Re: [Qemu-devel] [PATCH 1 of 2] [UPDATE] DisplayState interface change

[Anthony Liguori <anthony@codemonkey.ws>]

Paul Brook wrote:
>> There are really two parts to this patch.  One part is a completely
>> mechanical conversion of things like ds->linesize to
>> ds->surface->linesize.  The second is the guys of the API change.
>>
>> If you introduce accessors to access things like linesize from
>> DisplayState, so something like:
>>
>> int ds_get_linesize(DisplayState *ds) {
>>      return ds->linesize;
>> }
>>     
>
> I think this sort of thing should be going via the console structures.
> i.e. vga emulation deals entirely with a QEMUConsole.  sdl/vnc deal entirely 
> with a DisplayState, and qemu mediates in between.
>   

I'm not sure.  You can't really hide the fact that there are two VGA 
cards to the SDL/VNC backend.  I think a model where each emulated VGA 
device had a DisplayState, and each DisplayState had an SDL/VNC backend 
connected to it could work.

That way, you could have two VNC servers for each of the two VGA 
devices.  SDL is unworkable for multiple monitors AFAIK so it's kind of 
a moot point.

Where you thinking that we would try to have a QEMUConsole with each VGA 
card and then multiplex the QEMUConsoles on a single DisplayState?  
Couldn't we do the same by multiplexing two DisplayState's onto one 
DisplayState?

I don't think required interface for QEMUConsole would be much different 
than DisplayState so it's unclear to me if two structures are really needed.

Anyway, Stefano, please add accessors in some form to help with review 
of this patch.  If we decide we need a QEMUConsole abstraction, it's 
easy enough to update the patch to modify some of the accessors.

Regards,

Anthony Liguori

> Paul
>