Bug 700885

Summary: firefox crashes when run over nxclient [@ nsCOMPtr_base::assign_with_AddRef]
Product: [Fedora] Fedora Reporter: Tom Horsley <horsley1953>
Component: firefoxAssignee: Martin Stransky <stransky>
Status: CLOSED WORKSFORME QA Contact: Fedora Extras Quality Assurance <extras-qa>
Severity: medium Docs Contact:
Priority: unspecified    
Version: 14CC: gecko-bugs-nobody, stransky
Target Milestone: ---Keywords: Triaged
Target Release: ---   
Hardware: x86_64   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2012-02-27 07:26:03 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Attachments:
Description Flags
gdb session backtracing all the threads none

Description Tom Horsley 2011-04-29 18:11:46 UTC
Description of problem:

This is wacky, but I figured I might as well report it.

I'm running the nomachine nxclient-3.4.0-7.x86_64 at home to talk to a
freenx server at work (fedora 14 x86_64 on both systems). I'm running
absolutely all of the nxclient traffic through an ssh tunnel.

When, and only when, I'm running nxclient at home, a firefox, also run at
home flakes out. In the past it has only been very very slow responding to
clicks in the bookmarks sidebar (clicks in the main browser window have
no slowdown), but today not only does it get slow, but it also starts
aborting at random when I click on the sidebar bookmarks.

I ran it with ulimit -c unlimited, generated a giant core file:
-r--r--r-- 1 tom users 232972288 Apr 29 13:22 core.15733
and loaded a bunch (though by no means all) of debuginfo files
and created a backtrace of all the threads in gdb (which I will
attach to this bug).


Version-Release number of selected component (if applicable):
firefox-3.6.16-1.fc14.x86_64


How reproducible:
It is random, but a very few minutes of clicking around the bookmarks
in the sidebar usually gets it to happen.

With no nxclient running, operation is totally smooth and no aborts
seem to happen (I can't even begin to imagine how the two could
interact or even tell each other they exists in order for the behaviour
to change, but I'm just reporting what I'm seeing).

Steps to Reproduce:
1.see above
2.
3.
  
Actual results:
slow sidebar response, occasional abort

Expected results:
just as snappy with or without nxclient running

Additional info:
I have 8 gig of memory on this system, and the vast portion of it is unused
even with firefox and nxclient running, so I'm not running out of memory
or hitting oom killer or anything like that.

Comment 1 Tom Horsley 2011-04-29 18:12:33 UTC
Created attachment 495829 [details]
gdb session backtracing all the threads

Comment 2 Tom Horsley 2011-05-10 14:28:57 UTC
Actually, that description just added is wrong. Firefox is not being run over nxclient, it is merely being run on the same machine that is also running nxclient along side of firefox (which makes the bug far more mysterious since I have no idea how firefox can be aware than I'm running nxclient and act different as a result).

In fact, a firefox run inside the NX session on the remote machine works fine and has none of the mysterious delays or aborts.

Comment 3 Martin Stransky 2012-02-24 15:29:07 UTC
Can you reliably reproduce the crash? With the latest packages (FF10 & Fedora 15/16).

Comment 4 Tom Horsley 2012-02-24 16:14:00 UTC
Nope it stopped crashing and started working again at some point, so whatever
it was must have gotten fixed.

Comment 5 Martin Stransky 2012-02-27 07:26:03 UTC
Okay, thanks.