Re: [Qemu-devel] 答复： Re: [PATCH v2] object: Add 'help' option for all available backends and properties

[Markus Armbruster]

"Daniel P. Berrange" <address@hidden> writes:

> On Thu, Sep 22, 2016 at 01:12:22PM +0200, Markus Armbruster wrote:
>> "Daniel P. Berrange" <address@hidden> writes:
>> 
>> > On Thu, Sep 22, 2016 at 10:36:45AM +0200, Markus Armbruster wrote:
>> >> Don't make up a description in user_creatable_help_func(), improve the
>> >> description infrastructure and its use so you get more useful ones
>> >> there.
>> >> 
>> >> The existing description infrastructure is just Property member
>> >> description and object_property_set_description().  Rarely used, so
>> >> description is generally null.
>> >> 
>> >> Calling object_property_set_description() more often could be helpful,
>> >> but to come up with a sensible description string, you need to know what
>> >> the property does.  Needs to be left to people actually familiar with
>> >> the objects.
>> >> 
>> >> Aside: historically, we add properties to *instances*.  All the property
>> >> meta-data gets duplicated for every instance, including property
>> >> descriptions.  This is more flexible than adding the meta-data to the
>> >> class.  The flexibility is rarely needed, but the price in wasted memory
>> >> is always paid.  Only since commit 16bf7f5, we can add it to classes.
>> >> Adding lots of helpful property descriptions would increase the cost of
>> >> instance properties further.
>> >
>> > FWIW, we could easily optimize handling of description strings by
>> > applying the same trick that GLib has done for GObject property
>> > descriptions.
>> >
>> > Almost certainly every call to object_property_set_description is
>> > going to be passing a string literal, not a dynamically allocated
>> > string. So we take advantage of that and in fact mandate that it
>> > is a string literal, and thus avoid the strdup() of description.
>> >
>> > We can place a fun game to enforce this at compile time thus:
>> >
>> >  - Rename object_property_set_description() to
>> >    object_property_set_description_internal()
>> >
>> >  - Add the macro
>> >
>> >   #define  object_property_set_description(obj, name, desc, errp) \
>> >      object_property_set_description_internal(obj, name, "" desc "", errp)
>> 
>> Cute :)
>> 
>> > None the less, we really should make an effort to switch things
>> > over to use class properties instead of instance properties, as
>> > its going to save us allocating a 64 byte struct per property
>> > per instance
>> 
>> Yes, please.
>> 
>> Related: a way to define a bunch of properties as *data*, i.e. an array
>> of property descriptions, commonly static.  Reasoning about static data
>> is so much easier than reasoning about code.
>
> IMHO we should go further and leverage QAPI schema to auto-generate all
> the tedious boilerplate code for QOM objects
>
> eg, consider the crypto/secret.c object file.
>
> We could declare it as
>
> { 'object': 'QCryptoSecret',
>   'parent': 'Object',
>   'properties': {
>      'format': 'QCryptoSecretFormat',
>      'data': 'str',
>      'file': 'str',
>      'keyid': 'str',
>      'iv': 'str'
>      } }
>
> Based on that it would have enough knowledge to generate
>
>  - struct QCryptoSecret  definition + typedef
>  - struct QCryptoSecretClass definition + typedef
>  - TYPE_CRYPTO_SECRET macro
>  - QCRYPT_SECRET() cast macro
>  - Setters & getters aka
>      qcrypto_secret_prop_set_format
>      qcrypto_secret_prop_get_format
>      qcrypto_secret_prop_set_data
>      qcrypto_secret_prop_get_data
>      qcrypto_secret_prop_set_file
>      qcrypto_secret_prop_get_file
>      qcrypto_secret_prop_set_keyid
>      qcrypto_secret_prop_get_keyid
>      qcrypto_secret_prop_set_iv
>      qcrypto_secret_prop_get_iv
>  - qcrypto_secret_finalize
>  - qcrypto_secret_class_init
>  - TypeInfo qcrypto_secret_info variable
>  - qcrypto_secret_register_types() method
>  - type_init(qcrypto_secret_register_types);
>
> That'd massively reduce the work to create new objects, to just
> filling in the semantically useful logic

Promising idea.  Sadly, the existing QAPI queue is eating all my QAPI
cycles.