* Visible changes:
- When converting rationals to floats, ECL now rounds instead of using the
routine in GMP, which truncates.
(loop for type in '(short-float single-float double-float long-float)
for i from 60 to 70
nconc (loop with value = (ash 1 i)
with expected = (coerce 1 type)
for j from 0 to 10
for x = (- value j)
for r = (/ (1- x) x)
for f1 = (coerce r type)
for f2 = (- (coerce (- r) type))
unless (and (= f1 expected) (= f2 expected))
Before it produced a list of rationals it failed to round.
Instituto de Física Fundamental, CSIC
c/ Serrano, 113b, Madrid 28006 (Spain)