[PATCH] Introduce keyboard grabbing protocol

[Olivier Fourdan]

Hey,

----- Original Message -----
> On 30 August 2016 at 14:05, Olivier Fourdan <ofourdan at redhat.com> wrote:
> >> Xwayland should probably use a private protocol, like EGL, ideally
> >> completely hidden and pid-restricted.
> >
> > That's the point, I initially thought of a private protocol, but then
> > realized it could be useful outside of the Xwayland use case as well.
> >
> > If not, I'd be happy to drop this patch and return to the private approach
> > :)
> >
> > Speaking of which, that was the idea behind those patches:
> >
> > https://patchwork.freedesktop.org/patch/105319/
> > https://patchwork.freedesktop.org/patch/103815/
> > https://patchwork.freedesktop.org/patch/105421/
> 
> Hmmmm. I really, desperately, want to avoid keyboard grabs as much as
> possible. But I guess Alt-Tab in a VM is the one usecase which can't
> route around this. Out of interest, do you know if that works on
> VMWare/VirtualBox/etc on OS X and Windows? If they don't, then we
> don't have to either, and we can just make this private to XWayland.
> If they do, then I guess we probably need to follow suit as well.

According to VMWare docs available online, it has this feature on Windows, it uses a special key combo to release the mouse and keyboard input when needed.

And according to Virtualbox manual, Virtualbox has that too: https://www.virtualbox.org/manual/ch01.html#keyb_mouse_normal

> If we do though, I'd prefer to see two protocols: one for native
> clients which demands a matching wl_seat serial, and another for
> XWayland which doesn't have that restriction.

This has an additional benefit, we can work on a "private" protocol for Xwayland (to keep backward compatibility with existing X11 applications) and decide later to do the same for native clients, if needed.
 
> Note that this is basically just about the concept; I haven't looked
> at the exact detail of the wording yet.

Cheers,
Olivier