luajit: bug fix - prevent target CFLAGS from being used in host tool compile.

The luajit amalgamation compile starts with building a host tool, and then
uses it to build itself.  However, when CFLAGS is specified, as opposed to
TARGET_CFLAGS, then it is used for both HOST and TARGET builds.  So if you
add something target specific into 'Target Optimizations' (for example,
'-mfpu=neon -mfloat-abi=softfp' for ARM Cortex-A8), then it gets into the
host tool compile, which then fails (because my build machine is not an
ARM Cortext-A8).  This can be fixed by using TARGET_CFLAGS and
TARGET_LDFLAGS, instead of CFLAGS and LDFLAGS, respectively.

Signed-off-by: Danomi Manchego <danomimanchego123@gmail.com>
---
 package/luajit/luajit.mk |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Acked-by: Francois Perrad <francois.perrad@gadz.org>

2012/8/15 Danomi Manchego <danomimanchego123@gmail.com>:
> The luajit amalgamation compile starts with building a host tool, and then
> uses it to build itself.  However, when CFLAGS is specified, as opposed to
> TARGET_CFLAGS, then it is used for both HOST and TARGET builds.  So if you
> add something target specific into 'Target Optimizations' (for example,
> '-mfpu=neon -mfloat-abi=softfp' for ARM Cortex-A8), then it gets into the
> host tool compile, which then fails (because my build machine is not an
> ARM Cortext-A8).  This can be fixed by using TARGET_CFLAGS and
> TARGET_LDFLAGS, instead of CFLAGS and LDFLAGS, respectively.
>
> Signed-off-by: Danomi Manchego <danomimanchego123@gmail.com>
> ---
>  package/luajit/luajit.mk |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/package/luajit/luajit.mk b/package/luajit/luajit.mk
> index e4c46d8..913bc0c 100644
> --- a/package/luajit/luajit.mk
> +++ b/package/luajit/luajit.mk
> @@ -42,8 +42,8 @@ define LUAJIT_BUILD_CMDS
>                 TARGET_LD="$(TARGET_CC)" \
>                 TARGET_AR="$(TARGET_AR) rcus" \
>                 TARGET_STRIP="$(TARGET_STRIP)" \
> -               CFLAGS="$(TARGET_CFLAGS)" \
> -               LDFLAGS="$(TARGET_LDFLAGS)" \
> +               TARGET_CFLAGS="$(TARGET_CFLAGS)" \
> +               TARGET_LDFLAGS="$(TARGET_LDFLAGS)" \
>                 HOST_CC="$(LUAJIT_HOST_CC)" \
>                 HOST_CFLAGS="$(HOST_CFLAGS)" \
>                 HOST_LDFLAGS="$(HOST_LDFLAGS)" \
> --
> 1.7.9.5
>
> _______________________________________________
> buildroot mailing list
> buildroot@busybox.net
> http://lists.busybox.net/mailman/listinfo/buildroot

Le Tue, 14 Aug 2012 21:25:30 -0400,
Danomi Manchego <danomimanchego123@gmail.com> a écrit :

> The luajit amalgamation compile starts with building a host tool, and then
> uses it to build itself.  However, when CFLAGS is specified, as opposed to
> TARGET_CFLAGS, then it is used for both HOST and TARGET builds.  So if you
> add something target specific into 'Target Optimizations' (for example,
> '-mfpu=neon -mfloat-abi=softfp' for ARM Cortex-A8), then it gets into the
> host tool compile, which then fails (because my build machine is not an
> ARM Cortext-A8).  This can be fixed by using TARGET_CFLAGS and
> TARGET_LDFLAGS, instead of CFLAGS and LDFLAGS, respectively.
> 
> Signed-off-by: Danomi Manchego <danomimanchego123@gmail.com>

Applied, thanks. It fixes build issues like
http://autobuild.buildroot.net/results/022b73a989734d800225c91f419f18f4fc31c264/build-end.log.

Danomi, could you share what is your Buildroot configuration that uses
-mfpu=neon -mfloat-abi=softfp ? I might be interested in adding such a
configuration to the autobuilders.

Thanks!

Thomas

Hi Thomas,

> It fixes build issues like
>
http://autobuild.buildroot.net/results/022b73a989734d800225c91f419f18f4fc31c264/build-end.log.

This link doesn't work for me, but I'll take your word for it.

> could you share what is your Buildroot configuration that uses
> -mfpu=neon -mfloat-abi=softfp ?
0
Right now, we are on a rather old version of buildroot, but looking forward
to updating to the latest when 2012-08 is officially released.  So it's
probably best to wait until then.

There are a couple of reasons why we added these two flags to "Target
Optimizations".  For one, the options that are pre-baked into the external
toolchain wrapper don't capture enabling neon optimization.  But, more
importantly, I wasn't able to get ffmpeg to recognize that neon
optimizations were applicable until I (1) added the --cpu=cortex-a8 (for
which I submitted a patch, never got any comments,
http://patchwork.ozlabs.org/patch/154307/), and (2) passed in the softfp
option as part of the --extra-cflags specified during the package configure
step (for which I never actually submitted a patch).  Hopefully, ffmpeg is
unique in this regard ...

Danomi -