HAL battery event flood? Help

[Sean]

On Sun, 2007-10-21 at 16:05 -0700, Sean wrote:
> On Sun, 2007-10-21 at 12:55 -0700, Sean wrote: 
> > On Sun, 2007-10-21 at 19:28 +0200, Danny Kukawka wrote:
> > > On Sonntag, 21. Oktober 2007, Sean wrote:
> > > > Hello, Was wondering if anyone could provide some advice on an issue I
> > > > am seeing with Hal on a fresh gutsy install. The symptom is when I start
> > > > my laptop with a second battery installed the hald process consumes like
> > > > 40% of my CPU continuously.  Running hald in with verbose output shows
> > > > that it is getting hammered with battery events, like 100 per second.
> > > > My first reaction is that this is what is causing the CPU usage as I do
> > > > not see this with one battery. What I don't understand is hald the
> > > > 'victim' of something else gone haywire or the cause :)
> > > >
> > > [...]
> > > >
> > > > Any advice is appreciated!
> > > 
> > > I would say the kernel or your hardware is may the problem. Start acpi_listen 
> > > and see how many events you get from your battery. The HAL log show you only 
> > > that HAL get several events per second from your battery. HAL have to react 
> > > on them, what may cause the high load.
> > > 
> > > Danny
> > 
> > Thanks for the quick reply.  I agree that HAL reacting on all of those
> > events seems to be the cause of the high load.  I would have leaned
> > towards hardware or kernel too, except I didn't see this issue when
> > using Hal 0.5.8.1.  To verify, I just forced the install of hal 5.8.1
> > (leaving everything else as-is) and the issue goes away.  Maybe this
> > older version doesn't latch on to the dbus channel that is doing the
> > flooding in the latter version.  
> > 
> > Here is what I've found so far.  If I leave hal 5.9.1 in place, I see
> > the flood of battery events coming from acpi_listen.  However, if I run
> > hal 5.8.1, the output of acpi_listen is what it should be and no floods
> > jsut normal trickle of events.  
> > 
> > Any idea what in HAL could change the behavior of events that you see in
> > acpi_listen?
> > 
> > Thanks again,
> > Sean
> 
> Made some small debugging progress.  I have hal 5.9.1 installed now
> again, and found that deleting /usr/lib/hal/hald-addon-acpi 'fixes' the
> immediate issue.  No more flooding of battery events, no huge CPU draw.
> acpi_listen now shows the normal amount of battery events for when I
> unplug and plug in the ac adapter, etc.  Not sure what functionality I
> lose, but seems to indicates something in this binary is not liking my
> system. Any clues?
> 
> - Sean

Does anyone have any additional troubleshooting I can do?  Not running
the hald-addon-acpi script makes acpi_listen events come across as
normal.  As soon as this is enable the flood (100's per second) of
battery events start.  Is there something this script does to add more
event 'watches' in to acpi?  Just shooting in the dark at this point. I
understand that the root of the problem may not be hald, but it seems
something in there is the trigger.

Many thanks,
- Sean