On Wed, 06 Jul 2005 12:31:11 +0200, Mattias Ellert wrote: >On Sun, 2005-07-03 at 15:10 +0200, Henning Meier-Geinitz wrote: > >> getopt.c is compiled and linked unconditionally on all platforms as >> far as I can see. As this works on all the other platforms there must >> be something different on OS/2. >> >> You could try to move getopt and getopt_long to AC_REPLACE_FUNCS in >> configure.in and modify lib/Makefile.in accordingly. But I haven't >> checked all the implications of that. Maybe linking the getopt code >> was done intentionally. > >You shouldn't have to do this. If the compiler finds that the glibc has >the right getopt version all the code in getopt.c is removed by the >#ifndef ELIDE_CODE conditional and it creates an empty getopt.o.
But the compiler I have available isn't smart enough to decide... I'm quite shure that the getopt-code included in libc that comes with the gcc335b5-compiler is good enough for SANE. What's the trigger for the compiler to decide? in sane-backends/lib/getopt.c getWe have: ... #define GETOPT_INTERFACE_VERSION 2 #if !defined _LIBC && defined __GLIBC__ && __GLIBC__ >= 2 # include <gnu-versions.h> # if _GNU_GETOPT_INTERFACE_VERSION == GETOPT_INTERFACE_VERSION # define ELIDE_CODE # endif #endif #ifndef ELIDE_CODE ... but _GNU_GETOPT_INTERFACE_VERSION isn't defined in the headers that come with my compiler. Should it? and when yes where? Franz
