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                 CallBack(name, filename, fixed_partnum(partnum),
 258                         disposition, decoded,
 259                         content_type, charset, bytes_decoded, "binary", id, userdata);
 260         }
 261
 262         free(decoded);
 263 }
 264
 265 /*
 266  * this is the extract of mime_decode which can be called if 'dont_decode' was set;
 267  * to save the cpu intense process of decoding to the time when it realy wants the content.
 268  * returns:
 269  *   - > 0 we decoded something, its on *decoded, you need to free it.
 270  *   - = 0 no need to decode stuff. *decoded will be NULL.
 271  *   - < 0 an error occured, either an unknown encoding, or alloc failed. no need to free.
 272  */
 273 int mime_decode_now (char *part_start,
 274                      size_t length,
 275                      char *encoding,
 276                      char **decoded,
 277                      size_t *bytes_decoded)
 278 {
 279         *bytes_decoded = 0;
 280         *decoded = NULL;
 281         /* Some encodings aren't really encodings */
 282         if (!strcasecmp(encoding, "7bit"))
 283                 strcpy(encoding, "");
 284         if (!strcasecmp(encoding, "8bit"))
 285                 strcpy(encoding, "");
 286         if (!strcasecmp(encoding, "binary"))
 287                 strcpy(encoding, "");
 288
 289         /* If this part is not encoded, send as-is */
 290         if (strlen(encoding) == 0) {
 291                 return 0;
 292         }
 293
 294
 295         /* Fail if we hit an unknown encoding. */
 296         if ((strcasecmp(encoding, "base64"))
 297             && (strcasecmp(encoding, "quoted-printable"))) {
 298                 return -1;
 299         }
 300
 301         /*
 302          * Allocate a buffer for the decoded data.  The output buffer is slightly
 303          * larger than the input buffer; this assumes that the decoded data
 304          * will never be significantly larger than the encoded data.  This is a
 305          * safe assumption with base64, uuencode, and quoted-printable.
 306          */
 307         *decoded = malloc(length + 32768);
 308         if (decoded == NULL) {
 309                 return -1;
 310         }
 311
 312         if (!strcasecmp(encoding, "base64")) {
 313                 *bytes_decoded = CtdlDecodeBase64(*decoded, part_start, length);
 314                 return 1;
 315         }
 316         else if (!strcasecmp(encoding, "quoted-printable")) {
 317                 *bytes_decoded = CtdlDecodeQuotedPrintable(*decoded, part_start, length);
 318                 return 1;
 319         }
 320         return -1;
 321 }
 322
 323 typedef enum _eIntMimeHdrs {
 324         boundary,
 325         startary,
 326         endary,
 327         content_type,
 328         charset,
 329         encoding,
 330         content_type_name,
 331         content_disposition_name,
 332         filename,
 333         disposition,
 334         id,
 335         eMax /* don't move ! */
 336 } eIntMimeHdrs;
 337
 338 typedef struct _CBufStr {
 339         char Key[SIZ];
 340         long len;
 341 }CBufStr;
 342
 343 typedef struct _interesting_mime_headers {
 344         CBufStr b[eMax];
 345         long content_length;
 346         long is_multipart;
 347 } interesting_mime_headers;
 348
 349
 350 static void FlushInterestingMimes(interesting_mime_headers *m)
 351 {
 352         int i;
 353
 354         for (i = 0; i < eMax; i++) {
 355              m->b[i].Key[0] = '\0';
 356              m->b[i].len = 0;
 357         }
 358         m->content_length = -1;
 359 }
 360 static interesting_mime_headers *InitInterestingMimes(void)
 361 {
 362         interesting_mime_headers *m;
 363         m = (interesting_mime_headers*) malloc( sizeof(interesting_mime_headers));
 364
 365         FlushInterestingMimes(m);
 366
 367         return m;
 368 }
 369
 370
 371 static long parse_MimeHeaders(interesting_mime_headers *m,
 372                               char** pcontent_start,
 373                               char *content_end)
 374 {
 375         char buf[SIZ];
 376         char header[SIZ];
 377         long headerlen;
 378         char *ptr, *pch;
 379         int buflen = 0;
 380         int i;
 381
 382         /* Learn interesting things from the headers */
 383         ptr = *pcontent_start;
 384         *header = '\0';
 385         headerlen = 0;
 386         do {
 387                 ptr = memreadlinelen(ptr, buf, SIZ, &buflen);
 388
 389                 for (i = 0; i < buflen; ++i) {
 390                         if (isspace(buf[i])) {
 391                                 buf[i] = ' ';
 392                         }
 393                 }
 394
 395                 if (!isspace(buf[0]) && (headerlen > 0)) {
 396                         if (!strncasecmp(header, "Content-type:", 13)) {
 397                                 memcpy (m->b[content_type].Key, &header[13], headerlen - 12);
 398                                 m->b[content_type].len = striplt (m->b[content_type].Key);
 399
 400                                 m->b[content_type_name].len = extract_key(m->b[content_type_name].Key, CKEY(m->b[content_type]), HKEY("name"), '=');
 401                                 m->b[charset].len           = extract_key(m->b[charset].Key,           CKEY(m->b[content_type]), HKEY("charset"), '=');
 402                                 m->b[boundary].len          = extract_key(m->b[boundary].Key,          header,       headerlen,  HKEY("boundary"), '=');
 403
 404                                 /* Deal with weird headers */
 405                                 pch = strchr(m->b[content_type].Key, ' ');
 406                                 if (pch != NULL) {
 407                                         *pch = '\0';
 408                                         m->b[content_type].len = m->b[content_type].Key - pch;
 409                                 }
 410                                 pch = strchr(m->b[content_type].Key, ';');
 411                                 if (pch != NULL) {
 412                                         *pch = '\0';
 413                                         m->b[content_type].len = m->b[content_type].Key - pch;
 414                                 }
 415                         }
 416                         else if (!strncasecmp(header, "Content-Disposition:", 20)) {
 417                                 memcpy (m->b[disposition].Key, &header[20], headerlen - 19);
 418                                 m->b[disposition].len = striplt(m->b[disposition].Key);
 419
 420                                 m->b[content_disposition_name].len = extract_key(m->b[content_disposition_name].Key, CKEY(m->b[disposition]), HKEY("name"), '=');
 421                                 m->b[filename].len                 = extract_key(m->b[filename].Key,                 CKEY(m->b[disposition]), HKEY("filename"), '=');
 422                                 pch = strchr(m->b[disposition].Key, ';');
 423                                 if (pch != NULL) *pch = '\0';
 424                                 m->b[disposition].len = striplt(m->b[disposition].Key);
 425                         }
 426                         else if (!strncasecmp(header, "Content-ID:", 11)) {
 427                                 memcpy(m->b[id].Key, &header[11], headerlen);
 428                                 striplt(m->b[id].Key);
 429                                 m->b[id].len = stripallbut(m->b[id].Key, '<', '>');
 430                         }
 431                         else if (!strncasecmp(header, "Content-length: ", 15)) {
 432                                 char *clbuf;
 433                                 clbuf = &header[15];
 434                                 while (isspace(*clbuf))
 435                                         clbuf ++;
 436                                 m->content_length = (size_t) atol(clbuf);
 437                         }
 438                         else if (!strncasecmp(header, "Content-transfer-encoding: ", 26)) {
 439                                 memcpy(m->b[encoding].Key, &header[26], headerlen - 26);
 440                                 m->b[encoding].len = striplt(m->b[encoding].Key);
 441                         }
 442                         *header = '\0';
 443                         headerlen = 0;
 444                 }
 445                 if ((headerlen + buflen + 2) < SIZ) {
 446                         memcpy(&header[headerlen], buf, buflen);
 447                         headerlen += buflen;
 448                         header[headerlen] = '\0';
 449                 }
 450                 if (ptr >= content_end) {
 451                         return -1;
 452                 }
 453         } while ((!IsEmptyStr(buf)) && (*ptr != 0));
 454
 455         m->is_multipart = m->b[boundary].len != 0;
 456         *pcontent_start = ptr;
 457
 458         return 0;
 459 }
 460
 461
 462 static int IsAsciiEncoding(interesting_mime_headers *m)
 463 {
 464         if ((m->b[encoding].len != 0) &&
 465             (strcmp(m->b[encoding].Key, "binary") == 0))
 466                 return 0;
 467         else
 468                 return 1;
 469 }
 470
 471 static char *FindNextContent(char *ptr,
 472                              char *content_end,
 473                              interesting_mime_headers *SubMimeHeaders,
 474                              interesting_mime_headers *m)
 475 {
 476         char *next_boundary;
 477         char  tmp;
 478
 479         if (IsAsciiEncoding(SubMimeHeaders)) {
 480                 tmp = *content_end;
 481                 *content_end = '\0';
 482
 483                 /**
 484                  * ok, if we have a content length of the mime part,
 485                  * try skipping the content on the search for the next
 486                  * boundary. since we don't trust the content_length
 487                  * to be all accurate, and suspect it to lose one digit
 488                  * per line with a line length of 80 chars, we need
 489                  * to start searching a little before..
 490                  */
 491
 492                 if ((SubMimeHeaders->content_length != -1) &&
 493                     (SubMimeHeaders->content_length > 10))
 494                 {
 495                         char *pptr;
 496                         long lines;
 497
 498                         lines = SubMimeHeaders->content_length / 80;
 499                         pptr = ptr + SubMimeHeaders->content_length - lines - 10;
 500                         if (pptr < content_end)
 501                                 ptr = pptr;
 502                 }
 503
 504                 next_boundary = strstr(ptr, m->b[startary].Key);
 505                 *content_end = tmp;
 506         }
 507         else {
 508                 char *srch;
 509                 /**
 510                  * ok, if we have a content length of the mime part,
 511                  * try skipping the content on the search for the next
 512                  * boundary. since we don't trust the content_length
 513                  * to be all accurate, start searching a little before..
 514                  */
 515
 516                 if ((SubMimeHeaders->content_length != -1) &&
 517                     (SubMimeHeaders->content_length > 10))
 518                 {
 519                         char *pptr;
 520                         pptr = ptr + SubMimeHeaders->content_length - 10;
 521                         if (pptr < content_end)
 522                                 ptr = pptr;
 523                 }
 524                 next_boundary = NULL;
 525                 for (srch=ptr; srch<content_end; ++srch) {
 526                         if (!memcmp(srch,
 527                                     m->b[startary].Key,
 528                                     m->b[startary].len))
 529                         {
 530                                 next_boundary = srch;
 531                                 srch = content_end;
 532                         }
 533                 }
 534
 535         }
 536         return next_boundary;
 537 }
 538
 539 /*
 540  * Break out the components of a multipart message
 541  * (This function expects to be fed HEADERS + CONTENT)
 542  * Note: NULL can be supplied as content_end; in this case, the message is
 543  * considered to have ended when the parser encounters a 0x00 byte.
 544  */
 545 static void recurseable_mime_parser(char *partnum,
 546                                     char *content_start, char *content_end,
 547                                     MimeParserCallBackType CallBack,
 548                                     MimeParserCallBackType PreMultiPartCallBack,
 549                                     MimeParserCallBackType PostMultiPartCallBack,
 550                                     void *userdata,
 551                                     int dont_decode,
 552                                     interesting_mime_headers *m)
 553 {
 554         interesting_mime_headers *SubMimeHeaders;
 555         char     *ptr;
 556         char     *part_start;
 557         char     *part_end = NULL;
 558         char     *evaluate_crlf_ptr = NULL;
 559         char     *next_boundary;
 560         char      nested_partnum[256];
 561         int       crlf_in_use = 0;
 562         int       part_seq = 0;
 563         CBufStr  *chosen_name;
 564
 565
 566         /* If this is a multipart message, then recursively process it */
 567         ptr = content_start;
 568         part_start = NULL;
 569         if (m->is_multipart) {
 570
 571                 /* Tell the client about this message's multipartedness */
 572                 if (PreMultiPartCallBack != NULL) {
 573                         PreMultiPartCallBack("",
 574                                              "",
 575                                              partnum,
 576                                              "",
 577                                              NULL,
 578                                              m->b[content_type].Key,
 579                                              m->b[charset].Key,
 580                                              0,
 581                                              m->b[encoding].Key,
 582                                              m->b[id].Key,
 583                                              userdata);
 584                 }
 585
 586                 /* Figure out where the boundaries are */
 587                 m->b[startary].len = snprintf(m->b[startary].Key, SIZ, "--%s", m->b[boundary].Key);
 588                 SubMimeHeaders = InitInterestingMimes ();
 589                 if (*ptr == '\r')
 590                         ptr ++;
 591                 if (*ptr == '\n')
 592                         ptr ++;
 593                 if (strncmp(ptr, m->b[startary].Key, m->b[startary].len) == 0)
 594                         ptr += m->b[startary].len;
 595                 if (*ptr == '\r')
 596                         ptr ++;
 597                 if (*ptr == '\n')
 598                         ptr ++;
 599                 part_start = NULL;
 600                 do {
 601
 602                         if (parse_MimeHeaders(SubMimeHeaders, &ptr, content_end) != 0)
 603                                 break;
 604                         part_start = ptr;
 605
 606                         next_boundary = FindNextContent(ptr,
 607                                                         content_end,
 608                                                         SubMimeHeaders,
 609                                                         m);
 610                         if ((next_boundary != NULL) &&
 611                             (next_boundary - part_start < 3))
 612                                 continue;
 613
 614                         if ( (part_start != NULL) && (next_boundary != NULL) ) {
 615                                 part_end = next_boundary;
 616                                 --part_end;             /* omit the trailing LF */
 617                                 if (crlf_in_use) {
 618                                         --part_end;     /* omit the trailing CR */
 619                                 }
 620
 621                                 if (!IsEmptyStr(partnum)) {
 622                                         snprintf(nested_partnum,
 623                                                  sizeof nested_partnum,
 624                                                  "%s.%d", partnum,
 625                                                  ++part_seq);
 626                                 }
 627                                 else {
 628                                         snprintf(nested_partnum,
 629                                                  sizeof nested_partnum,
 630                                                  "%d", ++part_seq);
 631                                 }
 632                                 recurseable_mime_parser(nested_partnum,
 633                                                         part_start,
 634                                                         part_end,
 635                                                         CallBack,
 636                                                         PreMultiPartCallBack,
 637                                                         PostMultiPartCallBack,
 638                                                         userdata,
 639                                                         dont_decode,
 640                                                         SubMimeHeaders);
 641                         }
 642
 643                         if (next_boundary != NULL) {
 644                                 /* If we pass out of scope, don't attempt to
 645                                  * read past the end boundary. */
 646                                 if ((*(next_boundary + m->b[startary].len + 1) == '-') &&
 647                                     (*(next_boundary + m->b[startary].len + 2) == '-') ){
 648                                         ptr = content_end;
 649                                 }
 650                                 else {
 651                                         /* Set up for the next part. */
 652                                         part_start = strstr(next_boundary, "\n");
 653
 654                                         /* Determine whether newlines are LF or CRLF */
 655                                         evaluate_crlf_ptr = part_start;
 656                                         --evaluate_crlf_ptr;
 657                                         if ((*evaluate_crlf_ptr == '\r') &&
 658                                             (*(evaluate_crlf_ptr + 1) == '\n'))
 659                                         {
 660                                                 crlf_in_use = 1;
 661                                         }
 662                                         else {
 663                                                 crlf_in_use = 0;
 664                                         }
 665
 666                                         /* Advance past the LF ... now we're in the next part */
 667                                         ++part_start;
 668                                         ptr = part_start;
 669                                 }
 670                         }
 671                         else {
 672                                 /* Invalid end of multipart.  Bail out! */
 673                                 ptr = content_end;
 674                         }
 675                         FlushInterestingMimes(SubMimeHeaders);
 676                 } while ( (ptr < content_end) && (next_boundary != NULL) );
 677
 678                 free(SubMimeHeaders);
 679
 680                 if (PostMultiPartCallBack != NULL) {
 681                         PostMultiPartCallBack("",
 682                                               "",
 683                                               partnum,
 684                                               "",
 685                                               NULL,
 686                                               m->b[content_type].Key,
 687                                               m->b[charset].Key,
 688                                               0,
 689                                               m->b[encoding].Key,
 690                                               m->b[id].Key,
 691                                               userdata);
 692                 }
 693         } /* If it's not a multipart message, then do something with it */
 694         else {
 695                 size_t length;
 696                 part_start = ptr;
 697                 length = content_end - part_start;
 698                 ptr = part_end = content_end;
 699
 700
 701                 /* The following code will truncate the MIME part to the size
 702                  * specified by the Content-length: header.   We have commented it
 703                  * out because these headers have a tendency to be wrong.
 704                  *
 705                  *      if ( (content_length > 0) && (length > content_length) ) {
 706                  *              length = content_length;
 707                  *      }
 708                  */
 709
 710                 /* Sometimes the "name" field is tacked on to Content-type,
 711                  * and sometimes it's tacked on to Content-disposition.  Use
 712                  * whichever one we have.
 713                  */
 714                 if (m->b[content_disposition_name].len > m->b[content_type_name].len) {
 715                         chosen_name = &m->b[content_disposition_name];
 716                 }
 717                 else {
 718                         chosen_name = &m->b[content_type_name];
 719                 }
 720
 721                 /* Ok, we've got a non-multipart part here, so do something with it.
 722                  */
 723                 mime_decode(partnum,
 724                             part_start,
 725                             length,
 726                             m->b[content_type].Key,
 727                             m->b[charset].Key,
 728                             m->b[encoding].Key,
 729                             m->b[disposition].Key,
 730                             m->b[id].Key,
 731                             chosen_name->Key,
 732                             m->b[filename].Key,
 733                             CallBack,
 734                             NULL, NULL,
 735                             userdata,
 736                             dont_decode
 737                         );
 738
 739                 /*
 740                  * Now if it's an encapsulated message/rfc822 then we have to recurse into it
 741                  */
 742                 if (!strcasecmp(&m->b[content_type].Key[0], "message/rfc822")) {
 743
 744                         if (PreMultiPartCallBack != NULL) {
 745                                 PreMultiPartCallBack("",
 746                                                      "",
 747                                                      partnum,
 748                                                      "",
 749                                                      NULL,
 750                                                      m->b[content_type].Key,
 751                                                      m->b[charset].Key,
 752                                                      0,
 753                                                      m->b[encoding].Key,
 754                                                      m->b[id].Key,
 755                                                      userdata);
 756                         }
 757                         if (CallBack != NULL) {
 758                                 if (strlen(partnum) > 0) {
 759                                         snprintf(nested_partnum,
 760                                                  sizeof nested_partnum,
 761                                                  "%s.%d", partnum,
 762                                                  ++part_seq);
 763                                 }
 764                                 else {
 765                                         snprintf(nested_partnum,
 766                                                  sizeof nested_partnum,
 767                                                  "%d", ++part_seq);
 768                                 }
 769                                 the_mime_parser(nested_partnum,
 770                                                 part_start,
 771                                                 part_end,
 772                                                 CallBack,
 773                                                 PreMultiPartCallBack,
 774                                                 PostMultiPartCallBack,
 775                                                 userdata,
 776                                                 dont_decode
 777                                         );
 778                         }
 779                         if (PostMultiPartCallBack != NULL) {
 780                                 PostMultiPartCallBack("",
 781                                                       "",
 782                                                       partnum,
 783                                                       "",
 784                                                       NULL,
 785                                                       m->b[content_type].Key,
 786                                                       m->b[charset].Key,
 787                                                       0,
 788                                                       m->b[encoding].Key,
 789                                                       m->b[id].Key,
 790                                                       userdata);
 791                         }
 792
 793
 794                 }
 795
 796         }
 797
 798 }
 799
 800 /*
 801  * Break out the components of a multipart message
 802  * (This function expects to be fed HEADERS + CONTENT)
 803  * Note: NULL can be supplied as content_end; in this case, the message is
 804  * considered to have ended when the parser encounters a 0x00 byte.
 805  */
 806 void the_mime_parser(char *partnum,
 807                      char *content_start, char *content_end,
 808                      MimeParserCallBackType CallBack,
 809                      MimeParserCallBackType PreMultiPartCallBack,
 810                      MimeParserCallBackType PostMultiPartCallBack,
 811                      void *userdata,
 812                      int dont_decode)
 813 {
 814         interesting_mime_headers *m;
 815
 816         /* If the caller didn't supply an endpointer, generate one by measure */
 817         if (content_end == NULL) {
 818                 content_end = &content_start[strlen(content_start)];
 819         }
 820
 821         m = InitInterestingMimes();
 822
 823         if (!parse_MimeHeaders(m, &content_start, content_end))
 824         {
 825
 826                 recurseable_mime_parser(partnum,
 827                                         content_start, content_end,
 828                                         CallBack,
 829                                         PreMultiPartCallBack,
 830                                         PostMultiPartCallBack,
 831                                         userdata,
 832                                         dont_decode,
 833                                         m);
 834         }
 835         free(m);
 836 }
 837
 838 /*
 839  * Entry point for the MIME parser.
 840  * (This function expects to be fed HEADERS + CONTENT)
 841  * Note: NULL can be supplied as content_end; in this case, the message is
 842  * considered to have ended when the parser encounters a 0x00 byte.
 843  */
 844 void mime_parser(char *content_start,
 845                  char *content_end,
 846                  MimeParserCallBackType CallBack,
 847                  MimeParserCallBackType PreMultiPartCallBack,
 848                  MimeParserCallBackType PostMultiPartCallBack,
 849                  void *userdata,
 850                  int dont_decode)
 851 {
 852
 853         the_mime_parser("", content_start, content_end,
 854                         CallBack,
 855                         PreMultiPartCallBack,
 856                         PostMultiPartCallBack,
 857                         userdata, dont_decode);
 858 }
 859
 860
 861
 862
 863
 864
 865 typedef struct _MimeGuess {
 866         const char *Pattern;
 867         size_t PatternLen;
 868         long PatternOffset;
 869         const char *MimeString;
 870 } MimeGuess;
 871
 872 MimeGuess MyMimes [] = {
 873         {
 874                 "GIF",
 875                 3,
 876                 0,
 877                 "image/gif"
 878         },
 879         {
 880                 "\xff\xd8",
 881                 2,
 882                 0,
 883                 "image/jpeg"
 884         },
 885         {
 886                 "\x89PNG",
 887                 4,
 888                 0,
 889                 "image/png"
 890         },
 891         { // last...
 892                 "",
 893                 0,
 894                 0,
 895                 ""
 896         }
 897 };
 898
 899
 900 const char *GuessMimeType(const char *data, size_t dlen)
 901 {
 902         int MimeIndex = 0;
 903
 904         while (MyMimes[MimeIndex].PatternLen != 0)
 905         {
 906                 if ((MyMimes[MimeIndex].PatternLen +
 907                      MyMimes[MimeIndex].PatternOffset < dlen) &&
 908                     strncmp(MyMimes[MimeIndex].Pattern,
 909                             &data[MyMimes[MimeIndex].PatternOffset],
 910                             MyMimes[MimeIndex].PatternLen) == 0)
 911                 {
 912                         return MyMimes[MimeIndex].MimeString;
 913                 }
 914                 MimeIndex ++;
 915         }
 916         /*
 917          * ok, our simple minded algorythm didn't find anything,
 918          * let the big chegger try it, he wil default to application/octet-stream
 919          */
 920         return (xdg_mime_get_mime_type_for_data(data, dlen));
 921 }
 922
 923
 924 const char* GuessMimeByFilename(const char *what, size_t len)
 925 {
 926         /* we know some hardcoded on our own, try them... */
 927         if ((len > 3) && !strncasecmp(&what[len - 4], ".gif", 4))
 928                 return "image/gif";
 929         else if ((len > 2) && !strncasecmp(&what[len - 3], ".js", 3))
 930                 return  "text/javascript";
 931         else if ((len > 3) && !strncasecmp(&what[len - 4], ".txt", 4))
 932                 return "text/plain";
 933         else if ((len > 3) && !strncasecmp(&what[len - 4], ".css", 4))
 934                 return "text/css";
 935         else if ((len > 3) && !strncasecmp(&what[len - 4], ".htc", 4))
 936                 return "text/x-component";
 937         else if ((len > 3) && !strncasecmp(&what[len - 4], ".jpg", 4))
 938                 return "image/jpeg";
 939         else if ((len > 3) && !strncasecmp(&what[len - 4], ".png", 4))
 940                 return "image/png";
 941         else if ((len > 3) && !strncasecmp(&what[len - 4], ".ico", 4))
 942                 return "image/x-icon";
 943         else if ((len > 3) && !strncasecmp(&what[len - 4], ".vcf", 4))
 944                 return "text/x-vcard";
 945         else if ((len > 4) && !strncasecmp(&what[len - 5], ".html", 5))
 946                 return "text/html";
 947         else if ((len > 3) && !strncasecmp(&what[len - 4], ".htm", 4))
 948                 return "text/html";
 949         else if ((len > 3) && !strncasecmp(&what[len - 4], ".wml", 4))
 950                 return "text/vnd.wap.wml";
 951         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wmls", 5))
 952                 return "text/vnd.wap.wmlscript";
 953         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wmlc", 5))
 954                 return "application/vnd.wap.wmlc";
 955         else if ((len > 5) && !strncasecmp(&what[len - 6], ".wmlsc", 6))
 956                 return "application/vnd.wap.wmlscriptc";
 957         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wbmp", 5))
 958                 return "image/vnd.wap.wbmp";
 959         else
 960                 /* and let xdgmime do the fallback. */
 961                 return xdg_mime_get_mime_type_from_file_name(what);
 962 }
 963
 964 static HashList *IconHash = NULL;
 965
 966 typedef struct IconName IconName;
 967
 968 struct IconName {
 969         char *FlatName;
 970         char *FileName;
 971 };
 972
 973 static void DeleteIcon(void *IconNamePtr)
 974 {
 975         IconName *Icon = (IconName*) IconNamePtr;
 976         free(Icon->FlatName);
 977         free(Icon->FileName);
 978         free(Icon);
 979 }
 980
 981 /*
 982 static const char *PrintFlat(void *IconNamePtr)
 983 {
 984         IconName *Icon = (IconName*) IconNamePtr;
 985         return Icon->FlatName;
 986 }
 987 static const char *PrintFile(void *IconNamePtr)
 988 {
 989         IconName *Icon = (IconName*) IconNamePtr;
 990         return Icon->FileName;
 991 }
 992 */
 993
 994 #define GENSTR "x-generic"
 995 #define IGNORE_PREFIX_1 "gnome-mime"
 996 int LoadIconDir(const char *DirName)
 997 {
 998         DIR *filedir = NULL;
 999         struct dirent *filedir_entry;
1000         int d_namelen;
1001         int d_without_ext;
1002         IconName *Icon;
1003
1004         filedir = opendir (DirName);
1005         IconHash = NewHash(1, NULL);
1006         if (filedir == NULL) {
1007                 return 0;
1008         }
1009
1010         while ((filedir_entry = readdir(filedir)))
1011         {
1012                 char *MinorPtr;
1013                 char *PStart;
1014 #ifdef _DIRENT_HAVE_D_NAMELEN
1015                 d_namelen = filedir_entry->d_namelen;
1016 #else
1017                 d_namelen = strlen(filedir_entry->d_name);
1018 #endif
1019                 d_without_ext = d_namelen;
1020                 while ((d_without_ext > 0) && (filedir_entry->d_name[d_without_ext] != '.'))
1021                         d_without_ext --;
1022                 if ((d_without_ext == 0) || (d_namelen < 3))
1023                         continue;
1024
1025                 if ((sizeof(IGNORE_PREFIX_1) < d_namelen) &&
1026                     (strncmp(IGNORE_PREFIX_1,
1027                              filedir_entry->d_name,
1028                              sizeof(IGNORE_PREFIX_1) - 1) == 0)) {
1029                         PStart = filedir_entry->d_name + sizeof(IGNORE_PREFIX_1);
1030                         d_without_ext -= sizeof(IGNORE_PREFIX_1);
1031                 }
1032                 else {
1033                         PStart = filedir_entry->d_name;
1034                 }
1035                 Icon = malloc(sizeof(IconName));
1036
1037                 Icon->FileName = malloc(d_namelen + 1);
1038                 memcpy(Icon->FileName, filedir_entry->d_name, d_namelen + 1);
1039
1040                 Icon->FlatName = malloc(d_without_ext + 1);
1041                 memcpy(Icon->FlatName, PStart, d_without_ext);
1042                 Icon->FlatName[d_without_ext] = '\0';
1043                 /* Try to find Minor type in image-jpeg */
1044                 MinorPtr = strchr(Icon->FlatName, '-');
1045                 if (MinorPtr != NULL) {
1046                         size_t MinorLen;
1047                         MinorLen = 1 + d_without_ext - (MinorPtr - Icon->FlatName + 1);
1048                         if ((MinorLen == sizeof(GENSTR)) &&
1049                             (strncmp(MinorPtr + 1, GENSTR, sizeof(GENSTR)) == 0)) {
1050                                 /* ok, we found a generic filename. cut the generic. */
1051                                 *MinorPtr = '\0';
1052                                 d_without_ext = d_without_ext - (MinorPtr - Icon->FlatName);
1053                         }
1054                         else { /* Map the major / minor separator to / */
1055                                 *MinorPtr = '/';
1056                         }
1057                 }
1058
1059 //              PrintHash(IconHash, PrintFlat, PrintFile);
1060 //              printf("%s - %s\n", Icon->FlatName, Icon->FileName);
1061                 Put(IconHash, Icon->FlatName, d_without_ext, Icon, DeleteIcon);
1062 //              PrintHash(IconHash, PrintFlat, PrintFile);
1063         }
1064         closedir(filedir);
1065         return 1;
1066 }
1067
1068 const char *GetIconFilename(char *MimeType, size_t len)
1069 {
1070         void *vIcon;
1071         IconName *Icon;
1072
1073         if(IconHash == NULL)
1074                 return NULL;
1075
1076         GetHash(IconHash, MimeType, len, &vIcon), Icon = (IconName*) vIcon;
1077         /* didn't find the exact mimetype? try major only. */
1078         if (Icon == NULL) {
1079                 char * pMinor;
1080                 pMinor = strchr(MimeType, '/');
1081                 if (pMinor != NULL) {
1082                         *pMinor = '\0';
1083                         GetHash(IconHash, MimeType, pMinor - MimeType, &vIcon),
1084                                 Icon = (IconName*) vIcon;
1085                 }
1086         }
1087         if (Icon == NULL) {
1088                 return NULL;
1089         }
1090
1091         /*printf("Getting: [%s] == [%s] -> [%s]\n", MimeType, Icon->FlatName, Icon->FileName);*/
1092         return Icon->FileName;
1093 }
1094
1095 void ShutDownLibCitadelMime(void)
1096 {
1097         DeleteHash(&IconHash);
1098 }