Re: [Qemu-devel] [PATCH 1/2] object: recommend a few type check macros

[Peter Maydell]

On 21 September 2018 at 12:13, Marc-André Lureau
<address@hidden> wrote:
> I sometime regret that we have to resort to long
> object{_class}_dynamic_cast() calls instead of having a shorter and
> more readable macros available, similar to the one recommended by
> GObject (https://developer.gnome.org/gobject/stable/gtype-conventions.html).
>
> Signed-off-by: Marc-André Lureau <address@hidden>
> ---
>  include/qom/object.h | 25 ++++++++++++++++---------
>  1 file changed, 16 insertions(+), 9 deletions(-)
>
> diff --git a/include/qom/object.h b/include/qom/object.h
> index f0b0bf39cc..c16e0bc91e 100644
> --- a/include/qom/object.h
> +++ b/include/qom/object.h
> @@ -101,24 +101,31 @@ typedef struct InterfaceInfo InterfaceInfo;
>   *   </programlisting>
>   * </example>
>   *
> - * Every type has an #ObjectClass associated with it.  #ObjectClass 
> derivatives
> - * are instantiated dynamically but there is only ever one instance for any
> - * given type.  The #ObjectClass typically holds a table of function pointers
> - * for the virtual methods implemented by this type.
> + * Every type has an #ObjectClass associated with it.  #ObjectClass
> + * derivatives are instantiated dynamically but there is only ever one
> + * instance for any given type.  The #ObjectClass typically holds a
> + * table of function pointers for the virtual methods implemented by
> + * this type. You can cast an #ObjectClass to a subclass (or
> + * base-class) type using object_class_dynamic_cast().
>   *
> - * Using object_new(), a new #Object derivative will be instantiated.  You 
> can
> - * cast an #Object to a subclass (or base-class) type using
> - * object_dynamic_cast().  You typically want to define macro wrappers around
> - * OBJECT_CHECK() and OBJECT_CLASS_CHECK() to make it easier to convert to a
> - * specific type:
> + * Using object_new(), a new #Object derivative will be instantiated.
> + * You can cast an #Object to a subclass (or base-class) type using
> + * object_dynamic_cast().
> + *
> + * You typically want to define macro wrappers to make it easier to
> + * handle casting:
>   *
>   * <example>
>   *   <title>Typecasting macros</title>
>   *   <programlisting>
>   *    #define MY_DEVICE_GET_CLASS(obj) \
>   *       OBJECT_GET_CLASS(MyDeviceClass, obj, TYPE_MY_DEVICE)
> + *    #define IS_MY_DEVICE_CLASS(klass) \
> + *       object_class_dynamic_cast(OBJECT_CLASS(klass), TYPE_MY_DEVICE)
>   *    #define MY_DEVICE_CLASS(klass) \
>   *       OBJECT_CLASS_CHECK(MyDeviceClass, klass, TYPE_MY_DEVICE)
> + *    #define IS_MY_DEVICE(obj) \
> + *       object_dynamic_cast(OBJECT(obj), TYPE_MY_DEVICE)
>   *    #define MY_DEVICE(obj) \
>   *       OBJECT_CHECK(MyDevice, obj, TYPE_MY_DEVICE)
>   *   </programlisting>

I think in general this is a good idea, but I have a couple of
questions:

(1) is this new macro going to become one of the "standard set"
that every QOM object type should provide, like FOO_CLASS
and FOO_GET_CLASS, as noted in
https://wiki.qemu.org/Documentation/QOMConventions
or is it an "on-demand, added to classes where somebody cares" ?

(2) Naming. IS_FOO() implies a boolean return, but
object_dynamic_cast() is "return NULL, or a suitably cast
pointer to the object". (An awful lot of our actual uses
in tree really only seem to want the bool, but I think the
idea was to avoid having the overhead of doing the cast
operation twice.)

thanks
-- PMM