[GHC] #11370: Redundant superclass warnings being included in -Wall destroys the "3 Release Policy"

[GHC]

#11370: Redundant superclass warnings being included in -Wall destroys the "3
Release Policy"
-------------------------------------+-------------------------------------
        Reporter:  ekmett            |                Owner:
            Type:  bug               |               Status:  new
        Priority:  highest           |            Milestone:  8.0.1
       Component:  Compiler (Type    |              Version:  7.10.3
  checker)                           |
      Resolution:                    |             Keywords:
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  Incorrect         |  Unknown/Multiple
  warning at compile-time            |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:  #11369, #11429    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by ekmett):

 > Does this sentence have an implicit “without losing 3-release
 compatibility”?

 No.

 I mean explicitly that there is code that I have in the wild that can only
 work around this by turning off the warning. If you write any class that
 only supplies a law then it will always appear to be a "redundant
 constraint". GHC doesn't know about the laws.

 Examples from real math: Upgrading a braided monoidal category to a
 symmetric monoidal category, or needing to know that the Cayley-Dickson
 construction that yields the complex numbers from the reals needs a the
 ring structure it builds over to be trivially involutive for the lifted
 ring structure to be commutative.

 These will simply always trigger this warning. In some cases not just for
 my library but for users of my library.

 Now, I've already patched most of that code to include an explicit, messy

 {{{#!hs
 #if __GLASGOW_HASKELL__ >= 800
 {-# OPTIONS_GHC -fno-warn-redundant-constraints #-}
 #endif
 }}}

 But it is telling to me that this is the first situation where I've had to
 hack in explicit `{-# OPTIONS_GHC #-}` to work around a warning going
 haywire.

 At present the only other code I can think of where I turn on explicit
 `{-# OPTIONS_GHC #-}` is to hack around changes in sharing behavior across
 GHC versions that breaks particularly tricky uses of `{-# NOINLINE #-}`.
 To get `reflection` to work across a wide enough array of GHC versions I
 wind up with a cocktail that looks like:

 {{{#!hs
 {-# OPTIONS_GHC -fno-cse -fno-full-laziness -fno-float-in #-}
 }}}

 but that isn't hacking around warnings, but rather buggy code motion that
 changes in behavior across GHC versions.

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11370#comment:27>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler