#!/usr/local/bin/perl -wT
#
# W3C Link Checker
# by Hugo Haas <hugo@w3.org>
# (c) 1999-2011 World Wide Web Consortium
# based on Renaud Bruyeron's checklink.pl
#
# This program is licensed under the W3C(r) Software License:
#       http://www.w3.org/Consortium/Legal/copyright-software
#
# The documentation is at:
#       http://validator.w3.org/docs/checklink.html
#
# See the Mercurial interface at:
#       http://dvcs.w3.org/hg/link-checker/
#
# An online version is available at:
#       http://validator.w3.org/checklink
#
# Comments and suggestions should be sent to the www-validator mailing list:
#       www-validator@w3.org (with 'checklink' in the subject)
#       http://lists.w3.org/Archives/Public/www-validator/ (archives)
#
# Small modifications in March 2020 by Karl Berry <karl@freefriends.org>
# (contributed under the same license, or public domain if you prefer).
# I started from https://metacpan.org/release/W3C-LinkChecker, version 4.81.
# - (&simple_request)   ignore "Argument isn't numeric" warnings.
# - (%Opts, &check_uri) new option --exclude-url-file; see --help message.
# - (&parse_arguments)  allow multiple -X options.
# - (&check_uri)        missing argument to hprintf.
# - (&hprintf)          avoid useless warnings when undef is returned.
# The ideas are (1) to avoid rechecking every url during development,
# and (2) to make the exclude list easier to maintain,
# and (3) to eliminate useless warnings from the code,
#
# For GNU Automake, this program is used by the checklinkx target
# in doc/local.mk to check the (html output of) automake manual.

use strict;
use 5.008;

# Get rid of potentially unsafe and unneeded environment variables.
delete(@ENV{qw(IFS CDPATH ENV BASH_ENV)});
$ENV{PATH} = undef;

# ...but we want PERL5?LIB honored even in taint mode, see perlsec, perl5lib,
# http://www.mail-archive.com/cpan-testers-discuss%40perl.org/msg01064.html
use Config qw(%Config);
use lib map { /(.*)/ }
    defined($ENV{PERL5LIB}) ? split(/$Config{path_sep}/, $ENV{PERL5LIB}) :
    defined($ENV{PERLLIB})  ? split(/$Config{path_sep}/, $ENV{PERLLIB}) :
                              ();

# -----------------------------------------------------------------------------

package W3C::UserAgent;

use LWP::RobotUA 1.19 qw();
use LWP::UserAgent qw();
use Net::HTTP::Methods 5.833 qw();    # >= 5.833 for 4kB cookies (#6678)

# if 0, ignore robots exclusion (useful for testing)
use constant USE_ROBOT_UA => 1;

if (USE_ROBOT_UA) {
    @W3C::UserAgent::ISA = qw(LWP::RobotUA);
}
else {
    @W3C::UserAgent::ISA = qw(LWP::UserAgent);
}

sub new
{
    my $proto = shift;
    my $class = ref($proto) || $proto;
    my ($name, $from, $rules) = @_;

    # For security/privacy reasons, if $from was not given, do not send it.
    # Cheat by defining something for the constructor, and resetting it later.
    my $from_ok = $from;
    $from ||= 'www-validator@w3.org';

    my $self;
    if (USE_ROBOT_UA) {
        $self = $class->SUPER::new($name, $from, $rules);
    }
    else {
        my %cnf;
        @cnf{qw(agent from)} = ($name, $from);
        $self = LWP::UserAgent->new(%cnf);
        $self = bless $self, $class;
    }

    $self->from(undef) unless $from_ok;

    $self->env_proxy();

    $self->allow_private_ips(1);

    $self->protocols_forbidden([qw(mailto javascript)]);

    return $self;
}

sub allow_private_ips
{
    my $self = shift;
    if (@_) {
        $self->{Checklink_allow_private_ips} = shift;
        if (!$self->{Checklink_allow_private_ips}) {

            # Pull in dependencies
            require Net::IP;
            require Socket;
            require Net::hostent;
        }
    }
    return $self->{Checklink_allow_private_ips};
}

sub redirect_progress_callback
{
    my $self = shift;
    $self->{Checklink_redirect_callback} = shift if @_;
    return $self->{Checklink_redirect_callback};
}

sub simple_request
{
    my $self = shift;

    my $response = $self->ip_disallowed($_[0]->uri());

    # RFC 2616, section 15.1.3
    $_[0]->remove_header("Referer")
        if ($_[0]->referer() &&
        (!$_[0]->uri()->secure() && URI->new($_[0]->referer())->secure()));

    $response ||= do {
        local $SIG{__WARN__} =
            sub {    # Suppress RobotRules warnings, rt.cpan.org #18902
                     # Suppress "Argument isn't numeric" warnings, see below.
            warn($_[0])
              if ($_[0]
                  && $_[0] !~ /^RobotRules/
                  && $_[0] !~ /^Argument .* isn't numeric.*Response\.pm/
                 );
            };

        # @@@ Why not just $self->SUPER::simple_request? [--unknown]
        # --- Indeed. Further, why use simple_request in the first place?
        # It is not part of the UserAgent UI. I believe this can result
        # in warnings like:
        #   Argument "0, 0, 0, 0" isn't numeric in numeric gt (>) at
        #   /usr/local/lib/perl5/site_perl/5.30.2/HTTP/Response.pm line 261.
        # when checking, e.g.,
        #   https://metacpan.org/pod/distribution/Test-Harness/bin/prove
        # For testing, here is a three-line html file to check that url:
        #   <html><head><title>X</title></head><body>
        #   <p><a href="https://metacpan.org/pod/release/MSCHWERN/Test-Simple-0.98_05/lib/Test/More.pm">prove</a></p>
        #   </body></html>
        # I have been unable to reproduce the warning with a test program
        # checking that url using $ua->request(), or other UserAgent
        # functions, even after carefully reproducing all the headers
        # that checklink sends in the request. --karl@freefriends.org.

        $self->W3C::UserAgent::SUPER::simple_request(@_);
    };

    if (!defined($self->{FirstResponse})) {
        $self->{FirstResponse} = $response->code();
        $self->{FirstMessage} = $response->message() || '(no message)';
    }

    return $response;
}

sub redirect_ok
{
    my ($self, $request, $response) = @_;

    if (my $callback = $self->redirect_progress_callback()) {

        # @@@ TODO: when an LWP internal robots.txt request gets redirected,
        # this will a bit confusingly fire for it too.  Would need a robust
        # way to determine whether the request is such a LWP "internal
        # robots.txt" one.
        &$callback($request->method(), $request->uri());
    }

    return 0 unless $self->SUPER::redirect_ok($request, $response);

    if (my $res = $self->ip_disallowed($request->uri())) {
        $response->previous($response->clone());
        $response->request($request);
        $response->code($res->code());
        $response->message($res->message());
        return 0;
    }

    return 1;
}

#
# Checks whether we're allowed to retrieve the document based on its IP
# address.  Takes an URI object and returns a HTTP::Response containing the
# appropriate status and error message if the IP was disallowed, 0
# otherwise.  URIs without hostname or IP address are always allowed,
# including schemes where those make no sense (eg. data:, often javascript:).
#
sub ip_disallowed
{
    my ($self, $uri) = @_;
    return 0 if $self->allow_private_ips();    # Short-circuit

    my $hostname = undef;
    eval { $hostname = $uri->host() };    # Not all URIs implement host()...
    return 0 unless $hostname;

    my $addr = my $iptype = my $resp = undef;
    if (my $host = Net::hostent::gethostbyname($hostname)) {
        $addr = Socket::inet_ntoa($host->addr()) if $host->addr();
        if ($addr && (my $ip = Net::IP->new($addr))) {
            $iptype = $ip->iptype();
        }
    }
    if ($iptype && $iptype ne 'PUBLIC') {
        $resp = HTTP::Response->new(403,
            'Checking non-public IP address disallowed by link checker configuration'
        );
        $resp->header('Client-Warning', 'Internal response');
    }
    return $resp;
}

# -----------------------------------------------------------------------------

package W3C::LinkChecker;

use vars qw($AGENT $PACKAGE $PROGRAM $VERSION $REVISION
    $DocType $Head $Accept $ContentTypes %Cfg $CssUrl);

use CSS::DOM 0.09 qw();    # >= 0.09 for many bugfixes
use CSS::DOM::Constants qw(:rule);
use CSS::DOM::Style qw();
use CSS::DOM::Util qw();
use Encode qw();
use HTML::Entities qw();
use HTML::Parser 3.40 qw();    # >= 3.40 for utf8_mode()
use HTTP::Headers::Util qw();
use HTTP::Message 5.827 qw();    # >= 5.827 for content_charset()
use HTTP::Request 5.814 qw();    # >= 5.814 for accept_decodable()
use HTTP::Response 1.50 qw();    # >= 1.50 for decoded_content()
use Time::HiRes qw();
use URI 1.53 qw();               # >= 1.53 for secure()
use URI::Escape qw();
use URI::Heuristic qw();

# @@@ Needs also W3C::UserAgent but can't use() it here.

use constant RC_ROBOTS_TXT          => -1;
use constant RC_DNS_ERROR           => -2;
use constant RC_IP_DISALLOWED       => -3;
use constant RC_PROTOCOL_DISALLOWED => -4;

use constant LINE_UNKNOWN => -1;

use constant MP2 =>
    (exists($ENV{MOD_PERL_API_VERSION}) && $ENV{MOD_PERL_API_VERSION} >= 2);

# Tag=>attribute mapping of things we treat as links.
# Note: meta/@http-equiv gets special treatment, see start() for details.
use constant LINK_ATTRS => {
    a => ['href'],

    # base/@href intentionally not checked
    # http://www.w3.org/mid/200802091439.27764.ville.skytta%40iki.fi
    area       => ['href'],
    audio      => ['src'],
    blockquote => ['cite'],
    body       => ['background'],
    command    => ['icon'],

    # button/@formaction not checked (side effects)
    del => ['cite'],

    # @pluginspage, @pluginurl, @href: pre-HTML5 proprietary
    embed => ['href', 'pluginspage', 'pluginurl', 'src'],

    # form/@action not checked (side effects)
    frame  => ['longdesc', 'src'],
    html   => ['manifest'],
    iframe => ['longdesc', 'src'],
    img    => ['longdesc', 'src'],

    # input/@action, input/@formaction not checked (side effects)
    input  => ['src'],
    ins    => ['cite'],
    link   => ['href'],
    object => ['data'],
    q      => ['cite'],
    script => ['src'],
    source => ['src'],
    track  => ['src'],
    video  => ['src', 'poster'],
};

# Tag=>[separator, attributes] mapping of things we treat as lists of links.
use constant LINK_LIST_ATTRS => {
    a      => [qr/\s+/,    ['ping']],
    applet => [qr/[\s,]+/, ['archive']],
    area   => [qr/\s+/,    ['ping']],
    head   => [qr/\s+/,    ['profile']],
    object => [qr/\s+/,    ['archive']],
};

# TBD/TODO:
# - applet/@code?
# - bgsound/@src?
# - object/@classid?
# - isindex/@action?
# - layer/@background,@src?
# - ilayer/@background?
# - table,tr,td,th/@background?
# - xmp/@href?

@W3C::LinkChecker::ISA = qw(HTML::Parser);

BEGIN {

    # Version info
    $PACKAGE  = 'W3C Link Checker';
    $PROGRAM  = 'W3C-checklink';
    $VERSION  = '4.81';
    $REVISION = sprintf('version %s (c) 1999-2011 W3C', $VERSION);
    $AGENT    = sprintf(
        '%s/%s %s',
        $PROGRAM, $VERSION,
        (   W3C::UserAgent::USE_ROBOT_UA ? LWP::RobotUA->_agent() :
                LWP::UserAgent->_agent()
        )
    );

    # Pull in mod_perl modules if applicable.
    eval {
        local $SIG{__DIE__} = undef;
        require Apache2::RequestUtil;
    } if MP2();

    my @content_types = qw(
        text/html
        application/xhtml+xml;q=0.9
        application/vnd.wap.xhtml+xml;q=0.6
    );
    $Accept = join(', ', @content_types, '*/*;q=0.5');
    push(@content_types, 'text/css', 'text/html-sandboxed');
    my $re = join('|', map { s/;.*//; quotemeta } @content_types);
    $ContentTypes = qr{\b(?:$re)\b}io;

    # Regexp for matching URL values in CSS.
    $CssUrl = qr/(?:\s|^)url\(\s*(['"]?)(.*?)\1\s*\)(?=\s|$)/;

    #
    # Read configuration.  If the W3C_CHECKLINK_CFG environment variable has
    # been set or the default contains a non-empty file, read it.  Otherwise,
    # skip silently.
    #
    my $defaultconfig = '/etc/w3c/checklink.conf';
    if ($ENV{W3C_CHECKLINK_CFG} || -s $defaultconfig) {

        require Config::General;
        Config::General->require_version(2.06);    # Need 2.06 for -SplitPolicy

        my $conffile = $ENV{W3C_CHECKLINK_CFG} || $defaultconfig;
        eval {
            my %config_opts = (
                -ConfigFile        => $conffile,
                -SplitPolicy       => 'equalsign',
                -AllowMultiOptions => 'no',
            );
            %Cfg = Config::General->new(%config_opts)->getall();
        };
        if ($@) {
            die <<"EOF";
Failed to read configuration from '$conffile':
$@
EOF
        }
    }
    $Cfg{Markup_Validator_URI} ||= 'http://validator.w3.org/check?uri=%s';
    $Cfg{CSS_Validator_URI} ||=
        'http://jigsaw.w3.org/css-validator/validator?uri=%s';
    $Cfg{Doc_URI} ||= 'http://validator.w3.org/docs/checklink.html';

    # Untaint config params that are used as the format argument to (s)printf(),
    # Perl 5.10 does not want to see that in taint mode.
    ($Cfg{Markup_Validator_URI}) = ($Cfg{Markup_Validator_URI} =~ /^(.*)$/);
    ($Cfg{CSS_Validator_URI})    = ($Cfg{CSS_Validator_URI}    =~ /^(.*)$/);

    $DocType =
        '<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">';
    my $css_url = URI->new_abs('linkchecker.css', $Cfg{Doc_URI});
    my $js_url  = URI->new_abs('linkchecker.js',  $Cfg{Doc_URI});
    $Head =
        sprintf(<<'EOF', HTML::Entities::encode($AGENT), $css_url, $js_url);
<meta http-equiv="Content-Script-Type" content="text/javascript" />
<meta name="generator" content="%s" />
<link rel="stylesheet" type="text/css" href="%s" />
<script type="text/javascript" src="%s"></script>
EOF

    # Trusted environment variables that need laundering in taint mode.
    for (qw(NNTPSERVER NEWSHOST)) {
        ($ENV{$_}) = ($ENV{$_} =~ /^(.*)$/) if $ENV{$_};
    }

    # Use passive FTP by default, see Net::FTP(3).
    $ENV{FTP_PASSIVE} = 1 unless exists($ENV{FTP_PASSIVE});
}

# Autoflush
$| = 1;

# Different options specified by the user
my $cmdline = !($ENV{GATEWAY_INTERFACE} && $ENV{GATEWAY_INTERFACE} =~ /^CGI/);
my %Opts = (
    Command_Line    => $cmdline,
    Quiet           => 0,
    Summary_Only    => 0,
    Verbose         => 0,
    Progress        => 0,
    HTML            => 0,
    Timeout         => 30,
    Redirects       => 1,
    Dir_Redirects   => 1,
    Accept_Language => $cmdline ? undef : $ENV{HTTP_ACCEPT_LANGUAGE},
    Cookies         => undef,
    No_Referer      => 0,
    Hide_Same_Realm => 0,
    Depth                    => 0,             # < 0 means unlimited recursion.
    Sleep_Time               => 1,
    Connection_Cache_Size    => 2,
    Max_Documents            => 150,           # For the online version.
    User                     => undef,
    Password                 => undef,
    Base_Locations           => [],
    Exclude                  => undef,
    Exclude_Docs             => undef,
    Exclude_Url_File         => undef,
    Suppress_Redirect        => [],
    Suppress_Redirect_Prefix => [],
    Suppress_Redirect_Regexp => [],
    Suppress_Temp_Redirects  => 1,
    Suppress_Broken          => [],
    Suppress_Fragment        => [],
    Masquerade               => 0,
    Masquerade_From          => '',
    Masquerade_To            => '',
    Trusted                  => $Cfg{Trusted},
    Allow_Private_IPs => defined($Cfg{Allow_Private_IPs}) ?
        $Cfg{Allow_Private_IPs} :
        $cmdline,
);
undef $cmdline;

# Global variables
# What URI's did we process? (used for recursive mode)
my %processed;

# Result of the HTTP query
my %results;

# List of redirects
my %redirects;

# Count of the number of documents checked
my $doc_count = 0;

# Time stamp
my $timestamp = &get_timestamp();

# Per-document header; undefined if already printed.  See print_doc_header().
my $doc_header;

&parse_arguments() if $Opts{Command_Line};

my $ua = W3C::UserAgent->new($AGENT);    # @@@ TODO: admin address

$ua->conn_cache({total_capacity => $Opts{Connection_Cache_Size}});
if ($ua->can('delay')) {
    $ua->delay($Opts{Sleep_Time} / 60);
}
$ua->timeout($Opts{Timeout});

# Set up cookie stash if requested
if (defined($Opts{Cookies})) {
    require HTTP::Cookies;
    my $cookie_file = $Opts{Cookies};
    if ($cookie_file eq 'tmp') {
        $cookie_file = undef;
    }
    elsif ($cookie_file =~ /^(.*)$/) {
        $cookie_file = $1;    # untaint
    }
    $ua->cookie_jar(HTTP::Cookies->new(file => $cookie_file, autosave => 1));
}
eval { $ua->allow_private_ips($Opts{Allow_Private_IPs}); };
if ($@) {
    die <<"EOF";
Allow_Private_IPs is false; this feature requires the Net::IP, Socket, and
Net::hostent modules:
$@
EOF
}

# Add configured forbidden protocols
if ($Cfg{Forbidden_Protocols}) {
    my $forbidden = $ua->protocols_forbidden();
    push(@$forbidden, split(/[,\s]+/, lc($Cfg{Forbidden_Protocols})));
    $ua->protocols_forbidden($forbidden);
}

if ($Opts{Command_Line}) {

    require Text::Wrap;
    Text::Wrap->import('wrap');

    require URI::file;

    &usage(1) unless scalar(@ARGV);

    $Opts{_Self_URI} = 'http://validator.w3.org/checklink';   # For HTML output

    &ask_password() if ($Opts{User} && !$Opts{Password});

    if (!$Opts{Summary_Only}) {
        printf("%s %s\n", $PACKAGE, $REVISION) unless $Opts{HTML};
    }
    else {
        $Opts{Verbose}  = 0;
        $Opts{Progress} = 0;
    }

    # Populate data for print_form()
    my %params = (
        summary            => $Opts{Summary_Only},
        hide_redirects     => !$Opts{Redirects},
        hide_type          => $Opts{Dir_Redirects} ? 'dir' : 'all',
        no_accept_language => !(
            defined($Opts{Accept_Language}) && $Opts{Accept_Language} eq 'auto'
        ),
        no_referer => $Opts{No_Referer},
        recursive  => ($Opts{Depth} != 0),
        depth      => $Opts{Depth},
    );

    my $check_num = 1;
    my @bases     = @{$Opts{Base_Locations}};
    for my $uri (@ARGV) {

        # Reset base locations so that previous URI's given on the command line
        # won't affect the recursion scope for this URI (see check_uri())
        @{$Opts{Base_Locations}} = @bases;

        # Transform the parameter into a URI
        $uri = &urize($uri);
        $params{uri} = $uri;
        &check_uri(\%params, $uri, $check_num, $Opts{Depth}, undef, undef, 1);
        $check_num++;
    }
    undef $check_num;

    if ($Opts{HTML}) {
        &html_footer();
    }
    elsif ($doc_count > 0 && !$Opts{Summary_Only}) {
        printf("\n%s\n", &global_stats());
    }

}
else {

    require CGI;
    require CGI::Carp;
    CGI::Carp->import(qw(fatalsToBrowser));
    require CGI::Cookie;

    # file: URIs are not allowed in CGI mode
    my $forbidden = $ua->protocols_forbidden();
    push(@$forbidden, 'file');
    $ua->protocols_forbidden($forbidden);

    my $query = CGI->new();

    for my $param ($query->param()) {
        my @values = map { Encode::decode_utf8($_) } $query->param($param);
        $query->param($param, @values);
    }

    # Set a few parameters in CGI mode
    $Opts{Verbose}   = 0;
    $Opts{Progress}  = 0;
    $Opts{HTML}      = 1;
    $Opts{_Self_URI} = $query->url(-relative => 1);

    # Backwards compatibility
    my $uri = undef;
    if ($uri = $query->param('url')) {
        $query->param('uri', $uri) unless $query->param('uri');
        $query->delete('url');
    }
    $uri = $query->param('uri');

    if (!$uri) {
        &html_header('', undef);    # Set cookie only from results page.
        my %cookies = CGI::Cookie->fetch();
        &print_form(scalar($query->Vars()), $cookies{$PROGRAM}, 1);
        &html_footer();
        exit;
    }

    # Backwards compatibility
    if ($query->param('hide_dir_redirects')) {
        $query->param('hide_redirects', 'on');
        $query->param('hide_type',      'dir');
        $query->delete('hide_dir_redirects');
    }

    $Opts{Summary_Only} = 1 if $query->param('summary');

    if ($query->param('hide_redirects')) {
        $Opts{Dir_Redirects} = 0;
        if (my $type = $query->param('hide_type')) {
            $Opts{Redirects} = 0 if ($type ne 'dir');
        }
        else {
            $Opts{Redirects} = 0;
        }
    }

    $Opts{Accept_Language} = undef if $query->param('no_accept_language');
    $Opts{No_Referer} = $query->param('no_referer');

    $Opts{Depth} = -1 if ($query->param('recursive') && $Opts{Depth} == 0);
    if (my $depth = $query->param('depth')) {

        # @@@ Ignore invalid depth silently for now.
        $Opts{Depth} = $1 if ($depth =~ /(-?\d+)/);
    }

    # Save, clear or leave cookie as is.
    my $cookie = undef;
    if (my $action = $query->param('cookie')) {
        if ($action eq 'clear') {

            # Clear the cookie.
            $cookie = CGI::Cookie->new(-name => $PROGRAM);
            $cookie->value({clear => 1});
            $cookie->expires('-1M');
        }
        elsif ($action eq 'set') {

            # Set the options.
            $cookie = CGI::Cookie->new(-name => $PROGRAM);
            my %options = $query->Vars();
            delete($options{$_})
                for qw(url uri check cookie);    # Non-persistent.
            $cookie->value(\%options);
        }
    }
    if (!$cookie) {
        my %cookies = CGI::Cookie->fetch();
        $cookie = $cookies{$PROGRAM};
    }

    # Always refresh cookie expiration time.
    $cookie->expires('+1M') if ($cookie && !$cookie->expires());

    # All Apache configurations don't set HTTP_AUTHORIZATION for CGI scripts.
    # If we're under mod_perl, there is a way around it...
    eval {
        local $SIG{__DIE__} = undef;
        my $auth =
            Apache2::RequestUtil->request()->headers_in()->{Authorization};
        $ENV{HTTP_AUTHORIZATION} = $auth if $auth;
    } if (MP2() && !$ENV{HTTP_AUTHORIZATION});

    $uri =~ s/^\s+//g;
    if ($uri =~ /:/) {
        $uri = URI->new($uri);
    }
    else {
        if ($uri =~ m|^//|) {
            $uri = URI->new("http:$uri");
        }
        else {
            local $ENV{URL_GUESS_PATTERN} = '';
            my $guess = URI::Heuristic::uf_uri($uri);
            if ($guess->scheme() && $ua->is_protocol_supported($guess)) {
                $uri = $guess;
            }
            else {
                $uri = URI->new("http://$uri");
            }
        }
    }
    $uri = $uri->canonical();
    $query->param("uri", $uri);

    &check_uri(scalar($query->Vars()), $uri, 1, $Opts{Depth}, $cookie);
    undef $query;    # Not needed any more.
    &html_footer();
}

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

################################
# Command line and usage stuff #
################################

sub parse_arguments ()
{
    require Encode::Locale;
    Encode::Locale::decode_argv();

    require Getopt::Long;
    Getopt::Long->require_version(2.17);
    Getopt::Long->import('GetOptions');
    Getopt::Long::Configure('bundling', 'no_ignore_case');
    my $masq = '';
    my @locs = ();

    GetOptions(
        'help|h|?' => sub { usage(0) },
        'q|quiet'  => sub {
            $Opts{Quiet}        = 1;
            $Opts{Summary_Only} = 1;
        },
        's|summary' => \$Opts{Summary_Only},
        'b|broken'  => sub {
            $Opts{Redirects}     = 0;
            $Opts{Dir_Redirects} = 0;
        },
        'e|dir-redirects' => sub { $Opts{Dir_Redirects} = 0; },
        'v|verbose'       => \$Opts{Verbose},
        'i|indicator'     => \$Opts{Progress},
        'H|html'          => \$Opts{HTML},
        'r|recursive'     => sub {
            $Opts{Depth} = -1
                if $Opts{Depth} == 0;
        },
        'l|location=s'                => \@locs,
        'X|exclude=s@'                => \@{$Opts{Exclude}},
        'exclude-docs=s@'             => \@{$Opts{Exclude_Docs}},
        'exclude-url-file=s'          => \$Opts{Exclude_Url_File},
        'suppress-redirect=s@'        => \@{$Opts{Suppress_Redirect}},
        'suppress-redirect-prefix=s@' => \@{$Opts{Suppress_Redirect_Prefix}},
        'suppress-temp-redirects'     => \$Opts{Suppress_Temp_Redirects},
        'suppress-broken=s@'          => \@{$Opts{Suppress_Broken}},
        'suppress-fragment=s@'        => \@{$Opts{Suppress_Fragment}},
        'u|user=s'                    => \$Opts{User},
        'p|password=s'                => \$Opts{Password},
        't|timeout=i'                 => \$Opts{Timeout},
        'C|connection-cache=i'        => \$Opts{Connection_Cache_Size},
        'S|sleep=i'                   => \$Opts{Sleep_Time},
        'L|languages=s'               => \$Opts{Accept_Language},
        'c|cookies=s'                 => \$Opts{Cookies},
        'R|no-referer'                => \$Opts{No_Referer},
        'D|depth=i'                   => sub {
            $Opts{Depth} = $_[1]
                unless $_[1] == 0;
        },
        'd|domain=s'      => \$Opts{Trusted},
        'masquerade=s'    => \$masq,
        'hide-same-realm' => \$Opts{Hide_Same_Realm},
        'V|version'       => \&version,
        ) ||
        usage(1);

    if ($masq) {
        $Opts{Masquerade} = 1;
        my @masq = split(/\s+/, $masq);
        if (scalar(@masq) != 2 ||
            !defined($masq[0]) ||
            $masq[0] !~ /\S/ ||
            !defined($masq[1]) ||
            $masq[1] !~ /\S/)
        {
            usage(1,
                "Error: --masquerade takes two whitespace separated URIs.");
        }
        else {
            require URI::file;
            $Opts{Masquerade_From} = $masq[0];
            my $u = URI->new($masq[1]);
            $Opts{Masquerade_To} =
                $u->scheme() ? $u : URI::file->new_abs($masq[1]);
        }
    }

    if ($Opts{Accept_Language} && $Opts{Accept_Language} eq 'auto') {
        $Opts{Accept_Language} = &guess_language();
    }

    if (($Opts{Sleep_Time} || 0) < 1) {
        warn(
            "*** Warning: minimum allowed sleep time is 1 second, resetting.\n"
        );
        $Opts{Sleep_Time} = 1;
    }

    push(@{$Opts{Base_Locations}}, map { URI->new($_)->canonical() } @locs);

    $Opts{Depth} = -1 if ($Opts{Depth} == 0 && @locs);

    for my $i (0 .. $#{$Opts{Exclude_Docs}}) {
        eval { $Opts{Exclude_Docs}->[$i] = qr/$Opts{Exclude_Docs}->[$i]/; };
        &usage(1, "Error in exclude-docs regexp: $@") if $@;
    }
    if (defined($Opts{Trusted})) {
        eval { $Opts{Trusted} = qr/$Opts{Trusted}/io; };
        &usage(1, "Error in trusted domains regexp: $@") if $@;
    }

    # Sanity-check error-suppression arguments
    for my $i (0 .. $#{$Opts{Suppress_Redirect}}) {
        ${$Opts{Suppress_Redirect}}[$i] =~ s/ /->/;
        my $sr_arg = ${$Opts{Suppress_Redirect}}[$i];
        if ($sr_arg !~ /.->./) {
            &usage(1,
                "Bad suppress-redirect argument, should contain \"->\": $sr_arg"
            );
        }
    }
    for my $i (0 .. $#{$Opts{Suppress_Redirect_Prefix}}) {
        my $srp_arg = ${$Opts{Suppress_Redirect_Prefix}}[$i];
        $srp_arg =~ s/ /->/;
        if ($srp_arg !~ /^(.*)->(.*)$/) {
            &usage(1,
                "Bad suppress-redirect-prefix argument, should contain \"->\": $srp_arg"
            );
        }

        # Turn prefixes into a regexp.
        ${$Opts{Suppress_Redirect_Prefix}}[$i] = qr/^\Q$1\E(.*)->\Q$2\E\1$/ism;
    }
    for my $i (0 .. $#{$Opts{Suppress_Broken}}) {
        ${$Opts{Suppress_Broken}}[$i] =~ s/ /:/;
        my $sb_arg = ${$Opts{Suppress_Broken}}[$i];
        if ($sb_arg !~ /^(-1|[0-9]+):./) {
            &usage(1,
                "Bad suppress-broken argument, should be prefixed by a numeric response code: $sb_arg"
            );
        }
    }
    for my $sf_arg (@{$Opts{Suppress_Fragment}}) {
        if ($sf_arg !~ /.#./) {
            &usage(1,
                "Bad suppress-fragment argument, should contain \"#\": $sf_arg"
            );
        }
    }

    if ($#{$Opts{Exclude}} > 0) {
      # convert $Opts{Exclude} array into regexp by parenthesizing
      # each and inserting alternations between.
      my $exclude_rx = join("|", map { "($_)" } @{$Opts{Exclude}});
      #
      # For the sake of the rest of the program, pretend the option
      # was that string all along.
      $Opts{Exclude} = $exclude_rx;
    }

    if ($Opts{Exclude_Url_File}) {
        # The idea is that if the specified file exists, we read it and
        # treat it as a list of excludes. If the file doesn't exist, we
        # write it with all the urls that were successful. That way, we
        # can avoid re-checking them on every run, and it can be removed
        # externally (from cron) to get re-updated.
        #
        # We distinguish the cases here, and either add to
        # $Opts{Exclude} if reading, or setting Exclude_File_Write in
        # %Opts if writing (even though it is not really an option,
        # but it's the most convenient place).
        if (-s $Opts{Exclude_Url_File}) {
            open (my $xf, "$Opts{Exclude_Url_File}")
            || &usage(1, "Could not open $Opts{Exclude_Url_File}"
                         . " for reading: $!");
            my @xf = ();
            while (<$xf>) {
                chomp;
                # the file is urls, not regexps, so quotemeta.
                push (@xf, "(" . quotemeta($_) . ")");
            }
            my $xf_rx = join ("|", @xf);
            if ($Opts{Exclude}) {
                $Opts{Exclude} .= "|$xf_rx";
            } else {
                $Opts{Exclude} = $xf_rx;
            }
        } else {
            open ($Opts{Exclude_File_Write}, ">$Opts{Exclude_Url_File}")
            || &usage(1,
                     "Could not open $Opts{Exclude_Url_File} for writing: $!");
            # we write on a successful retrieve, and don't bother closing.
        }
    }

    # Precompile/error-check final list of regular expressions
    if (defined($Opts{Exclude})) {
        eval { $Opts{Exclude} = qr/$Opts{Exclude}/o; };
        &usage(1, "Error in exclude regexp $Opts{Exclude}: $@") if $@;
    }

    return;
}

sub version ()
{
    print "$PACKAGE $REVISION\n";
    exit 0;
}

sub usage ()
{
    my ($exitval, $msg) = @_;
    $exitval = 0 unless defined($exitval);
    $msg ||= '';
    $msg =~ s/[\r\n]*$/\n\n/ if $msg;

    die($msg) unless $Opts{Command_Line};

    my $trust = defined($Cfg{Trusted}) ? $Cfg{Trusted} : 'same host only';

    select(STDERR) if $exitval;
    print "$msg$PACKAGE $REVISION

Usage: checklink <options> <uris>
Options:
 -s, --summary              Result summary only.
 -b, --broken               Show only the broken links, not the redirects.
 -e, --directory            Hide directory redirects, for example
                            http://www.w3.org/TR -> http://www.w3.org/TR/
 -r, --recursive            Check the documents linked from the first one.
 -D, --depth N              Check the documents linked from the first one to
                            depth N (implies --recursive).
 -l, --location URI         Scope of the documents checked in recursive mode
                            (implies --recursive).  Can be specified multiple
                            times.  If not specified, the default eg. for
                            http://www.w3.org/TR/html4/Overview.html
                            would be http://www.w3.org/TR/html4/
 -X, --exclude REGEXP       Do not check links whose full, canonical URIs
                            match REGEXP; also limits recursion the same way
                            as --exclude-docs with the same regexp would.
                            This option may be specified multiple times.
 --exclude-docs REGEXP      In recursive mode, do not check links in documents
                            whose full, canonical URIs match REGEXP.  This
                            option may be specified multiple times.
 --exclude-url-file FILE    If FILE exists, treat each line as a string
                            specifying another exclude; quotemeta is called
                            to make them regexps. If FILE does not exist,
                            open it for writing and write each checked url
                            which gets a 200 response to it.
 --suppress-redirect URI->URI  Do not report a redirect from the first to the
                            second URI.  This option may be specified multiple
                            times.
 --suppress-redirect-prefix URI->URI  Do not report a redirect from a child of
                            the first URI to the same child of the second URI.
                            This option may be specified multiple times.
 --suppress-temp-redirects  Suppress warnings about temporary redirects.
 --suppress-broken CODE:URI  Do not report a broken link with the given CODE.
                            CODE is HTTP response, or -1 for robots exclusion.
                            This option may be specified multiple times.
 --suppress-fragment URI    Do not report the given broken fragment URI.
                            A fragment URI contains \"#\".  This option may be
                            specified multiple times.
 -L, --languages LANGS      Accept-Language header to send.  The special value
                            'auto' causes autodetection from the environment.
 -c, --cookies FILE         Use cookies, load/save them in FILE.  The special
                            value 'tmp' causes non-persistent use of cookies.
 -R, --no-referer           Do not send the Referer HTTP header.
 -q, --quiet                No output if no errors are found (implies -s).
 -v, --verbose              Verbose mode.
 -i, --indicator            Show percentage of lines processed while parsing.
 -u, --user USERNAME        Specify a username for authentication.
 -p, --password PASSWORD    Specify a password.
 --hide-same-realm          Hide 401's that are in the same realm as the
                            document checked.
 -S, --sleep SECS           Sleep SECS seconds between requests to each server
                            (default and minimum: 1 second).
 -t, --timeout SECS         Timeout for requests in seconds (default: 30).
 -d, --domain DOMAIN        Regular expression describing the domain to which
                            authentication information will be sent
                            (default: $trust).
 --masquerade \"BASE1 BASE2\" Masquerade base URI BASE1 as BASE2.  See the
                            manual page for more information.
 -H, --html                 HTML output.
 -?, -h, --help             Show this message and exit.
 -V, --version              Output version information and exit.

See \"perldoc LWP\" for information about proxy server support,
\"perldoc Net::FTP\" for information about various environment variables
affecting FTP connections and \"perldoc Net::NNTP\" for setting a default
NNTP server for news: URIs.

The W3C_CHECKLINK_CFG environment variable can be used to set the
configuration file to use.  See details in the full manual page, it can
be displayed with: perldoc checklink

More documentation at: $Cfg{Doc_URI}
Please send bug reports and comments to the www-validator mailing list:
  www-validator\@w3.org (with 'checklink' in the subject)
  Archives are at: http://lists.w3.org/Archives/Public/www-validator/
";
    exit $exitval;
}

sub ask_password ()
{
    eval {
        local $SIG{__DIE__} = undef;
        require Term::ReadKey;
        Term::ReadKey->require_version(2.00);
        Term::ReadKey->import(qw(ReadMode));
    };
    if ($@) {
        warn('Warning: Term::ReadKey 2.00 or newer not available, ' .
                "password input disabled.\n");
        return;
    }
    printf(STDERR 'Enter the password for user %s: ', $Opts{User});
    ReadMode('noecho', *STDIN);
    chomp($Opts{Password} = <STDIN>);
    ReadMode('restore', *STDIN);
    print(STDERR "ok.\n");
    return;
}

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

###########################################################################
# Guess an Accept-Language header based on the $LANG environment variable #
###########################################################################

sub guess_language ()
{
    my $lang = $ENV{LANG} or return;

    $lang =~ s/[\.@].*$//;    # en_US.UTF-8, fi_FI@euro...

    return 'en' if ($lang eq 'C' || $lang eq 'POSIX');

    my $res = undef;
    eval {
        require Locale::Language;
        if (my $tmp = Locale::Language::language2code($lang)) {
            $lang = $tmp;
        }
        if (my ($l, $c) = (lc($lang) =~ /^([a-z]+)(?:[-_]([a-z]+))?/)) {
            if (Locale::Language::code2language($l)) {
                $res = $l;
                if ($c) {
                    require Locale::Country;
                    $res .= "-$c" if Locale::Country::code2country($c);
                }
            }
        }
    };
    return $res;
}

############################
# Transform foo into a URI #
############################

sub urize ($)
{
    my $arg  = shift;
    my $uarg = URI::Escape::uri_unescape($arg);
    my $uri;
    if (-d $uarg) {

        # look for an "index" file in dir, return it if found
        require File::Spec;
        for my $index (map { File::Spec->catfile($uarg, $_) }
            qw(index.html index.xhtml index.htm index.xhtm))
        {
            if (-e $index) {
                $uri = URI::file->new_abs($index);
                last;
            }
        }

        # return dir itself if an index file was not found
        $uri ||= URI::file->new_abs($uarg);
    }
    elsif ($uarg =~ /^[.\/\\]/ || -e $uarg) {
        $uri = URI::file->new_abs($uarg);
    }
    else {
        my $newuri = URI->new($arg);
        if ($newuri->scheme()) {
            $uri = $newuri;
        }
        else {
            local $ENV{URL_GUESS_PATTERN} = '';
            $uri = URI::Heuristic::uf_uri($arg);
            $uri = URI::file->new_abs($uri) unless $uri->scheme();
        }
    }
    return $uri->canonical();
}

########################################
# Check for broken links in a resource #
########################################

sub check_uri (\%\$$$$;\$$)
{
    my ($params, $uri, $check_num, $depth, $cookie, $referer, $is_start) = @_;
    $is_start ||= ($check_num == 1);

    my $start = $Opts{Summary_Only} ? 0 : &get_timestamp();

    # Get and parse the document
    my $response = &get_document(
        'GET',   $uri,    $doc_count, \%redirects, $referer,
        $cookie, $params, $check_num, $is_start
    );

    # Can we check the resource? If not, we exit here...
    return if defined($response->{Stop});

    if ($Opts{HTML}) {
        &html_header($uri, $cookie) if ($check_num == 1);
        &print_form($params, $cookie, $check_num) if $is_start;
    }

    if ($is_start) { # Starting point of a new check, eg. from the command line
          # Use the first URI as the recursion base unless specified otherwise.
        push(@{$Opts{Base_Locations}}, $response->{absolute_uri}->canonical())
            unless @{$Opts{Base_Locations}};
    }
    else {

        # Before fetching the document, we don't know if we'll be within the
        # recursion scope or not (think redirects).
        if (!&in_recursion_scope($response->{absolute_uri})) {
            hprintf("Not in recursion scope: %s\n", $response->{absolute_uri})
                if ($Opts{Verbose});
            $response->content("");
            return;
        }
    }

    # Define the document header, and perhaps print it.
    # (It might still be defined if the previous document had no errors;
    # just redefine it in that case.)

    if ($check_num != 1) {
        if ($Opts{HTML}) {
            $doc_header = "\n<hr />\n";
        }
        else {
            $doc_header = "\n" . ('-' x 40) . "\n";
        }
    }

    if ($Opts{HTML}) {
        $doc_header .=
            ("<h2>\nProcessing\t" . &show_url($response->{absolute_uri}) .
                "\n</h2>\n\n");
    }
    else {
        $doc_header .= "\nProcessing\t$response->{absolute_uri}\n\n";
    }

    if (!$Opts{Quiet}) {
        print_doc_header();
    }

    # We are checking a new document
    $doc_count++;

    my $result_anchor = 'results' . $doc_count;

    if ($check_num == 1 && !$Opts{HTML} && !$Opts{Summary_Only}) {
        my $s = $Opts{Sleep_Time} == 1 ? '' : 's';
        my $acclang = $Opts{Accept_Language} || '(not sent)';
        my $send_referer = $Opts{No_Referer} ? 'not sent' : 'sending';
        my $cookies = 'not used';
        if (defined($Opts{Cookies})) {
            $cookies = 'used, ';
            if ($Opts{Cookies} eq 'tmp') {
                $cookies .= 'non-persistent';
            }
            else {
                $cookies .= "file $Opts{Cookies}";
            }
        }
        printf(
            <<'EOF', $Accept, $acclang, $send_referer, $cookies, $Opts{Sleep_Time}, $s);

Settings used:
- Accept: %s
- Accept-Language: %s
- Referer: %s
- Cookies: %s
- Sleeping %d second%s between requests to each server
EOF
        printf("- Excluding links matching %s\n", $Opts{Exclude})
            if defined($Opts{Exclude});
        printf("- Excluding links in documents whose URIs match %s\n",
            join(', ', @{$Opts{Exclude_Docs}}))
            if @{$Opts{Exclude_Docs}};
    }

    if ($Opts{HTML}) {
        if (!$Opts{Summary_Only}) {
            my $accept       = &encode($Accept);
            my $acclang      = &encode($Opts{Accept_Language} || '(not sent)');
            my $send_referer = $Opts{No_Referer} ? 'not sent' : 'sending';
            my $s            = $Opts{Sleep_Time} == 1 ? '' : 's';
            printf(
                <<'EOF', $accept, $acclang, $send_referer, $Opts{Sleep_Time}, $s);
<div class="settings">
Settings used:
 <ul>
  <li><tt><a href="http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.1">Accept</a></tt>: %s</li>
  <li><tt><a href="http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.4">Accept-Language</a></tt>: %s</li>
  <li><tt><a href="http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.36">Referer</a></tt>: %s</li>
  <li>Sleeping %d second%s between requests to each server</li>
 </ul>
</div>
EOF
            printf("<p>Go to <a href=\"#%s\">the results</a>.</p>\n",
                $result_anchor);
            my $esc_uri = URI::Escape::uri_escape($response->{absolute_uri},
                "^A-Za-z0-9.");
            print "<p>For reliable link checking results, check ";

            if (!$response->{IsCss}) {
                printf("<a href=\"%s\">HTML validity</a> and ",
                    &encode(sprintf($Cfg{Markup_Validator_URI}, $esc_uri)));
            }
            printf(
                "<a href=\"%s\">CSS validity</a> first.</p>
<p>Back to the <a accesskey=\"1\" href=\"%s\">link checker</a>.</p>\n",
                &encode(sprintf($Cfg{CSS_Validator_URI}, $esc_uri)),
                &encode($Opts{_Self_URI})
            );

            printf(<<'EOF', $result_anchor);
<div class="progress" id="progress%s">
<h3>Status: <span></span></h3>
<div class="progressbar"><div></div></div>
<pre>
EOF
        }
    }

    if ($Opts{Summary_Only} && !$Opts{Quiet}) {
        print '<p>' if $Opts{HTML};
        print 'This may take some time';
        print "... (<a href=\"$Cfg{Doc_URI}#wait\">why?</a>)</p>"
            if $Opts{HTML};
        print " if the document has many links to check.\n" unless $Opts{HTML};
    }

    # Record that we have processed this resource
    $processed{$response->{absolute_uri}} = 1;

    # Parse the document
    my $p =
        &parse_document($uri, $response->base(), $response, 1, ($depth != 0));
    my $base = URI->new($p->{base});

    # Check anchors
    ###############

    print "Checking anchors...\n" unless $Opts{Summary_Only};

    my %errors;
    while (my ($anchor, $lines) = each(%{$p->{Anchors}})) {
        if (!length($anchor)) {

            # Empty IDREF's are not allowed
            $errors{$anchor} = 1;
        }
        else {
            my $times = 0;
            $times += $_ for values(%$lines);

            # They should appear only once
            $errors{$anchor} = 1 if ($times > 1);
        }
    }
    print " done.\n" unless $Opts{Summary_Only};

    # Check links
    #############

    &hprintf("Recording all the links found: %d\n",
        scalar(keys %{$p->{Links}}))
        if ($Opts{Verbose});
    my %links;
    my %hostlinks;

    # Record all the links found
    while (my ($link, $lines) = each(%{$p->{Links}})) {
        my $link_uri = URI->new($link);
        my $abs_link_uri = URI->new_abs($link_uri, $base);

        if ($Opts{Masquerade}) {
            if ($abs_link_uri =~ m|^\Q$Opts{Masquerade_From}\E|) {
                print_doc_header();
                printf("processing %s in base %s\n",
                    $abs_link_uri, $Opts{Masquerade_To});
                my $nlink = $abs_link_uri;
                $nlink =~ s|^\Q$Opts{Masquerade_From}\E|$Opts{Masquerade_To}|;
                $abs_link_uri = URI->new($nlink);
            }
        }

        my $canon_uri = URI->new($abs_link_uri->canonical());
        my $fragment  = $canon_uri->fragment(undef);
        if (!defined($Opts{Exclude}) || $canon_uri !~ $Opts{Exclude}) {
            if (!exists($links{$canon_uri})) {
                my $hostport;
                $hostport = $canon_uri->host_port()
                    if $canon_uri->can('host_port');
                $hostport = '' unless defined $hostport;
                push(@{$hostlinks{$hostport}}, $canon_uri);
            }
            for my $line_num (keys(%$lines)) {
                if (!defined($fragment) || !length($fragment)) {

                    # Document without fragment
                    $links{$canon_uri}{location}{$line_num} = 1;
                }
                else {

                    # Resource with a fragment
                    $links{$canon_uri}{fragments}{$fragment}{$line_num} = 1;
                }
            }
        } else {
          hprintf("excluded via options: %s\n", $canon_uri)
              if ($Opts{Verbose});
        }
    }

    my @order = &distribute_links(\%hostlinks);
    undef %hostlinks;

    # Build the list of broken URI's

    my $nlinks = scalar(@order);

    &hprintf("Checking %d links to build list of broken URI's\n", $nlinks)
        if ($Opts{Verbose});

    my %broken;
    my $link_num = 0;
    for my $u (@order) {
        my $ulinks = $links{$u};

        if ($Opts{Summary_Only}) {

            # Hack: avoid browser/server timeouts in summary only CGI mode, bug 896
            print ' ' if ($Opts{HTML} && !$Opts{Command_Line});
        }
        else {
            &hprintf("\nChecking link %s\n", $u);
            my $progress = ($link_num / $nlinks) * 100;
            printf(
                '<script type="text/javascript">show_progress("%s", "Checking link %s", "%.1f%%");</script>',
                $result_anchor, &encode($u), $progress)
                if (!$Opts{Command_Line} &&
                $Opts{HTML} &&
                !$Opts{Summary_Only});
        }
        $link_num++;

        # Check that a link is valid
        &check_validity($uri, $u, ($depth != 0 && &in_recursion_scope($u)),
            \%links, \%redirects);
        &hprintf("\tReturn code: %s\n", $results{$u}{location}{code})
            if ($Opts{Verbose});
        if ($Opts{Exclude_File_Write} && $results{$u}{location}{code} == 200) {
            my $fh = $Opts{Exclude_File_Write};
            print $fh ("$u\n");
        }
        if ($results{$u}{location}{success}) {

            # Even though it was not broken, we might want to display it
            # on the results page (e.g. because it required authentication)
            $broken{$u}{location} = 1
                if ($results{$u}{location}{display} >= 400);

            # List the broken fragments
            while (my ($fragment, $lines) = each(%{$ulinks->{fragments}})) {

                my $fragment_ok = $results{$u}{fragments}{$fragment};

                if ($Opts{Verbose}) {
                    my @line_nums = sort { $a <=> $b } keys(%$lines);
                    &hprintf(
                        "\t\t%s %s - Line%s: %s\n",
                        $fragment,
                        $fragment_ok             ? 'OK' : 'Not found',
                        (scalar(@line_nums) > 1) ? 's'  : '',
                        join(', ', @line_nums)
                    );
                }

                # A broken fragment?
                $broken{$u}{fragments}{$fragment} += 2 unless $fragment_ok;
            }
        }
        elsif (!($Opts{Quiet} && &informational($results{$u}{location}{code})))
        {

            # Couldn't find the document
            $broken{$u}{location} = 1;

            # All the fragments associated are hence broken
            for my $fragment (keys %{$ulinks->{fragments}}) {
                $broken{$u}{fragments}{$fragment}++;
            }
        }
    }
    &hprintf(
        "\nProcessed in %s seconds.\n",
        &time_diff($start, &get_timestamp())
    ) unless $Opts{Summary_Only};
    printf(
        '<script type="text/javascript">show_progress("%s", "Done. Document processed in %s seconds.", "100%%");</script>',
        $result_anchor, &time_diff($start, &get_timestamp()))
        if ($Opts{HTML} && !$Opts{Summary_Only});

    # Display results
    if ($Opts{HTML} && !$Opts{Summary_Only}) {
        print("</pre>\n</div>\n");
        printf("<h2><a name=\"%s\">Results</a></h2>\n", $result_anchor);
    }
    print "\n" unless $Opts{Quiet};

    &links_summary(\%links, \%results, \%broken, \%redirects);
    &anchors_summary($p->{Anchors}, \%errors);

    # Do we want to process other documents?
    if ($depth != 0) {

        for my $u (map { URI->new($_) } keys %links) {

            next unless $results{$u}{location}{success};    # Broken link?

            next unless &in_recursion_scope($u);

            # Do we understand its content type?
            next unless ($results{$u}{location}{type} =~ $ContentTypes);

            # Have we already processed this URI?
            next if &already_processed($u, $uri);

            # Do the job
            print "\n" unless $Opts{Quiet};
            if ($Opts{HTML}) {
                if (!$Opts{Command_Line}) {
                    if ($doc_count == $Opts{Max_Documents}) {
                        print(
                            "<hr />\n<p><strong>Maximum number of documents ($Opts{Max_Documents}) reached!</strong></p>\n"
                        );
                    }
                    if ($doc_count >= $Opts{Max_Documents}) {
                        $doc_count++;
                        print("<p>Not checking <strong>$u</strong></p>\n");
                        $processed{$u} = 1;
                        next;
                    }
                }
            }

            # This is an inherently recursive algorithm, so Perl's warning is not
            # helpful.  You may wish to comment this out when debugging, though.
            no warnings 'recursion';

            if ($depth < 0) {
                &check_uri($params, $u, 0, -1, $cookie, $uri);
            }
            else {
                &check_uri($params, $u, 0, $depth - 1, $cookie, $uri);
            }
        }
    }
    return;
}

###############################################################
# Distribute links based on host:port to avoid RobotUA delays #
###############################################################

sub distribute_links(\%)
{
    my $hostlinks = shift;

    # Hosts ordered by weight (number of links), descending
    my @order =
        sort { scalar(@{$hostlinks->{$b}}) <=> scalar(@{$hostlinks->{$a}}) }
        keys %$hostlinks;

    # All link list flattened into one, in host weight order
    my @all;
    push(@all, @{$hostlinks->{$_}}) for @order;

    return @all if (scalar(@order) < 2);

    # Indexes and chunk size for "zipping" the end result list
    my $num = scalar(@{$hostlinks->{$order[0]}});
    my @indexes = map { $_ * $num } (0 .. $num - 1);

    # Distribute them
    my @result;
    while (my @chunk = splice(@all, 0, $num)) {
        @result[@indexes] = @chunk;
        @indexes = map { $_ + 1 } @indexes;
    }

    # Weed out undefs
    @result = grep(defined, @result);

    return @result;
}

##########################################
# Decode Content-Encodings in a response #
##########################################

sub decode_content ($)
{
    my $response = shift;
    my $error    = undef;

    my $docref = $response->decoded_content(ref => 1);
    if (defined($docref)) {
        utf8::encode($$docref);
        $response->content_ref($docref);

        # Remove Content-Encoding so it won't be decoded again later.
        $response->remove_header('Content-Encoding');
    }
    else {
        my $ce = $response->header('Content-Encoding');
        $ce = defined($ce) ? "'$ce'" : 'undefined';
        my $ct = $response->header('Content-Type');
        $ct = defined($ct) ? "'$ct'" : 'undefined';
        my $request_uri = $response->request->url;

        my $cs = $response->content_charset();
        $cs = defined($cs) ? "'$cs'" : 'unknown';
        $error =
            "Error decoding document at <$request_uri>, Content-Type $ct, " .
            "Content-Encoding $ce, content charset $cs: '$@'";
    }
    return $error;
}

#######################################
# Get and parse a resource to process #
#######################################

sub get_document ($\$$;\%\$$$$$)
{
    my ($method, $uri,    $in_recursion, $redirects, $referer,
        $cookie, $params, $check_num,    $is_start
    ) = @_;

    # $method contains the HTTP method the use (GET or HEAD)
    # $uri object contains the identifier of the resource
    # $in_recursion is > 0 if we are in recursion mode (i.e. it is at least
    #                        the second resource checked)
    # $redirects is a pointer to the hash containing the map of the redirects
    # $referer is the URI object of the referring document
    # $cookie, $params, $check_num, and $is_start are for printing HTTP headers
    #                  and the form if $in_recursion == 0 and not authenticating

    # Get the resource
    my $response;
    if (defined($results{$uri}{response}) &&
        !($method eq 'GET' && $results{$uri}{method} eq 'HEAD'))
    {
        $response = $results{$uri}{response};
    }
    else {
        $response = &get_uri($method, $uri, $referer);
        &record_results($uri, $method, $response, $referer);
        &record_redirects($redirects, $response);
    }
    if (!$response->is_success()) {
        if (!$in_recursion) {

            # Is it too late to request authentication?
            if ($response->code() == 401) {
                &authentication($response, $cookie, $params, $check_num,
                    $is_start);
            }
            else {
                if ($Opts{HTML}) {
                    &html_header($uri, $cookie) if ($check_num == 1);
                    &print_form($params, $cookie, $check_num) if $is_start;
                    print "<p>", &status_icon($response->code());
                }
                &hprintf("\nError: %d %s\n",
                    $response->code(), $response->message() || '(no message)');
                print "</p>\n" if $Opts{HTML};
            }
        }
        $response->{Stop} = 1;
        $response->content("");
        return ($response);
    }

    # What is the URI of the resource that we are processing by the way?
    my $base_uri    = $response->base();
    my $request_uri = URI->new($response->request->url);
    $response->{absolute_uri} = $request_uri->abs($base_uri);

    # Can we parse the document?
    my $failed_reason;
    my $ct = $response->header('Content-Type');
    if (!$ct || $ct !~ $ContentTypes) {
        $failed_reason = "Content-Type for <$request_uri> is " .
            (defined($ct) ? "'$ct'" : 'undefined');
    }
    else {
        $failed_reason = decode_content($response);
    }
    if ($failed_reason) {

        # No, there is a problem...
        if (!$in_recursion) {
            if ($Opts{HTML}) {
                &html_header($uri, $cookie) if ($check_num == 1);
                &print_form($params, $cookie, $check_num) if $is_start;
                print "<p>", &status_icon(406);

            }
            &hprintf("Can't check links: %s.\n", $failed_reason);
            print "</p>\n" if $Opts{HTML};
        }
        $response->{Stop} = 1;
        $response->content("");
    }

    # Ok, return the information
    return ($response);
}

#########################################################
# Check whether a URI is within the scope of recursion. #
#########################################################

sub in_recursion_scope (\$)
{
    my ($uri) = @_;
    return 0 unless $uri;

    my $candidate = $uri->canonical();

    return 0 if (defined($Opts{Exclude}) && $candidate =~ $Opts{Exclude});

    for my $excluded_doc (@{$Opts{Exclude_Docs}}) {
        return 0 if ($candidate =~ $excluded_doc);
    }

    for my $base (@{$Opts{Base_Locations}}) {
        my $rel = $candidate->rel($base);
        next if ($candidate eq $rel);    # Relative path not possible?
        next if ($rel =~ m|^(\.\.)?/|);  # Relative path upwards?
        return 1;
    }

    return 0;    # We always have at least one base location, but none matched.
}

#################################
# Check for content type match. #
#################################

sub is_content_type ($$)
{
    my ($candidate, $type) = @_;
    return 0 unless ($candidate && $type);
    my @v = HTTP::Headers::Util::split_header_words($candidate);
    return scalar(@v) ? $type eq lc($v[0]->[0]) : 0;
}

##################################################
# Check whether a URI has already been processed #
##################################################

sub already_processed (\$\$)
{
    my ($uri, $referer) = @_;

    # Don't be verbose for that part...
    my $summary_value = $Opts{Summary_Only};
    $Opts{Summary_Only} = 1;

    # Do a GET: if it fails, we stop, if not, the results are cached
    my $response = &get_document('GET', $uri, 1, undef, $referer);

    # ... but just for that part
    $Opts{Summary_Only} = $summary_value;

    # Can we process the resource?
    return -1 if defined($response->{Stop});

    # Have we already processed it?
    return 1 if defined($processed{$response->{absolute_uri}->as_string()});

    # It's not processed yet and it is processable: return 0
    return 0;
}

############################
# Get the content of a URI #
############################

sub get_uri ($\$;\$$\%$$$$)
{

    # Here we have a lot of extra parameters in order not to lose information
    # if the function is called several times (401's)
    my ($method, $uri,   $referer, $start, $redirects,
        $code,   $realm, $message, $auth
    ) = @_;

    # $method contains the method used
    # $uri object contains the target of the request
    # $referer is the URI object of the referring document
    # $start is a timestamp (not defined the first time the function is called)
    # $redirects is a map of redirects
    # $code is the first HTTP return code
    # $realm is the realm of the request
    # $message is the HTTP message received
    # $auth equals 1 if we want to send out authentication information

    # For timing purposes
    $start = &get_timestamp() unless defined($start);

    # Prepare the query

    # Do we want printouts of progress?
    my $verbose_progress =
        !($Opts{Summary_Only} || (!$doc_count && $Opts{HTML}));

    &hprintf("%s %s ", $method, $uri) if $verbose_progress;

    my $request = HTTP::Request->new($method, $uri);

    $request->header('Accept-Language' => $Opts{Accept_Language})
        if $Opts{Accept_Language};
    $request->header('Accept', $Accept);
    $request->accept_decodable();

    # Are we providing authentication info?
    if ($auth && $request->url()->host() =~ $Opts{Trusted}) {
        if (defined($ENV{HTTP_AUTHORIZATION})) {
            $request->header(Authorization => $ENV{HTTP_AUTHORIZATION});
        }
        elsif (defined($Opts{User}) && defined($Opts{Password})) {
            $request->authorization_basic($Opts{User}, $Opts{Password});
        }
    }

    # Tell the user agent if we want progress reports for redirects or not.
    $ua->redirect_progress_callback(sub { &hprintf("\n-> %s %s ", @_); })
        if $verbose_progress;

    # Set referer
    $request->referer($referer) if (!$Opts{No_Referer} && $referer);

    # Telling caches in the middle we want a fresh copy (Bug 4998)
    $request->header(Cache_Control => "max-age=0");

    # Do the query
    my $response = $ua->request($request);

    # Get the results
    # Record the very first response
    if (!defined($code)) {
        ($code, $message) = delete(@$ua{qw(FirstResponse FirstMessage)});
    }

    # Authentication requested?
    if ($response->code() == 401 &&
        !defined($auth) &&
        (defined($ENV{HTTP_AUTHORIZATION}) ||
            (defined($Opts{User}) && defined($Opts{Password})))
        )
    {

        # Set host as trusted domain unless we already have one.
        if (!$Opts{Trusted}) {
            my $re = sprintf('^%s$', quotemeta($response->base()->host()));
            $Opts{Trusted} = qr/$re/io;
        }

        # Deal with authentication and avoid loops
        if (!defined($realm) &&
            $response->www_authenticate() =~ /Basic realm=\"([^\"]+)\"/)
        {
            $realm = $1;
        }

        print "\n" if $verbose_progress;
        return &get_uri($method, $response->request()->url(),
            $referer, $start, $redirects, $code, $realm, $message, 1);
    }

    # @@@ subtract robot delay from the "fetched in" time?
    &hprintf(" fetched in %s seconds\n", &time_diff($start, &get_timestamp()))
        if $verbose_progress;

    $response->{IsCss} =
        is_content_type($response->content_type(), "text/css");
    $response->{Realm} = $realm if defined($realm);

    return $response;
}

#########################################
# Record the results of an HTTP request #
#########################################

sub record_results (\$$$$)
{
    my ($uri, $method, $response, $referer) = @_;
    $results{$uri}{referer}        = $referer;
    $results{$uri}{response}       = $response;
    $results{$uri}{method}         = $method;
    $results{$uri}{location}{code} = $response->code();
    $results{$uri}{location}{code} = RC_ROBOTS_TXT()
        if ($results{$uri}{location}{code} == 403 &&
        $response->message() =~ /Forbidden by robots\.txt/);
    $results{$uri}{location}{code} = RC_IP_DISALLOWED()
        if ($results{$uri}{location}{code} == 403 &&
        $response->message() =~ /non-public IP/);
    $results{$uri}{location}{code} = RC_DNS_ERROR()
        if ($results{$uri}{location}{code} == 500 &&
        $response->message() =~ /Bad hostname '[^\']*'/);
    $results{$uri}{location}{code} = RC_PROTOCOL_DISALLOWED()
        if ($results{$uri}{location}{code} == 500 &&
        $response->message() =~ /Access to '[^\']*' URIs has been disabled/);
    $results{$uri}{location}{type}    = $response->header('Content-type');
    $results{$uri}{location}{display} = $results{$uri}{location}{code};

    # Rewind, check for the original code and message.
    for (my $tmp = $response->previous(); $tmp; $tmp = $tmp->previous()) {
        $results{$uri}{location}{orig}         = $tmp->code();
        $results{$uri}{location}{orig_message} = $tmp->message() ||
            '(no message)';
    }
    $results{$uri}{location}{success} = $response->is_success();

    # If a suppressed broken link, fill the data structure like a typical success.
    # print STDERR "success? " . $results{$uri}{location}{success} . ": $uri\n";
    if (!$results{$uri}{location}{success}) {
        my $code = $results{$uri}{location}{code};
        my $match = grep { $_ eq "$code:$uri" } @{$Opts{Suppress_Broken}};
        if ($match) {
            $results{$uri}{location}{success} = 1;
            $results{$uri}{location}{code}    = 100;
            $results{$uri}{location}{display} = 100;
        }
    }

    # Stores the authentication information
    if (defined($response->{Realm})) {
        $results{$uri}{location}{realm} = $response->{Realm};
        $results{$uri}{location}{display} = 401 unless $Opts{Hide_Same_Realm};
    }

    # What type of broken link is it? (stored in {record} - the {display}
    #              information is just for visual use only)
    if ($results{$uri}{location}{display} == 401 &&
        $results{$uri}{location}{code} == 404)
    {
        $results{$uri}{location}{record} = 404;
    }
    else {
        $results{$uri}{location}{record} = $results{$uri}{location}{display};
    }

    # Did it fail?
    $results{$uri}{location}{message} = $response->message() || '(no message)';
    if (!$results{$uri}{location}{success}) {
        &hprintf(
            "Error: %d %s\n",
            $results{$uri}{location}{code},
            $results{$uri}{location}{message}
        ) if ($Opts{Verbose});
    }
    return;
}

####################
# Parse a document #
####################

sub parse_document (\$\$$$$)
{
    my ($uri, $base_uri, $response, $links, $rec_needs_links) = @_;

    print("parse_document($uri, $base_uri, ..., $links, $rec_needs_links)\n")
        if $Opts{Verbose};

    my $p;

    if (defined($results{$uri}{parsing})) {

        # We have already done the job. Woohoo!
        $p->{base}    = $results{$uri}{parsing}{base};
        $p->{Anchors} = $results{$uri}{parsing}{Anchors};
        $p->{Links}   = $results{$uri}{parsing}{Links};
        return $p;
    }

    $p = W3C::LinkChecker->new();
    $p->{base} = $base_uri;

    my $stype = $response->header("Content-Style-Type");
    $p->{style_is_css} = !$stype || is_content_type($stype, "text/css");

    my $start;
    if (!$Opts{Summary_Only}) {
        $start = &get_timestamp();
        print("Parsing...\n");
    }

    # Content-Encoding etc already decoded in get_document().
    my $docref = $response->content_ref();

    # Count lines beforehand if needed (for progress indicator, or CSS while
    # we don't get any line context out of the parser).  In case of HTML, the
    # actual final number of lines processed shown is populated by our
    # end_document handler.
    $p->{Total} = ($$docref =~ tr/\n//)
        if ($response->{IsCss} || $Opts{Progress});

    # We only look for anchors if we are not interested in the links
    # obviously, or if we are running a recursive checking because we
    # might need this information later
    $p->{only_anchors} = !($links || $rec_needs_links);

    if ($response->{IsCss}) {

        # Parse as CSS

        $p->parse_css($$docref, LINE_UNKNOWN());
    }
    else {

        # Parse as HTML

        # Transform <?xml:stylesheet ...?> into <xml:stylesheet ...> for parsing
        # Processing instructions are not parsed by process, but in this case
        # it should be. It's expensive, it's horrible, but it's the easiest way
        # for right now.
        $$docref =~ s/\<\?(xml:stylesheet.*?)\?\>/\<$1\>/
            unless $p->{only_anchors};

        $p->xml_mode(1) if ($response->content_type() =~ /\+xml$/);

        $p->parse($$docref)->eof();
    }

    $response->content("");

    if (!$Opts{Summary_Only}) {
        my $stop = &get_timestamp();
        print "\r" if $Opts{Progress};
        &hprintf(" done (%d lines in %s seconds).\n",
            $p->{Total}, &time_diff($start, $stop));
    }

    # Save the results before exiting
    $results{$uri}{parsing}{base}    = $p->{base};
    $results{$uri}{parsing}{Anchors} = $p->{Anchors};
    $results{$uri}{parsing}{Links}   = $p->{Links};

    return $p;
}

####################################
# Constructor for W3C::LinkChecker #
####################################

sub new
{
    my $p = HTML::Parser::new(@_, api_version => 3);
    $p->utf8_mode(1);

    # Set up handlers

    $p->handler(start => 'start', 'self, tagname, attr, line');
    $p->handler(end   => 'end',   'self, tagname, line');
    $p->handler(text  => 'text',  'self, dtext, line');
    $p->handler(
        declaration => sub {
            my $self = shift;
            $self->declaration(substr($_[0], 2, -1));
        },
        'self, text, line'
    );
    $p->handler(end_document => 'end_document', 'self, line');
    if ($Opts{Progress}) {
        $p->handler(default => 'parse_progress', 'self, line');
        $p->{last_percentage} = 0;
    }

    # Check <a [..] name="...">?
    $p->{check_name} = 1;

    # Check <[..] id="..">?
    $p->{check_id} = 1;

    # Don't interpret comment loosely
    $p->strict_comment(1);

    return $p;
}

#################################################
# Record or return  the doctype of the document #
#################################################

sub doctype
{
    my ($self, $dc) = @_;
    return $self->{doctype} unless $dc;
    $_ = $self->{doctype} = $dc;

    # What to look for depending on the doctype

    # Check for <a name="...">?
    $self->{check_name} = 0
        if m%^-//(W3C|WAPFORUM)//DTD XHTML (Basic|Mobile) %;

    # Check for <* id="...">?
    $self->{check_id} = 0
        if (m%^-//IETF//DTD HTML [23]\.0//% || m%^-//W3C//DTD HTML 3\.2//%);

    # Enable XML mode (XHTML, XHTML Mobile, XHTML-Print, XHTML+RDFa, ...)
    $self->xml_mode(1) if (m%^-//(W3C|WAPFORUM)//DTD XHTML[ \-\+]%);

    return;
}

###################################
# Print parse progress indication #
###################################

sub parse_progress
{
    my ($self, $line) = @_;
    return unless defined($line) && $line > 0 && $self->{Total} > 0;

    my $percentage = int($line / $self->{Total} * 100);
    if ($percentage != $self->{last_percentage}) {
        printf("\r%4d%%", $percentage);
        $self->{last_percentage} = $percentage;
    }

    return;
}

#############################
# Extraction of the anchors #
#############################

sub get_anchor
{
    my ($self, $tag, $attr) = @_;

    my $anchor = $self->{check_id} ? $attr->{id} : undef;
    if ($self->{check_name} && ($tag eq 'a')) {

        # @@@@ In XHTML, <a name="foo" id="foo"> is mandatory
        # Force an error if it's not the case (or if id's and name's values
        #                                      are different)
        # If id is defined, name if defined must have the same value
        $anchor ||= $attr->{name};
    }

    return $anchor;
}

#############################
# W3C::LinkChecker handlers #
#############################

sub add_link
{
    my ($self, $uri, $base, $line) = @_;
    if (defined($uri)) {

        # Remove repeated slashes after the . or .. in relative links, to avoid
        # duplicated checking or infinite recursion.
        $uri =~ s|^(\.\.?/)/+|$1|o;
        $uri = Encode::decode_utf8($uri);
        $uri = URI->new_abs($uri, $base) if defined($base);
        $self->{Links}{$uri}{defined($line) ? $line : LINE_UNKNOWN()}++;
    }
    return;
}

sub start
{
    my ($self, $tag, $attr, $line) = @_;
    $line = LINE_UNKNOWN() unless defined($line);

    # Anchors
    my $anchor = $self->get_anchor($tag, $attr);
    $self->{Anchors}{$anchor}{$line}++ if defined($anchor);

    # Links
    if (!$self->{only_anchors}) {

        my $tag_local_base = undef;

        # Special case: base/@href
        # @@@TODO: The reason for handling <base href> ourselves is that LWP's
        # head parsing magic fails at least for responses that have
        # Content-Encodings: https://rt.cpan.org/Ticket/Display.html?id=54361
        if ($tag eq 'base') {

            # Ignore <base> with missing/empty href.
            $self->{base} = $attr->{href}
                if (defined($attr->{href}) && length($attr->{href}));
        }

        # Special case: meta[@http-equiv=Refresh]/@content
        elsif ($tag eq 'meta') {
            if ($attr->{'http-equiv'} &&
                lc($attr->{'http-equiv'}) eq 'refresh')
            {
                my $content = $attr->{content};
                if ($content && $content =~ /.*?;\s*(?:url=)?(.+)/i) {
                    $self->add_link($1, undef, $line);
                }
            }
        }

        # Special case: tags that have "local base"
        elsif ($tag eq 'applet' || $tag eq 'object') {
            if (my $codebase = $attr->{codebase}) {

                # Applet codebases are directories, append trailing slash
                # if it's not there so that new_abs does the right thing.
                $codebase .= "/" if ($tag eq 'applet' && $codebase !~ m|/$|);

                # TODO: HTML 4 spec says applet/@codebase may only point to
                # subdirs of the directory containing the current document.
                # Should we do something about that?
                $tag_local_base = URI->new_abs($codebase, $self->{base});
            }
        }

        # Link attributes:
        if (my $link_attrs = LINK_ATTRS()->{$tag}) {
            for my $la (@$link_attrs) {
                $self->add_link($attr->{$la}, $tag_local_base, $line);
            }
        }

        # List of links attributes:
        if (my $link_attrs = LINK_LIST_ATTRS()->{$tag}) {
            my ($sep, $attrs) = @$link_attrs;
            for my $la (@$attrs) {
                if (defined(my $value = $attr->{$la})) {
                    for my $link (split($sep, $value)) {
                        $self->add_link($link, $tag_local_base, $line);
                    }
                }
            }
        }

        # Inline CSS:
        delete $self->{csstext};
        if ($tag eq 'style') {
            $self->{csstext} = ''
                if ((!$attr->{type} && $self->{style_is_css}) ||
                is_content_type($attr->{type}, "text/css"));
        }
        elsif ($self->{style_is_css} && (my $style = $attr->{style})) {
            $style = CSS::DOM::Style::parse($style);
            $self->parse_style($style, $line);
        }
    }

    $self->parse_progress($line) if $Opts{Progress};
    return;
}

sub end
{
    my ($self, $tagname, $line) = @_;

    $self->parse_css($self->{csstext}, $line) if ($tagname eq 'style');
    delete $self->{csstext};

    $self->parse_progress($line) if $Opts{Progress};
    return;
}

sub parse_css
{
    my ($self, $css, $line) = @_;
    return unless $css;

    my $sheet = CSS::DOM::parse($css);
    for my $rule (@{$sheet->cssRules()}) {
        if ($rule->type() == IMPORT_RULE()) {
            $self->add_link($rule->href(), $self->{base}, $line);
        }
        elsif ($rule->type == STYLE_RULE()) {
            $self->parse_style($rule->style(), $line);
        }
    }
    return;
}

sub parse_style
{
    my ($self, $style, $line) = @_;
    return unless $style;

    for (my $i = 0, my $len = $style->length(); $i < $len; $i++) {
        my $prop = $style->item($i);
        my $val  = $style->getPropertyValue($prop);

        while ($val =~ /$CssUrl/go) {
            my $url = CSS::DOM::Util::unescape($2);
            $self->add_link($url, $self->{base}, $line);
        }
    }

    return;
}

sub declaration
{
    my ($self, $text, $line) = @_;

    # Extract the doctype
    my @declaration = split(/\s+/, $text, 4);
    if ($#declaration >= 3 &&
        $declaration[0] eq 'DOCTYPE' &&
        lc($declaration[1]) eq 'html')
    {

        # Parse the doctype declaration
        if ($text =~
            m/^DOCTYPE\s+html\s+(?:PUBLIC\s+"([^"]+)"|SYSTEM)(\s+"([^"]+)")?\s*$/i
            )
        {

            # Store the doctype
            $self->doctype($1) if $1;

            # If there is a link to the DTD, record it
            $self->add_link($3, undef, $line)
                if (!$self->{only_anchors} && $3);
        }
    }

    $self->text($text) unless $self->{only_anchors};

    return;
}

sub text
{
    my ($self, $text, $line) = @_;
    $self->{csstext} .= $text if defined($self->{csstext});
    $self->parse_progress($line) if $Opts{Progress};
    return;
}

sub end_document
{
    my ($self, $line) = @_;
    $self->{Total} = $line;
    delete $self->{csstext};
    return;
}

################################
# Check the validity of a link #
################################

sub check_validity (\$\$$\%\%)
{
    my ($referer, $uri, $want_links, $links, $redirects) = @_;

    # $referer is the URI object of the document checked
    # $uri is the URI object of the target that we are verifying
    # $want_links is true if we're interested in links in the target doc
    # $links is a hash of the links in the documents checked
    # $redirects is a map of the redirects encountered

    # Get the document with the appropriate method: GET if there are
    # fragments to check or links are wanted, HEAD is enough otherwise.
    my $fragments = $links->{$uri}{fragments} || {};
    my $method = ($want_links || %$fragments) ? 'GET' : 'HEAD';

    my $response;
    my $being_processed = 0;
    if (!defined($results{$uri}) ||
        ($method eq 'GET' && $results{$uri}{method} eq 'HEAD'))
    {
        $being_processed = 1;
        $response = &get_uri($method, $uri, $referer);

        # Get the information back from get_uri()
        &record_results($uri, $method, $response, $referer);

        # Record the redirects
        &record_redirects($redirects, $response);
    }
    elsif (!($Opts{Summary_Only} || (!$doc_count && $Opts{HTML}))) {
        my $ref = $results{$uri}{referer};
        &hprintf("Already checked%s\n", $ref ? ", referrer $ref" : ".");
    }

    # We got the response of the HTTP request. Stop here if it was a HEAD.
    return if ($method eq 'HEAD');

    # There are fragments. Parse the document.
    my $p;
    if ($being_processed) {

        # Can we really parse the document?
        if (!defined($results{$uri}{location}{type}) ||
            $results{$uri}{location}{type} !~ $ContentTypes)
        {
            &hprintf("Can't check content: Content-Type for '%s' is '%s'.\n",
                $uri, $results{$uri}{location}{type})
                if ($Opts{Verbose});
            $response->content("");
            return;
        }

        # Do it then
        if (my $error = decode_content($response)) {
            &hprintf("%s\n.", $error);
        }

        # @@@TODO: this isn't the best thing to do if a decode error occurred
        $p =
            &parse_document($uri, $response->base(), $response, 0,
            $want_links);
    }
    else {

        # We already had the information
        $p->{Anchors} = $results{$uri}{parsing}{Anchors};
    }

    # Check that the fragments exist
    for my $fragment (keys %$fragments) {
        if (defined($p->{Anchors}{$fragment}) ||
            &escape_match($fragment, $p->{Anchors}) ||
            grep { $_ eq "$uri#$fragment" } @{$Opts{Suppress_Fragment}})
        {
            $results{$uri}{fragments}{$fragment} = 1;
        }
        else {
            $results{$uri}{fragments}{$fragment} = 0;
        }
    }
    return;
}

sub escape_match ($\%)
{
    my ($a, $hash) = (URI::Escape::uri_unescape($_[0]), $_[1]);
    for my $b (keys %$hash) {
        return 1 if ($a eq URI::Escape::uri_unescape($b));
    }
    return 0;
}

##########################
# Ask for authentication #
##########################

sub authentication ($;$$$$)
{
    my ($response, $cookie, $params, $check_num, $is_start) = @_;

    my $realm = '';
    if ($response->www_authenticate() =~ /Basic realm=\"([^\"]+)\"/) {
        $realm = $1;
    }

    if ($Opts{Command_Line}) {
        printf STDERR <<'EOF', $response->request()->url(), $realm;

Authentication is required for %s.
The realm is "%s".
Use the -u and -p options to specify a username and password and the -d option
to specify trusted domains.
EOF
    }
    else {

        printf(
            "Status: 401 Authorization Required\nWWW-Authenticate: %s\n%sConnection: close\nContent-Language: en\nContent-Type: text/html; charset=utf-8\n\n",
            $response->www_authenticate(),
            $cookie ? "Set-Cookie: $cookie\n" : "",
        );

        printf(
            "%s
<html lang=\"en\" xmlns=\"http://www.w3.org/1999/xhtml\" xml:lang=\"en\">
<head>
<title>W3C Link Checker: 401 Authorization Required</title>
%s</head>
<body>", $DocType, $Head
        );
        &banner(': 401 Authorization Required');
        &print_form($params, $cookie, $check_num) if $is_start;
        printf(
            '<p>
  %s
  You need "%s" access to <a href="%s">%s</a> to perform link checking.<br />
',
            &status_icon(401),
            &encode($realm), (&encode($response->request()->url())) x 2
        );

        my $host = $response->request()->url()->host();
        if ($Opts{Trusted} && $host !~ $Opts{Trusted}) {
            printf <<'EOF', &encode($Opts{Trusted}), &encode($host);
  This service has been configured to send authentication only to hostnames
  matching the regular expression <code>%s</code>, but the hostname
  <code>%s</code> does not match it.
EOF
        }

        print "</p>\n";
    }
    return;
}

##################
# Get statistics #
##################

sub get_timestamp ()
{
    return pack('LL', Time::HiRes::gettimeofday());
}

sub time_diff ($$)
{
    my @start = unpack('LL', $_[0]);
    my @stop  = unpack('LL', $_[1]);
    for ($start[1], $stop[1]) {
        $_ /= 1_000_000;
    }
    return (sprintf("%.2f", ($stop[0] + $stop[1]) - ($start[0] + $start[1])));
}

########################
# Handle the redirects #
########################

# Record the redirects in a hash
sub record_redirects (\%$)
{
    my ($redirects, $response) = @_;
    for (my $prev = $response->previous(); $prev; $prev = $prev->previous()) {

        # Check for redirect match.
        my $from = $prev->request()->url();
        my $to   = $response->request()->url();  # same on every loop iteration
        my $from_to = $from . '->' . $to;
        my $match = grep { $_ eq $from_to } @{$Opts{Suppress_Redirect}};

        # print STDERR "Result $match of redirect checking $from_to\n";
        if ($match) { next; }

        $match = grep { $from_to =~ /$_/ } @{$Opts{Suppress_Redirect_Prefix}};

        # print STDERR "Result $match of regexp checking $from_to\n";
        if ($match) { next; }

        my $c = $prev->code();
        if ($Opts{Suppress_Temp_Redirects} && ($c == 307 || $c == 302)) {
            next;
        }

        $redirects->{$prev->request()->url()} = $response->request()->url();
    }
    return;
}

# Determine if a request is redirected
sub is_redirected ($%)
{
    my ($uri, %redirects) = @_;
    return (defined($redirects{$uri}));
}

# Get a list of redirects for a URI
sub get_redirects ($%)
{
    my ($uri, %redirects) = @_;
    my @history = ($uri);
    my %seen    = ($uri => 1);    # for tracking redirect loops
    my $loop    = 0;
    while ($redirects{$uri}) {
        $uri = $redirects{$uri};
        push(@history, $uri);
        if ($seen{$uri}) {
            $loop = 1;
            last;
        }
        else {
            $seen{$uri}++;
        }
    }
    return ($loop, @history);
}

####################################################
# Tool for sorting the unique elements of an array #
####################################################

sub sort_unique (@)
{
    my %saw;
    @saw{@_} = ();
    return (sort { $a <=> $b } keys %saw);
}

#####################
# Print the results #
#####################

sub line_number ($)
{
    my $line = shift;
    return $line if ($line >= 0);
    return "(N/A)";
}

sub http_rc ($)
{
    my $rc = shift;
    return $rc if ($rc >= 0);
    return "(N/A)";
}

# returns true if the given code is informational
sub informational ($)
{
    my $rc = shift;
    return $rc == RC_ROBOTS_TXT() ||
        $rc == RC_IP_DISALLOWED() ||
        $rc == RC_PROTOCOL_DISALLOWED();
}

sub anchors_summary (\%\%)
{
    my ($anchors, $errors) = @_;

    # Number of anchors found.
    my $n = scalar(keys(%$anchors));
    if (!$Opts{Quiet}) {
        if ($Opts{HTML}) {
            print("<h3>Anchors</h3>\n<p>");
        }
        else {
            print("Anchors\n\n");
        }
        &hprintf("Found %d anchor%s.\n", $n, ($n == 1) ? '' : 's');
        print("</p>\n") if $Opts{HTML};
    }

    # List of the duplicates, if any.
    my @errors = keys %{$errors};
    if (!scalar(@errors)) {
        print("<p>Valid anchors!</p>\n")
            if (!$Opts{Quiet} && $Opts{HTML} && $n);
        return;
    }
    undef $n;

    print_doc_header();
    print('<p>') if $Opts{HTML};
    print('List of duplicate and empty anchors');
    print <<'EOF' if $Opts{HTML};
</p>
<table class="report" border="1" summary="List of duplicate and empty anchors.">
<thead>
<tr>
<th>Anchor</th>
<th>Lines</th>
</tr>
</thead>
<tbody>
EOF
    print("\n");

    for my $anchor (@errors) {
        my $format;
        my @unique = &sort_unique(
            map { line_number($_) }
                keys %{$anchors->{$anchor}}
        );
        if ($Opts{HTML}) {
            $format = "<tr><td class=\"broken\">%s</td><td>%s</td></tr>\n";
        }
        else {
            my $s = (scalar(@unique) > 1) ? 's' : '';
            $format = "\t%s\tLine$s: %s\n";
        }
        printf($format,
            &encode(length($anchor) ? $anchor : 'Empty anchor'),
            join(', ', @unique));
    }

    print("</tbody>\n</table>\n") if $Opts{HTML};

    return;
}

sub show_link_report (\%\%\%\%\@;$\%)
{
    my ($links, $results, $broken, $redirects, $urls, $codes, $todo) = @_;

    print("\n<dl class=\"report\">") if $Opts{HTML};
    print("\n") if (!$Opts{Quiet});

    # Process each URL
    my ($c, $previous_c);
    for my $u (@$urls) {
        my @fragments = keys %{$broken->{$u}{fragments}};

        # Did we get a redirect?
        my $redirected = &is_redirected($u, %$redirects);

        # List of lines
        my @total_lines;
        push(@total_lines, keys(%{$links->{$u}{location}}));
        for my $f (@fragments) {
            push(@total_lines, keys(%{$links->{$u}{fragments}{$f}}))
                unless ($f eq $u && defined($links->{$u}{$u}{LINE_UNKNOWN()}));
        }

        my ($redirect_loop, @redirects_urls) = get_redirects($u, %$redirects);
        my $currloc = $results->{$u}{location};

        # Error type
        $c = &code_shown($u, $results);

        # What to do
        my $whattodo;
        my $redirect_too;
        if ($todo) {
            if ($u =~ m/^javascript:/) {
                if ($Opts{HTML}) {
                    $whattodo =
                        'You must change this link: people using a browser without JavaScript support
will <em>not</em> be able to follow this link. See the
<a href="http://www.w3.org/TR/WAI-WEBCONTENT/#tech-scripts">Web Content
Accessibility Guidelines on the use of scripting on the Web</a> and the
<a href="http://www.w3.org/TR/WCAG10-HTML-TECHS/#directly-accessible-scripts">techniques
on how to solve this</a>.';
                }
                else {
                    $whattodo =
                        'Change this link: people using a browser without JavaScript support will not be able to follow this link.';
                }
            }
            elsif ($c == RC_ROBOTS_TXT()) {
                $whattodo =
                    'The link was not checked due to robots exclusion ' .
                    'rules. Check the link manually.';
            }
            elsif ($redirect_loop) {
                $whattodo =
                    'Retrieving the URI results in a redirect loop, that should be '
                    . 'fixed.  Examine the redirect sequence to see where the loop '
                    . 'occurs.';
            }
            else {
                $whattodo = $todo->{$c};
            }
        }
        elsif (defined($redirects{$u})) {

            # Redirects
            if (($u . '/') eq $redirects{$u}) {
                $whattodo =
                    'The link is missing a trailing slash, and caused a redirect. Adding the trailing slash would speed up browsing.';
            }
            elsif ($c == 307 || $c == 302) {
                $whattodo =
                    'This is a temporary redirect. Update the link if you believe it makes sense, or leave it as is.';
            }
            elsif ($c == 301) {
                $whattodo =
                    'This is a permanent redirect. The link should be updated.';
            }
        }

        my @unique = &sort_unique(map { line_number($_) } @total_lines);
        my $lines_list = join(', ', @unique);
        my $s = (scalar(@unique) > 1) ? 's' : '';
        undef @unique;

        my @http_codes = ($currloc->{code});
        unshift(@http_codes, $currloc->{orig}) if $currloc->{orig};
        @http_codes = map { http_rc($_) } @http_codes;

        if ($Opts{HTML}) {

            # Style stuff
            my $idref = '';
            if ($codes && (!defined($previous_c) || ($c != $previous_c))) {
                $idref      = ' id="d' . $doc_count . 'code_' . $c . '"';
                $previous_c = $c;
            }

            # Main info
            for (@redirects_urls) {
                $_ = &show_url($_);
            }

            # HTTP message
            my $http_message;
            if ($currloc->{message}) {
                $http_message = &encode($currloc->{message});
                if ($c == 404 || $c == 500) {
                    $http_message =
                        '<span class="broken">' . $http_message . '</span>';
                }
            }
            my $redirmsg =
                $redirect_loop ? ' <em>redirect loop detected</em>' : '';
            printf("
<dt%s>%s <span class='msg_loc'>Line%s: %s</span> %s</dt>
<dd class='responsecode'><strong>Status</strong>: %s %s %s</dd>
<dd class='message_explanation'><p>%s %s</p></dd>\n",

                # Anchor for return codes
                $idref,

                # Color
                &status_icon($c),
                $s,

                # List of lines
                $lines_list,

                # List of redirects
                $redirected ?
                    join(' redirected to ', @redirects_urls) . $redirmsg :
                    &show_url($u),

                # Realm
                defined($currloc->{realm}) ?
                    sprintf('Realm: %s<br />', &encode($currloc->{realm})) :
                    '',

                # HTTP original message
                # defined($currloc->{orig_message})
                # ? &encode($currloc->{orig_message}).
                # ' <span title="redirected to">-&gt;</span> '
                # : '',

                # Response code chain
                join(
                    ' <span class="redirected_to" title="redirected to">-&gt;</span> ',
                    map { &encode($_) } @http_codes),

                # HTTP final message
                $http_message,

                # What to do
                $whattodo,

                # Redirect too?
                $redirect_too ?
                    sprintf(' <span %s>%s</span>',
                    &bgcolor(301), $redirect_too) :
                    '',
            );
            if ($#fragments >= 0) {
                printf("<dd>Broken fragments: <ul>\n");
            }
        }
        else {
            my $redirmsg = $redirect_loop ? ' redirect loop detected' : '';
            printf(
                "\n%s\t%s\n  Code: %s %s\n%s\n",

                # List of redirects
                $redirected ? join("\n-> ", @redirects_urls) . $redirmsg : $u,

                # List of lines
                $lines_list ? sprintf("\n%6s: %s", "Line$s", $lines_list) : '',

                # Response code chain
                join(' -> ', @http_codes),

                # HTTP message
                $currloc->{message} || '',

                # What to do
                wrap(' To do: ', '        ', $whattodo)
            );
            if ($#fragments >= 0) {
                if ($currloc->{code} == 200) {
                    print("The following fragments need to be fixed:\n");
                }
                else {
                    print("Fragments:\n");
                }
            }
        }

        # Fragments
        for my $f (@fragments) {
            my @unique_lines =
                &sort_unique(keys %{$links->{$u}{fragments}{$f}});
            my $plural = (scalar(@unique_lines) > 1) ? 's' : '';
            my $unique_lines = join(', ', @unique_lines);
            if ($Opts{HTML}) {
                printf("<li>%s<em>#%s</em> (line%s %s)</li>\n",
                    &encode($u), &encode($f), $plural, $unique_lines);
            }
            else {
                printf("\t%-30s\tLine%s: %s\n", $f, $plural, $unique_lines);
            }
        }

        print("</ul></dd>\n") if ($Opts{HTML} && scalar(@fragments));
    }

    # End of the table
    print("</dl>\n") if $Opts{HTML};

    return;
}

sub code_shown ($$)
{
    my ($u, $results) = @_;

    if ($results->{$u}{location}{record} == 200) {
        return $results->{$u}{location}{orig} ||
            $results->{$u}{location}{record};
    }
    else {
        return $results->{$u}{location}{record};
    }
}

sub links_summary (\%\%\%\%)
{

    # Advices to fix the problems

    my %todo = (
        200 =>
            'Some of the links to this resource point to broken URI fragments (such as index.html#fragment).',
        300 =>
            'This often happens when a typo in the link gets corrected automatically by the server. For the sake of performance, the link should be fixed.',
        301 =>
            'This is a permanent redirect. The link should be updated to point to the more recent URI.',
        302 =>
            'This is a temporary redirect. Update the link if you believe it makes sense, or leave it as is.',
        303 =>
            'This rare status code points to a "See Other" resource. There is generally nothing to be done.',
        307 =>
            'This is a temporary redirect. Update the link if you believe it makes sense, or leave it as is.',
        400 =>
            'This is usually the sign of a malformed URL that cannot be parsed by the server. Check the syntax of the link.',
        401 =>
            "The link is not public and the actual resource is only available behind authentication. If not already done, you could specify it.",
        403 =>
            'The link is forbidden! This needs fixing. Usual suspects: a missing index.html or Overview.html, or a missing ACL.',
        404 =>
            'The link is broken. Double-check that you have not made any typo, or mistake in copy-pasting. If the link points to a resource that no longer exists, you may want to remove or fix the link.',
        405 =>
            'The server does not allow HTTP HEAD requests, which prevents the Link Checker to check the link automatically.  Check the link manually.',
        406 =>
            "The server isn't capable of responding according to the Accept* headers sent. This is likely to be a server-side issue with negotiation.",
        407 => 'The link is a proxy, but requires Authentication.',
        408 => 'The request timed out.',
        410 => 'The resource is gone. You should remove this link.',
        415 => 'The media type is not supported.',
        500 => 'This is a server side problem. Check the URI.',
        501 =>
            'Could not check this link: method not implemented or scheme not supported.',
        503 =>
            'The server cannot service the request, for some unknown reason.',

        # Non-HTTP codes:
        RC_ROBOTS_TXT() => sprintf(
            'The link was not checked due to %srobots exclusion rules%s. Check the link manually, and see also the link checker %sdocumentation on robots exclusion%s.',
            $Opts{HTML} ? (
                '<a href="http://www.robotstxt.org/robotstxt.html">', '</a>',
                "<a href=\"$Cfg{Doc_URI}#bot\">",                     '</a>'
                ) : ('') x 4
        ),
        RC_DNS_ERROR() =>
            'The hostname could not be resolved. Check the link for typos.',
        RC_IP_DISALLOWED() =>
            sprintf(
            'The link resolved to a %snon-public IP address%s, and this link checker instance has been configured to not access such addresses. This may be a real error or just a quirk of the name resolver configuration on the server where the link checker runs. Check the link manually, in particular its hostname/IP address.',
            $Opts{HTML} ?
                ('<a href="http://www.ietf.org/rfc/rfc1918.txt">', '</a>') :
                ('') x 2),
        RC_PROTOCOL_DISALLOWED() =>
            'Accessing links with this URI scheme has been disabled in link checker.',
    );
    my %priority = (
        410 => 1,
        404 => 2,
        403 => 5,
        200 => 10,
        300 => 15,
        401 => 20
    );

    my ($links, $results, $broken, $redirects) = @_;

    # List of the broken links
    my @urls              = keys %{$broken};
    my @dir_redirect_urls = ();
    if ($Opts{Redirects}) {

        # Add the redirected URI's to the report
        for my $l (keys %$redirects) {
            next
                unless (defined($results->{$l}) &&
                defined($links->{$l}) &&
                !defined($broken->{$l}));

            # Check whether we have a "directory redirect"
            # e.g. http://www.w3.org/TR -> http://www.w3.org/TR/
            my ($redirect_loop, @redirects) = get_redirects($l, %$redirects);
            if ($#redirects == 1) {
                push(@dir_redirect_urls, $l);
                next;
            }
            push(@urls, $l);
        }
    }

    # Broken links and redirects
    if ($#urls < 0) {
        if (!$Opts{Quiet}) {
            print_doc_header();
            if ($Opts{HTML}) {
                print "<h3>Links</h3>\n<p>Valid links!</p>\n";
            }
            else {
                print "\nValid links.\n";
            }
        }
    }
    else {
        print_doc_header();
        print('<h3>') if $Opts{HTML};
        print("\nList of broken links and other issues");

        #print(' and redirects') if $Opts{Redirects};

        # Sort the URI's by HTTP Code
        my %code_summary;
        my @idx;
        for my $u (@urls) {
            if (defined($results->{$u}{location}{record})) {
                my $c = &code_shown($u, $results);
                $code_summary{$c}++;
                push(@idx, $c);
            }
        }
        my @sorted = @urls[
            sort {
                defined($priority{$idx[$a]}) ?
                    defined($priority{$idx[$b]}) ?
                    $priority{$idx[$a]} <=> $priority{$idx[$b]} :
                        -1 :
                    defined($priority{$idx[$b]}) ? 1 :
                    $idx[$a] <=> $idx[$b]
            } 0 .. $#idx
        ];
        @urls = @sorted;
        undef(@sorted);
        undef(@idx);

        if ($Opts{HTML}) {

            # Print a summary
            print <<'EOF';
</h3>
<p><em>There are issues with the URLs listed below. The table summarizes the
issues and suggested actions by HTTP response status code.</em></p>
<table class="report" border="1" summary="List of issues and suggested actions.">
<thead>
<tr>
<th>Code</th>
<th>Occurrences</th>
<th>What to do</th>
</tr>
</thead>
<tbody>
EOF
            for my $code (sort(keys(%code_summary))) {
                printf('<tr%s>', &bgcolor($code));
                printf('<td><a href="#d%scode_%s">%s</a></td>',
                    $doc_count, $code, http_rc($code));
                printf('<td>%s</td>', $code_summary{$code});
                printf('<td>%s</td>', $todo{$code});
                print "</tr>\n";
            }
            print "</tbody>\n</table>\n";
        }
        else {
            print(':');
        }
        &show_link_report($links, $results, $broken, $redirects, \@urls, 1,
            \%todo);
    }

    # Show directory redirects
    if ($Opts{Dir_Redirects} && ($#dir_redirect_urls > -1)) {
        print_doc_header();
        print('<h3>') if $Opts{HTML};
        print("\nList of redirects");
        print(
            "</h3>\n<p>The links below are not broken, but the document does not use the exact URL, and the links were redirected. It may be a good idea to link to the final location, for the sake of speed.</p>"
        ) if $Opts{HTML};
        &show_link_report($links, $results, $broken, $redirects,
            \@dir_redirect_urls);
    }

    return;
}

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

################
# Global stats #
################

sub global_stats ()
{
    my $stop = &get_timestamp();
    my $n_docs =
        ($doc_count <= $Opts{Max_Documents}) ? $doc_count :
                                               $Opts{Max_Documents};
    return sprintf(
        'Checked %d document%s in %s seconds.',
        $n_docs,
        ($n_docs == 1) ? '' : 's',
        &time_diff($timestamp, $stop)
    );
}

##################
# HTML interface #
##################

sub html_header ($$)
{
    my ($uri, $cookie) = @_;

    my $title = defined($uri) ? $uri : '';
    $title = ': ' . $title if ($title =~ /\S/);

    my $headers = '';
    if (!$Opts{Command_Line}) {
        $headers .= "Cache-Control: no-cache\nPragma: no-cache\n" if $uri;
        $headers .= "Content-Type: text/html; charset=utf-8\n";
        $headers .= "Set-Cookie: $cookie\n"                       if $cookie;

        # mod_perl 1.99_05 doesn't seem to like it if the "\n\n" isn't in the same
        # print() statement as the last header
        $headers .= "Content-Language: en\n\n";
    }

    my $onload = $uri ? '' :
          ' onload="if(document.getElementById){document.getElementById(\'uri_1\').focus()}"';

    print $headers, $DocType, "
<html lang=\"en\" xmlns=\"http://www.w3.org/1999/xhtml\" xml:lang=\"en\">
<head>
<title>W3C Link Checker", &encode($title), "</title>
",      $Head,   "</head>
<body", $onload, '>';
    &banner($title);
    return;
}

sub banner ($)
{
    my $tagline = "Check links and anchors in Web pages or full Web sites";

    printf(
        <<'EOF', URI->new_abs("../images/no_w3c.png", $Cfg{Doc_URI}), $tagline);
<div id="banner"><h1 id="title"><a href="http://www.w3.org/" title="W3C"><img alt="W3C" id="logo" src="%s" width="110" height="61" /></a>
<a href="checklink"><span>Link Checker</span></a></h1>
<p id="tagline">%s</p></div>
<div id="main">
EOF
    return;
}

sub status_icon($)
{
    my ($code) = @_;
    my $icon_type;
    my $r = HTTP::Response->new($code);
    if ($r->is_success()) {
        $icon_type = 'error'
            ; # if is success but reported, it's because of broken frags => error
    }
    elsif (&informational($code)) {
        $icon_type = 'info';
    }
    elsif ($code == 300) {
        $icon_type = 'info';
    }
    elsif ($code == 401) {
        $icon_type = 'error';
    }
    elsif ($r->is_redirect()) {
        $icon_type = 'warning';
    }
    elsif ($r->is_error()) {
        $icon_type = 'error';
    }
    else {
        $icon_type = 'error';
    }
    return sprintf('<span class="err_type"><img src="%s" alt="%s" /></span>',
        URI->new_abs("../images/info_icons/$icon_type.png", $Cfg{Doc_URI}),
        $icon_type);
}

sub bgcolor ($)
{
    my ($code) = @_;
    my $class;
    my $r = HTTP::Response->new($code);
    if ($r->is_success()) {
        return '';
    }
    elsif ($code == RC_ROBOTS_TXT() || $code == RC_IP_DISALLOWED()) {
        $class = 'dubious';
    }
    elsif ($code == 300) {
        $class = 'multiple';
    }
    elsif ($code == 401) {
        $class = 'unauthorized';
    }
    elsif ($r->is_redirect()) {
        $class = 'redirect';
    }
    elsif ($r->is_error()) {
        $class = 'broken';
    }
    else {
        $class = 'broken';
    }
    return (' class="' . $class . '"');
}

sub show_url ($)
{
    my ($url) = @_;
    return sprintf('<a href="%s">%s</a>', (&encode($url)) x 2);
}

sub html_footer ()
{
    printf("<p>%s</p>\n", &global_stats())
        if ($doc_count > 0 && !$Opts{Quiet});
    if (!$doc_count) {
        print <<'EOF';
<div class="intro">
  <p>
    This Link Checker looks for issues in links, anchors and referenced objects
    in a Web page, CSS style sheet, or recursively on a whole Web site. For
    best results, it is recommended to first ensure that the documents checked
    use Valid <a href="http://validator.w3.org/">(X)HTML Markup</a> and
    <a href="http://jigsaw.w3.org/css-validator/">CSS</a>. The Link Checker is
    part of the W3C's <a href="http://www.w3.org/QA/Tools/">validators and
    Quality Web tools</a>.
  </p>
</div>
EOF
    }
    printf(<<'EOF', $Cfg{Doc_URI}, $Cfg{Doc_URI}, $PACKAGE, $REVISION);
</div><!-- main -->
<ul class="navbar" id="menu">
  <li><a href="%s" accesskey="3" title="Documentation for this Link Checker Service">Docs</a></li>
  <li><a href="http://search.cpan.org/dist/W3C-LinkChecker/" accesskey="2" title="Download the source / Install this service">Download</a></li>
  <li><a href="%s#csb" title="feedback: comments, suggestions and bugs" accesskey="4">Feedback</a></li>
  <li><a href="http://validator.w3.org/" title="Validate your markup with the W3C Markup Validation Service">Validator</a></li>
</ul>
<div>
<address>
%s<br /> %s
</address>
</div>
</body>
</html>
EOF
    return;
}

sub print_form (\%$$)
{
    my ($params, $cookie, $check_num) = @_;

    # Split params on \0, see CGI's docs on Vars()
    while (my ($key, $value) = each(%$params)) {
        if ($value) {
            my @vals = split(/\0/, $value, 2);
            $params->{$key} = $vals[0];
        }
    }

    # Override undefined values from the cookie, if we got one.
    my $valid_cookie = 0;
    if ($cookie) {
        my %cookie_values = $cookie->value();
        if (!$cookie_values{clear})
        {    # XXX no easy way to check if cookie expired?
            $valid_cookie = 1;
            while (my ($key, $value) = each(%cookie_values)) {
                $params->{$key} = $value unless defined($params->{$key});
            }
        }
    }

    my $chk = ' checked="checked"';
    $params->{hide_type} = 'all' unless $params->{hide_type};

    my $requested_uri = &encode($params->{uri} || '');
    my $sum = $params->{summary}        ? $chk : '';
    my $red = $params->{hide_redirects} ? $chk : '';
    my $all = ($params->{hide_type} ne 'dir') ? $chk : '';
    my $dir = $all                            ? ''   : $chk;
    my $acc = $params->{no_accept_language}   ? $chk : '';
    my $ref = $params->{no_referer}           ? $chk : '';
    my $rec = $params->{recursive}            ? $chk : '';
    my $dep = &encode($params->{depth} || '');

    my $cookie_options = '';
    if ($valid_cookie) {
        $cookie_options = "
    <label for=\"cookie1_$check_num\"><input type=\"radio\" id=\"cookie1_$check_num\" name=\"cookie\" value=\"nochanges\" checked=\"checked\" /> Don't modify saved options</label>
    <label for=\"cookie2_$check_num\"><input type=\"radio\" id=\"cookie2_$check_num\" name=\"cookie\" value=\"set\" /> Save these options</label>
    <label for=\"cookie3_$check_num\"><input type=\"radio\" id=\"cookie3_$check_num\" name=\"cookie\" value=\"clear\" /> Clear saved options</label>";
    }
    else {
        $cookie_options = "
    <label for=\"cookie_$check_num\"><input type=\"checkbox\" id=\"cookie_$check_num\" name=\"cookie\" value=\"set\" /> Save options in a <a href=\"http://www.w3.org/Protocols/rfc2109/rfc2109\">cookie</a></label>";
    }

    print "<form action=\"", $Opts{_Self_URI},
        "\" method=\"get\" onsubmit=\"return uriOk($check_num)\" accept-charset=\"UTF-8\">
<p><label for=\"uri_$check_num\">Enter the address (<a href=\"http://www.w3.org/Addressing/\">URL</a>)
of a document that you would like to check:</label></p>
<p><input type=\"text\" size=\"50\" id=\"uri_$check_num\" name=\"uri\" value=\"",
        $requested_uri, "\" /></p>
<fieldset id=\"extra_opt_uri_$check_num\" class=\"moreoptions\">
	<legend class=\"toggletext\">More Options</legend>
	<div class=\"options\">
  <p>
    <label for=\"summary_$check_num\"><input type=\"checkbox\" id=\"summary_$check_num\" name=\"summary\" value=\"on\"",
        $sum, " /> Summary only</label>
    <br />
    <label for=\"hide_redirects_$check_num\"><input type=\"checkbox\" id=\"hide_redirects_$check_num\" name=\"hide_redirects\" value=\"on\"",
        $red,
        " /> Hide <a href=\"http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html#sec10.3\">redirects</a>:</label>
    <label for=\"hide_type_all_$check_num\"><input type=\"radio\" id=\"hide_type_all_$check_num\" name=\"hide_type\" value=\"all\"",
        $all, " /> all</label>
    <label for=\"hide_type_dir_$check_num\"><input type=\"radio\" id=\"hide_type_dir_$check_num\" name=\"hide_type\" value=\"dir\"",
        $dir, " /> for directories only</label>
    <br />
    <label for=\"no_accept_language_$check_num\"><input type=\"checkbox\" id=\"no_accept_language_$check_num\" name=\"no_accept_language\" value=\"on\"",
        $acc,
        " /> Don't send the <tt><a href=\"http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.4\">Accept-Language</a></tt> header</label>
    <br />
    <label for=\"no_referer_$check_num\"><input type=\"checkbox\" id=\"no_referer_$check_num\" name=\"no_referer\" value=\"on\"",
        $ref,
        " /> Don't send the <tt><a href=\"http://www.w3.org/Protocols/rfc2616/rfc2616-sec14.html#sec14.36\">Referer</a></tt> header</label>
    <br />
    <label title=\"Check linked documents recursively (maximum: ",
        $Opts{Max_Documents},
        " documents)\" for=\"recursive_$check_num\"><input type=\"checkbox\" id=\"recursive_$check_num\" name=\"recursive\" value=\"on\"",
        $rec, " /> Check linked documents recursively</label>,
    <label title=\"Depth of the recursion (-1 is the default and means unlimited)\" for=\"depth_$check_num\">recursion depth: <input type=\"text\" size=\"3\" maxlength=\"3\" id=\"depth_$check_num\" name=\"depth\" value=\"",
        $dep, "\" /></label>
    <br /><br />", $cookie_options, "
  </p>
  </div>
</fieldset>
<p class=\"submit_button\"><input type=\"submit\" name=\"check\" value=\"Check\" /></p>
</form>
<div class=\"intro\" id=\"don_program\"></div>
<script type=\"text/javascript\" src=\"http://www.w3.org/QA/Tools/don_prog.js\"></script>
";
    return;
}

sub encode (@)
{
    return $Opts{HTML} ? HTML::Entities::encode(@_) : @_;
}

sub hprintf (@)
{
    print_doc_header();
    if (!$Opts{HTML}) {
        # can have undef values here; avoid useless warning. E.g.,
        #   Error: -1 Forbidden by robots.txt
        #   Use of uninitialized value $_[2] in printf at /usr/local/bin/checklink line 3245.
        # and
        #   Error: 404 File `/u/karl/gnu/src/akarl/doc/dejagnu.html' does not exist
        #   Use of uninitialized value $_[2] in printf at /usr/local/bin/checklink line 3245.
        my @args = ();
        for my $a (@_) {
          push (@args, defined $a ? $a : ""),
        }
        printf(@args);
    }
    else {
        print HTML::Entities::encode(sprintf($_[0], @_[1 .. @_ - 1]));
    }
    return;
}

# Print the document header, if it hasn't been printed already.
# This is invoked before most other output operations, in order
# to enable quiet processing that doesn't clutter the output with
# "Processing..." messages when nothing else will be reported.
sub print_doc_header ()
{
    if (defined($doc_header)) {
        print $doc_header;
        undef($doc_header);
    }
}

# Local Variables:
# mode: perl
# indent-tabs-mode: nil
# cperl-indent-level: 4
# cperl-continued-statement-offset: 4
# cperl-brace-offset: -4
# perl-indent-level: 4
# End:
# ex: ts=4 sw=4 et