Re: [RFC/Proposal] Cocoa.h header

[Alexander Malmberg]

Fred Kiefer wrote:
> Alexander Malmberg wrote:
[snip]
> > Well, fwiw, I think it's Wrong to put darwin system header 
> > compatibility stuff in Cocoa.h. If anyone is really interested in 
> > knowing why, I could make an attempt to explain. Otherwise, suffice to 
> > say that:
>
> Not sure why you call the implicit inclusion of math.h in Foundation.h 
> on Cocoa systems "darwin system header compatibility".

I haven't said anything about Foundation.h on Cocoa. I'm talking about 
core/gui/Headers/Cocoa/Cocoa.h, which (presumably), for compatibility 
reasons, includes symbols defined in darwin system headers that happen 
to be visible after including Cocoa.h on mosx, hence "darwin system 
header compatibility" stuff.

[snip]
> > 3. If you insist on defining the symbols yourself, at least get the 
> > rounding right for M_PI.
>
> This one falls back on you :-) I stole the definition from NSBezierPath, 
> where PI instead of M_PI is used (which needs correction!), and if I 
> remember correctly this was one of your first contributions to GNUstep 
> years ago.

It is better to check cvs history than to try to remember. The 
definition of PI in NSBezierPath.m was added by Enrico Sersale 1999-12-21.

> You only focused on the definition of M_PI, which shouldn't do to much 
> harm as it is protected by an #ifdef. You didn't mention the other bit, 
> which I see as being much more important, the handling of the compiler 
> specific bool type. GNUstep will keep on supporting gcc releases where 
> this type is not present, so we need to offer some sort of 
> compatibility. What are your choices here?

I don't really have anything to add beyond what I already said in 1.; in 
short, "don't do that".

- Alexander Malmberg