puzzle with bfloat and radcan



I hadn't thought about it so much, but it is not so hard to fool bfloat.

fpprec:30

bfloat(g(10^41+1)-g(10^41));

returns 0.0b0