Re: fixing old_snapshot_threshold's time->xid mapping

On Fri, Sep 25, 2020 at 2:00 PM Michael Paquier <michael(at)paquier(dot)xyz> wrote:
> On Thu, Sep 24, 2020 at 03:46:14PM -0400, Robert Haas wrote:
> > Committed.
>
> Cool, thanks.

+1

> > Thomas, with respect to your part of this patch set, I wonder if we
> > can make the functions that you're using to write tests safe enough
> > that we could add them to contrib/old_snapshot and let users run them
> > if they want. As you have them, they are hedged around with vague and
> > scary warnings, but is that really justified? And if so, can it be
> > fixed? It would be nicer not to end up with two loadable modules here,
> > and maybe the right sorts of functions could even have some practical
> > use.

Yeah, you may be right. I am thinking about that. In the meantime,
here is a rebase. A quick recap of these remaining patches:

0001 replaces the current "magic test mode" that didn't really test
anything with a new test mode that verifies pruning and STO behaviour.
0002 fixes a separate bug that Andres reported: the STO XID map
suffers from wraparound-itis.
0003 adds a simple smoke test for Robert's commit 55b7e2f4. Before
that fix landed, it failed.

> I have switched this item as waiting on author in the CF app then, as
> we are not completely done yet.

Thanks. For the record, I think there is still one more complaint
from Andres that remains unaddressed even once these are in the tree:
there are thought to be some more places that lack
TestForOldSnapshot() calls.