RE: [picocontainer-dev] NanoWAR and session madness in Tomcat 5

> Note that the original implementation has the containerKiller variable
> as a class transient field but was moved to a member variable in
> killContainer() method to fix http://jira.codehaus.org/browse/NANOWAR-14
> 
> Can you please try to apply this simple fix and test in Tomcat 5?

Spoke too soon.  It doesn't throw exceptions, so that's good.  But container
behavior will not survive a restart, ie session objects are recreated
without their previous state (the session bound event clobbers the old
container).  


Update:

Well, it still throws exceptions too.  It didn't in Jake's examples, which
was progress.  But it does still try to serialize the App container too.
(At least it doesn't try to serialize KILLER_HELPER, so that's a huge
improvement).

I've attached NanoWarOptionalSession.zip to NANOWAR-15 as a workaround.  (It
leaves sessions completely out of the picture).