On 2021-Jan-19, Michael Paquier wrote: > And while reviewing the thing, I have spotted that there is a specific > path for pg_default_acl in RemoveRoleFromObjectACL() that has zero > coverage. This can be triggered with DROP OWNED BY, and it is > actually safe to run as long as this is done in a separate transaction > to avoid any interactions with parallel regression sessions. > privileges.sql already has similar tests, so I'd like to add some > coverage as per the attached (the duplicated role name is wanted).
Heh, interesting case. Added coverage is good, so +1. Since the role regress_priv_user2 is "private" to the privileges.sql script, there's no danger of a concurrent test getting the added lines in trouble AFAICS. > +SELECT count(*) FROM pg_shdepend > + WHERE deptype = 'a' AND > + refobjid = 'regress_priv_user2'::regrole AND > + classid = 'pg_default_acl'::regclass; > + count > +------- > + 5 > +(1 row) Shrug. Seems sufficient. -- Álvaro Herrera Valdivia, Chile
