Re: __weak definition and problems with Obj-C

[christos%astron.com@localhost (Christos Zoulas)]

In article <552C306C.2000504%libero.it@localhost>,
Riccardo Mottola  <riccardo.mottola%libero.it@localhost> wrote:
>Hi,
>
>when using current NetBSD and compiling GNUstep (current objective-c 
>stuff from svn, not packages) I get for about any obj-c file these 
>redefinition warnings:
>
>  Compiling file main.m ...
>In file included from 
>/System/Library/Headers/GNUstepBase/GSVersionMacros.h:219:0,
>                  from /System/Library/Headers/AppKit/AppKit.h:32,
>                  from main.m:25:
>/System/Library/Headers/GNUstepBase/GSConfig.h:417:0: warning: "__weak" 
>redefined [enabled by default]
>  #    define __weak
>  ^
>In file included from /usr/include/sys/cdefs.h:60:0,
>                  from /usr/include/inttypes.h:35,
>                  from /System/Library/Headers/GNUstepBase/GSConfig.h:154,
>                  from 
>/System/Library/Headers/GNUstepBase/GSVersionMacros.h:219,
>                  from /System/Library/Headers/AppKit/AppKit.h:32,
>                  from main.m:25:
>/usr/include/sys/cdefs_elf.h:68:0: note: this is the location of the 
>previous definition
>  #define __weak __attribute__((__weak__))
>  ^
>
>on 6.1.5 instead everything works fine there. On 6.1.5, I see this in 
>the header file instead:
>
>#define __weak_reference(sym)   __attribute__((__weakref__(#sym)))
>

It depends what GNUStep considers itself to be; a system facility
or a user program. If GNUStep thinks it owns the __ namespace, and
can define macros in that namespace unconditionally, it should try
defining __cplusplus, __FILE__, __LINE__, etc. too and see if anything
breaks.

I sympathize about __weak, but I don't understand what is it trying to
do here because I don't have the context in GSConfig.h. It looks to
me like it is part of a nested cpp if statement, which perhaps is going
the wrong way because it does not know how to handle NetBSD or it fails
to detect NetBSD. Nevertheless, defining symbols in the __ space should
be protected or avoided.

Don't blame NetBSD though, because these are system headers...

christos