[ovirt-devel] hibernate's internal PersistentBag sent to FE

[Alexander Wels]

On Monday, June 01, 2015 09:33:07 AM Tomas Jelinek wrote:
> ----- Original Message -----
> 
> > From: "Alexander Wels" <awels at redhat.com>
> > To: devel at ovirt.org
> > Cc: "Tomas Jelinek" <tjelinek at redhat.com>, "Liran Zelkha"
> > <lzelkha at redhat.com> Sent: Monday, June 1, 2015 3:16:34 PM
> > Subject: Re: [ovirt-devel] hibernate's internal PersistentBag sent to FE
> > 
> > On Monday, June 01, 2015 09:08:50 AM Tomas Jelinek wrote:
> > > Hey all,
> > > 
> > > since the org.ovirt.engine.core.common.job.Job/Step... has been moved to
> > > use
> > > the JPA we have a problem on frontend. The problem is that the
> > > @OneToMany
> > > annotations results in a List which is of type PersistentBag. When we
> > > send
> > > this to Frontend it fails during deserialization. It actually fails
> > > quite
> > > bad because the FE already has an ui-override of it which is not correct
> > > resulting in a ton of NPEs in development mode.
> > > 
> > > So, there are 2 nasty fixes I have made where none of them should be
> > > merged
> > > but demonstrate the possibilities: 1: extend the FE to be able to work
> > > with
> > > the PersistentBag (https://gerrit.ovirt.org/#/c/41682/) not really good
> > > solution since the PersistenBag is an internal Hibernate class which is
> > > really not meant to be passed around
> > > 
> > > 2: fix on the backend to not send the PersistentBag but an ArrayList.
> > > This
> > > is only a PoC fixed on a command we face the problem
> > > (https://gerrit.ovirt.org/#/c/41797/) Obviously this is not going to
> > > work
> > > for other commands accessing the same Job nor for other entities.
> > > 
> > > So, the first option is generic but very very bad. The second option
> > > should
> > > be used but not sure how to do this in a cheep way (e.g. without using
> > > reflection to deep traverse everything sent back to frontend checking if
> > > it
> > > does not have a PersistentBag in it.
> > 
> > Tomas,
> > 
> > Thanks, I was investigating the same issue, I noticed it last Friday just
> > before leaving, so I was investigating the problem to see what was going
> > on. You are right we should not be sending PersistentBag to the frontend
> > at all. So how about we do a combination of [1] and [2], but instead of
> > delegating in [1] we actually simple throw an exception stating don't
> > sent PersistentBag to the front end. that way anyone inadvertently using
> > it will be notified immediately (since their code won't work).
> 
> Throwing the exception would help us in debugging but the main question is
> how will we make it work? Since we are planning to move more and more to
> JPA so we will face this issue more and more often. Solving it one by one
> on backend in each command is not going to work.

How about something like this [1]? It appears to use some aspects to translate 
the hibernate internal classes into normal java util classes.

[1] https://code.google.com/p/dehibernator/


> > Alexander
> > 
> > > Any better ideas?
> > > Thanx,
> > > Tomas
> > > _______________________________________________
> > > Devel mailing list
> > > Devel at ovirt.org
> > > http://lists.ovirt.org/mailman/listinfo/devel