On Thu, Oct 13, 2022 at 6:10 PM Andres Freund <and...@anarazel.de> wrote: > My point here is a lot more mundane. The code essentially does > _hash_pageinit(), overwriting the whole page, and *then* conditionally > acquires a cleanup lock. It simply is bogus code.
I understood that that was what you meant. It's easy to see why this code is broken, but to me it seems related to having too much confidence in what is possible while relying on cleanup locks. That's just my take. -- Peter Geoghegan