[GHC] #15370: Typed hole panic on GHC 8.6 (tcTyVarDetails)

[GHC]

#15370: Typed hole panic on GHC 8.6 (tcTyVarDetails)
-------------------------------------+-------------------------------------
        Reporter:  RyanGlScott       |                Owner:  Tritlo
            Type:  bug               |               Status:  new
        Priority:  high              |            Milestone:  8.6.1
       Component:  Compiler (Type    |              Version:  8.4.3
  checker)                           |             Keywords:  TypeInType,
      Resolution:                    |  TypedHoles
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Compile-time      |  Unknown/Multiple
  crash or panic                     |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonpj):

 I have a hypothesis about the assertion failure in comment:4

 I see in `TcHoleErrors.tcCheckHoleFit`
 {{{
      ; let w_rel_cts = addSimples wanted relevantCts
 ...
      ; let w_givens = foldr setWC w_rel_cts outermost_first
 ...
      ; rem <- runTcSDeriveds $ simpl_top w_givens
 }}}
 So the same `relevantCts` are used in multiple runs of `runTcSDerived`.
 You are careful to freshen the `ic_binds` of the implicatations,
 but if `relevantCts` contains any equalities they will have `HoleDest`,
 a coercion hole that is filled in directly.  You really need to clone
 those holes too.  Use `TcMType.cloneWanted` as in `cloneWC`.

 As it happens, this `cloneWanted` is overkill, and I'm improving it
 in an upcoming patch, but for now I think it'll do what you need.

 Does that make sense?  Best to verify that this is indeed what is
 happening.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15370#comment:10>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler