On Wed, Mar 10, 2010 at 11:55 AM, Dmitri Dolguikh dmitri@redhat.com wrote:
My idea writing those java functional tests was to have tests setup the *complete* test environment. But if they don't catch all issues (what were the issues in jboss environment?) then they are certainly less useful than python-based ones.
Lots of deployment issues in both tomcat and jboss won't be caught by these, it also doesn't keep us properly in tune with exactly what's getting serialized in and out as the Java objects kind of hide it all. I've got much more confidence if we can point the suite at a particular deployment and see the results, so I'd argue the nosetests (or replacement) *must* exist, and thus it probably makes more sense to just do one functional test suite instead of two.
I don't have a strong preference on which framework to use, I only want tests to be easy and quick to setup (otherwise *less* tests will be written).
Me either, right now I understand the nosetests and I don't understand Cucumber, but if we do a POC and it looks easy to write and maintain then it's fine with me.
Devan