[GHC] #15558: Locally handling an empty constraint

[GHC]

#15558: Locally handling an empty constraint
-------------------------------------+-------------------------------------
        Reporter:  Ericson2314       |                Owner:  (none)
            Type:  task              |               Status:  closed
        Priority:  normal            |            Milestone:  ⊥
       Component:  Compiler          |              Version:  8.0.1
      Resolution:  fixed             |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Incorrect         |  Unknown/Multiple
  error/warning at compile-time      |            Test Case:  gadt/T15558
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by goldfire):

 I think we need an `impossible` that has a type like `Absurdity => a`,
 where `Absurdity` is a special constraint that the type checker can
 satisfy only when an absurdity is in the context (that is, in inaccessible
 code). If an absurdity is used to prove `Absurdity`, then the
 "inaccessible code" warning is squelched. This would need custom support
 for the `Absurdity` prover, but I don't actually think it would be all
 that hard.

 I'm not sure what the Core evidence would look like. Perhaps we need
 `contra`, tucked without comment in the appendix of the
 [https://repository.brynmawr.edu/cgi/viewcontent.cgi?referer=&httpsredir=1&article=1010&context=compsci_pubs
 Coercible paper]. The `contra` there is to allow us to assert that `case`
 matches are total. GHC's Core, though, doesn't require `case` to be total,
 meaning that this has never been implemented.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15558#comment:14>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler