From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andres Freund <andres@anarazel.de>
cc: gomer94@yandex.ru, pgsql-bugs@postgresql.org
Subject: Re: BUG #14867: Cascade drop type error
In-reply-to: <20171023161726.5hr6rs56hxn67zz2@alap3.anarazel.de>
References: <20171023150118.1477.19174@wrigleys.postgresql.org>
 <4421.1508774964@sss.pgh.pa.us>
 <20171023161726.5hr6rs56hxn67zz2@alap3.anarazel.de>
Comments: In-reply-to Andres Freund <andres@anarazel.de>
	message dated "Mon, 23 Oct 2017 09:17:26 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <5378.1508776432.1@sss.pgh.pa.us>
Content-Transfer-Encoding: quoted-printable
Date: Mon, 23 Oct 2017 12:33:52 -0400
Message-ID: <5379.1508776432@sss.pgh.pa.us>
Precedence: bulk
Sender: pgsql-bugs-owner@postgresql.org

Andres Freund <andres@anarazel.de> writes:
> On 2017-10-23 12:09:24 -0400, Tom Lane wrote:
>> Cute.  Type my_type isn't exposed as a dependency of the view,
>> because it's only referenced internally in the expression tree
>> not as a result column type.  We can fix that easily enough by
>> teaching dependency.c to log the result type of a FieldSelect
>> as a dependency.

> That does remind me of patch 0001 in
> http://archives.postgresql.org/message-id/20170314224706.nxvtapenky6eom3z=
%40alap3.anarazel.de

It is in the same part of the code, but seems directed at a different
goal (which I disapproved of, IIRC).  But we definitely need the
principle that any type OID appearing in an expression tree needs to
be logged, except in cases where there's necessarily an indirect
dependency (eg, a function or operator's result type doesn't need to
be logged separately).

			regards, tom lane


--=20
Sent via pgsql-bugs mailing list (pgsql-bugs@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-bugs