[Alexander Lakhin <exclusion@gmail.com>]

Hi,

25.04.2024 08:00, Bertrand Drouvot wrote:
>
>> though concurrent create/drop operations can still produce
>> the "cache lookup failed" error, which is probably okay, except that it is
>> an INTERNAL_ERROR, which assumed to be not easily triggered by users.
> I did not see any of those "cache lookup failed" during my testing with/without
> your script. During which test(s) did you see those with v3 applied?

You can try, for example, table-trigger, or other tests that check for
"cache lookup failed" psql output only (maybe you'll need to increase the
iteration count). For instance, I've got (with v4 applied):
2024-04-25 05:48:08.102 UTC [3638763] ERROR:  cache lookup failed for function 20893
2024-04-25 05:48:08.102 UTC [3638763] STATEMENT:  CREATE TRIGGER modified_c1 BEFORE UPDATE OF c ON t
         FOR EACH ROW WHEN (OLD.c <> NEW.c) EXECUTE PROCEDURE trigger_func('modified_c');

Or with function-function:
2024-04-25 05:52:31.390 UTC [3711897] ERROR:  cache lookup failed for function 32190 at character 54
2024-04-25 05:52:31.390 UTC [3711897] STATEMENT:  CREATE FUNCTION f1() RETURNS int LANGUAGE SQL RETURN f() + 1;
--
2024-04-25 05:52:37.639 UTC [3720011] ERROR:  cache lookup failed for function 34465 at character 54
2024-04-25 05:52:37.639 UTC [3720011] STATEMENT:  CREATE FUNCTION f1() RETURNS int LANGUAGE SQL RETURN f() + 1;

> Attached v4, simply adding more tests to v3.

Thank you for working on this!

Best regards,
Alexander