package Pod2::Html;
use strict;
require Exporter;

use vars qw($VERSION @ISA @EXPORT);
$VERSION = 1.03;
@ISA = qw(Exporter);
@EXPORT = qw(pod2html htmlify);

use Carp;
use Config;
use Cwd;
use File::Spec::Unix;
use Getopt::Long;
use Pod::Functions;

use locale;	# make \w work right in non-ASCII lands


my $html = '';
my $cache_ext = $^O eq 'VMS' ? ".tmp" : ".x~~";
my $dircache = "pod2htmd$cache_ext";
my $itemcache = "pod2htmi$cache_ext";

my @begin_stack = ();		# begin/end stack

my @libpods = ();	    	# files to search for links from C<> directives
my $htmlroot = "/";	    	# http-server base directory from which all
				#   relative paths in $podpath stem.
my $htmldir = "";		# The directory to which the html pages
				# will (eventually) be written.
my $htmlfile = "";		# write to stdout by default
my $htmlfileurl = "" ;		# The url that other files would use to
				# refer to this file.  This is only used
				# to make relative urls that point to
				# other files.
my $podfile = "";		# read from stdin by default
my @podpath = ();		# list of directories containing library pods.
my $podroot = ".";		# filesystem base directory from which all
				#   relative paths in $podpath stem.
my $css = '';                   # Cascading style sheet
my $recurse = 1;		# recurse on subdirectories in $podpath.
my $quiet = 0;			# not quiet by default
my $verbose = 0;		# not verbose by default
my $doindex = 1;   	    	# non-zero if we should generate an index
my $backlink = '';              # text for "back to top" links
my $listlevel = 0;		# current list depth
my @listend = ();		# the text to use to end the list.
my $after_lpar = 0;             # set to true after a par in an =item
my $ignore = 1;			# whether or not to format text.  we don't
				#   format text until we hit our first pod
				#   directive.

my %items_named = ();		# for the multiples of the same item in perlfunc
my @items_seen = ();
my $netscape = 0;		# whether or not to use netscape directives.
my $title;			# title to give the pod(s)
my $header = 0;			# produce block header/footer
my $top = 1;			# true if we are at the top of the doc.  used
				#   to prevent the first <HR> directive.
my $paragraph;			# which paragraph we're processing (used
				#   for error messages)
my $ptQuote = 0;                # status of double-quote conversion
my %pages = ();			# associative array used to find the location
				#   of pages referenced by L<> links.
my %sections = ();		# sections within this page
my %items = ();			# associative array used to find the location
				#   of =item directives referenced by C<> links
my %local_items = ();           # local items - avoid destruction of %items
my $Is83;                       # is dos with short filenames (8.3)

sub init_globals {
$dircache = "pod2htmd$cache_ext";
$itemcache = "pod2htmi$cache_ext";

@begin_stack = ();		# begin/end stack

@libpods = ();	    	# files to search for links from C<> directives
$htmlroot = "/";	    	# http-server base directory from which all
				#   relative paths in $podpath stem.
$htmldir = "";	    	# The directory to which the html pages
				# will (eventually) be written.
$htmlfile = "";		# write to stdout by default
$podfile = "";		# read from stdin by default
@podpath = ();		# list of directories containing library pods.
$podroot = ".";		# filesystem base directory from which all
				#   relative paths in $podpath stem.
$css = '';                   # Cascading style sheet
$recurse = 1;		# recurse on subdirectories in $podpath.
$quiet = 0;		# not quiet by default
$verbose = 0;		# not verbose by default
$doindex = 1;   	    	# non-zero if we should generate an index
$backlink = '';		# text for "back to top" links
$listlevel = 0;		# current list depth
@listend = ();		# the text to use to end the list.
$after_lpar = 0;        # set to true after a par in an =item
$ignore = 1;			# whether or not to format text.  we don't
				#   format text until we hit our first pod
				#   directive.

@items_seen = ();
%items_named = ();
$netscape = 0;		# whether or not to use netscape directives.
$header = 0;			# produce block header/footer
$title = '';			# title to give the pod(s)
$top = 1;			# true if we are at the top of the doc.  used
				#   to prevent the first <HR> directive.
$paragraph = '';			# which paragraph we're processing (used
				#   for error messages)
%sections = ();		# sections within this page

# These are not reinitialised here but are kept as a cache.
# See get_cache and related cache management code.
#%pages = ();			# associative array used to find the location
				#   of pages referenced by L<> links.
#%items = ();			# associative array used to find the location
				#   of =item directives referenced by C<> links
%local_items = ();
$Is83=$^O eq 'dos';
}

#
# clean_data: global clean-up of pod data
#
sub clean_data($){
    my( $dataref ) = @_;
    my $i;
    for( $i = 0; $i <= $#$dataref; $i++ ){
	${$dataref}[$i] =~ s/\s+\Z//;

        # have a look for all-space lines
	if( ${$dataref}[$i] =~ /^\s+$/m ){
	    my @chunks = split( /^\s+$/m, ${$dataref}[$i] );
	    splice( @$dataref, $i, 1, @chunks );
	}
    }
}


sub pod2html {
    local(@ARGV) = @_;
    local($/);
    local $_;

    init_globals();

    $Is83 = 0 if (defined (&Dos::UseLFN) && Dos::UseLFN());

    # cache of %pages and %items from last time we ran pod2html

    #undef $opt_help if defined $opt_help;

    # parse the command-line parameters
    parse_command_line();

    # set some variables to their default values if necessary
    local *POD;
    unless (@ARGV && $ARGV[0]) { 
	$podfile  = "-" unless $podfile;	# stdin
	open(POD, "<$podfile")
		|| die "$0: cannot open $podfile file for input: $!\n";
    } else {
	$podfile = $ARGV[0];  # XXX: might be more filenames
	*POD = *ARGV;
    } 
    $htmlfile = "-" unless $htmlfile;	# stdout
    $htmlroot = "" if $htmlroot eq "/";	# so we don't get a //
    $htmldir =~ s#/\z## ;               # so we don't get a //
    if (  $htmlroot eq ''
       && defined( $htmldir ) 
       && $htmldir ne ''
       && substr( $htmlfile, 0, length( $htmldir ) ) eq $htmldir 
       ) 
    {
	# Set the 'base' url for this file, so that we can use it
	# as the location from which to calculate relative links 
	# to other files. If this is '', then absolute links will
	# be used throughout.
        $htmlfileurl= "$htmldir/" . substr( $htmlfile, length( $htmldir ) + 1);
    }

    # read the pod a paragraph at a time
    warn "Scanning for sections in input file(s)\n" if $verbose;
    $/ = "";
    my @poddata  = <POD>;
    close(POD);
    clean_data( \@poddata );

    # scan the pod for =head[1-6] directives and build an index
    my $index = scan_headings(\%sections, @poddata);

    unless($index) {
	warn "No headings in $podfile\n" if $verbose;
    }

    # open the output file
#    open(HTML, ">$htmlfile")
#	    || die "$0: cannot open $htmlfile file for output: $!\n";

    # put a title in the HTML file if one wasn't specified
    if ($title eq '') {
	TITLE_SEARCH: {
	    for (my $i = 0; $i < @poddata; $i++) { 
		if ($poddata[$i] =~ /^=head1\s*NAME\b/m) {
		    for my $para ( @poddata[$i, $i+1] ) { 
			last TITLE_SEARCH
			    if ($title) = $para =~ /(\S+\s+-+.*\S)/s;
		    }
		} 

	    } 
	}
    }
    if (!$title and $podfile =~ /\.pod\z/) {
	# probably a split pod so take first =head[12] as title
	for (my $i = 0; $i < @poddata; $i++) { 
	    last if ($title) = $poddata[$i] =~ /^=head[12]\s*(.*)/;
	} 
	warn "adopted '$title' as title for $podfile\n"
	    if $verbose and $title;
    } 
    if ($title) {
	$title =~ s/\s*\(.*\)//;
    } else {
	warn "$0: no title for $podfile" unless $quiet;
	$podfile =~ /^(.*)(\.[^.\/]+)?\z/s;
	$title = ($podfile eq "-" ? 'No Title' : $1);
	warn "using $title" if $verbose;
    }

    my $block = $header ? <<END_OF_BLOCK : '';
<TABLE BORDER=0 CELLPADDING=0 CELLSPACING=0 WIDTH=100%>
<TR><TD CLASS=block VALIGN=MIDDLE WIDTH=100% BGCOLOR="#cccccc">
<FONT SIZE=+1><STRONG><P CLASS=block>&nbsp;$title</P></STRONG></FONT>
</TD></TR>
</TABLE>
END_OF_BLOCK

    $html .= $block;

    # load/reload/validate/cache %pages and %items
    get_cache($dircache, $itemcache, \@podpath, $podroot, $recurse);

    # scan the pod for =item directives
    scan_items( \%local_items, "", @poddata);

    # put an index at the top of the file.  note, if $doindex is 0 we
    # still generate an index, but surround it with an html comment.
    # that way some other program can extract it if desired.
    $index =~ s/--+/-/g;
    $html .=  "<A NAME=\"__index__\"></A>\n";
    $html .=  "<!-- INDEX BEGIN -->\n";
    $html .=  "<!--\n" unless $doindex;
    $html .=  $index;
    $html .=  "-->\n" unless $doindex;
    $html .=  "<!-- INDEX END -->\n\n";
    $html .=  "<HR>\n" if $doindex and $index;

    # now convert this file
    my $after_item;             # set to true after an =item
    warn "Converting input file $podfile\n" if $verbose;
    foreach my $i (0..$#poddata){
        $ptQuote = 0; # status of quote conversion

	$_ = $poddata[$i];
	$paragraph = $i+1;
	if (/^(=.*)/s) {	# is it a pod directive?
	    $ignore = 0;
	    $after_item = 0;
	    $_ = $1;
	    if (/^=begin\s+(\S+)\s*(.*)/si) {# =begin
		process_begin($1, $2);
	    } elsif (/^=end\s+(\S+)\s*(.*)/si) {# =end
		process_end($1, $2);
	    } elsif (/^=cut/) {			# =cut
		process_cut();
	    } elsif (/^=pod/) {			# =pod
		process_pod();
	    } else {
		next if @begin_stack && $begin_stack[-1] ne 'html';

		if (/^=(head[1-6])\s+(.*\S)/s) {	# =head[1-6] heading
		    process_head( $1, $2, $doindex && $index );
		} elsif (/^=item\s*(.*\S)?/sm) {	# =item text
		    warn "$0: $podfile: =item without bullet, number or text"
		       . " in paragraph $paragraph.\n" if !$quiet and (!defined($1) or $1 eq '');
		    process_item( $1 );
		    $after_item = 1;
		} elsif (/^=over\s*(.*)/) {		# =over N
		    process_over();
		} elsif (/^=back/) {		# =back
		    process_back();
		} elsif (/^=for\s+(\S+)\s*(.*)/si) {# =for
		    process_for($1,$2);
		} else {
		    /^=(\S*)\s*/;
		    warn "$0: $podfile: unknown pod directive '$1' in "
		       . "paragraph $paragraph.  ignoring.\n" unless $quiet;
		}
	    }
	    $top = 0;
	}
	else {
	    next if $ignore;
	    next if @begin_stack && $begin_stack[-1] ne 'html';
	    my $text = $_;
	    if( $text =~ /\A\s+/ ){
		process_pre( \$text );
	        $html .=  "<PRE>\n$text</PRE>\n";

	    } else {
		process_text( \$text );

		# experimental: check for a paragraph where all lines
		# have some ...\t...\t...\n pattern
		if( $text =~ /\t/ ){
		    my @lines = split( "\n", $text );
		    if( @lines > 1 ){
			my $all = 2;
			foreach my $line ( @lines ){
			    if( $line =~ /\S/ && $line !~ /\t/ ){
				$all--;
				last if $all == 0;
			    }
			}
			if( $all > 0 ){
			    $text =~ s/\t+/<TD>/g;
			    $text =~ s/^/<TR><TD>/gm;
			    $text = '<TABLE CELLSPACING=0 CELLPADDING=0>' .
                                    $text . '</TABLE>';
			}
		    }
		}
		## end of experimental

		if( $after_item ){
		    $html .=  "$text\n";
		    $after_lpar = 1;
		} else {
		    $html .=  "<P>$text</P>\n";
		}
	    }
	    $after_item = 0;
	}
    }

    # finish off any pending directives
    finish_list();

    # link to page index
    $html .=  "<P><A HREF=\"#__index__\"><SMALL>$backlink</SMALL></A></P>\n"
	if $doindex and $index and $backlink;

    $html .=  $block;

    # close the html file
return $html;

    warn "Finished\n" if $verbose;
}

##############################################################################

my $usage;			# see below
sub usage {
    my $podfile = shift;
    warn "$0: $podfile: @_\n" if @_;
    die $usage;
}

$usage =<<END_OF_USAGE;
Usage:  $0 --help --htmlroot=<name> --infile=<name> --outfile=<name>
           --podpath=<name>:...:<name> --podroot=<name>
           --libpods=<name>:...:<name> --recurse --verbose --index
           --netscape --norecurse --noindex

  --backlink     - set text for "back to top" links (default: none).
  --css          - stylesheet URL
  --flush        - flushes the item and directory caches.
  --[no]header   - produce block header/footer (default is no headers).
  --help         - prints this message.
  --htmldir      - directory for resulting HTML files.
  --htmlroot     - http-server base directory from which all relative paths
                   in podpath stem (default is /).
  --[no]index    - generate an index at the top of the resulting html
                   (default behaviour).
  --infile       - filename for the pod to convert (input taken from stdin
                   by default).
  --libpods      - colon-separated list of pages to search for =item pod
                   directives in as targets of C<> and implicit links (empty
                   by default).  note, these are not filenames, but rather
                   page names like those that appear in L<> links.
  --[no]netscape - will use netscape html directives when applicable.
                   (default is not to use them).
  --outfile      - filename for the resulting html file (output sent to
                   stdout by default).
  --podpath      - colon-separated list of directories containing library
                   pods (empty by default).
  --podroot      - filesystem base directory from which all relative paths
                   in podpath stem (default is .).
  --[no]quiet    - supress some benign warning messages (default is off).
  --[no]recurse  - recurse on those subdirectories listed in podpath
                   (default behaviour).
  --title        - title that will appear in resulting html file.
  --[no]verbose  - self-explanatory (off by default).

END_OF_USAGE

sub parse_command_line {
    my ($opt_backlink,$opt_css,$opt_flush,$opt_header,$opt_help,$opt_htmldir,
	$opt_htmlroot,$opt_index,$opt_infile,$opt_libpods,$opt_netscape,
	$opt_outfile,$opt_podpath,$opt_podroot,$opt_quiet,$opt_recurse,
	$opt_title,$opt_verbose);

    unshift @ARGV, split ' ', $Config{pod2html} if $Config{pod2html};
    my $result = GetOptions(
			    'backlink=s' => \$opt_backlink,
			    'css=s'      => \$opt_css,
			    'flush'      => \$opt_flush,
			    'header!'    => \$opt_header,
			    'help'       => \$opt_help,
			    'htmldir=s'  => \$opt_htmldir,
			    'htmlroot=s' => \$opt_htmlroot,
			    'index!'     => \$opt_index,
			    'infile=s'   => \$opt_infile,
			    'libpods=s'  => \$opt_libpods,
			    'netscape!'  => \$opt_netscape,
			    'outfile=s'  => \$opt_outfile,
			    'podpath=s'  => \$opt_podpath,
			    'podroot=s'  => \$opt_podroot,
			    'quiet!'     => \$opt_quiet,
			    'recurse!'   => \$opt_recurse,
			    'title=s'    => \$opt_title,
			    'verbose!'   => \$opt_verbose,
			   );
    usage("-", "invalid parameters") if not $result;

    usage("-") if defined $opt_help;	# see if the user asked for help
    $opt_help = "";			# just to make -w shut-up.

    @podpath  = split(":", $opt_podpath) if defined $opt_podpath;
    @libpods  = split(":", $opt_libpods) if defined $opt_libpods;

    $backlink = $opt_backlink if defined $opt_backlink;
    $css      = $opt_css      if defined $opt_css;
    $header   = $opt_header   if defined $opt_header;
    $htmldir  = $opt_htmldir  if defined $opt_htmldir;
    $htmlroot = $opt_htmlroot if defined $opt_htmlroot;
    $doindex  = $opt_index    if defined $opt_index;
    $podfile  = $opt_infile   if defined $opt_infile;
    $netscape = $opt_netscape if defined $opt_netscape;
    $htmlfile = $opt_outfile  if defined $opt_outfile;
    $podroot  = $opt_podroot  if defined $opt_podroot;
    $quiet    = $opt_quiet    if defined $opt_quiet;
    $recurse  = $opt_recurse  if defined $opt_recurse;
    $title    = $opt_title    if defined $opt_title;
    $verbose  = $opt_verbose  if defined $opt_verbose;

    warn "Flushing item and directory caches\n"
	if $opt_verbose && defined $opt_flush;
    unlink($dircache, $itemcache) if defined $opt_flush;
}


my $saved_cache_key;

sub get_cache {
    my($dircache, $itemcache, $podpath, $podroot, $recurse) = @_;
    my @cache_key_args = @_;

    # A first-level cache:
    # Don't bother reading the cache files if they still apply
    # and haven't changed since we last read them.

    my $this_cache_key = cache_key(@cache_key_args);

    return if $saved_cache_key and $this_cache_key eq $saved_cache_key;

    # load the cache of %pages and %items if possible.  $tests will be
    # non-zero if successful.
    my $tests = 0;
    if (-f $dircache && -f $itemcache) {
	warn "scanning for item cache\n" if $verbose;
	$tests = load_cache($dircache, $itemcache, $podpath, $podroot);
    }

    # if we didn't succeed in loading the cache then we must (re)build
    #  %pages and %items.
    if (!$tests) {
	warn "scanning directories in pod-path\n" if $verbose;
	scan_podpath($podroot, $recurse, 0);
    }
    $saved_cache_key = cache_key(@cache_key_args);
}

sub cache_key {
    my($dircache, $itemcache, $podpath, $podroot, $recurse) = @_;
    return join('!', $dircache, $itemcache, $recurse,
	@$podpath, $podroot, stat($dircache), stat($itemcache));
}

#
# load_cache - tries to find if the caches stored in $dircache and $itemcache
#  are valid caches of %pages and %items.  if they are valid then it loads
#  them and returns a non-zero value.
#
sub load_cache {
    my($dircache, $itemcache, $podpath, $podroot) = @_;
    my($tests);
    local $_;

    $tests = 0;

    open(CACHE, "<$itemcache") ||
	die "$0: error opening $itemcache for reading: $!\n";
    $/ = "\n";

    # is it the same podpath?
    $_ = <CACHE>;
    chomp($_);
    $tests++ if (join(":", @$podpath) eq $_);

    # is it the same podroot?
    $_ = <CACHE>;
    chomp($_);
    $tests++ if ($podroot eq $_);

    # load the cache if its good
    if ($tests != 2) {
	close(CACHE);
	return 0;
    }

    warn "loading item cache\n" if $verbose;
    while (<CACHE>) {
	/(.*?) (.*)$/;
	$items{$1} = $2;
    }
    close(CACHE);

    warn "scanning for directory cache\n" if $verbose;
    open(CACHE, "<$dircache") ||
	die "$0: error opening $dircache for reading: $!\n";
    $/ = "\n";
    $tests = 0;

    # is it the same podpath?
    $_ = <CACHE>;
    chomp($_);
    $tests++ if (join(":", @$podpath) eq $_);

    # is it the same podroot?
    $_ = <CACHE>;
    chomp($_);
    $tests++ if ($podroot eq $_);

    # load the cache if its good
    if ($tests != 2) {
	close(CACHE);
	return 0;
    }

    warn "loading directory cache\n" if $verbose;
    while (<CACHE>) {
	/(.*?) (.*)$/;
	$pages{$1} = $2;
    }

    close(CACHE);

    return 1;
}

#
# scan_podpath - scans the directories specified in @podpath for directories,
#  .pod files, and .pm files.  it also scans the pod files specified in
#  @libpods for =item directives.
#
sub scan_podpath {
    my($podroot, $recurse, $append) = @_;
    my($pwd, $dir);
    my($libpod, $dirname, $pod, @files, @poddata);

    unless($append) {
	%items = ();
	%pages = ();
    }

    # scan each directory listed in @podpath
    $pwd = getcwd();
    chdir($podroot)
	|| die "$0: error changing to directory $podroot: $!\n";
    foreach $dir (@podpath) {
	scan_dir($dir, $recurse);
    }

    # scan the pods listed in @libpods for =item directives
    foreach $libpod (@libpods) {
	# if the page isn't defined then we won't know where to find it
	# on the system.
	next unless defined $pages{$libpod} && $pages{$libpod};

	# if there is a directory then use the .pod and .pm files within it.
	# NOTE: Only finds the first so-named directory in the tree.
#	if ($pages{$libpod} =~ /([^:]*[^(\.pod|\.pm)]):/) {
	if ($pages{$libpod} =~ /([^:]*(?<!\.pod)(?<!\.pm)):/) {
	    #  find all the .pod and .pm files within the directory
	    $dirname = $1;
	    opendir(DIR, $dirname) ||
		die "$0: error opening directory $dirname: $!\n";
	    @files = grep(/(\.pod|\.pm)\z/ && ! -d $_, readdir(DIR));
	    closedir(DIR);

	    # scan each .pod and .pm file for =item directives
	    foreach $pod (@files) {
		open(POD, "<$dirname/$pod") ||
		    die "$0: error opening $dirname/$pod for input: $!\n";
		@poddata = <POD>;
		close(POD);
		clean_data( \@poddata );

		scan_items( \%items, "$dirname/$pod", @poddata);
	    }

	    # use the names of files as =item directives too.
### Don't think this should be done this way - confuses issues.(WL)
###	    foreach $pod (@files) {
###		$pod =~ /^(.*)(\.pod|\.pm)$/;
###		$items{$1} = "$dirname/$1.html" if $1;
###	    }
	} elsif ($pages{$libpod} =~ /([^:]*\.pod):/ ||
		 $pages{$libpod} =~ /([^:]*\.pm):/) {
	    # scan the .pod or .pm file for =item directives
	    $pod = $1;
	    open(POD, "<$pod") ||
		die "$0: error opening $pod for input: $!\n";
	    @poddata = <POD>;
	    close(POD);
	    clean_data( \@poddata );

	    scan_items( \%items, "$pod", @poddata);
	} else {
	    warn "$0: shouldn't be here (line ".__LINE__."\n" unless $quiet;
	}
    }
    @poddata = ();	# clean-up a bit

    chdir($pwd)
	|| die "$0: error changing to directory $pwd: $!\n";

    # cache the item list for later use
    warn "caching items for later use\n" if $verbose;
    open(CACHE, ">$itemcache") ||
	die "$0: error open $itemcache for writing: $!\n";

    print CACHE join(":", @podpath) . "\n$podroot\n";
    foreach my $key (keys %items) {
	print CACHE "$key $items{$key}\n";
    }

    close(CACHE);

    # cache the directory list for later use
    warn "caching directories for later use\n" if $verbose;
    open(CACHE, ">$dircache") ||
	die "$0: error open $dircache for writing: $!\n";

    print CACHE join(":", @podpath) . "\n$podroot\n";
    foreach my $key (keys %pages) {
	print CACHE "$key $pages{$key}\n";
    }

    close(CACHE);
}

#
# scan_dir - scans the directory specified in $dir for subdirectories, .pod
#  files, and .pm files.  notes those that it finds.  this information will
#  be used later in order to figure out where the pages specified in L<>
#  links are on the filesystem.
#
sub scan_dir {
    my($dir, $recurse) = @_;
    my($t, @subdirs, @pods, $pod, $dirname, @dirs);
    local $_;

    @subdirs = ();
    @pods = ();

    opendir(DIR, $dir) ||
	die "$0: error opening directory $dir: $!\n";
    while (defined($_ = readdir(DIR))) {
	if (-d "$dir/$_" && $_ ne "." && $_ ne "..") {	    # directory
	    $pages{$_}  = "" unless defined $pages{$_};
	    $pages{$_} .= "$dir/$_:";
	    push(@subdirs, $_);
	} elsif (/\.pod\z/) {	    	    	    	    # .pod
	    s/\.pod\z//;
	    $pages{$_}  = "" unless defined $pages{$_};
	    $pages{$_} .= "$dir/$_.pod:";
	    push(@pods, "$dir/$_.pod");
	} elsif (/\.html\z/) { 	    	    	    	    # .html
	    s/\.html\z//;
	    $pages{$_}  = "" unless defined $pages{$_};
	    $pages{$_} .= "$dir/$_.pod:";
	} elsif (/\.pm\z/) { 	    	    	    	    # .pm
	    s/\.pm\z//;
	    $pages{$_}  = "" unless defined $pages{$_};
	    $pages{$_} .= "$dir/$_.pm:";
	    push(@pods, "$dir/$_.pm");
	}
    }
    closedir(DIR);

    # recurse on the subdirectories if necessary
    if ($recurse) {
	foreach my $subdir (@subdirs) {
	    scan_dir("$dir/$subdir", $recurse);
	}
    }
}

#
# scan_headings - scan a pod file for head[1-6] tags, note the tags, and
#  build an index.
#
sub scan_headings {
    my($sections, @data) = @_;
    my($tag, $which_head, $otitle, $listdepth, $index);

    # here we need	local $ignore = 0;
    #  unfortunately, we can't have it, because $ignore is lexical
    $ignore = 0;

    $listdepth = 0;
    $index = "";

    # scan for =head directives, note their name, and build an index
    #  pointing to each of them.
    foreach my $line (@data) {
	if ($line =~ /^=(head)([1-6])\s+(.*)/) {
	    ($tag, $which_head, $otitle) = ($1,$2,$3);

            my $title = depod( $otitle );
            my $name = htmlify( $title );
	    $$sections{$name} = 1;
	    $title = process_text( \$otitle );

	    while ($which_head != $listdepth) {
		if ($which_head > $listdepth) {
		    $index .= "\n" . ("\t" x $listdepth) . "<UL>\n";
		    $listdepth++;
		} elsif ($which_head < $listdepth) {
		    $listdepth--;
		    $index .= "\n" . ("\t" x $listdepth) . "</UL>\n";
		}
	    }

	    $index .= "\n" . ("\t" x $listdepth) . "<LI>" .
	              "<A HREF=\"#" . $name . "\">" .
		      $title . "</A></LI>";
	}
    }

    # finish off the lists
    while ($listdepth--) {
	$index .= "\n" . ("\t" x $listdepth) . "</UL>\n";
    }

    # get rid of bogus lists
    $index =~ s,\t*<UL>\s*</UL>\n,,g;

    $ignore = 1;	# restore old value;

    return $index;
}

#
# scan_items - scans the pod specified by $pod for =item directives.  we
#  will use this information later on in resolving C<> links.
#
sub scan_items {
    my( $itemref, $pod, @poddata ) = @_;
    my($i, $item);
    local $_;

    $pod =~ s/\.pod\z//;
    $pod .= ".html" if $pod;

    foreach $i (0..$#poddata) {
	my $txt = depod( $poddata[$i] );

	# figure out what kind of item it is.
	# Build string for referencing this item.
	if ( $txt =~ /\A=item\s+\*\s*(.*)\Z/s ) { # bullet
	    next unless $1;
	    $item = $1;
        } elsif( $txt =~ /\A=item\s+(?>\d+\.?)\s*(.*)\Z/s ) { # numbered list
	    $item = $1;
	} elsif( $txt =~ /\A=item\s+(.*)\Z/s ) { # plain item
	    $item = $1;
	} else {
	    next;
	}
	my $fid = fragment_id( $item );
	$$itemref{$fid} = "$pod" if $fid;
    }
}

#
# process_head - convert a pod head[1-6] tag and convert it to HTML format.
#
sub process_head {
    my($tag, $heading, $hasindex) = @_;

    # figure out the level of the =head
    $tag =~ /head([1-6])/;
    my $level = $1;

    if( $listlevel ){
	warn "$0: $podfile: unterminated list at =head in paragraph $paragraph.  ignoring.\n" unless $quiet;
        while( $listlevel ){
            process_back();
        }
    }

    $html .=  "<P>\n";
    if( $level == 1 && ! $top ){
	$html .=  "<A HREF=\"#__index__\"><SMALL>$backlink</SMALL></A>\n"
	    if $hasindex and $backlink;
	$html .=  "<HR>\n"
    }

    my $name = htmlify( depod( $heading ) );
    my $convert = process_text( \$heading );
    $html .=  "<H$level><A NAME=\"$name\">$convert</A></H$level>\n";
}


#
# emit_item_tag - print an =item's text
# Note: The global $EmittedItem is used for inhibiting self-references.
#
my $EmittedItem;

sub emit_item_tag($$$){
    my( $otext, $text, $compact ) = @_;
    my $item = fragment_id( $text );

    $EmittedItem = $item;
    ### print STDERR "emit_item_tag=$item ($text)\n";

    $html .=  '<STRONG>';
    if ($items_named{$item}++) {
	$html .=  process_text( \$otext );
    } else {
	my $name = 'item_' . $item;
	$html .=  qq{<A NAME="$name">}, process_text( \$otext ), '</A>';
    }
    $html .=  "$text</STRONG><BR>\n";
    undef( $EmittedItem );
}

sub emit_li {
    my( $tag ) = @_;
    if( $items_seen[$listlevel]++ == 0 ){
	push( @listend, "</$tag>" );
	$html .=  "<$tag>\n";
    }
    $html .=  $tag eq 'DL' ? '<DT>' : '<LI>';
}

#
# process_item - convert a pod item tag and convert it to HTML format.
#
sub process_item {
    my( $otext ) = @_;

    # lots of documents start a list without doing an =over.  this is
    # bad!  but, the proper thing to do seems to be to just assume
    # they did do an =over.  so warn them once and then continue.
    if( $listlevel == 0 ){
	warn "$0: $podfile: unexpected =item directive in paragraph $paragraph.  ignoring.\n" unless $quiet;
	process_over();
    }

    # formatting: insert a paragraph if preceding item has >1 paragraph
    if( $after_lpar ){
	$html .=  "<P></P>\n";
	$after_lpar = 0;
    }

    # remove formatting instructions from the text
    my $text = depod( $otext );

    # all the list variants:
    if( $text =~ /\A\*/ ){ # bullet
        emit_li( 'UL' );
	if ($text =~ /\A\*\s+(.+)\Z/s ) { # with additional text
	    my $tag = $1;
	    $otext =~ s/\A\*\s+//;
	    emit_item_tag( $otext, $tag, 1 );
	}

    } elsif( $text =~ /\A\d+/ ){ # numbered list
	emit_li( 'OL' );
	if ($text =~ /\A(?>\d+\.?)\s*(.+)\Z/s ) { # with additional text
	    my $tag = $1;
	    $otext =~ s/\A\d+\.?\s*//;
	    emit_item_tag( $otext, $tag, 1 );
	}

    } else {			# definition list
	emit_li( 'DL' );
	if ($text =~ /\A(.+)\Z/s ){ # should have text
	    emit_item_tag( $otext, $text, 1 );
	}
       $html .=  '<DD>';
    }
    $html .=  "\n";
}

#
# process_over - process a pod over tag and start a corresponding HTML list.
#
sub process_over {
    # start a new list
    $listlevel++;
    push( @items_seen, 0 );
    $after_lpar = 0;
}

#
# process_back - process a pod back tag and convert it to HTML format.
#
sub process_back {
    if( $listlevel == 0 ){
	warn "$0: $podfile: unexpected =back directive in paragraph $paragraph.  ignoring.\n" unless $quiet;
	return;
    }

    # close off the list.  note, I check to see if $listend[$listlevel] is
    # defined because an =item directive may have never appeared and thus
    # $listend[$listlevel] may have never been initialized.
    $listlevel--;
    if( defined $listend[$listlevel] ){
	$html .=  '<P></P>' if $after_lpar;
	$html .=  $listend[$listlevel];
        $html .=  "\n";
        pop( @listend );
    }
    $after_lpar = 0;

    # clean up item count
    pop( @items_seen );
}

#
# process_cut - process a pod cut tag, thus start ignoring pod directives.
#
sub process_cut {
    $ignore = 1;
}

#
# process_pod - process a pod pod tag, thus stop ignoring pod directives
# until we see a corresponding cut.
#
sub process_pod {
    # no need to set $ignore to 0 cause the main loop did it
}

#
# process_for - process a =for pod tag.  if it's for html, spit
# it out verbatim, if illustration, center it, otherwise ignore it.
#
sub process_for {
    my($whom, $text) = @_;
    if ( $whom =~ /^(pod2)?html$/i) {
	$html .=  $text;
    } elsif ($whom =~ /^illustration$/i) {
        1 while chomp $text;
	for my $ext (qw[.png .gif .jpeg .jpg .tga .pcl .bmp]) {
	  $text .= $ext, last if -r "$text$ext";
	}
        $html .=  qq{<p align = "center"><img src = "$text" alt = "$text illustration"></p>};
    }
}

#
# process_begin - process a =begin pod tag.  this pushes
# whom we're beginning on the begin stack.  if there's a
# begin stack, we only print if it us.
#
sub process_begin {
    my($whom, $text) = @_;
    $whom = lc($whom);
    push (@begin_stack, $whom);
    if ( $whom =~ /^(pod2)?html$/) {
	$html .=  $text if $text;
    }
}

#
# process_end - process a =end pod tag.  pop the
# begin stack.  die if we're mismatched.
#
sub process_end {
    my($whom, $text) = @_;
    $whom = lc($whom);
    if ($begin_stack[-1] ne $whom ) {
	die "Unmatched begin/end at chunk $paragraph\n"
    } 
    pop( @begin_stack );
}

#
# process_pre - indented paragraph, made into <PRE></PRE>
#
sub process_pre {
    my( $text ) = @_;
    my( $rest );
    return if $ignore;

    $rest = $$text;

    # insert spaces in place of tabs
    $rest =~ s#.*#
	    my $line = $&;
	    1 while $line =~ s/\t+/' ' x (length($&) * 8 - length($`) % 8)/e;
	    $line;
	#eg;

    # convert some special chars to HTML escapes
    $rest =~ s/&/&amp;/g;
    $rest =~ s/</&lt;/g;
    $rest =~ s/>/&gt;/g;
    $rest =~ s/"/&quot;/g;

    # try and create links for all occurrences of perl.* within
    # the preformatted text.
    $rest =~ s{
	         (\s*)(perl\w+)
	      }{
		 if ( defined $pages{$2} ){	# is a link
		     qq($1<A HREF="$htmlroot/$pages{$2}">$2</A>);
		 } elsif (defined $pages{dosify($2)}) {	# is a link
		     qq($1<A HREF="$htmlroot/$pages{dosify($2)}">$2</A>);
		 } else {
		     "$1$2";
		 }
	      }xeg;
     $rest =~ s{
		 (<A\ HREF="?) ([^>:]*:)? ([^>:]*) \.pod: ([^>:]*:)?
               }{
                  my $url ;
                  if ( $htmlfileurl ne '' ){
		     # Here, we take advantage of the knowledge 
		     # that $htmlfileurl ne '' implies $htmlroot eq ''.
		     # Since $htmlroot eq '', we need to prepend $htmldir
		     # on the fron of the link to get the absolute path
		     # of the link's target. We check for a leading '/'
		     # to avoid corrupting links that are #, file:, etc.
		     my $old_url = $3 ;
		     $old_url = "$htmldir$old_url" if $old_url =~ m{^\/};
 		     $url = relativize_url( "$old_url.html", $htmlfileurl );
	          } else {
		     $url = "$3.html" ;
		  }
		  "$1$url" ;
	       }xeg;

    # Look for embedded URLs and make them into links.  We don't
    # relativize them since they are best left as the author intended.

    my $urls = '(' . join ('|', qw{
                http
                telnet
		mailto
		news
                gopher
                file
                wais
                ftp
            } ) 
        . ')';
  
    my $ltrs = '\w';
    my $gunk = '/#~:.?+=&%@!\-';
    my $punc = '.:?\-';
    my $any  = "${ltrs}${gunk}${punc}";

    $rest =~ s{
        \b                          # start at word boundary
        (                           # begin $1  {
          $urls     :               # need resource and a colon
	  (?!:)                     # Ignore File::, among others.
          [$any] +?                 # followed by on or more
                                    #  of any valid character, but
                                    #  be conservative and take only
                                    #  what you need to....
        )                           # end   $1  }
        (?=                         # look-ahead non-consumptive assertion
                [$punc]*            # either 0 or more puntuation
                [^$any]             #   followed by a non-url char
            |                       # or else
                $                   #   then end of the string
        )
      }{<A HREF="$1">$1</A>}igox;

    # text should be as it is (verbatim)
    $$text = $rest;
}


#
# pure text processing
#
# pure_text/inIS_text: differ with respect to automatic C<> recognition.
# we don't want this to happen within IS
#
sub pure_text($){
    my $text = shift();
    process_puretext( $text, \$ptQuote, 1 );
}

sub inIS_text($){
    my $text = shift();
    process_puretext( $text, \$ptQuote, 0 );
}

#
# process_puretext - process pure text (without pod-escapes) converting
#  double-quotes and handling implicit C<> links.
#
sub process_puretext {
    my($text, $quote, $notinIS) = @_;

    ## Guessing at func() or [$@%&]*var references in plain text is destined
    ## to produce some strange looking ref's. uncomment to disable:
    ## $notinIS = 0;

    my(@words, $lead, $trail);

    # convert double-quotes to single-quotes
    if( $$quote && $text =~ s/"/''/s ){
        $$quote = 0;
    }
    while ($text =~ s/"([^"]*)"/``$1''/sg) {};
    $$quote = 1 if $text =~ s/"/``/s;

    # keep track of leading and trailing white-space
    $lead  = ($text =~ s/\A(\s+)//s ? $1 : "");
    $trail = ($text =~ s/(\s+)\Z//s ? $1 : "");

    # split at space/non-space boundaries
    @words = split( /(?<=\s)(?=\S)|(?<=\S)(?=\s)/, $text );

    # process each word individually
    foreach my $word (@words) {
	# skip space runs
 	next if $word =~ /^\s*$/;
	# see if we can infer a link
	if( $notinIS && $word =~ /^(\w+)\((.*)\)$/ ) {
	    # has parenthesis so should have been a C<> ref
            ## try for a pagename (perlXXX(1))?
            my( $func, $args ) = ( $1, $2 );
            if( $args =~ /^\d+$/ ){
                my $url = page_sect( $word, '' );
                if( defined $url ){
                    $word = "<A HREF=\"$url\">the $word manpage</A>";
                    next;
                }
            }
            ## try function name for a link, append tt'ed argument list
            $word = emit_C( $func, '', "($args)");

#### disabled. either all (including $\W, $\w+{.*} etc.) or nothing.
##      } elsif( $notinIS && $word =~ /^[\$\@%&*]+\w+$/) {
##	    # perl variables, should be a C<> ref
##	    $word = emit_C( $word );

	} elsif ($word =~ m,^\w+://\w,) {
	    # looks like a URL
            # Don't relativize it: leave it as the author intended
	    $word = qq(<A HREF="$word">$word</A>);
	} elsif ($word =~ /[\w.-]+\@[\w-]+\.\w/) {
	    # looks like an e-mail address
	    my ($w1, $w2, $w3) = ("", $word, "");
	    ($w1, $w2, $w3) = ("(", $1, ")$2") if $word =~ /^\((.*?)\)(,?)/;
	    ($w1, $w2, $w3) = ("&lt;", $1, "&gt;$2") if $word =~ /^<(.*?)>(,?)/;
	    $word = qq($w1<A HREF="mailto:$w2">$w2</A>$w3);
	} elsif ($word !~ /[a-z]/ && $word =~ /[A-Z]/) {  # all uppercase?
	    $word = html_escape($word) if $word =~ /["&<>]/;
	    $word = "\n<FONT SIZE=-1>$word</FONT>" if $netscape;
	} else { 
	    $word = html_escape($word) if $word =~ /["&<>]/;
	}
    }

    # put everything back together
    return $lead . join( '', @words ) . $trail;
}


#
# process_text - handles plaintext that appears in the input pod file.
# there may be pod commands embedded within the text so those must be
# converted to html commands.
#

sub process_text1($$;$$);
sub pattern ($) { $_[0] ? '[^\S\n]+'.('>' x ($_[0] + 1)) : '>' }
sub closing ($) { local($_) = shift; (defined && s/\s+$//) ? length : 0 }

sub process_text {
    return if $ignore;
    my( $tref ) = @_;
    my $res = process_text1( 0, $tref );
    $$tref = $res;
}

sub process_text1($$;$$){
    my( $lev, $rstr, $func, $closing ) = @_;
    my $res = '';

    unless (defined $func) {
	$func = '';
	$lev++;
    }

    if( $func eq 'B' ){
	# B<text> - boldface
	$res = '<STRONG>' . process_text1( $lev, $rstr ) . '</STRONG>';

    } elsif( $func eq 'C' ){
	# C<code> - can be a ref or <CODE></CODE>
	# need to extract text
	my $par = go_ahead( $rstr, 'C', $closing );

	## clean-up of the link target
        my $text = depod( $par );

	### my $x = $par =~ /[BI]</ ? 'yes' : 'no' ;
        ### print STDERR "-->call emit_C($par) lev=$lev, par with BI=$x\n"; 

	$res = emit_C( $text, $lev > 1 || ($par =~ /[BI]</) );

    } elsif( $func eq 'E' ){
	# E<x> - convert to character
	$$rstr =~ s/^([^>]*)>//;
	my $escape = $1;
	$escape =~ s/^(\d+|X[\dA-F]+)$/#$1/i;
	$res = "&$escape;";

    } elsif( $func eq 'F' ){
	# F<filename> - italizice
	$res = '<EM>' . process_text1( $lev, $rstr ) . '</EM>';

    } elsif( $func eq 'I' ){
	# I<text> - italizice
	$res = '<EM>' . process_text1( $lev, $rstr ) . '</EM>';

    } elsif( $func eq 'L' ){
	# L<link> - link
	## L<text|cross-ref> => produce text, use cross-ref for linking 
	## L<cross-ref> => make text from cross-ref
	## need to extract text
	my $par = go_ahead( $rstr, 'L', $closing );

        # some L<>'s that shouldn't be:
	# a) full-blown URL's are emitted as-is
        if( $par =~ m{^\w+://}s ){
	    return make_URL_href( $par );
	}
        # b) C<...> is stripped and treated as C<>
        if( $par =~ /^C<(.*)>$/ ){
	    my $text = depod( $1 );
 	    return emit_C( $text, $lev > 1 || ($par =~ /[BI]</) );
	}

	# analyze the contents
	$par =~ s/\n/ /g;   # undo word-wrapped tags
        my $opar = $par;
	my $linktext;
	if( $par =~ s{^([^|]+)\|}{} ){
	    $linktext = $1;
	}
    
	# make sure sections start with a /
	$par =~ s{^"}{/"};

	my( $page, $section, $ident );

	# check for link patterns
	if( $par =~ m{^([^/]+?)/(?!")(.*?)$} ){     # name/ident
            # we've got a name/ident (no quotes) 
            ( $page, $ident ) = ( $1, $2 );
            ### print STDERR "--> L<$par> to page $page, ident $ident\n";

	} elsif( $par =~ m{^(.*?)/"?(.*?)"?$} ){ # [name]/"section"
            # even though this should be a "section", we go for ident first
	    ( $page, $ident ) = ( $1, $2 );
            ### print STDERR "--> L<$par> to page $page, section $section\n";

	} elsif( $par =~ /\s/ ){  # this must be a section with missing quotes
	    ( $page, $section ) = ( '', $par );
            ### print STDERR "--> L<$par> to void page, section $section\n";

        } else {
	    ( $page, $section ) = ( $par, '' );
            ### print STDERR "--> L<$par> to page $par, void section\n";
	}

        # now, either $section or $ident is defined. the convoluted logic
        # below tries to resolve L<> according to what the user specified.
        # failing this, we try to find the next best thing...
        my( $url, $ltext, $fid );

        RESOLVE: {
            if( defined $ident ){
                ## try to resolve $ident as an item
	        ( $url, $fid ) = coderef( $page, $ident );
                if( $url ){
                    if( ! defined( $linktext ) ){
                        $linktext = $ident;
                        $linktext .= " in " if $ident && $page;
                        $linktext .= "the $page manpage" if $page;
                    }
                    ###  print STDERR "got coderef url=$url\n";
                    last RESOLVE;
                }
                ## no luck: go for a section (auto-quoting!)
                $section = $ident;
            }
            ## now go for a section
            my $htmlsection = htmlify( $section );
 	    $url = page_sect( $page, $htmlsection );
            if( $url ){
                if( ! defined( $linktext ) ){
                    $linktext = $section;
                    $linktext .= " in " if $section && $page;
                    $linktext .= "the $page manpage" if $page;
                }
                ### print STDERR "got page/section url=$url\n";
                last RESOLVE;
            }
            ## no luck: go for an ident 
            if( $section ){
                $ident = $section;
            } else {
                $ident = $page;
                $page  = undef();
            }
            ( $url, $fid ) = coderef( $page, $ident );
            if( $url ){
                if( ! defined( $linktext ) ){
                    $linktext = $ident;
                    $linktext .= " in " if $ident && $page;
                    $linktext .= "the $page manpage" if $page;
                }
                ### print STDERR "got section=>coderef url=$url\n";
                last RESOLVE;
            }

            # warning; show some text.
            $linktext = $opar unless defined $linktext;
            warn "$0: $podfile: cannot resolve L<$opar> in paragraph $paragraph." unless $quiet;
        }

        # now we have an URL or just plain code
        $$rstr = $linktext . '>' . $$rstr;
        if( defined( $url ) ){
            $res = "<A HREF=\"$url\">" . process_text1( $lev, $rstr ) . '</A>';
        } else {
	    $res = '<EM>' . process_text1( $lev, $rstr ) . '</EM>';
        }

    } elsif( $func eq 'S' ){
	# S<text> - non-breaking spaces
	$res = process_text1( $lev, $rstr );
	$res =~ s/ /&nbsp;/g;

    } elsif( $func eq 'X' ){
	# X<> - ignore
	$$rstr =~ s/^[^>]*>//;

    } elsif( $func eq 'Z' ){
	# Z<> - empty 
	warn "$0: $podfile: invalid X<> in paragraph $paragraph."
	    unless $$rstr =~ s/^>// or $quiet;

    } else {
        my $term = pattern $closing;
	while( $$rstr =~ s/\A(.*?)(([BCEFILSXZ])<(<+[^\S\n]+)?|$term)//s ){
	    # all others: either recurse into new function or
	    # terminate at closing angle bracket(s)
	    my $pt = $1;
            $pt .= $2 if !$3 &&  $lev == 1;
	    $res .= $lev == 1 ? pure_text( $pt ) : inIS_text( $pt );
	    return $res if !$3 && $lev > 1;
            if( $3 ){
		$res .= process_text1( $lev, $rstr, $3, closing $4 );
 	    }
	}
	if( $lev == 1 ){
	    $res .= pure_text( $$rstr );
	} else {
	    warn "$0: $podfile: undelimited $func<> in paragraph $paragraph." unless $quiet;
	}
    }
    return $res;
}

#
# go_ahead: extract text of an IS (can be nested)
#
sub go_ahead($$$){
    my( $rstr, $func, $closing ) = @_;
    my $res = '';
    my @closing = ($closing);
    while( $$rstr =~
      s/\A(.*?)(([BCEFILSXZ])<(<+[^\S\n]+)?|@{[pattern $closing[0]]})//s ){
	$res .= $1;
	unless( $3 ){
	    shift @closing;
	    return $res unless @closing;
	} else {
	    unshift @closing, closing $4;
	}
	$res .= $2;
    }
    warn "$0: $podfile: undelimited $func<> in paragraph $paragraph." unless $quiet;
    return $res;
}

#
# emit_C - output result of C<text>
#    $text is the depod-ed text
#
sub emit_C($;$$){
    my( $text, $nocode, $args ) = @_;
    $args = '' unless defined $args;
    my $res;
    my( $url, $fid ) = coderef( undef(), $text );

    # need HTML-safe text
    my $linktext = html_escape( "$text$args" );

    if( defined( $url ) &&
        (!defined( $EmittedItem ) || $EmittedItem ne $fid ) ){
	$res = "<A HREF=\"$url\"><CODE>$linktext</CODE></A>";
    } elsif( 0 && $nocode ){
	$res = $linktext;
    } else {
	$res = "<CODE>$linktext</CODE>";
    }
    return $res;
}

#
# html_escape: make text safe for HTML
#
sub html_escape {
    my $rest = $_[0];
    $rest   =~ s/&/&amp;/g;
    $rest   =~ s/</&lt;/g;
    $rest   =~ s/>/&gt;/g;
    $rest   =~ s/"/&quot;/g;
    return $rest;
} 


#
# dosify - convert filenames to 8.3
#
sub dosify {
    my($str) = @_;
    return lc($str) if $^O eq 'VMS';     # VMS just needs casing
    if ($Is83) {
        $str = lc $str;
        $str =~ s/(\.\w+)/substr ($1,0,4)/ge;
        $str =~ s/(\w+)/substr ($1,0,8)/ge;
    }
    return $str;
}

#
# page_sect - make an URL from the text of a L<>
#
sub page_sect($$) {
    my( $page, $section ) = @_;
    my( $linktext, $page83, $link);	# work strings

    # check if we know that this is a section in this page
    if (!defined $pages{$page} && defined $sections{$page}) {
	$section = $page;
	$page = "";
        ### print STDERR "reset page='', section=$section\n";
    }

    $page83=dosify($page);
    $page=$page83 if (defined $pages{$page83});
    if ($page eq "") {
	$link = "#" . htmlify( $section );
    } elsif ( $page =~ /::/ ) {
	$page =~ s,::,/,g;
	# Search page cache for an entry keyed under the html page name,
	# then look to see what directory that page might be in.  NOTE:
	# this will only find one page. A better solution might be to produce
	# an intermediate page that is an index to all such pages.
	my $page_name = $page ;
	$page_name =~ s,^.*/,,s ;
	if ( defined( $pages{ $page_name } ) && 
	     $pages{ $page_name } =~ /([^:]*$page)\.(?:pod|pm):/ 
	   ) {
	    $page = $1 ;
	}
	else {
	    # NOTE: This branch assumes that all A::B pages are located in
	    # $htmlroot/A/B.html . This is often incorrect, since they are
	    # often in $htmlroot/lib/A/B.html or such like. Perhaps we could
	    # analyze the contents of %pages and figure out where any
	    # cousins of A::B are, then assume that.  So, if A::B isn't found,
	    # but A::C is found in lib/A/C.pm, then A::B is assumed to be in
	    # lib/A/B.pm. This is also limited, but it's an improvement.
	    # Maybe a hints file so that the links point to the correct places
	    # nonetheless?

	}
	$link = "$htmlroot/$page.html";
	$link .= "#" . htmlify( $section ) if ($section);
    } elsif (!defined $pages{$page}) {
	$link = "";
    } else {
	$section = htmlify( $section ) if $section ne "";
        ### print STDERR "...section=$section\n";

	# if there is a directory by the name of the page, then assume that an
	# appropriate section will exist in the subdirectory
#	if ($section ne "" && $pages{$page} =~ /([^:]*[^(\.pod|\.pm)]):/) {
	if ($section ne "" && $pages{$page} =~ /([^:]*(?<!\.pod)(?<!\.pm)):/) {
	    $link = "$htmlroot/$1/$section.html";
            ### print STDERR "...link=$link\n";

	# since there is no directory by the name of the page, the section will
	# have to exist within a .html of the same name.  thus, make sure there
	# is a .pod or .pm that might become that .html
	} else {
	    $section = "#$section" if $section;
            ### print STDERR "...section=$section\n";

	    # check if there is a .pod with the page name
	    if ($pages{$page} =~ /([^:]*)\.pod:/) {
		$link = "$htmlroot/$1.html$section";
	    } elsif ($pages{$page} =~ /([^:]*)\.pm:/) {
		$link = "$htmlroot/$1.html$section";
	    } else {
		$link = "";
	    }
	}
    }

    if ($link) {
	# Here, we take advantage of the knowledge that $htmlfileurl ne ''
	# implies $htmlroot eq ''. This means that the link in question
	# needs a prefix of $htmldir if it begins with '/'. The test for
	# the initial '/' is done to avoid '#'-only links, and to allow
	# for other kinds of links, like file:, ftp:, etc.
        my $url ;
        if (  $htmlfileurl ne '' ) {
            $link = "$htmldir$link" if $link =~ m{^/}s;
            $url = relativize_url( $link, $htmlfileurl );
# print( "  b: [$link,$htmlfileurl,$url]\n" );
	}
	else {
            $url = $link ;
	}
	return $url;

    } else {
	return undef();
    }
}

#
# relativize_url - convert an absolute URL to one relative to a base URL.
# Assumes both end in a filename.
#
sub relativize_url {
    my ($dest,$source) = @_ ;

    my ($dest_volume,$dest_directory,$dest_file) = 
        File::Spec::Unix->splitpath( $dest ) ;
    $dest = File::Spec::Unix->catpath( $dest_volume, $dest_directory, '' ) ;

    my ($source_volume,$source_directory,$source_file) = 
        File::Spec::Unix->splitpath( $source ) ;
    $source = File::Spec::Unix->catpath( $source_volume, $source_directory, '' ) ;

    my $rel_path = '' ;
    if ( $dest ne '' ) {
       $rel_path = File::Spec::Unix->abs2rel( $dest, $source ) ;
    }

    if ( $rel_path ne ''                && 
         substr( $rel_path, -1 ) ne '/' &&
         substr( $dest_file, 0, 1 ) ne '#' 
        ) {
        $rel_path .= "/$dest_file" ;
    }
    else {
        $rel_path .= "$dest_file" ;
    }

    return $rel_path ;
}


#
# coderef - make URL from the text of a C<>
#
sub coderef($$){
    my( $page, $item ) = @_;
    my( $url );

    my $fid = fragment_id( $item );
    if( defined( $page ) ){
	# we have been given a $page...
	$page =~ s{::}{/}g;

	# Do we take it? Item could be a section!
	my $base = $items{$fid} || "";
	$base =~ s{[^/]*/}{};
	if( $base ne "$page.html" ){
            ###   print STDERR "coderef( $page, $item ): items{$fid} = $items{$fid} = $base => discard page!\n";
	    $page = undef();
	}

    } else {
        # no page - local items precede cached items
	if( defined( $fid ) ){
	    if(  exists $local_items{$fid} ){
		$page = $local_items{$fid};
	    } else {
		$page = $items{$fid};
	    }
	}
    }

    # if there was a pod file that we found earlier with an appropriate
    # =item directive, then create a link to that page.
    if( defined $page ){
	if( $page ){
            if( exists $pages{$page} and $pages{$page} =~ /([^:.]*)\.[^:]*:/){
		$page = $1 . '.html';
	    }
	    my $link = "$htmlroot/$page#item_$fid";

	    # Here, we take advantage of the knowledge that $htmlfileurl
	    # ne '' implies $htmlroot eq ''.
	    if (  $htmlfileurl ne '' ) {
		$link = "$htmldir$link" ;
		$url = relativize_url( $link, $htmlfileurl ) ;
	    } else {
		$url = $link ;
	    }
	} else {
	    $url = "#item_" . $fid;
	}

	confess "url has space: $url" if $url =~ /"[^"]*\s[^"]*"/;
    }       
    return( $url, $fid );
}



#
# Adapted from Nick Ing-Simmons' PodToHtml package.
sub relative_url {
    my $source_file = shift ;
    my $destination_file = shift;

    my $source = URI::file->new_abs($source_file);
    my $uo = URI::file->new($destination_file,$source)->abs;
    return $uo->rel->as_string;
}


#
# finish_list - finish off any pending HTML lists.  this should be called
# after the entire pod file has been read and converted.
#
sub finish_list {
    while ($listlevel > 0) {
	$html .=  "</DL>\n";
	$listlevel--;
    }
}

#
# htmlify - converts a pod section specification to a suitable section
# specification for HTML. Note that we keep spaces and special characters
# except ", ? (Netscape problem) and the hyphen (writer's problem...).
#
sub htmlify {
    my( $heading) = @_;
    $heading =~ s/(\s+)/ /g;
    $heading =~ s/\s+\Z//;
    $heading =~ s/\A\s+//;
    # The hyphen is a disgrace to the English language.
    $heading =~ s/[-"?]//g;
    $heading = lc( $heading );
    return $heading;
}

#
# depod - convert text by eliminating all interior sequences
# Note: can be called with copy or modify semantics
#
my %E2c;
$E2c{lt}     = '<';
$E2c{gt}     = '>';
$E2c{sol}    = '/';
$E2c{verbar} = '|';
$E2c{amp}    = '&'; # in Tk's pods

sub depod1($;$$);

sub depod($){
    my $string;
    if( ref( $_[0] ) ){
	$string =  ${$_[0]};
        ${$_[0]} = depod1( \$string );
    } else {
	$string =  $_[0];
        depod1( \$string );
    }    
}

sub depod1($;$$){
  my( $rstr, $func, $closing ) = @_;
  my $res = '';
  return $res unless defined $$rstr;
  if( ! defined( $func ) ){
      # skip to next begin of an interior sequence
      while( $$rstr =~ s/\A(.*?)([BCEFILSXZ])<(<+[^\S\n]+)?// ){
         # recurse into its text
	  $res .= $1 . depod1( $rstr, $2, closing $3);
      }
      $res .= $$rstr;
  } elsif( $func eq 'E' ){
      # E<x> - convert to character
      $$rstr =~ s/^([^>]*)>//;
      $res .= $E2c{$1} || "";
  } elsif( $func eq 'X' ){
      # X<> - ignore
      $$rstr =~ s/^[^>]*>//;
  } elsif( $func eq 'Z' ){
      # Z<> - empty 
      $$rstr =~ s/^>//;
  } else {
      # all others: either recurse into new function or
      # terminate at closing angle bracket
      my $term = pattern $closing;
      while( $$rstr =~ s/\A(.*?)(([BCEFILSXZ])<(<+[^\S\n]+)?|$term)// ){
	  $res .= $1;
	  last unless $3;
          $res .= depod1( $rstr, $3, closing $4 );
      }
      ## If we're here and $2 ne '>': undelimited interior sequence.
      ## Ignored, as this is called without proper indication of where we are.
      ## Rely on process_text to produce diagnostics.
  }
  return $res;
}

#
# fragment_id - construct a fragment identifier from:
#   a) =item text
#   b) contents of C<...>
#
my @hc;
sub fragment_id {
    my $text = shift();
    $text =~ s/\s+\Z//s;
    if( $text ){
	# a method or function?
	return $1 if $text =~ /(\w+)\s*\(/;
	return $1 if $text =~ /->\s*(\w+)\s*\(?/;

	# a variable name?
	return $1 if $text =~ /^([$@%*]\S+)/;

	# some pattern matching operator?
	return $1 if $text =~ m|^(\w+/).*/\w*$|;

	# fancy stuff... like "do { }"
	return $1 if $text =~ m|^(\w+)\s*{.*}$|;

	# honour the perlfunc manpage: func [PAR[,[ ]PAR]...]
	# and some funnies with ... Module ...
	return $1 if $text =~ m{^([a-z\d]+)(\s+[A-Z\d,/& ]+)?$};
	return $1 if $text =~ m{^([a-z\d]+)\s+Module(\s+[A-Z\d,/& ]+)?$};

	# text? normalize!
	$text =~ s/\s+/_/sg;
	$text =~ s{(\W)}{
         defined( $hc[ord($1)] ) ? $hc[ord($1)]
                 : ( $hc[ord($1)] = sprintf( "%%%02X", ord($1) ) ) }gxe;
        $text = substr( $text, 0, 50 );
    } else {
	return undef();
    }
}

#
# make_URL_href - generate HTML href from URL
# Special treatment for CGI queries.
#
sub make_URL_href($){
    my( $url ) = @_;
    if( $url !~ 
        s{^(http:[-\w/#~:.+=&%@!]+)(\?.*)$}{<A HREF="$1$2">$1</A>}i ){
        $url = "<A HREF=\"$url\">$url</A>";
    }
    return $url;
}

1;