Consistently truncate non-key suffix columns.

petergeoghegan · petergeoghegan · commit 4b42a89938ac · 2020-03-30T12:03:59.000-07:00
INCLUDE indexes failed to have their non-key attributes physically truncated away in certain rare cases. This led to physically larger pivot tuples that contained useless non-key attribute values. The impact on users should be negligible, but this is still clearly a regression (Postgres 11 supports INCLUDE indexes, and yet was not affected). The bug appeared in commit dd299df, which introduced "true" suffix truncation of key attributes. Discussion: https://postgr.es/m/CAH2-Wz=E8pkV9ivRSFHtv812H5ckf8s1-yhx61_WrJbKccGcrQ@mail.gmail.com Backpatch: 12-, where "true" suffix truncation was introduced.
diff --git a/src/backend/access/nbtree/nbtutils.c b/src/backend/access/nbtree/nbtutils.c
@@ -2205,7 +2205,8 @@ _bt_truncate(Relation rel, IndexTuple lastleft, IndexTuple firstright,
 	{
 		IndexTuple	tidpivot;
 
-		pivot = index_truncate_tuple(itupdesc, firstright, keepnatts);
+		pivot = index_truncate_tuple(itupdesc, firstright,
+									 Min(keepnatts, nkeyatts));
 
 		if (BTreeTupleIsPosting(pivot))
 		{
@@ -2250,6 +2251,10 @@ _bt_truncate(Relation rel, IndexTuple lastleft, IndexTuple firstright,
 		/*
 		 * No truncation was possible, since key attributes are all equal.
 		 * It's necessary to add a heap TID attribute to the new pivot tuple.
+		 *
+		 * This path is only taken when rel is not an INCLUDE index.  It
+		 * avoids a second palloc0() by avoiding the index_truncate_tuple()
+		 * call completely.
 		 */
 		Assert(natts == nkeyatts);
 		newsize = IndexTupleSize(firstright) + MAXALIGN(sizeof(ItemPointerData));
