Re: [eclipselink-users] PROGRESS: Sv: Re: Value of DiscriminatorValue no

[Andrei Ilitchev <andrei.ilitchev@xxxxxxxxxx>]

The problem is that method is called by a ForeignReferenceMapping 
passing its reference descriptor:

The problem arises when the actual referenced object is of derived type.

That's obviously a bug.

To fix this, descriptor should be substituted for the correct one:

public ObjectChangeSet findOrCreateLocalObjectChangeSet(Object 
entityClone, ClassDescriptor descriptor, boolean isNew){
    ObjectChangeSet changes = 
(ObjectChangeSet)this.getObjectChangeSetForClone(entityClone);
    if (changes == null) {
    	if (descriptor.hasInheritance() && descriptor.getJavaClass() != 
entityClone.getClass()) {
            descriptor = 
descriptor.getInheritancePolicy().getDescriptor(entityClone.getClass());
    	}
        if (descriptor.isAggregateDescriptor()) {
            changes = new AggregateObjectChangeSet(CacheId.EMPTY, 
descriptor, entityClone, this, isNew);
        } else {
            changes = new 
ObjectChangeSet(descriptor.getObjectBuilder().extractPrimaryKeyFromObject(entityClone, 
session), descriptor, entityClone, this, isNew);
        }
        changes.setIsAggregate(descriptor.isDescriptorTypeAggregate());
        this.addObjectChangeSetForIdentity(changes, entityClone);
    }
    return changes;
}

Please log a bug.

Thanks,
Andrei

On 8/20/2013 4:14 PM, Andreas Joseph Krogh wrote:
> På tirsdag 20. august 2013 kl. 18:52:25, skrev Andreas Joseph Krogh
> <andreak@xxxxxxxxxxxx <mailto:andreak@xxxxxxxxxxxx>>:
>
>     På tirsdag 20. august 2013 kl. 18:10:28, skrev Andreas Joseph Krogh
>     <andreak@xxxxxxxxxxxx <mailto:andreak@xxxxxxxxxxxx>>:
>
>         På tirsdag 20. august 2013 kl. 17:28:39, skrev Christopher
>         Delahunt <christopher.delahunt@xxxxxxxxxx
>         <mailto:christopher.delahunt@xxxxxxxxxx>>:
>
>             Hello Andreas,
>
>             Can you provide the stack trace from the debugger where you
>             noticed the
>                ObjectBuilder.buildRow was called on the EmailPart
>             ObjectBuilder/Descriptor?  Short of reproducing it locally,
>             that might
>             show the code path taken and point to how the wrong
>             objectbuilder/descriptor is getting used.  If you are able
>             to debug
>             further, I'd go back from that point and see how that
>             ObjectBuilder was
>             looked up.
>
>         I think I'm on to something.
>
>     This is where UnitOfWorkChangeSet.cloneToObjectChangeSet is
>     populated with the wrong ObjectChangeSet:
>
>     *public**ObjectChangeSet*findOrCreateLocalObjectChangeSet(*Object*entityClone,*ClassDescriptor*descriptor,*boolean*isNew){
>          *ObjectChangeSet*changes = (*ObjectChangeSet*)*this*.getObjectChangeSetForClone(entityClone);
>          *if*(changes ==*null*) {
>              *if*(descriptor.isAggregateDescriptor()) {
>                  changes =*new*AggregateObjectChangeSet(*CacheId*./*EMPTY*/, descriptor, entityClone,*this*, isNew);
>              }*else*{
>                  changes =*new*ObjectChangeSet(descriptor.getObjectBuilder().extractPrimaryKeyFromObject(entityClone,*session*), descriptor, entityClone,*this*, isNew);
>              }
>              changes.setIsAggregate(descriptor.isDescriptorTypeAggregate());
>              *this*.addObjectChangeSetForIdentity(changes, entityClone);
>          }
>          *return*changes;
>     }
>
>     The value of the descriptor is:
>     RelationalDescriptor(no.officenet.origo.email.domain.model.entities.EmailPart
>     --> [DatabaseTable(origo_email_part)])
>     and not EmailSinglePart as I would expect.
>     I need to figure out why the descriptor sometimes is of the correct
>     type when adding to the same collection (EmailMessage.rootPartList).
>
> I have an isolated test-case in JAVA now. Will file bug and post link.
> --
> Andreas Joseph Krogh <andreak@xxxxxxxxxxxx>      mob: +47 909 56 963
> Senior Software Developer / CTO - OfficeNet AS - http://www.officenet.no
> Public key: http://home.officenet.no/~andreak/public_key.asc
>
>
> _______________________________________________
> eclipselink-users mailing list
> eclipselink-users@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/eclipselink-users