Re: [Qemu-devel] [PATCH 0 of 7] [UPDATE] DisplayState interface change

[Paul Brook]

> >>> - do not use is_active_console, use is_graphic_console instead.
> >>
> >> This is wrong. There may be multiple graphic consoles.
> >
> > I thought we wanted to move to a model in which a DisplayState
> > corresponds to a single QEMUConsole and a single graphic device.
> > This is the reason for Anthony to ask me this update, if I am not
> > mistaken.
>
> I need to look more closely at the patches (and I will tomorrow), but to
> me, the use of is_active_console or is_graphics_console is a red
> herring.  Nothing should never need to know whether it's the "active"
> console.  Such a concept simply shouldn't exist.
>
> I think this is Paul's point about multiple graphics consoles.  If you
> have two graphics consoles, than the concept of active console doesn't
> make very much sense.  Is that what you were thinking Paul?

Something like that, yes.

My understanding was that the plan is to have one DisplayState per image 
source. At that point the device doesn't care whether it's "active" or not 
(inactive DisplayStates will be the equivalent of /dev/null).
If we still have a single DisplayState shared by multiple sources then 
is_graphics_console is not a sufficient check to tell whether the VGA device 
is active.

I'm not sure that removing QEMUConsole without adding multiple DisplayStates 
is feasible.

Paul