Limit or Bug?

The following bit of SQL uses the mod() function to determine modulus 2 of
two large numbers

13:37:20 SQL>set echo on
13:37:24 SQL>@mod_test
13:37:25 SQL>
13:37:25 SQL>
13:37:25 SQL>
13:37:25 SQL>select mod(power(2,127)-1,2) from dual
13:37:25   2  /

MOD(POWER(2,127)-1,2)
---------------------
                    1

1 row selected.

13:37:25 SQL>
13:37:25 SQL>select mod(power(2,128)-1,2) from dual
13:37:25   2  /

MOD(POWER(2,128)-1,2)
---------------------
                   -1

1 row selected.

13:37:25 SQL>

--------------------------------------------------

Notice that mod(power(2,128)-1,2) returns an incorrect answer, while the
answer for mod(power(2,127)-1,2) is correct.

Does someone here know why?

Is it a limit?

Or is it a bug?

Please supply an explanation and/or URL for your answer.  :)

My guess (yes Alex, it's a guess, or rather, a hunch) is that this is
related to two's complement binary numbers, but I haven't been able
to (yet) find any explanation for this.

-- 
Jared Still
Certifiable Oracle DBA and Part Time Perl Evangelist

Other related posts: