Re: Thoughts on an xinput extension?

[Jason Miller]

On Tue, 2 Jun 2020 17:41:45 -0700, Jason Miller <jason@milr.com> wrote:
> I was frustrated by the fact that since the pointer-grab is global, you can't
> inject pointer events with e.g. xdotool from button maps.
> 
> I hacked-up a demo that allows grabbing via xinput, which only grabs a 
> specific
> device which means pointer events can still be sent to the window.
> 
> My current syntax is only tested in the guile format; it allows specifying an
> xinput device by name as a modifier and looks like:
> 
> ("d:ELECOM ELECOM TrackBall Mouse" "b:9")
> 
> Which will map to button 9 of the first xinput device named "d:ELECOM ELECOM 
> TrackBall Mouse" 
> 
> Is this something that interests anyone?  If so I can make a more polished 
> patch.
> 
> Still TBD:
> 
> 1. I haven't looked at the non-guile parser at all; would introducing spaces 
> in
> modifier names cause a grammar ambiguity?  If not we would need a way of
> quoting xinput device names, since spaces are relatively common.
> 
> 2. Should this be just for pointers?  As implemented it currently pointer-only
> as I haven't had any issues with injecting keystrokes from keygrabs.  Maybe in
> corner-cases from chording it can happen?  Anyone run into keyboard issues 
> with
> grabs affecting your ability to send keystrokes?  If we do allow keyboards as
> well, then the syntax will need tweaking as many programmable mice may 
> register
> both a pointer and a keyboard with the same name.
> (maybe p:<devname> and k:<devname> ?)
> 
> -Jason