big float to float conversion



(%i12) fpprec : 50$
(%i13)  x : 1.8514804991703517693765213431502813b-7;
(%o13) 1.8514804991703517693765213431502813b-7

Convert x to a double. Shouldn't this be 1.8514804991703518*10^-7
(last digit is wrong)>

(%i14) y : float(x);
(%o14) 1.8514804991703517*10^-7

The relative difference seems too big too:

(%i15) abs(x-y)/abs(x);
(%o15) 4.6687728105236312161420230492805183214979733851685b-17

Barton