Fix multistage aggregation plan targetlists

If there are aggregation queries with aliases same as the table actual
columns and they are propagated further from subqueries and grouping is
applied on the column alias it may result in inconsistent targetlists
for aggregation plan causing crash.

	CREATE TABLE t1 (a int) DISTRIBUTED RANDOMLY;
	SELECT substr(a, 2) as a
	FROM
		(SELECT ('-'||a)::varchar as a
			FROM (SELECT a FROM t1) t2
		) t3
	GROUP BY a;

src/backend/cdb/cdbgroup.c

@@ -3486,9 +3486,14 @@ Node* deconstruct_expr_mutator(Node *node, MppGroupContext *ctx)
 	/* If the given expression is a grouping expression, replace it with
 	 * a Var node referring to the (lower) preliminary aggregation's
 	 * target list.
+	 *
+	 * While building subplan targetlist we flatten (deduplicate) the
+	 * targetlist ignoring RelabelType node.
+	 * Including RelabelType will cause inconsistent top level target list
+	 * and final target list for aggregation plans.
 	 */
-	tle = tlist_member(node, ctx->grps_tlist);
-	if ( tle != NULL )
+	tle = tlist_member_ignore_relabel(node, ctx->grps_tlist);
+	if( tle != NULL )
 	{
 		return  (Node*) makeVar(grp_varno, tle->resno,
 								exprType((Node*)tle->expr),

src/test/regress/expected/bfv_aggregate.out

@@ -1468,6 +1468,24 @@ from mtup1 where c0 = 'foo' group by c0, c1 limit 10;
  foo | 2015-09-1.1 |         5670
 (1 row)
 
+-- MPP-29042 Multistage aggregation plans should have consistent targetlists in
+-- case of same column aliases and grouping on them.
+DROP TABLE IF EXISTS t1;
+NOTICE:  table "t1" does not exist, skipping
+CREATE TABLE t1 (a varchar) DISTRIBUTED RANDOMLY;
+INSERT INTO t1 VALUES ('aaaaaaa');
+INSERT INTO t1 VALUES ('aaaaaaa');
+INSERT INTO t1 VALUES ('bbbbbbb');
+INSERT INTO t1 VALUES ('bbbbbbb');
+INSERT INTO t1 VALUES ('bbbbb');
+SELECT substr(a, 1) as a FROM (SELECT ('-'||a)::varchar as a FROM (SELECT a FROM t1) t2) t3 GROUP BY a ORDER BY a;
+    a     
+----------
+ -aaaaaaa
+ -bbbbb
+ -bbbbbbb
+(3 rows)
+
 -- CLEANUP
 set client_min_messages='warning';
 drop schema bfv_aggregate cascade;

src/test/regress/expected/bfv_aggregate_optimizer.out

@@ -1467,6 +1467,24 @@ from mtup1 where c0 = 'foo' group by c0, c1 limit 10;
  foo | 2015-09-1.1 |         5670
 (1 row)
 
+-- MPP-29042 Multistage aggregation plans should have consistent targetlists in
+-- case of same column aliases and grouping on them.
+DROP TABLE IF EXISTS t1;
+NOTICE:  table "t1" does not exist, skipping
+CREATE TABLE t1 (a varchar) DISTRIBUTED RANDOMLY;
+INSERT INTO t1 VALUES ('aaaaaaa');
+INSERT INTO t1 VALUES ('aaaaaaa');
+INSERT INTO t1 VALUES ('bbbbbbb');
+INSERT INTO t1 VALUES ('bbbbbbb');
+INSERT INTO t1 VALUES ('bbbbb');
+SELECT substr(a, 1) as a FROM (SELECT ('-'||a)::varchar as a FROM (SELECT a FROM t1) t2) t3 GROUP BY a ORDER BY a;
+    a     
+----------
+ -aaaaaaa
+ -bbbbb
+ -bbbbbbb
+(3 rows)
+
 -- CLEANUP
 set client_min_messages='warning';
 drop schema bfv_aggregate cascade;

src/test/regress/sql/bfv_aggregate.sql

@@ -1351,6 +1351,16 @@ select c0, c1, array_length(ARRAY[
  SUM(c4 % 5670), SUM(c4 % 5671)], 1)
 from mtup1 where c0 = 'foo' group by c0, c1 limit 10;
 
+-- MPP-29042 Multistage aggregation plans should have consistent targetlists in
+-- case of same column aliases and grouping on them.
+DROP TABLE IF EXISTS t1;
+CREATE TABLE t1 (a varchar) DISTRIBUTED RANDOMLY;
+INSERT INTO t1 VALUES ('aaaaaaa');
+INSERT INTO t1 VALUES ('aaaaaaa');
+INSERT INTO t1 VALUES ('bbbbbbb');
+INSERT INTO t1 VALUES ('bbbbbbb');
+INSERT INTO t1 VALUES ('bbbbb');
+SELECT substr(a, 1) as a FROM (SELECT ('-'||a)::varchar as a FROM (SELECT a FROM t1) t2) t3 GROUP BY a ORDER BY a;
 
 -- CLEANUP
 set client_min_messages='warning';