- Haskell-Cafe - Haskell.org

ffi call works fine in ghci but not as ghc compiledsegfault
by Miroslav Karpis 17 Mar '14

17 Mar '14

Hi, please can you help me with following? I have a call to an external dll (via ffi) which if executed from ghci works fine. If I compile my code with 'ghc -o fm Mycode.hs -L. -lextdll' and run it I get 'Segmentation fault/access violation in generated code'. main = do let param = "FilePath" let value = "C:/dev/misc/haskell/services/FM" result <- liftIO $ FM.setmodulestring param value return "done" setmodulestring :: String -> String -> IO CInt setmodulestring param value = do let cParamLength = fromIntegral $ length param ::CInt cValueLength = fromIntegral $ length value ::CInt setVarInArray = (-1)::CInt alloca $ \cParam -> do alloca $ \cValue -> do result <- c_setmodulestring cParam cParamLength cValue cValueLength setVarInArray return result If I try also with following, the behaviour is the same: setmodulestring2 :: String -> String -> IO CInt setmodulestring2 param value = do cParam <- newCWString param cValue <- newCWString value let cParamLength = fromIntegral $ length param ::CInt cValueLength = fromIntegral $ length value ::CInt setVarInArray = (-1)::CInt result <- c_setmodulestring cParam cParamLength cValue cValueLength setVarInArray free cParam free cValue return res Any comments/ideas more than appreciated. Cheers, Miro

6 14

ANN: LogicGrowsOnTrees-1.1.0.2
by Gregory Crosswhite 17 Mar '14

17 Mar '14

Dear Haskellers, First, for those eager to skip directly to the more detailed documentation you can find an FAQ after the list of files at http://github.com/gcross/LogicGrowsOnTrees as well as a tutorial and users guide in respectively TUTORIAL.md<http://github.com/gcross/LogicGrowsOnTrees/blob/master/TUTORIAL.md>and USERS_GUIDE.md<http://github.com/gcross/LogicGrowsOnTrees/blob/master/USERS_GUIDE.md>in the list of files. The Hackage page with the reference documentation is at: http://hackage.haskell.org/package/LogicGrowsOnTrees Like many other packages on Hackage, LogicGrowsOnTrees provides an implementation of logic programming using MonadPlus; in this sense it is nothing new. What sets it apart is that it has been designed from the beginning to work in a distributed environment, allowing it to be parallelized over large numbers of processors with no shared memory. The benchmarks I have run (using the N-Queens problem with 17-19 queens) showed essentially perfect speed-up all the way up to 256 cores, and the only reason why this number is not larger is because I haven't had the opportunity to run tests on a larger cluster. Furthermore, I know from personal experience that complicated interfaces can take even the most powerful system and make it impotent, so I have tried hard to design interfaces that are easy to use and well-documented with a tutorial and several examples. There are three packages that are sibling to this: - LogicGrowsOnTrees-processes<http://hackage.haskell.org/package/LogicGrowsOnTrees-processes> - LogicGrowsOnTrees-network<http://hackage.haskell.org/package/LogicGrowsOnTrees-network> - LogicGrowsOnTrees-MPI<http://hackage.haskell.org/package/LogicGrowsOnTrees-MPI> The base package (LogicGrowsOnTrees<http://hackage.haskell.org/package/LogicGrowsOnTrees>) provides an "adapter" that allows one to use threads for parallelization; the other packages above provide adapters for, respectively: processes, a network, and MPI. Each one of these packages provides specialized functionality that can be used directly, but there is also an abstraction over all of them that can be used such that the only difference between the adapters for end-users is a single line of code. For more information, see the links at the top of the post. Thank you for your attention! :-) - Greg

1 0

Re: [Haskell-cafe] ffi call works fine in ghci but not as ghc compiledsegfault
by Miro Karpis 16 Mar '14

16 Mar '14

still no luck ;-( Everything works well under ghci but if I compile it with ghc I get seg. fault ghc version 7.6.3 I found out that when I load my ghci without any .o and .hi files ghci function calls works fine. If I then compile my code with ghc and restart my ghci (.o and .hi files now exists) the ghci crashes. I have tried compiling with: ghc -O --make Service.hs -L. -lmodel ghc --make Service.hs -L. -lmodel ghc -o service Service.hs -L. -lmodel I have tried to call ffi function with: setmodulestring1 param value = do let param = cParamLength = fromIntegral $ length param ::CInt cValueLength = fromIntegral $ length value ::CInt setVarInArray = (-1)::CInt B.useAsCString (B.pack "FilePath") $ \cParam -> do B.useAsCString (B.pack "C:/dev/misc/haskell/services/FM") $ \cValue -> do result <- c_setmodulestring cParam cParamLength cValue cValueLength setVarInArray return result or with this: setmodulestring :: String -> String -> IO CInt setmodulestring param value = do let param = "FilePath" let value = "C:/dev/misc/haskell/services/FM" cParam <- newCString param cValue <- newCString value let cParamLength = fromIntegral $ length param ::CInt cValueLength = fromIntegral $ length value ::CInt setVarInArray = (-1)::CInt result <- c_setmodulestring cParam cParamLength cValue cValueLength setVarInArray free cParam free cValue return result my main is: main = do let param = "FilePath" let value = "C:/dev/misc/haskell/services/FM" result <- liftIO $ FM.setmodulestring param value return "done" and still no luck - works in ghci and crashes as exe interfacing the same library from a c++ code: extern "C" int __stdcall setmodulestring(char* param,unsigned int length, char* valuein,unsigned int valuelength,int index); interfacing it from Haskell: foreign import ccall "setmodulestring" c_setmodulestring :: CString -> CInt -> CString -> CInt -> CInt -> IO CInt please any ideas? On Sat, Mar 15, 2014 at 4:48 PM, Miro Karpis <miroslav.karpis(a)gmail.com>wrote: > thank you very much so far and sorry for my not 100% understanding ;-). > The dll is an external software - I have no control over it. > > This is how I interface it from c++ program: > > extern "C" int __stdcall setmodulestring(char* param,unsigned int length, > char* valuein,unsigned int valuelength,int index); > > > And this is how I'm trying to interface it from Haskell (with the seg-fault): > > foreign import ccall "setmodulestring" c_setmodulestring :: Ptr Char -> CInt -> Ptr Char -> CInt -> CInt -> IO CInt > > m. > > > On Sat, Mar 15, 2014 at 3:32 AM, Brandon Allbery <allbery.b(a)gmail.com>wrote: > >> On Fri, Mar 14, 2014 at 7:29 PM, Branimir Maksimovic < >> branimir.maksimovic(a)gmail.com> wrote: >> >>> You are passing wchar_t* and function expects char*. >>> >> >> FWIW they asked this before (possibly on a different list) and, absent >> the C information, I pointed out that Win32 API functions use CWString. >> This was apparently taken to mean that CWString was the only option.... >> >> Some awareness of one's platform conventions tends to be helpful when >> doing FFI. It seems to be absent here. :/ >> >> -- >> brandon s allbery kf8nh sine nomine >> associates >> allbery.b(a)gmail.com >> ballbery(a)sinenomine.net >> unix, openafs, kerberos, infrastructure, xmonad >> http://sinenomine.net >> > >

1 0

Save the date: BayHac '14
by Mark Lentczner 15 Mar '14

15 Mar '14

The Silicon Valley / San Francisco Bay Area Haskell Hackaton's 4th year! Save the date for a weekend of Haskell hacking: *BayHac '14* *May 16th ~ 18th,* *Hacker Dojo* *Mountain View, California* Details and registration will follow in few weeks. Wiki page for event: BayHac2014 - HaskellWiki<http://www.haskell.org/haskellwiki/BayHac2014> — Jonathan Fischoff & Mark Lentczner

1 0

overlapping/Incoherent closed type families
by Silvio Frischknecht 15 Mar '14

15 Mar '14

Hi I have been playing around a bit with closed type families. However, I somehow always bump my head at the fact that things usually doesn't work for Num without specifying the type. Here is an example. {-# LANGUAGE FlexibleInstances #-} {-# LANGUAGE FlexibleContexts #-} {-# LANGUAGE TypeFamilies #-} {-# LANGUAGE DataKinds #-} {-# LANGUAGE UndecidableInstances #-} {-# LANGUAGE OverlappingInstances #-} {-# LANGUAGE IncoherentInstances #-} module Main where import Data.Typeable type family UnMaybed a where UnMaybed (Maybe a) = a UnMaybed a = a class UnMaybe x where unMaybe :: x -> UnMaybed x instance UnMaybe (Maybe a) where unMaybe (Just a) = a instance (UnMaybed a ~ a) => UnMaybe a where unMaybe a = a main = do print $ unMaybe 'c' print $ unMaybe (1::Int) print $ unMaybe (Just 1) print $ unMaybe 1 -- this line does not compile everything except the last line will compile. ../Example.hs:23:17: Occurs check: cannot construct the infinite type: s0 ~ UnMaybed s0 The type variable ‘s0’ is ambiguous In the second argument of ‘($)’, namely ‘unMaybe 1’ In a stmt of a 'do' block: print $ unMaybe 1 Now I know this is because numbers are polymorphic and (Maybe a) could be an instance of Num. I think for normal overlapping typeclasses this dilemma can be solved by using the IncoherentInstances PRAGMA. Anyway, I wanted to ask if there is a way to make this work in type families? I also thought about specifying Num explicitly in UnMaybed type family UnMaybed a where unMaybed (Num a => a) = a UnMaybed (Maybe a) = a UnMaybed a = a This compiles but i think the first case will never be matched this is probably a bug. Silvio

4 4

HaL-9 - regional Haskell meeting in Halle/Saale, Germany, 2014-06-20 - Save the date
by Henning Thielemann 15 Mar '14

15 Mar '14

Save the date for our local Haskell meeting in Halle/Saale, Germany, presenting tutorials, talks, demonstrations ... everything you like. Workshop language is German (mainly), and English (by request). Switching to German: ------------------------------------------------- Was: Haskell-Treffen HaL-9 Wann: Freitag, 2014-06-20 Wo: Institut für Informatik an der Martin-Luther-Universität in Halle an der Saale Der offizielle Aufruf zum Einreichen von Beiträgen folgt demnächst. Mit besten Grüßen Henning Thielemann

1 0

I seem to constantly abuse TypeFamilies; what do i really want?
by Brandon Simmons 15 Mar '14

15 Mar '14

I've used TypeFamilies numerous times in code and library APIs to get more powerful and creative type-checking, and I seem to always use it in the same way that ends up feeling like abuse. For instance here's an example I started to sketch up for a talk I'm giving; this is to be a type-checked RPN calculator, used e.g. like `eval () (1,(2,((+),())))`: ``` {-# LANGUAGE TypeFamilies , MultiParamTypeClasses , FlexibleInstances , UndecidableInstances #-} type family EvaledStack x stack type instance EvaledStack Int st = (Int,st) type instance EvaledStack (Int -> x) (Int,st) = EvaledStack x st type family FinalStack string initialStack type instance FinalStack () st = st type instance FinalStack (x,xs) st = FinalStack xs (EvaledStack x st) class EvalStep x stack where evalStep :: x -> stack -> EvaledStack x stack instance (EvalStep x st)=> EvalStep (Int -> x) (Int,st) where evalStep f (int,st) = evalStep (f int) st instance EvalStep Int st where evalStep int st = (int,st) class Eval string initialStack where eval :: initialStack -> string -> FinalStack string initialStack instance Eval () st where eval st () = st instance (EvalStep x st, Eval xs (EvaledStack x st))=> Eval (x,xs) st where eval st (x, xs) = eval (evalStep x st) xs ``` The code above is just a WIP, but notice several things: 1) It's intended to be "closed" but I can't express that 2) I need to use UndecidableInstances for the nested type family instances, even though the recursion I'm doing is simple 3) The classes are "ugly" with arbitrary instance heads 4) If I want to support polymorphic operators/operands I need to use OverlappingInstances, which is another layer of hack 5) Users get an unhelpful error from the type-checker if their RPN expression is ill-typed I think what I'm trying to do is fundamentally pretty simple, but I only have the tools to do it in the very ad-hoc way I've described. I think the new closed type families help me here, but I'm wondering: - have there been any proposals or discussions about this use case, or a name given to it? - do closed type families provide an elegant solution and I just don't realize it yet? - do other people find themselves using this pattern as well, or have I just gotten caught up in a strange way of abusing these extensions? Thanks a lot, Brandon

5 5

π-day
by Henk-Jan van Tuyl 14 Mar '14

14 Mar '14

Happy Pi Day! [0] Henk-Jan van Tuyl [0] http://en.wikipedia.org/wiki/Pi_day -- Folding@home What if you could share your unused computer power to help find a cure? In just 5 minutes you can join the world's biggest networked computer and get us closer sooner. Watch the video. http://folding.stanford.edu/ http://Van.Tuyl.eu/ http://members.chello.nl/hjgtuyl/tourdemonad.html Haskell programming --

5 5

looking for alex+happy examples that uses location annotated tokens and location information in err msgs
by Ömer Sinan Ağacan 14 Mar '14

14 Mar '14

Does anyone have any example alex+happy programs that uses location location annotated tokens and location information of token while reporting parse errors? I already wrote lexer and my function type is something like: > runLexer :: String -> [(Token, AlexPosn)] where AlexPosn type contains location information. But I have no idea how can I use that in happy to parse and use location information in parse error messages. There are lots of examples in interwebs but as far as I can see none of them works on location-annotated tokens. Thanks.. --- Ömer Sinan Ağacan http://osa1.net

4 12

Re: [Haskell-cafe] GSOC Project Ideas
by Michel Kuhlmann 14 Mar '14

14 Mar '14

I would like to see gitit2 [1] replacing gitit [2]. gitit2 uses yesod as a foundation; currently not all features of gitit are implemented yet; despite from that, it is not up-to-date with current yesod (not compiling). The old gitit has quite some bugs; and it is not easy to extent it. -Michel [1]: https://github.com/jgm/gitit2.git [2]: https://github.com/jgm/gitit.git

2 1