Re: Methods as dictionary keys using invocation

[Stefan Urbanek]

On 2003-07-06 19:27:29 +0200 Richard Frith-Macdonald <address@hidden> wrote:

> On Sunday, July 6, 2003, at 10:53 AM, Stefan Urbanek wrote:
>
> > Hi,
> >
> > I have an object (prototype) containing a dictionary. I want to access keys in 
> > that dictionary using messages. So if i do:
> >       [proto someKey];
> > it is analogous to:
> >       [[proto dictionary] objectForKey:@"someKey"];
> >
> > However, this used to work some time ago, but now I am getting:
> >       vacall: va_start type 6 and va_return type 14 disagree.
>
> Try commenting out the call to gs_find_best_typed_sel() in 
> gs_objc_msg_forward() in GSFFCallInvocation.m

thanks very much, it helped. can this be fixed on CVS?

> Ideally, the runtime should be passing gs_objc_msg_forward() the correct type 
> information.
> As the objc runtime doesn't necessarily do so, the gs_find_best_typed_sel() 
> guesses,
> and is probably guessing wrong.
> I don't know whether it's generally better to guess or not ... the ideal 
> solution is to modify
> gcc, but that's not going to happen fast.

I think, all requirements for the Objective-C compiler and runtime should  be 
collected somewhere, even those small or 'stupid-from-some-point-of-view' ones 
for further consideration. The reason is, that gnustep/objective-c programmers 
are adapting to old objective-c, however requirements grow. I do not say that 
all things should be imlpemented, but i would like to suggest, that objective-c 
should be revisited by removing obsolete and unused functionality and adding 
new required functionality.

What do you think about such collection of objective c requirements? Who can 
make such collection/specification better than active objective-c users?

Stefan