[GHC] #15757: Coercion Variable Almost Devoid Checking in ForAllCo

[GHC]

#15757: Coercion Variable Almost Devoid Checking in ForAllCo
-------------------------------------+-------------------------------------
           Reporter:  ningning       |             Owner:  (none)
               Type:  task           |            Status:  new
           Priority:  normal         |         Milestone:
          Component:  Compiler       |           Version:
           Keywords:                 |  Operating System:  Unknown/Multiple
       Architecture:                 |   Type of failure:  None/Unknown
  Unknown/Multiple                   |
          Test Case:                 |        Blocked By:
           Blocking:                 |   Related Tickets:  #15497
Differential Rev(s):                 |         Wiki Page:
-------------------------------------+-------------------------------------
 In
 [https://repository.brynmawr.edu/cgi/viewcontent.cgi?article=1074&context=compsci_pubs/
 Richard's thesis] Section 5.8.5.2, there is a restriction on where a
 coercion variable can appear in `ForAllCo` for the sake of consistency:
 the coercion variable can appear nowhere except in coherence coercions.

 Currently this restriction is missing in coercion quantification (see
 #15497). The goal of this ticket is to add the missing restriction.

 After discussion with Richard, we restate the restriction a little bit:
 the coercion variable can appear nowhere except in `GRefl` and `Refl`.
 Allowing it to appear in `Refl` should not break consistency.

 Notice that this also means `liftCoSubst` might fail when it lifts a
 `ForAllTy` over a coercion variable to a `ForAllCo`.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15757>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler