Re: pg_restore with --disable-triggers discards ENABLE ALWAYS
Tom Lane <tgl@sss.pgh.pa.us>
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Laurenz Albe <laurenz.albe@cybertec.at>
Cc: Duncan Sands <duncan.sands@deepbluecap.com>,
pgsql-bugs@lists.postgresql.org
Date: 2024-09-12T20:47:07Z
Lists: pgsql-bugs
Laurenz Albe <laurenz.albe@cybertec.at> writes: > On Thu, 2024-09-12 at 10:27 +0200, Duncan Sands wrote: >> pg_restore --data-only --disable-triggers --dbname duncan dump.custom >> ^ Observe that "Triggers firing always" has disappeared. > This looks like a user error to me. > If you restore with --data-only, the table and constraint definitions > are not restored at all. So the table "test_table" in database "duncan" > must already have existed before the restore, and the trigger was already > defined like that. The given recipe seems incomplete, but I think Duncan has put his finger on a shortcoming. pg_restore with --disable-triggers will issue ALTER TABLE foo DISABLE TRIGGER ALL; and later ALTER TABLE foo ENABLE TRIGGER ALL; and if you read the fine print in the ALTER TABLE man page, you'll discover that ENABLE TRIGGER sets the triggers' replication mode to the default (origin only). I'm not sure why somebody thought that replication mode shouldn't be stored separately from enable/disable, but it isn't: there's just one four-valued state field. We could probably add code to make pg_dump individually re-enable the table's triggers with the appropriate state(s), but I can't muster too much enthusiasm for writing that myself. regards, tom lane