Scott,

This is fine. All Jetty needs is a definite policy from JBoss - either
load all the classes/jars from the war, or don't load any of it - Jetty
is able to handle both situations.

If you don't crack open the wars, then we will probably go a step
further and add a switch to our service mbean to allow the specification 
of servlet 2.3 compliance (delegate "system" classes) or Java 2 
compliance (always delegate) for class loading.

... all this assuming, of course, that Marc's unified class loaders 
don't achieve world domination!

Jan

> To fix the current problems with class loading in wars I
> have dropped the parsing of war subcomponents by the
> MainDeployer. The problem was that the jars found in
> wars were being broken out and loaded as deployment units
> and added to the JBoss ServiceLibraries. The WEB-INF/classes
> directory was not added correctly. When a class in the some.jar
> (WEB-INF/lib/some.jar) tried to load a class from
> WEB-INF/classes using Marc's favorite Class.forName(String),
> this foreced used of the JBoss classloader for the some.jar
> subdeployment and this knew nothing about WEB-INF/classes
> and so failed.
> 
> Instead of heading further into the abyss and also adding the
> WEB-INF/classes to the ServiceLibraries, I don't want to
> deploy any subcomponents of wars. Let the servlet container
> handle this. In addition to their own WEB-INF/classes
> reload semantics, they also have to deal with the recommended
> non-standard classloader behavior defined in the 2.3 servlet
> spec:
> 
[snip]
> 
> The only downside to throwing out the parsing of the war subcomponents
> is the potential loss of jar manifest classpaths. The servlet container is
> supposed to handle this as well so this needs to be tested.







_______________________________________________
Jboss-development mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/jboss-development

Reply via email to