[hibernate-dev] Weird backref issue within the AttributeFactory (Was Re: Some code questions)

[Emmanuel Bernard]

OK I've fixed this problem but only to discover another one :(
http://opensource.atlassian.com/projects/hibernate/browse/HHH-4805

Basically, Adam makes use of generics to declare its components. Since HEM's metamodel generator does not use the XLayer that resolve these generics to their value, a TypeVariable object is returned instead of a Class.

I'll start working on the conversion to use the XLayer and hope it resolves this problem. But that's probably a longer process.

Damn you Adam, damn you!

On 14 janv. 2010, at 15:35, Steve Ebersole wrote:

> Ah, backrefs!
> 
> Just disregarding them makes the most sense for sure from the JPA
> perspective since they are "virtual attributes" and solely a hibernate
> construct.
> 
> On Thu, 2010-01-14 at 11:29 +0100, Emmanuel Bernard wrote:
>> On 13 janv. 2010, at 20:37, Hardy Ferentschik wrote:
>>> I have a couple of question regarding some recent code changes. The first one seems to be the cause of a
>>> test failure in VersionsJoinTableRangeComponentNamingTest (Envers). It actually causes a NullPointerException.
>>> Have a look at: http://fisheye.jboss.org/browse/Hibernate/core/trunk/annotations/src/main/java/org/hibernate/cfg/AbstractPropertyHolder.java?r=18518#l254
>> 
>> Right, I fixed the problem, sorry about that.
>> Though it uncovered a weird problem.
>> When I run the entire envers test suite, I pass with flags up.
>> When I specifically run VersionsJoinTableRangeComponentNamingTest
>> I've got a 
>> java.lang.IllegalArgumentException: Cannot determine java-type from given member [null]
>> 	at org.hibernate.ejb.metamodel.AttributeFactory$BaseAttributeMetadata.<init>(AttributeFactory.java:591)
>> 	at org.hibernate.ejb.metamodel.AttributeFactory$SingularAttributeMetadataImpl.<init>(AttributeFactory.java:671)
>> 	at org.hibernate.ejb.metamodel.AttributeFactory$SingularAttributeMetadataImpl.<init>(AttributeFactory.java:661)
>> 	at org.hibernate.ejb.metamodel.AttributeFactory.determineAttributeMetadata(AttributeFactory.java:542)
>> 	at org.hibernate.ejb.metamodel.AttributeFactory.buildAttribute(AttributeFactory.java:81)
>> 	at org.hibernate.ejb.metamodel.MetadataContext.wrapUp(MetadataContext.java:177)
>> 	at org.hibernate.ejb.metamodel.MetamodelImpl.buildMetamodel(MetamodelImpl.java:66)
>> 	at org.hibernate.ejb.EntityManagerFactoryImpl.<init>(EntityManagerFactoryImpl.java:79)
>> 	at org.hibernate.ejb.Ejb3Configuration.buildEntityManagerFactory(Ejb3Configuration.java:752)
>> 	at org.hibernate.envers.test.AbstractEntityTest.init(AbstractEntityTest.java:94)
>> 	at org.hibernate.envers.test.AbstractEntityTest.init(AbstractEntityTest.java:82)
>> 
>> The member is null because the "property" has a backref getter and hence no member is returned by NORMAL_MEMBER_RESOLVER. I think we should ignore backref properties
>> 
>> Does that make sense? If we agree, I will go and apply a fix.
>> 
>> http://opensource.atlassian.com/projects/hibernate/browse/HHH-4797
>> 
>> 
>>> 
>>> 
>>> Also have a look at http://fisheye.jboss.org/browse/Hibernate/core/trunk/annotations/src/main/java/org/hibernate/cfg/AnnotationConfiguration.java?r=18506#l896
>>> You create a AssertionFailure, but then don't throw it? I guess this is just missing a 'throw' before the new.
>> 
>> yep, fixed.
> -- 
> Steve Ebersole <steve at hibernate.org>
> Hibernate.org
>