[GHC] #8851: Standalone deriving and GND behave differently

[GHC]

#8851: Standalone deriving and GND behave differently
--------------------------------------------+------------------------------
        Reporter:  simonpj                  |            Owner:  goldfire
            Type:  bug                      |           Status:  new
        Priority:  normal                   |        Milestone:
       Component:  Compiler (Type checker)  |          Version:  7.8.1-rc1
      Resolution:                           |         Keywords:
Operating System:  Unknown/Multiple         |     Architecture:
 Type of failure:  None/Unknown             |  Unknown/Multiple
       Test Case:                           |       Difficulty:  Unknown
        Blocking:                           |       Blocked By:
                                            |  Related Tickets:
--------------------------------------------+------------------------------

Comment (by simonpj):

 Aha. Right.  Suppose we have
 {{{
 class C r where
   op :: forall a. theta => op_ty

 newtype T a = MkT rep_ty deriving( C )
 }}}
 Then the both newtype deriving and standalone deriving should succeed if
 this will typecheck
 {{{
 instance C rep_ty => C (T a) where
   op = coerce (op :: op_ty[rep_ty/r]
 }}}
 When will this typecheck?  Notice that we instantiate `op` (like any other
 variable occurrence) at its occurrence site, which throws the instantiated
 `theta` into the constraints to be solved.

 Answer: when
  * `Coercible (op_ty[rep_ty/r]) (op_ty[N a/r])` holds
  * Assuming `C rep_ty` and `theta[T a/r]`, we can prove `theta[rep_ty/r]`

 In this case, `theta[rep_ty/r]` is simply `Monad IdrisInlineParser` which
 holds from a top-level instance, not from any of the local "assuming.."
 parts.

 So I think the `deriving` mechanism should check this second bullet.

 I believe that "standalone deriving" is deliberately
 [http://www.haskell.org/ghc/docs/latest/html/users_guide/deriving.html
 #stand-alone-deriving more gung-ho]: it simply generates the boilerplate
 and tries to typecheck it.  The `deriving` clause makes checks up-front so
 that the generated code is certain to typecheck.  We just need to fix
 those up-front checks.

 Thanks for looking at this Richard.

 Simon

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8851#comment:5>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler