Fix bug in clean shutdown of walsender that pg_receiving is connecting to.

On clean shutdown, walsender waits for all WAL to be replicated to a standby,
and exits. It determined whether that replication had been completed by
checking whether its sent location had been equal to a standby's flush
location. Unfortunately this condition never becomes true when the standby
such as pg_receivexlog which always returns an invalid flush location is
connecting to walsender, and then walsender waits forever.

This commit changes walsender so that it just checks a standby's write
location if a flush location is invalid.

Back-patch to 9.1 where enough infrastructure for this exists.

Author: MasaoFujii

src/backend/replication/walsender.c

@@ -808,9 +808,20 @@ WalSndLoop(void)
 			 */
 			if (walsender_ready_to_stop)
 			{
+				XLogRecPtr	replicatedPtr;
+
 				/* ... let's just be real sure we're caught up ... */
 				XLogSend(output_message, &caughtup);
-				if (caughtup && XLByteEQ(sentPtr, MyWalSnd->flush) &&
+
+				/*
+				 * Check a write location to see whether all the WAL have
+				 * successfully been replicated if this walsender is connecting
+				 * to a standby such as pg_receivexlog which always returns
+				 * an invalid flush location. Otherwise, check a flush location.
+				 */
+				replicatedPtr = XLogRecPtrIsInvalid(MyWalSnd->flush) ?
+					MyWalSnd->write : MyWalSnd->flush;
+				if (caughtup && XLByteEQ(sentPtr, replicatedPtr) &&
 					!pq_is_send_pending())
 				{
 					walsender_shutdown_requested = true;