Issue #2487.  math.ldexp(x, n) raised OverflowError when n was large and
negative; fix to return an (appropriately signed) zero instead.
3 files changed