From: Peter Geoghegan <pg@bowt.ie>
Date: Wed, 29 Jul 2020 23:00:54 +0000 (-0700)
Subject: Backpatch tuplesort.c assertion.
X-Git-Tag: REL_11_9~22
X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=87eb25535d0db49fec9b3ddb2f1c4dfd97c8805e;p=thirdparty%2Fpostgresql.git

Backpatch tuplesort.c assertion.

Backpatch an assertion (that was originally added to Postgres 12 by
commit dd299df8189) that seems broadly useful.  The assertion can detect
violations of the HOT invariant (i.e. no two index tuples can point to
the same heap TID) when CREATE INDEX somehow incorrectly allows that to
take place.

For example, a IndexBuildHeapScan/heapam_index_build_range_scan bug
might result in two tuples that both point to the same heap TID.  If
these two tuples also happen to be duplicates, the assertion will fail.

Discussion: https://postgr.es/m/CAH2-WzmBxu4o=pMsniur+bwHqCGCmV_AOLkuK6BuU7ngA6evqw@mail.gmail.com
Backpatch: 9.5-11 only
---

diff --git a/src/backend/utils/sort/tuplesort.c b/src/backend/utils/sort/tuplesort.c
index 549baf4b4f6..f770b570eb6 100644
--- a/src/backend/utils/sort/tuplesort.c
+++ b/src/backend/utils/sort/tuplesort.c
@@ -4076,6 +4076,9 @@ comparetup_index_btree(const SortTuple *a, const SortTuple *b,
 			return (pos1 < pos2) ? -1 : 1;
 	}
 
+	/* ItemPointer values should never be equal */
+	Assert(false);
+
 	return 0;
 }
 
@@ -4128,6 +4131,9 @@ comparetup_index_hash(const SortTuple *a, const SortTuple *b,
 			return (pos1 < pos2) ? -1 : 1;
 	}
 
+	/* ItemPointer values should never be equal */
+	Assert(false);
+
 	return 0;
 }