Found this problem while playing with elliptic integrals an bigfloats:
(%i1) display2d:false;
(%o1) false
(%i2) gamma(3/4)/gamma(1/4);
(%o2) gamma(3/4)/gamma(1/4)
(%i3) float(%);
(%o3) .3379891200336423
(%i4) bfloat(%o2);
(%o4) 4.125843750410604b-5
%o4 is WAY off. But, we have
(%i5) bfloat(gamma(3/4));
(%o5) 1.2254167024651776452b0
(%i6) bfloat(gamma(1/4));
(%o6) 3.6256099082219083112b0
(%i7) %o5/%o6;
(%o7) 3.379891200336424b-1
So something is messed with with the computation of the ratio.
Ray