Re: bk commit into 5.1 tree (acurtis:1.2555) BUG#29875

[Antony T Curtis]

On Fri, 2007-07-20 at 09:44 +0200, Ingo Strüwing wrote:
> Hi Antony,
> 
> OK to push from me. But please consider the below suggestions.
> 
> antony@stripped wrote:
> ...
> > ChangeSet@stripped, 2007-07-18 12:22:05-07:00, acurtis@stripped +3 -0
> >   Bug#29875
> >     "Disable transaction support in Federated storage engine"
> >     Minimal patch to disable Federated's transactions until they can be fixed.
> ...
> > +  /*
> > +    Support for transactions disabled until WL#2952 fixes it.
> > +  */
> > +  federated_hton->commit= 0;
> > +  federated_hton->rollback= 0;
> > +
> 
> I was a bit confused that you did not remove the former assignments to
> these function pointers. I suggest to do similar as in ::external_lock:
> 
>   /*
>     Support for transactions disabled until WL#2952 fixes it.
>   */
> #ifdef XXX_SUPERCEDED_BY_WL2952
>   federated_hton->commit= federated_commit;
>   federated_hton->rollback= federated_rollback;
> #else
>   federated_hton->commit= 0;
>   federated_hton->rollback= 0;
> #endif

If I did that, I would receive compiler warnings that federated_commit
and federated_rollback were declared and not used.

> ...
> 
> It would also be great to see one or two small tests that show how
> attempts to use transactions with FEDERATED are rejected.

federated_transactions was disabled due to change of results. I suppose
I can put in a few quick/simple tests.

Regards,

-- 
Antony T Curtis, Senior Software Developer
MySQL Inc, www.mysql.com
SIP: 4468@stripped