[weld-issues] [JBoss JIRA] Updated: (WELD-712) Regression in Extension proxy method invocations

Sivakumar Thyagarajan (JIRA) jira-events at lists.jboss.org
Thu Oct 7 05:50:39 EDT 2010


     [ https://jira.jboss.org/browse/WELD-712?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sivakumar Thyagarajan updated WELD-712:
---------------------------------------

    Attachment: relevant-jersey-sources.tar.bz2


Relevant jersey source files attached for convenience

> Regression in Extension proxy method invocations
> ------------------------------------------------
>
>                 Key: WELD-712
>                 URL: https://jira.jboss.org/browse/WELD-712
>             Project: Weld
>          Issue Type: Bug
>          Components: Proxies, Reflection layer
>    Affects Versions: 1.1.0.Beta1
>         Environment: Latest GlassFish that has 1.1.0-01-glassfish build of Weld integrated. To reproduce the problem the following system property needs to be set to bypass the Jersey workaround for this issue.
>         <jvm-options>-Dcom.sun.jersey.server.impl.cdi.lookupExtensionInBeanManager=true</jvm-options>
>            Reporter: Sivakumar Thyagarajan
>             Fix For: 1.1.0.Beta2
>
>         Attachments: relevant-jersey-sources.tar.bz2
>
>
> Glassfish issue 13131 [1],  involving the Jersey portable extension, notes that there has been a regression in the last integration of Weld (1.1.0-01-glassfish).
> It appears that invocations on observer methods of container lifecycle events are handled in the proxy and delegated to the proxied instance of the Extension, whereas invocations of non-observer methods on a Extension that is obtained from BeanManager, are directed towards the non-proxied instance of the Extension. This behaviour appears to have been changed after Weld 1.0
> More details:
> The Jersey extension has a private initialization method (initialize() [2]) that inits the state of the extension during beforeBeanDiscovery. I notice that there are two instances of the Extension being created (because the extension is being proxied) -- let us call it actual_instance1 and proxy_instance_2.
> beforeBeanDiscovery observer method is called on proxy_instance_2 and this method call in turn delegates to actual_instance1 and initialize() method is called on actual_instance1. So now actual_instance1's state is initialized correctly.
> However later when the Extension is looked up through the BeanManager elsewhere [3], the proxy_instance_2 is returned. When a non-observer method (lateInitialize [4]) is called on this instance proxy_instance_2, this method call goes to proxy_instance_2 and is not forwarded to actual_instance_1. Since the state is not initialized in proxy_instance_2, this call fails.
> Please see the attached text file for stack traces related to the various stages discussed above and also contact me should you need to reproduce this.
> [1] http://glassfish.dev.java.net/issues/show_bug.cgi?id=13131
> [2] intialize method in https://jersey.dev.java.net/svn/jersey/trunk/jersey/jersey-server/src/main/java/com/sun/jersey/server/impl/cdi/CDIExtension.java
> [3] constructor in CDIComponentProviderFactory looks up the Extension using the BeanManager https://jersey.dev.java.net/svn/jersey/trunk/jersey/jersey-server/src/main/java/com/sun/jersey/server/impl/cdi/CDIComponentProviderFactory.java
> [4] lateInitialize in https://jersey.dev.java.net/svn/jersey/trunk/jersey/jersey-server/src/main/java/com/sun/jersey/server/impl/cdi/CDIExtension.java 

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira

        


More information about the weld-issues mailing list