Rafal Somla wrote:
> Øystein Grøvlen wrote:
>> STATUS
>> ======
>>
>> One request for change.
>>
>> REQUESTS
>> ========
>>
>> 1. If it is a requirement that duplicate dbs should give error, I
>> think the test needs to fail if it does not. Otherwise we would
>> not detect regressions. Hence, I think this should be change to
>> expect an error:
> >
>> > +--error 0, ER_NONUNIQ_DB
>> > +BACKUP DATABASE bup_default, bup_default TO "bup_default1.bak";
>>
>
> The fact that correct error is given is tested in backup_errors test.
> Inside backup test, I only set-up a trap for a faulty server as it was
> before this patch.
OK, I understand.
>
>>
>> COMMENTARY
>> ===========
>>
>> I do not really see the point in adding tests for something that
>> should not work in case it starts working. It complicates the test to
>> handle something that may never be needed.
>>
>
> This breaks my heart because I included this test scenario in
> expectation of your request for something along these lines ;) But if
> you don't see the point of such test, then I'm fine with leaving backup
> test unchanged and using only the backup_errors test. Do you concur?
Yes, I think the changes to backup_errors should be sufficient testing
for this.
--
Øystein
