Hello,
Andrei Elkin a écrit, Le 27.04.2011 17:52:
> That looks good. Approved.
Thank you!
Independently of this patch, your remarks triggered some more thoughts;
I observe (in 5.1, didn't test other versions) that this test:
--source include/have_innodb.inc
create table t1(a int auto_increment primary key, b char(1)) engine=innodb;
connect (con1,localhost,root,,test);
connection default;
send insert into t1 values
(null,'A'),(null,'A'),(1025,'A'),(null+sleep(5),'A');
connection con1;
sleep 1;
insert into t1 values(null,'B');
connection default;
--error 1062
reap;
select * from t1;
drop table t1;
yields this result:
create table t1(a int auto_increment primary key, b char(1)) engine=innodb;
insert into t1 values (null,'A'),(null,'A'),(1025,'A'),(null+sleep(5),'A');
insert into t1 values(null,'B');
ERROR 23000: Duplicate entry '1026' for key 'PRIMARY'
select * from t1;
a b
1026 B
drop table t1;
Notice how the "reap" gets "dup entry" error: the 4-row INSERT fails.
This is no replication bug (only the second INSERT was committed and
sent to binlog).
I think this is no bug at all; by specifying an explicit value the user
is bound to get dup key errors sometimes anyway in collision with
autoinc values from other threads. Just wanted to share this, I don't
plan to file a bug. I am using all default values of options.
