336994 - Crash when window gets destroyed on SVGZoom event

Status: RESOLVED FIXED

(Core :: SVG, defect)

Description

[Martijn Wargers (dead)]

See upcoming testcase, which crashes Mozilla on load. Also crashes in Firefox1.5.0.3.
Marking security sensitive, just to be sure.

Talkback ID: TB18394613W
0x00000000
nsSVGSVGElement::DidModifySVGObservable   nsSVGValue::NotifyObservers   nsSVGValue::DidModify   XPTC_InvokeByIndex   XPCWrappedNative::CallMethod

Comment 1

[Martijn Wargers (dead)]

Attachment: testcase, crashes on load

The iframe with the svg has this code:

<script xmlns="http://www.w3.org/1999/xhtml">
window.addEventListener('SVGZoom', doe, true);
function doe(e) {
var x= parent.document.getElementsByTagName('iframe')[0];
x.parentNode.removeChild(x);
}
setTimeout(doe2, 1000);

function doe2() {
document.documentElement.currentScale = 2;
}
</script>
</svg>

Comment 2

[Martijn Wargers (dead)]

Attachment: Backtrace from debug build

Comment 3

[Martijn Wargers (dead)]

Attachment: testcase2

Same crash seems to happen with SVGScroll event.

Comment 4

[Martijn Wargers (dead)]

This now got worksforme between 2007-01-04 and 2007-01-05:
http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=all&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2007-01-04+04&maxdate=2007-01-05+09&cvsroot=%2Fcvsroot
It seems that this was somehow fixed by bug 333078?
So I guess this bug can be resolved then, right?

Comment 5

[Boris Zbarsky [:bzbarsky]]

I don't think so.  If it's just something that's hidden because of when the cycle collector happens to run or not run, that's not really good enough.  In my opinion, of course.

In any case, the fix for this bug is obvious.  DO NOT hold a weak ref to an object you're dispatching an event on.  Just don't do it.  So fix nsSVGSVGElement::DidModifySVGObservable to hold a strong ref to the presshell.  ;)  And then probably audit the rest of the code that calls that method.

Comment 6

[Olli Pettay [:smaug][[e-mail]]]

Attachment: proposed patch

Keep strong ref to presShell.
Went quickly through the code and this really should be enough.
The patch applies to branch and trunk.

Comment 7

[Olli Pettay [:smaug][[e-mail]]]

Checked in

Comment 8

[Daniel Veditz [:dveditz]]

Comment on attachment 256027 [details] [diff] [review]
proposed patch

Is this patch appropriate for 1.8.0 as well? Firefox 1.5.0.x is also affected by this bug.

Comment 9

[Daniel Veditz [:dveditz]]

Comment on attachment 256027 [details] [diff] [review]
proposed patch

approved for 1.8.0.12 and 1.8.1.4, a=dveditz for release-drivers

Comment 10

[Marcia Knous [:marcia]]

verified on the 1.8 branch using Mozilla/5.0 (Windows; U; Windows NT 6.0; en-US; rv:1.8.1.4pre) Gecko/2007050804 BonEcho/2.0.0.4pre. No crash for me, but i do still see the iframe in the page. Martijn attributes that to a test case error. Adding branch fixed keyword.

Comment 11

[Bob Clary [:bc] (inactive)]

crash test landed
http://hg.mozilla.org/mozilla-central/rev/8334854afb84

Comment 12

[Boris Zbarsky [:bzbarsky]]

The crashtest seems like a no-op to me: it finishes after 500ms, but the subframe doesn't do its dirty work until 1s after its code runs... So the test is not exercising the code.