Dennis Bjorklund <db(at)zigo(dot)dhs(dot)org> writes:
> On Sat, 17 Apr 2004, Tom Lane wrote:
>> *some* set of inputs. (Also, I have been harboring some notions of
>> supporting cross-type hash joins for integer types, which will not work
>> unless small int8 values hash the same as int4 etc.)
> The simple solution would be to always extend integers to 64 bits (or
> whatever the biggest integer is) before calculating the hash.
That creates portability issues though. We do not depend on there being
a 64-bit-int type for anything except int8 itself, and I don't want to
start doing so.
regards, tom lane