Andres Löh wrote:
Yes, because these are superclasses of Unbox. So can't you simply say this:
newtype Foo = Foo Int deriving (Eq, Show, Unbox, M.MVector MVector, G.Vector Vector)
I don't think this should work. This is the basic set up:
[...]
I don't understand what GeneralizedNewtypeDeriving does with deriving(G.Vector Vector) in this situation but it can't possibly do anything sensible without a data instance Vector Foo. The fact that it doesn't fail is just a bug, IMO.
I've actually started wondering the same after I posted this. I tried to use -ddump-deriv to get at the code, but GHC seems to be cheating and doesn't show what it generates. I guess I understand newtype-deriving less than I thought ...
I don't really understand it, either, and try to stay well away from it :-) It simply doesn't seem to work with GADTs, type families etc. I've created a ticket: http://hackage.haskell.org/trac/ghc/ticket/6147. Roman