Chao Li <[email protected]> writes:
> On Jun 23, 2026, at 10:11, Chao Li <[email protected]> wrote:
>> Then I checked why the sanity check didn’t report the issue. I found that
>> the check used != for the comparison. When the final argument type is not an
>> array type, the subquery returns NULL, and NULL != provariadic evaluates to
>> NULL rather than true. I fixed the check to use IS DISTINCT FROM. After
>> that, it reported the mismatch:
Ouch. Yes, that regression test is clearly one brick shy of a load.
We need to update the documentation for these functions too, and maybe
some other places.
regards, tom lane
