maria-developers team mailing list archive

[Sergei Golubchik <serg@xxxxxxxxxxx>]

Hi, Oleksandr!

On Aug 05, Oleksandr Byelkin wrote:
> On 12.06.15 13:09, Sergei Golubchik wrote:
> > Hi, Sanja!
> >
> > On May 06, sanja@xxxxxxxxxxx wrote:
> >> revision-id: 3eceb51e697d5a514e70c374a06e7eef6b5a37a0
> >> parent(s): aa5095627e2619bdad7916d33d1016802a84a9e1
> >> committer: Oleksandr Byelkin
> >> branch nick: server
> >> timestamp: 2015-05-06 13:19:22 +0200
> >> message:
> >>
> >> MDEV-7990: ERROR 1526 when procedure executed for second time ALTER
> >> TABLE partition ... pMAX values less than MAXVALUE
> >>
> >> Added cleaning partition data which are saved in LEX structure to make
> >> possible correct re-execution of a SP or PS using the code.
> > No, this is wrong. See prep_alter_part_table(). The first thing it does
> > it to copy thd->lex->part_info to thd->work_part_info. It tries to do a
> > deep copy to be able to modify it freely, but apparently the copy isn't
> > deep enough. I see two options
> > 1. Do a deeper copy. Duplicate all value lists.
> > 2. Make part_column_list_val elements immutable. That would be
> >     preferred, of course, but I'm not sure it's possible.
> What is wrong? I modified clone procedure which used for getting 
> work_part_info deeper copy which is following case 1.

That's fine, but there should be no cleanup. The original design was to
create a deep copy of the structure. Then this copy can be modified
arbitrarily and later discarded without affecting the original.

But your cleanup, as far as I can see, is like doing the changes in the
original and then restoring the structure back to its initial state.
So, part of the structure is deep-copied but the leaves are modified in
place and are later "cleaned up". You've got two different approaches
applied to different levels of the part_info - and that was what I
didn't like.

Regards,
Sergei