From fc974626cd060271c39480ea5ac960b8db903699 Mon Sep 17 00:00:00 2001
From: David Steele <david@pgmasters.net>
Date: Sun, 17 Mar 2019 08:35:40 +0400
Subject: [PATCH] Add a note regarding verifying checkpoint against replay
 position.

---
 lib/pgBackRest/Db.pm | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/lib/pgBackRest/Db.pm b/lib/pgBackRest/Db.pm
index 273492280..f0e9da0f9 100644
--- a/lib/pgBackRest/Db.pm
+++ b/lib/pgBackRest/Db.pm
@@ -999,6 +999,10 @@ sub replayWait
     $self->executeSql('checkpoint', undef, false);
 
     # On PostgreSQL >= 9.6 the checkpoint location can be verified
+    #
+    # ??? We have seen one instance where this check failed.  Is there any chance that the replayed position could be ahead of the
+    # checkpoint recorded in pg_control?  It seems possible, so in the C version of this add a loop to keep checking pg_control
+    # until the checkpoint has been recorded.
     my $strCheckpointLSN = undef;
 
     if ($self->{strDbVersion} >= PG_VERSION_96)