Chuck Bell wrote:
> Oops. Sent email too soon. There were 2 other things....
>
> 1) The test fails with the latest pull from mysql-6.0-backup on Windows.
> Comments suggest that the code should have failed, but in this case it
> passed causing a test failure.
>
> main.backup_multi_blocks [ fail ]
>
> mysqltest: At line 22: query 'RESTORE FROM 'test.ba'' succeeded - should
> have failed with errno 1698...
Ah, of course. It only fails on 64 bit platforms. Any ideas if it is
possible to say that if a statement fails it should fail with errno so
and so, but it is also OK that it succeeds? If not, I just will have to
leave out the test case from this patch and add it when fixing 36624.
>
> 2) Do not use hard-coded error values for the --error command. The
> following:
>
>> +--error 1698
>
> ...should be:
>
> --error ER_BACKUP_READ_SUMMARY
>
> This allows us to alter the errmsg.txt file and generate new error messages
> without breaking tests. This happens most frequenly in merges from other
> trees.
Good point. Will do.
--
Øystein