From the lang.reference,
3.5 Type Conversion (Promotion):
... If the result of an integer *, -, or + operation is outside the integer range, the operation will be done in double precision and the result will be double precision.
Does not seem to be the case though:
BSD> ? 65536*65536, 65536*32768
0 -2147483648
BSD> ? 2^30, 2^31, 2^32
1073741824 -2147483648 0
So either documentation or the implementation must be wrong. Which one?
