September 2009 - jhc - Haskell.org

darcs patch: make type of System.Info.compilerVersion match that of...
by David Roundy 02 Sep '09

02 Sep '09

Hi John, I don't like that ghc changed (somewhere around ghc 6) the type of compilerVersion, but it does seem best for jhc to match it, which means adding the Data.Version module. David Wed Sep 2 08:44:18 EDT 2009 David Roundy <droundy(a)darcs.net> * make type of System.Info.compilerVersion match that of ghc.

1 0

feature requests for franchise's benefit
by David Roundy 02 Sep '09

02 Sep '09

I've been looking into how to teach franchise to compile using jhc, and there are a few features that would make this much easier. The first is that I'd like a way to ask jhc what files a given compile will depend on. Something like gcc -M, although I don't care to have make syntax in the output. I don't really care about the format of the output, so long as it's easy to extract a list of files from it. Ideally, this would output *all* dependencies, including installed .hl files. I'm imagining this is the sort of output that could (hypothetically) be extracted from strace run on a successful compile. Incidentally, I'm thinking of enfranchise the libraries that come with jhc, which would allow them to be rebuilt automatically when the source is changed, rather than the current scheme of "rm *.hl; make libs". The above won't be so helpful for unsuccessful compiles, as will happen if certain modules are missing (e.g. they need to be generated from a .hsc file), so a second feature request would be to have a way of finding out from jhc what modules it can't find. These modules might either be generated files (e.g. from a .hs.in file or a .hsc file, or a user-written rule), or they might be defined in a package that needs to be included. For ghc, I get this information by parsing its error messages. It's not a very good solution, but works. I'd like a better solution with jhc, perhaps a flag to specify a file to output missing modules to? And finally, a nice way to find out which modules are exported by which packages. Perhaps something as simple as a --show-exports flag: jhc --no-auto -pbase-1.0 --show-exports Another interesting option for this feature would be to allow --show-ho to accept a package name such as base-1.0 and look up its .hl file using the same path that -p would. Then if the output of --show-ho were stable (and parsable), that would be sufficient. -- David Roundy

1 0

Fwd: [Haskell-cafe] Slow IO?
by Gwern Branwen 02 Sep '09

02 Sep '09

---------- Forwarded message ---------- From: Mark Wotton <mwotton(a)gmail.com> Date: Wed, Sep 2, 2009 at 1:08 AM Subject: Re: [Haskell-cafe] Slow IO? To: Eugene Kirpichov <ekirpichov(a)gmail.com> Cc: Steve <stevech1097(a)yahoo.com.au>, haskell-cafe(a)haskell.org On 02/09/2009, at 2:26 PM, Eugene Kirpichov wrote: >> > > I've got a Centrino Duo 2000 (I'm on a notebook), Ubuntu 9.04 and ghc 6.10.2. > > However, we have not set up on what exact input file we're using :) > I'm using one where it is written "10000000 3" and then 10000000 lines > of "999999999" follow. > > Also, I wonder what one'd get if one compiled this program with jhc, > but I don't know whether jhc is able to compile Data.ByteString. It couldn't last time I tried - choked on some INLINE pragmas. Might not be a massive job, but there aren't enough hours in the day... mark _______________________________________________ Haskell-Cafe mailing list Haskell-Cafe(a)haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe -- gwern

1 0

darcs patch: fix bug when 'do' expressions used with infix expressi...
by John Meacham 01 Sep '09

01 Sep '09

Tue Sep 1 11:08:31 PDT 2009 John Meacham <john(a)repetae.net> * fix bug when 'do' expressions used with infix expressions, add test case to regression

1 0

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