Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
# Test of replication of stored procedures (WL#2146 for MySQL 5.0)
|
2005-11-10 17:50:51 +01:00
|
|
|
# Modified by WL#2971.
|
|
|
|
|
|
|
|
# Note that in the .opt files we still use the old variable name
|
|
|
|
# log-bin-trust-routine-creators so that this test checks that it's
|
|
|
|
# still accepted (this test also checks that the new name is
|
|
|
|
# accepted). The old name could be removed in 5.1 or 6.0.
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
|
|
|
|
source include/master-slave.inc;
|
|
|
|
|
|
|
|
# we need a db != test, where we don't have automatic grants
|
2005-11-10 17:50:51 +01:00
|
|
|
--disable_warnings
|
|
|
|
drop database if exists mysqltest1;
|
|
|
|
--enable_warnings
|
|
|
|
create database mysqltest1;
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
use mysqltest1;
|
|
|
|
create table t1 (a varchar(100));
|
|
|
|
sync_slave_with_master;
|
|
|
|
use mysqltest1;
|
|
|
|
|
|
|
|
# ********************** PART 1 : STORED PROCEDURES ***************
|
|
|
|
|
|
|
|
# Does the same proc as on master get inserted into mysql.proc ?
|
|
|
|
# (same definer, same properties...)
|
|
|
|
|
|
|
|
connection master;
|
|
|
|
|
|
|
|
delimiter |;
|
|
|
|
|
2005-11-10 17:50:51 +01:00
|
|
|
# Stored procedures don't have the limitations that functions have
|
|
|
|
# regarding binlogging: it's ok to create a procedure as not
|
|
|
|
# deterministic and updating data, while it's not ok to create such a
|
|
|
|
# function. We test this.
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
|
2005-11-10 17:50:51 +01:00
|
|
|
create procedure foo()
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
begin
|
|
|
|
declare b int;
|
|
|
|
set b = 8;
|
|
|
|
insert into t1 values (b);
|
|
|
|
insert into t1 values (unix_timestamp());
|
|
|
|
end|
|
|
|
|
delimiter ;|
|
|
|
|
|
|
|
|
# we replace columns having times
|
|
|
|
# (even with fixed timestamp displayed time may changed based on TZ)
|
|
|
|
--replace_result localhost.localdomain localhost 127.0.0.1 localhost
|
|
|
|
--replace_column 13 # 14 #
|
|
|
|
select * from mysql.proc where name='foo' and db='mysqltest1';
|
|
|
|
sync_slave_with_master;
|
2005-11-10 17:50:51 +01:00
|
|
|
# You will notice in the result that the definer does not match what
|
|
|
|
# it is on master, it is a known bug on which Alik is working
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
--replace_result localhost.localdomain localhost 127.0.0.1 localhost
|
|
|
|
--replace_column 13 # 14 #
|
|
|
|
select * from mysql.proc where name='foo' and db='mysqltest1';
|
|
|
|
|
|
|
|
connection master;
|
|
|
|
# see if timestamp used in SP on slave is same as on master
|
|
|
|
set timestamp=1000000000;
|
|
|
|
call foo();
|
|
|
|
select * from t1;
|
|
|
|
sync_slave_with_master;
|
|
|
|
select * from t1;
|
|
|
|
|
2005-11-10 17:50:51 +01:00
|
|
|
# Now a SP which is not updating tables
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
|
|
|
|
connection master;
|
|
|
|
delete from t1;
|
|
|
|
create procedure foo2()
|
|
|
|
select * from mysqltest1.t1;
|
|
|
|
call foo2();
|
|
|
|
|
2005-11-10 17:50:51 +01:00
|
|
|
# check that this is allowed (it's not for functions):
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
alter procedure foo2 contains sql;
|
|
|
|
|
|
|
|
# SP with definer's right
|
|
|
|
|
|
|
|
drop table t1;
|
|
|
|
create table t1 (a int);
|
|
|
|
create table t2 like t1;
|
|
|
|
|
|
|
|
create procedure foo3()
|
|
|
|
deterministic
|
|
|
|
insert into t1 values (15);
|
|
|
|
|
|
|
|
# let's create a non-privileged user
|
|
|
|
grant CREATE ROUTINE, EXECUTE on mysqltest1.* to "zedjzlcsjhd"@127.0.0.1;
|
|
|
|
grant SELECT on mysqltest1.t1 to "zedjzlcsjhd"@127.0.0.1;
|
|
|
|
grant SELECT, INSERT on mysqltest1.t2 to "zedjzlcsjhd"@127.0.0.1;
|
|
|
|
|
2006-01-04 15:25:32 +01:00
|
|
|
# ToDo: BUG#14931: There is a race between the last grant binlogging, and
|
|
|
|
# the binlogging in the new connection made below, causing sporadic test
|
|
|
|
# failures due to switched statement order in binlog. To fix this we do
|
|
|
|
# SELECT 1 in the first connection before starting the second, ensuring
|
|
|
|
# that binlogging is done in the expected order.
|
|
|
|
# Please remove this SELECT 1 when BUG#14931 is fixed.
|
|
|
|
SELECT 1;
|
|
|
|
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
connect (con1,127.0.0.1,zedjzlcsjhd,,mysqltest1,$MASTER_MYPORT,);
|
|
|
|
connection con1;
|
|
|
|
|
2005-11-10 17:50:51 +01:00
|
|
|
# this routine will fail in the second INSERT because of privileges
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
delimiter |;
|
|
|
|
create procedure foo4()
|
|
|
|
deterministic
|
|
|
|
begin
|
|
|
|
insert into t2 values(3);
|
|
|
|
insert into t1 values (5);
|
|
|
|
end|
|
|
|
|
|
|
|
|
delimiter ;|
|
|
|
|
|
2005-05-05 18:28:54 +02:00
|
|
|
# I add ,0 so that it does not print the error in the test output,
|
|
|
|
# because this error is hostname-dependent
|
2005-09-01 17:01:23 +02:00
|
|
|
--error 1142,0
|
2005-11-10 17:50:51 +01:00
|
|
|
call foo4(); # invoker has no INSERT grant on table t1 => failure
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
|
|
|
|
connection master;
|
|
|
|
call foo3(); # success (definer == root)
|
|
|
|
show warnings;
|
|
|
|
|
2005-09-01 17:01:23 +02:00
|
|
|
--error 1142,0
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
call foo4(); # definer's rights => failure
|
|
|
|
|
|
|
|
# we test replication of ALTER PROCEDURE
|
|
|
|
alter procedure foo4 sql security invoker;
|
|
|
|
call foo4(); # invoker's rights => success
|
|
|
|
show warnings;
|
|
|
|
|
2005-11-10 17:50:51 +01:00
|
|
|
# Note that half-failed procedure calls are ok with binlogging;
|
|
|
|
# if we compare t2 on master and slave we see they are identical:
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
|
|
|
|
select * from t1;
|
|
|
|
select * from t2;
|
|
|
|
sync_slave_with_master;
|
|
|
|
select * from t1;
|
|
|
|
select * from t2;
|
|
|
|
|
2005-11-10 17:50:51 +01:00
|
|
|
# Let's check another failing-in-the-middle procedure
|
|
|
|
connection master;
|
|
|
|
delete from t2;
|
|
|
|
alter table t2 add unique (a);
|
|
|
|
|
|
|
|
drop procedure foo4;
|
|
|
|
delimiter |;
|
|
|
|
create procedure foo4()
|
|
|
|
deterministic
|
|
|
|
begin
|
|
|
|
insert into t2 values(20),(20);
|
|
|
|
end|
|
|
|
|
|
|
|
|
delimiter ;|
|
|
|
|
|
|
|
|
--error 1062
|
|
|
|
call foo4();
|
|
|
|
show warnings;
|
|
|
|
|
|
|
|
select * from t2;
|
|
|
|
sync_slave_with_master;
|
|
|
|
# check that this failed-in-the-middle replicated right:
|
|
|
|
select * from t2;
|
|
|
|
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
# Test of DROP PROCEDURE
|
|
|
|
|
|
|
|
--replace_result localhost.localdomain localhost 127.0.0.1 localhost
|
|
|
|
--replace_column 13 # 14 #
|
|
|
|
select * from mysql.proc where name="foo4" and db='mysqltest1';
|
|
|
|
connection master;
|
|
|
|
drop procedure foo4;
|
|
|
|
select * from mysql.proc where name="foo4" and db='mysqltest1';
|
|
|
|
sync_slave_with_master;
|
|
|
|
select * from mysql.proc where name="foo4" and db='mysqltest1';
|
|
|
|
|
|
|
|
# ********************** PART 2 : FUNCTIONS ***************
|
|
|
|
|
|
|
|
connection master;
|
|
|
|
drop procedure foo;
|
|
|
|
drop procedure foo2;
|
|
|
|
drop procedure foo3;
|
|
|
|
|
|
|
|
delimiter |;
|
2005-11-10 17:50:51 +01:00
|
|
|
# check that needs "deterministic"
|
|
|
|
--error 1418
|
|
|
|
create function fn1(x int)
|
|
|
|
returns int
|
|
|
|
begin
|
|
|
|
insert into t1 values (x);
|
|
|
|
return x+2;
|
|
|
|
end|
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
create function fn1(x int)
|
|
|
|
returns int
|
|
|
|
deterministic
|
|
|
|
begin
|
|
|
|
insert into t1 values (x);
|
|
|
|
return x+2;
|
|
|
|
end|
|
|
|
|
|
|
|
|
delimiter ;|
|
|
|
|
delete t1,t2 from t1,t2;
|
|
|
|
select fn1(20);
|
|
|
|
insert into t2 values(fn1(21));
|
|
|
|
select * from t1;
|
|
|
|
select * from t2;
|
|
|
|
sync_slave_with_master;
|
|
|
|
select * from t1;
|
|
|
|
select * from t2;
|
|
|
|
|
|
|
|
connection master;
|
|
|
|
delimiter |;
|
|
|
|
|
|
|
|
drop function fn1;
|
|
|
|
|
|
|
|
create function fn1()
|
|
|
|
returns int
|
2005-11-10 17:50:51 +01:00
|
|
|
no sql
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
begin
|
|
|
|
return unix_timestamp();
|
|
|
|
end|
|
2005-11-10 17:50:51 +01:00
|
|
|
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
delimiter ;|
|
2005-11-10 17:50:51 +01:00
|
|
|
# check that needs "deterministic"
|
|
|
|
--error 1418
|
|
|
|
alter function fn1 contains sql;
|
|
|
|
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
delete from t1;
|
|
|
|
set timestamp=1000000000;
|
|
|
|
insert into t1 values(fn1());
|
|
|
|
|
2005-11-10 17:50:51 +01:00
|
|
|
connection con1;
|
|
|
|
|
|
|
|
delimiter |;
|
|
|
|
--error 1419 # only full-global-privs user can create a function
|
|
|
|
create function fn2()
|
|
|
|
returns int
|
|
|
|
no sql
|
|
|
|
begin
|
|
|
|
return unix_timestamp();
|
|
|
|
end|
|
|
|
|
delimiter ;|
|
|
|
|
connection master;
|
|
|
|
# test old variable name:
|
|
|
|
set global log_bin_trust_routine_creators=1;
|
|
|
|
# now use new name:
|
|
|
|
set global log_bin_trust_function_creators=0;
|
|
|
|
set global log_bin_trust_function_creators=1;
|
|
|
|
# slave needs it too otherwise will not execute what master allowed:
|
|
|
|
connection slave;
|
|
|
|
set global log_bin_trust_function_creators=1;
|
|
|
|
|
|
|
|
connection con1;
|
|
|
|
|
|
|
|
delimiter |;
|
|
|
|
create function fn2()
|
|
|
|
returns int
|
|
|
|
no sql
|
|
|
|
begin
|
|
|
|
return unix_timestamp();
|
|
|
|
end|
|
|
|
|
delimiter ;|
|
|
|
|
|
|
|
|
connection master;
|
|
|
|
|
|
|
|
# Now a function which is supposed to not update tables
|
|
|
|
# as it's "reads sql data", so should not give error even if
|
|
|
|
# non-deterministic.
|
|
|
|
|
|
|
|
delimiter |;
|
|
|
|
create function fn3()
|
|
|
|
returns int
|
|
|
|
not deterministic
|
|
|
|
reads sql data
|
|
|
|
begin
|
|
|
|
return 0;
|
|
|
|
end|
|
|
|
|
delimiter ;|
|
|
|
|
|
|
|
|
select fn3();
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
--replace_result localhost.localdomain localhost 127.0.0.1 localhost
|
|
|
|
--replace_column 13 # 14 #
|
|
|
|
select * from mysql.proc where db='mysqltest1';
|
|
|
|
select * from t1;
|
|
|
|
|
|
|
|
sync_slave_with_master;
|
|
|
|
use mysqltest1;
|
|
|
|
select * from t1;
|
|
|
|
--replace_result localhost.localdomain localhost 127.0.0.1 localhost
|
|
|
|
--replace_column 13 # 14 #
|
|
|
|
select * from mysql.proc where db='mysqltest1';
|
|
|
|
|
2005-11-10 17:50:51 +01:00
|
|
|
# Let's check a failing-in-the-middle function
|
|
|
|
connection master;
|
|
|
|
delete from t2;
|
|
|
|
alter table t2 add unique (a);
|
|
|
|
|
|
|
|
drop function fn1;
|
|
|
|
|
|
|
|
delimiter |;
|
Fix for BUG#14769 "Function fails to replicate if fails half-way (slave stops)":
if the function, invoked in a non-binlogged caller (e.g. SELECT, DO), failed half-way on the master,
slave would stop and complain that error code between him and master mismatch.
To solve this, when a stored function is invoked in a non-binlogged caller (e.g. SELECT, DO), we binlog the function
call as SELECT instead of as DO (see revision comment of sp_head.cc for more).
And: minor wording change in the help text.
This cset will cause conflicts in 5.1, I'll merge.
mysql-test/r/rpl_sp.result:
result update
mysql-test/t/rpl_sp-slave.opt:
bug just fixed so option not needed
mysql-test/t/rpl_sp.test:
test for more half-failed functions with DO and SELECT, to test the bug of this changeset.
cleanup at the end.
sql/mysqld.cc:
function -> stored function (change suggested by Paul)
sql/sp_head.cc:
When a function updates data and is called from a non-binlogged statement (SELECT, DO), we binlog it
as SELECT myfunc(), and not DO myfunc() like before.
2006-02-18 17:26:30 +01:00
|
|
|
create function fn1(x int)
|
2005-11-10 17:50:51 +01:00
|
|
|
returns int
|
|
|
|
begin
|
Fix for BUG#14769 "Function fails to replicate if fails half-way (slave stops)":
if the function, invoked in a non-binlogged caller (e.g. SELECT, DO), failed half-way on the master,
slave would stop and complain that error code between him and master mismatch.
To solve this, when a stored function is invoked in a non-binlogged caller (e.g. SELECT, DO), we binlog the function
call as SELECT instead of as DO (see revision comment of sp_head.cc for more).
And: minor wording change in the help text.
This cset will cause conflicts in 5.1, I'll merge.
mysql-test/r/rpl_sp.result:
result update
mysql-test/t/rpl_sp-slave.opt:
bug just fixed so option not needed
mysql-test/t/rpl_sp.test:
test for more half-failed functions with DO and SELECT, to test the bug of this changeset.
cleanup at the end.
sql/mysqld.cc:
function -> stored function (change suggested by Paul)
sql/sp_head.cc:
When a function updates data and is called from a non-binlogged statement (SELECT, DO), we binlog it
as SELECT myfunc(), and not DO myfunc() like before.
2006-02-18 17:26:30 +01:00
|
|
|
insert into t2 values(x),(x);
|
2005-11-10 17:50:51 +01:00
|
|
|
return 10;
|
|
|
|
end|
|
|
|
|
|
|
|
|
delimiter ;|
|
|
|
|
|
Fix for BUG#14769 "Function fails to replicate if fails half-way (slave stops)":
if the function, invoked in a non-binlogged caller (e.g. SELECT, DO), failed half-way on the master,
slave would stop and complain that error code between him and master mismatch.
To solve this, when a stored function is invoked in a non-binlogged caller (e.g. SELECT, DO), we binlog the function
call as SELECT instead of as DO (see revision comment of sp_head.cc for more).
And: minor wording change in the help text.
This cset will cause conflicts in 5.1, I'll merge.
mysql-test/r/rpl_sp.result:
result update
mysql-test/t/rpl_sp-slave.opt:
bug just fixed so option not needed
mysql-test/t/rpl_sp.test:
test for more half-failed functions with DO and SELECT, to test the bug of this changeset.
cleanup at the end.
sql/mysqld.cc:
function -> stored function (change suggested by Paul)
sql/sp_head.cc:
When a function updates data and is called from a non-binlogged statement (SELECT, DO), we binlog it
as SELECT myfunc(), and not DO myfunc() like before.
2006-02-18 17:26:30 +01:00
|
|
|
do fn1(100);
|
2005-11-10 17:50:51 +01:00
|
|
|
|
|
|
|
--error 1062
|
Fix for BUG#14769 "Function fails to replicate if fails half-way (slave stops)":
if the function, invoked in a non-binlogged caller (e.g. SELECT, DO), failed half-way on the master,
slave would stop and complain that error code between him and master mismatch.
To solve this, when a stored function is invoked in a non-binlogged caller (e.g. SELECT, DO), we binlog the function
call as SELECT instead of as DO (see revision comment of sp_head.cc for more).
And: minor wording change in the help text.
This cset will cause conflicts in 5.1, I'll merge.
mysql-test/r/rpl_sp.result:
result update
mysql-test/t/rpl_sp-slave.opt:
bug just fixed so option not needed
mysql-test/t/rpl_sp.test:
test for more half-failed functions with DO and SELECT, to test the bug of this changeset.
cleanup at the end.
sql/mysqld.cc:
function -> stored function (change suggested by Paul)
sql/sp_head.cc:
When a function updates data and is called from a non-binlogged statement (SELECT, DO), we binlog it
as SELECT myfunc(), and not DO myfunc() like before.
2006-02-18 17:26:30 +01:00
|
|
|
select fn1(20);
|
2005-11-10 17:50:51 +01:00
|
|
|
|
|
|
|
select * from t2;
|
|
|
|
sync_slave_with_master;
|
|
|
|
|
|
|
|
# check that this failed-in-the-middle replicated right:
|
|
|
|
select * from t2;
|
|
|
|
|
|
|
|
# ********************** PART 3 : TRIGGERS ***************
|
2005-05-06 18:52:19 +02:00
|
|
|
|
|
|
|
connection con1;
|
2005-09-01 17:01:23 +02:00
|
|
|
--error 1227
|
2005-05-06 18:52:19 +02:00
|
|
|
create trigger trg before insert on t1 for each row set new.a= 10;
|
|
|
|
|
|
|
|
connection master;
|
|
|
|
delete from t1;
|
|
|
|
# TODO: when triggers can contain an update, test that this update
|
|
|
|
# does not go into binlog.
|
|
|
|
# I'm not setting user vars in the trigger, because replication of user vars
|
|
|
|
# would take care of propagating the user var's value to slave, so even if
|
|
|
|
# the trigger was not executed on slave it would not be discovered.
|
|
|
|
create trigger trg before insert on t1 for each row set new.a= 10;
|
|
|
|
insert into t1 values (1);
|
|
|
|
select * from t1;
|
|
|
|
sync_slave_with_master;
|
|
|
|
select * from t1;
|
|
|
|
|
|
|
|
connection master;
|
|
|
|
delete from t1;
|
2005-07-19 18:06:49 +02:00
|
|
|
drop trigger trg;
|
2005-05-06 18:52:19 +02:00
|
|
|
insert into t1 values (1);
|
|
|
|
select * from t1;
|
|
|
|
sync_slave_with_master;
|
|
|
|
select * from t1;
|
|
|
|
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
|
Fix for BUG#19725 "Calls to SF in other database are not replicated
correctly in some cases".
In short, calls to a stored function located in another database
than the default database, may fail to replicate if the call was made
by SET, SELECT, or DO.
Longer: when a stored function is called from a statement which does not go
to binlog ("SET @a=somedb.myfunc()", "SELECT somedb.myfunc()",
"DO somedb.myfunc()"), this crafted statement is binlogged:
"SELECT myfunc();" (accompanied with a mention of the default database
if there is one). So, if "somedb" is not the default database,
the slave would fail to find myfunc(). The fix is to specify the
function's database name in the crafted binlogged statement, like this:
"SELECT somedb.myfunc();". Test added in rpl_sp.test.
mysql-test/r/rpl_sp.result:
Because I moved the SHOW BINLOG EVENTS down a bit, big portions of its
output move. Also, the function's database name appears in
SELECT statements.
mysql-test/t/rpl_sp.test:
Adding test for BUG#19725.
Moving the SHOW BINLOG EVENTS down, it is run at the very end to
test everything.
sql/sp_head.cc:
When binlogging a "SELECT myfunc()" (when a stored function is executed
inside a statement which does not go to the binlog (like a SET,
SELECT, DO), we need to write "SELECT db_of_myfunc().myfunc()",
because the function may be in a database which is not the default
database.
2007-01-08 22:01:06 +01:00
|
|
|
# ********************** PART 4 : RELATED FIXED BUGS ***************
|
|
|
|
|
|
|
|
|
2005-10-12 21:42:51 +02:00
|
|
|
#
|
|
|
|
# Test for bug #13969 "Routines which are replicated from master can't be
|
|
|
|
# executed on slave".
|
|
|
|
#
|
|
|
|
connection master;
|
|
|
|
create procedure foo()
|
|
|
|
not deterministic
|
|
|
|
reads sql data
|
|
|
|
select * from t1;
|
|
|
|
sync_slave_with_master;
|
|
|
|
# This should not fail
|
|
|
|
call foo();
|
|
|
|
connection master;
|
|
|
|
drop procedure foo;
|
|
|
|
sync_slave_with_master;
|
|
|
|
|
|
|
|
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
# Clean up
|
|
|
|
connection master;
|
|
|
|
drop function fn1;
|
|
|
|
drop database mysqltest1;
|
|
|
|
drop user "zedjzlcsjhd"@127.0.0.1;
|
2005-11-17 01:15:10 +01:00
|
|
|
use test;
|
Approximative fixes for BUG#2610,2611,9100 i.e. WL#2146 binlogging/replication of routines (stored procs and functions).
Approximative, because it's using our binlogging way (what we call "query"-level) and this is not as good as record-level binlog (5.1) would be. It imposes several
limitations to routines, and has caveats (which I'll document, and for which the server will try to issue errors but that is not always possible).
Reason I don't propagate caller info to the binlog as planned is that on master and slave
users may be different; even with that some caveats would remain.
mysql-test/mysql-test-run.sh:
In the testsuite we know what we do, we are not creating nasty routines, and breaking binlog is ok except in rpl_sp.
mysql-test/r/blackhole.result:
Updating results now that 4.1 has been merged
mysql-test/valgrind.supp:
Some suppressions for Valgrind (useful on my machine Suse 9.1);
this is just adding to the already existing suppressions of pthread and dl.
sql/item_func.cc:
Don't binlog the substatements when executing a function. If the function
is declared to modify data and does not complete, warning "broken binlog".
Note that SELECT myfunc() will not be binlogged even if myfunc() updates data (will be documented);
but INSERT INTO t VALUES(myfunc()) will be binlogged (what decides is if the caller
gets binlogged; the function changes nothing to binlogging).
sql/log_event.cc:
Just making functions which can be re-used when we binlog more strings
in status_vars in Query_log_event (e.g. one day "user", "host").
sql/log_event.h:
comment
sql/mysql_priv.h:
--log-bin-trust-routine-creators
sql/mysqld.cc:
--log-bin-trust-routine-creators
sql/set_var.cc:
--log-bin-trust-routine-creators
sql/share/errmsg.txt:
error messages to warn about problems with routines and binlog
sql/slave.cc:
If in a routine, replication table inclusion/exclusion rules always answer "replicate!" (see comment in code).
sql/sp.cc:
If binlog is on: errors if one wants to create a non-deterministic update routine
(repeatability problem - note that the test is not perfect for functions) or does not have SUPER (because routines can easily
be made to destroy slave's data with just CREATE ROUTINE and EXECUTE priv on master).
--log-bin-trust-routine-creators removes these errors.
Binlogging of CREATE PROCEDURE|FUNCTION.
sql/sql_acl.cc:
No thd==0 in tables_ok().
sql/sql_parse.cc:
Binlogging of CALL (and not of the substatements of the SP).
If SP returns error, we don't binlog it (see comment); we push warning in this case.
Binlogging of ALTER|DROP PROCEDURE|FUNCTION with safety messages.
2005-05-05 14:20:53 +02:00
|
|
|
sync_slave_with_master;
|
2005-11-17 01:15:10 +01:00
|
|
|
use test;
|
|
|
|
|
|
|
|
#
|
|
|
|
# Bug#14077 "Failure to replicate a stored function with a cursor":
|
|
|
|
# verify that stored routines with cursors work on slave.
|
|
|
|
#
|
|
|
|
connection master;
|
|
|
|
--disable_warnings
|
|
|
|
drop function if exists f1;
|
|
|
|
--enable_warnings
|
|
|
|
delimiter |;
|
|
|
|
create function f1() returns int reads sql data
|
|
|
|
begin
|
|
|
|
declare var integer;
|
|
|
|
declare c cursor for select a from v1;
|
|
|
|
open c;
|
|
|
|
fetch c into var;
|
|
|
|
close c;
|
|
|
|
return var;
|
|
|
|
end|
|
|
|
|
delimiter ;|
|
|
|
|
create view v1 as select 1 as a;
|
|
|
|
create table t1 (a int);
|
|
|
|
insert into t1 (a) values (f1());
|
|
|
|
select * from t1;
|
|
|
|
drop view v1;
|
|
|
|
drop function f1;
|
|
|
|
sync_slave_with_master;
|
|
|
|
connection slave;
|
|
|
|
select * from t1;
|
|
|
|
|
2006-01-20 15:21:39 +01:00
|
|
|
#
|
|
|
|
# Bug#16621 "INSERTs in Stored Procedures causes data corruption in the Binary
|
|
|
|
# Log for 5.0.18"
|
|
|
|
#
|
|
|
|
|
|
|
|
# Prepare environment.
|
|
|
|
|
|
|
|
connection master;
|
|
|
|
|
|
|
|
--disable_warnings
|
|
|
|
DROP PROCEDURE IF EXISTS p1;
|
|
|
|
DROP TABLE IF EXISTS t1;
|
|
|
|
--enable_warnings
|
|
|
|
|
|
|
|
# Test case.
|
|
|
|
|
|
|
|
CREATE TABLE t1(col VARCHAR(10));
|
|
|
|
|
|
|
|
CREATE PROCEDURE p1(arg VARCHAR(10))
|
|
|
|
INSERT INTO t1 VALUES(arg);
|
|
|
|
|
|
|
|
CALL p1('test');
|
|
|
|
|
|
|
|
SELECT * FROM t1;
|
|
|
|
|
|
|
|
sync_slave_with_master;
|
|
|
|
connection slave;
|
|
|
|
|
|
|
|
SELECT * FROM t1;
|
|
|
|
|
|
|
|
# Cleanup.
|
|
|
|
|
|
|
|
connection master;
|
|
|
|
|
|
|
|
DROP PROCEDURE p1;
|
|
|
|
|
2006-07-28 00:49:18 +02:00
|
|
|
|
|
|
|
#
|
|
|
|
# BUG#20438: CREATE statements for views, stored routines and triggers can be
|
|
|
|
# not replicable.
|
|
|
|
#
|
|
|
|
|
|
|
|
--echo
|
|
|
|
--echo ---> Test for BUG#20438
|
|
|
|
|
|
|
|
# Prepare environment.
|
|
|
|
|
|
|
|
--echo
|
|
|
|
--echo ---> Preparing environment...
|
|
|
|
--echo ---> connection: master
|
|
|
|
--connection master
|
|
|
|
|
|
|
|
--disable_warnings
|
|
|
|
DROP PROCEDURE IF EXISTS p1;
|
|
|
|
DROP FUNCTION IF EXISTS f1;
|
|
|
|
--enable_warnings
|
|
|
|
|
|
|
|
--echo
|
|
|
|
--echo ---> Synchronizing slave with master...
|
|
|
|
|
|
|
|
--save_master_pos
|
|
|
|
--connection slave
|
|
|
|
--sync_with_master
|
|
|
|
|
|
|
|
--echo
|
|
|
|
--echo ---> connection: master
|
|
|
|
--connection master
|
|
|
|
|
|
|
|
# Test.
|
|
|
|
|
|
|
|
--echo
|
|
|
|
--echo ---> Creating procedure...
|
|
|
|
|
|
|
|
/*!50003 CREATE PROCEDURE p1() SET @a = 1 */;
|
|
|
|
|
|
|
|
/*!50003 CREATE FUNCTION f1() RETURNS INT RETURN 0 */;
|
|
|
|
|
|
|
|
--echo
|
|
|
|
--echo ---> Checking on master...
|
|
|
|
|
|
|
|
SHOW CREATE PROCEDURE p1;
|
|
|
|
SHOW CREATE FUNCTION f1;
|
|
|
|
|
|
|
|
--echo
|
|
|
|
--echo ---> Synchronizing slave with master...
|
|
|
|
|
|
|
|
--save_master_pos
|
|
|
|
--connection slave
|
|
|
|
--sync_with_master
|
|
|
|
|
|
|
|
--echo ---> connection: master
|
|
|
|
|
|
|
|
--echo
|
|
|
|
--echo ---> Checking on slave...
|
|
|
|
|
|
|
|
SHOW CREATE PROCEDURE p1;
|
|
|
|
SHOW CREATE FUNCTION f1;
|
|
|
|
|
|
|
|
# Cleanup.
|
|
|
|
|
|
|
|
--echo
|
|
|
|
--echo ---> connection: master
|
|
|
|
--connection master
|
|
|
|
|
|
|
|
--echo
|
|
|
|
--echo ---> Cleaning up...
|
|
|
|
|
|
|
|
DROP PROCEDURE p1;
|
|
|
|
DROP FUNCTION f1;
|
|
|
|
|
|
|
|
--save_master_pos
|
|
|
|
--connection slave
|
|
|
|
--sync_with_master
|
|
|
|
--connection master
|
|
|
|
|
|
|
|
|
2005-11-17 01:15:10 +01:00
|
|
|
# cleanup
|
|
|
|
connection master;
|
|
|
|
drop table t1;
|
Fix for BUG#14769 "Function fails to replicate if fails half-way (slave stops)":
if the function, invoked in a non-binlogged caller (e.g. SELECT, DO), failed half-way on the master,
slave would stop and complain that error code between him and master mismatch.
To solve this, when a stored function is invoked in a non-binlogged caller (e.g. SELECT, DO), we binlog the function
call as SELECT instead of as DO (see revision comment of sp_head.cc for more).
And: minor wording change in the help text.
This cset will cause conflicts in 5.1, I'll merge.
mysql-test/r/rpl_sp.result:
result update
mysql-test/t/rpl_sp-slave.opt:
bug just fixed so option not needed
mysql-test/t/rpl_sp.test:
test for more half-failed functions with DO and SELECT, to test the bug of this changeset.
cleanup at the end.
sql/mysqld.cc:
function -> stored function (change suggested by Paul)
sql/sp_head.cc:
When a function updates data and is called from a non-binlogged statement (SELECT, DO), we binlog it
as SELECT myfunc(), and not DO myfunc() like before.
2006-02-18 17:26:30 +01:00
|
|
|
sync_slave_with_master;
|
2006-11-15 10:23:27 +01:00
|
|
|
|
2006-11-28 16:03:53 +01:00
|
|
|
#
|
|
|
|
# Bug22043: MySQL don't add "USE <DATABASE>" before "DROP PROCEDURE IF EXISTS"
|
|
|
|
#
|
Fix for BUG#19725 "Calls to SF in other database are not replicated
correctly in some cases".
In short, calls to a stored function located in another database
than the default database, may fail to replicate if the call was made
by SET, SELECT, or DO.
Longer: when a stored function is called from a statement which does not go
to binlog ("SET @a=somedb.myfunc()", "SELECT somedb.myfunc()",
"DO somedb.myfunc()"), this crafted statement is binlogged:
"SELECT myfunc();" (accompanied with a mention of the default database
if there is one). So, if "somedb" is not the default database,
the slave would fail to find myfunc(). The fix is to specify the
function's database name in the crafted binlogged statement, like this:
"SELECT somedb.myfunc();". Test added in rpl_sp.test.
mysql-test/r/rpl_sp.result:
Because I moved the SHOW BINLOG EVENTS down a bit, big portions of its
output move. Also, the function's database name appears in
SELECT statements.
mysql-test/t/rpl_sp.test:
Adding test for BUG#19725.
Moving the SHOW BINLOG EVENTS down, it is run at the very end to
test everything.
sql/sp_head.cc:
When binlogging a "SELECT myfunc()" (when a stored function is executed
inside a statement which does not go to the binlog (like a SET,
SELECT, DO), we need to write "SELECT db_of_myfunc().myfunc()",
because the function may be in a database which is not the default
database.
2007-01-08 22:01:06 +01:00
|
|
|
|
2006-11-28 16:03:53 +01:00
|
|
|
connection master;
|
|
|
|
--disable_warnings
|
|
|
|
drop database if exists mysqltest;
|
|
|
|
drop database if exists mysqltest2;
|
|
|
|
--enable_warnings
|
|
|
|
create database mysqltest;
|
|
|
|
create database mysqltest2;
|
|
|
|
use mysqltest2;
|
|
|
|
create table t ( t integer );
|
|
|
|
create procedure mysqltest.test() begin end;
|
|
|
|
insert into t values ( 1 );
|
|
|
|
--error ER_WRONG_DB_NAME
|
|
|
|
create procedure `\\`.test() begin end;
|
Fix for BUG#19725 "Calls to SF in other database are not replicated
correctly in some cases".
In short, calls to a stored function located in another database
than the default database, may fail to replicate if the call was made
by SET, SELECT, or DO.
Longer: when a stored function is called from a statement which does not go
to binlog ("SET @a=somedb.myfunc()", "SELECT somedb.myfunc()",
"DO somedb.myfunc()"), this crafted statement is binlogged:
"SELECT myfunc();" (accompanied with a mention of the default database
if there is one). So, if "somedb" is not the default database,
the slave would fail to find myfunc(). The fix is to specify the
function's database name in the crafted binlogged statement, like this:
"SELECT somedb.myfunc();". Test added in rpl_sp.test.
mysql-test/r/rpl_sp.result:
Because I moved the SHOW BINLOG EVENTS down a bit, big portions of its
output move. Also, the function's database name appears in
SELECT statements.
mysql-test/t/rpl_sp.test:
Adding test for BUG#19725.
Moving the SHOW BINLOG EVENTS down, it is run at the very end to
test everything.
sql/sp_head.cc:
When binlogging a "SELECT myfunc()" (when a stored function is executed
inside a statement which does not go to the binlog (like a SET,
SELECT, DO), we need to write "SELECT db_of_myfunc().myfunc()",
because the function may be in a database which is not the default
database.
2007-01-08 22:01:06 +01:00
|
|
|
|
|
|
|
#
|
|
|
|
# BUG#19725: Calls to stored function in other database are not
|
|
|
|
# replicated correctly in some cases
|
|
|
|
#
|
|
|
|
|
|
|
|
connection master;
|
|
|
|
delimiter |;
|
|
|
|
create function f1 () returns int
|
|
|
|
begin
|
|
|
|
insert into t values (1);
|
|
|
|
return 0;
|
|
|
|
end|
|
|
|
|
delimiter ;|
|
|
|
|
sync_slave_with_master;
|
|
|
|
# Let us test if we don't forget to binlog the function's database
|
|
|
|
connection master;
|
|
|
|
use mysqltest;
|
|
|
|
set @a:= mysqltest2.f1();
|
|
|
|
sync_slave_with_master;
|
|
|
|
connection master;
|
|
|
|
|
|
|
|
|
|
|
|
# Final inspection which verifies how all statements of this test file
|
|
|
|
# were written to the binary log.
|
|
|
|
--replace_column 2 # 5 #
|
|
|
|
show binlog events in 'master-bin.000001' from 98;
|
|
|
|
|
|
|
|
|
|
|
|
# Restore log_bin_trust_function_creators to its original value.
|
|
|
|
# This is a cleanup for all parts above where we tested stored
|
|
|
|
# functions and triggers.
|
|
|
|
set global log_bin_trust_function_creators=0;
|
|
|
|
connection master;
|
|
|
|
set global log_bin_trust_function_creators=0;
|
|
|
|
|
2006-11-28 16:03:53 +01:00
|
|
|
# Clean up
|
|
|
|
drop database mysqltest;
|
|
|
|
drop database mysqltest2;
|