[GHC] #11982: Typechecking fails for parallel monad comprehensions with polymorphic let (GHC 7.10.3 through 8.6.3)

[GHC]

#11982: Typechecking fails for parallel monad comprehensions with polymorphic let
(GHC 7.10.3 through 8.6.3)
-------------------------------------+-------------------------------------
        Reporter:  simonpj           |                Owner:  josefs
            Type:  bug               |               Status:  new
        Priority:  normal            |            Milestone:
       Component:  Compiler          |              Version:  7.10.3
      Resolution:                    |             Keywords:  ApplicativeDo
Operating System:  Unknown/Multiple  |         Architecture:
 Type of failure:  GHC rejects       |  Unknown/Multiple
  valid program                      |            Test Case:
      Blocked By:                    |             Blocking:
 Related Tickets:                    |  Differential Rev(s):
       Wiki Page:                    |
-------------------------------------+-------------------------------------

Comment (by simonmar):

 Actually I do think that `ApplicativeDo` has a similar problem to parallel
 list comprehensions. It's true that in the example given in comment:5
 above we could fix `ApplicativeDo` to do a slightly different
 transformation and not run into the problem. But that won't work in
 general, e.g.

 {{{
 test :: IO (Int, Char, Int)
 test = do
   x <- return 'a'
   y <- return 'a'
   let f | y == 'c' = id | otherwise = id
   z <- return (f 3)
   return (f 3, f 'a', z)
 }}}

 This is accepted without `ApplicativeDo`, but rejected with it, because we
 need `f` to be polymorphic.  The renamer produces:

 {{{
 ==================== Renamer ====================
 Test.test :: IO (Int, Char, Int)
 Test.test
   = do x_a3Dl <- return 'a' |
        (f_a3Dn, z_a3Do) <- do y_a3Dm <- return 'a'
                               let f_a3Dn
                                     | y_a3Dm == 'c' = id
                                     | otherwise = id
                               z_a3Do <- return (f_a3Dn 3)
                               (f_a3Dn, z_a3Do)
        return (f_a3Dn 3, f_a3Dn 'a', z_a3Do)
 }}}

 so we would need to instantiate the tuple type with polymorphic arguments.

 Simon - I vaguely remember that we discussed this when reviewing the code
 for ApplicativeDo and we decided to punt on it at the time, but
 unfortunately I can't find any record of it, do you remember by any
 chance?

-- 
Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11982#comment:19>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler