openerp-expert-framework team mailing list archive

[Antony Lesuisse <al@xxxxxxxxxxx>]

The api looks ok, even if i think we should discourage the use of savepoints. 
@chs what do you think?

On 03/13/2014 08:05 PM, Valentin LAB wrote:
> Here's a MP to propose a simple way to add transaction where it could be
> needed without breaking legacy code:
>
> https://code.launchpad.net/~0k.io/openobject-server/trunk-new-prevent-cr-commit-to-allow-super-transaction/+merge/210884
>
>
> If this is okay, it's trivial to encompass openerp's initialize / update
> sequence in one go.
>
> Any comments ?
>
>
> On 13/03/14 15:16, Valentin LAB wrote:
> > Hi,
> >
> > I've noticed with great pleasure the introduction of a small savepoint
> > facility in the "sql_db.py" quite recently, and I looked how transaction
> > where used in the initialisation / update process.
> >
> > I don't understand why ``_auto_init`` uses ``cr.commit()`` every now and
> > them. What is the exact purpose of these ?
> >
> > Wouldn't it be much better NOT to use them at all ?
> > And if there are some hidden benefits, what about replacing them with
> > savepoints ?
> >
> > My main concern is about having a clean initialisation / update process:
> > we have issues quite often with modules that are not correctly updated,
> > and they are marked (and commited) wrongly in unsatisfactory states.
> > This leaves the whole database often in a intermediary state of which I
> > can't find any justification. We then face unique issues depending on
> > which modules failed and from which path we perform multi-update, and
> > often have to enter the weird multiple-updates process.
> >
> > I was changing the code to use savepoints to support sort of
> > "nested-commit", allowing to leave the database untouched if something
> > went wrong before the full initialisation occured.
> >
> > But why are these commits there in the first place ? What am I missing ?
> >
> > Thanks for your enlightenments,