Peter Eisentraut <peter.eisentraut@enterprisedb.com> writes:
> On 29.11.22 22:34, Tom Lane wrote:
>> Concretely, it seems like something like the attached could be
>> useful, independently of the other change.
> Yes, right now you can easily declare things that don't make sense.
> Cross-checks like these look useful.
Checking my notes from awhile back, there was one other cross-check
that I thought was pretty high-priority: verifying that array_size
fields precede their array fields. Without that, a read function
will fail entirely, and a compare function might index off the
end of an array depending on which array-size field it chooses
to believe. It seems like an easy mistake to make, too.
I added that and pushed.
regards, tom lane