June 2005 - Glasgow-haskell-users

`tested-with' in package
by Serge D. Mechveliani 09 Jun '05

09 Jun '05

Who knows, please, what is the matter with the `tested-with' field in a package specification in ghc-6.4 ? Among the attempts tested-with: ghc-6.4, 5.02 tested-with: ghc-6.4 tested-with: GHC-6.4 tested-with: "GHC-6.4" tested-with: ghc tested-with: GHC only the last one satisfies the `configure' command runghc Setup.hs configire --ghc Other variants yield " Line 12: Parse of field 'tested-with' failed. " Thank you in advance for explanation. ----------------- Serge Mechveliani mechvel(a)botik.ru

1 0

How to open device as a IO Handle
by Lei Pan 09 Jun '05

09 Jun '05

Hi, there, Could anyone tell me how to open a block device (e.g., /dev/hda1) into a IO Handle by using GHC (version > 5.0)? I want to open a block/char device and return a handle. Apparently, openFile function can't do the job. It will be very appreciated if someone could give me a hint. An earlier post pointed out the problem, but got no reply. http://www.haskell.org/pipermail/glasgow-haskell-users/2001-November/002624… Something I played in GHCI-6.2.2 *Main> openFile "/dev/hdb2" ReadMode >>= \hin -> hGetPosn hin >>= \pos -> print pos *** Exception: /dev/hdb2: openFile: unsupported operation (unknown file type) (I already set /dev/hdb2 globally read/write) Cheers, Lei

1 0

impressions on ghc-6.4
by Serge D. Mechveliani 08 Jun '05

08 Jun '05

Dear GHC developers and users, Here are some impressions on ghc-6.4. Cabal ----- It looks good. Thanks to developers. Long ago I asked for the building tool in GHC which would allow to avoid composing Makefile-s, with their difficult syntax and odd-looking system commands: `ar', `ld', and such. Now, Cabal does what I asked for. Remarks: * maybe, `version' should not be mandatory in the package specification? * " version: 5.02 " (imagine "ghc-5.02") converts to "5.0.2" . Could it allow digit sequences separated by period? -O, memory management --------------------- In ghc-6.2.2 the test for dumatel-1.02 runs 2 times faster when Dumatel is built under -O. And in ghc-6.4, the time does not differ. Then, I tried the benchmark with certain mixture of list operations: user-written maximum and mergeSort, `reverse' and `zip' of GHC -- find enclosed the source. -O is 1.3 faster than -Onot. The speed looks like in ghc-5.02 (I think, this is all right). 5.02 was able to run `main' for n = 55000 in 200 Kb memory. 6.4 needs 2600 Kb. Probably, 5.02 is distinguished among other versions in memory management. Regards, ----------------- Serge Mechveliani mechvel(a)botik.ru *************************************************************** -- the result of `main' must be n-1. main = putStr $ shows (test 55000 ) "\n" -- n -- to edit test :: Int -> Int test n = max' $ sort' $ zipWith (-) (reverse xs) xs where xs = [1..n] max' :: Ord a => [a] -> a max' [x] = x max' (x:y:xs) = case compare x y of LT -> max' (y:xs) _ -> max' (x:xs) sort' :: Ord a => [a] -> [a] sort' xs = s $ mergePairs [[x] | x <- xs] where s [] = [] s [xs] = xs s xss = s $ mergePairs xss mergePairs (xs:ys:zss) = (merge xs ys):(mergePairs zss) mergePairs xss = xss merge :: Ord a => [a] -> [a] -> [a] merge [] ys = ys merge xs [] = xs merge (x:xs) (y:ys) = case compare x y of GT -> y:(merge (x:xs) ys) _ -> x:(merge xs (y:ys))

1 0

RE: GHC hang
by Simon Peyton-Jones 07 Jun '05

07 Jun '05

There is definitely a performance bug somewhere in GHC, and we're collecting cases that show it up. It'd be ideal if you, or someone else, could snip out a standalone module or two that demonstrates the problem, rather than all of WASH. If not, we'll try to tackle it directly. Simon | -----Original Message----- | From: glasgow-haskell-users-bounces(a)haskell.org [mailto:glasgow-haskell-users- | bounces(a)haskell.org] On Behalf Of John Goerzen | Sent: 06 June 2005 21:21 | To: glasgow-haskell-users(a)haskell.org | Subject: Re: GHC hang | | On 2005-06-06, Simon Peyton-Jones <simonpj(a)microsoft.com> wrote: | > Lots of people seem to have had problems compiling Wash with GHC 6.4. | | Another person told me, perhaps in private, that it just takes GHC 15 | minutes to compile the specific file in question on amd64. It turns out | that there is not a hang bug with this. | | There were various other bugs with Wash, and I think I have got it to at | least a buildable state. See my announcement on that topic in this list | for more. | | -- John | | _______________________________________________ | Glasgow-haskell-users mailing list | Glasgow-haskell-users(a)haskell.org | http://www.haskell.org/mailman/listinfo/glasgow-haskell-users

3 2

RE: ghc-pkg, recompilation check
by Simon Marlow 07 Jun '05

07 Jun '05

On 06 June 2005 15:42, Serge D. Mechveliani wrote: > And in 6.4, make dm > reports > ... > " Reading package info from stdin... done. > ghc-pkg: invalid package identifier: > " The syntax of package descriptions has changed. Documentation for the new format is here: http://www.haskell.org/ghc/docs/latest/html/users_guide/packages.html#in stalled-pkg-info > Packages always were difficult. And in 6.4 they are changed again ... I'd urge you to look at using Cabal for your library, I think it'll simplify things for you a lot. In particular, you won't have to deal with ghc-pkg and you won't have to write any Makefiles. Cabal is specifically designed to make life easier for people like you! I know changes are painful, but in this case they weren't gratuitous. All the changes were to support Cabal, and we believe this is the right direction to go. If it all pans out, we won't need to make any big changes again in the future. > Question 2: recompilation > ------------------------- > > Why does it recompile both modules when after the first make dm > we remove the .a library and repeat make dm > ? > This recompilation bites when there are many modules. Unfortunately your use of -ddump-hi-diffs has forced recompilation. This is a bug: we changed it so that all -ddump-* flags forced recompilation because in most cases some intermediate output has been requested. However, -ddump-hi-diffs should be an exception. Thanks for the report - just turn off -ddump-hi-diffs. Cheers, Simon

1 0

RE: Simplifier output explanation needed
by Simon Peyton-Jones 07 Jun '05

07 Jun '05

Yes $s is for specialised versions of (usually overloaded) functions. If you are using -02 you can also get specialised versions from so-called constructor specialisation, when the function is specialised for a particular argument pattern. I don't know why the strictness annotations disappear, but the specialised function probably has a different type, and different arguments, than the un-specialised one, so that might be part of it. This stuff is not documented, I'm afraid. I'll volunteer to explain what you are seeing (if you send me a small example) if you volunteer to write some documentation. Seriously. Simon | -----Original Message----- | From: glasgow-haskell-users-bounces(a)haskell.org [mailto:glasgow-haskell-users- | bounces(a)haskell.org] On Behalf Of Jens Fisseler | Sent: 01 June 2005 13:54 | To: glasgow-haskell-users(a)haskell.org | Subject: Simplifier output explanation needed | | Hi everbody! | | I'm currently trying to optimize some code, in particular trying to make | some functions as strict as possible. To do this, I quite often look for | strictness annotations either in the interface files or the the | simplifier output. Doing this, something odd occured to me: inserting a | '$!' into a function call, trying to enforce eager evaluation, the | simplifier output changed. Ok, I should expect this, but I don't | understand the change. The function name got '$s' as a prefix and the | strictness annotation ('Str: DmdType SSL') vanished. | | So my question is what all those different prefixes mean. I've stumbled | across at least '$s' and '$w' (worker wrapper?). Why did the strictness | annotation disappear? Is there any documentation that can enlighten me? | | Regards, | | Jens | _______________________________________________ | Glasgow-haskell-users mailing list | Glasgow-haskell-users(a)haskell.org | http://www.haskell.org/mailman/listinfo/glasgow-haskell-users

1 0

RE: GHC hang
by Simon Peyton-Jones 06 Jun '05

06 Jun '05

Lots of people seem to have had problems compiling Wash with GHC 6.4. As far as we know, they are all caused by the fact that the package-configuration file format changed in 6.4, and you need to fix that before it'll compile cleanly. Maybe someone can make available a version of Wash that works, rather than have everyone duplicate effort? (In GHC's defence, the package spec is still in flux, esp with Cabal coming on-stream. I know it's inconvenient to have changes.) However, the problems other people have been having are of the form "can't find interface file definition" rather than an indefinite hang. That's undoubtedly odd. You could tar up the complete tree and send it to us to see if we can reproduce what's going on. Certainly, GHC should never just hang. But it might be easier just to fix the package spec. I think the problem is that you need to list all the modules in a package, whereas you didn't before. Simon | -----Original Message----- | From: glasgow-haskell-users-bounces(a)haskell.org [mailto:glasgow-haskell-users- | bounces(a)haskell.org] On Behalf Of John Goerzen | Sent: 31 May 2005 21:36 | To: glasgow-haskell-users(a)haskell.org | Subject: GHC hang | | Hi, | | I'm trying to compile WashNGo 2.3.1 on my amd64 machine running GHC 6.4. | When compiling HTMLMonad98.hs, ghc hangs. It's been running over 5 | minutes now with no output or other activity. | | What can I do to help the GHC folks debug this? | | _______________________________________________ | Glasgow-haskell-users mailing list | Glasgow-haskell-users(a)haskell.org | http://www.haskell.org/mailman/listinfo/glasgow-haskell-users

3 2

ANN: Wash for GHC 6.4, Cabalized
by John Goerzen 06 Jun '05

06 Jun '05

Hello, I have made an, as far as I can tell, working copy of WASH for GHC 6.4. You can get it with: darcs get http://darcs.complete.org/local-branches/WashNGo Here are the basic things I did: * Applied the GHC 6.4 compatibility patch I found on one of these lists * Converted the build system to use Cabal. auto* is GONE. * Due to a limitation of Cabal, build only one package named WASH. The default Wash, even with the above compatibility patch, doesn't actually let me build programs that use it correctly. This one seems to work for me. Also, I have noted this in the README, but I couldn't figure out a way to run wash2hs automatically. You'll need to add -F -pgmF wash2hs in the GHC options for any file that needs wash2hs treatment. Problems? Please darcs send me patches. Thanks, -- John

1 0

ghc-pkg, recompilation check
by Serge D. Mechveliani 06 Jun '05

06 Jun '05

Dear GHC developers and users, Could you, please, help with ghc-pkg ? (in ghc-6.4) The contrived testing project is of the two small modules M1.hs, M2.hs. (find enclosed the source). `Makefile' is set so that make dm should compile the modules and put *.hi, *.o modules and .a, .o libraries to $(s)/export/. It also needs to create a package dm-1.0 (in 6.2.2 it was dm ). A particular command in Makefile is ghc-pkg -f $(s)/dm-1.0.conf -u -g , which (as I recall) makes the .o library from .a one. It worked in 6.2.2 -- with dm instead of dm-1.0, with `=' instead of `:' and with package_deps = ["data"] in the package specification. And in 6.4, make dm reports ... " Reading package info from stdin... done. ghc-pkg: invalid package identifier: " Packages always were difficult. And in 6.4 they are changed again ... Question 2: recompilation ------------------------- Why does it recompile both modules when after the first make dm we remove the .a library and repeat make dm ? This recompilation bites when there are many modules. Thank you in advance for explanation. ------------------ Serge Mechveliani mechvel(a)botik.ru ------------------------------------------------------------ module M1 -- file M1.hs where tuple31 :: (a, b, c) -> a tuple31 (x, _ ,_) = x --------------------------------- file M2.hs module M2 where import qualified Data.Set as Set (Set(..), singleton) import M1 (tuple31) f :: Set.Set Bool f = Set.singleton $ tuple31 (True, False, False) -- Makefile ------------------------------------------------------ ghcBinDir = /usr/bin s = /home/mechvel/ghc/6.4/pkgTest e = $(s)/export ghc = $(ghcBinDir)/ghc ghcpkg = $(ghcBinDir)/ghc-pkg RANLIB = ar -s HCFlags = -odir $(e) -hidir $(e) -ddump-hi-diffs pack = Package {name : ["\"dm-1.0\""], \ import_dirs : ["\"$(e)\""], \ source_dirs : [], \ library_dirs : ["\"$(e)\""], \ hs_libraries : ["\"HSdm\""], \ extra_libraries : [], \ include_dirs : [], \ c_includes : [], \ package_deps : [], \ extra_ghc_opts : [], \ extra_cc_opts : [], \ extra_ld_opts : [] } # of 6.2.2: name = ["\"dm\""], \ # package_deps = [\"data\"], \ obj: if [ ! -d $(e) ]; then mkdir $(e); fi $(ghc) $(HCFlags) --make M2 -package-name dm-1.0 dm: obj rm -f $(e)/libHSdm* $(e)/HSdm* ar -qc $(e)/libHSdm.a $(wildcard $(e)/*.o) $(RANLIB) $(e)/libHSdm.a $(RANLIB) $(e)/libHSdm.a echo $(pack) | $(ghcpkg) -f $(s)/dm-1.0.conf -u -g $(ghcpkg) -f $(s)/dm-1.0.conf -l ------------------------------------------------------------------ make dm yields ----------------------------------------------------------- if [ ! -d /home/mechvel/ghc/6.4/pkgTest/export ]; then mkdir /home/mechvel/ghc/6.4/pkgTest/export; fi /usr/bin/ghc -odir /home/mechvel/ghc/6.4/pkgTest/export -hidir /home/mechvel/ghc/6.4/pkgTest/export -ddump-hi-diffs --make M2 -package-name dm-1.0 Chasing modules from: M2 Source file changed or recompilation check turned off Compiling M1 ( ./M1.hs, /home/mechvel/ghc/6.4/pkgTest/export/M1.o ) No old interface file Source file changed or recompilation check turned off Compiling M2 ( M2.hs, /home/mechvel/ghc/6.4/pkgTest/export/M2.o ) No old interface file rm -f /home/mechvel/ghc/6.4/pkgTest/export/libHSdm* /home/mechvel/ghc/6.4/pkgTest/export/HSdm* ar -qc /home/mechvel/ghc/6.4/pkgTest/export/libHSdm.a /home/mechvel/ghc/6.4/pkgTest/export/M1.o /home/mechvel/ghc/6.4/pkgTest/export/M2.o ar -s /home/mechvel/ghc/6.4/pkgTest/export/libHSdm.a ar -s /home/mechvel/ghc/6.4/pkgTest/export/libHSdm.a echo Package {name : ["\"dm-1.0\""], import_dirs : ["\"/home/mechvel/ghc/6.4/pkgTest/export\""], source_dirs : [], library_dirs : ["\"/home/mechvel/ghc/6.4/pkgTest/export\""], hs_libraries : ["\"HSdm\""], extra_libraries : [], include_dirs : [], c_includes : [], package_deps : [], extra_ghc_opts : [], extra_cc_opts : [], extra_ld_opts : [] } | /usr/bin/ghc-pkg -f /home/mechvel/ghc/6.4/pkgTest/dm-1.0.conf -u -g Reading package info from stdin... done. ghc-pkg: invalid package identifier: make: *** [dm] Error 1 ----------------------------------------------------------

1 0

RE: recompilation check in 6.4
by Simon Marlow 06 Jun '05

06 Jun '05

Hi Serge, On 05 June 2005 11:58, Serge D. Mechveliani wrote: > for the first time compiles all the modules before Foo in > 50 seconds. > Then, I fix an error in Foo.hs and repeat "make dm". > > The options are set so that ghc-6.2.2 gets to Foo.hs in less > than 1 second when doing this repeated `make'. > It does consider each module before Foo and each time decides that > there is no need in recompilation. It keeps on going this way: > -------- > Considering whether compilation is required for Prelude0: > Skipping Prelude0 ( ./Prelude0.hs, /foo/source/export/Prelude0.o ) > -------- > and so on, by the scheme > "Considering ..XX -- {Checking usages .. --} Skipping XX". > > The compiler applies this way: > -------------------- > /foo/ghc/6.2.2/inst/bin/ghc -fglasgow-exts > -fallow-undecidable-instances > -fallow-overlapping-instances -fno-warn-overlapping-patterns > -fwarn-unused-binds -fwarn-unused-matches -fwarn-unused-imports > -idemotest > -odir /home/mechvel/dm/1.02/dm/source/export > -hidir /home/mechvel/dm/1.02/dm/source/export > -ddump-hi-diffs > --make DExport -package-name dm > -------------------- > > Now, when ghc-6.4 does such a repeated `make', it re-compiles > all the `previous' modules under the above options: I'm afraid I don't know what the problem might be. Probably GHC is looking in the wrong place for .hi files or .o files, or both. > --------- > Chasing modules from: DExport > Source file changed or recompilation check turned off > Compiling Prelude0 ( ./Prelude0.hs, /.../source/export/Prelude0.o ^^^^^^^ in particul, that looks very strange! Did you edit that by hand, or is that the verbatim GHC output? (copying the verbatim output is usually better, to avoid confusion). Cheers, Simon

1 0