Horton

Re: BUG #16644: null value for defaults in OLD variable for trigger

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Amit Langote <amitlangote09@gmail.com>
Cc: fedor_erastov@mail.ru, PostgreSQL mailing lists <pgsql-bugs@lists.postgresql.org>, Andres Freund <andres@anarazel.de>, Andrew Dunstan <andrew@dunslane.net>
Date: 2020-10-26T03:19:04Z
Lists: pgsql-bugs 
Amit Langote <amitlangote09@gmail.com> writes:
> On Mon, Oct 26, 2020 at 5:53 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Moreover, it's more correct, even disregarding the problem
>> at hand, because the tlist isn't a perfectly accurate depiction of
>> the relation rowtype: ExecCleanTypeFromTL will not derive the correct
>> info for dropped columns.

> Hmm, I don't understand.  Isn't it the planner's job to make the
> targetlist correctly account for dropped columns; what
> expand_targetlist() does?

Yes, there are columns in the tlist to match them, but ExecCleanTypeFromTL
cannot mark those columns as "attisdropped".  The column data type
likely won't be right either.  The latter shouldn't matter, if the
column is being filled with a null ... but I'm a bit surprised that
we've gotten away this long with not being honest about attisdropped.

			regards, tom lane

Commits

  1. In INSERT/UPDATE, use the table's real tuple descriptor as target.

  2. Fix corner case for a BEFORE ROW UPDATE trigger returning OLD.

  3. Use slots in trigger infrastructure, except for the actual invocation.