I agree with Dan that depending on a SNAPSHOT version is probably not the
best, especially when it is a major part running our tests. This could have
problems when no developers change anything but JUnit releases a new
snapshot and the nightly build uses the newest release resulting in new
failures.

I would be in favor of a workaround, if we have consensus, until the actual
4.13 release.

--Mark

On Tue, May 10, 2016 at 4:26 PM, Dan Smith <[email protected]> wrote:

> Looking at the fix
> <
> https://github.com/junit-team/junit4/commit/37a8aaba8a817b4bebbeebcd645d304601a0c8f0
> >
> a little bit, I think we can work around this with the current version of
> junit with something like this. We just need to create our own runner
> factory that will return the correct annotations.
>
> @Category(UnitTest.class)
> @RunWith(Parameterized.class)
> @UseParametersRunnerFactory(MyRunnerFactory.class)
> public class YourTest {...}
>
>
> public class MyRunnerFactory implements ParametersRunnerFactory {
>
>     @Override public Runner createRunnerForTestWithParameters(final
> TestWithParameters test)
>       throws InitializationError
>     {
>       return new BlockJUnit4ClassRunnerWithParameters(test) {
>         @Override protected Annotation[] getRunnerAnnotations() {
>           return getTestClass().getAnnotations();
>         }
>       };
>     }
> }
>
> -Dan
>
>
>
> On Tue, May 10, 2016 at 3:58 PM, Dan Smith <[email protected]> wrote:
>
> > If we switch to a snapshot release, that means the version of junit will
> > be a moving target, right? So if someone checks a breaking change into
> > junit all of our builds will fail. Or worse, not fail. I'm not sure if
> > that's state we want to be in.
> >
> > -Dan
> >
> > On Tue, May 10, 2016 at 2:54 PM, Jens Deppe <[email protected]>
> wrote:
> >
> >> Kirk recently found that test classes annotated with both @Category and
> >> @Parameterized (both standard JUnit annotations) result in tests being
> >> skipped [1]
> >>
> >> It appears that JUnit 4.13-SNAPSHOT fixes the issue.
> >>
> >> Does anyone have any aversion to using a 'SNAPSHOT' release for junit?
> >>
> >> An alternative is to use JunitParamsRunner [2]. However I've already run
> >> into a problem with it in that params are not exposed to @Before methods
> >> which means that any @Before code, using parameters, would need to be
> >> explicitly called from every test method.
> >>
> >> Thoughts? Comments?
> >>
> >> --Jens
> >>
> >>
> >> [1] https://issues.apache.org/jira/browse/GEODE-1350
> >> [2] http://pragmatists.github.io/JUnitParams/
> >>
> >
> >
>

Reply via email to