The already existing AC_CANONICAL_HOST is insufficient in that it does not
detect the actual cpu instruction set to which the C compiler compiles.
E.g., it is quite common that "uname -m" prints sparc64 while gcc compile
32-bit binaries by default.
It would be nice if gnulib offered a module containing one m4 file with one
exported macro which would cache and substitute @host_cpu_abi@ (or
@host_cpu_instructionset@).
Specifically, clisp and libffcall both have this (originally by Bruno, modified
by me):
=============================================================
dnl FFCALL_CACHE_EGREP_CPP(what,variable,condition)
AC_DEFUN([FFCALL_CACHE_EGREP_CPP],
[AC_CACHE_CHECK([for $1], [$2], [AC_EGREP_CPP(yes,
[#if $3
yes
#endif
], [$2=yes], [$2=no])])])
dnl FFCALL_SET_CPU_ABI(what,variable,condition,yes_abi,no_abi)
AC_DEFUN([FFCALL_SET_CPU_ABI],
[FFCALL_CACHE_EGREP_CPP([$1],[$2],[$3])
if test $$2 = yes; then
host_cpu_abi=$4
else
host_cpu_abi=$5
fi])
AC_DEFUN([FFCALL_CANONICAL_HOST_CPU],
[AC_REQUIRE([AC_CANONICAL_HOST])AC_REQUIRE([AC_PROG_CC])
case "$host_cpu" in
changequote(,)dnl
i[4567]86 )
host_cpu_abi=i386
;;
alphaev[4-8] | alphaev56 | alphapca5[67] | alphaev6[78] )
host_cpu_abi=alpha
;;
hppa1.0 | hppa1.1 | hppa2.0* | hppa64 )
host_cpu_abi=hppa
;;
rs6000 )
host_cpu_abi=powerpc
;;
c1 | c2 | c32 | c34 | c38 | c4 )
host_cpu_abi=convex
;;
changequote([,])dnl
arm* )
FFCALL_SET_CPU_ABI([ARMel], ffcall_cv_host_armel,
[defined(__ARMEL__)],armel,arm)
;;
mips* )
dnl We should also check for (_MIPS_SZPTR == 64), but gcc keeps this at 32.
FFCALL_CACHE_EGREP_CPP([64-bit MIPS], ffcall_cv_host_mips64,
[defined(_MIPS_SZLONG) && (_MIPS_SZLONG == 64)])
if test $ffcall_cv_host_mips64 = yes; then
host_cpu_abi=mips64
else
dnl Strictly speaking, the MIPS ABI (-32 or -n32) is independent from the CPU
dnl identification (-mips[12] or -mips[34]). But -n32 is commonly used together
dnl with -mips3, and it's easier to test the CPU identification.
FFCALL_SET_CPU_ABI([MIPS with n32 ABI], ffcall_cv_host_mipsn32,
[__mips >= 3], mipsn32, mips)
fi
;;
dnl On powerpc64 systems, the C compiler may still be generating 32-bit code.
powerpc64 )
FFCALL_SET_CPU_ABI([64-bit PowerPC], ffcall_cv_host_powerpc64,
[defined(__powerpc64__) || defined(_ARCH_PPC64)], powerpc64, powerpc)
;;
dnl UltraSPARCs running Linux have `uname -m` = "sparc64", but the C compiler
dnl still generates 32-bit code.
sparc | sparc64 )
FFCALL_SET_CPU_ABI([64-bit SPARC], ffcall_cv_host_sparc64,
[defined(__sparcv9) || defined(__arch64__)], sparc64, sparc)
;;
dnl On x86_64 systems, the C compiler may still be generating 32-bit code.
x86_64 )
FFCALL_SET_CPU_ABI([64-bit x86_64], ffcall_cv_host_x86_64,
[defined(__LP64__) || defined(__x86_64__) || defined(__amd64__)],
x86_64, i386)
;;
*)
host_cpu_abi=$host_cpu
;;
esac
AC_SUBST(host_cpu_abi)
dnl was AC_DEFINE_UNQUOTED(__${host_cpu}__) but KAI C++ 3.2d doesn't like this
cat >> confdefs.h <<EOF
#ifndef __${host_cpu_abi}__
#define __${host_cpu_abi}__ 1
#endif
EOF
])
=============================================================
It would be nice if you could incorporate it into gnulib.
Thanks.
Sam
