From 4fdbf9af5184d033f07010a1ab573fb329e7b8cd Mon Sep 17 00:00:00 2001
From: Tom Lane <tgl@sss.pgh.pa.us>
Date: Thu, 20 Jan 2022 17:28:07 -0500
Subject: [PATCH] Tighten TAP tests' tracking of postmaster state some more.

Commits 6c4a8903b et al. had a couple of deficiencies:

* The logic I added to Cluster::start to see if a PID file is present
could be fooled by a stale PID file left over from a previous
postmaster.  To fix, if we're not sure whether we expect to find a
running postmaster or not, validate the PID using "kill 0".

* 017_shm.pl has a loop in which it just issues repeated Cluster::start
calls; this will fail if some invocation fails but leaves self->_pid
set.  Per buildfarm results, the above fix is not enough to make this
safe: we might have "validated" a PID for a postmaster that exits
immediately after we look.  Hence, match each failed start call with
a stop call that will get us back to the self->_pid == undef state.
Add a fail_ok option to Cluster::stop to make this work.

Discussion: https://postgr.es/m/CA+hUKGKV6fOHvfiPt8=dOKzvswjAyLoFoJF1iQXMNpi7+hD1JQ@mail.gmail.com
---
 src/test/perl/PostgreSQL/Test/Cluster.pm | 40 ++++++++++++++++++++----
 src/test/recovery/t/017_shm.pl           |  3 ++
 2 files changed, 37 insertions(+), 6 deletions(-)

diff --git a/src/test/perl/PostgreSQL/Test/Cluster.pm b/src/test/perl/PostgreSQL/Test/Cluster.pm
index b7d4c24553..265f3ae657 100644
--- a/src/test/perl/PostgreSQL/Test/Cluster.pm
+++ b/src/test/perl/PostgreSQL/Test/Cluster.pm
@@ -896,23 +896,40 @@ Note: if the node is already known stopped, this does nothing.
 However, if we think it's running and it's not, it's important for
 this to fail.  Otherwise, tests might fail to detect server crashes.
 
+With optional extra param fail_ok => 1, returns 0 for failure
+instead of bailing out.
+
 =cut
 
 sub stop
 {
-	my ($self, $mode) = @_;
-	my $port   = $self->port;
+	my ($self, $mode, %params) = @_;
 	my $pgdata = $self->data_dir;
 	my $name   = $self->name;
+	my $ret;
 
 	local %ENV = $self->_get_env();
 
 	$mode = 'fast' unless defined $mode;
-	return unless defined $self->{_pid};
+	return 1 unless defined $self->{_pid};
+
 	print "### Stopping node \"$name\" using mode $mode\n";
-	PostgreSQL::Test::Utils::system_or_bail('pg_ctl', '-D', $pgdata, '-m', $mode, 'stop');
+	$ret = PostgreSQL::Test::Utils::system_log('pg_ctl', '-D', $pgdata,
+		'-m', $mode, 'stop');
+
+	if ($ret != 0)
+	{
+		print "# pg_ctl stop failed: $ret\n";
+
+		# Check to see if we still have a postmaster or not.
+		$self->_update_pid(-1);
+
+		BAIL_OUT("pg_ctl stop failed") unless $params{fail_ok};
+		return 0;
+	}
+
 	$self->_update_pid(0);
-	return;
+	return 1;
 }
 
 =pod
@@ -1142,9 +1159,20 @@ sub _update_pid
 	if (open my $pidfile, '<', $self->data_dir . "/postmaster.pid")
 	{
 		chomp($self->{_pid} = <$pidfile>);
-		print "# Postmaster PID for node \"$name\" is $self->{_pid}\n";
 		close $pidfile;
 
+		# If we aren't sure what to expect, validate the PID using kill().
+		# This protects against stale PID files left by crashed postmasters.
+		if ($is_running == -1 && kill(0, $self->{_pid}) == 0)
+		{
+			print
+			  "# Stale postmaster.pid file for node \"$name\": PID $self->{_pid} no longer exists\n";
+			$self->{_pid} = undef;
+			return;
+		}
+
+		print "# Postmaster PID for node \"$name\" is $self->{_pid}\n";
+
 		# If we found a pidfile when there shouldn't be one, complain.
 		BAIL_OUT("postmaster.pid unexpectedly present") if $is_running == 0;
 		return;
diff --git a/src/test/recovery/t/017_shm.pl b/src/test/recovery/t/017_shm.pl
index 288041d6d7..d57679b531 100644
--- a/src/test/recovery/t/017_shm.pl
+++ b/src/test/recovery/t/017_shm.pl
@@ -207,6 +207,9 @@ sub poll_start
 		# Wait 0.1 second before retrying.
 		usleep(100_000);
 
+		# Clean up in case the start attempt just timed out or some such.
+		$node->stop('fast', fail_ok => 1);
+
 		$attempts++;
 	}