huge memleak in *webkit based browsers

AndyRTR · 2012-04-09 15:30:56

I'm using Xxxterm with testing libwebkit (gtk2 based) and get huge memleaks browsing certain pages. Can you try e.g. that one:

http://forum.xda-developers.com/forumdisplay.php?f=1407 - with no JS or any other filters?

It fills all memory almost instantly. I also had this with Facebook when it wants to autorefresh the page sometimes. It seems like a nasty JavaScript bug in new webkit to me. Can somebody confirm thist before I report it to the webkit tracker?

-Andy

hadrons123 · 2012-04-09 18:42:16

I 'm using chrome and I disabled JS. Tried your link. no memory leaks.

AndyRTR · 2012-04-09 18:57:04

Please try a libwebkit (gtk) 1.8.x based browser (Luakit, Xxxterm, Epiphany,....) and try with all features enabled. I assume Google's internal webkit releases are far from what the community gets back.

hadrons123 · 2012-04-09 19:03:35

Tried epiphany too with JS disabled. No memory leaks.

AndyRTR · 2012-04-09 19:07:14

Sure, With JS disabled nothing happens. Enable JS and it floods all memory here withing a few seconds. Just tested with removing that forum from my JS whitelist that is JS disabled again and so it works well.

jasonwryan · 2012-04-09 19:07:25

I don't get it on the page you linked to, but on http://unix.stackexchange.com/ - where they have also recently introduced a javascript 'new activity' feature, I see this issue.

Vimprobable2: 1.0.1
libwebkit: 1.6.3
libsoup: 2.36.1

Gcool · 2012-04-09 20:51:31

Same issue here.

luakit: 2012.03.25-1
libwebkit: 1.8.0-2

jass0 · 2012-04-14 01:16:39

I opened both your links using midori and kept them open while writing this message. Javascript enabled. No memory leaks

midori-0.4.4-1
libwebkit-1.8.0-2

Last edited by jass0 (2012-04-14 01:20:08)

Earnestly · 2012-04-14 01:30:28

dwb-hg : 1191-1
libwebkit : 1.6.3-1
libsoup : 2.36.1-1
gtk2 : 2.24.10-3

No memory leaks

Last edited by Earnestly (2012-04-14 01:31:32)

skottish · 2012-04-14 14:11:01

jasonwryan wrote:

I don't get it on the page you linked to, but on http://unix.stackexchange.com/ - where they have also recently introduced a javascript 'new activity' feature, I see this issue.
Vimprobable2: 1.0.1
libwebkit: 1.6.3
libsoup: 2.36.1

Hiya Jason.

I get a runaway process with this site, but I'm also not seeing memory leaks with dwb-hg anywhere that I've seen yet.

aksr · 2012-04-29 07:07:27

Same issue here. I thought it's dwb's problem.., until I tried vimprobable2 ..

Jristz · 2012-04-29 08:03:46

Midori 0.4.5 / 0.4.5a-36-g2d5762d
Libwebkit 1.8.1 /gtk2 gtk3
gtk2 2.24.10-3/gtk3 3.4.1

xf86-video-vesa / libegl/libgl/libglapi/libgles

All give me a partialy freeze (Switch to tt1 (30sec) login (20sec) kill midori solved)

Extra info:
and html5-youtube videos give me a slowdown to the system (al take the dowble time but the video run perfectly)

Last edited by Jristz (2012-04-30 00:18:50)

AndyRTR · 2012-04-29 08:32:55

I'm also affected by those freezes and have to login from another device to killall xxxterm. Probably also libwebkit related.

skottish · 2012-04-30 00:16:23

libwebkit based browsers are behaving differently on my different machines. I haven't seen the memory issues that you started this thread about, but on my laptop using dwb I get 100% core usage after visiting some sites (of course that can still be memory leak related). I went to test this on another webkit browser, so I tried Midori which segfaulted within seconds. My workstation doesn't suffer from either of these problems. The laptop has an Intel GPU and the workstation has a nVidia GPU. I don't follow webkit development closely. Are they using hardware acceleration for anything now?

jass0 · 2012-05-05 12:25:49

I gave this another try. I'm running midori from testing on a Eee PC 900 (has an intel gpu) with 2 GB of memory.

Its single core processor is running at 900 MHz so it's definately not a powerful laptop, but I don't get any lag or unresponsiveness when visiting those URLs (cpu usage jumps though). HOWEVER a short while after entering the xda-developers link, about 1 GB more memory is allocated and then freed properly after ~5 seconds.

Last edited by jass0 (2012-05-05 12:33:24)

Trilby · 2012-05-05 12:31:12

I suspect skottish may be on to something. I have recently started experiencing this (or a related) issue on webkit browsers on my work computer (intel video) but not on my home computer (ati video).

Edit: I also haven't noticed this on another machine I use with nvidia (nouveau driver), but I also don't use that one enough to know if the problem isn't there, or just hasn't been spotted yet.

Another edit: I just saw this was in the [testing] forum. I am not using testing, so unless the relevant version of webkit just came to [extra] (or whichever repo it's in) then my observation may be unreleated.

Last edited by Trilby (2012-05-05 13:02:28)

skottish · 2012-05-05 17:53:00

Trilby wrote:

Another edit: I just saw this was in the [testing] forum. I am not using testing, so unless the relevant version of webkit just came to [extra] (or whichever repo it's in) then my observation may be unreleated.

It started with 1.8.0, so yes, the one in extra is on point.

*** EDIT ***

Right after I posted this, I started to see it on the workstation as well. I noticed some lagging on pages and checked htop and found that I was using 2GB of RAM with nothing but dwb running. I had never seen this on this machine. If it is related to hardware acceleration, then either something happened with the newer nVidia releases or I've just been lucky.

Last edited by skottish (2012-05-06 16:57:14)

Arch Linux

#1 2012-04-09 15:30:56

huge memleak in *webkit based browsers

#2 2012-04-09 18:42:16

Re: huge memleak in *webkit based browsers

#3 2012-04-09 18:57:04

Re: huge memleak in *webkit based browsers

#4 2012-04-09 19:03:35

Re: huge memleak in *webkit based browsers

#5 2012-04-09 19:07:14

Re: huge memleak in *webkit based browsers

#6 2012-04-09 19:07:25

Re: huge memleak in *webkit based browsers

#7 2012-04-09 20:51:31

Re: huge memleak in *webkit based browsers

#8 2012-04-14 01:16:39

Re: huge memleak in *webkit based browsers

#9 2012-04-14 01:30:28

Re: huge memleak in *webkit based browsers

#10 2012-04-14 14:11:01

Re: huge memleak in *webkit based browsers

#11 2012-04-29 07:07:27

Re: huge memleak in *webkit based browsers

#12 2012-04-29 08:03:46

Re: huge memleak in *webkit based browsers

#13 2012-04-29 08:32:55

Re: huge memleak in *webkit based browsers

#14 2012-04-30 00:16:23

Re: huge memleak in *webkit based browsers

#15 2012-05-05 12:25:49

Re: huge memleak in *webkit based browsers

#16 2012-05-05 12:31:12

Re: huge memleak in *webkit based browsers

#17 2012-05-05 17:53:00

Re: huge memleak in *webkit based browsers

Board footer