Multitouch type A protocol clarifications

[Daniel Drake <dsd@laptop.org>]

Hi,

I'm working with a prototype for the XO-4 touchscreen laptop (with
both trackpad + touchscreen). The touchscreen in this test unit is a
raydium device, using this driver:
http://dev.laptop.org/git/olpc-kernel/plain/drivers/input/touchscreen/raydium_ts.c?h=arm-3.0-wip

We understand that it is a type A device, since the hardware just
sends a list of touch points without tracking them. If I touch with
finger 1, it starts sending that as the first list entry. Then I touch
with finger 2, it starts reporting that in the second list entry. Then
I release finger 1, and finger 2 jumps into list entry 1.

The driver at the above link is wrong/confused since it reports
tracking IDs, and reading
http://www.kernel.org/doc/Documentation/input/multi-touch-protocol.txt
I see that tracking IDs are only for type B devices. Also, in the
following tests, I removed the pressure reporting, just to make things
more simplistic.

Removing also the BTN_TOUCH reporting, I end up with what is suggested
in multi-touch-protocol.txt as the example for type A. As reported
with evtest, two fingers are placed:

type 3 (EV_ABS), code 53 (ABS_MT_POSITION_X), value 678
type 3 (EV_ABS), code 54 (ABS_MT_POSITION_Y), value 498
++++++++++++++ SYN_MT_REPORT ++++++++++++
type 3 (EV_ABS), code 53 (ABS_MT_POSITION_X), value 215
type 3 (EV_ABS), code 54 (ABS_MT_POSITION_Y), value 288
++++++++++++++ SYN_MT_REPORT ++++++++++++
-------------- SYN_REPORT ------------

First finger is released:

type 3 (EV_ABS), code 53 (ABS_MT_POSITION_X), value 215
type 3 (EV_ABS), code 54 (ABS_MT_POSITION_Y), value 288
++++++++++++++ SYN_MT_REPORT ++++++++++++
-------------- SYN_REPORT ------------

The other finger is released:

++++++++++++++ SYN_MT_REPORT ++++++++++++
-------------- SYN_REPORT ------------



Or, to simplify things, just a single finger quick touch and release:

type 3 (EV_ABS), code 53 (ABS_MT_POSITION_X), value 487
type 3 (EV_ABS), code 54 (ABS_MT_POSITION_Y), value 343
++++++++++++++ SYN_MT_REPORT ++++++++++++
-------------- SYN_REPORT ------------
++++++++++++++ SYN_MT_REPORT ++++++++++++
-------------- SYN_REPORT ------------



That represents all of the events - I'm not reporting BTN_TOUCH,
pressure, tracking ID, ... (neither the event nor the capability). And
I think I'm following whats suggested in the documentation. But under
this setup, X (server 1.13, xf86-input-evdev-2.7.3, mtdev-1.1.2)
doesn't even consider the device as an input device (doesn't appear in
output of "xinput" command), and therefore the touchscreen is
unresponsive under X. I have to readd this line to get X to pick it
up:

	input->keybit[BIT_WORD(BTN_TOUCH)] = BIT_MASK(BTN_TOUCH);

(note that is just advertising the BTN_TOUCH capability - not actually
reporting any events from it)

Is that an X bug or is there room for improvement in the documentation
here? Right now the documentation makes me think that BTN_TOUCH is
optional.

Should I be reporting the BTN_TOUCH capability? Should I be sending
BTN_TOUCH events to report if there are any active touch points or
not?



At this point, whether or not I report BTN_TOUCH in the event stream,
I'm facing the problem that X (as reported by "xinput test-xi2") never
reports a TouchEnd event, not even for a simple touch. And, after
touching the screen for the first time, if I move the mouse cursor
around the screen with the touchpad, X reports that the first mouse
button is held down, when its not (no active touch, touchpad's mouse
button is not pressed).

Is that an X problem, or is there some further events I need to be passing?

Thanks,
Daniel