Re: Leaky at-spi-registryd update

["Nolan Darilek" <nolan thewordnerd info>]

On 02/17/2009 08:12 AM, Willie Walker wrote:
> Hi Nolan:
>
> I wonder if some sort of popup attack might be happening in Firefox 
> and it's causing degradation in at-spi-registryd (just a guess).  If 
> you kill all firefox-related processes on your machine when the spike 
> happens, do things end up getting back to normal?
No. That was what I tried for a long time before I ran system-monitor 
and saw registryd at 300M memory use. From an earlier message:

It's as if the memory use just hangs around beneath 1%, then as soon as 
it breaks beyond that, it's entirely uncontained and nothing can stop 
it. Killing apps can make it shrink, but never back to what it was, and 
it soon starts growing again. And I can't for the life of me come up 
with one task that reliably gets it started.

So no matter what I kill, once it's spiked it's gone and nothing brings 
it back. Killing everything and returning to an idle desktop brings 
memory use down a bit, but it never even gets near 1% again.

> In addition, try running Orca from a console and redirecting the 
> output to a file: orca > orca.out 2>&1
>
> When you notice the problem crop up, give Firefox focus and press 
> Orca+Ctrl+Alt+PageDown.  This tells Orca to dump the hierarchy of the 
> current app to the console.  The output will be caught in orca.out and 
> might be informative.  Note that you might want to check the contents 
> of orca.out for any private information before sending it to someone.
Good idea, but that too reveals nothing interesting, not even with other 
apps. Everything shown is just what I'd expect to see. :(

The power supply in that box died last night. Since I'm going to have to 
open it anyway, I'll upgrade its 512 megs of RAM to a gig and see if 
that helps. And since no one else on multiple lists seems to have seen 
this behavior, I'm going to call it bad hardware until proven otherwise.

Thanks for the suggestions.
> Nolan Darilek wrote:
> > On 02/16/2009 12:04 PM, Willie Walker wrote:
> > > Hi Nolan:
> > >
> > > If one of the suspects is FF, I wonder if we might be able to focus 
> > > on it a little more.  For example, I wonder if it might be possible 
> > > that it is some of the web sites you typically visit and/or maybe 
> > > some usage patterns (e.g., frequently opening/closing multiple FF 
> > > tabs or windows).
> > That's one of the many things I've thought of. Unfortunately, try 
> > though I might, I just can't seem to come up with a correlation.
> >
> > Last night I noticed at-spi-registryd use start climbing when I was 
> > using the totem plugin to listen to a streamed MP3, thought I might 
> > have finally figured out the cause, only the memory use kept spiking 
> > after I'd closed the tab, and in fact, kept growing even when I did 
> > nothing with FF. Just switching between apps could cause its memory 
> > use to grow.
> >
> > It's as if the memory use just hangs around beneath 1%, then as soon 
> > as it breaks beyond that, it's entirely uncontained and nothing can 
> > stop it. Killing apps can make it shrink, but never back to what it 
> > was, and it soon starts growing again. And I can't for the life of me 
> > come up with one task that reliably gets it started.
> >
> > > While still respecting your privacy, are you able to share some of 
> > > the typical pages you visit in a day?
> > Unfortunately, no. Most of my repeatedly-visited sites are 
> > password-protected--LJ, a work-related site, etc. Beyond that it is 
> > mostly random. The only site I regularly visit that isn't 
> > password-protected is news.thewordnerd.info, my feedreader, but I 
> > don't notice at-spi-registryd spikes correlating with visiting that.