[GHC] #15135: Overlapping typeclass instance selection depends on the optimisation level

[GHC]

#15135: Overlapping typeclass instance selection depends on the optimisation level
-------------------------------------+-------------------------------------
        Reporter:  nicuveo           |                Owner:  (none)
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:  8.6.1
       Component:  Compiler          |              Version:  8.4.2
      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):

 You are on thin ice here.   In `getInt`:
 {{{
 getInt :: A a => a -> Int
 getInt x = fromMaybe 0 $ someValue x
 }}}
 we have a "wanted" constraint `A a`. How can we solve it?  In two ways:
 * From the instance
 * From the `A a =>` given to `getInt`.

 GHC chooses to solve it from the instance.  But then you later overlap the
 instance, so that decision was arguably wrong.  But the same thing would
 happen if you tried to infer a type for `getInt`:
 {{{
 getInt x = fromMaybe 0 $ someValue x
 }}}
 Again, GHC will use the instance and infer
 {{{
 getInt :: a -> Int
 }}}
 If you want to signal to GHC that the instance might be overlapped, use
 `{-# OVERLAPPABLE #-}`; and then you'll always get 42.


 I think it's arguable that an instance should only be overlappable if it
 says `{-# OVERLAPPABLE #-}`.  But that's not our current spec.

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/15135#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler