On Fri, Mar 03, 2017 at 06:23:15PM +0100, Cédric Le Goater wrote:
> On 03/03/2017 05:54 PM, Thomas Huth wrote:
> > On 03.03.2017 15:49, Cédric Le Goater wrote:
> >> On 03/03/2017 03:13 PM, Thomas Huth wrote:
> >>> On 03.03.2017 14:37, Cédric Le Goater wrote:
> >>>> Also use an 'Object *' under the sPAPR machine to hold the RTC
> >>>> object.
> >>>
> >>> The change from TYPE_SYS_BUS_DEVICE to TYPE_DEVICE is certainly a good
> >>> idea! But what's the advantage of using Object* instead of DeviceState*
> >>> in sPAPRMachineState ?
> >>
> >> it makes spapr_rtc_create() a little simpler.  
> >>
> >> We could go even further and use a sPAPRRTCState under sPAPRMachineState 
> >> that we  would initialize with object_initialize(). 
> > 
> > I think a sPAPRRTCState* would make more sense here - if you just see an
> > Object* and are not familiar with the code, you wonder what this pointer
> > is all about (and you then have to cast it to something different if you
> > want to do anything with it) ... so IMHO either a DeviceState* or
> > sPAPRRTCState* is the better choice here.
> 
> I think having a non-pointer is a better for the object lifecycle 
> but I don't have a strong opinion on that. We will see what Dave 
> prefers.

You seem to be talking at cross purposes.  AFAICT Thomas is not
talking about the difference between a pointer and a non-pointer, he's
talking about the difference between a generic Object (pointer or
otherwise) and a concrete type.

I agree with Thomas that a specific type is preferable.  I'm not
particularly bothered by the difference between pointer and
non-pointer.

In any case, this has missed the cutoff for qemu-2.9, I'm afraid.

-- 
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson