'make check' fails on PPC64

[Fuzzy Logic]

Okay, I tried again, but with:

configure --enable-cxx --build=none-unknown-linux-gnu

to build with just the plain vanilla.

make succeeds.

make check fails with an ld error.

last few lines:

ranlib .libs/libtests.a
creating libtests.la
(cd .libs && rm -f libtests.la && ln -s ../libtests.la libtests.la)
gcc -DHAVE_CONFIG_H -I. -I. -I.. -I..  -DNO_ASM  -O3 -c t-bswap.c
/bin/sh ../libtool --mode=link gcc  -O3   -o t-bswap  t-bswap.o
libtests.la ../libgmp.la
gcc -O3 -o .libs/t-bswap t-bswap.o  ./.libs/libtests.a
/root/updates/gmp-4.2.1/.libs/libgmp.so ../.libs/libgmp.so
/root/updates/gmp-4.2.1/.libs/libgmp.so: undefined reference to `__gmpn_add_nc'
/root/updates/gmp-4.2.1/.libs/libgmp.so: undefined reference to `__gmpn_sub_nc'
collect2: ld returned 1 exit status
make[4]: *** [t-bswap] Error 1
make[4]: Leaving directory `/root/updates/gmp-4.2.1/tests'
make[3]: *** [check-am] Error 2
make[3]: Leaving directory `/root/updates/gmp-4.2.1/tests'
make[2]: *** [check-recursive] Error 1
make[2]: Leaving directory `/root/updates/gmp-4.2.1/tests'
make[1]: *** [check-recursive] Error 1
make[1]: Leaving directory `/root/updates/gmp-4.2.1'
make: *** [check] Error 2

This doesn't seem like a compiler bug. Same everything except the
option I used for the configure script.

On 15 Feb 2007 23:40:30 +0100, Torbjorn Granlund wrote:
> "Fuzzy Logic" <fuzzymo at gmail.com> writes:
>
>   So, I ask again, Torbjorn. Does this sound like a compiler bug, a
>   configure bug, or "other"?
>
> Sure, like with all failures like that, I'd bet and arm and a leg that
> it is a compiler bug.
>
> It might seem that GMP is unusually unlucky with compilers, that it
> triggers more bugs than other packages.  I believe that is an
> illusion.  What really is happening, is that GMP has a test suite that
> will find almost every miscompilation in the library.  Try a coverage
> analysis program and you'll see; the GMP test suite covers every line
> of the GMP C code.
>
> It seems that some people see "make check" failures in GMP as a GMP
> quality problem.  It is the contrary, of course!
>
> We would all get shivers if we knew how many miscompiled binaries we
> had in our systems.  In most cases, they will go unnoticed, but since
> sshd, openssl, our firewall, etc, of course also are miscompiled, we
> cannot know anything about the real security of any system.
>
> --
> Torbjörn
>


-- 
Latin: Dum spiro spero.
English: While I breathe, I hope.