On Fri, Feb 24, 2012 at 10:08:35AM +0100, Henrik Rydberg wrote:

quoted

quoted

The most striking effect is the ability to better retain a
drag. Although the statement was made in light of possible
(1-strong,1-weak) packets, it should help in the 2-weak case too.

The bigger question is why is this needed in kernel. The original
hysteresis with BTN_TOUCH was done for sole benefit of mousedev so that
we could allow somewhat better transition from standard PS/2 mode into
native Synaptics mode with absolute coordinates at time when barely
anyone had Synaptics X driver installed. This was, what, 10 years ago?

The semi-mt behavior is obviously a special case, where userspace
relies on the reported number of fingers to transition between one
touch and two touches. There is no pressure information sent to
userspace in this case, so the situation is in fact quite similar to
the ancient mousedev situation.

We still report ABS_PRESSURE but maybe we should report ABS_MT_PRESSURE
as well?

-- 
Dmitry