Re: [Qemu-devel] [PULL v2 8/9] input-linux: refine mouse detection

[Gerd Hoffmann]

On Do, 2016-04-14 at 09:42 +0200, Ladi Prosek wrote:
> On Wed, Apr 13, 2016 at 5:45 PM, Gerd Hoffmann <address@hidden> wrote:
> > Read absolute and relative axis information, only classify
> > devices as mouse/tablet in case the x axis is present.
> 
> I, too, had to come up with a heuristic to classify input devices in
> my guest driver and what I ended up with is different.
> 
> For example my Dell keyboard has two endpoints, one with a bunch of
> keys and LEDs, so it would be classified as a keyboard. The second one
> with special keys (KEY_MUTE, KEY_WWW, KEY_BACK, ..) *and* with a mouse
> (REL_X, REL_Y, REL_WHEEL, BTN_LEFT, ...). The reason for this are the
> zoom in/out buttons. Pressing them generates Ctrl down on the first
> endpoint and mouse wheel up/down on the second one. Releasing them
> then translates to Ctrl up. Crazy.

So they are building os-specific hotkeys into the hardware.  Crazy
indeed.

It's not obvious though why the second function has keyboard keys.
Possibly they want the first function look like a pretty standard
keyboard without any extra fluff, to sidestep compatibility issues with
old software not expecting that.

> So I wouldn't use exclusive OR when classifying because there are
> combo devices out there. Maybe anything with an EV_KEY (minus BTN_*)
> would be a keyboard?

Right now each device type has its own callback function, we have to
reorganize that to support a device being classified as both mouse and
keyboard.  Worth considering, but not 2.6 material.

There surely is more room for improvements, not only in input-linux but
in the input system in general and in the virtual input devices, for
example to support all those extra keys on multimedia keyboards.

cheers,
  Gerd