divappr tests

[Torbjorn Granlund]

nisse at lysator.liu.se (Niels Möller) writes:

  I also have some concerns over the divappr interface (basically, to
  produce qn quotient words it should need only qn+1 limbs of N and at
  most qn+1 limbs of d, I'll have to get the details down in a separate
  mail).
  
Absolutely, and that is why I called the code "silly" a few days ago.

Fixing this is the next item on http://gmplib.org/devel/GMP44.html I
intend to address.

-- 
Torbjörn