preprocessor, 'inline', and Apple's gcc

[Rod Kirkpatrick]

Hiya

Bit of  newbie question here prompted by the fact that i just got the  
mail below from someone running OSX :-)

Im on a G4 running 10.4.11 and XCode 2.5

At the moment all Id ilke to do is write a Cocoa based app that uses  
GMP for the math, using the static library.

Ive read the mailings list and tried all the good things suggested  
but still come up with linking errors similar to those described below.
If I start with an empty Cocoa app with a single action I can declare

mpz_t   N;

but the compiler balks when it hits

mpz_init (N);

Have you got any suggestions please. Ive tried so many different ways  
of compiling the library and app Im going nuts.
I have no idea how to apply the suggestion below.

Rod



On 17/03/2008, at 8:35 AM, David Monniaux wrote:

> If one attempts linking multi-module programs using GMP on MacOS X  
> 10.5,
> using Apple's modified gcc 4.0.1 with the -std=c99, one gets errors
> about duplicate symbols such as __gmpz_abs. The reason is that, in  
> this
> mode, Apple's gcc applies C99 semantics for the 'inline' keyword.
> Accordingly, gmp.h tests whether this is the case (via the  
> preprocessor
> symbol __GNUC_STDC_INLINE_) and acts accordingly.
>
> Problem: contrary to the "official" versions of gcc that have the same
> behavior, Apple's gcc does not define __GNUC_STDC_INLINE__ in order to
> signal that it enforces the new standard C behavior for 'inline' when
> run in C99 mode.
>
> This breaks linking certain programs such as APRON
> (http://apron.cri.ensmp.fr/library/).
>
> Proposed fix:
> * test whether we're on gcc or compilers claiming to be compatible  
> with it
> ** test whether we're on Apple's gcc
> * test whether we're on a non-gcc C99-compliant compiler
> * do other special cases
>
> Your opinion?
>
> #ifdef __GNUC__
> /* Apple's cc applies C99 semantics in C99 mode but does not define
>    __GNUC_STDC_INLINE__ */
>
> #ifdef __APPLE_CC__
> #if defined(__STDC__) && defined(__STDC_VERSION__) && __STDC_VERSION__
>> = 199901L
> #define __GMP_EXTERN_INLINE inline
> #else
> #define __GMP_EXTERN_INLINE      extern __inline__
> #endif
> #else /* __APPLE_CC__ */
> #ifdef __GNUC_STDC_INLINE__
> #define __GMP_EXTERN_INLINE extern __inline__ __attribute__
> ((__gnu_inline__))
> #else
> #define __GMP_EXTERN_INLINE      extern __inline__
> #endif
> #define __GMP_INLINE_PROTOTYPES  1
> #endif /* __APPLE_CC__ */
>
> #elif defined(__STDC__) && __STDC__ >= 199901L /* __GNUC__*/
> #define __GMP_EXTERN_INLINE inline
>
> /* SCO OpenUNIX 8 cc supports "static inline foo()" but not in -Xc  
> strict
>    ANSI mode (__STDC__ is 1 in that mode).  Inlining only actually  
> takes
>    place under -O.  Without -O "foo" seems to be emitted whether  
> it's used
>    or not, which is wasteful.  "extern inline foo()" isn't useful, the
>    "extern" is apparently ignored, so foo is inlined if possible  
> but also
>    emitted as a global, which causes multiple definition errors when
>    building a shared libgmp.  */
> #elif defined(__SCO_VERSION__)
> #if __SCO_VERSION__ > 400000000 && __STDC__ != 1 \
>   && ! defined (__GMP_EXTERN_INLINE)
> #define __GMP_EXTERN_INLINE  static inline
> #endif
> #endif
>
> _______________________________________________
> gmp-discuss mailing list
> gmp-discuss at swox.com
> https://gmplib.org/mailman/listinfo/gmp-discuss

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://gmplib.org/list-archives/gmp-discuss/attachments/20080320/9d7dfe2c/attachment.html