January 2007 - Hugs-Bugs

[Hugs] #30: Windows and Hugs, fails to complete build of base
by Hugs 21 Mar '07

21 Mar '07

#30: Windows and Hugs, fails to complete build of base ---------------------+------------------------------------------------------ Reporter: neil | Owner: nobody Type: defect | Status: new Priority: blocker | Milestone: next release Component: hugs | Version: 200605 Keywords: | ---------------------+------------------------------------------------------ Under Windows/Mingw32, building hugs and the base package gives the following issue: ..\..\src\ffihugs -98 -P..\..\hugsdir\packages\base; -i"HsBase.h" ..\..\hugsdir\packages\base\System\Process\Internals.hs -Iinclude -I..\..\ghc\includes cbits\execvpe.c Skipping base package This bug is probably related to: > ERROR "C:\Program > Files\WinHugs\packages\base\System\Process\Internals.hs":747 - > Undefined variable "addMVarFinalizer" Ross says: Hmm, it uses addMVarFinalizer under Windows, and Hugs doesn't have that. Have to think about that one. This build failure causes the Regex module not to be built, since it comes after in the list of builds. -- Ticket URL: <http://hackage.haskell.org/trac/hugs/ticket/30> Hugs <http://www.haskell.org/hugs/> Hugs 98, an interpreter for Haskell

1 3

[Hugs] #43: WinHugs installer does not provide "Install for All Users" option
by Hugs 21 Mar '07

21 Mar '07

#43: WinHugs installer does not provide "Install for All Users" option ------------------------+--------------------------------------------------- Reporter: briansmith | Owner: neil Type: defect | Status: new Priority: major | Milestone: Component: winhugs | Version: 200609 Keywords: | ------------------------+--------------------------------------------------- The WinHugs installer does not provide an "Install for All Users" option. If the user is logged in with a limited account (which is the recommended practice in Windows), they will have to use the "Run as" command to run WinHugs as an administrator. But, the installer will then install the shortcuts to WinHugs in the Administrator account's Startup menu, but not any other users'. -- Ticket URL: <http://hackage.haskell.org/trac/hugs/ticket/43> Hugs <http://www.haskell.org/hugs/> Hugs 98, an interpreter for Haskell

1 3

[Hugs] #50: Disparity between GHCi and Hugs: Outstanding context : Show b
by Hugs 21 Mar '07

21 Mar '07

#50: Disparity between GHCi and Hugs: Outstanding context : Show b -----------------------------------+---------------------------------------- Reporter: guest | Owner: nobody Type: defect | Status: new Priority: minor | Milestone: next release Component: hugs | Version: 200503 Keywords: top level overloading | -----------------------------------+---------------------------------------- Possibly Hugs bug, or possible GHCi feature? {{{ showln = (++ "\n") . show main = interact (showln . length) }}} In GHCi: {{{ $ ghci Z.hs *Main> :t showln showln :: Int -> [Char] }}} In Hugs (Version: March 2005): {{{ ERROR "Z.hs":1 - Unresolved top-level overloading *** Binding : showln *** Outstanding context : Show b }}} I'm not sure which is right. -- Don -- Ticket URL: <http://hackage.haskell.org/trac/hugs/ticket/50> Hugs <http://www.haskell.org/hugs/> Hugs 98, an interpreter for Haskell

1 2

[Hugs] #35: WinHugs: foreground/background colours not from Windows
by Hugs 21 Mar '07

21 Mar '07

#35: WinHugs: foreground/background colours not from Windows ---------------------+------------------------------------------------------ Reporter: neil | Owner: neil Type: defect | Status: new Priority: minor | Milestone: Component: winhugs | Version: 200605 Keywords: | ---------------------+------------------------------------------------------ I just started using the new WinHugs GUI you developed and maintain, and I have a small issue with its appearance. I usually try to tone down all bright colors on the desktop, and hence the default window background color on my machines are not white but some shade of light gray. The issue is then that text portions seem to always have a white background while the rest of the WinHugs window conforms to the global color setting. I attach a window-shot for illustration. I suggest that it would be much nicer if the global background color setting was honored by the written text portion as well. --- Good point! The issue is to do with escape character, you can programatically change the background color of the text by outputting some magic, and this means that by default the background gets "forced" to white, rather than getting its default setting. I think the solution is to interpret white = the users background colour, black = the users foreground colour, and honour all other colours. I'll hack that up in the next few days and the next release should have this fixed. -- Ticket URL: <http://hackage.haskell.org/trac/hugs/ticket/35> Hugs <http://www.haskell.org/hugs/> Hugs 98, an interpreter for Haskell

1 2

[Hugs] #52: [PATCHES] port hugs98-Sep2006 to AIX 4.3.3, AIX 5.2, HPUX 11.0, SUNOS 5.9
by Hugs 21 Mar '07

21 Mar '07

#52: [PATCHES] port hugs98-Sep2006 to AIX 4.3.3, AIX 5.2, HPUX 11.0, SUNOS 5.9 -------------------------+-------------------------------------------------- Reporter: guest | Owner: nobody Type: enhancement | Status: new Priority: major | Milestone: Component: hugs | Version: 200609 Keywords: port | -------------------------+-------------------------------------------------- Sorry if this is a duplicate. I sent it to hugs-bugs(a)haskell.org but it just seems to have disappeared into the ether. Here is a shell script that I used to build the Sep 2006 release for: AIX 4.3.3 (IBM 5.0.2.0 compiler) AIX 5.2 (IBM 6.0.0.7 compiler) HPUX 11.0 (A.11.00.13 compiler) Redhat 9 x86 Redhat AS 4 x86 (CentOS actually) Solaris 9 (6.0u2 compiler) There are some source patches in the script as well as some settings that are required for proper compilation. I have a framework that runs the script in the unpacked source tree to do the build and install. Some autoconf work needs to be done for proper configuration of inline functions and shared library creation. As far as the latter goes you might consider using libtool, which is designed for that sort of thing. The current approach of using the compiler for shared library creation is not supported by HPUX 11.0 -- the link needs to be done by the linker directly. So at the very least the compiles and links need to be split into two separate commands. Some documentation for the OPTFLAGS and PTHREAD_CC variables would be nice. There is still an issue in the SUNOS build related to a couple of symbols in System/Posix/Internals.so. Hugs runs properly for what I am doing so I haven't spent time on this yet: runhugs: Error occurred ERROR "libraries/bootlib/System/Posix/Internals.hs" - Error while importing DLL "libraries/bootlib/System/Posix/Internals.so": ld.so.1: ffihugs: fatal: relocation error: file libraries/bootlib/System/Posix/Internals.so: symbol __hscore_readdir: referenced symbol not found There is also a question about the make_bootlib script -- what is "cat -s" intended to do? It is not clear to me that it is portable -- the AIX man page states that -S does what -s used to do, for example. Here is a summary of what I had to do for the various platforms to get hugs to compile and install: - AIX The AIX preprocessor fails when it sees invalid # directives. (Note that it also allows # directives to have white space or comments before the # -- it looked like some of the Haskell code in Hugs thinks spaces will prevent interference from the preprocessor.) The proper inline keyword is __inline. The _LARGE_FILES symbol collides with _LARGE_FILES_API, which latter is turned on automatically when long long is enabled (in at least one failed compilation, anyway). The hugs code has // comments in it -- this is not valid for older C compilers. Increase -qmaxmem so optimization completes without warnings. Turn on alloca support. Use -G compiler option for proper shared library support. - HPUX Added +Z to generate position independent code for shared library use. Added +DAportable so the resulting binaries can run on older PARISC version processors. Pass -b option to linker when building shared libraries. Added wrapper around ld for proper shared library linking. HPUX 11, at least the version I have, does not support linking a shared library via the compiler -- you are supposed to use the linker directly. I made a wrapper that throws away the crt0.o argument that the compiler passes to the linker and that seems to fix the problem (that particular file is not used for dynamic links and is not position-independent anyway so will cause a link failure). The HPUX compiler I have does not appear to support the inline keyword. - LINUX no changes necessary - SUNOS The SUNOS compiler I have (6.0u2) has some sort of problem with inline functions as used by Hugs in shared libraries. I converted them to "static" to work around whatever was going on. Perhaps "extern inline" is needed or some such. Set the -G option for linking shared libraries. -- Joe Buehler jbuehler(a)spirentcom.com {{{ #!/bin/ksh HOST="$1" OS="$2" OS_VERSION="$3" OS_VERSION_MINOR="$4" patch_aix() { /usr/local/bin/patch -p0 -N -b -Vnumbered <<\EOF --- libraries/tools/make_bootlib.~1~ 2006-05-19 17:36:17.000000000 -0400 +++ libraries/tools/make_bootlib 2007-01-09 12:50:58.000000000 -0500 @@ -62,10 +62,19 @@ # portability we run the preprocessor on a .c file. cpp_input=$tmpdir/cppinput.c - cp "$1" $cpp_input + # AIX C/C++ compiler version 6.0 fails on illegal preprocessor directives + sed ' + s/^$[ ]*#[ ]*[-!]$/ELIMINATE THIS STRING PLEASE\1/ + s/^$[ ][ ]*#[ ]*[A-Z]$/ELIMINATE THIS STRING PLEASE\1/ + s/^$[ ][ ]*#[ ]*osthreads$/ELIMINATE THIS STRING PLEASE\1/ + ' "$1" >$cpp_input # gcc-3.3 on MacOS X 10.3 is reported to add #pragma - $cpp $cpp_flags $cpp_input | grep -v '^#' | cat -s + $cpp $cpp_flags $cpp_input | + sed ' + /^#/d + s/^ELIMINATE THIS STRING PLEASE// + ' | cat -s } # internal Hugs modules --- packages/base/include/HsBase.h.~1~ 2006-09-20 18:01:52.000000000 -0400 +++ packages/base/include/HsBase.h 2007-01-09 13:10:35.000000000 -0500 @@ -220,6 +220,8 @@ #ifndef INLINE # if defined(_MSC_VER) # define INLINE extern __inline +# elif defined(__xlC__) +# define INLINE __inline # elif defined(__GNUC__) # define INLINE extern inline # else EOF } patch_hpux() { /usr/local/bin/patch -p0 -N -b -Vnumbered <<\EOF --- packages/base/include/HsBase.h.~1~ 2006-09-20 18:01:52.000000000 -0400 +++ packages/base/include/HsBase.h 2007-01-09 13:10:35.000000000 -0500 @@ -220,6 +220,8 @@ #ifndef INLINE # if defined(_MSC_VER) # define INLINE extern __inline +# elif defined(__hpux) +# define INLINE static # elif defined(__GNUC__) # define INLINE extern inline # else EOF } patch_sunos() { /usr/local/bin/patch -p0 -N -b -Vnumbered <<\EOF --- packages/base/include/HsBase.h.~1~ 2006-09-20 18:01:52.000000000 -0400 +++ packages/base/include/HsBase.h 2007-01-09 13:10:35.000000000 -0500 @@ -220,6 +220,8 @@ #ifndef INLINE # if defined(_MSC_VER) # define INLINE extern __inline +# elif defined(__sun) +# define INLINE static # elif defined(__GNUC__) # define INLINE extern inline # else EOF } export CC=cc export PTHREAD_CC=cc export CFLAGS=-O export OPTFLAGS=-O case "$OS" in aix) # This is the compiler I like to use under AIX. CC=/usr/vacpp/bin/xlC_r PTHREAD_CC=/usr/vacpp/bin/xlC_r # Enable large files -- this is implied by "long long" so we turn it on globally. CFLAGS="$CFLAGS -D_LARGE_FILE_API" # Some of the C code actually has C++ commments in it! CFLAGS="$CFLAGS -qcpluscmt" # Increase memory available for optimization. CFLAGS="$CFLAGS -qmaxmem=32768" # For alloca support CFLAGS="$CFLAGS -ma" # This version of HUGS does not know how to make AIX shared objects. export ac_cv_dll_flags=-G # The _LARGE_FILES and _LARGE_FILE_API symbols are incompatible -- # defining both causes system header file conflicts (AIX 5.2.0.0) export ac_cv_sys_large_files=no patch_aix ;; hpux) # This version of HUGS does not know how to make HPUX shared objects. CFLAGS="$CFLAGS +Z +DAportable -tl,$PWD/myld" export ac_cv_dll_flags="-Wl,-b" cat >myld <<-\EOF && chmod 755 myld #!/bin/ksh I=0 J=0 SHARED= #echo "LINK: $0 $*" >&2 for ARG; do case "$ARG" in */crt0.o) ARGV[$I]="$ARG" I=$(expr $I + 1) ;; -b) SHARED=1 ARGV[$I]="$ARG" ARGV_SHARED[$J]="$ARG" I=$(expr $I + 1) J=$(expr $J + 1) ;; *) ARGV[$I]="$ARG" ARGV_SHARED[$J]="$ARG" I=$(expr $I + 1) J=$(expr $J + 1) ;; esac done if [ "$SHARED" != "" ]; then set -x ld "${ARGV_SHARED[@]}" else set -x ld "${ARGV[@]}" fi EOF patch_hpux ;; linux) ;; sunos) # This version of HUGS does not know how to make SUNOS shared objects. export ac_cv_dll_flags=-G patch_sunos ;; esac && ./configure \ CC="$CC" \ PTHREAD_CC="$CC" \ CFLAGS="$CFLAGS" \ OPTFLAGS="$OPTFLAGS" \ --disable-large-banner \ --with-pthreads \ && gmake && ( gmake verbosecheck || true ) && gmake install && true }}} -- Ticket URL: <http://hackage.haskell.org/trac/hugs/ticket/52> Hugs <http://www.haskell.org/hugs/> Hugs 98, an interpreter for Haskell

1 1

[Hugs] #37: Equal Ints are not equal
by Hugs 18 Mar '07

18 Mar '07

#37: Equal Ints are not equal ---------------------+------------------------------------------------------ Reporter: guest | Owner: nobody Type: defect | Status: new Priority: blocker | Milestone: Component: hugs | Version: current Keywords: | ---------------------+------------------------------------------------------ Hi, a few days ago I wrote a program which handles my own integers. Then I check my own implementation against the normal Int implementation in Hugs. Thereby I wonder about inequality of Ints which are definitely equal. This is a test program I wrote after I recognized that hugs has probably a bug in the inequality check of Ints. data MInt = Zero | Succ MInt | Pred MInt deriving Show tn :: Int -> MInt tn x | x<0 = Pred (tn (x+1)) tn 0 = Zero tn (n+1) = Succ (tn n) ti :: MInt -> Int ti Zero = 0 ti (Succ x) = 1+(ti x) ti (Pred x) = (ti x) -1 testi :: (MInt -> MInt -> MInt) -> (Int -> Int -> Int) -> Int -> Int -> Bool testi f g x y = (ti (f (tn x) (tn y))) /= (g x y) myMul x y = tn ((ti x) * (ti y)) test = [(x,y,ti (myMul (tn x) (tn y)),x * y)| x<-[-100..100], y<-([-100..(-1)]++[1..100]),(testi myMul (*) x y) ] the list "test" should be empty in any case but unfortunately hugs returns this (copy of the console): __ __ __ __ ____ ___ _________________________________________ || || || || || || ||__ Hugs 98: Based on the Haskell 98 standard ||___|| ||__|| ||__|| __|| Copyright (c) 1994-2003 ||---|| ___|| World Wide Web: http://haskell.org/hugs || || Report bugs to: hugs-bugs(a)haskell.org || || Version: November 2003 _________________________________________ Haskell 98 mode: Restart with command line option -98 to enable extensions Type :? for help Main> test [(-100,80,-8000,-8000),(-100,99,-9900,-9900), (-99,58,-5742,-5742),(-99,83,-8217,-8217), (-98,71,-6958,-6958),(-98,86,-8428,-8428), (-97,46,-4462,-4462),(-97,87,-8439,-8439), (-96,76,-7296,-7296),(-95,88,-8360,-8360),(-94,73,-6862,-6862) ERROR - Control stack overflow Main> I have checked this program with Hugs on different plattforms: Gentoo Linux x86 64-Bit and Suse Linux 32-Bit. On both plattforms the same bug occurs with different Ints for each run (i.e. the list contains different tuples). I have also checked my program with other Haskell implementations and there it returns an empty list as expected. Best regards, Stephan Swiderski (swiderski(a)informatik.rwth-aachen.de) P.S.: We have tested Sep2006 with similar results: __ __ __ __ ____ ___ _________________________________________ || || || || || || ||__ Hugs 98: Based on the Haskell 98 standard ||___|| ||__|| ||__|| __|| Copyright (c) 1994-2005 ||---|| ___|| World Wide Web: http://haskell.org/hugs || || Bugs: http://hackage.haskell.org/trac/hugs || || Version: September 2006 _________________________________________ Haskell 98 mode: Restart with command line option -98 to enable extensions Type :? for help Main> :e Main> test [(-100,90,-9000,-9000),(-98,40,-3920,-3920),(-97,63,-6111,-6111),(-97,69,-6693,- 6693) ERROR - Control stack overflow -- Ticket URL: <http://hackage.haskell.org/trac/hugs/ticket/37> Hugs <http://www.haskell.org/hugs/> Hugs 98, an interpreter for Haskell

2 7

Undefined variable bug
by John Tromp 27 Feb '07

27 Feb '07

I just found out that I can circumvent the problem by adding a line import Char to my loaded file. regards, -john

2 1

[Hugs] #53: qualified + hiding hides too much
by Hugs 23 Jan '07

23 Jan '07

#53: qualified + hiding hides too much ---------------------------+------------------------------------------------ Reporter: ross | Owner: nobody Type: defect | Status: new Priority: major | Milestone: Component: module system | Version: 200609 Keywords: | ---------------------------+------------------------------------------------ Given the following modules: {{{ module A where f = 1 g = 2 }}} {{{ module B where import qualified A hiding (f) h = A.g }}} Hugs fails with {{{ ERROR "B.hs":3 - Undefined qualified variable "A.g" }}} (reported by Iavor Diatchki) This affects variables, but not types or data constructors. -- Ticket URL: <http://hackage.haskell.org/trac/hugs/ticket/53> Hugs <http://www.haskell.org/hugs/> Hugs 98, an interpreter for Haskell

1 1

module system bug
by Iavor Diatchki 22 Jan '07

22 Jan '07

Hello, There appears to be a problem when combinig "hiding" and "qualified" imports. Here is an example: module A where x = True module Main(main) where import qualified A hiding () main = print A.x When loading the second module (Main) Hugs (September 2006) complains that A.x is not in scope. It appears that this happens whenever there is a 'hiding' clause (it does not have to be empty) but it does not happen if the import is not qualified. -Iavor

1 0

Building Hugs on PPC Mac OS X 10.4 message: No rule to make target `../hsc2hs/Main.hs', needed by `../hugsdir/programs/hsc2hs/Main.hs'.
by Thorkil Naur 08 Jan '07

08 Jan '07

Hello all Hugs'ers, As a well-behaved boy, I try to prepare for the Hac07 event by, among other things, installing Hugs. I follow the directions on http://hackage.haskell.org/trac/hugs for getting and building a development version, cvs for hugs98, darcs for the libraries etc. All this happens on a PPC Mac OS X 10.4 with a ghc-6.4.1 installed from binary. Unfortunately, I am now stuck in the "make". Initially, I got the messages ... configure: creating ./config.status config.status: creating include/HsBaseConfig.h find: -wholename: unknown expression primary find: -wholename: unknown expression primary find: -wholename: unknown expression primary ERROR: unable to locate Prelude along new path: "libraries/bootlib" - ignoring it. ERROR: unable to locate Prelude along new path: "libraries/bootlib" - ignoring it. ERROR: unable to locate Prelude along new path: "libraries/bootlib" - ignoring it. ERROR: unable to locate Prelude along new path: "libraries/bootlib" - ignoring it. ERROR: unable to locate Prelude along new path: "libraries/bootlib" - ignoring it. ERROR: unable to locate Prelude along new path: "libraries/bootlib" - ignoring it. ERROR: unable to locate Prelude along new path: "libraries/bootlib" - ignoring it. echo timestamp for bootlib >bootlib/.stamp cd ../cpphs; HUGSFLAGS=-P../libraries/bootlib HUGSDIR=../hugsdir ../src/runhugs -98 Setup.hs configure --verbose --hugs --prefix='/usr/local' --scratchdir='../hugsdir/packages/cpphs' --with-compiler=../src/ffihugs ERROR: unable to locate Prelude along new path: "../libraries/bootlib" - ignoring it. make[1]: *** [../hugsdir/programs/cpphs/Main.hs] Error 1 make: *** [all] Error 2 and those "find: -wholename: unknown expression primary" messages made me a bit concerned: I investigated "man find" and found that the OS X 10.4 find does not have a -wholename option. Has this been tried before, i.e. boot-strapping hugs98 on a Mac OS X 10.4? In any case, on a Linux machine, "man find" says that -path is the same as -wholename, so I replaced -wholename by -path in "hugs98/libraries/tools/make_bootlib" (which is the only place I found -wholename) and tried again. And this time, I am stuck with the messages ... echo timestamp for bootlib >bootlib/.stamp cd ../cpphs; HUGSFLAGS=-P../libraries/bootlib HUGSDIR=../hugsdir ../src/runhugs -98 Setup.hs configure --verbose --hugs --prefix='/usr/local' --scratchdir='../hugsdir/packages/cpphs' --with-compiler=../src/ffihugs Configuring cpphs-1.3... configure: looking for package tool: hugs near compiler in ../src configure: found package tool in ../src/hugs configure: Using install prefix: /usr/local configure: Binaries installed in: /usr/local/bin configure: Libraries installed in: /usr/local/lib/hugs/packages/cpphs configure: Private binaries installed in: /usr/local/libexec configure: Data files installed in: /usr/local/share/cpphs-1.3 configure: Using compiler: ../src/ffihugs configure: Compiler flavor: Hugs configure: Compiler version: configure: Using package tool: ../src/hugs configure: Using ar found on system at: /usr/bin/ar configure: No haddock found configure: Using ld found on system at: /usr/bin/ld configure: No pfesetup found configure: Using ranlib found on system at: /usr/bin/ranlib configure: Using runghc found on system at: /usr/local/bin/runghc configure: No runhugs found configure: Using tar found on system at: /usr/bin/tar configure: Using happy: /usr/local/bin/happy configure: Using alex: /usr/local/bin/alex configure: Using hsc2hs: /usr/local/bin/hsc2hs configure: No c2hs found configure: No cpphs found configure: No greencard found cd ../cpphs; HUGSFLAGS=-P../libraries/bootlib HUGSDIR=../hugsdir ../src/runhugs -98 Setup.hs build --verbose Preprocessing library cpphs-1.3... Preprocessing executables for cpphs-1.3... Building cpphs-1.3... /bin/rm -f -r ../hugsdir/packages/cpphs/autogen mkdir -p ../hugsdir/programs /bin/rm -f -r ../hugsdir/programs/cpphs mv ../hugsdir/packages/cpphs/programs/cpphs ../hugsdir/programs rmdir ../hugsdir/packages/cpphs/programs (echo '#! /bin/sh'; \ echo "rootdir='`cd ..; pwd`'"; \ echo '$rootdir/src/runhugs -P$rootdir/libraries/bootlib $rootdir/hugsdir/programs/cpphs/Main.hs "$@"') >tools/cpphs chmod 755 tools/cpphs make[1]: *** No rule to make target `../hsc2hs/Main.hs', needed by `../hugsdir/programs/hsc2hs/Main.hs'. Stop. make: *** [all] Error 2 With my limited capacity, I have not managed to find a way around this obstacle. Your help to overcome this would be most welcome. I hope that I am not wasting your time. I will gladly supply additional information. And one last piece of information: I never tried to install Hugs from some binary distribution, that could very well work. Best regards Thorkil

1 0