Serge, Alfranio, hello.
> Serge Kozlov wrote:
>> Alfranio Correia пишет:
>>> Hi Serge,
>>>
>>> If the patch aims to provide more information on the failure, I can
>>> agree
>>
>> Yes, the patch will provide that localizes place of failure.
>
> Let's wait for the second reviewer.
> If he approves I have no problem in doing so.
>
As far as I have understand, there is still no idea why the test failed in
PB. The patch does not show that.
Considering that the test failures ceased for more than two months
(whereas previosly they were some per month) I think we can think the
bug is over. Therefore, I would suggest to mark the bug as Can-t-Repeat.
> Cheers.
>>
>>
>> although still
>>> have doubts if it will provide it.
>>>
>>> Cheers.
>>>
>>> Serge Kozlov wrote:
>>>> Alfranio Correia пишет:
>>>>> Hi Serge,
>>>>>
>>>>> See comments in-line.
>>>>>
>>>>> Serge Kozlov wrote:
>>>>>> #At file:///home/ksm/sun/repo/bug38077/mysql-5.1-bugteam/ based
> on
>>>>>> revid:timothy.smith@stripped
>>>>>>
>>>>>> 2853 Serge Kozlov 2009-04-02
>>>>>> Bug#38077.
>>>>>> 1. Replace waiting of SQL thread stop by waiting of SQL
> error
>>>>>> on slave and stopped SQL thread.
that's ok.
>>>>>> 2. Remove debug code because it already implemented in
> MTR2.
>>>>>> modified:
>>>>>> mysql-test/suite/rpl/t/rpl_incident.test
that's ok.
>>>>>>
>>>>>> === modified file 'mysql-test/suite/rpl/t/rpl_incident.test'
>>>>>> --- a/mysql-test/suite/rpl/t/rpl_incident.test 2008-08-14
>>>>>> 09:38:22 +0000
>>>>>> +++ b/mysql-test/suite/rpl/t/rpl_incident.test 2009-04-02
>>>>>> 19:33:02 +0000
>>>>>> @@ -14,42 +14,13 @@ REPLACE INTO t1 VALUES (4);
>>>>>> SELECT * FROM t1;
>>>>>>
>>>>>> connection slave;
>>>>>> -source include/wait_for_slave_sql_to_stop.inc;
>>>>>> +# Wait until SQL thread stops with error LOST_EVENT on master
>>>>>> +let $slave_sql_errno= 1590;
>>>>>> +source include/wait_for_slave_sql_error.inc;
>>>>>>
>>>>> You are waiting until the sql thread stops due to an specific
>>>>> error. Before that, you were just waiting for any error.
I would answer this reasonable question with that previous unconditional waiting
was too coarse.
regards,
Andrei
>>>> No, currently test waits stop of sql thread. This code will pass even
>>>> if slave stopped by unknown reasons: PB error or something else.
>>>> Waiting of specific error guaranties that test will be interrupted at
>>>> this point if behavior of slave is another than we expected.
>>>>
>>>>> Unfortunately, in the context of this test case, I can't see the
>>>>> difference to avoid having the failures in pushbuild.
>>>>>
>>>>> Can you explain what I am missing?
>>>>>
>>>>> Cheers.
>>>>>
>>>>>>
>>>>>> # The 4 should not be inserted into the table, since the
> incident
>>>>>> log
>>>>>> # event should have stop the slave.
>>>>>> --echo **** On Slave ****
>>>>>> -#### BEGIN DEBUG INFO ADDED BY SVEN 2008-07-18 -- SEE BUG#38077
> ####
>>>>>> -let $tables= query_get_value(SHOW TABLES, Tables_in_test, 1);
>>>>>> -if (`SELECT '$tables' != 't1'`)
>>>>>> -{
>>>>>> - --echo **** TEST CASE BUG! PRINTING DEBUG INFO! ****
>>>>>> - --echo **** Dear developer, if you see this in the output of
> a
>>>>>> test
>>>>>> - --echo **** case run, please add all the information below as
> a
>>>>>> - --echo **** comment to BUG#38077. If it's a pushbuild
> failure,
>>>>>> please
>>>>>> - --echo **** include a link to the push page.
>>>>>> - --echo **** Thank you! /Sven
>>>>>> - SHOW BINLOG EVENTS;
>>>>>> - --echo **** master binlog ****
>>>>>> - --error 0,1
>>>>>> - --exec $MYSQL_BINLOG --hexdump
>>>>>> $MYSQLTEST_VARDIR/log/master-bin.000001
>>>>>> - --echo **** slave binlog ****
>>>>>> - --error 0,1
>>>>>> - --exec $MYSQL_BINLOG --hexdump
>>>>>> $MYSQLTEST_VARDIR/log/slave-bin.000001
>>>>>> - --echo **** slave status ****
>>>>>> - query_vertical SHOW SLAVE STATUS;
>>>>>> - --echo **** slave's master status ****
>>>>>> - SHOW MASTER STATUS;
>>>>>> - --echo **** slave binlog events ****
>>>>>> - --echo [on master]
>>>>>> - connection master;
>>>>>> - --echo **** master status ****
>>>>>> - SHOW MASTER STATUS;
>>>>>> - --echo **** master binlog events ****
>>>>>> - SHOW BINLOG EVENTS;
>>>>>> - exit;
>>>>>> -}
>>>>>> -#### END DEBUG INFO ####
>>>>>> SELECT * FROM t1;
>>>>>>
>>>>>> --replace_result $MASTER_MYPORT MASTER_PORT
>>>>>>
