Laurenz Albe <laurenz.albe@xxxxxxxxxxx> writes:
> On Tue, 2023-01-17 at 12:46 +0200, Kouber Saparev wrote:
>> Any ideas how we could proceed any further?
>> And btw what is the entry above: objid = 1202633909 + deptype = i?
> There was probably a dependency *missing*: you get this error message if
> the database wants to delete the dependency on the old schema and add
> one on the new schema, but it cannot find the former.
> This is data corruption, but considering the way you delete catalog entries,
> I am not surprised.
Indeed. Personally, when I want to look into pg_depend, I invariably
use pg_describe_object() to make sense of the entries.
regression=# create type foo as enum ('a','b');
CREATE TYPE
regression=# create schema bar;
CREATE SCHEMA
regression=# alter type public.foo set schema bar;
ALTER TYPE
regression=# create schema baz;
CREATE SCHEMA
regression=# alter type bar.foo set schema baz;
ALTER TYPE
regression=# select pg_describe_object(classid, objid, objsubid) as obj, pg_describe_object(refclassid, refobjid, refobjsubid) as ref, deptype from pg_depend where refobjid = 'baz.foo'::regtype;
obj | ref | deptype
----------------+--------------+---------
type baz.foo[] | type baz.foo | i
(1 row)
regression=# select pg_describe_object(classid, objid, objsubid) as obj, pg_describe_object(refclassid, refobjid, refobjsubid) as ref, deptype from pg_depend where objid = 'baz.foo'::regtype;
obj | ref | deptype
--------------+------------+---------
type baz.foo | schema baz | n
(1 row)
If you'd done it like that, you would probably have figured out fairly
quickly that you were looking at the wrong end of the dependency
relationships. See also
https://www.postgresql.org/docs/current/catalog-pg-depend.html
regards, tom lane
