On Tue, Jul 6, 2021 at 3:12 PM Mark Dilger <mark.dilger@enterprisedb.com> wrote:
> Thanks, Peter, for drawing my attention to this. I had already been following this thread, but had not yet thought
aboutthe problem in terms of amcheck.
>
> I will investigate possible solutions in verify_heapam().
Thanks! Great that we might be able to make a whole class of bugs
detectable with the new amcheck stuff. Glad that I didn't forget about
amcheck myself -- I almost forgot.
When I was working on the btree amcheck code, I looked for interesting
historical bugs and made sure that I could detect them. That seems
even more important with heapam. I wouldn't be surprised if one or two
important invariants were missed, in part because the heapam design
didn't have invariants as a starting point.
--
Peter Geoghegan