Play with touchscreen Part II

"Marco Trevisan (Treviño)" mail at 3v1n0.net
Wed Nov 5 05:44:38 CET 2008


Tick Chen wrote:
> Hi List,
>   From the first observation of touch screen, we found the up/down event jitter phenomenon. 
> If you touch the touchscreen lightly, you will get very unstable data,
> and many up/down events in a very short time.
> In the driver it already average the data within a threshold. But the
> extra up/down events will still makes user space libraries think they get
> clicks. This causes annoying problems, and make the GUI hard to use. 
> 
> Thanks for Andy's advise and the inspiration from Dima's patch, we found
> that can just ignore those extra events with a little bit delay of sending up
> envent. If we found it's noise, we can just throw them away. 

Recently I've built the stable kernel image, and I've added also your
touch screen fix, however it - in cooperation with the Dima's soft-tap
patch - creates some problems.

I didn't find this issue until I started playing numpty physics with it,
also if before I got some strange things (like the Qtopia softbar menu
poping-up while writing a SMS), drawing some lines on numptyphysics
shows better the issue: practically as soon as I start drawing a line, I
get two lines, one starting from a random position to the point where I
started drawing, the other is the one I'm currently writing.

How to reproduce:
 - Load numptyphysics (I've tested the 0.2 package in portrait mode)
 - Draw something
 - pop-up the illume keyboard and hit a key (i.e. "r" to reset)
 - Draw another thing
Now you'll see that your new line will be "double", practically it draws
both the line from the past focussed point and the one that you're
currently drawing...


LATE TEST: after writing this mail I've found that using your
touch_test.py in the same way, doesn't draw any wrong line.
So maybe it's just a numptyphysics bug that considers the last touch
before losing the focus (to hit the keyboard) as the starting point of a
new move that ends in the point in which you start drawing.

So is this an application bug, or does it hide something more?

Thanks...

PS: Using an older kernel - without the ticks and Dima's patches - I
can't reproduce this issue at all.

-- 
Treviño's World - Life and Linux
http://www.3v1n0.net/




More information about the devel mailing list