Re: GCC libobjc vs. GNUstep libobjc

[Nicola Pero]

> > Can someone give me a quick overview of why, exactally, it is that when
> > you build gnustep on win32 you must use the gnustep libobjc while under a
> > 'real' *nix derived OS the GCC-supplied objc runtime is sufficient?
> > 
> > Is something terminally broken?

Good question.

The GCC-supplied objc runtime does not really work on Windows.  If I
remember correctly, it's an autoconf/libtool problem which will be fixed
in future autoconf/libtool releases.


> > Maybe this is a dumb question, but do the GCC maintainers know about it?
> > If so, might we need to make more noise about it?

If you are interested in it, and you don't have it, asking for it politely
on the GCC mailing list will help.  (It won't change anything in the short
term, but might remind people it's an outstanding issue).


> GCC's libobjc doesn't export all the functions/variables that we need. 
> I'm currently trying to put together a patch for GCC that will fix this, 
> but right now there seems to be some kind of bug that prevents me from 
> doing this in a simple way.

There is a GCC patch from me to fix GCC's libobjc def file - has been
pending for months/ages - Stan Sheb's pretty absent, and he needs to
approve patches before they can be applied, which is why it's pending.

I've not pursued that much, because, as someone from Omnigroup trying to
use the patched GCC libobjc for real on win32 found out (check the mailing
lists), there are much more serious problems with autoconf/libtool which
make it sound a bit hopeless until libtool win32 support is
fixed/enhanced.

This if I remember correctly - I never seriously attempted running stuff
on Windows, so I don't know.

But make sure to track the past GCC mailing list discussion about GCC
libobjc on win32.

If there are any new patches or developments in this area, let me know as
I can try to sponsor the patches speeding up their inclusion into GCC.