Rather than the strong-form semi-strong weak-form classification, I think it's more useful to think about two distinct notions.
First, is it possible to use information to generate excess returns by trading on the market? Second, are prices "correct" in the allocative sense?
On the first, the evidence is fairly supportive of market efficiency. The major chink in the armour of this proposition is the performance of some kinds of hedge funds. E.g. Renaissance is very very good and they are only working through mathematical models. If such evidence piles on, it would suggest there are small chinks in market efficiency of this sort.
The second distinct notion is "Are prices capable of going wrong in the sense of resource allocation"? It seems that the hypothesis is on weaker ground here.