154967 - Crash by filling up memory while browsing google.com searches

Status: RESOLVED WORKSFORME

(Core :: XPCOM, defect)

Description

[Silviu]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.0.0) Gecko/20020529
BuildID:    2002052918

When I'm searching using www.google.com after a few searches (sometimes the
first sometimes after a few more) Mozilla starts to fill up the memory and is
killed by the linux kernel when the memory is full

Reproducible: Always

Steps to Reproduce:
1.Search on www.google.com
2.
3.

Actual Results:  Mozilla started after the first search (sometimes only after a
few more) to use more and more memory (as reported by top) and when the memory
was full mozilla-bin was killed by the linux kernel

Expected Results:  IMHO Mozilla should work fine on www.google.com (older
versions were working fine)

Due to the active nature of the www.google.com search results I am unable to
isolate the problem. Also the fact that sometimes doing exactly the same search
doesn't necessarily crash mozilla isn't helping me to isolate the problem.
But it is reproductible 100% as I am not able to do more than 5-8 searches
(maximum) on www.google.com.
The only message I get is from the linux kernel:
bluebox kernel: Out of Memory: Killed process xxxx (mozilla-bin).
Mozilla dies with absolutely no message.
Additional Information:
I am using the default theme of Mozilla.
The system is: Slackware 8.0 with several updates. Kernel 2.4.17.
Hardware Duron 650Mhz with 196Mb RAM and 200Mb Swap

Comment 1

[Thomas Sjögren]

wfm . Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.1a) Gecko/20020610

Comment 2

[Adam Hauner]

Silviu: If you are crashing in Mozilla the best thing you can do to help the
developers fix your bug is to attach a stacktrace. If you're not building
yourself you are not out of luck. Mozilla releases nightly and milestone builds
with Netscape's Talkback (you can get latest build on:
http://ftp.mozilla.org/pub/mozilla/nightly/latest/) Talkback should catch most
crashes and offer to send in a crash report. Developers can retrieve that crash
report and attach it to your bug report if you provide either the Incident ID
(you can get it by running the talkback program from /components/talkback/).
Thanks for your help in testing Mozilla and reporting bugs.

Comment 3

[timeless]

I can't expect a stack for this, if the linux kernel kills a runaway or leaky
process because it needs memory, then it has better things to do than allow
someone to gather a stack.

i'm sending this to xpcom in hopes that a unix guru will have ideas about how to
make the memory pressure stuff work on linux. the core problem you're actually
experiencing might be a leak, but it'd be helpful if you managed to log your
activities.  a really crummy way would be NSPR_LOG_MODULES=nsHttp:1
NSPR_LOG_FILE=somefile (i'm hoping 1 does something useful like record pages
being loaded, 5 means log everything)

Comment 4

[Alec Flett]

my only guess is this is related to the processing of search results by the
search service.. however if this is a runaway allocation (i.e. continuously
allocating more and more in a loop) then the memory pressure observer won't help
here (it will just make the process live an extra 1/2 second or so before being
killed)

this is on a pretty old build (about 2 months!) - can anyone reproduce this on a
recent build?

Comment 5

[ABi]

Bug 163802 is very similar.
I also cannot use google any more.
I even deleted google.src in /searchplugins but it did not help!

Comment 6

[Vadim Kutsyy]

I have simular problem on WIN2K 2002101612.  Ussualy it happens if I glock
"groups" after mozilla search.  It takes up 97% CPU, and doesn't do anything
else.  I can not gat stachtree either.

Comment 7

[John Mertic]

I am having the same issue with Mozilla 1.2.1 release on Win2k SP3. Google
searches ( especially on Google Groups ) can take all of the system resources
when try to load search results. Freezes Mozilla, have to kill and then restart.

Comment 8

[Brendan]

I'm not quite sure whether my problem is the same, but it is very similar to
this one and rather than start a new bug I thought I'd just comment on my
experience here. 

When I try to search on Google Mozilla freezes, and my whole system becomes very
slow. I usually end up having to kill the process. 
I am running RH 7.3/KDE3 and Mozilla 1.2.1.  I was previously using Mozilla
1.??? and had the same experience. When I upgraded to 1.2.1 Google searching
worked fine for a couple of days and I breathed a sigh of relief.  A day or two
ago the freezing started again. 

I now have to use Konquerer to do a search and cut and paste anything
particularly interesting into Moz. 

Brendan 
    

Comment 9

[Richard Körber]

I can confirm this bug: Mozilla 1.3b (Build ID: 2003021008), Linux RedHat 7.3
(freshly installed, Kernel 2.4.18), 256MB RAM, 512MB Swap.

A lot of searches are working fine on google.de, but some show this behaviour
reproducably. On my system, top shows that the X process keeps allocating all
available memory, while Mozilla only requires some more memory. When the memory
has entirely been filled up, Google's result page is shown correctly, and I can
continue browsing. Anyhow the memory keeps being filled up until I quit Mozilla.

On an older system (a tweaked RedHat 7.2, Mozilla 1.3a), quite the very same
problem occured to me. The difference was that Mozilla killed itself when the
memory has been filled up. I am not sure if X or the mozilla-bin process
allocated all the memory. This system is not available for testing any more, though.

I have the impression that this bug is depending on the entered search terms,
and only occurs when searching for web pages or images (not usenet). As soon as
I have found a search term where I can 100% reproduce this, I will report it here.

Maybe interesting to you: I am using the latest versions of these Mozilla
extensions: optimoz, bannerblind, refspoof (all from mozdev) and tabextensions
(http://white.sakura.ne.jp/~piro/xul/_tabextensions.html.en).

@Silviu: what extensions are you using?

Comment 10

[Richard Körber]

Okay, I have found one! :-) Going to this URL always triggers this bug for me:

http://www.google.de/search?q=K%C3%B6ln+Semesterbeitrag+2003&hl=de&lr=&ie=UTF-8&oe=UTF-8&start=10&sa=N

You can reach this page like this:
* Go to http://www.google.de
* Enter the search term "Köln Semesterbeitrag 2003" (without quotes), then click
"Google-Suche" below.
* Proceed to the 2nd result page by clicking "Vorwärts" on the bottom of the page.
* Mozilla will now show this behaviour.

Can anyone confirm?

Comment 11

[Richard Körber]

Yeap, I have found the cause for this crash... :-)

I could boil the problem down to a page containing just one single UTF-8
character (0xC2 0x93):

  http://www.shredzone.de/mozilla/utf8.html

Opera 6.11 crashes with similar symptoms too, when trying to display this page.
It seems like X is filling up all the memory when requested to render this
character.

Mozilla cannot be blamed for this one IMO.

Comment 12

[Kai Lahmann (is there, where MNG is)]

Reporter: Can you reproduce this bug with a recent build of Mozilla (for
example, 1.4 RC1)? If so, then please comment again with details. If not, then
please resolve this bug as WORKSFORME. Thanks.

Comment 13

[Silviu]

I cannot reproduce this problem anymore. The last configuration on witch I could reproduce this 
bug was: 
Durn 650 Mhz, VIA KT133 motherboard, 128 Mb RAM, slackware 8.1 with upgraded kernel to 
2.4.19 and X to 4.3.0. Mozilla was 1.3a. 
Upgraded to (all at once) 1200Mhz athlon, 256 Mb RAM, slackware 9.0 with Mozilla 1.3 and the 
bug wen away. Since my upgrade was "all at once" I cannot point witch of this solved my 
problem. 

Comment 14

[Richard Körber]

Sorry for replying that late... The bug is still acute for me. I experience this
issue at least once a day. I have tested with Mozilla up to the official 1.4
release.

I cannot really track down the cause of it. It seems to be a problem with XFree.
I have two Linux machines here, both are nearly identically installed (RedHat
7.3, Ximian Desktop 2).

On machine A this issue often shows up and can always be reproduced. On machine
B this bug does NOT show up at all. Now, when I remotely start Mozilla on
machine A and having it displayed on machine B, everything works fine too...

Even more strange, Opera also shows this behaviour on that machine A, but Galeon
is working fine.

I really hope someone can give me a hint how I could get rid of that nasty bug,
because it is really becoming a major pain in the ass for me... :)