Import wiredtiger: bb8086c4ed645854f19554198f391ea8c821f603 from branch mongodb-4.2

ref: 67066a329d..bb8086c4ed
for: 4.1.10

WT-4614       Trigger sweep immediately to reclaim space after a table drop
WT-4631       Always clear the read timestamp for transactions
WT-4636       Fix strace in syscall test
WT-4650       Coverity complaint: Constant expression result (CONSTANT_EXPRESSION_RESULT)
WT-4668       lint

12 files changed, 199 insertions, 64 deletions

diff --git a/src/third_party/wiredtiger/dist/s_define.list b/src/third_party/wiredtiger/dist/s_define.list
index f199900e860..d585c1e268d 100644
--- a/src/third_party/wiredtiger/dist/s_define.list
+++ b/src/third_party/wiredtiger/dist/s_define.list
@@ -72,6 +72,7 @@ WT_TRACK_OP_END
 WT_TRACK_OP_INIT
 WT_TRET_ERROR_OK
 WT_UPDATE_SIZE
+WT_USE_OPENAT
 WT_WITH_LOCK_NOWAIT
 WT_WITH_LOCK_WAIT
 __F
diff --git a/src/third_party/wiredtiger/dist/s_string.ok b/src/third_party/wiredtiger/dist/s_string.ok
index c2e70ba2646..bc265efb34f 100644
--- a/src/third_party/wiredtiger/dist/s_string.ok
+++ b/src/third_party/wiredtiger/dist/s_string.ok
@@ -156,6 +156,7 @@ Fsync
 Fuerst
 GCC
 GIDs
+GLIBC
 Gcc
 Geoff
 GetEnvironmentVariableA
@@ -1052,6 +1053,7 @@ online
 onpage
 oo
 opcode
+openat
 opendir
 openfile
 oplist
@@ -1214,6 +1216,7 @@ stdin
 stdout
 stepp
 str
+strace
 strcmp
 strdup
 strerror
diff --git a/src/third_party/wiredtiger/import.data b/src/third_party/wiredtiger/import.data
index c9b3aed9f2d..340e9ea128f 100644
--- a/src/third_party/wiredtiger/import.data
+++ b/src/third_party/wiredtiger/import.data
@@ -1,5 +1,5 @@
 {
-    "commit": "67066a329d365cb1de6e198184016d4a9d280679", 
+    "commit": "bb8086c4ed645854f19554198f391ea8c821f603", 
     "github": "wiredtiger/wiredtiger.git", 
     "vendor": "wiredtiger", 
     "branch": "mongodb-4.2"
diff --git a/src/third_party/wiredtiger/src/btree/bt_vrfy_dsk.c b/src/third_party/wiredtiger/src/btree/bt_vrfy_dsk.c
index 844da3f041b..ee6cd904aec 100644
--- a/src/third_party/wiredtiger/src/btree/bt_vrfy_dsk.c
+++ b/src/third_party/wiredtiger/src/btree/bt_vrfy_dsk.c
@@ -43,7 +43,7 @@ static int __verify_dsk_row(
 } while (0)
 
 #define	WT_RET_VRFY(session, ...)					\
-	WT_RET_VRFY_RETVAL(session, 0, __VA_ARGS__);
+	WT_RET_VRFY_RETVAL(session, 0, __VA_ARGS__)
 
 /*
  * WT_CELL_FOREACH_VRFY --
diff --git a/src/third_party/wiredtiger/src/cache/cache_las.c b/src/third_party/wiredtiger/src/cache/cache_las.c
index d7175c32275..905c24dceae 100644
--- a/src/third_party/wiredtiger/src/cache/cache_las.c
+++ b/src/third_party/wiredtiger/src/cache/cache_las.c
@@ -1142,15 +1142,17 @@ __wt_las_sweep(WT_SESSION_IMPL *session)
 			    session, saved_key, las_key.data, las_key.size));
 
 			/*
-			 * Never expect an entry with prepare locked state or
-			 * with durable timestamp as max timestamp or with
-			 * in-progress prepare state and non-zero durable
-			 * timestamp. In all other cases the durable timestamp
-			 * is higher or same as the las timestamp.
+			 * Expect an update entry with:
+			 *	1. not in a prepare locked state
+			 *	2. durable timestamp as not max timestamp.
+			 *	3. for an in-progress prepared update, durable
+			 *	timestamp should be zero.
+			 *	4. no restriction on durable timestamp value
+			 *	for other updates.
 			 */
 			WT_ASSERT(session,
-			    prepare_state != WT_PREPARE_LOCKED ||
-			    durable_timestamp != WT_TS_MAX ||
+			    prepare_state != WT_PREPARE_LOCKED &&
+			    durable_timestamp != WT_TS_MAX &&
 			    (prepare_state != WT_PREPARE_INPROGRESS ||
 			    durable_timestamp == 0));
 
diff --git a/src/third_party/wiredtiger/src/conn/conn_sweep.c b/src/third_party/wiredtiger/src/conn/conn_sweep.c
index d07d32c41c9..7b755c28854 100644
--- a/src/third_party/wiredtiger/src/conn/conn_sweep.c
+++ b/src/third_party/wiredtiger/src/conn/conn_sweep.c
@@ -279,6 +279,7 @@ __sweep_server(void *arg)
 	uint64_t last, now;
 	uint64_t last_las_sweep_id, min_sleep, oldest_id, sweep_interval;
 	u_int dead_handles;
+	bool cv_signalled;
 
 	session = arg;
 	conn = S2C(session);
@@ -298,12 +299,13 @@ __sweep_server(void *arg)
 		/* Wait until the next event. */
 		if (FLD_ISSET(conn->timing_stress_flags,
 		    WT_TIMING_STRESS_AGGRESSIVE_SWEEP))
-			__wt_cond_wait(session, conn->sweep_cond,
+			__wt_cond_wait_signal(session, conn->sweep_cond,
 			    min_sleep * 100 * WT_THOUSAND,
-			    __sweep_server_run_chk);
+			    __sweep_server_run_chk, &cv_signalled);
 		else
-			__wt_cond_wait(session, conn->sweep_cond,
-			    min_sleep * WT_MILLION, __sweep_server_run_chk);
+			__wt_cond_wait_signal(session,
+			    conn->sweep_cond, min_sleep * WT_MILLION,
+			    __sweep_server_run_chk, &cv_signalled);
 
 		/* Check if we're quitting or being reconfigured. */
 		if (!__sweep_server_run_chk(session))
@@ -339,7 +341,7 @@ __sweep_server(void *arg)
 		 * less frequently than the lookaside table by default and the
 		 * frequency is controlled by a user setting.
 		 */
-		if (now - last < sweep_interval)
+		if (!cv_signalled && (now - last < sweep_interval))
 			continue;
 		WT_STAT_CONN_INCR(session, dh_sweeps);
 		/*
diff --git a/src/third_party/wiredtiger/src/include/wt_internal.h b/src/third_party/wiredtiger/src/include/wt_internal.h
index 87437cd9ed3..4d03a40b860 100644
--- a/src/third_party/wiredtiger/src/include/wt_internal.h
+++ b/src/third_party/wiredtiger/src/include/wt_internal.h
@@ -337,6 +337,16 @@ typedef uint64_t wt_timestamp_t;
 #elif defined(_MSC_VER)
 #include "msvc.h"
 #endif
+/*
+ * GLIBC 2.26 and later use the openat syscall to implement open.
+ * Set this flag so that our strace tests know to expect this.
+ */
+#ifdef __GLIBC_PREREQ
+#if __GLIBC_PREREQ(2, 26)
+#define	WT_USE_OPENAT 1
+#endif
+#endif
+
 #include "hardware.h"
 #include "swap.h"
 
diff --git a/src/third_party/wiredtiger/src/meta/meta_track.c b/src/third_party/wiredtiger/src/meta/meta_track.c
index 8012eafe650..f411b9675eb 100644
--- a/src/third_party/wiredtiger/src/meta/meta_track.c
+++ b/src/third_party/wiredtiger/src/meta/meta_track.c
@@ -249,6 +249,7 @@ __wt_meta_track_off(WT_SESSION_IMPL *session, bool need_sync, bool unroll)
 	WT_META_TRACK *trk, *trk_orig;
 	WT_SESSION_IMPL *ckpt_session;
 	int saved_ret;
+	bool did_drop;
 
 	saved_ret = 0;
 
@@ -323,14 +324,20 @@ err:	/*
 	 * Undo any tracked operations on failure.
 	 * Apply any tracked operations post-commit.
 	 */
+	did_drop = false;
 	if (unroll || ret != 0) {
 		saved_ret = ret;
 		ret = 0;
-		while (--trk >= trk_orig)
+		while (--trk >= trk_orig) {
+			did_drop = did_drop || trk->op == WT_ST_DROP_COMMIT;
 			WT_TRET(__meta_track_unroll(session, trk));
+		}
 	} else
-		for (; trk_orig < trk; trk_orig++)
+		for (; trk_orig < trk; trk_orig++) {
+			did_drop = did_drop ||
+			    trk_orig->op == WT_ST_DROP_COMMIT;
 			WT_TRET(__meta_track_apply(session, trk_orig));
+		}
 
 	if (F_ISSET(session, WT_SESSION_SCHEMA_TXN)) {
 		F_CLR(session, WT_SESSION_SCHEMA_TXN);
@@ -346,6 +353,13 @@ err:	/*
 #endif
 	}
 
+	/*
+	 * Wake up the sweep thread: particularly for the in-memory
+	 * storage engine, we want to reclaim space immediately.
+	 */
+	if (did_drop && S2C(session)->sweep_cond != NULL)
+		__wt_cond_signal(session, S2C(session)->sweep_cond);
+
 	if (ret != 0)
 		WT_PANIC_RET(session, ret,
 		    "failed to apply or unroll all tracked operations");
diff --git a/src/third_party/wiredtiger/src/txn/txn_timestamp.c b/src/third_party/wiredtiger/src/txn/txn_timestamp.c
index 594be12e6b1..f4e340535ef 100644
--- a/src/third_party/wiredtiger/src/txn/txn_timestamp.c
+++ b/src/third_party/wiredtiger/src/txn/txn_timestamp.c
@@ -150,6 +150,20 @@ __wt_txn_parse_timestamp(WT_SESSION_IMPL *session, const char *name,
 }
 
 /*
+ * __txn_get_read_timestamp --
+ *	Get the read timestamp from the transaction. Additionally
+ *	return bool to specify whether the transaction has set
+ *	clear read queue flag.
+ */
+static bool
+__txn_get_read_timestamp(
+    WT_TXN *txn, wt_timestamp_t *read_timestampp)
+{
+	WT_ORDERED_READ(*read_timestampp, txn->read_timestamp);
+	return (!txn->clear_read_q);
+}
+
+/*
  * __txn_get_pinned_timestamp --
  *	Calculate the current pinned timestamp.
  */
@@ -160,7 +174,7 @@ __txn_get_pinned_timestamp(
 	WT_CONNECTION_IMPL *conn;
 	WT_TXN *txn;
 	WT_TXN_GLOBAL *txn_global;
-	wt_timestamp_t tmp_ts;
+	wt_timestamp_t tmp_read_ts, tmp_ts;
 	bool include_oldest, txn_has_write_lock;
 
 	conn = S2C(session);
@@ -189,15 +203,18 @@ __txn_get_pinned_timestamp(
 	TAILQ_FOREACH(txn, &txn_global->read_timestamph, read_timestampq) {
 		/*
 		 * Skip any transactions on the queue that are not active.
+		 * Copy out value of read timestamp to prevent possible
+		 * race where a transaction resets its read timestamp while
+		 * we traverse the queue.
 		 */
-		if (txn->clear_read_q)
+		if (!__txn_get_read_timestamp(txn, &tmp_read_ts))
 			continue;
 		/*
 		 * A zero timestamp is possible here only when the oldest
 		 * timestamp is not accounted for.
 		 */
-		if (tmp_ts == 0 || txn->read_timestamp < tmp_ts)
-			tmp_ts = txn->read_timestamp;
+		if (tmp_ts == 0 || tmp_read_ts < tmp_ts)
+			tmp_ts = tmp_read_ts;
 		/*
 		 * We break on the first active txn on the list.
 		 */
@@ -638,22 +655,18 @@ __wt_txn_commit_timestamp_validate(WT_SESSION_IMPL *session, const char *name,
 	 * Compare against the prepare timestamp of the current transaction.
 	 * Return an error if the given timestamp is older than the prepare
 	 * timestamp.
+	 *
+	 * If roundup timestamps is configured, the commit timestamp will be
+	 * rounded up to the prepare timestamp.
 	 */
-	if (F_ISSET(txn, WT_TXN_PREPARE) && ts < txn->prepare_timestamp) {
-		/*
-		 * If roundup timestamps is configured, the commit timestamp
-		 * will be rounded up to the prepare timestamp.
-		 */
-		if (F_ISSET(txn, WT_TXN_TS_ROUND_PREPARED))
-			ts = txn->prepare_timestamp;
-		else {
-			__wt_timestamp_to_string(
-			    txn->prepare_timestamp, ts_string[0]);
-			WT_RET_MSG(session, EINVAL,
-			    "%s timestamp %.*s older than the prepare "
-			    "timestamp %s for this transaction",
-			    name, (int)cval->len, cval->str, ts_string[0]);
-		}
+	if (F_ISSET(txn, WT_TXN_PREPARE) &&
+	    !F_ISSET(txn, WT_TXN_TS_ROUND_PREPARED) &&
+	    ts < txn->prepare_timestamp) {
+		__wt_timestamp_to_string(txn->prepare_timestamp, ts_string[0]);
+		WT_RET_MSG(session, EINVAL,
+		    "%s timestamp %.*s older than the prepare "
+		    "timestamp %s for this transaction",
+		    name, (int)cval->len, cval->str, ts_string[0]);
 	}
 
 	if (F_ISSET(txn, WT_TXN_HAS_TS_DURABLE) &&
@@ -754,7 +767,7 @@ __wt_txn_parse_prepare_timestamp(
 	WT_CONFIG_ITEM cval;
 	WT_TXN *prev, *txn;
 	WT_TXN_GLOBAL *txn_global;
-	wt_timestamp_t oldest_ts, timestamp;
+	wt_timestamp_t oldest_ts, timestamp, tmp_timestamp;
 	char ts_string[2][WT_TS_INT_STRING_SIZE];
 
 	txn = &session->txn;
@@ -779,18 +792,23 @@ __wt_txn_parse_prepare_timestamp(
 	prev = TAILQ_LAST(
 	    &txn_global->read_timestamph, __wt_txn_rts_qh);
 	while (prev != NULL) {
-		/* Skip self and non-active transactions. */
-		if (prev->clear_read_q || prev == txn) {
+		/*
+		 * Skip self and non-active transactions. Copy out value of
+		 * read timestamp to prevent possible race where a transaction
+		 * resets its read timestamp while we traverse the queue.
+		 */
+		if (!__txn_get_read_timestamp(prev, &tmp_timestamp) ||
+		    prev == txn) {
 			prev = TAILQ_PREV(
 			    prev, __wt_txn_rts_qh, read_timestampq);
 			continue;
 		}
 
-		if (prev->read_timestamp >= timestamp) {
+		if (tmp_timestamp >= timestamp) {
 			__wt_readunlock(session,
 			    &txn_global->read_timestamp_rwlock);
 			__wt_timestamp_to_string(
-			    prev->read_timestamp, ts_string[0]);
+			    tmp_timestamp, ts_string[0]);
 			WT_RET_MSG(session, EINVAL,
 			    "prepare timestamp %.*s must be greater than the "
 			    "latest active read timestamp %s ",
@@ -1076,6 +1094,7 @@ __wt_txn_set_read_timestamp(WT_SESSION_IMPL *session)
 {
 	WT_TXN *qtxn, *txn, *txn_tmp;
 	WT_TXN_GLOBAL *txn_global;
+	wt_timestamp_t tmp_timestamp;
 	uint64_t walked;
 
 	txn = &session->txn;
@@ -1127,11 +1146,14 @@ __wt_txn_set_read_timestamp(WT_SESSION_IMPL *session)
 		 */
 		qtxn = TAILQ_LAST(
 		     &txn_global->read_timestamph, __wt_txn_rts_qh);
-		while (qtxn != NULL &&
-		    qtxn->read_timestamp > txn->read_timestamp) {
-			++walked;
-			qtxn = TAILQ_PREV(
-			    qtxn, __wt_txn_rts_qh, read_timestampq);
+		while (qtxn != NULL) {
+			if (!__txn_get_read_timestamp(qtxn, &tmp_timestamp) ||
+			    tmp_timestamp > txn->read_timestamp) {
+				++walked;
+				qtxn = TAILQ_PREV(qtxn,
+				    __wt_txn_rts_qh, read_timestampq);
+			} else
+				break;
 		}
 		if (qtxn == NULL) {
 			TAILQ_INSERT_HEAD(&txn_global->read_timestamph,
@@ -1165,9 +1187,10 @@ __wt_txn_clear_read_timestamp(WT_SESSION_IMPL *session)
 
 	txn = &session->txn;
 
-	if (!F_ISSET(txn, WT_TXN_PUBLIC_TS_READ))
+	if (!F_ISSET(txn, WT_TXN_PUBLIC_TS_READ)) {
+		txn->read_timestamp = WT_TS_NONE;
 		return;
-
+	}
 #ifdef HAVE_DIAGNOSTIC
 	{
 	WT_TXN_GLOBAL *txn_global;
@@ -1188,6 +1211,7 @@ __wt_txn_clear_read_timestamp(WT_SESSION_IMPL *session)
 	 */
 	WT_PUBLISH(txn->clear_read_q, true);
 	WT_PUBLISH(txn->flags, flags);
+	txn->read_timestamp = WT_TS_NONE;
 }
 
 /*
diff --git a/src/third_party/wiredtiger/test/suite/test_timestamp16.py b/src/third_party/wiredtiger/test/suite/test_timestamp16.py
new file mode 100644
index 00000000000..259503b7d26
--- /dev/null
+++ b/src/third_party/wiredtiger/test/suite/test_timestamp16.py
@@ -0,0 +1,73 @@
+#!/usr/bin/env python
+#
+# Public Domain 2014-2019 MongoDB, Inc.
+# Public Domain 2008-2014 WiredTiger, Inc.
+#
+# This is free and unencumbered software released into the public domain.
+#
+# Anyone is free to copy, modify, publish, use, compile, sell, or
+# distribute this software, either in source code form or as a compiled
+# binary, for any purpose, commercial or non-commercial, and by any
+# means.
+#
+# In jurisdictions that recognize copyright laws, the author or authors
+# of this software dedicate any and all copyright interest in the
+# software to the public domain. We make this dedication for the benefit
+# of the public at large and to the detriment of our heirs and
+# successors. We intend this dedication to be an overt act of
+# relinquishment in perpetuity of all present and future rights to this
+# software under copyright law.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+# EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
+# IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+# OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+# ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+# OTHER DEALINGS IN THE SOFTWARE.
+#
+# test_timestamp16.py
+#   Test to ensure read timestamp is properly cleared at the
+#   end of a txn.
+#
+
+import random
+from suite_subprocess import suite_subprocess
+import wiredtiger, wttest
+from wtscenario import make_scenarios
+
+def timestamp_str(t):
+    return '%x' % t
+
+class test_timestamp16(wttest.WiredTigerTestCase, suite_subprocess):
+    tablename = 'test_timestamp16'
+    uri = 'table:' + tablename
+
+    def test_read_timestamp_cleared(self):
+        # Ensure that the read timestamp doesn't move our checkpoint.
+        self.session.create(self.uri, 'key_format=i,value_format=i')
+        self.session.begin_transaction('read_timestamp=100')
+        self.session.rollback_transaction()
+        self.session.checkpoint('use_timestamp=true')
+        self.assertTimestampsEqual('0',
+            self.conn.query_timestamp('get=last_checkpoint'))
+
+        # Set a stable and make sure that we still checkpoint at
+        # the stable.
+        self.conn.set_timestamp('stable_timestamp=1')
+        self.session.begin_transaction('read_timestamp=100')
+        self.session.rollback_transaction()
+        self.session.checkpoint('use_timestamp=true')
+        self.assertTimestampsEqual('1',
+            self.conn.query_timestamp('get=last_checkpoint'))
+
+        # Finally make sure that commit also resets the read timestamp.
+        self.session.create(self.uri, 'key_format=i,value_format=i')
+        self.session.begin_transaction('read_timestamp=150')
+        self.session.commit_transaction()
+        self.session.checkpoint('use_timestamp=true')
+        self.assertTimestampsEqual('1',
+            self.conn.query_timestamp('get=last_checkpoint'))
+
+if __name__ == '__main__':
+    wttest.run()
\ No newline at end of file
diff --git a/src/third_party/wiredtiger/test/syscall/syscall.py b/src/third_party/wiredtiger/test/syscall/syscall.py
index 6d5f4bc4c10..0fafc56fe45 100644
--- a/src/third_party/wiredtiger/test/syscall/syscall.py
+++ b/src/third_party/wiredtiger/test/syscall/syscall.py
@@ -156,7 +156,7 @@ defines_used = [
     'HAVE_FTRUNCATE', 'O_ACCMODE', 'O_APPEND', 'O_ASYNC',
     'O_CLOEXEC', 'O_CREAT', 'O_EXCL', 'O_EXLOCK', 'O_NOATIME',
     'O_NOFOLLOW', 'O_NONBLOCK', 'O_RDONLY', 'O_RDWR', 'O_SHLOCK',
-    'O_TRUNC', 'O_WRONLY' ]
+    'O_TRUNC', 'O_WRONLY', 'WT_USE_OPENAT' ]
 
 ################################################################
 
diff --git a/src/third_party/wiredtiger/test/syscall/wt2336_base/base.run b/src/third_party/wiredtiger/test/syscall/wt2336_base/base.run
index 0dc46548c5b..b51e24a2c06 100644
--- a/src/third_party/wiredtiger/test/syscall/wt2336_base/base.run
+++ b/src/third_party/wiredtiger/test/syscall/wt2336_base/base.run
@@ -32,11 +32,11 @@
  */
 #ifdef __linux__
 SYSTEM("Linux");
-#define OPEN_EXISTING(name, flags)		open(name, flags)
+#define OPEN_EXISTING(name, flags)		OPEN(name, flags)
 #else /* __linux__ */
 SYSTEM("Darwin");
 #define O_NOATIME 0
-#define OPEN_EXISTING(name, flags)		open(name, flags, 0)
+#define OPEN_EXISTING(name, flags)		OPEN(name, flags, 0)
 #endif /* __linux__ */
 
 #ifdef HAVE_FTRUNCATE
@@ -49,14 +49,20 @@ SYSTEM("Darwin");
 #define FTRUNCATE(fd, len)			/* do nothing */
 #endif
 
-TRACE("close,fdatasync,fsync,ftruncate,open,pwrite64,rename");
+#ifdef WT_USE_OPENAT
+#define OPEN(...)				openat(AT_FDCWD, __VA_ARGS__)
+#else
+#define OPEN(...)				open(__VA_ARGS__)
+#endif
+
+TRACE("close,fdatasync,fsync,ftruncate,open,openat,pwrite64,rename");
 RUN("");
 ...
 OUTPUT("--------------wiredtiger_open");
 // lock == 3
-lock = open("./WiredTiger.lock", O_RDWR|O_CREAT|O_CLOEXEC, 0666);
+lock = OPEN("./WiredTiger.lock", O_RDWR|O_CREAT|O_CLOEXEC, 0666);
 pwrite64(lock, "WiredTiger lock file\n", 0x15, 0x0);
-fd = open("./WiredTiger", O_RDWR|O_CREAT|O_CLOEXEC, 0666);
+fd = OPEN("./WiredTiger", O_RDWR|O_CREAT|O_CLOEXEC, 0666);
 pwrite64(fd, "WiredTiger\nWiredTiger"..., ...);
 #ifdef __linux__
 fdatasync(fd);
@@ -65,7 +71,7 @@ close(fd);
 
 ...  // On Linux, there are calls to open and read "/proc/meminfo" here.
 
-fd = open("./WiredTiger.basecfg.set", O_RDWR|O_CREAT|O_EXCL|O_CLOEXEC, 0666);
+fd = OPEN("./WiredTiger.basecfg.set", O_RDWR|O_CREAT|O_EXCL|O_CLOEXEC, 0666);
 pwrite64(fd, "# Do not modify this file."..., ...);
 #ifdef __linux__
 fdatasync(fd);
@@ -74,15 +80,15 @@ close(fd);
 rename("./WiredTiger.basecfg.set", "./WiredTiger.basecfg");
 
 #ifdef __linux__
-dir = open("./", O_RDONLY); 
+dir = OPEN("./", O_RDONLY|O_CLOEXEC);
 fdatasync(dir);
 close(dir);
 #endif
 
-fd = open("./WiredTiger.wt", O_RDWR|O_CREAT|O_EXCL|O_NOATIME|O_CLOEXEC, 0666);
+fd = OPEN("./WiredTiger.wt", O_RDWR|O_CREAT|O_EXCL|O_NOATIME|O_CLOEXEC, 0666);
 
 #ifdef __linux__
-dir = open("./", O_RDONLY); 
+dir = OPEN("./", O_RDONLY|O_CLOEXEC);
 fdatasync(dir);
 close(dir);
 #endif /* __linux__ */
@@ -96,7 +102,7 @@ close(fd);
 wt = OPEN_EXISTING("./WiredTiger.wt\0", O_RDWR|O_NOATIME|O_CLOEXEC);
 FTRUNCATE(wt, 0x1000);
 
-fd = open("./WiredTiger.turtle.set\0", O_RDWR|O_CREAT|O_EXCL|O_CLOEXEC, 0666);
+fd = OPEN("./WiredTiger.turtle.set\0", O_RDWR|O_CREAT|O_EXCL|O_CLOEXEC, 0666);
 pwrite64(fd, "WiredTiger version string\nWiredTiger"..., ...);
 #ifdef __linux__
 fdatasync(fd);
@@ -106,10 +112,10 @@ rename("./WiredTiger.turtle.set", "./WiredTiger.turtle");
 
 ...  // There is a second open of turtle here, is it important?
 
-fd = open("./WiredTigerLAS.wt", O_RDWR|O_CREAT|O_EXCL|O_NOATIME|O_CLOEXEC, 0666);
+fd = OPEN("./WiredTigerLAS.wt", O_RDWR|O_CREAT|O_EXCL|O_NOATIME|O_CLOEXEC, 0666);
 
 #ifdef __linux__
-dir = open("./", O_RDONLY); 
+dir = OPEN("./", O_RDONLY|O_CLOEXEC);
 fdatasync(dir);
 close(dir);
 #endif /* __linux__ */
@@ -134,7 +140,7 @@ fdatasync(wt);
 fd = OPEN_EXISTING("./WiredTiger.turtle", O_RDWR|O_CLOEXEC);
 
 close(fd);
-fd = open("./WiredTiger.turtle.set", O_RDWR|O_CREAT|O_EXCL|O_CLOEXEC, 0666);
+fd = OPEN("./WiredTiger.turtle.set", O_RDWR|O_CREAT|O_EXCL|O_CLOEXEC, 0666);
 pwrite64(fd, "WiredTiger version string\nWiredTiger"..., ...);
 #ifdef __linux__
 fdatasync(fd);
@@ -142,7 +148,7 @@ fdatasync(fd);
 close(fd);
 rename("./WiredTiger.turtle.set", "./WiredTiger.turtle");
 #ifdef __linux__
-dir = open("./", O_RDONLY); 
+dir = OPEN("./", O_RDONLY|O_CLOEXEC);
 fdatasync(dir);
 close(dir);
 fdatasync(wt);
@@ -151,9 +157,9 @@ fdatasync(wt);
 OUTPUT("--------------open_session");
 OUTPUT("--------------create");
 
-hello = open("./hello.wt", O_RDWR|O_CREAT|O_EXCL|O_NOATIME|O_CLOEXEC, 0666);
+hello = OPEN("./hello.wt", O_RDWR|O_CREAT|O_EXCL|O_NOATIME|O_CLOEXEC, 0666);
 #ifdef __linux__
-dir = open("./", O_RDONLY); 
+dir = OPEN("./", O_RDONLY|O_CLOEXEC);
 fdatasync(dir);
 close(dir);
 #endif /* __linux__ */