Initial commit of stock (IIRC) gitweb.
authorAaron Taylor <ataylor@subgeniuskitty.com>
Thu, 5 Nov 2020 02:54:17 +0000 (02:54 +0000)
committerAaron Taylor <ataylor@subgeniuskitty.com>
Thu, 5 Nov 2020 02:54:17 +0000 (02:54 +0000)
gitweb.cgi [new file with mode: 0755]
static/git-favicon.png [new file with mode: 0644]
static/git-logo.png [new file with mode: 0644]
static/gitweb.css [new file with mode: 0644]
static/gitweb.js [new file with mode: 0644]

diff --git a/gitweb.cgi b/gitweb.cgi
new file mode 100755 (executable)
index 0000000..16021cc
--- /dev/null
@@ -0,0 +1,8429 @@
+#!/usr/bin/perl
+
+# gitweb - simple web interface to track changes in git repositories
+#
+# (C) 2005-2006, Kay Sievers <kay.sievers@vrfy.org>
+# (C) 2005, Christian Gierke
+#
+# This program is licensed under the GPLv2
+
+use 5.008;
+use strict;
+use warnings;
+use CGI qw(:standard :escapeHTML -nosticky);
+use CGI::Util qw(unescape);
+use CGI::Carp qw(fatalsToBrowser set_message);
+use Encode;
+use Fcntl ':mode';
+use File::Find qw();
+use File::Basename qw(basename);
+use Time::HiRes qw(gettimeofday tv_interval);
+binmode STDOUT, ':utf8';
+
+if (!defined($CGI::VERSION) || $CGI::VERSION < 4.08) {
+       eval 'sub CGI::multi_param { CGI::param(@_) }'
+}
+
+our $t0 = [ gettimeofday() ];
+our $number_of_git_cmds = 0;
+
+BEGIN {
+       CGI->compile() if $ENV{'MOD_PERL'};
+}
+
+our $version = "2.11.0";
+
+our ($my_url, $my_uri, $base_url, $path_info, $home_link);
+sub evaluate_uri {
+       our $cgi;
+
+       our $my_url = $cgi->url();
+       our $my_uri = $cgi->url(-absolute => 1);
+
+       # Base URL for relative URLs in gitweb ($logo, $favicon, ...),
+       # needed and used only for URLs with nonempty PATH_INFO
+       our $base_url = $my_url;
+
+       # When the script is used as DirectoryIndex, the URL does not contain the name
+       # of the script file itself, and $cgi->url() fails to strip PATH_INFO, so we
+       # have to do it ourselves. We make $path_info global because it's also used
+       # later on.
+       #
+       # Another issue with the script being the DirectoryIndex is that the resulting
+       # $my_url data is not the full script URL: this is good, because we want
+       # generated links to keep implying the script name if it wasn't explicitly
+       # indicated in the URL we're handling, but it means that $my_url cannot be used
+       # as base URL.
+       # Therefore, if we needed to strip PATH_INFO, then we know that we have
+       # to build the base URL ourselves:
+       our $path_info = decode_utf8($ENV{"PATH_INFO"});
+       if ($path_info) {
+               # $path_info has already been URL-decoded by the web server, but
+               # $my_url and $my_uri have not. URL-decode them so we can properly
+               # strip $path_info.
+               $my_url = unescape($my_url);
+               $my_uri = unescape($my_uri);
+               if ($my_url =~ s,\Q$path_info\E$,, &&
+                   $my_uri =~ s,\Q$path_info\E$,, &&
+                   defined $ENV{'SCRIPT_NAME'}) {
+                       $base_url = $cgi->url(-base => 1) . $ENV{'SCRIPT_NAME'};
+               }
+       }
+
+       # target of the home link on top of all pages
+       our $home_link = $my_uri || "/";
+}
+
+# core git executable to use
+# this can just be "git" if your webserver has a sensible PATH
+our $GIT = "/usr/bin/git";
+
+# absolute fs-path which will be prepended to the project path
+#our $projectroot = "/pub/scm";
+our $projectroot = "/pub/git";
+
+# fs traversing limit for getting project list
+# the number is relative to the projectroot
+our $project_maxdepth = 2007;
+
+# string of the home link on top of all pages
+our $home_link_str = "projects";
+
+# extra breadcrumbs preceding the home link
+our @extra_breadcrumbs = ();
+
+# name of your site or organization to appear in page titles
+# replace this with something more descriptive for clearer bookmarks
+our $site_name = ""
+                 || ($ENV{'SERVER_NAME'} || "Untitled") . " Git";
+
+# html snippet to include in the <head> section of each page
+our $site_html_head_string = "";
+# filename of html text to include at top of each page
+our $site_header = "";
+# html text to include at home page
+our $home_text = "indextext.html";
+# filename of html text to include at bottom of each page
+our $site_footer = "";
+
+# URI of stylesheets
+our @stylesheets = ("static/gitweb.css");
+# URI of a single stylesheet, which can be overridden in GITWEB_CONFIG.
+our $stylesheet = undef;
+# URI of GIT logo (72x27 size)
+our $logo = "static/git-logo.png";
+# URI of GIT favicon, assumed to be image/png type
+our $favicon = "static/git-favicon.png";
+# URI of gitweb.js (JavaScript code for gitweb)
+our $javascript = "static/gitweb.js";
+
+# URI and label (title) of GIT logo link
+#our $logo_url = "http://www.kernel.org/pub/software/scm/git/docs/";
+#our $logo_label = "git documentation";
+our $logo_url = "http://git-scm.com/";
+our $logo_label = "git homepage";
+
+# source of projects list
+our $projects_list = "";
+
+# the width (in characters) of the projects list "Description" column
+our $projects_list_description_width = 80;
+
+# group projects by category on the projects list
+# (enabled if this variable evaluates to true)
+our $projects_list_group_categories = 0;
+
+# default category if none specified
+# (leave the empty string for no category)
+our $project_list_default_category = "";
+
+# default order of projects list
+# valid values are none, project, descr, owner, and age
+our $default_projects_order = "project";
+
+# show repository only if this file exists
+# (only effective if this variable evaluates to true)
+our $export_ok = "";
+
+# don't generate age column on the projects list page
+our $omit_age_column = 0;
+
+# don't generate information about owners of repositories
+our $omit_owner=0;
+
+# show repository only if this subroutine returns true
+# when given the path to the project, for example:
+#    sub { return -e "$_[0]/git-daemon-export-ok"; }
+our $export_auth_hook = undef;
+
+# only allow viewing of repositories also shown on the overview page
+our $strict_export = "";
+
+# list of git base URLs used for URL to where fetch project from,
+# i.e. full URL is "$git_base_url/$project"
+our @git_base_url_list = grep { $_ ne '' } ("");
+
+# default blob_plain mimetype and default charset for text/plain blob
+our $default_blob_plain_mimetype = 'text/plain';
+our $default_text_plain_charset  = undef;
+
+# file to use for guessing MIME types before trying /etc/mime.types
+# (relative to the current git repository)
+our $mimetypes_file = undef;
+
+# assume this charset if line contains non-UTF-8 characters;
+# it should be valid encoding (see Encoding::Supported(3pm) for list),
+# for which encoding all byte sequences are valid, for example
+# 'iso-8859-1' aka 'latin1' (it is decoded without checking, so it
+# could be even 'utf-8' for the old behavior)
+our $fallback_encoding = 'latin1';
+
+# rename detection options for git-diff and git-diff-tree
+# - default is '-M', with the cost proportional to
+#   (number of removed files) * (number of new files).
+# - more costly is '-C' (which implies '-M'), with the cost proportional to
+#   (number of changed files + number of removed files) * (number of new files)
+# - even more costly is '-C', '--find-copies-harder' with cost
+#   (number of files in the original tree) * (number of new files)
+# - one might want to include '-B' option, e.g. '-B', '-M'
+our @diff_opts = ('-M'); # taken from git_commit
+
+# Disables features that would allow repository owners to inject script into
+# the gitweb domain.
+our $prevent_xss = 0;
+
+# Path to the highlight executable to use (must be the one from
+# http://www.andre-simon.de due to assumptions about parameters and output).
+# Useful if highlight is not installed on your webserver's PATH.
+# [Default: highlight]
+our $highlight_bin = "/usr/bin/highlight";
+
+# information about snapshot formats that gitweb is capable of serving
+our %known_snapshot_formats = (
+       # name => {
+       #       'display' => display name,
+       #       'type' => mime type,
+       #       'suffix' => filename suffix,
+       #       'format' => --format for git-archive,
+       #       'compressor' => [compressor command and arguments]
+       #                       (array reference, optional)
+       #       'disabled' => boolean (optional)}
+       #
+       'tgz' => {
+               'display' => 'tar.gz',
+               'type' => 'application/x-gzip',
+               'suffix' => '.tar.gz',
+               'format' => 'tar',
+               'compressor' => ['gzip', '-n']},
+
+       'tbz2' => {
+               'display' => 'tar.bz2',
+               'type' => 'application/x-bzip2',
+               'suffix' => '.tar.bz2',
+               'format' => 'tar',
+               'compressor' => ['bzip2']},
+
+       'txz' => {
+               'display' => 'tar.xz',
+               'type' => 'application/x-xz',
+               'suffix' => '.tar.xz',
+               'format' => 'tar',
+               'compressor' => ['xz'],
+               'disabled' => 1},
+
+       'zip' => {
+               'display' => 'zip',
+               'type' => 'application/x-zip',
+               'suffix' => '.zip',
+               'format' => 'zip'},
+);
+
+# Aliases so we understand old gitweb.snapshot values in repository
+# configuration.
+our %known_snapshot_format_aliases = (
+       'gzip'  => 'tgz',
+       'bzip2' => 'tbz2',
+       'xz'    => 'txz',
+
+       # backward compatibility: legacy gitweb config support
+       'x-gzip' => undef, 'gz' => undef,
+       'x-bzip2' => undef, 'bz2' => undef,
+       'x-zip' => undef, '' => undef,
+);
+
+# Pixel sizes for icons and avatars. If the default font sizes or lineheights
+# are changed, it may be appropriate to change these values too via
+# $GITWEB_CONFIG.
+our %avatar_size = (
+       'default' => 16,
+       'double'  => 32
+);
+
+# Used to set the maximum load that we will still respond to gitweb queries.
+# If server load exceed this value then return "503 server busy" error.
+# If gitweb cannot determined server load, it is taken to be 0.
+# Leave it undefined (or set to 'undef') to turn off load checking.
+our $maxload = 300;
+
+# configuration for 'highlight' (http://www.andre-simon.de/)
+# match by basename
+our %highlight_basename = (
+       #'Program' => 'py',
+       #'Library' => 'py',
+       'SConstruct' => 'py', # SCons equivalent of Makefile
+       'Makefile' => 'make',
+);
+# match by extension
+our %highlight_ext = (
+       # main extensions, defining name of syntax;
+       # see files in /usr/share/highlight/langDefs/ directory
+       (map { $_ => $_ } qw(py rb java css js tex bib xml awk bat ini spec tcl sql)),
+       # alternate extensions, see /etc/highlight/filetypes.conf
+       (map { $_ => 'c'   } qw(c h)),
+       (map { $_ => 'sh'  } qw(sh bash zsh ksh)),
+       (map { $_ => 'cpp' } qw(cpp cxx c++ cc)),
+       (map { $_ => 'php' } qw(php php3 php4 php5 phps)),
+       (map { $_ => 'pl'  } qw(pl perl pm)), # perhaps also 'cgi'
+       (map { $_ => 'make'} qw(make mak mk)),
+       (map { $_ => 'xml' } qw(xml xhtml html htm)),
+);
+
+# You define site-wide feature defaults here; override them with
+# $GITWEB_CONFIG as necessary.
+our %feature = (
+       # feature => {
+       #       'sub' => feature-sub (subroutine),
+       #       'override' => allow-override (boolean),
+       #       'default' => [ default options...] (array reference)}
+       #
+       # if feature is overridable (it means that allow-override has true value),
+       # then feature-sub will be called with default options as parameters;
+       # return value of feature-sub indicates if to enable specified feature
+       #
+       # if there is no 'sub' key (no feature-sub), then feature cannot be
+       # overridden
+       #
+       # use gitweb_get_feature(<feature>) to retrieve the <feature> value
+       # (an array) or gitweb_check_feature(<feature>) to check if <feature>
+       # is enabled
+
+       # Enable the 'blame' blob view, showing the last commit that modified
+       # each line in the file. This can be very CPU-intensive.
+
+       # To enable system wide have in $GITWEB_CONFIG
+       # $feature{'blame'}{'default'} = [1];
+       # To have project specific config enable override in $GITWEB_CONFIG
+       # $feature{'blame'}{'override'} = 1;
+       # and in project config gitweb.blame = 0|1;
+       'blame' => {
+               'sub' => sub { feature_bool('blame', @_) },
+               'override' => 0,
+               'default' => [0]},
+
+       # Enable the 'snapshot' link, providing a compressed archive of any
+       # tree. This can potentially generate high traffic if you have large
+       # project.
+
+       # Value is a list of formats defined in %known_snapshot_formats that
+       # you wish to offer.
+       # To disable system wide have in $GITWEB_CONFIG
+       # $feature{'snapshot'}{'default'} = [];
+       # To have project specific config enable override in $GITWEB_CONFIG
+       # $feature{'snapshot'}{'override'} = 1;
+       # and in project config, a comma-separated list of formats or "none"
+       # to disable.  Example: gitweb.snapshot = tbz2,zip;
+       'snapshot' => {
+               'sub' => \&feature_snapshot,
+               'override' => 0,
+               'default' => ['tgz']},
+
+       # Enable text search, which will list the commits which match author,
+       # committer or commit text to a given string.  Enabled by default.
+       # Project specific override is not supported.
+       #
+       # Note that this controls all search features, which means that if
+       # it is disabled, then 'grep' and 'pickaxe' search would also be
+       # disabled.
+       'search' => {
+               'override' => 0,
+               'default' => [1]},
+
+       # Enable grep search, which will list the files in currently selected
+       # tree containing the given string. Enabled by default. This can be
+       # potentially CPU-intensive, of course.
+       # Note that you need to have 'search' feature enabled too.
+
+       # To enable system wide have in $GITWEB_CONFIG
+       # $feature{'grep'}{'default'} = [1];
+       # To have project specific config enable override in $GITWEB_CONFIG
+       # $feature{'grep'}{'override'} = 1;
+       # and in project config gitweb.grep = 0|1;
+       'grep' => {
+               'sub' => sub { feature_bool('grep', @_) },
+               'override' => 0,
+               'default' => [1]},
+
+       # Enable the pickaxe search, which will list the commits that modified
+       # a given string in a file. This can be practical and quite faster
+       # alternative to 'blame', but still potentially CPU-intensive.
+       # Note that you need to have 'search' feature enabled too.
+
+       # To enable system wide have in $GITWEB_CONFIG
+       # $feature{'pickaxe'}{'default'} = [1];
+       # To have project specific config enable override in $GITWEB_CONFIG
+       # $feature{'pickaxe'}{'override'} = 1;
+       # and in project config gitweb.pickaxe = 0|1;
+       'pickaxe' => {
+               'sub' => sub { feature_bool('pickaxe', @_) },
+               'override' => 0,
+               'default' => [1]},
+
+       # Enable showing size of blobs in a 'tree' view, in a separate
+       # column, similar to what 'ls -l' does.  This cost a bit of IO.
+
+       # To disable system wide have in $GITWEB_CONFIG
+       # $feature{'show-sizes'}{'default'} = [0];
+       # To have project specific config enable override in $GITWEB_CONFIG
+       # $feature{'show-sizes'}{'override'} = 1;
+       # and in project config gitweb.showsizes = 0|1;
+       'show-sizes' => {
+               'sub' => sub { feature_bool('showsizes', @_) },
+               'override' => 0,
+               'default' => [1]},
+
+       # Make gitweb use an alternative format of the URLs which can be
+       # more readable and natural-looking: project name is embedded
+       # directly in the path and the query string contains other
+       # auxiliary information. All gitweb installations recognize
+       # URL in either format; this configures in which formats gitweb
+       # generates links.
+
+       # To enable system wide have in $GITWEB_CONFIG
+       # $feature{'pathinfo'}{'default'} = [1];
+       # Project specific override is not supported.
+
+       # Note that you will need to change the default location of CSS,
+       # favicon, logo and possibly other files to an absolute URL. Also,
+       # if gitweb.cgi serves as your indexfile, you will need to force
+       # $my_uri to contain the script name in your $GITWEB_CONFIG.
+       'pathinfo' => {
+               'override' => 0,
+               'default' => [0]},
+
+       # Make gitweb consider projects in project root subdirectories
+       # to be forks of existing projects. Given project $projname.git,
+       # projects matching $projname/*.git will not be shown in the main
+       # projects list, instead a '+' mark will be added to $projname
+       # there and a 'forks' view will be enabled for the project, listing
+       # all the forks. If project list is taken from a file, forks have
+       # to be listed after the main project.
+
+       # To enable system wide have in $GITWEB_CONFIG
+       # $feature{'forks'}{'default'} = [1];
+       # Project specific override is not supported.
+       'forks' => {
+               'override' => 0,
+               'default' => [0]},
+
+       # Insert custom links to the action bar of all project pages.
+       # This enables you mainly to link to third-party scripts integrating
+       # into gitweb; e.g. git-browser for graphical history representation
+       # or custom web-based repository administration interface.
+
+       # The 'default' value consists of a list of triplets in the form
+       # (label, link, position) where position is the label after which
+       # to insert the link and link is a format string where %n expands
+       # to the project name, %f to the project path within the filesystem,
+       # %h to the current hash (h gitweb parameter) and %b to the current
+       # hash base (hb gitweb parameter); %% expands to %.
+
+       # To enable system wide have in $GITWEB_CONFIG e.g.
+       # $feature{'actions'}{'default'} = [('graphiclog',
+       #       '/git-browser/by-commit.html?r=%n', 'summary')];
+       # Project specific override is not supported.
+       'actions' => {
+               'override' => 0,
+               'default' => []},
+
+       # Allow gitweb scan project content tags of project repository,
+       # and display the popular Web 2.0-ish "tag cloud" near the projects
+       # list.  Note that this is something COMPLETELY different from the
+       # normal Git tags.
+
+       # gitweb by itself can show existing tags, but it does not handle
+       # tagging itself; you need to do it externally, outside gitweb.
+       # The format is described in git_get_project_ctags() subroutine.
+       # You may want to install the HTML::TagCloud Perl module to get
+       # a pretty tag cloud instead of just a list of tags.
+
+       # To enable system wide have in $GITWEB_CONFIG
+       # $feature{'ctags'}{'default'} = [1];
+       # Project specific override is not supported.
+
+       # In the future whether ctags editing is enabled might depend
+       # on the value, but using 1 should always mean no editing of ctags.
+       'ctags' => {
+               'override' => 0,
+               'default' => [0]},
+
+       # The maximum number of patches in a patchset generated in patch
+       # view. Set this to 0 or undef to disable patch view, or to a
+       # negative number to remove any limit.
+
+       # To disable system wide have in $GITWEB_CONFIG
+       # $feature{'patches'}{'default'} = [0];
+       # To have project specific config enable override in $GITWEB_CONFIG
+       # $feature{'patches'}{'override'} = 1;
+       # and in project config gitweb.patches = 0|n;
+       # where n is the maximum number of patches allowed in a patchset.
+       'patches' => {
+               'sub' => \&feature_patches,
+               'override' => 0,
+               'default' => [16]},
+
+       # Avatar support. When this feature is enabled, views such as
+       # shortlog or commit will display an avatar associated with
+       # the email of the committer(s) and/or author(s).
+
+       # Currently available providers are gravatar and picon.
+       # If an unknown provider is specified, the feature is disabled.
+
+       # Gravatar depends on Digest::MD5.
+       # Picon currently relies on the indiana.edu database.
+
+       # To enable system wide have in $GITWEB_CONFIG
+       # $feature{'avatar'}{'default'} = ['<provider>'];
+       # where <provider> is either gravatar or picon.
+       # To have project specific config enable override in $GITWEB_CONFIG
+       # $feature{'avatar'}{'override'} = 1;
+       # and in project config gitweb.avatar = <provider>;
+       'avatar' => {
+               'sub' => \&feature_avatar,
+               'override' => 0,
+               'default' => ['']},
+
+       # Enable displaying how much time and how many git commands
+       # it took to generate and display page.  Disabled by default.
+       # Project specific override is not supported.
+       'timed' => {
+               'override' => 0,
+               'default' => [0]},
+
+       # Enable turning some links into links to actions which require
+       # JavaScript to run (like 'blame_incremental').  Not enabled by
+       # default.  Project specific override is currently not supported.
+       'javascript-actions' => {
+               'override' => 0,
+               'default' => [0]},
+
+       # Enable and configure ability to change common timezone for dates
+       # in gitweb output via JavaScript.  Enabled by default.
+       # Project specific override is not supported.
+       'javascript-timezone' => {
+               'override' => 0,
+               'default' => [
+                       'local',     # default timezone: 'utc', 'local', or '(-|+)HHMM' format,
+                                    # or undef to turn off this feature
+                       'gitweb_tz', # name of cookie where to store selected timezone
+                       'datetime',  # CSS class used to mark up dates for manipulation
+               ]},
+
+       # Syntax highlighting support. This is based on Daniel Svensson's
+       # and Sham Chukoury's work in gitweb-xmms2.git.
+       # It requires the 'highlight' program present in $PATH,
+       # and therefore is disabled by default.
+
+       # To enable system wide have in $GITWEB_CONFIG
+       # $feature{'highlight'}{'default'} = [1];
+
+       'highlight' => {
+               'sub' => sub { feature_bool('highlight', @_) },
+               'override' => 0,
+               'default' => [1]},
+
+       # Enable displaying of remote heads in the heads list
+
+       # To enable system wide have in $GITWEB_CONFIG
+       # $feature{'remote_heads'}{'default'} = [1];
+       # To have project specific config enable override in $GITWEB_CONFIG
+       # $feature{'remote_heads'}{'override'} = 1;
+       # and in project config gitweb.remoteheads = 0|1;
+       'remote_heads' => {
+               'sub' => sub { feature_bool('remote_heads', @_) },
+               'override' => 0,
+               'default' => [0]},
+
+       # Enable showing branches under other refs in addition to heads
+
+       # To set system wide extra branch refs have in $GITWEB_CONFIG
+       # $feature{'extra-branch-refs'}{'default'} = ['dirs', 'of', 'choice'];
+       # To have project specific config enable override in $GITWEB_CONFIG
+       # $feature{'extra-branch-refs'}{'override'} = 1;
+       # and in project config gitweb.extrabranchrefs = dirs of choice
+       # Every directory is separated with whitespace.
+
+       'extra-branch-refs' => {
+               'sub' => \&feature_extra_branch_refs,
+               'override' => 0,
+               'default' => []},
+);
+
+sub gitweb_get_feature {
+       my ($name) = @_;
+       return unless exists $feature{$name};
+       my ($sub, $override, @defaults) = (
+               $feature{$name}{'sub'},
+               $feature{$name}{'override'},
+               @{$feature{$name}{'default'}});
+       # project specific override is possible only if we have project
+       our $git_dir; # global variable, declared later
+       if (!$override || !defined $git_dir) {
+               return @defaults;
+       }
+       if (!defined $sub) {
+               warn "feature $name is not overridable";
+               return @defaults;
+       }
+       return $sub->(@defaults);
+}
+
+# A wrapper to check if a given feature is enabled.
+# With this, you can say
+#
+#   my $bool_feat = gitweb_check_feature('bool_feat');
+#   gitweb_check_feature('bool_feat') or somecode;
+#
+# instead of
+#
+#   my ($bool_feat) = gitweb_get_feature('bool_feat');
+#   (gitweb_get_feature('bool_feat'))[0] or somecode;
+#
+sub gitweb_check_feature {
+       return (gitweb_get_feature(@_))[0];
+}
+
+
+sub feature_bool {
+       my $key = shift;
+       my ($val) = git_get_project_config($key, '--bool');
+
+       if (!defined $val) {
+               return ($_[0]);
+       } elsif ($val eq 'true') {
+               return (1);
+       } elsif ($val eq 'false') {
+               return (0);
+       }
+}
+
+sub feature_snapshot {
+       my (@fmts) = @_;
+
+       my ($val) = git_get_project_config('snapshot');
+
+       if ($val) {
+               @fmts = ($val eq 'none' ? () : split /\s*[,\s]\s*/, $val);
+       }
+
+       return @fmts;
+}
+
+sub feature_patches {
+       my @val = (git_get_project_config('patches', '--int'));
+
+       if (@val) {
+               return @val;
+       }
+
+       return ($_[0]);
+}
+
+sub feature_avatar {
+       my @val = (git_get_project_config('avatar'));
+
+       return @val ? @val : @_;
+}
+
+sub feature_extra_branch_refs {
+       my (@branch_refs) = @_;
+       my $values = git_get_project_config('extrabranchrefs');
+
+       if ($values) {
+               $values = config_to_multi ($values);
+               @branch_refs = ();
+               foreach my $value (@{$values}) {
+                       push @branch_refs, split /\s+/, $value;
+               }
+       }
+
+       return @branch_refs;
+}
+
+# checking HEAD file with -e is fragile if the repository was
+# initialized long time ago (i.e. symlink HEAD) and was pack-ref'ed
+# and then pruned.
+sub check_head_link {
+       my ($dir) = @_;
+       my $headfile = "$dir/HEAD";
+       return ((-e $headfile) ||
+               (-l $headfile && readlink($headfile) =~ /^refs\/heads\//));
+}
+
+sub check_export_ok {
+       my ($dir) = @_;
+       return (check_head_link($dir) &&
+               (!$export_ok || -e "$dir/$export_ok") &&
+               (!$export_auth_hook || $export_auth_hook->($dir)));
+}
+
+# process alternate names for backward compatibility
+# filter out unsupported (unknown) snapshot formats
+sub filter_snapshot_fmts {
+       my @fmts = @_;
+
+       @fmts = map {
+               exists $known_snapshot_format_aliases{$_} ?
+                      $known_snapshot_format_aliases{$_} : $_} @fmts;
+       @fmts = grep {
+               exists $known_snapshot_formats{$_} &&
+               !$known_snapshot_formats{$_}{'disabled'}} @fmts;
+}
+
+sub filter_and_validate_refs {
+       my @refs = @_;
+       my %unique_refs = ();
+
+       foreach my $ref (@refs) {
+               die_error(500, "Invalid ref '$ref' in 'extra-branch-refs' feature") unless (is_valid_ref_format($ref));
+               # 'heads' are added implicitly in get_branch_refs().
+               $unique_refs{$ref} = 1 if ($ref ne 'heads');
+       }
+       return sort keys %unique_refs;
+}
+
+# If it is set to code reference, it is code that it is to be run once per
+# request, allowing updating configurations that change with each request,
+# while running other code in config file only once.
+#
+# Otherwise, if it is false then gitweb would process config file only once;
+# if it is true then gitweb config would be run for each request.
+our $per_request_config = 1;
+
+# read and parse gitweb config file given by its parameter.
+# returns true on success, false on recoverable error, allowing
+# to chain this subroutine, using first file that exists.
+# dies on errors during parsing config file, as it is unrecoverable.
+sub read_config_file {
+       my $filename = shift;
+       return unless defined $filename;
+       # die if there are errors parsing config file
+       if (-e $filename) {
+               do $filename;
+               die $@ if $@;
+               return 1;
+       }
+       return;
+}
+
+our ($GITWEB_CONFIG, $GITWEB_CONFIG_SYSTEM, $GITWEB_CONFIG_COMMON);
+sub evaluate_gitweb_config {
+       our $GITWEB_CONFIG = $ENV{'GITWEB_CONFIG'} || "gitweb_config.perl";
+       our $GITWEB_CONFIG_SYSTEM = $ENV{'GITWEB_CONFIG_SYSTEM'} || "/etc/gitweb.conf";
+       our $GITWEB_CONFIG_COMMON = $ENV{'GITWEB_CONFIG_COMMON'} || "/etc/gitweb-common.conf";
+
+       # Protect against duplications of file names, to not read config twice.
+       # Only one of $GITWEB_CONFIG and $GITWEB_CONFIG_SYSTEM is used, so
+       # there possibility of duplication of filename there doesn't matter.
+       $GITWEB_CONFIG = ""        if ($GITWEB_CONFIG eq $GITWEB_CONFIG_COMMON);
+       $GITWEB_CONFIG_SYSTEM = "" if ($GITWEB_CONFIG_SYSTEM eq $GITWEB_CONFIG_COMMON);
+
+       # Common system-wide settings for convenience.
+       # Those settings can be ovverriden by GITWEB_CONFIG or GITWEB_CONFIG_SYSTEM.
+       read_config_file($GITWEB_CONFIG_COMMON);
+
+       # Use first config file that exists.  This means use the per-instance
+       # GITWEB_CONFIG if exists, otherwise use GITWEB_SYSTEM_CONFIG.
+       read_config_file($GITWEB_CONFIG) and return;
+       read_config_file($GITWEB_CONFIG_SYSTEM);
+}
+
+# Get loadavg of system, to compare against $maxload.
+# Currently it requires '/proc/loadavg' present to get loadavg;
+# if it is not present it returns 0, which means no load checking.
+sub get_loadavg {
+       if( -e '/proc/loadavg' ){
+               open my $fd, '<', '/proc/loadavg'
+                       or return 0;
+               my @load = split(/\s+/, scalar <$fd>);
+               close $fd;
+
+               # The first three columns measure CPU and IO utilization of the last one,
+               # five, and 10 minute periods.  The fourth column shows the number of
+               # currently running processes and the total number of processes in the m/n
+               # format.  The last column displays the last process ID used.
+               return $load[0] || 0;
+       }
+       # additional checks for load average should go here for things that don't export
+       # /proc/loadavg
+
+       return 0;
+}
+
+# version of the core git binary
+our $git_version;
+sub evaluate_git_version {
+       our $git_version = qx("$GIT" --version) =~ m/git version (.*)$/ ? $1 : "unknown";
+       $number_of_git_cmds++;
+}
+
+sub check_loadavg {
+       if (defined $maxload && get_loadavg() > $maxload) {
+               die_error(503, "The load average on the server is too high");
+       }
+}
+
+# ======================================================================
+# input validation and dispatch
+
+# input parameters can be collected from a variety of sources (presently, CGI
+# and PATH_INFO), so we define an %input_params hash that collects them all
+# together during validation: this allows subsequent uses (e.g. href()) to be
+# agnostic of the parameter origin
+
+our %input_params = ();
+
+# input parameters are stored with the long parameter name as key. This will
+# also be used in the href subroutine to convert parameters to their CGI
+# equivalent, and since the href() usage is the most frequent one, we store
+# the name -> CGI key mapping here, instead of the reverse.
+#
+# XXX: Warning: If you touch this, check the search form for updating,
+# too.
+
+our @cgi_param_mapping = (
+       project => "p",
+       action => "a",
+       file_name => "f",
+       file_parent => "fp",
+       hash => "h",
+       hash_parent => "hp",
+       hash_base => "hb",
+       hash_parent_base => "hpb",
+       page => "pg",
+       order => "o",
+       searchtext => "s",
+       searchtype => "st",
+       snapshot_format => "sf",
+       extra_options => "opt",
+       search_use_regexp => "sr",
+       ctag => "by_tag",
+       diff_style => "ds",
+       project_filter => "pf",
+       # this must be last entry (for manipulation from JavaScript)
+       javascript => "js"
+);
+our %cgi_param_mapping = @cgi_param_mapping;
+
+# we will also need to know the possible actions, for validation
+our %actions = (
+       "blame" => \&git_blame,
+       "blame_incremental" => \&git_blame_incremental,
+       "blame_data" => \&git_blame_data,
+       "blobdiff" => \&git_blobdiff,
+       "blobdiff_plain" => \&git_blobdiff_plain,
+       "blob" => \&git_blob,
+       "blob_plain" => \&git_blob_plain,
+       "commitdiff" => \&git_commitdiff,
+       "commitdiff_plain" => \&git_commitdiff_plain,
+       "commit" => \&git_commit,
+       "forks" => \&git_forks,
+       "heads" => \&git_heads,
+       "history" => \&git_history,
+       "log" => \&git_log,
+       "patch" => \&git_patch,
+       "patches" => \&git_patches,
+       "remotes" => \&git_remotes,
+       "rss" => \&git_rss,
+       "atom" => \&git_atom,
+       "search" => \&git_search,
+       "search_help" => \&git_search_help,
+       "shortlog" => \&git_shortlog,
+       "summary" => \&git_summary,
+       "tag" => \&git_tag,
+       "tags" => \&git_tags,
+       "tree" => \&git_tree,
+       "snapshot" => \&git_snapshot,
+       "object" => \&git_object,
+       # those below don't need $project
+       "opml" => \&git_opml,
+       "project_list" => \&git_project_list,
+       "project_index" => \&git_project_index,
+);
+
+# finally, we have the hash of allowed extra_options for the commands that
+# allow them
+our %allowed_options = (
+       "--no-merges" => [ qw(rss atom log shortlog history) ],
+);
+
+# fill %input_params with the CGI parameters. All values except for 'opt'
+# should be single values, but opt can be an array. We should probably
+# build an array of parameters that can be multi-valued, but since for the time
+# being it's only this one, we just single it out
+sub evaluate_query_params {
+       our $cgi;
+
+       while (my ($name, $symbol) = each %cgi_param_mapping) {
+               if ($symbol eq 'opt') {
+                       $input_params{$name} = [ map { decode_utf8($_) } $cgi->multi_param($symbol) ];
+               } else {
+                       $input_params{$name} = decode_utf8($cgi->param($symbol));
+               }
+       }
+}
+
+# now read PATH_INFO and update the parameter list for missing parameters
+sub evaluate_path_info {
+       return if defined $input_params{'project'};
+       return if !$path_info;
+       $path_info =~ s,^/+,,;
+       return if !$path_info;
+
+       # find which part of PATH_INFO is project
+       my $project = $path_info;
+       $project =~ s,/+$,,;
+       while ($project && !check_head_link("$projectroot/$project")) {
+               $project =~ s,/*[^/]*$,,;
+       }
+       return unless $project;
+       $input_params{'project'} = $project;
+
+       # do not change any parameters if an action is given using the query string
+       return if $input_params{'action'};
+       $path_info =~ s,^\Q$project\E/*,,;
+
+       # next, check if we have an action
+       my $action = $path_info;
+       $action =~ s,/.*$,,;
+       if (exists $actions{$action}) {
+               $path_info =~ s,^$action/*,,;
+               $input_params{'action'} = $action;
+       }
+
+       # list of actions that want hash_base instead of hash, but can have no
+       # pathname (f) parameter
+       my @wants_base = (
+               'tree',
+               'history',
+       );
+
+       # we want to catch, among others
+       # [$hash_parent_base[:$file_parent]..]$hash_parent[:$file_name]
+       my ($parentrefname, $parentpathname, $refname, $pathname) =
+               ($path_info =~ /^(?:(.+?)(?::(.+))?\.\.)?([^:]+?)?(?::(.+))?$/);
+
+       # first, analyze the 'current' part
+       if (defined $pathname) {
+               # we got "branch:filename" or "branch:dir/"
+               # we could use git_get_type(branch:pathname), but:
+               # - it needs $git_dir
+               # - it does a git() call
+               # - the convention of terminating directories with a slash
+               #   makes it superfluous
+               # - embedding the action in the PATH_INFO would make it even
+               #   more superfluous
+               $pathname =~ s,^/+,,;
+               if (!$pathname || substr($pathname, -1) eq "/") {
+                       $input_params{'action'} ||= "tree";
+                       $pathname =~ s,/$,,;
+               } else {
+                       # the default action depends on whether we had parent info
+                       # or not
+                       if ($parentrefname) {
+                               $input_params{'action'} ||= "blobdiff_plain";
+                       } else {
+                               $input_params{'action'} ||= "blob_plain";
+                       }
+               }
+               $input_params{'hash_base'} ||= $refname;
+               $input_params{'file_name'} ||= $pathname;
+       } elsif (defined $refname) {
+               # we got "branch". In this case we have to choose if we have to
+               # set hash or hash_base.
+               #
+               # Most of the actions without a pathname only want hash to be
+               # set, except for the ones specified in @wants_base that want
+               # hash_base instead. It should also be noted that hand-crafted
+               # links having 'history' as an action and no pathname or hash
+               # set will fail, but that happens regardless of PATH_INFO.
+               if (defined $parentrefname) {
+                       # if there is parent let the default be 'shortlog' action
+                       # (for http://git.example.com/repo.git/A..B links); if there
+                       # is no parent, dispatch will detect type of object and set
+                       # action appropriately if required (if action is not set)
+                       $input_params{'action'} ||= "shortlog";
+               }
+               if ($input_params{'action'} &&
+                   grep { $_ eq $input_params{'action'} } @wants_base) {
+                       $input_params{'hash_base'} ||= $refname;
+               } else {
+                       $input_params{'hash'} ||= $refname;
+               }
+       }
+
+       # next, handle the 'parent' part, if present
+       if (defined $parentrefname) {
+               # a missing pathspec defaults to the 'current' filename, allowing e.g.
+               # someproject/blobdiff/oldrev..newrev:/filename
+               if ($parentpathname) {
+                       $parentpathname =~ s,^/+,,;
+                       $parentpathname =~ s,/$,,;
+                       $input_params{'file_parent'} ||= $parentpathname;
+               } else {
+                       $input_params{'file_parent'} ||= $input_params{'file_name'};
+               }
+               # we assume that hash_parent_base is wanted if a path was specified,
+               # or if the action wants hash_base instead of hash
+               if (defined $input_params{'file_parent'} ||
+                       grep { $_ eq $input_params{'action'} } @wants_base) {
+                       $input_params{'hash_parent_base'} ||= $parentrefname;
+               } else {
+                       $input_params{'hash_parent'} ||= $parentrefname;
+               }
+       }
+
+       # for the snapshot action, we allow URLs in the form
+       # $project/snapshot/$hash.ext
+       # where .ext determines the snapshot and gets removed from the
+       # passed $refname to provide the $hash.
+       #
+       # To be able to tell that $refname includes the format extension, we
+       # require the following two conditions to be satisfied:
+       # - the hash input parameter MUST have been set from the $refname part
+       #   of the URL (i.e. they must be equal)
+       # - the snapshot format MUST NOT have been defined already (e.g. from
+       #   CGI parameter sf)
+       # It's also useless to try any matching unless $refname has a dot,
+       # so we check for that too
+       if (defined $input_params{'action'} &&
+               $input_params{'action'} eq 'snapshot' &&
+               defined $refname && index($refname, '.') != -1 &&
+               $refname eq $input_params{'hash'} &&
+               !defined $input_params{'snapshot_format'}) {
+               # We loop over the known snapshot formats, checking for
+               # extensions. Allowed extensions are both the defined suffix
+               # (which includes the initial dot already) and the snapshot
+               # format key itself, with a prepended dot
+               while (my ($fmt, $opt) = each %known_snapshot_formats) {
+                       my $hash = $refname;
+                       unless ($hash =~ s/(\Q$opt->{'suffix'}\E|\Q.$fmt\E)$//) {
+                               next;
+                       }
+                       my $sfx = $1;
+                       # a valid suffix was found, so set the snapshot format
+                       # and reset the hash parameter
+                       $input_params{'snapshot_format'} = $fmt;
+                       $input_params{'hash'} = $hash;
+                       # we also set the format suffix to the one requested
+                       # in the URL: this way a request for e.g. .tgz returns
+                       # a .tgz instead of a .tar.gz
+                       $known_snapshot_formats{$fmt}{'suffix'} = $sfx;
+                       last;
+               }
+       }
+}
+
+our ($action, $project, $file_name, $file_parent, $hash, $hash_parent, $hash_base,
+     $hash_parent_base, @extra_options, $page, $searchtype, $search_use_regexp,
+     $searchtext, $search_regexp, $project_filter);
+sub evaluate_and_validate_params {
+       our $action = $input_params{'action'};
+       if (defined $action) {
+               if (!is_valid_action($action)) {
+                       die_error(400, "Invalid action parameter");
+               }
+       }
+
+       # parameters which are pathnames
+       our $project = $input_params{'project'};
+       if (defined $project) {
+               if (!is_valid_project($project)) {
+                       undef $project;
+                       die_error(404, "No such project");
+               }
+       }
+
+       our $project_filter = $input_params{'project_filter'};
+       if (defined $project_filter) {
+               if (!is_valid_pathname($project_filter)) {
+                       die_error(404, "Invalid project_filter parameter");
+               }
+       }
+
+       our $file_name = $input_params{'file_name'};
+       if (defined $file_name) {
+               if (!is_valid_pathname($file_name)) {
+                       die_error(400, "Invalid file parameter");
+               }
+       }
+
+       our $file_parent = $input_params{'file_parent'};
+       if (defined $file_parent) {
+               if (!is_valid_pathname($file_parent)) {
+                       die_error(400, "Invalid file parent parameter");
+               }
+       }
+
+       # parameters which are refnames
+       our $hash = $input_params{'hash'};
+       if (defined $hash) {
+               if (!is_valid_refname($hash)) {
+                       die_error(400, "Invalid hash parameter");
+               }
+       }
+
+       our $hash_parent = $input_params{'hash_parent'};
+       if (defined $hash_parent) {
+               if (!is_valid_refname($hash_parent)) {
+                       die_error(400, "Invalid hash parent parameter");
+               }
+       }
+
+       our $hash_base = $input_params{'hash_base'};
+       if (defined $hash_base) {
+               if (!is_valid_refname($hash_base)) {
+                       die_error(400, "Invalid hash base parameter");
+               }
+       }
+
+       our @extra_options = @{$input_params{'extra_options'}};
+       # @extra_options is always defined, since it can only be (currently) set from
+       # CGI, and $cgi->param() returns the empty array in array context if the param
+       # is not set
+       foreach my $opt (@extra_options) {
+               if (not exists $allowed_options{$opt}) {
+                       die_error(400, "Invalid option parameter");
+               }
+               if (not grep(/^$action$/, @{$allowed_options{$opt}})) {
+                       die_error(400, "Invalid option parameter for this action");
+               }
+       }
+
+       our $hash_parent_base = $input_params{'hash_parent_base'};
+       if (defined $hash_parent_base) {
+               if (!is_valid_refname($hash_parent_base)) {
+                       die_error(400, "Invalid hash parent base parameter");
+               }
+       }
+
+       # other parameters
+       our $page = $input_params{'page'};
+       if (defined $page) {
+               if ($page =~ m/[^0-9]/) {
+                       die_error(400, "Invalid page parameter");
+               }
+       }
+
+       our $searchtype = $input_params{'searchtype'};
+       if (defined $searchtype) {
+               if ($searchtype =~ m/[^a-z]/) {
+                       die_error(400, "Invalid searchtype parameter");
+               }
+       }
+
+       our $search_use_regexp = $input_params{'search_use_regexp'};
+
+       our $searchtext = $input_params{'searchtext'};
+       our $search_regexp = undef;
+       if (defined $searchtext) {
+               if (length($searchtext) < 2) {
+                       die_error(403, "At least two characters are required for search parameter");
+               }
+               if ($search_use_regexp) {
+                       $search_regexp = $searchtext;
+                       if (!eval { qr/$search_regexp/; 1; }) {
+                               (my $error = $@) =~ s/ at \S+ line \d+.*\n?//;
+                               die_error(400, "Invalid search regexp '$search_regexp'",
+                                         esc_html($error));
+                       }
+               } else {
+                       $search_regexp = quotemeta $searchtext;
+               }
+       }
+}
+
+# path to the current git repository
+our $git_dir;
+sub evaluate_git_dir {
+       our $git_dir = "$projectroot/$project" if $project;
+}
+
+our (@snapshot_fmts, $git_avatar, @extra_branch_refs);
+sub configure_gitweb_features {
+       # list of supported snapshot formats
+       our @snapshot_fmts = gitweb_get_feature('snapshot');
+       @snapshot_fmts = filter_snapshot_fmts(@snapshot_fmts);
+
+       # check that the avatar feature is set to a known provider name,
+       # and for each provider check if the dependencies are satisfied.
+       # if the provider name is invalid or the dependencies are not met,
+       # reset $git_avatar to the empty string.
+       our ($git_avatar) = gitweb_get_feature('avatar');
+       if ($git_avatar eq 'gravatar') {
+               $git_avatar = '' unless (eval { require Digest::MD5; 1; });
+       } elsif ($git_avatar eq 'picon') {
+               # no dependencies
+       } else {
+               $git_avatar = '';
+       }
+
+       our @extra_branch_refs = gitweb_get_feature('extra-branch-refs');
+       @extra_branch_refs = filter_and_validate_refs (@extra_branch_refs);
+}
+
+sub get_branch_refs {
+       return ('heads', @extra_branch_refs);
+}
+
+# custom error handler: 'die <message>' is Internal Server Error
+sub handle_errors_html {
+       my $msg = shift; # it is already HTML escaped
+
+       # to avoid infinite loop where error occurs in die_error,
+       # change handler to default handler, disabling handle_errors_html
+       set_message("Error occurred when inside die_error:\n$msg");
+
+       # you cannot jump out of die_error when called as error handler;
+       # the subroutine set via CGI::Carp::set_message is called _after_
+       # HTTP headers are already written, so it cannot write them itself
+       die_error(undef, undef, $msg, -error_handler => 1, -no_http_header => 1);
+}
+set_message(\&handle_errors_html);
+
+# dispatch
+sub dispatch {
+       if (!defined $action) {
+               if (defined $hash) {
+                       $action = git_get_type($hash);
+                       $action or die_error(404, "Object does not exist");
+               } elsif (defined $hash_base && defined $file_name) {
+                       $action = git_get_type("$hash_base:$file_name");
+                       $action or die_error(404, "File or directory does not exist");
+               } elsif (defined $project) {
+                       $action = 'summary';
+               } else {
+                       $action = 'project_list';
+               }
+       }
+       if (!defined($actions{$action})) {
+               die_error(400, "Unknown action");
+       }
+       if ($action !~ m/^(?:opml|project_list|project_index)$/ &&
+           !$project) {
+               die_error(400, "Project needed");
+       }
+       $actions{$action}->();
+}
+
+sub reset_timer {
+       our $t0 = [ gettimeofday() ]
+               if defined $t0;
+       our $number_of_git_cmds = 0;
+}
+
+our $first_request = 1;
+sub run_request {
+       reset_timer();
+
+       evaluate_uri();
+       if ($first_request) {
+               evaluate_gitweb_config();
+               evaluate_git_version();
+       }
+       if ($per_request_config) {
+               if (ref($per_request_config) eq 'CODE') {
+                       $per_request_config->();
+               } elsif (!$first_request) {
+                       evaluate_gitweb_config();
+               }
+       }
+       check_loadavg();
+
+       # $projectroot and $projects_list might be set in gitweb config file
+       $projects_list ||= $projectroot;
+
+       evaluate_query_params();
+       evaluate_path_info();
+       evaluate_and_validate_params();
+       evaluate_git_dir();
+
+       configure_gitweb_features();
+
+       dispatch();
+}
+
+our $is_last_request = sub { 1 };
+our ($pre_dispatch_hook, $post_dispatch_hook, $pre_listen_hook);
+our $CGI = 'CGI';
+our $cgi;
+sub configure_as_fcgi {
+       require CGI::Fast;
+       our $CGI = 'CGI::Fast';
+
+       my $request_number = 0;
+       # let each child service 100 requests
+       our $is_last_request = sub { ++$request_number > 100 };
+}
+sub evaluate_argv {
+       my $script_name = $ENV{'SCRIPT_NAME'} || $ENV{'SCRIPT_FILENAME'} || __FILE__;
+       configure_as_fcgi()
+               if $script_name =~ /\.fcgi$/;
+
+       return unless (@ARGV);
+
+       require Getopt::Long;
+       Getopt::Long::GetOptions(
+               'fastcgi|fcgi|f' => \&configure_as_fcgi,
+               'nproc|n=i' => sub {
+                       my ($arg, $val) = @_;
+                       return unless eval { require FCGI::ProcManager; 1; };
+                       my $proc_manager = FCGI::ProcManager->new({
+                               n_processes => $val,
+                       });
+                       our $pre_listen_hook    = sub { $proc_manager->pm_manage()        };
+                       our $pre_dispatch_hook  = sub { $proc_manager->pm_pre_dispatch()  };
+                       our $post_dispatch_hook = sub { $proc_manager->pm_post_dispatch() };
+               },
+       );
+}
+
+sub run {
+       evaluate_argv();
+
+       $first_request = 1;
+       $pre_listen_hook->()
+               if $pre_listen_hook;
+
+ REQUEST:
+       while ($cgi = $CGI->new()) {
+               $pre_dispatch_hook->()
+                       if $pre_dispatch_hook;
+
+               run_request();
+
+               $post_dispatch_hook->()
+                       if $post_dispatch_hook;
+               $first_request = 0;
+
+               last REQUEST if ($is_last_request->());
+       }
+
+ DONE_GITWEB:
+       1;
+}
+
+run();
+
+if (defined caller) {
+       # wrapped in a subroutine processing requests,
+       # e.g. mod_perl with ModPerl::Registry, or PSGI with Plack::App::WrapCGI
+       return;
+} else {
+       # pure CGI script, serving single request
+       exit;
+}
+
+## ======================================================================
+## action links
+
+# possible values of extra options
+# -full => 0|1      - use absolute/full URL ($my_uri/$my_url as base)
+# -replay => 1      - start from a current view (replay with modifications)
+# -path_info => 0|1 - don't use/use path_info URL (if possible)
+# -anchor => ANCHOR - add #ANCHOR to end of URL, implies -replay if used alone
+sub href {
+       my %params = @_;
+       # default is to use -absolute url() i.e. $my_uri
+       my $href = $params{-full} ? $my_url : $my_uri;
+
+       # implicit -replay, must be first of implicit params
+       $params{-replay} = 1 if (keys %params == 1 && $params{-anchor});
+
+       $params{'project'} = $project unless exists $params{'project'};
+
+       if ($params{-replay}) {
+               while (my ($name, $symbol) = each %cgi_param_mapping) {
+                       if (!exists $params{$name}) {
+                               $params{$name} = $input_params{$name};
+                       }
+               }
+       }
+
+       my $use_pathinfo = gitweb_check_feature('pathinfo');
+       if (defined $params{'project'} &&
+           (exists $params{-path_info} ? $params{-path_info} : $use_pathinfo)) {
+               # try to put as many parameters as possible in PATH_INFO:
+               #   - project name
+               #   - action
+               #   - hash_parent or hash_parent_base:/file_parent
+               #   - hash or hash_base:/filename
+               #   - the snapshot_format as an appropriate suffix
+
+               # When the script is the root DirectoryIndex for the domain,
+               # $href here would be something like http://gitweb.example.com/
+               # Thus, we strip any trailing / from $href, to spare us double
+               # slashes in the final URL
+               $href =~ s,/$,,;
+
+               # Then add the project name, if present
+               $href .= "/".esc_path_info($params{'project'});
+               delete $params{'project'};
+
+               # since we destructively absorb parameters, we keep this
+               # boolean that remembers if we're handling a snapshot
+               my $is_snapshot = $params{'action'} eq 'snapshot';
+
+               # Summary just uses the project path URL, any other action is
+               # added to the URL
+               if (defined $params{'action'}) {
+                       $href .= "/".esc_path_info($params{'action'})
+                               unless $params{'action'} eq 'summary';
+                       delete $params{'action'};
+               }
+
+               # Next, we put hash_parent_base:/file_parent..hash_base:/file_name,
+               # stripping nonexistent or useless pieces
+               $href .= "/" if ($params{'hash_base'} || $params{'hash_parent_base'}
+                       || $params{'hash_parent'} || $params{'hash'});
+               if (defined $params{'hash_base'}) {
+                       if (defined $params{'hash_parent_base'}) {
+                               $href .= esc_path_info($params{'hash_parent_base'});
+                               # skip the file_parent if it's the same as the file_name
+                               if (defined $params{'file_parent'}) {
+                                       if (defined $params{'file_name'} && $params{'file_parent'} eq $params{'file_name'}) {
+                                               delete $params{'file_parent'};
+                                       } elsif ($params{'file_parent'} !~ /\.\./) {
+                                               $href .= ":/".esc_path_info($params{'file_parent'});
+                                               delete $params{'file_parent'};
+                                       }
+                               }
+                               $href .= "..";
+                               delete $params{'hash_parent'};
+                               delete $params{'hash_parent_base'};
+                       } elsif (defined $params{'hash_parent'}) {
+                               $href .= esc_path_info($params{'hash_parent'}). "..";
+                               delete $params{'hash_parent'};
+                       }
+
+                       $href .= esc_path_info($params{'hash_base'});
+                       if (defined $params{'file_name'} && $params{'file_name'} !~ /\.\./) {
+                               $href .= ":/".esc_path_info($params{'file_name'});
+                               delete $params{'file_name'};
+                       }
+                       delete $params{'hash'};
+                       delete $params{'hash_base'};
+               } elsif (defined $params{'hash'}) {
+                       $href .= esc_path_info($params{'hash'});
+                       delete $params{'hash'};
+               }
+
+               # If the action was a snapshot, we can absorb the
+               # snapshot_format parameter too
+               if ($is_snapshot) {
+                       my $fmt = $params{'snapshot_format'};
+                       # snapshot_format should always be defined when href()
+                       # is called, but just in case some code forgets, we
+                       # fall back to the default
+                       $fmt ||= $snapshot_fmts[0];
+                       $href .= $known_snapshot_formats{$fmt}{'suffix'};
+                       delete $params{'snapshot_format'};
+               }
+       }
+
+       # now encode the parameters explicitly
+       my @result = ();
+       for (my $i = 0; $i < @cgi_param_mapping; $i += 2) {
+               my ($name, $symbol) = ($cgi_param_mapping[$i], $cgi_param_mapping[$i+1]);
+               if (defined $params{$name}) {
+                       if (ref($params{$name}) eq "ARRAY") {
+                               foreach my $par (@{$params{$name}}) {
+                                       push @result, $symbol . "=" . esc_param($par);
+                               }
+                       } else {
+                               push @result, $symbol . "=" . esc_param($params{$name});
+                       }
+               }
+       }
+       $href .= "?" . join(';', @result) if scalar @result;
+
+       # final transformation: trailing spaces must be escaped (URI-encoded)
+       $href =~ s/(\s+)$/CGI::escape($1)/e;
+
+       if ($params{-anchor}) {
+               $href .= "#".esc_param($params{-anchor});
+       }
+
+       return $href;
+}
+
+
+## ======================================================================
+## validation, quoting/unquoting and escaping
+
+sub is_valid_action {
+       my $input = shift;
+       return undef unless exists $actions{$input};
+       return 1;
+}
+
+sub is_valid_project {
+       my $input = shift;
+
+       return unless defined $input;
+       if (!is_valid_pathname($input) ||
+               !(-d "$projectroot/$input") ||
+               !check_export_ok("$projectroot/$input") ||
+               ($strict_export && !project_in_list($input))) {
+               return undef;
+       } else {
+               return 1;
+       }
+}
+
+sub is_valid_pathname {
+       my $input = shift;
+
+       return undef unless defined $input;
+       # no '.' or '..' as elements of path, i.e. no '.' or '..'
+       # at the beginning, at the end, and between slashes.
+       # also this catches doubled slashes
+       if ($input =~ m!(^|/)(|\.|\.\.)(/|$)!) {
+               return undef;
+       }
+       # no null characters
+       if ($input =~ m!\0!) {
+               return undef;
+       }
+       return 1;
+}
+
+sub is_valid_ref_format {
+       my $input = shift;
+
+       return undef unless defined $input;
+       # restrictions on ref name according to git-check-ref-format
+       if ($input =~ m!(/\.|\.\.|[\000-\040\177 ~^:?*\[]|/$)!) {
+               return undef;
+       }
+       return 1;
+}
+
+sub is_valid_refname {
+       my $input = shift;
+
+       return undef unless defined $input;
+       # textual hashes are O.K.
+       if ($input =~ m/^[0-9a-fA-F]{40}$/) {
+               return 1;
+       }
+       # it must be correct pathname
+       is_valid_pathname($input) or return undef;
+       # check git-check-ref-format restrictions
+       is_valid_ref_format($input) or return undef;
+       return 1;
+}
+
+# decode sequences of octets in utf8 into Perl's internal form,
+# which is utf-8 with utf8 flag set if needed.  gitweb writes out
+# in utf-8 thanks to "binmode STDOUT, ':utf8'" at beginning
+sub to_utf8 {
+       my $str = shift;
+       return undef unless defined $str;
+
+       if (utf8::is_utf8($str) || utf8::decode($str)) {
+               return $str;
+       } else {
+               return decode($fallback_encoding, $str, Encode::FB_DEFAULT);
+       }
+}
+
+# quote unsafe chars, but keep the slash, even when it's not
+# correct, but quoted slashes look too horrible in bookmarks
+sub esc_param {
+       my $str = shift;
+       return undef unless defined $str;
+       $str =~ s/([^A-Za-z0-9\-_.~()\/:@ ]+)/CGI::escape($1)/eg;
+       $str =~ s/ /\+/g;
+       return $str;
+}
+
+# the quoting rules for path_info fragment are slightly different
+sub esc_path_info {
+       my $str = shift;
+       return undef unless defined $str;
+
+       # path_info doesn't treat '+' as space (specially), but '?' must be escaped
+       $str =~ s/([^A-Za-z0-9\-_.~();\/;:@&= +]+)/CGI::escape($1)/eg;
+
+       return $str;
+}
+
+# quote unsafe chars in whole URL, so some characters cannot be quoted
+sub esc_url {
+       my $str = shift;
+       return undef unless defined $str;
+       $str =~ s/([^A-Za-z0-9\-_.~();\/;?:@&= ]+)/CGI::escape($1)/eg;
+       $str =~ s/ /\+/g;
+       return $str;
+}
+
+# quote unsafe characters in HTML attributes
+sub esc_attr {
+
+       # for XHTML conformance escaping '"' to '&quot;' is not enough
+       return esc_html(@_);
+}
+
+# replace invalid utf8 character with SUBSTITUTION sequence
+sub esc_html {
+       my $str = shift;
+       my %opts = @_;
+
+       return undef unless defined $str;
+
+       $str = to_utf8($str);
+       $str = $cgi->escapeHTML($str);
+       if ($opts{'-nbsp'}) {
+               $str =~ s/ /&nbsp;/g;
+       }
+       $str =~ s|([[:cntrl:]])|(($1 ne "\t") ? quot_cec($1) : $1)|eg;
+       return $str;
+}
+
+# quote control characters and escape filename to HTML
+sub esc_path {
+       my $str = shift;
+       my %opts = @_;
+
+       return undef unless defined $str;
+
+       $str = to_utf8($str);
+       $str = $cgi->escapeHTML($str);
+       if ($opts{'-nbsp'}) {
+               $str =~ s/ /&nbsp;/g;
+       }
+       $str =~ s|([[:cntrl:]])|quot_cec($1)|eg;
+       return $str;
+}
+
+# Sanitize for use in XHTML + application/xml+xhtml (valid XML 1.0)
+sub sanitize {
+       my $str = shift;
+
+       return undef unless defined $str;
+
+       $str = to_utf8($str);
+       $str =~ s|([[:cntrl:]])|(index("\t\n\r", $1) != -1 ? $1 : quot_cec($1))|eg;
+       return $str;
+}
+
+# Make control characters "printable", using character escape codes (CEC)
+sub quot_cec {
+       my $cntrl = shift;
+       my %opts = @_;
+       my %es = ( # character escape codes, aka escape sequences
+               "\t" => '\t',   # tab            (HT)
+               "\n" => '\n',   # line feed      (LF)
+               "\r" => '\r',   # carrige return (CR)
+               "\f" => '\f',   # form feed      (FF)
+               "\b" => '\b',   # backspace      (BS)
+               "\a" => '\a',   # alarm (bell)   (BEL)
+               "\e" => '\e',   # escape         (ESC)
+               "\013" => '\v', # vertical tab   (VT)
+               "\000" => '\0', # nul character  (NUL)
+       );
+       my $chr = ( (exists $es{$cntrl})
+                   ? $es{$cntrl}
+                   : sprintf('\%2x', ord($cntrl)) );
+       if ($opts{-nohtml}) {
+               return $chr;
+       } else {
+               return "<span class=\"cntrl\">$chr</span>";
+       }
+}
+
+# Alternatively use unicode control pictures codepoints,
+# Unicode "printable representation" (PR)
+sub quot_upr {
+       my $cntrl = shift;
+       my %opts = @_;
+
+       my $chr = sprintf('&#%04d;', 0x2400+ord($cntrl));
+       if ($opts{-nohtml}) {
+               return $chr;
+       } else {
+               return "<span class=\"cntrl\">$chr</span>";
+       }
+}
+
+# git may return quoted and escaped filenames
+sub unquote {
+       my $str = shift;
+
+       sub unq {
+               my $seq = shift;
+               my %es = ( # character escape codes, aka escape sequences
+                       't' => "\t",   # tab            (HT, TAB)
+                       'n' => "\n",   # newline        (NL)
+                       'r' => "\r",   # return         (CR)
+                       'f' => "\f",   # form feed      (FF)
+                       'b' => "\b",   # backspace      (BS)
+                       'a' => "\a",   # alarm (bell)   (BEL)
+                       'e' => "\e",   # escape         (ESC)
+                       'v' => "\013", # vertical tab   (VT)
+               );
+
+               if ($seq =~ m/^[0-7]{1,3}$/) {
+                       # octal char sequence
+                       return chr(oct($seq));
+               } elsif (exists $es{$seq}) {
+                       # C escape sequence, aka character escape code
+                       return $es{$seq};
+               }
+               # quoted ordinary character
+               return $seq;
+       }
+
+       if ($str =~ m/^"(.*)"$/) {
+               # needs unquoting
+               $str = $1;
+               $str =~ s/\\([^0-7]|[0-7]{1,3})/unq($1)/eg;
+       }
+       return $str;
+}
+
+# escape tabs (convert tabs to spaces)
+sub untabify {
+       my $line = shift;
+
+       while ((my $pos = index($line, "\t")) != -1) {
+               if (my $count = (8 - ($pos % 8))) {
+                       my $spaces = ' ' x $count;
+                       $line =~ s/\t/$spaces/;
+               }
+       }
+
+       return $line;
+}
+
+sub project_in_list {
+       my $project = shift;
+       my @list = git_get_projects_list();
+       return @list && scalar(grep { $_->{'path'} eq $project } @list);
+}
+
+## ----------------------------------------------------------------------
+## HTML aware string manipulation
+
+# Try to chop given string on a word boundary between position
+# $len and $len+$add_len. If there is no word boundary there,
+# chop at $len+$add_len. Do not chop if chopped part plus ellipsis
+# (marking chopped part) would be longer than given string.
+sub chop_str {
+       my $str = shift;
+       my $len = shift;
+       my $add_len = shift || 10;
+       my $where = shift || 'right'; # 'left' | 'center' | 'right'
+
+       # Make sure perl knows it is utf8 encoded so we don't
+       # cut in the middle of a utf8 multibyte char.
+       $str = to_utf8($str);
+
+       # allow only $len chars, but don't cut a word if it would fit in $add_len
+       # if it doesn't fit, cut it if it's still longer than the dots we would add
+       # remove chopped character entities entirely
+
+       # when chopping in the middle, distribute $len into left and right part
+       # return early if chopping wouldn't make string shorter
+       if ($where eq 'center') {
+               return $str if ($len + 5 >= length($str)); # filler is length 5
+               $len = int($len/2);
+       } else {
+               return $str if ($len + 4 >= length($str)); # filler is length 4
+       }
+
+       # regexps: ending and beginning with word part up to $add_len
+       my $endre = qr/.{$len}\w{0,$add_len}/;
+       my $begre = qr/\w{0,$add_len}.{$len}/;
+
+       if ($where eq 'left') {
+               $str =~ m/^(.*?)($begre)$/;
+               my ($lead, $body) = ($1, $2);
+               if (length($lead) > 4) {
+                       $lead = " ...";
+               }
+               return "$lead$body";
+
+       } elsif ($where eq 'center') {
+               $str =~ m/^($endre)(.*)$/;
+               my ($left, $str)  = ($1, $2);
+               $str =~ m/^(.*?)($begre)$/;
+               my ($mid, $right) = ($1, $2);
+               if (length($mid) > 5) {
+                       $mid = " ... ";
+               }
+               return "$left$mid$right";
+
+       } else {
+               $str =~ m/^($endre)(.*)$/;
+               my $body = $1;
+               my $tail = $2;
+               if (length($tail) > 4) {
+                       $tail = "... ";
+               }
+               return "$body$tail";
+       }
+}
+
+# takes the same arguments as chop_str, but also wraps a <span> around the
+# result with a title attribute if it does get chopped. Additionally, the
+# string is HTML-escaped.
+sub chop_and_escape_str {
+       my ($str) = @_;
+
+       my $chopped = chop_str(@_);
+       $str = to_utf8($str);
+       if ($chopped eq $str) {
+               return esc_html($chopped);
+       } else {
+               $str =~ s/[[:cntrl:]]/?/g;
+               return $cgi->span({-title=>$str}, esc_html($chopped));
+       }
+}
+
+# Highlight selected fragments of string, using given CSS class,
+# and escape HTML.  It is assumed that fragments do not overlap.
+# Regions are passed as list of pairs (array references).
+#
+# Example: esc_html_hl_regions("foobar", "mark", [ 0, 3 ]) returns
+# '<span class="mark">foo</span>bar'
+sub esc_html_hl_regions {
+       my ($str, $css_class, @sel) = @_;
+       my %opts = grep { ref($_) ne 'ARRAY' } @sel;
+       @sel     = grep { ref($_) eq 'ARRAY' } @sel;
+       return esc_html($str, %opts) unless @sel;
+
+       my $out = '';
+       my $pos = 0;
+
+       for my $s (@sel) {
+               my ($begin, $end) = @$s;
+
+               # Don't create empty <span> elements.
+               next if $end <= $begin;
+
+               my $escaped = esc_html(substr($str, $begin, $end - $begin),
+                                      %opts);
+
+               $out .= esc_html(substr($str, $pos, $begin - $pos), %opts)
+                       if ($begin - $pos > 0);
+               $out .= $cgi->span({-class => $css_class}, $escaped);
+
+               $pos = $end;
+       }
+       $out .= esc_html(substr($str, $pos), %opts)
+               if ($pos < length($str));
+
+       return $out;
+}
+
+# return positions of beginning and end of each match
+sub matchpos_list {
+       my ($str, $regexp) = @_;
+       return unless (defined $str && defined $regexp);
+
+       my @matches;
+       while ($str =~ /$regexp/g) {
+               push @matches, [$-[0], $+[0]];
+       }
+       return @matches;
+}
+
+# highlight match (if any), and escape HTML
+sub esc_html_match_hl {
+       my ($str, $regexp) = @_;
+       return esc_html($str) unless defined $regexp;
+
+       my @matches = matchpos_list($str, $regexp);
+       return esc_html($str) unless @matches;
+
+       return esc_html_hl_regions($str, 'match', @matches);
+}
+
+
+# highlight match (if any) of shortened string, and escape HTML
+sub esc_html_match_hl_chopped {
+       my ($str, $chopped, $regexp) = @_;
+       return esc_html_match_hl($str, $regexp) unless defined $chopped;
+
+       my @matches = matchpos_list($str, $regexp);
+       return esc_html($chopped) unless @matches;
+
+       # filter matches so that we mark chopped string
+       my $tail = "... "; # see chop_str
+       unless ($chopped =~ s/\Q$tail\E$//) {
+               $tail = '';
+       }
+       my $chop_len = length($chopped);
+       my $tail_len = length($tail);
+       my @filtered;
+
+       for my $m (@matches) {
+               if ($m->[0] > $chop_len) {
+                       push @filtered, [ $chop_len, $chop_len + $tail_len ] if ($tail_len > 0);
+                       last;
+               } elsif ($m->[1] > $chop_len) {
+                       push @filtered, [ $m->[0], $chop_len + $tail_len ];
+                       last;
+               }
+               push @filtered, $m;
+       }
+
+       return esc_html_hl_regions($chopped . $tail, 'match', @filtered);
+}
+
+## ----------------------------------------------------------------------
+## functions returning short strings
+
+# CSS class for given age value (in seconds)
+sub age_class {
+       my $age = shift;
+
+       if (!defined $age) {
+               return "noage";
+       } elsif ($age < 60*60*2) {
+               return "age0";
+       } elsif ($age < 60*60*24*2) {
+               return "age1";
+       } else {
+               return "age2";
+       }
+}
+
+# convert age in seconds to "nn units ago" string
+sub age_string {
+       my $age = shift;
+       my $age_str;
+
+       if ($age > 60*60*24*365*2) {
+               $age_str = (int $age/60/60/24/365);
+               $age_str .= " years ago";
+       } elsif ($age > 60*60*24*(365/12)*2) {
+               $age_str = int $age/60/60/24/(365/12);
+               $age_str .= " months ago";
+       } elsif ($age > 60*60*24*7*2) {
+               $age_str = int $age/60/60/24/7;
+               $age_str .= " weeks ago";
+       } elsif ($age > 60*60*24*2) {
+               $age_str = int $age/60/60/24;
+               $age_str .= " days ago";
+       } elsif ($age > 60*60*2) {
+               $age_str = int $age/60/60;
+               $age_str .= " hours ago";
+       } elsif ($age > 60*2) {
+               $age_str = int $age/60;
+               $age_str .= " min ago";
+       } elsif ($age > 2) {
+               $age_str = int $age;
+               $age_str .= " sec ago";
+       } else {
+               $age_str .= " right now";
+       }
+       return $age_str;
+}
+
+use constant {
+       S_IFINVALID => 0030000,
+       S_IFGITLINK => 0160000,
+};
+
+# submodule/subproject, a commit object reference
+sub S_ISGITLINK {
+       my $mode = shift;
+
+       return (($mode & S_IFMT) == S_IFGITLINK)
+}
+
+# convert file mode in octal to symbolic file mode string
+sub mode_str {
+       my $mode = oct shift;
+
+       if (S_ISGITLINK($mode)) {
+               return 'm---------';
+       } elsif (S_ISDIR($mode & S_IFMT)) {
+               return 'drwxr-xr-x';
+       } elsif (S_ISLNK($mode)) {
+               return 'lrwxrwxrwx';
+       } elsif (S_ISREG($mode)) {
+               # git cares only about the executable bit
+               if ($mode & S_IXUSR) {
+                       return '-rwxr-xr-x';
+               } else {
+                       return '-rw-r--r--';
+               };
+       } else {
+               return '----------';
+       }
+}
+
+# convert file mode in octal to file type string
+sub file_type {
+       my $mode = shift;
+
+       if ($mode !~ m/^[0-7]+$/) {
+               return $mode;
+       } else {
+               $mode = oct $mode;
+       }
+
+       if (S_ISGITLINK($mode)) {
+               return "submodule";
+       } elsif (S_ISDIR($mode & S_IFMT)) {
+               return "directory";
+       } elsif (S_ISLNK($mode)) {
+               return "symlink";
+       } elsif (S_ISREG($mode)) {
+               return "file";
+       } else {
+               return "unknown";
+       }
+}
+
+# convert file mode in octal to file type description string
+sub file_type_long {
+       my $mode = shift;
+
+       if ($mode !~ m/^[0-7]+$/) {
+               return $mode;
+       } else {
+               $mode = oct $mode;
+       }
+
+       if (S_ISGITLINK($mode)) {
+               return "submodule";
+       } elsif (S_ISDIR($mode & S_IFMT)) {
+               return "directory";
+       } elsif (S_ISLNK($mode)) {
+               return "symlink";
+       } elsif (S_ISREG($mode)) {
+               if ($mode & S_IXUSR) {
+                       return "executable";
+               } else {
+                       return "file";
+               };
+       } else {
+               return "unknown";
+       }
+}
+
+
+## ----------------------------------------------------------------------
+## functions returning short HTML fragments, or transforming HTML fragments
+## which don't belong to other sections
+
+# format line of commit message.
+sub format_log_line_html {
+       my $line = shift;
+
+       $line = esc_html($line, -nbsp=>1);
+       $line =~ s{
+        \b
+        (
+            # The output of "git describe", e.g. v2.10.0-297-gf6727b0
+            # or hadoop-20160921-113441-20-g094fb7d
+            (?<!-) # see strbuf_check_tag_ref(). Tags can't start with -
+            [A-Za-z0-9.-]+
+            (?!\.) # refs can't end with ".", see check_refname_format()
+            -g[0-9a-fA-F]{7,40}
+            |
+            # Just a normal looking Git SHA1
+            [0-9a-fA-F]{7,40}
+        )
+        \b
+    }{
+               $cgi->a({-href => href(action=>"object", hash=>$1),
+                                       -class => "text"}, $1);
+       }egx;
+
+       return $line;
+}
+
+# format marker of refs pointing to given object
+
+# the destination action is chosen based on object type and current context:
+# - for annotated tags, we choose the tag view unless it's the current view
+#   already, in which case we go to shortlog view
+# - for other refs, we keep the current view if we're in history, shortlog or
+#   log view, and select shortlog otherwise
+sub format_ref_marker {
+       my ($refs, $id) = @_;
+       my $markers = '';
+
+       if (defined $refs->{$id}) {
+               foreach my $ref (@{$refs->{$id}}) {
+                       # this code exploits the fact that non-lightweight tags are the
+                       # only indirect objects, and that they are the only objects for which
+                       # we want to use tag instead of shortlog as action
+                       my ($type, $name) = qw();
+                       my $indirect = ($ref =~ s/\^\{\}$//);
+                       # e.g. tags/v2.6.11 or heads/next
+                       if ($ref =~ m!^(.*?)s?/(.*)$!) {
+                               $type = $1;
+                               $name = $2;
+                       } else {
+                               $type = "ref";
+                               $name = $ref;
+                       }
+
+                       my $class = $type;
+                       $class .= " indirect" if $indirect;
+
+                       my $dest_action = "shortlog";
+
+                       if ($indirect) {
+                               $dest_action = "tag" unless $action eq "tag";
+                       } elsif ($action =~ /^(history|(short)?log)$/) {
+                               $dest_action = $action;
+                       }
+
+                       my $dest = "";
+                       $dest .= "refs/" unless $ref =~ m!^refs/!;
+                       $dest .= $ref;
+
+                       my $link = $cgi->a({
+                               -href => href(
+                                       action=>$dest_action,
+                                       hash=>$dest
+                               )}, esc_html($name));
+
+                       $markers .= " <span class=\"".esc_attr($class)."\" title=\"".esc_attr($ref)."\">" .
+                               $link . "</span>";
+               }
+       }
+
+       if ($markers) {
+               return ' <span class="refs">'. $markers . '</span>';
+       } else {
+               return "";
+       }
+}
+
+# format, perhaps shortened and with markers, title line
+sub format_subject_html {
+       my ($long, $short, $href, $extra) = @_;
+       $extra = '' unless defined($extra);
+
+       if (length($short) < length($long)) {
+               $long =~ s/[[:cntrl:]]/?/g;
+               return $cgi->a({-href => $href, -class => "list subject",
+                               -title => to_utf8($long)},
+                      esc_html($short)) . $extra;
+       } else {
+               return $cgi->a({-href => $href, -class => "list subject"},
+                      esc_html($long)) . $extra;
+       }
+}
+
+# Rather than recomputing the url for an email multiple times, we cache it
+# after the first hit. This gives a visible benefit in views where the avatar
+# for the same email is used repeatedly (e.g. shortlog).
+# The cache is shared by all avatar engines (currently gravatar only), which
+# are free to use it as preferred. Since only one avatar engine is used for any
+# given page, there's no risk for cache conflicts.
+our %avatar_cache = ();
+
+# Compute the picon url for a given email, by using the picon search service over at
+# http://www.cs.indiana.edu/picons/search.html
+sub picon_url {
+       my $email = lc shift;
+       if (!$avatar_cache{$email}) {
+               my ($user, $domain) = split('@', $email);
+               $avatar_cache{$email} =
+                       "//www.cs.indiana.edu/cgi-pub/kinzler/piconsearch.cgi/" .
+                       "$domain/$user/" .
+                       "users+domains+unknown/up/single";
+       }
+       return $avatar_cache{$email};
+}
+
+# Compute the gravatar url for a given email, if it's not in the cache already.
+# Gravatar stores only the part of the URL before the size, since that's the
+# one computationally more expensive. This also allows reuse of the cache for
+# different sizes (for this particular engine).
+sub gravatar_url {
+       my $email = lc shift;
+       my $size = shift;
+       $avatar_cache{$email} ||=
+               "//www.gravatar.com/avatar/" .
+                       Digest::MD5::md5_hex($email) . "?s=";
+       return $avatar_cache{$email} . $size;
+}
+
+# Insert an avatar for the given $email at the given $size if the feature
+# is enabled.
+sub git_get_avatar {
+       my ($email, %opts) = @_;
+       my $pre_white  = ($opts{-pad_before} ? "&nbsp;" : "");
+       my $post_white = ($opts{-pad_after}  ? "&nbsp;" : "");
+       $opts{-size} ||= 'default';
+       my $size = $avatar_size{$opts{-size}} || $avatar_size{'default'};
+       my $url = "";
+       if ($git_avatar eq 'gravatar') {
+               $url = gravatar_url($email, $size);
+       } elsif ($git_avatar eq 'picon') {
+               $url = picon_url($email);
+       }
+       # Other providers can be added by extending the if chain, defining $url
+       # as needed. If no variant puts something in $url, we assume avatars
+       # are completely disabled/unavailable.
+       if ($url) {
+               return $pre_white .
+                      "<img width=\"$size\" " .
+                           "class=\"avatar\" " .
+                           "src=\"".esc_url($url)."\" " .
+                           "alt=\"\" " .
+                      "/>" . $post_white;
+       } else {
+               return "";
+       }
+}
+
+sub format_search_author {
+       my ($author, $searchtype, $displaytext) = @_;
+       my $have_search = gitweb_check_feature('search');
+
+       if ($have_search) {
+               my $performed = "";
+               if ($searchtype eq 'author') {
+                       $performed = "authored";
+               } elsif ($searchtype eq 'committer') {
+                       $performed = "committed";
+               }
+
+               return $cgi->a({-href => href(action=>"search", hash=>$hash,
+                               searchtext=>$author,
+                               searchtype=>$searchtype), class=>"list",
+                               title=>"Search for commits $performed by $author"},
+                               $displaytext);
+
+       } else {
+               return $displaytext;
+       }
+}
+
+# format the author name of the given commit with the given tag
+# the author name is chopped and escaped according to the other
+# optional parameters (see chop_str).
+sub format_author_html {
+       my $tag = shift;
+       my $co = shift;
+       my $author = chop_and_escape_str($co->{'author_name'}, @_);
+       return "<$tag class=\"author\">" .
+              format_search_author($co->{'author_name'}, "author",
+                      git_get_avatar($co->{'author_email'}, -pad_after => 1) .
+                      $author) .
+              "</$tag>";
+}
+
+# format git diff header line, i.e. "diff --(git|combined|cc) ..."
+sub format_git_diff_header_line {
+       my $line = shift;
+       my $diffinfo = shift;
+       my ($from, $to) = @_;
+
+       if ($diffinfo->{'nparents'}) {
+               # combined diff
+               $line =~ s!^(diff (.*?) )"?.*$!$1!;
+               if ($to->{'href'}) {
+                       $line .= $cgi->a({-href => $to->{'href'}, -class => "path"},
+                                        esc_path($to->{'file'}));
+               } else { # file was deleted (no href)
+                       $line .= esc_path($to->{'file'});
+               }
+       } else {
+               # "ordinary" diff
+               $line =~ s!^(diff (.*?) )"?a/.*$!$1!;
+               if ($from->{'href'}) {
+                       $line .= $cgi->a({-href => $from->{'href'}, -class => "path"},
+                                        'a/' . esc_path($from->{'file'}));
+               } else { # file was added (no href)
+                       $line .= 'a/' . esc_path($from->{'file'});
+               }
+               $line .= ' ';
+               if ($to->{'href'}) {
+                       $line .= $cgi->a({-href => $to->{'href'}, -class => "path"},
+                                        'b/' . esc_path($to->{'file'}));
+               } else { # file was deleted
+                       $line .= 'b/' . esc_path($to->{'file'});
+               }
+       }
+
+       return "<div class=\"diff header\">$line</div>\n";
+}
+
+# format extended diff header line, before patch itself
+sub format_extended_diff_header_line {
+       my $line = shift;
+       my $diffinfo = shift;
+       my ($from, $to) = @_;
+
+       # match <path>
+       if ($line =~ s!^((copy|rename) from ).*$!$1! && $from->{'href'}) {
+               $line .= $cgi->a({-href=>$from->{'href'}, -class=>"path"},
+                                      esc_path($from->{'file'}));
+       }
+       if ($line =~ s!^((copy|rename) to ).*$!$1! && $to->{'href'}) {
+               $line .= $cgi->a({-href=>$to->{'href'}, -class=>"path"},
+                                esc_path($to->{'file'}));
+       }
+       # match single <mode>
+       if ($line =~ m/\s(\d{6})$/) {
+               $line .= '<span class="info"> (' .
+                        file_type_long($1) .
+                        ')</span>';
+       }
+       # match <hash>
+       if ($line =~ m/^index [0-9a-fA-F]{40},[0-9a-fA-F]{40}/) {
+               # can match only for combined diff
+               $line = 'index ';
+               for (my $i = 0; $i < $diffinfo->{'nparents'}; $i++) {
+                       if ($from->{'href'}[$i]) {
+                               $line .= $cgi->a({-href=>$from->{'href'}[$i],
+                                                 -class=>"hash"},
+                                                substr($diffinfo->{'from_id'}[$i],0,7));
+                       } else {
+                               $line .= '0' x 7;
+                       }
+                       # separator
+                       $line .= ',' if ($i < $diffinfo->{'nparents'} - 1);
+               }
+               $line .= '..';
+               if ($to->{'href'}) {
+                       $line .= $cgi->a({-href=>$to->{'href'}, -class=>"hash"},
+                                        substr($diffinfo->{'to_id'},0,7));
+               } else {
+                       $line .= '0' x 7;
+               }
+
+       } elsif ($line =~ m/^index [0-9a-fA-F]{40}..[0-9a-fA-F]{40}/) {
+               # can match only for ordinary diff
+               my ($from_link, $to_link);
+               if ($from->{'href'}) {
+                       $from_link = $cgi->a({-href=>$from->{'href'}, -class=>"hash"},
+                                            substr($diffinfo->{'from_id'},0,7));
+               } else {
+                       $from_link = '0' x 7;
+               }
+               if ($to->{'href'}) {
+                       $to_link = $cgi->a({-href=>$to->{'href'}, -class=>"hash"},
+                                          substr($diffinfo->{'to_id'},0,7));
+               } else {
+                       $to_link = '0' x 7;
+               }
+               my ($from_id, $to_id) = ($diffinfo->{'from_id'}, $diffinfo->{'to_id'});
+               $line =~ s!$from_id\.\.$to_id!$from_link..$to_link!;
+       }
+
+       return $line . "<br/>\n";
+}
+
+# format from-file/to-file diff header
+sub format_diff_from_to_header {
+       my ($from_line, $to_line, $diffinfo, $from, $to, @parents) = @_;
+       my $line;
+       my $result = '';
+
+       $line = $from_line;
+       #assert($line =~ m/^---/) if DEBUG;
+       # no extra formatting for "^--- /dev/null"
+       if (! $diffinfo->{'nparents'}) {
+               # ordinary (single parent) diff
+               if ($line =~ m!^--- "?a/!) {
+                       if ($from->{'href'}) {
+                               $line = '--- a/' .
+                                       $cgi->a({-href=>$from->{'href'}, -class=>"path"},
+                                               esc_path($from->{'file'}));
+                       } else {
+                               $line = '--- a/' .
+                                       esc_path($from->{'file'});
+                       }
+               }
+               $result .= qq!<div class="diff from_file">$line</div>\n!;
+
+       } else {
+               # combined diff (merge commit)
+               for (my $i = 0; $i < $diffinfo->{'nparents'}; $i++) {
+                       if ($from->{'href'}[$i]) {
+                               $line = '--- ' .
+                                       $cgi->a({-href=>href(action=>"blobdiff",
+                                                            hash_parent=>$diffinfo->{'from_id'}[$i],
+                                                            hash_parent_base=>$parents[$i],
+                                                            file_parent=>$from->{'file'}[$i],
+                                                            hash=>$diffinfo->{'to_id'},
+                                                            hash_base=>$hash,
+                                                            file_name=>$to->{'file'}),
+                                                -class=>"path",
+                                                -title=>"diff" . ($i+1)},
+                                               $i+1) .
+                                       '/' .
+                                       $cgi->a({-href=>$from->{'href'}[$i], -class=>"path"},
+                                               esc_path($from->{'file'}[$i]));
+                       } else {
+                               $line = '--- /dev/null';
+                       }
+                       $result .= qq!<div class="diff from_file">$line</div>\n!;
+               }
+       }
+
+       $line = $to_line;
+       #assert($line =~ m/^\+\+\+/) if DEBUG;
+       # no extra formatting for "^+++ /dev/null"
+       if ($line =~ m!^\+\+\+ "?b/!) {
+               if ($to->{'href'}) {
+                       $line = '+++ b/' .
+                               $cgi->a({-href=>$to->{'href'}, -class=>"path"},
+                                       esc_path($to->{'file'}));
+               } else {
+                       $line = '+++ b/' .
+                               esc_path($to->{'file'});
+               }
+       }
+       $result .= qq!<div class="diff to_file">$line</div>\n!;
+
+       return $result;
+}
+
+# create note for patch simplified by combined diff
+sub format_diff_cc_simplified {
+       my ($diffinfo, @parents) = @_;
+       my $result = '';
+
+       $result .= "<div class=\"diff header\">" .
+                  "diff --cc ";
+       if (!is_deleted($diffinfo)) {
+               $result .= $cgi->a({-href => href(action=>"blob",
+                                                 hash_base=>$hash,
+                                                 hash=>$diffinfo->{'to_id'},
+                                                 file_name=>$diffinfo->{'to_file'}),
+                                   -class => "path"},
+                                  esc_path($diffinfo->{'to_file'}));
+       } else {
+               $result .= esc_path($diffinfo->{'to_file'});
+       }
+       $result .= "</div>\n" . # class="diff header"
+                  "<div class=\"diff nodifferences\">" .
+                  "Simple merge" .
+                  "</div>\n"; # class="diff nodifferences"
+
+       return $result;
+}
+
+sub diff_line_class {
+       my ($line, $from, $to) = @_;
+
+       # ordinary diff
+       my $num_sign = 1;
+       # combined diff
+       if ($from && $to && ref($from->{'href'}) eq "ARRAY") {
+               $num_sign = scalar @{$from->{'href'}};
+       }
+
+       my @diff_line_classifier = (
+               { regexp => qr/^\@\@{$num_sign} /, class => "chunk_header"},
+               { regexp => qr/^\\/,               class => "incomplete"  },
+               { regexp => qr/^ {$num_sign}/,     class => "ctx" },
+               # classifier for context must come before classifier add/rem,
+               # or we would have to use more complicated regexp, for example
+               # qr/(?= {0,$m}\+)[+ ]{$num_sign}/, where $m = $num_sign - 1;
+               { regexp => qr/^[+ ]{$num_sign}/,   class => "add" },
+               { regexp => qr/^[- ]{$num_sign}/,   class => "rem" },
+       );
+       for my $clsfy (@diff_line_classifier) {
+               return $clsfy->{'class'}
+                       if ($line =~ $clsfy->{'regexp'});
+       }
+
+       # fallback
+       return "";
+}
+
+# assumes that $from and $to are defined and correctly filled,
+# and that $line holds a line of chunk header for unified diff
+sub format_unidiff_chunk_header {
+       my ($line, $from, $to) = @_;
+
+       my ($from_text, $from_start, $from_lines, $to_text, $to_start, $to_lines, $section) =
+               $line =~ m/^\@{2} (-(\d+)(?:,(\d+))?) (\+(\d+)(?:,(\d+))?) \@{2}(.*)$/;
+
+       $from_lines = 0 unless defined $from_lines;
+       $to_lines   = 0 unless defined $to_lines;
+
+       if ($from->{'href'}) {
+               $from_text = $cgi->a({-href=>"$from->{'href'}#l$from_start",
+                                    -class=>"list"}, $from_text);
+       }
+       if ($to->{'href'}) {
+               $to_text   = $cgi->a({-href=>"$to->{'href'}#l$to_start",
+                                    -class=>"list"}, $to_text);
+       }
+       $line = "<span class=\"chunk_info\">@@ $from_text $to_text @@</span>" .
+               "<span class=\"section\">" . esc_html($section, -nbsp=>1) . "</span>";
+       return $line;
+}
+
+# assumes that $from and $to are defined and correctly filled,
+# and that $line holds a line of chunk header for combined diff
+sub format_cc_diff_chunk_header {
+       my ($line, $from, $to) = @_;
+
+       my ($prefix, $ranges, $section) = $line =~ m/^(\@+) (.*?) \@+(.*)$/;
+       my (@from_text, @from_start, @from_nlines, $to_text, $to_start, $to_nlines);
+
+       @from_text = split(' ', $ranges);
+       for (my $i = 0; $i < @from_text; ++$i) {
+               ($from_start[$i], $from_nlines[$i]) =
+                       (split(',', substr($from_text[$i], 1)), 0);
+       }
+
+       $to_text   = pop @from_text;
+       $to_start  = pop @from_start;
+       $to_nlines = pop @from_nlines;
+
+       $line = "<span class=\"chunk_info\">$prefix ";
+       for (my $i = 0; $i < @from_text; ++$i) {
+               if ($from->{'href'}[$i]) {
+                       $line .= $cgi->a({-href=>"$from->{'href'}[$i]#l$from_start[$i]",
+                                         -class=>"list"}, $from_text[$i]);
+               } else {
+                       $line .= $from_text[$i];
+               }
+               $line .= " ";
+       }
+       if ($to->{'href'}) {
+               $line .= $cgi->a({-href=>"$to->{'href'}#l$to_start",
+                                 -class=>"list"}, $to_text);
+       } else {
+               $line .= $to_text;
+       }
+       $line .= " $prefix</span>" .
+                "<span class=\"section\">" . esc_html($section, -nbsp=>1) . "</span>";
+       return $line;
+}
+
+# process patch (diff) line (not to be used for diff headers),
+# returning HTML-formatted (but not wrapped) line.
+# If the line is passed as a reference, it is treated as HTML and not
+# esc_html()'ed.
+sub format_diff_line {
+       my ($line, $diff_class, $from, $to) = @_;
+
+       if (ref($line)) {
+               $line = $$line;
+       } else {
+               chomp $line;
+               $line = untabify($line);
+
+               if ($from && $to && $line =~ m/^\@{2} /) {
+                       $line = format_unidiff_chunk_header($line, $from, $to);
+               } elsif ($from && $to && $line =~ m/^\@{3}/) {
+                       $line = format_cc_diff_chunk_header($line, $from, $to);
+               } else {
+                       $line = esc_html($line, -nbsp=>1);
+               }
+       }
+
+       my $diff_classes = "diff";
+       $diff_classes .= " $diff_class" if ($diff_class);
+       $line = "<div class=\"$diff_classes\">$line</div>\n";
+
+       return $line;
+}
+
+# Generates undef or something like "_snapshot_" or "snapshot (_tbz2_ _zip_)",
+# linked.  Pass the hash of the tree/commit to snapshot.
+sub format_snapshot_links {
+       my ($hash) = @_;
+       my $num_fmts = @snapshot_fmts;
+       if ($num_fmts > 1) {
+               # A parenthesized list of links bearing format names.
+               # e.g. "snapshot (_tar.gz_ _zip_)"
+               return "snapshot (" . join(' ', map
+                       $cgi->a({
+                               -href => href(
+                                       action=>"snapshot",
+                                       hash=>$hash,
+                                       snapshot_format=>$_
+                               )
+                       }, $known_snapshot_formats{$_}{'display'})
+               , @snapshot_fmts) . ")";
+       } elsif ($num_fmts == 1) {
+               # A single "snapshot" link whose tooltip bears the format name.
+               # i.e. "_snapshot_"
+               my ($fmt) = @snapshot_fmts;
+               return
+                       $cgi->a({
+                               -href => href(
+                                       action=>"snapshot",
+                                       hash=>$hash,
+                                       snapshot_format=>$fmt
+                               ),
+                               -title => "in format: $known_snapshot_formats{$fmt}{'display'}"
+                       }, "snapshot");
+       } else { # $num_fmts == 0
+               return undef;
+       }
+}
+
+## ......................................................................
+## functions returning values to be passed, perhaps after some
+## transformation, to other functions; e.g. returning arguments to href()
+
+# returns hash to be passed to href to generate gitweb URL
+# in -title key it returns description of link
+sub get_feed_info {
+       my $format = shift || 'Atom';
+       my %res = (action => lc($format));
+       my $matched_ref = 0;
+
+       # feed links are possible only for project views
+       return unless (defined $project);
+       # some views should link to OPML, or to generic project feed,
+       # or don't have specific feed yet (so they should use generic)
+       return if (!$action || $action =~ /^(?:tags|heads|forks|tag|search)$/x);
+
+       my $branch = undef;
+       # branches refs uses 'refs/' + $get_branch_refs()[x] + '/' prefix
+       # (fullname) to differentiate from tag links; this also makes
+       # possible to detect branch links
+       for my $ref (get_branch_refs()) {
+               if ((defined $hash_base && $hash_base =~ m!^refs/\Q$ref\E/(.*)$!) ||
+                   (defined $hash      && $hash      =~ m!^refs/\Q$ref\E/(.*)$!)) {
+                       $branch = $1;
+                       $matched_ref = $ref;
+                       last;
+               }
+       }
+       # find log type for feed description (title)
+       my $type = 'log';
+       if (defined $file_name) {
+               $type  = "history of $file_name";
+               $type .= "/" if ($action eq 'tree');
+               $type .= " on '$branch'" if (defined $branch);
+       } else {
+               $type = "log of $branch" if (defined $branch);
+       }
+
+       $res{-title} = $type;
+       $res{'hash'} = (defined $branch ? "refs/$matched_ref/$branch" : undef);
+       $res{'file_name'} = $file_name;
+
+       return %res;
+}
+
+## ----------------------------------------------------------------------
+## git utility subroutines, invoking git commands
+
+# returns path to the core git executable and the --git-dir parameter as list
+sub git_cmd {
+       $number_of_git_cmds++;
+       return $GIT, '--git-dir='.$git_dir;
+}
+
+# quote the given arguments for passing them to the shell
+# quote_command("command", "arg 1", "arg with ' and ! characters")
+# => "'command' 'arg 1' 'arg with '\'' and '\!' characters'"
+# Try to avoid using this function wherever possible.
+sub quote_command {
+       return join(' ',
+               map { my $a = $_; $a =~ s/(['!])/'\\$1'/g; "'$a'" } @_ );
+}
+
+# get HEAD ref of given project as hash
+sub git_get_head_hash {
+       return git_get_full_hash(shift, 'HEAD');
+}
+
+sub git_get_full_hash {
+       return git_get_hash(@_);
+}
+
+sub git_get_short_hash {
+       return git_get_hash(@_, '--short=7');
+}
+
+sub git_get_hash {
+       my ($project, $hash, @options) = @_;
+       my $o_git_dir = $git_dir;
+       my $retval = undef;
+       $git_dir = "$projectroot/$project";
+       if (open my $fd, '-|', git_cmd(), 'rev-parse',
+           '--verify', '-q', @options, $hash) {
+               $retval = <$fd>;
+               chomp $retval if defined $retval;
+               close $fd;
+       }
+       if (defined $o_git_dir) {
+               $git_dir = $o_git_dir;
+       }
+       return $retval;
+}
+
+# get type of given object
+sub git_get_type {
+       my $hash = shift;
+
+       open my $fd, "-|", git_cmd(), "cat-file", '-t', $hash or return;
+       my $type = <$fd>;
+       close $fd or return;
+       chomp $type;
+       return $type;
+}
+
+# repository configuration
+our $config_file = '';
+our %config;
+
+# store multiple values for single key as anonymous array reference
+# single values stored directly in the hash, not as [ <value> ]
+sub hash_set_multi {
+       my ($hash, $key, $value) = @_;
+
+       if (!exists $hash->{$key}) {
+               $hash->{$key} = $value;
+       } elsif (!ref $hash->{$key}) {
+               $hash->{$key} = [ $hash->{$key}, $value ];
+       } else {
+               push @{$hash->{$key}}, $value;
+       }
+}
+
+# return hash of git project configuration
+# optionally limited to some section, e.g. 'gitweb'
+sub git_parse_project_config {
+       my $section_regexp = shift;
+       my %config;
+
+       local $/ = "\0";
+
+       open my $fh, "-|", git_cmd(), "config", '-z', '-l',
+               or return;
+
+       while (my $keyval = <$fh>) {
+               chomp $keyval;
+               my ($key, $value) = split(/\n/, $keyval, 2);
+
+               hash_set_multi(\%config, $key, $value)
+                       if (!defined $section_regexp || $key =~ /^(?:$section_regexp)\./o);
+       }
+       close $fh;
+
+       return %config;
+}
+
+# convert config value to boolean: 'true' or 'false'
+# no value, number > 0, 'true' and 'yes' values are true
+# rest of values are treated as false (never as error)
+sub config_to_bool {
+       my $val = shift;
+
+       return 1 if !defined $val;             # section.key
+
+       # strip leading and trailing whitespace
+       $val =~ s/^\s+//;
+       $val =~ s/\s+$//;
+
+       return (($val =~ /^\d+$/ && $val) ||   # section.key = 1
+               ($val =~ /^(?:true|yes)$/i));  # section.key = true
+}
+
+# convert config value to simple decimal number
+# an optional value suffix of 'k', 'm', or 'g' will cause the value
+# to be multiplied by 1024, 1048576, or 1073741824
+sub config_to_int {
+       my $val = shift;
+
+       # strip leading and trailing whitespace
+       $val =~ s/^\s+//;
+       $val =~ s/\s+$//;
+
+       if (my ($num, $unit) = ($val =~ /^([0-9]*)([kmg])$/i)) {
+               $unit = lc($unit);
+               # unknown unit is treated as 1
+               return $num * ($unit eq 'g' ? 1073741824 :
+                              $unit eq 'm' ?    1048576 :
+                              $unit eq 'k' ?       1024 : 1);
+       }
+       return $val;
+}
+
+# convert config value to array reference, if needed
+sub config_to_multi {
+       my $val = shift;
+
+       return ref($val) ? $val : (defined($val) ? [ $val ] : []);
+}
+
+sub git_get_project_config {
+       my ($key, $type) = @_;
+
+       return unless defined $git_dir;
+
+       # key sanity check
+       return unless ($key);
+       # only subsection, if exists, is case sensitive,
+       # and not lowercased by 'git config -z -l'
+       if (my ($hi, $mi, $lo) = ($key =~ /^([^.]*)\.(.*)\.([^.]*)$/)) {
+               $lo =~ s/_//g;
+               $key = join(".", lc($hi), $mi, lc($lo));
+               return if ($lo =~ /\W/ || $hi =~ /\W/);
+       } else {
+               $key = lc($key);
+               $key =~ s/_//g;
+               return if ($key =~ /\W/);
+       }
+       $key =~ s/^gitweb\.//;
+
+       # type sanity check
+       if (defined $type) {
+               $type =~ s/^--//;
+               $type = undef
+                       unless ($type eq 'bool' || $type eq 'int');
+       }
+
+       # get config
+       if (!defined $config_file ||
+           $config_file ne "$git_dir/config") {
+               %config = git_parse_project_config('gitweb');
+               $config_file = "$git_dir/config";
+       }
+
+       # check if config variable (key) exists
+       return unless exists $config{"gitweb.$key"};
+
+       # ensure given type
+       if (!defined $type) {
+               return $config{"gitweb.$key"};
+       } elsif ($type eq 'bool') {
+               # backward compatibility: 'git config --bool' returns true/false
+               return config_to_bool($config{"gitweb.$key"}) ? 'true' : 'false';
+       } elsif ($type eq 'int') {
+               return config_to_int($config{"gitweb.$key"});
+       }
+       return $config{"gitweb.$key"};
+}
+
+# get hash of given path at given ref
+sub git_get_hash_by_path {
+       my $base = shift;
+       my $path = shift || return undef;
+       my $type = shift;
+
+       $path =~ s,/+$,,;
+
+       open my $fd, "-|", git_cmd(), "ls-tree", $base, "--", $path
+               or die_error(500, "Open git-ls-tree failed");
+       my $line = <$fd>;
+       close $fd or return undef;
+
+       if (!defined $line) {
+               # there is no tree or hash given by $path at $base
+               return undef;
+       }
+
+       #'100644 blob 0fa3f3a66fb6a137f6ec2c19351ed4d807070ffa  panic.c'
+       $line =~ m/^([0-9]+) (.+) ([0-9a-fA-F]{40})\t/;
+       if (defined $type && $type ne $2) {
+               # type doesn't match
+               return undef;
+       }
+       return $3;
+}
+
+# get path of entry with given hash at given tree-ish (ref)
+# used to get 'from' filename for combined diff (merge commit) for renames
+sub git_get_path_by_hash {
+       my $base = shift || return;
+       my $hash = shift || return;
+
+       local $/ = "\0";
+
+       open my $fd, "-|", git_cmd(), "ls-tree", '-r', '-t', '-z', $base
+               or return undef;
+       while (my $line = <$fd>) {
+               chomp $line;
+
+               #'040000 tree 595596a6a9117ddba9fe379b6b012b558bac8423  gitweb'
+               #'100644 blob e02e90f0429be0d2a69b76571101f20b8f75530f  gitweb/README'
+               if ($line =~ m/(?:[0-9]+) (?:.+) $hash\t(.+)$/) {
+                       close $fd;
+                       return $1;
+               }
+       }
+       close $fd;
+       return undef;
+}
+
+## ......................................................................
+## git utility functions, directly accessing git repository
+
+# get the value of config variable either from file named as the variable
+# itself in the repository ($GIT_DIR/$name file), or from gitweb.$name
+# configuration variable in the repository config file.
+sub git_get_file_or_project_config {
+       my ($path, $name) = @_;
+
+       $git_dir = "$projectroot/$path";
+       open my $fd, '<', "$git_dir/$name"
+               or return git_get_project_config($name);
+       my $conf = <$fd>;
+       close $fd;
+       if (defined $conf) {
+               chomp $conf;
+       }
+       return $conf;
+}
+
+sub git_get_project_description {
+       my $path = shift;
+       return git_get_file_or_project_config($path, 'description');
+}
+
+sub git_get_project_category {
+       my $path = shift;
+       return git_get_file_or_project_config($path, 'category');
+}
+
+
+# supported formats:
+# * $GIT_DIR/ctags/<tagname> file (in 'ctags' subdirectory)
+#   - if its contents is a number, use it as tag weight,
+#   - otherwise add a tag with weight 1
+# * $GIT_DIR/ctags file, each line is a tag (with weight 1)
+#   the same value multiple times increases tag weight
+# * `gitweb.ctag' multi-valued repo config variable
+sub git_get_project_ctags {
+       my $project = shift;
+       my $ctags = {};
+
+       $git_dir = "$projectroot/$project";
+       if (opendir my $dh, "$git_dir/ctags") {
+               my @files = grep { -f $_ } map { "$git_dir/ctags/$_" } readdir($dh);
+               foreach my $tagfile (@files) {
+                       open my $ct, '<', $tagfile
+                               or next;
+                       my $val = <$ct>;
+                       chomp $val if $val;
+                       close $ct;
+
+                       (my $ctag = $tagfile) =~ s#.*/##;
+                       if ($val =~ /^\d+$/) {
+                               $ctags->{$ctag} = $val;
+                       } else {
+                               $ctags->{$ctag} = 1;
+                       }
+               }
+               closedir $dh;
+
+       } elsif (open my $fh, '<', "$git_dir/ctags") {
+               while (my $line = <$fh>) {
+                       chomp $line;
+                       $ctags->{$line}++ if $line;
+               }
+               close $fh;
+
+       } else {
+               my $taglist = config_to_multi(git_get_project_config('ctag'));
+               foreach my $tag (@$taglist) {
+                       $ctags->{$tag}++;
+               }
+       }
+
+       return $ctags;
+}
+
+# return hash, where keys are content tags ('ctags'),
+# and values are sum of weights of given tag in every project
+sub git_gather_all_ctags {
+       my $projects = shift;
+       my $ctags = {};
+
+       foreach my $p (@$projects) {
+               foreach my $ct (keys %{$p->{'ctags'}}) {
+                       $ctags->{$ct} += $p->{'ctags'}->{$ct};
+               }
+       }
+
+       return $ctags;
+}
+
+sub git_populate_project_tagcloud {
+       my $ctags = shift;
+
+       # First, merge different-cased tags; tags vote on casing
+       my %ctags_lc;
+       foreach (keys %$ctags) {
+               $ctags_lc{lc $_}->{count} += $ctags->{$_};
+               if (not $ctags_lc{lc $_}->{topcount}
+                   or $ctags_lc{lc $_}->{topcount} < $ctags->{$_}) {
+                       $ctags_lc{lc $_}->{topcount} = $ctags->{$_};
+                       $ctags_lc{lc $_}->{topname} = $_;
+               }
+       }
+
+       my $cloud;
+       my $matched = $input_params{'ctag'};
+       if (eval { require HTML::TagCloud; 1; }) {
+               $cloud = HTML::TagCloud->new;
+               foreach my $ctag (sort keys %ctags_lc) {
+                       # Pad the title with spaces so that the cloud looks
+                       # less crammed.
+                       my $title = esc_html($ctags_lc{$ctag}->{topname});
+                       $title =~ s/ /&nbsp;/g;
+                       $title =~ s/^/&nbsp;/g;
+                       $title =~ s/$/&nbsp;/g;
+                       if (defined $matched && $matched eq $ctag) {
+                               $title = qq(<span class="match">$title</span>);
+                       }
+                       $cloud->add($title, href(project=>undef, ctag=>$ctag),
+                                   $ctags_lc{$ctag}->{count});
+               }
+       } else {
+               $cloud = {};
+               foreach my $ctag (keys %ctags_lc) {
+                       my $title = esc_html($ctags_lc{$ctag}->{topname}, -nbsp=>1);
+                       if (defined $matched && $matched eq $ctag) {
+                               $title = qq(<span class="match">$title</span>);
+                       }
+                       $cloud->{$ctag}{count} = $ctags_lc{$ctag}->{count};
+                       $cloud->{$ctag}{ctag} =
+                               $cgi->a({-href=>href(project=>undef, ctag=>$ctag)}, $title);
+               }
+       }
+       return $cloud;
+}
+
+sub git_show_project_tagcloud {
+       my ($cloud, $count) = @_;
+       if (ref $cloud eq 'HTML::TagCloud') {
+               return $cloud->html_and_css($count);
+       } else {
+               my @tags = sort { $cloud->{$a}->{'count'} <=> $cloud->{$b}->{'count'} } keys %$cloud;
+               return
+                       '<div id="htmltagcloud"'.($project ? '' : ' align="center"').'>' .
+                       join (', ', map {
+                               $cloud->{$_}->{'ctag'}
+                       } splice(@tags, 0, $count)) .
+                       '</div>';
+       }
+}
+
+sub git_get_project_url_list {
+       my $path = shift;
+
+       $git_dir = "$projectroot/$path";
+       open my $fd, '<', "$git_dir/cloneurl"
+               or return wantarray ?
+               @{ config_to_multi(git_get_project_config('url')) } :
+                  config_to_multi(git_get_project_config('url'));
+       my @git_project_url_list = map { chomp; $_ } <$fd>;
+       close $fd;
+
+       return wantarray ? @git_project_url_list : \@git_project_url_list;
+}
+
+sub git_get_projects_list {
+       my $filter = shift || '';
+       my $paranoid = shift;
+       my @list;
+
+       if (-d $projects_list) {
+               # search in directory
+               my $dir = $projects_list;
+               # remove the trailing "/"
+               $dir =~ s!/+$!!;
+               my $pfxlen = length("$dir");
+               my $pfxdepth = ($dir =~ tr!/!!);
+               # when filtering, search only given subdirectory
+               if ($filter && !$paranoid) {
+                       $dir .= "/$filter";
+                       $dir =~ s!/+$!!;
+               }
+
+               File::Find::find({
+                       follow_fast => 1, # follow symbolic links
+                       follow_skip => 2, # ignore duplicates
+                       dangling_symlinks => 0, # ignore dangling symlinks, silently
+                       wanted => sub {
+                               # global variables
+                               our $project_maxdepth;
+                               our $projectroot;
+                               # skip project-list toplevel, if we get it.
+                               return if (m!^[/.]$!);
+                               # only directories can be git repositories
+                               return unless (-d $_);
+                               # don't traverse too deep (Find is super slow on os x)
+                               # $project_maxdepth excludes depth of $projectroot
+                               if (($File::Find::name =~ tr!/!!) - $pfxdepth > $project_maxdepth) {
+                                       $File::Find::prune = 1;
+                                       return;
+                               }
+
+                               my $path = substr($File::Find::name, $pfxlen + 1);
+                               # paranoidly only filter here
+                               if ($paranoid && $filter && $path !~ m!^\Q$filter\E/!) {
+                                       next;
+                               }
+                               # we check related file in $projectroot
+                               if (check_export_ok("$projectroot/$path")) {
+                                       push @list, { path => $path };
+                                       $File::Find::prune = 1;
+                               }
+                       },
+               }, "$dir");
+
+       } elsif (-f $projects_list) {
+               # read from file(url-encoded):
+               # 'git%2Fgit.git Linus+Torvalds'
+               # 'libs%2Fklibc%2Fklibc.git H.+Peter+Anvin'
+               # 'linux%2Fhotplug%2Fudev.git Greg+Kroah-Hartman'
+               open my $fd, '<', $projects_list or return;
+       PROJECT:
+               while (my $line = <$fd>) {
+                       chomp $line;
+                       my ($path, $owner) = split ' ', $line;
+                       $path = unescape($path);
+                       $owner = unescape($owner);
+                       if (!defined $path) {
+                               next;
+                       }
+                       # if $filter is rpovided, check if $path begins with $filter
+                       if ($filter && $path !~ m!^\Q$filter\E/!) {
+                               next;
+                       }
+                       if (check_export_ok("$projectroot/$path")) {
+                               my $pr = {
+                                       path => $path
+                               };
+                               if ($owner) {
+                                       $pr->{'owner'} = to_utf8($owner);
+                               }
+                               push @list, $pr;
+                       }
+               }
+               close $fd;
+       }
+       return @list;
+}
+
+# written with help of Tree::Trie module (Perl Artistic License, GPL compatibile)
+# as side effects it sets 'forks' field to list of forks for forked projects
+sub filter_forks_from_projects_list {
+       my $projects = shift;
+
+       my %trie; # prefix tree of directories (path components)
+       # generate trie out of those directories that might contain forks
+       foreach my $pr (@$projects) {
+               my $path = $pr->{'path'};
+               $path =~ s/\.git$//;      # forks of 'repo.git' are in 'repo/' directory
+               next if ($path =~ m!/$!); # skip non-bare repositories, e.g. 'repo/.git'
+               next unless ($path);      # skip '.git' repository: tests, git-instaweb
+               next unless (-d "$projectroot/$path"); # containing directory exists
+               $pr->{'forks'} = [];      # there can be 0 or more forks of project
+
+               # add to trie
+               my @dirs = split('/', $path);
+               # walk the trie, until either runs out of components or out of trie
+               my $ref = \%trie;
+               while (scalar @dirs &&
+                      exists($ref->{$dirs[0]})) {
+                       $ref = $ref->{shift @dirs};
+               }
+               # create rest of trie structure from rest of components
+               foreach my $dir (@dirs) {
+                       $ref = $ref->{$dir} = {};
+               }
+               # create end marker, store $pr as a data
+               $ref->{''} = $pr if (!exists $ref->{''});
+       }
+
+       # filter out forks, by finding shortest prefix match for paths
+       my @filtered;
+ PROJECT:
+       foreach my $pr (@$projects) {
+               # trie lookup
+               my $ref = \%trie;
+       DIR:
+               foreach my $dir (split('/', $pr->{'path'})) {
+                       if (exists $ref->{''}) {
+                               # found [shortest] prefix, is a fork - skip it
+                               push @{$ref->{''}{'forks'}}, $pr;
+                               next PROJECT;
+                       }
+                       if (!exists $ref->{$dir}) {
+                               # not in trie, cannot have prefix, not a fork
+                               push @filtered, $pr;
+                               next PROJECT;
+                       }
+                       # If the dir is there, we just walk one step down the trie.
+                       $ref = $ref->{$dir};
+               }
+               # we ran out of trie
+               # (shouldn't happen: it's either no match, or end marker)
+               push @filtered, $pr;
+       }
+
+       return @filtered;
+}
+
+# note: fill_project_list_info must be run first,
+# for 'descr_long' and 'ctags' to be filled
+sub search_projects_list {
+       my ($projlist, %opts) = @_;
+       my $tagfilter  = $opts{'tagfilter'};
+       my $search_re = $opts{'search_regexp'};
+
+       return @$projlist
+               unless ($tagfilter || $search_re);
+
+       # searching projects require filling to be run before it;
+       fill_project_list_info($projlist,
+                              $tagfilter  ? 'ctags' : (),
+                              $search_re ? ('path', 'descr') : ());
+       my @projects;
+ PROJECT:
+       foreach my $pr (@$projlist) {
+
+               if ($tagfilter) {
+                       next unless ref($pr->{'ctags'}) eq 'HASH';
+                       next unless
+                               grep { lc($_) eq lc($tagfilter) } keys %{$pr->{'ctags'}};
+               }
+
+               if ($search_re) {
+                       next unless
+                               $pr->{'path'} =~ /$search_re/ ||
+                               $pr->{'descr_long'} =~ /$search_re/;
+               }
+
+               push @projects, $pr;
+       }
+
+       return @projects;
+}
+
+our $gitweb_project_owner = undef;
+sub git_get_project_list_from_file {
+
+       return if (defined $gitweb_project_owner);
+
+       $gitweb_project_owner = {};
+       # read from file (url-encoded):
+       # 'git%2Fgit.git Linus+Torvalds'
+       # 'libs%2Fklibc%2Fklibc.git H.+Peter+Anvin'
+       # 'linux%2Fhotplug%2Fudev.git Greg+Kroah-Hartman'
+       if (-f $projects_list) {
+               open(my $fd, '<', $projects_list);
+               while (my $line = <$fd>) {
+                       chomp $line;
+                       my ($pr, $ow) = split ' ', $line;
+                       $pr = unescape($pr);
+                       $ow = unescape($ow);
+                       $gitweb_project_owner->{$pr} = to_utf8($ow);
+               }
+               close $fd;
+       }
+}
+
+sub git_get_project_owner {
+       my $project = shift;
+       my $owner;
+
+       return undef unless $project;
+       $git_dir = "$projectroot/$project";
+
+       if (!defined $gitweb_project_owner) {
+               git_get_project_list_from_file();
+       }
+
+       if (exists $gitweb_project_owner->{$project}) {
+               $owner = $gitweb_project_owner->{$project};
+       }
+       if (!defined $owner){
+               $owner = git_get_project_config('owner');
+       }
+       if (!defined $owner) {
+               $owner = get_file_owner("$git_dir");
+       }
+
+       return $owner;
+}
+
+sub git_get_last_activity {
+       my ($path) = @_;
+       my $fd;
+
+       $git_dir = "$projectroot/$path";
+       open($fd, "-|", git_cmd(), 'for-each-ref',
+            '--format=%(committer)',
+            '--sort=-committerdate',
+            '--count=1',
+            map { "refs/$_" } get_branch_refs ()) or return;
+       my $most_recent = <$fd>;
+       close $fd or return;
+       if (defined $most_recent &&
+           $most_recent =~ / (\d+) [-+][01]\d\d\d$/) {
+               my $timestamp = $1;
+               my $age = time - $timestamp;
+               return ($age, age_string($age));
+       }
+       return (undef, undef);
+}
+
+# Implementation note: when a single remote is wanted, we cannot use 'git
+# remote show -n' because that command always work (assuming it's a remote URL
+# if it's not defined), and we cannot use 'git remote show' because that would
+# try to make a network roundtrip. So the only way to find if that particular
+# remote is defined is to walk the list provided by 'git remote -v' and stop if
+# and when we find what we want.
+sub git_get_remotes_list {
+       my $wanted = shift;
+       my %remotes = ();
+
+       open my $fd, '-|' , git_cmd(), 'remote', '-v';
+       return unless $fd;
+       while (my $remote = <$fd>) {
+               chomp $remote;
+               $remote =~ s!\t(.*?)\s+\((\w+)\)$!!;
+               next if $wanted and not $remote eq $wanted;
+               my ($url, $key) = ($1, $2);
+
+               $remotes{$remote} ||= { 'heads' => () };
+               $remotes{$remote}{$key} = $url;
+       }
+       close $fd or return;
+       return wantarray ? %remotes : \%remotes;
+}
+
+# Takes a hash of remotes as first parameter and fills it by adding the
+# available remote heads for each of the indicated remotes.
+sub fill_remote_heads {
+       my $remotes = shift;
+       my @heads = map { "remotes/$_" } keys %$remotes;
+       my @remoteheads = git_get_heads_list(undef, @heads);
+       foreach my $remote (keys %$remotes) {
+               $remotes->{$remote}{'heads'} = [ grep {
+                       $_->{'name'} =~ s!^$remote/!!
+                       } @remoteheads ];
+       }
+}
+
+sub git_get_references {
+       my $type = shift || "";
+       my %refs;
+       # 5dc01c595e6c6ec9ccda4f6f69c131c0dd945f8c refs/tags/v2.6.11
+       # c39ae07f393806ccf406ef966e9a15afc43cc36a refs/tags/v2.6.11^{}
+       open my $fd, "-|", git_cmd(), "show-ref", "--dereference",
+               ($type ? ("--", "refs/$type") : ()) # use -- <pattern> if $type
+               or return;
+
+       while (my $line = <$fd>) {
+               chomp $line;
+               if ($line =~ m!^([0-9a-fA-F]{40})\srefs/($type.*)$!) {
+                       if (defined $refs{$1}) {
+                               push @{$refs{$1}}, $2;
+                       } else {
+                               $refs{$1} = [ $2 ];
+                       }
+               }
+       }
+       close $fd or return;
+       return \%refs;
+}
+
+sub git_get_rev_name_tags {
+       my $hash = shift || return undef;
+
+       open my $fd, "-|", git_cmd(), "name-rev", "--tags", $hash
+               or return;
+       my $name_rev = <$fd>;
+       close $fd;
+
+       if ($name_rev =~ m|^$hash tags/(.*)$|) {
+               return $1;
+       } else {
+               # catches also '$hash undefined' output
+               return undef;
+       }
+}
+
+## ----------------------------------------------------------------------
+## parse to hash functions
+
+sub parse_date {
+       my $epoch = shift;
+       my $tz = shift || "-0000";
+
+       my %date;
+       my @months = ("Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec");
+       my @days = ("Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat");
+       my ($sec, $min, $hour, $mday, $mon, $year, $wday, $yday) = gmtime($epoch);
+       $date{'hour'} = $hour;
+       $date{'minute'} = $min;
+       $date{'mday'} = $mday;
+       $date{'day'} = $days[$wday];
+       $date{'month'} = $months[$mon];
+       $date{'rfc2822'}   = sprintf "%s, %d %s %4d %02d:%02d:%02d +0000",
+                            $days[$wday], $mday, $months[$mon], 1900+$year, $hour ,$min, $sec;
+       $date{'mday-time'} = sprintf "%d %s %02d:%02d",
+                            $mday, $months[$mon], $hour ,$min;
+       $date{'iso-8601'}  = sprintf "%04d-%02d-%02dT%02d:%02d:%02dZ",
+                            1900+$year, 1+$mon, $mday, $hour ,$min, $sec;
+
+       my ($tz_sign, $tz_hour, $tz_min) =
+               ($tz =~ m/^([-+])(\d\d)(\d\d)$/);
+       $tz_sign = ($tz_sign eq '-' ? -1 : +1);
+       my $local = $epoch + $tz_sign*((($tz_hour*60) + $tz_min)*60);
+       ($sec, $min, $hour, $mday, $mon, $year, $wday, $yday) = gmtime($local);
+       $date{'hour_local'} = $hour;
+       $date{'minute_local'} = $min;
+       $date{'tz_local'} = $tz;
+       $date{'iso-tz'} = sprintf("%04d-%02d-%02d %02d:%02d:%02d %s",
+                                 1900+$year, $mon+1, $mday,
+                                 $hour, $min, $sec, $tz);
+       return %date;
+}
+
+sub parse_tag {
+       my $tag_id = shift;
+       my %tag;
+       my @comment;
+
+       open my $fd, "-|", git_cmd(), "cat-file", "tag", $tag_id or return;
+       $tag{'id'} = $tag_id;
+       while (my $line = <$fd>) {
+               chomp $line;
+               if ($line =~ m/^object ([0-9a-fA-F]{40})$/) {
+                       $tag{'object'} = $1;
+               } elsif ($line =~ m/^type (.+)$/) {
+                       $tag{'type'} = $1;
+               } elsif ($line =~ m/^tag (.+)$/) {
+                       $tag{'name'} = $1;
+               } elsif ($line =~ m/^tagger (.*) ([0-9]+) (.*)$/) {
+                       $tag{'author'} = $1;
+                       $tag{'author_epoch'} = $2;
+                       $tag{'author_tz'} = $3;
+                       if ($tag{'author'} =~ m/^([^<]+) <([^>]*)>/) {
+                               $tag{'author_name'}  = $1;
+                               $tag{'author_email'} = $2;
+                       } else {
+                               $tag{'author_name'} = $tag{'author'};
+                       }
+               } elsif ($line =~ m/--BEGIN/) {
+                       push @comment, $line;
+                       last;
+               } elsif ($line eq "") {
+                       last;
+               }
+       }
+       push @comment, <$fd>;
+       $tag{'comment'} = \@comment;
+       close $fd or return;
+       if (!defined $tag{'name'}) {
+               return
+       };
+       return %tag
+}
+
+sub parse_commit_text {
+       my ($commit_text, $withparents) = @_;
+       my @commit_lines = split '\n', $commit_text;
+       my %co;
+
+       pop @commit_lines; # Remove '\0'
+
+       if (! @commit_lines) {
+               return;
+       }
+
+       my $header = shift @commit_lines;
+       if ($header !~ m/^[0-9a-fA-F]{40}/) {
+               return;
+       }
+       ($co{'id'}, my @parents) = split ' ', $header;
+       while (my $line = shift @commit_lines) {
+               last if $line eq "\n";
+               if ($line =~ m/^tree ([0-9a-fA-F]{40})$/) {
+                       $co{'tree'} = $1;
+               } elsif ((!defined $withparents) && ($line =~ m/^parent ([0-9a-fA-F]{40})$/)) {
+                       push @parents, $1;
+               } elsif ($line =~ m/^author (.*) ([0-9]+) (.*)$/) {
+                       $co{'author'} = to_utf8($1);
+                       $co{'author_epoch'} = $2;
+                       $co{'author_tz'} = $3;
+                       if ($co{'author'} =~ m/^([^<]+) <([^>]*)>/) {
+                               $co{'author_name'}  = $1;
+                               $co{'author_email'} = $2;
+                       } else {
+                               $co{'author_name'} = $co{'author'};
+                       }
+               } elsif ($line =~ m/^committer (.*) ([0-9]+) (.*)$/) {
+                       $co{'committer'} = to_utf8($1);
+                       $co{'committer_epoch'} = $2;
+                       $co{'committer_tz'} = $3;
+                       if ($co{'committer'} =~ m/^([^<]+) <([^>]*)>/) {
+                               $co{'committer_name'}  = $1;
+                               $co{'committer_email'} = $2;
+                       } else {
+                               $co{'committer_name'} = $co{'committer'};
+                       }
+               }
+       }
+       if (!defined $co{'tree'}) {
+               return;
+       };
+       $co{'parents'} = \@parents;
+       $co{'parent'} = $parents[0];
+
+       foreach my $title (@commit_lines) {
+               $title =~ s/^    //;
+               if ($title ne "") {
+                       $co{'title'} = chop_str($title, 80, 5);
+                       # remove leading stuff of merges to make the interesting part visible
+                       if (length($title) > 50) {
+                               $title =~ s/^Automatic //;
+                               $title =~ s/^merge (of|with) /Merge ... /i;
+                               if (length($title) > 50) {
+                                       $title =~ s/(http|rsync):\/\///;
+                               }
+                               if (length($title) > 50) {
+                                       $title =~ s/(master|www|rsync)\.//;
+                               }
+                               if (length($title) > 50) {
+                                       $title =~ s/kernel.org:?//;
+                               }
+                               if (length($title) > 50) {
+                                       $title =~ s/\/pub\/scm//;
+                               }
+                       }
+                       $co{'title_short'} = chop_str($title, 50, 5);
+                       last;
+               }
+       }
+       if (! defined $co{'title'} || $co{'title'} eq "") {
+               $co{'title'} = $co{'title_short'} = '(no commit message)';
+       }
+       # remove added spaces
+       foreach my $line (@commit_lines) {
+               $line =~ s/^    //;
+       }
+       $co{'comment'} = \@commit_lines;
+
+       my $age = time - $co{'committer_epoch'};
+       $co{'age'} = $age;
+       $co{'age_string'} = age_string($age);
+       my ($sec, $min, $hour, $mday, $mon, $year, $wday, $yday) = gmtime($co{'committer_epoch'});
+       if ($age > 60*60*24*7*2) {
+               $co{'age_string_date'} = sprintf "%4i-%02u-%02i", 1900 + $year, $mon+1, $mday;
+               $co{'age_string_age'} = $co{'age_string'};
+       } else {
+               $co{'age_string_date'} = $co{'age_string'};
+               $co{'age_string_age'} = sprintf "%4i-%02u-%02i", 1900 + $year, $mon+1, $mday;
+       }
+       return %co;
+}
+
+sub parse_commit {
+       my ($commit_id) = @_;
+       my %co;
+
+       local $/ = "\0";
+
+       open my $fd, "-|", git_cmd(), "rev-list",
+               "--parents",
+               "--header",
+               "--max-count=1",
+               $commit_id,
+               "--",
+               or die_error(500, "Open git-rev-list failed");
+       %co = parse_commit_text(<$fd>, 1);
+       close $fd;
+
+       return %co;
+}
+
+sub parse_commits {
+       my ($commit_id, $maxcount, $skip, $filename, @args) = @_;
+       my @cos;
+
+       $maxcount ||= 1;
+       $skip ||= 0;
+
+       local $/ = "\0";
+
+       open my $fd, "-|", git_cmd(), "rev-list",
+               "--header",
+               @args,
+               ("--max-count=" . $maxcount),
+               ("--skip=" . $skip),
+               @extra_options,
+               $commit_id,
+               "--",
+               ($filename ? ($filename) : ())
+               or die_error(500, "Open git-rev-list failed");
+       while (my $line = <$fd>) {
+               my %co = parse_commit_text($line);
+               push @cos, \%co;
+       }
+       close $fd;
+
+       return wantarray ? @cos : \@cos;
+}
+
+# parse line of git-diff-tree "raw" output
+sub parse_difftree_raw_line {
+       my $line = shift;
+       my %res;
+
+       # ':100644 100644 03b218260e99b78c6df0ed378e59ed9205ccc96d 3b93d5e7cc7f7dd4ebed13a5cc1a4ad976fc94d8 M   ls-files.c'
+       # ':100644 100644 7f9281985086971d3877aca27704f2aaf9c448ce bc190ebc71bbd923f2b728e505408f5e54bd073a M   rev-tree.c'
+       if ($line =~ m/^:([0-7]{6}) ([0-7]{6}) ([0-9a-fA-F]{40}) ([0-9a-fA-F]{40}) (.)([0-9]{0,3})\t(.*)$/) {
+               $res{'from_mode'} = $1;
+               $res{'to_mode'} = $2;
+               $res{'from_id'} = $3;
+               $res{'to_id'} = $4;
+               $res{'status'} = $5;
+               $res{'similarity'} = $6;
+               if ($res{'status'} eq 'R' || $res{'status'} eq 'C') { # renamed or copied
+                       ($res{'from_file'}, $res{'to_file'}) = map { unquote($_) } split("\t", $7);
+               } else {
+                       $res{'from_file'} = $res{'to_file'} = $res{'file'} = unquote($7);
+               }
+       }
+       # '::100755 100755 100755 60e79ca1b01bc8b057abe17ddab484699a7f5fdb 94067cc5f73388f33722d52ae02f44692bc07490 94067cc5f73388f33722d52ae02f44692bc07490 MR git-gui/git-gui.sh'
+       # combined diff (for merge commit)
+       elsif ($line =~ s/^(::+)((?:[0-7]{6} )+)((?:[0-9a-fA-F]{40} )+)([a-zA-Z]+)\t(.*)$//) {
+               $res{'nparents'}  = length($1);
+               $res{'from_mode'} = [ split(' ', $2) ];
+               $res{'to_mode'} = pop @{$res{'from_mode'}};
+               $res{'from_id'} = [ split(' ', $3) ];
+               $res{'to_id'} = pop @{$res{'from_id'}};
+               $res{'status'} = [ split('', $4) ];
+               $res{'to_file'} = unquote($5);
+       }
+       # 'c512b523472485aef4fff9e57b229d9d243c967f'
+       elsif ($line =~ m/^([0-9a-fA-F]{40})$/) {
+               $res{'commit'} = $1;
+       }
+
+       return wantarray ? %res : \%res;
+}
+
+# wrapper: return parsed line of git-diff-tree "raw" output
+# (the argument might be raw line, or parsed info)
+sub parsed_difftree_line {
+       my $line_or_ref = shift;
+
+       if (ref($line_or_ref) eq "HASH") {
+               # pre-parsed (or generated by hand)
+               return $line_or_ref;
+       } else {
+               return parse_difftree_raw_line($line_or_ref);
+       }
+}
+
+# parse line of git-ls-tree output
+sub parse_ls_tree_line {
+       my $line = shift;
+       my %opts = @_;
+       my %res;
+
+       if ($opts{'-l'}) {
+               #'100644 blob 0fa3f3a66fb6a137f6ec2c19351ed4d807070ffa   16717  panic.c'
+               $line =~ m/^([0-9]+) (.+) ([0-9a-fA-F]{40}) +(-|[0-9]+)\t(.+)$/s;
+
+               $res{'mode'} = $1;
+               $res{'type'} = $2;
+               $res{'hash'} = $3;
+               $res{'size'} = $4;
+               if ($opts{'-z'}) {
+                       $res{'name'} = $5;
+               } else {
+                       $res{'name'} = unquote($5);
+               }
+       } else {
+               #'100644 blob 0fa3f3a66fb6a137f6ec2c19351ed4d807070ffa  panic.c'
+               $line =~ m/^([0-9]+) (.+) ([0-9a-fA-F]{40})\t(.+)$/s;
+
+               $res{'mode'} = $1;
+               $res{'type'} = $2;
+               $res{'hash'} = $3;
+               if ($opts{'-z'}) {
+                       $res{'name'} = $4;
+               } else {
+                       $res{'name'} = unquote($4);
+               }
+       }
+
+       return wantarray ? %res : \%res;
+}
+
+# generates _two_ hashes, references to which are passed as 2 and 3 argument
+sub parse_from_to_diffinfo {
+       my ($diffinfo, $from, $to, @parents) = @_;
+
+       if ($diffinfo->{'nparents'}) {
+               # combined diff
+               $from->{'file'} = [];
+               $from->{'href'} = [];
+               fill_from_file_info($diffinfo, @parents)
+                       unless exists $diffinfo->{'from_file'};
+               for (my $i = 0; $i < $diffinfo->{'nparents'}; $i++) {
+                       $from->{'file'}[$i] =
+                               defined $diffinfo->{'from_file'}[$i] ?
+                                       $diffinfo->{'from_file'}[$i] :
+                                       $diffinfo->{'to_file'};
+                       if ($diffinfo->{'status'}[$i] ne "A") { # not new (added) file
+                               $from->{'href'}[$i] = href(action=>"blob",
+                                                          hash_base=>$parents[$i],
+                                                          hash=>$diffinfo->{'from_id'}[$i],
+                                                          file_name=>$from->{'file'}[$i]);
+                       } else {
+                               $from->{'href'}[$i] = undef;
+                       }
+               }
+       } else {
+               # ordinary (not combined) diff
+               $from->{'file'} = $diffinfo->{'from_file'};
+               if ($diffinfo->{'status'} ne "A") { # not new (added) file
+                       $from->{'href'} = href(action=>"blob", hash_base=>$hash_parent,
+                                              hash=>$diffinfo->{'from_id'},
+                                              file_name=>$from->{'file'});
+               } else {
+                       delete $from->{'href'};
+               }
+       }
+
+       $to->{'file'} = $diffinfo->{'to_file'};
+       if (!is_deleted($diffinfo)) { # file exists in result
+               $to->{'href'} = href(action=>"blob", hash_base=>$hash,
+                                    hash=>$diffinfo->{'to_id'},
+                                    file_name=>$to->{'file'});
+       } else {
+               delete $to->{'href'};
+       }
+}
+
+## ......................................................................
+## parse to array of hashes functions
+
+sub git_get_heads_list {
+       my ($limit, @classes) = @_;
+       @classes = get_branch_refs() unless @classes;
+       my @patterns = map { "refs/$_" } @classes;
+       my @headslist;
+
+       open my $fd, '-|', git_cmd(), 'for-each-ref',
+               ($limit ? '--count='.($limit+1) : ()), '--sort=-committerdate',
+               '--format=%(objectname) %(refname) %(subject)%00%(committer)',
+               @patterns
+               or return;
+       while (my $line = <$fd>) {
+               my %ref_item;
+
+               chomp $line;
+               my ($refinfo, $committerinfo) = split(/\0/, $line);
+               my ($hash, $name, $title) = split(' ', $refinfo, 3);
+               my ($committer, $epoch, $tz) =
+                       ($committerinfo =~ /^(.*) ([0-9]+) (.*)$/);
+               $ref_item{'fullname'}  = $name;
+               my $strip_refs = join '|', map { quotemeta } get_branch_refs();
+               $name =~ s!^refs/($strip_refs|remotes)/!!;
+               $ref_item{'name'} = $name;
+               # for refs neither in 'heads' nor 'remotes' we want to
+               # show their ref dir
+               my $ref_dir = (defined $1) ? $1 : '';
+               if ($ref_dir ne '' and $ref_dir ne 'heads' and $ref_dir ne 'remotes') {
+                   $ref_item{'name'} .= ' (' . $ref_dir . ')';
+               }
+
+               $ref_item{'id'}    = $hash;
+               $ref_item{'title'} = $title || '(no commit message)';
+               $ref_item{'epoch'} = $epoch;
+               if ($epoch) {
+                       $ref_item{'age'} = age_string(time - $ref_item{'epoch'});
+               } else {
+                       $ref_item{'age'} = "unknown";
+               }
+
+               push @headslist, \%ref_item;
+       }
+       close $fd;
+
+       return wantarray ? @headslist : \@headslist;
+}
+
+sub git_get_tags_list {
+       my $limit = shift;
+       my @tagslist;
+
+       open my $fd, '-|', git_cmd(), 'for-each-ref',
+               ($limit ? '--count='.($limit+1) : ()), '--sort=-creatordate',
+               '--format=%(objectname) %(objecttype) %(refname) '.
+               '%(*objectname) %(*objecttype) %(subject)%00%(creator)',
+               'refs/tags'
+               or return;
+       while (my $line = <$fd>) {
+               my %ref_item;
+
+               chomp $line;
+               my ($refinfo, $creatorinfo) = split(/\0/, $line);
+               my ($id, $type, $name, $refid, $reftype, $title) = split(' ', $refinfo, 6);
+               my ($creator, $epoch, $tz) =
+                       ($creatorinfo =~ /^(.*) ([0-9]+) (.*)$/);
+               $ref_item{'fullname'} = $name;
+               $name =~ s!^refs/tags/!!;
+
+               $ref_item{'type'} = $type;
+               $ref_item{'id'} = $id;
+               $ref_item{'name'} = $name;
+               if ($type eq "tag") {
+                       $ref_item{'subject'} = $title;
+                       $ref_item{'reftype'} = $reftype;
+                       $ref_item{'refid'}   = $refid;
+               } else {
+                       $ref_item{'reftype'} = $type;
+                       $ref_item{'refid'}   = $id;
+               }
+
+               if ($type eq "tag" || $type eq "commit") {
+                       $ref_item{'epoch'} = $epoch;
+                       if ($epoch) {
+                               $ref_item{'age'} = age_string(time - $ref_item{'epoch'});
+                       } else {
+                               $ref_item{'age'} = "unknown";
+                       }
+               }
+
+               push @tagslist, \%ref_item;
+       }
+       close $fd;
+
+       return wantarray ? @tagslist : \@tagslist;
+}
+
+## ----------------------------------------------------------------------
+## filesystem-related functions
+
+sub get_file_owner {
+       my $path = shift;
+
+       my ($dev, $ino, $mode, $nlink, $st_uid, $st_gid, $rdev, $size) = stat($path);
+       my ($name, $passwd, $uid, $gid, $quota, $comment, $gcos, $dir, $shell) = getpwuid($st_uid);
+       if (!defined $gcos) {
+               return undef;
+       }
+       my $owner = $gcos;
+       $owner =~ s/[,;].*$//;
+       return to_utf8($owner);
+}
+
+# assume that file exists
+sub insert_file {
+       my $filename = shift;
+
+       open my $fd, '<', $filename;
+       print map { to_utf8($_) } <$fd>;
+       close $fd;
+}
+
+## ......................................................................
+## mimetype related functions
+
+sub mimetype_guess_file {
+       my $filename = shift;
+       my $mimemap = shift;
+       -r $mimemap or return undef;
+
+       my %mimemap;
+       open(my $mh, '<', $mimemap) or return undef;
+       while (<$mh>) {
+               next if m/^#/; # skip comments
+               my ($mimetype, @exts) = split(/\s+/);
+               foreach my $ext (@exts) {
+                       $mimemap{$ext} = $mimetype;
+               }
+       }
+       close($mh);
+
+       $filename =~ /\.([^.]*)$/;
+       return $mimemap{$1};
+}
+
+sub mimetype_guess {
+       my $filename = shift;
+       my $mime;
+       $filename =~ /\./ or return undef;
+
+       if ($mimetypes_file) {
+               my $file = $mimetypes_file;
+               if ($file !~ m!^/!) { # if it is relative path
+                       # it is relative to project
+                       $file = "$projectroot/$project/$file";
+               }
+               $mime = mimetype_guess_file($filename, $file);
+       }
+       $mime ||= mimetype_guess_file($filename, '/etc/mime.types');
+       return $mime;
+}
+
+sub blob_mimetype {
+       my $fd = shift;
+       my $filename = shift;
+
+       if ($filename) {
+               my $mime = mimetype_guess($filename);
+               $mime and return $mime;
+       }
+
+       # just in case
+       return $default_blob_plain_mimetype unless $fd;
+
+       if (-T $fd) {
+               return 'text/plain';
+       } elsif (! $filename) {
+               return 'application/octet-stream';
+       } elsif ($filename =~ m/\.png$/i) {
+               return 'image/png';
+       } elsif ($filename =~ m/\.gif$/i) {
+               return 'image/gif';
+       } elsif ($filename =~ m/\.jpe?g$/i) {
+               return 'image/jpeg';
+       } else {
+               return 'application/octet-stream';
+       }
+}
+
+sub blob_contenttype {
+       my ($fd, $file_name, $type) = @_;
+
+       $type ||= blob_mimetype($fd, $file_name);
+       if ($type eq 'text/plain' && defined $default_text_plain_charset) {
+               $type .= "; charset=$default_text_plain_charset";
+       }
+
+       return $type;
+}
+
+# guess file syntax for syntax highlighting; return undef if no highlighting
+# the name of syntax can (in the future) depend on syntax highlighter used
+sub guess_file_syntax {
+       my ($highlight, $file_name) = @_;
+       return undef unless ($highlight && defined $file_name);
+       my $basename = basename($file_name, '.in');
+       return $highlight_basename{$basename}
+               if exists $highlight_basename{$basename};
+
+       $basename =~ /\.([^.]*)$/;
+       my $ext = $1 or return undef;
+       return $highlight_ext{$ext}
+               if exists $highlight_ext{$ext};
+
+       return undef;
+}
+
+# run highlighter and return FD of its output,
+# or return original FD if no highlighting
+sub run_highlighter {
+       my ($fd, $highlight, $syntax) = @_;
+       return $fd unless ($highlight);
+
+       close $fd;
+       my $syntax_arg = (defined $syntax) ? "--syntax $syntax" : "--force";
+       open $fd, quote_command(git_cmd(), "cat-file", "blob", $hash)." | ".
+                 quote_command($^X, '-CO', '-MEncode=decode,FB_DEFAULT', '-pse',
+                   '$_ = decode($fe, $_, FB_DEFAULT) if !utf8::decode($_);',
+                   '--', "-fe=$fallback_encoding")." | ".
+                 quote_command($highlight_bin).
+                 " --replace-tabs=8 --fragment $syntax_arg |"
+               or die_error(500, "Couldn't open file or run syntax highlighter");
+       return $fd;
+}
+
+## ======================================================================
+## functions printing HTML: header, footer, error page
+
+sub get_page_title {
+       my $title = to_utf8($site_name);
+
+       unless (defined $project) {
+               if (defined $project_filter) {
+                       $title .= " - projects in '" . esc_path($project_filter) . "'";
+               }
+               return $title;
+       }
+       $title .= " - " . to_utf8($project);
+
+       return $title unless (defined $action);
+       $title .= "/$action"; # $action is US-ASCII (7bit ASCII)
+
+       return $title unless (defined $file_name);
+       $title .= " - " . esc_path($file_name);
+       if ($action eq "tree" && $file_name !~ m|/$|) {
+               $title .= "/";
+       }
+
+       return $title;
+}
+
+sub get_content_type_html {
+       # require explicit support from the UA if we are to send the page as
+       # 'application/xhtml+xml', otherwise send it as plain old 'text/html'.
+       # we have to do this because MSIE sometimes globs '*/*', pretending to
+       # support xhtml+xml but choking when it gets what it asked for.
+       if (defined $cgi->http('HTTP_ACCEPT') &&
+           $cgi->http('HTTP_ACCEPT') =~ m/(,|;|\s|^)application\/xhtml\+xml(,|;|\s|$)/ &&
+           $cgi->Accept('application/xhtml+xml') != 0) {
+               return 'application/xhtml+xml';
+       } else {
+               return 'text/html';
+       }
+}
+
+sub print_feed_meta {
+       if (defined $project) {
+               my %href_params = get_feed_info();
+               if (!exists $href_params{'-title'}) {
+                       $href_params{'-title'} = 'log';
+               }
+
+               foreach my $format (qw(RSS Atom)) {
+                       my $type = lc($format);
+                       my %link_attr = (
+                               '-rel' => 'alternate',
+                               '-title' => esc_attr("$project - $href_params{'-title'} - $format feed"),
+                               '-type' => "application/$type+xml"
+                       );
+
+                       $href_params{'extra_options'} = undef;
+                       $href_params{'action'} = $type;
+                       $link_attr{'-href'} = href(%href_params);
+                       print "<link ".
+                             "rel=\"$link_attr{'-rel'}\" ".
+                             "title=\"$link_attr{'-title'}\" ".
+                             "href=\"$link_attr{'-href'}\" ".
+                             "type=\"$link_attr{'-type'}\" ".
+                             "/>\n";
+
+                       $href_params{'extra_options'} = '--no-merges';
+                       $link_attr{'-href'} = href(%href_params);
+                       $link_attr{'-title'} .= ' (no merges)';
+                       print "<link ".
+                             "rel=\"$link_attr{'-rel'}\" ".
+                             "title=\"$link_attr{'-title'}\" ".
+                             "href=\"$link_attr{'-href'}\" ".
+                             "type=\"$link_attr{'-type'}\" ".
+                             "/>\n";
+               }
+
+       } else {
+               printf('<link rel="alternate" title="%s projects list" '.
+                      'href="%s" type="text/plain; charset=utf-8" />'."\n",
+                      esc_attr($site_name), href(project=>undef, action=>"project_index"));
+               printf('<link rel="alternate" title="%s projects feeds" '.
+                      'href="%s" type="text/x-opml" />'."\n",
+                      esc_attr($site_name), href(project=>undef, action=>"opml"));
+       }
+}
+
+sub print_header_links {
+       my $status = shift;
+
+       # print out each stylesheet that exist, providing backwards capability
+       # for those people who defined $stylesheet in a config file
+       if (defined $stylesheet) {
+               print '<link rel="stylesheet" type="text/css" href="'.esc_url($stylesheet).'"/>'."\n";
+       } else {
+               foreach my $stylesheet (@stylesheets) {
+                       next unless $stylesheet;
+                       print '<link rel="stylesheet" type="text/css" href="'.esc_url($stylesheet).'"/>'."\n";
+               }
+       }
+       print_feed_meta()
+               if ($status eq '200 OK');
+       if (defined $favicon) {
+               print qq(<link rel="shortcut icon" href=").esc_url($favicon).qq(" type="image/png" />\n);
+       }
+}
+
+sub print_nav_breadcrumbs_path {
+       my $dirprefix = undef;
+       while (my $part = shift) {
+               $dirprefix .= "/" if defined $dirprefix;
+               $dirprefix .= $part;
+               print $cgi->a({-href => href(project => undef,
+                                            project_filter => $dirprefix,
+                                            action => "project_list")},
+                             esc_html($part)) . " / ";
+       }
+}
+
+sub print_nav_breadcrumbs {
+       my %opts = @_;
+
+       for my $crumb (@extra_breadcrumbs, [ $home_link_str => $home_link ]) {
+               print $cgi->a({-href => esc_url($crumb->[1])}, $crumb->[0]) . " / ";
+       }
+       if (defined $project) {
+               my @dirname = split '/', $project;
+               my $projectbasename = pop @dirname;
+               print_nav_breadcrumbs_path(@dirname);
+               print $cgi->a({-href => href(action=>"summary")}, esc_html($projectbasename));
+               if (defined $action) {
+                       my $action_print = $action ;
+                       if (defined $opts{-action_extra}) {
+                               $action_print = $cgi->a({-href => href(action=>$action)},
+                                       $action);
+                       }
+                       print " / $action_print";
+               }
+               if (defined $opts{-action_extra}) {
+                       print " / $opts{-action_extra}";
+               }
+               print "\n";
+       } elsif (defined $project_filter) {
+               print_nav_breadcrumbs_path(split '/', $project_filter);
+       }
+}
+
+sub print_search_form {
+       if (!defined $searchtext) {
+               $searchtext = "";
+       }
+       my $search_hash;
+       if (defined $hash_base) {
+               $search_hash = $hash_base;
+       } elsif (defined $hash) {
+               $search_hash = $hash;
+       } else {
+               $search_hash = "HEAD";
+       }
+       my $action = $my_uri;
+       my $use_pathinfo = gitweb_check_feature('pathinfo');
+       if ($use_pathinfo) {
+               $action .= "/".esc_url($project);
+       }
+       print $cgi->start_form(-method => "get", -action => $action) .
+             "<div class=\"search\">\n" .
+             (!$use_pathinfo &&
+             $cgi->input({-name=>"p", -value=>$project, -type=>"hidden"}) . "\n") .
+             $cgi->input({-name=>"a", -value=>"search", -type=>"hidden"}) . "\n" .
+             $cgi->input({-name=>"h", -value=>$search_hash, -type=>"hidden"}) . "\n" .
+             $cgi->popup_menu(-name => 'st', -default => 'commit',
+                              -values => ['commit', 'grep', 'author', 'committer', 'pickaxe']) .
+             " " . $cgi->a({-href => href(action=>"search_help"),
+                            -title => "search help" }, "?") . " search:\n",
+             $cgi->textfield(-name => "s", -value => $searchtext, -override => 1) . "\n" .
+             "<span title=\"Extended regular expression\">" .
+             $cgi->checkbox(-name => 'sr', -value => 1, -label => 're',
+                            -checked => $search_use_regexp) .
+             "</span>" .
+             "</div>" .
+             $cgi->end_form() . "\n";
+}
+
+sub git_header_html {
+       my $status = shift || "200 OK";
+       my $expires = shift;
+       my %opts = @_;
+
+       my $title = get_page_title();
+       my $content_type = get_content_type_html();
+       print $cgi->header(-type=>$content_type, -charset => 'utf-8',
+                          -status=> $status, -expires => $expires)
+               unless ($opts{'-no_http_header'});
+       my $mod_perl_version = $ENV{'MOD_PERL'} ? " $ENV{'MOD_PERL'}" : '';
+       print <<EOF;
+<?xml version="1.0" encoding="utf-8"?>
+<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en-US" lang="en-US">
+<!-- git web interface version $version, (C) 2005-2006, Kay Sievers <kay.sievers\@vrfy.org>, Christian Gierke -->
+<!-- git core binaries version $git_version -->
+<head>
+<meta http-equiv="content-type" content="$content_type; charset=utf-8"/>
+<meta name="generator" content="gitweb/$version git/$git_version$mod_perl_version"/>
+<meta name="robots" content="index, nofollow"/>
+<title>$title</title>
+EOF
+       # the stylesheet, favicon etc urls won't work correctly with path_info
+       # unless we set the appropriate base URL
+       if ($ENV{'PATH_INFO'}) {
+               print "<base href=\"".esc_url($base_url)."\" />\n";
+       }
+       print_header_links($status);
+
+       if (defined $site_html_head_string) {
+               print to_utf8($site_html_head_string);
+       }
+
+       print "</head>\n" .
+             "<body>\n";
+
+       if (defined $site_header && -f $site_header) {
+               insert_file($site_header);
+       }
+
+       print "<div class=\"page_header\">\n";
+       if (defined $logo) {
+               print $cgi->a({-href => esc_url($logo_url),
+                              -title => $logo_label},
+                             $cgi->img({-src => esc_url($logo),
+                                        -width => 72, -height => 27,
+                                        -alt => "git",
+                                        -class => "logo"}));
+       }
+       print_nav_breadcrumbs(%opts);
+       print "</div>\n";
+
+       my $have_search = gitweb_check_feature('search');
+       if (defined $project && $have_search) {
+               print_search_form();
+       }
+}
+
+sub git_footer_html {
+       my $feed_class = 'rss_logo';
+
+       print "<div class=\"page_footer\">\n";
+       if (defined $project) {
+               my $descr = git_get_project_description($project);
+               if (defined $descr) {
+                       print "<div class=\"page_footer_text\">" . esc_html($descr) . "</div>\n";
+               }
+
+               my %href_params = get_feed_info();
+               if (!%href_params) {
+                       $feed_class .= ' generic';
+               }
+               $href_params{'-title'} ||= 'log';
+
+               foreach my $format (qw(RSS Atom)) {
+                       $href_params{'action'} = lc($format);
+                       print $cgi->a({-href => href(%href_params),
+                                     -title => "$href_params{'-title'} $format feed",
+                                     -class => $feed_class}, $format)."\n";
+               }
+
+       } else {
+               print $cgi->a({-href => href(project=>undef, action=>"opml",
+                                            project_filter => $project_filter),
+                             -class => $feed_class}, "OPML") . " ";
+               print $cgi->a({-href => href(project=>undef, action=>"project_index",
+                                            project_filter => $project_filter),
+                             -class => $feed_class}, "TXT") . "\n";
+       }
+       print "</div>\n"; # class="page_footer"
+
+       if (defined $t0 && gitweb_check_feature('timed')) {
+               print "<div id=\"generating_info\">\n";
+               print 'This page took '.
+                     '<span id="generating_time" class="time_span">'.
+                     tv_interval($t0, [ gettimeofday() ]).
+                     ' seconds </span>'.
+                     ' and '.
+                     '<span id="generating_cmd">'.
+                     $number_of_git_cmds.
+                     '</span> git commands '.
+                     " to generate.\n";
+               print "</div>\n"; # class="page_footer"
+       }
+
+       if (defined $site_footer && -f $site_footer) {
+               insert_file($site_footer);
+       }
+
+       print qq!<script type="text/javascript" src="!.esc_url($javascript).qq!"></script>\n!;
+       if (defined $action &&
+           $action eq 'blame_incremental') {
+               print qq!<script type="text/javascript">\n!.
+                     qq!startBlame("!. href(action=>"blame_data", -replay=>1) .qq!",\n!.
+                     qq!           "!. href() .qq!");\n!.
+                     qq!</script>\n!;
+       } else {
+               my ($jstimezone, $tz_cookie, $datetime_class) =
+                       gitweb_get_feature('javascript-timezone');
+
+               print qq!<script type="text/javascript">\n!.
+                     qq!window.onload = function () {\n!;
+               if (gitweb_check_feature('javascript-actions')) {
+                       print qq!       fixLinks();\n!;
+               }
+               if ($jstimezone && $tz_cookie && $datetime_class) {
+                       print qq!       var tz_cookie = { name: '$tz_cookie', expires: 14, path: '/' };\n!. # in days
+                             qq!       onloadTZSetup('$jstimezone', tz_cookie, '$datetime_class');\n!;
+               }
+               print qq!};\n!.
+                     qq!</script>\n!;
+       }
+
+       print "</body>\n" .
+             "</html>";
+}
+
+# die_error(<http_status_code>, <error_message>[, <detailed_html_description>])
+# Example: die_error(404, 'Hash not found')
+# By convention, use the following status codes (as defined in RFC 2616):
+# 400: Invalid or missing CGI parameters, or
+#      requested object exists but has wrong type.
+# 403: Requested feature (like "pickaxe" or "snapshot") not enabled on
+#      this server or project.
+# 404: Requested object/revision/project doesn't exist.
+# 500: The server isn't configured properly, or
+#      an internal error occurred (e.g. failed assertions caused by bugs), or
+#      an unknown error occurred (e.g. the git binary died unexpectedly).
+# 503: The server is currently unavailable (because it is overloaded,
+#      or down for maintenance).  Generally, this is a temporary state.
+sub die_error {
+       my $status = shift || 500;
+       my $error = esc_html(shift) || "Internal Server Error";
+       my $extra = shift;
+       my %opts = @_;
+
+       my %http_responses = (
+               400 => '400 Bad Request',
+               403 => '403 Forbidden',
+               404 => '404 Not Found',
+               500 => '500 Internal Server Error',
+               503 => '503 Service Unavailable',
+       );
+       git_header_html($http_responses{$status}, undef, %opts);
+       print <<EOF;
+<div class="page_body">
+<br /><br />
+$status - $error
+<br />
+EOF
+       if (defined $extra) {
+               print "<hr />\n" .
+                     "$extra\n";
+       }
+       print "</div>\n";
+
+       git_footer_html();
+       goto DONE_GITWEB
+               unless ($opts{'-error_handler'});
+}
+
+## ----------------------------------------------------------------------
+## functions printing or outputting HTML: navigation
+
+sub git_print_page_nav {
+       my ($current, $suppress, $head, $treehead, $treebase, $extra) = @_;
+       $extra = '' if !defined $extra; # pager or formats
+
+       my @navs = qw(summary shortlog log commit commitdiff tree);
+       if ($suppress) {
+               @navs = grep { $_ ne $suppress } @navs;
+       }
+
+       my %arg = map { $_ => {action=>$_} } @navs;
+       if (defined $head) {
+               for (qw(commit commitdiff)) {
+                       $arg{$_}{'hash'} = $head;
+               }
+               if ($current =~ m/^(tree | log | shortlog | commit | commitdiff | search)$/x) {
+                       for (qw(shortlog log)) {
+                               $arg{$_}{'hash'} = $head;
+                       }
+               }
+       }
+
+       $arg{'tree'}{'hash'} = $treehead if defined $treehead;
+       $arg{'tree'}{'hash_base'} = $treebase if defined $treebase;
+
+       my @actions = gitweb_get_feature('actions');
+       my %repl = (
+               '%' => '%',
+               'n' => $project,         # project name
+               'f' => $git_dir,         # project path within filesystem
+               'h' => $treehead || '',  # current hash ('h' parameter)
+               'b' => $treebase || '',  # hash base ('hb' parameter)
+       );
+       while (@actions) {
+               my ($label, $link, $pos) = splice(@actions,0,3);
+               # insert
+               @navs = map { $_ eq $pos ? ($_, $label) : $_ } @navs;
+               # munch munch
+               $link =~ s/%([%nfhb])/$repl{$1}/g;
+               $arg{$label}{'_href'} = $link;
+       }
+
+       print "<div class=\"page_nav\">\n" .
+               (join " | ",
+                map { $_ eq $current ?
+                      $_ : $cgi->a({-href => ($arg{$_}{_href} ? $arg{$_}{_href} : href(%{$arg{$_}}))}, "$_")
+                } @navs);
+       print "<br/>\n$extra<br/>\n" .
+             "</div>\n";
+}
+
+# returns a submenu for the nagivation of the refs views (tags, heads,
+# remotes) with the current view disabled and the remotes view only
+# available if the feature is enabled
+sub format_ref_views {
+       my ($current) = @_;
+       my @ref_views = qw{tags heads};
+       push @ref_views, 'remotes' if gitweb_check_feature('remote_heads');
+       return join " | ", map {
+               $_ eq $current ? $_ :
+               $cgi->a({-href => href(action=>$_)}, $_)
+       } @ref_views
+}
+
+sub format_paging_nav {
+       my ($action, $page, $has_next_link) = @_;
+       my $paging_nav;
+
+
+       if ($page > 0) {
+               $paging_nav .=
+                       $cgi->a({-href => href(-replay=>1, page=>undef)}, "first") .
+                       " &sdot; " .
+                       $cgi->a({-href => href(-replay=>1, page=>$page-1),
+                                -accesskey => "p", -title => "Alt-p"}, "prev");
+       } else {
+               $paging_nav .= "first &sdot; prev";
+       }
+
+       if ($has_next_link) {
+               $paging_nav .= " &sdot; " .
+                       $cgi->a({-href => href(-replay=>1, page=>$page+1),
+                                -accesskey => "n", -title => "Alt-n"}, "next");
+       } else {
+               $paging_nav .= " &sdot; next";
+       }
+
+       return $paging_nav;
+}
+
+## ......................................................................
+## functions printing or outputting HTML: div
+
+sub git_print_header_div {
+       my ($action, $title, $hash, $hash_base) = @_;
+       my %args = ();
+
+       $args{'action'} = $action;
+       $args{'hash'} = $hash if $hash;
+       $args{'hash_base'} = $hash_base if $hash_base;
+
+       print "<div class=\"header\">\n" .
+             $cgi->a({-href => href(%args), -class => "title"},
+             $title ? $title : $action) .
+             "\n</div>\n";
+}
+
+sub format_repo_url {
+       my ($name, $url) = @_;
+       return "<tr class=\"metadata_url\"><td>$name</td><td>$url</td></tr>\n";
+}
+
+# Group output by placing it in a DIV element and adding a header.
+# Options for start_div() can be provided by passing a hash reference as the
+# first parameter to the function.
+# Options to git_print_header_div() can be provided by passing an array
+# reference. This must follow the options to start_div if they are present.
+# The content can be a scalar, which is output as-is, a scalar reference, which
+# is output after html escaping, an IO handle passed either as *handle or
+# *handle{IO}, or a function reference. In the latter case all following
+# parameters will be taken as argument to the content function call.
+sub git_print_section {
+       my ($div_args, $header_args, $content);
+       my $arg = shift;
+       if (ref($arg) eq 'HASH') {
+               $div_args = $arg;
+               $arg = shift;
+       }
+       if (ref($arg) eq 'ARRAY') {
+               $header_args = $arg;
+               $arg = shift;
+       }
+       $content = $arg;
+
+       print $cgi->start_div($div_args);
+       git_print_header_div(@$header_args);
+
+       if (ref($content) eq 'CODE') {
+               $content->(@_);
+       } elsif (ref($content) eq 'SCALAR') {
+               print esc_html($$content);
+       } elsif (ref($content) eq 'GLOB' or ref($content) eq 'IO::Handle') {
+               print <$content>;
+       } elsif (!ref($content) && defined($content)) {
+               print $content;
+       }
+
+       print $cgi->end_div;
+}
+
+sub format_timestamp_html {
+       my $date = shift;
+       my $strtime = $date->{'rfc2822'};
+
+       my (undef, undef, $datetime_class) =
+               gitweb_get_feature('javascript-timezone');
+       if ($datetime_class) {
+               $strtime = qq!<span class="$datetime_class">$strtime</span>!;
+       }
+
+       my $localtime_format = '(%02d:%02d %s)';
+       if ($date->{'hour_local'} < 6) {
+               $localtime_format = '(<span class="atnight">%02d:%02d</span> %s)';
+       }
+       $strtime .= ' ' .
+                   sprintf($localtime_format,
+                           $date->{'hour_local'}, $date->{'minute_local'}, $date->{'tz_local'});
+
+       return $strtime;
+}
+
+# Outputs the author name and date in long form
+sub git_print_authorship {
+       my $co = shift;
+       my %opts = @_;
+       my $tag = $opts{-tag} || 'div';
+       my $author = $co->{'author_name'};
+
+       my %ad = parse_date($co->{'author_epoch'}, $co->{'author_tz'});
+       print "<$tag class=\"author_date\">" .
+             format_search_author($author, "author", esc_html($author)) .
+             " [".format_timestamp_html(\%ad)."]".
+             git_get_avatar($co->{'author_email'}, -pad_before => 1) .
+             "</$tag>\n";
+}
+
+# Outputs table rows containing the full author or committer information,
+# in the format expected for 'commit' view (& similar).
+# Parameters are a commit hash reference, followed by the list of people
+# to output information for. If the list is empty it defaults to both
+# author and committer.
+sub git_print_authorship_rows {
+       my $co = shift;
+       # too bad we can't use @people = @_ || ('author', 'committer')
+       my @people = @_;
+       @people = ('author', 'committer') unless @people;
+       foreach my $who (@people) {
+               my %wd = parse_date($co->{"${who}_epoch"}, $co->{"${who}_tz"});
+               print "<tr><td>$who</td><td>" .
+                     format_search_author($co->{"${who}_name"}, $who,
+                                          esc_html($co->{"${who}_name"})) . " " .
+                     format_search_author($co->{"${who}_email"}, $who,
+                                          esc_html("<" . $co->{"${who}_email"} . ">")) .
+                     "</td><td rowspan=\"2\">" .
+                     git_get_avatar($co->{"${who}_email"}, -size => 'double') .
+                     "</td></tr>\n" .
+                     "<tr>" .
+                     "<td></td><td>" .
+                     format_timestamp_html(\%wd) .
+                     "</td>" .
+                     "</tr>\n";
+       }
+}
+
+sub git_print_page_path {
+       my $name = shift;
+       my $type = shift;
+       my $hb = shift;
+
+
+       print "<div class=\"page_path\">";
+       print $cgi->a({-href => href(action=>"tree", hash_base=>$hb),
+                     -title => 'tree root'}, to_utf8("[$project]"));
+       print " / ";
+       if (defined $name) {
+               my @dirname = split '/', $name;
+               my $basename = pop @dirname;
+               my $fullname = '';
+
+               foreach my $dir (@dirname) {
+                       $fullname .= ($fullname ? '/' : '') . $dir;
+                       print $cgi->a({-href => href(action=>"tree", file_name=>$fullname,
+                                                    hash_base=>$hb),
+                                     -title => $fullname}, esc_path($dir));
+                       print " / ";
+               }
+               if (defined $type && $type eq 'blob') {
+                       print $cgi->a({-href => href(action=>"blob_plain", file_name=>$file_name,
+                                                    hash_base=>$hb),
+                                     -title => $name}, esc_path($basename));
+               } elsif (defined $type && $type eq 'tree') {
+                       print $cgi->a({-href => href(action=>"tree", file_name=>$file_name,
+                                                    hash_base=>$hb),
+                                     -title => $name}, esc_path($basename));
+                       print " / ";
+               } else {
+                       print esc_path($basename);
+               }
+       }
+       print "<br/></div>\n";
+}
+
+sub git_print_log {
+       my $log = shift;
+       my %opts = @_;
+
+       if ($opts{'-remove_title'}) {
+               # remove title, i.e. first line of log
+               shift @$log;
+       }
+       # remove leading empty lines
+       while (defined $log->[0] && $log->[0] eq "") {
+               shift @$log;
+       }
+
+       # print log
+       my $skip_blank_line = 0;
+       foreach my $line (@$log) {
+               if ($line =~ m/^\s*([A-Z][-A-Za-z]*-[Bb]y|C[Cc]): /) {
+                       if (! $opts{'-remove_signoff'}) {
+                               print "<span class=\"signoff\">" . esc_html($line) . "</span><br/>\n";
+                               $skip_blank_line = 1;
+                       }
+                       next;
+               }
+
+               if ($line =~ m,\s*([a-z]*link): (https?://\S+),i) {
+                       if (! $opts{'-remove_signoff'}) {
+                               print "<span class=\"signoff\">" . esc_html($1) . ": " .
+                                       "<a href=\"" . esc_html($2) . "\">" . esc_html($2) . "</a>" .
+                                       "</span><br/>\n";
+                               $skip_blank_line = 1;
+                       }
+                       next;
+               }
+
+               # print only one empty line
+               # do not print empty line after signoff
+               if ($line eq "") {
+                       next if ($skip_blank_line);
+                       $skip_blank_line = 1;
+               } else {
+                       $skip_blank_line = 0;
+               }
+
+               print format_log_line_html($line) . "<br/>\n";
+       }
+
+       if ($opts{'-final_empty_line'}) {
+               # end with single empty line
+               print "<br/>\n" unless $skip_blank_line;
+       }
+}
+
+# return link target (what link points to)
+sub git_get_link_target {
+       my $hash = shift;
+       my $link_target;
+
+       # read link
+       open my $fd, "-|", git_cmd(), "cat-file", "blob", $hash
+               or return;
+       {
+               local $/ = undef;
+               $link_target = <$fd>;
+       }
+       close $fd
+               or return;
+
+       return $link_target;
+}
+
+# given link target, and the directory (basedir) the link is in,
+# return target of link relative to top directory (top tree);
+# return undef if it is not possible (including absolute links).
+sub normalize_link_target {
+       my ($link_target, $basedir) = @_;
+
+       # absolute symlinks (beginning with '/') cannot be normalized
+       return if (substr($link_target, 0, 1) eq '/');
+
+       # normalize link target to path from top (root) tree (dir)
+       my $path;
+       if ($basedir) {
+               $path = $basedir . '/' . $link_target;
+       } else {
+               # we are in top (root) tree (dir)
+               $path = $link_target;
+       }
+
+       # remove //, /./, and /../
+       my @path_parts;
+       foreach my $part (split('/', $path)) {
+               # discard '.' and ''
+               next if (!$part || $part eq '.');
+               # handle '..'
+               if ($part eq '..') {
+                       if (@path_parts) {
+                               pop @path_parts;
+                       } else {
+                               # link leads outside repository (outside top dir)
+                               return;
+                       }
+               } else {
+                       push @path_parts, $part;
+               }
+       }
+       $path = join('/', @path_parts);
+
+       return $path;
+}
+
+# print tree entry (row of git_tree), but without encompassing <tr> element
+sub git_print_tree_entry {
+       my ($t, $basedir, $hash_base, $have_blame) = @_;
+
+       my %base_key = ();
+       $base_key{'hash_base'} = $hash_base if defined $hash_base;
+
+       # The format of a table row is: mode list link.  Where mode is
+       # the mode of the entry, list is the name of the entry, an href,
+       # and link is the action links of the entry.
+
+       print "<td class=\"mode\">" . mode_str($t->{'mode'}) . "</td>\n";
+       if (exists $t->{'size'}) {
+               print "<td class=\"size\">$t->{'size'}</td>\n";
+       }
+       if ($t->{'type'} eq "blob") {
+               print "<td class=\"list\">" .
+                       $cgi->a({-href => href(action=>"blob", hash=>$t->{'hash'},
+                                              file_name=>"$basedir$t->{'name'}", %base_key),
+                               -class => "list"}, esc_path($t->{'name'}));
+               if (S_ISLNK(oct $t->{'mode'})) {
+                       my $link_target = git_get_link_target($t->{'hash'});
+                       if ($link_target) {
+                               my $norm_target = normalize_link_target($link_target, $basedir);
+                               if (defined $norm_target) {
+                                       print " -> " .
+                                             $cgi->a({-href => href(action=>"object", hash_base=>$hash_base,
+                                                                    file_name=>$norm_target),
+                                                      -title => $norm_target}, esc_path($link_target));
+                               } else {
+                                       print " -> " . esc_path($link_target);
+                               }
+                       }
+               }
+               print "</td>\n";
+               print "<td class=\"link\">";
+               print $cgi->a({-href => href(action=>"blob", hash=>$t->{'hash'},
+                                            file_name=>"$basedir$t->{'name'}", %base_key)},
+                             "blob");
+               if ($have_blame) {
+                       print " | " .
+                             $cgi->a({-href => href(action=>"blame", hash=>$t->{'hash'},
+                                                    file_name=>"$basedir$t->{'name'}", %base_key)},
+                                     "blame");
+               }
+               if (defined $hash_base) {
+                       print " | " .
+                             $cgi->a({-href => href(action=>"history", hash_base=>$hash_base,
+                                                    hash=>$t->{'hash'}, file_name=>"$basedir$t->{'name'}")},
+                                     "history");
+               }
+               print " | " .
+                       $cgi->a({-href => href(action=>"blob_plain", hash_base=>$hash_base,
+                                              file_name=>"$basedir$t->{'name'}")},
+                               "raw");
+               print "</td>\n";
+
+       } elsif ($t->{'type'} eq "tree") {
+               print "<td class=\"list\">";
+               print $cgi->a({-href => href(action=>"tree", hash=>$t->{'hash'},
+                                            file_name=>"$basedir$t->{'name'}",
+                                            %base_key)},
+                             esc_path($t->{'name'}));
+               print "</td>\n";
+               print "<td class=\"link\">";
+               print $cgi->a({-href => href(action=>"tree", hash=>$t->{'hash'},
+                                            file_name=>"$basedir$t->{'name'}",
+                                            %base_key)},
+                             "tree");
+               if (defined $hash_base) {
+                       print " | " .
+                             $cgi->a({-href => href(action=>"history", hash_base=>$hash_base,
+                                                    file_name=>"$basedir$t->{'name'}")},
+                                     "history");
+               }
+               print "</td>\n";
+       } else {
+               # unknown object: we can only present history for it
+               # (this includes 'commit' object, i.e. submodule support)
+               print "<td class=\"list\">" .
+                     esc_path($t->{'name'}) .
+                     "</td>\n";
+               print "<td class=\"link\">";
+               if (defined $hash_base) {
+                       print $cgi->a({-href => href(action=>"history",
+                                                    hash_base=>$hash_base,
+                                                    file_name=>"$basedir$t->{'name'}")},
+                                     "history");
+               }
+               print "</td>\n";
+       }
+}
+
+## ......................................................................
+## functions printing large fragments of HTML
+
+# get pre-image filenames for merge (combined) diff
+sub fill_from_file_info {
+       my ($diff, @parents) = @_;
+
+       $diff->{'from_file'} = [ ];
+       $diff->{'from_file'}[$diff->{'nparents'} - 1] = undef;
+       for (my $i = 0; $i < $diff->{'nparents'}; $i++) {
+               if ($diff->{'status'}[$i] eq 'R' ||
+                   $diff->{'status'}[$i] eq 'C') {
+                       $diff->{'from_file'}[$i] =
+                               git_get_path_by_hash($parents[$i], $diff->{'from_id'}[$i]);
+               }
+       }
+
+       return $diff;
+}
+
+# is current raw difftree line of file deletion
+sub is_deleted {
+       my $diffinfo = shift;
+
+       return $diffinfo->{'to_id'} eq ('0' x 40);
+}
+
+# does patch correspond to [previous] difftree raw line
+# $diffinfo  - hashref of parsed raw diff format
+# $patchinfo - hashref of parsed patch diff format
+#              (the same keys as in $diffinfo)
+sub is_patch_split {
+       my ($diffinfo, $patchinfo) = @_;
+
+       return defined $diffinfo && defined $patchinfo
+               && $diffinfo->{'to_file'} eq $patchinfo->{'to_file'};
+}
+
+
+sub git_difftree_body {
+       my ($difftree, $hash, @parents) = @_;
+       my ($parent) = $parents[0];
+       my $have_blame = gitweb_check_feature('blame');
+       print "<div class=\"list_head\">\n";
+       if ($#{$difftree} > 10) {
+               print(($#{$difftree} + 1) . " files changed:\n");
+       }
+       print "</div>\n";
+
+       print "<table class=\"" .
+             (@parents > 1 ? "combined " : "") .
+             "diff_tree\">\n";
+
+       # header only for combined diff in 'commitdiff' view
+       my $has_header = @$difftree && @parents > 1 && $action eq 'commitdiff';
+       if ($has_header) {
+               # table header
+               print "<thead><tr>\n" .
+                      "<th></th><th></th>\n"; # filename, patchN link
+               for (my $i = 0; $i < @parents; $i++) {
+                       my $par = $parents[$i];
+                       print "<th>" .
+                             $cgi->a({-href => href(action=>"commitdiff",
+                                                    hash=>$hash, hash_parent=>$par),
+                                      -title => 'commitdiff to parent number ' .
+                                                 ($i+1) . ': ' . substr($par,0,7)},
+                                     $i+1) .
+                             "&nbsp;</th>\n";
+               }
+               print "</tr></thead>\n<tbody>\n";
+       }
+
+       my $alternate = 1;
+       my $patchno = 0;
+       foreach my $line (@{$difftree}) {
+               my $diff = parsed_difftree_line($line);
+
+               if ($alternate) {
+                       print "<tr class=\"dark\">\n";
+               } else {
+                       print "<tr class=\"light\">\n";
+               }
+               $alternate ^= 1;
+
+               if (exists $diff->{'nparents'}) { # combined diff
+
+                       fill_from_file_info($diff, @parents)
+                               unless exists $diff->{'from_file'};
+
+                       if (!is_deleted($diff)) {
+                               # file exists in the result (child) commit
+                               print "<td>" .
+                                     $cgi->a({-href => href(action=>"blob", hash=>$diff->{'to_id'},
+                                                            file_name=>$diff->{'to_file'},
+                                                            hash_base=>$hash),
+                                             -class => "list"}, esc_path($diff->{'to_file'})) .
+                                     "</td>\n";
+                       } else {
+                               print "<td>" .
+                                     esc_path($diff->{'to_file'}) .
+                                     "</td>\n";
+                       }
+
+                       if ($action eq 'commitdiff') {
+                               # link to patch
+                               $patchno++;
+                               print "<td class=\"link\">" .
+                                     $cgi->a({-href => href(-anchor=>"patch$patchno")},
+                                             "patch") .
+                                     " | " .
+                                     "</td>\n";
+                       }
+
+                       my $has_history = 0;
+                       my $not_deleted = 0;
+                       for (my $i = 0; $i < $diff->{'nparents'}; $i++) {
+                               my $hash_parent = $parents[$i];
+                               my $from_hash = $diff->{'from_id'}[$i];
+                               my $from_path = $diff->{'from_file'}[$i];
+                               my $status = $diff->{'status'}[$i];
+
+                               $has_history ||= ($status ne 'A');
+                               $not_deleted ||= ($status ne 'D');
+
+                               if ($status eq 'A') {
+                                       print "<td  class=\"link\" align=\"right\"> | </td>\n";
+                               } elsif ($status eq 'D') {
+                                       print "<td class=\"link\">" .
+                                             $cgi->a({-href => href(action=>"blob",
+                                                                    hash_base=>$hash,
+                                                                    hash=>$from_hash,
+                                                                    file_name=>$from_path)},
+                                                     "blob" . ($i+1)) .
+                                             " | </td>\n";
+                               } else {
+                                       if ($diff->{'to_id'} eq $from_hash) {
+                                               print "<td class=\"link nochange\">";
+                                       } else {
+                                               print "<td class=\"link\">";
+                                       }
+                                       print $cgi->a({-href => href(action=>"blobdiff",
+                                                                    hash=>$diff->{'to_id'},
+                                                                    hash_parent=>$from_hash,
+                                                                    hash_base=>$hash,
+                                                                    hash_parent_base=>$hash_parent,
+                                                                    file_name=>$diff->{'to_file'},
+                                                                    file_parent=>$from_path)},
+                                                     "diff" . ($i+1)) .
+                                             " | </td>\n";
+                               }
+                       }
+
+                       print "<td class=\"link\">";
+                       if ($not_deleted) {
+                               print $cgi->a({-href => href(action=>"blob",
+                                                            hash=>$diff->{'to_id'},
+                                                            file_name=>$diff->{'to_file'},
+                                                            hash_base=>$hash)},
+                                             "blob");
+                               print " | " if ($has_history);
+                       }
+                       if ($has_history) {
+                               print $cgi->a({-href => href(action=>"history",
+                                                            file_name=>$diff->{'to_file'},
+                                                            hash_base=>$hash)},
+                                             "history");
+                       }
+                       print "</td>\n";
+
+                       print "</tr>\n";
+                       next; # instead of 'else' clause, to avoid extra indent
+               }
+               # else ordinary diff
+
+               my ($to_mode_oct, $to_mode_str, $to_file_type);
+               my ($from_mode_oct, $from_mode_str, $from_file_type);
+               if ($diff->{'to_mode'} ne ('0' x 6)) {
+                       $to_mode_oct = oct $diff->{'to_mode'};
+                       if (S_ISREG($to_mode_oct)) { # only for regular file
+                               $to_mode_str = sprintf("%04o", $to_mode_oct & 0777); # permission bits
+                       }
+                       $to_file_type = file_type($diff->{'to_mode'});
+               }
+               if ($diff->{'from_mode'} ne ('0' x 6)) {
+                       $from_mode_oct = oct $diff->{'from_mode'};
+                       if (S_ISREG($from_mode_oct)) { # only for regular file
+                               $from_mode_str = sprintf("%04o", $from_mode_oct & 0777); # permission bits
+                       }
+                       $from_file_type = file_type($diff->{'from_mode'});
+               }
+
+               if ($diff->{'status'} eq "A") { # created
+                       my $mode_chng = "<span class=\"file_status new\">[new $to_file_type";
+                       $mode_chng   .= " with mode: $to_mode_str" if $to_mode_str;
+                       $mode_chng   .= "]</span>";
+                       print "<td>";
+                       print $cgi->a({-href => href(action=>"blob", hash=>$diff->{'to_id'},
+                                                    hash_base=>$hash, file_name=>$diff->{'file'}),
+                                     -class => "list"}, esc_path($diff->{'file'}));
+                       print "</td>\n";
+                       print "<td>$mode_chng</td>\n";
+                       print "<td class=\"link\">";
+                       if ($action eq 'commitdiff') {
+                               # link to patch
+                               $patchno++;
+                               print $cgi->a({-href => href(-anchor=>"patch$patchno")},
+                                             "patch") .
+                                     " | ";
+                       }
+                       print $cgi->a({-href => href(action=>"blob", hash=>$diff->{'to_id'},
+                                                    hash_base=>$hash, file_name=>$diff->{'file'})},
+                                     "blob");
+                       print "</td>\n";
+
+               } elsif ($diff->{'status'} eq "D") { # deleted
+                       my $mode_chng = "<span class=\"file_status deleted\">[deleted $from_file_type]</span>";
+                       print "<td>";
+                       print $cgi->a({-href => href(action=>"blob", hash=>$diff->{'from_id'},
+                                                    hash_base=>$parent, file_name=>$diff->{'file'}),
+                                      -class => "list"}, esc_path($diff->{'file'}));
+                       print "</td>\n";
+                       print "<td>$mode_chng</td>\n";
+                       print "<td class=\"link\">";
+                       if ($action eq 'commitdiff') {
+                               # link to patch
+                               $patchno++;
+                               print $cgi->a({-href => href(-anchor=>"patch$patchno")},
+                                             "patch") .
+                                     " | ";
+                       }
+                       print $cgi->a({-href => href(action=>"blob", hash=>$diff->{'from_id'},
+                                                    hash_base=>$parent, file_name=>$diff->{'file'})},
+                                     "blob") . " | ";
+                       if ($have_blame) {
+                               print $cgi->a({-href => href(action=>"blame", hash_base=>$parent,
+                                                            file_name=>$diff->{'file'})},
+                                             "blame") . " | ";
+                       }
+                       print $cgi->a({-href => href(action=>"history", hash_base=>$parent,
+                                                    file_name=>$diff->{'file'})},
+                                     "history");
+                       print "</td>\n";
+
+               } elsif ($diff->{'status'} eq "M" || $diff->{'status'} eq "T") { # modified, or type changed
+                       my $mode_chnge = "";
+                       if ($diff->{'from_mode'} != $diff->{'to_mode'}) {
+                               $mode_chnge = "<span class=\"file_status mode_chnge\">[changed";
+                               if ($from_file_type ne $to_file_type) {
+                                       $mode_chnge .= " from $from_file_type to $to_file_type";
+                               }
+                               if (($from_mode_oct & 0777) != ($to_mode_oct & 0777)) {
+                                       if ($from_mode_str && $to_mode_str) {
+                                               $mode_chnge .= " mode: $from_mode_str->$to_mode_str";
+                                       } elsif ($to_mode_str) {
+                                               $mode_chnge .= " mode: $to_mode_str";
+                                       }
+                               }
+                               $mode_chnge .= "]</span>\n";
+                       }
+                       print "<td>";
+                       print $cgi->a({-href => href(action=>"blob", hash=>$diff->{'to_id'},
+                                                    hash_base=>$hash, file_name=>$diff->{'file'}),
+                                     -class => "list"}, esc_path($diff->{'file'}));
+                       print "</td>\n";
+                       print "<td>$mode_chnge</td>\n";
+                       print "<td class=\"link\">";
+                       if ($action eq 'commitdiff') {
+                               # link to patch
+                               $patchno++;
+                               print $cgi->a({-href => href(-anchor=>"patch$patchno")},
+                                             "patch") .
+                                     " | ";
+                       } elsif ($diff->{'to_id'} ne $diff->{'from_id'}) {
+                               # "commit" view and modified file (not onlu mode changed)
+                               print $cgi->a({-href => href(action=>"blobdiff",
+                                                            hash=>$diff->{'to_id'}, hash_parent=>$diff->{'from_id'},
+                                                            hash_base=>$hash, hash_parent_base=>$parent,
+                                                            file_name=>$diff->{'file'})},
+                                             "diff") .
+                                     " | ";
+                       }
+                       print $cgi->a({-href => href(action=>"blob", hash=>$diff->{'to_id'},
+                                                    hash_base=>$hash, file_name=>$diff->{'file'})},
+                                      "blob") . " | ";
+                       if ($have_blame) {
+                               print $cgi->a({-href => href(action=>"blame", hash_base=>$hash,
+                                                            file_name=>$diff->{'file'})},
+                                             "blame") . " | ";
+                       }
+                       print $cgi->a({-href => href(action=>"history", hash_base=>$hash,
+                                                    file_name=>$diff->{'file'})},
+                                     "history");
+                       print "</td>\n";
+
+               } elsif ($diff->{'status'} eq "R" || $diff->{'status'} eq "C") { # renamed or copied
+                       my %status_name = ('R' => 'moved', 'C' => 'copied');
+                       my $nstatus = $status_name{$diff->{'status'}};
+                       my $mode_chng = "";
+                       if ($diff->{'from_mode'} != $diff->{'to_mode'}) {
+                               # mode also for directories, so we cannot use $to_mode_str
+                               $mode_chng = sprintf(", mode: %04o", $to_mode_oct & 0777);
+                       }
+                       print "<td>" .
+                             $cgi->a({-href => href(action=>"blob", hash_base=>$hash,
+                                                    hash=>$diff->{'to_id'}, file_name=>$diff->{'to_file'}),
+                                     -class => "list"}, esc_path($diff->{'to_file'})) . "</td>\n" .
+                             "<td><span class=\"file_status $nstatus\">[$nstatus from " .
+                             $cgi->a({-href => href(action=>"blob", hash_base=>$parent,
+                                                    hash=>$diff->{'from_id'}, file_name=>$diff->{'from_file'}),
+                                     -class => "list"}, esc_path($diff->{'from_file'})) .
+                             " with " . (int $diff->{'similarity'}) . "% similarity$mode_chng]</span></td>\n" .
+                             "<td class=\"link\">";
+                       if ($action eq 'commitdiff') {
+                               # link to patch
+                               $patchno++;
+                               print $cgi->a({-href => href(-anchor=>"patch$patchno")},
+                                             "patch") .
+                                     " | ";
+                       } elsif ($diff->{'to_id'} ne $diff->{'from_id'}) {
+                               # "commit" view and modified file (not only pure rename or copy)
+                               print $cgi->a({-href => href(action=>"blobdiff",
+                                                            hash=>$diff->{'to_id'}, hash_parent=>$diff->{'from_id'},
+                                                            hash_base=>$hash, hash_parent_base=>$parent,
+                                                            file_name=>$diff->{'to_file'}, file_parent=>$diff->{'from_file'})},
+                                             "diff") .
+                                     " | ";
+                       }
+                       print $cgi->a({-href => href(action=>"blob", hash=>$diff->{'to_id'},
+                                                    hash_base=>$parent, file_name=>$diff->{'to_file'})},
+                                     "blob") . " | ";
+                       if ($have_blame) {
+                               print $cgi->a({-href => href(action=>"blame", hash_base=>$hash,
+                                                            file_name=>$diff->{'to_file'})},
+                                             "blame") . " | ";
+                       }
+                       print $cgi->a({-href => href(action=>"history", hash_base=>$hash,
+                                                   file_name=>$diff->{'to_file'})},
+                                     "history");
+                       print "</td>\n";
+
+               } # we should not encounter Unmerged (U) or Unknown (X) status
+               print "</tr>\n";
+       }
+       print "</tbody>" if $has_header;
+       print "</table>\n";
+}
+
+# Print context lines and then rem/add lines in a side-by-side manner.
+sub print_sidebyside_diff_lines {
+       my ($ctx, $rem, $add) = @_;
+
+       # print context block before add/rem block
+       if (@$ctx) {
+               print join '',
+                       '<div class="chunk_block ctx">',
+                               '<div class="old">',
+                               @$ctx,
+                               '</div>',
+                               '<div class="new">',
+                               @$ctx,
+                               '</div>',
+                       '</div>';
+       }
+
+       if (!@$add) {
+               # pure removal
+               print join '',
+                       '<div class="chunk_block rem">',
+                               '<div class="old">',
+                               @$rem,
+                               '</div>',
+                       '</div>';
+       } elsif (!@$rem) {
+               # pure addition
+               print join '',
+                       '<div class="chunk_block add">',
+                               '<div class="new">',
+                               @$add,
+                               '</div>',
+                       '</div>';
+       } else {
+               print join '',
+                       '<div class="chunk_block chg">',
+                               '<div class="old">',
+                               @$rem,
+                               '</div>',
+                               '<div class="new">',
+                               @$add,
+                               '</div>',
+                       '</div>';
+       }
+}
+
+# Print context lines and then rem/add lines in inline manner.
+sub print_inline_diff_lines {
+       my ($ctx, $rem, $add) = @_;
+
+       print @$ctx, @$rem, @$add;
+}
+
+# Format removed and added line, mark changed part and HTML-format them.
+# Implementation is based on contrib/diff-highlight
+sub format_rem_add_lines_pair {
+       my ($rem, $add, $num_parents) = @_;
+
+       # We need to untabify lines before split()'ing them;
+       # otherwise offsets would be invalid.
+       chomp $rem;
+       chomp $add;
+       $rem = untabify($rem);
+       $add = untabify($add);
+
+       my @rem = split(//, $rem);
+       my @add = split(//, $add);
+       my ($esc_rem, $esc_add);
+       # Ignore leading +/- characters for each parent.
+       my ($prefix_len, $suffix_len) = ($num_parents, 0);
+       my ($prefix_has_nonspace, $suffix_has_nonspace);
+
+       my $shorter = (@rem < @add) ? @rem : @add;
+       while ($prefix_len < $shorter) {
+               last if ($rem[$prefix_len] ne $add[$prefix_len]);
+
+               $prefix_has_nonspace = 1 if ($rem[$prefix_len] !~ /\s/);
+               $prefix_len++;
+       }
+
+       while ($prefix_len + $suffix_len < $shorter) {
+               last if ($rem[-1 - $suffix_len] ne $add[-1 - $suffix_len]);
+
+               $suffix_has_nonspace = 1 if ($rem[-1 - $suffix_len] !~ /\s/);
+               $suffix_len++;
+       }
+
+       # Mark lines that are different from each other, but have some common
+       # part that isn't whitespace.  If lines are completely different, don't
+       # mark them because that would make output unreadable, especially if
+       # diff consists of multiple lines.
+       if ($prefix_has_nonspace || $suffix_has_nonspace) {
+               $esc_rem = esc_html_hl_regions($rem, 'marked',
+                       [$prefix_len, @rem - $suffix_len], -nbsp=>1);
+               $esc_add = esc_html_hl_regions($add, 'marked',
+                       [$prefix_len, @add - $suffix_len], -nbsp=>1);
+       } else {
+               $esc_rem = esc_html($rem, -nbsp=>1);
+               $esc_add = esc_html($add, -nbsp=>1);
+       }
+
+       return format_diff_line(\$esc_rem, 'rem'),
+              format_diff_line(\$esc_add, 'add');
+}
+
+# HTML-format diff context, removed and added lines.
+sub format_ctx_rem_add_lines {
+       my ($ctx, $rem, $add, $num_parents) = @_;
+       my (@new_ctx, @new_rem, @new_add);
+       my $can_highlight = 0;
+       my $is_combined = ($num_parents > 1);
+
+       # Highlight if every removed line has a corresponding added line.
+       if (@$add > 0 && @$add == @$rem) {
+               $can_highlight = 1;
+
+               # Highlight lines in combined diff only if the chunk contains
+               # diff between the same version, e.g.
+               #
+               #    - a
+               #   -  b
+               #    + c
+               #   +  d
+               #
+               # Otherwise the highlightling would be confusing.
+               if ($is_combined) {
+                       for (my $i = 0; $i < @$add; $i++) {
+                               my $prefix_rem = substr($rem->[$i], 0, $num_parents);
+                               my $prefix_add = substr($add->[$i], 0, $num_parents);
+
+                               $prefix_rem =~ s/-/+/g;
+
+                               if ($prefix_rem ne $prefix_add) {
+                                       $can_highlight = 0;
+                                       last;
+                               }
+                       }
+               }
+       }
+
+       if ($can_highlight) {
+               for (my $i = 0; $i < @$add; $i++) {
+                       my ($line_rem, $line_add) = format_rem_add_lines_pair(
+                               $rem->[$i], $add->[$i], $num_parents);
+                       push @new_rem, $line_rem;
+                       push @new_add, $line_add;
+               }
+       } else {
+               @new_rem = map { format_diff_line($_, 'rem') } @$rem;
+               @new_add = map { format_diff_line($_, 'add') } @$add;
+       }
+
+       @new_ctx = map { format_diff_line($_, 'ctx') } @$ctx;
+
+       return (\@new_ctx, \@new_rem, \@new_add);
+}
+
+# Print context lines and then rem/add lines.
+sub print_diff_lines {
+       my ($ctx, $rem, $add, $diff_style, $num_parents) = @_;
+       my $is_combined = $num_parents > 1;
+
+       ($ctx, $rem, $add) = format_ctx_rem_add_lines($ctx, $rem, $add,
+               $num_parents);
+
+       if ($diff_style eq 'sidebyside' && !$is_combined) {
+               print_sidebyside_diff_lines($ctx, $rem, $add);
+       } else {
+               # default 'inline' style and unknown styles
+               print_inline_diff_lines($ctx, $rem, $add);
+       }
+}
+
+sub print_diff_chunk {
+       my ($diff_style, $num_parents, $from, $to, @chunk) = @_;
+       my (@ctx, @rem, @add);
+
+       # The class of the previous line.
+       my $prev_class = '';
+
+       return unless @chunk;
+
+       # incomplete last line might be among removed or added lines,
+       # or both, or among context lines: find which
+       for (my $i = 1; $i < @chunk; $i++) {
+               if ($chunk[$i][0] eq 'incomplete') {
+                       $chunk[$i][0] = $chunk[$i-1][0];
+               }
+       }
+
+       # guardian
+       push @chunk, ["", ""];
+
+       foreach my $line_info (@chunk) {
+               my ($class, $line) = @$line_info;
+
+               # print chunk headers
+               if ($class && $class eq 'chunk_header') {
+                       print format_diff_line($line, $class, $from, $to);
+                       next;
+               }
+
+               ## print from accumulator when have some add/rem lines or end
+               # of chunk (flush context lines), or when have add and rem
+               # lines and new block is reached (otherwise add/rem lines could
+               # be reordered)
+               if (!$class || ((@rem || @add) && $class eq 'ctx') ||
+                   (@rem && @add && $class ne $prev_class)) {
+                       print_diff_lines(\@ctx, \@rem, \@add,
+                                        $diff_style, $num_parents);
+                       @ctx = @rem = @add = ();
+               }
+
+               ## adding lines to accumulator
+               # guardian value
+               last unless $line;
+               # rem, add or change
+               if ($class eq 'rem') {
+                       push @rem, $line;
+               } elsif ($class eq 'add') {
+                       push @add, $line;
+               }
+               # context line
+               if ($class eq 'ctx') {
+                       push @ctx, $line;
+               }
+
+               $prev_class = $class;
+       }
+}
+
+sub git_patchset_body {
+       my ($fd, $diff_style, $difftree, $hash, @hash_parents) = @_;
+       my ($hash_parent) = $hash_parents[0];
+
+       my $is_combined = (@hash_parents > 1);
+       my $patch_idx = 0;
+       my $patch_number = 0;
+       my $patch_line;
+       my $diffinfo;
+       my $to_name;
+       my (%from, %to);
+       my @chunk; # for side-by-side diff
+
+       print "<div class=\"patchset\">\n";
+
+       # skip to first patch
+       while ($patch_line = <$fd>) {
+               chomp $patch_line;
+
+               last if ($patch_line =~ m/^diff /);
+       }
+
+ PATCH:
+       while ($patch_line) {
+
+               # parse "git diff" header line
+               if ($patch_line =~ m/^diff --git (\"(?:[^\\\"]*(?:\\.[^\\\"]*)*)\"|[^ "]*) (.*)$/) {
+                       # $1 is from_name, which we do not use
+                       $to_name = unquote($2);
+                       $to_name =~ s!^b/!!;
+               } elsif ($patch_line =~ m/^diff --(cc|combined) ("?.*"?)$/) {
+                       # $1 is 'cc' or 'combined', which we do not use
+                       $to_name = unquote($2);
+               } else {
+                       $to_name = undef;
+               }
+
+               # check if current patch belong to current raw line
+               # and parse raw git-diff line if needed
+               if (is_patch_split($diffinfo, { 'to_file' => $to_name })) {
+                       # this is continuation of a split patch
+                       print "<div class=\"patch cont\">\n";
+               } else {
+                       # advance raw git-diff output if needed
+                       $patch_idx++ if defined $diffinfo;
+
+                       # read and prepare patch information
+                       $diffinfo = parsed_difftree_line($difftree->[$patch_idx]);
+
+                       # compact combined diff output can have some patches skipped
+                       # find which patch (using pathname of result) we are at now;
+                       if ($is_combined) {
+                               while ($to_name ne $diffinfo->{'to_file'}) {
+                                       print "<div class=\"patch\" id=\"patch". ($patch_idx+1) ."\">\n" .
+                                             format_diff_cc_simplified($diffinfo, @hash_parents) .
+                                             "</div>\n";  # class="patch"
+
+                                       $patch_idx++;
+                                       $patch_number++;
+
+                                       last if $patch_idx > $#$difftree;
+                                       $diffinfo = parsed_difftree_line($difftree->[$patch_idx]);
+                               }
+                       }
+
+                       # modifies %from, %to hashes
+                       parse_from_to_diffinfo($diffinfo, \%from, \%to, @hash_parents);
+
+                       # this is first patch for raw difftree line with $patch_idx index
+                       # we index @$difftree array from 0, but number patches from 1
+                       print "<div class=\"patch\" id=\"patch". ($patch_idx+1) ."\">\n";
+               }
+
+               # git diff header
+               #assert($patch_line =~ m/^diff /) if DEBUG;
+               #assert($patch_line !~ m!$/$!) if DEBUG; # is chomp-ed
+               $patch_number++;
+               # print "git diff" header
+               print format_git_diff_header_line($patch_line, $diffinfo,
+                                                 \%from, \%to);
+
+               # print extended diff header
+               print "<div class=\"diff extended_header\">\n";
+       EXTENDED_HEADER:
+               while ($patch_line = <$fd>) {
+                       chomp $patch_line;
+
+                       last EXTENDED_HEADER if ($patch_line =~ m/^--- |^diff /);
+
+                       print format_extended_diff_header_line($patch_line, $diffinfo,
+                                                              \%from, \%to);
+               }
+               print "</div>\n"; # class="diff extended_header"
+
+               # from-file/to-file diff header
+               if (! $patch_line) {
+                       print "</div>\n"; # class="patch"
+                       last PATCH;
+               }
+               next PATCH if ($patch_line =~ m/^diff /);
+               #assert($patch_line =~ m/^---/) if DEBUG;
+
+               my $last_patch_line = $patch_line;
+               $patch_line = <$fd>;
+               chomp $patch_line;
+               #assert($patch_line =~ m/^\+\+\+/) if DEBUG;
+
+               print format_diff_from_to_header($last_patch_line, $patch_line,
+                                                $diffinfo, \%from, \%to,
+                                                @hash_parents);
+
+               # the patch itself
+       LINE:
+               while ($patch_line = <$fd>) {
+                       chomp $patch_line;
+
+                       next PATCH if ($patch_line =~ m/^diff /);
+
+                       my $class = diff_line_class($patch_line, \%from, \%to);
+
+                       if ($class eq 'chunk_header') {
+                               print_diff_chunk($diff_style, scalar @hash_parents, \%from, \%to, @chunk);
+                               @chunk = ();
+                       }
+
+                       push @chunk, [ $class, $patch_line ];
+               }
+
+       } continue {
+               if (@chunk) {
+                       print_diff_chunk($diff_style, scalar @hash_parents, \%from, \%to, @chunk);
+                       @chunk = ();
+               }
+               print "</div>\n"; # class="patch"
+       }
+
+       # for compact combined (--cc) format, with chunk and patch simplification
+       # the patchset might be empty, but there might be unprocessed raw lines
+       for (++$patch_idx if $patch_number > 0;
+            $patch_idx < @$difftree;
+            ++$patch_idx) {
+               # read and prepare patch information
+               $diffinfo = parsed_difftree_line($difftree->[$patch_idx]);
+
+               # generate anchor for "patch" links in difftree / whatchanged part
+               print "<div class=\"patch\" id=\"patch". ($patch_idx+1) ."\">\n" .
+                     format_diff_cc_simplified($diffinfo, @hash_parents) .
+                     "</div>\n";  # class="patch"
+
+               $patch_number++;
+       }
+
+       if ($patch_number == 0) {
+               if (@hash_parents > 1) {
+                       print "<div class=\"diff nodifferences\">Trivial merge</div>\n";
+               } else {
+                       print "<div class=\"diff nodifferences\">No differences found</div>\n";
+               }
+       }
+
+       print "</div>\n"; # class="patchset"
+}
+
+# . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
+
+sub git_project_search_form {
+       my ($searchtext, $search_use_regexp) = @_;
+
+       my $limit = '';
+       if ($project_filter) {
+               $limit = " in '$project_filter/'";
+       }
+
+       print "<div class=\"projsearch\">\n";
+       print $cgi->start_form(-method => 'get', -action => $my_uri) .
+             $cgi->hidden(-name => 'a', -value => 'project_list')  . "\n";
+       print $cgi->hidden(-name => 'pf', -value => $project_filter). "\n"
+               if (defined $project_filter);
+       print $cgi->textfield(-name => 's', -value => $searchtext,
+                             -title => "Search project by name and description$limit",
+                             -size => 60) . "\n" .
+             "<span title=\"Extended regular expression\">" .
+             $cgi->checkbox(-name => 'sr', -value => 1, -label => 're',
+                            -checked => $search_use_regexp) .
+             "</span>\n" .
+             $cgi->submit(-name => 'btnS', -value => 'Search') .
+             $cgi->end_form() . "\n" .
+             $cgi->a({-href => href(project => undef, searchtext => undef,
+                                    project_filter => $project_filter)},
+                     esc_html("List all projects$limit")) . "<br />\n";
+       print "</div>\n";
+}
+
+# entry for given @keys needs filling if at least one of keys in list
+# is not present in %$project_info
+sub project_info_needs_filling {
+       my ($project_info, @keys) = @_;
+
+       # return List::MoreUtils::any { !exists $project_info->{$_} } @keys;
+       foreach my $key (@keys) {
+               if (!exists $project_info->{$key}) {
+                       return 1;
+               }
+       }
+       return;
+}
+
+# fills project list info (age, description, owner, category, forks, etc.)
+# for each project in the list, removing invalid projects from
+# returned list, or fill only specified info.
+#
+# Invalid projects are removed from the returned list if and only if you
+# ask 'age' or 'age_string' to be filled, because they are the only fields
+# that run unconditionally git command that requires repository, and
+# therefore do always check if project repository is invalid.
+#
+# USAGE:
+# * fill_project_list_info(\@project_list, 'descr_long', 'ctags')
+#   ensures that 'descr_long' and 'ctags' fields are filled
+# * @project_list = fill_project_list_info(\@project_list)
+#   ensures that all fields are filled (and invalid projects removed)
+#
+# NOTE: modifies $projlist, but does not remove entries from it
+sub fill_project_list_info {
+       my ($projlist, @wanted_keys) = @_;
+       my @projects;
+       my $filter_set = sub { return @_; };
+       if (@wanted_keys) {
+               my %wanted_keys = map { $_ => 1 } @wanted_keys;
+               $filter_set = sub { return grep { $wanted_keys{$_} } @_; };
+       }
+
+       my $show_ctags = gitweb_check_feature('ctags');
+ PROJECT:
+       foreach my $pr (@$projlist) {
+               if (project_info_needs_filling($pr, $filter_set->('age', 'age_string'))) {
+                       my (@activity) = git_get_last_activity($pr->{'path'});
+                       unless (@activity) {
+                               next PROJECT;
+                       }
+                       ($pr->{'age'}, $pr->{'age_string'}) = @activity;
+               }
+               if (project_info_needs_filling($pr, $filter_set->('descr', 'descr_long'))) {
+                       my $descr = git_get_project_description($pr->{'path'}) || "";
+                       $descr = to_utf8($descr);
+                       $pr->{'descr_long'} = $descr;
+                       $pr->{'descr'} = chop_str($descr, $projects_list_description_width, 5);
+               }
+               if (project_info_needs_filling($pr, $filter_set->('owner'))) {
+                       $pr->{'owner'} = git_get_project_owner("$pr->{'path'}") || "";
+               }
+               if ($show_ctags &&
+                   project_info_needs_filling($pr, $filter_set->('ctags'))) {
+                       $pr->{'ctags'} = git_get_project_ctags($pr->{'path'});
+               }
+               if ($projects_list_group_categories &&
+                   project_info_needs_filling($pr, $filter_set->('category'))) {
+                       my $cat = git_get_project_category($pr->{'path'}) ||
+                                                          $project_list_default_category;
+                       $pr->{'category'} = to_utf8($cat);
+               }
+
+               push @projects, $pr;
+       }
+
+       return @projects;
+}
+
+sub sort_projects_list {
+       my ($projlist, $order) = @_;
+
+       sub order_str {
+               my $key = shift;
+               return sub { $a->{$key} cmp $b->{$key} };
+       }
+
+       sub order_num_then_undef {
+               my $key = shift;
+               return sub {
+                       defined $a->{$key} ?
+                               (defined $b->{$key} ? $a->{$key} <=> $b->{$key} : -1) :
+                               (defined $b->{$key} ? 1 : 0)
+               };
+       }
+
+       my %orderings = (
+               project => order_str('path'),
+               descr => order_str('descr_long'),
+               owner => order_str('owner'),
+               age => order_num_then_undef('age'),
+       );
+
+       my $ordering = $orderings{$order};
+       return defined $ordering ? sort $ordering @$projlist : @$projlist;
+}
+
+# returns a hash of categories, containing the list of project
+# belonging to each category
+sub build_projlist_by_category {
+       my ($projlist, $from, $to) = @_;
+       my %categories;
+
+       $from = 0 unless defined $from;
+       $to = $#$projlist if (!defined $to || $#$projlist < $to);
+
+       for (my $i = $from; $i <= $to; $i++) {
+               my $pr = $projlist->[$i];
+               push @{$categories{ $pr->{'category'} }}, $pr;
+       }
+
+       return wantarray ? %categories : \%categories;
+}
+
+# print 'sort by' <th> element, generating 'sort by $name' replay link
+# if that order is not selected
+sub print_sort_th {
+       print format_sort_th(@_);
+}
+
+sub format_sort_th {
+       my ($name, $order, $header) = @_;
+       my $sort_th = "";
+       $header ||= ucfirst($name);
+
+       if ($order eq $name) {
+               $sort_th .= "<th>$header</th>\n";
+       } else {
+               $sort_th .= "<th>" .
+                           $cgi->a({-href => href(-replay=>1, order=>$name),
+                                    -class => "header"}, $header) .
+                           "</th>\n";
+       }
+
+       return $sort_th;
+}
+
+sub git_project_list_rows {
+       my ($projlist, $from, $to, $check_forks) = @_;
+
+       $from = 0 unless defined $from;
+       $to = $#$projlist if (!defined $to || $#$projlist < $to);
+
+       my $alternate = 1;
+       for (my $i = $from; $i <= $to; $i++) {
+               my $pr = $projlist->[$i];
+
+               if ($alternate) {
+                       print "<tr class=\"dark\">\n";
+               } else {
+                       print "<tr class=\"light\">\n";
+               }
+               $alternate ^= 1;
+
+               if ($check_forks) {
+                       print "<td>";
+                       if ($pr->{'forks'}) {
+                               my $nforks = scalar @{$pr->{'forks'}};
+                               if ($nforks > 0) {
+                                       print $cgi->a({-href => href(project=>$pr->{'path'}, action=>"forks"),
+                                                      -title => "$nforks forks"}, "+");
+                               } else {
+                                       print $cgi->span({-title => "$nforks forks"}, "+");
+                               }
+                       }
+                       print "</td>\n";
+               }
+               print "<td>" . $cgi->a({-href => href(project=>$pr->{'path'}, action=>"summary"),
+                                       -class => "list"},
+                                      esc_html_match_hl($pr->{'path'}, $search_regexp)) .
+                     "</td>\n" .
+                     "<td>" . $cgi->a({-href => href(project=>$pr->{'path'}, action=>"summary"),
+                                       -class => "list",
+                                       -title => $pr->{'descr_long'}},
+                                       $search_regexp
+                                       ? esc_html_match_hl_chopped($pr->{'descr_long'},
+                                                                   $pr->{'descr'}, $search_regexp)
+                                       : esc_html($pr->{'descr'})) .
+                     "</td>\n";
+               unless ($omit_owner) {
+                       print "<td><i>" . chop_and_escape_str($pr->{'owner'}, 15) . "</i></td>\n";
+               }
+               unless ($omit_age_column) {
+                       print "<td class=\"". age_class($pr->{'age'}) . "\">" .
+                           (defined $pr->{'age_string'} ? $pr->{'age_string'} : "No commits") . "</td>\n";
+               }
+               print"<td class=\"link\">" .
+                     $cgi->a({-href => href(project=>$pr->{'path'}, action=>"summary")}, "summary")   . " | " .
+                     $cgi->a({-href => href(project=>$pr->{'path'}, action=>"shortlog")}, "shortlog") . " | " .
+                     $cgi->a({-href => href(project=>$pr->{'path'}, action=>"log")}, "log") . " | " .
+                     $cgi->a({-href => href(project=>$pr->{'path'}, action=>"tree")}, "tree") .
+                     ($pr->{'forks'} ? " | " . $cgi->a({-href => href(project=>$pr->{'path'}, action=>"forks")}, "forks") : '') .
+                     "</td>\n" .
+                     "</tr>\n";
+       }
+}
+
+sub git_project_list_body {
+       # actually uses global variable $project
+       my ($projlist, $order, $from, $to, $extra, $no_header) = @_;
+       my @projects = @$projlist;
+
+       my $check_forks = gitweb_check_feature('forks');
+       my $show_ctags  = gitweb_check_feature('ctags');
+       my $tagfilter = $show_ctags ? $input_params{'ctag'} : undef;
+       $check_forks = undef
+               if ($tagfilter || $search_regexp);
+
+       # filtering out forks before filling info allows to do less work
+       @projects = filter_forks_from_projects_list(\@projects)
+               if ($check_forks);
+       # search_projects_list pre-fills required info
+       @projects = search_projects_list(\@projects,
+                                        'search_regexp' => $search_regexp,
+                                        'tagfilter'  => $tagfilter)
+               if ($tagfilter || $search_regexp);
+       # fill the rest
+       my @all_fields = ('descr', 'descr_long', 'ctags', 'category');
+       push @all_fields, ('age', 'age_string') unless($omit_age_column);
+       push @all_fields, 'owner' unless($omit_owner);
+       @projects = fill_project_list_info(\@projects, @all_fields);
+
+       $order ||= $default_projects_order;
+       $from = 0 unless defined $from;
+       $to = $#projects if (!defined $to || $#projects < $to);
+
+       # short circuit
+       if ($from > $to) {
+               print "<center>\n".
+                     "<b>No such projects found</b><br />\n".
+                     "Click ".$cgi->a({-href=>href(project=>undef)},"here")." to view all projects<br />\n".
+                     "</center>\n<br />\n";
+               return;
+       }
+
+       @projects = sort_projects_list(\@projects, $order);
+
+       if ($show_ctags) {
+               my $ctags = git_gather_all_ctags(\@projects);
+               my $cloud = git_populate_project_tagcloud($ctags);
+               print git_show_project_tagcloud($cloud, 64);
+       }
+
+       print "<table class=\"project_list\">\n";
+       unless ($no_header) {
+               print "<tr>\n";
+               if ($check_forks) {
+                       print "<th></th>\n";
+               }
+               print_sort_th('project', $order, 'Project');
+               print_sort_th('descr', $order, 'Description');
+               print_sort_th('owner', $order, 'Owner') unless $omit_owner;
+               print_sort_th('age', $order, 'Last Change') unless $omit_age_column;
+               print "<th></th>\n" . # for links
+                     "</tr>\n";
+       }
+
+       if ($projects_list_group_categories) {
+               # only display categories with projects in the $from-$to window
+               @projects = sort {$a->{'category'} cmp $b->{'category'}} @projects[$from..$to];
+               my %categories = build_projlist_by_category(\@projects, $from, $to);
+               foreach my $cat (sort keys %categories) {
+                       unless ($cat eq "") {
+                               print "<tr>\n";
+                               if ($check_forks) {
+                                       print "<td></td>\n";
+                               }
+                               print "<td class=\"category\" colspan=\"5\">".esc_html($cat)."</td>\n";
+                               print "</tr>\n";
+                       }
+
+                       git_project_list_rows($categories{$cat}, undef, undef, $check_forks);
+               }
+       } else {
+               git_project_list_rows(\@projects, $from, $to, $check_forks);
+       }
+
+       if (defined $extra) {
+               print "<tr>\n";
+               if ($check_forks) {
+                       print "<td></td>\n";
+               }
+               print "<td colspan=\"5\">$extra</td>\n" .
+                     "</tr>\n";
+       }
+       print "</table>\n";
+}
+
+sub git_log_body {
+       # uses global variable $project
+       my ($commitlist, $from, $to, $refs, $extra) = @_;
+
+       $from = 0 unless defined $from;
+       $to = $#{$commitlist} if (!defined $to || $#{$commitlist} < $to);
+
+       for (my $i = 0; $i <= $to; $i++) {
+               my %co = %{$commitlist->[$i]};
+               next if !%co;
+               my $commit = $co{'id'};
+               my $ref = format_ref_marker($refs, $commit);
+               git_print_header_div('commit',
+                              "<span class=\"age\">$co{'age_string'}</span>" .
+                              esc_html($co{'title'}) . $ref,
+                              $commit);
+               print "<div class=\"title_text\">\n" .
+                     "<div class=\"log_link\">\n" .
+                     $cgi->a({-href => href(action=>"commit", hash=>$commit)}, "commit") .
+                     " | " .
+                     $cgi->a({-href => href(action=>"commitdiff", hash=>$commit)}, "commitdiff") .
+                     " | " .
+                     $cgi->a({-href => href(action=>"tree", hash=>$commit, hash_base=>$commit)}, "tree") .
+                     "<br/>\n" .
+                     "</div>\n";
+                     git_print_authorship(\%co, -tag => 'span');
+                     print "<br/>\n</div>\n";
+
+               print "<div class=\"log_body\">\n";
+               git_print_log($co{'comment'}, -final_empty_line=> 1);
+               print "</div>\n";
+       }
+       if ($extra) {
+               print "<div class=\"page_nav\">\n";
+               print "$extra\n";
+               print "</div>\n";
+       }
+}
+
+sub git_shortlog_body {
+       # uses global variable $project
+       my ($commitlist, $from, $to, $refs, $extra) = @_;
+
+       $from = 0 unless defined $from;
+       $to = $#{$commitlist} if (!defined $to || $#{$commitlist} < $to);
+
+       print "<table class=\"shortlog\">\n";
+       my $alternate = 1;
+       for (my $i = $from; $i <= $to; $i++) {
+               my %co = %{$commitlist->[$i]};
+               my $commit = $co{'id'};
+               my $ref = format_ref_marker($refs, $commit);
+               if ($alternate) {
+                       print "<tr class=\"dark\">\n";
+               } else {
+                       print "<tr class=\"light\">\n";
+               }
+               $alternate ^= 1;
+               # git_summary() used print "<td><i>$co{'age_string'}</i></td>\n" .
+               print "<td title=\"$co{'age_string_age'}\"><i>$co{'age_string_date'}</i></td>\n" .
+                     format_author_html('td', \%co, 10) . "<td>";
+               print format_subject_html($co{'title'}, $co{'title_short'},
+                                         href(action=>"commit", hash=>$commit), $ref);
+               print "</td>\n" .
+                     "<td class=\"link\">" .
+                     $cgi->a({-href => href(action=>"commit", hash=>$commit)}, "commit") . " | " .
+                     $cgi->a({-href => href(action=>"commitdiff", hash=>$commit)}, "commitdiff") . " | " .
+                     $cgi->a({-href => href(action=>"tree", hash=>$commit, hash_base=>$commit)}, "tree");
+               my $snapshot_links = format_snapshot_links($commit);
+               if (defined $snapshot_links) {
+                       print " | " . $snapshot_links;
+               }
+               print "</td>\n" .
+                     "</tr>\n";
+       }
+       if (defined $extra) {
+               print "<tr>\n" .
+                     "<td colspan=\"4\">$extra</td>\n" .
+                     "</tr>\n";
+       }
+       print "</table>\n";
+}
+
+sub git_history_body {
+       # Warning: assumes constant type (blob or tree) during history
+       my ($commitlist, $from, $to, $refs, $extra,
+           $file_name, $file_hash, $ftype) = @_;
+
+       $from = 0 unless defined $from;
+       $to = $#{$commitlist} unless (defined $to && $to <= $#{$commitlist});
+
+       print "<table class=\"history\">\n";
+       my $alternate = 1;
+       for (my $i = $from; $i <= $to; $i++) {
+               my %co = %{$commitlist->[$i]};
+               if (!%co) {
+                       next;
+               }
+               my $commit = $co{'id'};
+
+               my $ref = format_ref_marker($refs, $commit);
+
+               if ($alternate) {
+                       print "<tr class=\"dark\">\n";
+               } else {
+                       print "<tr class=\"light\">\n";
+               }
+               $alternate ^= 1;
+               print "<td title=\"$co{'age_string_age'}\"><i>$co{'age_string_date'}</i></td>\n" .
+       # shortlog:   format_author_html('td', \%co, 10)
+                     format_author_html('td', \%co, 15, 3) . "<td>";
+               # originally git_history used chop_str($co{'title'}, 50)
+               print format_subject_html($co{'title'}, $co{'title_short'},
+                                         href(action=>"commit", hash=>$commit), $ref);
+               print "</td>\n" .
+                     "<td class=\"link\">" .
+                     $cgi->a({-href => href(action=>$ftype, hash_base=>$commit, file_name=>$file_name)}, $ftype) . " | " .
+                     $cgi->a({-href => href(action=>"commitdiff", hash=>$commit)}, "commitdiff");
+
+               if ($ftype eq 'blob') {
+                       my $blob_current = $file_hash;
+                       my $blob_parent  = git_get_hash_by_path($commit, $file_name);
+                       if (defined $blob_current && defined $blob_parent &&
+                                       $blob_current ne $blob_parent) {
+                               print " | " .
+                                       $cgi->a({-href => href(action=>"blobdiff",
+                                                              hash=>$blob_current, hash_parent=>$blob_parent,
+                                                              hash_base=>$hash_base, hash_parent_base=>$commit,
+                                                              file_name=>$file_name)},
+                                               "diff to current");
+                       }
+               }
+               print "</td>\n" .
+                     "</tr>\n";
+       }
+       if (defined $extra) {
+               print "<tr>\n" .
+                     "<td colspan=\"4\">$extra</td>\n" .
+                     "</tr>\n";
+       }
+       print "</table>\n";
+}
+
+sub git_tags_body {
+       # uses global variable $project
+       my ($taglist, $from, $to, $extra) = @_;
+       $from = 0 unless defined $from;
+       $to = $#{$taglist} if (!defined $to || $#{$taglist} < $to);
+
+       print "<table class=\"tags\">\n";
+       my $alternate = 1;
+       for (my $i = $from; $i <= $to; $i++) {
+               my $entry = $taglist->[$i];
+               my %tag = %$entry;
+               my $comment = $tag{'subject'};
+               my $comment_short;
+               if (defined $comment) {
+                       $comment_short = chop_str($comment, 30, 5);
+               }
+               if ($alternate) {
+                       print "<tr class=\"dark\">\n";
+               } else {
+                       print "<tr class=\"light\">\n";
+               }
+               $alternate ^= 1;
+               if (defined $tag{'age'}) {
+                       print "<td><i>$tag{'age'}</i></td>\n";
+               } else {
+                       print "<td></td>\n";
+               }
+               print "<td>" .
+                     $cgi->a({-href => href(action=>$tag{'reftype'}, hash=>$tag{'refid'}),
+                              -class => "list name"}, esc_html($tag{'name'})) .
+                     "</td>\n" .
+                     "<td>";
+               if (defined $comment) {
+                       print format_subject_html($comment, $comment_short,
+                                                 href(action=>"tag", hash=>$tag{'id'}));
+               }
+               print "</td>\n" .
+                     "<td class=\"selflink\">";
+               if ($tag{'type'} eq "tag") {
+                       print $cgi->a({-href => href(action=>"tag", hash=>$tag{'id'})}, "tag");
+               } else {
+                       print "&nbsp;";
+               }
+               print "</td>\n" .
+                     "<td class=\"link\">" . " | " .
+                     $cgi->a({-href => href(action=>$tag{'reftype'}, hash=>$tag{'refid'})}, $tag{'reftype'});
+               if ($tag{'reftype'} eq "commit") {
+                       print " | " . $cgi->a({-href => href(action=>"shortlog", hash=>$tag{'fullname'})}, "shortlog") .
+                             " | " . $cgi->a({-href => href(action=>"log", hash=>$tag{'fullname'})}, "log");
+               } elsif ($tag{'reftype'} eq "blob") {
+                       print " | " . $cgi->a({-href => href(action=>"blob_plain", hash=>$tag{'refid'})}, "raw");
+               }
+               print "</td>\n" .
+                     "</tr>";
+       }
+       if (defined $extra) {
+               print "<tr>\n" .
+                     "<td colspan=\"5\">$extra</td>\n" .
+                     "</tr>\n";
+       }
+       print "</table>\n";
+}
+
+sub git_heads_body {
+       # uses global variable $project
+       my ($headlist, $head_at, $from, $to, $extra) = @_;
+       $from = 0 unless defined $from;
+       $to = $#{$headlist} if (!defined $to || $#{$headlist} < $to);
+
+       print "<table class=\"heads\">\n";
+       my $alternate = 1;
+       for (my $i = $from; $i <= $to; $i++) {
+               my $entry = $headlist->[$i];
+               my %ref = %$entry;
+               my $curr = defined $head_at && $ref{'id'} eq $head_at;
+               if ($alternate) {
+                       print "<tr class=\"dark\">\n";
+               } else {
+                       print "<tr class=\"light\">\n";
+               }
+               $alternate ^= 1;
+               print "<td><i>$ref{'age'}</i></td>\n" .
+                     ($curr ? "<td class=\"current_head\">" : "<td>") .
+                     $cgi->a({-href => href(action=>"shortlog", hash=>$ref{'fullname'}),
+                              -class => "list name"},esc_html($ref{'name'})) .
+                     "</td>\n" .
+                     "<td class=\"link\">" .
+                     $cgi->a({-href => href(action=>"shortlog", hash=>$ref{'fullname'})}, "shortlog") . " | " .
+                     $cgi->a({-href => href(action=>"log", hash=>$ref{'fullname'})}, "log") . " | " .
+                     $cgi->a({-href => href(action=>"tree", hash=>$ref{'fullname'}, hash_base=>$ref{'fullname'})}, "tree") .
+                     "</td>\n" .
+                     "</tr>";
+       }
+       if (defined $extra) {
+               print "<tr>\n" .
+                     "<td colspan=\"3\">$extra</td>\n" .
+                     "</tr>\n";
+       }
+       print "</table>\n";
+}
+
+# Display a single remote block
+sub git_remote_block {
+       my ($remote, $rdata, $limit, $head) = @_;
+
+       my $heads = $rdata->{'heads'};
+       my $fetch = $rdata->{'fetch'};
+       my $push = $rdata->{'push'};
+
+       my $urls_table = "<table class=\"projects_list\">\n" ;
+
+       if (defined $fetch) {
+               if ($fetch eq $push) {
+                       $urls_table .= format_repo_url("URL", $fetch);
+               } else {
+                       $urls_table .= format_repo_url("Fetch URL", $fetch);
+                       $urls_table .= format_repo_url("Push URL", $push) if defined $push;
+               }
+       } elsif (defined $push) {
+               $urls_table .= format_repo_url("Push URL", $push);
+       } else {
+               $urls_table .= format_repo_url("", "No remote URL");
+       }
+
+       $urls_table .= "</table>\n";
+
+       my $dots;
+       if (defined $limit && $limit < @$heads) {
+               $dots = $cgi->a({-href => href(action=>"remotes", hash=>$remote)}, "...");
+       }
+
+       print $urls_table;
+       git_heads_body($heads, $head, 0, $limit, $dots);
+}
+
+# Display a list of remote names with the respective fetch and push URLs
+sub git_remotes_list {
+       my ($remotedata, $limit) = @_;
+       print "<table class=\"heads\">\n";
+       my $alternate = 1;
+       my @remotes = sort keys %$remotedata;
+
+       my $limited = $limit && $limit < @remotes;
+
+       $#remotes = $limit - 1 if $limited;
+
+       while (my $remote = shift @remotes) {
+               my $rdata = $remotedata->{$remote};
+               my $fetch = $rdata->{'fetch'};
+               my $push = $rdata->{'push'};
+               if ($alternate) {
+                       print "<tr class=\"dark\">\n";
+               } else {
+                       print "<tr class=\"light\">\n";
+               }
+               $alternate ^= 1;
+               print "<td>" .
+                     $cgi->a({-href=> href(action=>'remotes', hash=>$remote),
+                              -class=> "list name"},esc_html($remote)) .
+                     "</td>";
+               print "<td class=\"link\">" .
+                     (defined $fetch ? $cgi->a({-href=> $fetch}, "fetch") : "fetch") .
+                     " | " .
+                     (defined $push ? $cgi->a({-href=> $push}, "push") : "push") .
+                     "</td>";
+
+               print "</tr>\n";
+       }
+
+       if ($limited) {
+               print "<tr>\n" .
+                     "<td colspan=\"3\">" .
+                     $cgi->a({-href => href(action=>"remotes")}, "...") .
+                     "</td>\n" . "</tr>\n";
+       }
+
+       print "</table>";
+}
+
+# Display remote heads grouped by remote, unless there are too many
+# remotes, in which case we only display the remote names
+sub git_remotes_body {
+       my ($remotedata, $limit, $head) = @_;
+       if ($limit and $limit < keys %$remotedata) {
+               git_remotes_list($remotedata, $limit);
+       } else {
+               fill_remote_heads($remotedata);
+               while (my ($remote, $rdata) = each %$remotedata) {
+                       git_print_section({-class=>"remote", -id=>$remote},
+                               ["remotes", $remote, $remote], sub {
+                                       git_remote_block($remote, $rdata, $limit, $head);
+                               });
+               }
+       }
+}
+
+sub git_search_message {
+       my %co = @_;
+
+       my $greptype;
+       if ($searchtype eq 'commit') {
+               $greptype = "--grep=";
+       } elsif ($searchtype eq 'author') {
+               $greptype = "--author=";
+       } elsif ($searchtype eq 'committer') {
+               $greptype = "--committer=";
+       }
+       $greptype .= $searchtext;
+       my @commitlist = parse_commits($hash, 101, (100 * $page), undef,
+                                      $greptype, '--regexp-ignore-case',
+                                      $search_use_regexp ? '--extended-regexp' : '--fixed-strings');
+
+       my $paging_nav = '';
+       if ($page > 0) {
+               $paging_nav .=
+                       $cgi->a({-href => href(-replay=>1, page=>undef)},
+                               "first") .
+                       " &sdot; " .
+                       $cgi->a({-href => href(-replay=>1, page=>$page-1),
+                                -accesskey => "p", -title => "Alt-p"}, "prev");
+       } else {
+               $paging_nav .= "first &sdot; prev";
+       }
+       my $next_link = '';
+       if ($#commitlist >= 100) {
+               $next_link =
+                       $cgi->a({-href => href(-replay=>1, page=>$page+1),
+                                -accesskey => "n", -title => "Alt-n"}, "next");
+               $paging_nav .= " &sdot; $next_link";
+       } else {
+               $paging_nav .= " &sdot; next";
+       }
+
+       git_header_html();
+
+       git_print_page_nav('','', $hash,$co{'tree'},$hash, $paging_nav);
+       git_print_header_div('commit', esc_html($co{'title'}), $hash);
+       if ($page == 0 && !@commitlist) {
+               print "<p>No match.</p>\n";
+       } else {
+               git_search_grep_body(\@commitlist, 0, 99, $next_link);
+       }
+
+       git_footer_html();
+}
+
+sub git_search_changes {
+       my %co = @_;
+
+       local $/ = "\n";
+       open my $fd, '-|', git_cmd(), '--no-pager', 'log', @diff_opts,
+               '--pretty=format:%H', '--no-abbrev', '--raw', "-S$searchtext",
+               ($search_use_regexp ? '--pickaxe-regex' : ())
+                       or die_error(500, "Open git-log failed");
+
+       git_header_html();
+
+       git_print_page_nav('','', $hash,$co{'tree'},$hash);
+       git_print_header_div('commit', esc_html($co{'title'}), $hash);
+
+       print "<table class=\"pickaxe search\">\n";
+       my $alternate = 1;
+       undef %co;
+       my @files;
+       while (my $line = <$fd>) {
+               chomp $line;
+               next unless $line;
+
+               my %set = parse_difftree_raw_line($line);
+               if (defined $set{'commit'}) {
+                       # finish previous commit
+                       if (%co) {
+                               print "</td>\n" .
+                                     "<td class=\"link\">" .
+                                     $cgi->a({-href => href(action=>"commit", hash=>$co{'id'})},
+                                             "commit") .
+                                     " | " .
+                                     $cgi->a({-href => href(action=>"tree", hash=>$co{'tree'},
+                                                            hash_base=>$co{'id'})},
+                                             "tree") .
+                                     "</td>\n" .
+                                     "</tr>\n";
+                       }
+
+                       if ($alternate) {
+                               print "<tr class=\"dark\">\n";
+                       } else {
+                               print "<tr class=\"light\">\n";
+                       }
+                       $alternate ^= 1;
+                       %co = parse_commit($set{'commit'});
+                       my $author = chop_and_escape_str($co{'author_name'}, 15, 5);
+                       print "<td title=\"$co{'age_string_age'}\"><i>$co{'age_string_date'}</i></td>\n" .
+                             "<td><i>$author</i></td>\n" .
+                             "<td>" .
+                             $cgi->a({-href => href(action=>"commit", hash=>$co{'id'}),
+                                     -class => "list subject"},
+                                     chop_and_escape_str($co{'title'}, 50) . "<br/>");
+               } elsif (defined $set{'to_id'}) {
+                       next if ($set{'to_id'} =~ m/^0{40}$/);
+
+                       print $cgi->a({-href => href(action=>"blob", hash_base=>$co{'id'},
+                                                    hash=>$set{'to_id'}, file_name=>$set{'to_file'}),
+                                     -class => "list"},
+                                     "<span class=\"match\">" . esc_path($set{'file'}) . "</span>") .
+                             "<br/>\n";
+               }
+       }
+       close $fd;
+
+       # finish last commit (warning: repetition!)
+       if (%co) {
+               print "</td>\n" .
+                     "<td class=\"link\">" .
+                     $cgi->a({-href => href(action=>"commit", hash=>$co{'id'})},
+                             "commit") .
+                     " | " .
+                     $cgi->a({-href => href(action=>"tree", hash=>$co{'tree'},
+                                            hash_base=>$co{'id'})},
+                             "tree") .
+                     "</td>\n" .
+                     "</tr>\n";
+       }
+
+       print "</table>\n";
+
+       git_footer_html();
+}
+
+sub git_search_files {
+       my %co = @_;
+
+       local $/ = "\n";
+       open my $fd, "-|", git_cmd(), 'grep', '-n', '-z',
+               $search_use_regexp ? ('-E', '-i') : '-F',
+               $searchtext, $co{'tree'}
+                       or die_error(500, "Open git-grep failed");
+
+       git_header_html();
+
+       git_print_page_nav('','', $hash,$co{'tree'},$hash);
+       git_print_header_div('commit', esc_html($co{'title'}), $hash);
+
+       print "<table class=\"grep_search\">\n";
+       my $alternate = 1;
+       my $matches = 0;
+       my $lastfile = '';
+       my $file_href;
+       while (my $line = <$fd>) {
+               chomp $line;
+               my ($file, $lno, $ltext, $binary);
+               last if ($matches++ > 1000);
+               if ($line =~ /^Binary file (.+) matches$/) {
+                       $file = $1;
+                       $binary = 1;
+               } else {
+                       ($file, $lno, $ltext) = split(/\0/, $line, 3);
+                       $file =~ s/^$co{'tree'}://;
+               }
+               if ($file ne $lastfile) {
+                       $lastfile and print "</td></tr>\n";
+                       if ($alternate++) {
+                               print "<tr class=\"dark\">\n";
+                       } else {
+                               print "<tr class=\"light\">\n";
+                       }
+                       $file_href = href(action=>"blob", hash_base=>$co{'id'},
+                                         file_name=>$file);
+                       print "<td class=\"list\">".
+                               $cgi->a({-href => $file_href, -class => "list"}, esc_path($file));
+                       print "</td><td>\n";
+                       $lastfile = $file;
+               }
+               if ($binary) {
+                       print "<div class=\"binary\">Binary file</div>\n";
+               } else {
+                       $ltext = untabify($ltext);
+                       if ($ltext =~ m/^(.*)($search_regexp)(.*)$/i) {
+                               $ltext = esc_html($1, -nbsp=>1);
+                               $ltext .= '<span class="match">';
+                               $ltext .= esc_html($2, -nbsp=>1);
+                               $ltext .= '</span>';
+                               $ltext .= esc_html($3, -nbsp=>1);
+                       } else {
+                               $ltext = esc_html($ltext, -nbsp=>1);
+                       }
+                       print "<div class=\"pre\">" .
+                               $cgi->a({-href => $file_href.'#l'.$lno,
+                                       -class => "linenr"}, sprintf('%4i', $lno)) .
+                               ' ' .  $ltext . "</div>\n";
+               }
+       }
+       if ($lastfile) {
+               print "</td></tr>\n";
+               if ($matches > 1000) {
+                       print "<div class=\"diff nodifferences\">Too many matches, listing trimmed</div>\n";
+               }
+       } else {
+               print "<div class=\"diff nodifferences\">No matches found</div>\n";
+       }
+       close $fd;
+
+       print "</table>\n";
+
+       git_footer_html();
+}
+
+sub git_search_grep_body {
+       my ($commitlist, $from, $to, $extra) = @_;
+       $from = 0 unless defined $from;
+       $to = $#{$commitlist} if (!defined $to || $#{$commitlist} < $to);
+
+       print "<table class=\"commit_search\">\n";
+       my $alternate = 1;
+       for (my $i = $from; $i <= $to; $i++) {
+               my %co = %{$commitlist->[$i]};
+               if (!%co) {
+                       next;
+               }
+               my $commit = $co{'id'};
+               if ($alternate) {
+                       print "<tr class=\"dark\">\n";
+               } else {
+                       print "<tr class=\"light\">\n";
+               }
+               $alternate ^= 1;
+               print "<td title=\"$co{'age_string_age'}\"><i>$co{'age_string_date'}</i></td>\n" .
+                     format_author_html('td', \%co, 15, 5) .
+                     "<td>" .
+                     $cgi->a({-href => href(action=>"commit", hash=>$co{'id'}),
+                              -class => "list subject"},
+                             chop_and_escape_str($co{'title'}, 50) . "<br/>");
+               my $comment = $co{'comment'};
+               foreach my $line (@$comment) {
+                       if ($line =~ m/^(.*?)($search_regexp)(.*)$/i) {
+                               my ($lead, $match, $trail) = ($1, $2, $3);
+                               $match = chop_str($match, 70, 5, 'center');
+                               my $contextlen = int((80 - length($match))/2);
+                               $contextlen = 30 if ($contextlen > 30);
+                               $lead  = chop_str($lead,  $contextlen, 10, 'left');
+                               $trail = chop_str($trail, $contextlen, 10, 'right');
+
+                               $lead  = esc_html($lead);
+                               $match = esc_html($match);
+                               $trail = esc_html($trail);
+
+                               print "$lead<span class=\"match\">$match</span>$trail<br />";
+                       }
+               }
+               print "</td>\n" .
+                     "<td class=\"link\">" .
+                     $cgi->a({-href => href(action=>"commit", hash=>$co{'id'})}, "commit") .
+                     " | " .
+                     $cgi->a({-href => href(action=>"commitdiff", hash=>$co{'id'})}, "commitdiff") .
+                     " | " .
+                     $cgi->a({-href => href(action=>"tree", hash=>$co{'tree'}, hash_base=>$co{'id'})}, "tree");
+               print "</td>\n" .
+                     "</tr>\n";
+       }
+       if (defined $extra) {
+               print "<tr>\n" .
+                     "<td colspan=\"3\">$extra</td>\n" .
+                     "</tr>\n";
+       }
+       print "</table>\n";
+}
+
+## ======================================================================
+## ======================================================================
+## actions
+
+sub git_project_list {
+       my $order = $input_params{'order'};
+       if (defined $order && $order !~ m/none|project|descr|owner|age/) {
+               die_error(400, "Unknown order parameter");
+       }
+
+       my @list = git_get_projects_list($project_filter, $strict_export);
+       if (!@list) {
+               die_error(404, "No projects found");
+       }
+
+       git_header_html();
+       if (defined $home_text && -f $home_text) {
+               print "<div class=\"index_include\">\n";
+               insert_file($home_text);
+               print "</div>\n";
+       }
+
+       git_project_search_form($searchtext, $search_use_regexp);
+       git_project_list_body(\@list, $order);
+       git_footer_html();
+}
+
+sub git_forks {
+       my $order = $input_params{'order'};
+       if (defined $order && $order !~ m/none|project|descr|owner|age/) {
+               die_error(400, "Unknown order parameter");
+       }
+
+       my $filter = $project;
+       $filter =~ s/\.git$//;
+       my @list = git_get_projects_list($filter);
+       if (!@list) {
+               die_error(404, "No forks found");
+       }
+
+       git_header_html();
+       git_print_page_nav('','');
+       git_print_header_div('summary', "$project forks");
+       git_project_list_body(\@list, $order);
+       git_footer_html();
+}
+
+sub git_project_index {
+       my @projects = git_get_projects_list($project_filter, $strict_export);
+       if (!@projects) {
+               die_error(404, "No projects found");
+       }
+
+       print $cgi->header(
+               -type => 'text/plain',
+               -charset => 'utf-8',
+               -content_disposition => 'inline; filename="index.aux"');
+
+       foreach my $pr (@projects) {
+               if (!exists $pr->{'owner'}) {
+                       $pr->{'owner'} = git_get_project_owner("$pr->{'path'}");
+               }
+
+               my ($path, $owner) = ($pr->{'path'}, $pr->{'owner'});
+               # quote as in CGI::Util::encode, but keep the slash, and use '+' for ' '
+               $path  =~ s/([^a-zA-Z0-9_.\-\/ ])/sprintf("%%%02X", ord($1))/eg;
+               $owner =~ s/([^a-zA-Z0-9_.\-\/ ])/sprintf("%%%02X", ord($1))/eg;
+               $path  =~ s/ /\+/g;
+               $owner =~ s/ /\+/g;
+
+               print "$path $owner\n";
+       }
+}
+
+sub git_summary {
+       my $descr = git_get_project_description($project) || "none";
+       my %co = parse_commit("HEAD");
+       my %cd = %co ? parse_date($co{'committer_epoch'}, $co{'committer_tz'}) : ();
+       my $head = $co{'id'};
+       my $remote_heads = gitweb_check_feature('remote_heads');
+
+       my $owner = git_get_project_owner($project);
+
+       my $refs = git_get_references();
+       # These get_*_list functions return one more to allow us to see if
+       # there are more ...
+       my @taglist  = git_get_tags_list(16);
+       my @headlist = git_get_heads_list(16);
+       my %remotedata = $remote_heads ? git_get_remotes_list() : ();
+       my @forklist;
+       my $check_forks = gitweb_check_feature('forks');
+
+       if ($check_forks) {
+               # find forks of a project
+               my $filter = $project;
+               $filter =~ s/\.git$//;
+               @forklist = git_get_projects_list($filter);
+               # filter out forks of forks
+               @forklist = filter_forks_from_projects_list(\@forklist)
+                       if (@forklist);
+       }
+
+       git_header_html();
+       git_print_page_nav('summary','', $head);
+
+       print "<div class=\"title\">&nbsp;</div>\n";
+       print "<table class=\"projects_list\">\n" .
+             "<tr id=\"metadata_desc\"><td>description</td><td>" . esc_html($descr) . "</td></tr>\n";
+        if ($owner and not $omit_owner) {
+               print  "<tr id=\"metadata_owner\"><td>owner</td><td>" . esc_html($owner) . "</td></tr>\n";
+        }
+       if (defined $cd{'rfc2822'}) {
+               print "<tr id=\"metadata_lchange\"><td>last change</td>" .
+                     "<td>".format_timestamp_html(\%cd)."</td></tr>\n";
+       }
+
+       # use per project git URL list in $projectroot/$project/cloneurl
+       # or make project git URL from git base URL and project name
+       my $url_tag = "URL";
+       my @url_list = git_get_project_url_list($project);
+       @url_list = map { "$_/$project" } @git_base_url_list unless @url_list;
+       foreach my $git_url (@url_list) {
+               next unless $git_url;
+               print format_repo_url($url_tag, $git_url);
+               $url_tag = "";
+       }
+
+       # Tag cloud
+       my $show_ctags = gitweb_check_feature('ctags');
+       if ($show_ctags) {
+               my $ctags = git_get_project_ctags($project);
+               if (%$ctags) {
+                       # without ability to add tags, don't show if there are none
+                       my $cloud = git_populate_project_tagcloud($ctags);
+                       print "<tr id=\"metadata_ctags\">" .
+                             "<td>content tags</td>" .
+                             "<td>".git_show_project_tagcloud($cloud, 48)."</td>" .
+                             "</tr>\n";
+               }
+       }
+
+       print "</table>\n";
+
+       # If XSS prevention is on, we don't include README.html.
+       # TODO: Allow a readme in some safe format.
+       if (!$prevent_xss && -s "$projectroot/$project/README.html") {
+               print "<div class=\"title\">readme</div>\n" .
+                     "<div class=\"readme\">\n";
+               insert_file("$projectroot/$project/README.html");
+               print "\n</div>\n"; # class="readme"
+       }
+
+       # we need to request one more than 16 (0..15) to check if
+       # those 16 are all
+       my @commitlist = $head ? parse_commits($head, 17) : ();
+       if (@commitlist) {
+               git_print_header_div('shortlog');
+               git_shortlog_body(\@commitlist, 0, 15, $refs,
+                                 $#commitlist <=  15 ? undef :
+                                 $cgi->a({-href => href(action=>"shortlog")}, "..."));
+       }
+
+       if (@taglist) {
+               git_print_header_div('tags');
+               git_tags_body(\@taglist, 0, 15,
+                             $#taglist <=  15 ? undef :
+                             $cgi->a({-href => href(action=>"tags")}, "..."));
+       }
+
+       if (@headlist) {
+               git_print_header_div('heads');
+               git_heads_body(\@headlist, $head, 0, 15,
+                              $#headlist <= 15 ? undef :
+                              $cgi->a({-href => href(action=>"heads")}, "..."));
+       }
+
+       if (%remotedata) {
+               git_print_header_div('remotes');
+               git_remotes_body(\%remotedata, 15, $head);
+       }
+
+       if (@forklist) {
+               git_print_header_div('forks');
+               git_project_list_body(\@forklist, 'age', 0, 15,
+                                     $#forklist <= 15 ? undef :
+                                     $cgi->a({-href => href(action=>"forks")}, "..."),
+                                     'no_header');
+       }
+
+       git_footer_html();
+}
+
+sub git_tag {
+       my %tag = parse_tag($hash);
+
+       if (! %tag) {
+               die_error(404, "Unknown tag object");
+       }
+
+       my $head = git_get_head_hash($project);
+       git_header_html();
+       git_print_page_nav('','', $head,undef,$head);
+       git_print_header_div('commit', esc_html($tag{'name'}), $hash);
+       print "<div class=\"title_text\">\n" .
+             "<table class=\"object_header\">\n" .
+             "<tr>\n" .
+             "<td>object</td>\n" .
+             "<td>" . $cgi->a({-class => "list", -href => href(action=>$tag{'type'}, hash=>$tag{'object'})},
+                              $tag{'object'}) . "</td>\n" .
+             "<td class=\"link\">" . $cgi->a({-href => href(action=>$tag{'type'}, hash=>$tag{'object'})},
+                                             $tag{'type'}) . "</td>\n" .
+             "</tr>\n";
+       if (defined($tag{'author'})) {
+               git_print_authorship_rows(\%tag, 'author');
+       }
+       print "</table>\n\n" .
+             "</div>\n";
+       print "<div class=\"page_body\">";
+       my $comment = $tag{'comment'};
+       foreach my $line (@$comment) {
+               chomp $line;
+               print esc_html($line, -nbsp=>1) . "<br/>\n";
+       }
+       print "</div>\n";
+       git_footer_html();
+}
+
+sub git_blame_common {
+       my $format = shift || 'porcelain';
+       if ($format eq 'porcelain' && $input_params{'javascript'}) {
+               $format = 'incremental';
+               $action = 'blame_incremental'; # for page title etc
+       }
+
+       # permissions
+       gitweb_check_feature('blame')
+               or die_error(403, "Blame view not allowed");
+
+       # error checking
+       die_error(400, "No file name given") unless $file_name;
+       $hash_base ||= git_get_head_hash($project);
+       die_error(404, "Couldn't find base commit") unless $hash_base;
+       my %co = parse_commit($hash_base)
+               or die_error(404, "Commit not found");
+       my $ftype = "blob";
+       if (!defined $hash) {
+               $hash = git_get_hash_by_path($hash_base, $file_name, "blob")
+                       or die_error(404, "Error looking up file");
+       } else {
+               $ftype = git_get_type($hash);
+               if ($ftype !~ "blob") {
+                       die_error(400, "Object is not a blob");
+               }
+       }
+
+       my $fd;
+       if ($format eq 'incremental') {
+               # get file contents (as base)
+               open $fd, "-|", git_cmd(), 'cat-file', 'blob', $hash
+                       or die_error(500, "Open git-cat-file failed");
+       } elsif ($format eq 'data') {
+               # run git-blame --incremental
+               open $fd, "-|", git_cmd(), "blame", "--incremental",
+                       $hash_base, "--", $file_name
+                       or die_error(500, "Open git-blame --incremental failed");
+       } else {
+               # run git-blame --porcelain
+               open $fd, "-|", git_cmd(), "blame", '-p',
+                       $hash_base, '--', $file_name
+                       or die_error(500, "Open git-blame --porcelain failed");
+       }
+       binmode $fd, ':utf8';
+
+       # incremental blame data returns early
+       if ($format eq 'data') {
+               print $cgi->header(
+                       -type=>"text/plain", -charset => "utf-8",
+                       -status=> "200 OK");
+               local $| = 1; # output autoflush
+               while (my $line = <$fd>) {
+                       print to_utf8($line);
+               }
+               close $fd
+                       or print "ERROR $!\n";
+
+               print 'END';
+               if (defined $t0 && gitweb_check_feature('timed')) {
+                       print ' '.
+                             tv_interval($t0, [ gettimeofday() ]).
+                             ' '.$number_of_git_cmds;
+               }
+               print "\n";
+
+               return;
+       }
+
+       # page header
+       git_header_html();
+       my $formats_nav =
+               $cgi->a({-href => href(action=>"blob", -replay=>1)},
+                       "blob") .
+               " | ";
+       if ($format eq 'incremental') {
+               $formats_nav .=
+                       $cgi->a({-href => href(action=>"blame", javascript=>0, -replay=>1)},
+                               "blame") . " (non-incremental)";
+       } else {
+               $formats_nav .=
+                       $cgi->a({-href => href(action=>"blame_incremental", -replay=>1)},
+                               "blame") . " (incremental)";
+       }
+       $formats_nav .=
+               " | " .
+               $cgi->a({-href => href(action=>"history", -replay=>1)},
+                       "history") .
+               " | " .
+               $cgi->a({-href => href(action=>$action, file_name=>$file_name)},
+                       "HEAD");
+       git_print_page_nav('','', $hash_base,$co{'tree'},$hash_base, $formats_nav);
+       git_print_header_div('commit', esc_html($co{'title'}), $hash_base);
+       git_print_page_path($file_name, $ftype, $hash_base);
+
+       # page body
+       if ($format eq 'incremental') {
+               print "<noscript>\n<div class=\"error\"><center><b>\n".
+                     "This page requires JavaScript to run.\n Use ".
+                     $cgi->a({-href => href(action=>'blame',javascript=>0,-replay=>1)},
+                             'this page').
+                     " instead.\n".
+                     "</b></center></div>\n</noscript>\n";
+
+               print qq!<div id="progress_bar" style="width: 100%; background-color: yellow"></div>\n!;
+       }
+
+       print qq!<div class="page_body">\n!;
+       print qq!<div id="progress_info">... / ...</div>\n!
+               if ($format eq 'incremental');
+       print qq!<table id="blame_table" class="blame" width="100%">\n!.
+             #qq!<col width="5.5em" /><col width="2.5em" /><col width="*" />\n!.
+             qq!<thead>\n!.
+             qq!<tr><th>Commit</th><th>Line</th><th>Data</th></tr>\n!.
+             qq!</thead>\n!.
+             qq!<tbody>\n!;
+
+       my @rev_color = qw(light dark);
+       my $num_colors = scalar(@rev_color);
+       my $current_color = 0;
+
+       if ($format eq 'incremental') {
+               my $color_class = $rev_color[$current_color];
+
+               #contents of a file
+               my $linenr = 0;
+       LINE:
+               while (my $line = <$fd>) {
+                       chomp $line;
+                       $linenr++;
+
+                       print qq!<tr id="l$linenr" class="$color_class">!.
+                             qq!<td class="sha1"><a href=""> </a></td>!.
+                             qq!<td class="linenr">!.
+                             qq!<a class="linenr" href="">$linenr</a></td>!;
+                       print qq!<td class="pre">! . esc_html($line) . "</td>\n";
+                       print qq!</tr>\n!;
+               }
+
+       } else { # porcelain, i.e. ordinary blame
+               my %metainfo = (); # saves information about commits
+
+               # blame data
+       LINE:
+               while (my $line = <$fd>) {
+                       chomp $line;
+                       # the header: <SHA-1> <src lineno> <dst lineno> [<lines in group>]
+                       # no <lines in group> for subsequent lines in group of lines
+                       my ($full_rev, $orig_lineno, $lineno, $group_size) =
+                          ($line =~ /^([0-9a-f]{40}) (\d+) (\d+)(?: (\d+))?$/);
+                       if (!exists $metainfo{$full_rev}) {
+                               $metainfo{$full_rev} = { 'nprevious' => 0 };
+                       }
+                       my $meta = $metainfo{$full_rev};
+                       my $data;
+                       while ($data = <$fd>) {
+                               chomp $data;
+                               last if ($data =~ s/^\t//); # contents of line
+                               if ($data =~ /^(\S+)(?: (.*))?$/) {
+                                       $meta->{$1} = $2 unless exists $meta->{$1};
+                               }
+                               if ($data =~ /^previous /) {
+                                       $meta->{'nprevious'}++;
+                               }
+                       }
+                       my $short_rev = substr($full_rev, 0, 8);
+                       my $author = $meta->{'author'};
+                       my %date =
+                               parse_date($meta->{'author-time'}, $meta->{'author-tz'});
+                       my $date = $date{'iso-tz'};
+                       if ($group_size) {
+                               $current_color = ($current_color + 1) % $num_colors;
+                       }
+                       my $tr_class = $rev_color[$current_color];
+                       $tr_class .= ' boundary' if (exists $meta->{'boundary'});
+                       $tr_class .= ' no-previous' if ($meta->{'nprevious'} == 0);
+                       $tr_class .= ' multiple-previous' if ($meta->{'nprevious'} > 1);
+                       print "<tr id=\"l$lineno\" class=\"$tr_class\">\n";
+                       if ($group_size) {
+                               print "<td class=\"sha1\"";
+                               print " title=\"". esc_html($author) . ", $date\"";
+                               print " rowspan=\"$group_size\"" if ($group_size > 1);
+                               print ">";
+                               print $cgi->a({-href => href(action=>"commit",
+                                                            hash=>$full_rev,
+                                                            file_name=>$file_name)},
+                                             esc_html($short_rev));
+                               if ($group_size >= 2) {
+                                       my @author_initials = ($author =~ /\b([[:upper:]])\B/g);
+                                       if (@author_initials) {
+                                               print "<br />" .
+                                                     esc_html(join('', @author_initials));
+                                               #           or join('.', ...)
+                                       }
+                               }
+                               print "</td>\n";
+                       }
+                       # 'previous' <sha1 of parent commit> <filename at commit>
+                       if (exists $meta->{'previous'} &&
+                           $meta->{'previous'} =~ /^([a-fA-F0-9]{40}) (.*)$/) {
+                               $meta->{'parent'} = $1;
+                               $meta->{'file_parent'} = unquote($2);
+                       }
+                       my $linenr_commit =
+                               exists($meta->{'parent'}) ?
+                               $meta->{'parent'} : $full_rev;
+                       my $linenr_filename =
+                               exists($meta->{'file_parent'}) ?
+                               $meta->{'file_parent'} : unquote($meta->{'filename'});
+                       my $blamed = href(action => 'blame',
+                                         file_name => $linenr_filename,
+                                         hash_base => $linenr_commit);
+                       print "<td class=\"linenr\">";
+                       print $cgi->a({ -href => "$blamed#l$orig_lineno",
+                                       -class => "linenr" },
+                                     esc_html($lineno));
+                       print "</td>";
+                       print "<td class=\"pre\">" . esc_html($data) . "</td>\n";
+                       print "</tr>\n";
+               } # end while
+
+       }
+
+       # footer
+       print "</tbody>\n".
+             "</table>\n"; # class="blame"
+       print "</div>\n";   # class="blame_body"
+       close $fd
+               or print "Reading blob failed\n";
+
+       git_footer_html();
+}
+
+sub git_blame {
+       git_blame_common();
+}
+
+sub git_blame_incremental {
+       git_blame_common('incremental');
+}
+
+sub git_blame_data {
+       git_blame_common('data');
+}
+
+sub git_tags {
+       my $head = git_get_head_hash($project);
+       git_header_html();
+       git_print_page_nav('','', $head,undef,$head,format_ref_views('tags'));
+       git_print_header_div('summary', $project);
+
+       my @tagslist = git_get_tags_list();
+       if (@tagslist) {
+               git_tags_body(\@tagslist);
+       }
+       git_footer_html();
+}
+
+sub git_heads {
+       my $head = git_get_head_hash($project);
+       git_header_html();
+       git_print_page_nav('','', $head,undef,$head,format_ref_views('heads'));
+       git_print_header_div('summary', $project);
+
+       my @headslist = git_get_heads_list();
+       if (@headslist) {
+               git_heads_body(\@headslist, $head);
+       }
+       git_footer_html();
+}
+
+# used both for single remote view and for list of all the remotes
+sub git_remotes {
+       gitweb_check_feature('remote_heads')
+               or die_error(403, "Remote heads view is disabled");
+
+       my $head = git_get_head_hash($project);
+       my $remote = $input_params{'hash'};
+
+       my $remotedata = git_get_remotes_list($remote);
+       die_error(500, "Unable to get remote information") unless defined $remotedata;
+
+       unless (%$remotedata) {
+               die_error(404, defined $remote ?
+                       "Remote $remote not found" :
+                       "No remotes found");
+       }
+
+       git_header_html(undef, undef, -action_extra => $remote);
+       git_print_page_nav('', '',  $head, undef, $head,
+               format_ref_views($remote ? '' : 'remotes'));
+
+       fill_remote_heads($remotedata);
+       if (defined $remote) {
+               git_print_header_div('remotes', "$remote remote for $project");
+               git_remote_block($remote, $remotedata->{$remote}, undef, $head);
+       } else {
+               git_print_header_div('summary', "$project remotes");
+               git_remotes_body($remotedata, undef, $head);
+       }
+
+       git_footer_html();
+}
+
+sub git_blob_plain {
+       my $type = shift;
+       my $expires;
+
+       if (!defined $hash) {
+               if (defined $file_name) {
+                       my $base = $hash_base || git_get_head_hash($project);
+                       $hash = git_get_hash_by_path($base, $file_name, "blob")
+                               or die_error(404, "Cannot find file");
+               } else {
+                       die_error(400, "No file name defined");
+               }
+       } elsif ($hash =~ m/^[0-9a-fA-F]{40}$/) {
+               # blobs defined by non-textual hash id's can be cached
+               $expires = "+1d";
+       }
+
+       open my $fd, "-|", git_cmd(), "cat-file", "blob", $hash
+               or die_error(500, "Open git-cat-file blob '$hash' failed");
+
+       # content-type (can include charset)
+       $type = blob_contenttype($fd, $file_name, $type);
+
+       # "save as" filename, even when no $file_name is given
+       my $save_as = "$hash";
+       if (defined $file_name) {
+               $save_as = $file_name;
+       } elsif ($type =~ m/^text\//) {
+               $save_as .= '.txt';
+       }
+
+       # With XSS prevention on, blobs of all types except a few known safe
+       # ones are served with "Content-Disposition: attachment" to make sure
+       # they don't run in our security domain.  For certain image types,
+       # blob view writes an <img> tag referring to blob_plain view, and we
+       # want to be sure not to break that by serving the image as an
+       # attachment (though Firefox 3 doesn't seem to care).
+       my $sandbox = $prevent_xss &&
+               $type !~ m!^(?:text/[a-z]+|image/(?:gif|png|jpeg))(?:[ ;]|$)!;
+
+       # serve text/* as text/plain
+       if ($prevent_xss &&
+           ($type =~ m!^text/[a-z]+\b(.*)$! ||
+            ($type =~ m!^[a-z]+/[a-z]\+xml\b(.*)$! && -T $fd))) {
+               my $rest = $1;
+               $rest = defined $rest ? $rest : '';
+               $type = "text/plain$rest";
+       }
+
+       print $cgi->header(
+               -type => $type,
+               -expires => $expires,
+               -content_disposition =>
+                       ($sandbox ? 'attachment' : 'inline')
+                       . '; filename="' . $save_as . '"');
+       local $/ = undef;
+       binmode STDOUT, ':raw';
+       print <$fd>;
+       binmode STDOUT, ':utf8'; # as set at the beginning of gitweb.cgi
+       close $fd;
+}
+
+sub git_blob {
+       my $expires;
+
+       if (!defined $hash) {
+               if (defined $file_name) {
+                       my $base = $hash_base || git_get_head_hash($project);
+                       $hash = git_get_hash_by_path($base, $file_name, "blob")
+                               or die_error(404, "Cannot find file");
+               } else {
+                       die_error(400, "No file name defined");
+               }
+       } elsif ($hash =~ m/^[0-9a-fA-F]{40}$/) {
+               # blobs defined by non-textual hash id's can be cached
+               $expires = "+1d";
+       }
+
+       my $have_blame = gitweb_check_feature('blame');
+       open my $fd, "-|", git_cmd(), "cat-file", "blob", $hash
+               or die_error(500, "Couldn't cat $file_name, $hash");
+       my $mimetype = blob_mimetype($fd, $file_name);
+       # use 'blob_plain' (aka 'raw') view for files that cannot be displayed
+       if ($mimetype !~ m!^(?:text/|image/(?:gif|png|jpeg)$)! && -B $fd) {
+               close $fd;
+               return git_blob_plain($mimetype);
+       }
+       # we can have blame only for text/* mimetype
+       $have_blame &&= ($mimetype =~ m!^text/!);
+
+       my $highlight = gitweb_check_feature('highlight');
+       my $syntax = guess_file_syntax($highlight, $file_name);
+       $fd = run_highlighter($fd, $highlight, $syntax);
+
+       git_header_html(undef, $expires);
+       my $formats_nav = '';
+       if (defined $hash_base && (my %co = parse_commit($hash_base))) {
+               if (defined $file_name) {
+                       if ($have_blame) {
+                               $formats_nav .=
+                                       $cgi->a({-href => href(action=>"blame", -replay=>1)},
+                                               "blame") .
+                                       " | ";
+                       }
+                       $formats_nav .=
+                               $cgi->a({-href => href(action=>"history", -replay=>1)},
+                                       "history") .
+                               " | " .
+                               $cgi->a({-href => href(action=>"blob_plain", -replay=>1)},
+                                       "raw") .
+                               " | " .
+                               $cgi->a({-href => href(action=>"blob",
+                                                      hash_base=>"HEAD", file_name=>$file_name)},
+                                       "HEAD");
+               } else {
+                       $formats_nav .=
+                               $cgi->a({-href => href(action=>"blob_plain", -replay=>1)},
+                                       "raw");
+               }
+               git_print_page_nav('','', $hash_base,$co{'tree'},$hash_base, $formats_nav);
+               git_print_header_div('commit', esc_html($co{'title'}), $hash_base);
+       } else {
+               print "<div class=\"page_nav\">\n" .
+                     "<br/><br/></div>\n" .
+                     "<div class=\"title\">".esc_html($hash)."</div>\n";
+       }
+       git_print_page_path($file_name, "blob", $hash_base);
+       print "<div class=\"page_body\">\n";
+       if ($mimetype =~ m!^image/!) {
+               print qq!<img class="blob" type="!.esc_attr($mimetype).qq!"!;
+               if ($file_name) {
+                       print qq! alt="!.esc_attr($file_name).qq!" title="!.esc_attr($file_name).qq!"!;
+               }
+               print qq! src="! .
+                     href(action=>"blob_plain", hash=>$hash,
+                          hash_base=>$hash_base, file_name=>$file_name) .
+                     qq!" />\n!;
+       } else {
+               my $nr;
+               while (my $line = <$fd>) {
+                       chomp $line;
+                       $nr++;
+                       $line = untabify($line);
+                       printf qq!<div class="pre"><a id="l%i" href="%s#l%i" class="linenr">%4i</a> %s</div>\n!,
+                              $nr, esc_attr(href(-replay => 1)), $nr, $nr,
+                              $highlight ? sanitize($line) : esc_html($line, -nbsp=>1);
+               }
+       }
+       close $fd
+               or print "Reading blob failed.\n";
+       print "</div>";
+       git_footer_html();
+}
+
+sub git_tree {
+       if (!defined $hash_base) {
+               $hash_base = "HEAD";
+       }
+       if (!defined $hash) {
+               if (defined $file_name) {
+                       $hash = git_get_hash_by_path($hash_base, $file_name, "tree");
+               } else {
+                       $hash = $hash_base;
+               }
+       }
+       die_error(404, "No such tree") unless defined($hash);
+
+       my $show_sizes = gitweb_check_feature('show-sizes');
+       my $have_blame = gitweb_check_feature('blame');
+
+       my @entries = ();
+       {
+               local $/ = "\0";
+               open my $fd, "-|", git_cmd(), "ls-tree", '-z',
+                       ($show_sizes ? '-l' : ()), @extra_options, $hash
+                       or die_error(500, "Open git-ls-tree failed");
+               @entries = map { chomp; $_ } <$fd>;
+               close $fd
+                       or die_error(404, "Reading tree failed");
+       }
+
+       my $refs = git_get_references();
+       my $ref = format_ref_marker($refs, $hash_base);
+       git_header_html();
+       my $basedir = '';
+       if (defined $hash_base && (my %co = parse_commit($hash_base))) {
+               my @views_nav = ();
+               if (defined $file_name) {
+                       push @views_nav,
+                               $cgi->a({-href => href(action=>"history", -replay=>1)},
+                                       "history"),
+                               $cgi->a({-href => href(action=>"tree",
+                                                      hash_base=>"HEAD", file_name=>$file_name)},
+                                       "HEAD"),
+               }
+               my $snapshot_links = format_snapshot_links($hash);
+               if (defined $snapshot_links) {
+                       # FIXME: Should be available when we have no hash base as well.
+                       push @views_nav, $snapshot_links;
+               }
+               git_print_page_nav('tree','', $hash_base, undef, undef,
+                                  join(' | ', @views_nav));
+               git_print_header_div('commit', esc_html($co{'title'}) . $ref, $hash_base);
+       } else {
+               undef $hash_base;
+               print "<div class=\"page_nav\">\n";
+               print "<br/><br/></div>\n";
+               print "<div class=\"title\">".esc_html($hash)."</div>\n";
+       }
+       if (defined $file_name) {
+               $basedir = $file_name;
+               if ($basedir ne '' && substr($basedir, -1) ne '/') {
+                       $basedir .= '/';
+               }
+               git_print_page_path($file_name, 'tree', $hash_base);
+       }
+       print "<div class=\"page_body\">\n";
+       print "<table class=\"tree\">\n";
+       my $alternate = 1;
+       # '..' (top directory) link if possible
+       if (defined $hash_base &&
+           defined $file_name && $file_name =~ m![^/]+$!) {
+               if ($alternate) {
+                       print "<tr class=\"dark\">\n";
+               } else {
+                       print "<tr class=\"light\">\n";
+               }
+               $alternate ^= 1;
+
+               my $up = $file_name;
+               $up =~ s!/?[^/]+$!!;
+               undef $up unless $up;
+               # based on git_print_tree_entry
+               print '<td class="mode">' . mode_str('040000') . "</td>\n";
+               print '<td class="size">&nbsp;</td>'."\n" if $show_sizes;
+               print '<td class="list">';
+               print $cgi->a({-href => href(action=>"tree",
+                                            hash_base=>$hash_base,
+                                            file_name=>$up)},
+                             "..");
+               print "</td>\n";
+               print "<td class=\"link\"></td>\n";
+
+               print "</tr>\n";
+       }
+       foreach my $line (@entries) {
+               my %t = parse_ls_tree_line($line, -z => 1, -l => $show_sizes);
+
+               if ($alternate) {
+                       print "<tr class=\"dark\">\n";
+               } else {
+                       print "<tr class=\"light\">\n";
+               }
+               $alternate ^= 1;
+
+               git_print_tree_entry(\%t, $basedir, $hash_base, $have_blame);
+
+               print "</tr>\n";
+       }
+       print "</table>\n" .
+             "</div>";
+       git_footer_html();
+}
+
+sub sanitize_for_filename {
+    my $name = shift;
+
+    $name =~ s!/!-!g;
+    $name =~ s/[^[:alnum:]_.-]//g;
+
+    return $name;
+}
+
+sub snapshot_name {
+       my ($project, $hash) = @_;
+
+       # path/to/project.git  -> project
+       # path/to/project/.git -> project
+       my $name = to_utf8($project);
+       $name =~ s,([^/])/*\.git$,$1,;
+       $name = sanitize_for_filename(basename($name));
+
+       my $ver = $hash;
+       if ($hash =~ /^[0-9a-fA-F]+$/) {
+               # shorten SHA-1 hash
+               my $full_hash = git_get_full_hash($project, $hash);
+               if ($full_hash =~ /^$hash/ && length($hash) > 7) {
+                       $ver = git_get_short_hash($project, $hash);
+               }
+       } elsif ($hash =~ m!^refs/tags/(.*)$!) {
+               # tags don't need shortened SHA-1 hash
+               $ver = $1;
+       } else {
+               # branches and other need shortened SHA-1 hash
+               my $strip_refs = join '|', map { quotemeta } get_branch_refs();
+               if ($hash =~ m!^refs/($strip_refs|remotes)/(.*)$!) {
+                       my $ref_dir = (defined $1) ? $1 : '';
+                       $ver = $2;
+
+                       $ref_dir = sanitize_for_filename($ref_dir);
+                       # for refs neither in heads nor remotes we want to
+                       # add a ref dir to archive name
+                       if ($ref_dir ne '' and $ref_dir ne 'heads' and $ref_dir ne 'remotes') {
+                               $ver = $ref_dir . '-' . $ver;
+                       }
+               }
+               $ver .= '-' . git_get_short_hash($project, $hash);
+       }
+       # special case of sanitization for filename - we change
+       # slashes to dots instead of dashes
+       # in case of hierarchical branch names
+       $ver =~ s!/!.!g;
+       $ver =~ s/[^[:alnum:]_.-]//g;
+
+       # name = project-version_string
+       $name = "$name-$ver";
+
+       return wantarray ? ($name, $name) : $name;
+}
+
+sub exit_if_unmodified_since {
+       my ($latest_epoch) = @_;
+       our $cgi;
+
+       my $if_modified = $cgi->http('IF_MODIFIED_SINCE');
+       if (defined $if_modified) {
+               my $since;
+               if (eval { require HTTP::Date; 1; }) {
+                       $since = HTTP::Date::str2time($if_modified);
+               } elsif (eval { require Time::ParseDate; 1; }) {
+                       $since = Time::ParseDate::parsedate($if_modified, GMT => 1);
+               }
+               if (defined $since && $latest_epoch <= $since) {
+                       my %latest_date = parse_date($latest_epoch);
+                       print $cgi->header(
+                               -last_modified => $latest_date{'rfc2822'},
+                               -status => '304 Not Modified');
+                       goto DONE_GITWEB;
+               }
+       }
+}
+
+sub git_snapshot {
+       my $format = $input_params{'snapshot_format'};
+       if (!@snapshot_fmts) {
+               die_error(403, "Snapshots not allowed");
+       }
+       # default to first supported snapshot format
+       $format ||= $snapshot_fmts[0];
+       if ($format !~ m/^[a-z0-9]+$/) {
+               die_error(400, "Invalid snapshot format parameter");
+       } elsif (!exists($known_snapshot_formats{$format})) {
+               die_error(400, "Unknown snapshot format");
+       } elsif ($known_snapshot_formats{$format}{'disabled'}) {
+               die_error(403, "Snapshot format not allowed");
+       } elsif (!grep($_ eq $format, @snapshot_fmts)) {
+               die_error(403, "Unsupported snapshot format");
+       }
+
+       my $type = git_get_type("$hash^{}");
+       if (!$type) {
+               die_error(404, 'Object does not exist');
+       }  elsif ($type eq 'blob') {
+               die_error(400, 'Object is not a tree-ish');
+       }
+
+       my ($name, $prefix) = snapshot_name($project, $hash);
+       my $filename = "$name$known_snapshot_formats{$format}{'suffix'}";
+
+       my %co = parse_commit($hash);
+       exit_if_unmodified_since($co{'committer_epoch'}) if %co;
+
+       my $cmd = quote_command(
+               git_cmd(), 'archive',
+               "--format=$known_snapshot_formats{$format}{'format'}",
+               "--prefix=$prefix/", $hash);
+       if (exists $known_snapshot_formats{$format}{'compressor'}) {
+               $cmd .= ' | ' . quote_command(@{$known_snapshot_formats{$format}{'compressor'}});
+       }
+
+       $filename =~ s/(["\\])/\\$1/g;
+       my %latest_date;
+       if (%co) {
+               %latest_date = parse_date($co{'committer_epoch'}, $co{'committer_tz'});
+       }
+
+       print $cgi->header(
+               -type => $known_snapshot_formats{$format}{'type'},
+               -content_disposition => 'inline; filename="' . $filename . '"',
+               %co ? (-last_modified => $latest_date{'rfc2822'}) : (),
+               -status => '200 OK');
+
+       open my $fd, "-|", $cmd
+               or die_error(500, "Execute git-archive failed");
+       binmode STDOUT, ':raw';
+       print <$fd>;
+       binmode STDOUT, ':utf8'; # as set at the beginning of gitweb.cgi
+       close $fd;
+}
+
+sub git_log_generic {
+       my ($fmt_name, $body_subr, $base, $parent, $file_name, $file_hash) = @_;
+
+       my $head = git_get_head_hash($project);
+       if (!defined $base) {
+               $base = $head;
+       }
+       if (!defined $page) {
+               $page = 0;
+       }
+       my $refs = git_get_references();
+
+       my $commit_hash = $base;
+       if (defined $parent) {
+               $commit_hash = "$parent..$base";
+       }
+       my @commitlist =
+               parse_commits($commit_hash, 101, (100 * $page),
+                             defined $file_name ? ($file_name, "--full-history") : ());
+
+       my $ftype;
+       if (!defined $file_hash && defined $file_name) {
+               # some commits could have deleted file in question,
+               # and not have it in tree, but one of them has to have it
+               for (my $i = 0; $i < @commitlist; $i++) {
+                       $file_hash = git_get_hash_by_path($commitlist[$i]{'id'}, $file_name);
+                       last if defined $file_hash;
+               }
+       }
+       if (defined $file_hash) {
+               $ftype = git_get_type($file_hash);
+       }
+       if (defined $file_name && !defined $ftype) {
+               die_error(500, "Unknown type of object");
+       }
+       my %co;
+       if (defined $file_name) {
+               %co = parse_commit($base)
+                       or die_error(404, "Unknown commit object");
+       }
+
+
+       my $paging_nav = format_paging_nav($fmt_name, $page, $#commitlist >= 100);
+       my $next_link = '';
+       if ($#commitlist >= 100) {
+               $next_link =
+                       $cgi->a({-href => href(-replay=>1, page=>$page+1),
+                                -accesskey => "n", -title => "Alt-n"}, "next");
+       }
+       my $patch_max = gitweb_get_feature('patches');
+       if ($patch_max && !defined $file_name) {
+               if ($patch_max < 0 || @commitlist <= $patch_max) {
+                       $paging_nav .= " &sdot; " .
+                               $cgi->a({-href => href(action=>"patches", -replay=>1)},
+                                       "patches");
+               }
+       }
+
+       git_header_html();
+       git_print_page_nav($fmt_name,'', $hash,$hash,$hash, $paging_nav);
+       if (defined $file_name) {
+               git_print_header_div('commit', esc_html($co{'title'}), $base);
+       } else {
+               git_print_header_div('summary', $project)
+       }
+       git_print_page_path($file_name, $ftype, $hash_base)
+               if (defined $file_name);
+
+       $body_subr->(\@commitlist, 0, 99, $refs, $next_link,
+                    $file_name, $file_hash, $ftype);
+
+       git_footer_html();
+}
+
+sub git_log {
+       git_log_generic('log', \&git_log_body,
+                       $hash, $hash_parent);
+}
+
+sub git_commit {
+       $hash ||= $hash_base || "HEAD";
+       my %co = parse_commit($hash)
+           or die_error(404, "Unknown commit object");
+
+       my $parent  = $co{'parent'};
+       my $parents = $co{'parents'}; # listref
+
+       # we need to prepare $formats_nav before any parameter munging
+       my $formats_nav;
+       if (!defined $parent) {
+               # --root commitdiff
+               $formats_nav .= '(initial)';
+       } elsif (@$parents == 1) {
+               # single parent commit
+               $formats_nav .=
+                       '(parent: ' .
+                       $cgi->a({-href => href(action=>"commit",
+                                              hash=>$parent)},
+                               esc_html(substr($parent, 0, 7))) .
+                       ')';
+       } else {
+               # merge commit
+               $formats_nav .=
+                       '(merge: ' .
+                       join(' ', map {
+                               $cgi->a({-href => href(action=>"commit",
+                                                      hash=>$_)},
+                                       esc_html(substr($_, 0, 7)));
+                       } @$parents ) .
+                       ')';
+       }
+       if (gitweb_check_feature('patches') && @$parents <= 1) {
+               $formats_nav .= " | " .
+                       $cgi->a({-href => href(action=>"patch", -replay=>1)},
+                               "patch");
+       }
+
+       if (!defined $parent) {
+               $parent = "--root";
+       }
+       my @difftree;
+       open my $fd, "-|", git_cmd(), "diff-tree", '-r', "--no-commit-id",
+               @diff_opts,
+               (@$parents <= 1 ? $parent : '-c'),
+               $hash, "--"
+               or die_error(500, "Open git-diff-tree failed");
+       @difftree = map { chomp; $_ } <$fd>;
+       close $fd or die_error(404, "Reading git-diff-tree failed");
+
+       # non-textual hash id's can be cached
+       my $expires;
+       if ($hash =~ m/^[0-9a-fA-F]{40}$/) {
+               $expires = "+1d";
+       }
+       my $refs = git_get_references();
+       my $ref = format_ref_marker($refs, $co{'id'});
+
+       git_header_html(undef, $expires);
+       git_print_page_nav('commit', '',
+                          $hash, $co{'tree'}, $hash,
+                          $formats_nav);
+
+       if (defined $co{'parent'}) {
+               git_print_header_div('commitdiff', esc_html($co{'title'}) . $ref, $hash);
+       } else {
+               git_print_header_div('tree', esc_html($co{'title'}) . $ref, $co{'tree'}, $hash);
+       }
+       print "<div class=\"title_text\">\n" .
+             "<table class=\"object_header\">\n";
+       git_print_authorship_rows(\%co);
+       print "<tr><td>commit</td><td class=\"sha1\">$co{'id'}</td></tr>\n";
+       print "<tr>" .
+             "<td>tree</td>" .
+             "<td class=\"sha1\">" .
+             $cgi->a({-href => href(action=>"tree", hash=>$co{'tree'}, hash_base=>$hash),
+                      class => "list"}, $co{'tree'}) .
+             "</td>" .
+             "<td class=\"link\">" .
+             $cgi->a({-href => href(action=>"tree", hash=>$co{'tree'}, hash_base=>$hash)},
+                     "tree");
+       my $snapshot_links = format_snapshot_links($hash);
+       if (defined $snapshot_links) {
+               print " | " . $snapshot_links;
+       }
+       print "</td>" .
+             "</tr>\n";
+
+       foreach my $par (@$parents) {
+               print "<tr>" .
+                     "<td>parent</td>" .
+                     "<td class=\"sha1\">" .
+                     $cgi->a({-href => href(action=>"commit", hash=>$par),
+                              class => "list"}, $par) .
+                     "</td>" .
+                     "<td class=\"link\">" .
+                     $cgi->a({-href => href(action=>"commit", hash=>$par)}, "commit") .
+                     " | " .
+                     $cgi->a({-href => href(action=>"commitdiff", hash=>$hash, hash_parent=>$par)}, "diff") .
+                     "</td>" .
+                     "</tr>\n";
+       }
+       print "</table>".
+             "</div>\n";
+
+       print "<div class=\"page_body\">\n";
+       git_print_log($co{'comment'});
+       print "</div>\n";
+
+       git_difftree_body(\@difftree, $hash, @$parents);
+
+       git_footer_html();
+}
+
+sub git_object {
+       # object is defined by:
+       # - hash or hash_base alone
+       # - hash_base and file_name
+       my $type;
+
+       # - hash or hash_base alone
+       if ($hash || ($hash_base && !defined $file_name)) {
+               my $object_id = $hash || $hash_base;
+
+               open my $fd, "-|", quote_command(
+                       git_cmd(), 'cat-file', '-t', $object_id) . ' 2> /dev/null'
+                       or die_error(404, "Object does not exist");
+               $type = <$fd>;
+               defined $type && chomp $type;
+               close $fd
+                       or die_error(404, "Object does not exist");
+
+       # - hash_base and file_name
+       } elsif ($hash_base && defined $file_name) {
+               $file_name =~ s,/+$,,;
+
+               system(git_cmd(), "cat-file", '-e', $hash_base) == 0
+                       or die_error(404, "Base object does not exist");
+
+               # here errors should not happen
+               open my $fd, "-|", git_cmd(), "ls-tree", $hash_base, "--", $file_name
+                       or die_error(500, "Open git-ls-tree failed");
+               my $line = <$fd>;
+               close $fd;
+
+               #'100644 blob 0fa3f3a66fb6a137f6ec2c19351ed4d807070ffa  panic.c'
+               unless ($line && $line =~ m/^([0-9]+) (.+) ([0-9a-fA-F]{40})\t/) {
+                       die_error(404, "File or directory for given base does not exist");
+               }
+               $type = $2;
+               $hash = $3;
+       } else {
+               die_error(400, "Not enough information to find object");
+       }
+
+       print $cgi->redirect(-uri => href(action=>$type, -full=>1,
+                                         hash=>$hash, hash_base=>$hash_base,
+                                         file_name=>$file_name),
+                            -status => '302 Found');
+}
+
+sub git_blobdiff {
+       my $format = shift || 'html';
+       my $diff_style = $input_params{'diff_style'} || 'inline';
+
+       my $fd;
+       my @difftree;
+       my %diffinfo;
+       my $expires;
+
+       # preparing $fd and %diffinfo for git_patchset_body
+       # new style URI
+       if (defined $hash_base && defined $hash_parent_base) {
+               if (defined $file_name) {
+                       # read raw output
+                       open $fd, "-|", git_cmd(), "diff-tree", '-r', @diff_opts,
+                               $hash_parent_base, $hash_base,
+                               "--", (defined $file_parent ? $file_parent : ()), $file_name
+                               or die_error(500, "Open git-diff-tree failed");
+                       @difftree = map { chomp; $_ } <$fd>;
+                       close $fd
+                               or die_error(404, "Reading git-diff-tree failed");
+                       @difftree
+                               or die_error(404, "Blob diff not found");
+
+               } elsif (defined $hash &&
+                        $hash =~ /[0-9a-fA-F]{40}/) {
+                       # try to find filename from $hash
+
+                       # read filtered raw output
+                       open $fd, "-|", git_cmd(), "diff-tree", '-r', @diff_opts,
+                               $hash_parent_base, $hash_base, "--"
+                               or die_error(500, "Open git-diff-tree failed");
+                       @difftree =
+                               # ':100644 100644 03b21826... 3b93d5e7... M     ls-files.c'
+                               # $hash == to_id
+                               grep { /^:[0-7]{6} [0-7]{6} [0-9a-fA-F]{40} $hash/ }
+                               map { chomp; $_ } <$fd>;
+                       close $fd
+                               or die_error(404, "Reading git-diff-tree failed");
+                       @difftree
+                               or die_error(404, "Blob diff not found");
+
+               } else {
+                       die_error(400, "Missing one of the blob diff parameters");
+               }
+
+               if (@difftree > 1) {
+                       die_error(400, "Ambiguous blob diff specification");
+               }
+
+               %diffinfo = parse_difftree_raw_line($difftree[0]);
+               $file_parent ||= $diffinfo{'from_file'} || $file_name;
+               $file_name   ||= $diffinfo{'to_file'};
+
+               $hash_parent ||= $diffinfo{'from_id'};
+               $hash        ||= $diffinfo{'to_id'};
+
+               # non-textual hash id's can be cached
+               if ($hash_base =~ m/^[0-9a-fA-F]{40}$/ &&
+                   $hash_parent_base =~ m/^[0-9a-fA-F]{40}$/) {
+                       $expires = '+1d';
+               }
+
+               # open patch output
+               open $fd, "-|", git_cmd(), "diff-tree", '-r', @diff_opts,
+                       '-p', ($format eq 'html' ? "--full-index" : ()),
+                       $hash_parent_base, $hash_base,
+                       "--", (defined $file_parent ? $file_parent : ()), $file_name
+                       or die_error(500, "Open git-diff-tree failed");
+       }
+
+       # old/legacy style URI -- not generated anymore since 1.4.3.
+       if (!%diffinfo) {
+               die_error('404 Not Found', "Missing one of the blob diff parameters")
+       }
+
+       # header
+       if ($format eq 'html') {
+               my $formats_nav =
+                       $cgi->a({-href => href(action=>"blobdiff_plain", -replay=>1)},
+                               "raw");
+               $formats_nav .= diff_style_nav($diff_style);
+               git_header_html(undef, $expires);
+               if (defined $hash_base && (my %co = parse_commit($hash_base))) {
+                       git_print_page_nav('','', $hash_base,$co{'tree'},$hash_base, $formats_nav);
+                       git_print_header_div('commit', esc_html($co{'title'}), $hash_base);
+               } else {
+                       print "<div class=\"page_nav\"><br/>$formats_nav<br/></div>\n";
+                       print "<div class=\"title\">".esc_html("$hash vs $hash_parent")."</div>\n";
+               }
+               if (defined $file_name) {
+                       git_print_page_path($file_name, "blob", $hash_base);
+               } else {
+                       print "<div class=\"page_path\"></div>\n";
+               }
+
+       } elsif ($format eq 'plain') {
+               print $cgi->header(
+                       -type => 'text/plain',
+                       -charset => 'utf-8',
+                       -expires => $expires,
+                       -content_disposition => 'inline; filename="' . "$file_name" . '.patch"');
+
+               print "X-Git-Url: " . $cgi->self_url() . "\n\n";
+
+       } else {
+               die_error(400, "Unknown blobdiff format");
+       }
+
+       # patch
+       if ($format eq 'html') {
+               print "<div class=\"page_body\">\n";
+
+               git_patchset_body($fd, $diff_style,
+                                 [ \%diffinfo ], $hash_base, $hash_parent_base);
+               close $fd;
+
+               print "</div>\n"; # class="page_body"
+               git_footer_html();
+
+       } else {
+               while (my $line = <$fd>) {
+                       $line =~ s!a/($hash|$hash_parent)!'a/'.esc_path($diffinfo{'from_file'})!eg;
+                       $line =~ s!b/($hash|$hash_parent)!'b/'.esc_path($diffinfo{'to_file'})!eg;
+
+                       print $line;
+
+                       last if $line =~ m!^\+\+\+!;
+               }
+               local $/ = undef;
+               print <$fd>;
+               close $fd;
+       }
+}
+
+sub git_blobdiff_plain {
+       git_blobdiff('plain');
+}
+
+# assumes that it is added as later part of already existing navigation,
+# so it returns "| foo | bar" rather than just "foo | bar"
+sub diff_style_nav {
+       my ($diff_style, $is_combined) = @_;
+       $diff_style ||= 'inline';
+
+       return "" if ($is_combined);
+
+       my @styles = (inline => 'inline', 'sidebyside' => 'side by side');
+       my %styles = @styles;
+       @styles =
+               @styles[ map { $_ * 2 } 0..$#styles/2 ];
+
+       return join '',
+               map { " | ".$_ }
+               map {
+                       $_ eq $diff_style ? $styles{$_} :
+                       $cgi->a({-href => href(-replay=>1, diff_style => $_)}, $styles{$_})
+               } @styles;
+}
+
+sub git_commitdiff {
+       my %params = @_;
+       my $format = $params{-format} || 'html';
+       my $diff_style = $input_params{'diff_style'} || 'inline';
+
+       my ($patch_max) = gitweb_get_feature('patches');
+       if ($format eq 'patch') {
+               die_error(403, "Patch view not allowed") unless $patch_max;
+       }
+
+       $hash ||= $hash_base || "HEAD";
+       my %co = parse_commit($hash)
+           or die_error(404, "Unknown commit object");
+
+       # choose format for commitdiff for merge
+       if (! defined $hash_parent && @{$co{'parents'}} > 1) {
+               $hash_parent = '--cc';
+       }
+       # we need to prepare $formats_nav before almost any parameter munging
+       my $formats_nav;
+       if ($format eq 'html') {
+               $formats_nav =
+                       $cgi->a({-href => href(action=>"commitdiff_plain", -replay=>1)},
+                               "raw");
+               if ($patch_max && @{$co{'parents'}} <= 1) {
+                       $formats_nav .= " | " .
+                               $cgi->a({-href => href(action=>"patch", -replay=>1)},
+                                       "patch");
+               }
+               $formats_nav .= diff_style_nav($diff_style, @{$co{'parents'}} > 1);
+
+               if (defined $hash_parent &&
+                   $hash_parent ne '-c' && $hash_parent ne '--cc') {
+                       # commitdiff with two commits given
+                       my $hash_parent_short = $hash_parent;
+                       if ($hash_parent =~ m/^[0-9a-fA-F]{40}$/) {
+                               $hash_parent_short = substr($hash_parent, 0, 7);
+                       }
+                       $formats_nav .=
+                               ' (from';
+                       for (my $i = 0; $i < @{$co{'parents'}}; $i++) {
+                               if ($co{'parents'}[$i] eq $hash_parent) {
+                                       $formats_nav .= ' parent ' . ($i+1);
+                                       last;
+                               }
+                       }
+                       $formats_nav .= ': ' .
+                               $cgi->a({-href => href(-replay=>1,
+                                                      hash=>$hash_parent, hash_base=>undef)},
+                                       esc_html($hash_parent_short)) .
+                               ')';
+               } elsif (!$co{'parent'}) {
+                       # --root commitdiff
+                       $formats_nav .= ' (initial)';
+               } elsif (scalar @{$co{'parents'}} == 1) {
+                       # single parent commit
+                       $formats_nav .=
+                               ' (parent: ' .
+                               $cgi->a({-href => href(-replay=>1,
+                                                      hash=>$co{'parent'}, hash_base=>undef)},
+                                       esc_html(substr($co{'parent'}, 0, 7))) .
+                               ')';
+               } else {
+                       # merge commit
+                       if ($hash_parent eq '--cc') {
+                               $formats_nav .= ' | ' .
+                                       $cgi->a({-href => href(-replay=>1,
+                                                              hash=>$hash, hash_parent=>'-c')},
+                                               'combined');
+                       } else { # $hash_parent eq '-c'
+                               $formats_nav .= ' | ' .
+                                       $cgi->a({-href => href(-replay=>1,
+                                                              hash=>$hash, hash_parent=>'--cc')},
+                                               'compact');
+                       }
+                       $formats_nav .=
+                               ' (merge: ' .
+                               join(' ', map {
+                                       $cgi->a({-href => href(-replay=>1,
+                                                              hash=>$_, hash_base=>undef)},
+                                               esc_html(substr($_, 0, 7)));
+                               } @{$co{'parents'}} ) .
+                               ')';
+               }
+       }
+
+       my $hash_parent_param = $hash_parent;
+       if (!defined $hash_parent_param) {
+               # --cc for multiple parents, --root for parentless
+               $hash_parent_param =
+                       @{$co{'parents'}} > 1 ? '--cc' : $co{'parent'} || '--root';
+       }
+
+       # read commitdiff
+       my $fd;
+       my @difftree;
+       if ($format eq 'html') {
+               open $fd, "-|", git_cmd(), "diff-tree", '-r', @diff_opts,
+                       "--no-commit-id", "--patch-with-raw", "--full-index",
+                       $hash_parent_param, $hash, "--"
+                       or die_error(500, "Open git-diff-tree failed");
+
+               while (my $line = <$fd>) {
+                       chomp $line;
+                       # empty line ends raw part of diff-tree output
+                       last unless $line;
+                       push @difftree, scalar parse_difftree_raw_line($line);
+               }
+
+       } elsif ($format eq 'plain') {
+               open $fd, "-|", git_cmd(), "diff-tree", '-r', @diff_opts,
+                       '-p', $hash_parent_param, $hash, "--"
+                       or die_error(500, "Open git-diff-tree failed");
+       } elsif ($format eq 'patch') {
+               # For commit ranges, we limit the output to the number of
+               # patches specified in the 'patches' feature.
+               # For single commits, we limit the output to a single patch,
+               # diverging from the git-format-patch default.
+               my @commit_spec = ();
+               if ($hash_parent) {
+                       if ($patch_max > 0) {
+                               push @commit_spec, "-$patch_max";
+                       }
+                       push @commit_spec, '-n', "$hash_parent..$hash";
+               } else {
+                       if ($params{-single}) {
+                               push @commit_spec, '-1';
+                       } else {
+                               if ($patch_max > 0) {
+                                       push @commit_spec, "-$patch_max";
+                               }
+                               push @commit_spec, "-n";
+                       }
+                       push @commit_spec, '--root', $hash;
+               }
+               open $fd, "-|", git_cmd(), "format-patch", @diff_opts,
+                       '--encoding=utf8', '--stdout', @commit_spec
+                       or die_error(500, "Open git-format-patch failed");
+       } else {
+               die_error(400, "Unknown commitdiff format");
+       }
+
+       # non-textual hash id's can be cached
+       my $expires;
+       if ($hash =~ m/^[0-9a-fA-F]{40}$/) {
+               $expires = "+1d";
+       }
+
+       # write commit message
+       if ($format eq 'html') {
+               my $refs = git_get_references();
+               my $ref = format_ref_marker($refs, $co{'id'});
+
+               git_header_html(undef, $expires);
+               git_print_page_nav('commitdiff','', $hash,$co{'tree'},$hash, $formats_nav);
+               git_print_header_div('commit', esc_html($co{'title'}) . $ref, $hash);
+               print "<div class=\"title_text\">\n" .
+                     "<table class=\"object_header\">\n";
+               git_print_authorship_rows(\%co);
+               print "</table>".
+                     "</div>\n";
+               print "<div class=\"page_body\">\n";
+               if (@{$co{'comment'}} > 1) {
+                       print "<div class=\"log\">\n";
+                       git_print_log($co{'comment'}, -final_empty_line=> 1, -remove_title => 1);
+                       print "</div>\n"; # class="log"
+               }
+
+       } elsif ($format eq 'plain') {
+               my $refs = git_get_references("tags");
+               my $tagname = git_get_rev_name_tags($hash);
+               my $filename = basename($project) . "-$hash.patch";
+
+               print $cgi->header(
+                       -type => 'text/plain',
+                       -charset => 'utf-8',
+                       -expires => $expires,
+                       -content_disposition => 'inline; filename="' . "$filename" . '"');
+               my %ad = parse_date($co{'author_epoch'}, $co{'author_tz'});
+               print "From: " . to_utf8($co{'author'}) . "\n";
+               print "Date: $ad{'rfc2822'} ($ad{'tz_local'})\n";
+               print "Subject: " . to_utf8($co{'title'}) . "\n";
+
+               print "X-Git-Tag: $tagname\n" if $tagname;
+               print "X-Git-Url: " . $cgi->self_url() . "\n\n";
+
+               foreach my $line (@{$co{'comment'}}) {
+                       print to_utf8($line) . "\n";
+               }
+               print "---\n\n";
+       } elsif ($format eq 'patch') {
+               my $filename = basename($project) . "-$hash.patch";
+
+               print $cgi->header(
+                       -type => 'text/plain',
+                       -charset => 'utf-8',
+                       -expires => $expires,
+                       -content_disposition => 'inline; filename="' . "$filename" . '"');
+       }
+
+       # write patch
+       if ($format eq 'html') {
+               my $use_parents = !defined $hash_parent ||
+                       $hash_parent eq '-c' || $hash_parent eq '--cc';
+               git_difftree_body(\@difftree, $hash,
+                                 $use_parents ? @{$co{'parents'}} : $hash_parent);
+               print "<br/>\n";
+
+               git_patchset_body($fd, $diff_style,
+                                 \@difftree, $hash,
+                                 $use_parents ? @{$co{'parents'}} : $hash_parent);
+               close $fd;
+               print "</div>\n"; # class="page_body"
+               git_footer_html();
+
+       } elsif ($format eq 'plain') {
+               local $/ = undef;
+               print <$fd>;
+               close $fd
+                       or print "Reading git-diff-tree failed\n";
+       } elsif ($format eq 'patch') {
+               local $/ = undef;
+               print <$fd>;
+               close $fd
+                       or print "Reading git-format-patch failed\n";
+       }
+}
+
+sub git_commitdiff_plain {
+       git_commitdiff(-format => 'plain');
+}
+
+# format-patch-style patches
+sub git_patch {
+       git_commitdiff(-format => 'patch', -single => 1);
+}
+
+sub git_patches {
+       git_commitdiff(-format => 'patch');
+}
+
+sub git_history {
+       git_log_generic('history', \&git_history_body,
+                       $hash_base, $hash_parent_base,
+                       $file_name, $hash);
+}
+
+sub git_search {
+       $searchtype ||= 'commit';
+
+       # check if appropriate features are enabled
+       gitweb_check_feature('search')
+               or die_error(403, "Search is disabled");
+       if ($searchtype eq 'pickaxe') {
+               # pickaxe may take all resources of your box and run for several minutes
+               # with every query - so decide by yourself how public you make this feature
+               gitweb_check_feature('pickaxe')
+                       or die_error(403, "Pickaxe search is disabled");
+       }
+       if ($searchtype eq 'grep') {
+               # grep search might be potentially CPU-intensive, too
+               gitweb_check_feature('grep')
+                       or die_error(403, "Grep search is disabled");
+       }
+
+       if (!defined $searchtext) {
+               die_error(400, "Text field is empty");
+       }
+       if (!defined $hash) {
+               $hash = git_get_head_hash($project);
+       }
+       my %co = parse_commit($hash);
+       if (!%co) {
+               die_error(404, "Unknown commit object");
+       }
+       if (!defined $page) {
+               $page = 0;
+       }
+
+       if ($searchtype eq 'commit' ||
+           $searchtype eq 'author' ||
+           $searchtype eq 'committer') {
+               git_search_message(%co);
+       } elsif ($searchtype eq 'pickaxe') {
+               git_search_changes(%co);
+       } elsif ($searchtype eq 'grep') {
+               git_search_files(%co);
+       } else {
+               die_error(400, "Unknown search type");
+       }
+}
+
+sub git_search_help {
+       git_header_html();
+       git_print_page_nav('','', $hash,$hash,$hash);
+       print <<EOT;
+<p><strong>Pattern</strong> is by default a normal string that is matched precisely (but without
+regard to case, except in the case of pickaxe). However, when you check the <em>re</em> checkbox,
+the pattern entered is recognized as the POSIX extended
+<a href="http://en.wikipedia.org/wiki/Regular_expression">regular expression</a> (also case
+insensitive).</p>
+<dl>
+<dt><b>commit</b></dt>
+<dd>The commit messages and authorship information will be scanned for the given pattern.</dd>
+EOT
+       my $have_grep = gitweb_check_feature('grep');
+       if ($have_grep) {
+               print <<EOT;
+<dt><b>grep</b></dt>
+<dd>All files in the currently selected tree (HEAD unless you are explicitly browsing
+    a different one) are searched for the given pattern. On large trees, this search can take
+a while and put some strain on the server, so please use it with some consideration. Note that
+due to git-grep peculiarity, currently if regexp mode is turned off, the matches are
+case-sensitive.</dd>
+EOT
+       }
+       print <<EOT;
+<dt><b>author</b></dt>
+<dd>Name and e-mail of the change author and date of birth of the patch will be scanned for the given pattern.</dd>
+<dt><b>committer</b></dt>
+<dd>Name and e-mail of the committer and date of commit will be scanned for the given pattern.</dd>
+EOT
+       my $have_pickaxe = gitweb_check_feature('pickaxe');
+       if ($have_pickaxe) {
+               print <<EOT;
+<dt><b>pickaxe</b></dt>
+<dd>All commits that caused the string to appear or disappear from any file (changes that
+added, removed or "modified" the string) will be listed. This search can take a while and
+takes a lot of strain on the server, so please use it wisely. Note that since you may be
+interested even in changes just changing the case as well, this search is case sensitive.</dd>
+EOT
+       }
+       print "</dl>\n";
+       git_footer_html();
+}
+
+sub git_shortlog {
+       git_log_generic('shortlog', \&git_shortlog_body,
+                       $hash, $hash_parent);
+}
+
+## ......................................................................
+## feeds (RSS, Atom; OPML)
+
+sub git_feed {
+       my $format = shift || 'atom';
+       my $have_blame = gitweb_check_feature('blame');
+
+       # Atom: http://www.atomenabled.org/developers/syndication/
+       # RSS:  http://www.notestips.com/80256B3A007F2692/1/NAMO5P9UPQ
+       if ($format ne 'rss' && $format ne 'atom') {
+               die_error(400, "Unknown web feed format");
+       }
+
+       # log/feed of current (HEAD) branch, log of given branch, history of file/directory
+       my $head = $hash || 'HEAD';
+       my @commitlist = parse_commits($head, 150, 0, $file_name);
+
+       my %latest_commit;
+       my %latest_date;
+       my $content_type = "application/$format+xml";
+       if (defined $cgi->http('HTTP_ACCEPT') &&
+                $cgi->Accept('text/xml') > $cgi->Accept($content_type)) {
+               # browser (feed reader) prefers text/xml
+               $content_type = 'text/xml';
+       }
+       if (defined($commitlist[0])) {
+               %latest_commit = %{$commitlist[0]};
+               my $latest_epoch = $latest_commit{'committer_epoch'};
+               exit_if_unmodified_since($latest_epoch);
+               %latest_date = parse_date($latest_epoch, $latest_commit{'committer_tz'});
+       }
+       print $cgi->header(
+               -type => $content_type,
+               -charset => 'utf-8',
+               %latest_date ? (-last_modified => $latest_date{'rfc2822'}) : (),
+               -status => '200 OK');
+
+       # Optimization: skip generating the body if client asks only
+       # for Last-Modified date.
+       return if ($cgi->request_method() eq 'HEAD');
+
+       # header variables
+       my $title = "$site_name - $project/$action";
+       my $feed_type = 'log';
+       if (defined $hash) {
+               $title .= " - '$hash'";
+               $feed_type = 'branch log';
+               if (defined $file_name) {
+                       $title .= " :: $file_name";
+                       $feed_type = 'history';
+               }
+       } elsif (defined $file_name) {
+               $title .= " - $file_name";
+               $feed_type = 'history';
+       }
+       $title .= " $feed_type";
+       $title = esc_html($title);
+       my $descr = git_get_project_description($project);
+       if (defined $descr) {
+               $descr = esc_html($descr);
+       } else {
+               $descr = "$project " .
+                        ($format eq 'rss' ? 'RSS' : 'Atom') .
+                        " feed";
+       }
+       my $owner = git_get_project_owner($project);
+       $owner = esc_html($owner);
+
+       #header
+       my $alt_url;
+       if (defined $file_name) {
+               $alt_url = href(-full=>1, action=>"history", hash=>$hash, file_name=>$file_name);
+       } elsif (defined $hash) {
+               $alt_url = href(-full=>1, action=>"log", hash=>$hash);
+       } else {
+               $alt_url = href(-full=>1, action=>"summary");
+       }
+       print qq!<?xml version="1.0" encoding="utf-8"?>\n!;
+       if ($format eq 'rss') {
+               print <<XML;
+<rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/">
+<channel>
+XML
+               print "<title>$title</title>\n" .
+                     "<link>$alt_url</link>\n" .
+                     "<description>$descr</description>\n" .
+                     "<language>en</language>\n" .
+                     # project owner is responsible for 'editorial' content
+                     "<managingEditor>$owner</managingEditor>\n";
+               if (defined $logo || defined $favicon) {
+                       # prefer the logo to the favicon, since RSS
+                       # doesn't allow both
+                       my $img = esc_url($logo || $favicon);
+                       print "<image>\n" .
+                             "<url>$img</url>\n" .
+                             "<title>$title</title>\n" .
+                             "<link>$alt_url</link>\n" .
+                             "</image>\n";
+               }
+               if (%latest_date) {
+                       print "<pubDate>$latest_date{'rfc2822'}</pubDate>\n";
+                       print "<lastBuildDate>$latest_date{'rfc2822'}</lastBuildDate>\n";
+               }
+               print "<generator>gitweb v.$version/$git_version</generator>\n";
+       } elsif ($format eq 'atom') {
+               print <<XML;
+<feed xmlns="http://www.w3.org/2005/Atom">
+XML
+               print "<title>$title</title>\n" .
+                     "<subtitle>$descr</subtitle>\n" .
+                     '<link rel="alternate" type="text/html" href="' .
+                     $alt_url . '" />' . "\n" .
+                     '<link rel="self" type="' . $content_type . '" href="' .
+                     $cgi->self_url() . '" />' . "\n" .
+                     "<id>" . href(-full=>1) . "</id>\n" .
+                     # use project owner for feed author
+                     "<author><name>$owner</name></author>\n";
+               if (defined $favicon) {
+                       print "<icon>" . esc_url($favicon) . "</icon>\n";
+               }
+               if (defined $logo) {
+                       # not twice as wide as tall: 72 x 27 pixels
+                       print "<logo>" . esc_url($logo) . "</logo>\n";
+               }
+               if (! %latest_date) {
+                       # dummy date to keep the feed valid until commits trickle in:
+                       print "<updated>1970-01-01T00:00:00Z</updated>\n";
+               } else {
+                       print "<updated>$latest_date{'iso-8601'}</updated>\n";
+               }
+               print "<generator version='$version/$git_version'>gitweb</generator>\n";
+       }
+
+       # contents
+       for (my $i = 0; $i <= $#commitlist; $i++) {
+               my %co = %{$commitlist[$i]};
+               my $commit = $co{'id'};
+               # we read 150, we always show 30 and the ones more recent than 48 hours
+               if (($i >= 20) && ((time - $co{'author_epoch'}) > 48*60*60)) {
+                       last;
+               }
+               my %cd = parse_date($co{'author_epoch'}, $co{'author_tz'});
+
+               # get list of changed files
+               open my $fd, "-|", git_cmd(), "diff-tree", '-r', @diff_opts,
+                       $co{'parent'} || "--root",
+                       $co{'id'}, "--", (defined $file_name ? $file_name : ())
+                       or next;
+               my @difftree = map { chomp; $_ } <$fd>;
+               close $fd
+                       or next;
+
+               # print element (entry, item)
+               my $co_url = href(-full=>1, action=>"commitdiff", hash=>$commit);
+               if ($format eq 'rss') {
+                       print "<item>\n" .
+                             "<title>" . esc_html($co{'title'}) . "</title>\n" .
+                             "<author>" . esc_html($co{'author'}) . "</author>\n" .
+                             "<pubDate>$cd{'rfc2822'}</pubDate>\n" .
+                             "<guid isPermaLink=\"true\">$co_url</guid>\n" .
+                             "<link>$co_url</link>\n" .
+                             "<description>" . esc_html($co{'title'}) . "</description>\n" .
+                             "<content:encoded>" .
+                             "<![CDATA[\n";
+               } elsif ($format eq 'atom') {
+                       print "<entry>\n" .
+                             "<title type=\"html\">" . esc_html($co{'title'}) . "</title>\n" .
+                             "<updated>$cd{'iso-8601'}</updated>\n" .
+                             "<author>\n" .
+                             "  <name>" . esc_html($co{'author_name'}) . "</name>\n";
+                       if ($co{'author_email'}) {
+                               print "  <email>" . esc_html($co{'author_email'}) . "</email>\n";
+                       }
+                       print "</author>\n" .
+                             # use committer for contributor
+                             "<contributor>\n" .
+                             "  <name>" . esc_html($co{'committer_name'}) . "</name>\n";
+                       if ($co{'committer_email'}) {
+                               print "  <email>" . esc_html($co{'committer_email'}) . "</email>\n";
+                       }
+                       print "</contributor>\n" .
+                             "<published>$cd{'iso-8601'}</published>\n" .
+                             "<link rel=\"alternate\" type=\"text/html\" href=\"$co_url\" />\n" .
+                             "<id>$co_url</id>\n" .
+                             "<content type=\"xhtml\" xml:base=\"" . esc_url($my_url) . "\">\n" .
+                             "<div xmlns=\"http://www.w3.org/1999/xhtml\">\n";
+               }
+               my $comment = $co{'comment'};
+               print "<pre>\n";
+               foreach my $line (@$comment) {
+                       $line = esc_html($line);
+                       print "$line\n";
+               }
+               print "</pre><ul>\n";
+               foreach my $difftree_line (@difftree) {
+                       my %difftree = parse_difftree_raw_line($difftree_line);
+                       next if !$difftree{'from_id'};
+
+                       my $file = $difftree{'file'} || $difftree{'to_file'};
+
+                       print "<li>" .
+                             "[" .
+                             $cgi->a({-href => href(-full=>1, action=>"blobdiff",
+                                                    hash=>$difftree{'to_id'}, hash_parent=>$difftree{'from_id'},
+                                                    hash_base=>$co{'id'}, hash_parent_base=>$co{'parent'},
+                                                    file_name=>$file, file_parent=>$difftree{'from_file'}),
+                                     -title => "diff"}, 'D');
+                       if ($have_blame) {
+                               print $cgi->a({-href => href(-full=>1, action=>"blame",
+                                                            file_name=>$file, hash_base=>$commit),
+                                             -title => "blame"}, 'B');
+                       }
+                       # if this is not a feed of a file history
+                       if (!defined $file_name || $file_name ne $file) {
+                               print $cgi->a({-href => href(-full=>1, action=>"history",
+                                                            file_name=>$file, hash=>$commit),
+                                             -title => "history"}, 'H');
+                       }
+                       $file = esc_path($file);
+                       print "] ".
+                             "$file</li>\n";
+               }
+               if ($format eq 'rss') {
+                       print "</ul>]]>\n" .
+                             "</content:encoded>\n" .
+                             "</item>\n";
+               } elsif ($format eq 'atom') {
+                       print "</ul>\n</div>\n" .
+                             "</content>\n" .
+                             "</entry>\n";
+               }
+       }
+
+       # end of feed
+       if ($format eq 'rss') {
+               print "</channel>\n</rss>\n";
+       } elsif ($format eq 'atom') {
+               print "</feed>\n";
+       }
+}
+
+sub git_rss {
+       git_feed('rss');
+}
+
+sub git_atom {
+       git_feed('atom');
+}
+
+sub git_opml {
+       my @list = git_get_projects_list($project_filter, $strict_export);
+       if (!@list) {
+               die_error(404, "No projects found");
+       }
+
+       print $cgi->header(
+               -type => 'text/xml',
+               -charset => 'utf-8',
+               -content_disposition => 'inline; filename="opml.xml"');
+
+       my $title = esc_html($site_name);
+       my $filter = " within subdirectory ";
+       if (defined $project_filter) {
+               $filter .= esc_html($project_filter);
+       } else {
+               $filter = "";
+       }
+       print <<XML;
+<?xml version="1.0" encoding="utf-8"?>
+<opml version="1.0">
+<head>
+  <title>$title OPML Export$filter</title>
+</head>
+<body>
+<outline text="git RSS feeds">
+XML
+
+       foreach my $pr (@list) {
+               my %proj = %$pr;
+               my $head = git_get_head_hash($proj{'path'});
+               if (!defined $head) {
+                       next;
+               }
+               $git_dir = "$projectroot/$proj{'path'}";
+               my %co = parse_commit($head);
+               if (!%co) {
+                       next;
+               }
+
+               my $path = esc_html(chop_str($proj{'path'}, 25, 5));
+               my $rss  = href('project' => $proj{'path'}, 'action' => 'rss', -full => 1);
+               my $html = href('project' => $proj{'path'}, 'action' => 'summary', -full => 1);
+               print "<outline type=\"rss\" text=\"$path\" title=\"$path\" xmlUrl=\"$rss\" htmlUrl=\"$html\"/>\n";
+       }
+       print <<XML;
+</outline>
+</body>
+</opml>
+XML
+}
diff --git a/static/git-favicon.png b/static/git-favicon.png
new file mode 100644 (file)
index 0000000..4fa44bb
Binary files /dev/null and b/static/git-favicon.png differ
diff --git a/static/git-logo.png b/static/git-logo.png
new file mode 100644 (file)
index 0000000..fdaf7b7
Binary files /dev/null and b/static/git-logo.png differ
diff --git a/static/gitweb.css b/static/gitweb.css
new file mode 100644 (file)
index 0000000..db1b045
--- /dev/null
@@ -0,0 +1,686 @@
+body {
+       font-family: sans-serif;
+       font-size: large;
+       border: solid #d9d8d1;
+       border-width: 1px;
+       margin: 10px;
+       background-color: #ffffff;
+       color: #000000;
+}
+
+a {
+       color: #0000cc;
+}
+
+a:hover, a:visited, a:active {
+       color: #880000;
+}
+
+span.cntrl {
+       border: dashed #aaaaaa;
+       border-width: 1px;
+       padding: 0px 2px 0px 2px;
+       margin:  0px 2px 0px 2px;
+}
+
+img.logo {
+       float: right;
+       border-width: 0px;
+}
+
+img.avatar {
+       vertical-align: middle;
+}
+
+img.blob {
+       max-height: 100%;
+       max-width: 100%;
+}
+
+a.list img.avatar {
+       border-style: none;
+}
+
+div.page_header {
+       height: 25px;
+       padding: 8px;
+       font-size: 150%;
+       font-weight: bold;
+       background-color: #d9d8d1;
+}
+
+div.page_header a:visited, a.header {
+       color: #0000cc;
+}
+
+div.page_header a:hover {
+       color: #880000;
+}
+
+div.page_nav {
+       padding: 8px;
+}
+
+div.page_nav a:visited {
+       color: #0000cc;
+}
+
+div.page_path {
+       padding: 8px;
+       font-weight: bold;
+       border: solid #d9d8d1;
+       border-width: 0px 0px 1px;
+}
+
+div.page_footer {
+       height: 22px;
+       padding: 4px 8px;
+       background-color: #d9d8d1;
+}
+
+div.page_footer_text {
+       line-height: 22px;
+       float: left;
+       color: #555555;
+       font-style: italic;
+}
+
+div#generating_info {
+       margin: 4px;
+       font-size: smaller;
+       text-align: center;
+       color: #505050;
+}
+
+div.page_body {
+       padding: 8px;
+       font-family: monospace;
+}
+
+div.title, a.title {
+       display: block;
+       padding: 6px 8px;
+       font-weight: bold;
+       background-color: #edece6;
+       text-decoration: none;
+       color: #000000;
+}
+
+div.readme {
+       padding: 8px;
+}
+
+a.title:hover {
+       background-color: #d9d8d1;
+}
+
+div.title_text {
+       padding: 6px 0px;
+       border: solid #d9d8d1;
+       border-width: 0px 0px 1px;
+       font-family: monospace;
+}
+
+div.log_body {
+       padding: 8px 8px 8px 150px;
+}
+
+span.age {
+       position: relative;
+       float: left;
+       width: 142px;
+       font-style: italic;
+}
+
+span.signoff {
+       color: #888888;
+}
+
+div.log_link {
+       padding: 0px 8px;
+       font-size: 70%;
+       font-family: sans-serif;
+       font-style: normal;
+       position: relative;
+       float: left;
+       width: 136px;
+}
+
+div.list_head {
+       padding: 6px 8px 4px;
+       border: solid #d9d8d1;
+       border-width: 1px 0px 0px;
+       font-style: italic;
+}
+
+.author_date, .author {
+       font-style: italic;
+}
+
+div.author_date {
+       padding: 8px;
+       border: solid #d9d8d1;
+       border-width: 0px 0px 1px 0px;
+}
+
+a.list {
+       text-decoration: none;
+       color: #000000;
+}
+
+a.subject, a.name {
+       font-weight: bold;
+}
+
+table.tags a.subject {
+       font-weight: normal;
+}
+
+a.list:hover {
+       text-decoration: underline;
+       color: #880000;
+}
+
+a.text {
+       text-decoration: none;
+       color: #0000cc;
+}
+
+a.text:visited {
+       text-decoration: none;
+       color: #880000;
+}
+
+a.text:hover {
+       text-decoration: underline;
+       color: #880000;
+}
+
+table {
+       padding: 8px 4px;
+       border-spacing: 0;
+}
+
+table.diff_tree {
+       font-family: monospace;
+}
+
+table.combined.diff_tree th {
+       text-align: center;
+}
+
+table.combined.diff_tree td {
+       padding-right: 24px;
+}
+
+table.combined.diff_tree th.link,
+table.combined.diff_tree td.link {
+       padding: 0px 2px;
+}
+
+table.combined.diff_tree td.nochange a {
+       color: #6666ff;
+}
+
+table.combined.diff_tree td.nochange a:hover,
+table.combined.diff_tree td.nochange a:visited {
+       color: #d06666;
+}
+
+table.blame {
+       border-collapse: collapse;
+}
+
+table.blame td {
+       padding: 0px 5px;
+       font-size: 100%;
+       vertical-align: top;
+}
+
+th {
+       padding: 2px 5px;
+       font-size: 100%;
+       text-align: left;
+}
+
+/* do not change row style on hover for 'blame' view */
+tr.light,
+table.blame .light:hover {
+       background-color: #ffffff;
+}
+
+tr.dark,
+table.blame .dark:hover {
+       background-color: #f6f6f0;
+}
+
+/* currently both use the same, but it can change */
+tr.light:hover,
+tr.dark:hover {
+       background-color: #edece6;
+}
+
+/* boundary commits in 'blame' view */
+/* and commits without "previous" */
+tr.boundary td.sha1,
+tr.no-previous td.linenr {
+       font-weight: bold;
+}
+
+/* for 'blame_incremental', during processing */
+tr.color1 { background-color: #f6fff6; }
+tr.color2 { background-color: #f6f6ff; }
+tr.color3 { background-color: #fff6f6; }
+
+td {
+       padding: 10px 20px;
+       font-size: 100%;
+       vertical-align: top;
+}
+
+td.link, td.selflink {
+       padding: 2px 5px;
+       font-family: sans-serif;
+       font-size: 70%;
+}
+
+td.selflink {
+       padding-right: 0px;
+}
+
+td.sha1 {
+       font-family: monospace;
+}
+
+.error {
+       color: red;
+       background-color: yellow;
+}
+
+td.current_head {
+       text-decoration: underline;
+}
+
+td.category {
+       background-color: #d9d8d1;
+       border-top: 1px solid #000000;
+       border-left: 1px solid #000000;
+       font-weight: bold;
+}
+
+table.diff_tree span.file_status.new {
+       color: #008000;
+}
+
+table.diff_tree span.file_status.deleted {
+       color: #c00000;
+}
+
+table.diff_tree span.file_status.moved,
+table.diff_tree span.file_status.mode_chnge {
+       color: #777777;
+}
+
+table.diff_tree span.file_status.copied {
+  color: #70a070;
+}
+
+/* noage: "No commits" */
+table.project_list td.noage {
+       color: #808080;
+       font-style: italic;
+}
+
+/* age2: 60*60*24*2 <= age */
+table.project_list td.age2, table.blame td.age2 {
+       font-style: italic;
+}
+
+/* age1: 60*60*2 <= age < 60*60*24*2 */
+table.project_list td.age1 {
+       color: #009900;
+       font-style: italic;
+}
+
+table.blame td.age1 {
+       color: #009900;
+       background: transparent;
+}
+
+/* age0: age < 60*60*2 */
+table.project_list td.age0 {
+       color: #009900;
+       font-style: italic;
+       font-weight: bold;
+}
+
+table.blame td.age0 {
+       color: #009900;
+       background: transparent;
+       font-weight: bold;
+}
+
+td.pre, div.pre, div.diff {
+       font-family: monospace;
+       font-size: 12px;
+       white-space: pre;
+}
+
+td.mode {
+       font-family: monospace;
+}
+
+/* progress of blame_interactive */
+div#progress_bar {
+       height: 2px;
+       margin-bottom: -2px;
+       background-color: #d8d9d0;
+}
+div#progress_info {
+       float: right;
+       text-align: right;
+}
+
+/* format of (optional) objects size in 'tree' view */
+td.size {
+       font-family: monospace;
+       text-align: right;
+}
+
+/* styling of diffs (patchsets): commitdiff and blobdiff views */
+div.diff.header,
+div.diff.extended_header {
+       white-space: normal;
+}
+
+div.diff.header {
+       font-weight: bold;
+
+       background-color: #edece6;
+
+       margin-top: 4px;
+       padding: 4px 0px 2px 0px;
+       border: solid #d9d8d1;
+       border-width: 1px 0px 1px 0px;
+}
+
+div.diff.header a.path {
+       text-decoration: underline;
+}
+
+div.diff.extended_header,
+div.diff.extended_header a.path,
+div.diff.extended_header a.hash {
+       color: #777777;
+}
+
+div.diff.extended_header .info {
+       color: #b0b0b0;
+}
+
+div.diff.extended_header {
+       background-color: #f6f5ee;
+       padding: 2px 0px 2px 0px;
+}
+
+div.diff a.list,
+div.diff a.path,
+div.diff a.hash {
+       text-decoration: none;
+}
+
+div.diff a.list:hover,
+div.diff a.path:hover,
+div.diff a.hash:hover {
+       text-decoration: underline;
+}
+
+div.diff.to_file a.path,
+div.diff.to_file {
+       color: #007000;
+}
+
+div.diff.add {
+       color: #008800;
+}
+
+div.diff.add span.marked {
+       background-color: #aaffaa;
+}
+
+div.diff.from_file a.path,
+div.diff.from_file {
+       color: #aa0000;
+}
+
+div.diff.rem {
+       color: #cc0000;
+}
+
+div.diff.rem span.marked {
+       background-color: #ffaaaa;
+}
+
+div.diff.chunk_header a,
+div.diff.chunk_header {
+       color: #990099;
+}
+
+div.diff.chunk_header {
+       border: dotted #ffe0ff;
+       border-width: 1px 0px 0px 0px;
+       margin-top: 2px;
+}
+
+div.diff.chunk_header span.chunk_info {
+       background-color: #ffeeff;
+}
+
+div.diff.chunk_header span.section {
+       color: #aa22aa;
+}
+
+div.diff.incomplete {
+       color: #cccccc;
+}
+
+div.diff.nodifferences {
+       font-weight: bold;
+       color: #600000;
+}
+
+/* side-by-side diff */
+div.chunk_block {
+       overflow: hidden;
+}
+
+div.chunk_block div.old {
+       float: left;
+       width: 50%;
+       overflow: hidden;
+}
+
+div.chunk_block div.new {
+       margin-left: 50%;
+       width: 50%;
+}
+
+div.chunk_block.rem div.old div.diff.rem {
+       background-color: #fff5f5;
+}
+div.chunk_block.add div.new div.diff.add {
+       background-color: #f8fff8;
+}
+div.chunk_block.chg div     div.diff {
+       background-color: #fffff0;
+}
+div.chunk_block.ctx div     div.diff.ctx {
+       color: #404040;
+}
+
+
+div.index_include {
+       border: solid #d9d8d1;
+       border-width: 0px 0px 1px;
+       padding: 12px 8px;
+}
+
+div.search {
+       font-size: 100%;
+       font-weight: normal;
+       margin: 4px 8px;
+       float: right;
+       top: 56px;
+       right: 12px
+}
+
+div.projsearch {
+       text-align: center;
+       margin: 20px 0px;
+}
+
+div.projsearch form {
+       margin-bottom: 2px;
+}
+
+td.linenr {
+       text-align: right;
+}
+
+a.linenr {
+       color: #999999;
+       text-decoration: none
+}
+
+a.rss_logo {
+       float: right;
+       padding: 3px 5px;
+       line-height: 10px;
+       border: 1px solid;
+       border-color: #fcc7a5 #7d3302 #3e1a01 #ff954e;
+       color: #ffffff;
+       background-color: #ff6600;
+       font-weight: bold;
+       font-family: sans-serif;
+       font-size: 70%;
+       text-align: center;
+       text-decoration: none;
+}
+
+a.rss_logo:hover {
+       background-color: #ee5500;
+}
+
+a.rss_logo.generic {
+       background-color: #ff8800;
+}
+
+a.rss_logo.generic:hover {
+       background-color: #ee7700;
+}
+
+span.refs span {
+       padding: 0px 4px;
+       font-size: 70%;
+       font-weight: normal;
+       border: 1px solid;
+       background-color: #ffaaff;
+       border-color: #ffccff #ff00ee #ff00ee #ffccff;
+}
+
+span.refs span a {
+       text-decoration: none;
+       color: inherit;
+}
+
+span.refs span a:hover {
+       text-decoration: underline;
+}
+
+span.refs span.indirect {
+       font-style: italic;
+}
+
+span.refs span.ref {
+       background-color: #aaaaff;
+       border-color: #ccccff #0033cc #0033cc #ccccff;
+}
+
+span.refs span.tag {
+       background-color: #ffffaa;
+       border-color: #ffffcc #ffee00 #ffee00 #ffffcc;
+}
+
+span.refs span.head {
+       background-color: #aaffaa;
+       border-color: #ccffcc #00cc33 #00cc33 #ccffcc;
+}
+
+span.atnight {
+       color: #cc0000;
+}
+
+span.match {
+       color: #e00000;
+}
+
+div.binary {
+       font-style: italic;
+}
+
+div.remote {
+       margin: .5em;
+       border: 1px solid #d9d8d1;
+       display: inline-block;
+}
+
+/* JavaScript-based timezone manipulation */
+
+.popup { /* timezone selection UI */
+       position: absolute;
+       /* "top: 0; right: 0;" would be better, if not for bugs in browsers */
+       top: 0; left: 0;
+       border: 1px solid;
+       padding: 2px;
+       background-color: #f0f0f0;
+       font-style: normal;
+       color: #000000;
+       cursor: auto;
+}
+
+.close-button { /* close timezone selection UI without selecting */
+       /* float doesn't work within absolutely positioned container,
+        * if width of container is not set explicitly */
+       /* float: right; */
+       position: absolute;
+       top: 0px; right: 0px;
+       border:  1px solid green;
+       margin:  1px 1px 1px 1px;
+       padding-bottom: 2px;
+       width:     12px;
+       height:    10px;
+       font-size:  9px;
+       font-weight: bold;
+       text-align: center;
+       background-color: #fff0f0;
+       cursor: pointer;
+}
+
+
+/* Style definition generated by highlight 2.4.5, http://www.andre-simon.de/ */
+
+/* Highlighting theme definition: */
+
+.num    { color:#2928ff; }
+.esc    { color:#ff00ff; }
+.str    { color:#ff0000; }
+.dstr   { color:#818100; }
+.slc    { color:#838183; font-style:italic; }
+.com    { color:#838183; font-style:italic; }
+.dir    { color:#008200; }
+.sym    { color:#000000; }
+.line   { color:#555555; }
+.kwa    { color:#000000; font-weight:bold; }
+.kwb    { color:#830000; }
+.kwc    { color:#000000; font-weight:bold; }
+.kwd    { color:#010181; }
diff --git a/static/gitweb.js b/static/gitweb.js
new file mode 100644 (file)
index 0000000..a971109
--- /dev/null
@@ -0,0 +1,1579 @@
+// Copyright (C) 2007, Fredrik Kuivinen <frekui@gmail.com>
+//               2007, Petr Baudis <pasky@suse.cz>
+//          2008-2011, Jakub Narebski <jnareb@gmail.com>
+
+/**
+ * @fileOverview Generic JavaScript code (helper functions)
+ * @license GPLv2 or later
+ */
+
+
+/* ============================================================ */
+/* ............................................................ */
+/* Padding */
+
+/**
+ * pad INPUT on the left with STR that is assumed to have visible
+ * width of single character (for example nonbreakable spaces),
+ * to WIDTH characters
+ *
+ * example: padLeftStr(12, 3, '\u00A0') == '\u00A012'
+ *          ('\u00A0' is nonbreakable space)
+ *
+ * @param {Number|String} input: number to pad
+ * @param {Number} width: visible width of output
+ * @param {String} str: string to prefix to string, defaults to '\u00A0'
+ * @returns {String} INPUT prefixed with STR x (WIDTH - INPUT.length)
+ */
+function padLeftStr(input, width, str) {
+       var prefix = '';
+       if (typeof str === 'undefined') {
+               ch = '\u00A0'; // using '&nbsp;' doesn't work in all browsers
+       }
+
+       width -= input.toString().length;
+       while (width > 0) {
+               prefix += str;
+               width--;
+       }
+       return prefix + input;
+}
+
+/**
+ * Pad INPUT on the left to WIDTH, using given padding character CH,
+ * for example padLeft('a', 3, '_') is '__a'
+ *             padLeft(4, 2) is '04' (same as padLeft(4, 2, '0'))
+ *
+ * @param {String} input: input value converted to string.
+ * @param {Number} width: desired length of output.
+ * @param {String} ch: single character to prefix to string, defaults to '0'.
+ *
+ * @returns {String} Modified string, at least SIZE length.
+ */
+function padLeft(input, width, ch) {
+       var s = input + "";
+       if (typeof ch === 'undefined') {
+               ch = '0';
+       }
+
+       while (s.length < width) {
+               s = ch + s;
+       }
+       return s;
+}
+
+
+/* ............................................................ */
+/* Handling browser incompatibilities */
+
+/**
+ * Create XMLHttpRequest object in cross-browser way
+ * @returns XMLHttpRequest object, or null
+ */
+function createRequestObject() {
+       try {
+               return new XMLHttpRequest();
+       } catch (e) {}
+       try {
+               return window.createRequest();
+       } catch (e) {}
+       try {
+               return new ActiveXObject("Msxml2.XMLHTTP");
+       } catch (e) {}
+       try {
+               return new ActiveXObject("Microsoft.XMLHTTP");
+       } catch (e) {}
+
+       return null;
+}
+
+
+/**
+ * Insert rule giving specified STYLE to given SELECTOR at the end of
+ * first CSS stylesheet.
+ *
+ * @param {String} selector: CSS selector, e.g. '.class'
+ * @param {String} style: rule contents, e.g. 'background-color: red;'
+ */
+function addCssRule(selector, style) {
+       var stylesheet = document.styleSheets[0];
+
+       var theRules = [];
+       if (stylesheet.cssRules) {     // W3C way
+               theRules = stylesheet.cssRules;
+       } else if (stylesheet.rules) { // IE way
+               theRules = stylesheet.rules;
+       }
+
+       if (stylesheet.insertRule) {    // W3C way
+               stylesheet.insertRule(selector + ' { ' + style + ' }', theRules.length);
+       } else if (stylesheet.addRule) { // IE way
+               stylesheet.addRule(selector, style);
+       }
+}
+
+
+/* ............................................................ */
+/* Support for legacy browsers */
+
+/**
+ * Provides getElementsByClassName method, if there is no native
+ * implementation of this method.
+ *
+ * NOTE that there are limits and differences compared to native
+ * getElementsByClassName as defined by e.g.:
+ *   https://developer.mozilla.org/en/DOM/document.getElementsByClassName
+ *   http://www.whatwg.org/specs/web-apps/current-work/multipage/dom.html#dom-getelementsbyclassname
+ *   http://www.whatwg.org/specs/web-apps/current-work/multipage/dom.html#dom-document-getelementsbyclassname
+ *
+ * Namely, this implementation supports only single class name as
+ * argument and not set of space-separated tokens representing classes,
+ * it returns Array of nodes rather than live NodeList, and has
+ * additional optional argument where you can limit search to given tags
+ * (via getElementsByTagName).
+ *
+ * Based on
+ *   http://code.google.com/p/getelementsbyclassname/
+ *   http://www.dustindiaz.com/getelementsbyclass/
+ *   http://stackoverflow.com/questions/1818865/do-we-have-getelementsbyclassname-in-javascript
+ *
+ * See also http://ejohn.org/blog/getelementsbyclassname-speed-comparison/
+ *
+ * @param {String} class: name of _single_ class to find
+ * @param {String} [taghint] limit search to given tags
+ * @returns {Node[]} array of matching elements
+ */
+if (!('getElementsByClassName' in document)) {
+       document.getElementsByClassName = function (classname, taghint) {
+               taghint = taghint || "*";
+               var elements = (taghint === "*" && document.all) ?
+                              document.all :
+                              document.getElementsByTagName(taghint);
+               var pattern = new RegExp("(^|\\s)" + classname + "(\\s|$)");
+               var matches= [];
+               for (var i = 0, j = 0, n = elements.length; i < n; i++) {
+                       var el= elements[i];
+                       if (el.className && pattern.test(el.className)) {
+                               // matches.push(el);
+                               matches[j] = el;
+                               j++;
+                       }
+               }
+               return matches;
+       };
+} // end if
+
+
+/* ............................................................ */
+/* unquoting/unescaping filenames */
+
+/**#@+
+ * @constant
+ */
+var escCodeRe = /\\([^0-7]|[0-7]{1,3})/g;
+var octEscRe = /^[0-7]{1,3}$/;
+var maybeQuotedRe = /^\"(.*)\"$/;
+/**#@-*/
+
+/**
+ * unquote maybe C-quoted filename (as used by git, i.e. it is
+ * in double quotes '"' if there is any escape character used)
+ * e.g. 'aa' -> 'aa', '"a\ta"' -> 'a   a'
+ *
+ * @param {String} str: git-quoted string
+ * @returns {String} Unquoted and unescaped string
+ *
+ * @globals escCodeRe, octEscRe, maybeQuotedRe
+ */
+function unquote(str) {
+       function unq(seq) {
+               var es = {
+                       // character escape codes, aka escape sequences (from C)
+                       // replacements are to some extent JavaScript specific
+                       t: "\t",   // tab            (HT, TAB)
+                       n: "\n",   // newline        (NL)
+                       r: "\r",   // return         (CR)
+                       f: "\f",   // form feed      (FF)
+                       b: "\b",   // backspace      (BS)
+                       a: "\x07", // alarm (bell)   (BEL)
+                       e: "\x1B", // escape         (ESC)
+                       v: "\v"    // vertical tab   (VT)
+               };
+
+               if (seq.search(octEscRe) !== -1) {
+                       // octal char sequence
+                       return String.fromCharCode(parseInt(seq, 8));
+               } else if (seq in es) {
+                       // C escape sequence, aka character escape code
+                       return es[seq];
+               }
+               // quoted ordinary character
+               return seq;
+       }
+
+       var match = str.match(maybeQuotedRe);
+       if (match) {
+               str = match[1];
+               // perhaps str = eval('"'+str+'"'); would be enough?
+               str = str.replace(escCodeRe,
+                       function (substr, p1, offset, s) { return unq(p1); });
+       }
+       return str;
+}
+
+/* end of common-lib.js */
+// Copyright (C) 2007, Fredrik Kuivinen <frekui@gmail.com>
+//               2007, Petr Baudis <pasky@suse.cz>
+//          2008-2011, Jakub Narebski <jnareb@gmail.com>
+
+/**
+ * @fileOverview Datetime manipulation: parsing and formatting
+ * @license GPLv2 or later
+ */
+
+
+/* ............................................................ */
+/* parsing and retrieving datetime related information */
+
+/**
+ * used to extract hours and minutes from timezone info, e.g '-0900'
+ * @constant
+ */
+var tzRe = /^([+\-])([0-9][0-9])([0-9][0-9])$/;
+
+/**
+ * convert numeric timezone +/-ZZZZ to offset from UTC in seconds
+ *
+ * @param {String} timezoneInfo: numeric timezone '(+|-)HHMM'
+ * @returns {Number} offset from UTC in seconds for timezone
+ *
+ * @globals tzRe
+ */
+function timezoneOffset(timezoneInfo) {
+       var match = tzRe.exec(timezoneInfo);
+       var tz_sign = (match[1] === '-' ? -1 : +1);
+       var tz_hour = parseInt(match[2],10);
+       var tz_min  = parseInt(match[3],10);
+
+       return tz_sign*(((tz_hour*60) + tz_min)*60);
+}
+
+/**
+ * return local (browser) timezone as offset from UTC in seconds
+ *
+ * @returns {Number} offset from UTC in seconds for local timezone
+ */
+function localTimezoneOffset() {
+       // getTimezoneOffset returns the time-zone offset from UTC,
+       // in _minutes_, for the current locale
+       return ((new Date()).getTimezoneOffset() * -60);
+}
+
+/**
+ * return local (browser) timezone as numeric timezone '(+|-)HHMM'
+ *
+ * @returns {String} locat timezone as -/+ZZZZ
+ */
+function localTimezoneInfo() {
+       var tzOffsetMinutes = (new Date()).getTimezoneOffset() * -1;
+
+       return formatTimezoneInfo(0, tzOffsetMinutes);
+}
+
+
+/**
+ * Parse RFC-2822 date into a Unix timestamp (into epoch)
+ *
+ * @param {String} date: date in RFC-2822 format, e.g. 'Thu, 21 Dec 2000 16:01:07 +0200'
+ * @returns {Number} epoch i.e. seconds since '00:00:00 1970-01-01 UTC'
+ */
+function parseRFC2822Date(date) {
+       // Date.parse accepts the IETF standard (RFC 1123 Section 5.2.14 and elsewhere)
+       // date syntax, which is defined in RFC 2822 (obsoletes RFC 822)
+       // and returns number of _milli_seconds since January 1, 1970, 00:00:00 UTC
+       return Date.parse(date) / 1000;
+}
+
+
+/* ............................................................ */
+/* formatting date */
+
+/**
+ * format timezone offset as numerical timezone '(+|-)HHMM' or '(+|-)HH:MM'
+ *
+ * @param {Number} hours:    offset in hours, e.g. 2 for '+0200'
+ * @param {Number} [minutes] offset in minutes, e.g. 30 for '-4030';
+ *                           it is split into hours if not 0 <= minutes < 60,
+ *                           for example 1200 would give '+0100';
+ *                           defaults to 0
+ * @param {String} [sep] separator between hours and minutes part,
+ *                       default is '', might be ':' for W3CDTF (rfc-3339)
+ * @returns {String} timezone in '(+|-)HHMM' or '(+|-)HH:MM' format
+ */
+function formatTimezoneInfo(hours, minutes, sep) {
+       minutes = minutes || 0; // to be able to use formatTimezoneInfo(hh)
+       sep = sep || ''; // default format is +/-ZZZZ
+
+       if (minutes < 0 || minutes > 59) {
+               hours = minutes > 0 ? Math.floor(minutes / 60) : Math.ceil(minutes / 60);
+               minutes = Math.abs(minutes - 60*hours); // sign of minutes is sign of hours
+               // NOTE: this works correctly because there is no UTC-00:30 timezone
+       }
+
+       var tzSign = hours >= 0 ? '+' : '-';
+       if (hours < 0) {
+               hours = -hours; // sign is stored in tzSign
+       }
+
+       return tzSign + padLeft(hours, 2, '0') + sep + padLeft(minutes, 2, '0');
+}
+
+/**
+ * translate 'utc' and 'local' to numerical timezone
+ * @param {String} timezoneInfo: might be 'utc' or 'local' (browser)
+ */
+function normalizeTimezoneInfo(timezoneInfo) {
+       switch (timezoneInfo) {
+       case 'utc':
+               return '+0000';
+       case 'local': // 'local' is browser timezone
+               return localTimezoneInfo();
+       }
+       return timezoneInfo;
+}
+
+
+/**
+ * return date in local time formatted in iso-8601 like format
+ * 'yyyy-mm-dd HH:MM:SS +/-ZZZZ' e.g. '2005-08-07 21:49:46 +0200'
+ *
+ * @param {Number} epoch: seconds since '00:00:00 1970-01-01 UTC'
+ * @param {String} timezoneInfo: numeric timezone '(+|-)HHMM'
+ * @returns {String} date in local time in iso-8601 like format
+ */
+function formatDateISOLocal(epoch, timezoneInfo) {
+       // date corrected by timezone
+       var localDate = new Date(1000 * (epoch +
+               timezoneOffset(timezoneInfo)));
+       var localDateStr = // e.g. '2005-08-07'
+               localDate.getUTCFullYear()                 + '-' +
+               padLeft(localDate.getUTCMonth()+1, 2, '0') + '-' +
+               padLeft(localDate.getUTCDate(),    2, '0');
+       var localTimeStr = // e.g. '21:49:46'
+               padLeft(localDate.getUTCHours(),   2, '0') + ':' +
+               padLeft(localDate.getUTCMinutes(), 2, '0') + ':' +
+               padLeft(localDate.getUTCSeconds(), 2, '0');
+
+       return localDateStr + ' ' + localTimeStr + ' ' + timezoneInfo;
+}
+
+/**
+ * return date in local time formatted in rfc-2822 format
+ * e.g. 'Thu, 21 Dec 2000 16:01:07 +0200'
+ *
+ * @param {Number} epoch: seconds since '00:00:00 1970-01-01 UTC'
+ * @param {String} timezoneInfo: numeric timezone '(+|-)HHMM'
+ * @param {Boolean} [padDay] e.g. 'Sun, 07 Aug' if true, 'Sun, 7 Aug' otherwise
+ * @returns {String} date in local time in rfc-2822 format
+ */
+function formatDateRFC2882(epoch, timezoneInfo, padDay) {
+       // A short textual representation of a month, three letters
+       var months = ["Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug", "Sep", "Oct", "Nov", "Dec"];
+       // A textual representation of a day, three letters
+       var days = ["Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"];
+       // date corrected by timezone
+       var localDate = new Date(1000 * (epoch +
+               timezoneOffset(timezoneInfo)));
+       var localDateStr = // e.g. 'Sun, 7 Aug 2005' or 'Sun, 07 Aug 2005'
+               days[localDate.getUTCDay()] + ', ' +
+               (padDay ? padLeft(localDate.getUTCDate(),2,'0') : localDate.getUTCDate()) + ' ' +
+               months[localDate.getUTCMonth()] + ' ' +
+               localDate.getUTCFullYear();
+       var localTimeStr = // e.g. '21:49:46'
+               padLeft(localDate.getUTCHours(),   2, '0') + ':' +
+               padLeft(localDate.getUTCMinutes(), 2, '0') + ':' +
+               padLeft(localDate.getUTCSeconds(), 2, '0');
+
+       return localDateStr + ' ' + localTimeStr + ' ' + timezoneInfo;
+}
+
+/* end of datetime.js */
+/**
+ * @fileOverview Accessing cookies from JavaScript
+ * @license GPLv2 or later
+ */
+
+/*
+ * Based on subsection "Cookies in JavaScript" of "Professional
+ * JavaScript for Web Developers" by Nicholas C. Zakas and cookie
+ * plugin from jQuery (dual licensed under the MIT and GPL licenses)
+ */
+
+
+/**
+ * Create a cookie with the given name and value,
+ * and other optional parameters.
+ *
+ * @example
+ *   setCookie('foo', 'bar'); // will be deleted when browser exits
+ *   setCookie('foo', 'bar', { expires: new Date(Date.parse('Jan 1, 2012')) });
+ *   setCookie('foo', 'bar', { expires: 7 }); // 7 days = 1 week
+ *   setCookie('foo', 'bar', { expires: 14, path: '/' });
+ *
+ * @param {String} sName:    Unique name of a cookie (letters, numbers, underscores).
+ * @param {String} sValue:   The string value stored in a cookie.
+ * @param {Object} [options] An object literal containing key/value pairs
+ *                           to provide optional cookie attributes.
+ * @param {String|Number|Date} [options.expires] Either literal string to be used as cookie expires,
+ *                            or an integer specifying the expiration date from now on in days,
+ *                            or a Date object to be used as cookie expiration date.
+ *                            If a negative value is specified or a date in the past),
+ *                            the cookie will be deleted.
+ *                            If set to null or omitted, the cookie will be a session cookie
+ *                            and will not be retained when the browser exits.
+ * @param {String} [options.path] Restrict access of a cookie to particular directory
+ *                               (default: path of page that created the cookie).
+ * @param {String} [options.domain] Override what web sites are allowed to access cookie
+ *                                  (default: domain of page that created the cookie).
+ * @param {Boolean} [options.secure] If true, the secure attribute of the cookie will be set
+ *                                   and the cookie would be accessible only from secure sites
+ *                                   (cookie transmission will require secure protocol like HTTPS).
+ */
+function setCookie(sName, sValue, options) {
+       options = options || {};
+       if (sValue === null) {
+               sValue = '';
+               option.expires = 'delete';
+       }
+
+       var sCookie = sName + '=' + encodeURIComponent(sValue);
+
+       if (options.expires) {
+               var oExpires = options.expires, sDate;
+               if (oExpires === 'delete') {
+                       sDate = 'Thu, 01 Jan 1970 00:00:00 GMT';
+               } else if (typeof oExpires === 'string') {
+                       sDate = oExpires;
+               } else {
+                       var oDate;
+                       if (typeof oExpires === 'number') {
+                               oDate = new Date();
+                               oDate.setTime(oDate.getTime() + (oExpires * 24 * 60 * 60 * 1000)); // days to ms
+                       } else {
+                               oDate = oExpires;
+                       }
+                       sDate = oDate.toGMTString();
+               }
+               sCookie += '; expires=' + sDate;
+       }
+
+       if (options.path) {
+               sCookie += '; path=' + (options.path);
+       }
+       if (options.domain) {
+               sCookie += '; domain=' + (options.domain);
+       }
+       if (options.secure) {
+               sCookie += '; secure';
+       }
+       document.cookie = sCookie;
+}
+
+/**
+ * Get the value of a cookie with the given name.
+ *
+ * @param {String} sName: Unique name of a cookie (letters, numbers, underscores)
+ * @returns {String|null} The string value stored in a cookie
+ */
+function getCookie(sName) {
+       var sRE = '(?:; )?' + sName + '=([^;]*);?';
+       var oRE = new RegExp(sRE);
+       if (oRE.test(document.cookie)) {
+               return decodeURIComponent(RegExp['$1']);
+       } else {
+               return null;
+       }
+}
+
+/**
+ * Delete cookie with given name
+ *
+ * @param {String} sName:    Unique name of a cookie (letters, numbers, underscores)
+ * @param {Object} [options] An object literal containing key/value pairs
+ *                           to provide optional cookie attributes.
+ * @param {String} [options.path]   Must be the same as when setting a cookie
+ * @param {String} [options.domain] Must be the same as when setting a cookie
+ */
+function deleteCookie(sName, options) {
+       options = options || {};
+       options.expires = 'delete';
+
+       setCookie(sName, '', options);
+}
+
+/* end of cookies.js */
+// Copyright (C) 2007, Fredrik Kuivinen <frekui@gmail.com>
+//               2007, Petr Baudis <pasky@suse.cz>
+//          2008-2011, Jakub Narebski <jnareb@gmail.com>
+
+/**
+ * @fileOverview Detect if JavaScript is enabled, and pass it to server-side
+ * @license GPLv2 or later
+ */
+
+
+/* ============================================================ */
+/* Manipulating links */
+
+/**
+ * used to check if link has 'js' query parameter already (at end),
+ * and other reasons to not add 'js=1' param at the end of link
+ * @constant
+ */
+var jsExceptionsRe = /[;?]js=[01](#.*)?$/;
+
+/**
+ * Add '?js=1' or ';js=1' to the end of every link in the document
+ * that doesn't have 'js' query parameter set already.
+ *
+ * Links with 'js=1' lead to JavaScript version of given action, if it
+ * exists (currently there is only 'blame_incremental' for 'blame')
+ *
+ * To be used as `window.onload` handler
+ *
+ * @globals jsExceptionsRe
+ */
+function fixLinks() {
+       var allLinks = document.getElementsByTagName("a") || document.links;
+       for (var i = 0, len = allLinks.length; i < len; i++) {
+               var link = allLinks[i];
+               if (!jsExceptionsRe.test(link)) {
+                       link.href = link.href.replace(/(#|$)/,
+                               (link.href.indexOf('?') === -1 ? '?' : ';') + 'js=1$1');
+               }
+       }
+}
+
+/* end of javascript-detection.js */
+// Copyright (C) 2011, John 'Warthog9' Hawley <warthog9@eaglescrag.net>
+//               2011, Jakub Narebski <jnareb@gmail.com>
+
+/**
+ * @fileOverview Manipulate dates in gitweb output, adjusting timezone
+ * @license GPLv2 or later
+ */
+
+/**
+ * Get common timezone, add UI for changing timezones, and adjust
+ * dates to use requested common timezone.
+ *
+ * This function is called during onload event (added to window.onload).
+ *
+ * @param {String} tzDefault: default timezone, if there is no cookie
+ * @param {Object} tzCookieInfo: object literal with info about cookie to store timezone
+ * @param {String} tzCookieInfo.name: name of cookie to store timezone
+ * @param {String} tzClassName: denotes elements with date to be adjusted
+ */
+function onloadTZSetup(tzDefault, tzCookieInfo, tzClassName) {
+       var tzCookieTZ = getCookie(tzCookieInfo.name, tzCookieInfo);
+       var tz = tzDefault;
+
+       if (tzCookieTZ) {
+               // set timezone to value saved in a cookie
+               tz = tzCookieTZ;
+               // refresh cookie, so its expiration counts from last use of gitweb
+               setCookie(tzCookieInfo.name, tzCookieTZ, tzCookieInfo);
+       }
+
+       // add UI for changing timezone
+       addChangeTZ(tz, tzCookieInfo, tzClassName);
+
+       // server-side of gitweb produces datetime in UTC,
+       // so if tz is 'utc' there is no need for changes
+       var nochange = tz === 'utc';
+
+       // adjust dates to use specified common timezone
+       fixDatetimeTZ(tz, tzClassName, nochange);
+}
+
+
+/* ...................................................................... */
+/* Changing dates to use requested timezone */
+
+/**
+ * Replace RFC-2822 dates contained in SPAN elements with tzClassName
+ * CSS class with equivalent dates in given timezone.
+ *
+ * @param {String} tz: numeric timezone in '(-|+)HHMM' format, or 'utc', or 'local'
+ * @param {String} tzClassName: specifies elements to be changed
+ * @param {Boolean} nochange: markup for timezone change, but don't change it
+ */
+function fixDatetimeTZ(tz, tzClassName, nochange) {
+       // sanity check, method should be ensured by common-lib.js
+       if (!document.getElementsByClassName) {
+               return;
+       }
+
+       // translate to timezone in '(-|+)HHMM' format
+       tz = normalizeTimezoneInfo(tz);
+
+       // NOTE: result of getElementsByClassName should probably be cached
+       var classesFound = document.getElementsByClassName(tzClassName, "span");
+       for (var i = 0, len = classesFound.length; i < len; i++) {
+               var curElement = classesFound[i];
+
+               curElement.title = 'Click to change timezone';
+               if (!nochange) {
+                       // we use *.firstChild.data (W3C DOM) instead of *.innerHTML
+                       // as the latter doesn't always work everywhere in every browser
+                       var epoch = parseRFC2822Date(curElement.firstChild.data);
+                       var adjusted = formatDateRFC2882(epoch, tz);
+
+                       curElement.firstChild.data = adjusted;
+               }
+       }
+}
+
+
+/* ...................................................................... */
+/* Adding triggers, generating timezone menu, displaying and hiding */
+
+/**
+ * Adds triggers for UI to change common timezone used for dates in
+ * gitweb output: it marks up and/or creates item to click to invoke
+ * timezone change UI, creates timezone UI fragment to be attached,
+ * and installs appropriate onclick trigger (via event delegation).
+ *
+ * @param {String} tzSelected: pre-selected timezone,
+ *                             'utc' or 'local' or '(-|+)HHMM'
+ * @param {Object} tzCookieInfo: object literal with info about cookie to store timezone
+ * @param {String} tzClassName: specifies elements to install trigger
+ */
+function addChangeTZ(tzSelected, tzCookieInfo, tzClassName) {
+       // make link to timezone UI discoverable
+       addCssRule('.'+tzClassName + ':hover',
+                  'text-decoration: underline; cursor: help;');
+
+       // create form for selecting timezone (to be saved in a cookie)
+       var tzSelectFragment = document.createDocumentFragment();
+       tzSelectFragment = createChangeTZForm(tzSelectFragment,
+                                             tzSelected, tzCookieInfo, tzClassName);
+
+       // event delegation handler for timezone selection UI (clicking on entry)
+       // see http://www.nczonline.net/blog/2009/06/30/event-delegation-in-javascript/
+       // assumes that there is no existing document.onclick handler
+       document.onclick = function onclickHandler(event) {
+               //IE doesn't pass in the event object
+               event = event || window.event;
+
+               //IE uses srcElement as the target
+               var target = event.target || event.srcElement;
+
+               switch (target.className) {
+               case tzClassName:
+                       // don't display timezone menu if it is already displayed
+                       if (tzSelectFragment.childNodes.length > 0) {
+                               displayChangeTZForm(target, tzSelectFragment);
+                       }
+                       break;
+               } // end switch
+       };
+}
+
+/**
+ * Create DocumentFragment with UI for changing common timezone in
+ * which dates are shown in.
+ *
+ * @param {DocumentFragment} documentFragment: where attach UI
+ * @param {String} tzSelected: default (pre-selected) timezone
+ * @param {Object} tzCookieInfo: object literal with info about cookie to store timezone
+ * @returns {DocumentFragment}
+ */
+function createChangeTZForm(documentFragment, tzSelected, tzCookieInfo, tzClassName) {
+       var div = document.createElement("div");
+       div.className = 'popup';
+
+       /* '<div class="close-button" title="(click on this box to close)">X</div>' */
+       var closeButton = document.createElement('div');
+       closeButton.className = 'close-button';
+       closeButton.title = '(click on this box to close)';
+       closeButton.appendChild(document.createTextNode('X'));
+       closeButton.onclick = closeTZFormHandler(documentFragment, tzClassName);
+       div.appendChild(closeButton);
+
+       /* 'Select timezone: <br clear="all">' */
+       div.appendChild(document.createTextNode('Select timezone: '));
+       var br = document.createElement('br');
+       br.clear = 'all';
+       div.appendChild(br);
+
+       /* '<select name="tzoffset">
+        *    ...
+        *    <option value="-0700">UTC-07:00</option>
+        *    <option value="-0600">UTC-06:00</option>
+        *    ...
+        *  </select>' */
+       var select = document.createElement("select");
+       select.name = "tzoffset";
+       //select.style.clear = 'all';
+       select.appendChild(generateTZOptions(tzSelected));
+       select.onchange = selectTZHandler(documentFragment, tzCookieInfo, tzClassName);
+       div.appendChild(select);
+
+       documentFragment.appendChild(div);
+
+       return documentFragment;
+}
+
+
+/**
+ * Hide (remove from DOM) timezone change UI, ensuring that it is not
+ * garbage collected and that it can be re-enabled later.
+ *
+ * @param {DocumentFragment} documentFragment: contains detached UI
+ * @param {HTMLSelectElement} target: select element inside of UI
+ * @param {String} tzClassName: specifies element where UI was installed
+ * @returns {DocumentFragment} documentFragment
+ */
+function removeChangeTZForm(documentFragment, target, tzClassName) {
+       // find containing element, where we appended timezone selection UI
+       // `target' is somewhere inside timezone menu
+       var container = target.parentNode, popup = target;
+       while (container &&
+              container.className !== tzClassName) {
+               popup = container;
+               container = container.parentNode;
+       }
+       // safety check if we found correct container,
+       // and if it isn't deleted already
+       if (!container || !popup ||
+           container.className !== tzClassName ||
+           popup.className     !== 'popup') {
+               return documentFragment;
+       }
+
+       // timezone selection UI was appended as last child
+       // see also displayChangeTZForm function
+       var removed = popup.parentNode.removeChild(popup);
+       if (documentFragment.firstChild !== removed) { // the only child
+               // re-append it so it would be available for next time
+               documentFragment.appendChild(removed);
+       }
+       // all of inline style was added by this script
+       // it is not really needed to remove it, but it is a good practice
+       container.removeAttribute('style');
+
+       return documentFragment;
+}
+
+
+/**
+ * Display UI for changing common timezone for dates in gitweb output.
+ * To be used from 'onclick' event handler.
+ *
+ * @param {HTMLElement} target: where to install/display UI
+ * @param {DocumentFragment} tzSelectFragment: timezone selection UI
+ */
+function displayChangeTZForm(target, tzSelectFragment) {
+       // for absolute positioning to be related to target element
+       target.style.position = 'relative';
+       target.style.display = 'inline-block';
+
+       // show/display UI for changing timezone
+       target.appendChild(tzSelectFragment);
+}
+
+
+/* ...................................................................... */
+/* List of timezones for timezone selection menu */
+
+/**
+ * Generate list of timezones for creating timezone select UI
+ *
+ * @returns {Object[]} list of e.g. { value: '+0100', descr: 'GMT+01:00' }
+ */
+function generateTZList() {
+       var timezones = [
+               { value: "utc",   descr: "UTC/GMT"},
+               { value: "local", descr: "Local (per browser)"}
+       ];
+
+       // generate all full hour timezones (no fractional timezones)
+       for (var x = -12, idx = timezones.length; x <= +14; x++, idx++) {
+               var hours = (x >= 0 ? '+' : '-') + padLeft(x >=0 ? x : -x, 2);
+               timezones[idx] = { value: hours + '00', descr: 'UTC' + hours + ':00'};
+               if (x === 0) {
+                       timezones[idx].descr = 'UTC\u00B100:00'; // 'UTC&plusmn;00:00'
+               }
+       }
+
+       return timezones;
+}
+
+/**
+ * Generate <options> elements for timezone select UI
+ *
+ * @param {String} tzSelected: default timezone
+ * @returns {DocumentFragment} list of options elements to appendChild
+ */
+function generateTZOptions(tzSelected) {
+       var elems = document.createDocumentFragment();
+       var timezones = generateTZList();
+
+       for (var i = 0, len = timezones.length; i < len; i++) {
+               var tzone = timezones[i];
+               var option = document.createElement("option");
+               if (tzone.value === tzSelected) {
+                       option.defaultSelected = true;
+               }
+               option.value = tzone.value;
+               option.appendChild(document.createTextNode(tzone.descr));
+
+               elems.appendChild(option);
+       }
+
+       return elems;
+}
+
+
+/* ...................................................................... */
+/* Event handlers and/or their generators */
+
+/**
+ * Create event handler that select timezone and closes timezone select UI.
+ * To be used as $('select[name="tzselect"]').onchange handler.
+ *
+ * @param {DocumentFragment} tzSelectFragment: timezone selection UI
+ * @param {Object} tzCookieInfo: object literal with info about cookie to store timezone
+ * @param {String} tzCookieInfo.name: name of cookie to save result of selection
+ * @param {String} tzClassName: specifies element where UI was installed
+ * @returns {Function} event handler
+ */
+function selectTZHandler(tzSelectFragment, tzCookieInfo, tzClassName) {
+       //return function selectTZ(event) {
+       return function (event) {
+               event = event || window.event;
+               var target = event.target || event.srcElement;
+
+               var selected = target.options.item(target.selectedIndex);
+               removeChangeTZForm(tzSelectFragment, target, tzClassName);
+
+               if (selected) {
+                       selected.defaultSelected = true;
+                       setCookie(tzCookieInfo.name, selected.value, tzCookieInfo);
+                       fixDatetimeTZ(selected.value, tzClassName);
+               }
+       };
+}
+
+/**
+ * Create event handler that closes timezone select UI.
+ * To be used e.g. as $('.closebutton').onclick handler.
+ *
+ * @param {DocumentFragment} tzSelectFragment: timezone selection UI
+ * @param {String} tzClassName: specifies element where UI was installed
+ * @returns {Function} event handler
+ */
+function closeTZFormHandler(tzSelectFragment, tzClassName) {
+       //return function closeTZForm(event) {
+       return function (event) {
+               event = event || window.event;
+               var target = event.target || event.srcElement;
+
+               removeChangeTZForm(tzSelectFragment, target, tzClassName);
+       };
+}
+
+/* end of adjust-timezone.js */
+// Copyright (C) 2007, Fredrik Kuivinen <frekui@gmail.com>
+//               2007, Petr Baudis <pasky@suse.cz>
+//          2008-2011, Jakub Narebski <jnareb@gmail.com>
+
+/**
+ * @fileOverview JavaScript side of Ajax-y 'blame_incremental' view in gitweb
+ * @license GPLv2 or later
+ */
+
+/* ============================================================ */
+/*
+ * This code uses DOM methods instead of (nonstandard) innerHTML
+ * to modify page.
+ *
+ * innerHTML is non-standard IE extension, though supported by most
+ * browsers; however Firefox up to version 1.5 didn't implement it in
+ * a strict mode (application/xml+xhtml mimetype).
+ *
+ * Also my simple benchmarks show that using elem.firstChild.data =
+ * 'content' is slightly faster than elem.innerHTML = 'content'.  It
+ * is however more fragile (text element fragment must exists), and
+ * less feature-rich (we cannot add HTML).
+ *
+ * Note that DOM 2 HTML is preferred over generic DOM 2 Core; the
+ * equivalent using DOM 2 Core is usually shown in comments.
+ */
+
+
+/* ............................................................ */
+/* utility/helper functions (and variables) */
+
+var projectUrl; // partial query + separator ('?' or ';')
+
+// 'commits' is an associative map. It maps SHA1s to Commit objects.
+var commits = {};
+
+/**
+ * constructor for Commit objects, used in 'blame'
+ * @class Represents a blamed commit
+ * @param {String} sha1: SHA-1 identifier of a commit
+ */
+function Commit(sha1) {
+       if (this instanceof Commit) {
+               this.sha1 = sha1;
+               this.nprevious = 0; /* number of 'previous', effective parents */
+       } else {
+               return new Commit(sha1);
+       }
+}
+
+/* ............................................................ */
+/* progress info, timing, error reporting */
+
+var blamedLines = 0;
+var totalLines  = '???';
+var div_progress_bar;
+var div_progress_info;
+
+/**
+ * Detects how many lines does a blamed file have,
+ * This information is used in progress info
+ *
+ * @returns {Number|String} Number of lines in file, or string '...'
+ */
+function countLines() {
+       var table =
+               document.getElementById('blame_table') ||
+               document.getElementsByTagName('table')[0];
+
+       if (table) {
+               return table.getElementsByTagName('tr').length - 1; // for header
+       } else {
+               return '...';
+       }
+}
+
+/**
+ * update progress info and length (width) of progress bar
+ *
+ * @globals div_progress_info, div_progress_bar, blamedLines, totalLines
+ */
+function updateProgressInfo() {
+       if (!div_progress_info) {
+               div_progress_info = document.getElementById('progress_info');
+       }
+       if (!div_progress_bar) {
+               div_progress_bar = document.getElementById('progress_bar');
+       }
+       if (!div_progress_info && !div_progress_bar) {
+               return;
+       }
+
+       var percentage = Math.floor(100.0*blamedLines/totalLines);
+
+       if (div_progress_info) {
+               div_progress_info.firstChild.data  = blamedLines + ' / ' + totalLines +
+                       ' (' + padLeftStr(percentage, 3, '\u00A0') + '%)';
+       }
+
+       if (div_progress_bar) {
+               //div_progress_bar.setAttribute('style', 'width: '+percentage+'%;');
+               div_progress_bar.style.width = percentage + '%';
+       }
+}
+
+
+var t_interval_server = '';
+var cmds_server = '';
+var t0 = new Date();
+
+/**
+ * write how much it took to generate data, and to run script
+ *
+ * @globals t0, t_interval_server, cmds_server
+ */
+function writeTimeInterval() {
+       var info_time = document.getElementById('generating_time');
+       if (!info_time || !t_interval_server) {
+               return;
+       }
+       var t1 = new Date();
+       info_time.firstChild.data += ' + (' +
+               t_interval_server + ' sec server blame_data / ' +
+               (t1.getTime() - t0.getTime())/1000 + ' sec client JavaScript)';
+
+       var info_cmds = document.getElementById('generating_cmd');
+       if (!info_time || !cmds_server) {
+               return;
+       }
+       info_cmds.firstChild.data += ' + ' + cmds_server;
+}
+
+/**
+ * show an error message alert to user within page (in progress info area)
+ * @param {String} str: plain text error message (no HTML)
+ *
+ * @globals div_progress_info
+ */
+function errorInfo(str) {
+       if (!div_progress_info) {
+               div_progress_info = document.getElementById('progress_info');
+       }
+       if (div_progress_info) {
+               div_progress_info.className = 'error';
+               div_progress_info.firstChild.data = str;
+       }
+}
+
+/* ............................................................ */
+/* coloring rows during blame_data (git blame --incremental) run */
+
+/**
+ * used to extract N from 'colorN', where N is a number,
+ * @constant
+ */
+var colorRe = /\bcolor([0-9]*)\b/;
+
+/**
+ * return N if <tr class="colorN">, otherwise return null
+ * (some browsers require CSS class names to begin with letter)
+ *
+ * @param {HTMLElement} tr: table row element to check
+ * @param {String} tr.className: 'class' attribute of tr element
+ * @returns {Number|null} N if tr.className == 'colorN', otherwise null
+ *
+ * @globals colorRe
+ */
+function getColorNo(tr) {
+       if (!tr) {
+               return null;
+       }
+       var className = tr.className;
+       if (className) {
+               var match = colorRe.exec(className);
+               if (match) {
+                       return parseInt(match[1], 10);
+               }
+       }
+       return null;
+}
+
+var colorsFreq = [0, 0, 0];
+/**
+ * return one of given possible colors (currently least used one)
+ * example: chooseColorNoFrom(2, 3) returns 2 or 3
+ *
+ * @param {Number[]} arguments: one or more numbers
+ *        assumes that  1 <= arguments[i] <= colorsFreq.length
+ * @returns {Number} Least used color number from arguments
+ * @globals colorsFreq
+ */
+function chooseColorNoFrom() {
+       // choose the color which is least used
+       var colorNo = arguments[0];
+       for (var i = 1; i < arguments.length; i++) {
+               if (colorsFreq[arguments[i]-1] < colorsFreq[colorNo-1]) {
+                       colorNo = arguments[i];
+               }
+       }
+       colorsFreq[colorNo-1]++;
+       return colorNo;
+}
+
+/**
+ * given two neighbor <tr> elements, find color which would be different
+ * from color of both of neighbors; used to 3-color blame table
+ *
+ * @param {HTMLElement} tr_prev
+ * @param {HTMLElement} tr_next
+ * @returns {Number} color number N such that
+ * colorN != tr_prev.className && colorN != tr_next.className
+ */
+function findColorNo(tr_prev, tr_next) {
+       var color_prev = getColorNo(tr_prev);
+       var color_next = getColorNo(tr_next);
+
+
+       // neither of neighbors has color set
+       // THEN we can use any of 3 possible colors
+       if (!color_prev && !color_next) {
+               return chooseColorNoFrom(1,2,3);
+       }
+
+       // either both neighbors have the same color,
+       // or only one of neighbors have color set
+       // THEN we can use any color except given
+       var color;
+       if (color_prev === color_next) {
+               color = color_prev; // = color_next;
+       } else if (!color_prev) {
+               color = color_next;
+       } else if (!color_next) {
+               color = color_prev;
+       }
+       if (color) {
+               return chooseColorNoFrom((color % 3) + 1, ((color+1) % 3) + 1);
+       }
+
+       // neighbors have different colors
+       // THEN there is only one color left
+       return (3 - ((color_prev + color_next) % 3));
+}
+
+/* ............................................................ */
+/* coloring rows like 'blame' after 'blame_data' finishes */
+
+/**
+ * returns true if given row element (tr) is first in commit group
+ * to be used only after 'blame_data' finishes (after processing)
+ *
+ * @param {HTMLElement} tr: table row
+ * @returns {Boolean} true if TR is first in commit group
+ */
+function isStartOfGroup(tr) {
+       return tr.firstChild.className === 'sha1';
+}
+
+/**
+ * change colors to use zebra coloring (2 colors) instead of 3 colors
+ * concatenate neighbor commit groups belonging to the same commit
+ *
+ * @globals colorRe
+ */
+function fixColorsAndGroups() {
+       var colorClasses = ['light', 'dark'];
+       var linenum = 1;
+       var tr, prev_group;
+       var colorClass = 0;
+       var table =
+               document.getElementById('blame_table') ||
+               document.getElementsByTagName('table')[0];
+
+       while ((tr = document.getElementById('l'+linenum))) {
+       // index origin is 0, which is table header; start from 1
+       //while ((tr = table.rows[linenum])) { // <- it is slower
+               if (isStartOfGroup(tr, linenum, document)) {
+                       if (prev_group &&
+                           prev_group.firstChild.firstChild.href ===
+                                   tr.firstChild.firstChild.href) {
+                               // we have to concatenate groups
+                               var prev_rows = prev_group.firstChild.rowSpan || 1;
+                               var curr_rows =         tr.firstChild.rowSpan || 1;
+                               prev_group.firstChild.rowSpan = prev_rows + curr_rows;
+                               //tr.removeChild(tr.firstChild);
+                               tr.deleteCell(0); // DOM2 HTML way
+                       } else {
+                               colorClass = (colorClass + 1) % 2;
+                               prev_group = tr;
+                       }
+               }
+               var tr_class = tr.className;
+               tr.className = tr_class.replace(colorRe, colorClasses[colorClass]);
+               linenum++;
+       }
+}
+
+
+/* ============================================================ */
+/* main part: parsing response */
+
+/**
+ * Function called for each blame entry, as soon as it finishes.
+ * It updates page via DOM manipulation, adding sha1 info, etc.
+ *
+ * @param {Commit} commit: blamed commit
+ * @param {Object} group: object representing group of lines,
+ *                        which blame the same commit (blame entry)
+ *
+ * @globals blamedLines
+ */
+function handleLine(commit, group) {
+       /*
+          This is the structure of the HTML fragment we are working
+          with:
+
+          <tr id="l123" class="">
+            <td class="sha1" title=""><a href=""> </a></td>
+            <td class="linenr"><a class="linenr" href="">123</a></td>
+            <td class="pre"># times (my ext3 doesn&#39;t).</td>
+          </tr>
+       */
+
+       var resline = group.resline;
+
+       // format date and time string only once per commit
+       if (!commit.info) {
+               /* e.g. 'Kay Sievers, 2005-08-07 21:49:46 +0200' */
+               commit.info = commit.author + ', ' +
+                       formatDateISOLocal(commit.authorTime, commit.authorTimezone);
+       }
+
+       // color depends on group of lines, not only on blamed commit
+       var colorNo = findColorNo(
+               document.getElementById('l'+(resline-1)),
+               document.getElementById('l'+(resline+group.numlines))
+       );
+
+       // loop over lines in commit group
+       for (var i = 0; i < group.numlines; i++, resline++) {
+               var tr = document.getElementById('l'+resline);
+               if (!tr) {
+                       break;
+               }
+               /*
+                       <tr id="l123" class="">
+                         <td class="sha1" title=""><a href=""> </a></td>
+                         <td class="linenr"><a class="linenr" href="">123</a></td>
+                         <td class="pre"># times (my ext3 doesn&#39;t).</td>
+                       </tr>
+               */
+               var td_sha1  = tr.firstChild;
+               var a_sha1   = td_sha1.firstChild;
+               var a_linenr = td_sha1.nextSibling.firstChild;
+
+               /* <tr id="l123" class=""> */
+               var tr_class = '';
+               if (colorNo !== null) {
+                       tr_class = 'color'+colorNo;
+               }
+               if (commit.boundary) {
+                       tr_class += ' boundary';
+               }
+               if (commit.nprevious === 0) {
+                       tr_class += ' no-previous';
+               } else if (commit.nprevious > 1) {
+                       tr_class += ' multiple-previous';
+               }
+               tr.className = tr_class;
+
+               /* <td class="sha1" title="?" rowspan="?"><a href="?">?</a></td> */
+               if (i === 0) {
+                       td_sha1.title = commit.info;
+                       td_sha1.rowSpan = group.numlines;
+
+                       a_sha1.href = projectUrl + 'a=commit;h=' + commit.sha1;
+                       if (a_sha1.firstChild) {
+                               a_sha1.firstChild.data = commit.sha1.substr(0, 8);
+                       } else {
+                               a_sha1.appendChild(
+                                       document.createTextNode(commit.sha1.substr(0, 8)));
+                       }
+                       if (group.numlines >= 2) {
+                               var fragment = document.createDocumentFragment();
+                               var br   = document.createElement("br");
+                               var match = commit.author.match(/\b([A-Z])\B/g);
+                               if (match) {
+                                       var text = document.createTextNode(
+                                                       match.join(''));
+                               }
+                               if (br && text) {
+                                       var elem = fragment || td_sha1;
+                                       elem.appendChild(br);
+                                       elem.appendChild(text);
+                                       if (fragment) {
+                                               td_sha1.appendChild(fragment);
+                                       }
+                               }
+                       }
+               } else {
+                       //tr.removeChild(td_sha1); // DOM2 Core way
+                       tr.deleteCell(0); // DOM2 HTML way
+               }
+
+               /* <td class="linenr"><a class="linenr" href="?">123</a></td> */
+               var linenr_commit =
+                       ('previous' in commit ? commit.previous : commit.sha1);
+               var linenr_filename =
+                       ('file_parent' in commit ? commit.file_parent : commit.filename);
+               a_linenr.href = projectUrl + 'a=blame_incremental' +
+                       ';hb=' + linenr_commit +
+                       ';f='  + encodeURIComponent(linenr_filename) +
+                       '#l' + (group.srcline + i);
+
+               blamedLines++;
+
+               //updateProgressInfo();
+       }
+}
+
+// ----------------------------------------------------------------------
+
+/**#@+
+ * @constant
+ */
+var sha1Re = /^([0-9a-f]{40}) ([0-9]+) ([0-9]+) ([0-9]+)/;
+var infoRe = /^([a-z-]+) ?(.*)/;
+var endRe  = /^END ?([^ ]*) ?(.*)/;
+/**@-*/
+
+var curCommit = new Commit();
+var curGroup  = {};
+
+/**
+ * Parse output from 'git blame --incremental [...]', received via
+ * XMLHttpRequest from server (blamedataUrl), and call handleLine
+ * (which updates page) as soon as blame entry is completed.
+ *
+ * @param {String[]} lines: new complete lines from blamedata server
+ *
+ * @globals commits, curCommit, curGroup, t_interval_server, cmds_server
+ * @globals sha1Re, infoRe, endRe
+ */
+function processBlameLines(lines) {
+       var match;
+
+       for (var i = 0, len = lines.length; i < len; i++) {
+
+               if ((match = sha1Re.exec(lines[i]))) {
+                       var sha1 = match[1];
+                       var srcline  = parseInt(match[2], 10);
+                       var resline  = parseInt(match[3], 10);
+                       var numlines = parseInt(match[4], 10);
+
+                       var c = commits[sha1];
+                       if (!c) {
+                               c = new Commit(sha1);
+                               commits[sha1] = c;
+                       }
+                       curCommit = c;
+
+                       curGroup.srcline = srcline;
+                       curGroup.resline = resline;
+                       curGroup.numlines = numlines;
+
+               } else if ((match = infoRe.exec(lines[i]))) {
+                       var info = match[1];
+                       var data = match[2];
+                       switch (info) {
+                       case 'filename':
+                               curCommit.filename = unquote(data);
+                               // 'filename' information terminates the entry
+                               handleLine(curCommit, curGroup);
+                               updateProgressInfo();
+                               break;
+                       case 'author':
+                               curCommit.author = data;
+                               break;
+                       case 'author-time':
+                               curCommit.authorTime = parseInt(data, 10);
+                               break;
+                       case 'author-tz':
+                               curCommit.authorTimezone = data;
+                               break;
+                       case 'previous':
+                               curCommit.nprevious++;
+                               // store only first 'previous' header
+                               if (!'previous' in curCommit) {
+                                       var parts = data.split(' ', 2);
+                                       curCommit.previous    = parts[0];
+                                       curCommit.file_parent = unquote(parts[1]);
+                               }
+                               break;
+                       case 'boundary':
+                               curCommit.boundary = true;
+                               break;
+                       } // end switch
+
+               } else if ((match = endRe.exec(lines[i]))) {
+                       t_interval_server = match[1];
+                       cmds_server = match[2];
+
+               } else if (lines[i] !== '') {
+                       // malformed line
+
+               } // end if (match)
+
+       } // end for (lines)
+}
+
+/**
+ * Process new data and return pointer to end of processed part
+ *
+ * @param {String} unprocessed: new data (from nextReadPos)
+ * @param {Number} nextReadPos: end of last processed data
+ * @return {Number} end of processed data (new value for nextReadPos)
+ */
+function processData(unprocessed, nextReadPos) {
+       var lastLineEnd = unprocessed.lastIndexOf('\n');
+       if (lastLineEnd !== -1) {
+               var lines = unprocessed.substring(0, lastLineEnd).split('\n');
+               nextReadPos += lastLineEnd + 1 /* 1 == '\n'.length */;
+
+               processBlameLines(lines);
+       } // end if
+
+       return nextReadPos;
+}
+
+/**
+ * Handle XMLHttpRequest errors
+ *
+ * @param {XMLHttpRequest} xhr: XMLHttpRequest object
+ * @param {Number} [xhr.pollTimer] ID of the timeout to clear
+ *
+ * @globals commits
+ */
+function handleError(xhr) {
+       errorInfo('Server error: ' +
+               xhr.status + ' - ' + (xhr.statusText || 'Error contacting server'));
+
+       if (typeof xhr.pollTimer === "number") {
+               clearTimeout(xhr.pollTimer);
+               delete xhr.pollTimer;
+       }
+       commits = {}; // free memory
+}
+
+/**
+ * Called after XMLHttpRequest finishes (loads)
+ *
+ * @param {XMLHttpRequest} xhr: XMLHttpRequest object
+ * @param {Number} [xhr.pollTimer] ID of the timeout to clear
+ *
+ * @globals commits
+ */
+function responseLoaded(xhr) {
+       if (typeof xhr.pollTimer === "number") {
+               clearTimeout(xhr.pollTimer);
+               delete xhr.pollTimer;
+       }
+
+       fixColorsAndGroups();
+       writeTimeInterval();
+       commits = {}; // free memory
+}
+
+/**
+ * handler for XMLHttpRequest onreadystatechange event
+ * @see startBlame
+ *
+ * @param {XMLHttpRequest} xhr: XMLHttpRequest object
+ * @param {Number} xhr.prevDataLength: previous value of xhr.responseText.length
+ * @param {Number} xhr.nextReadPos: start of unread part of xhr.responseText
+ * @param {Number} [xhr.pollTimer] ID of the timeout (to reset or cancel)
+ * @param {Boolean} fromTimer: if handler was called from timer
+ */
+function handleResponse(xhr, fromTimer) {
+
+       /*
+        * xhr.readyState
+        *
+        *  Value  Constant (W3C)    Description
+        *  -------------------------------------------------------------------
+        *  0      UNSENT            open() has not been called yet.
+        *  1      OPENED            send() has not been called yet.
+        *  2      HEADERS_RECEIVED  send() has been called, and headers
+        *                           and status are available.
+        *  3      LOADING           Downloading; responseText holds partial data.
+        *  4      DONE              The operation is complete.
+        */
+
+       if (xhr.readyState !== 4 && xhr.readyState !== 3) {
+               return;
+       }
+
+       // the server returned error
+       // try ... catch block is to work around bug in IE8
+       try {
+               if (xhr.readyState === 3 && xhr.status !== 200) {
+                       return;
+               }
+       } catch (e) {
+               return;
+       }
+       if (xhr.readyState === 4 && xhr.status !== 200) {
+               handleError(xhr);
+               return;
+       }
+
+       // In konqueror xhr.responseText is sometimes null here...
+       if (xhr.responseText === null) {
+               return;
+       }
+
+
+       // extract new whole (complete) lines, and process them
+       if (xhr.prevDataLength !== xhr.responseText.length) {
+               xhr.prevDataLength = xhr.responseText.length;
+               var unprocessed = xhr.responseText.substring(xhr.nextReadPos);
+               xhr.nextReadPos = processData(unprocessed, xhr.nextReadPos);
+       }
+
+       // did we finish work?
+       if (xhr.readyState === 4) {
+               responseLoaded(xhr);
+               return;
+       }
+
+       // if we get from timer, we have to restart it
+       // otherwise onreadystatechange gives us partial response, timer not needed
+       if (fromTimer) {
+               setTimeout(function () {
+                       handleResponse(xhr, true);
+               }, 1000);
+
+       } else if (typeof xhr.pollTimer === "number") {
+               clearTimeout(xhr.pollTimer);
+               delete xhr.pollTimer;
+       }
+}
+
+// ============================================================
+// ------------------------------------------------------------
+
+/**
+ * Incrementally update line data in blame_incremental view in gitweb.
+ *
+ * @param {String} blamedataUrl: URL to server script generating blame data.
+ * @param {String} bUrl: partial URL to project, used to generate links.
+ *
+ * Called from 'blame_incremental' view after loading table with
+ * file contents, a base for blame view.
+ *
+ * @globals t0, projectUrl, div_progress_bar, totalLines
+*/
+function startBlame(blamedataUrl, bUrl) {
+
+       var xhr = createRequestObject();
+       if (!xhr) {
+               errorInfo('ERROR: XMLHttpRequest not supported');
+               return;
+       }
+
+       t0 = new Date();
+       projectUrl = bUrl + (bUrl.indexOf('?') === -1 ? '?' : ';');
+       if ((div_progress_bar = document.getElementById('progress_bar'))) {
+               //div_progress_bar.setAttribute('style', 'width: 100%;');
+               div_progress_bar.style.cssText = 'width: 100%;';
+       }
+       totalLines = countLines();
+       updateProgressInfo();
+
+       /* add extra properties to xhr object to help processing response */
+       xhr.prevDataLength = -1;  // used to detect if we have new data
+       xhr.nextReadPos = 0;      // where unread part of response starts
+
+       xhr.onreadystatechange = function () {
+               handleResponse(xhr, false);
+       };
+
+       xhr.open('GET', blamedataUrl);
+       xhr.setRequestHeader('Accept', 'text/plain');
+       xhr.send(null);
+
+       // not all browsers call onreadystatechange event on each server flush
+       // poll response using timer every second to handle this issue
+       xhr.pollTimer = setTimeout(function () {
+               handleResponse(xhr, true);
+       }, 1000);
+}
+
+/* end of blame_incremental.js */