Re: 2.6.24: false double-clicks from USB mouse

["Ray Lee" <ray-lk@madrabbit.org>]

On Dec 7, 2007 10:32 AM, Dmitry Torokhov <dmitry.torokhov@gmail.com> wrote:
> On Dec 7, 2007 12:59 PM, Ray Lee <ray-lk@madrabbit.org> wrote:
> > On Dec 2, 2007 2:07 PM, Jiri Kosina <jkosina@suse.cz> wrote:
> > > Thanks. To sum up this longish thread:
> > >
> > > - Mark seems to be able to reproduce the problem quite easily; I was not
> > >   successful reproducing this no matter how hard I tried, and I also
> > >   didn't receive any similar bugreports from anyone else
> > > - we are currently waiting for Mark to provide HID_DEBUG (and preferably
> > >   also usbmon) output from the situation where multiple clicks are
> > >   being generated incorrectly
> >
> > Okay, I'm having a problem with 2.6.24-rc4 that I didn't with 2.6.23-rc6.
> >
> > When the problem hits, my mouse (synaptics touchpad) is randomly
> > moving about and clicking things whenever I have my finger on the
> > trackpad. This behavior will persists for five to ten minutes (is
> > happening now), and seems to be triggered by load (watching online
> > videos in flash player, for example -- this matches Mark's experience
> > with a make -j4).
> >
> > I *think* the touchpad is connected via ps2, which looks like it'd
> > clear usb of any blame. It's /dev/input/event6 at any rate.
> >
> > This has been happening about once a day, and seems to start because
> > of high CPU usage. ONce the CPU is idle, it still happens though, so
> > I'm not sure why it clears up.
> >
> > Any clues for where I should hunt from here?
> >
>
> Can you try replacing drivers/input/ and include/linux/input.h from
> 2.6.23-rc6 and see if it works or not? That should give us idea if
> input locking changes are to blame.

I'm hitting a number of build errors, trying to compile a hybrid
between the two.

git checkout v2.6.24-rc4
git checkout v2.6.23 include/linux/input.h drivers/input
make
# fails with kconfig issues
git checkout v2.6.24-rc4 drivers/input/misc/Kconfig
make
# fails
...adding to include/linux/input.h:
#define KEY_CNT                   (KEY_MAX+1)

takes the build a little farther until it dies on more missing
definitions. Doing a checkout of 2.6.24-rc4's include/linux/input.h
leads to different build errors.

Looking at git log for drivers/input, it appears the patches you're
talking about are the ones starting with 8006479c9b? If so, perhaps I
should just build the revision right before that commit and test that?

Ray