[GHC] #10141: Kind inference regression in closed type families

[GHC]

#10141: Kind inference regression in closed type families
-------------------------------------+-------------------------------------
        Reporter:  goldfire          |                   Owner:  goldfire
            Type:  bug               |                  Status:  closed
        Priority:  highest           |               Milestone:  7.10.1
       Component:  Compiler          |                 Version:  7.10.1-rc2
      Resolution:  invalid           |                Keywords:
Operating System:  Unknown/Multiple  |            Architecture:
 Type of failure:  None/Unknown      |  Unknown/Multiple
      Blocked By:                    |               Test Case:
 Related Tickets:                    |                Blocking:
                                     |  Differential Revisions:
-------------------------------------+-------------------------------------
Changes (by goldfire):

 * status:  new => closed
 * resolution:   => invalid


Comment:

 This is by design. As I discovered en route to the solution to #9200, the
 old kind inference for closed type families was bogus. For the type family
 above, what should the return kind be? `k`, `k2`, and `*` are all viable
 options. Thus, this type family has no principal kind and should be
 rejected.

 Another was of looking at this is that `G` does not have a CUSK.
 Therefore, its right-hand side must use kind variables uniformly -- which
 it doesn't, because the first two equations must specialize `k` to `*`.

 So, closing as invalid.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10141#comment:2>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler