timo at drick.de
Mon Jul 14 16:15:04 CEST 2008
Ok i know my reports about the problem are not detailed engouh to localize
But i would like to publish one more finding.
The Problem described in the last mail do only occure if ONE PROCESS has
both accels open.
The process only read from one event!
i do not tested with different threads but this would complicate my
apllication and is not an good option.
Also python test code produce this problem if the code have both events
So my first assumption is wrong that short intervals of reading occure this
Please could anyone reproduce this?
On Sun, 13 Jul 2008 14:42:39 +0200, Timo Drick <timo at drick.de> wrote:
> Ok i have some new findings again.
> I write a c application to read the accelerometers and some times the
> fread command blocks after reading some data. Some times it do not stop
> and read forever. If the fread commands blocks then no data could be
> read from the /dev/input/event also from an other process. I use a
> normal blocking read command. But it looks like this occurse if the
> fread is called too short intervals. So this could be an bug in the
> accelerometer driver. (This problem do not occure with hexdump. I think
> hexdump is too slow)
> I think neod will also read the accelerometers and this problem occurse.
> So i have only the event that does not block.
> regards Timo
More information about the openmoko-devel