February 2018 - 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 24 Jul '18

24 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] #14333: GHC doesn't use the fact that Coercible is symmetric
by GHC 23 Jul '18

23 Jul '18

#14333: GHC doesn't use the fact that Coercible is symmetric -------------------------------------+------------------------------------- Reporter: Iceland_jack | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.1 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: -------------------------------------+------------------------------------- {{{#!hs {-# Language ScopedTypeVariables, GADTs, DeriveGeneric #-} import GHC.Generics import Data.Type.Coercion data AB = A | B deriving (Generic) data XY = X | Y deriving (Generic) data This ab xy where At :: This AB XY foo :: This ab xy -> Coercion (Rep ab a) (Rep xy a) foo At = Coercion bar :: forall ab xy a. This ab xy -> Coercion (Rep ab a) (Rep xy a) bar at | Coercion <- foo at :: Coercion (Rep ab a) (Rep xy a) = Coercion }}} This compiles on 8.2.1 and 8.3.20170920 but flipping the arguments to `Coercion` fails {{{#!hs -- ... -- • Could not deduce: Coercible (Rep xy a) (Rep ab a) -- arising from a use of ‘Coercion’ -- from the context: Coercible (Rep ab a) (Rep xy a) -- bound by a pattern with constructor: bar :: forall ab xy a. This ab xy -> Coercion (Rep xy a) (Rep ab a) bar at | Coercion <- foo at :: Coercion (Rep ab a) (Rep xy a) = Coercion }}} Conceptually there should be an (`Undecidable`) superclass constraint for `Coercible` {{{#!hs class Coercible b a => Coercible a b }}} -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14333> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 16

[GHC] #14414: Profiled program runs 2.5x faster than non-profiled
by GHC 23 Jul '18

23 Jul '18

#14414: Profiled program runs 2.5x faster than non-profiled -------------------------------------+------------------------------------- Reporter: Fuuzetsu | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.1 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: -------------------------------------+------------------------------------- I was looking at benchmarks game (attached as fasta.ghc-2.hs). I have found that with the flags give there, this program on GHC 8.2.1 runs in about a second with `-prof -fprof-auto` and 2.5 seconds without! To run without profiling: {{{ ghc --make -Wall -fforce-recomp -fllvm -O2 -XBangPatterns -threaded -rtsopts -XOverloadedStrings fasta.ghc-2.hs -o fasta.ghc-2.ghc_run && ./fasta.ghc-2.ghc_run +RTS -N4 -s -RTS 250000 > /dev/null }}} Same program with profiling {{{ ghc --make -fforce-recomp -fllvm -prof -fprof-auto -O2 -XBangPatterns -threaded -rtsopts -XOverloadedStrings fasta.ghc-2.hs -o fasta.ghc-2.ghc_run && ./fasta.ghc-2.ghc_run +RTS -N4 -p -s -RTS 250000 > /dev/null }}} I also attach Core outputs for both profiled and unprofiled version. To me this seems very strange: profiled version is somehow faster. Perhaps what's worse is that this means that there's some optimisation GHC is performing when profiling is not on that makes the program a lot slower than it could be! This program is not minimised. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14414> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 16

[GHC] #12714: T9405 fails on Windows
by GHC 22 Jul '18

22 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 21 Jul '18

21 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] #14242: Ticks and join points don't play well
by GHC 21 Jul '18

21 Jul '18

#14242: Ticks and join points don't play well -------------------------------------+------------------------------------- Reporter: bgamari | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.2.1 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: -------------------------------------+------------------------------------- Currently the presence of ticks appear to inhibit our ability to discover join points as described by this comment in `OccurAnal.occAnal`, {{{#!hs usage_lam = markAllNonTailCalled (markAllInsideLam usage) -- TODO There may be ways to make ticks and join points play -- nicer together, but right now there are problems: -- let j x = ... in tick<t> (j 1) -- Making j a join point may cause the simplifier to drop t -- (if the tick is put into the continuation). So we don't -- count j 1 as a tail call. }}} This causes spurious warnings when compiling with a `DEBUG` compiler and `-g` or profiling enabled. -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14242> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 3

[GHC] #14388: GHC Panic
by GHC 20 Jul '18

20 Jul '18

#14388: GHC Panic -------------------------------------+------------------------------------- Reporter: aferrandi | Owner: (none) Type: bug | Status: new Priority: normal | Milestone: Component: Compiler | Version: 8.0.2 Keywords: IfaceTyLit | Operating System: Linux panic | Architecture: | Type of failure: Compile-time Unknown/Multiple | crash or panic Test Case: | Blocked By: Blocking: | Related Tickets: Differential Rev(s): | Wiki Page: -------------------------------------+------------------------------------- Building the project exmap I,ve got: Information:stack: Building 'exmap' ... Information:stack: /usr/bin/stack build . --stack-yaml /home/andrea/prj/exmap/server/stack.yaml "--ghc-options=-Wall -Werror" exmap-0.1.0.0: configure (exe) Configuring exmap-0.1.0.0... exmap-0.1.0.0: build (exe) Preprocessing executable 'exmap' for exmap-0.1.0.0... ghc: panic! (the 'impossible' happened) (GHC version 8.0.2 for x86_64-unknown-linux): get IfaceTyLit 255 Information:stack: Please report this as a GHC bug: http://www.haskell.org/ghc/reportabug Information:stack: -- While building package exmap-0.1.0.0 using: /home/andrea/.stack/setup-exe-cache/x86_64-linux/Cabal- simple_mPHDZzAJ_1.24.2.0_ghc-8.0.2 --builddir=.stack- work/dist/x86_64-linux/Cabal-1.24.2.0 build exe:exmap --ghc-options " -ddump-hi -ddump-to-file" Process exited with code: ExitFailure 1 Information:25.10.17 16:19 - Compilation completed with 1 error and 0 warnings in 1s 946ms Error:stack: Stack build failed with nonzero exit status -- Ticket URL: <http://ghc.haskell.org/trac/ghc/ticket/14388> GHC <http://www.haskell.org/ghc/> The Glasgow Haskell Compiler

1 5

[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