-- When dealing with the deriving clause
-- co1 : N [(b,b)] ~ R1:N (b,b)
-- co2 : R1:N (b,b) ~ Tree (b,b)
+-- co : N [(b,b)] ~ Tree (b,b)
genDerivBinds :: SrcSpan -> FixityEnv -> Class -> TyCon -> (LHsBinds RdrName, DerivAuxBinds)
genDerivBinds loc fix_env clas tycon
-- see Note [Newtype deriving superclasses] in TcDeriv.lhs
tc_inst_decl2 dfun_id (NewTypeDerived coi)
- = do { let rigid_info = InstSkol
- origin = SigOrigin rigid_info
- inst_ty = idType dfun_id
+ = do { let rigid_info = InstSkol
+ origin = SigOrigin rigid_info
+ inst_ty = idType dfun_id
+ inst_tvs = fst (tcSplitForAllTys inst_ty)
; (inst_tvs', theta, inst_head_ty) <- tcSkolSigType rigid_info inst_ty
-- inst_head_ty is a PredType
(rep_ty, wrapper)
= case coi of
IdCo -> (last_ty, idHsWrapper)
- ACo co -> (snd (coercionKind co), WpCast (mk_full_coercion co))
+ ACo co -> (snd (coercionKind co'), WpCast (mk_full_coercion co'))
+ where
+ co' = substTyWith inst_tvs (mkTyVarTys inst_tvs') co
+ -- NB: the free variable of coi are bound by the
+ -- universally quantified variables of the dfun_id
+ -- This is weird, and maybe we should make NewTypeDerived
+ -- carry a type-variable list too; but it works fine
-----------------------
-- mk_full_coercion
this_dict dfun_id
prag_fn monobinds
; (meth_exprs, meth_binds) <- tcExtendTyVarEnv inst_tyvars' $
- mapAndUnzipM tc_meth op_items
+ mapAndUnzipM tc_meth op_items
-- Figure out bindings for the superclass context
-- Don't include this_dict in the 'givens', else