cf documentation



The 'cf' (continued fraction) documentation should mention that cf
expansions are exact for rational numbers, but are cut off at ratepsilon
(for floats) and 10^-fpprec (for bigfloats).

              -s