Re: [gtk-vnc-devel] PATCH: make keyboard & pointer grabs configurable

[Anthony Liguori <anthony codemonkey ws>]

Daniel P. Berrange wrote:
> On Mon, Jul 09, 2007 at 09:08:22AM -0500, Anthony Liguori wrote:
>   
> > Daniel P. Berrange wrote:
> >     
> > > > One of the obvious reasons for that is that there is one "release grab"
> > > > hotkey only.
> > > >    
> > > >         
> > > Two primary scenarios I have
> > >
> > >  - Grab both  keyboard & mouse at same time - upon first click - in this
> > >    case there's  only a single ungrab hotkey needed
> > >  - Grab keyboard only when mouse is over the VNC widget - there's no
> > >    ungrab hotkey needed at all - simply move the mouse outside the
> > >    widget
> > >  
> > >       
> > So this should be an option within gtk-vnc instead of being something 
> > implemented externally to it.  I agree that you really don't want 
> > keyboard grab exposed as a separate option except for this one very 
> > special case.
> >     
>
> The trouble is there's no way for the VNC widget to determine which grab
> mode it needs to operate in.

Ignore my earlier comments.  I didn't read your patch thoroughly enough 
as it's doing what I think should be done.  Should make sure to get 
caffeine in my system before looking at patches :-)

Regards,

Anthony Liguori

>  Although you added support for QEMU to run
> in relative vs  absolute mode, that is not in any currently supported
> releases of any virt product. So all VNC servers are running in absolute
> mode on the wire, and may or may not be converting into relative mode
> before passing data off to the guest. 
>
> Virt-manager can (partially) tell whether the remote server is at least
> capable of passing absolute pointer co-ords to the guest (eg looking at
> the guest config to see if a USB tablet is configured, or seeing if the
> paravirt framebuffer is in use). Even then it can't tell whether the
> guest is actually configured to use the device in absolute mode - the
> X server may be accessing the tablet via /dev/input/mice & thus still get
> the relative co-ords. It all really rather sucks, so I don't see any
> way to reliably determine whether to grab keyboard only, or mouse+keyboard
> from within the widget.
>
> Dan.
>
>