Date: Thu, 21 Mar 2019 17:49:28 -0300
From: Alvaro Herrera <alvherre@2ndquadrant.com>
To: Jesper Pedersen <jesper.pedersen@redhat.com>
Cc: Amit Langote <Langote_Amit_f8@lab.ntt.co.jp>,
	Pg Hackers <pgsql-hackers@postgresql.org>
Subject: Re: partitioned tables referenced by FKs
Message-ID: <20190321204928.GA17535@alvherre.pgsql>
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="0F1p//8PRICkK4MW"
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20190321201704.GA8280@alvherre.pgsql>
User-Agent: Mutt/1.9.4 (2018-02-28)
Precedence: bulk


--0F1p//8PRICkK4MW
Content-Type: text/plain; charset=iso-8859-1
Content-Disposition: inline
Content-Transfer-Encoding: 8bit

On 2019-Mar-21, Alvaro Herrera wrote:

> I figured this out.  It's actually a bug in pg11, whereby we're setting
> a dependency wrongly.  If you try to do pg_describe_object() the
> pg_depend entries for tables set up the way the regression test does it,
> it'll fail with a "cache lookup failed for constraint XYZ".  In other
> words, pg_depend contains bogus data :-(

Oh, actually, it's not a problem in PG11 ... or at least, it's not THIS
problem in pg11.  The bug was introduced by 1d92a0c9f7dd which ended the
DEPENDENCY_INTERNAL_AUTO saga, so the bug is only in master, and that
explains why I hadn't seen the problem before with my posted patch
series.

I think the attached is a better solution, which I'll go push shortly.

-- 
Álvaro Herrera                https://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services

--0F1p//8PRICkK4MW
Content-Type: text/x-diff; charset=us-ascii
Content-Disposition: attachment; filename="shadow.patch"

diff --git a/src/backend/catalog/index.c b/src/backend/catalog/index.c
index c339a2bb779..cb2c0010171 100644
--- a/src/backend/catalog/index.c
+++ b/src/backend/catalog/index.c
@@ -1247,7 +1247,7 @@ index_constraint_create(Relation heapRelation,
 {
 	Oid			namespaceId = RelationGetNamespace(heapRelation);
 	ObjectAddress myself,
-				referenced;
+				idxaddr;
 	Oid			conOid;
 	bool		deferrable;
 	bool		initdeferred;
@@ -1341,15 +1341,9 @@ index_constraint_create(Relation heapRelation,
 	 * Note that the constraint has a dependency on the table, so we don't
 	 * need (or want) any direct dependency from the index to the table.
 	 */
-	myself.classId = RelationRelationId;
-	myself.objectId = indexRelationId;
-	myself.objectSubId = 0;
-
-	referenced.classId = ConstraintRelationId;
-	referenced.objectId = conOid;
-	referenced.objectSubId = 0;
-
-	recordDependencyOn(&myself, &referenced, DEPENDENCY_INTERNAL);
+	ObjectAddressSet(myself, ConstraintRelationId, conOid);
+	ObjectAddressSet(idxaddr, RelationRelationId, indexRelationId);
+	recordDependencyOn(&idxaddr, &myself, DEPENDENCY_INTERNAL);
 
 	/*
 	 * Also, if this is a constraint on a partition, give it partition-type
@@ -1357,7 +1351,8 @@ index_constraint_create(Relation heapRelation,
 	 */
 	if (OidIsValid(parentConstraintId))
 	{
-		ObjectAddressSet(myself, ConstraintRelationId, conOid);
+		ObjectAddress	referenced;
+
 		ObjectAddressSet(referenced, ConstraintRelationId, parentConstraintId);
 		recordDependencyOn(&myself, &referenced, DEPENDENCY_PARTITION_PRI);
 		ObjectAddressSet(referenced, RelationRelationId,
@@ -1444,7 +1439,7 @@ index_constraint_create(Relation heapRelation,
 		table_close(pg_index, RowExclusiveLock);
 	}
 
-	return referenced;
+	return myself;
 }
 
 /*

--0F1p//8PRICkK4MW--