Last part of WL#1062: better replication of timezones: no more use

of SET ONE_SHOT; storing tz info directly in event (if this info is needed),
it's now allowed to have different global tz on master and slave.


client/mysqlbinlog.cc:
  we need MAX_TIME_ZONE_NAME_LENGTH when processing log_event.h, and it's declared in mysql_priv.h
mysql-test/r/rpl_timezone.result:
  result update
mysql-test/t/rpl_timezone-slave.opt:
  Now that we can have different global value of timezone on master and slave, let's test it.
mysql-test/t/rpl_timezone.test:
  Tests of the new replication of timezones: checking the output of mysqlbinlog,
  replication of CONVERT_TZ().
sql/ha_innodb.cc:
  No very fast shutdown on Netware (anyway it's disabled on all platforms,
  but this is so that we don't forget to keep it disabled on Netware in the future).
sql/log.cc:
  No more need to write SET ONE_SHOT to binlog for character set and timezone
  (as we store this info  directly nin the Query_log_event now).
sql/log_event.cc:
  Exclude ::write() methods if MYSQL_CLIENT.
  Storing timezone info in the Query_log_event in master. Re-reading it in slave.
  Small code cleanups. I plan to not store the end 0 of catalog in binlog
  events soon.
sql/log_event.h:
  replication of time zones: a place for tz info in Query_log_event,
  in LAST_EVENT_INFO. Plus if we are compiling a client, we don't need
  the ::write() methods, so keeping them out (of mysqlbinlog.cc;
  keeping them in, resulted in problem that mysqlbinlog does not know Timezone
  structure).
sql/mysql_priv.h:
  moving this define from tztime.h (tztime.h has things which are
  too much for a client like mysqlbinlog).
sql/set_var.cc:
  It's now allowed to change global value of charset or timezone even if using binlogging
  or if being a slave.
  Making CONVERT_TZ(,,@@session.time_zone) replicate.
sql/set_var.h:
  these ::check()s are not needed anymore (changing global charset
  or timezone is now allowed even if binlogging or slave)
sql/slave.cc:
  No more need to check for same global timezone if master is 5.x
  (ok, strictly speaking if it is > 5.0.3 but this is alpha).
sql/slave.h:
  a function to wrap settings of charset to default.
sql/tztime.cc:
  Adaptation of my_tz_find() to the case where it's not called from inside
  a query (i.e. cannot join its tz tables to the query's ones): this variant
  opens the tz tables itself, reads from them, and closes them. This is presently
  only used by the slave SQL thread (when it sets the tz before executing a query).
sql/tztime.h:
  declaration of new function, plus moving symbol to mysql_priv.h
  for easier usage in mysqlbinlog (Dmitri, pardon me).
BitKeeper/etc/logging_ok:
  Logging to logging@openlogging.org accepted

mysql-test/t/rpl_timezone.test

@@ -3,21 +3,25 @@ source include/master-slave.inc;
 
 # Some preparations
 let $VERSION=`select version()`;
+set timestamp=100000000; # for fixed output of mysqlbinlog
 create table t1 (t timestamp);
 create table t2 (t char(32));
 
+connection slave;
+select @@time_zone;
+
 #
 # Let us check how well replication works when we are saving datetime
 # value in TIMESTAMP field.
 #
 connection master;
 select @@time_zone;
+insert into t1 values ('20050101000000'), ('20050611093902');
 set time_zone='UTC';
 insert into t1 values ('20040101000000'), ('20040611093902');
 select * from t1;
-# On slave we still in 'Europe/Moscow' so we should see equivalent but
-# textually different values.
 sync_slave_with_master;
+set time_zone='UTC';
 select * from t1;
 
 # Let us check also that setting of time_zone back to default also works
@@ -28,12 +32,11 @@ set time_zone='Europe/Moscow';
 insert into t1 values ('20040101000000'), ('20040611093902');
 select * from t1;
 sync_slave_with_master;
+set time_zone='Europe/Moscow';
 select * from t1;
 connection master;
-# We should not see SET ONE_SHOT time_zone before second insert
---replace_result $VERSION VERSION
---replace_column 2 # 5 #
-show binlog events;
+--replace_result $MYSQL_TEST_DIR MYSQL_TEST_DIR
+--exec $MYSQL_BINLOG --short-form $MYSQL_TEST_DIR/var/log/master-bin.000001 
 
 #
 # Now let us check how well we replicate statments reading TIMESTAMP fields
@@ -54,10 +57,6 @@ delete from t2;
 set timestamp=1000072000;
 insert into t2 values (current_timestamp), (current_date), (current_time);
 sync_slave_with_master;
-# Values in ouput of these to queries should differ because we are in
-# in 'MET' on master and in 'Europe/Moscow on slave...
-set timestamp=1000072000;
-select current_timestamp, current_date, current_time;
 select * from t2;
 
 #
@@ -73,13 +72,24 @@ sync_slave_with_master;
 select * from t2;
 
 #
-# Let us check that we are not allowing to set global time_zone with
+# Let us check that we are allowing to set global time_zone with
 # replication
 #
 connection master;
---error 1387
 set global time_zone='MET';
 
+#
+# Let us see if CONVERT_TZ(@@time_zone) replicates
+#
+delete from t2;
+set time_zone='UTC';
+insert into t2 values(convert_tz('2004-01-01 00:00:00','MET',@@time_zone));
+insert into t2 values(convert_tz('2005-01-01 00:00:00','MET','Japan'));
+select * from t2;
+sync_slave_with_master;
+select * from t2;
+
 # Clean up
+connection master;
 drop table t1, t2;
 sync_slave_with_master;