Tom Lane writes:
> Our alternatives would be to add enough ORDER BY clauses to constrain
> the results (which I don't want to do because it would limit the set
> of join plans testable by the tests), or to add another result file
> for this test. I'd be in favor of the latter if there were more than
> one platform showing this behavior, or if I thought that UnixWare's
> behavior wouldn't change again next time SCO decides to tweak qsort.
Fair enough. I just thought, reading your commit message, that you had
set out to completely fix it. I say we leave it as is for now.
--
Peter Eisentraut peter_e(at)gmx(dot)net