[jboss-dev-forums] [Design of EJB 3.0] - Re: Plugin is Broken

[wolfc]

It doesn't say that we need to be able to build AS against snapshots. It says that we need to be backwards compatible all the time. The reason for this is that you can't patch everything (or even anything) in AS, because that's outside of our control.

Let's reiterate the (informal) procedure:
I wanted to make an SPI with persistence unit resolving (Jan 13th) http://viewvc.jboss.org/cgi-bin/viewvc.cgi/jbossas/projects/ejb3/trunk/core/src/main/java/org/jboss/injection/InjectionContainer.java?r1=75613&r2=75612&pathrev=75613 which involved AS, EJB3 and JPA.

Step one: deprecate InjectionContainer.getPersistenceUnitDeployment and add ExtendedInjectionContainer (the true SPI)
Step two: release the SPI
Step three: change all components to match
Step four: release all
Then step five: remove said method
Especially step four is tricky, because we don't control most bits.
I should add step seven: clean up the residual mess

Now the true problem, as you already spotted, is that ejb3 module has a bean definition which relies on proxy internals.

I pointed at with EAR code needing a patch (or MC) as another example, I just hadn't expected the problem to appear from this angle.

Now the passing of the TCK tests again proves the need for this transition component, because no TCK means no release and that would have deadlocked the situation.

So work on the assumption that you'll not get a patch in AS.

View the original post : http://www.jboss.org/index.html?module=bb&op=viewtopic&p=4218007#4218007

Reply to the post : http://www.jboss.org/index.html?module=bb&op=posting&mode=reply&p=4218007