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  *
   6  * This program is open source software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License as published by
   8  * the Free Software Foundation; either version 3 of the License, or
   9  * (at your option) any later version.
  10  *
  11  * This program is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14  * GNU General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU General Public License
  17  * along with this program; if not, write to the Free Software
  18  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  19  */
  20
  21 #include <stdlib.h>
  22 #include <unistd.h>
  23 #include <stdio.h>
  24 #include <signal.h>
  25 #include <sys/types.h>
  26 #include <ctype.h>
  27 #include <string.h>
  28 #include <sys/stat.h>
  29 #include <sys/types.h>
  30 #include <dirent.h>
  31 #include <errno.h>
  32
  33 #include "xdgmime/xdgmime.h"
  34 #include "libcitadel.h"
  35 #include "libcitadellocal.h"
  36
  37 const unsigned char FromHexTable [256] = {
  38         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //  0
  39         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 10
  40         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 20
  41         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 30
  42         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x00, 0x01, // 40
  43         0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0xFF, 0xFF, // 50
  44         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0A, 0x0B, 0x0C, 0x0D, 0x0E, // 60
  45         0x0F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 70
  46         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, // 80
  47         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0x0A, 0x0B, 0x0C, // 90
  48         0x0D, 0x0E, 0x0F, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //100
  49         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //110
  50         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //120
  51         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //130
  52         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //140
  53         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //150
  54         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //160
  55         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //170
  56         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //180
  57         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //190
  58         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //200
  59         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //210
  60         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //220
  61         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //230
  62         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, //240
  63         0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF                          //250
  64 };
  65
  66
  67 long extract_key(char *target, char *source, long sourcelen, char *key, long keylen, char KeyEnd)
  68 {
  69         char *sptr, *ptr = NULL;
  70         int double_quotes = 0;
  71         long RealKeyLen = keylen;
  72
  73         sptr = source;
  74
  75         while (sptr != NULL)
  76         {
  77                 ptr = bmstrcasestr_len(sptr, sourcelen - (sptr - source),
  78                                        key, keylen);
  79                 if(ptr != NULL)
  80                 {
  81                         while (isspace(*(ptr + RealKeyLen)))
  82                                 RealKeyLen ++;
  83                         if (*(ptr + RealKeyLen) == KeyEnd)
  84                         {
  85                                 sptr = NULL;
  86                                 RealKeyLen ++;
  87                         }
  88                         else
  89                         {
  90                                 sptr = ptr + RealKeyLen + 1;
  91                         }
  92                 }
  93                 else
  94                         sptr = ptr;
  95         }
  96         if (ptr == NULL) {
  97                 *target = '\0';
  98                 return 0;
  99         }
 100         strcpy(target, (ptr + RealKeyLen));
 101
 102         for (ptr=target; (*ptr != 0); ptr++) {
 103
 104                 /* A semicolon means we've hit the end of the key, unless we're inside double quotes */
 105                 if ( (double_quotes != 1) && (*ptr == ';')) {
 106                         *ptr = 0;
 107                 }
 108
 109                 /* if we find double quotes, we've got a great set of string boundaries */
 110                 if (*ptr == '\"') {
 111                         ++double_quotes;
 112                         if (double_quotes == 1) {
 113                                 strcpy(ptr, ptr+1);
 114                         }
 115                         else {
 116                                 *ptr = 0;
 117                         }
 118                 }
 119         }
 120         *ptr = '\0';
 121         return ptr - target;
 122 }
 123
 124
 125 /*
 126  * For non-multipart messages, we need to generate a quickie partnum of "1"
 127  * to return to callback functions.  Some callbacks demand it.
 128  */
 129 char *fixed_partnum(char *supplied_partnum) {
 130         if (supplied_partnum == NULL) return "1";
 131         if (strlen(supplied_partnum)==0) return "1";
 132         return supplied_partnum;
 133 }
 134
 135
 136 static inline unsigned int _decode_hex(const char *Source)
 137 {
 138         unsigned int ret = '?';
 139         unsigned char LO_NIBBLE;
 140         unsigned char HI_NIBBLE;
 141
 142         HI_NIBBLE = FromHexTable[(unsigned char) *Source];
 143         LO_NIBBLE = FromHexTable[(unsigned char) *(Source+1)];
 144
 145         if ((LO_NIBBLE == 0xFF) || (LO_NIBBLE == 0xFF))
 146                 return ret;
 147         ret = HI_NIBBLE;
 148         ret = ret << 4;
 149         ret = ret | LO_NIBBLE;
 150         return ret;
 151 }
 152
 153 unsigned int decode_hex(char *Source) {return _decode_hex(Source);}
 154
 155 /*
 156  * Convert "quoted-printable" to binary.  Returns number of bytes decoded.
 157  * according to RFC2045 section 6.7
 158  */
 159 int CtdlDecodeQuotedPrintable(char *decoded, char *encoded, int sourcelen) {
 160         unsigned int ch;
 161         int decoded_length = 0;
 162         int pos = 0;
 163
 164         while (pos < sourcelen)
 165         {
 166                 if (*(encoded + pos) == '=')
 167                 {
 168                         pos ++;
 169                         if (*(encoded + pos) == '\n')
 170                         {
 171                                 pos ++;
 172                         }
 173                         else if (*(encoded + pos) == '\r')
 174                         {
 175                                 pos ++;
 176                                 if (*(encoded + pos) == '\n')
 177                                         pos++;
 178                         }
 179                         else
 180                         {
 181                                 ch = _decode_hex(&encoded[pos]);
 182                                 pos += 2;
 183                                 decoded[decoded_length++] = ch;
 184                         }
 185                 }
 186                 else
 187                 {
 188                         decoded[decoded_length++] = encoded[pos];
 189                         pos += 1;
 190                 }
 191         }
 192         decoded[decoded_length] = 0;
 193         return(decoded_length);
 194 }
 195
 196
 197 /*
 198  * Given a message or message-part body and a length, handle any necessary
 199  * decoding and pass the request up the stack.
 200  */
 201 void mime_decode(char *partnum,
 202                  char *part_start, size_t length,
 203                  char *content_type, char *charset, char *encoding,
 204                  char *disposition,
 205                  char *id,
 206                  char *name, char *filename,
 207                  MimeParserCallBackType CallBack,
 208                  MimeParserCallBackType PreMultiPartCallBack,
 209                  MimeParserCallBackType PostMultiPartCallBack,
 210                  void *userdata,
 211                  int dont_decode)
 212 {
 213
 214         char *decoded;
 215         size_t bytes_decoded = 0;
 216
 217         /* Some encodings aren't really encodings */
 218         if (!strcasecmp(encoding, "7bit"))
 219                 *encoding = '\0';
 220         if (!strcasecmp(encoding, "8bit"))
 221                 *encoding = '\0';
 222         if (!strcasecmp(encoding, "binary"))
 223                 *encoding = '\0';
 224         if (!strcasecmp(encoding, "ISO-8859-1"))
 225                 *encoding = '\0';
 226
 227         /* If this part is not encoded, send as-is */
 228         if ( (strlen(encoding) == 0) || (dont_decode)) {
 229                 if (CallBack != NULL) {
 230                         CallBack(name,
 231                                  filename,
 232                                  fixed_partnum(partnum),
 233                                  disposition,
 234                                  part_start,
 235                                  content_type,
 236                                  charset,
 237                                  length,
 238                                  encoding,
 239                                  id,
 240                                  userdata);
 241                         }
 242                 return;
 243         }
 244
 245         /* Fail silently if we hit an unknown encoding. */
 246         if ((strcasecmp(encoding, "base64"))
 247             && (strcasecmp(encoding, "quoted-printable"))) {
 248                 return;
 249         }
 250
 251         /*
 252          * Allocate a buffer for the decoded data.  The output buffer is slightly
 253          * larger than the input buffer; this assumes that the decoded data
 254          * will never be significantly larger than the encoded data.  This is a
 255          * safe assumption with base64, uuencode, and quoted-printable.
 256          */
 257         decoded = malloc(length + 32768);
 258         if (decoded == NULL) {
 259                 return;
 260         }
 261
 262         if (!strcasecmp(encoding, "base64")) {
 263                 bytes_decoded = CtdlDecodeBase64(decoded, part_start, length);
 264         }
 265         else if (!strcasecmp(encoding, "quoted-printable")) {
 266                 bytes_decoded = CtdlDecodeQuotedPrintable(decoded, part_start, length);
 267         }
 268
 269         if (bytes_decoded > 0) if (CallBack != NULL) {
 270                         char encoding_buf[SIZ];
 271
 272                         strcpy(encoding_buf, "binary");
 273                         CallBack(name,
 274                                  filename,
 275                                  fixed_partnum(partnum),
 276                                  disposition,
 277                                  decoded,
 278                                  content_type,
 279                                  charset,
 280                                  bytes_decoded,
 281                                  encoding_buf,
 282                                  id,
 283                                  userdata);
 284         }
 285
 286         free(decoded);
 287 }
 288
 289 /*
 290  * this is the extract of mime_decode which can be called if 'dont_decode' was set;
 291  * to save the cpu intense process of decoding to the time when it realy wants the content.
 292  * returns:
 293  *   - > 0 we decoded something, its on *decoded, you need to free it.
 294  *   - = 0 no need to decode stuff. *decoded will be NULL.
 295  *   - < 0 an error occured, either an unknown encoding, or alloc failed. no need to free.
 296  */
 297 int mime_decode_now (char *part_start,
 298                      size_t length,
 299                      char *encoding,
 300                      char **decoded,
 301                      size_t *bytes_decoded)
 302 {
 303         *bytes_decoded = 0;
 304         *decoded = NULL;
 305         /* Some encodings aren't really encodings */
 306         if (!strcasecmp(encoding, "7bit"))
 307                 *encoding = '\0';
 308         if (!strcasecmp(encoding, "8bit"))
 309                 *encoding = '\0';
 310         if (!strcasecmp(encoding, "binary"))
 311                 *encoding = '\0';
 312
 313         /* If this part is not encoded, send as-is */
 314         if (strlen(encoding) == 0) {
 315                 return 0;
 316         }
 317
 318
 319         /* Fail if we hit an unknown encoding. */
 320         if ((strcasecmp(encoding, "base64"))
 321             && (strcasecmp(encoding, "quoted-printable"))) {
 322                 return -1;
 323         }
 324
 325         /*
 326          * Allocate a buffer for the decoded data.  The output buffer is slightly
 327          * larger than the input buffer; this assumes that the decoded data
 328          * will never be significantly larger than the encoded data.  This is a
 329          * safe assumption with base64, uuencode, and quoted-printable.
 330          */
 331         *decoded = malloc(length + 32768);
 332         if (decoded == NULL) {
 333                 return -1;
 334         }
 335
 336         if (!strcasecmp(encoding, "base64")) {
 337                 *bytes_decoded = CtdlDecodeBase64(*decoded, part_start, length);
 338                 return 1;
 339         }
 340         else if (!strcasecmp(encoding, "quoted-printable")) {
 341                 *bytes_decoded = CtdlDecodeQuotedPrintable(*decoded, part_start, length);
 342                 return 1;
 343         }
 344         return -1;
 345 }
 346
 347 typedef enum _eIntMimeHdrs {
 348         boundary,
 349         startary,
 350         endary,
 351         content_type,
 352         charset,
 353         encoding,
 354         content_type_name,
 355         content_disposition_name,
 356         filename,
 357         disposition,
 358         id,
 359         eMax /* don't move ! */
 360 } eIntMimeHdrs;
 361
 362 typedef struct _CBufStr {
 363         char Key[SIZ];
 364         long len;
 365 }CBufStr;
 366
 367 typedef struct _interesting_mime_headers {
 368         CBufStr b[eMax];
 369         long content_length;
 370         long is_multipart;
 371 } interesting_mime_headers;
 372
 373
 374 static void FlushInterestingMimes(interesting_mime_headers *m)
 375 {
 376         int i;
 377
 378         for (i = 0; i < eMax; i++) {
 379              m->b[i].Key[0] = '\0';
 380              m->b[i].len = 0;
 381         }
 382         m->content_length = -1;
 383 }
 384 static interesting_mime_headers *InitInterestingMimes(void)
 385 {
 386         interesting_mime_headers *m;
 387         m = (interesting_mime_headers*) malloc( sizeof(interesting_mime_headers));
 388
 389         FlushInterestingMimes(m);
 390
 391         return m;
 392 }
 393
 394
 395 static long parse_MimeHeaders(interesting_mime_headers *m,
 396                               char** pcontent_start,
 397                               char *content_end)
 398 {
 399         char buf[SIZ];
 400         char header[SIZ];
 401         long headerlen;
 402         char *ptr, *pch;
 403         int buflen = 0;
 404         int i;
 405
 406         /* Learn interesting things from the headers */
 407         ptr = *pcontent_start;
 408         *header = '\0';
 409         headerlen = 0;
 410         do {
 411                 ptr = memreadlinelen(ptr, buf, SIZ, &buflen);
 412
 413                 for (i = 0; i < buflen; ++i) {
 414                         if (isspace(buf[i])) {
 415                                 buf[i] = ' ';
 416                         }
 417                 }
 418
 419                 if (!isspace(buf[0]) && (headerlen > 0)) {
 420                         if (!strncasecmp(header, "Content-type:", 13)) {
 421                                 memcpy (m->b[content_type].Key, &header[13], headerlen - 12);
 422                                 m->b[content_type].Key[headerlen - 12] = '\0';
 423                                 m->b[content_type].len = striplt (m->b[content_type].Key);
 424
 425                                 m->b[content_type_name].len = extract_key(m->b[content_type_name].Key, CKEY(m->b[content_type]), HKEY("name"), '=');
 426                                 m->b[charset].len           = extract_key(m->b[charset].Key,           CKEY(m->b[content_type]), HKEY("charset"), '=');
 427                                 m->b[boundary].len          = extract_key(m->b[boundary].Key,          header,       headerlen,  HKEY("boundary"), '=');
 428
 429                                 /* Deal with weird headers */
 430                                 pch = strchr(m->b[content_type].Key, ' ');
 431                                 if (pch != NULL) {
 432                                         *pch = '\0';
 433                                         m->b[content_type].len = m->b[content_type].Key - pch;
 434                                 }
 435                                 pch = strchr(m->b[content_type].Key, ';');
 436                                 if (pch != NULL) {
 437                                         *pch = '\0';
 438                                         m->b[content_type].len = m->b[content_type].Key - pch;
 439                                 }
 440                         }
 441                         else if (!strncasecmp(header, "Content-Disposition:", 20)) {
 442                                 memcpy (m->b[disposition].Key, &header[20], headerlen - 19);
 443                                 m->b[disposition].Key[headerlen - 19] = '\0';
 444                                 m->b[disposition].len = striplt(m->b[disposition].Key);
 445
 446                                 m->b[content_disposition_name].len = extract_key(m->b[content_disposition_name].Key, CKEY(m->b[disposition]), HKEY("name"), '=');
 447                                 m->b[filename].len                 = extract_key(m->b[filename].Key,                 CKEY(m->b[disposition]), HKEY("filename"), '=');
 448                                 pch = strchr(m->b[disposition].Key, ';');
 449                                 if (pch != NULL) *pch = '\0';
 450                                 m->b[disposition].len = striplt(m->b[disposition].Key);
 451                         }
 452                         else if (!strncasecmp(header, "Content-ID:", 11)) {
 453                                 memcpy(m->b[id].Key, &header[11], headerlen - 11);
 454                                 m->b[id].Key[headerlen - 11] = '\0';
 455                                 striplt(m->b[id].Key);
 456                                 m->b[id].len = stripallbut(m->b[id].Key, '<', '>');
 457                         }
 458                         else if (!strncasecmp(header, "Content-length: ", 15)) {
 459                                 char *clbuf;
 460                                 clbuf = &header[15];
 461                                 while (isspace(*clbuf))
 462                                         clbuf ++;
 463                                 m->content_length = (size_t) atol(clbuf);
 464                         }
 465                         else if (!strncasecmp(header, "Content-transfer-encoding: ", 26)) {
 466                                 memcpy(m->b[encoding].Key, &header[26], headerlen - 26);
 467                                 m->b[encoding].Key[headerlen - 26] = '\0';
 468                                 m->b[encoding].len = striplt(m->b[encoding].Key);
 469                         }
 470                         *header = '\0';
 471                         headerlen = 0;
 472                 }
 473                 if ((headerlen + buflen + 2) < SIZ) {
 474                         memcpy(&header[headerlen], buf, buflen);
 475                         headerlen += buflen;
 476                         header[headerlen] = '\0';
 477                 }
 478                 if (ptr >= content_end) {
 479                         return -1;
 480                 }
 481         } while ((!IsEmptyStr(buf)) && (*ptr != 0));
 482
 483         m->is_multipart = m->b[boundary].len != 0;
 484         *pcontent_start = ptr;
 485
 486         return 0;
 487 }
 488
 489
 490 static int IsAsciiEncoding(interesting_mime_headers *m)
 491 {
 492
 493         if ((m->b[encoding].len != 0) &&
 494             (strcasecmp(m->b[encoding].Key, "base64") == 0))
 495                 return 1;
 496         if ((m->b[encoding].len != 0) &&
 497             (strcmp(m->b[encoding].Key, "quoted-printable") == 0))
 498                 return 1;
 499
 500         return 0;
 501 }
 502
 503 static char *FindNextContent(char *ptr,
 504                              char *content_end,
 505                              interesting_mime_headers *SubMimeHeaders,
 506                              interesting_mime_headers *m)
 507 {
 508         char *next_boundary;
 509         char  tmp;
 510
 511         if (IsAsciiEncoding(SubMimeHeaders)) {
 512                 tmp = *content_end;
 513                 *content_end = '\0';
 514
 515                 /**
 516                  * ok, if we have a content length of the mime part,
 517                  * try skipping the content on the search for the next
 518                  * boundary. since we don't trust the content_length
 519                  * to be all accurate, and suspect it to lose one digit
 520                  * per line with a line length of 80 chars, we need
 521                  * to start searching a little before..
 522                  */
 523
 524                 if ((SubMimeHeaders->content_length != -1) &&
 525                     (SubMimeHeaders->content_length > 10))
 526                 {
 527                         char *pptr;
 528                         long lines;
 529
 530                         lines = SubMimeHeaders->content_length / 80;
 531                         pptr = ptr + SubMimeHeaders->content_length - lines - 10;
 532                         if (pptr < content_end)
 533                                 ptr = pptr;
 534                 }
 535
 536                 next_boundary = strstr(ptr, m->b[startary].Key);
 537                 *content_end = tmp;
 538         }
 539         else {
 540                 char *srch;
 541                 /**
 542                  * ok, if we have a content length of the mime part,
 543                  * try skipping the content on the search for the next
 544                  * boundary. since we don't trust the content_length
 545                  * to be all accurate, start searching a little before..
 546                  */
 547
 548                 if ((SubMimeHeaders->content_length != -1) &&
 549                     (SubMimeHeaders->content_length > 10))
 550                 {
 551                         char *pptr;
 552                         pptr = ptr + SubMimeHeaders->content_length - 10;
 553                         if (pptr < content_end)
 554                                 ptr = pptr;
 555                 }
 556
 557
 558                 srch = next_boundary = NULL;
 559                 for (srch = memchr(ptr, '-',  content_end - ptr);
 560                      (srch != NULL) && (srch < content_end);
 561                      srch = memchr(srch, '-',  content_end - srch))
 562                 {
 563                         if (!memcmp(srch,
 564                                     m->b[startary].Key,
 565                                     m->b[startary].len))
 566                         {
 567                                 next_boundary = srch;
 568                                 srch = content_end;
 569                         }
 570                         else srch ++;
 571
 572                 }
 573
 574         }
 575         return next_boundary;
 576 }
 577
 578 /*
 579  * Break out the components of a multipart message
 580  * (This function expects to be fed HEADERS + CONTENT)
 581  * Note: NULL can be supplied as content_end; in this case, the message is
 582  * considered to have ended when the parser encounters a 0x00 byte.
 583  */
 584 static void recurseable_mime_parser(char *partnum,
 585                                     char *content_start, char *content_end,
 586                                     MimeParserCallBackType CallBack,
 587                                     MimeParserCallBackType PreMultiPartCallBack,
 588                                     MimeParserCallBackType PostMultiPartCallBack,
 589                                     void *userdata,
 590                                     int dont_decode,
 591                                     interesting_mime_headers *m)
 592 {
 593         interesting_mime_headers *SubMimeHeaders;
 594         char     *ptr;
 595         char     *part_start;
 596         char     *part_end = NULL;
 597         char     *evaluate_crlf_ptr = NULL;
 598         char     *next_boundary;
 599         char      nested_partnum[256];
 600         int       crlf_in_use = 0;
 601         int       part_seq = 0;
 602         CBufStr  *chosen_name;
 603
 604
 605         /* If this is a multipart message, then recursively process it */
 606         ptr = content_start;
 607         part_start = NULL;
 608         if (m->is_multipart) {
 609
 610                 /* Tell the client about this message's multipartedness */
 611                 if (PreMultiPartCallBack != NULL) {
 612                         PreMultiPartCallBack("",
 613                                              "",
 614                                              partnum,
 615                                              "",
 616                                              NULL,
 617                                              m->b[content_type].Key,
 618                                              m->b[charset].Key,
 619                                              0,
 620                                              m->b[encoding].Key,
 621                                              m->b[id].Key,
 622                                              userdata);
 623                 }
 624
 625                 /* Figure out where the boundaries are */
 626                 m->b[startary].len = snprintf(m->b[startary].Key, SIZ, "--%s", m->b[boundary].Key);
 627                 SubMimeHeaders = InitInterestingMimes ();
 628
 629                 while ((*ptr == '\r') || (*ptr == '\n')) ptr ++;
 630
 631                 if (strncmp(ptr, m->b[startary].Key, m->b[startary].len) == 0)
 632                         ptr += m->b[startary].len;
 633
 634                 while ((*ptr == '\r') || (*ptr == '\n')) ptr ++;
 635
 636                 part_start = NULL;
 637                 do {
 638                         char *optr;
 639
 640                         optr = ptr;
 641                         if (parse_MimeHeaders(SubMimeHeaders, &ptr, content_end) != 0)
 642                                 break;
 643                         if ((ptr - optr > 2) &&
 644                             (*(ptr - 2) == '\r'))
 645                                 crlf_in_use = 1;
 646
 647                         part_start = ptr;
 648
 649                         next_boundary = FindNextContent(ptr,
 650                                                         content_end,
 651                                                         SubMimeHeaders,
 652                                                         m);
 653                         if ((next_boundary != NULL) &&
 654                             (next_boundary - part_start < 3)) {
 655                                 FlushInterestingMimes(SubMimeHeaders);
 656
 657                                 continue;
 658                         }
 659
 660                         if ( (part_start != NULL) && (next_boundary != NULL) ) {
 661                                 part_end = next_boundary;
 662                                 --part_end;             /* omit the trailing LF */
 663                                 if (crlf_in_use) {
 664                                         --part_end;     /* omit the trailing CR */
 665                                 }
 666
 667                                 if (!IsEmptyStr(partnum)) {
 668                                         snprintf(nested_partnum,
 669                                                  sizeof nested_partnum,
 670                                                  "%s.%d", partnum,
 671                                                  ++part_seq);
 672                                 }
 673                                 else {
 674                                         snprintf(nested_partnum,
 675                                                  sizeof nested_partnum,
 676                                                  "%d", ++part_seq);
 677                                 }
 678                                 recurseable_mime_parser(nested_partnum,
 679                                                         part_start,
 680                                                         part_end,
 681                                                         CallBack,
 682                                                         PreMultiPartCallBack,
 683                                                         PostMultiPartCallBack,
 684                                                         userdata,
 685                                                         dont_decode,
 686                                                         SubMimeHeaders);
 687                         }
 688
 689                         if (next_boundary != NULL) {
 690                                 /* If we pass out of scope, don't attempt to
 691                                  * read past the end boundary. */
 692                                 if ((*(next_boundary + m->b[startary].len) == '-') &&
 693                                     (*(next_boundary + m->b[startary].len + 1) == '-') ){
 694                                         ptr = content_end;
 695                                 }
 696                                 else {
 697                                         /* Set up for the next part. */
 698                                         part_start = strstr(next_boundary, "\n");
 699
 700                                         /* Determine whether newlines are LF or CRLF */
 701                                         evaluate_crlf_ptr = part_start;
 702                                         --evaluate_crlf_ptr;
 703                                         if ((*evaluate_crlf_ptr == '\r') &&
 704                                             (*(evaluate_crlf_ptr + 1) == '\n'))
 705                                         {
 706                                                 crlf_in_use = 1;
 707                                         }
 708                                         else {
 709                                                 crlf_in_use = 0;
 710                                         }
 711
 712                                         /* Advance past the LF ... now we're in the next part */
 713                                         ++part_start;
 714                                         ptr = part_start;
 715                                 }
 716                         }
 717                         else {
 718                                 /* Invalid end of multipart.  Bail out! */
 719                                 ptr = content_end;
 720                         }
 721                         FlushInterestingMimes(SubMimeHeaders);
 722                 } while ( (ptr < content_end) && (next_boundary != NULL) );
 723
 724                 free(SubMimeHeaders);
 725
 726                 if (PostMultiPartCallBack != NULL) {
 727                         PostMultiPartCallBack("",
 728                                               "",
 729                                               partnum,
 730                                               "",
 731                                               NULL,
 732                                               m->b[content_type].Key,
 733                                               m->b[charset].Key,
 734                                               0,
 735                                               m->b[encoding].Key,
 736                                               m->b[id].Key,
 737                                               userdata);
 738                 }
 739         } /* If it's not a multipart message, then do something with it */
 740         else {
 741                 size_t length;
 742                 part_start = ptr;
 743                 length = content_end - part_start;
 744                 ptr = part_end = content_end;
 745
 746
 747                 /* The following code will truncate the MIME part to the size
 748                  * specified by the Content-length: header.   We have commented it
 749                  * out because these headers have a tendency to be wrong.
 750                  *
 751                  *      if ( (content_length > 0) && (length > content_length) ) {
 752                  *              length = content_length;
 753                  *      }
 754                  */
 755
 756                 /* Sometimes the "name" field is tacked on to Content-type,
 757                  * and sometimes it's tacked on to Content-disposition.  Use
 758                  * whichever one we have.
 759                  */
 760                 if (m->b[content_disposition_name].len > m->b[content_type_name].len) {
 761                         chosen_name = &m->b[content_disposition_name];
 762                 }
 763                 else {
 764                         chosen_name = &m->b[content_type_name];
 765                 }
 766
 767                 /* Ok, we've got a non-multipart part here, so do something with it.
 768                  */
 769                 mime_decode(partnum,
 770                             part_start,
 771                             length,
 772                             m->b[content_type].Key,
 773                             m->b[charset].Key,
 774                             m->b[encoding].Key,
 775                             m->b[disposition].Key,
 776                             m->b[id].Key,
 777                             chosen_name->Key,
 778                             m->b[filename].Key,
 779                             CallBack,
 780                             NULL, NULL,
 781                             userdata,
 782                             dont_decode
 783                         );
 784
 785                 /*
 786                  * Now if it's an encapsulated message/rfc822 then we have to recurse into it
 787                  */
 788                 if (!strcasecmp(&m->b[content_type].Key[0], "message/rfc822")) {
 789
 790                         if (PreMultiPartCallBack != NULL) {
 791                                 PreMultiPartCallBack("",
 792                                                      "",
 793                                                      partnum,
 794                                                      "",
 795                                                      NULL,
 796                                                      m->b[content_type].Key,
 797                                                      m->b[charset].Key,
 798                                                      0,
 799                                                      m->b[encoding].Key,
 800                                                      m->b[id].Key,
 801                                                      userdata);
 802                         }
 803                         if (CallBack != NULL) {
 804                                 if (strlen(partnum) > 0) {
 805                                         snprintf(nested_partnum,
 806                                                  sizeof nested_partnum,
 807                                                  "%s.%d", partnum,
 808                                                  ++part_seq);
 809                                 }
 810                                 else {
 811                                         snprintf(nested_partnum,
 812                                                  sizeof nested_partnum,
 813                                                  "%d", ++part_seq);
 814                                 }
 815                                 the_mime_parser(nested_partnum,
 816                                                 part_start,
 817                                                 part_end,
 818                                                 CallBack,
 819                                                 PreMultiPartCallBack,
 820                                                 PostMultiPartCallBack,
 821                                                 userdata,
 822                                                 dont_decode
 823                                         );
 824                         }
 825                         if (PostMultiPartCallBack != NULL) {
 826                                 PostMultiPartCallBack("",
 827                                                       "",
 828                                                       partnum,
 829                                                       "",
 830                                                       NULL,
 831                                                       m->b[content_type].Key,
 832                                                       m->b[charset].Key,
 833                                                       0,
 834                                                       m->b[encoding].Key,
 835                                                       m->b[id].Key,
 836                                                       userdata);
 837                         }
 838
 839
 840                 }
 841
 842         }
 843
 844 }
 845
 846 /*
 847  * Break out the components of a multipart message
 848  * (This function expects to be fed HEADERS + CONTENT)
 849  * Note: NULL can be supplied as content_end; in this case, the message is
 850  * considered to have ended when the parser encounters a 0x00 byte.
 851  */
 852 void the_mime_parser(char *partnum,
 853                      char *content_start, char *content_end,
 854                      MimeParserCallBackType CallBack,
 855                      MimeParserCallBackType PreMultiPartCallBack,
 856                      MimeParserCallBackType PostMultiPartCallBack,
 857                      void *userdata,
 858                      int dont_decode)
 859 {
 860         interesting_mime_headers *m;
 861
 862         /* If the caller didn't supply an endpointer, generate one by measure */
 863         if (content_end == NULL) {
 864                 content_end = &content_start[strlen(content_start)];
 865         }
 866
 867         m = InitInterestingMimes();
 868
 869         if (!parse_MimeHeaders(m, &content_start, content_end))
 870         {
 871
 872                 recurseable_mime_parser(partnum,
 873                                         content_start, content_end,
 874                                         CallBack,
 875                                         PreMultiPartCallBack,
 876                                         PostMultiPartCallBack,
 877                                         userdata,
 878                                         dont_decode,
 879                                         m);
 880         }
 881         free(m);
 882 }
 883
 884 /*
 885  * Entry point for the MIME parser.
 886  * (This function expects to be fed HEADERS + CONTENT)
 887  * Note: NULL can be supplied as content_end; in this case, the message is
 888  * considered to have ended when the parser encounters a 0x00 byte.
 889  */
 890 void mime_parser(char *content_start,
 891                  char *content_end,
 892                  MimeParserCallBackType CallBack,
 893                  MimeParserCallBackType PreMultiPartCallBack,
 894                  MimeParserCallBackType PostMultiPartCallBack,
 895                  void *userdata,
 896                  int dont_decode)
 897 {
 898
 899         the_mime_parser("", content_start, content_end,
 900                         CallBack,
 901                         PreMultiPartCallBack,
 902                         PostMultiPartCallBack,
 903                         userdata, dont_decode);
 904 }
 905
 906
 907
 908
 909
 910
 911 typedef struct _MimeGuess {
 912         const char *Pattern;
 913         size_t PatternLen;
 914         long PatternOffset;
 915         const char *MimeString;
 916 } MimeGuess;
 917
 918 MimeGuess MyMimes [] = {
 919         {
 920                 "GIF",
 921                 3,
 922                 0,
 923                 "image/gif"
 924         },
 925         {
 926                 "\xff\xd8",
 927                 2,
 928                 0,
 929                 "image/jpeg"
 930         },
 931         {
 932                 "\x89PNG",
 933                 4,
 934                 0,
 935                 "image/png"
 936         },
 937         { // last...
 938                 "",
 939                 0,
 940                 0,
 941                 ""
 942         }
 943 };
 944
 945
 946 const char *GuessMimeType(const char *data, size_t dlen)
 947 {
 948         int MimeIndex = 0;
 949
 950         while (MyMimes[MimeIndex].PatternLen != 0)
 951         {
 952                 if ((MyMimes[MimeIndex].PatternLen +
 953                      MyMimes[MimeIndex].PatternOffset < dlen) &&
 954                     strncmp(MyMimes[MimeIndex].Pattern,
 955                             &data[MyMimes[MimeIndex].PatternOffset],
 956                             MyMimes[MimeIndex].PatternLen) == 0)
 957                 {
 958                         return MyMimes[MimeIndex].MimeString;
 959                 }
 960                 MimeIndex ++;
 961         }
 962         /*
 963          * ok, our simple minded algorythm didn't find anything,
 964          * let the big chegger try it, he wil default to application/octet-stream
 965          */
 966         return (xdg_mime_get_mime_type_for_data(data, dlen));
 967 }
 968
 969
 970 const char* GuessMimeByFilename(const char *what, size_t len)
 971 {
 972         /* we know some hardcoded on our own, try them... */
 973         if ((len > 3) && !strncasecmp(&what[len - 4], ".gif", 4))
 974                 return "image/gif";
 975         else if ((len > 2) && !strncasecmp(&what[len - 3], ".js", 3))
 976                 return  "text/javascript";
 977         else if ((len > 3) && !strncasecmp(&what[len - 4], ".txt", 4))
 978                 return "text/plain";
 979         else if ((len > 3) && !strncasecmp(&what[len - 4], ".css", 4))
 980                 return "text/css";
 981         else if ((len > 3) && !strncasecmp(&what[len - 4], ".htc", 4))
 982                 return "text/x-component";
 983         else if ((len > 3) && !strncasecmp(&what[len - 4], ".jpg", 4))
 984                 return "image/jpeg";
 985         else if ((len > 4) && !strncasecmp(&what[len - 5], ".jpeg", 5))
 986                 return "image/jpeg";
 987         else if ((len > 3) && !strncasecmp(&what[len - 4], ".png", 4))
 988                 return "image/png";
 989         else if ((len > 3) && !strncasecmp(&what[len - 4], ".ico", 4))
 990                 return "image/x-icon";
 991         else if ((len > 3) && !strncasecmp(&what[len - 4], ".vcf", 4))
 992                 return "text/x-vcard";
 993         else if ((len > 4) && !strncasecmp(&what[len - 5], ".html", 5))
 994                 return "text/html";
 995         else if ((len > 3) && !strncasecmp(&what[len - 4], ".htm", 4))
 996                 return "text/html";
 997         else if ((len > 3) && !strncasecmp(&what[len - 4], ".wml", 4))
 998                 return "text/vnd.wap.wml";
 999         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wmls", 5))
1000                 return "text/vnd.wap.wmlscript";
1001         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wmlc", 5))
1002                 return "application/vnd.wap.wmlc";
1003         else if ((len > 5) && !strncasecmp(&what[len - 6], ".wmlsc", 6))
1004                 return "application/vnd.wap.wmlscriptc";
1005         else if ((len > 4) && !strncasecmp(&what[len - 5], ".wbmp", 5))
1006                 return "image/vnd.wap.wbmp";
1007         else
1008                 /* and let xdgmime do the fallback. */
1009                 return xdg_mime_get_mime_type_from_file_name(what);
1010 }
1011
1012 static HashList *IconHash = NULL;
1013
1014 typedef struct IconName IconName;
1015
1016 struct IconName {
1017         char *FlatName;
1018         char *FileName;
1019 };
1020
1021 static void DeleteIcon(void *IconNamePtr)
1022 {
1023         IconName *Icon = (IconName*) IconNamePtr;
1024         free(Icon->FlatName);
1025         free(Icon->FileName);
1026         free(Icon);
1027 }
1028
1029 /*
1030 static const char *PrintFlat(void *IconNamePtr)
1031 {
1032         IconName *Icon = (IconName*) IconNamePtr;
1033         return Icon->FlatName;
1034 }
1035 static const char *PrintFile(void *IconNamePtr)
1036 {
1037         IconName *Icon = (IconName*) IconNamePtr;
1038         return Icon->FileName;
1039 }
1040 */
1041
1042 #define GENSTR "x-generic"
1043 #define IGNORE_PREFIX_1 "gnome-mime"
1044 int LoadIconDir(const char *DirName)
1045 {
1046         DIR *filedir = NULL;
1047         struct dirent *filedir_entry;
1048         int d_namelen;
1049         int d_without_ext;
1050         IconName *Icon;
1051
1052         filedir = opendir (DirName);
1053         IconHash = NewHash(1, NULL);
1054         if (filedir == NULL) {
1055                 return 0;
1056         }
1057
1058         while ((filedir_entry = readdir(filedir)))
1059         {
1060                 char *MinorPtr;
1061                 char *PStart;
1062 #ifdef _DIRENT_HAVE_D_NAMLEN
1063                 d_namelen = filedir_entry->d_namlen;
1064 #else
1065                 d_namelen = strlen(filedir_entry->d_name);
1066 #endif
1067                 d_without_ext = d_namelen;
1068                 while ((d_without_ext > 0) && (filedir_entry->d_name[d_without_ext] != '.'))
1069                         d_without_ext --;
1070                 if ((d_without_ext == 0) || (d_namelen < 3))
1071                         continue;
1072
1073                 if ((sizeof(IGNORE_PREFIX_1) < d_namelen) &&
1074                     (strncmp(IGNORE_PREFIX_1,
1075                              filedir_entry->d_name,
1076                              sizeof(IGNORE_PREFIX_1) - 1) == 0)) {
1077                         PStart = filedir_entry->d_name + sizeof(IGNORE_PREFIX_1);
1078                         d_without_ext -= sizeof(IGNORE_PREFIX_1);
1079                 }
1080                 else {
1081                         PStart = filedir_entry->d_name;
1082                 }
1083                 Icon = malloc(sizeof(IconName));
1084
1085                 Icon->FileName = malloc(d_namelen + 1);
1086                 memcpy(Icon->FileName, filedir_entry->d_name, d_namelen + 1);
1087
1088                 Icon->FlatName = malloc(d_without_ext + 1);
1089                 memcpy(Icon->FlatName, PStart, d_without_ext);
1090                 Icon->FlatName[d_without_ext] = '\0';
1091                 /* Try to find Minor type in image-jpeg */
1092                 MinorPtr = strchr(Icon->FlatName, '-');
1093                 if (MinorPtr != NULL) {
1094                         size_t MinorLen;
1095                         MinorLen = 1 + d_without_ext - (MinorPtr - Icon->FlatName + 1);
1096                         if ((MinorLen == sizeof(GENSTR)) &&
1097                             (strncmp(MinorPtr + 1, GENSTR, sizeof(GENSTR)) == 0)) {
1098                                 /* ok, we found a generic filename. cut the generic. */
1099                                 *MinorPtr = '\0';
1100                                 d_without_ext = d_without_ext - (MinorPtr - Icon->FlatName);
1101                         }
1102                         else { /* Map the major / minor separator to / */
1103                                 *MinorPtr = '/';
1104                         }
1105                 }
1106
1107 //              PrintHash(IconHash, PrintFlat, PrintFile);
1108 //              printf("%s - %s\n", Icon->FlatName, Icon->FileName);
1109                 Put(IconHash, Icon->FlatName, d_without_ext, Icon, DeleteIcon);
1110 //              PrintHash(IconHash, PrintFlat, PrintFile);
1111         }
1112         closedir(filedir);
1113         return 1;
1114 }
1115
1116 const char *GetIconFilename(char *MimeType, size_t len)
1117 {
1118         void *vIcon;
1119         IconName *Icon;
1120
1121         if(IconHash == NULL)
1122                 return NULL;
1123
1124         GetHash(IconHash, MimeType, len, &vIcon), Icon = (IconName*) vIcon;
1125         /* didn't find the exact mimetype? try major only. */
1126         if (Icon == NULL) {
1127                 char * pMinor;
1128                 pMinor = strchr(MimeType, '/');
1129                 if (pMinor != NULL) {
1130                         *pMinor = '\0';
1131                         GetHash(IconHash, MimeType, pMinor - MimeType, &vIcon),
1132                                 Icon = (IconName*) vIcon;
1133                 }
1134         }
1135         if (Icon == NULL) {
1136                 return NULL;
1137         }
1138
1139         /*printf("Getting: [%s] == [%s] -> [%s]\n", MimeType, Icon->FlatName, Icon->FileName);*/
1140         return Icon->FileName;
1141 }
1142
1143 void ShutDownLibCitadelMime(void)
1144 {
1145         DeleteHash(&IconHash);
1146 }