Date: Wed, 21 Oct 2020 14:02:37 -0300
From: Alvaro Herrera <alvherre@alvh.no-ip.org>
To: Justin Pryzby <pryzby@telsasoft.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, pgsql-hackers@postgresql.org,
	Amit Langote <amitlangote09@gmail.com>
Subject: Re: CREATE TABLE .. PARTITION OF fails to preserve tgenabled for
 inherited row triggers
Message-ID: <20201021170237.GA17401@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20201021115453.GL9241@telsasoft.com>
User-Agent: Mutt/1.10.1 (2018-07-13)
Precedence: bulk

On 2020-Oct-21, Justin Pryzby wrote:

> I came up with this, which probably needs more than a little finesse.

Hmm, there are two important changes needed on this: 1) it must not emit
CREATE lines for the child triggers; only the ALTER TABLE ONLY
<partition> lines to set disable state on the partition are needed.  2)
tgparentid does not exist prior to pg13, so you need some additional
trick to cover that case.

Also, I think the multipartition case is broken: if grandparent has
trigger enabled, parent has trigger disabled and child trigger set to
always, is that dumped correctly?  I think the right way to do this is
change only the partitions that differ from the topmost partitioned
table -- not their immediate parents; and use ONLY to ensure they don't
affect downstream children.

Change 1 also means that the test with the "this shouldn't ever get
emitted" comment remains unchanged.

I'm not sure how to tackle change 2.  You need to search pg_depend for
entries with classid=pg_trigger and refclass=pg_trigger ... (commit
1fa846f1c9af might give some clue)