high cpu + OOM with adblock plus



7 years ago
5 years ago


(Reporter: dvrsn, Unassigned)


SeaMonkey 2.16 Branch
Dependency tree / graph

Firefox Tracking Flags

(Not tracked)


User Agent: Mozilla/5.0 (X11; Linux i686; rv:19.0) Gecko/20100101 Firefox/19.0 SeaMonkey/2.16
Build ID: 20130130231528

Steps to reproduce:

Installed adblock plus
Added subscription to "Easylist"
Visited page: http://www.openforum.com/articles/top-10-ways-to-be-gracious/

Actual results:

cpu usage goes to 100%
memory usage steadily rises (leading to eventual error ###!!! ABORT: OOM: file /builds/slave/rel-c-beta-lnx-bld/build/mozilla/xpcom/string/src/nsTSubstring.cpp, line 533 )
seamokney appears to be constantly open() or stat64() adblockplus "patterns.ini" files according to strace

Expected results:

none of that stuff, especially not the crashing.
This appears to affect FireFox 18/Win7 also.
Blocks: abp
May be related to Bug 837845
> Added subscription to "Easylist"

Appears that the simple act of installing ABP automatically adds a subscription.
(That must be somewhat new?  In the past you always had to actually do it yourself or at least on prompt to install one.)

> cpu usage goes to 100%

(Actually I suppose it will be 100% of 1 core.)

> appears to affect FireFox 18/Win7 also

I'm not seeing FF 18 being affected (XP)?
Could you confirm?

> Bug 837845 - Multi-GB memory spike when using regular expressions

That bug says "status-firefox18: unaffected" which (you would think) would also mean that SeaMonkey 2.15 would also be unaffected.

Yet this bug certainly does affect SeaMonkey 2.15.
Note that after this (CPU jump) starts, if you Disable ABP, it stops immediately, but if ABP is disabled initially, you'll need to refresh the openforum.com page for it to start.

adblock_plus-2.0.2-sm+tb+fx+fn.xpi does not cause this issue.
adblock_plus-2.1-an+sm+fx+tb-linux.xpi & versions later do cause this issue.
I'm not seeing FF being affected, 18 or otherwise?

Mozilla/5.0 (Windows NT 5.1; rv:21.0) Gecko/20130206 Firefox/21.0
SeaMonkey 2.14.1 is not affected.
Mozilla/5.0 (Windows NT 5.1; rv:17.0) Gecko/20100101 Firefox/17.0 SeaMonkey/2.14.1
Build identifier: 20121129191119

SeaMonkey 2.15 & later are.
Mozilla/5.0 (Windows NT 5.1; rv:18.0) Gecko/20100101 Firefox/18.0 SeaMonkey/2.15
Build identifier: 20130105222038

> adblock_plus-2.0.2-sm+tb+fx+fn.xpi does not cause this issue.

I'm wrong on that, it does.

2.0.2 is not restartless (which I didn't realize at the time) & when I installed it, it "appeared" to install (correctly) & not cause an issue.  Rechecking, removing later versions of ABP first, then installing 2.0.2 & restarting, the issue does manifest itself.  (Also disabling ABP 2.0.2 does not cause CPU usage to subside like it does with the restartless versions.  You must disable, then Quit, then restart.)
> Bug 837845 - Multi-GB memory spike when using regular expressions

SeaMonkey 2.15.2 is not affected.
SeaMonkey 2.17a2 (Aurora) is affected so this bug is different from 837845.
Not a regression range, but closer then earlier.
(Is that apples to apples?  One -central, a1, one -aurora, a2?)

Mozilla/5.0 (Windows NT 6.1; WOW64; rv:18.0) Gecko/18.0 Firefox/18.0 SeaMonkey/2.15a1
Build identifier: 20120907003006


Mozilla/5.0 (Windows NT 6.1; WOW64; rv:18.0) Gecko/18.0 Firefox/18.0 SeaMonkey/2.15a2
Build identifier: 20121022013003

Still not see FF being affected.
Reporter, do you still see this problem?
Flags: needinfo?(dvrsn)
Whiteboard: closeme WFM 2013-06-01
Yes, I still see this problem on win7, using a build tagged 
in about:buildconfig (sorry, I don't know where to find better build info)
Flags: needinfo?(dvrsn)
In about:support
Whiteboard: closeme WFM 2013-06-01
Almost one year later, Jeff, do you still have this problem?
Flags: needinfo?(dvrsn)
Whiteboard: closeme INCO 2014-06-01

SeaMonkey 2.25
Adblock Plus 2.6
EasyList filter 5-1-2014
Win XP

URL: http://www.openforum.com/articles/top-10-ways-to-be-gracious/
Resolved per whiteboard
Closed: 5 years ago
Flags: needinfo?(dvrsn)
Resolution: --- → INCOMPLETE
Whiteboard: closeme INCO 2014-06-01
You need to log in before you can comment on or make changes to this bug.