From mboxrd@z Thu Jan  1 00:00:00 1970
From: Alex =?utf-8?Q?Benn=C3=A9e?= <alex.bennee@linaro.org>
Subject: Re: [PATCH kvm-unit-tests] configure: fix non-newlib cross-compiling
Date: Mon, 16 Jan 2017 15:22:00 +0000
Message-ID: <874m0zuj1j.fsf@linaro.org>
References: <20170116132139.13839-1-drjones@redhat.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit
Cc: kvm@vger.kernel.org, pbonzini@redhat.com
To: Andrew Jones <drjones@redhat.com>
Return-path: <kvm-owner@vger.kernel.org>
Received: from mail-wm0-f41.google.com ([74.125.82.41]:38065 "EHLO
        mail-wm0-f41.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org
        with ESMTP id S1750886AbdAPPWE (ORCPT <rfc822;kvm@vger.kernel.org>);
        Mon, 16 Jan 2017 10:22:04 -0500
Received: by mail-wm0-f41.google.com with SMTP id r144so179688988wme.1
        for <kvm@vger.kernel.org>; Mon, 16 Jan 2017 07:22:04 -0800 (PST)
In-reply-to: <20170116132139.13839-1-drjones@redhat.com>
Sender: kvm-owner@vger.kernel.org
List-ID: <kvm.vger.kernel.org>


Andrew Jones <drjones@redhat.com> writes:

> Commit 529046c3 "libcflat: add PRI(dux)32 format types" applies a
> detection trick that requires native compiling or cross-compiling
> with newlib, an embedded system C standard library. We can avoid
> the new dependency with a different trick.
>
> Cc: Alex Bennée <alex.bennee@linaro.org>
> Signed-off-by: Andrew Jones <drjones@redhat.com>
> ---
>  configure | 14 +++++---------
>  1 file changed, 5 insertions(+), 9 deletions(-)
>
> diff --git a/configure b/configure
> index 127868ce3090..3045dd129bda 100755
> --- a/configure
> +++ b/configure
> @@ -12,6 +12,7 @@ host=$arch
>  cross_prefix=
>  endian=""
>  pretty_print_stacks=yes
> +u32_long=
>
>  usage() {
>      cat <<-EOF
> @@ -110,16 +111,11 @@ if [ -f $testdir/run ]; then
>  fi
>
>  # check if uint32_t needs a long format modifier
> -cat << EOF > lib_test.c
> -#include <inttypes.h>
> +cat << EOF > lib-test.c
> +__UINT32_TYPE__
>  EOF
> -
> -$cross_prefix$cc lib_test.c -E | grep "typedef" | grep "long" | grep "uint32_t" &> /dev/null
> -exit=$?
> -if [ $exit -eq 0 ]; then
> -    u32_long=true
> -fi
> -rm -f lib_test.c
> +u32_long=$($cross_prefix$cc -E lib-test.c | awk '! /^#/ && $2 == "long" {print "yes"}')
> +rm -f lib-test.c

Hmm this fails on my system as the -E output is:

  15:18 alex@overdrive01/aarch64  [kvm-unit-tests-32bit.git/review/drew-config@github] >arm-none-eabi-gcc -E test.c
  # 1 "test.c"
  # 1 "<built-in>"
  # 1 "<command-line>"
  # 1 "test.c"
  long unsigned int

Changing the line to:

  u32_long=$($cross_prefix$cc -E lib-test.c | awk '! /^#/ && $1 == "long" {print "yes"}')

Fixes the detection for me but I wonder if there are other corner cases
to catch?

The rest of the compile fails later due to un-merged format fixes for
pci_setup_msi which I guess is expected.

>
>  # check for dependent 32 bit libraries
>  if [ "$arch" != "arm" ]; then


--
Alex Bennée