Teodor Sigaev <teodor@sigaev.ru> writes:
>> The alternative seems to be to forbid uninitialized pad bytes within
>> Datums. That's not very pleasant to contemplate either, since it'll
>> forever be vulnerable to sins of omission.
> Hmm, we can add to palloc random filling of allocated memory with
> --enable-cassert. It'll allow to catch such bugs in most cases.
Based on the fact that this problem went undetected for so long,
I'm not real convinced of that. The difficulty is exactly that
not very many system behaviors will fail obviously if nodetrees
that should be seen as equal are not.
Perhaps we could add some additional tests to help exercise it?
But how do you get them applied to all datatypes?
regards, tom lane