mpn_gcdext bug in v4.3.0?

[Torbjorn Granlund]

shoup at cims.nyu.edu writes:

  I have been running a number of tests with mpn_gcdext,
  and came across what seems to be a bug.
  
Thanks for the bug report!

This is indeed GMP 4.3.0 bug.

In some strict sense, mpn_gcdext is correct, since it does not work in a
way disallowed by the documentation.

But this is not the intended behaviour, and we will change the code to
properly normalise the returned cofactor.  We will also clarify the
documentation, so that it promises normalisation.

It is worth noticing the mpz_gcdext does not cope with mpn_gcdext's
behaviour in 4.3.0.

We are working on a fix.

This bug will prompt a GMP 4.3.1 release shortly.

-- 
Torbjörn