fix for BUG#1331: "Unexistent user variable is not replicated".

When an update query is to be written to the binlog, and it reads unset
user variables (example: INSERT INTO t VALUEs(@a) where @a does not exist),
we create the variable like if it had been set with SET before, and we loop
(i.e. we redo the steps of logging, now that the variable exists).


mysql-test/r/rpl_user_variables.result:
  result update.
mysql-test/t/rpl_user_variables.test:
  test to see if unset user variables are written to the binlog
sql/item_func.cc:
  When an update query is to be written to the binlog, and it reads unset
  user variables (example: INSERT INTO t VALUEs(@a) where @a does not exist),
  we create the variable like if it had been set with SET before, and we loop
  (i.e. we redo the steps of logging, now that the variable exists).
This commit is contained in:
unknown 2003-10-02 10:31:37 +02:00
commit 031b6b5fbb
3 changed files with 144 additions and 48 deletions

View file

@ -29,13 +29,16 @@ insert into t1 values (@i1), (@i2), (@i3), (@i4);
insert into t1 values (@r1), (@r2);
insert into t1 values (@s1), (@s2), (@s3), (@s4), (@s5);
insert into t1 values (@n1);
insert into t1 values (@n2);
insert into t1 values (@n2); # not explicitely set before
insert into t1 values (@a:=0), (@a:=@a+1), (@a:=@a+1);
insert into t1 values (@a+(@b:=@a+1));
set @q:='abc';
insert t1 values (@q), (@q:=concat(@q, 'n1')), (@q:=concat(@q, 'n2'));
set @a:=5;
insert into t1 values (@a),(@a);
connection master1; # see if variable is reset in binlog when thread changes
insert into t1 values (@a),(@a),(@a*5);
select * from t1;
save_master_pos;
connection slave;
sync_with_master;