Xinput button mapping and middle button emulation

[Peter Hutterer]

On Thu, Oct 13, 2016 at 01:31:06PM +0200, Tobia Conforto wrote:
> Yes, I probably misused a couple of terms.
> 
> As you say, the middle button emulation happens in the evdev Xorg
> driver, in fact it's controlled by the "Emulate3Buttons" option. The
> kernel simply generates the MSC_SCAN and BTN_* codes that I pasted.
> 
> So, as far as I can tell, there are two ways to get what I need (which
> is middle button emulation A+B outputting button number 2, while
> having the physical button C outputting some other number instead of
> 2.)
> 
> Option 1.
> Have the Xorg driver (evdev or another?) remap button C from 2 to 8,
> while still providing an emulation for A+B to number 2. This is
> clearly something that an Xorg driver could do, but I don't know
> whether there is an existing driver that can be configured to do so,
> or if I would need to patch / write one. (Hence this thread.)
> 
> By the way, this could arguably be considered a bug in the
> Emulate3Buttons feature: if I remap button 2 to 8, because it's
> physically in a different position on the trackball, the middle button
> emulation should still output number 2, which is what is needed to
> paste stuff.

hmm, yeah, after looking at the code the ButtonMapping option merely sets
the X logical button mapping on startup, the one that is changed by all the
x client tools. There is no internal button mapping.

I think what you could do is map MB emulation to button 10 (that is
driver-internal) and then init the button mapping so that button 10 is
logical button 2 and button 2 is logical button 8. 

> Option 2.
> Change the kernel mapping between MSC_SCAN and BTN_* codes, so that
> button C won't output BTN_MIDDLE in the first place. Again, this is
> something the kernel can clearly do, but I don't know where that
> mapping is stored, and whether it can be changed without patching the
> kernel or writing an input driver.

on most usb devices it's handled by the HID code, so I don't think you can
arbitrarily remap without extra efforts. never tried though.

Cheers,
   Peter