Hi David,
Le 6 janv. 2010 à 23:29, David Chisnall a écrit :
> On 6 Jan 2010, at 15:06, 荒田 実樹 wrote:
>
>> - class_getSuperclass directly refer to cls->super_class
>> (class_get_super_class doesn't work on metaclass).
class_getSuperclass() seems to be correctly implemented in
ObjectiveC2.framework but not in libobjc2.
In libobjc2, it calls class_get_super_class() which doesn't work with
a metaclass as argument.
I think several similar new libobjc2 functions are currently broken
when you pass a metaclass because they call the old libobjc1 API which
doesn't accept metaclasses as classes.
Unless libobjc2 has changed the semantic of CLS_ISCLASS, CLS_ISCLASS
(aMetaclass) counterintuitively returns false. Since the old libobjc1
API checks the argument with CLS_ISCLASS in functions such as
class_get_super_class() or class_get_instance_size(), new APIs such as
class_getSuperclass() and class_getInstanceSize() are not equivalent.
This problem probably exists with few other functions in addition the
two I mentionned.
On Mac OS X, the following is valid and returns two different class
objects:
class_getSuperclass(objc_getClass("NSCountedSet"));
class_getSuperclass(objc_getMetaClass("NSCountedSet"));
Similarly:
class_getInstanceSize(objc_getClass("NSCountedSet")); // returns 12
class_getInstanceSize(objc_getMetaClass("NSCountedSet")); // returns 48
> Hmm. For Apple-compatibility we should be setting it to NSObject. I
> wonder if that would break GNUstep...
>
> By the way, if you've got a decent set of tests for this API then they
> would be very helpful...
May be you can adapt the Apple test suite:
http://www.opensource.apple.com/source/objc4/objc4-437.1/test/
Cheers,
Quentin.
_______________________________________________
Etoile-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/etoile-discuss
