Re: [Chicken-users] Objective-C interface (sort of...)

[felix winkelmann]

On 11/24/05, Thomas Chust <address@hidden> wrote:

>    * One could get rid of the lazy-ffi dependency (which is currently pre-
>      venting me from testing the egg) by using the objc_msgSendv family of
>      functions instead of the objc_msgSend family

libffi should work fine on OS X, so I don't see a problem here.
objc_msgSendv() might not solve our problems, since we have to construct
arguments by hand.

>    * To create an NSMethodSignature you should be able to just use the
>      - (NSMethodSignature *)methodSignatureForSelector:(SEL)aSelector
>      or
>      + (NSMethodSignature
> *)instanceMethodSignatureForSelector:(SEL)aSelector
>      methods provided by the NSObject class.

But we have to return a valid signature. If the selector represents
something on the Scheme side, I see no way to construct the argument
types of a method signature, since we have to type-information (or we
can provide a list of type-specifications, which is probably the only
way to handle this).

>    * When I tried to create new ObjC classes from Scheme, which is really
>      necessary when you want to do any useful GUI Programming for example,
>      I ran into big trouble as the ObjC runtime functions apparently do
>      *not* behave as documented in that area. For example, the supplied
>      usage examples in the Apple Documentation do not work at all. Looking
>      around on the net, I found out, that the PyObjC developers have been
>      complaining about the same problem. At that point I decided to give up
>      the project for lack of time to reverse engineer Apples runtime and to
>      study PyObjC's guts.
>

Urgh. That doesn't sound too good. Another thing is that I find integration of
external tools (like chicken) with Xcode quite hard to accomplish.
The Mac really wants to be programmed in ObjC...


cheers,
felix