[GHC] #13766: Confusing "redundant pattern match" in 8.0, no warning at all in 8.2

[GHC]

#13766: Confusing "redundant pattern match" in 8.0, no warning at all in 8.2
-------------------------------------+-------------------------------------
        Reporter:  edsko             |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  8.0.1
      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:                    |
-------------------------------------+-------------------------------------
Changes (by RyanGlScott):

 * cc: mpickering (added)


Comment:

 Hoo boy, that's pretty scary. To make it worse, because of this bug you
 can write blatantly nonsensical things like this:

 {{{#!hs
 {-# LANGUAGE FlexibleInstances #-}
 {-# LANGUAGE TypeFamilies #-}
 {-# OPTIONS_GHC -Wall -Woverlapping-patterns #-}

 module T where

 class C a where
   c :: a

 instance (a ~ Bool, a ~ Int) => C a where
   c = 42 + True
 }}}

 And it'll compile on GHC 8.2.1-rc2 with no warnings. (I haven't found a
 way to actually //use// that bogus `c` definition, but still.)

 The commit that caused this behavior is
 adb565aa74582969bbcc3b411d6d518b1c76c3cf (Don't return empty initial
 uncovered set for an unsat context). Matthew, do you know what is going on
 here?

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/13766#comment:3>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler