Move *HibernateSession to a Proxy implementation

[epbernard]

I am trying to implement http://jira.jboss.com/jira/browse/EJBTHREE-714 and avoid too tight dependency between the container and Hibernate.

I would like to use a proxy based implementation for TransactionScopedHibernateSession, ExtendedHibernateSession rather than a tradtional implementation approach. It will avoid breakage when a method is added to either
org.hibernate.classic.Session.class
org.hibernate.engine.SessionImplementor.class
org.hibernate.jdbc.JDBCContext.Context.class
org.hibernate.event.EventSource.class

I am looking for something similar to http://fisheye.jboss.org/browse/Hibernate/branches/Branch_3_2/Hibernate3/src/org/hibernate/context/ThreadLocalSessionContext.java?r=10125
method #wrap()

Does anyone see something bad in this strategy?

[bill.burke]

Lol, yeah, you're right....Not sure why I did it that way to begin with. Either I was just being stupid, or at one time I had to intercept specific methods.

[brian.stansberry]

For fine-grained replication of SFSBs that have an XPC, we'd talked about fine-grained replication of the EM. Will this make that impossible?

[epbernard]

No it should not have any impact. The proxy can be ignored by the "serialization", you want to replicate the underlying EM

[epbernard]

Where is ExtendedPersistenceContext.getFactory() (return an ManagedEntityManagerFactory) used?

EJB 3.0 does not seem to use it and it conflicts with o.h.e.SessionImplementor.getFactory() (returns a SessionFactoryImplementor)

Can I get rid of it?

[bill.burke]

looks like an artifact. I'll remove it and put in the proxy implementation too. Just let me get a clean checkout. My wc is really old.

[epbernard]

I am almost done actually already
I am passing the tests
stateful is having issues, I need to investigate

[epbernard]

Done in both head and 4.2
I also removed
ExtendedPersistenceContext.getFactory()

I used a JDK Proxy generation approach if someone want to use Javassist, feel free.