I'd like to suggest a patch for reloption regression tests.
This patch tests case, that can be rarely met in actual life: when reloptions
have some illegal option set (as a result of malfunction or extension
downgrade or something), and user tries to remove this option by using RESET.
Current postgres behaviour is to actually remove this option.
Like:
UPDATE pg_class SET reloptions = '{illegal_option=4}'
WHERE oid = 'reloptions_test'::regclass;
ALTER TABLE reloptions_test RESET (illegal_option);
Why this should be tested:
1. It is what postgres actually do now.
2. This behaviour is reasonable. DB User can fix problem without updating
pg_class, having rights to change his own table.
3. Better to get test alarm, if this behavior is accidentally changed.
--
Nikolay Shaplov aka Nataraj
Fuzzing Engineer at Postgres Professional
Matrix IM: @dhyan:nataraj.su