Re: [GHC] #12881: GHC 8.0.2 regression involving OVERLAP annotations

28 Nov 2016


      #12881: GHC 8.0.2 regression involving OVERLAP annotations
-------------------------------------+-------------------------------------
        Reporter:  RyanGlScott       |                Owner:
            Type:  bug               |               Status:  new
        Priority:  high              |            Milestone:  8.0.2
       Component:  Compiler          |              Version:  8.0.2-rc1
      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:                    |
-------------------------------------+-------------------------------------

@@ -16,15 +16,0 @@
- {{{
- $ /opt/ghc/8.0.2/bin/ghc Bug.hs
- [1 of 1] Compiling Bug              ( Bug.hs, Bug.o )
-
- Bug.hs:9:10: error:
-     • Overlapping instances for Arbitrary Int
-         arising from a use of ‘Bug.$dmshrink’
-       Matching instances:
-         instance Arbitrary a -- Defined at Bug.hs:8:10
-         instance Arbitrary Int -- Defined at Bug.hs:9:10
-     • In the expression: Bug.$dmshrink @Int
-       In an equation for ‘shrink’: shrink = Bug.$dmshrink @Int
-       In the instance declaration for ‘Arbitrary Int’
- }}}
-

New description:

 `quickcheck-combinators-0.0.1` fails to build with GHC 8.0.2-rc1 (but does
 build with GHC 8.0.1) due to this issue. Here is a simplified example:

 {{{#!hs
 {-# LANGUAGE FlexibleInstances #-}
 module Bug where

 class Arbitrary a where
   shrink :: a -> [a]
   shrink _ = []

 instance Arbitrary a
 instance Arbitrary Int
 }}}

 Is this expected? If so, we should make a note of this in the 8.0.2
 release notes.

--

Comment (by simonpj):

 Well there really is an issue here.  Suppose you added to the program in
 the Description
 {{{
 foo :: [Int] -> Int
 foo x = shrink x
 }}}
 then you'd ''expect'' to get the warning
 {{{
 T12881.hs:15:9: error:
     • Overlapping instances for Arbitrary Int
         arising from a use of ‘shrink’
       Matching instances:
         instance Arbitrary a -- Defined at T12881.hs:8:10
         instance Arbitrary Int -- Defined at T12881.hs:10:10
     • In the expression: shrink x
       In an equation for ‘foo’: foo x = shrink x
 }}}
 So, without the pragmas, or the global `-XOverlappingInstances`, the
 instance declarations for `Arbitrary` are effectively useless.

 Now, it's true that 8.0.2 is reporting that problem a bit more eagerly
 than before.  I suppose we could switch on `-XOverlappingInstances` when
 typechecking the default methods in each instance declaration.  But that
 feels like sweeping the real problem under the carpet: those instances
 really are useless unless you allow overlapping.  So the right solution is
 to add those pragmas.

 We could mitigate for 8.0.3 (as suggested in the previous para) but I
 worry that the same thing would happen in 8.2.

 Advice?

--
Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/12881#comment:6
GHC http://www.haskell.org/ghc/
The Glasgow Haskell Compiler

    