Hi John, That's pretty interesting, the issue caused by the fact that in your test the Application class instances are Weld proxies, while in orginal test case they aren't. As such, the @ApplicationPath annotations happen to be lost for Weld proxies, which causes address conflicts for JAXRSServerFactoryBeans. I will try to figure out what are the differences ...
Thanks! Best Regards, Andriy Redko JDA> Hi Andriy, JDA> Actually I had come to that conclusion a couple of days ago. So I have to JDA> wonder, was the test ever valid? I know that JAX-RS 2 specified that it was JDA> OK to have multiple application impls per deployment, so I figure that JDA> should work. JDA> And if this is the issue, how come it was fine with how CXF built it before? JDA> John JDA> On Sun, Aug 7, 2016 at 2:44 PM Andriy Redko <drr...@gmail.com> wrote: >> Hey John, >> The exception in question comes from the fact that your deployment >> descriptor >> contains 2 applications, BookStoreApplication and >> BookStoreCustomApplication: >> return ShrinkWrap.create(WebArchive.class) >> .addClasses(BookStoreApplication.class, >> BookStoreCustomApplication.class, >> BookStoreService.class, BookStore.class) >> It is surely possible to have more than one JAX-RS application but in CDI >> context they >> both try to bind to root path "/" (@ApplicationPath is taken into account >> but later). In this >> case both JAXRSServerFactoryBean instances try to registered themselved >> under "/" and exception >> is being raised :( >> Thanks. >> Best Regards, >> Andriy Redko >> JDA> So I'm running into an interesting issue. I'm not sure if I found a >> bug, >> JDA> or if I've gone way far into the deep end here. >> JDA> I'm getting this exception: https://paste.apache.org/8WEp >> JDA> You can see my changes here: >> JDA> https://github.com/johnament/cxf/tree/arquillian >> JDA> I suspect what's happening is that CXF is being initialized twice. >> Now >> JDA> that I've enabled CDI, the static initialization and the CDI >> initialization >> JDA> are both being triggered. Not sure if you have any ideas. >> JDA> John >> JDA> On Tue, Aug 2, 2016 at 9:46 PM John D. Ament <johndam...@apache.org> >> wrote: >> >> Hi Andriy, >> >> Thanks for the prompt reply. >> >> On Tue, Aug 2, 2016 at 9:28 PM Andriy Redko <drr...@gmail.com> wrote: >> >>> Hi John, >> >>> Thanks a lot for your fixes, much appreciated and of great value for >> CDI >> >>> users. To answer a couple of your questions / concerns. >> >>> *JDA> First, its assuming that Weld is the only testable container. >> *This >> >>> is very true. The reason for that is Weld was the only one implementing >> >>> CDI 1.2 at the moment the CXF/CDI integration was done. OpenWebBeans >> were >> >>> behind but there is no obstacles or objects to have a test suite(s) >> >>> against >> >>> it as well, it's been a while OpenWebBeans implements 1.2. >> >> I just realized that the CDI integration has been around for all of the >> >> 3.x line. I only found out about it in the last couple of weeks. >> >> With that said, I'd like to try to put together a test suite. I'll send >> >> it over when ready. If you guys like it, its yours. >> >> I created https://issues.apache.org/jira/browse/CXF-6988 >> >>> *JDA> Second, its always doing classpath scanning. *This is also >> true, >> >>> as there was an intention to test exactly the way it is >> >>> going to be used. The suite also tests against Tomcat and Jetty, >> embedded >> >>> and >> >>> WAR based deployments. With that being said, if Arquillian allows to >> >>> simplify >> >>> the test structure while opening more opportuties to test different >> >>> scenarios >> >>> (including the ones we already have), it would be great in my opinion. >> >> Sounds good. Some of the things I'd like to make sure are proved out: >> >> - Testing both Weld and OWB >> >> - Running different parts of the test per deployment. >> >> - Ensuring the various tests with Jetty/Jetty Embedded/Tomcat >> >>> Thanks. >> >>> Best Regards, >> >>> Andriy Redko >> >>> *JDA> Hi, JDA> Long time user, first time contributor to CXF. Though >> I'm >> >>> no stranger to JDA> the ASF by any long shot. JDA> I was looking at >> putting >> >>> in some fixes for issues I reported. First one JDA> was a non-problem. >> >>> However, when trying to figure out how to add tests to JDA> ensure that >> >>> empty application class applications work fine (CXF-6986), I JDA> >> realized >> >>> that the current testing structure in systest wouldn't work. *JDA> >> >>> >> https://github.com/apache/cxf/blob/master/systests/cdi/src/test/java/org/apache/cxf/systest/jaxrs/cdi/AbstractCDITest.java >> >>> < >> https://github.com/apache/cxf/blob/master/systests/cdi/src/test/java/org/apache/cxf/systest/jaxrs/cdi/AbstractCDITest.java >> > >> >>> *JDA> It looks like this test code is doing a few odd things. First, >> its >> >>> JDA> assuming that Weld is the only testable container. The ASF >> actually >> >>> hosts JDA> the other CDI impl, OpenWebBeans. Second, its always doing >> >>> classpath JDA> scanning. This strategy would mean I need a separate >> module >> >>> to test my JDA> feature, which is a little odd. JDA> I was wondering if >> >>> there was any interest in converting this to an JDA> arquillian based >> >>> test? The test code could be platform inspecific, JDA> allowing tests >> to >> >>> be created for both CDI impls, improving compatibility. JDA> WDYT? >> JDA> in >> >>> case it helps understand the problem, my proposed changes can be seen >> JDA> >> >>> here: *https://github.com/apache/cxf/pull/150 >> >>> *JDA> - John *