← Back to team overview

multi-touch-dev team mailing list archive

Re: Testing M-t without HW

 

On 10/22/2010 04:29 AM, Chase Douglas wrote:
On Fri, 2010-10-22 at 07:42 +1000, Peter Hutterer wrote:
On Thu, Oct 21, 2010 at 12:30:37PM -0400, Chase Douglas wrote:
On Thu, 2010-10-21 at 15:46 +0200, Henrik Sandklef wrote:
The new version of GNU Xnee should have support for MT:

     http://ftp.gnu.org/gnu/xnee/xnee-3.07.tar.gz
     http://ftp.gnu.org/gnu/xnee/xnee-3.07.tar.gz.sig

Cool!

Can you describe how this differs from evtest-capture recording as Peter
wrote here:

http://who-t.blogspot.com/2009/11/evtest-capture-and-replaying-events.html

it's on a different level. evtest-capture records kernel events and replays
them through uinput.

xnee records X events (through the RECORD extension) and then replays them
through (I think the Xtest extension). No kernel involved.

That's what I had thought as well, which confused me. There's no MT
support in X yet, so how does Xnee now have MT support?

Sorry, my bad:

  Xnee can record XI events


We (Canonical) haven't had a chance yet to build a test harness for our
utouch stack, but it's going to be one of my main tasks for the next
couple weeks. I'll be looking into Xnee, evtest-capture, and the test
tool that Henrik has been hacking up.


If Xnee is at the X level it may
be too high up for what we want.

Xnee is at the X level. It records XI events (and can replay them either to X (using XTest, as Peter said) or "beta beta beta" to the kernel using swinput (Linux module)).

So if some X/kernel code, for example a gesture recognition extension, "steals" some possible events (e.g MotionNotify) at the kernel level before creation, Xnee will only record the events created by that code.

.... so I guess Xnee will be pretty useless here

 We need something we can feed events
through as though they came from a real device, with all the quirks that
a real device may offer, like ghosted or dropped touches.

-- Chase






References