Configure touchpad for touchscreen driver

[Peter Hutterer]

On Mon, May 08, 2017 at 03:33:33AM +0200, Martin Kepplinger wrote:
> Am 2017-05-08 um 02:40 schrieb Peter Hutterer:
> > On Mon, May 08, 2017 at 10:39:32AM +1000, Peter Hutterer wrote:
> >> On Fri, May 05, 2017 at 05:30:51PM +0200, Martin Kepplinger wrote:
> >>> Hi,
> >>>
> >>> I'm doing a touchscreen xorg input driver. 
> >>
> >> First question: why? :)
> >> we generally discourage people from doing that, instead it's better to write
> >> a kernel driver and use the existing evdev/libinput drivers.
> 
> I'm reviving / rewriting xf86-input-tslib, the main sources now living
> at https://github.com/merge/xf86-input-tslib
> 
> With tslib, I *am* somewhat pushing towards having evdev kernel drivers.
> (Sadly, there are *real* userspace drivers in tslib because there's no
> proper driver, but I don't have the hardware to write them; even more
> sadly, there are companies that still ship tslib drivers in their BSPs
> instead of kernel drivers) but that's a different thing. tslib itself
> lives at http://tslib.org
> 
> So the real reason I need tslib is the filter stack and easy to use
> graphical calibration (and the little dependencies and portability are a
> bonus too).
> 
> All these "drivers" :) To me, xf86-input-tslib is no driver. tslib (or
> the libts library) isn't either. It's either a userspace driver as tslib
> module or a proper kernel driver. But, oh well :)
> 
> 
> I'm getting off track here, but libinput is really interesting in that
> it does gesture recognition too. Do you somehow sync that work with gtk+
> or others who do that too?

libinput only does gesture recognition on touchpads where toolkits can't
meaningfully provide gestures. it doesn't do it on touchscreens where the
toolkits have the better context, so for touchscreens we merely forward the
events as-is.

Cheers,
   Peter
> 
> 
> >>
> >>> For testing purposes, I'd like to configure my laptop's touchpad for the
> >>> driver to use. It's totally comparable and I've done it for other stuff
> >>> for rough testing. I append the evtest output below. Should be fine.
> >>>
> >>> I'm unsure how to override all other touchpad configs in my system.
> >>>
> >>>  * is it enough to say "MatchIsTouchpad" instead of touchscreen? There's
> >>> obviously only one touchpad. Ideally I think I should omit
> >>> "MatchIsTouchpad" and say only "Device=/dev/input/event1" or whatever
> >>> number it is? Would that be possible?
> >>
> >> MatchIsTouchpad maps to the udev property ID_INPUT_TOUCHPAD, your homework
> >> is to figure out what MatchIsTouchscreen maps to ;)
> >> so doing that would be enough, but you could also add a
> >> MatchProduct "ETPS/2 Elantech Touchpad" to only match on that touchpad.
> >>
> >> Don't match on device nodes as they can change during reboots.
> >>
> >>>  * My driver config installs to /usr/local/share/X11/... Will this
> >>> *ever* override /usr/share/X11 configs? Even if I call it
> >>> "00-mydriver.conf" in /usr/local/
> >>
> >> the server is usually configured to look into /usr/share/X11/ and /etc/X11,
> >> but not elsewhere. So anything you put in /usr/local is ignored. The start
> >> of the xorg.log should list where the server is looking for configs.
> >>
> >> for local configurations, you're supposed to put them into /etc/ anyway,
> >> distribution-provided config snippets go into /usr/local.
> > 
> > make that:
> >     distribution-provided config snippets go into /usr/share.
> 
> Sure. thanks Peter for getting back at me. I actually had figured it out
> and have a pretty nice dev setup.
> 
> > 
> > Cheers,
> >    Peter
> > 
>