The following procedure is suggested for installation of Hugs on a Unix
machine.  (Precompiled binaries are provided for common PC platforms,
although similar installation procedures can be used for these machines
if you have a suitable C compiler.)

Short version (for Unix-like environments):

    cd Hugs98/src/unix
    ./configure --prefix=$HOME
    cd ..
    make install  
    make clean
    $HOME/bin/hugs $HOME/share/hugs/demos/Say
    putStr (say "  /Hugs")
    :quit


Long version:

 0) Choose a directory (or directories) for the Hugs files to go in.
    In the following, we'll assume:

    /usr/local/solaris/bin             user executables
    /usr/local/solaris/lib/hugs        architecture dependent files
    /usr/local/lib/hugs                architecture independent files

    Check that these directories have appropriate permission settings.  

 1) Run the configure script

    On Unix, MacOS X, or other platform providing sh, sed and related tools:

      cd Hugs98/src/unix
      ./configure --exec-prefix=/usr/local/solaris --datadir=/usr/local/lib
      -- This tests the capabilities of your C compiler and OS
      -- and arranges to install binaries in the directories places.

      Read Hugs98/src/unix/INSTALL to find out about other configuration
      options.  Hugs specific options are listed at the end of this file.

    On Windows:

      Pick one of the following configurations:

        msc:     Microsoft Visual C++ 
        winhugs: Microsoft Visual C++, Hugs for Windows
        bcc32:   32 bit Borland C
        djgpp2:  DJGPP2

      cd Hugs98\src\{msc,winhugs,bcc32,djgpp2}
      config.bat

        If you haven't used Visual C++ or Borland C from the command line
        before, you might have to define some environment variables (in
        autoexec.bat or using "Control Panel|System") before you can use
        the compiler - check your compiler documentation.

      make hugs.exe   (or nmake hugs.exe with Visual C++)

      If you prefer to use an integrated development environment,
      run "make hugs.exe -n" to get a list of the files we compile
      and add them to your project.

    On MacOS (pre MacOS X):

      Open CodeWarrior Pro 5 and choose [File->Import Project...].
      Select the file ":hugs98:src:machugs:hugs.mcp.xml".  When the file
      dialog appears, name the new project "hugs.mcp", and save it in the
      same folder as the imported xml file.  Then choose [Project->Make].
      The Hugs 98 application will be created in the hugs98 folder.

      Preferences in the form of a Hugs command line string can be saved
      in a file "Hugs Preferences" in the same folder as the application.

    On another platform:

      If you can't run the configure script, pick one of the Windows
      configurations and use it as a basis for constructing the
      following files 

        Hugs98/src/Makefile
        Hugs98/src/options.h   # Configuration options
        Hugs98/src/config.h    # Description of your machine/compiler/OS

 2) Build Hugs and associated files

    cd ..
    make

    At this point, you might like to run a few tests to make sure 
    everything's working.

      ./hugs -P../lib
      -- Run a few tests like 1+2, [1..], etc.  Use ":quit" to exit

    On Windows, you have to "move hugs.exe .." before running it.

 3) Install Hugs in chosen directories

      make install

    (Note that the permissions of the installed files will be affected by
    your "umask" setting.  If you want the installed files to be world
    readable, you need to set umask accordingly.)

    Try a few simple tests:

      /usr/local/solaris/bin/hugs
      [1..10]
      :quit

      cat > echo <<EOF
      #! /usr/local/solaris/bin/runhugs +l

      > module Main(main) where
      > import System(getArgs)
      >
      > main = do { args <- getArgs; putStrLn (unwords args) }
      EOF
      chmod 755 echo
      ./echo Hello World

    Administrators of sites with multiple architectures often prefer to
    install a shell script in /usr/local/bin which will select an 
    appropriate binary according to the architecture on which it is
    run rather than a binary.  Configuration should be the same as above
    (use a different --bindir argument for each architecture but, if
    you want, save a small amount of space by using the same --datadir)
    and use a shell script like the following.

      #! /bin/sh

      HW_OS=${HW_OS-`/usr/local/gnu/bin/hw_os`}
      BINDIR=/local/lib/Hugs1.4/$HW_OS
      exec $BINDIR/hugs $*

    This kind of script would also be a good place to set system-wide
    default options (eg to select an editor or to switch on the
    "show types" option).

 4) Cleanup after yourself

    You can now run "make clean" to delete all machine-generated files.  
    If you ran "make install", you could delete the entire Hugs source
    tree - but you might want to keep the Hugs98/doc directory.

Hugs specific configuration options:

 --with-readline

   Use command line editor if available.

   We highly recommend this option.  However, it changes Hugs' user
   interface in a way that will cause every single regression test
   to fail.  
   (On Unix, regression tests are run by "make install; make check".)

   There are two choices of command line editor that can be used with Hugs:

     GNU readline:      usual GNU sources (e.g. bash distribution)
     editline:          (comp.sources.misc, vol 31, issue 71)

   As you might expect, the GNU library is a bit larger but more powerful.

 --enable-timer

   Time how long each evaluation takes.

   Timing is included for the purpose of benchmarking the Hugs interpreter,
   comparing its performance across a variety of different machines, and
   with other systems for similar languages.
   
   It would be somewhat foolish to try to use the timings produced in this
   way for any other purpose.  In particular, using timings to compare the
   performance of different versions of an algorithm is likely to give very
   misleading results.  The current implementation of Hugs as an interpreter,
   without any significant optimizations, means that there are much more
   significant overheads than can be accounted for by small variations in
   Hugs code.

 --enable-profiling

   Gather statistics about heap allocation during evaluation.
   Statistics are written to a file profile.hp which may be viewed
   using the hp2ps program.

   This option makes Hugs use much more memory and run much slower.
   The ":set -d" command can be used to reduce the time overhead by
   controlling the frequency with which statistics are gathered.

 --with-nmake

   Try to generate a Makefile that will work with Microsoft's nmake.

 --disable-large-banner

   Print a single-line startup banner instead of the 9 line banner.
   (This option will also cause the "make check" regression tests to fail.)

 --with-gui

   Used when generating Hugs for Windows.  Only works with Borland C++

 --enable-internal-prims

   Enable experimental features used in Hugs98/lib/hugs/HugsInternals.hs

 --enable-stack-dumps

   Enable printing of the top and bottom few objects on the stack when
   stack overflow happens.  This feature is currently (Sept'97) just a
   proof of concept.  We welcome suggestions (and/or code) to make it 
   useful for people who don't have an intimate knowledge of how the
   G machine operates.

 --enable-debug
 --enable-tag-checks
 --enable-lint

   For use when debugging Hugs.

 --with-preprocessor

   This is an experimental feature and may change in future versions.

   Enable the use of a preprocessor for processing Haskell source files
   before compiling them with Hugs.  When configured with preprocessing
   on, you can use the "-F" option to specify which preprocessor to use.
   For example, if your preprocessor is in /users/JFH/bin/hscpp, you might
   say

     :set -F"/users/JFH/bin/hscpp"

   If you have perl and gcc installed on your machine, the following
   script provides a simple cpp-like preprocessor.

     eval "exec perl -S $0 $*"
	  if $running_under_some_random_shell;
     #
     # Reads CPP output and turns #line things into appropriate Haskell
     # pragmas.  This program is derived from the "hscpp" script
     # distributed with the Glasgow Haskell Compiler.
     #
     $Cpp = 'gcc -E -xc -traditional';
     open(INPIPE, "$Cpp @ARGV |") || die "Can't open C pre-processor pipe\n";
     while (<INPIPE>) {
     # line directives come in flavo[u]rs:
     #   s/^#\s*line\s+\d+$/\{\-# LINE \-\}/;   IGNORE THIS ONE FOR NOW
	 s/^#\s*line\s+(\d+)\s+(\".+\")$/\{\-# LINE \1 \2 \-\}/;
	 s/^#\s*(\d+)\s+(\".*\").*/\{\-# LINE \1 \2 \-\}/;
	 print $_;
     }
     close(INPIPE) || exit(1); # exit is so we reflect any errors.
     exit(0);

   Note that Hugs currently ignores the {-# LINE _ _ #-} pragmas so error
   messages will refer to the wrong line numbers.