Alfranio Correia wrote:
> Hi Sven,
>
> Sven Sandberg wrote:
>> Hi Alfranio,
>>
>> Great work with refactoring and adding comments! The comments you added
>> are very clear, making the test *much* easier to follow. I'm very happy
>> with this.
>>
>> Just one simple problem:
>>
>> (S1) Since there is no sync_slave_with_master in
>> rpl_implicit_commit_binlog.test, there is a race in the test (so it will
>> fail sporadically in pushbuild). See details inline.
>
> I know that. It fails sporadically.
>
> I thought about disabling the test case and re-enabling it as soon as the bug
> was fixed. However, I can do what you suggested.
>
> Is this is the only comment you have?
Yes, I think the test case is very good now, only this race shold be fixed.
/Sven
>
>
> Cheers.
>
>
>>> --echo
>>>
> ###################################################################################
>>>
>>> --echo # CLEAN
>>> --echo
>>>
> ###################################################################################
>>>
>>> connection master;
>>>
>>> SET SQL_LOG_BIN=0;
>>> DROP TABLE tt_1;
>>> DROP TABLE tt_2;
>>> SET SQL_LOG_BIN=1;
>>>
>>> connection slave;
>>>
>>> SET SQL_LOG_BIN=0;
>>> DROP TABLE tt_1;
>>> DROP TABLE tt_2;
>> (S1) There is a race between the slave sql thread and the connection
>> issuing 'drop table', since we don't do sync_slave_with_master. If the
>> slave sql thread has not received 'create table' yet, then it will fail
>> on 'DROP TABLE'.
>>
>> There is also a race between the slave sql thread and check-testcase.
>> For example, if the slave has executed 'create view' but not 'drop view'
>> when check-testcase runs, then check-testcase will report an error.
>>
>> I would suggest that you run the entire test with SQL_LOG_BIN=0 on
>> master until BUG#46572 is fixed.
>>
>>> SET SQL_LOG_BIN=1;
>>>
>>> #Re-enable this after fixing failures for BUG#46572
>>> #sync_slave_with_master;
>>
>>
>>
>>
--
Sven Sandberg, Software Engineer
MySQL AB, www.mysql.com
