There seems to be an issue with dueling byte-code instrumenting/weaving/enhancing between Emma and OpenJPA. I found this old issue for OpenJPA https://issues.apache.org/jira/browse/OPENJPA-1095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12991027and that points over to a SERP issue which is where they decided the problem was.
BTW ... I am using the terms "instrument", "enhance" and "weave" willy-nilly as they mean almost the same thing in this context. It is the process of modifying the byte code of a class after javac compiles and and before "java" executes it. Anyway ... the short form is that if OpenJPA attempts to enhance a class after Emma has instrumented it, an error message says "IllegalArgumentException: 5" with a stack trace.Then the tests fail when they run with a message saying they aren't enhanced. Because the OpenJPA enhance error message is so limited, nobody seems to know whether its just some classes or all of them that have the issue. All of the unit tests complain about the lack of enhancement. There are other comments here and there I found by Googling that seem to say there is a similar problem with Cobertura but it only happens depending on the order of the "instrumenting". I don't recall if it fails when Cobertura instruments first or if OpenJPA has to enhance first to have it fail. Emma instruments at "compile" time. Cobertura instruments at run time. My problems are, obviously, with running JUnit 3 tests inside Maven. I have read that they run twice--once without Emma instrumentation and once with. OpenJPA can enhance at compile time. There is a maven plugin to do that. Or, if not already enhanced, it will happen at run time, perhaps when the class loads. (It depends on the JDK version. I'm using JDK 5 or 6) Finally the question ... what would I have to do to make sure that OpenJPA enhances the classes before Emma instruments them. I don't know if this will make my tests run correctly but its the only thing I have come up with to try. It's not clear to me which phase of the lifecycle is the one in which the Emma instrumentation takes place. if I know that, I might be able to do the OpenJPA enhancing prior to that point. I'm also wondering about a typical way to do the OpenJPA enhancing after comiple and before running tests. This should make my tests run faster since there would be no enhancing going for every one of the hundreds of tests. I/m assuming if there are classes marked with @Entity they get enhanced if loaded even if no actual database activity goes on in the test. (This question came up as I thought about how Emma instruments one time and Coburtura instruments when every test runs so Emma runs tests faster.) If it doesn't work to reverse the enhancing order with Emma, my next plan is to see what I can do with Cobertura, though I have found Emma easier to use. Thanks -- Lee Meador Sent from gmail. My real email address is lee AT leemeador.com
