On Wed, Jul 28, 2010 at 7:55 AM, Eric Covener <cove...@gmail.com> wrote: > On Wed, Jul 28, 2010 at 1:40 AM, Gurkan Erdogdu <gurkanerdo...@yahoo.com> > wrote: >>>>>Good catch Eric! >> Not mean that current logic is wrong. >> >> Bean API types are local interfaces of the EJB bean . We check all >> injection >> fields at deployment time for validation. If there is no validation error, >> using >> of EJB local interfaces are correct. >> >> >> What is the problem that you think about? > > My concern is that at runtime, each time we perform injection various > threads are poking around at BaseEjbBean.iface when really all they > need to is use the iface on the stack to create their > proxies/instances. This is because there is not actually one iface > per BaseEjbBean (EJB class), just one per injection point. >
Although now I'm not sure if we can properly know the injected iface when it's time to get the instance, since we are funneled through the Contextual.getInstance() interface. But stashing away the injected iface in the BaseEjbBean in between proxy creation and obtaining the instance does not seem threadsafe. -- Eric Covener cove...@gmail.com
