[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:  new
        Priority:  normal            |               Milestone:  7.12.1
       Component:  Compiler          |                 Version:  7.10.1-rc2
      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've added a (not-yet-pushed) test case, but it's hard to see how to
 improve error messages. The problem is that the place that generates the
 kind errors is far from where we know whether or not we have a CUSK. I
 attempted to use `addErrCtxt` to include a note about CUSKs before kind-
 checking, but this note got appended to lots of errors that are clearly
 unrelated to CUSKs -- this is the wrong approach. I think the right
 approach is to somehow record in `TcM` that we're kind-checking a
 declaration which could potentially have a CUSK but in fact does not and
 then look there before reporting certain errors. But this is terribly
 heavy.

 Perhaps after my branch is merged, a way forward will present itself.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10141#comment:9>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler