host-cpu-c-abi: detect ILP32 ABI on IA-64

[Bruno Haible]

The ILP32 ABI on IA-64 is like the x32 ABI on x86_64, or the n32 ABI on mips64.
It exists only on HP-UX.


2017-07-30  Bruno Haible  <address@hidden>

        host-cpu-c-abi: Detect ILP32 ABI on IA-64 HP-UX.
        * m4/host-cpu-c-abi.m4 (gl_HOST_CPU_C_ABI): Distinguish ia64-ilp32
        from ia64. For arm64, test only __aarch64__, as __ARM_64BIT_STATE and
        __ARM_PCS_AAPCS64 are not reliable indicators: they are not defined on
        Ubuntu 14.04 (gcc 4.8.4) and Debian 8 (gcc 4.9.2).

diff --git a/m4/host-cpu-c-abi.m4 b/m4/host-cpu-c-abi.m4
index 3017f6c..01c5a66 100644
--- a/m4/host-cpu-c-abi.m4
+++ b/m4/host-cpu-c-abi.m4
@@ -1,4 +1,4 @@
-# host-cpu-c-abi.m4 serial 5
+# host-cpu-c-abi.m4 serial 6
 dnl Copyright (C) 2002-2017 Free Software Foundation, Inc.
 dnl This file is free software; the Free Software Foundation
 dnl gives unlimited permission to copy and/or distribute it,
@@ -95,7 +95,7 @@ changequote([,])dnl
          # - aarch64 instruction set, 32-bit pointers, 32-bit 'long': 
arm64-ilp32.
          # - 32-bit instruction set, 32-bit pointers, 32-bit 'long': arm or 
armhf.
          AC_EGREP_CPP([yes],
-           [#if defined(__aarch64__) || defined(__ARM_64BIT_STATE) || 
defined(__ARM_PCS_AAPCS64)
+           [#if defined __aarch64__
             yes
             #endif],
            [AC_EGREP_CPP([yes],
@@ -106,13 +106,13 @@ changequote([,])dnl
               [gl_cv_host_cpu_c_abi=arm64])],
            [# Don't distinguish little-endian and big-endian arm, since they
             # don't require different machine code for simple operations and
-            # since the user can distinguish them through the preprocessot
+            # since the user can distinguish them through the preprocessor
             # defines __ARMEL__ vs. __ARMEB__.
             # But distinguish arm which passes floating-point arguments and
             # return values in integer registers (r0, r1, ...) - this is
             # gcc -mfloat-abi=soft or gcc -mfloat-abi=softfp - from arm which
             # passes them in float registers (s0, s1, ...) and double registers
-            # (d0, d1, ...) - rhis is gcc -mfloat-abi=hard. GCC 4.6 or newer
+            # (d0, d1, ...) - this is gcc -mfloat-abi=hard. GCC 4.6 or newer
             # sets the preprocessor defines __ARM_PCS (for the first case) and
             # __ARM_PCS_VFP (for the second case), but older GCC does not.
             echo 'double ddd; void func (double dd) { ddd = dd; }' > conftest.c
@@ -138,6 +138,17 @@ changequote([,])dnl
            [gl_cv_host_cpu_c_abi=hppa])
          ;;
 
+       ia64* )
+         # On ia64 on HP-UX, the C compiler may be generating 64-bit code or
+         # 32-bit code. In the latter case, it defines _ILP32.
+         AC_EGREP_CPP([yes],
+           [#if defined _ILP32
+            yes
+            #endif],
+           [gl_cv_host_cpu_c_abi=ia64-ilp32],
+           [gl_cv_host_cpu_c_abi=ia64])
+         ;;
+
        mips* )
          # We should also check for (_MIPS_SZPTR == 64), but gcc keeps this
          # at 32.
@@ -262,6 +273,9 @@ EOF
 #ifndef __hppa64__
 #undef __hppa64__
 #endif
+#ifndef __ia64_ilp32__
+#undef __ia64_ilp32__
+#endif
 #ifndef __ia64__
 #undef __ia64__
 #endif