On Mon, 17 Dec 2007 20:18:23 +0100
Joerg Sonnenberger <joerg@britannica.bec.de> wrote:

> On Mon, Dec 17, 2007 at 12:22:50PM -0500, Steven M. Bellovin wrote:
> > On Mon, 17 Dec 2007 13:50:11 +0100
> > Joerg Sonnenberger <joerg@britannica.bec.de> wrote:
> > 
> > > On Sat, Dec 15, 2007 at 06:42:24PM +0000, Steven M. Bellovin
> > > wrote:
> > > > I think I'm still seeing some delays, though I haven't monitored
> > > > that closely.  I did run 'envstat -i 5' for about 45 minutes,
> > > > with no failures of battery charge level or state.
> > > 
> > > Battery reading can take a long time. That was part of the
> > > problem.
> > 
> > How long should be considered normal?
> 
> With broken ECs, it was taking over a second to send the interrupt
> (the original problem here). Even without that, I have times in the
> hundred milisecond range already. Note that T42 and R52 had a bunch
> of very broken EC firmware releases. I could kill the power button
> completely by spinning on the EC :-(

Wonderful (I'm typing this on a T42....).  That said, hundreds of
millseconds doesn't account for the near-20-second times I've seen.
> 
> > > envstat will normally take care of that, but the data will be
> > > slightly stale.
> > > 
> > How stale? 
> 
> IIRC it updates it between 10 and 15sec, normally.
> 
Certainly acceptable for something like that.

 



		--Steve Bellovin, http://www.cs.columbia.edu/~smb