Re: Fw: maybey an error in mpf_cmp

[Kevin Ryde]

address@hidden (Paul Zimmermann) writes:
>
>   mpf_out_str (stdout, 2, x->_mp_size * mp_bits_per_limb, x);

mpf_out_str never prints more digits than the user requested
precision, even when _mp_size is bigger, which is what's happening
here.  x is 3 limbs, y is 4 limbs.

There's notes about the possible extra limb of precision in the
internals chapter of the manual.  Usually it does only good, but it's
a bit of a trap I guess.  In general testing two floats for exact
equality of course is a bad idea.  mpf_eq would perhaps be the
function to use, if it worked like it's meant to :(.

mpfr does a better job with all this.