Re: [RFC] Property-based testing for Postgres

On Mon, Nov 2, 2026 at 5:33 PM Jeff Davis <pgsql(at)j-davis(dot)com> wrote:
> The timing of your proposal is not ideal, because AI is already finding
> a lot of bugs, so finding more might be a thankless exercise until this
> wave passes. So, if it doesn't get a lot of attention that may be the
> reason why.

I conservatively estimate that I have 25 unreported correctness bugs
found by AI. No doubt some of those are invalid issues, but most
aren't.

Most of the worst bugs have already been fixed. But there are still at
least a handful of fairly bad ones. Or at least issues that have nasty
consequences when users encounter them. Some of these will be hard to
fix because they involve pretty fundamental assumptions. For example,
it's possible for deletion of LP_DEAD bits to delete items that are
still required by a transaction that started during recovery, after
recovery ends. See the attached repro.

I find it extraordinary that we're living in a world where I have an
unpublished B roll of bugs that includes issues as tricky and
potentially severe as this one. But we are. Naturally this has
significantly changed how I assess what to work on.

--
Peter Geoghegan