should raise a divide-by-zero floating point exception for x = +-0
and an invalid floating point exception for x < 0 including x = -Inf.
Update the code to raise the exception and update the documentation
with hopefully better description of the behavior.
Reviewed by: bde (code only)
186f620727