195714 - moz-opacity crashes latest 1.3 Branch

Status: RESOLVED DUPLICATE of bug 192469

(SeaMonkey :: General, defect)

Description

[Markus Hübner]

Using 1.3 branch build 2003030205 on winxp pro sp1 I crash on an intranet page.
[testcase coming up]

TB17677196H

Comment 1

[Markus Hübner]

Testcase is ready.
I'm having a DIV element with a SPAN inside it and a TABLE inside the SPAN. 
When changing this DIV's moz-opacity for the first time, Mozilla becomes 
unstable. When changing it for the second time, Mozilla 1.3 totally crashes. 
Works fine in trunk build 2003022709, as well as 1.2 and NS7

Comment 2

[Manfred Schneiderbauer]

See also the inline comments in the testcase's source.

Comment 3

[José Jeria]

WFM 2003030208, Windows 2000 SP1

Comment 4

[Boris Zbarsky [:bzbarsky]]

Has someone asked for the talkback stack yet?  Or should I do it?

Comment 5

[David Bradley]

Attachment: Stack for TB17677196H

It's crashing on the ret, so looks like stack corruption is a good bet. This
isn't a virtual call (Assuming the line # is accurate), so doesn't look like a
case of invoking a method on a dead object.

Comment 6

[Zibi Braniecki [:zbraniecki][:gandalf]]

Confirming. 2003030208 1.3 trunk. WinXP.
Will another TB help?

Comment 7

[Zibi Braniecki [:zbraniecki][:gandalf]]

This bug does not influence 1.4 trunks.

Comment 8

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

I already fixed this in 1.4.

Personally I don't think this should block the 1.3 release. It only happens on
pages which use -moz-opacity or background-position:fixed on an inline element
which contains a block element. That is not very common.

If someone wants to nominate for 1.3 and see what other drivers think, feel
free. (But nominate bug 192469 where the fix is.) The fix is very safe because
it will only change our behavior in the rare situation mentioned above.

*** This bug has been marked as a duplicate of 192469 ***

Comment 9

[Manfred Schneiderbauer]

I think this is a major issue anyway. It appears to me that creators of dynamic 
web applications (like myself) would have to make heavy workarounds to make 
things in 1.3 work that already worked in 1.2. The only alternative would be to 
just leave out certain features for Mozilla users that actually should be 
enabled for use. To me, this is not really acceptable... in this case, I would 
either have to turn off the opacity effect on Moz (which would make things 
worse looking than on IE) or to construct the object for Mozilla in a different 
way (which is ofcourse a way to handle it, but will not be necessary for 
earlier or later versions, so it's just a new burden for the web developer)... 
it would be great to see that fixed.

Comment 10

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

Using a TABLE inside a SPAN is not valid HTML, so you really should be using
different structure.

But feel free to nominate bug 192469 for 1.3final.

Comment 11

[Zibi Braniecki [:zbraniecki][:gandalf]]

Nominating 192469 as 1.3 blocker.

I cannot get Comment #8. There are really, many pages which use -moz-opacity
since it's in Mozilla for long, long time. It's quite widely used on
http://dhtmlcentral.com http://alladyn.art.pl and many others.

Second thing is that as i understand we already _have_ patch for this and it's
extremly safe. So what we're gaining leaving our third public, final relase with
crash on DHTML pages????

Comment 12

[Manfred Schneiderbauer]

Uhm... what exactly is not valid when placing a TABLE inside a SPAN? A SPAN is 
a container element for just everything.

Comment 13

[José Jeria]

A block element cannot be inside a inline element.

Comment 14

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

This bug does not affect all uses of -moz-opacity. It only affects -moz-opacity
applied to an inline element containing a block (which is invalid HTML). So even
if there are many sites using -moz-opacity, it's not clear that this will cause
a crash on any of them. In fact, given that we've had the bug for nearly three
months and I've seen exactly two bug reports filed, it appears that this is not
a big problem in practice.

Having said that, I'm still happy to put this fix into 1.3 if Asa or some other
driver approves.

Comment 15

[Zibi Braniecki [:zbraniecki][:gandalf]]

Ok. So still, if Mozilla works with this invalid HTML, then we have to handle
it. And for sure we shouldnt react in crash in final ver.

Comment 16

[Manfred Schneiderbauer]

I extended the testcase on http://www.world-direct.com/mozilla-opacity-crash-
testcase/ with a fading element to demonstrate another problem which is already 
well-known since a year or so: fading will flicker and is extremely slow.

Comment 17

[Markus Hübner]

Filed a sperate one - bug 195883 about the performance problem.

Comment 18

[Asa Dotzler [:asa]]

not a blocker.