Is it intentional that forward references in unittests aren't

[bearophile]

retard:
> What does this mean? Can't you write a wrapper struct/class inside the 
> block if you don't want to expose S and S2 outside the unittest block.

Seen from a high level a unittest section != function. But in D they are implemented as a kind-of-function anyway. So unittests in D2 can show some unwanted characteristics that are present in functions. This means that they are an abstraction that "leaks" (a bit):
http://en.wikipedia.org/wiki/Leaky_abstraction
But I have not criticized this design choice because overall the alternatives can be worse.

Bye,
bearophile