With the mips-linux-gnu-gcc-5 cross compiler on Debian, host-cpu-c-abi reports the n32 ABI, when it's in fact the 32 (a.k.a. o32) ABI. This fixes it.
2017-07-18 Bruno Haible <br...@clisp.org> host-cpu-c-abi: Fix detection of MIPS ABI. * m4/host-cpu-c-abi.m4 (gl_HOST_CPU_C_ABI): For MIPS, really test the ABI, not the CPU instruction set. diff --git a/m4/host-cpu-c-abi.m4 b/m4/host-cpu-c-abi.m4 index a3a3d43..3017f6c 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 4 +# host-cpu-c-abi.m4 serial 5 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, @@ -146,12 +146,12 @@ changequote([,])dnl yes #endif], [gl_cv_host_cpu_c_abi=mips64], - [# Strictly speaking, the MIPS ABI (-32 or -n32) is independent - # from the CPU identification (-mips[12] or -mips[34]). But -n32 - # is commonly used together with -mips3, and it's easier to test - # the CPU identification. + [# In the n32 ABI, _ABIN32 is defined, _ABIO32 is not defined, and + # _MIPS_SIM == _ABIN32. + # In the 32 ABI, _ABIO32 is defined, _ABIN32 is not defined, and + # _MIPS_SIM == _ABIO32. AC_EGREP_CPP([yes], - [#if __mips >= 3 + [#if defined _ABIN32 yes #endif], [gl_cv_host_cpu_c_abi=mipsn32],