July 2002 - Hugs-Bugs

OSX
by Michel Schellekens 25 Oct '02

25 Oct '02

After downloading the OSX version, and typing "hugs" to start up the language, I obtain the following error: [schellekens:~] mpcs% hugs dyld: hugs can't open library: /usr/lib/libncurses.5.dylib (No such file or directory, errno = 2) Any suggestions? Michel.

2 1

FFI for SPARC [was: FFI testers wanted]
by Sven Panne 03 Aug '02

03 Aug '02

Using cut-n-paste from GHC's sources (esp. Adjustor.c) quite generously, I've hacked together the missing FFI parts for SPARC. Attached is a torture test, too, which is again based on GHC stuff (ffi009.hs). Cheers, S. diff -r -u hugs98-ffi-09072002c.orig/src/builtin.c hugs98-ffi-09072002c/src/builtin.c --- hugs98-ffi-09072002c.orig/src/builtin.c Tue Jul 9 13:05:26 2002 +++ hugs98-ffi-09072002c/src/builtin.c Tue Jul 23 13:04:30 2002 @@ -2239,6 +2239,8 @@ HugsStablePtr stable; #if defined(__ppc__) char code[13*4]; +#elif defined(__sparc__) && defined(__GNUC__) + char code[44]; #else char code[16]; #endif @@ -2324,6 +2326,65 @@ } __asm__ volatile ("sync\n\tisync"); } + } +#elif defined(__sparc__) && defined(__GNUC__) + /* Mostly cut-n-pasted from GHC's Adjustor.c: + + <00>: 9C23A008 sub %sp, 8, %sp ! make room for %o4/%o5 in caller's frame + <04>: DA23A060 st %o5, [%sp + 96] ! shift registers by 2 positions + <08>: D823A05C st %o4, [%sp + 92] + <0C>: 9A10000B mov %o3, %o5 + <10>: 9810000A mov %o2, %o4 + <14>: 96100009 mov %o1, %o3 + <18>: 94100008 mov %o0, %o2 + <1C>: 13000000 sethi %hi(app), %o1 ! load up app (1 of 2) + <20>: 11000000 sethi %hi(s), %o0 ! load up s (1 of 2) + <24>: 81C26000 jmp %o1 + %lo(app) ! jump to app (load 2 of 2) + <28>: 90122000 or %o0, %lo(), %o0 ! load up s (2 of 2, delay slot) + + ccall'ing on SPARC is easy, because we are quite lucky to push a + multiple of 8 bytes (1 word stable pointer + 1 word dummy arg) in front + of the existing arguments (note that %sp must stay double-word aligned at + all times, see ABI spec at http://www.sparc.org/standards/psABI3rd.pdf) + To do this, we extend the *caller's* stack frame by 2 words and shift + the output registers used for argument passing (%o0 - %o5, we are a + *leaf* procedure because of the tail-jump) by 2 positions. This makes + room in %o0 and %o1 for the additinal arguments, namely the stable + pointer and a dummy (used for destination addr of jump on SPARC). This + shouldn't cause any problems for a C-like caller: alloca is implemented + similarly, and local variables should be accessed via %fp, not %sp. In + a nutshell: This should work! (Famous last words! :-) + */ + { + unsigned long *adj_code = (unsigned long *)pc; + adj_code[ 0] = 0x9C23A008UL; /* sub %sp, 8, %sp */ + adj_code[ 1] = 0xDA23A060UL; /* st %o5, [%sp + 96] */ + adj_code[ 2] = 0xD823A05CUL; /* st %o4, [%sp + 92] */ + adj_code[ 3] = 0x9A10000BUL; /* mov %o3, %o5 */ + adj_code[ 4] = 0x9810000AUL; /* mov %o2, %o4 */ + adj_code[ 5] = 0x96100009UL; /* mov %o1, %o3 */ + adj_code[ 6] = 0x94100008UL; /* mov %o0, %o2 */ + adj_code[ 7] = 0x13000000UL; /* sethi %hi(app), %o1 */ + adj_code[ 7] |= ((unsigned long)app) >> 10; + adj_code[ 8] = 0x11000000UL; /* sethi %hi(s), %o0 */ + adj_code[ 8] |= ((unsigned long)s) >> 10; + adj_code[ 9] = 0x81C26000UL; /* jmp %o1 + %lo(app) */ + adj_code[ 9] |= ((unsigned long)app) & 0x000003FFUL; + adj_code[10] = 0x90122000UL; /* or %o0, %lo(s), %o0 */ + adj_code[10] |= ((unsigned long)s) & 0x000003FFUL; + + /* flush cache */ + asm("flush %0" : : "r" (adj_code )); + asm("flush %0" : : "r" (adj_code + 2)); + asm("flush %0" : : "r" (adj_code + 4)); + asm("flush %0" : : "r" (adj_code + 6)); + asm("flush %0" : : "r" (adj_code + 10)); + + /* max. 5 instructions latency, and we need at >= 1 for returning */ + asm("nop"); + asm("nop"); + asm("nop"); + asm("nop"); } #else ERRMSG(0) "Foreign import wrapper is not supported on this architecture" diff -r -u hugs98-ffi-09072002c.orig/src/ffi.c hugs98-ffi-09072002c/src/ffi.c --- hugs98-ffi-09072002c.orig/src/ffi.c Sat Jul 6 12:52:00 2002 +++ hugs98-ffi-09072002c/src/ffi.c Tue Jul 23 11:54:05 2002 @@ -412,7 +412,13 @@ } fprintf(out,"("); if (extraArg) { +#ifdef __sparc__ + /* On SPARC we need an additional dummy argument due to stack alignment + restrictions, see the comment in mkThunk in builtin.c. */ + fprintf(out,"HugsStablePtr fun1, void* unusedArg"); +#else fprintf(out,"HugsStablePtr fun1"); +#endif if (nonNull(argTys)) { fprintf(out,", "); } import Foreign import Random -------------------------------------------------------------------------------- foreign import ccall "dynamic" callFun5I :: FunPtr (Int -> Int -> Int -> Int -> Int -> Int) -> (Int -> Int -> Int -> Int -> Int -> Int) foreign import ccall "wrapper" mkFun5I :: (Int -> Int -> Int -> Int -> Int -> Int) -> IO (FunPtr (Int -> Int -> Int -> Int -> Int -> Int)) manyArgs5I :: (Int -> Int -> Int -> Int -> Int -> Int) manyArgs5I a1 a2 a3 a4 a5 = (((a1 * 31 + a2) * 31 + a3) * 31 + a4) * 31 + a5 test5I :: IO () test5I = do a1 <- randomIO a2 <- randomIO a3 <- randomIO a4 <- randomIO a5 <- randomIO funAddr <- mkFun5I manyArgs5I print (callFun5I funAddr a1 a2 a3 a4 a5 == manyArgs5I a1 a2 a3 a4 a5) freeHaskellFunPtr funAddr -------------------------------------------------------------------------------- foreign import ccall "dynamic" callFun6D :: FunPtr (Double -> Double -> Double -> Double -> Double -> Double -> Double) -> (Double -> Double -> Double -> Double -> Double -> Double -> Double) foreign import ccall "wrapper" mkFun6D :: (Double -> Double -> Double -> Double -> Double -> Double -> Double) -> IO (FunPtr (Double -> Double -> Double -> Double -> Double -> Double -> Double)) manyArgs6D :: Double -> Double -> Double -> Double -> Double -> Double -> Double manyArgs6D a1 a2 a3 a4 a5 a6 = ((((a1 * 31 + a2) * 31 + a3) * 31 + a4) * 31 + a5) * 31 + a6 test6D :: IO () test6D = do a1 <- randomIO a2 <- randomIO a3 <- randomIO a4 <- randomIO a5 <- randomIO a6 <- randomIO funAddr <- mkFun6D manyArgs6D print (callFun6D funAddr a1 a2 a3 a4 a5 a6 == manyArgs6D a1 a2 a3 a4 a5 a6) freeHaskellFunPtr funAddr -------------------------------------------------------------------------------- foreign import ccall "dynamic" callFun11M :: FunPtr (Int -> Double -> Float -> Char -> Int -> Int -> Float -> Int -> Char -> Double -> Int -> Double) -> (Int -> Double -> Float -> Char -> Int -> Int -> Float -> Int -> Char -> Double -> Int -> Double) foreign import ccall "wrapper" mkFun11M :: (Int -> Double -> Float -> Char -> Int -> Int -> Float -> Int -> Char -> Double -> Int -> Double) -> IO (FunPtr (Int -> Double -> Float -> Char -> Int -> Int -> Float -> Int -> Char -> Double -> Int -> Double)) manyArgs11M :: Int -> Double -> Float -> Char -> Int -> Int -> Float -> Int -> Char -> Double -> Int -> Double manyArgs11M a1 a2 a3 a4 a5 a6 a7 a8 a9 a10 a11 = (((((((((fromIntegral a1 * 31 + a2) * 31 + realToFrac a3) * 31 + fromIntegral (fromEnum a4)) * 31 + fromIntegral a5) * 31 + fromIntegral a6) * 31 + realToFrac a7) * 31 + fromIntegral a8) * 31 + fromIntegral (fromEnum a9)) * 31 + a10) * 31 + fromIntegral a11 test11M :: IO () test11M = do a1 <- randomIO a2 <- randomIO a3 <- randomIO a4 <- randomIO a5 <- randomIO a6 <- randomIO a7 <- randomIO a8 <- randomIO a9 <- randomIO a10 <- randomIO a11 <- randomIO funAddr <- mkFun11M manyArgs11M print (callFun11M funAddr a1 a2 a3 a4 a5 a6 a7 a8 a9 a10 a11 == manyArgs11M a1 a2 a3 a4 a5 a6 a7 a8 a9 a10 a11) freeHaskellFunPtr funAddr -------------------------------------------------------------------------------- rep :: String -> IO () -> IO () rep msg tst = do putStrLn ("Testing " ++ msg ++ "...") sequence_ (replicate 10 tst) main :: IO () main = do setStdGen (mkStdGen 4711) rep "5 Int arguments" test5I rep "6 Double arguments" test6D rep "11 mixed arguments" test11M

2 1

(no subject)
by TewCottage＠aol.com 31 Jul '02

31 Jul '02

Dear Sir/Madam Please could you give me the web address of where I would be able to find Haskell 1.3. Yours faithfully Andrew McLennan

2 1

Your message to Hugs-Users awaits moderator approval
by hugs-users-admin＠haskell.org 30 Jul '02

30 Jul '02

Your mail to 'Hugs-Users' with the subject A special funny website Is being held until the list moderator can review it for approval. The reason it is being held: Message body is too big: 130909 bytes but there's a limit of 40 KB Either the message will get posted to the list, or you will receive notification of the moderator's decision.

1 0

round doesn't seem to work correctly
by Martin Erwig 30 Jul '02

30 Jul '02

Hi, I am using Hugs (Version: December 2001) on Unix (Solaris). I think that round might not be working correctly for numbers that end with "2.5". For example, Prelude> round 2.5 2 Prelude> round 72.500 72 Prelude> round 11112.5 11112 Thanks, Martin

2 1

input.c: nextLine EOF bug + patch
by Sven M. Hallberg 29 Jul '02

29 Jul '02

Moin, I've just stumbled across a bug in Hugs98 Dec2001 on my PowerPC running Debian GNU/Linux. The bug is in src/input.c, line 516: arctic:~/build/deb/hugs98-98.200112/src% make input.o gcc -c -g -O2 input.c input.c: In function `nextLine': input.c:516: warning: comparison is always false due to limited range of data type arctic:~/build/deb/hugs98-98.200112/src% gcc --version 2.95.4 The comparison in question is the check for EOF, which is #defined as -1 in my stdio.h. This sent my Hugs into an infinite loop when loading the Prelude. The attached patch fixes the problem by storing the result of fgetc() in an int and using that for the comparison with EOF. Happy hacking, Sven Moritz

1 0

Forall quantified record fields.
by Ulf Norell 29 Jul '02

29 Jul '02

Hugs will produce funny looking results when using forall quantified fields in records. For instance newtype Foo = Foo { foo :: forall a. a -> a } Main> foo (Foo {foo = id}) 42 Foo_Foo id 42 Main> $$ + 1 INTERNAL ERROR: Bignum expected Main> foo (Foo {foo = id}) 42 + (1 :: Int) 1 Main> foo (Foo {foo = id}) True Program error: {instShow_v81_v1123 0 (Foo_Foo id Bool_True)} Everything works fine though, if you don't use the record update syntax: Main> foo (Foo id) True True / Ulf Norell

1 0

BUG in input.c
by Menaka Lashitha Bandara 29 Jul '02

29 Jul '02

Hi Guys, Here's a patch for a bug that was uncovered on my machine. Hugs did work for some time, and then suddenly stopped working (ie, wouldn't Parse Prelude.hs, and eats up CPU). Basically, lineBuffer in input.c was declared type (char), and EOF on my machine is -1, so it's outside the range of (char). The bug occurs line 516 of input.c (original). It works ok on x86, so I've put a little preprocessor routine to adapt depending on the value of EOF. There's one of my comments there, and if you feel that it taints hugs code, feel free to remove it. Take care, keep up the good work :). \LaShI -- Debian GNU/Linux 3.0 (Sid) (ibook2,powerpc), kernel 2.4.18-benh-xfs <lashi(a)optusnet.com.au> 1874 5597 C754 5209 0A7D 4DA6 9112 5D0E CC1F 67BC

1 0

An Urgent Request for Help -- Problem in doing the installation of Hugs.
by krishnan raman 26 Jul '02

26 Jul '02

July 25, 2002 HUGS people: I would greatly appreciate your help in getting HUGS installed properly very soon. BACKGROUND: 1) My system: Pentium PC with WIN NT 4 -- plenty of disk space and RAM. [ After I get this right, I will have to do the same in another PC elsewhere with Windows 95. ] 2) My background -- Experienced programmer. I have also installed all kinds of software in the past. 3) I am installing everything in D:\HUGS98. Curreently, I have in it the following: hugs98-Dec2001.msi; hugs98-Dec2001.zip; GraphicsLib.msi; SOE.msi; win32exes.zip ------------------------------------------------------- 4) I try to do the setup by double-clicking on hugs98-Dec2001.msi. The Setup process starts off as expected.t After the first few steps, I get the message: "The Installer was interrupted before Hugs98 � December 2001 could be installed. You need to restart the installer to try again." I have tried this a few trimes, The same problem. NO CLUE TO WHAT IS MISSING OR WRONG> WHY IS THIS OCCURRING ? HOW CAN I GET AROUND IT ? The setup must be the key step. ================================= 5) If I try to run hugs.exe or winhugs.exe, it comes up with the expected Hugs Screen, with an Error Message: " Prelude not found on current path: "{Hugs}\lib;{Hugs\lib\hugs; {Hugs\lib\exts; ;{Hugs\lib\win32 FATAL ERROR: Unable to load prelude This tells me that a successful setup would have installed the lib and subdirectories in the paths above, and the executable it would have found the prelude. Is the Setup process looking for something in some default location ? This happens in some kinds of software. Could you please tell me soon what I need to do to get the setup to go through correctly ? [ Also, Could you please explain: WHAT IS the "Prelude" ? ] I am keen on getting working on Haskell, using Hugs, I would greatly appreciate your reply as soon as possible. Thank you. K. Raman ramank0(a)yahoo.com P.S. Is there a telephone number at which I can call you, if need be ? If so, during what hours ? --------------------------------- Do You Yahoo!? Yahoo! Health - Feel better, live better

2 1

Problems...
by Pedro L. P. Dias 25 Jul '02

25 Jul '02

Dear Sir: I installed Hugs in my computer, and I'm reading "A Gentle Introduction to Haskell 98" downloaded from Haskell website. I start Hugs, and, at the line command "prelude>", when I digit, for example, "add x y = x + y", an error message appear: "...unexpected '='". Also, when I digit (other example) "inc n = n + 1", the error message is "undefined variable inc", or, sometimes, for the same digited line command, "undefined variable n". I'm just typing what is showed in that manual... Could you, please, help me? Thank you very much. Best regards, Pedro. Rio de Janeiro, Brazil.

2 1