Cofactor canonicalisation of mpn_gcdext

[Torbjorn Granlund]

Marc Glisse <marc.glisse at inria.fr> writes:

  On Thu, 5 May 2011, Torbjorn Granlund wrote:
  
  > Now I could reproduce it also on shell, using GMP_CHECK_RANDOMIZE.
  
  Ah, I didn't even suspect the seeding as the cause for this difference
  of behavior. Thanks for the tip, I'll try and remember it next time.

I don't think it was, actually.  The sequence of random number for the
gmp tests will be the same on all platforms.

There is some undefined behaviour in the code.  Perhaps memory
corruption, perhaps something else.

-- 
Torbjörn