Re: libmudflap/mf-hooks2.c portability breakage

[kaih at khms dot westfalen dot de (Kai Henningsen)]

gerald@pfeifer.com (Gerald Pfeifer)  wrote on 29.07.03 in <Pine.BSF.4.56.0307291422240.68102@acrux.dbai.tuwien.ac.at>:

> FreeBSD fails to bootstrap tree-ssa branch, because libmudflap/mf-hooks2.c
> contains
>
>   /* These attempt to coax various unix flavours to declare all our
>      needed tidbits in the system headers.  */
>   #define _POSIX_SOURCE  <<---
>   #define _GNU_SOURCE
>   #define _XOPEN_SOURCE
>   #define _BSD_TYPES
>   #define __EXTENSIONS__
>
> and afterwards includes <sys/types.h>, which on FreeBSD 4.8 has the
> following
>
>   #ifndef _POSIX_SOURCE
>   typedef unsigned char   u_char;
>   typedef unsigned short  u_short;
>   typedef unsigned int    u_int;
>   typedef unsigned long   u_long;
>   typedef unsigned short  ushort;         /* Sys V compatibility */
>   typedef unsigned int    uint;           /* Sys V compatibility */
>   #endif
>
> so, in the end, the following part of <sys/socket.h> which is included a
> bit fails to compile (as u_char is undefined):
>
>   typedef u_char          sa_family_t;
>
> Defining _POSIX_SOURCE breaks compilation on FreeBSD.

That would seem to be a (pretty obvious) bug in FreeBSD. After all, the  
whole intention of _POSIX_SOURCE is to make visible bits for POSIX not in  
ANSI/ISO C, such as the socket stuff.

Does it also break with the more modern

#define _POSIX_C_SOURCE 2001<whatever>

?

(And would not this be a candidate for fixincludes?)

MfG Kai