git.marmaro.de Git - mmh/blob - sbr/fmt_rfc2047.c

   1 /*
   2 ** fmt_rfc2047.c -- decode RFC-2047 header format
   3 **
   4 ** This code is Copyright (c) 2002, by the authors of nmh.  See the
   5 ** COPYRIGHT file in the root directory of the nmh distribution for
   6 ** complete copyright information.
   7 */
   8
   9 #include <h/mh.h>
  10 #include <h/utils.h>
  11 #ifdef HAVE_ICONV
  12 #  include <iconv.h>
  13 #  include <errno.h>
  14 #endif
  15
  16 static signed char hexindex[] = {
  17         -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  18         -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  19         -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  20          0, 1, 2, 3, 4, 5, 6, 7, 8, 9,-1,-1,-1,-1,-1,-1,
  21         -1,10,11,12,13,14,15,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  22         -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  23         -1,10,11,12,13,14,15,-1,-1,-1,-1,-1,-1,-1,-1,-1,
  24         -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1
  25 };
  26
  27 static signed char index_64[128] = {
  28         -1,-1,-1,-1, -1,-1,-1,-1, -1,-1,-1,-1, -1,-1,-1,-1,
  29         -1,-1,-1,-1, -1,-1,-1,-1, -1,-1,-1,-1, -1,-1,-1,-1,
  30         -1,-1,-1,-1, -1,-1,-1,-1, -1,-1,-1,62, -1,-1,-1,63,
  31         52,53,54,55, 56,57,58,59, 60,61,-1,-1, -1,-1,-1,-1,
  32         -1, 0, 1, 2,  3, 4, 5, 6,  7, 8, 9,10, 11,12,13,14,
  33         15,16,17,18, 19,20,21,22, 23,24,25,-1, -1,-1,-1,-1,
  34         -1,26,27,28, 29,30,31,32, 33,34,35,36, 37,38,39,40,
  35         41,42,43,44, 45,46,47,48, 49,50,51,-1, -1,-1,-1,-1
  36 };
  37
  38 #define char64(c) (((unsigned char) (c) > 127) ? -1 : index_64[(unsigned char) (c)])
  39
  40 static int
  41 unqp(unsigned char byte1, unsigned char byte2)
  42 {
  43         if (hexindex[byte1] == -1 || hexindex[byte2] == -1)
  44                 return -1;
  45         return (hexindex[byte1] << 4 | hexindex[byte2]);
  46 }
  47
  48 /* Check if character is linear whitespace */
  49 #define is_lws(c)  ((c) == ' ' || (c) == '\t' || (c) == '\n')
  50
  51
  52 /*
  53 ** Decode the string as a RFC-2047 header field
  54 */
  55
  56 /* Add character to the destination buffer, and bomb out if it fills up */
  57 #define ADDCHR(C) do { *q++ = (C); dstlen--; if (!dstlen) goto buffull; } while (0)
  58
  59 int
  60 decode_rfc2047(char *str, char *dst, size_t dstlen)
  61 {
  62         char *p, *q, *pp;
  63         char *startofmime, *endofmime;
  64         int c, quoted_printable;
  65         int encoding_found = 0;  /* did we decode anything? */
  66         int between_encodings = 0;  /* are we between two encodings? */
  67         int equals_pending = 0;  /* is there a '=' pending?  */
  68         int whitespace = 0;  /* how much whitespace between encodings? */
  69 #ifdef HAVE_ICONV
  70         int use_iconv = 0;  /* are we converting encoding with iconv? */
  71         iconv_t cd = NULL;
  72         int fromutf8 = 0;
  73         char *saveq, *convbuf = NULL;
  74         size_t savedstlen;
  75 #endif
  76
  77         if (!str)
  78                 return 0;
  79
  80         /*
  81         ** Do a quick and dirty check for the '=' character.
  82         ** This should quickly eliminate many cases.
  83         */
  84         if (!strchr(str, '='))
  85                 return 0;
  86
  87         for (p = str, q = dst; *p; p++) {
  88
  89                 /* reset iconv */
  90 #ifdef HAVE_ICONV
  91                 if (use_iconv) {
  92                         iconv_close(cd);
  93                         use_iconv = 0;
  94                 }
  95 #endif
  96                 /*
  97                 ** If we had an '=' character pending from
  98                 ** last iteration, then add it first.
  99                 */
 100                 if (equals_pending) {
 101                         ADDCHR('=');
 102                         equals_pending = 0;
 103                         between_encodings = 0;  /* we have added non-whitespace text */
 104                 }
 105
 106                 if (*p != '=') {
 107                         /* count linear whitespace while between encodings */
 108                         if (between_encodings && is_lws(*p))
 109                                 whitespace++;
 110                         else
 111                                 between_encodings = 0;  /* we have added non-whitespace text */
 112                         ADDCHR(*p);
 113                         continue;
 114                 }
 115
 116                 equals_pending = 1;  /* we have a '=' pending */
 117
 118                 /* Check for initial =? */
 119                 if (*p == '=' && p[1] && p[1] == '?' && p[2]) {
 120                         startofmime = p + 2;
 121
 122                         /* Scan ahead for the next '?' character */
 123                         for (pp = startofmime; *pp && *pp != '?'; pp++)
 124                                 ;
 125
 126                         if (!*pp)
 127                                 continue;
 128                         *pp = '\0';
 129
 130                         /* Check if character set can be handled natively */
 131                         if (!is_native_charset(startofmime)) {
 132 #ifdef HAVE_ICONV
 133                                 /* .. it can't. We'll use iconv then. */
 134                                 cd = iconv_open(get_charset(), startofmime);
 135                                 fromutf8 = !mh_strcasecmp(startofmime, "UTF-8");
 136                                 *pp = '?';
 137                                 if (cd == (iconv_t)-1)
 138                                         continue;
 139                                 use_iconv = 1;
 140 #else
 141                                 *pp = '?';
 142                                 continue;
 143 #endif
 144                         }
 145
 146                         *pp = '?';
 147                         startofmime = pp + 1;
 148
 149                         /* Check for valid encoding type */
 150                         if (*startofmime != 'B' && *startofmime != 'b' &&
 151                                 *startofmime != 'Q' && *startofmime != 'q')
 152                                 continue;
 153
 154                         /* Is encoding quoted printable or base64? */
 155                         quoted_printable = (*startofmime == 'Q' || *startofmime == 'q');
 156                         startofmime++;
 157
 158                         /* Check for next '?' character */
 159                         if (*startofmime != '?')
 160                                 continue;
 161                         startofmime++;
 162
 163                         /*
 164                         ** Scan ahead for the ending ?=
 165                         **
 166                         ** While doing this, we will also check if encoded
 167                         ** word has any embedded linear whitespace.
 168                         */
 169                         endofmime = NULL;
 170                         for (pp = startofmime; *pp && *(pp+1); pp++) {
 171                                 if (is_lws(*pp)) {
 172                                         break;
 173                                 } else if (*pp == '?' && pp[1] == '=') {
 174                                         endofmime = pp;
 175                                         break;
 176                                 }
 177                         }
 178                         if (is_lws(*pp) || endofmime == NULL)
 179                                 continue;
 180
 181                         /*
 182                         ** We've found an encoded word, so we can drop
 183                         ** the '=' that was pending
 184                         */
 185                         equals_pending = 0;
 186
 187                         /*
 188                         ** If we are between two encoded words separated
 189                         ** only by linear whitespace, then we ignore
 190                         ** the whitespace.  We will roll back the buffer
 191                         ** the number of whitespace characters we've seen
 192                         ** since last encoded word.
 193                         */
 194                         if (between_encodings) {
 195                                 q -= whitespace;
 196                                 dstlen += whitespace;
 197                         }
 198
 199 #ifdef HAVE_ICONV
 200                         /*
 201                         ** empty encoded text. This ensures that we don't
 202                         ** malloc 0 bytes but skip on to the end
 203                         */
 204                         if (endofmime == startofmime && use_iconv) {
 205                                 use_iconv = 0;
 206                                 iconv_close(cd);
 207                         }
 208
 209                         if (use_iconv) {
 210                                 saveq = q;
 211                                 savedstlen = dstlen;
 212                                 q = convbuf = mh_xcalloc(endofmime - startofmime, sizeof(char));
 213                         }
 214 /*
 215 ** ADDCHR2 is for adding characters when q is or might be convbuf:
 216 ** in this case on buffer-full we want to run iconv before returning.
 217 ** I apologise for the dreadful name.
 218 */
 219 # define ADDCHR2(C) do { *q++ = (C); dstlen--; if (!dstlen) goto iconvbuffull; } while (0)
 220 #else
 221 # define ADDCHR2(C) ADDCHR(C)
 222 #endif
 223
 224                         /* Now decode the text */
 225                         if (quoted_printable) {
 226                                 for (pp = startofmime; pp < endofmime; pp++) {
 227                                         if (*pp == '=') {
 228                                                 c = unqp(pp[1], pp[2]);
 229                                                 if (c == -1)
 230                                                         continue;
 231                                                 if (c != 0)
 232                                                         *q++ = c;
 233                                                 pp += 2;
 234                                         } else if (*pp == '_') {
 235                                                 ADDCHR2(' ');
 236                                         } else {
 237                                                 ADDCHR2(*pp);
 238                                         }
 239                                 }
 240                         } else {
 241                                 /* base64 */
 242                                 int c1, c2, c3, c4;
 243                                 c1 = c2 = c3 = c4 = -1;
 244
 245                                 pp = startofmime;
 246                                 while (pp < endofmime) {
 247                                         /* 6 + 2 bits */
 248                                         while ((pp < endofmime) &&
 249                                                         ((c1 = char64(*pp))
 250                                                         == -1)) {
 251                                                 pp++;
 252                                         }
 253                                         if (pp < endofmime) {
 254                                                 pp++;
 255                                         }
 256                                         while ((pp < endofmime) &&
 257                                                         ((c2 = char64(*pp))
 258                                                         == -1)) {
 259                                                 pp++;
 260                                         }
 261                                         if (pp < endofmime && c1 != -1 && c2 != -1) {
 262                                                 ADDCHR2((c1 << 2) | (c2 >> 4));
 263                                                 pp++;
 264                                         }
 265                                         /* 4 + 4 bits */
 266                                         while ((pp < endofmime) &&
 267                                                         ((c3 = char64(*pp))
 268                                                         == -1)) {
 269                                                 pp++;
 270                                         }
 271                                         if (pp < endofmime && c2 != -1 && c3 != -1) {
 272                                                 ADDCHR2(((c2 & 0xF) << 4) | (c3 >> 2));
 273                                                 pp++;
 274                                         }
 275                                         /* 2 + 6 bits */
 276                                         while ((pp < endofmime) &&
 277                                                         ((c4 = char64(*pp))
 278                                                         == -1)) {
 279                                                 pp++;
 280                                         }
 281                                         if (pp < endofmime && c3 != -1 && c4 != -1) {
 282                                                 ADDCHR2(((c3 & 0x3) << 6) | (c4));
 283                                                 pp++;
 284                                         }
 285                                 }
 286                         }
 287
 288 #ifdef HAVE_ICONV
 289                 iconvbuffull:
 290                         /*
 291                         ** NB that the string at convbuf is not necessarily
 292                         ** NUL terminated here:
 293                         ** q points to the first byte after the valid part.
 294                         */
 295                         /* Convert to native character set */
 296                         if (use_iconv) {
 297                                 size_t inbytes = q - convbuf;
 298                                 ICONV_CONST char *start = convbuf;
 299
 300                                 while (inbytes) {
 301                                         if (iconv(cd, &start, &inbytes, &saveq, &savedstlen) ==
 302                                                 (size_t)-1) {
 303                                                 if (errno != EILSEQ)
 304                                                         break;
 305                                                 /*
 306                                                 ** character couldn't be
 307                                                 ** converted. we output a
 308                                                 ** `?'  and try to carry on
 309                                                 ** which won't work if either
 310                                                 ** encoding was stateful
 311                                                 */
 312                                                 iconv(cd, 0, 0, &saveq, &savedstlen);
 313                                                 if (!savedstlen)
 314                                                         break;
 315                                                 *saveq++ = '?';
 316                                                 savedstlen--;
 317                                                 if (!savedstlen)
 318                                                         break;
 319                                                 /* skip to next input character */
 320                                                 if (fromutf8) {
 321                                                         inbytes--;
 322                                                         start++;
 323                                                         while ((start < q) && ((*start & 192) == 128)) {
 324                                                                 start++;
 325                                                                 inbytes--;
 326                                                         }
 327                                                 } else
 328                                                         start++, inbytes--;
 329                                                 if (start >= q)
 330                                                         break;
 331                                         }
 332                                 }
 333                                 q = saveq;
 334                                 /*
 335                                 ** Stop now if (1) we hit the end of the
 336                                 ** buffer trying to do MIME decoding and
 337                                 ** have just iconv-converted a partial
 338                                 ** string or (2) our iconv-conversion hit
 339                                 ** the end of the buffer.
 340                                 */
 341                                 if (!dstlen || !savedstlen)
 342                                         goto buffull;
 343                                 dstlen = savedstlen;
 344                                 mh_free0(&convbuf);
 345                         }
 346 #endif
 347
 348                         /*
 349                         ** Now that we are done decoding this particular
 350                         ** encoded word, advance string to trailing '='.
 351                         */
 352                         p = endofmime + 1;
 353
 354                         encoding_found = 1;  /* we found (at least 1) encoded word */
 355                         between_encodings = 1;  /* we have just decoded something */
 356                         whitespace = 0;  /* re-initialize amount of whitespace */
 357                 }
 358         }
 359 #ifdef HAVE_ICONV
 360         if (use_iconv) iconv_close(cd);
 361 #endif
 362
 363         /* If an equals was pending at end of string, add it now. */
 364         if (equals_pending)
 365                 ADDCHR('=');
 366         *q = '\0';
 367
 368         return encoding_found;
 369
 370   buffull:
 371         /*
 372         ** q is currently just off the end of the buffer,
 373         ** so rewind to NUL terminate
 374         */
 375         q--;
 376         *q = '\0';
 377         return encoding_found;
 378 }