Re: [Springframework-developer] Serializable support

[Tim M. <tec...@mo...>]

MacMahon, Mark wrote:

> Hi All,
> 
<snip>

> 
> I am also getting problems passivating the SFSB due to the fact that
> AbstractEnterpriseBean stores its logger as a member rather than a class
> variable:
> Is this convenience really worth sacrificing serializable support?
> 
> abstract class AbstractEnterpriseBean implements EnterpriseBean 
> {
> 
> 	/** Logger, available to subclasses */
> 	protected final Log logger = LogFactory.getLog(getClass());
> 	
> 	//Change to static to support serializable
> 	//private final static Log logger =
> LogFactory.getLog(AbstractEnterpriseBean.class);
> 	
> }
> 

I am working along with Mark on this issue and now have some further 
points to note:

I see that there is now a note on the "logger" variable that it should 
be reinitialised during activation for a stateful session bean. This has 
been implemented (logger set to null during passivation) but we now 
receive the following error during passivation:

javax.ejb.EJBException: Could not passivate; failed to save state; 
CausedByException is:
	org.springframework.aop.framework.Cglib2AopProxy
	at 
org.jboss.ejb.plugins.StatefulSessionFilePersistenceManager.passivateSession(StatefulSessionFilePersistenceManager.java:378)
	at 
org.jboss.ejb.plugins.StatefulSessionInstanceCache.passivate(StatefulSessionInstanceCache.java:85)

I have even set the logger on the Cglib2AopProxy to a static variable 
but this did not help.

I do reiterate Mark's comments about the loggers used in the Spring 
code. It would be much easier to use these classes if the loggers were 
static. It may require a little more code placed in to initialise a 
logger for the subclasses but the resulting stability would be worth it.

Currently it looks like we'll have to stop using Spring's stateful 
session bean until it can be passivated successfully.

Once again, thanks for a great framework and looking forward to 1.0!

Tim McAuley