Re: pg15b2: large objects lost on upgrade

> On Aug 3, 2022, at 10:14 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>
> Robert Haas <robertmhaas(at)gmail(dot)com> writes:
>>> On Tue, Aug 2, 2022 at 3:51 PM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>>> I also think that ">=" is a sufficient requirement.
>
>> I don't really like this approach. Imagine that the code got broken in
>> such a way that relfrozenxid and relminmxid were set to a value chosen
>> at random - say, the contents of 4 bytes of unallocated memory that
>> contained random garbage. Well, right now, the chances that this would
>> cause a test failure are nearly 100%. With this change, they'd be
>> nearly 0%.
>
> If you have a different solution that you can implement by, say,
> tomorrow, then go for it. But I want to see some fix in there
> within about 24 hours, because 15beta3 wraps on Monday and we
> will need at least a few days to see if the buildfarm is actually
> stable with whatever solution is applied.

Yeah, I would argue that the current proposal
guards against the false positives as they currently stand.

I do think Robert raises a fair point, but I wonder
if another test would catch that? I don’t want to
say “this would never happen” because, well,
it could happen. But AIUI this would probably
manifest itself in other places too?

> A possible compromise is to allow new values that are between
> old value and old-value-plus-a-few-dozen.

Well, that’s kind of deterministic :-) I’m OK
with that tweak, where “OK” means not thrilled,
but I don’t see a better way to get more granular
details (at least through my phone searches).

I can probably have a tweak for this in a couple
of hours if and when I’m on plane wifi.

Jonathan