March 2016 - ghc-tickets

[GHC] #11642: Typechecker doesn't use evidence available from pattern synonym?
by GHC 16 Mar '16

16 Mar '16

#11642: Typechecker doesn't use evidence available from pattern synonym? -------------------------------------+------------------------------------- Reporter: bgamari | Owner: Type: bug | Status: new Priority: high | Milestone: 8.0.1 Component: Compiler | Version: 7.10.3 (Type checker) | Keywords: | 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: -------------------------------------+------------------------------------- Either that or I am missing something... Consider the following, {{{#!hs {-# LANGUAGE GADTs, PolyKinds, PatternSynonyms, RankNTypes, TypeOperators, ViewPatterns #-} module Test where data TypeRep (a :: k) where TypeCon :: String -> TypeRep a TypeApp :: TypeRep f -> TypeRep x -> TypeRep (f x) data TypeRepX where TypeRepX :: TypeRep a -> TypeRepX data (a :: k1) :~~: (b :: k2) where HRefl :: a :~~: a trArrow :: TypeRep (->) trArrow = undefined eqTypeRep :: TypeRep (a :: k1) -> TypeRep (b :: k2) -> Maybe (a :~~: b) eqTypeRep = undefined typeRepKind :: forall (k :: *). forall (a :: k). TypeRep a -> TypeRep k typeRepKind = undefined pattern TRFun :: forall fun. () => forall arg res. (fun ~ (arg -> res)) => TypeRep arg -> TypeRep res -> TypeRep fun pattern TRFun arg res <- TypeApp (TypeApp (eqTypeRep trArrow -> Just HRefl) arg) res buildApp :: TypeRepX -> TypeRepX -> TypeRepX buildApp (TypeRepX f) (TypeRepX x) = case typeRepKind f of TRFun arg _ -> case eqTypeRep arg x of Just HRefl -> TypeRepX $ TypeApp f x }}} This fails with, {{{ $ ghc Test.hs -fprint-explicit-kinds [1 of 1] Compiling Test ( Test.hs, Test.o ) Test.hs:38:30: error: • Expected kind ‘TypeRep (k1 -> res) a’, but ‘f’ has kind ‘TypeRep k a’ • In the first argument of ‘TypeApp’, namely ‘f’ In the second argument of ‘($)’, namely ‘TypeApp f x’ In the expression: TypeRepX $ TypeApp f x • Relevant bindings include arg :: TypeRep * arg (bound at Test.hs:35:11) }}} That is, the typechecker doesn't believe that `f`'s type (why is it saying "kind" here?), `TypeRep k a`, will unify with `TypeRep (k1 -> res) a`, despite the `TRFun` pattern match, which should have brought into scope that `k ~ (arg -> res)`. This was tested with a recent snapshot from `ghc-8.0` (23baff798aca5856650508ad0f7727045efe3680). Am I missing something here or is this a bug? -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11642> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 11

[GHC] #11709: Merge some TypeInType fixes
by GHC 16 Mar '16

16 Mar '16

#11709: Merge some TypeInType fixes -------------------------------------+------------------------------------- Reporter: goldfire | Owner: Type: bug | Status: new Priority: highest | Milestone: 8.0.1 Component: Compiler | Version: 8.1 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: -------------------------------------+------------------------------------- There are a few more commits to merge, that didn't fit with a single ticket: * de4df6b41a227f527e9eb77733cd6c87b069d3d0 * 3f5d1a13f112f34d992f6b74656d64d95a3f506d * 6c768fcf0d6749bf5029baf6b7f99271b48b1037 * 18fbfa39104b92a05061ec5f6f5bf3b84b462605 -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11709> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 3

[GHC] #11648: assertPprPanic, called at compiler/types/TyCoRep.hs:1932
by GHC 16 Mar '16

16 Mar '16

#11648: assertPprPanic, called at compiler/types/TyCoRep.hs:1932 -------------------------------------+------------------------------------- Reporter: thomie | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.1 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: -------------------------------------+------------------------------------- As mentioned in https://mail.haskell.org/pipermail/ghc- devs/2016-February/011455.html, the following tests currently hit an assert when the compiler is built with -DDEBUG: {{{ patsyn/should_compile MoreEx [exit code non-0] (normal) patsyn/should_compile T11224b [exit code non-0] (normal) polykinds MonoidsTF [exit code non-0] (normal) polykinds T11480b [exit code non-0] (normal) polykinds T11523 [exit code non-0] (normal) }}} {{{ Compile failed (status 256) errors were: ghc-stage2: panic! (the 'impossible' happened) (GHC version 8.1.20160218 for x86_64-unknown-linux): ASSERT failed! CallStack (from HasCallStack): assertPprPanic, called at compiler/types/TyCoRep.hs:1932:56 in ghc:TyCoRep checkValidSubst, called at compiler/types/TyCoRep.hs:1991:17 in ghc:TyCoRep substTys, called at compiler/types/TyCoRep.hs:2012:14 in ghc:TyCoRep substTheta, called at compiler/typecheck/TcPatSyn.hs:255:20 in ghc:TcPatSyn in_scope InScope {d_ap0 c_apv} tenv [ap1 :-> c_apv[tau:5]] tenvFVs [aps :-> t_aps[tau:1], apv :-> c_apv[tau:5]] cenv [] cenvFVs [] tys [] cos [] }}} Phab:D1951 contained a stopgap patch, but since it wasn't accepted, I'll just mark the tests as expect_broken for this ticket, so that other developers aren't bothered by this issue. The offending commit is b5292557dcf2e3844b4837172230575d40a8917e. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11648> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 25

[GHC] #11614: document TypeInType
by GHC 16 Mar '16

16 Mar '16

#11614: document TypeInType -------------------------------------+------------------------------------- Reporter: rwbarton | Owner: goldfire Type: task | Status: new Priority: highest | Milestone: 8.0.1 Component: Documentation | Version: 8.0.1-rc2 Keywords: TypeInType | Operating System: Unknown/Multiple Architecture: | Type of failure: Documentation Unknown/Multiple | bug Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- (Just to keep track of the status for 8.0.1.) -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11614> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #11699: Type families mistakingly report kind variables as unbound type variables
by GHC 16 Mar '16

16 Mar '16

#11699: Type families mistakingly report kind variables as unbound type variables -------------------------------------+------------------------------------- Reporter: mniip | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1-rc2 Keywords: | 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: -------------------------------------+------------------------------------- GHC verifies that if some equation of a type family binds a type variable, that this type variable actually exists and doesn't disappear through type synonym/family application (#7536). However this also mistakingly catches kind variables that aren't present in the type family head. Simplest test case: {{{#!hs {-# LANGUAGE TypeFamilies, PolyKinds #-} type family F a where F (f a) = f a }}} As seen on 8.0.1-rc2 and 7.10.2: {{{ ../File.hs:3:23: Family instance purports to bind type variable ‘k1’ but the real LHS (expanding synonyms) is: F (f a) = ... In the equations for closed type family ‘F’ In the type family declaration for ‘F’ }}} The culprit seems to be in `exactTyCoVarsOfType`, which doesn't grab kind variables from a type variable's kind, even though `tyCoVarsOfType` does. Now, I'm not too sure whether this is a "GHC rejects valid program", or "Incorrect warning at compile time", as I'm not sure if type families like the aforementioned `F` are actually okay. (Are they?) -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11699> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 7

[GHC] #11401: No match in record selector ctev_dest
by GHC 16 Mar '16

16 Mar '16

#11401: No match in record selector ctev_dest -------------------------------------+------------------------------------- Reporter: Lemming | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1-rc1 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: -------------------------------------+------------------------------------- With the llvm-tf package I got the following problem: {{{ $ cat RecordSelectorCtevDest.hs {-# LANGUAGE TypeFamilies #-} {-# LANGUAGE MultiParamTypeClasses #-} {-# LANGUAGE FlexibleInstances #-} module RecordSelectorCtevDest where import Data.Word (Word32, ) import Foreign.Ptr (Ptr, ) newtype Value a = Value a newtype Function a = Function a newtype CodeGenFunction r a = CodeGenFunction a bind :: CodeGenFunction r a -> (a -> CodeGenFunction r b) -> CodeGenFunction r b bind (CodeGenFunction a) k = k a class (f ~ CalledFunction g, r ~ CallerResult g, g ~ CallerFunction f r) => CallArgs f g r where type CalledFunction g :: * type CallerResult g :: * type CallerFunction f r :: * call :: Function f -> g instance CallArgs (IO a) (CodeGenFunction r (Value a)) r where type CalledFunction (CodeGenFunction r (Value a)) = IO a type CallerResult (CodeGenFunction r (Value a)) = r type CallerFunction (IO a) r = CodeGenFunction r (Value a) call = undefined instance CallArgs b b' r => CallArgs (a -> b) (Value a -> b') r where type CalledFunction (Value a -> b') = a -> CalledFunction b' type CallerResult (Value a -> b') = CallerResult b' type CallerFunction (a -> b) r = Value a -> CallerFunction b r call = undefined test :: Function (IO (Ptr a)) -> Function (Ptr a -> IO Word32) -> CodeGenFunction Word32 (Value Word32) test start fill = bind (call start) (call fill) $ ghci-8.0.0.20160109 RecordSelectorCtevDest.hs GHCi, version 8.0.0.20160109: http://www.haskell.org/ghc/ :? for help [1 of 1] Compiling RecordSelectorCtevDest ( RecordSelectorCtevDest.hs, interpreted ) *** Exception: No match in record selector ctev_dest }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11401> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 16

[GHC] #11407: -XTypeInType uses up all memory when used in data family instance
by GHC 16 Mar '16

16 Mar '16

#11407: -XTypeInType uses up all memory when used in data family instance -------------------------------------+------------------------------------- Reporter: RyanGlScott | Owner: Type: bug | Status: new Priority: high | Milestone: Component: Compiler | Version: 8.1 (Type checker) | Keywords: TypeInType | Operating System: Unknown/Multiple Architecture: | Type of failure: Compile-time Unknown/Multiple | performance bug Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Compiling this code with GHC 8.1 causes all my memory to be used very quickly: {{{#!hs {-# LANGUAGE TypeFamilies #-} {-# LANGUAGE TypeInType #-} module TypeInTypeEatsAllMemory where import Data.Kind type Const a b = a data family UhOh (f :: k1) (a :: k2) (b :: k3) data instance UhOh (f :: * -> * -> *) (a :: forall a) (b :: Const * a) = UhOh }}} On the other hand, this code compiles without issue: {{{#!hs {-# LANGUAGE TypeInType #-} module TypeInTypeEatsAllMemory where import Data.Kind type Const a b = a data UhOh (f :: * -> * -> *) (a :: forall a) (b :: Const * a) = UhOh }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11407> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 7

Re: [GHC] #11518: Test TcCoercibleFail hangs with substitution sanity checks enabled
by GHC 14 Mar '16

14 Mar '16

#11518: Test TcCoercibleFail hangs with substitution sanity checks enabled -------------------------------------+------------------------------------- Reporter: niteria | Owner: niteria Type: bug | Status: new Priority: normal | Milestone: Component: Compiler (Type | Version: 8.1 checker) | Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Compile-time | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by goldfire): @thomie: Great -- thanks! -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11518#comment:21> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0

[GHC] #11705: Back port of LlvmCodeGen fix broke 8.0
by GHC 14 Mar '16

14 Mar '16

#11705: Back port of LlvmCodeGen fix broke 8.0 -------------------------------------+------------------------------------- Reporter: erikd | Owner: erikd Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.1-rc2 Keywords: | Operating System: Unknown/Multiple Architecture: | Type of failure: Building GHC Unknown/Multiple | failed Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Commit 1e7764ce in the ghc-8.0 branch (back ported from master) fails to compile because it assumes the prelude of ghc-7.10. Specifically: {{{ compiler/llvmGen/LlvmCodeGen.hs:131:12: Not in scope: ‘pure’ compiler/llvmGen/LlvmCodeGen.hs:140:34: Not in scope: ‘<$>’ Perhaps you meant one of these: ‘<+>’ (imported from Outputable), ‘<>’ (imported from Outputable) }}} etc. The fix should be pretty trivial. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11705> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 2

Re: [GHC] #11518: Test TcCoercibleFail hangs with substitution sanity checks enabled
by GHC 14 Mar '16

14 Mar '16

#11518: Test TcCoercibleFail hangs with substitution sanity checks enabled -------------------------------------+------------------------------------- Reporter: niteria | Owner: niteria Type: bug | Status: new Priority: normal | Milestone: Component: Compiler (Type | Version: 8.1 checker) | Resolution: | Keywords: Operating System: Unknown/Multiple | Architecture: Type of failure: Compile-time | Unknown/Multiple performance bug | Test Case: Blocked By: | Blocking: Related Tickets: | Differential Rev(s): Wiki Page: | -------------------------------------+------------------------------------- Comment (by thomie): Replying to [comment:19 goldfire]: > Can we somehow tell the testsuite to skip this one in this scenario until we have a better workaround? Add the following [wiki:Building/RunningTests/Adding#Thesetupfield setup function] to the test definition in `all.T`: {{{ when(compiler_debugged(), skip) }}} I made that change in 49c55e68aae9841c166430ae566b0d9bdc03c99d. You might need a `git pull`. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11518#comment:20> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 0