On Mon, Nov 24, 2025 at 2:03 PM Nathan Bossart <[email protected]> wrote: > I think 0002 was more complicated than necessary. Here's a second try.
I haven't done a full review of this and I'm not sure whether you want me to spend more time on it, but something I notice about this version is that the PG_CATCH() blocks only contain dshash_delete_entry() calls. That seems good, because that means that all the other cleanup is being handled by transaction abort (assuming that the patch isn't buggy, which I haven't attempted to verify). However, it does make me wonder if we could also find a way to postpone adding the dshash entries so that we don't need to do anything in PG_CATCH() blocks at all. I'm guessing that the reason why that doesn't easily work is because you're relying on those locks to prevent multiple backends from doing the same initialization? -- Robert Haas EDB: http://www.enterprisedb.com
