Re: Odd 9.4, 9.3 buildfarm failure on s390x

I wrote:
> Andres Freund <andres(at)anarazel(dot)de> writes:
>> On 2018-10-01 12:13:57 -0400, Tom Lane wrote:
>>> (2) Drop the restriction. This'd require at least changing the
>>> DESC correction, and maybe other things. I'm not sure what the
>>> odds would be of finding everyplace we need to check.

>> (2) seems more maintainable to me (or perhaps less unmaintainable). It's
>> infrastructure, rather than every datatype + support out there...

> I guess we could set up some testing infrastructure: hack int4cmp
> and/or a couple other popular comparators so that they *always*
> return INT_MIN, 0, or INT_MAX, and then see what falls over.

Here's a draft patch against HEAD for this.

I looked for problem spots by (a) testing with the STRESS_SORT_INT_MIN
option I added in nbtcompare.c, (b) grepping for "x = -x" type code,
and (c) grepping for "return -x" type code. (b) and (c) found several
places that (a) didn't, which does not give me a warm feeling about
whether I have found quite everything.

I changed a couple of places where things might've been safe but
I didn't feel like chasing the calls to prove it (e.g. imath.c),
and contrariwise I left a *very* small number of places alone
because they were inverting the result of a specific function
that is defined to return 1/0/-1 and nothing else.

regards, tom lane