658845 - "ABORT: Reset() expects non-null URI for referenced elem"

Status: RESOLVED FIXED

Description

[Jesse Ruderman]

Created attachment 534272 [details]
testcase (asserts fatally when loaded)

###!!! ABORT: Reset() expects non-null URI for referenced elem: 'aURI', file content/base/src/nsReferencedElement.cpp, line 55

This abort was added in http://hg.mozilla.org/mozilla-central/rev/655514007ebd (bug 308590 patch 5).

Comment 1

[Jesse Ruderman]

Created attachment 534273 [details]
stack trace

Comment 2

[Daniel Holbert [:dholbert]]

The assert may very well be bogus.

Comment 3

[Daniel Holbert [:dholbert]]

Created attachment 534278 [details] [diff] [review]
fix

(Confirmed in debug m-c build.)

Yeah, I think the assert is bogus.

Previously (before bug 308590), nsReferencedElement::Reset() began like so:
> 81   Unlink();
> 82 
> 83   nsCOMPtr<nsIURL> url = do_QueryInterface(aURI);
> 84   if (!url)
> 85     return;

I mistakenly assumed that line 84 there was simply a check for "Is aURI a nsIURL?", but turns out it was *also* serving the purpose of checking "Is aURI non-null".

Attached patch removes the bogus ABORT_IF_FALSE and restores a null-check early-return.  Also includes this bug's patch as a crash-test.

Comment 4

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Comment on attachment 534278 [details] [diff] [review]
fix

Review of attachment 534278 [details] [diff] [review]:
-----------------------------------------------------------------

Comment 5

[Daniel Holbert [:dholbert]]

http://hg.mozilla.org/projects/cedar/rev/96fae135f592

Comment 6

[Matt Brubeck (:mbrubeck)]

http://hg.mozilla.org/mozilla-central/rev/96fae135f592

Comment 7

[Daniel Holbert [:dholbert]]

Comment on attachment 534278 [details] [diff] [review]
fix

Requesting approval to land on aurora, as a followup to bug 308590.  

This is a trivial patch to fix a null-pointer crash. Includes crashtest.

Comment 8

[Daniel Holbert [:dholbert]]

(In reply to comment #7)
> This is a trivial patch to fix a null-pointer crash. Includes crashtest.

(a _new_ null-pointer crash, that is)

Comment 9

[Daniel Holbert [:dholbert]]

Gah, sorry -- I'm failing today at remembering whether things landed before or after the last aurora merge.  (I thought the merge was longer ago than it actually was.)

The checkin in comment 6 beat the merge, so this is already in aurora.