class Bodypart * Message::bodypart( const EString & s, bool create )
{
uint b = 0;
Bodypart * bp = 0;
while ( b < s.length() ) {
uint e = b;
while ( s[e] >= '0' && s[e] <= '9' )
e++;
if ( e < s.length() && s[e] != '.' )
return 0;
bool inrange = false;
uint n = s.mid( b, e-b ).number( &inrange );
b = e + 1;
if ( !inrange || n == 0 )
return 0;
List<Bodypart> * c = children();
if ( bp )
c = bp->children();
List<Bodypart>::Iterator i( c );
while ( i && i->number() < n )
++i;
if ( i && i->number() == n ) {
if ( n == 1 && !i->header() ) {
// it's possible that i doesn't have a header of its
// own, and that the parent message's header functions
// as such. link it in if that's the case.
Header * h = header();
if ( bp && bp->message() )
h = bp->message()->header();
if ( h && ( !h->contentType() ||
h->contentType()->type() != "multipart" ) )
i->setHeader( h );
}
bp = i;
}
else if ( create ) {
Bodypart * child = 0;
if ( bp )
child = new Bodypart( n, bp );
else
child = new Bodypart( n, this );
c->insert( i, child );
bp = child;
}
else {
return 0;
}
}
return bp;
}
Bodypart * Bodypart::parseBodypart( uint start, uint end,
const EString & rfc2822,
Header * h, Multipart * parent )
{
if ( rfc2822[start] == 13 )
start++;
if ( rfc2822[start] == 10 )
start++;
Bodypart * bp = new Bodypart;
bp->setParent( parent );
bp->setHeader( h );
EString body;
if ( end > start )
body = rfc2822.mid( start, end-start );
if ( !body.contains( '=' ) ) {
// sometimes people send c-t-e: q-p _and_ c-t-e: 7bit or 8bit.
// if they are equivalent we can accept it.
uint i = 0;
bool any = false;
HeaderField * f = 0;
while ( (f=h->field(HeaderField::ContentTransferEncoding,i)) != 0 ) {
if ( ((ContentTransferEncoding*)f)->encoding() == EString::QP )
any = true;
i++;
}
if ( any && i > 1 )
h->removeField( HeaderField::ContentTransferEncoding );
}
EString::Encoding e = EString::Binary;
ContentTransferEncoding * cte = h->contentTransferEncoding();
if ( cte )
e = cte->encoding();
if ( !body.isEmpty() ) {
if ( e == EString::Base64 || e == EString::Uuencode )
body = body.decoded( e );
else
body = body.crlf().decoded( e );
}
ContentType * ct = h->contentType();
if ( !ct ) {
switch ( h->defaultType() ) {
case Header::TextPlain:
h->add( "Content-Type", "text/plain" );
break;
case Header::MessageRfc822:
h->add( "Content-Type", "message/rfc822" );
break;
}
ct = h->contentType();
}
if ( ct->type() == "text" ) {
bool specified = false;
bool unknown = false;
Codec * c = 0;
if ( ct ) {
EString csn = ct->parameter( "charset" );
if ( csn.lower() == "default" )
csn = "";
if ( !csn.isEmpty() )
specified = true;
c = Codec::byName( csn );
if ( !c )
unknown = true;
if ( c && c->name().lower() == "us-ascii" ) {
// Some MTAs appear to say this in case there is no
// Content-Type field - without checking whether the
// body actually is ASCII. If it isn't, we'd better
// call our charset guesser.
(void)c->toUnicode( body );
if ( !c->valid() )
specified = false;
// Not pretty.
}
}
if ( !c )
c = new AsciiCodec;
bp->d->hasText = true;
bp->d->text = c->toUnicode( body.crlf() );
if ( c->name() == "GB2312" || c->name() == "ISO-2022-JP" ||
c->name() == "KS_C_5601-1987" ) {
// undefined code point usage in GB2312 spam is much too
// common. (GB2312 spam is much too common, but that's
// another matter.) Gb2312Codec turns all undefined code
// points into U+FFFD, so here, we can take the unicode
// form and say it's the canonical form. when a client
// later reads the message, it gets the text in unicode,
// including U+FFFD.
bool bad = !c->valid();
// the header may contain some unencoded gb2312. we bang
// it by hand, ignoring errors.
List<HeaderField>::Iterator hf( h->fields() );
while ( hf ) {
if ( !hf->valid() &&
hf->type() == HeaderField::Subject ) {
// is it right to bang only Subject?
c->reset();
hf->setValue( c->toUnicode( hf->unparsedValue() ) );
}
++hf;
}
// if the body was bad, we prefer the (unicode) in
// bp->d->text and pretend it arrived as UTF-8:
if ( bad ) {
c = new Utf8Codec;
body = c->fromUnicode( bp->d->text );
}
}
if ( ( !specified && ( !c->wellformed() ||
ct->subtype() == "html" ) ) ||
( specified && ( !c->valid() ) ) ) {
Codec * g = 0;
if ( ct->subtype() == "html" )
g = guessHtmlCodec( body );
else
g = guessTextCodec( body );
UString guessed;
if ( g )
guessed = g->toUnicode( body.crlf() );
if ( !g ) {
// if we couldn't guess anything, keep what we had if
// it's valid or explicitly specified, else use
// unknown-8bit.
if ( !specified && !c->valid() ) {
c = new Unknown8BitCodec;
bp->d->text = c->toUnicode( body.crlf() );
}
}
else {
// if we could guess something, is our guess better
// than what we had?
if ( g->wellformed() && !c->wellformed() ) {
c = g;
bp->d->text = guessed;
}
}
}
if ( specified && c->state() == Codec::Invalid ) {
// the codec was specified, and the specified codec
// resulted in an error, but did not abort conversion. we
// respond by forgetting the error, using the conversion
// result (probably including one or more U+FFFD) and
// labelling the message as UTF-8.
c = new Utf8Codec;
body = c->fromUnicode( bp->d->text );
}
else if ( !specified && c->state() == Codec::Invalid ) {
// the codec was not specified, and we couldn't find
// anything. we call it unknown-8bit.
c = new Unknown8BitCodec;
bp->d->text = c->toUnicode( body );
}
// if we ended up using a 16-bit codec and were using q-p, we
// need to reevaluate without any trailing CRLF
if ( e == EString::QP && c->name().startsWith( "UTF-16" ) )
bp->d->text = c->toUnicode( body.stripCRLF() );
if ( !c->valid() && bp->d->error.isEmpty() ) {
bp->d->error = "Could not convert body to Unicode";
if ( specified ) {
EString cs;
if ( ct )
cs = ct->parameter( "charset" );
if ( cs.isEmpty() )
cs = c->name();
bp->d->error.append( " from " + cs );
}
if ( specified && unknown )
bp->d->error.append( ": Character set not implemented" );
else if ( !c->error().isEmpty() )
bp->d->error.append( ": " + c->error() );
}
if ( c->name().lower() != "us-ascii" )
ct->addParameter( "charset", c->name().lower() );
else if ( ct )
ct->removeParameter( "charset" );
body = c->fromUnicode( bp->d->text );
bool qp = body.needsQP();
if ( cte ) {
if ( !qp ) {
h->removeField( HeaderField::ContentTransferEncoding );
cte = 0;
}
else if ( cte->encoding() != EString::QP ) {
cte->setEncoding( EString::QP );
}
}
else if ( qp ) {
h->add( "Content-Transfer-Encoding", "quoted-printable" );
cte = h->contentTransferEncoding();
}
}
else {
bp->d->data = body;
if ( ct->type() != "multipart" && ct->type() != "message" ) {
e = EString::Base64;
// there may be exceptions. cases where some format really
// needs another content-transfer-encoding:
if ( ct->type() == "application" &&
ct->subtype().startsWith( "pgp-" ) &&
!body.needsQP() ) {
// seems some PGP things need "Version: 1" unencoded
e = EString::Binary;
}
else if ( ct->type() == "application" &&
ct->subtype() == "octet-stream" &&
body.contains( "BEGIN PGP MESSAGE" ) ) {
// mutt cannot handle PGP in base64 (what a crock)
e = EString::Binary;
}
// change c-t-e to match the encoding decided above
if ( e == EString::Binary ) {
h->removeField( HeaderField::ContentTransferEncoding );
cte = 0;
}
else if ( cte ) {
cte->setEncoding( e );
}
else {
h->add( "Content-Transfer-Encoding", "base64" );
cte = h->contentTransferEncoding();
}
}
}
if ( ct->type() == "multipart" ) {
parseMultipart( start, end, rfc2822,
ct->parameter( "boundary" ),
ct->subtype() == "digest",
bp->children(), bp, false );
}
else if ( ct->type() == "message" && ct->subtype() == "rfc822" ) {
// There are sometimes blank lines before the message.
while ( rfc2822[start] == 13 || rfc2822[start] == 10 )
start++;
Message * m = new Message;
m->setParent( bp );
m->parse( rfc2822.mid( start, end-start ) );
List<Bodypart>::Iterator it( m->children() );
while ( it ) {
bp->children()->append( it );
it->setParent( bp );
++it;
}
bp->setMessage( m );
body = m->rfc822( false );
}
bp->d->numBytes = body.length();
if ( cte )
body = body.encoded( cte->encoding(), 72 );
bp->d->numEncodedBytes = body.length();
if ( bp->d->hasText ||
( ct->type() == "message" && ct->subtype() == "rfc822" ) ) {
uint n = 0;
uint i = 0;
uint l = body.length();
while ( i < l ) {
if ( body[i] == '\n' )
n++;
i++;
}
if ( l && body[l-1] != '\n' )
n++;
bp->setNumEncodedLines( n );
}
h->simplify();
return bp;
}
void SieveCommand::parse( const EString & previous )
{
if ( identifier().isEmpty() )
setError( "Command name is empty" );
bool test = false;
bool blk = false;
EString i = identifier();
if ( i == "if" || i == "elsif" ) {
test = true;
blk = true;
if ( i == "elsif" && previous != "if" && previous != "elsif" )
setError( "elsif is only permitted after if/elsif" );
}
else if ( i == "else" ) {
blk = true;
if ( previous != "if" && previous != "elsif" )
setError( "else is only permitted after if/elsif" );
}
else if ( i == "require" ) {
arguments()->numberRemainingArguments();
UStringList::Iterator i( arguments()->takeStringList( 1 ) );
EStringList a;
EStringList e;
while ( i ) {
if ( supportedExtensions()->contains( i->ascii() ) )
a.append( i->ascii().quoted() );
else
e.append( i->ascii().quoted() );
++i;
}
if ( !e.isEmpty() )
setError( "Each string must be a supported "
"sieve extension. "
"These are not: " + e.join( ", " ) );
if ( !d->require )
setError( "require is only permitted as the first command." );
else if ( parent() )
parent()->addExtensions( &a );
}
else if ( i == "stop" ) {
// nothing needed
}
else if ( i == "reject" ) {
require( "reject" );
if ( arguments()->arguments()->isEmpty() ) {
// we accept reject without reason
}
else {
// if there is an argument, it must be a string
arguments()->numberRemainingArguments();
(void)arguments()->takeString( 1 );
}
}
else if ( i == "ereject" ) {
require( "reject" );
arguments()->numberRemainingArguments();
(void)arguments()->takeString( 1 );
}
else if ( i == "fileinto" ) {
require( "fileinto" );
if ( arguments()->findTag( ":copy" ) )
require( "copy" );
if ( arguments()->findTag( ":flags" ) ) {
require( "imap4flags" );
(void)arguments()->takeTaggedStringList( ":copy" );
}
arguments()->numberRemainingArguments();
UString mailbox = arguments()->takeString( 1 );
UString p;
p.append( "/" );
p.append( mailbox );
if ( !Mailbox::validName( mailbox ) && !Mailbox::validName( p ) ) {
setError( "Expected mailbox name, but got: " + mailbox.utf8() );
}
else if ( mailbox.startsWith( "INBOX." ) ) {
// a sieve script which wants to reference a
// mailbox called INBOX.X must use lower case
// (inbox.x).
UString aox =
UStringList::split( '.', mailbox.mid( 6 ) )->join( "/" );
setError( mailbox.utf8().quoted() +
" is Cyrus syntax. Archiveopteryx uses " +
aox.utf8().quoted() );
}
}
else if ( i == "redirect" ) {
(void)arguments()->findTag( ":copy" );
arguments()->numberRemainingArguments();
EString s = arguments()->takeString( 1 ).utf8();
AddressParser ap( s );
ap.assertSingleAddress();
if ( !ap.error().isEmpty() )
setError( "Expected one normal address (local@domain), but got: "
+ s );
}
else if ( i == "keep" ) {
// nothing needed
}
else if ( i == "discard" ) {
// nothing needed
}
else if ( i == "vacation" ) {
// vacation [":days" number] [":subject" string]
// [":from" string] [":addresses" string-list]
// [":mime"] [":handle" string] <reason: string>
require( "vacation" );
// :days
uint days = 7;
if ( arguments()->findTag( ":days" ) )
days = arguments()->takeTaggedNumber( ":days" );
if ( days < 1 || days > 365 )
arguments()->tagError( ":days", "Number must be 1..365" );
// :subject
(void)arguments()->takeTaggedString( ":subject" );
// anything is acceptable, right?
// :from
if ( arguments()->findTag( ":from" ) ) {
parseAsAddress( arguments()->takeTaggedString( ":from" ),
":from" );
// XXX we don't enforce its being a local address.
}
// :addresses
if ( arguments()->findTag( ":addresses" ) ) {
UStringList * addresses
= arguments()->takeTaggedStringList( ":addresses" );
UStringList::Iterator i( addresses );
while ( i ) {
parseAsAddress( *i, ":addresses" );
++i;
}
}
// :mime
bool mime = false;
if ( arguments()->findTag( ":mime" ) )
mime = true;
// :handle
(void)arguments()->takeTaggedString( ":handle" );
// reason
arguments()->numberRemainingArguments();
UString reason = arguments()->takeString( 1 );
if ( mime ) {
if ( !reason.isAscii() )
setError( ":mime bodies must be all-ASCII, "
"8-bit text is not permitted" ); // so says the RFC
EString x = reason.utf8();
uint i = 0;
Header * h = Message::parseHeader( i, x.length(),
x, Header::Mime );
Bodypart * bp = Bodypart::parseBodypart( i, x.length(),
x, h, 0 );
if ( !h->error().isEmpty() )
setError( "While parsing MIME header: " + h->error() );
else if ( !bp->error().isEmpty() )
setError( "While parsing MIME bodypart: " + bp->error() );
List<HeaderField>::Iterator f( h->fields() );
while ( f ) {
if ( !f->name().startsWith( "Content-" ) )
setError( "Header field not permitted: " + f->name() );
++f;
}
if ( bp->children()->isEmpty() && bp->text().isEmpty() )
setError( "Vacation reply does not contain any text" );
}
else {
if ( reason.isEmpty() )
setError( "Empty vacation text does not make sense" );
}
}
else if ( i == "setflag" ||
i == "addflags" ||
i == "removeflag" ) {
arguments()->numberRemainingArguments();
(void)arguments()->takeStringList( 1 );
}
else if ( i == "notify" ) {
require( "enotify" );
UString from;
if ( arguments()->findTag( ":from" ))
from = arguments()->takeTaggedString( ":from" );
UString importance;
importance.append( "2" );
if ( arguments()->findTag( ":importance" ) )
importance = arguments()->takeTaggedString( ":from" );
uint c = importance[0];
if ( c < '1' || c > '3' )
arguments()->tagError( ":importance",
"Importance must be 1, 2 or 3" );
UStringList * options;
if ( arguments()->findTag( ":options" ) )
options = arguments()->takeTaggedStringList( ":options" );
UString message;
if ( arguments()->findTag( ":message" ) )
message = arguments()->takeTaggedString( ":message" );
arguments()->numberRemainingArguments();
UString method = arguments()->takeString( 1 );
SieveNotifyMethod * m
= new SieveNotifyMethod( method,
arguments()->takeArgument( 1 ),
this );
if ( m->valid() ) {
if ( arguments()->findTag( ":from" ) )
m->setFrom( from, arguments()->findTag( ":from" ) );
if ( arguments()->findTag( ":message" ) )
m->setMessage( message, arguments()->findTag( ":message" ) );
}
}
else {
setError( "Command unknown: " + identifier() );
}
arguments()->flagUnparsedAsBad();
if ( test ) {
// we must have a test
if ( !arguments() || arguments()->tests()->count() != 1 )
setError( "Command " + identifier() + " requires one test" );
if ( arguments() ) {
List<SieveTest>::Iterator i( arguments()->tests() );
while ( i ) {
i->parse();
if ( blk && block() ) {
if ( i->ihaveFailed() )
block()->setIhaveFailed();
else
block()->addExtensions( i->addedExtensions() );
}
++i;
}
}
}
else {
// we cannot have a test
if ( arguments() && arguments()->tests()->isEmpty() ) {
List<SieveTest>::Iterator i( arguments()->tests() );
while ( i ) {
i->setError( "Command " + identifier() +
" does not use tests" );
++i;
}
}
}
if ( blk ) {
// we must have a subsidiary block
if ( !block() ) {
setError( "Command " + identifier() +
" requires a subsidiary {..} block" );
}
else {
EString prev;
List<SieveCommand>::Iterator i( block()->commands() );
while ( i ) {
i->parse( prev );
prev = i->identifier();
++i;
}
}
}
else {
// we cannot have a subsidiary block
if ( block() )
block()->setError( "Command " + identifier() +
" does not use a subsidiary command block" );
// in this case we don't even bother syntax-checking the test
// or block
}
}