Re: Failing property setters + hardwired devices + -global = a bad day

From: Paolo Bonzini <pbonzini@redhat.com>
To: Markus Armbruster <armbru@redhat.com>, qemu-devel@nongnu.org
Cc: "Kevin Wolf" <kwolf@redhat.com>,
	"Peter Maydell" <peter.maydell@linaro.org>,
	"Daniel P. Berrangé" <berrange@redhat.com>,
	"Eduardo Habkost" <ehabkost@redhat.com>,
	"Jason Wang" <jasowang@redhat.com>,
	"Max Reitz" <mreitz@redhat.com>,
	"Marc-André Lureau" <marcandre.lureau@redhat.com>
Subject: Re: Failing property setters + hardwired devices + -global = a bad day
Date: Wed, 29 Apr 2020 17:39:20 +0200	[thread overview]
Message-ID: <1915d102-9528-9c31-8a8c-a84f7842fbe9@redhat.com> (raw)
In-Reply-To: <87mu6uia5i.fsf@dusky.pond.sub.org>

On 29/04/20 17:28, Markus Armbruster wrote:
> When an object gets created, its memory is zeroed, and then any class
> properties with an ->init() are initialized with
> 
>     obj->init(obj, prop);
> 
> We have just one ->init():
> 
>     void object_property_init_defval(Object *obj, ObjectProperty *prop)
> 
> It initializes from ->defval using ->set().
> 
> Aside: feels overengineered, but let's move on.
> 
> For TYPE_DEVICE objects, "static" properties get initialized the same.
> 
> Aside: if I know what "static" means, I'll be hanged.

Originally these were the only properties that were part of the class
rather than the object (so, not dynamic --> static).

> I'd love to deprecate -global wholesale, but we can't as long as we
> don't have better means to configure onboard devices.  Can we deprecate
> its use with backend properties at least?

I wouldn't mind deprecating -global wholesale, leaving the global/compat
props code only for internal usage.

Paolo