libcitadel/lib/mime_parser.c

   1 /*
   2  * This is the MIME parser for Citadel.
   3  *
   4  * Copyright (c) 1998-2010 by the citadel.org development team.
   5  * This code is distributed under the GNU General Public License v3.
   6  *
   7  */
   8
   9 #include <stdlib.h>
  10 #include <unistd.h>
  11 #include <stdio.h>
  12 #include <signal.h>
  13 #include <sys/types.h>
  14 #include <ctype.h>
  15 #include <string.h>
  16 #include <sys/stat.h>
  17 #include <sys/types.h>
  18 #include <dirent.h>
  19 #include <errno.h>
  20
  21 #include "xdgmime/xdgmime.h"
  22 #include "libcitadel.h"
  23 #include "libcitadellocal.h"
  24
  25 const unsigned char FromHexTable [256] = {
  26         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //  0
  27         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 10
  28         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 20
  29         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 30
  30         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x00, 0x01, // 40
  31         0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0xFF, 0xFF, // 50
  32         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, // 60
  33         0x0F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 70
  34         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 80
  35         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0A, 0x0B, 0x0C, // 90
  36         0x0D, 0x0E, 0x0F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //100
  37         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //110
  38         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //120
  39         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //130
  40         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //140
  41         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //150
  42         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //160
  43         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //170
  44         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //180
  45         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //190
  46         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //200
  47         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //210
  48         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //220
  49         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //230
  50         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //240
  51         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF                          //250
  52 };
  53
  54
  55 long extract_key(char *target, char *source, long sourcelen, char *key, long keylen, char KeyEnd)
  56 {
  57         char *sptr, *ptr = NULL;
  58         int double_quotes = 0;
  59         long RealKeyLen = keylen;
  60
  61         sptr = source;
  62
  63         while (sptr != NULL)
  64         {
  65                 ptr = bmstrcasestr_len(sptr, sourcelen - (sptr - source),
  66                                        key, keylen);
  67                 if(ptr != NULL)
  68                 {
  69                         while (isspace(*(ptr + RealKeyLen)))
  70                                 RealKeyLen ++;
  71                         if (*(ptr + RealKeyLen) == KeyEnd)
  72                         {
  73                                 sptr = NULL;
  74                                 RealKeyLen ++;
  75                         }
  76                         else
  77                         {
  78                                 sptr = ptr + RealKeyLen + 1;
  79                         }
  80                 }
  81                 else
  82                         sptr = ptr;
  83         }
  84         if (ptr == NULL) {
  85                 *target = '\0';
  86                 return 0;
  87         }
  88         strcpy(target, (ptr + RealKeyLen));
  89
  90         for (ptr=target; (*ptr != 0); ptr++) {
  91
  92                 /* A semicolon means we've hit the end of the key, unless we're inside double quotes */
  93                 if ( (double_quotes != 1) && (*ptr == ';')) {
  94                         *ptr = 0;
  95                 }
  96
  97                 /* if we find double quotes, we've got a great set of string boundaries */
  98                 if (*ptr == '\"') {
  99                         ++double_quotes;
 100                         if (double_quotes == 1) {
 101                                 strcpy(ptr, ptr+1);
 102                         }
 103                         else {
 104                                 *ptr = 0;
 105                         }
 106                 }
 107         }
 108         *ptr = '\0';
 109         return ptr - target;
 110 }
 111
 112
 113 /*
 114  * For non-multipart messages, we need to generate a quickie partnum of "1"
 115  * to return to callback functions.  Some callbacks demand it.
 116  */
 117 char *fixed_partnum(char *supplied_partnum) {
 118         if (supplied_partnum == NULL) return "1";
 119         if (strlen(supplied_partnum)==0) return "1";
 120         return supplied_partnum;
 121 }
 122
 123
 124 static inline unsigned int _decode_hex(const char *Source)
 125 {
 126         int ret = '?';
 127         unsigned char LO_NIBBLE;
 128         unsigned char HI_NIBBLE;
 129
 130         HI_NIBBLE = FromHexTable[(unsigned char) *Source];
 131         LO_NIBBLE = FromHexTable[(unsigned char) *(Source+1)];
 132
 133         if ((LO_NIBBLE == 0xFF) || (LO_NIBBLE == 0xFF))
 134                 return ret;
 135         ret = HI_NIBBLE;
 136         ret = ret << 4;
 137         ret = ret | LO_NIBBLE;
 138         return ret;
 139 }
 140
 141 unsigned int decode_hex(char *Source) {return _decode_hex(Source);}
 142
 143 /*
 144  * Convert "quoted-printable" to binary.  Returns number of bytes decoded.
 145  * according to RFC2045 section 6.7
 146  */
 147 int CtdlDecodeQuotedPrintable(char *decoded, char *encoded, int sourcelen) {
 148         unsigned int ch;
 149         int decoded_length = 0;
 150         int pos = 0;
 151
 152         while (pos < sourcelen)
 153         {
 154                 if (*(encoded + pos) == '=')
 155                 {
 156                         pos ++;
 157                         if (*(encoded + pos) == '\n')
 158                         {
 159                                 pos ++;
 160                         }
 161                         else if (*(encoded + pos) == '\r')
 162                         {
 163                                 pos ++;
 164                                 if (*(encoded + pos) == '\n')
 165                                         pos++;
 166                         }
 167                         else
 168                         {
 169                                 ch = 0;
 170                                 ch = _decode_hex(&encoded[pos]);
 171                                 pos += 2;
 172                                 decoded[decoded_length++] = ch;
 173                         }
 174                 }
 175                 else
 176                 {
 177                         decoded[decoded_length++] = encoded[pos];
 178                         pos += 1;
 179                 }
 180         }
 181         decoded[decoded_length] = 0;
 182         return(decoded_length);
 183 }
 184
 185
 186 /*
 187  * Given a message or message-part body and a length, handle any necessary
 188  * decoding and pass the request up the stack.
 189  */
 190 void mime_decode(char *partnum,
 191                  char *part_start, size_t length,
 192                  char *content_type, char *charset, char *encoding,
 193                  char *disposition,
 194                  char *id,
 195                  char *name, char *filename,
 196                  MimeParserCallBackType CallBack,
 197                  MimeParserCallBackType PreMultiPartCallBack,
 198                  MimeParserCallBackType PostMultiPartCallBack,
 199                  void *userdata,
 200                  int dont_decode)
 201 {
 202
 203         char *decoded;
 204         size_t bytes_decoded = 0;
 205
 206         /* Some encodings aren't really encodings */
 207         if (!strcasecmp(encoding, "7bit"))
 208                 strcpy(encoding, "");
 209         if (!strcasecmp(encoding, "8bit"))
 210                 strcpy(encoding, "");
 211         if (!strcasecmp(encoding, "binary"))
 212                 strcpy(encoding, "");
 213
 214         /* If this part is not encoded, send as-is */
 215         if ( (strlen(encoding) == 0) || (dont_decode)) {
 216                 if (CallBack != NULL) {
 217                         CallBack(name,
 218                                  filename,
 219                                  fixed_partnum(partnum),
 220                                  disposition,
 221                                  part_start,
 222                                  content_type,
 223                                  charset,
 224                                  length,
 225                                  encoding,
 226                                  id,
 227                                  userdata);
 228                         }
 229                 return;
 230         }
 231
 232         /* Fail silently if we hit an unknown encoding. */
 233         if ((strcasecmp(encoding, "base64"))
 234             && (strcasecmp(encoding, "quoted-printable"))) {
 235                 return;
 236         }
 237
 238         /*
 239          * Allocate a buffer for the decoded data.  The output buffer is slightly
 240          * larger than the input buffer; this assumes that the decoded data
 241          * will never be significantly larger than the encoded data.  This is a
 242          * safe assumption with base64, uuencode, and quoted-printable.
 243          */
 244         decoded = malloc(length + 32768);
 245         if (decoded == NULL) {
 246                 return;
 247         }
 248
 249         if (!strcasecmp(encoding, "base64")) {
 250                 bytes_decoded = CtdlDecodeBase64(decoded, part_start, length);
 251         }
 252         else if (!strcasecmp(encoding, "quoted-printable")) {
 253                 bytes_decoded = CtdlDecodeQuotedPrintable(decoded, part_start, length);
 254         }
 255
 256         if (bytes_decoded > 0) if (CallBack != NULL) {
 257                         char encoding_buf[SIZ];
 258
 259                         strcpy(encoding_buf, "binary");
 260                         CallBack(name,
 261                                  filename,
 262                                  fixed_partnum(partnum),
 263                                  disposition,
 264                                  decoded,
 265                                  content_type,
 266                                  charset,
 267                                  bytes_decoded,
 268                                  encoding_buf,
 269                                  id,
 270                                  userdata);
 271         }
 272
 273         free(decoded);
 274 }
 275
 276 /*
 277  * this is the extract of mime_decode which can be called if 'dont_decode' was set;
 278  * to save the cpu intense process of decoding to the time when it realy wants the content.
 279  * returns:
 280  *   - > 0 we decoded something, its on *decoded, you need to free it.
 281  *   - = 0 no need to decode stuff. *decoded will be NULL.
 282  *   - < 0 an error occured, either an unknown encoding, or alloc failed. no need to free.
 283  */
 284 int mime_decode_now (char *part_start,
 285                      size_t length,
 286                      char *encoding,
 287                      char **decoded,
 288                      size_t *bytes_decoded)
 289 {
 290         *bytes_decoded = 0;
 291         *decoded = NULL;
 292         /* Some encodings aren't really encodings */
 293         if (!strcasecmp(encoding, "7bit"))
 294                 strcpy(encoding, "");
 295         if (!strcasecmp(encoding, "8bit"))
 296                 strcpy(encoding, "");
 297         if (!strcasecmp(encoding, "binary"))
 298                 strcpy(encoding, "");
 299
 300         /* If this part is not encoded, send as-is */
 301         if (strlen(encoding) == 0) {
 302                 return 0;
 303         }
 304
 305
 306         /* Fail if we hit an unknown encoding. */
 307         if ((strcasecmp(encoding, "base64"))
 308             && (strcasecmp(encoding, "quoted-printable"))) {
 309                 return -1;
 310         }
 311
 312         /*
 313          * Allocate a buffer for the decoded data.  The output buffer is slightly
 314          * larger than the input buffer; this assumes that the decoded data
 315          * will never be significantly larger than the encoded data.  This is a
 316          * safe assumption with base64, uuencode, and quoted-printable.
 317          */
 318         *decoded = malloc(length + 32768);
 319         if (decoded == NULL) {
 320                 return -1;
 321         }
 322
 323         if (!strcasecmp(encoding, "base64")) {
 324                 *bytes_decoded = CtdlDecodeBase64(*decoded, part_start, length);
 325                 return 1;
 326         }
 327         else if (!strcasecmp(encoding, "quoted-printable")) {
 328                 *bytes_decoded = CtdlDecodeQuotedPrintable(*decoded, part_start, length);
 329                 return 1;
 330         }
 331         return -1;
 332 }
 333
 334 typedef enum _eIntMimeHdrs {
 335         boundary,
 336         startary,
 337         endary,
 338         content_type,
 339         charset,
 340         encoding,
 341         content_type_name,
 342         content_disposition_name,
 343         filename,
 344         disposition,
 345         id,
 346         eMax /* don't move ! */
 347 } eIntMimeHdrs;
 348
 349 typedef struct _CBufStr {
 350         char Key[SIZ];
 351         long len;
 352 }CBufStr;
 353
 354 typedef struct _interesting_mime_headers {
 355         CBufStr b[eMax];
 356         long content_length;
 357         long is_multipart;
 358 } interesting_mime_headers;
 359
 360
 361 static void FlushInterestingMimes(interesting_mime_headers *m)
 362 {
 363         int i;
 364
 365         for (i = 0; i < eMax; i++) {
 366              m->b[i].Key[0] = '\0';
 367              m->b[i].len = 0;
 368         }
 369         m->content_length = -1;
 370 }
 371 static interesting_mime_headers *InitInterestingMimes(void)
 372 {
 373         interesting_mime_headers *m;
 374         m = (interesting_mime_headers*) malloc( sizeof(interesting_mime_headers));
 375
 376         FlushInterestingMimes(m);
 377
 378         return m;
 379 }
 380
 381
 382 static long parse_MimeHeaders(interesting_mime_headers *m,
 383                               char** pcontent_start,
 384                               char *content_end)
 385 {
 386         char buf[SIZ];
 387         char header[SIZ];
 388         long headerlen;
 389         char *ptr, *pch;
 390         int buflen = 0;
 391         int i;
 392
 393         /* Learn interesting things from the headers */
 394         ptr = *pcontent_start;
 395         *header = '\0';
 396         headerlen = 0;
 397         do {
 398                 ptr = memreadlinelen(ptr, buf, SIZ, &buflen);
 399
 400                 for (i = 0; i < buflen; ++i) {
 401                         if (isspace(buf[i])) {
 402                                 buf[i] = ' ';
 403                         }
 404                 }
 405
 406                 if (!isspace(buf[0]) && (headerlen > 0)) {
 407                         if (!strncasecmp(header, "Content-type:", 13)) {
 408                                 memcpy (m->b[content_type].Key, &header[13], headerlen - 12);
 409                                 m->b[content_type].len = striplt (m->b[content_type].Key);
 410
 411                                 m->b[content_type_name].len = extract_key(m->b[content_type_name].Key, CKEY(m->b[content_type]), HKEY("name"), '=');
 412                                 m->b[charset].len           = extract_key(m->b[charset].Key,           CKEY(m->b[content_type]), HKEY("charset"), '=');
 413                                 m->b[boundary].len          = extract_key(m->b[boundary].Key,          header,       headerlen,  HKEY("boundary"), '=');
 414
 415                                 /* Deal with weird headers */
 416                                 pch = strchr(m->b[content_type].Key, ' ');
 417                                 if (pch != NULL) {
 418                                         *pch = '\0';
 419                                         m->b[content_type].len = m->b[content_type].Key - pch;
 420                                 }
 421                                 pch = strchr(m->b[content_type].Key, ';');
 422                                 if (pch != NULL) {
 423                                         *pch = '\0';
 424                                         m->b[content_type].len = m->b[content_type].Key - pch;
 425                                 }
 426                         }
 427                         else if (!strncasecmp(header, "Content-Disposition:", 20)) {
 428                                 memcpy (m->b[disposition].Key, &header[20], headerlen - 19);
 429                                 m->b[disposition].len = striplt(m->b[disposition].Key);
 430
 431                                 m->b[content_disposition_name].len = extract_key(m->b[content_disposition_name].Key, CKEY(m->b[disposition]), HKEY("name"), '=');
 432                                 m->b[filename].len                 = extract_key(m->b[filename].Key,                 CKEY(m->b[disposition]), HKEY("filename"), '=');
 433                                 pch = strchr(m->b[disposition].Key, ';');
 434                                 if (pch != NULL) *pch = '\0';
 435                                 m->b[disposition].len = striplt(m->b[disposition].Key);
 436                         }
 437                         else if (!strncasecmp(header, "Content-ID:", 11)) {
 438                                 memcpy(m->b[id].Key, &header[11], headerlen);
 439                                 striplt(m->b[id].Key);
 440                                 m->b[id].len = stripallbut(m->b[id].Key, '<', '>');
 441                         }
 442                         else if (!strncasecmp(header, "Content-length: ", 15)) {
 443                                 char *clbuf;
 444                                 clbuf = &header[15];
 445                                 while (isspace(*clbuf))
 446                                         clbuf ++;
 447                                 m->content_length = (size_t) atol(clbuf);
 448                         }
 449                         else if (!strncasecmp(header, "Content-transfer-encoding: ", 26)) {
 450                                 memcpy(m->b[encoding].Key, &header[26], headerlen - 26);
 451                                 m->b[encoding].len = striplt(m->b[encoding].Key);
 452                         }
 453                         *header = '\0';
 454                         headerlen = 0;
 455                 }
 456                 if ((headerlen + buflen + 2) < SIZ) {
 457                         memcpy(&header[headerlen], buf, buflen);
 458                         headerlen += buflen;
 459                         header[headerlen] = '\0';
 460                 }
 461                 if (ptr >= content_end) {
 462                         return -1;
 463                 }
 464         } while ((!IsEmptyStr(buf)) && (*ptr != 0));
 465
 466         m->is_multipart = m->b[boundary].len != 0;
 467         *pcontent_start = ptr;
 468
 469         return 0;
 470 }
 471
 472
 473 static int IsAsciiEncoding(interesting_mime_headers *m)
 474 {
 475
 476         if ((m->b[encoding].len != 0) &&
 477             (strcasecmp(m->b[encoding].Key, "base64") == 0))
 478                 return 1;
 479         if ((m->b[encoding].len != 0) &&
 480             (strcmp(m->b[encoding].Key, "quoted-printable") == 0))
 481                 return 1;
 482
 483         return 0;
 484 }
 485
 486 static char *FindNextContent(char *ptr,
 487                              char *content_end,
 488                              interesting_mime_headers *SubMimeHeaders,
 489                              interesting_mime_headers *m)
 490 {
 491         char *next_boundary;
 492         char  tmp;
 493
 494         if (IsAsciiEncoding(SubMimeHeaders)) {
 495                 tmp = *content_end;
 496                 *content_end = '\0';
 497
 498                 /**
 499                  * ok, if we have a content length of the mime part,
 500                  * try skipping the content on the search for the next
 501                  * boundary. since we don't trust the content_length
 502                  * to be all accurate, and suspect it to lose one digit
 503                  * per line with a line length of 80 chars, we need
 504                  * to start searching a little before..
 505                  */
 506
 507                 if ((SubMimeHeaders->content_length != -1) &&
 508                     (SubMimeHeaders->content_length > 10))
 509                 {
 510                         char *pptr;
 511                         long lines;
 512
 513                         lines = SubMimeHeaders->content_length / 80;
 514                         pptr = ptr + SubMimeHeaders->content_length - lines - 10;
 515                         if (pptr < content_end)
 516                                 ptr = pptr;
 517                 }
 518
 519                 next_boundary = strstr(ptr, m->b[startary].Key);
 520                 *content_end = tmp;
 521         }
 522         else {
 523                 char *srch;
 524                 /**
 525                  * ok, if we have a content length of the mime part,
 526                  * try skipping the content on the search for the next
 527                  * boundary. since we don't trust the content_length
 528                  * to be all accurate, start searching a little before..
 529                  */
 530
 531                 if ((SubMimeHeaders->content_length != -1) &&
 532                     (SubMimeHeaders->content_length > 10))
 533                 {
 534                         char *pptr;
 535                         pptr = ptr + SubMimeHeaders->content_length - 10;
 536                         if (pptr < content_end)
 537                                 ptr = pptr;
 538                 }
 539
 540
 541                 srch = next_boundary = NULL;
 542                 for (srch = memchr(ptr, '-',  content_end - srch);
 543                      (srch != NULL) && (srch < content_end);
 544                      srch = memchr(srch, '-',  content_end - srch))
 545                 {
 546                         if (!memcmp(srch,
 547                                     m->b[startary].Key,
 548                                     m->b[startary].len))
 549                         {
 550                                 next_boundary = srch;
 551                                 srch = content_end;
 552                         }
 553                         else srch ++;
 554
 555                 }
 556
 557         }
 558         return next_boundary;
 559 }
 560
 561 /*
 562  * Break out the components of a multipart message
 563  * (This function expects to be fed HEADERS + CONTENT)
 564  * Note: NULL can be supplied as content_end; in this case, the message is
 565  * considered to have ended when the parser encounters a 0x00 byte.
 566  */
 567 static void recurseable_mime_parser(char *partnum,
 568                                     char *content_start, char *content_end,
 569                                     MimeParserCallBackType CallBack,
 570                                     MimeParserCallBackType PreMultiPartCallBack,
 571                                     MimeParserCallBackType PostMultiPartCallBack,
 572                                     void *userdata,
 573                                     int dont_decode,
 574                                     interesting_mime_headers *m)
 575 {
 576         interesting_mime_headers *SubMimeHeaders;
 577         char     *ptr;
 578         char     *part_start;
 579         char     *part_end = NULL;
 580         char     *evaluate_crlf_ptr = NULL;
 581         char     *next_boundary;
 582         char      nested_partnum[256];
 583         int       crlf_in_use = 0;
 584         int       part_seq = 0;
 585         CBufStr  *chosen_name;
 586
 587
 588         /* If this is a multipart message, then recursively process it */
 589         ptr = content_start;
 590         part_start = NULL;
 591         if (m->is_multipart) {
 592
 593                 /* Tell the client about this message's multipartedness */
 594                 if (PreMultiPartCallBack != NULL) {
 595                         PreMultiPartCallBack("",
 596                                              "",
 597                                              partnum,
 598                                              "",
 599                                              NULL,
 600                                              m->b[content_type].Key,
 601                                              m->b[charset].Key,
 602                                              0,
 603                                              m->b[encoding].Key,
 604                                              m->b[id].Key,
 605                                              userdata);
 606                 }
 607
 608                 /* Figure out where the boundaries are */
 609                 m->b[startary].len = snprintf(m->b[startary].Key, SIZ, "--%s", m->b[boundary].Key);
 610                 SubMimeHeaders = InitInterestingMimes ();
 611                 if (*ptr == '\r')
 612                         ptr ++;
 613                 if (*ptr == '\n')
 614                         ptr ++;
 615                 if (strncmp(ptr, m->b[startary].Key, m->b[startary].len) == 0)
 616                         ptr += m->b[startary].len;
 617                 if (*ptr == '\r')
 618                         ptr ++;
 619                 if (*ptr == '\n')
 620                         ptr ++;
 621                 part_start = NULL;
 622                 do {
 623                         char *optr;
 624
 625                         optr = ptr;
 626                         if (parse_MimeHeaders(SubMimeHeaders, &ptr, content_end) != 0)
 627                                 break;
 628                         if ((ptr - optr > 2) &&
 629                             (*(ptr - 2) == '\r'))
 630                                 crlf_in_use = 1;
 631
 632                         part_start = ptr;
 633
 634                         next_boundary = FindNextContent(ptr,
 635                                                         content_end,
 636                                                         SubMimeHeaders,
 637                                                         m);
 638                         if ((next_boundary != NULL) &&
 639                             (next_boundary - part_start < 3))
 640                                 continue;
 641
 642                         if ( (part_start != NULL) && (next_boundary != NULL) ) {
 643                                 part_end = next_boundary;
 644                                 --part_end;             /* omit the trailing LF */
 645                                 if (crlf_in_use) {
 646                                         --part_end;     /* omit the trailing CR */
 647                                 }
 648
 649                                 if (!IsEmptyStr(partnum)) {
 650                                         snprintf(nested_partnum,
 651                                                  sizeof nested_partnum,
 652                                                  "%s.%d", partnum,
 653                                                  ++part_seq);
 654                                 }
 655                                 else {
 656                                         snprintf(nested_partnum,
 657                                                  sizeof nested_partnum,
 658                                                  "%d", ++part_seq);
 659                                 }
 660                                 recurseable_mime_parser(nested_partnum,
 661                                                         part_start,
 662                                                         part_end,
 663                                                         CallBack,
 664                                                         PreMultiPartCallBack,
 665                                                         PostMultiPartCallBack,
 666                                                         userdata,
 667                                                         dont_decode,
 668                                                         SubMimeHeaders);
 669                         }
 670
 671                         if (next_boundary != NULL) {
 672                                 /* If we pass out of scope, don't attempt to
 673                                  * read past the end boundary. */
 674                                 if ((*(next_boundary + m->b[startary].len + 1) == '-') &&
 675                                     (*(next_boundary + m->b[startary].len + 2) == '-') ){
 676                                         ptr = content_end;
 677                                 }
 678                                 else {
 679                                         /* Set up for the next part. */
 680                                         part_start = strstr(next_boundary, "\n");
 681
 682                                         /* Determine whether newlines are LF or CRLF */
 683                                         evaluate_crlf_ptr = part_start;
 684                                         --evaluate_crlf_ptr;
 685                                         if ((*evaluate_crlf_ptr == '\r') &&
 686                                             (*(evaluate_crlf_ptr + 1) == '\n'))
 687                                         {
 688                                                 crlf_in_use = 1;
 689                                         }
 690                                         else {
 691                                                 crlf_in_use = 0;
 692                                         }
 693
 694                                         /* Advance past the LF ... now we're in the next part */
 695                                         ++part_start;
 696                                         ptr = part_start;
 697                                 }
 698                         }
 699                         else {
 700                                 /* Invalid end of multipart.  Bail out! */
 701                                 ptr = content_end;
 702                         }
 703                         FlushInterestingMimes(SubMimeHeaders);
 704                 } while ( (ptr < content_end) && (next_boundary != NULL) );
 705
 706                 free(SubMimeHeaders);
 707
 708                 if (PostMultiPartCallBack != NULL) {
 709                         PostMultiPartCallBack("",
 710                                               "",
 711                                               partnum,
 712                                               "",
 713                                               NULL,
 714                                               m->b[content_type].Key,
 715                                               m->b[charset].Key,
 716                                               0,
 717                                               m->b[encoding].Key,
 718                                               m->b[id].Key,
 719                                               userdata);
 720                 }
 721         } /* If it's not a multipart message, then do something with it */
 722         else {
 723                 size_t length;
 724                 part_start = ptr;
 725                 length = content_end - part_start;
 726                 ptr = part_end = content_end;
 727
 728
 729                 /* The following code will truncate the MIME part to the size
 730                  * specified by the Content-length: header.   We have commented it
 731                  * out because these headers have a tendency to be wrong.
 732                  *
 733                  *      if ( (content_length > 0) && (length > content_length) ) {
 734                  *              length = content_length;
 735                  *      }
 736                  */
 737
 738                 /* Sometimes the "name" field is tacked on to Content-type,
 739                  * and sometimes it's tacked on to Content-disposition.  Use
 740                  * whichever one we have.
 741                  */
 742                 if (m->b[content_disposition_name].len > m->b[content_type_name].len) {
 743                         chosen_name = &m->b[content_disposition_name];
 744                 }
 745                 else {
 746                         chosen_name = &m->b[content_type_name];
 747                 }
 748
 749                 /* Ok, we've got a non-multipart part here, so do something with it.
 750                  */
 751                 mime_decode(partnum,
 752                             part_start,
 753                             length,
 754                             m->b[content_type].Key,
 755                             m->b[charset].Key,
 756                             m->b[encoding].Key,
 757                             m->b[disposition].Key,
 758                             m->b[id].Key,
 759                             chosen_name->Key,
 760                             m->b[filename].Key,
 761                             CallBack,
 762                             NULL, NULL,
 763                             userdata,
 764                             dont_decode
 765                         );
 766
 767                 /*
 768                  * Now if it's an encapsulated message/rfc822 then we have to recurse into it
 769                  */
 770                 if (!strcasecmp(&m->b[content_type].Key[0], "message/rfc822")) {
 771
 772                         if (PreMultiPartCallBack != NULL) {
 773                                 PreMultiPartCallBack("",
 774                                                      "",
 775                                                      partnum,
 776                                                      "",
 777                                                      NULL,
 778                                                      m->b[content_type].Key,
 779                                                      m->b[charset].Key,
 780                                                      0,
 781                                                      m->b[encoding].Key,
 782                                                      m->b[id].Key,
 783                                                      userdata);
 784                         }
 785                         if (CallBack != NULL) {
 786                                 if (strlen(partnum) > 0) {
 787                                         snprintf(nested_partnum,
 788                                                  sizeof nested_partnum,
 789                                                  "%s.%d", partnum,
 790                                                  ++part_seq);
 791                                 }
 792                                 else {
 793                                         snprintf(nested_partnum,
 794                                                  sizeof nested_partnum,
 795                                                  "%d", ++part_seq);
 796                                 }
 797                                 the_mime_parser(nested_partnum,
 798                                                 part_start,
 799                                                 part_end,
 800                                                 CallBack,
 801                                                 PreMultiPartCallBack,
 802                                                 PostMultiPartCallBack,
 803                                                 userdata,
 804                                                 dont_decode
 805                                         );
 806                         }
 807                         if (PostMultiPartCallBack != NULL) {
 808                                 PostMultiPartCallBack("",
 809                                                       "",
 810                                                       partnum,
 811                                                       "",
 812                                                       NULL,
 813                                                       m->b[content_type].Key,
 814                                                       m->b[charset].Key,
 815                                                       0,
 816                                                       m->b[encoding].Key,
 817                                                       m->b[id].Key,
 818                                                       userdata);
 819                         }
 820
 821
 822                 }
 823
 824         }
 825
 826 }
 827
 828 /*
 829  * Break out the components of a multipart message
 830  * (This function expects to be fed HEADERS + CONTENT)
 831  * Note: NULL can be supplied as content_end; in this case, the message is
 832  * considered to have ended when the parser encounters a 0x00 byte.
 833  */
 834 void the_mime_parser(char *partnum,
 835                      char *content_start, char *content_end,
 836                      MimeParserCallBackType CallBack,
 837                      MimeParserCallBackType PreMultiPartCallBack,
 838                      MimeParserCallBackType PostMultiPartCallBack,
 839                      void *userdata,
 840                      int dont_decode)
 841 {
 842         interesting_mime_headers *m;
 843
 844         /* If the caller didn't supply an endpointer, generate one by measure */
 845         if (content_end == NULL) {
 846                 content_end = &content_start[strlen(content_start)];
 847         }
 848
 849         m = InitInterestingMimes();
 850
 851         if (!parse_MimeHeaders(m, &content_start, content_end))
 852         {
 853
 854                 recurseable_mime_parser(partnum,
 855                                         content_start, content_end,
 856                                         CallBack,
 857                                         PreMultiPartCallBack,
 858                                         PostMultiPartCallBack,
 859                                         userdata,
 860                                         dont_decode,
 861                                         m);
 862         }
 863         free(m);
 864 }
 865
 866 /*
 867  * Entry point for the MIME parser.
 868  * (This function expects to be fed HEADERS + CONTENT)
 869  * Note: NULL can be supplied as content_end; in this case, the message is
 870  * considered to have ended when the parser encounters a 0x00 byte.
 871  */
 872 void mime_parser(char *content_start,
 873                  char *content_end,
 874                  MimeParserCallBackType CallBack,
 875                  MimeParserCallBackType PreMultiPartCallBack,
 876                  MimeParserCallBackType PostMultiPartCallBack,
 877                  void *userdata,
 878                  int dont_decode)
 879 {
 880
 881         the_mime_parser("", content_start, content_end,
 882                         CallBack,
 883                         PreMultiPartCallBack,
 884                         PostMultiPartCallBack,
 885                         userdata, dont_decode);
 886 }
 887
 888
 889
 890
 891
 892
 893 typedef struct _MimeGuess {
 894         const char *Pattern;
 895         size_t PatternLen;
 896         long PatternOffset;
 897         const char *MimeString;
 898 } MimeGuess;
 899
 900 MimeGuess MyMimes [] = {
 901         {
 902                 "GIF",
 903                 3,
 904                 0,
 905                 "image/gif"
 906         },
 907         {
 908                 "\xff\xd8",
 909                 2,
 910                 0,
 911                 "image/jpeg"
 912         },
 913         {
 914                 "\x89PNG",
 915                 4,
 916                 0,
 917                 "image/png"
 918         },
 919         { // last...
 920                 "",
 921                 0,
 922                 0,
 923                 ""
 924         }
 925 };
 926
 927
 928 const char *GuessMimeType(const char *data, size_t dlen)
 929 {
 930         int MimeIndex = 0;
 931
 932         while (MyMimes[MimeIndex].PatternLen != 0)
 933         {
 934                 if ((MyMimes[MimeIndex].PatternLen +
 935                      MyMimes[MimeIndex].PatternOffset < dlen) &&
 936                     strncmp(MyMimes[MimeIndex].Pattern,
 937                             &data[MyMimes[MimeIndex].PatternOffset],
 938                             MyMimes[MimeIndex].PatternLen) == 0)
 939                 {
 940                         return MyMimes[MimeIndex].MimeString;
 941                 }
 942                 MimeIndex ++;
 943         }
 944         /*
 945          * ok, our simple minded algorythm didn't find anything,
 946          * let the big chegger try it, he wil default to application/octet-stream
 947          */
 948         return (xdg_mime_get_mime_type_for_data(data, dlen));
 949 }
 950
 951
 952 const char* GuessMimeByFilename(const char *what, size_t len)
 953 {
 954         /* we know some hardcoded on our own, try them... */
 955         if ((len > 3) && !strncasecmp(&what[len - 4], ".gif", 4))
 956                 return "image/gif";
 957         else if ((len > 2) && !strncasecmp(&what[len - 3], ".js", 3))
 958                 return  "text/javascript";
 959         else if ((len > 3) && !strncasecmp(&what[len - 4], ".txt", 4))
 960                 return "text/plain";
 961         else if ((len > 3) && !strncasecmp(&what[len - 4], ".css", 4))
 962                 return "text/css";
 963         else if ((len > 3) && !strncasecmp(&what[len - 4], ".htc", 4))
 964                 return "text/x-component";
 965         else if ((len > 3) && !strncasecmp(&what[len - 4], ".jpg", 4))
 966                 return "image/jpeg";
 967         else if ((len > 3) && !strncasecmp(&what[len - 4], ".png", 4))
 968                 return "image/png";
 969         else if ((len > 3) && !strncasecmp(&what[len - 4], ".ico", 4))
 970                 return "image/x-icon";
 971         else if ((len > 3) && !strncasecmp(&what[len - 4], ".vcf", 4))
 972                 return "text/x-vcard";
 973         else if ((len > 4) && !strncasecmp(&what[len - 5], ".html", 5))
 974                 return "text/html";
 975         else if ((len > 3) && !strncasecmp(&what[len - 4], ".htm", 4))
 976                 return "text/html";
 977         else if ((len > 3) && !strncasecmp(&what[len - 4], ".wml", 4))
 978                 return "text/vnd.wap.wml";
 979         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wmls", 5))
 980                 return "text/vnd.wap.wmlscript";
 981         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wmlc", 5))
 982                 return "application/vnd.wap.wmlc";
 983         else if ((len > 5) && !strncasecmp(&what[len - 6], ".wmlsc", 6))
 984                 return "application/vnd.wap.wmlscriptc";
 985         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wbmp", 5))
 986                 return "image/vnd.wap.wbmp";
 987         else
 988                 /* and let xdgmime do the fallback. */
 989                 return xdg_mime_get_mime_type_from_file_name(what);
 990 }
 991
 992 static HashList *IconHash = NULL;
 993
 994 typedef struct IconName IconName;
 995
 996 struct IconName {
 997         char *FlatName;
 998         char *FileName;
 999 };
1000
1001 static void DeleteIcon(void *IconNamePtr)
1002 {
1003         IconName *Icon = (IconName*) IconNamePtr;
1004         free(Icon->FlatName);
1005         free(Icon->FileName);
1006         free(Icon);
1007 }
1008
1009 /*
1010 static const char *PrintFlat(void *IconNamePtr)
1011 {
1012         IconName *Icon = (IconName*) IconNamePtr;
1013         return Icon->FlatName;
1014 }
1015 static const char *PrintFile(void *IconNamePtr)
1016 {
1017         IconName *Icon = (IconName*) IconNamePtr;
1018         return Icon->FileName;
1019 }
1020 */
1021
1022 #define GENSTR "x-generic"
1023 #define IGNORE_PREFIX_1 "gnome-mime"
1024 int LoadIconDir(const char *DirName)
1025 {
1026         DIR *filedir = NULL;
1027         struct dirent *filedir_entry;
1028         int d_namelen;
1029         int d_without_ext;
1030         IconName *Icon;
1031
1032         filedir = opendir (DirName);
1033         IconHash = NewHash(1, NULL);
1034         if (filedir == NULL) {
1035                 return 0;
1036         }
1037
1038         while ((filedir_entry = readdir(filedir)))
1039         {
1040                 char *MinorPtr;
1041                 char *PStart;
1042 #ifdef _DIRENT_HAVE_D_NAMELEN
1043                 d_namelen = filedir_entry->d_namelen;
1044 #else
1045                 d_namelen = strlen(filedir_entry->d_name);
1046 #endif
1047                 d_without_ext = d_namelen;
1048                 while ((d_without_ext > 0) && (filedir_entry->d_name[d_without_ext] != '.'))
1049                         d_without_ext --;
1050                 if ((d_without_ext == 0) || (d_namelen < 3))
1051                         continue;
1052
1053                 if ((sizeof(IGNORE_PREFIX_1) < d_namelen) &&
1054                     (strncmp(IGNORE_PREFIX_1,
1055                              filedir_entry->d_name,
1056                              sizeof(IGNORE_PREFIX_1) - 1) == 0)) {
1057                         PStart = filedir_entry->d_name + sizeof(IGNORE_PREFIX_1);
1058                         d_without_ext -= sizeof(IGNORE_PREFIX_1);
1059                 }
1060                 else {
1061                         PStart = filedir_entry->d_name;
1062                 }
1063                 Icon = malloc(sizeof(IconName));
1064
1065                 Icon->FileName = malloc(d_namelen + 1);
1066                 memcpy(Icon->FileName, filedir_entry->d_name, d_namelen + 1);
1067
1068                 Icon->FlatName = malloc(d_without_ext + 1);
1069                 memcpy(Icon->FlatName, PStart, d_without_ext);
1070                 Icon->FlatName[d_without_ext] = '\0';
1071                 /* Try to find Minor type in image-jpeg */
1072                 MinorPtr = strchr(Icon->FlatName, '-');
1073                 if (MinorPtr != NULL) {
1074                         size_t MinorLen;
1075                         MinorLen = 1 + d_without_ext - (MinorPtr - Icon->FlatName + 1);
1076                         if ((MinorLen == sizeof(GENSTR)) &&
1077                             (strncmp(MinorPtr + 1, GENSTR, sizeof(GENSTR)) == 0)) {
1078                                 /* ok, we found a generic filename. cut the generic. */
1079                                 *MinorPtr = '\0';
1080                                 d_without_ext = d_without_ext - (MinorPtr - Icon->FlatName);
1081                         }
1082                         else { /* Map the major / minor separator to / */
1083                                 *MinorPtr = '/';
1084                         }
1085                 }
1086
1087 //              PrintHash(IconHash, PrintFlat, PrintFile);
1088 //              printf("%s - %s\n", Icon->FlatName, Icon->FileName);
1089                 Put(IconHash, Icon->FlatName, d_without_ext, Icon, DeleteIcon);
1090 //              PrintHash(IconHash, PrintFlat, PrintFile);
1091         }
1092         closedir(filedir);
1093         return 1;
1094 }
1095
1096 const char *GetIconFilename(char *MimeType, size_t len)
1097 {
1098         void *vIcon;
1099         IconName *Icon;
1100
1101         if(IconHash == NULL)
1102                 return NULL;
1103
1104         GetHash(IconHash, MimeType, len, &vIcon), Icon = (IconName*) vIcon;
1105         /* didn't find the exact mimetype? try major only. */
1106         if (Icon == NULL) {
1107                 char * pMinor;
1108                 pMinor = strchr(MimeType, '/');
1109                 if (pMinor != NULL) {
1110                         *pMinor = '\0';
1111                         GetHash(IconHash, MimeType, pMinor - MimeType, &vIcon),
1112                                 Icon = (IconName*) vIcon;
1113                 }
1114         }
1115         if (Icon == NULL) {
1116                 return NULL;
1117         }
1118
1119         /*printf("Getting: [%s] == [%s] -> [%s]\n", MimeType, Icon->FlatName, Icon->FileName);*/
1120         return Icon->FileName;
1121 }
1122
1123 void ShutDownLibCitadelMime(void)
1124 {
1125         DeleteHash(&IconHash);
1126 }