED(rv) && xpc_qsSameResult(debug_result, result)', file dom_quickstubs.cpp, line 13553

(gdb) p debug_rv
$1 = 2147500034
(gdb) p rv
$2 = 0

Comment 1

[Jesse Ruderman]

Attachment: stack trace

Comment 2

[Boris Zbarsky [:bzbarsky]]

debug_rv there is NS_ERROR_NO_INTERFACE.

And the reason is that nsHTMLDocument::GetBody will happily return the first element with tagname "frameset" it finds, no matter what the namespace, in IsHTML() documents.  However the COM getter will fail to QI this to nsIDOMHTMLElement, obviously.

The testcase should be ending up with a null body, but isn't, so this is a correctness bug.  Presumably regression from bug 560462.

Can we just use the GetElementsByTagNameNS call there for IsHTML()?  Seems like we should be able to.

Comment 3

[Peter Van der Beken [:peterv]]

Attachment: v1

Comment 4

[Boris Zbarsky [:bzbarsky]]

Can we add a correctness test too (testing that document.body is null there)?

Comment 5

[Jesse Ruderman]

s/too/instead?

Comment 6

[Peter Van der Beken [:peterv]]

Attachment: v1

Comment 7

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 451958 [details] [diff] [review]
v1

>+  isnot(elem, null,
>+        "HTMLDocument.body should be an html element with nodeName 'BODY' " +
>+        "and a child of the documentElement");

That doesn't make sense; there's no doc.body in sight there.  Should there be a separate assert that is(elem, doc.body)?

r=bzbarsky with that.

Comment 8

[Peter Van der Beken [:peterv]]

Ugh, that what I wanted to assert, yes.

Comment 9

[Peter Van der Beken [:peterv]]

http://hg.mozilla.org/mozilla-central/rev/94fe0dd33355