Emmanuel Bernard, Thanks for the reply. I can understand the need to raise the awareness of the problem. But, to be honest, when this situation was first discovered by Hibernate BV, there should have been a bigger push back on Weld to either correct their implementation or to get the specs updated. I don't know exactly when this was first discovered and worked around, but now we're just proliferating the problem. I realize that we're probably only talking about two implementations for the Bean Validation spec, but we really shouldn't be requiring special code just to satisfy Weld's unique processing that is not spec compliant.

We are going to be requesting an update to Apache Bean Validation as well. But, just like this discussion, we're at the mercy of that development team. So, if Apache BV says "no, we're going to stay spec compliant" and not put special code in to support Weld, can we really blame them? And, since their main dance partner is OpenWebBeans, they don't really have the incentive to support this unique Weld scenario. But, we'll try to convince them that it's worthwhile to support the RI...

Once we have that JIRA logged, then do you feel that we have raised the awareness of this problem sufficiently well to exclude the testcase? Leaving the testcase has forced this conversation, I agree. But, it also prevents the successful passing of the TCK just because we're using an alternate spec-compliant implementation. Sounds kind of strange, doesn't it?