Xi Wang <xi.wang@gmail.com> writes:
> The reality is that C compilers are not friendly to postcondition
> checking; they consider signed integer overflow as undefined behavior,
> so they do whatever they want to do. Even workaround options like
> -fwrapv are often broken, not to mention that they may not even have
> those options.
I think it's probably past time that we stopped guessing about this
sort of thing and added some regression test cases for it. I'm
planning to add cases like this:
-- check sane handling of INT_MIN overflow cases
SELECT (-2147483648)::int4 * (-1)::int4;
SELECT (-2147483648)::int4 / (-1)::int4;
SELECT (-2147483648)::int4 % (-1)::int4;
regards, tom lane