On 13 May 2002 10:37:22, [EMAIL PROTECTED] wrote:
> The touchscreen is *NOT* a mouse.  It should definately *NOT*
> be called /dev/mouse....  It does not return motion: it returns
> position, in a funky format that we intend to abolish...

I hope this isn't a general condemnation of touch screens
appearing via /dev/mouse.

The Semtech ScreenCoder used in the ViewSonic ViewPad 1000
rather tightly couples the interaction of the mouse and touch
screen.  Upon startup, the touch screen acts as the controller
for an external mouse, and generates its own relative motion
packets, which it merges with those from the external mouse.
In this mode, it's obvious that /dev/mouse (actually /dev/psaux)
is appropriate.

Upon receipt of a special knocking sequence, the controller
shifts to generating absolute mode packets for the touch
screen, while preserving relative motion packets from the mouse.

On April 11, I submitted a set of patches, one for the 4.2.0
release and one for a CVS extraction done April 8, which
generates this knocking sequence and then disambiguates the
packets arriving via the interface.  So if one wishes to
use absolute mode, one simply declares the protocol to be
"ScreenCoderPS/2" instead of "PS/2".  (One can also use
"ScreenCoderIMPS/2" and "ScreenCoderEXPPS/2" instead of
"IMPS/2" and "EXPPS/2".)

(I'm investigating two enhancements: the first to better
control calibration and the second to switch between
relative and absolute position mode at run-time.  I'll post
progress reports later.)

Of course if Jim is writing about a touch screen which has
a distinct data path to the the computer, what I've written
is only remotely tangential to the topic.

Randolph Bentson
[EMAIL PROTECTED]
[EMAIL PROTECTED]

_______________________________________________
Xpert mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/xpert

Reply via email to