Jon Pretty

@JonathanChayat They may well be legitimate type errors, though! Maybe def foo[T, S <: R[_ >: T <: A]] would help in that case. Not sure. :/

@propensive First of all, Thanks! This is progress - however, now implicit search stops looking at companion objects for evidence.

@propensive I've revised my question on SO so that the evidence is in the companion of A (which is more realistic).

@propensive Do you understand why in def foo[T, TT >: T](x : T)(implicit ev: R[TT]) TT gets inferred the same as T?

@JonathanChayat I'd guess because there's a unique possible `ev`, and the type is inferred from the resolved implicit. Sound plausible?

@propensive I don't think so. It seems the type gets inferred before implicit search (making TT as specific as possible - meaning T)

@JonathanChayat Hmm... If TT isn't mentioned in the first parameter block, it should be completely unconstrained in the second block.

@propensive That is why it is so surprising...