PAXEXAM-674 Configurations and probe builders do not need to be static

[vincent-zurczak]

By default, JUnit parameterized tests with PAX Exam need static methods for configuration and probe builders. Notice that this is not mentioned in the documentation. When these methods are not static, you get a NullPointerException, as mentioned in PAXEXAM-674.

This patch allows to use non-static methods for configurations and probe builders.
The test class is instantiated with every configuration. And every test class instance comes with its own OSGi configuration.

This was tested with parameterized tests on Apache Karaf 4.0.2.
This patch also comes with support of custom names for parameterized tests (the name attribute of the Parameters annotation).

I asked for some feedback on JIRA, but nothing yet.
I hope more will come with this PR.

[hwellmann]

Thanks for the patch - but we need some more info to apply it.

• Any fix for a bug should come with an integration test that fails without the fix and passes with the fix. I don't see any tests in this PR.
• Did you run the integration test suite locally?
• Please respect the coding and formatting style. There are Eclipse formatter definitions in https://github.com/ops4j/org.ops4j.pax.exam2/blob/master/assets/EclipseJavaFormatter.xml

[vincent-zurczak]

Hi. Thanks for the feedback.
I will add an integration test and format the code as expected.

[vincent-zurczak]

Hi. I have updated my patch.
First, having non-static methods for the PAX configuration only makes sense for the PerMethod strategy. I fixed it, although I took a strong hypothesis, that is to say I expect the test class to be annotated (default PAX settings are ignored). This could easily be fixed but I preferred limiting my modifications.

FYI, the use case I wanted to cover works perfectly (PerMethod strategy in Karaf, with parameterized configuration). This use case is illustrated with a new integration test. And I formatted the code with the right formatter.

However, I got some troubles with integration tests.
Those with Java EE never succeeded on my machine, before and after my modifications. And it seems my patch broke the Parameterized test. I get a NPE in the afterClass() method. There is probably something wrong with the run() method, but I could not determine what exactly. It is possible that I do not run the before and after class methods in all the staged reactors. To be honest, things are getting confused in my mind about this. If you have some clue or hint to give me, that would be helpful. 🆘

[vincent-zurczak]

OK. I thought about it recently, and I think I will submit a new runner for tests that use a parameterized OSGi configuration. It will only suit PerMethod strategies with a real container. It is much more simple than upgrading the existing Parameterized runner (which already covers several use cases).

I will do it as soon as a I have some time.