${\color{red}Regressions (1)}$

fast_default

QUERY:          DELETE FROM leader;
RECEIVED ERROR: table not found: test_trigger() (errno 1146) (sqlstate HY000)

Changes look good. Have you already looked into the regression error?

Changes look good. Have you already looked into the regression error?

Haven't looked into the regression yet, but I suspect it's due to handling deletion cases better now. It may be deleting the function and then deleting the table, and some step in table deletion may be looking for the function and erroring when it can't find it. Previously, there were cases where we could delete something and it never actually deletes, so this error wouldn't happen in that case (assuming the test just deletes things and doesn't verify that stuff is gone).

This is just an assumption though, would have to look at the test to actually find out.

It's due to the function schema serialization location change. We used to forcefully write all functions to pg_catalog (and therefore forcefully delete from pg_catalog too), but now functions can be written to any schema. We only check pg_catalog and the current schema for now though, so if the schema changes, then you can't access those functions without explicitly stating the schema. This is a known issue though (functions not using the search path), so this is a test regression but an overall improvement to actual usage.

Put another way, I didn't fix the issue of function namespace searching, I just made it a bit better than it was before, but it's still on the list of things that need to actually be fixed.