- jhc - Haskell.org

darcs patch: add regression test cases for known bugs (and 2 more)
by John Meacham 01 Sep '09

01 Sep '09

Mon Aug 31 21:50:45 PDT 2009 John Meacham <john(a)repetae.net> * add regression test cases for known bugs Mon Aug 31 22:28:49 PDT 2009 John Meacham <john(a)repetae.net> * fix bug where Grin.DeadCode would sometimes rename some variables non-uniquely Mon Aug 31 23:37:47 PDT 2009 John Meacham <john(a)repetae.net> tagged 0.7.2

1 0

ANNOUNCE: jhc 0.7.2
by John Meacham 01 Sep '09

01 Sep '09

Hi, I am happy to announce a new development release of jhc, it mainly is a bug fix release, but it also re-enables and adds several new optimizations. Installation: http://repetae.net/computer/jhc/building.shtml A few of the changes: * added an --annotate-source option, which follows jhcs normal rules for reading code, but then just spits it back out, pre-processed and annotated with meta-info. this source is suitable for feeding to something like a documentation system or other code meta-tools. * --list-libraries is now in machine and human readable JSON/YAML format. * replaced the old quadratic grin simplifier with an improved linear time one. * case return unboxing, case merging, superfluous return grin optimizations added. * Some major bug fixes in the more advanced Grin NodeAnalysis, making it stable enough to enable by default. This enables the all so important apply/eval inlining as well as other important optimizations. See the effects with -dgrin-final :) John -- John Meacham - ⑆repetae.net⑆john⑈ - http://notanumber.net/

1 0

jhc-0.7.1, Segmentation fault
by gsan＠stillpsycho.net 01 Sep '09

01 Sep '09

Hi, This is my first time using jhc; I installed the jhc-0.7.1 tarball. Below program gives a segfault at runtime: > module Main (main) where > data Tree = Leaf | Branch [Tree] > mkTree :: Int -> Tree > mkTree 0 = Leaf > mkTree _ = Branch [mkTree 0] > leaves :: Tree -> Int > leaves Leaf = 1 > leaves (Branch ts) = sum $ map leaves ts > main :: IO () > main = print . leaves . mkTree $ 1 $ rm -R ~/.jhc/ && jhc -o jhctest Test.lhs && ./jhctest jhc 0.7.1 (0.7.0-13) <snip> Running: gcc '-std=gnu99' -D_GNU_SOURCE '-falign-functions=4' -ffast-math -Wshadow -Wextra -Wall -Wno-unused-parameter -o jhctest jhctest_code.c -DNDEBUG -O3 Segmentation fault $ uname -a Linux 2.6.30-gentoo-r4 #3 SMP x86_64 Intel(R) Core(TM)2 CPU 6300 @ 1.86GHz GenuineIntel GNU/Linux And with "-f debug -d c": $ ./jhctest jhctest: jhctest_code.c:809: promote: Assertion `!(((uintptr_t)(s)) & 0x1)' failed. Aborted It throws an exception: static inline wptr_t A_STD A_UNUSED A_HOT promote(sptr_t s) { assert(!ISLAZY(s)); // <- *HERE* assert(jhc_valid_whnf((wptr_t)s)); return (wptr_t)s; } -- Gökhan San

2 2

darcs patch: perform apply analysis to transform appl... (and 3 more)
by John Meacham 01 Sep '09

01 Sep '09

Mon Aug 31 03:07:51 PDT 2009 John Meacham <john(a)repetae.net> * perform apply analysis to transform applications into direct calls to functions Mon Aug 31 05:13:00 PDT 2009 John Meacham <john(a)repetae.net> * fix bug in UnionSolve that caused it to not always propegate changes caused by lte constraints properly Mon Aug 31 05:15:14 PDT 2009 John Meacham <john(a)repetae.net> * cleanups for node analysis Mon Aug 31 18:36:13 PDT 2009 John Meacham <john(a)repetae.net> * fix bug where constant partial applications would not propegate constraints properly

1 0

darcs patch: remove old quadratic grin simplifier in ... (and 9 more)
by John Meacham 31 Aug '09

31 Aug '09

Tue Aug 25 21:19:02 PDT 2009 John Meacham <john(a)repetae.net> * remove old quadratic grin simplifier in favor of newer linear(ish) one. re-enable several grin optimizations. Thu Aug 27 01:56:26 PDT 2009 John Meacham <john(a)repetae.net> * perform case return unboxing when on the simplifiers upwards pass to have it converge in fewer iterations Thu Aug 27 03:43:36 PDT 2009 John Meacham <john(a)repetae.net> * add new grin optimizatons, case merging and getting rid of superfluous returns Thu Aug 27 18:08:25 PDT 2009 John Meacham <john(a)repetae.net> * fix issue in return unboxing optimization that would sometimes re-create constants on the heap Thu Aug 27 21:17:12 PDT 2009 John Meacham <john(a)repetae.net> * remove old 'Store' grin operation in favor of the StoreNode and Demote basic ops Fri Aug 28 00:47:55 PDT 2009 John Meacham <john(a)repetae.net> * after node analysis, modify function bodies. Fri Aug 28 02:47:36 PDT 2009 John Meacham <john(a)repetae.net> * redo node analysis to rewrite function return values as well Fri Aug 28 04:17:51 PDT 2009 John Meacham <john(a)repetae.net> * add unboxing of demotions Fri Aug 28 07:29:43 PDT 2009 John Meacham <john(a)repetae.net> * add ability to modify arguments of suspsended functions to the node analyzer. Sat Aug 29 17:09:29 PDT 2009 John Meacham <john(a)repetae.net> * add -fdebug to regression tests

1 0

confusion on library source
by David Roundy 31 Aug '09

31 Aug '09

Hello, I'm still looking into jhc, and am a little confused in trying to locate library source. Actually, I have a couple of questions. First, is there any way to query jhc to find out which libraries (or packages) export a given module? I'm getting errors like: Error: Module not found: Control.Concurrent Error: Module not found: Data.Version Error: Module not found: System.Exit Error: Module not found: System.Process I'm sure some of these just aren't present at all (since I've only got the five libraries that come with jhc), but I can't see how to be certain. And if I get more libraries, the puzzle seems likely only to get trickier. Of course, this would also be useful for teaching franchise to compile using jhc, since I'd like franchise to be able to guess which packages to install, so I won't have to do this in the future. Secondly, I'm very puzzled in looking at the lib/ source, as the libraries seem to use modules that are nowhere defined. For instance: $ cat lib/haskell98/System.hs module System ( ExitCode(ExitSuccess,ExitFailure), getArgs, getProgName, getEnv, system, exitWith, exitFailure ) where import System.Exit import System.Environment import System.Cmd So it looks like System.Exit should exist somewhere. But I can't for the life of me (or the grep of me...) figure out where. :( I had been thinking that adding System.Exit would be an easy way to contribute a bit towards jhc, and at the same time learn a bit about how jhc works. Is System.Exit defined somewhere? If so, how can I convince jhc to use it in my code? I am very puzzled. And finally, I think I've got a bug report. If I specify -p haskell98 on the jhc command line, I get an error like: $ jhc -p haskell98 enfranchise.hs jhc -p haskell98 enfranchise.hs jhc 0.7.1 (0.7.1-0) Finding Dependencies... Using Ho Cache: '/mnt/home/droundy/.jhc/cache' Module 'Array' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'CPUTime' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'Char' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'Complex' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'Directory' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'IO' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'Ix' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'List' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'Locale' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'Maybe' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'Monad' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'Random' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'Ratio' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] Module 'Time' is exported by multiple libraries: ["haskell98-1.0","haskell98-1.0"] There were conflicting modules! Perhaps a simple nub would suffice? Is there ever a reason to import the same library twice? I suppose you could argue that this was my mistake, and jhc rightly rebuked me, but in that case a simpler error message would seem appropriate. -- David Roundy

2 3

Control.Exception
by Mark Wotton 30 Aug '09

30 Aug '09

Is there something inherent in JHC which prohibits using exceptions? I noticed the Control.Exception module is a bit anaemic - my use case here is to catch pattern match failures. (Yeah, I know it's an ugly idiom, but if I'm calling from Ruby, I just want an automatic exception that packages the error back up into the ruby runtime.) cheers Mark

3 4

parse error on where with jhc
by David Roundy 29 Aug '09

29 Aug '09

The following code gives a parse error on jhc: configureFlagWithDefault :: String -> String -> String -> C () -> (String -> C ()) -> C FranchiseFlag configureFlagWithDefault n argname h defaultaction j = do whenC amConfiguring $ addHook n defaultaction return $ FF $ Option [] [n] (ReqArg (addHook n . j') argname) h where j' v = do putV $ "handling configure flag --"++n++" "++v; j v Evidently, jhc wants the where to be less indented than the do block. I've no idea how this compares with the language standard, but it certainly differs from ghc, so I thought it'd be worth mentioning. Of course, it's not really a nice way to indent... -- David Roundy

3 2

fail message sent to stdout?
by David Roundy 29 Aug '09

29 Aug '09

Hi, It seems that under jhc, the from fail (called in the IO monad) is sent to stdout rather than stderr. Is this a policy choice... and is it open to debate? I discovered this because a test script was failing due to the assumption that fail would print to stderr. Admittedly, it isn't in the standard, but it makes sense to me that it ought to go to stderr. -- David Roundy

2 1

darcs patch: make pretty printing of grin a bit prettier (and 5 more)
by John Meacham 29 Aug '09

29 Aug '09

Thu Aug 27 02:34:58 PDT 2009 John Meacham <john(a)repetae.net> * make pretty printing of grin a bit prettier Thu Aug 27 03:42:54 PDT 2009 John Meacham <john(a)repetae.net> * minor library optimization to Numeric.roundTo Fri Aug 28 00:42:37 PDT 2009 John Meacham <john(a)repetae.net> * some clean ups Fri Aug 28 05:26:04 PDT 2009 John Meacham <john(a)repetae.net> * fix export-import test on 64 bit systems Fri Aug 28 07:29:17 PDT 2009 John Meacham <john(a)repetae.net> * make some minor changes to typechecking code based on profiler output Sat Aug 29 16:04:29 PDT 2009 John Meacham <john(a)repetae.net> * fix bug in UnionSolve that flipped meet and join for the toppable type

1 0