[weld-issues] [JBoss JIRA] Commented: (WELD-787) CurrentInjectionPoint ThreadLocal leak with Weld Servlet

[Aslak Knutsen (JIRA)]

    [ https://issues.jboss.org/browse/WELD-787?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12570450#comment-12570450 ] 

Aslak Knutsen commented on WELD-787:
------------------------------------

Any of the current weld-servlet-test-tomcat test cases should do to trigger the issue.

The problem is that Tomcats WebappClassLoader, which is the class that clears the ThreadLocals, will only clear the local variables if the key or value is loaded by the WebappClassLoader.

In the weld-servlet-test-tomcat setup, Weld Servlet is loaded by AppCL, not the WebappClassLoader, so no leak is reported.

If you breakpoint: WebappClassloader.clearReferencesThreadLocals and loop until it handles Thread[http-127.0.0.1-8888-1,5,main]. you will see in WebappClassloader.clearThreadLocalMap that the variable CurrentInjectionPoint is ignored since it's not loaded by this(ClassLoader),

To trigger the leak with warning, any full War the bundles Weld-Servlet deployed on tomcat should do the trix.




> CurrentInjectionPoint ThreadLocal leak with Weld Servlet
> --------------------------------------------------------
>
>                 Key: WELD-787
>                 URL: https://issues.jboss.org/browse/WELD-787
>             Project: Weld
>          Issue Type: Bug
>          Components: Servlet Container Support
>    Affects Versions: 1.1.0.Beta2
>         Environment: Weld Servlet 1.1.0-SNAPSHOT / Beta2 / Beta1 on Tomcat 6 
>            Reporter: Aslak Knutsen
>             Fix For: 1.2.0.Beta1
>
>
> Nov 22, 2010 7:13:40 PM org.apache.catalina.loader.WebappClassLoader clearThreadLocalMap 
> SEVERE: The web application [/active1] created a ThreadLocal with key of type [null] (value [org.jboss.weld.injection.CurrentInjectionPoint$1 at 30b95f2]) and a value of type [java.util.Stack] (value [[]]) but failed to remove it when the web application was stopped. This is very likely to create a memory leak. 
> The CurrentInjectionPoint push/pop works, since the Stack object on the Thread is Empty when undeploy. 
> CurrentInjectionPoint.cleanup is called during the contextDestroyed lifecycle, but the event is executed on the main Thread, not the HTTP thread. This results in the Empty Stack leak. 
> ThreadLocals active during a HTTP request has to be cleaned up at the end of the Request. There are no guarantees that you will ever see that Thread again. CurrentInjectionPoint.cleanup should probably be moved to the requestDestroyed lifecycle.

-- 
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira