On 1/27/06, Ian Ruffell <[EMAIL PROTECTED]> wrote: > On Friday 27 January 2006 18:01, Michel Dänzer wrote: > > About a week ago, I switched from pmud to pbbuttonsd, and today I had a > > sudden shutdown again, on AC and with mostly full battery. Note that I > > don't mean to blame pbbuttonsd, ... > > If I'm on battery I've ended up stopping pbbuttonsd if I'm doing something > critical - not had a shutdown if I have not had pbbuttonsd running...
Same here. With feedback from Ben H. the working theory is that it's the PMU is responding incorrectly when it gets lots of queries. Normally pbbuttonsd sends lots of queries to the PMU because that's how it works... it has to poll the PMU frequently. Always has. And of course any time you type or use the trackpad or mouse, those signals have to go somewhere so the PMU knows you're using it. All this explains why it only fails when you're typing and/or using the trackpad or mouse. I have been intending to write a little script to "flood" the PMU with requests so maybe I can RELIABLY get it to fail. As we all know who have had the problem, it's utterly random and unpredictable. If I could get the thing to aways fail within a reasonably short period of time (less than five minutes would be very nice) then I could try different kernel options to try to narrow down exactly what's causing the problem. I was going to write a script that would send commands to query the PMU using the pbbcmd application. I suggested writing a shell script, and a response was received that using perl might be more appropriate. (I'm not knowledgeable in either, so I haven't tackled it yet.) If anyone can suggest a shell or perl script that would "flood" pbbcmd with innocuous requests, maybe we can create a test protocol.
