diff --git a/src/test/regress/expected/purge.out b/src/test/regress/expected/purge.out
deleted file mode 100644
index be4d2a1678..0000000000
--- a/src/test/regress/expected/purge.out
+++ /dev/null
@@ -1,28 +0,0 @@
-QUERY: PURGE hash_f8_heap BEFORE 'now';
-QUERY: SELECT count(*) AS has_10002 FROM hash_f8_heap[,] h;
-has_10002
----------
-    10002
-(1 row)
-
-QUERY: VACUUM hash_f8_heap;
-QUERY: SELECT count(*) AS has_10000 FROM hash_f8_heap[,] h;
-has_10000
----------
-    10000
-(1 row)
-
-QUERY: PURGE hash_i4_heap AFTER '@ 1 second ago';
-QUERY: SELECT count(*) AS has_10002 FROM hash_i4_heap[,] h;
-has_10002
----------
-    10002
-(1 row)
-
-QUERY: VACUUM hash_i4_heap;
-QUERY: SELECT count(*) AS has_10000 FROM hash_i4_heap[,] h;
-has_10000
----------
-    10000
-(1 row)
-
diff --git a/src/test/regress/sql/purge.sql b/src/test/regress/sql/purge.sql
deleted file mode 100644
index 37fee233df..0000000000
--- a/src/test/regress/sql/purge.sql
+++ /dev/null
@@ -1,30 +0,0 @@
---
--- PURGE
---
--- we did two updates on each of these 10K tables up above.  we should
--- therefore go from 10002 tuples (two of which are not visible without
--- using a time qual) to 10000.
---
--- vacuuming here also tests whether or not the hash index compaction
--- code works; this used to be commented out because the hash AM would
--- miss deleting a bunch of index tuples, which caused big problems when
--- you dereferenced the tids and found garbage..
---
--- absolute time
-PURGE hash_f8_heap BEFORE 'now';
-
-SELECT count(*) AS has_10002 FROM hash_f8_heap[,] h;
-
-VACUUM hash_f8_heap;
-
-SELECT count(*) AS has_10000 FROM hash_f8_heap[,] h;
-
--- relative time
-PURGE hash_i4_heap AFTER '@ 1 second ago';
-
-SELECT count(*) AS has_10002 FROM hash_i4_heap[,] h;
-
-VACUUM hash_i4_heap;
-
-SELECT count(*) AS has_10000 FROM hash_i4_heap[,] h;
-