Details
-
Improvement
-
Status: Closed
-
Major
-
Resolution: Fixed
-
2.0-dev/cvs, 2.0-M2, 2.0-FINAL
-
None
Description
The current implementation of handling deployment, redeployment and undeployment of Portlet Applications (PAs) contains
several workarounds to prevent file locking of jars and classes on Windows and to allow layout portlets to run under
the Jetspeed-2 context.
After almost a day debugging and testing I found the real cause of all the problems we try to fix:
our own URLClassLoaders we create to be able to access the portlet and portlet resource bundles in the PA.
As a test, I replaced our own URLClassLoaders with the contextClassLoader from the PA, retrieved during the init
of the JetspeedServletContainer. And viola, no more problem!
I'm going to drastically refactor the whole deployment implementation to be able to only use the PA contextClassLoaders.
The autodeployment from the deploy folder will be changed to infuse the PA war (if needed) and move it to the webapps
folder so the autodeployment of the application server will handle the real deployment.
Thus: the undeployment from the deploy folder for PAs will be gone!
Redeployment will still work as expected though.
For undeployment, an new undeployment feature in the PAM Portlet will be added.
Additionally, other PA management features like upload/start/stop/reload/ should be added soon too. We need a separate
issue for that though.
Decorators (and soon layouts too) will still be handled as is done right now (they don't involve classloaders).
For deployment of layout portlets I want to make the following changes:
- must be deployed as jar, not as war
- can only contain classes, no resources
- resources required by a layout portlet must be deployed separately with a decorators or layouts jar
- will not be deployed as web app but the jar will be loaded with a URLClassloader with the portal classloader
as parent: the portlets will be invoked through a JetspeedContainerServlet from the portal itself.
These changes will also fix the outstanding problem with running Jetspeed-2 under a different context!
(seeJS2-172andJS2-182)
I'm going to need several days to perform this refactoring so if anyone has problems with this plan please do comment within the next few days!