Re: pg_upgrade fails with an error "object doesn't exist"

> On 16 Jun 2025, at 10:59, Vaibhav Dalvi <vaibhav(dot)dalvi(at)enterprisedb(dot)com> wrote:

> It's OFF.
> postgres=# select version();
> version
> ----------------------------------------------------------------------------------------------------------------
> PostgreSQL 18beta1 on aarch64-unknown-linux-gnu, compiled by gcc (Ubuntu 14.2.0-4ubuntu2~24.04) 14.2.0, 64-bit
> (1 row)
> postgres=# show allow_system_table_mods ;
> allow_system_table_mods
> -------------------------
> off
> (1 row)
> postgres=# CREATE FUNCTION pg_catalog.nont_ext_func() RETURNS char AS $$ BEGIN return 'v'; END; $$ LANGUAGE plpgsql;
> CREATE FUNCTION

I stand corrected, I misremembered the extent to which we prohibit creation in
pg_catalog via that GUC. It still feels like a case of getting to keep both
pieces when breaking it, but I wonder if we shouldn't make it harder to break?

--
Daniel Gustafsson