[PATCH 08/18] Avoid unexpected jumps

[Takashi Iwai]

At Wed, 13 Oct 2010 08:29:24 +0200,
Henrik Rydberg wrote:
> 
> On 10/13/2010 08:01 AM, Takashi Iwai wrote:
> [...]
> 
> > The background of this patch is that the device sometimes doesn't report
> > the finger off from the pad, instead it sends z with a high value in
> > the last packet, then stops sending packets.  Thus the driver still
> > keeps tracking the position as a part of continuous pointer movement
> > when you touch again at a totally different point, which results in a
> > jump.
> 
> 
> I wonder if this is entirely correct... Making sure the device/driver produces
> correct data on the kernel level is the sane way to go. We must be able to trust
> the basic assumptions we put up, or we end up with patches like this one in
> every application.

OTOH, if the judgment of the pointer jump has to be done in a way like
this, user-space is more flexible for different parameters, etc, in
comparison with the kernel.


Takashi