replication updates and bugfixes. Still not perfect - there is some strange

memory corruption I can only repeat on one system.


mysql-test/r/multi_update.result:
  extended test trying to find why the old one was failing
mysql-test/r/rpl000015.result:
  result update
mysql-test/t/multi_update.test:
  updated test to do more intermediate result checks
mysql-test/t/rpl000014.test:
  cosmetic change
mysql-test/t/rpl000015-slave-master-info.opt:
  reduce connect retry trying to simulate a bug
mysql-test/t/rpl000015.test:
  cosmetic change
sql/lex.h:
  added RELAY_LOG_FILE and RELAY_LOG_POS to CHANGE MASTER TO
sql/log.cc:
  fixed replication bug
sql/log_event.cc:
  properly ignore replication errors according to error mask
sql/mysqld.cc:
  handle SIGFPE in the coredump handler
sql/slave.cc:
  do not kick slave if it has already exited and freed the structures
sql/sql_lex.cc:
  fixed initialization bug in CHANGE MASTER TO
sql/sql_lex.h:
  CHANGE MASTER TO .. RELAY_LOG_FILE=,RELAY_LOG_POS=
sql/sql_repl.cc:
  CHANGE MASTER TO .. RELAY_LOG_FILE=,RELAY_LOG_POS=
sql/sql_yacc.yy:
  CHANGE MASTER TO .. RELAY_LOG_FILE=,RELAY_LOG_POS=

1 files changed, 9 insertions, 1 deletions

diff --git a/sql/slave.cc b/sql/slave.cc
index 0ffc37eb975..1d701ce1f81 100644
--- a/sql/slave.cc
+++ b/sql/slave.cc
@@ -345,7 +345,13 @@ int terminate_slave_thread(THD* thd, pthread_mutex_t* term_lock,
     }
   }
   DBUG_ASSERT(thd != 0);
-  KICK_SLAVE(thd);
+  /* is is criticate to test if the slave is running. Otherwise, we might
+     be referening freed memory trying to kick it
+  */
+  if (*slave_running)
+  {
+    KICK_SLAVE(thd);
+  }
   while (*slave_running)
   {
     /* there is a small chance that slave thread might miss the first
@@ -367,7 +373,9 @@ int terminate_slave_thread(THD* thd, pthread_mutex_t* term_lock,
     DBUG_ASSERT_LOCK(cond_lock);
     pthread_cond_timedwait(term_cond, cond_lock, &abstime);
     if (*slave_running)
+    {
       KICK_SLAVE(thd);
+    }
   }
   if (term_lock)
     pthread_mutex_unlock(term_lock);