March 2018 - ghc-tickets

[GHC] #12540: RFC: Allow not quantifying every top-level quantifiee
by GHC 28 May '18

28 May '18

#12540: RFC: Allow not quantifying every top-level quantifiee -------------------------------------+------------------------------------- Reporter: Iceland_jack | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Minor nuisance, does this happen to anyone else? Currently Haskell has an all-nothing policy on quantified type variables, either you quantify none or all (fine, let's ignore non-prenex quantification `foo :: a -> forall b. b -> a`). Can we have a top-level quantification of only a subset of the free variables? Say I'm working on a function {{{#!hs reflected :: forall s m a. (Applicative m, Reifies s a) => TaggedT s m a }}} and I need to make ''s'' a scoped type variable, I always accidentally write {{{#!hs reflected :: forall s. (Applicative m, Reifies s a) => TaggedT s m a reflected = TagT . pure . reflect $ (Proxy :: Proxy s) }}} This causes GHC to complain that the other types — ''m'', ''a'' — are not in scope so and I have to add the remaining quantifiees I don't really care about (may have long names as well). It could be worse (''dramatization'') {{{#!hs ipartsOf :: forall i p f s t a. (Indexable [i] p, Functor f) => Traversing (Indexed i) f s t a a -> Over p f s t [a] [a] ipartsOf l = conjoined (\f s -> let b = inline l sell s in outs b <$> f (wins b)) (\f s -> let b = inline l sell s; (is, as) = unzip (pins b) in outs b <$> indexed f (is :: [i]) as) appendAssocAxiom :: forall p q r as bs cs. p as -> q bs -> r cs -> Dict ((as ++ (bs ++ cs)) ~ ((as ++ bs) ++ cs)) appendAssocAxiom _ _ _ = unsafeCoerce (Dict :: Dict (as ~ as)) }}} It would be nice to be nice to only have to specify the type one is interested in: {{{#!hs ipartsOf :: forall i. (Indexable [i] p, Functor f) => Traversing (Indexed i) f s t a a -> Over p f s t [a] [a] appendAssocAxiom :: forall as. p as -> q bs -> r cs -> Dict ((as ++ (bs ++ cs)) ~ ((as ++ bs) ++ cs)) }}} and have the others chosen in some way. This is not just useful for writing, but it makes it easier to read: If I see a type `forall i p f s t a. ...` any of them may appear in the function body, if I see a type `forall i. ...` I know only one is. Thoughts? -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12540> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 10

[GHC] #9427: Do finer-grained dependency analysis to infer more general kinds on type/class declarations
by GHC 27 May '18

27 May '18

#9427: Do finer-grained dependency analysis to infer more general kinds on type/class declarations -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.2 Keywords: | Operating System: Architecture: Unknown/Multiple | Unknown/Multiple Difficulty: Unknown | Type of failure: Blocked By: | None/Unknown Related Tickets: | Test Case: | Blocking: | Differential Revisions: -------------------------------------+------------------------------------- In the resolution to #9200, as detailed [wiki:GhcKinds/KindInference#Apossiblevariation here], we identified an area for improvement. This improvement was not implemented in the fix for #9200. This ticket will track just this improvement. (In brief: we can do better dependency analysis on mutually recursive type/class declarations to allow more kinds to generalize earlier. This would lessen the burden on programmers to put in kind annotations. See the wiki page.) A solid but unfinished attempt is [https://github.com/goldfirere/ghc/tree /re-sort-non-cusk-decls here]. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9427> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 4

[GHC] #10141: Kind inference regression in closed type families
by GHC 27 May '18

27 May '18

#10141: Kind inference regression in closed type families -------------------------------------+------------------------------------- Reporter: goldfire | Owner: goldfire Type: bug | Status: new Priority: highest | Milestone: 7.10.1 Component: Compiler | Version: 7.10.1-rc2 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Blocked By: Test Case: | Related Tickets: Blocking: | Differential Revisions: | -------------------------------------+------------------------------------- Take the following definition: {{{ type family G (a :: k) where G Int = Bool G Bool = Int G a = a }}} It compiles in 7.8.3, but not in 7.10.1 RC2. This makes me sad. I will fix. (Found by Jan Stolarek.) -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/10141> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 15

[GHC] #14662: Partial type signatures + mutual recursion = confusion
by GHC 26 May '18

26 May '18

#14662: Partial type signatures + mutual recursion = confusion -------------------------------------+------------------------------------- Reporter: goldfire | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.2 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- I'm trying to understand better how partial type signatures interact with mutual recursion. This is all in 8.4.1-alpha1. Example 1: {{{#!hs f :: forall a. _ -> a -> a f _ x = g True x g :: forall b. _ -> b -> b g _ x = f 'x' x }}} This works -- no problem. Example 2: {{{#!hs f :: forall a. _ -> a -> a f _ x = snd (g True 'a', x) g :: forall b. _ -> b -> b g _ x = f 'x' x }}} This fails, explaining that GHC inferred `g :: Bool -> a -> a` and that `a` doesn't match `Char` (in the second argument of the call site in the body of `f`). This is unsatisfactory because clearly `g` should be ''instantiated'' at `Char`. The manual does say that polymorphic recursion isn't available with partial type signatures, and I suppose this is an example of polymorphic (mutual) recursion. Example 3: {{{#!hs f :: forall a. _ -> a -> a f _ x = snd (g True 'a', x) where g :: forall b. _ -> b -> b g _ y = f 'x' y }}} This is accepted. This is the same example as the last one, but now `g` is local. It does not capture any variables (including type variables) from `f`, so it seems to me that it should be equivalent to Example 2. But somehow GHC is clever enough to accept. Example 4: {{{#!hs thdOf3 (_, _, c) = c f :: forall a. _ -> a -> a f _ x = thdOf3 (g True 'a', g False 5, x) where g :: forall b. _ -> b -> b g _ y = f 'x' y }}} This works, too. Note that `g` is applied at several different types, so it really is being generalized. Example 5: {{{#!hs f :: Int -> forall a. _ -> a -> a f n _ x = snd (g n True 'a', x) g :: Int -> forall b. _ -> b -> b g n _ x = f n 'x' x }}} This is accepted. This is the same as Example 2, but each function now takes an `Int`, which is simply passed back and forth. Evidently, when you write the type non-prenex, polymorphic recursion is OK. Example 6: {{{#!hs f :: Int -> forall a. _ -> a -> a f n _ x = snd (f n True 'x', x) }}} This is accepted, even though it's blatantly using polymorphic recursion. Example 7: {{{#!hs f :: forall a. _ -> a -> a f _ x = snd (f True 'x', x) }}} This is rejected as polymorphically recursive. -------------------------- Something is fishy here. It's not the explicit prenex `forall`s -- leaving those out doesn't change the behavior. I guess my big question is this: * If the user quantifies a partial type signature (either by using `forall`, or just using an out-of-scope type variable and using Haskell's implicit quantification), why can't we use polymorphic recursion with that variable? I understand why we can't use polymorphic recursion with wildcards. ----------------------------------- A little background for context: I'm struggling (in my work on #14066) with GHC's use of `SigTv`s for partial type signatures. I don't have a better suggestion, but `SigTv`s never make me feel good. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14662> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[GHC] #14777: panic
by GHC 26 May '18

26 May '18

#14777: panic -------------------------------------+------------------------------------- Reporter: zilinc | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.2 Keywords: | Operating System: Linux Architecture: | Type of failure: Compile-time Unknown/Multiple | crash or panic Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- The error message I got from GHC was: {{{ <no location info>: error: ghc: panic! (the 'impossible' happened) (GHC version 8.2.2 for x86_64-unknown-linux): idInfo t_XG0J Call stack: CallStack (from HasCallStack): prettyCurrentCallStack, called at compiler/utils/Outputable.hs:1133:58 in ghc:Outputable callStackDoc, called at compiler/utils/Outputable.hs:1137:37 in ghc:Outputable pprPanic, called at compiler/basicTypes/Var.hs:526:34 in ghc:Var Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug }}} How was it triggerred (in short): There's function definition, `f x = undefined`. If I change it to `f x = __impossible "Oops!"`, ghc crashes. The definition of `__impossible` is simply `__impossible s = error $ s ++ "some text"`. The code can be found here: https://github.com/NICTA/cogent/blob/92c40530cbcbac77469865e096dab3228a2fb9… and https://github.com/NICTA/cogent/blob/92c40530cbcbac77469865e096dab3228a2fb9… It seems that this bug is not limited to 8.2.2; it also happened in 8.2.1. I'm happy to provide more information, if someone can advice me what is needed. I'll also try to shrink it to a small test case. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14777> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 37

[GHC] #11767: Add @since annotations for base instances
by GHC 26 May '18

26 May '18

#11767: Add @since annotations for base instances -------------------------------------+------------------------------------- Reporter: bgamari | Owner: Type: task | Status: new Priority: normal | Milestone: 8.2.1 Component: Core | Version: 7.10.3 Libraries | Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Up until now we haven't been adding `@since` annotations for instances that we add to `base` (see Phab:D2051 for the discussion that spurred this ticket). We should try to get better coverage here. Unfortunately, we currently have no way to place Haddock comments on derived instances, which limits our ability consistently add these annotations. We'll need to work out some mechanism for this. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11767> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 16

[GHC] #14989: "Each block should be reachable from only one ProcPoint" compiling `integer` with `-fllvm`
by GHC 26 May '18

26 May '18

#14989: "Each block should be reachable from only one ProcPoint" compiling `integer` with `-fllvm` -------------------------------------+------------------------------------- Reporter: sgraf | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.5 Keywords: | Operating System: Windows Architecture: x86_64 | Type of failure: Compile-time (amd64) | crash or panic Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- The attached file is from reduced from NoFib's `spectral/integer`. When I compile this with `-O2 -fllvm`, I get the following panic on HEAD: {{{ ghc-stage2.exe: panic! (the 'impossible' happened) (GHC version 8.5.20180329 for x86_64-unknown-mingw32): Each block should be reachable from only one ProcPoint }}} FWIW, I don't have the LLVM toolchain installed, but this is still in GHC's backend. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14989> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 14

[GHC] #12447: Pretty-printing of equality `~` without parentheses
by GHC 26 May '18

26 May '18

#12447: Pretty-printing of equality `~` without parentheses -------------------------------------+------------------------------------- Reporter: Iceland_jack | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Based off on [https://github.com/ekmett/constraints/blob/master/src/Data/Constraint/Defer… Data.Constraint.Deferrable] {{{#!hs {-# Language RankNTypes, ConstraintKinds #-} import Data.Typeable class Deferrable p where deferEither :: proxy p -> (p => r) -> Either String r instance (Typeable a, Typeable b) => Deferrable (a ~ b) where deferEither = undefined }}} `PolyKinds` aren't enabled so `deferEither @(_ ~ _)` is not enough to select the `Deferrable (a ~ b)` instance, but it is displayed without parentheses {{{ $ ghci -XTypeApplications -ignore-dot-ghci tyiS.hs GHCi, version 8.0.1: http://www.haskell.org/ghc/ :? for help [1 of 1] Compiling Main ( tyiS.hs, interpreted ) Ok, modules loaded: Main. *Main> :t deferEither @(_ ~ _) deferEither @(_ ~ _) :: Deferrable t ~ t1 => proxy t ~ t1 -> (t ~ t1 => r) -> Either String r }}} Instead of a preferable {{{#!hs deferEither @(_ ~ _) :: Deferrable (t ~ t1) => proxy (t ~ t1) -> (t ~ t1 => r) -> Either String r }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12447> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #11540: ghc accepts non-standard type without language extension
by GHC 26 May '18

26 May '18

#11540: ghc accepts non-standard type without language extension -------------------------------------+------------------------------------- Reporter: augustss | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.4 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: None/Unknown Unknown/Multiple | Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Ghc accepts this {{{ f :: Eq a => Eq b => a -> b -> Bool f a b = a==a && b==b }}} The nested contexts should require some language extension. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11540> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #14938: Pattern matching on GADT does not refine type family parameters
by GHC 26 May '18

26 May '18

#14938: Pattern matching on GADT does not refine type family parameters -------------------------------------+------------------------------------- Reporter: kcsongor | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.2 (Type checker) | Keywords: GADTs, | Operating System: Unknown/Multiple TypeFamilies | Architecture: | Type of failure: GHC rejects Unknown/Multiple | valid program Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Compiling the following code {{{#!hs type family R (x :: k) (y :: k) (prf :: x :~: y) :: x :~: y where R x y Refl = Refl }}} fails with the error {{{#!txt Temp.hs:49:9: error: • Expected kind ‘x :~: y’, but ‘Refl’ has kind ‘x :~: x’ • In the third argument of ‘R’, namely ‘Refl’ In the type family declaration for ‘R’ | 49 | R x y Refl = Refl | ^^^^ }}} where `Refl` is defined as {{{#!hs data a :~: b where Refl :: a :~: a }}} I would expect pattern-matching on `Refl` to bring the equality `x ~ y` into scope. It seems like it not only doesn't do that, but the pattern match itself is rejected. (both 8.2.2 and HEAD) -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14938> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 10