Re: [Qemu-devel] [PATCH 03/14] qdev: add qdev_add_properties

[Anthony Liguori]

On 05/01/2012 03:43 PM, Andreas Färber wrote:
> Am 01.05.2012 22:37, schrieb Anthony Liguori:
> > On 05/01/2012 02:05 PM, Andreas Färber wrote:
> > > Am 01.05.2012 20:18, schrieb Anthony Liguori:
> > > > This allows a base class to easily add properties.
> > > >
> > > > Signed-off-by: Anthony Liguori<address@hidden>
> > >
> > > Implementation looks okay but /me not so happy with it: This conflicts
> > > with the move of the qdev static property infrastructure from
> > > DeviceState to Object.
> > >
> > > Consider rebasing this onto part of Paolo's series and call it
> > > object_add_properties?
> >
> > Eh?  There's nothing object_ about these properties and there's no way
> > I'm willing to put legacy properties in object...
> >
> > So I'm not quite sure what you're suggesting.
>
> You just suggested to Peter using qdev_add_properties() in new QOM ARM
> classes of his.
>
> I'd rather not propagate using qdev_* functions in new QOM code because
> either it remains forever or renaming becomes another touch-all series.
>
> In Paolo's series the Property* concept is moved from qdev to QOM; thus
> if it's in Object we usually use an object_ prefix.
> Not too long ago you were willing to merge the large part of Paolo's
> series which included this code movement, so if you don't want that
> after all you should communicate that openly as a reply there. :)

Legacy properties != static properties.

qdev_add_properties adds both legacy and static properties.  I'm happy to put 
static properties into object but not legacy properties.

So qdev_add_properties is going to stick around even after Paolo's changes.

Base classes are free to call object_property_add_static directly and avoid 
qdev_add_properties entirely but we need the later for backwards compat.

Regards,

Anthony Liguori

> Andreas