Bart De VOS pointed out that removing absolute paths from the libc linker scripts is plainly wrong.
It dates from dawn ages of the original crosstool code, and is not well explained. At that time, binutils might not understand the sysroot stuff, and it was necessary to remove absolute paths in that case.
/trunk/scripts/build/libc/glibc.sh | 14 2 12 0 ++------------
1 file changed, 2 insertions(+), 12 deletions(-)
1 # Taken from http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config.gcc.diff?r1=1.3&r2=1.4
3 # Configuration powerpc-host_apple-darwin7.3.0 not supported
4 # when configuring gcc-2.95 on Mac OS X for i686 target
5 # Also create xm-darwin.h, seems to be required, else we get the error
6 # In file included from .../gcc-2.95.3/gcc/gencheck.c:21:
7 # hconfig.h:11:30: rs6000/xm-darwin.h: No such file or directory
8 # make[1]: *** [gencheck.o] Error 1
11 --- gcc-2.95.3/gcc/configure.old 2004-03-24 12:17:44.000000000 -0800
12 +++ gcc-2.95.3/gcc/configure 2004-03-24 17:14:38.000000000 -0800
13 @@ -5079,6 +5079,10 @@
14 tmake_file=rs6000/t-beos
15 xmake_file=rs6000/x-beos
18 + xm_file="rs6000/xm-rs6000.h rs6000/xm-darwin.h"
19 + xmake_file=rs6000/x-darwin
21 powerpc-*-sysv* | powerpc-*-elf*)
22 tm_file=rs6000/sysv4.h
23 xm_file="xm-siglist.h rs6000/xm-sysv4.h"
24 --- /dev/null 2003-01-30 02:24:37.000000000 -0800
25 +++ gcc-3.0.4/gcc/config/rs6000/xm-darwin.h 2000-11-20 19:02:09.000000000 -0800
27 +/* Undo the USG definition in xm-rs6000.h, Darwin is a BSD flavor. */
31 +/* Override the usual setting, since Apple's GCC has lame bugs and
32 + can't handle the initializers. Someday the bugs will be fixed and
33 + we can get rid of this silliness. */
35 +#define HAVE_DESIGNATED_INITIALIZERS 0