260240 - hangs in specific site with undetected document.all support

Status: RESOLVED WORKSFORME

(SeaMonkey :: General, defect)

Description

[Sekundes]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.0; rv:1.7.3) Gecko/20040913 Firefox/0.10
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.0; rv:1.7.3) Gecko/20040913 Firefox/0.10

It seems related to bug 256644, which loops while document.all is being used in
DHTML events.

Reproducible: Always
Steps to Reproduce:
1. load http://www.gamebase.com.tw/.
2. wait until the ads shown.
3. scroll the scrollbar.

Actual Results:  
Firefox hangs.

Comment 1

[Bill Mason]

This can't be from bug 256644, as a build from the day before that bug's
resolution (2004082209 PC/WinXP) hangs as well.

Confirm hang, though.

Comment 2

[Nile]

Odd, I'm getting an error. The page is in chinese except for this bit:

HTTP 錯誤 500-13 - 伺服器太忙碌
Internet Information Services

Blindly assuming that this is a 500 error, something's wrong on their side?

Comment 3

[Hermann Schwab]

it even was freezing windows98 taskmanager, could start taskmanager, but not use
the mouse or the keyboard to kill the selected mozilla.

Maybe I´ve found a bug in the js:

view-source:http://www.gamebase.com.tw/

function HtmlEncode(text)
{
    return text.replace(/&/g, '&amp').replace(/'/g, '&quot;').replace(/</g,
'&lt;').replace(/>/g, '&gt;');
}

Isn´t there a semicolon missing in the first replace function?
is replace(/&/g, '&amp') correct, not replace(/&/g, '&amp;')?

Comment 4

[Asaf Romano (gone)]

I'm wondering if the hang is related to bug's summary.

Comment 5

[Sekundes]

(In reply to comment #2)
> Odd, I'm getting an error. The page is in chinese except for this bit:
> 
> HTTP 錯誤 500-13 - 伺服器太忙碌
> Internet Information Services
> 
> Blindly assuming that this is a 500 error, something's wrong on their side?

It's 500-13 error. To my knowledge, it means the server is too busy.

(In reply to comment #3)
> it even was freezing windows98 taskmanager, could start taskmanager, but not use
> the mouse or the keyboard to kill the selected mozilla.
> 
> Maybe I´ve found a bug in the js:
> 
> view-source:http://www.gamebase.com.tw/
> 
> function HtmlEncode(text)
> {
>     return text.replace(/&/g, '&amp').replace(/'/g, '&quot;').replace(/</g,
> '&lt;').replace(/>/g, '&gt;');
> }
> 
> Isn´t there a semicolon missing in the first replace function?
> is replace(/&/g, '&amp') correct, not replace(/&/g, '&amp;')?

It's strange that I don't see the code you provided.
Prehaps it's random ads code.

(In reply to comment #4)
> I'm wondering if the hang is related to bug's summary.
I just get this hang in Firefox 1.0PR but not 0.9 so I pretend it's related to
document.all support. Sorry if it isn't.
However, with bug 256644 fixed makes JavaScript Console shown full of errors and
hangs for a while to open it due to too many errors.

Comment 6

[Hermann Schwab]

(In reply to comment #5)
The page I´ve been seeing was a short script, splitting and retrieving and
merging URLS, I didn´t quite understand, assumed that will build the page via
document.write. At the end was some MS IIS eror messages,  But I can´t read
chinese, and don´t have chinese fonts installed.
I´m sorry that I didn´t save the page. 

I assume 

function HtmlEncode(text)
{
    return text.replace(/&/g, '&amp').replace(/'/g, '&quot;').replace(/</g,
'&lt;').replace(/>/g, '&gt;');
}

will replace & with "&amp" and not with "&amp;" wouldn´t that produce errors by
not giving the correct parameters?
 
As js tends to be reused, I wonder if that function is used elsewhere on the site.

I installed LiveHttpheaders, opened it, and loaded the site.
A lot of headers were passing by, then a script, one more gif, and mozilla or
the computer was hanging, don´t remember exactly.
The script could have been this one, but I´m not sure about that:
http://210.242.69.247/adjs.php?clientid=16
this will document.write a link with image, then a hidden div containing a
webbug. The webbug will be loaded as image, 
... width='0' height='0' alt='' style='width: 0px; height: 0px;'>

My preference for document.all:
browser.dom.document.all.disabled user set boolean true

Comment 7

[Brendan Eich [:brendan]]

bugs.mano, don't reassign undiagnosed bugs to the JavaScript Engine.

The undetected document.all support is part of the DOM: Level 0, anyway, but I
am moving this bug back to Browser-General till we get better diagnosis.

/be

Comment 8

[Johnny Stenback (:jst)]

For the record, I did look into this a while ago (forgot to comment in the bug,
apparently), and it seems like this isn't really a hang, it just takes a long
long long time to get out of what this site is doing. This site ends up firing
timeouts (probably interval timeouts) and doing something when those timeouts
run that end up hogging the CPU real bad, and it's really hard to get out of.
And what they're doing is something that's time consuming, yet doesn't get
caught by the branch callback protection code. I wish I knew more, but that's
what I've figured out so far. In no case did I see this being a real hang in any
of the new code that was added for the undetectable document.all support, I'm
guessing it's the site that now executes different code in mozilla due to the
fact that we now support document.all.

Comment 9

[Sekundes]

It is basically the dupe of bug 261633, but since the web authors have changed
its code, I'd mark it as WFM.