6.0.0a -Fails to build on my powerpc box (debian wheezy)

[Richard Biener]

On Wed, 26 Mar 2014, Torbjorn Granlund wrote:

> [Moving bug report from gmp-discuss and gmp-devel.]
> 
>   libtool: compile:  ../mpn/m4-ccas --m4=m4 gcc -std=gnu99 -c
>   -DHAVE_CONFIG_H -I. -I.. -D__GMP_WITHIN_GMP -I.. -DOPERATION_gcd_1
>   -m64 -O3 gcd_1.asm  -fPIC -DPIC -o .libs/gcd_1.o
>   gcc -std=gnu99 -c -DHAVE_CONFIG_H -I. -I.. -D__GMP_WITHIN_GMP
>   -I.. -DOPERATION_gcd_1 -m64 -O3 tmp-gcd_1.s -fPIC -DPIC -o
>   .libs/gcd_1.o
>   tmp-gcd_1.s: Assembler messages:
>   tmp-gcd_1.s:94: Error: unsupported relocation against
>   BMOD_1_TO_MOD_1_THRESHOLD
>   make[2]: *** [gcd_1.lo] Error 1
>   
> With Richard Biener's analysis this was easy to fix:
> 
> https://gmplib.org/repo/gmp/raw-rev/4a6d258b467f

Thanks, that fixes it.

> We might want to improve GMP's config.guess to better recognise the
> affected CPUs.  We have very broad testing for powerpc64, but no system
> was identified as a plain powerpc64-*.
> 
> I am curious about the contents of /prec/cpuinfo for the problem systems.

processor: 0
cpu: POWER7 (raw), altivec supported
clock: 3300.000000MHz

note that the builds run inside KVM virtual machines (so you see what
qemu tells the system it should masquerade as).

gmp concludes at the end of configure

[   44s]   Version:           GNU MP 6.0.0
[   44s]   Host type:         power7-unknown-linux-gnu
[   44s]   ABI:               mode64
[   44s]   Install prefix:    /usr
[   44s]   Compiler:          gcc -std=gnu99
[   44s]   Static libraries:  yes
[   44s]   Shared libraries:  yes


Richard.