#!/usr/bin/perl -w
# Copyright © 2001-2012 Jamie Zawinski <jwz@jwz.org>.
#
# Permission to use, copy, modify, distribute, and sell this software and its
# documentation for any purpose is hereby granted without fee, provided that
# the above copyright notice appear in all copies and that both that
# copyright notice and this permission notice appear in supporting
# documentation.  No representations are made about the suitability of this
# software for any purpose.  It is provided "as is" without express or 
# implied warranty.
#
# This program chooses a random file from under the given directory, and
# prints its name.  The file will be an image file whose dimensions are
# larger than a certain minimum size.
#
# If the directory is a URL, it is assumed to be an RSS or Atom feed.
# The images from that feed will be downloaded, cached, and selected from
# at random.  The feed will be re-polled periodically, as needed.
#
# The various xscreensaver hacks that manipulate images ("jigsaw", etc.) get
# the image to manipulate by running the "xscreensaver-getimage" program.
#
# Under X11, the "xscreensaver-getimage" program invokes this script,
# depending on the value of the "chooseRandomImages" and "imageDirectory"
# settings in the ~/.xscreensaver file (or .../app-defaults/XScreenSaver).
# The screen savers invoke "xscreensaver-getimage" via utils/grabclient.c,
# which then invokes this script.
#
# Under Cocoa, this script lives inside the .saver bundle, and is invoked
# directly from utils/grabclient.c.
#
# Created: 12-Apr-01.

require 5;
#use diagnostics;	# Fails on some MacOS 10.5 systems
use strict;

use POSIX;
use Fcntl;

use Fcntl ':flock'; # import LOCK_* constants

use POSIX ':fcntl_h';				# S_ISDIR was here in Perl 5.6
import Fcntl ':mode' unless defined &S_ISUID;	# but it is here in Perl 5.8
	# but in Perl 5.10, both of these load, and cause errors!
	# So we have to check for S_ISUID instead of S_ISDIR?  WTF?

use bytes;  # Larry can take Unicode and shove it up his ass sideways.
            # Perl 5.8.0 causes us to start getting incomprehensible
            # errors about UTF-8 all over the place without this.

use Digest::MD5 qw(md5_base64);

# Some Linux systems don't install LWP by default!
# Only error out if we're actually loading a URL instead of local data.
BEGIN { eval 'use LWP::Simple;' }


my $progname = $0; $progname =~ s@.*/@@g;
my $version = q{ $Revision: 1.34 $ }; $version =~ s/^[^0-9]+([0-9.]+).*$/$1/;

my $verbose = 0;

# Whether to use MacOS X's Spotlight to generate the list of files.
# When set to -1, uses Spotlight if "mdfind" exists.
#
# (In my experience, this isn't actually any faster, and might not find
# everything if your Spotlight index is out of date, which happens often.)
#
my $use_spotlight_p = 0;

# Whether to cache the results of the last run.
#
my $cache_p = 1;

# Regenerate the cache if it is older than this many seconds.
#
my $cache_max_age = 60 * 60 * 3;   # 3 hours

# Re-poll RSS/Atom feeds when local copy is older than this many seconds.
#
my $feed_max_age = $cache_max_age;


# This matches files that we are allowed to use as images (case-insensitive.)
# Anything not matching this is ignored.  This is so you can point your
# imageDirectory at directory trees that have things other than images in
# them, but it assumes that you gave your images sensible file extensions.
#
my @good_extensions = ('jpg', 'jpeg', 'pjpeg', 'pjpg', 'png', 'gif',
                       'tif', 'tiff', 'xbm', 'xpm');
my $good_file_re = '\.(' . join("|", @good_extensions) . ')$';

# This matches file extensions that might occur in an image directory,
# and that are never used in the name of a subdirectory.  This is an
# optimization that prevents us from having to stat() those files to
# tell whether they are directories or not.  (It speeds things up a
# lot.  Don't give your directories stupid names.)
#
my @nondir_extensions = ('ai', 'bmp', 'bz2', 'cr2', 'crw', 'db',
   'dmg', 'eps', 'gz', 'hqx', 'htm', 'html', 'icns', 'ilbm', 'mov',
   'nef', 'pbm', 'pdf', 'pl', 'ppm', 'ps', 'psd', 'sea', 'sh', 'shtml',
   'tar', 'tgz', 'thb', 'txt', 'xcf', 'xmp', 'Z', 'zip' );
my $nondir_re = '\.(' . join("|", @nondir_extensions) . ')$';


# JPEG, GIF, and PNG files that are are smaller than this are rejected:
# this is so that you can use an image directory that contains both big
# images and thumbnails, and have it only select the big versions.
#
my $min_image_width  = 255;
my $min_image_height = 255;

my @all_files = ();         # list of "good" files we've collected
my %seen_inodes;            # for breaking recursive symlink loops

# For diagnostic messages:
#
my $dir_count = 1;          # number of directories seen
my $stat_count = 0;	    # number of files/dirs stat'ed
my $skip_count_unstat = 0;  # number of files skipped without stat'ing
my $skip_count_stat = 0;    # number of files skipped after stat

sub find_all_files($);
sub find_all_files($) {
  my ($dir) = @_;

  print STDERR "$progname:  + reading dir $dir/...\n" if ($verbose > 1);

  my $dd;
  if (! opendir ($dd, $dir)) {
    print STDERR "$progname: couldn't open $dir: $!\n" if ($verbose);
    return;
  }
  my @files = readdir ($dd);
  closedir ($dd);

  my @dirs = ();

  foreach my $file (@files) {
    next if ($file =~ m/^\./);      # silently ignore dot files/dirs

    if ($file =~ m/[~%\#]$/) {      # ignore backup files (and dirs...)
      $skip_count_unstat++;
      print STDERR "$progname:  - skip file  $file\n" if ($verbose > 1);
    }

    $file = "$dir/$file";

    if ($file =~ m/$good_file_re/io) {
      #
      # Assume that files ending in .jpg exist and are not directories.
      #
      push @all_files, $file;
      print STDERR "$progname:  - found file $file\n" if ($verbose > 1);

    } elsif ($file =~ m/$nondir_re/io) {
      #
      # Assume that files ending in .html are not directories.
      #
      $skip_count_unstat++;
      print STDERR "$progname: -- skip file  $file\n" if ($verbose > 1);

    } else {
      #
      # Now we need to stat the file to see if it's a subdirectory.
      #
      # Note: we could use the trick of checking "nlinks" on the parent
      # directory to see if this directory contains any subdirectories,
      # but that would exclude any symlinks to directories.
      #
      my @st = stat($file);
      my ($dev,$ino,$mode,$nlink,$uid,$gid,$rdev,$size,
          $atime,$mtime,$ctime,$blksize,$blocks) = @st;

      $stat_count++;

      if ($#st == -1) {
        if ($verbose) {
          my $ll = readlink $file;
          if (defined ($ll)) {
            print STDERR "$progname: + dangling symlink: $file -> $ll\n";
          } else {
            print STDERR "$progname: + unreadable: $file\n";
          }
        }
        next;
      }

      next if ($seen_inodes{"$dev:$ino"}); # break symlink loops
      $seen_inodes{"$dev:$ino"} = 1;

      if (S_ISDIR($mode)) {
        push @dirs, $file;
        $dir_count++;
        print STDERR "$progname:  + found dir  $file\n" if ($verbose > 1);

      } else {
        $skip_count_stat++;
        print STDERR "$progname:  + skip file  $file\n" if ($verbose > 1);
      }
    }
  }

  foreach (@dirs) {
    find_all_files ($_);
  }
}


sub spotlight_all_files($) {
  my ($dir) = @_;

  my @terms = ();
  # "public.image" matches all (indexed) images, including Photoshop, etc.
#  push @terms, "kMDItemContentTypeTree == 'public.image'";
  foreach (@good_extensions) {

    # kMDItemFSName hits the file system every time: much worse than "find".
#    push @terms, "kMDItemFSName == '*.$_'";

    # kMDItemDisplayName matches against the name in the Spotlight index,
    # but won't find files that (for whatever reason) didn't get indexed.
    push @terms, "kMDItemDisplayName == '*.$_'";
  }

  $dir =~ s@([^-_/a-z\d.,])@\\$1@gsi;  # quote for sh
  my $cmd = "mdfind -onlyin $dir \"" . join (' || ', @terms) . "\"";

  print STDERR "$progname: executing: $cmd\n" if ($verbose > 1);
  @all_files = split (/[\r\n]+/, `$cmd`);
}


# If we're using cacheing, read the cache file and return its contents,
# if any.  This also holds an exclusive lock on the cache file, which 
# has the additional benefit that if two copies of this program are
# running at once, one will wait for the other, instead of both of
# them spanking the same file system at the same time.
#
my $cache_fd = undef;
my $cache_file_name = undef;
my $read_cache_p = 0;

sub read_cache($) {
  my ($dir) = @_;

  return () unless ($cache_p);

  my $dd = "$ENV{HOME}/Library/Caches";    # MacOS location
  if (-d $dd) {
    $cache_file_name = "$dd/org.jwz.xscreensaver.getimage.cache";
  } elsif (-d "$ENV{HOME}/tmp") {
    $cache_file_name = "$ENV{HOME}/tmp/.xscreensaver-getimage.cache";
  } else {
    $cache_file_name = "$ENV{HOME}/.xscreensaver-getimage.cache";
  }

  print STDERR "$progname: awaiting lock: $cache_file_name\n"
    if ($verbose > 1);

  my $file = $cache_file_name;
  open ($cache_fd, '+>>', $file) || error ("unable to write $file: $!");
  flock ($cache_fd, LOCK_EX)     || error ("unable to lock $file: $!");
  seek ($cache_fd, 0, 0)         || error ("unable to rewind $file: $!");

  my $mtime = (stat($cache_fd))[9];

  if ($mtime + $cache_max_age < time) {
    print STDERR "$progname: cache is too old\n" if ($verbose);
    return ();
  }

  my $odir = <$cache_fd>;
  $odir =~ s/[\r\n]+$//s if defined ($odir);
  if (!defined ($odir) || ($dir ne $odir)) {
    print STDERR "$progname: cache is for $odir, not $dir\n"
      if ($verbose && $odir);
    return ();
  }

  my @files = ();
  while (<$cache_fd>) { 
    s/[\r\n]+$//s;
    push @files, "$odir/$_";
  }

  print STDERR "$progname: " . ($#files+1) . " files in cache\n"
    if ($verbose);

  $read_cache_p = 1;
  return @files;
}


sub write_cache($) {
  my ($dir) = @_;

  return unless ($cache_p);

  # If we read the cache, just close it without rewriting it.
  # If we didn't read it, then write it now.

  if (! $read_cache_p) {

    truncate ($cache_fd, 0) ||
      error ("unable to truncate $cache_file_name: $!");
    seek ($cache_fd, 0, 0) ||
      error ("unable to rewind $cache_file_name: $!");

    if ($#all_files >= 0) {
      print $cache_fd "$dir\n";
      foreach (@all_files) {
        my $f = $_; # stupid Perl. do this to avoid modifying @all_files!
        $f =~ s@^\Q$dir\L/@@so || die;  # remove $dir from front
        print $cache_fd "$f\n";
      }
    }

    print STDERR "$progname: cached " . ($#all_files+1) . " files\n"
      if ($verbose);
  }

  flock ($cache_fd, LOCK_UN) ||
    error ("unable to unlock $cache_file_name: $!");
  close ($cache_fd);
  $cache_fd = undef;
}


sub html_unquote($) {
  my ($h) = @_;
  my %ent = ( 'amp' => '&', 'lt' => '<', 'gt' => '>', 
              'quot' => '"', 'apos' => "'" );
  $h =~ s/(&(\#)?([[:alpha:]\d]+);?)/
    {
     my ($o, $c) = ($1, $3);
     if (! defined($2)) {
       $c = $ent{$c};			# for &lt;
     } else {
       if ($c =~ m@^x([\dA-F]+)$@si) {	# for &#x41;
         $c = chr(hex($1));
       } elsif ($c =~ m@^\d+$@si) {	# for &#65;
         $c = chr($c);
       } else {
         $c = undef;
       }
     }
     ($c || $o);
    }
   /gexi;
  return $h;
}



sub set_proxy($) {
  my ($ua) = @_;

  if (!defined($ENV{http_proxy}) && !defined($ENV{HTTP_PROXY})) {
    my $proxy_data = `scutil --proxy 2>/dev/null`;
    my ($server) = ($proxy_data =~ m/\bHTTPProxy\s*:\s*([^\s]+)/s);
    my ($port)   = ($proxy_data =~ m/\bHTTPPort\s*:\s*([^\s]+)/s);
    if ($server) {
      # Note: this ignores the "ExceptionsList".
      $ENV{http_proxy} = "http://" . $server . ($port ? ":$port" : "") . "/";
      print STDERR "$progname: MacOS proxy: $ENV{http_proxy}\n"
        if ($verbose > 2)
      }
  }

  $ua->env_proxy();
}


sub init_lwp() {
  if (! defined ($LWP::Simple::ua)) {
    error ("\n\n\tPerl is broken. Do this to repair it:\n" .
           "\n\tsudo cpan LWP::Simple\n");
  }
  set_proxy ($LWP::Simple::ua);
}


# Returns a list of the image enclosures in the RSS or Atom feed.
# Elements of the list are references, [ "url", "guid" ].
#
sub parse_feed($);
sub parse_feed($) {
  my ($url) = @_;

  init_lwp();
  $LWP::Simple::ua->agent ("$progname/$version");
  $LWP::Simple::ua->timeout (10);  # bail sooner than the default of 3 minutes

  my $body = (LWP::Simple::get($url) || '');

  if ($body !~ m@^<\?xml\s@si) {
    # Not an RSS/Atom feed.  Try RSS autodiscovery.

    error ("not an RSS or Atom feed, or HTML: $url")
      unless ($body =~ m@<(HEAD|BODY|A|IMG)\b@si);

    # Find the first <link> with RSS or Atom in it, and use that instead.

    $body =~ s@<LINK\s+([^<>]*)>@{
      my $p = $1;
      if ($p =~ m! \b REL  \s* = \s* ['"]? alternate \b!six &&
          $p =~ m! \b TYPE \s* = \s* ['"]? application/(atom|rss) !six &&
          $p =~ m! \b HREF \s* = \s* ['"]  ( [^<>'"]+ ) !six
         ) {
        my $u2 = html_unquote ($1);
        print STDERR "$progname: found feed: $u2\n"
          if ($verbose);
        return parse_feed ($u2);
      }
      '';
    }@gsexi;

    error ("no RSS or Atom feed for HTML page: $url");
  }


  $body =~ s@(<ENTRY|<ITEM)@\001$1@gsi;
  my @items = split(/\001/, $body);
  shift @items;

  my @imgs = ();
  my %ids;

  foreach my $item (@items) {
    my $iurl = undef;
    my $id = undef;

    # First look for <link rel="enclosure" href="...">
    #
    if (! $iurl) {
      $item =~ s!(<LINK[^<>]*>)!{
        my $link = $1;
        my ($rel)  = ($link =~ m/\bREL\s*=\s*[\"\']?([^<>\'\"]+)/si);
        my ($type) = ($link =~ m/\bTYPE\s*=\s*[\"\']?([^<>\'\"]+)/si);
        my ($href) = ($link =~ m/\bHREF\s*=\s*[\"\']([^<>\'\"]+)/si);

        if ($rel && lc($rel) eq 'enclosure') {
          if ($type) {
            $href = undef unless ($type =~ m@^image/@si);  # omit videos
          }
          $iurl = html_unquote($href) if $href;
        }
        $link;
      }!gsexi;
    }

    # Then look for <media:content url="...">
    #
    if (! $iurl) {
      $item =~ s!(<MEDIA:CONTENT[^<>]*>)!{
        my $link = $1;
        my ($href) = ($link =~ m/\bURL\s*=\s*[\"\']([^<>\'\"]+)/si);
        $iurl = html_unquote($href) if $href;
        $link;
      }!gsexi;
    }

    # Then look for <enclosure url="..."/> 
    #
    if (! $iurl) {
      $item =~ s!(<ENCLOSURE[^<>]*>)!{
        my $link = $1;
        my ($type) = ($link =~ m/\bTYPE\s*=\s*[\"\']?([^<>\'\"]+)/si);
        my ($href) = ($link =~ m/\bURL\s*=\s*[\"\']([^<>\'\"]+)/si);
        $iurl = html_unquote($href)
          if ($href && $type && $type =~ m@^image/@si);  # omit videos
        $link;
      }!gsexi;
    }

    # Ok, maybe there's an image in the <url> field?
    #
    if (! $iurl) {
      $item =~ s!((<URL\b[^<>]*>)([^<>]*))!{
        my ($all, $u2) = ($1, $3);
        $iurl = html_unquote($u2) if ($u2 =~ m/$good_file_re/io);
        $all;
      }!gsexi;
    }

    # Then look for <description>... with an <img src="..."> inside.
    #
    if (! $iurl) {
      $item =~ s!(<description[^<>]*>.*?</description>)!{
        my $desc = $1;
        $desc = html_unquote($desc);
        my ($href) = ($desc =~ m@<IMG[^<>]*\bSRC=[\"\']?([^\"\'<>]+)@si);
        $iurl = $href if ($href);
        $desc;
      }!gsexi;
    }

    # Could also do <content:encoded>, but the above probably covers all
    # of the real-world possibilities.


    # Find a unique ID for this image, to defeat image farms.
    # First look for <id>...</id>
    ($id) = ($item =~ m!<ID\b[^<>]*>\s*([^<>]+?)\s*</ID>!si) unless $id;

    # Then look for <guid isPermaLink=...> ... </guid>
    ($id) = ($item =~ m!<GUID\b[^<>]*>\s*([^<>]+?)\s*</GUID>!si) unless $id;

    # Then look for <link> ... </link>
    ($id) = ($item =~ m!<LINK\b[^<>]*>\s*([^<>]+?)\s*</LINK>!si) unless $id;


    if ($iurl) {
      $id = $iurl unless $id;
      my $o = $ids{$id};
      if (! $o) {
        $ids{$id} = $iurl;
        my @P = ($iurl, $id);
        push @imgs, \@P;
      } elsif ($iurl ne $o) {
        print STDERR "$progname: WARNING: dup ID \"$id\"" .
                     " for \"$o\" and \"$iurl\"\n";
      }
    }
  }

  return @imgs;
}


# Like md5_base64 but uses filename-safe characters.
#
sub md5_file($) {
  my ($s) = @_;
  $s = md5_base64($s);
  $s =~ s@[/]@_@gs;
  $s =~ s@[+]@-@gs;
  return $s;
}


# Given the URL of an image, download it into the given directory
# and return the file name.
#
sub download_image($$$) {
  my ($url, $uid, $dir) = @_;

  my ($ext) = ($url =~ m@\.([a-z\d]+)$@si);
  my $file = md5_file ($uid);
  $file .= '.' . lc($ext) if $ext;

  # Don't bother doing If-Modified-Since to see if the URL has changed.
  # If we have already downloaded it, assume it's good.
  if (-f "$dir/$file") {
    print STDERR "$progname: exists: $dir/$file for $uid / $url\n" 
      if ($verbose > 1);
    return $file;
  }

  # Special-case kludge for Flickr:
  # Their RSS feeds sometimes include only the small versions of the images.
  # So if the URL ends in "s" (75x75), "t" (100x100) or "m" (240x240),then
  # munge it to be "b" (1024x1024).
  #
  $url =~ s@_[stm](\.[a-z]+)$@_b$1@si
    if ($url =~ m@^https?://[^/?#&]*?flickr\.com/@si);

  print STDERR "$progname: downloading: $dir/$file for $uid / $url\n" 
    if ($verbose > 1);
  init_lwp();
  $LWP::Simple::ua->agent ("$progname/$version");
  my $status = LWP::Simple::mirror ($url, "$dir/$file");
  if (!LWP::Simple::is_success ($status)) {
    print STDERR "$progname: error $status: $url\n";   # keep going
  }

  return $file;
}


sub mirror_feed($) {
  my ($url) = @_;

  if ($url !~ m/^https?:/si) {   # not a URL: local directory.
    return (undef, $url);
  }

  my $dir = "$ENV{HOME}/Library/Caches";    # MacOS location
  if (-d $dir) {
    $dir = "$dir/org.jwz.xscreensaver.feeds";
  } elsif (-d "$ENV{HOME}/tmp") {
    $dir = "$ENV{HOME}/tmp/.xscreensaver-feeds";
  } else {
    $dir = "$ENV{HOME}/.xscreensaver-feeds";
  }

  if (! -d $dir) {
    mkdir ($dir) || error ("mkdir $dir: $!");
    print STDERR "$progname: mkdir $dir/\n" if ($verbose);
  }

  # MD5 for directory name to use for cache of a feed URL.
  $dir .= '/' . md5_file ($url);

  if (! -d $dir) {
    mkdir ($dir) || error ("mkdir $dir: $!");
    print STDERR "$progname: mkdir $dir/ for $url\n" if ($verbose);
  }

  # At this point, we have the directory corresponding to this URL.
  # Now check to see if the files in it are up to date, and download
  # them if not.

  my $stamp = '.timestamp';
  my $lock = "$dir/$stamp";

  print STDERR "$progname: awaiting lock: $lock\n"
    if ($verbose > 1);

  my $mtime = ((stat($lock))[9]) || 0;

  my $lock_fd;
  open ($lock_fd, '+>>', $lock) || error ("unable to write $lock: $!");
  flock ($lock_fd, LOCK_EX)     || error ("unable to lock $lock: $!");
  seek ($lock_fd, 0, 0)         || error ("unable to rewind $lock: $!");

  my $poll_p = ($mtime + $feed_max_age < time);

  $poll_p = 1 unless ($cache_p);  # poll again now with --no-cache cmd line arg.

  # Even if the cache is young, let's make sure there are at least
  # a few files in it, and re-check if not.
  #
  if (! $poll_p) {
    my $count = 0;
    opendir (my $dirh, $dir) || error ("$dir: $!");
    foreach my $f (readdir ($dirh)) {
      next if ($f =~ m/^\./s);
      $count++;
      last;
    }
    closedir $dirh;

    if ($count <= 0) {
      print STDERR "$progname: no files in cache of $url\n" if ($verbose);
      $poll_p = 1;
    }
  }

  if ($poll_p) {

    print STDERR "$progname: loading $url\n" if ($verbose);

    my %files;
    opendir (my $dirh, $dir) || error ("$dir: $!");
    foreach my $f (readdir ($dirh)) {
      next if ($f eq '.' || $f eq '..');
      $files{$f} = 0;  # 0 means "file exists, should be deleted"
    }
    closedir $dirh;

    $files{$stamp} = 1;

    # Download each image currently in the feed.
    #
    my $count = 0;
    my @urls = parse_feed ($url);
    print STDERR "$progname: " . ($#urls + 1) . " images\n"
      if ($verbose > 1);
    foreach my $p (@urls) {
      my ($furl, $id) = @$p;
      my $f = download_image ($furl, $id, $dir);
      next unless $f;
      $files{$f} = 1;    # Got it, don't delete
      $count++;
    }

    print STDERR "$progname: empty feed: $url\n" if ($count <= 0);

    # Now delete any files that are no longer in the feed.
    # But if there was nothing in the feed (network failure?)
    # then don't blow away the old files.
    #
    my $kept = 0;
    foreach my $f (keys(%files)) {
      if ($count <= 0) {
        $kept++;
      } elsif ($files{$f}) {
        $kept++;
      } else {
        if (unlink ("$dir/$f")) {
          print STDERR "$progname: rm $dir/$f\n" if ($verbose > 1);
        } else {
          print STDERR "$progname: rm $dir/$f: $!\n";   # don't bail
        }
      }
    }

    # Both feed and cache are empty. No files at all.
    error ("empty feed: $url") if ($kept <= 1);

    $mtime = time();	# update the timestamp

  } else {

    # Not yet time to re-check the URL.
    print STDERR "$progname: using cache: $url\n" if ($verbose);

  }

  # Unlock and update the write date on the .timestamp file.
  #
  truncate ($lock_fd, 0) || error ("unable to truncate $lock: $!");
  seek ($lock_fd, 0, 0)  || error ("unable to rewind $lock: $!");
  utime ($mtime, $mtime, $lock_fd) || error ("unable to touch $lock: $!");
  flock ($lock_fd, LOCK_UN) || error ("unable to unlock $lock: $!");
  close ($lock_fd);
  $lock_fd = undef;
  print STDERR "$progname: unlocked $lock\n" if ($verbose > 1);

  # Don't bother using the imageDirectory cache.  We know that this directory
  # is flat, and we can assume that an RSS feed doesn't contain 100,000 images
  # like ~/Pictures/ might.
  #
  $cache_p = 0;

  # Return the URL and directory name of the files of that URL's local cache.
  #
  return ($url, $dir);
}


sub find_random_file($) {
  my ($dir) = @_;

  if ($use_spotlight_p == -1) {
    $use_spotlight_p = 0;
    if (-x '/usr/bin/mdfind') {
      $use_spotlight_p = 1;
    }
  }

  my $url;
  ($url, $dir) = mirror_feed ($dir);

  if ($url) {
    $use_spotlight_p = 0;
    print STDERR "$progname: $dir is cache for $url\n" if ($verbose > 1);
  }

  @all_files = read_cache ($dir);

  if ($#all_files >= 0) {
    # got it from the cache...

  } elsif ($use_spotlight_p) {
    print STDERR "$progname: spotlighting $dir...\n" if ($verbose);
    spotlight_all_files ($dir);
    print STDERR "$progname: found " . ($#all_files+1) .
                 " file" . ($#all_files == 0 ? "" : "s") .
                 " via Spotlight\n"
      if ($verbose);
  } else {
    print STDERR "$progname: recursively reading $dir...\n" if ($verbose);
    find_all_files ($dir);
    print STDERR "$progname: " .
                 "f=" . ($#all_files+1) . "; " .
                 "d=$dir_count; " .
                 "s=$stat_count; " .
                 "skip=${skip_count_unstat}+$skip_count_stat=" .
                  ($skip_count_unstat + $skip_count_stat) .
                 ".\n"
      if ($verbose);
  }

  write_cache ($dir);

#  @all_files = sort(@all_files);

  if ($#all_files < 0) {
    print STDERR "$progname: no files in $dir\n";
    exit 1;
  }

  my $max_tries = 50;
  for (my $i = 0; $i < $max_tries; $i++) {

    my $n = int (rand ($#all_files + 1));
    my $file = $all_files[$n];
    if (large_enough_p ($file)) {
      if (! $url) {
        $file =~ s@^\Q$dir\L/@@so || die;  # remove $dir from front
      }
      return $file;
    }
  }

  print STDERR "$progname: no suitable images in $dir " .
               "(after $max_tries tries)\n";

  # If we got here, blow away the cache.  Maybe it's stale.
  unlink $cache_file_name if $cache_file_name;

  exit 1;
}


sub large_enough_p($) {
  my ($file) = @_;

  my ($w, $h) = image_file_size ($file);

  if (!defined ($h)) {

    # Nonexistent files are obviously too small!
    # Already printed $verbose message about the file not existing.
    return 0 unless -f $file;

    print STDERR "$progname: $file: unable to determine image size\n"
      if ($verbose);
    # Assume that unknown files are of good sizes: this will happen if
    # they matched $good_file_re, but we don't have code to parse them.
    # (This will also happen if the file is junk...)
    return 1;
  }

  if ($w < $min_image_width || $h < $min_image_height) {
    print STDERR "$progname: $file: too small ($w x $h)\n" if ($verbose);
    return 0;
  }

  print STDERR "$progname: $file: $w x $h\n" if ($verbose);
  return 1;
}



# Given the raw body of a GIF document, returns the dimensions of the image.
#
sub gif_size($) {
  my ($body) = @_;
  my $type = substr($body, 0, 6);
  my $s;
  return () unless ($type =~ /GIF8[7,9]a/);
  $s = substr ($body, 6, 10);
  my ($a,$b,$c,$d) = unpack ("C"x4, $s);
  return (($b<<8|$a), ($d<<8|$c));
}

# Given the raw body of a JPEG document, returns the dimensions of the image.
#
sub jpeg_size($) {
  my ($body) = @_;
  my $i = 0;
  my $L = length($body);

  my $c1 = substr($body, $i, 1); $i++;
  my $c2 = substr($body, $i, 1); $i++;
  return () unless (ord($c1) == 0xFF && ord($c2) == 0xD8);

  my $ch = "0";
  while (ord($ch) != 0xDA && $i < $L) {
    # Find next marker, beginning with 0xFF.
    while (ord($ch) != 0xFF) {
      return () if (length($body) <= $i);
      $ch = substr($body, $i, 1); $i++;
    }
    # markers can be padded with any number of 0xFF.
    while (ord($ch) == 0xFF) {
      return () if (length($body) <= $i);
      $ch = substr($body, $i, 1); $i++;
    }

    # $ch contains the value of the marker.
    my $marker = ord($ch);

    if (($marker >= 0xC0) &&
        ($marker <= 0xCF) &&
        ($marker != 0xC4) &&
        ($marker != 0xCC)) {  # it's a SOFn marker
      $i += 3;
      return () if (length($body) <= $i);
      my $s = substr($body, $i, 4); $i += 4;
      my ($a,$b,$c,$d) = unpack("C"x4, $s);
      return (($c<<8|$d), ($a<<8|$b));

    } else {
      # We must skip variables, since FFs in variable names aren't
      # valid JPEG markers.
      return () if (length($body) <= $i);
      my $s = substr($body, $i, 2); $i += 2;
      my ($c1, $c2) = unpack ("C"x2, $s);
      my $length = ($c1 << 8) | $c2;
      return () if ($length < 2);
      $i += $length-2;
    }
  }
  return ();
}

# Given the raw body of a PNG document, returns the dimensions of the image.
#
sub png_size($) {
  my ($body) = @_;
  return () unless ($body =~ m/^\211PNG\r/s);
  my ($bits) = ($body =~ m/^.{12}(.{12})/s);
  return () unless defined ($bits);
  return () unless ($bits =~ /^IHDR/);
  my ($ign, $w, $h) = unpack("a4N2", $bits);
  return ($w, $h);
}


# Given the raw body of a GIF, JPEG, or PNG document, returns the dimensions
# of the image.
#
sub image_size($) {
  my ($body) = @_;
  return () if (length($body) < 10);
  my ($w, $h) = gif_size ($body);
  if ($w && $h) { return ($w, $h); }
  ($w, $h) = jpeg_size ($body);
  if ($w && $h) { return ($w, $h); }
  # #### TODO: need image parsers for TIFF, XPM, XBM.
  return png_size ($body);
}

# Returns the dimensions of the image file.
#
sub image_file_size($) {
  my ($file) = @_;
  my $in;
  if (! open ($in, '<', $file)) {
    print STDERR "$progname: $file: $!\n" if ($verbose);
    return ();
  }
  binmode ($in);  # Larry can take Unicode and shove it up his ass sideways.
  my $body = '';
  sysread ($in, $body, 1024 * 50);  # The first 50k should be enough.
  close $in;			    # (It's not for certain huge jpegs...
  return image_size ($body);	    # but we know they're huge!)
}


sub error($) {
  my ($err) = @_;
  print STDERR "$progname: $err\n";
  exit 1;
}

sub usage() {
  print STDERR "usage: $progname [--verbose] directory\n" .
  "       Prints the name of a randomly-selected image file.  The directory\n" .
  "       is searched recursively.  Images smaller than " .
         "${min_image_width}x${min_image_height} are excluded.\n" .
  "\n" .
  "       The directory may also be the URL of an RSS/Atom feed.  Enclosed\n" .
  "       images will be downloaded and cached locally.\n" .
  "\n";
  exit 1;
}

sub main() {
  my $dir = undef;

  while ($_ = $ARGV[0]) {
    shift @ARGV;
    if ($_ eq "--verbose") { $verbose++; }
    elsif (m/^-v+$/) { $verbose += length($_)-1; }
    elsif ($_ eq "--name") { }     # ignored, for compatibility
    elsif ($_ eq "--spotlight")    { $use_spotlight_p = 1; }
    elsif ($_ eq "--no-spotlight") { $use_spotlight_p = 0; }
    elsif ($_ eq "--cache")        { $cache_p = 1; }
    elsif ($_ eq "--no-cache")     { $cache_p = 0; }
    elsif (m/^-./) { usage; }
    elsif (!defined($dir)) { $dir = $_; }
    else { usage; }
  }

  usage unless (defined($dir));

  $dir =~ s@^feed:@http:@si;

  if ($dir =~ m/^https?:/si) {
    # ok
  } else {
    $dir =~ s@^~/@$ENV{HOME}/@s;     # allow literal "~/"
    $dir =~ s@/+$@@s;		   # omit trailing /

    if (! -d $dir) {
      print STDERR "$progname: $dir: not a directory or URL\n";
      usage;
    }
  }

  my $file = find_random_file ($dir);
  print STDOUT "$file\n";
}

main;
exit 0;