[GHC] #10789: Notify user when a kind mismatch holds up a type family reduction

[GHC]

#10789: Notify user when a kind mismatch holds up a type family reduction
-------------------------------------+-------------------------------------
        Reporter:  goldfire          |                Owner:  v0d1ch
            Type:  feature request   |               Status:  new
        Priority:  normal            |            Milestone:  8.8.1
       Component:  Compiler          |              Version:  7.10.2
      Resolution:                    |             Keywords:  newcomer,
                                     |  TypeFamilies
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #13192            |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 Sasa asks (on ghc-devs)

   With the help of @int_index I found the place in TcErrors.hs where the
 error printing occurs  and I think the check that I need to add will look
 similar to this one
 https://github.com/ghc/ghc/blob/master/compiler/typecheck/TcErrors.hs#L1935.

   So I guess that we need to check if one of the kinds of two types we are
 comparing defaults to * (or  Type if you will) and then add new warning
 that will be more descriptive as to why the failure happened. Maybe there
 is a way to check if what we are comparing are actually type families so
 that would make the job easier I guess.

 Richard replies

   I don't think the problem is particular to `Type` or defaulting.
 Instead, the problem is when
 * one of the two mismatched types is a type family application
 * and the type family has equations that pattern-match on an invisible
 parameter,
 * and it's that invisible-parameter matching that's gone awry.

   Now that I think about it, detecting these particular conditions might
 be tricky: you might need to edit code in `FamInstEnv` that does type
 family equation lookup to return diagnostic information if a match fails.
 (I would look at `reduceTyFamApp_maybe`, and perhaps it can return
 something more interesting than Nothing in the failure case.)

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10789#comment:18>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler