Fix bug#27072: RPM autostarts the server

This is the fix for 5.1, where only the behaviour on upgrade is changed:
If the server was stopped when the upgrade begins, we assume the
administrator is taking manual action, so we do not start the (new)
server at the end of the upgrade.
We still install the start/stop script, so it will be started on reboot.


support-files/mysql.spec.sh:
  In the "pre" section of the spec file, check the server status, and write
  the result to a file.
  In the "post" section, evaluate the status file, and start the server if
  it was running during status analysis.
  In 5.1, we start the server if there is no status file (which will happen
  on first installation, when there is no data directory yet).

1 files changed, 113 insertions, 7 deletions

diff --git a/support-files/mysql.spec.sh b/support-files/mysql.spec.sh
index 91848ffc8bd..21a4d0c52a0 100644
--- a/support-files/mysql.spec.sh
+++ b/support-files/mysql.spec.sh
@@ -1,4 +1,4 @@
-# Copyright (C) 2000-2008 MySQL AB, 2008-2010 Sun Microsystems, Inc.
+# Copyright (c) 2000, 2010, Oracle and/or its affiliates. All rights reserved.
 #
 # This program is free software; you can redistribute it and/or modify
 # it under the terms of the GNU General Public License as published by
@@ -610,6 +610,7 @@ touch $RBR%{_sysconfdir}/mysqlmanager.passwd
 ##############################################################################
 
 %pre server
+mysql_datadir=%{mysqldatadir}
 # Check if we can safely upgrade.  An upgrade is only safe if it's from one
 # of our RPMs in the same version family.
 
@@ -678,7 +679,74 @@ HERE
   fi
 fi
 
+# We assume that if there is exactly one ".pid" file,
+# it contains the valid PID of a running MySQL server.
+NR_PID_FILES=`ls $mysql_datadir/*.pid 2>/dev/null | wc -l`
+case $NR_PID_FILES in
+	0 ) SERVER_TO_START=''  ;;  # No "*.pid" file == no running server
+	1 ) SERVER_TO_START='true' ;;
+	* ) SERVER_TO_START=''      # Situation not clear
+	    SEVERAL_PID_FILES=true ;;
+esac
+# That logic may be debated: We might check whether it is non-empty,
+# contains exactly one number (possibly a PID), and whether "ps" finds it.
+# OTOH, if there is no such process, it means a crash without a cleanup -
+# is that a reason not to start a new server after upgrade?
+
+STATUS_FILE=$mysql_datadir/RPM_UPGRADE_MARKER
+
+if [ -f $STATUS_FILE ]; then
+	echo "Some previous upgrade was not finished:"
+	ls -ld $STATUS_FILE
+	echo "Please check its status, then do"
+	echo "    rm $STATUS_FILE"
+	echo "before repeating the MySQL upgrade."
+	exit 1
+elif [ -n "$SEVERAL_PID_FILES" ] ; then
+	echo "Your MySQL directory '$mysql_datadir' has more than one PID file:"
+	ls -ld $mysql_datadir/*.pid
+	echo "Please check which one (if any) corresponds to a running server"
+	echo "and delete all others before repeating the MySQL upgrade."
+	exit 1
+fi
+
+NEW_VERSION=%{mysql_version}-%{release}
+
+# The "pre" section code is also run on a first installation,
+# when there  is no data directory yet. Protect against error messages.
+if [ -d $mysql_datadir ] ; then
+	echo "MySQL RPM upgrade to version $NEW_VERSION"  > $STATUS_FILE
+	echo "'pre' step running at `date`"          >> $STATUS_FILE
+	echo                                         >> $STATUS_FILE
+	echo "ERR file(s):"                          >> $STATUS_FILE
+	ls -ltr $mysql_datadir/*.err                 >> $STATUS_FILE
+	echo                                         >> $STATUS_FILE
+	echo "Latest 'Version' line in latest file:" >> $STATUS_FILE
+	grep '^Version' `ls -tr $mysql_datadir/*.err | tail -1` | \
+		tail -1                              >> $STATUS_FILE
+	echo                                         >> $STATUS_FILE
+
+	if [ -n "$SERVER_TO_START" ] ; then
+		# There is only one PID file, race possibility ignored
+		echo "PID file:"                           >> $STATUS_FILE
+		ls -l   $mysql_datadir/*.pid               >> $STATUS_FILE
+		cat     $mysql_datadir/*.pid               >> $STATUS_FILE
+		echo                                       >> $STATUS_FILE
+		echo "Server process:"                     >> $STATUS_FILE
+		ps -fp `cat $mysql_datadir/*.pid`          >> $STATUS_FILE
+		echo                                       >> $STATUS_FILE
+		echo "SERVER_TO_START=$SERVER_TO_START"    >> $STATUS_FILE
+	else
+		# Take a note we checked it ...
+		echo "PID file:"                           >> $STATUS_FILE
+		ls -l   $mysql_datadir/*.pid               >> $STATUS_FILE 2>&1
+	fi
+fi
+
 # Shut down a previously installed server first
+# Note we *could* make that depend on $SERVER_TO_START, but we rather don't,
+# so a "stop" is attempted even if there is no PID file.
+# (Maybe the "stop" doesn't work then, but we might fix that in itself.)
 if [ -x %{_sysconfdir}/init.d/mysql ] ; then
 	%{_sysconfdir}/init.d/mysql stop > /dev/null 2>&1
 	echo "Giving mysqld 5 seconds to exit nicely"
@@ -687,17 +755,33 @@ fi
 
 %post server
 mysql_datadir=%{mysqldatadir}
+NEW_VERSION=%{mysql_version}-%{release}
+STATUS_FILE=$mysql_datadir/RPM_UPGRADE_MARKER
 
 # ----------------------------------------------------------------------
-# Create data directory if needed
+# Create data directory if needed, check whether upgrade or install
 # ----------------------------------------------------------------------
 if [ ! -d $mysql_datadir ] ; then mkdir -m 755 $mysql_datadir; fi
-if [ ! -d $mysql_datadir/mysql ] ; then mkdir $mysql_datadir/mysql; fi
+if [ -f $STATUS_FILE ] ; then
+	SERVER_TO_START=`grep '^SERVER_TO_START=' $STATUS_FILE | cut -c17-`
+else
+	SERVER_TO_START='true'   # This is for 5.1 only, to not change behavior
+fi
+# echo "Analyzed: SERVER_TO_START=$SERVER_TO_START"
+if [ ! -d $mysql_datadir/mysql ] ; then
+	mkdir $mysql_datadir/mysql;
+	echo "MySQL RPM installation of version $NEW_VERSION" >> $STATUS_FILE
+else
+	# If the directory exists, we may assume it is an upgrade.
+	echo "MySQL RPM upgrade to version $NEW_VERSION" >> $STATUS_FILE
+fi
 if [ ! -d $mysql_datadir/test ] ; then mkdir $mysql_datadir/test; fi
 
 # ----------------------------------------------------------------------
 # Make MySQL start/shutdown automatically when the machine does it.
 # ----------------------------------------------------------------------
+# NOTE: This still needs to be debated. Should we check whether these links
+# for the other run levels exist(ed) before the upgrade?
 # use insserv for older SuSE Linux versions
 if [ -x /sbin/insserv ] ; then
 	/sbin/insserv %{_sysconfdir}/init.d/mysql
@@ -741,11 +825,14 @@ chown -R %{mysqld_user}:%{mysqld_group} $mysql_datadir
 # ----------------------------------------------------------------------
 chmod -R og-rw $mysql_datadir/mysql
 
-# Restart in the same way that mysqld will be started normally.
-%{_sysconfdir}/init.d/mysql start
+# Was the server running before the upgrade? If so, restart the new one.
+if [ "$SERVER_TO_START" = "true" ] ; then
+	# Restart in the same way that mysqld will be started normally.
+	%{_sysconfdir}/init.d/mysql start
 
-# Allow mysqld_safe to start mysqld and print a message before we exit
-sleep 2
+	# Allow mysqld_safe to start mysqld and print a message before we exit
+	sleep 2
+fi
 
 #echo "Thank you for installing the MySQL Community Server! For Production
 #systems, we recommend MySQL Enterprise, which contains enterprise-ready
@@ -753,6 +840,14 @@ sleep 2
 #scheduled service packs and more.  Visit www.mysql.com/enterprise for more
 #information."
 
+# Collect an upgrade history ...
+echo "Upgrade/install finished at `date`"        >> $STATUS_FILE
+echo                                             >> $STATUS_FILE
+echo "====="                                     >> $STATUS_FILE
+STATUS_HISTORY=$mysql_datadir/RPM_UPGRADE_HISTORY
+cat $STATUS_FILE >> $STATUS_HISTORY
+rm  $STATUS_FILE
+  
 %if %{CLUSTER_BUILD}
 %post ndb-storage
 mysql_clusterdir=/var/lib/mysql-cluster
@@ -1038,6 +1133,17 @@ fi
 # merging BK trees)
 ##############################################################################
 %changelog
+
+* Tue Jun 15 2010 Joerg Bruehe <joerg.bruehe@sun.com>
+
+- Change the behaviour on upgrade:
+  *Iff* the server was stopped before the upgrade is started, this is taken as a
+  sign the administrator is handling that manually, and so the new server will
+  not be started automatically at the end of the upgrade.
+  The start/stop scripts will still be installed, so the server will be started
+  on the next machine boot.
+  This is the 5.1 version of fixing bug#27072 (RPM autostarting the server).
+
 * Mon Mar 01 2010 Joerg Bruehe <joerg.bruehe@sun.com>
 
 - Set "Oracle and/or its affiliates" as the vendor and copyright owner,