this is my last attempt to push CDI-494  forward. I'm not going to
describe the problem again, there is ML thread , issue comments ,
pull request comments, etc.
But, I've prepared a simple sample (arquillian test)  to:
1. Show the current behavior (Weld 1.1, Weld 2.3, OWB 1.2, OWB 1.6)
2. Demonstrate what exactly is not possible to write in a portable way today
Interesting note: it seems OWB 1.6 partialy supports the behaviour I
tend to prefer - i.e. a wildcard type is not considered an unresolvable
type variable (although it does not always infer the parameterized type
Feel free to comment the test, issue, etc. Any feedback is appreciated.
Red Hat, Czech Republic