#!/usr/bin/perl -w
# $Id: t-prot,v 1.253 2007/03/09 15:06:04 jochen Exp $

require 5.006;
use strict;
use Fcntl qw(O_EXCL O_WRONLY O_CREAT);
use Getopt::Mixed qw(nextOption);
use constant VER            => '2.4';
use constant REV            => '';
use constant REL            => q$Revision: 1.253 $=~m/(\d+(?:\.\d+)+)/;
# MTA expecting mail on STDIN
# (you might have to adjust this if using a different MTA)
use constant SENDMAIL       => '/usr/sbin/sendmail -oi';
# From <sysexits.h>
# (you might have to adjust those if your libc wants different values)
use constant EX_OK          =>  0;
use constant EX_USAGE       => 64;
use constant EX_DATAERR     => 65;
use constant EX_UNAVAILABLE => 69;
use constant EX_SOFTWARE    => 70;
use constant EX_IOERR       => 74;
use constant EX_BOUNCE      => EX_UNAVAILABLE;
use vars qw(
    $ad $ads $bigqn $bigqx $boun $check $check_ratio $cr $diff $elli
    $footers $ftr_ad $ftr_ml $hdrs $indent $kamm $kdiff $kminl $kmaxl
    $lax $lsig $maxsig $maxlines $mda $ml $ms $ms_smart $msg_quote
    $msg_ratio $mua $nohdr $ofile $pgpshort $pgpmove $pgpmovevrf $sig
    $sigint $sign $spass $spass_prefix $sysl $trad $trsp

    $gpg_WARNING $gpg_Warning $gpg_Cantcheck $gpg_aka $gpg_bad
    $gpg_expired $gpg_good $gpg_bug

    $mutt_attachment $mutt_contenttype $mutt_pgpsigned $mutt_beginsigned
    $mutt_pgpclearsigned $mutt_pgpclearsigstart $mutt_pgpencrypted
    $mutt_pgpoutstart $mutt_pgpoutend
);


# Please adjust these vals to your needs (they are no constants because
# command line can change them or they are used in rexexp's):
$0 =~ s!^.*/!!;
$maxsig         = 4;      # max. valid signature length
$maxlines       = undef;  # no limit of message lines
$indent         = '>';    # Indent string, regexp to identify a quoted line
$kminl          = 65;     # see decomb() for details
$kmaxl          = 80;
$kdiff          = 20;
$pgpshort       = 0;      # hide pgp key ids if set
$pgpmove        = 0;      # move pgp output to bottom if set
$pgpmovevrf     = 0;
$sign           = 1;      # max number of sigs tolerated, undef for no limit
$boun           = "Blocked by $0: This user does not accept TOFUed email. Please see <http://learn.to/quote/> and <http://www.escape.de/users/tolot/mutt/> for more info. Have a nice day!\n";
$ftr_ad         = undef;  # too hard on performance to be default
$ftr_ml         = undef;  # too hard on performance to be default
$ofile          = '-';    # use STDOUT if nothing is specified
$spass_prefix   = 'SPAM: ';
$check_ratio    = .75;    # 3/4 tofu is enough not to accept the message
$msg_quote      = "Blocked by $0: This message has been rejected because of a full quote. Please see http://learn.to/quote/ and http://www.escape.de/users/tolot/mutt/ for more info. Have a nice day!\n";
$msg_ratio      = "Blocked by $0: This message has been rejected because of excessive quoting. Please see http://learn.to/quote/ and http://www.escape.de/users/tolot/mutt/ for more info. Have a nice day!\n";
# end of user adjusted vals


# set the defaults to the C locale
$mutt_attachment        = '[-- Attachment #';
$mutt_contenttype       = '[-- Type: ';
$mutt_pgpsigned         = '[-- End of signed data --]';
$mutt_beginsigned       = '[-- The following data is signed --]';
$mutt_pgpclearsigstart  = '[-- BEGIN PGP SIGNED MESSAGE --]';
$mutt_pgpclearsigned    = '[-- END PGP SIGNED MESSAGE --]';
$mutt_pgpencrypted      = '[-- End of PGP/MIME encrypted data --]';
$mutt_pgpoutstart       = '[-- PGP output follows (current time:';
$mutt_pgpoutend         = '[-- End of PGP output --]';

# set the defaults to the C locale
$gpg_WARNING            = 'WARNING: ';
# (yes, the translations in gnupg's po files *are* braindamaged):
$gpg_Warning            = 'WARNING: ';
$gpg_Cantcheck          = 'Can\'t check signature: ';
$gpg_aka                = '                aka ';
$gpg_bad                = 'BAD signature from ';
$gpg_expired            = 'Note: This key has expired!';
$gpg_good               = 'Good signature from';
# (actually, this bugs me quite often since upgrading gpg from v1.0.7):
$gpg_bug                = '... this is a bug (';


# help(): print help text and exit with appropriate exit code
sub help {
    print "Usage: $0 [options] 
  -A=DIRECTORY    ad footer directory, treat ad footers as signature
  -a              remove ad footers; requires -A
  --bigq[=n[,x]]  remove all but x lines of quotes with more than n
                  lines; default is n=30 and x=10
  --body          input has no headers; does not work with --pgp-short;
                  multipart messages will not be detected
  -c              merge multiple blank lines
  --check[=FLAGS] check various criteria, print error message and quit;
                  see man page for details
  -d, --debug     print notice to syslog when bouncing; requires -p
  --diff          tolerate diffs appended *after* the signature
  -e              force ellipsis for excessive punctuation
  --ftr-ad        enable aggressive ad footer matching; requires -A
  --ftr-ml        enable aggressive mailing list footer matching; req. -L
  -h, --help      show this short help and exit
  -i=INFILE       file to be read; '-' for STDIN (default)
  -k              try to fix \"Kammquotes\"
  --kminl=n       min. line length for wrapped line; requires -k
  --kmaxl=n       max. line length for wrapped line; requires -k
  --kdiff=n       max. length difference between wrapped lines; req. -k
  -L=DIRECTORY    mailling list footer directory, treat mailing list
                  footers as signature
  -l              delete mailing list footer; requires -L
  --lax-security  use unsafe writing method; USE ON YOUR OWN RISK!
  --locale=LOCALE internationalization; currently only used with -Mmutt
  -M, --mua=MUA   turn on special treatment for some mail user agents
  -m              delete MS style TOFU; careful: might be too agressive
  --max-lines=x   maximum number of message lines
  --ms-smart      try to be smart with MS style TOFU; req. -Mmutt and -m
  -o=OUTFILE      file to be written to; '-' for STDOUT (default), 'NONE'
                  for no output at all
  -P=MESSAGE      user defined bounce message; requires -p
  -p[=ADDRESS]    redirect to ADDRESS if no TOFU was found
  --pgp-move      move pgp verification output to bottom; requires -Mmutt
  --pgp-move-vrf  move pgp output if verified and good; requires -Mmutt
  --pgp-short     hide non-relevant pgp key uids; requires -Mmutt
  -r              delete mail header lines
  -S[=n]          supress signatures with more than n lines; 
                  default is $maxsig if n not specified
  -s              delete signature
  --sigsmax[=n]   max number of sigs tolerated, no value for unlimited
  --spass         enable SpamAssassin workaround
  -t              delete traditional style TOFU
  -v, --version   show version string and exit
  -w              delete trailing whitespaces\n";
    exit(EX_USAGE);
}

# version(): print version info and exit with appropriate exit code
sub version {
    print "$0 v".VER.REV." (Rev. ".REL."), Jochen Striepe <t-prot\@tolot.escape.de>
Get the latest version at <http://www.escape.de/users/tolot/mutt/>\n";
    exit(EX_OK);
}

# sigint_handler(): what to do if we receive a single SIGINT
sub sigint_handler {
    $sigint = 1;
}

# remove_footers(): remove any trailing appearance of footers contained
# in the given directory.
sub remove_footers {
    my $L = shift;        # array of message lines
    my $S = shift;        # array to store removed lines in
    my $F = shift;        # footers dir name
    my $O = shift;        # remove only one footer?
    my $V = shift;        # allow footers match before end of message
    my $off;

    if (!defined $V) {
        for ($off = 0; $#$L>=$off && $$L[$#$L-$off] =~ /^\s*$/; $off++) {;};
    }

    if ($F && scalar(@$L)) {
        if (!opendir(DIR, $F))
            { print STDERR "Could not open $F: $!\n"; exit(EX_IOERR); }
        my @feet = grep { /^[^.]/ && -f "$F/$_" } readdir DIR;
        closedir DIR;

        foreach my $f (@feet) {
            if (!open(IN, "$F/$f"))
                { print STDERR "Could not open $F/$f: $!\n"; exit(EX_IOERR); }
            my @l = <IN>;
            close IN;

            if (!scalar(@l)) { next; }
            for (my $z=0; $z<=$#l; $z++) { chomp($l[$z]); }

            if (defined $V) {
                WIPE: for (my $z=scalar(@$L)-scalar(@l); $z>=0; $z--)
                {
                    if (scalar(@l)+$z<=scalar(@$L)) {
                        my $y = 0;
                        for(my $x=1; $x<=scalar(@l); $x++) {
                            if (index($$L[scalar(@$L)-$x-$z], $l[scalar(@l)-$x])!=0) {
                                $y = 1;
                            }
                        }
                        if (!$y) {
                            unshift(@$S, @$L[$#$L-$#l-$z..$#$L]);
                            splice(@$L, $#$L-$#l-$z);
                            while (scalar(@$L) && $$L[$#$L] =~ /^\s*$/) {
                                unshift(@$S, pop(@$L));
                            }
                            if ($O) { last; } else { goto WIPE; }
                        }
                    }
                }
            }
            else {
                while (scalar(@l)<=scalar(@$L)) {
                    for(my $x=1; $x<=scalar(@l); $x++) {
                        if (index($$L[scalar(@$L)-$x-$off], $l[scalar(@l)-$x])!=0) { 
                            goto FINISH;
                        }
                    }
                    unshift(@$S, @$L[$#$L-$off-$#l..$#$L]);
                    splice(@$L, $#$L-$off-$#l);
                    while (scalar(@$L) && $$L[$#$L] =~ /^\s*$/) {
                        unshift(@$S, pop(@$L));
                    }
                    if ($O) { last; }
                }
                FINISH:
            }
        }
    }
}

# decomb(): Try to detect and fix zig-zag shaped quoting (a.k.a. German
# "Kammquoting").
sub decomb {
    my $L = shift;        # array of message lines
    my $V = shift;        # array with verbatim list
    my $max = 0;        # plausible wraparound pos

    # We scan the whole message first for a plausible common maximum line
    # length where longer lines would be wrapped.
    for (my $x=0; $x<$#$L; $x++) {
        if ($$V[$x]!=1 && $max<length($$L[$x])) { $max = length($$L[$x]); }
    }

    # Next we see if there are plausible wraparounds.    
    for (my $x=0; $x+1<$#$L; $x++) {

        # OK, 
        # * it must not be verbatim,
        # * the 2nd line must not be quoted nor empty nor just 
        #   underlining some part of the line above (using '^')
        #   nor begin with a whitespace,
        # * the 1st line must not end with a hyphen,
        # * the 2nd line must not indicate content was deleted,
        # * the 2nd line must not be some mutt(1) commentary,
        # * there must not be a valid word wrap to produce a longer
        #   1st line (if not quoted),
        # * the 1st and 2nd line together must not be longer than some
        #   magical upper limit nor shorter than some magical lower
        #   bound nor nearly of the same length, and
        # * the 3rd line must not be empty.
        # With all that odds should be quite good that we have an
        # automatedly wrapped line. Please send a note if you have
        # additional good criteria. Thanks.
        if (($$V[$x]!=1 && $$V[$x+1]!=1) &&
            (index($$L[$x+1], $indent)!=0) &&
            ($$L[$x+1] !~ /^$/) &&
            ($$L[$x+1] !~ /^[\s^]/) &&
            ($$L[$x] !~ /-$/) &&
            ($$L[$x+1] !~ /^\Q[...]\E\s*$/) &&
            ($mua ne 'mutt' || $$L[$x+1] !~ /^(?:\e[^\a]+\a)?\[-- /) &&
            (length($$L[$x])+index($$L[$x+1], ' ')>$max ||
                (index($$L[$x+1], ' ')<0 && length($$L[$x])+length($$L[$x+1])>$max)) &&
            (length($$L[$x])+length($$L[$x+1])<$kmaxl) &&
            (length($$L[$x])+length($$L[$x+1])>$kminl) &&
            (length($$L[$x])-length($$L[$x+1])>$kdiff) &&
            ($$L[$x+2] !~ /^\s*$/))
        {
            $$L[$x] =~ s/\s*$/' ' . $$L[$x+1]/e;
            splice(@$L, $x+1, 1);
            splice(@$V, $x+1, 1);
        }
    }
}

# debigq(): Finds big quotes (more than $n lines quoted) and deletes all
# but $x lines of them.
sub debigq {
    my $L = shift;        # array of message lines
    my $V = shift;        # array with verbatim list
    my $k = 0;

    for (my $i=0; $i<=$#$L; $i++) {

        if ($$V[$i]) { 
            $k = 0;
            next;
        }

        if (index($$L[$i], $indent)==0) { $k++; } else { 
            if ($k>$bigqn) {
                my $x = $k-$bigqx;
                $i -= $k;

                $$L[$i] = "[---=| Quote block shrinked by $0: " .
                    "$x lines snipped |=---]\n";
                $i++;
                splice(@$L, $i, $x-1);
                splice(@$V, $i, $x-1);

                $i++;
            }
            $k = 0;
        }
    }
}

# pgp(): treat mutt(1)'s pgp/gpg output contained in signed or encrypted
# messages
sub pgp {

    sub verified {
        my $L = shift;    # message body
        my $X = shift;    # start line
        my $Z = shift;    # end line

		my $ok = 0;

        while ($X<$Z) {
            if (index($$L[$X], "gpg: $gpg_WARNING")==0 ||
                index($$L[$X], "gpg: $gpg_Warning")==0 ||
                index($$L[$X], "gpg: $gpg_bad")==0 ||
                index($$L[$X], "gpg: $gpg_Cantcheck")==0 ||
                index($$L[$X], "gpg: $gpg_expired")==0 ||
                index($$L[$X], "gpg: $gpg_bug")==0)
                { return 0; }
            if (index($$L[$X], "gpg: $gpg_good")==0)
                { $ok = 1; }
            $X++;
        }

        return $ok;
    }

    my $L = shift;        # message body
    my $V = shift;        # verbatim list
    my $H = shift;        # headers

    my @tmp = ();
    my $tmp = 0;

    for (my $x=0; $x<scalar(@$L); $x++) {
        if ($$V[$x]) { next; }

        if ($$L[$x]=~/^(?:\e[^\a]+\a)?(?:\Q$mutt_pgpoutstart\E)/o)
        {
            my $from;
            for (my $m=0; $m<scalar(@$H); $m++) {
                if (index($$H[$m], 'From:')==0) {
                    $from = $$H[$m];
                    $m++;
                    while (exists($$H[$m]) && $$H[$m] =~ /^\s/)
                        { $from .= $$H[$m]; $m++; }
                    last;
                }
            }
            ($from) = $from=~m/(\w[\w.+-]*@(?:[\w.+-]+\.)+[A-Za-z]{2,})/;

            my $uid = 1;

            for (my $i=$x+1; $i<scalar(@$L); $i++) {
                if ($pgpshort && index($$L[$i], "gpg: $gpg_aka")==0)
                    { $uid++; }

                if ($pgpshort && $uid>1 &&
                    index($$L[$i], "gpg: $gpg_aka")==0 &&
                    index($$L[$i], $from)<0)
                { 
                    splice(@$L, $i, 1);
                    splice(@$V, $i, 1);
                    $i--;
                }
                elsif ($$L[$i]=~/^(?:\e[^\a]+\a)?(?:\Q$mutt_pgpoutend\E)/o)
                {
                    if ($pgpmove ||
                        ($pgpmovevrf && (!$sigint) && verified($L, $x+1, $i)))
                    {
                        push(@{$tmp[++$tmp]}, "\n", @$L[$x..($i+1)]);
                        splice(@$L, $x, $i-$x+2);
                        splice(@$V, $x, $i-$x+2);
                        $i -= $#{$tmp[$tmp]}-2;
                    }
                    $x = $i;
                    last;
                }
            }
        }
        elsif ($tmp &&
            $$L[$x] =~ /^
                        (?:\e[^\a]+\a)?
                        (?:\Q$mutt_pgpencrypted\E  |
                           \Q$mutt_pgpclearsigned\E|
                           \Q$mutt_pgpsigned\E)
                       /ox)
        {
            splice(@$L, $x+1, 0, @{$tmp[$tmp]});
            for (my $i=$x; $i<scalar(@{$tmp[$tmp]}); $i++) {
                splice(@$V, $x+1, 0, (0));
            }
            $x += scalar(@{$tmp[$tmp--]});
            pop(@tmp);
        }
    }

    while ($tmp>0) {
        push(@$L, @{$tmp[$tmp--]});
        pop(@tmp);
        for (my $i=$#$V; $i<$#$L; $i++) { push(@$V, 0); }
    }
}

# write_msg(): output
sub write_msg {
    my $O = shift;
    my $l;

    if ((!$lax) && ($O =~ /^>(.*)/) && ($1 ne '-')) {
        if (!sysopen(OUT, $1, O_EXCL|O_CREAT|O_WRONLY)) { 
            print STDERR "Could not open $1: $!\n"; exit(EX_IOERR);
        }
    }
    elsif (!open(OUT, $O)) { 
        print STDERR "Could not open $O: $!\n"; exit(EX_IOERR);
    }
    while (scalar(@_)) {
        $l = shift;
        if (defined $l) {
            $^W = 0;
            print OUT @$l;
            $^W = 1;
        }
    }
    close OUT;
}

# process_msg(): This one proc does *everything* what has to be done with
# the lines of the message
sub process_msg {
    my $lines = shift;

    my ($j, $x, $verb) = (0, 0, 0);
    my (@ads, @hdr, @bo1, @bo2, @ftr, @sig, @vrb, @att) = 
        ((), (), (), (), (), (), (), (), ());

    # First, remove and store lines we might need later...
    # Remove headers:
    if (!$nohdr) {
        for ($x=0; $x<$#$lines; $x++) { if (@$lines[$x] =~ /^$/) { last; }; }

        # check body length and stop processing silently if msg is too long:
        if ((defined $maxlines) && (@$lines-$x > $maxlines)) {
            return;
        }

        @hdr = @$lines[0..$x];
        splice(@$lines, 0, $x+1);
    }
    elsif ((defined $maxlines) && (@$lines > $maxlines)) {
        # check body length and stop processing silently if msg is too long:
        return;
    }
    # remember the original body lines count
    my $linecount = scalar(@$lines);

    # Remove ML footers:
    remove_footers($lines, \@ftr, $footers, undef, $ftr_ml);

    # Remove ad footers:
    remove_footers($lines, \@ads, $ads, undef, $ftr_ad);

    # See if we have a multipart content type. If yes, see if it is already
    # ripped (e.g. by you MUA, assuming it does not get special treatment with
    # --mua=foo), otherwise only leave the first part if it is plain text (if
    # not, we are done - non-text messages are not our business).
    if ($mua ne 'mutt') { 
        for ($x=0; $x<scalar(@hdr); $x++) {
            if ($hdr[$x] =~ /^Content-[Tt]ype:\s+(.*)$/) {
                my $foo = $1;

                if ($foo =~ m!^multipart/!) {
                    undef $foo;

                    if ($hdr[$x] =~ /\Wboundary="([^"]+)"/i) { $foo = $1; }
                    else { 
                        for (my $z=1; $x+$z<@hdr && $hdr[$x+$z]=~/^\s/; $z++) {
                            if ($hdr[$x] =~ /\Wboundary="?([^"\s]+)"?$/i) { 
                                $foo = $1;
                                last;
                            }
                        }
                    }

                    if (defined $foo) {
                        for (my $x=0; $x<scalar(@$lines); $x++) {
                            if (index($$lines[$x], '--'.$foo)!=0) { next; }

                            my $bar = 'text/plain';
                            for ($x++; $x<@$lines && $$lines[$x]!~/^$/; $x++)
                            {
                                if ($$lines[$x] =~ /^Content-[Tt]ype:\s+(.*)/) { 
                                    $bar = $1;
                                }
                            }
                            if ($x>=scalar(@$lines)) { exit(EX_DATAERR); }

                            if ($bar =~ m!^text/plain!) {
                                my $z;
                                for ($z=1; $x+$z<@$lines; $z++) {
                                    if (index($$lines[$x+$z], '--'.$foo)==0) {
                                        last;
                                    }
                                }
                                if ($x+$z>=scalar(@$lines)) { exit(EX_DATAERR); }

                                @bo2 = @$lines[$x+$z..$#$lines];
                                splice(@$lines, $x+$z);
                                if ($$lines[$#$lines] =~ /^\s*$/) {
                                    unshift(@bo2, pop @$lines);
                                }
                                @bo1 = @$lines[0..$x];
                                splice(@$lines, 0, $x+1);

                                # remove mailing list and ad footers within this
                                # attachment:
                                remove_footers($lines, \@ftr, $footers, undef, $ftr_ml);
                                remove_footers($lines, \@ads, $ads, undef, $ftr_ad);

                                last;
                            }
                            else { 
                                write_msg(($mda?'|'.SENDMAIL." $mda":">$ofile"),
                                    ($hdrs?undef:\@hdr), $lines);
                                exit;
                            }
                        }
                    }
                }
                last;
            }
        } 
    }


    # Protect verbatims:
    $verb = 0;
    for ($x=0; $x<scalar(@$lines); $x++) {
        if ($$lines[$x] =~ /^\s*#v([+-])$/) { 
            $verb = $1 eq '+' ? 1 : 0;
            $vrb[$x] = 1;
        } else { $vrb[$x] = $verb; }
    }

    # Calculate quoting ratio (with respect to verbatims):
    if ($check && scalar(@$lines)) {
        my ($y, $z) = (0, 0);
        for ($x=0; $x<scalar(@$lines); $x++) {
            if (!$vrb[$x]) {
                $z++;
                if (index($$lines[$x], $indent)==0) { $y++; }
            }
        }
        $y = $y/$z;

        if ($y>=$check_ratio) {
            print $msg_ratio;
            exit EX_UNAVAILABLE;
        }
    }

    if ($mua eq 'mutt') {
        # See if we find pgp output generated by mutt before we scramble
        # the thing. If yes, see if we can beautify it.
        if ($pgpshort || $pgpmove || $pgpmovevrf) { pgp($lines, \@vrb, \@hdr); }

        # Remove all but the first attachment (if this is text/plain)
        # mutt did introduce (bah!). Remember, all this ugliness could
        # be replaced with a proper and clean edit_filter patch in 
        # mutt(1) itself...
        for ($x=$#$lines; $x>=0; $x--) {
            if ($vrb[$x]) { next; }
            # The following regexp's are quite ugly because for most users
            # these lines are coloured using termcap... (bah!)
            if (($$lines[$x] =~
                     /^(?:\e[^\a]+\a)?\Q$mutt_attachment\E(\d+)/o &&
                    (($1 ne '1') ||
                    ($x<$#$lines &&
                        $$lines[$x+1] !~ m!^
                            (?:\e[^\a]+\a)?
                            (?:\Q$mutt_contenttype\E)
                            (?:text/plain|application/pgp)
                        !ox))) ||
                ($$lines[$x] =~ /^
                    (?:\e[^\a]+\a)?
                    (?:\Q$mutt_pgpsigned\E     |
                       \Q$mutt_pgpclearsigned\E|
                       \Q$mutt_pgpencrypted\E)
                /ox))
            { 
                # Strip attachments to prepare further processing
                unshift(@att, @$lines[$x..$#$lines]);
                splice(@$lines, $x);
                # Try to fix trailing empty lines
                while (scalar(@$lines) && $$lines[$#$lines] =~
                    /^(?:\e[^\a]+\a)?\s*$/) { 
                    unshift(@att, pop(@$lines));
                }

                # Remove ML and ad footers within attachments:
                my @tmp;
                if ($ml) { remove_footers($lines, \@tmp, $footers, undef); }
                if ($ad) { remove_footers($lines, \@tmp, $ads,     undef); }
                $x = scalar(@$lines);
            }
        }

        # care about the rest
        if (scalar(@att)) {
            for ($x=0; $x<$#att; $x++) {
                if ($vrb[scalar(@$lines)+$x]) { next; }

                # Pipe message/rfc822 parts to another instance of
                # process_msg() for further processing.
                # Please note that we cannot see what a hierarchy the
                # original message had -- if there were message/rfc822
                # parts within other message/rfc822 parts constellations
                # can occur which we cannot resolve. Therefore we simply
                # do not even try to be smart. This should work for most
                # situations.
                # The following regexp is quite ugly because for most
                # users the line is coloured using termcap... (bah!)
                if ($att[$x] =~
                        /^(?:\e[^\a]+\a)?\Q$mutt_attachment\E\d+/o &&
                    $att[$x+1] =~ m!^
                        (?:\e[^\a]+\a)?
                        (?:\Q$mutt_contenttype\E) (?:message/rfc822|multipart/alternative)
                    !ox)
                {
                    $x += 2;
                    while ($att[$x] !~ /^\s*$/) { $x++; }
                    $x++;

                    my @tmp = @att[$x..$#att];
                    process_msg(\@tmp);
                    splice(@att, $x, scalar(@att)-$x, @tmp);
                    $x += scalar(@tmp);
                }
            }
        }
    }

    # Remove signature:
    if (scalar(@$lines)) { 
        my $sn = 0;
        my $chk_empty = 1;
        my $empty = 0;

        for ($x = $#$lines; $x>=0; $x--) {
            if (!$vrb[$x]) {
                if ($$lines[$x] =~ /^-- $/) {
                    if ($diff) {
                        for (my $i=1; $x+$i+1<scalar(@$lines); $i++) {
                            if ($$lines[$x+$i] =~ /^-{3}\s+\S/ &&
                                $$lines[$x+$i+1] =~ /^\+{3}\s+\S/)
                            {
                                $sig = 0;
                                unshift(@sig, @$lines[$x..$#$lines]);
                                splice(@$lines, $x);
                                last;
                            }
                        }
                        if (scalar(@sig)) {
                            if (defined($sign) && ++$sn==$sign) { last; } else { next; }
                        }
                    }

                    if ($sig || ($lsig && ($#$lines-$x-$empty>$lsig))) {
                        if ($lsig && !$sig) {
                            unshift(@sig, "[---=| Overlong signature removed by $0: " .
                                (scalar(@$lines)-$x) . " lines snipped |=---]\n");
                        }
                        splice(@$lines, $x);
                    }
                    else {
                        unshift(@sig, @$lines[$x..$#$lines]);
                        splice(@$lines, $x);
                    }
                    if (defined($sign) && ++$sn==$sign) { last; } else { next; }
                }
                # any trailing newlines?
                elsif ($chk_empty && $$lines[$x] =~ /^\s*$/) { $empty++; }
                elsif ($chk_empty) { $chk_empty = 0; }
            }
        }
    }

    # See if there is some Kammquoting to fix:
    if ($kamm) { decomb($lines, \@vrb); }

    # Now care about TOFU.
    # One common mispractice is M$ style TOFU:
    if ($ms) {
        # bloat this array if you want more internationalization:
        my @tofu = ('Original Message',
                    'Original-Nachricht',
                    'Ursprüngliche Nachricht',
                    'Ursprungliche Nachricht',
                    'Mensagem original',
                    'Ursprungligt meddelande',
                    'Oorspronkelijk bericht',
                    'Message d\'origine',
                    'Forwarded message',
                    'Weitergeleitete Nachricht / Forwarded Message');
        my $k = 0;    # any text above?
        my $tmp = 0;  # flagged if inside PGP output

        DONE: for ($x=0; $x<scalar(@$lines); $x++) { 
            if (!$vrb[$x]) {
                foreach my $tmp (@tofu) {
                    if ($$lines[$x] =~ /^-+\s?$tmp\s?-+/) { 
                        $x++;
                        $trad = 0;
                        $bigqn = 0;
                        last DONE; 
                    }
                }

                if ((!$k) && $$lines[$x] !~ /^\s*$/o &&
                    ((!$mua) ||
                     ($mua eq 'mutt' &&
                         $$lines[$x] !~
                             /^(?:\e[^\a]+\a)?(?:\Q$mutt_attachment\E)/o &&
                         $$lines[$x] !~
                             /^(?:\e[^\a]+\a)?(?:\Q$mutt_contenttype\E)/o)) &&
                    ((!$spass) || index($$lines[$x], $spass_prefix)!=0))
                {
                    if ($mua eq 'mutt' && (!$tmp) &&
                        $$lines[$x] =~
                             /^(?:\e[^\a]+\a)?(?:\Q$mutt_pgpoutstart\E)/o) {
                        $tmp = 1; 
                    } elsif ($mua eq 'mutt' && $tmp && 
                        ($$lines[$x] =~
                             /^(?:\e[^\a]+\a)?(?:\Q$mutt_beginsigned\E)/o ||
                         $$lines[$x] =~
                             /^(?:\e[^\a]+\a)?(?:\Q$mutt_pgpclearsigstart\E)/o)) {
                        $tmp = 0;
                    } elsif (!$tmp) {
                        $k = 1;
                    }
                }
            }
        }

        # try to avoid false positives and only delete m$ style tofu if
        # there is text above
        if ($k) {
            if (!$ms_smart) { goto CLEAN; }

            # first, see if there is pgp stuff inside the tofu:
            my $p = 0;    # levels of pgp signed parts

            for (my $i=$x+1; $i<scalar(@$lines); $i++) {
                if ($$lines[$i] =~
                        /^(?:\e[^\a]+\a)?(?:\Q$mutt_pgpclearsigstart\E)/o) {
                    $p++;
                }
            }
            if ($p) {
                STAIRS: for (my $i=0; $i<scalar(@att); $i++) {
                    if ($p==0 && $att[$i] =~ /^(?:\e[^\a]+\a)?\[-- /o) {
                        splice(@att, 0, $i);
                        unshift(@att, "\n");
                        goto CLEAN;
                    } elsif ($att[$i] =~
                            /^(?:\e[^\a]+\a)?(?:\Q$mutt_pgpclearsigned\E)/o) {
                        splice(@att, 0, $i+1);
                        $p--;
                        goto STAIRS;
                    }
                }
                splice(@att);
            }

            # now removing is safe:
            CLEAN: $j = scalar(@$lines)-$x;
            splice(@$lines, $x);
        }
    }

    # Nothing? Then try traditional TOFU (deleting M$ style TOFU is done
    # much more aggressively, so we won't need to search any more if we
    # did find some):
    if ($trad && (!$j) && !$vrb[$#$lines]) {
        if (scalar(@$lines) && $$lines[$#$lines] =~ /^\s*$/) { 
            unshift(@sig, pop(@$lines));
        }

        my $k;
        my $x = 1;

        for (my $i=$#$lines; $i>=0; $i--) {
            if ($$lines[$i] =~ /^$indent/o) {
                $j++;
                $k = $i;
            }
            elsif ($$lines[$i] !~ /^\s*$/) { last; }
        }

        if ($j) {
            # if there is no text above, we will assume the message is meant
            # as forwarding and therefore OK
            for (my $i=$k-1; $i>=0; $i--) {
                if ($$lines[$i] !~ /^\s*$/o) {
                   $x = 0;
                   last;
                }
            }
            if ($x) {
                $j = 0;
            } else {
                splice(@$lines, $k);
            }
        }
    }

    # OK, if we found TOFU, we will leave a message that we were here...
    if ($j) { 
        # make sendmail bounce if we shall be picky 
        # and indeed found something:
        if ($mda) { 
            if ($mda ne '1') { 
                print STDERR $boun;

                if ($sysl) {
                    eval { require Sys::Syslog; }; 
                    if ($@) { warn $@; } else {
                        Sys::Syslog::setlogsock('unix');
                        Sys::Syslog::openlog("$0[$$]", 'pid', 'mail');
                        Sys::Syslog::syslog('debug', 'bounced message %s', $hdr[0]);
                        Sys::Syslog::closelog();
                    }
                }
            }

            exit EX_BOUNCE;
        }

        # if we were invoked just for checking and indeed found something,
        # print out the error message and quit:
        if ($check) {
            print $msg_quote;
            exit EX_UNAVAILABLE;
        }

        push(@$lines, "[---=| TOFU protection by $0: " .
            "$j lines snipped |=---]\n");
    }
    elsif ($mda eq '1') { exit EX_OK; }

    # Care for huge blocks of quoted original message:
    if ($bigqn) { debigq($lines, \@vrb); }

    # Care for trailing whitespaces:
    if ($trsp) {
        for ($x=0; $x<scalar(@$lines); $x++) { 
            if (!$vrb[$x]) { $$lines[$x] =~ s/[\ \t]+$//; }
        }
    }

    # Care for punctuation abuse:
    if ($elli) {
        for ($x=0; $x<scalar(@$lines); $x++) { 
            if (!$vrb[$x]) { $$lines[$x] =~ s/(([.?!])\2\2)\2+/$1/g; }
        }
    }

    # (Nearly) at last care for multiple blank lines. (Do not do this
    # earlier -- the way it is done right now would screw up the verbatim
    # list)
    if ($cr) {
        my $t = 0;
        for ($x=scalar(@$lines)-1; $x>=0; $x--) {
            if ((!$vrb[$x]) && $$lines[$x] =~ /^\s*$/) { 
                if ($t<2) { $t++; } else { splice(@$lines, $x, 1); }
            }
            else { $t = 0; }
        }
    }

    # Everything changing the body is done now. Time to fix the line count
    # header so naive clients do not get confused. Just to be sure, append
    # the old line count to X-headers.
    my $l = scalar(@bo1) + scalar(@$lines) + scalar(@att) + scalar(@bo2) +
                (!$sig?scalar(@sig):0) + (!$ml?scalar(@ftr):0) + 
                (!$ad?scalar(@ads):0);
    if ($linecount-$l!=0) {
        for ($x=0; $x<scalar(@hdr); $x++) {
            if ($hdr[$x] =~ 
                s/^(Lines:\s+)(\d+)/$1.($2-$linecount+$l)/e)
            { 
                $hdr[$#hdr] = "X-Old-Lines: $2\n";
                push(@hdr, "\n");
            }
        }
    }

    # Finally, before leaving we put everything back in right order.
    unshift(@$lines, (!$hdrs?@hdr:()), @bo1);
    push(@$lines, (!$sig?@sig:()), (!$ad?@ads:()), (!$ml?@ftr:()), @att,
        @bo2);
}


# environment
my $locale = $ENV{'LC_ALL'}?$ENV{'LC_ALL'}:
                 ($ENV{'LC_MESSAGES'}?$ENV{'LC_MESSAGES'}:$ENV{'LANG'});

# command line switches
($ad, $ads, $bigqn, $bigqx, $check, $cr, $sysl, $diff, $elli, $footers, $lax,
    $ml, $ms, $ms_smart, $mda, $mua, $hdrs, $kamm, $lsig, $nohdr, $sig, $sigint,
    $spass, $trad, $trsp) =
    (0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
my $ifile   = '-';    # use STDIN if nothing specified

# get command line params:
Getopt::Mixed::init('a A=s c d e h i=s k L=s l m M=s o=s P=s p:s r S:i'.
    ' s t v w bigq:s body check:s debug>d diff ftr-ad ftr-ml help>h'.
    ' kminl=i kmaxl=i kdiff=i lax-security locale=s max-lines=i ms-smart'.
    ' mua>M pgp-short pgp-move pgp-move-vrf sigsmax:i spass version>v');
while (my ($opt, $val, $pretty) = nextOption()) {
    if    ($opt eq 'a')     { $ad = 1; }
    elsif ($opt eq 'A')     { $ads = $val; }
    elsif ($opt eq 'bigq')  { 
        if ($val !~ /^(?:(\d+)(?:,(\d+))?)?$/) { help(); }
        $bigqn = $1?$1:30; 
        $bigqx = $2?$2:10;
        if ($bigqn<=0 || $bigqx<=0 || $bigqn<=$bigqx) { help(); }
    }
    elsif ($opt eq 'body')  { $nohdr = 1; $hdrs = 1; }
    elsif ($opt eq 'c')     { $cr = 1; }
    elsif ($opt eq 'check') {
        $check = 1;
        while ($val && $val =~ s/^([^,\s]+)(?:,(\S+))?$/$2/) {
            my $foo = $1;

            if ($foo =~ /^ratio=(0?\.\d+)$/) {
                $check_ratio = $1?$1:1;
            }
        }
    }
    elsif ($opt eq 'd')     { $sysl = 1; }
    elsif ($opt eq 'diff')  { $diff = 1; }
    elsif ($opt eq 'e')     { $elli = 1; }
    elsif ($opt eq 'ftr-ad') { $ftr_ad = 1; $ad = 1; }
    elsif ($opt eq 'ftr-ml') { $ftr_ml = 1; $ml = 1; }
    elsif ($opt eq 'i')     { $ifile = $val; }
    elsif ($opt eq 'k')     { $kamm = 1; }
    elsif ($opt eq 'kminl') { $kminl = $val; $kamm = 1; }
    elsif ($opt eq 'kmaxl') { $kmaxl = $val; $kamm = 1; }
    elsif ($opt eq 'kdiff') { $kdiff = $val; $kamm = 1; }
    elsif ($opt eq 'L')     { $footers = $val; }
    elsif ($opt eq 'l')     { $ml = 1; }
    elsif ($opt eq 'lax-security') { $lax = 1; }
    elsif ($opt eq 'locale') { $locale = $val; }
    elsif ($opt eq 'm')     { $ms = 1; }
    elsif ($opt eq 'max-lines') { $maxlines = $val; }
    elsif ($opt eq 'ms-smart') { $ms_smart = 1; $ms = 1; }
    elsif ($opt eq 'M') {
        $mua = lc($val);

        if ($mua eq 'mutt') {
            # mutt still displays the message when ^C'ing pgp verification:
            $SIG{'INT'} = 'sigint_handler';
        }
    }
    elsif ($opt eq 'o')     { $ofile = $val; }
    elsif ($opt eq 'P')     { $boun = $val; }
    elsif ($opt eq 'p')     { $mda = $val ? $val : '1'; }
    elsif ($opt eq 'pgp-short') { $pgpshort = 1; }
    elsif ($opt eq 'pgp-move') { $pgpmove = 1; }
    elsif ($opt eq 'pgp-move-vrf') { $pgpmovevrf = 1; }
    elsif ($opt eq 'r')     { $hdrs = 1; }
    elsif ($opt eq 'S')     { $lsig = $val ? $val : $maxsig; }
    elsif ($opt eq 's')     { $sig = 1; }
    elsif ($opt eq 'sigsmax') { $sign = $val ? $val : undef; }
    elsif ($opt eq 'spass') { $spass = 1; }
    elsif ($opt eq 't')     { $trad = 1; }
    elsif ($opt eq 'v')     { version(); }
    elsif ($opt eq 'w')     { $trsp = 1; }
    else                    { help(); }
}
Getopt::Mixed::cleanup();
if (($ml && $footers eq '')||    # no -l without -L
    ($ad && $ads eq '')||        # no -a without -A
    ($nohdr && $pgpshort)||      # --body and --pgp-short are like oil and water
    ($ifile eq '')||             # no empty -i
    ($ofile eq ''))              # no empty -o
{ help(); }


if ($mua eq 'mutt') {
    if (defined $locale && $locale ne '' && $locale ne 'C' && $locale ne 'POSIX') {
        eval { require Locale::gettext; };
        if ($@) { warn $@; exit(EX_SOFTWARE); } else {
            Locale::gettext::textdomain('mutt');
            ($mutt_attachment) =
                Locale::gettext::gettext("[-- Attachment #%d") =~
                m/^([^%]*)/;
            ($mutt_contenttype) =
                Locale::gettext::gettext("[-- Type: %s/%s, Encoding: %s, Size: %s --]\n") =~
                m/^([^%]*)/;
            ($mutt_pgpsigned) =
                Locale::gettext::gettext("[-- End of signed data --]\n")  =~
                m/^(.*)\n/m;
            ($mutt_beginsigned) =
                Locale::gettext::gettext("[-- The following data is signed --]\n\n") =~
                m/^(.*)\n/m;
            ($mutt_pgpclearsigned) =
                Locale::gettext::gettext("[-- END PGP SIGNED MESSAGE --]\n") =~
                m/^(.*)\n/m;
            ($mutt_pgpclearsigstart) =
                Locale::gettext::gettext("[-- BEGIN PGP SIGNED MESSAGE --]\n") =~
                m/^(.*)\n/m;
            ($mutt_pgpencrypted) =
                Locale::gettext::gettext("[-- End of PGP/MIME encrypted data --]\n") =~
                m/^(.*)\n/m;
            ($mutt_pgpoutstart) =
                Locale::gettext::gettext("[-- PGP output follows (current time: %c) --]\n") =~
                m/^([^%]*)/;
            ($mutt_pgpoutend) =
                Locale::gettext::gettext("[-- End of PGP output --]\n") =~
                m/^(.*)\n/m;

            Locale::gettext::textdomain('gnupg');
            ($gpg_WARNING) =
                Locale::gettext::gettext("WARNING: using insecure random number generator!!\n") =~
                m/^([^:]*: )/;
            ($gpg_Warning) =
                Locale::gettext::gettext("WARNING: message was encrypted with a weak key in the symmetric cipher.\n") =~
                m/^([^:]*: )/;
            ($gpg_Cantcheck) =
                Locale::gettext::gettext("Can't check signature: %s\n") =~
                m/^([^%]*)/;
            ($gpg_aka) =
                Locale::gettext::gettext((' 'x16).'aka "%s"') =~
                m/^([^"«%]*)["«%]/;
            ($gpg_bad) =
                Locale::gettext::gettext('BAD signature from "%s"') =~
                m/^([^"«%]*)["«%]/;
            ($gpg_expired) =
                Locale::gettext::gettext("Note: This key has expired!\n") =~
                m/^(.*)\n/m;
            ($gpg_good) =
                Locale::gettext::gettext('Good signature from "%s"') =~
                m/^(([^"«%]*))["«%]/;
            ($gpg_bug) =
                Locale::gettext::gettext("... this is a bug (%s:%d:%s)\n") =~
                m/^([^%]*)/;
        }
    }

}
else {
    if ($ms_smart || $pgpshort || $pgpmove || $pgpmovevrf) { help(); }
}


# Read message:
if (!open(IN, $ifile))
    { print STDERR "Could not open $ifile: $!\n"; exit(EX_IOERR); }
my @message = <IN>;
close IN;

# this should be self-explanatory:
process_msg(\@message);

# Finally, print clean lines:
if ($ofile ne 'NONE') {
    write_msg(($mda?'|'.SENDMAIL." $mda":">$ofile"), \@message);
}

# vim600:set foldmethod=marker:
# eof