November 2016 - ghc-tickets

[GHC] #12736: Calling a complex Haskell function (obtained via FFI wrapper function) from MSVC 64-bit C code (passed in as FunPtr) can leave SSE2 registers in the XMM6-XMM15 range modified
by GHC 25 Jul '18

25 Jul '18

#12736: Calling a complex Haskell function (obtained via FFI wrapper function) from MSVC 64-bit C code (passed in as FunPtr) can leave SSE2 registers in the XMM6-XMM15 range modified -------------------------------------+------------------------------------- Reporter: bavism | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 (FFI) | Keywords: | Operating System: Windows ffi,registers,sse2,clobber,xmm | Architecture: x86_64 | Type of failure: Incorrect result (amd64) | at runtime Test Case: | Blocked By: https://github.com/bavis-m/raycast | Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- According to the [https://msdn.microsoft.com/en-us/library/9z1stfyw.aspx MSDN], in the Microsoft x64 architecture function calls must preserve the SSE2 registers in the range XMM6-XMM15. The Haskell FFI can produce a function pointer via dynamic wrapper that, when called from MSVC x64 C code, does not preserve these registers, causing further floating-point operations in the C code to fail. I can reproduce this error in [https://github.com/bavis-m/raycast this project], which is a DOOM-style raycasting engine written in Haskell, that imports a C DLL with glue for rendering and window management. The Haskell executable generates a FunPtr to a frame update function using the dynamic import mechanism, and passes this to a long-lived C function that runs the update loop. Any time this update function is called from the C loop, subsequent floating point operations produce incorrect results (in this case, the next operations compute a view matrix for the OpenGL window). The output on every frame showing the view matrix should be: {{{ viewM: 0.003125 0.000000 0.000000 -1.000000 0.000000 0.004167 0.000000 -1.000000 0.000000 0.000000 1.000000 0.000000 0.000000 0.000000 0.000000 1.000000 }}} Running the raycaster with the Release version of the DLL causes the value of this matrix to be corrupted. There is a patch provided (stub.patch in the root folder) that turns the Haskell update function into an empty stub. This causes the program to work. When stepping through the assembly code with this patch applied, I can see in the function prologue where the XMM registers are saved. Without the patch, these registers are not saved. Running the Debug version does not show this error; the register allocation must be different. I have been attempting to create a much simpler test case to reveal this code-generation issue, however it has been difficult. Even seemingly trivial changes can cause the bug to not show up, it is clearly dependent on the register allocation used internally to produce the assembly code. Instructions for building the project are in the readme. (You will need the Haskell Stack Tool, and Visual Studio 15). -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12736> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 4

[GHC] #8128: Standalone deriving fails for GADTs due to inaccessible code
by GHC 24 Jul '18

24 Jul '18

#8128: Standalone deriving fails for GADTs due to inaccessible code -------------------------------------+------------------------------------- Reporter: adamgundry | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler (Type | Version: 7.7 checker) | Operating System: Unknown/Multiple Keywords: | Type of failure: GHC rejects Architecture: Unknown/Multiple | valid program Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | -------------------------------------+------------------------------------- Consider the following: {{{ {-# LANGUAGE StandaloneDeriving, GADTs, FlexibleInstances #-} module StandaloneDerivingGADT where data T a where MkT1 :: T Int MkT2 :: (Bool -> Bool) -> T Bool deriving instance Show (T Int) }}} This gives the error: {{{ StandaloneDerivingGADT.hs:9:1: Couldn't match type ‛Int’ with ‛Bool’ Inaccessible code in a pattern with constructor MkT2 :: (Bool -> Bool) -> T Bool, in an equation for ‛showsPrec’ In the pattern: MkT2 b1 In an equation for ‛showsPrec’: showsPrec a (MkT2 b1) = showParen ((a >= 11)) ((.) (showString "MkT2 ") (showsPrec 11 b1)) When typechecking the code for ‛showsPrec’ in a standalone derived instance for ‛Show (T Int)’: To see the code I am typechecking, use -ddump-deriv In the instance declaration for ‛Show (T Int)’ }}} The derived instance declaration matches on all the constructors, even if they cannot possibly match. It should omit obviously inaccessible constructors so that this example is accepted. For reference, the derived code is: {{{ instance GHC.Show.Show (StandaloneDerivingGADT.T GHC.Types.Int) where GHC.Show.showsPrec _ StandaloneDerivingGADT.MkT1 = GHC.Show.showString "MkT1" GHC.Show.showsPrec a_aij (StandaloneDerivingGADT.MkT2 b1_aik) = GHC.Show.showParen ((a_aij GHC.Classes.>= 11)) ((GHC.Base..) (GHC.Show.showString "MkT2 ") (GHC.Show.showsPrec 11 b1_aik)) GHC.Show.showList = GHC.Show.showList__ (GHC.Show.showsPrec 0) }}} The same problem applies to other derivable classes (e.g. `Eq`). -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8128> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 15

[GHC] #8740: Code conditionally compiles
by GHC 24 Jul '18

24 Jul '18

#8740: Code conditionally compiles ----------------------------+---------------------------------------------- Reporter: | Owner: thomaseding | Status: new Type: bug | Milestone: Priority: normal | Version: 7.6.3 Component: Compiler | Operating System: MacOS X Keywords: | Type of failure: GHC rejects valid program Architecture: | Test Case: Unknown/Multiple | Blocking: Difficulty: Unknown | Blocked By: | Related Tickets: | ----------------------------+---------------------------------------------- {{{ {-# LANGUAGE GADTs #-} {-# LANGUAGE StandaloneDeriving #-} data Abstract data Reified data Player data Elect p a where ElectRefAsTypeOf :: Int -> Elect Abstract a -> Elect Abstract a ElectHandle :: a -> Elect Reified a Controller :: Elect Abstract Player Owner :: Elect Abstract Player You :: Elect Abstract Player deriving instance (Eq a) => Eq (Elect p a) deriving instance (Ord a) => Ord (Elect p a) }}} As is, the above code fails to compile. But if I move `ElectRefAsTypeOf` to be the last constructor for the GADT, the code does compile. If I remove one of the `Elect Abstract Player` constructors, the code still won't compile even if the `ElectRefAsTypeOf` is moved. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8740> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 8

[GHC] #12714: T9405 fails on Windows
by GHC 23 Jul '18

23 Jul '18

#12714: T9405 fails on Windows --------------------------------------+--------------------------------- Reporter: bgamari | Owner: Type: bug | Status: new Priority: normal | Milestone: 8.2.1 Component: Runtime System | Version: 8.0.1 Keywords: | Operating System: Windows Architecture: x86_64 (amd64) | Type of failure: None/Unknown Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: --------------------------------------+--------------------------------- The `T9405` testcase currently fails on Windows, {{{ }}} Contrary to what was suggested in ticket:12004#comment:3 it looks like the problem is that an empty `.ticky` file is produced. Even increasing the `sleep` time to 10 seconds (after bumping the sleep in `T9405.hs` accordingly) doesn't change this, so I suspect there is a runtime system bug here. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12714> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 7

[GHC] #9599: app runs 10 times faster when compiled with profilling information than without it
by GHC 22 Jul '18

22 Jul '18

#9599: app runs 10 times faster when compiled with profilling information than without it -------------------------------------+------------------------------------- Reporter: kgardas | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.8.2 Keywords: | Operating System: Solaris Architecture: x86 | Type of failure: Runtime Difficulty: Unknown | performance bug Blocked By: | Test Case: Related Tickets: | Blocking: | Differential Revisions: -------------------------------------+------------------------------------- Hello, I've found interesting application code which when compiled with {{{ -O2 -threaded --make }}} runs 10 times slower than the same code compiled with {{{ -O2 -threaded --make -prof -auto-all -caf-all -fforce-recomp }}} the application is compiled by GHC 7.8.2 on i386-solaris2 platform. The profiling application is then run as normal application, i.e. without +RTS -p. The application itself is my crude beginner Haskell code which takes Wikipedia dump (XML file) and tries wiki pages from there import into XWiki using REST API. It uses Data.Text package for lazy load of Wikipedia data, whole logic of selection of page/title/content is done on Data.Text and later saved into XPage which is using ByteString. The application also uses Simon Marlow's Async/async/wait code which is copied from Simon Marlow's Parallel and Concurrent Programming in Haskell http://chimera.labs.oreilly.com/books/1230000000929 App attached. My question is, what to do with it to help finding the issue? E.g. I assume it's issue when profiling app is that much faster than common optimized app. For your information, optimized app takes 9m50s to upload 1000 pages while profiled optimized app takes only 50s to upload the same amount of pages on my equipped with E5-2620 (2 GHz 6 core/12 threads xeon) Also the difference in time is that big that it's not caused by a noise in import of pages on XWiki side. I've tested several times of course. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/9599> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 4

[GHC] #12606: Linux (ELF) Support for "ghc -static -shared"
by GHC 19 Jul '18

19 Jul '18

#12606: Linux (ELF) Support for "ghc -static -shared" ----------------------------------------+--------------------------------- Reporter: tmobile | Owner: Type: feature request | Status: new Priority: normal | Milestone: ⊥ Component: Compiler | Version: 8.0.1 Keywords: | Operating System: Linux Architecture: Unknown/Multiple | Type of failure: None/Unknown Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: ----------------------------------------+--------------------------------- On some platforms it is possible to create a stand-alone shared library out of some object `Foo.o` including foreign exports with something like: `ghc -static -shared Foo.o ...` The resulting library will have `Foo`'s dependencies including `base` as well as the RTS statically linked in. On BSDs this is possible because generating position independent code is the default there, and on Windows this is possible because the dynamic loader uses some magic that doesn't require position independent code. Unfortunately on Linux this is not possible because `libHSbase`, `libHSrts`, etc. are not built with `-fPIC` unless one specifically asks for it when building GHC. As far as I know this is the only way to get `-static -shared` to work on this platform. There are use cases for compiling Haskell code into a single shared object that does not have additional runtime dependencies. Suppose one wishes to write a plugin for a program in Haskell. It would be nice to provide the plugin as a single shared object, rather than requiring GHC's libraries to be present at runtime as well. An easy solution might be adding a switch to the configure script or `build.mk` to make it easier to build GHC+libs with `-fPIC` on Linux. Ideally it would be possible to build and install both PIC and non-PIC libraries and have GHC choose the correct one at link time based on the flags and target platform, since PIC code carries a small space/performance overhead and isn't necessary on Linux in most cases. I don't know my way around GHC(outside of the GHC API) or the build system that well yet, but if adding extra build system flags and GHC link-time flag handling code to make this work is an acceptable solution I'm happy to spend the time to get this working. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12606> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 1

[GHC] #8581: Add support for explicitly-bidirectional pattern synonyms
by GHC 12 Jul '18

12 Jul '18

#8581: Add support for explicitly-bidirectional pattern synonyms ------------------------------------+------------------------------------- Reporter: cactus | Owner: cactus Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler | Version: Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: 5144 | Blocking: Related Tickets: | ------------------------------------+------------------------------------- Some patterns cannot, by themselves, be turned over into an expression, so they have to be defined as unidirectional. Maybe the most trivial example would be {{{ pattern P -> _ }}} Sometimes, however, it would be desirable to give an explicit way of turning these pattern synonyms into expressions. The PatternSynonyms wiki page has this example: {{{ import qualified Data.Sequence as Seq pattern Empty -> (Seq.viewl -> Seq.EmptyL) pattern x :< xs -> (Seq.viewl -> x Seq.:< xs) pattern xs :> x -> (Seq.viewr -> xs Seq.:> x) }}} It would make a ton of sense to be able to use this cons/snoc notation as "constructors" for `Seq`s. The proposed syntax for this {{{ pattern x :< xs -> (Seq.viewl -> x Seq.:< xs) where x :< xs = x Seq.<| xs }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8581> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 52

[GHC] #12050: Allow haddock comments on non-record types
by GHC 10 Jul '18

10 Jul '18

#12050: Allow haddock comments on non-record types -------------------------------------+------------------------------------- Reporter: bgamari | Owner: Type: feature | Status: new request | Priority: normal | Milestone: Component: Compiler | Version: 7.10.3 (Parser) | 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 variety of cases I have found myself wanting to place haddock documentation on fields of non-record data types rather often in the past. The syntax for this might look like, {{{#!hs data Foo = Foo -- ^ A foo Int -- ^ how big is the foo? String -- ^ the name of the foo }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/12050> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6

[GHC] #8822: Allow -- ^ Haddock syntax on record constructors
by GHC 10 Jul '18

10 Jul '18

#8822: Allow -- ^ Haddock syntax on record constructors -------------------------------------+------------------------------------- Reporter: Fuuzetsu | Owner: Type: feature request | Status: new Priority: normal | Milestone: Component: Compiler (Parser) | Version: 7.9 Keywords: | Operating System: Unknown/Multiple Architecture: Unknown/Multiple | Type of failure: None/Unknown Difficulty: Unknown | Test Case: Blocked By: | Blocking: Related Tickets: | -------------------------------------+------------------------------------- Original ticket is on Haddock Trac [http://trac.haskell.org/haddock/ticket/144 here] which I'm opening here as it requires lexer/parser changes. The idea is that given something like {{{ -- | A tree with labeled internal nodes data Tree = Empty -- ^ Leaf | Branch -- ^ Internal node { label :: a -- ^ Node label , subtrees :: [Tree] -- ^ List of subtrees } }}} we'd like it to parse as the user would expect instead of failing. We're currently forced to do something like {{{ -- | A tree with labeled internal nodes data Tree = Empty -- ^ Leaf | -- | Internal node Branch { label :: a -- ^ Node label , subtrees :: [Tree] -- ^ List of subtrees } }}} which is quite ugly. I can't think of a reason why the first form wouldn't be allowed. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/8822> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 8

[GHC] #11070: Type-level arithmetic of sized-types has weaker inference power than in 7.8
by GHC 10 Jul '18

10 Jul '18

#11070: Type-level arithmetic of sized-types has weaker inference power than in 7.8 -------------------------------------+------------------------------------- Reporter: cactus | Owner: Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 7.10.2 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: -------------------------------------+------------------------------------- Given the following definitions (just a copy of the relevant bits from sized-types 0.3.5.1, to keep this example self-contained): {{{ -- Copied from sized-types 0.3.5.1 data X0 = X0 data N1 = N1 data X0_ a = X0_ Integer data X1_ a = X1_ Integer type X1 = X1_ X0 type X2 = X0_ (X1_ X0) class Size ix instance Size X0 instance Size a => Size (X1_ a) instance Size a => Size (X0_ a) type family APP0 a type instance APP0 X0 = X0 type instance APP0 N1 = X0_ N1 type instance APP0 (X1_ a) = X0_ (X1_ a) type instance APP0 (X0_ a) = X0_ (X0_ a) type family APP1 a type instance APP1 X0 = X1_ X0 type instance APP1 N1 = N1 type instance APP1 (X1_ a) = X1_ (X1_ a) type instance APP1 (X0_ a) = X1_ (X0_ a) type family SUCC a type instance SUCC X0 = X1_ X0 type instance SUCC N1 = X0 type instance SUCC (X1_ a) = APP0 (SUCC a) type instance SUCC (X0_ a) = APP1 a type family ADD a b type instance ADD a X0 = a type instance ADD X0 a = a type instance ADD X0 N1 = N1 type instance ADD N1 N1 = APP0 N1 type instance ADD N1 (X1_ b) = APP0 b type instance ADD N1 (X0_ b) = APP1 (ADD N1 b) type instance ADD (X1_ a) N1 = APP0 a type instance ADD (X0_ a) N1 = APP1 (ADD a N1) type instance ADD (X1_ a) (X1_ b) = APP0 (SUCC (ADD a b)) type instance ADD (X1_ a) (X0_ b) = APP1 (ADD a b) type instance ADD (X0_ a) (X1_ b) = APP1 (ADD a b) type instance ADD (X0_ a) (X0_ b) = APP0 (ADD a b) type family NOT a type instance NOT X0 = N1 type instance NOT N1 = X0 type instance NOT (X1_ a) = APP0 (NOT a) type instance NOT (X0_ a) = APP1 (NOT a) type SUB a b = ADD a (SUCC (NOT b)) }}} The following module typechecks with GHC 7.8.3: {{{ data B w = B (&*) :: (Size n, Size n', Size (ADD n' n), n ~ SUB (ADD n' n) n', n' ~ SUB (ADD n' n) n) => [(a, B n)] -> [(b, B n')] -> [((a, b), B (ADD n' n))] mks &* args = undefined foo :: [((), B X1)] foo = [((), B)] bar :: [(((), ()), B X2)] bar = [((), B)] &* foo }}} However, it fails with GHC 7.10.2, with {{{ /tmp/GHCBug.hs:70:7: Couldn't match type ‘ADD X1 n0’ with ‘X0_ (X1_ X0)’ The type variable ‘n0’ is ambiguous Expected type: [(((), ()), B X2)] Actual type: [(((), ()), B (ADD X1 n0))] In the expression: [((), B)] &* foo In an equation for ‘bar’: bar = [((), B)] &* foo /tmp/GHCBug.hs:70:17: Occurs check: cannot construct the infinite type: n0 ~ ADD (ADD X1 n0) N1 The type variable ‘n0’ is ambiguous Expected type: SUB (ADD X1 n0) X1 Actual type: n0 In the expression: [((), B)] &* foo In an equation for ‘bar’: bar = [((), B)] &* foo Failed, modules loaded: none. }}} The workaround/solution is to change the definition of `bar`: {{{ bar :: [(((), ()), B X2)] bar = [((), B :: B X1)] &* foo }}} This second version typechecks with GHC 7.10.2. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/11070> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 6