On Thu, 2008-07-10 at 10:56 +0100, Andy Green wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Somebody in the thread at some point said: > | On the OM2007.2 images it is the neod which opens the accelerometers. No > | reason for doing so, even more as neod starts to eat 50% CPU, because > | the am's send both 100 events per second... > | > | If anybody is interested, I recompiled the neod without this and without > | the click-sound. Drop me a mail. > | > | No more probs with the am's and everything a lot faster. > > Well I am glad to hear it can be made stable, because I did believe I > fixed the instability last patch set.
Just for adding more information: I used one of the lastest builds from the openmoko repository and the accelerometer was running for two days using a slightly modified version of Thomas Wood's gta02.py which reads them every 100ms. Usually it ran for a couple of minutes before getting stuck reading /dev/input/event3. Once gta02.py was killed the output of cat /dev/input/event3 remained empty until a reboot. Now it works. > However I don't get why neod sucking on the input device too makes > trouble (other than the CPU load or whatever). The input layer in Linux > takes care about copying the input content around consistently so > everyone with the input device open gets one copy of what's happening. > Ie, the accel driver is not aware that there are multiple apps > listening. I guess it is a bug in the accel driver somehow still, but a > much less worrying one that the SPI sharing difficulties until now. :-)