Fix for BUG#2477 "Slave stop with error after master reboot if use HEAP tables":

when we open the HEAP table for the first time since server restart,
in hp_open(), we set a flag to propagate this info to the handler level
which then writes a DELETE FROM this_heap_table to the binlog.
It is not a perfect solution for the bug, because between the server start and 
the first open of the table, the slave still had old data in his table so
a SELECT on the slave may show wrong content. But if there is a --init-file
to populate the HEAP table on master as startup, then this is a safe fix
(I'll put a note about init-file in the HEAP section of the manual).


heap/hp_info.c:
  new info variable implicit_emptied
heap/hp_open.c:
  If this is the first open of the HEAP table, it means it is empty,
  so we mark it.
include/heap.h:
  new variables implicit_emptied
  (we need one in HEAPINFO for the hp_info() call).
sql/ha_heap.cc:
  report info to upper level
sql/handler.h:
  new info 'implicit_emptied' in the handler level; only HEAP uses it.
sql/sql_base.cc:
  When a HEAP table is opened for the first time, write a DELETE FROM to the binlog,
  for replication and mysqlbinlog|mysql.
  Monty: I added the
  entry->file->implicit_emptied= 0;

1 files changed, 47 insertions, 0 deletions

diff --git a/mysql-test/t/rpl_heap.test b/mysql-test/t/rpl_heap.test
new file mode 100644
index 00000000000..15f61918034
--- /dev/null
+++ b/mysql-test/t/rpl_heap.test
@@ -0,0 +1,47 @@
+# You must run this test with --manager.
+
+require_manager;
+
+# Don't know why, but using TCP/IP connections makes this test fail
+# with "Lost connection to MySQL server during query" when we
+# issue a query after the server restart.
+# Maybe this is something awkward in mysqltest or in the manager?
+# So we use sockets.
+connect (master,localhost,root,,test,0,master.sock);
+connect (slave,localhost,root,,test,0,slave.sock);
+
+connection master;
+reset master;
+drop table if exists t1;
+create table t1 (a int) type=HEAP;
+insert into t1 values(10);
+save_master_pos;
+show binlog events from 79;
+connection slave;
+reset slave;
+start slave;
+sync_with_master;
+show create table t1;
+select * from t1; # should be one row
+
+server_stop master;
+server_start master;
+
+connection master;
+select * from t1;
+# to check that DELETE is not written twice
+# (the LIMIT is to not use the query cache)
+select * from t1 limit 10;
+save_master_pos;
+show binlog events in 'master-bin.002' from 79;
+
+connection slave;
+sync_with_master;
+select * from t1; # should be empty
+
+# clean up
+connection master;
+drop table t1;
+save_master_pos;
+connection slave;
+sync_with_master;