Hema Sridharan wrote:
>> I think having a separate backup_charsets test suite is too
>> much granularity.
>> I'd define one common backup test suite where all the backup
>> tests go, including the character set related ones.
> Hema: This separate suite was created after consulting with Magnus, Lars,
> Also long back I sent an email to everyone asking suggestion about the same.
> There was a plan to move all the tests to backup suite, so Magnus said that
> there is no need to run all the
> backup tests with different character sets combinations. That's the reason
> to create separate backup_charsets after taking advice from him.
Ok, now I understand. I was afraid we will have a separate suite for each aspect
of backup which we are testing but now I see that there is a good reason for
separating charset testing.
>> This patch defines 2 test cases: backup_datatypes and
>> backup_partitions. But LLD of the WL talks about 5 test
>> cases. Moreover, neither of the test cases described in the
>> WL seems to correspond to one of the 2 test cases implemented
>> here. So, I would say that this patch does not implement the
>> design described in the WL... Am I wrong?
> Hema: No, you are correct. Actually I splitted some of the test cases in to
> separate tests to have good coverage.
> Also I thought that there is no specific test in backup that tests all
> datatypes and all partitions alone, so I created separate tests. This will
> help in running all datatypes/partitions with different combinations of
> character sets/collations/storage engines. This way we can have good test
Hmm, but then why is the new test included in the WL about testing different
character sets. If you need better tests for testing partitions and all
datatypes then it should go to a different (new?) WL. Including it in the
charsets WL is confusing.
>> I have also a question concerning the design. To me tests 2-4
>> from LLD seem to be basically the same. They all test correct
>> backup and restore of different objects with names using
>> various character sets. Then why not to define just one test
>> case for this? Is there any important difference between them
>> I'm missing?
> Hema: As far as testing of objects is concerned, WL#4222(Test of metadata
> consistency) tests all the objects. Chuck is reviewing the tests for this WL
> currently. Once I get approval to push from him, I will have to make slight
> modifications to the test so that the same test can be run for different
> combinations of csets and coll. I will let you know what modifications I am
> going to make once I do it. There are some tests are common for different
> There is no much difference between tests 2-4. I will modify the WL
> Please let me know if you have any questions further.
Then I'll wait for you to update the WL so that it describes the thing you are
going to finally implement. Let me know when it is ready and also if you create
any new patches.