AIX/xlc test failures for gmp-5.0.1

[Torbjorn Granlund]

Paul Zimmermann <Paul.Zimmermann at loria.fr> writes:

  I agree with Torbjörn it is most likely a compiler bug. Peter did also report
  a problem with MPFR on the same platform, which turned out to be a compiler
  bug. (By the way, Peter, you might also try without optimization.)
  
The compiler hits worse bugs without optimisation, unfortunately.
Bitand is broken, which affects GMP's nails code (even with a nails of
zero bits).

How did you track down the xlc bug?  Did you manage to make the compiler
emit usable debug symbols?

-- 
Torbjörn