[GHC] #11120: Missing type representations

[GHC]

#11120: Missing type representations
-------------------------------------+-------------------------------------
        Reporter:  goldfire          |                Owner:
            Type:  bug               |               Status:  patch
        Priority:  high              |            Milestone:  8.0.1
       Component:  Compiler          |              Version:  7.11
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):  Phab:D1769
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 We appear to be on to Phab:D1774 now, correct?  Change the Phab link?

 Like Richard I'm getting lost.  That Phab seems to be spiraling out of
 control.  Let's take one thing at at time.

 * Is `Proxy Char#` a valid type?  Not currently: we get
 {{{
 Illegal unlifted type: Char#
 }}}
   But there is no reason for this to be rejected!  `Proxy :: forall k. k
 ->*`, and I see no reason why we can't instantiate `k` with `TYPE
 Unlifted`.  Looking further, the error message comes not from the kind-
 unifier, but later in `TcValidity`; see `check_lifted`.  I think we can
 probably simply remove all calls to `check_lifted`.

 * Can you have `TypeRep Char#` or `typeRep @Char#`, or `Typeable Char#`.
 Again, you can't right now, but actually I think you could; they are all
 kind-polymorphic, and the instantiating kind could be `TYPE Unlifted`.

 * If we did, we might be faced with solving `Typeable Char#`. That would
 mean adding a type rep for `Char#`; but we could simply make it insoluble
 for now, and report a type error.

 * Could solving `Typeable Char#` come up in GHC today (i.e. without fixing
 `TcValidity`)?  Yes: consider `typeRep (undefined :: Proxy (Char# ->
 Int)`.  Nothing wrong with that on the face of it.  So we generate
 `Typeable (Char# -> Int)` constraint, and then decompose to `Typeable
 Char#` and `Typeable Int`.  And then we fail in solving `Typeable Char#`

 * What about ''promoted'' data constructors?
 {{{
 data CH = DCH Char#
 }}}
   Now suppose (comment:14) that we ask for `typeOf (Proxy :: Proxy 'DCH)`.
 This gives `Typeable 'DCH`.  At the moment I can't see why that would need
 a type-rep for `Char#` as comment:14 suggests.   Any ideas?

 So my proposal is:
 * Let's NOT (yet) have type-reps for primitive types
 * Let's fix the Typeable solver so that it fails gracefully on `Typeable
 Char#`.
 * And let's try the effect of removing the `check_unlifted` in
 `TcValidity`.

 Would that be a good start?

 That still leaves open the questions in comment:3.   But first things
 first.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11120#comment:19>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler