Re: BUG #17959: amcheck fails to find a matching index tuple for an invisible heap tuple

On Mon, Jun 5, 2023 at 12:29 AM PG Bug reporting form
<noreply(at)postgresql(dot)org> wrote:
> SELECT ctid, * FROM pg_depend WHERE refclassid = 0x0a37 AND refobjid =
> 0x4036 AND refobjsubid = 0
> doesn't return any rows.
>
> Shouldn't amcheck ignore invisible tuples?

It should -- so there must be a bug. This is a system catalog index,
so I wonder if this issue is in any way related to this known issue:

https://www.postgresql.org/message-id/CAH2-WzkjjCoq5Y4LeeHJcjYJVxGm3M3SAWZ0%3D6J8K1FPSC9K0w%40mail.gmail.com

(I've been meaning to get around to finally fixing it.)

Admittedly this is a fairly wild guess -- the details don't really
match. Even still, the fact that this is a system catalog index seems
very unlikely to be incidental to the problem. There are some
significant differences between how system indexes and other indexes
are built in heapam_index_build_range_scan(). Those differences seem
like they could easily be relevant.

--
Peter Geoghegan