Andres Freund <andres@anarazel.de> writes:
> I'm not quite following. Could you check if the same happens without
> -O2? Not because that'd be a solution, but to narrow down how this
> happens?
The committed test looks quite broken to me: it's missing some &
operators. Not sure how that translates into failing to fail the
configure test, but personally I'd have done this like
volatile PG_INT64_TYPE a = 1;
volatile PG_INT64_TYPE b = 1;
PG_INT64_TYPE result;
__builtin_mul_overflow(a, b, &result);
regards, tom lane