CLONE - Double-checked locking in org.jboss.weld.util.BeansClosure#getClosure is a cause for NPE ------------------------------------------------------------------------------------------------ Key: WELD-1087 URL: https://issues.jboss.org/browse/WELD-1087 Project: Weld Issue Type: Bug Affects Versions: 1.1.5.Final Environment: N/A Reporter: Anton Arhipov Assignee: Stuart Douglas Labels: cdi, jbossas_7.1, weld Application deployment fails with a following exception: 05:33:26,201 ERROR [org.jboss.msc.service.fail] (MSC service thread 1-7) MSC00001: Failed to start service jboss.deployment.unit."exp-cdi-injectApplication.war".WeldService: org.jboss.msc.service.StartException in service jboss.deployment.unit."exp-cdi-injectApplication.war".WeldService: java.lang.NullPointerException at org.jboss.as.weld.services.WeldService.start(WeldService.java:83) at org.jboss.msc.service.ServiceControllerImpl$StartTask.startService(ServiceControllerImpl.java:1811) [jboss-msc-1.0.2.GA.jar:1.0.2.GA] at org.jboss.msc.service.ServiceControllerImpl$StartTask.run(ServiceControllerImpl.java:1746) [jboss-msc-1.0.2.GA.jar:1.0.2.GA] at java.util.concurrent.ThreadPoolExecutor$Worker.runTask(ThreadPoolExecutor.java:886) [rt.jar:1.6.0_27] at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:908) [rt.jar:1.6.0_27] at java.lang.Thread.run(Thread.java:662) [rt.jar:1.6.0_27] Caused by: java.lang.NullPointerException at org.jboss.weld.bootstrap.BeanDeployment.createBeans(BeanDeployment.java:194) at org.jboss.weld.bootstrap.WeldBootstrap.deployBeans(WeldBootstrap.java:336) at org.jboss.as.weld.WeldContainer.start(WeldContainer.java:82) at org.jboss.as.weld.services.WeldService.start(WeldService.java:76) ... 5 more Looking at the code that belongs to weld-core-1.1.5.AS71.Final, we can see the following at the lines 193-194 of BeanDeployment: 193: BeansClosure closure = BeansClosure.getClosure(this.beanManager); 194: closure.addEnvironment(this.beanDeployer.getEnvironment()); Which leads to a conclusion that closure is null (the lines prior to the mentioned ones show that beanDeployer itself couldn't be null). Looking into BeansClosure#getClosure(): public static BeansClosure getClosure(BeanManagerImpl beanManager) { BeansClosure closure = (BeansClosure)closureMap.get(beanManager); if (closure == null) { synchronized (closureMap) { if (!closureMap.containsKey(beanManager)) { closure = new BeansClosure(); for (Iterable beanManagers : BeanManagers.getAccessibleClosure(beanManager)) { for (BeanManagerImpl accessibleBeanManager : beanManagers) { closureMap.put(accessibleBeanManager, closure); } } } } } return closure; } To prevent race conditions, maybe it makes sense to reside the full method body into synchronized block.