- Glasgow-haskell-users

can't manage to compile ghc6
by Isaac Jones 30 May '03

30 May '03

Greetings, I'm having trouble compiling ghc6. I have two versions of ghc installed at the moment, one is 5.04.2 from Debian, and another is a CVS snapshot from some time back that I've been using because it fixes a pet bug of mine. I can't seem to use either of these to build ghc6 from source on i386 linux. The failure seems to be occuring during the make boot stage. Is there any particular compiler I need to build ghc 6? See below for some compiling output. peace, isaac (no configure options except --prefix and --with-ghc for the two different versions) Using a somewhat old CVS version (I don't recall having trouble building this): ------------------------------------------------------------ /usr/local/bin/ghc -M -optdep-f -optdep.depend-BASE -osuf o -I../includes -H16m -O -iutils -ibasicTypes -itypes -ihsSyn -iprelude -irename -itypecheck -ideSugar -icoreSyn -ispecialise -isimplCore -istranal -istgSyn -isimplStg -icodeGen -iabsCSyn -imain -iprofiling -iparser -icprAnalysis -icompMan -indpFlatten -inativeGen -cpp -fglasgow-exts -Rghc-timing -I. -IcodeGen -InativeGen -Iparser -package concurrent -package posix -package util -recomp -Rghc-timing -H16M '-#include ... utils/UniqSet.lhs utils/Util.lhs main/DriverPipeline.hs:1008: warning: extra tokens at end of #endif directive main/SysTools.lhs: can't locate import `System.Posix.Internals' <<ghc: 83356908 bytes, 24 GCs, 80924/80924 avg/max bytes residency (1 samples), 16M in use, 0.01 INIT (0.00 elapsed), 0.50 MUT (2.41 elapsed), 0.07 GC (0.08 elapsed) :ghc>>make[2]: *** [depend] Error 1 make[1]: *** [boot] Error 1 make[1]: Leaving directory `/usr/local/ijones/usr/src/haskell/ghc/ghc-6.0/ghc' make: *** [build] Error 1 ------------------------------------------------------------ Using 5.04.3 from Debian: /usr/bin/ghc -M -optdep-f -optdep.depend-BASE -osuf o -I../includes -H16m -O -iutils -ibasicTypes -itypes -ihsSyn -iprelude -irename -itypecheck -ideSugar -icoreSyn -ispecialise -isimplCore -istranal -istgSyn -isimplStg -icodeGen -iabsCSyn -imain -iprofiling -iparser -icprAnalysis -icompMan -indpFlatten -inativeGen -cpp -fglasgow-exts -Rghc-timing -I. -IcodeGen -InativeGen -Iparser -package concurrent -package posix -package util -recomp -Rghc-timing -H16M '-#include ... nativeGen/MachCode.lhs:4404: warning: extra tokens at end of #endif directive nativeGen/MachCode.lhs:4460: warning: extra tokens at end of #endif directive nativeGen/MachCode.lhs:4509: warning: extra tokens at end of #endif directive nativeGen/MachMisc.lhs:80: unterminated argument list invoking macro "IF_ARCH_alpha" <<ghc: 325535444 bytes, 59 GCs, 122060/122060 avg/max bytes residency (1 samples), 16M in use, 0.01 INIT (0.00 elapsed), 1.66 MUT (4.81 elapsed), 0.08 GC (0.09 elapsed) :ghc>> make[2]: *** [depend] Error 1 make[1]: *** [boot] Error 1 make[1]: Leaving directory `/usr/local/ijones/usr/src/haskell/ghc/ghc-6.0/ghc' make: *** [build] Error 1

1 0

RE: ANNOUNCE: GHC version 6.0
by Simon Marlow 30 May '03

30 May '03

> There is a strange error message in ghc-pkg-6.0, viz. > > $ ghc-pkg-6.0 --list-packages > option `--list-packages' is ambiguous; could be one of: > -l --list-packages List packages in all config > files > -L --list-packages-local List packages in the > specified config file > > Hmm. Perhaps you should rename the long form of -L to > --list-local-packages. Done, thanks for the report. Cheers, Simon

1 0

ghc-split accepts building 6.0 on Windows
by Dominic Cooney 30 May '03

30 May '03

My environment's TEMP and TMP are C:\temp, but for some reason, my recent Cygwin install makes 'TEMP' and 'TMP' /cygdrive/c/temp/4; this duly gets passed to ghc-inplace as C:\temp\4. Then, in libraries\base, ghc-split complains that \4 mentions a non-existent group in a regexp: ../../ghc/compiler/ghc-inplace -H16m -O -fglasgow-exts -cpp -Iinclude -#include HsBase.h -Icbits/regex -funbox-strict-fields -package-name base -O -Rghc-timing -split-objs -c GHC/Base.lhs -o GHC/Base.o -ohi GHC/Base.hi Reference to nonexistent group in regex; marked by <-- HERE in m/"c:\temp\4 <--HERE \ghc936.split\.c"/ at d:\\users\\cooney\\ghc\\ghc-6.0\ghc\driver\split\ghc-split line 51, <TMPI> line 23. make[2]: *** [GHC/Base.o] Error 255 All I needed to do was editing ghc\driver\split\ghc-split so $Tmp_prefix = $ARGV[1] was instead ($Tmp_prefix = $ARGV[1]) =~ s/\\/\\\\/g; I don't know whether this habit of appending \4 is a delightful quirk of my Cygwin install, but FWIW for other Windows Server 2k3/Cygwin users out there... Dominic Cooney

1 0

RE: Ways and Build Tags for Optimisation
by Simon Marlow 30 May '03

30 May '03

> I'd like to include three different flavours of my HBase > package, with > these combinations of optimisation and profiling: > > release: optimised (with -O -fvia-C) > profiling: optimised and profiled > quick: neither > > The release and profiling flavours take forever to compile, so it's > useful to have a "quick" flavour to speed up the development > cycle. The > trouble is, any Haskell program using the release flavour of > HBase also > compiles very slowly, even if it itself does not use -O. If nothing > else, this may be due to the huge interface file for the root import: > > $ ls -Al HBase.*hi > -rw-r--r-- 1 ashley ashley 2117554 May 28 04:04 HBase.hi > -rw-r--r-- 1 ashley ashley 2119865 May 28 08:15 HBase.p_hi > -rw-r--r-- 1 ashley ashley 72669 May 28 16:20 HBase.q_hi Wow :-) There is some ancient support for "user-ways" in the compiler, but as far as I know it doesn't currently work. The only ways that the compiler knows about are the built-in ones (profiling, ticky, parallel etc.). I think the best way to achieve what you want is to have two completely separate packages, say hbase and hbase_q. The hbase package can have profiling support, but the hbase_q package need not. Cheers, Simon

2 1

GHC 6.0 Release: sparc-solaris2 binaries
by Hal Daume III 29 May '03

29 May '03

Greetings, A binary version of GHC6 is available for sparc-solaris2 machines at: http://www.isi.edu/~hdaume/ghc-6.0-sparc-solaris2.tar.bz2 17.5 mb I'd appreciate it if the maintainers could copy it and make it available locally off of the GHC web page so as to not kill bandwidth here :). Happy Haskelling! - Hal -- Hal Daume III | hdaume(a)isi.edu "Arrest this man, he talks in maths." | www.isi.edu/~hdaume

1 0

glasgow-haskell-users@haskell.org
by sms 29 May '03

29 May '03

1 0

Ways and Build Tags for Optimisation
by Ashley Yakeley 29 May '03

29 May '03

I'd like to include three different flavours of my HBase package, with these combinations of optimisation and profiling: release: optimised (with -O -fvia-C) profiling: optimised and profiled quick: neither The release and profiling flavours take forever to compile, so it's useful to have a "quick" flavour to speed up the development cycle. The trouble is, any Haskell program using the release flavour of HBase also compiles very slowly, even if it itself does not use -O. If nothing else, this may be due to the huge interface file for the root import: $ ls -Al HBase.*hi -rw-r--r-- 1 ashley ashley 2117554 May 28 04:04 HBase.hi -rw-r--r-- 1 ashley ashley 2119865 May 28 08:15 HBase.p_hi -rw-r--r-- 1 ashley ashley 72669 May 28 16:20 HBase.q_hi So what I'd like to do is include the "quick" version with suffices .q_hi and .q_o within the "hbase" package, just as profiling has .p_hi and .p_o. However, simply using -hisuf and -osuf seems not to be sufficient to force GHC to use those suffices when looking for files within HBase. I looked in fptools/ghc/compiler/main/DriverState.hs to see how GHC does this sort of thing for profiling: the file speaks of "ways" and "build tags" but I can't find any reference to these in the GHC manual. How would I use one of the "user ways"? -- Ashley Yakeley, Seattle WA

1 0

Why is this program leaking memory ?
by Ahn Ki-yung 29 May '03

29 May '03

I'm playing with the lazy abstract machine of 1997 Sestoft's paper. I implemented this with haskell using ghc 5.04.3. I lambda lifted the original input expression to prevent memory leak of "Lazy Abstarc Machine", and it works fine. I tested with the leaky program example of the Sestoft 97 paper. [kyagrd@OBIWAN transform]$ cat test.txt let ff = \n.let i=\x.x in ff i in ff ff I printed the 4 tuples (Heap, Exp, Env, Stack) each step. It goes on and on like this. [kyagrd@OBIWAN transform]$ ./main.exe < test.txt let "ff"=\"n".let "i"=\"x"."x" in "ff" "i" in "ff" "ff" let 1=\2.let 3=\4.4 in 1 3 in 1 1 let -1=\2.-1 -3 -3=\4.4 in -1 -1 let -1=\2.-1 -3 -3=\4.4 in -1 -1 let 0=\0.1 2 0=\0.0 in 0 0 ([],let 0=\0.1 2 0=\0.0 in 0 0,[],[]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],0 0,[2,1],[]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],0,[2,1],[2]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],\0.1 2,[2,1],[2]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],1 2,[2,2,1],[]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],1,[2,2,1],[1]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],\0.1 2,[2,1],[1]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],1 2,[1,2,1],[]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],1,[1,2,1],[1]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],\0.1 2,[2,1],[1]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],1 2,[1,2,1],[]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],1,[1,2,1],[1]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],\0.1 2,[2,1],[1]) ([(2,(\0.1 2,[2,1])),(1,(\0.0,[2,1]))],1 2,[1,2,1],[]) ... But Strangely the Haskell heap memory leaks if I omit printing every step but only print the result. [kyagrd@OBIWAN transform]$ ./main.exe < test.txt let "ff"=\"n".let "i"=\"x"."x" in "ff" "i" in "ff" "ff" let 1=\2.let 3=\4.4 in 1 3 in 1 1 let -1=\2.-1 -3 -3=\4.4 in -1 -1 let -1=\2.-1 -3 -3=\4.4 in -1 -1 let 0=\0.1 2 0=\0.0 in 0 0 c:\MyDoc\iFolder\kyagrd\LAZY\transform\main.exe: fatal error: RTS exhausted max heap size (268435456 bytes) I attach my source code except for the parser and lexer stuff. I switched between two main' and main fucntion. Why, in the Main module, "printNreduce" do not leak but while "printeval" leaks ? Can't understand this behavior. module Main where import Syntax import Parser import Lazyeval import MonadST printeval q = if q'==q then print q' else printeval q' where q' = reduce q printNreduce q = if q'==q then print q' else (print q >> printNreduce q') where q' = reduce q -- print every step does not leak main = do s <- getContents let e = parser s e' <- printNpreprocess e printNreduce ([],e',[],[]) -- print only result leak !! main' = do s <- getContents let e = parser s e' <- printNpreprocess e printeval ([],e',[],[]) printNpreprocess e = print e >> print e1 >> print e2 >> print e3 >> print e' >> return e' where ((ns,[]),e1) = evalST ([1..],[]) (uniqueify e) e2 = lambdalift negate e1 (ns3,e3) = evalST ns (normalize e2) e' = bruijnize e3 {- Syntax.hs preprocessing of lazy language for evaluation Ahn Ki-yung -} module Syntax where import MonadST import List data Lambda id = Var id | App (Lambda id) (Lambda id) | Lam id (Lambda id) | Let [(id,Lambda id)] (Lambda id) -- deriving (Eq, Ord) deriving (Eq, Ord, Read) instance Show a => Show (Lambda a) where show (Var s) = show s show (Lam s e) = '\\':show s ++ '.':show e show (App e e') = showParenExpr e ++ ' ' : showParenExpr e' where showParenExpr e@(Var s) = show e showParenExpr e = '(':show e++")" show (Let h e) = "let" ++ concat [' ':show s++'=':show e|(s,e)<-h] ++ " in " ++ show e instance Functor Lambda where fmap f (Var x) = Var (f x) fmap f (App e e') = App (fmap f e) (fmap f e') fmap f (Lam x e) = Lam (f x) (fmap f e) fmap f (Let ds e) = Let [(f x,fmap f e)|(x,e)<-ds] (fmap f e) transform (getId,putId,popId) = trans where newId x = putId x >> getId x trans (Var x) = do { x'<-getId x; return (Var x') } trans (App e e1) = do { e'<-trans e; e1'<-trans e1; return (App e' e1') } trans (Lam x e) = do { x'<-newId x; e'<-trans e; popId x; return (Lam x' e') } trans (Let ds e) = do let (xs,es) = unzip ds xs'<-mapM newId xs es'<-mapM trans es let ds' = zip xs' es' e'<-trans e mapM popId xs return (Let ds' e') uniqueify :: (Eq a, Eq b) => Lambda a -> StateTrans ([b],[(a,b)]) (Lambda b) uniqueify = transform (getId,putId,popId) where getId x = do (_,l) <- readST let Just (_,n) = find ((x==).fst) l return n putId x = do (n:ns,l) <- readST writeST (ns,(x,n):l) popId x = do (ns,l) <- readST writeST (ns,deleteBy (\p-> \q->fst p==fst q) (x,head ns) l) normalize :: Lambda a -> StateTrans [a] (Lambda a) normalize = normExpr where newvar = do (x:xs) <- readST writeST xs return x normExpr (Var x) = return (Var x) normExpr (Lam x e) = do e' <- normExpr e return (Lam x e') normExpr (App e (Var x)) = do e' <- normExpr e return (App e' (Var x)) normExpr (App e1 e2) = do x <- newvar e1' <- normExpr e1 e2' <- normExpr e2 return (Let [(x,e2')] (App e1' (Var x))) normExpr (Let ds e) = do let (vs,es) = unzip ds es' <- mapM normExpr es let ds' = zip vs es' e' <- normExpr e return (Let ds' e') (f,g)@@(x,y) = (f x, g y) -- assumes uniquified normalized e lambdalift topv e = Let ds (subsfree [] topv d' e') where d' = delcombs [] d ds = [ (topv x, foldr Lam (subsfree fv topv d' e) fv) | (x,fv,e)<-d' ] (d,fv,e') = llift e delcombs combs d = if combs/=combs' then delcombs combs' [(x,fv\\combs',e) | (x,fv,e)<-d] else d where combs' = [x | (x,[],_)<-d] -- assumes uniquified and no lets -- done llift subsfree bv topv d e = subs e where var' = Var . topv subs (Var x) = case find (\(y,_,_)->x==y && not (elem x bv)) d of Just (_,fv,_) -> foldl App (var' x) (map var' fv) _ -> Var x subs (App e e') = App (subs e) (subs e') subs (Lam x e) = Lam x (subs e) -- assumes uniquified llift (Var x) = ([], [x], Var x) llift (App e e') = (d1++d2, union fv1 fv2, App e1 e2) where (d1,fv1,e1) = llift e (d2,fv2,e2) = llift e' llift (Lam x e) = (d, fv\\[x], Lam x e') where (d, fv, e') = llift e llift (Let ds e) = (dd++d', fv, e') where (d, fv, e') = llift e xds = map ((id,llift)@@) ds dd = [ (x, fv\\[x], e) | (x,(d,fv,e))<-xds ] d' = foldr (++) d [d | (_,(d,_,_))<-xds] bruijnize e = bruijn [] e elemIndex' x xs = (\(Just i)->i) $ elemIndex x xs bruijn xs (Var x) = Var (elemIndex' x xs) bruijn xs (Lam x e) = Lam 0 (bruijn (x:xs) e) bruijn xs (App e e') = App (bruijn xs e) (bruijn xs e') bruijn xs (Let ds e) = Let [(0,bruijn xs' e) | e<-es] (bruijn xs' e) where (vs,es) = unzip ds xs' = vs ++ xs module Lazyeval where import Syntax import List data StackElem = Update Int | Point Int deriving Eq instance Show StackElem where show (Update i) = '#':show i show (Point i) = show i getHeap p = (\(Just x)->snd x) . find ((p==).fst) setHeap t@(p,e) = insertBy (mapF2 fst $ flip compare) t . deleteBy (mapF2 fst (==)) t where mapF2 g f2 x y = f2 (g x) (g y) reduce (h,App e (Var i),env,s) = (h,e,env,Point(env!!i):s) reduce (h,Lam _ e,env,Point p:s) = (h,e,p:env,s) reduce (h,Var i,env,s) = (h,e',env',s') where s' = case e' of Lam _ _ ->s; _->Update p:s (e',env') = getHeap p h p = env!!i reduce (h,Lam x e,env,Update p:s) = (setHeap (p,(Lam x e,env)) h,Lam x e,env,s) reduce (h,Let ds e,env,s) = (h',e,env',s) where es = snd (unzip ds) h' = newhs ++ h env' = newps ++ env newhs = zip newps [(e,env')|e<-es] newps = take n [m+n,m+n-1..] m = if null h then 0 else (head . fst . unzip) h n = length es reduce q = q fix f x = if x'==x then x else fix f x' where x' = f x eval = fix reduce module MonadST (readST, writeST, applyST, evalST, valueST, stateST, StateTrans) where data StateTrans s a = ST { st :: s -> (s,a) } instance Monad (StateTrans a) where return x = ST (\s -> (s, x)) m >>= f = ST (\s -> let (s',x) = st m s in st (f x) s') readST = ST (\s -> (s, s)) writeST s' = ST (\s -> (s', ())) applyST f = ST (\s -> (f s, ())) evalST s m = st m s valueST s = snd . evalST s stateST s = fst . evalST s

1 0

hsc2hs feature request
by Ross Paterson 28 May '03

28 May '03

The hsc2hs way of handling header files does not fit well with what is described by the FFI spec, making it difficult for non-GHC implementations to use .hsc files. Typical .hsc files may contain several #include lines, which turn into OPTIONS -#include pragmas. The portable way is to specify the header file in the foreign import, but only one header file is permitted there. Now if #def is used, hsc2hs generates a header file (as well as a C file) that would be suitable for this purpose. But #def is deprecated. So I propose that hsc2hs always generate this header file, as per the accompanying simple patch. That leaves the task of adding the header file to all the foreign imports, which could be done manually or automatically, but at least there would be a single header file to name. Index: ghc/utils/hsc2hs/Main.hs =================================================================== RCS file: /home/cvs/root/fptools/ghc/utils/hsc2hs/Main.hs,v retrieving revision 1.47 diff -u -r1.47 Main.hs --- ghc/utils/hsc2hs/Main.hs 20 May 2003 11:07:54 -0000 1.47 +++ ghc/utils/hsc2hs/Main.hs 28 May 2003 12:14:53 -0000 @@ -492,7 +492,7 @@ let specials = [(pos, key, arg) | Special pos key arg <- toks] let needsC = any (\(_, key, _) -> key == "def") specials - needsH = needsC + needsH = True let includeGuard = map fixChar outHName where

1 0

RE: Problems building ghc-current
by Simon Marlow 28 May '03

28 May '03

> The problem is that >ghc<, >green-card<, >hood< et al. are > subdirectories of >fptools<, that's why >cvs update -d< retrieves them > automatically. I guess a more appropriate repository layout would be > _flat_, like this: > > /cvs --+ > | > +-- CVSROOT > | > +-- fptools > | > +-- ghc > | > +-- libraries > | > +-- hslibs > | > +-- green-card > [...] Hmm, that sounds like a very good idea. I think perhaps we were fixated on having the same repo layout as the source tree when we originally checked everything into CVS. We could just rearrange the repository in place, and everything would still work, right? Well, perhaps the script that generates the CVS commit logs would have to be tweaked. Cheers, Simon

3 3