[GHC] #9495: Do What I Mean RULES for foldr2 look shady

[GHC]

#9495: Do What I Mean RULES for foldr2 look shady
-------------------------------------+-------------------------------------
              Reporter:  dfeuer      |            Owner:
                  Type:  bug         |           Status:  new
              Priority:  highest     |        Milestone:  7.8.4
             Component:              |          Version:  7.8.3
  libraries/base                     |         Keywords:
            Resolution:              |     Architecture:  Unknown/Multiple
      Operating System:              |       Difficulty:  Unknown
  Unknown/Multiple                   |       Blocked By:
       Type of failure:  Runtime     |  Related Tickets:
  crash                              |
             Test Case:              |
              Blocking:              |
Differential Revisions:              |
-------------------------------------+-------------------------------------

Comment (by dfeuer):

 Replying to [comment:6 simonpj]:
 > To broaden this a bit, it is worth remembering that ''the main
 `foldr/build` rule itself is unsound in the presence of `seq`''.  See, for
 example, the
 [http://www.haskell.org/haskellwiki/Correctness_of_short_cut_fusion#In_the_presence_of_seq
 Haskell wiki page].
 >
 > This is not nice.  The "right" solution is for `seq` to be an operation
 of a type class, something that was the case in Haskell originally, and
 then changed after ''extensive'' debate on the Haskell committee.
 >
 > So currently we are stuck in the unsatisfactory situation that certain
 optimisations, which have a generally very beneficial effect on
 performance, can change termination behaviour. Eta reduction/expansion is
 another.
 >
 > In that general context I don't have a strong opinion about the `foldr2`
 question.  I'd consult the Core Libraries Committee.
 >
 > Simon

 Without a doubt, these are not nice. But we generally keep the danger
 under control. We make the compiler prove that eta expansion is safe
 before applying it (I'm not sure what happens with eta reduction). We hide
 `build` away from the normal user libraries to allow us to pretend it will
 only be used by people who have read the necessary documentation (yes,
 that documentation may need some improvement). But `foldr2`, `zipWith`,
 and `zip` are completely exposed, right in the Prelude, along with a
 `filter` for them to crash with!

 David

--
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9495#comment:7>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler