272448 - 100% CPU Utilization After Period of Time on aa419.org

Status: RESOLVED WORKSFORME

(Firefox :: General, defect)

Description

[Ryan VanderMeulen [:RyanVM]]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7.5) Gecko/20041118 Firefox/1.0 (Community Edition Firefox P3V-X1)
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7.5) Gecko/20041118 Firefox/1.0 (Community Edition Firefox P3V-X1)

When I open the page in a new tab, things work fine for awhile. After a long
period of time (it took around an hour or so on my system), Firefox began to
consume 99% of the CPU cycles and the only way to kill the process was through
task manager.

Reproducible: Always
Steps to Reproduce:
1. Oepn http://www.aa419.org/ladvampire.html in a new tab
2. Let Firefox sit on the site for awhile
Actual Results:  
After a long period of time, Firefox will consume 99% of the CPU cycles.

Expected Results:  
Firefox should continue to consume ~50% of the cycles like it does when
initially on the page.

Comment 1

[Ryan VanderMeulen [:RyanVM]]

Also, memory usage rises continually the entire time the site is open.

Comment 2

[Daniel Wang]

Attachment: stress test 1 (perpetual js reload only)

stress test 1 - reload images only, no animation

Comment 3

[Daniel Wang]

Attachment: stress test 2 - reload images with animations

stress test 2 - this time, with animated images

Comment 4

[Daniel Wang]

worksforme
reporter (Ryan), can you test stress test 1, and then stress test 2 and see if
you can reproduce the problem?

Comment 5

[Ryan VanderMeulen [:RyanVM]]

I'm running the stress tests right now. For the record, it takes until about the
90,000 image mark on the real site before the browser dies for me. I'll let you
know my results on the two stress tests.

Comment 6

[Ryan VanderMeulen [:RyanVM]]

(In reply to comment #3)
> Created an attachment (id=167535)
> stress test 2 - reload images with animations
> 
> stress test 2 - this time, with animated images

OK, here's what I've got for ya. Sorry it took so long, but I'd been having
really bad connection problems lately.

Test 1:	Testcase #1
	Result: 99% CPU
	Starting Memory Usage: 14MB
	End Memory Usage: 223MB
	Time Run: 50min
	Images Loaded: 74707
	Average Image Loading Rate: 24/s
	Average CPU Utilization Prior to 99%: 20-30%

Test 2:	Testcase #2
	Result: 99% CPU
	Starting Memory Usage: 15MB
	End Memory Usage: 226MB
	Time Run: 60min
	Images Loaded: 74555
	Average Image Loading Rate: 20/s
	Average CPU Utilization Prior to 99%: 20-30%

Test 3:	aa419.org
	Result: 99% CPU
	Starting Memory Usage: 11MB
	End Memory Usage: 147MB
	Time Run: 60min
	Images Loaded: 74543
	Average Image Loading Rate: 20/s
	Average CPU Utilization Prior to 99%: 20-40%

It should be noted for all of these tests that the transition from 20-30% CPU
utilization to 99% was very abrupt. It wasn't a gradual climb in CPU utilization
(unlike the memory usage which grew continually the entire time). There were a
few spikes in the 80-90% range right before the end, but it wasn't a steady
increase.

And for the record, I have 1GB of RAM and system memory utilization when Firefox
dies is still well below that.

Comment 7

[François Gagné]

It could be interresting to check the number of GDI in use.

In the task manager you can add the "GDI Objects" colums, "View" Menu -> "Select Columns ..."

Comment 8

[Ryan VanderMeulen [:RyanVM]]

URL has changed. Updating accordingly.

Comment 9

[Wayne Mery (:wsmwk)]

(In reply to comment #7)
> It could be interresting to check the number of GDI in use.

http://aa419.org/vampire/ladvampire.php for 20+ hours with Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9a1) Gecko/20060531 Minefield/3.0a1

François  GDI does increase substantially. I started at about 100 and it's now 1,500+. What do you conclude from that?

cpu has been low and steady but varies between 5 and 15 percent. memory usage does not seem extreme.

Comment 10

[François Gagné]

1500 is a fair number, higher than normal usage but I don't think it's critical. My hypotheses was simply that there could have been a leak of GDI of objects by Firefox and thus causing the disproportionate CPU usage (which doesn't seems to be the case).

Comment 11

[Wayne Mery (:wsmwk)]

Ryan, do you still see the problem??

WFM now that comment 9 still true after (now) 3 days.  GDI is at 2,000+, so the rate of increase has decreased.  
Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.9a1) Gecko/20060531 Minefield/3.0a1

Comment 12

[Ryan VanderMeulen [:RyanVM]]

I'll try it out this weekend and let you know.

Comment 13

[Ryan VanderMeulen [:RyanVM]]

Yeah, with a recent Cairo trunk build, things seem to be good (though the memory usage creeping up is still evident). I'm going to resolve this WFM since there are plenty of other bugs out there regarding memory not being released.