[commit: ghc] wip/generalized-arrow: CoreLint: Check that there are no saturated (->) applications (89b2685)

[git at git.haskell.org]

Repository : ssh://git@git.haskell.org/ghc

On branch  : wip/generalized-arrow
Link       : http://ghc.haskell.org/trac/ghc/changeset/89b26858b53be665cc1de3c5f4b36170fd67d01e/ghc

>---------------------------------------------------------------

commit 89b26858b53be665cc1de3c5f4b36170fd67d01e
Author: Ben Gamari <ben at smart-cactus.org>
Date:   Fri Mar 25 13:28:40 2016 +0100

    CoreLint: Check that there are no saturated (->) applications


>---------------------------------------------------------------

89b26858b53be665cc1de3c5f4b36170fd67d01e
 compiler/coreSyn/CoreLint.hs | 12 ++++++++++++
 1 file changed, 12 insertions(+)

diff --git a/compiler/coreSyn/CoreLint.hs b/compiler/coreSyn/CoreLint.hs
index ef44aff..1c40049 100644
--- a/compiler/coreSyn/CoreLint.hs
+++ b/compiler/coreSyn/CoreLint.hs
@@ -127,6 +127,12 @@ Outstanding issues:
     --   may well be happening...);
 
 
+Note [Linting function types]
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+As described in Note [Representation of function types], all saturated
+applications of funTyCon are represented with ForAllTy. We check this
+invariant in lintType.
+
 Note [Linting type lets]
 ~~~~~~~~~~~~~~~~~~~~~~~~
 In the desugarer, it's very very convenient to be able to say (in effect)
@@ -1033,6 +1039,12 @@ lintType ty@(TyConApp tc tys)
   = lintType ty'   -- Expand type synonyms, so that we do not bogusly complain
                    --  about un-saturated type synonyms
 
+  -- We should never see a saturated application of funTyCon; such applications
+  -- should be represented with ForAllTy. See Note [Linting function types].
+  | isFunTyCon tc
+  , length tys == 4
+  = failWithL (hang (text "Saturated application of (->)") 2 (ppr ty))
+
   | isUnliftedTyCon tc || isTypeSynonymTyCon tc || isTypeFamilyTyCon tc
        -- Also type synonyms and type families
   , length tys < tyConArity tc