July 2016 - Haskell-Cafe

RFC: revisiting the simplified haskell-src-exts AST
by Bertram Felgenhauer 30 Jul '16

30 Jul '16

Dear all, the latest major release of haskell-src-exts (version 1.18) removed the annotation free variant of the Haskell AST. It occurred to me that using ghc's pattern synonyms it is possible to offer a light-weight simplified view on the annotated AST. For example, Language.Haskell.Exts.Syntax defines data Name l = Ident l String | Symbol l String and we can provide a simplified view with all annotations equal to () by doing {-# LANGUAGE PatternSynonyms #-} import qualified Language.Haskell.Exts.Syntax as H type Name = H.Name () pattern Ident a = H.Ident () a pattern Symbol a = H.Symbol () a Because the type isn't changed, no code has to be reimplemented to work with this AST. A first rough cut at an implementation can be found on github [1]. Before releasing it on hackage (which requires some more polishing) I'd appreciate feedback on the approach. Would you consider using such a package? Is there a better approach? Also any thoughts on usability (e.g., how to approach documentation) would be welcome. Cheers, Bertram [1] https://github.com/int-e/haskell-src-exts-simple P.S. This grew out of an effort to update lambdabot to HSE-1.18, which made me feel awful about all the superfluous and ugly () in the resulting code. I also have some numbers showing that the annotation-free view makes updating the code smoother: Using the annotation-free view, the update touches 75 lines: src/Lambdabot/Plugin/Haskell/Eval.hs | 18 +- src/Lambdabot/Plugin/Haskell/Pointful.hs | 201 +++++++++++++++---------------- src/Lambdabot/Plugin/Haskell/Pretty.hs | 14 +- src/Lambdabot/Plugin/Haskell/UnMtl.hs | 62 ++++----- src/Lambdabot/Plugin/Haskell/Undo.hs | 115 ++++++++--------- 5 files changed, 72 insertions(+), 75 deletions(-) Using the haskell-src-exts 1.18 AST, the update touches 208 lines: src/Lambdabot/Plugin/Haskell/Pointful.hs | 87 ++++++++++++++----------------- src/Lambdabot/Plugin/Haskell/Pretty.hs | 12 ++-- src/Lambdabot/Plugin/Haskell/UnMtl.hs | 10 +-- src/Lambdabot/Plugin/Haskell/Undo.hs | 34 ++++++------ 5 files changed, 204 insertions(+), 206 deletions(-)

1 0

ANN: network-2.6.3.0
by evan＠evan-borden.com 30 Jul '16

30 Jul '16

Good day, I am pleased to announce the release of network 2.6.3.0. This release is primarily focused on stability, but includes a few new APIs. Thank you to the contributors for their continued excellence. https://hackage.haskell.org/package/network-2.6.3.0 Release Notes, Version 2.6.3.0: * New maintainers: Evan Borden (@eborden) and Kazu Yamamoto (@kazu-yamamoto). The maintainer for a long period, Johan Tibell (@tibbe) stepped down. Thank you, Johan, for your hard work for a long time. * New APIs: ntohl, htonl,hostAddressToTuple{,6} and tupleToHostAddress{,6}. [#210](https://github.com/haskell/network/pull/210) * Added a Read instance for PortNumber. [#145]( https://github.com/haskell/network/pull/145) * We only set the IPV6_V6ONLY flag to 0 for stream and datagram socket types, as opposed to all of them. This makes it possible to use ICMPv6. [#180](https://github.com/haskell/network/pull/180) [#181](https://github.com/haskell/network/pull/181) * Work around GHC bug #12020. Socket errors no longer cause segfaults or hangs on Windows. [#192](https://github.com/haskell/network/pull/192) * Various documentation improvements and the deprecated pragmas. [#186](https://github.com/haskell/network/pull/186) [#201](https://github.com/haskell/network/issues/201) [#205](https://github.com/haskell/network/pull/205) [#206](https://github.com/haskell/network/pull/206) [#211](https://github.com/haskell/network/issues/211) * Various internal improvements. [#193](https://github.com/haskell/network/pull/193) [#200](https://github.com/haskell/network/pull/200)

2 2

Re: [Haskell-cafe] Breaking in library, and getting stacktrace
by Carlos D'Agostino 29 Jul '16

29 Jul '16

Without knowing too much about your project setup, and if I'm understanding correctly, see http://haddock.stackage.org/lts-5.1/base-4.8.2.0/Debug-Trace.html#v:traceSt… Keep in mind that you have to build with the corresponding flags in order to get it working correctly. Assuming you're using stack that would be: stack build --enable-profiling --library-profiling --ghc-options="-fprof-auto-rtsopts" Though I'm not sure if that's 100% what you're looking for since it doesn't break as much as that it logs with a stack trace. Hope it helps! > ---------------------------------------------------------------------- > > Message: 1 > Date: Thu, 28 Jul 2016 16:43:03 +0300 > From: Constantine Kharlamov <hi-angel(a)ya.ru> > To: haskell-cafe(a)haskell.org > Subject: [Haskell-cafe] Breaking in library, and getting stacktrace > Message-ID: <147361469713383(a)web5j.yandex.ru> > Content-Type: text/plain; charset="utf-8" > > An HTML attachment was scrubbed... > URL: < > http://mail.haskell.org/pipermail/haskell-cafe/attachments/20160728/b4652af… > > > > ------------------------------ > > Subject: Digest Footer > > _______________________________________________ > Haskell-Cafe mailing list > Haskell-Cafe(a)haskell.org > http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe > > > ------------------------------ > > End of Haskell-Cafe Digest, Vol 155, Issue 33 > ********************************************* > -- Carlos D'Agostino.

1 0

Breaking in library, and getting stacktrace
by Constantine Kharlamov 28 Jul '16

28 Jul '16

1 0

Question, re: upgrading Stack manually.
by David Banas 27 Jul '16

27 Jul '16

Hi Stackers, Is this a show stopper, wrt/ upgrading Stack manually, via: - stack unpack stack-1.1.2 - pushed <directory created, above> - stack setup - stack build Warning: /Users/dbanas/Documents/Projects/stack/upgrade_2016-07-27/stack-1.1.2/stack.yaml: Unrecognized field in ProjectAndConfigMonoid: nix ? Thanks, -db

2 1

good choice for random number generator ?
by briand＠aracnet.com 27 Jul '16

27 Jul '16

There's quite a few. Many are very old. I would have like to use vector-random and got this : Data/Vector/Random/Mersenne.hs:33:18: Could not find module ‘Data.Vector.Fusion.Stream’ It is a member of the hidden package ‘vector-0.10.12.3@vecto_1COyUuV1LrA1IjYnWfJnbs’. Perhaps you need to add ‘vector’ to the build-depends in your .cabal file. Use -v to see a list of the files searched for. Data/Vector/Random/Mersenne.hs:35:18: Could not find module ‘Data.Vector.Fusion.Stream.Size’ It is a member of the hidden package ‘vector-0.10.12.3@vecto_1COyUuV1LrA1IjYnWfJnbs’. Perhaps you need to add ‘vector’ to the build-depends in your .cabal file. Use -v to see a list of the files searched for. I wouldn't be opposed to fixing it, but I'm wondering at this point there isn't a better package to use instead. I'm looking for both integer and floating point random numbers, uniform and gaussian. gsl-random looks promising as does mwc-random. any other suggestions ? Thanks, Brian

6 10

How do I debug this RTS segfault?
by Lana Black 26 Jul '16

26 Jul '16

Hello, I have run into this RTS bug recently. In short, when executing multiple consequtive forks, after 500-600 or so the process is terminated by SIGSEGV. I know this kind of thing is totally artificial, but still. The problem I have is that I can't get any meaningful backtrace in gdb. For example, for threaded RTS I get this (gdb) bt #0 0x0000000000560d63 in base_GHCziEventziThread_ensureIOManagerIsRunning1_info () Backtrace stopped: Cannot access memory at address 0x7fffff7fcea0 For non-threaded RTS I get this (gdb) bt #0 0x00000000007138c9 in stg_makeStablePtrzh () Backtrace stopped: Cannot access memory at address 0x7fffff7fc720 Build command: ghc --make -O2 -g -fforce-recomp fork.hs Add threaded if needed. I was able to reproduce this bug with both GHC 7.10.3 and todays HEAD with the code below. >import System.Exit (exitSuccess) >import System.Posix.Process (forkProcess) > >fork_ n | n > 0 = processPid =<< forkProcess (fork_ $! n - 1) > | otherwise = putStrLn "I'm done!" > >processPid pid | pid > 0 = exitSuccess > | pid < 0 = putStrLn "OOOPS, forkProcess failed!" > | otherwise = pure () > >main = fork_ 1000 > With best regards.

4 5

Forcing the kind in data
by Michael Roth 26 Jul '16

26 Jul '16

Hi, if I have: data Foobar a b = Foobar it has kind: * -> * -> * How can I force the kind to: (* -> *) -> * -> * ? Thank you!

3 2

Different behaviour with -XAllowAmbiguousTypes in 7.10.3b and 8.0.1
by Max Amanshauser 26 Jul '16

26 Jul '16

Hi, while porting a library to Haskell, which deals with persisting finite state automata to various stores, depending on the user's choice and the instances provided for types s(tate) e(vent) a(ction), I ran into different behaviour in ghc 7.10.3b and 8.0.1 related to ambiguity checks. This is a minimised and somewhat contrived example: > -- testme.hs > > {-# LANGUAGE MultiParamTypeClasses #-} > {-# LANGUAGE FlexibleInstances #-} > {-# LANGUAGE AllowAmbiguousTypes #-} > > module TestMe where > > data MyStore = MyStore > > class FSMStore st s e a where > fsmRead :: st -> String -> s > > instance (Num s) => FSMStore MyStore s e a where > fsmRead st i = 23 > > get :: (FSMStore st s e a) => st -> String -> s > get st i = fsmRead st i With GHC 8.0.1 I get: > *TestMe> :l test.hs > [1 of 1] Compiling TestMe ( test.hs, interpreted ) > > test.hs:19:12: error: > • Could not deduce (FSMStore st s e0 a0) > arising from a use of ‘fsmRead’ > from the context: FSMStore st s e a > bound by the type signature for: > get :: FSMStore st s e a => st -> String -> s > at test.hs:18:1-47 > The type variables ‘e0’, ‘a0’ are ambiguous > Relevant bindings include > st :: st (bound at test.hs:19:5) > get :: st -> String -> s (bound at test.hs:19:1) > These potential instance exist: > instance Num s => FSMStore MyStore s e a > -- Defined at test.hs:12:10 > • In the expression: fsmRead st i > In an equation for ‘get’: get st i = fsmRead st i > Failed, modules loaded: none. However, when I remove the type signature for get: > Prelude> :l test.hs > [1 of 1] Compiling TestMe ( test.hs, interpreted ) > Ok, modules loaded: TestMe. > *TestMe> :t get > get :: FSMStore st s e a => st -> String -> s > *TestMe> get MyStore "asdf" > 23 GHC inferred the exact same type I provided, but this time it compiles successfully. When going back to GHC 7.10.3b, without type signature: > GHCi, version 7.10.3: http://www.haskell.org/ghc/ :? for help > :Prelude> :l test.hs > [1 of 1] Compiling TestMe ( test.hs, interpreted ) > > test.hs:19:1: > Could not deduce (FSMStore st s e0 a0) > from the context (FSMStore st s e a) > bound by the inferred type for ‘get’: > FSMStore st s e a => st -> String -> s > at test.hs:19:1-23 > The type variables ‘e0’, ‘a0’ are ambiguous > When checking that ‘get’ has the inferred type > get :: forall st s e a. FSMStore st s e a => st -> String -> s > Probable cause: the inferred type is ambiguous > Failed, modules loaded: none. So my questions are: *) What's with the different behaviour depending on whether the type sig is inferred or provided? *) Is GHC 7.10.3b or 8.0.1 closer to the correct behaviour w.r.t. -XAllowAmbiguousTypes? -- Regards, Max Amanshauser.

2 1

Use GHC API in standalone executable?
by Phyx 25 Jul '16

25 Jul '16

I guess this depends on what you want to do. I've seen various level of GHC API usage in standalone applications, but it just depends on how much of the compiler pipeline you want to use. Their are ways to override the paths or provide stubs for programs ghc expects to be there. I've packaged API calls into shared libraries myself before, in that case I was mostly accessing the compiler frontend.

1 0