Hakan Kuecuekyilmaz wrote:
> On Di, 2008-09-09 at 13:07 +0200, John Embretsen wrote:
[...]
>>> === renamed file 'mysql-test/suite/falcon_team/r/falcon_bug_24024.result'
> => 'mysql-test/suite/falcon/r/falcon_bug_24024.result'
>>> --- a/mysql-test/suite/falcon_team/r/falcon_bug_24024.result 2008-04-20
> 00:05:17 +0000
>>> +++ b/mysql-test/suite/falcon/r/falcon_bug_24024.result 2008-09-09 08:30:18
> +0000
>>> @@ -14,10 +16,13 @@ a b
>>> 1 NULL
>>> 7 NULL
>>> DROP TABLE t1;
>>> -!!! Error message has to be decided.
>>> -!!! Something like "Cannot drop locked table" or similar would be fine.
>>> -!!! But never the misleading ERROR 42S02: Unknown table 't1'
>>> -COMMIT;
>>> -# Switch to session default and disconnect session con1
>>> +ERROR 42S02: Unknown table 't1'
>>> +SHOW WARNINGS;
>>> +Level Code Message
>>> +Warning 178 Can't execute the given command because you have active locked
> tables or an active transaction
>> Although this change is OK, I don't really like the fact that a request
>> of never using the misleading "Unknown table" error message for this
>> issue is removed.
>
> I put back the comment.
All right.
[...]
>> So I'm OK with this particular change to the test, but I wonder if there
>> is a bug reported against the server regarding this?
>
> There is no bug report against the server regarding this, which I am
> aware of.
OK. I noticed that Bug#22972 had category "Server: Errors", so I added a
comment to that issue. Not sure if anyone will notice, though.
>> Please explain why you did the following change:
>>
>>> --- a/mysql-test/suite/falcon_team/t/falcon_bug_22189.test 2008-04-28
> 11:21:52 +0000
>>> +++ b/mysql-test/suite/falcon/t/falcon_bug_22189.test 2008-09-09 08:30:18
> +0000
>>> @@ -68,8 +68,10 @@ SELECT * FROM x1;
>>>
>>> --echo # Switch to connection conn1
>>> connection conn1;
>>> ---send UPDATE x1 SET x1 = 0, x2 = 5
>>> ---send INSERT INTO x1 VALUES (0,6)
>>> +--error ER_DUP_ENTRY
>>> +UPDATE x1 SET x1 = 0, x2 = 5;
>>> +--error ER_DUP_ENTRY
>>> +INSERT INTO x1 VALUES (0,6);
>
> I removed the two --send's, because the statements where not blocking
> anymore. Having two consecutive --send's can lead to race conditions.
> Removing the --send's makes the test more robust.
OK, I see. Useful to know, and now it is in the archives. Thanks.
>>> --echo # Switch to connection default
>>> connection default;
>>> @@ -78,8 +80,6 @@ ROLLBACK;
>>> --echo # Switch to connection conn1
>>> connection conn1;
>>> --error ER_DUP_ENTRY
>>> ---reap
>>> ---error ER_DUP_ENTRY
>>> INSERT INTO x1 VALUES (0,6);
>>> --error ER_DUP_ENTRY
>>> INSERT INTO x1 VALUES (0,6);
>
> Every --send needs a --reap, therefore I removed the --reap.
Obviously :)
>> Changes to falcon_bug_23945.[test|result] look good.
>>
>> This test is still problematic, since it causes subsequent tests in the
>> test suite to fail ("1005: Can't create table 'test.t1' (errno: 156)").
>>
>> Until that is fixed, we should at least add a note and/or echo statement
>> about that in the test case, or possibly even disable it.
>>
>> However, those things can be done later in a separate changeset if so
>> desired.
>
> Agreed. I think disabling it would be good, so that the other tests
> don't suffer from the side effect. I think we should do it in a separate
> commit.
OK, deal.
thanks,
--
John
