Role of static should be nominal

[edsko]

To see why, consider

    ``` haskell
    {-# OPTIONS_GHC -Wall #-}
    module Main (main) where

    import Control.Distributed.Static
    import Data.Rank1Dynamic
    import Data.Rank1Typeable
    import GHC.Prim (coerce)
    import GHC.Types (Coercible)
    import Unsafe.Coerce (unsafeCoerce)

    newtype T = T Bool
      deriving (Show)

    remoteTable :: RemoteTable
    remoteTable = registerStatic "$true" (toDynamic True)
                . registerStatic "$unsafeCoerce" (toDynamic (unsafeCoerce :: ANY1 -> ANY2))
                $ initRemoteTable
    staticUnsafeCoerce :: Static (a -> b)
    staticUnsafeCoerce = staticLabel "$unsafeCoerce"

    staticTrue :: Static Bool
    staticTrue = staticLabel "$true"

    staticCoerce :: Coercible a b => Static a -> Static b
    staticCoerce = staticApply staticUnsafeCoerce

    main :: IO ()
    main = do
        let a :: Either String Bool
            a = unstatic remoteTable staticTrue
        print a

        let b :: Either String T
            b = unstatic remoteTable (coerce staticTrue)
        print b

        let c :: Either String T
            c = unstatic remoteTable (staticCoerce staticTrue)
        print c

        putStrLn "done"

The definition of b will result in "Cannot unify Bool and T". For completeness, this example also shows how one might implement staticCoerce if one wished to do so; all this PR does is add a role annotation to static to rule out using coerce.

[mboes]

👍

[edsko]

I've rebased this on the latest master. @mboes , was there a reason why this wasn't merged?

[facundominguez]

I'm guessing nobody cared to press the merge button :)

[edsko]

Fair enough :) Thanks for merging :)