2009/7/13 Adam Murdoch <[email protected]> > > > Tom Eyckmans wrote: > >> Hi all, >> >> The behavior described in GRADLE-538 occures when there are no test >> classes, in this case both JUnit / TestNG tasks try to execute all classes >> as tests. >> >> In order to solve this I'd like to skip test execution (calling of the >> JUnit / TestNG Ant tasks). This results in an empty JUnit report and no >> report in case of TestNG. The TestNG Ant task doesn't allow it to be run for >> no test classes, the Ant JUnit task does. So I'd like to add a >> GradleException when there are no test classes with a boolean to disable the >> exception (failWhenNoTests) default true. >> > > The general approach of most of our tasks is to be forgiving of things like > this. This approach leads to a more flexible system because the tasks make > less assumptions about the world, they just deal with the world as they find > it.
Ok, that makes sense. > Why should the test task care that there are no executable test classes in > the test classes dir? Why is this a failure from it's point of view? The > task should deal happily with either case, I think. > > One case for an exception like this is that it performs a kind of > verification that the world matches the model. However, this kind of > checking belongs with the thing that owns and understands the model, that > is, the Java plugin, not the executable steps, like the test task. > > >> Currently when there are no test classes at all the build is ok in case of >> JUnit but in case of TestNG it fails, this is not really consistent, so we >> might want to straighten this out. >> >> This may cause a lot off builds to fail so I'd like to hear your optinions >> on this matter - to fail or not? >> > > Not > After reading this I really don't know why I wanted to make it a blocker luckily I'm aware of the more minds approach :) > > > Adam > > > --------------------------------------------------------------------- > To unsubscribe from this list, please visit: > > http://xircles.codehaus.org/manage_email > > >
