1 # This module is part of debbugs, and is released
2 # under the terms of the GPL version 2, or any later
3 # version at your option.
4 # See the file README and COPYING for more information.
6 # [Other people have contributed to this file; their copyrights should
8 # Copyright 2006 by Don Armstrong <don@donarmstrong.com>.
11 package Debbugs::MIME;
15 use base qw(Exporter);
16 use vars qw($VERSION @EXPORT_OK);
21 @EXPORT_OK = qw(parse decode_rfc1522 encode_rfc1522 convert_to_utf8 create_mime_message getmailbody);
28 use MIME::WordDecoder qw();
29 use Encode qw(decode encode encode_utf8 decode_utf8 is_utf8);
37 my $type = $entity->effective_type;
38 if ($type eq 'text/plain' or
39 ($type =~ m#text/# and $type ne 'text/html') or
40 $type eq 'application/pgp') {
41 return $entity->bodyhandle;
42 } elsif ($type eq 'multipart/alternative') {
43 # RFC 2046 says we should use the last part we recognize.
44 for my $part (reverse $entity->parts) {
45 my $ret = getmailbody($part);
49 # For other multipart types, we just pretend they're
50 # multipart/mixed and run through in order.
51 for my $part ($entity->parts) {
52 my $ret = getmailbody($part);
61 # header and decoded body respectively
62 my (@headerlines, @bodylines);
64 my $parser = new MIME::Parser;
65 mkdir "mime.tmp.$$", 0777;
66 $parser->output_under("mime.tmp.$$");
67 my $entity = eval { $parser->parse_data($_[0]) };
69 if ($entity and $entity->head->tags) {
70 @headerlines = @{$entity->head->header};
73 my $entity_body = getmailbody($entity);
74 @bodylines = $entity_body ? $entity_body->as_lines() : ();
77 # Legacy pre-MIME code, kept around in case MIME::Parser fails.
78 my @msg = split /\n/, $_[0];
81 for ($i = 0; $i <= $#msg; ++$i) {
84 while ($msg[$i + 1] =~ /^\s/) {
86 $_ .= "\n" . $msg[$i];
88 push @headerlines, $_;
91 @bodylines = @msg[$i .. $#msg];
94 rmtree "mime.tmp.$$", 0, 1;
97 shift @bodylines while @bodylines and $bodylines[0] !~ /\S/;
99 # Strip off RFC2440-style PGP clearsigning.
100 if (@bodylines and $bodylines[0] =~ /^-----BEGIN PGP SIGNED/) {
101 shift @bodylines while @bodylines and length $bodylines[0];
102 shift @bodylines while @bodylines and $bodylines[0] !~ /\S/;
103 for my $findsig (0 .. $#bodylines) {
104 if ($bodylines[$findsig] =~ /^-----BEGIN PGP SIGNATURE/) {
105 $#bodylines = $findsig - 1;
109 map { s/^- // } @bodylines;
112 return { header => [@headerlines], body => [@bodylines]};
115 =head2 create_mime_message
117 create_mime_message([To=>'don@debian.org'],$body,[$attach1, $attach2]);
119 Creates a MIME encoded message with headers given by the first
120 argument, and a message given by the second.
122 Optional attachments can be specified in the third arrayref argument.
124 Headers are passed directly to MIME::Entity::build, the message is the
127 Each of the elements of the attachment arrayref is attached as an
128 rfc822 message if it is a scalar or an arrayref; otherwise if it is a
129 hashref, the contents are passed as an argument to
134 sub create_mime_message{
135 my ($headers,$body,$attachments) = @_;
136 $attachments = [] if not defined $attachments;
138 die "The first argument to create_mime_message must be an arrayref" unless ref($headers) eq 'ARRAY';
139 die "The third argument to create_mime_message must be an arrayref" unless ref($attachments) eq 'ARRAY';
142 # MIME::Entity is stupid, and doesn't rfc1522 encode its headers, so we do it for it.
143 my $msg = MIME::Entity->build('Content-Type' => 'text/plain; charset=utf-8',
144 'Encoding' => 'quoted-printable',
145 (map{encode_rfc1522($_)} @{$headers}),
149 # Attach the attachments
150 for my $attachment (@{$attachments}) {
151 if (ref($attachment) eq 'HASH') {
152 $msg->attach(%{$attachment});
155 # This is *craptacular*, but because various MTAs
156 # (sendmail and exim4, at least) appear to eat From
157 # lines in message/rfc822 attachments, we need eat
158 # the entire From line ourselves so the MTA doesn't
159 # leave \n detrius around.
160 if (ref($attachment) eq 'ARRAY' and $attachment->[1] =~ /^From /) {
161 # make a copy so that we don't screw up anything
162 # that is expecting this arrayref to stay constant
163 $attachment = [@{$attachment}];
164 # remove the from line
165 splice @$attachment, 1, 1;
167 elsif (not ref($attachment)) {
168 # It's a scalar; remove the from line
169 $attachment =~ s/^(Received:[^\n]+\n)(From [^\n]+\n)/$1/s;
171 $msg->attach(Type => 'message/rfc822',
177 return $msg->as_string;
183 sub convert_to_utf8 {
184 my ($data, $charset) = @_;
185 # raw data just gets returned (that's the charset WordDecorder
186 # uses when it doesn't know what to do)
187 return $data if $charset eq 'raw' or is_utf8($data,1);
190 # this encode/decode madness is to make sure that the data
191 # really is valid utf8 and that the is_utf8 flag is off.
192 $result = encode("utf8",decode($charset,$data))
195 warn "Unable to decode charset; '$charset' and '$data': $@";
202 =head2 decode_rfc1522
204 decode_rfc1522('=?iso-8859-1?Q?D=F6n_Armstr=F3ng?= <don@donarmstrong.com>')
206 Turn RFC-1522 names into the UTF-8 equivalent.
211 # Set up the default RFC1522 decoder, which turns all charsets that
212 # are supported into the appropriate UTF-8 charset.
213 MIME::WordDecoder->default(new MIME::WordDecoder(
214 ['*' => \&convert_to_utf8,
218 sub decode_rfc1522 ($)
222 # this is craptacular, but leading space is hacked off by unmime.
224 my $leading_space = '';
225 $leading_space = $1 if $string =~ s/^(\s+)//;
226 # unmime calls the default MIME::WordDecoder handler set up at
227 # initialization time.
228 return $leading_space . MIME::WordDecoder::unmime($string);
231 =head2 encode_rfc1522
233 encode_rfc1522('Dön Armströng <don@donarmstrong.com>')
235 Encodes headers according to the RFC1522 standard by calling
236 MIME::Words::encode_mimeword on distinct words as appropriate.
240 # We cannot use MIME::Words::encode_mimewords because that function
241 # does not handle spaces properly at all.
243 sub encode_rfc1522 ($) {
246 # handle being passed undef properly
247 return undef if not defined $rawstr;
248 # We process words in reverse so we can preserve spacing between
249 # encoded words. This regex splits on word|nonword boundaries and
250 # nonword|nonword boundaries.
251 my @words = reverse split /(?:(?<=[\s\n])|(?=[\s\n]))/m, $rawstr;
253 my $previous_word_encoded = 0;
255 for my $word (@words) {
256 if ($word !~ m#[\x00-\x1F\x7F-\xFF]#o and $word ne ' ') {
257 $string = $word.$string;
258 $previous_word_encoded=0;
260 elsif ($word =~ /^[\s\n]$/) {
261 $string = $word.$string;
262 $previous_word_encoded = 0 if $word eq "\n";
265 my $encoded = MIME::Words::encode_mimeword($word, 'q', 'UTF-8');
266 # RFC 1522 mandates that segments be at most 76 characters
267 # long. If that's the case, we split the word up into 10
268 # character pieces and encode it. We must use the Encode
269 # magic here to avoid breaking on bit boundaries here.
270 if (length $encoded > 75) {
271 # Turn utf8 into the internal perl representation
272 # so . is a character, not a byte.
273 my $tempstr = decode_utf8($word,Encode::FB_DEFAULT);
275 # Strip it into 10 character long segments, and encode
277 # XXX It's possible that these segments are > 76 characters
278 while ($tempstr =~ s/(.{1,10})$//) {
279 # turn the character back into the utf8 representation.
280 my $tempword = encode_utf8($1);
281 # It may actually be better to eventually use
282 # the base64 encoding here, but I'm not sure
283 # if that's as widely supported as quoted
285 unshift @encoded, MIME::Words::encode_mimeword($tempword,'q','UTF-8');
287 $encoded = join(" ",@encoded);
288 # If the previous word was encoded, we must
289 # include a trailing _ that gets encoded as a
291 $encoded =~ s/\?\=$/_\?\=/ if $previous_word_encoded;
292 $string = $encoded.$string;
295 # If the previous word was encoded, we must
296 # include a trailing _ that gets encoded as a
298 $encoded =~ s/\?\=$/_\?\=/ if $previous_word_encoded;
299 $string = $encoded.$string;
301 $previous_word_encoded = 1;