[GHC] #9858: Typeable instance for datatype and its promoted constructor is the same

[GHC]

#9858: Typeable instance for datatype and its promoted constructor is the same
-------------------------------------+-------------------------------------
        Reporter:  dreixel           |                   Owner:  dreixel
            Type:  bug               |                  Status:  new
        Priority:  highest           |               Milestone:  7.10.1
       Component:  Compiler          |                 Version:  7.9
      Resolution:                    |                Keywords:
Operating System:  Unknown/Multiple  |            Architecture:
 Type of failure:  None/Unknown      |  Unknown/Multiple
      Blocked By:                    |               Test Case:
 Related Tickets:                    |                Blocking:
                                     |  Differential Revisions:
-------------------------------------+-------------------------------------

Comment (by goldfire):

 I see two different issues at work here:

 1. If we have `data A = A`, then (typeRep (Proxy :: Proxy A)) and (typeRep
 (Proxy :: Proxy 'A)) are equal. This is not a problem with kinds, but a
 problem with namespaces. To me, the salient difference between `A` and
 `'A` is not that their kinds are different, but that they represent
 totally different entities. Fixing this issue is simple, either by adding
 some namespace information to `TyCon` or putting some marker in the
 `tyConName` string.

 2. If we have `data B x` (no constructors), then `B :: forall k. k -> *`.
 The types `(B :: Bool -> *)` and `(B :: Ordering -> *)` are actually, in
 Core, `B Bool` and `B Ordering` -- in Core, kind arguments are explicit.
 Thus, the `TypeRep` for `B :: Bool -> *` could feasibly be `TypeRep
 <fingerprint> <B's TyCon> [<Bool's TypeRep>]`. The change here isn't to
 the structure of `TyCon`, but instead mandating that all kind arguments
 are explicit in `TypeRep`s, just like they are in Core. Indeed, it
 surprises me that this isn't already the case! I believe that making kind
 arguments explicit in `TypeRep`s will be a simplification.

 Personally, it seems adding kind information to `TypeRep` is not the best
 approach here.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9858#comment:19>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler