From: Tom Lane <tgl@sss.pgh.pa.us>
To: Simon Riggs <simon@2ndquadrant.com>, Michael Paquier <michael@paquier.xyz>
cc: pgsql-hackers@lists.postgresql.org
Subject: Why did commit 6271fceb8 enable debug logging for all TAP tests?
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <32458.1525657786.1@sss.pgh.pa.us>
Date: Sun, 06 May 2018 21:49:46 -0400
Message-ID: <32459.1525657786@sss.pgh.pa.us>
Precedence: bulk

I've been trying to trace down a slowdown in HEAD's regression tests
compared to the v10 branch, and one thing I discovered is that there was
a noticeable across-the-board slowdown in the TAP tests as a result of
commit 6271fceb8; as much as 10% for the "scripts" tests, though less in
other places.  The changes it made in xlog.c cannot explain that; it has
to be down to this:

diff --git a/src/test/perl/PostgresNode.pm b/src/test/perl/PostgresNode.pm
index 93faadc20e..80f68df246 100644
--- a/src/test/perl/PostgresNode.pm
+++ b/src/test/perl/PostgresNode.pm
@@ -419,6 +419,7 @@ sub init
 	print $conf "restart_after_crash = off\n";
 	print $conf "log_line_prefix = '%m [%p] %q%a '\n";
 	print $conf "log_statement = all\n";
+	print $conf "log_min_messages = debug1\n";
 	print $conf "log_replication_commands = on\n";
 	print $conf "wal_retrieve_retry_interval = '500ms'\n";
 	print $conf "port = $port\n";

Now, that was neither mentioned in the commit message nor justified
by any added test cases or scaffolding, so I'm assuming it was simply a
mistake and should be reverted.  Please confirm.

			regards, tom lane