[GHC] #14045: Data family instances must list all patterns of family, despite documentation's claims to the contrary

[GHC]

#14045: Data family instances must list all patterns of family, despite
documentation's claims to the contrary
-------------------------------------+-------------------------------------
        Reporter:  RyanGlScott       |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler (Type    |              Version:  8.3
  checker)                           |
      Resolution:                    |             Keywords:  TypeFamilies
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  GHC rejects       |            Test Case:  indexed-
  valid program                      |  types/should_compile/T14045,
                                     |  indexed-types/should_fail/T14045a
      Blocked By:                    |             Blocking:
 Related Tickets:  #12369            |  Differential Rev(s):  Phab:D3804
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by RyanGlScott):

 > Do you agree this is a bug?

 I certainly do. In the bowels of `TcDeriv`, we check that
 `newTyConEtadRhs` (the eta-reduced representation type) can accommodate
 the number of arguments from the instance we're deriving. But after doing
 some `pprTrace` scavenging, I discovered that the `newTyConEtadRhs` for
 `newtype instance T Int :: Type -> Type` is

 {{{#!hs
 ([a_a1vc], IO a_a1vc)
 }}}

 Whereas for `newtype instance T Int a :: Type`, it's:

 {{{#!hs
 ([], IO)
 }}}

 Something fishy is going on. The result of `newTyConEtadRhs` is ultimately
 being created in `mkNewTyConRhs`, which computes the eta-reduced type
 variables by checking one-by-one if a tycon tyvar matches the tyvar in the
 representation type in the corresponding position (see
 [https://git.haskell.org/ghc.git/blob/c13720c8c6047844f659ad4ce684946b80c99bee:/compiler/iface/BuildTyCl.hs#l89
 eta_reduce]).

 My guess is that this `a_a1vc` tyvar from the representation type doesn't
 match the corresponding tyvar from `newtype instance T Int :: Type ->
 Type` (what //would// the corresponding tyvar be, anyway?)

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14045#comment:14>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler