[jboss-dev] Integration of naming deployers failed

[Brian Stansberry]

At a minimum the smoke-tests target should be run before any non-trivial 
commit. That's why it exists!

On 01/12/2010 11:04 AM, Adrian Brock wrote:
> It didn't need a hudson run. The smoke tests show the problem.
> They take less than 5 mins to run before you get too itchy
> on the commit trigger. :-)
>
> $ cd testsuite/
> $ ./build.sh smoke-tests
>
> <snip/>
>
>      [junit] Running org.jboss.test.cts.test.CtsCmp2UnitTestCase
>      [junit] Tests run: 4, Failures: 0, Errors: 2, Time elapsed: 3.646
> sec
>      [junit] Test org.jboss.test.cts.test.CtsCmp2UnitTestCase FAILED
>
> <snip/>
>
>      [junit] Running org.jboss.test.web.test.WebIntegrationUnitTestCase
>      [junit] Tests run: 38, Failures: 0, Errors: 5, Time elapsed: 20.556
> sec
>      [junit] Test org.jboss.test.web.test.WebIntegrationUnitTestCase
> FAILED
>
>> From server.log:
>
> 2010-01-12 17:37:29,022 DEBUG
> [org.jboss.reloaded.naming.deployers.mc.MCJavaEEModule] (RMI TCP
> Connection(14)-127.0.0.1) Installed context org.jnp.interfaces.NamingCon
> text at 17d2f4 for JavaEE module cts-v1cmp, parentContext =
> org.jnp.interfaces.NamingContext at 1776efd
> 2010-01-12 17:37:29,039 ERROR
> [org.jboss.kernel.plugins.dependency.AbstractKernelController] (RMI TCP
> Connection(14)-127.0.0.1) Error installing to PreInstall: name=ja
> va:module state=Real: java.lang.IllegalStateException: Unable to
> register context name=java:module state=Real, name already exists:
> java:module
>          at
> org.jboss.dependency.plugins.AbstractController.registerControllerContext(AbstractController.java:1867)
>          at
> org.jboss.dependency.plugins.AbstractController.registerControllerContext(AbstractController.java:1760)
>          at
> org.jboss.dependency.plugins.ScopedController.addControllerContext(ScopedController.java:115)
>
> <snip/>
>
> 2010-01-12 17:40:45,350 DEBUG
> [org.jboss.deployers.structure.spi.helpers.AbstractDeploymentContext]
> (RMI TCP Connection(47)-127.0.0.1) Added component java:module to v
> fszip:/home/ejort/development/jboss-6.x/testsuite/output/lib/good-web.war/
> 2010-01-12 17:40:45,350 DEBUG
> [org.jboss.deployers.vfs.deployer.kernel.BeanMetaDataDeployer] (RMI TCP
> Connection(47)-127.0.0.1) Error during deploy: java:module: org.j
> boss.deployers.spi.DeploymentException: Error deploying: java:module
>          at
> org.jboss.deployers.spi.DeploymentException.rethrowAsDeploymentException(DeploymentException.java:49)
>          at
> org.jboss.deployers.vfs.deployer.kernel.BeanMetaDataDeployer.deploy(BeanMetaDataDeployer.java:125)
>          at
> org.jboss.deployers.vfs.deployer.kernel.BeanMetaDataDeployer.deploy(BeanMetaDataDeployer.java:51)
>          at
> org.jboss.deployers.spi.deployer.helpers.AbstractSimpleRealDeployer.internalDeploy(AbstractSimpleRealDeployer.java:62)
>
> On Tue, 2010-01-12 at 10:02 -0500, Bill Burke wrote:
>> Am I wrong in thinking that these naming commits were introduced to
>> trunk/ (because Hudson caught it)?  If so,  you really can't be
>> introducing errors like this.  Such fundamental changes need to go
>> through the automated test cycle before they are committed.  Otherwise
>> you're going to screw up everybody else that is working off of trunk.
>> Especially considering every "maven install" that is done pulls in new
>> snapshots.
>>
>> Ales Justin wrote:
>>> Yes, like I told you n-times, the api for scoped stuff is poor atm. ;-)
>>> To get around this, we used unique name and "real" alias via @Aliases
>>> annotation.
>>>
>>> Carlo de Wolf wrote:
>>>> Hmm, BeanMetaDataDeployer.undeploy doesn't take scoping into account and
>>>> will undeploy a 'random' context.
>>>>
>>>> Carlo
>>>>
>>>> On 01/12/2010 01:52 PM, Carlo de Wolf wrote:
>>>>> The initial runs on Hudson show no success. After a couple of
>>>>> deployments it goes poof with the following error:
>>>>>
>>


-- 
Brian Stansberry
Lead, AS Clustering
JBoss by Red Hat