multi-touch-dev team mailing list archive

[Rafi Rubin <rafi@xxxxxxxxxxxxxx>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 04/07/10 13:43, Duncan M. McGreggor wrote:
>
> On Apr 1, 2010, at 3:48 PM, Stéphane Chatty wrote:
>
>>
>> Le 1 avr. 10 à 00:09, Rafi wrote:
>>
>>> Just an FYI, Stéphane and I have been seeing an odd behavior where
>>> the n-trig
>>> hardware stops sending events.  The devices look like they should be
>>> functioning
>>> normally and just show no sign of activity (no events coming down
>>> from the usb
>>> device).
>>
>> I think I have a working solution. It has to do with a message
>> starting with 0x0a that I had observed the PC sending to the device
>> without being able to reproduce it correctly. Considering it as a
>> feature report and sending it as such with usbhid_submit_report seems
>> to do the trick.
>>
>> I'll make a few more tests and send a patch to Rafi for further testing.
>
> Any progress on this?
>
> d

Yes, Stephane submitted a patch yesterday.  In testing, I found two problems,
particularly with really old firmware.  The fixes are trivial, I just need to
test more firmware versions to make sure my fixes are safe.

Stephane's patch sends an activation message to the hardware.  I'm still not
clear why the things are so intermittent about needing that.  I also thought
that wake up would enable MT on some versions of the firmware, but I haven't
found any where that happens (I should ask my lab mate to test on his XT2).  I
have checked a fair number of firmwares (an annoying process that seems to take
10-30 minutes, if all goes well, per firmware).


On top of that I noticed, the single touch firmwares are emitting DOUBLETAP but
not TOUCH.  I thought I'd checked that more carefully.  The distinction is that
wacom uses DOUBLETAP and evdev ignores it.  This can be patched in the user
space side (either change the udev rules to single->wacom and multi->evdev, or
add DOUBLETAP in one spot to evdev), if that's preferable to trying to sneak in
another kernel change.

For at least one version of the firmware (shipped with the XP driver), the
hardware uses a different meaning for "CONFIDENCE" events.  There are three
events which are generally redundant with later versions of the firmware, and we
have been using CONFIDENCE to indicate if a finger is on the screen or not.  It
would seem that the older firmwares just uses TIPSWITCH and hold CONFIDENCE on.
 After I'm done testing the full range of firmwares at my disposal, I'll send in
a patch to switch that particular mapping for single touch.


My plan is to finish my testing tonight and hopefully keep these two changes
relatively close the Stephane's patch.


In other news, I've started writing a small tool to test touchscreen behavior.
I will post it once I finish the first couple tests and reports.  I've been
using drawing tools to verify single touch and pen are working, and to see how
noisy they are, but for multitouch, I really want to have a more systematic
tool.  Something that will be able to tell us how reliable contact and tracking
are, particularly as we run through a number of multi-finger actions.  Such a
tool will be helpful when developing and comparing drivers, not to mention
application dev.

Rafi
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAku8zaMACgkQwuRiAT9o609VgwCeKFMM7cbdyiyi+BKyP0BWFUSW
pHMAoN9DyjwrkHKO6+9Ymix7RqCm51ZH
=Z/ts
-----END PGP SIGNATURE-----