Re: bzr commit into mysql-6.0-backup branch (oystein.grovlen:2674) Bug#36795

[Øystein Grøvlen]

Thanks for the quick response, Chuck.  I will go ahead and make the name 
change and add the other combinations of concurrent backup and restore 
to the test.

--
Øystein


Chuck Bell wrote:
> Hi, 
> 
>>> I would use the "before_backup_data" breakpoint -- see 
>> kernel.cc @916. 
>>> This will ensure backup is in the middle (after metadata and before 
>>> data). That should be safe. The following is a suggestion 
>> for how to 
>>> set this up (see other tests that use breakpoints for ideas).
>> I used the after_backup_start_backup breakpoint.   That provoked a 
>> segfault withot the fix.  Any reason I should change to 
>> before_backup_data?
> 
> None at all.
> 
>>>> +  static Backup_restore_ctx *current;
>>> I am not thrilled with this variable name change. What was 
>> wrong with
>>> is_running? That made more contextual sense. Please let me know your
>>> thoughts on this.
>> Note that it is not just a name change.  I have changed the 
>> type, too. 
>> I feel that is_running is a good name for a boolean, but for 
>> a pointer. 
>>   It should describe what it points to.  I agree that current 
>> is a bit 
>> generic.  My first idea was current_backup, but then I 
>> realized it could 
>> be a restore to.  Maybe current_op or running_op would work?  Any 
>> suggestions?
> 
> I think current_op would help as well as a short descriptive comment on the
> declaration.
> 
> Chuck
> 
> 


-- 
Øystein Grøvlen, Senior Staff Engineer
Sun Microsystems, Database Group
Trondheim, Norway