2 * Copyright notice from original mutt:
3 * Copyright (C) 1996-2000 Michael R. Elkins <me@mutt.org>
4 * Copyright (C) 2000-2001 Edmund Grimley Evans <edmundo@rano.org>
6 * This file is part of mutt-ng, see http://www.muttng.org/.
7 * It's licensed under the GNU General Public License,
8 * please see the file GPL in the top level source directory.
15 #include <lib-lib/mem.h>
33 /* If you are debugging this file, comment out the following line. */
42 #define ENCWORD_LEN_MAX 75
43 #define ENCWORD_LEN_MIN 9 /* str_len ("=?.?.?.?=") */
45 #define HSPACE(x) ((x) == '\0' || (x) == ' ' || (x) == '\t')
47 #define CONTINUATION_BYTE(c) (((c) & 0xc0) == 0x80)
49 extern char RFC822Specials[];
51 typedef size_t (*encoder_t) (char *, const char *, size_t,
54 static size_t convert_string (const char *f, size_t flen,
55 const char *from, const char *to,
56 char **t, size_t * tlen)
63 cd = mutt_iconv_open (to, from, 0);
64 if (cd == (iconv_t) (-1))
67 ob = buf = p_new(char, obl);
68 n = my_iconv(cd, &f, &flen, &ob, &obl);
69 if (n == (size_t) (-1) || my_iconv(cd, 0, 0, &ob, &obl) == (size_t) (-1)) {
80 mem_realloc (&buf, ob - buf + 1);
87 char *mutt_choose_charset (const char *fromcode, const char *charsets,
88 char *u, size_t ulen, char **d, size_t * dlen)
90 char canonical_buff[LONG_STRING];
91 char *e = 0, *tocode = 0;
92 size_t elen = 0, bestn = 0;
95 for (p = charsets; p; p = q ? q + 1 : 0) {
101 n = q ? q - p : str_len (p);
104 /* Assume that we never need more than 12 characters of
105 encoded-text to encode a single character. */
106 n > (ENCWORD_LEN_MAX - ENCWORD_LEN_MIN + 2 - 12))
111 n = convert_string (u, ulen, fromcode, t, &s, &slen);
112 if (n == (size_t) (-1))
115 if (!tocode || n < bestn) {
140 mutt_canonical_charset (canonical_buff, sizeof (canonical_buff), tocode);
141 str_replace (&tocode, canonical_buff);
146 static size_t b_encoder (char *s, const char *d, size_t dlen,
151 memcpy (s, "=?", 2), s += 2;
152 memcpy (s, tocode, str_len (tocode)), s += str_len (tocode);
153 memcpy (s, "?B?", 3), s += 3;
157 else if (dlen == 1) {
158 *s++ = B64Chars[(*d >> 2) & 0x3f];
159 *s++ = B64Chars[(*d & 0x03) << 4];
164 else if (dlen == 2) {
165 *s++ = B64Chars[(*d >> 2) & 0x3f];
166 *s++ = B64Chars[((*d & 0x03) << 4) | ((d[1] >> 4) & 0x0f)];
167 *s++ = B64Chars[(d[1] & 0x0f) << 2];
172 *s++ = B64Chars[(*d >> 2) & 0x3f];
173 *s++ = B64Chars[((*d & 0x03) << 4) | ((d[1] >> 4) & 0x0f)];
174 *s++ = B64Chars[((d[1] & 0x0f) << 2) | ((d[2] >> 6) & 0x03)];
175 *s++ = B64Chars[d[2] & 0x3f];
179 memcpy (s, "?=", 2), s += 2;
183 static size_t q_encoder (char *s, const char *d, size_t dlen,
186 char hex[] = "0123456789ABCDEF";
189 memcpy (s, "=?", 2), s += 2;
190 memcpy (s, tocode, str_len (tocode)), s += str_len (tocode);
191 memcpy (s, "?Q?", 3), s += 3;
193 unsigned char c = *d++;
197 else if (c >= 0x7f || c < 0x20 || c == '_' || strchr (MimeSpecials, c)) {
199 *s++ = hex[(c & 0xf0) >> 4];
200 *s++ = hex[c & 0x0f];
205 memcpy (s, "?=", 2), s += 2;
210 * Return 0 if and set *encoder and *wlen if the data (d, dlen) could
211 * be converted to an encoded word of length *wlen using *encoder.
212 * Otherwise return an upper bound on the maximum length of the data
213 * which could be converted.
214 * The data is converted from fromcode (which must be stateless) to
215 * tocode, unless fromcode is 0, in which case the data is assumed to
216 * be already in tocode, which should be 8-bit and stateless.
218 static size_t try_block (const char *d, size_t dlen,
219 const char *fromcode, const char *tocode,
220 encoder_t * encoder, size_t * wlen)
222 char buf1[ENCWORD_LEN_MAX - ENCWORD_LEN_MIN + 1];
227 int count, len, len_b, len_q;
230 cd = mutt_iconv_open (tocode, fromcode, 0);
231 assert (cd != (iconv_t) (-1));
232 ib = d, ibl = dlen, ob = buf1, obl = sizeof (buf1) - str_len (tocode);
233 if (my_iconv(cd, &ib, &ibl, &ob, &obl) == (size_t) (-1) ||
234 my_iconv(cd, 0, 0, &ob, &obl) == (size_t) (-1)) {
235 assert (errno == E2BIG);
238 return (ib - d == dlen) ? dlen : ib - d + 1;
243 if (dlen > sizeof (buf1) - str_len (tocode))
244 return sizeof (buf1) - str_len (tocode) + 1;
245 memcpy (buf1, d, dlen);
250 for (p = buf1; p < ob; p++) {
251 unsigned char c = *p;
253 assert (strchr (MimeSpecials, '?'));
254 if (c >= 0x7f || c < 0x20 || *p == '_' ||
255 (c != ' ' && strchr (MimeSpecials, *p)))
259 len = ENCWORD_LEN_MIN - 2 + str_len (tocode);
260 len_b = len + (((ob - buf1) + 2) / 3) * 4;
261 len_q = len + (ob - buf1) + 2 * count;
263 /* Apparently RFC 1468 says to use B encoding for iso-2022-jp. */
264 if (!ascii_strcasecmp (tocode, "ISO-2022-JP"))
265 len_q = ENCWORD_LEN_MAX + 1;
267 if (len_b < len_q && len_b <= ENCWORD_LEN_MAX) {
268 *encoder = b_encoder;
272 else if (len_q <= ENCWORD_LEN_MAX) {
273 *encoder = q_encoder;
282 * Encode the data (d, dlen) into s using the encoder.
283 * Return the length of the encoded word.
285 static size_t encode_block (char *s, char *d, size_t dlen,
286 const char *fromcode, const char *tocode,
289 char buf1[ENCWORD_LEN_MAX - ENCWORD_LEN_MIN + 1];
293 size_t ibl, obl, n1, n2;
296 cd = mutt_iconv_open (tocode, fromcode, 0);
297 assert (cd != (iconv_t) (-1));
298 ib = d, ibl = dlen, ob = buf1, obl = sizeof (buf1) - str_len (tocode);
299 n1 = my_iconv(cd, &ib, &ibl, &ob, &obl);
300 n2 = my_iconv(cd, 0, 0, &ob, &obl);
301 assert (n1 != (size_t) (-1) && n2 != (size_t) (-1));
303 return (*encoder) (s, buf1, ob - buf1, tocode);
306 return (*encoder) (s, d, dlen, tocode);
310 * Discover how much of the data (d, dlen) can be converted into
311 * a single encoded word. Return how much data can be converted,
312 * and set the length *wlen of the encoded word and *encoder.
313 * We start in column col, which limits the length of the word.
315 static size_t choose_block (char *d, size_t dlen, int col,
316 const char *fromcode, const char *tocode,
317 encoder_t * encoder, size_t * wlen)
320 int utf8 = fromcode && !ascii_strcasecmp (fromcode, "UTF-8");
325 nn = try_block (d, n, fromcode, tocode, encoder, wlen);
326 if (!nn && (col + *wlen <= ENCWORD_LEN_MAX + 1 || n <= 1))
328 n = (nn ? nn : n) - 1;
331 while (n > 1 && CONTINUATION_BYTE (d[n]))
338 * Place the result of RFC-2047-encoding (d, dlen) into the dynamically
339 * allocated buffer (e, elen). The input data is in charset fromcode
340 * and is converted into a charset chosen from charsets.
341 * Return 1 if the conversion to UTF-8 failed, 2 if conversion from UTF-8
342 * failed, otherwise 0. If conversion failed, fromcode is assumed to be
343 * compatible with us-ascii and the original data is used.
344 * The input data is assumed to be a single line starting at column col;
345 * if col is non-zero, the preceding character was a space.
347 static int rfc2047_encode (const char *d, size_t dlen, int col,
348 const char *fromcode, const char *charsets,
349 char **e, size_t * elen, char *specials)
353 size_t bufpos, buflen;
354 char *u, *t0, *t1, *t;
356 size_t ulen, r, n, wlen;
360 const char *icode = "UTF-8";
362 /* Try to convert to UTF-8. */
363 if (convert_string (d, dlen, fromcode, icode, &u, &ulen)) {
366 u = p_dupstr(d, ulen = dlen);
369 /* Find earliest and latest things we must encode. */
370 s0 = s1 = t0 = t1 = 0;
371 for (t = u; t < u + ulen; t++) {
373 (*t == '=' && t[1] == '?' && (t == u || HSPACE (*(t - 1))))) {
378 else if (specials && strchr (specials, *t)) {
385 /* If we have something to encode, include RFC822 specials */
386 if (t0 && s0 && s0 < t0)
388 if (t1 && s1 && s1 > t1)
392 /* No encoding is required. */
398 /* Choose target charset. */
401 if ((tocode1 = mutt_choose_charset (icode, charsets, u, ulen, 0, 0)))
407 /* Hack to avoid labelling 8-bit data as us-ascii. */
408 if (!icode && mutt_is_us_ascii (tocode))
409 tocode = "unknown-8bit";
411 /* Adjust t0 for maximum length of line. */
412 t = u + (ENCWORD_LEN_MAX + 1) - col - ENCWORD_LEN_MIN;
419 /* Adjust t0 until we can encode a character after a space. */
420 for (; t0 > u; t0--) {
421 if (!HSPACE (*(t0 - 1)))
425 while (t < u + ulen && CONTINUATION_BYTE (*t))
427 if (!try_block (t0, t - t0, icode, tocode, &encoder, &wlen) &&
428 col + (t0 - u) + wlen <= ENCWORD_LEN_MAX + 1)
432 /* Adjust t1 until we can encode a character before a space. */
433 for (; t1 < u + ulen; t1++) {
438 while (CONTINUATION_BYTE (*t))
440 if (!try_block (t, t1 - t, icode, tocode, &encoder, &wlen) &&
441 1 + wlen + (u + ulen - t1) <= ENCWORD_LEN_MAX + 1)
445 /* We shall encode the region [t0,t1). */
447 /* Initialise the output buffer with the us-ascii prefix. */
449 buf = p_new(char, buflen);
451 memcpy (buf, u, t0 - u);
457 /* Find how much we can encode. */
458 n = choose_block (t, t1 - t, col, icode, tocode, &encoder, &wlen);
460 /* See if we can fit the us-ascii suffix, too. */
461 if (col + wlen + (u + ulen - t1) <= ENCWORD_LEN_MAX + 1)
465 while (CONTINUATION_BYTE (t[n]))
469 /* This should only happen in the really stupid case where the
470 only word that needs encoding is one character long, but
471 there is too much us-ascii stuff after it to use a single
472 encoded word. We add the next word to the encoded region
474 assert (t1 < u + ulen);
475 for (t1++; t1 < u + ulen && !HSPACE (*t1); t1++);
478 n = choose_block (t, n, col, icode, tocode, &encoder, &wlen);
481 /* Add to output buffer. */
482 #define LINEBREAK "\n\t"
483 if (bufpos + wlen + str_len (LINEBREAK) > buflen) {
484 buflen = bufpos + wlen + str_len (LINEBREAK);
485 mem_realloc (&buf, buflen);
487 r = encode_block (buf + bufpos, t, n, icode, tocode, encoder);
490 memcpy (buf + bufpos, LINEBREAK, str_len (LINEBREAK));
491 bufpos += str_len (LINEBREAK);
499 /* Add last encoded word and us-ascii suffix to buffer. */
500 buflen = bufpos + wlen + (u + ulen - t1);
501 mem_realloc (&buf, buflen + 1);
502 r = encode_block (buf + bufpos, t, t1 - t, icode, tocode, encoder);
505 memcpy (buf + bufpos, t1, u + ulen - t1);
517 void _rfc2047_encode_string (char **pd, int encode_specials, int col)
521 const char *charsets;
523 if (!Charset || !*pd)
526 charsets = SendCharset;
527 if (!charsets || !*charsets)
530 rfc2047_encode (*pd, str_len (*pd), col,
531 Charset, charsets, &e, &elen,
532 encode_specials ? RFC822Specials : NULL);
538 void rfc2047_encode_adrlist (ADDRESS * addr, const char *tag)
541 int col = tag ? str_len (tag) + 2 : 32;
545 _rfc2047_encode_string (&ptr->personal, 1, col);
550 static int rfc2047_decode_word (char *d, const char *s, size_t len)
552 const char *pp, *pp1;
555 int enc = 0, count = 0;
556 char *charset = NULL;
558 pd = d0 = p_new(char, str_len(s));
560 for (pp = s; (pp1 = strchr (pp, '?')); pp = pp1 + 1) {
564 /* ignore language specification a la RFC 2231 */
566 if ((t1 = memchr (pp, '*', t - pp)))
568 charset = p_dupstr(pp, t - pp);
571 if (toupper ((unsigned char) *pp) == 'Q')
572 enc = ENCQUOTEDPRINTABLE;
573 else if (toupper ((unsigned char) *pp) == 'B')
582 if (enc == ENCQUOTEDPRINTABLE) {
583 for (; pp < pp1; pp++) {
586 else if (*pp == '=' &&
587 (!(pp[1] & ~127) && hexval (pp[1]) != -1) &&
588 (!(pp[2] & ~127) && hexval (pp[2]) != -1)) {
589 *pd++ = (hexval (pp[1]) << 4) | hexval (pp[2]);
597 else if (enc == ENCBASE64) {
600 for (; pp < pp1; pp++) {
603 if ((*pp & ~127) || (c = base64val (*pp)) == -1)
607 *pd++ = b | (c >> k);
622 mutt_convert_string (&d0, charset, Charset, M_ICONV_HOOK_FROM);
623 strfcpy (d, d0, len);
630 * Find the start and end of the first encoded word in the string.
631 * We use the grammar in section 2 of RFC 2047, but the "encoding"
632 * must be B or Q. Also, we don't require the encoded word to be
633 * separated by linear-white-space (section 5(1)).
635 static const char *find_encoded_word (const char *s, const char **x)
640 while ((p = strstr (q, "=?"))) {
642 0x20 < *q && *q < 0x7f && !strchr ("()<>@,;:\"/[]?.=", *q); q++);
643 if (q[0] != '?' || !strchr ("BbQq", q[1]) || q[2] != '?')
645 for (q = q + 3; 0x20 <= *q && *q < 0x7f && *q != '?'; q++);
646 if (q[0] != '?' || q[1] != '=') {
658 /* return length of linear white space */
659 static size_t lwslen (const char *s, size_t n)
667 for (; p < s + n; p++)
668 if (!strchr (" \t\r\n", *p)) {
669 len = (size_t) (p - s);
672 if (strchr ("\r\n", *(p - 1))) /* LWS doesn't end with CRLF */
677 /* return length of linear white space : reverse */
678 static size_t lwsrlen (const char *s, size_t n)
680 const char *p = s + n - 1;
686 if (strchr ("\r\n", *p)) /* LWS doesn't end with CRLF */
690 if (!strchr (" \t\r\n", *p)) {
691 len = (size_t) (s + n - 1 - p);
697 /* try to decode anything that looks like a valid RFC2047 encoded
698 * header field, ignoring RFC822 parsing rules
700 void rfc2047_decode (char **pd)
704 int found_encoded = 0;
712 dlen = 4 * str_len (s); /* should be enough */
713 d = d0 = p_new(char, dlen + 1);
715 while (*s && dlen > 0) {
716 if (!(p = find_encoded_word (s, &q))) {
717 /* no encoded words */
718 if (!option (OPTSTRICTMIME)) {
720 if (found_encoded && (m = lwslen (s, n)) != 0) {
722 *d = ' ', d++, dlen--;
725 if (ascii_strcasecmp (AssumedCharset, "us-ascii")) {
730 if (mutt_convert_nonmime_string (&t) == 0) {
732 strncpy (d, t, tlen);
743 strncpy (d, s, dlen);
749 n = (size_t) (p - s);
750 /* ignore spaces between encoded words
751 * and linear white spaces between encoded word and *text */
752 if (!option (OPTSTRICTMIME)) {
753 if (found_encoded && (m = lwslen (s, n)) != 0) {
755 *d = ' ', d++, dlen--;
759 if ((m = n - lwsrlen (s, n)) != 0) {
766 *d = ' ', d++, dlen--;
769 else if (!found_encoded || strspn (s, " \t\r\n") != n) {
778 rfc2047_decode_word (d, p, dlen);
792 void rfc2047_decode_adrlist (ADDRESS * a)
796 rfc2047_decode (&a->personal);
801 void rfc2047_decode_envelope (ENVELOPE* e) {
806 /* do RFC2047 decoding */
807 rfc2047_decode_adrlist (e->from);
808 rfc2047_decode_adrlist (e->to);
809 rfc2047_decode_adrlist (e->cc);
810 rfc2047_decode_adrlist (e->bcc);
811 rfc2047_decode_adrlist (e->reply_to);
812 rfc2047_decode_adrlist (e->mail_followup_to);
813 rfc2047_decode_adrlist (e->return_path);
814 rfc2047_decode_adrlist (e->sender);
817 rfc2047_decode (&e->subject);
818 mutt_adjust_subject (e);