[GHC] #14154: Some cocktail of features causes GHC panic

[GHC]

#14154: Some cocktail of features causes GHC panic
-------------------------------------+-------------------------------------
        Reporter:  Iceland_jack      |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.3
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
                                     |  Unknown/Multiple
 Type of failure:  None/Unknown      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 >  Could we fix one case this with a well-placed zonk?

 I suppose we could always zonk before calling `typeKind`.  But I am deeply
 averse to doing this.

 * The constraint solver does zonking on the fly, as part of
 canonicalisation.  That is simple and nice.  I don't think we should
 ''ever'' call zonking when we are solving constraints.

 * I think the constraint solver indeed obeys the invariant that we never
 constructor an ill-kinded type, provided you ''don't'' zonk it!  But, as
 remarked above, `tcDataConPat` does form an ill-kinded type.  I think
 right solution is not do to so, which isn't difficult.

 What I'm upset about is that it's hard to be sure when we have nailed
 every case.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14154#comment:8>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler