[Debian-med-packaging] Bug#896843: FTBFS: incorrect CPPFLAG for powerpc*

[Adrian Bunk]

On Tue, Apr 24, 2018 at 03:24:11PM -0500, Barry Arndt wrote:
> Source: ctsim
> Version: 6.0.2
> Severity: serious
> Tags: l10n patch
> Justification: fails to build from source (but built successfully in the past)
> 
> Dear Maintainer,
> 
> FTBFS of ctsim-6.0.2 on ppc64el reported from unicamp.
> 
> The build log reported:
> g++: error: unrecognized command line option '-faltivec'; did you mean '-maltivec'
> 
> The correct flag for ppc64el (as well as powerpc*) is -maltivec.
> 
> I changed -faltivec in configure.ac to -maltivec as follows:
> 
> --- ctsim-6.0.2.orig/configure.ac
> +++ ctsim-6.0.2/configure.ac
> @@ -137,7 +137,7 @@ case $target_cpu in
>          CXXFLAGS="$CXXFLAGS $CPUEXT_FLAGS $SIMD_FLAGS"
>          ;;
>      powerpc*)
> -        ARCH_OPTION="-fno-common -faltivec";;
> +        ARCH_OPTION="-fno-common -maltivec";;
>      armv1*|armv2*|armv3*|armv4*|armv5*|armv6*)
>          ARCH_OPTION="-ffast-math";;
>      armv7*|armv8*)
> 
> I then ran autoconf.  After that, the package build properly.

This whole compiler options block should be nuked,
it also creates baseline violations on amd64 and i386.

> This fix should also fix the FTBFS of ctsim on 
> powerpc, powerpcspe, and ppc64.

None of these have AltiVec in the port baseline,
so that mustn't be done there.

(And ppc64el has it enabled by default, so no benefits from adding
 -maltivec there.)

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed