Hi folks, In the process of upgrading from Java 1.6 to Java 1.7 (1.7.0_03 to be precise) we updated Drools from 5.1.1 to 5.3.1 (as in 5.1.1 language level 1.7 is not supported). However with Drools 5.3.1 the system property "drools.dialect.mvel.strict" seems not to be applied all the time. I know this sounds fishy, but this is the behaviour:

In a (junit based) test we create the knowledgebase by loading multiple drl files. Each of them has the dialect 'mvel' set in its beginning. In our @BeforeClass of the test we do a: System.setProperty("drools.dialect.mvel.strict", "false"); and I also tried using a KnowledgeBuilderConfiguration with the appropriate settings. But still sometimes the test fails with a "unable to resolve method using strict-mode..." message. I say sometime because running the test multiple times via maven "mvn -Dtest=ClassicDroolsTest test" sometimes produces failure, sometimes produces success. It might by noteworthy that the method called in the test expects two parameters of type Double, but as it seems Drools is not seeing the covariant return type of the class which is actually used (the abstract base method has return type Object, the concrete method has return type Double). I tried to reproduce the issue with a simpler test setup to be able to post it here but did not succeed. I also applied the patch https://github.com/droolsjbpm/drools/commit/1c20018abeef8bee86f896ffff794... (comming from https://issues.jboss.org/browse/JBRULES-3163) to the Drools 5.1.1 sources, compiled them myself and everything works like a charm with it. However I cannot use this unofficial work around as I can't (and don't want) use a patched version of Drools in my environment. Also when I compile& run my sources with jdk 1.6.0_31 everything is fine. I'd really appreciate any help on this topic as I couldn't find any hints on this behaviour in the release notes from Drools versions> 5.1.1 Thanks Joerg -- View this message in context: http://drools.46999.n3.nabble.com/mvel-strict-mode-not-applied-in-drools-... Sent from the Drools: User forum mailing list archive at Nabble.com. _______________________________________________ rules-users mailing list rules-users(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users

Hi Mark, Thanks for your reply, but the row in the drl the error occurs on is in an eval.what do you mean by explicit? anyway this does not explain why it runs on jdk1.6 but not on 1.7.

Best Jörg sent from a mobile device On Apr 4, 2012 12:55 PM, "Mark Proctor" <mproctor(a)codehaus.org <mailto:mproctor@codehaus.org>> wrote: On 04/04/2012 11:04, korgen wrote: > Hi folks, > In the process of upgrading from Java 1.6 to Java 1.7 (1.7.0_03 to be > precise) we updated Drools from 5.1.1 to 5.3.1 (as in 5.1.1 language level > 1.7 is not supported). > > However with Drools 5.3.1 the system property "drools.dialect.mvel.strict" > seems not to be applied all the time. I know this sounds fishy, but this is > the behaviour: MVEL strictness is only applied to explicite eval, consequence, return values and accumulates. It's not applied to field constaints. We foudn some points, around collections, where type safeness for constraints is not enforced and have strengthened that. You can still get type unsafe execution, by using the typesfalse(false), for the specific pattern. We don't like to encourage lack of type safety, as it creates difficult to maintain systems and will not benefit from performance enhancements we do on bytecode generation. Mark > > In a (junit based) test we create the knowledgebase by loading multiple drl > files. Each of them has the dialect 'mvel' set in its beginning. > In our @BeforeClass of the test we do a: > System.setProperty("drools.dialect.mvel.strict", "false"); > > and I also tried using a KnowledgeBuilderConfiguration with the appropriate > settings. > But still sometimes the test fails with a "unable to resolve method using > strict-mode..." message. I say sometime because running the test multiple > times via maven "mvn -Dtest=ClassicDroolsTest test" sometimes produces > failure, sometimes produces success. > It might by noteworthy that the method called in the test expects two > parameters of type Double, but as it seems Drools is not seeing the > covariant return type of the class which is actually used (the abstract base > method has return type Object, the concrete method has return type Double). > > I tried to reproduce the issue with a simpler test setup to be able to post > it here but did not succeed. > > I also applied the patch > https://github.com/droolsjbpm/drools/commit/1c20018abeef8bee86f896ffff794... > (comming from https://issues.jboss.org/browse/JBRULES-3163) to the Drools > 5.1.1 sources, compiled them myself and everything works like a charm with > it. However I cannot use this unofficial work around as I can't (and don't > want) use a patched version of Drools in my environment. > Also when I compile& run my sources with jdk 1.6.0_31 everything is fine. > > I'd really appreciate any help on this topic as I couldn't find any hints on > this behaviour in the release notes from Drools versions> 5.1.1 > > Thanks > > Joerg > > -- > View this message in context: http://drools.46999.n3.nabble.com/mvel-strict-mode-not-applied-in-drools-... > Sent from the Drools: User forum mailing list archive at Nabble.com. > _______________________________________________ > rules-users mailing list > rules-users(a)lists.jboss.org <mailto:rules-users@lists.jboss.org> > https://lists.jboss.org/mailman/listinfo/rules-users _______________________________________________ rules-users mailing list rules-users(a)lists.jboss.org <mailto:rules-users@lists.jboss.org> https://lists.jboss.org/mailman/listinfo/rules-users _______________________________________________ rules-users mailing list rules-users(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users

@Jörg: >> On 04/04/2012 11:04, korgen wrote: >> we updated Drools from 5.1.1 to 5.3.1 (as in 5.1.1 >> language level >> > 1.7 is not supported). Why would Drools 5.1.1. not be usable with Java 1.7? Byte code is supposed to be upward compatible.

-W _______________________________________________ rules-users mailing list rules-users(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/rules-users