Horton

Re: Fixing findDependentObjects()'s dependency on scan order (regressions in DROP diagnostic messages)

Peter Geoghegan <pg@bowt.ie>

View thread View on postgr.es
From: Peter Geoghegan <pg@bowt.ie>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Alvaro Herrera <alvherre@2ndquadrant.com>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2019-01-18T01:15:23Z
Lists: pgsql-hackers 
On Thu, Jan 17, 2019 at 5:09 PM Peter Geoghegan <pg@bowt.ie> wrote:
> In the kludgey patch that I posted, the 4-byte value is manufactured
> artificially within a backend in descending order. That may have a
> slight advantage over object oid, even after the pg_depend correctness
> issues are addressed. A fixed order within a backend or originating
> transaction seems like it might avoid a few remaining instability
> issues. Not sure. I seem to recall there being some disagreement
> between you and Andres on this very point (is object oid a perfectly
> stable tie-breaker in practice?) on a similar thread from 2017.

Here are your remarks about it on that 2017 thread:
https://www.postgresql.org/message-id/11852.1501610262%40sss.pgh.pa.us

-- 
Peter Geoghegan

Commits

  1. Redesign the partition dependency mechanism.

  2. Fix trigger drop procedure

  3. Sort the dependent objects before recursing in findDependentObjects().

  4. Avoid sometimes printing both tables and their columns in DROP CASCADE.