[gatein-issues] [JBoss JIRA] Resolved: (GTNWSRP-218) Force use of JBoss EL on AS 6

Force use of JBoss EL on AS 6 ----------------------------- Key: GTNWSRP-218 URL: https://issues.jboss.org/browse/GTNWSRP-218 Project: GateIn WSRP Issue Type: Sub-task Components: Admin GUI Affects Versions: 2.1.0-Beta01 Reporter: Chris Laprun Assignee: Chris Laprun Fix For: 2.1.0-Beta02 Attachments: JSFRIPortlet.war The EL evaluation engine seems to have changed between AS 5 and AS 6 and results in improper evaluations of expressions leading to calls being made on objects that are not ready to respond to them. More specifically line 77 in /jsf/consumers/consumer.xhtml: rendered=#{!cons.refreshNeeded and cons.active and cons.producerInfo.registrationRequired} Basically, it doesn't seem like the EL expression: !cons.refreshNeeded and cons.active and cons.producerInfo.registrationRequired is properly interpreted (or at the very least, not interpreted like it was on AS 5). When the exception which causes the portlet to fail is thrown, the state is so that the expression should not be completely evaluated: cons.refreshNeeded is true so !cons.refreshNeeded is false, the evaluation of the expression should stop because the result is already known to be false before looking at the rest of the expression (which is the Java behavior, foo && bar results in bar never being evaluated if foo is false). Moreover, cons.active is also false (even though it should never be evaluated). The exception is thrown because, despite the appropriate guards (the first two expressions !cons.refreshNeeded and cons.active), the EL engine still evaluates the last clause when the state of the underlying objects is not ready for such an evaluation as the root exception clearly states (java.lang.IllegalStateException: Registration status not yet known: call refresh first!). For me, this is a bug in the EL evaluation engine. At the very least, this is a huge change in behavior which should be documented somewhere i.e. the fact that you cannot rely on lazy evaluation of expressions in "and" clauses but this seems to be too big a behavioral change (which would also go against how Java works) for it to be proper behavior.