#13549: GHC 8.2.1's typechecker rejects code generated by singletons that 8.0 accepts -------------------------------------+------------------------------------- Reporter: RyanGlScott | Owner: (none) Type: bug | Status: new Priority: high | Milestone: 8.2.1 Component: Compiler (Type | Version: 8.1 checker) | Resolution: | Keywords: TypeInType 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 RyanGlScott): Replying to [comment:9 RyanGlScott]:
And I think that's probably the limit of my ability to debug this :)
OK, I lied. I //was// able to discover a workaround by placing kind signatures at random places until I found a trick that worked reliably. Here's how you can fix each of the files I posted: {{{#!diff diff --git a/Bug.hs b/Bug.hs index d534fe0..52163b8 100644 --- a/Bug.hs +++ b/Bug.hs @@ -1127,7 +1127,7 @@ src/Data/Singletons/Prelude/List.hs:(260,3)-(806,4): Splicing declar lambda_a33gY xs0_a33gZ = let sPerms :: - forall arg_a33h0 arg_a33h1. + forall (arg_a33h0 :: [a_a337f]) arg_a33h1. Sing arg_a33h0 -> Sing arg_a33h1 -> Sing (Apply (Apply (Let6989586621679736931PermsSym1 xs0_a33a0) ar @@ -1153,7 +1153,7 @@ src/Data/Singletons/Prelude/List.hs:(260,3)-(806,4): Splicing declar lambda_a33ha t_a33hb ts_a33hc is_a33hd = let sInterleave' :: - forall arg_a33he arg_a33hf arg_a33hg. + forall (arg_a33he :: TyFun [a_a337f] k -> Type) arg_a33hf a Sing arg_a33he -> Sing arg_a33hf -> Sing arg_a33hg }}} {{{#!diff diff --git a/Bug2.hs b/Bug2.hs index 98fbbf0..75c4523 100644 --- a/Bug2.hs +++ b/Bug2.hs @@ -393,7 +393,7 @@ src/Data/Singletons/Prelude/List.hs:(260,3)-(806,4): Splicing declarat lambda_a33gY xs0_a33gZ = let sPerms :: - forall arg_a33h0 arg_a33h1. + forall (arg_a33h0 :: [k]) arg_a33h1. Sing arg_a33h0 -> Sing arg_a33h1 -> Sing (Apply (Apply (Let6989586621679736931PermsSym1 xs0_a33a0) ar @@ -419,7 +419,7 @@ src/Data/Singletons/Prelude/List.hs:(260,3)-(806,4): Splicing declarat lambda_a33ha t_a33hb ts_a33hc is_a33hd = let sInterleave' :: - forall arg_a33he arg_a33hf arg_a33hg. + forall (arg_a33he :: TyFun [k] j -> Type) arg_a33hf arg_a33 Sing arg_a33he -> Sing arg_a33hf -> Sing arg_a33hg }}} {{{#!diff diff --git a/Bug3.hs b/Bug3.hs index 77484b2..a286b43 100644 --- a/Bug3.hs +++ b/Bug3.hs @@ -115,7 +115,7 @@ lambda_a33gY _ lambda_a33ha _ _ _ = let sInterleave' :: - forall arg_a33he arg_a33hf arg_a33hg. + forall (arg_a33he :: TyFun [k] j -> Type) (arg_a33hf :: [y]) (arg_a33hg :: z). Sing arg_a33he -> Sing arg_a33hf -> Sing arg_a33hg }}} The `Bug3` fix is interesting, since I didn't have to specify the kind of a type variable in `sPerms`. But perhaps that's just because I had replaced so many things by `undefined` by that point that the kind inference behavior had changed. So now the question is: is this just another occurrence of #13555 in disguise? Or is there an actual bug here? -- Ticket URL: http://ghc.haskell.org/trac/ghc/ticket/13549#comment:10 GHC http://www.haskell.org/ghc/ The Glasgow Haskell Compiler