builtin-mailinfo.c

   1 /*
   2  * Another stupid program, this one parsing the headers of an
   3  * email to figure out authorship and subject
   4  */
   5 #include "cache.h"
   6 #include "builtin.h"
   7 #include "utf8.h"
   8
   9 static FILE *cmitmsg, *patchfile, *fin, *fout;
  10
  11 static int keep_subject;
  12 static const char *metainfo_charset;
  13 static char line[1000];
  14 static char name[1000];
  15 static char email[1000];
  16
  17 static enum  {
  18         TE_DONTCARE, TE_QP, TE_BASE64,
  19 } transfer_encoding;
  20 static enum  {
  21         TYPE_TEXT, TYPE_OTHER,
  22 } message_type;
  23
  24 static char charset[256];
  25 static int patch_lines;
  26 static char **p_hdr_data, **s_hdr_data;
  27
  28 #define MAX_HDR_PARSED 10
  29 #define MAX_BOUNDARIES 5
  30
  31 static char *sanity_check(char *name, char *email)
  32 {
  33         int len = strlen(name);
  34         if (len < 3 || len > 60)
  35                 return email;
  36         if (strchr(name, '@') || strchr(name, '<') || strchr(name, '>'))
  37                 return email;
  38         return name;
  39 }
  40
  41 static int bogus_from(char *line)
  42 {
  43         /* John Doe <johndoe> */
  44         char *bra, *ket, *dst, *cp;
  45
  46         /* This is fallback, so do not bother if we already have an
  47          * e-mail address.
  48          */
  49         if (*email)
  50                 return 0;
  51
  52         bra = strchr(line, '<');
  53         if (!bra)
  54                 return 0;
  55         ket = strchr(bra, '>');
  56         if (!ket)
  57                 return 0;
  58
  59         for (dst = email, cp = bra+1; cp < ket; )
  60                 *dst++ = *cp++;
  61         *dst = 0;
  62         for (cp = line; isspace(*cp); cp++)
  63                 ;
  64         for (bra--; isspace(*bra); bra--)
  65                 *bra = 0;
  66         cp = sanity_check(cp, email);
  67         strcpy(name, cp);
  68         return 1;
  69 }
  70
  71 static int handle_from(char *in_line)
  72 {
  73         char line[1000];
  74         char *at;
  75         char *dst;
  76
  77         strcpy(line, in_line);
  78         at = strchr(line, '@');
  79         if (!at)
  80                 return bogus_from(line);
  81
  82         /*
  83          * If we already have one email, don't take any confusing lines
  84          */
  85         if (*email && strchr(at+1, '@'))
  86                 return 0;
  87
  88         /* Pick up the string around '@', possibly delimited with <>
  89          * pair; that is the email part.  White them out while copying.
  90          */
  91         while (at > line) {
  92                 char c = at[-1];
  93                 if (isspace(c))
  94                         break;
  95                 if (c == '<') {
  96                         at[-1] = ' ';
  97                         break;
  98                 }
  99                 at--;
 100         }
 101         dst = email;
 102         for (;;) {
 103                 unsigned char c = *at;
 104                 if (!c || c == '>' || isspace(c)) {
 105                         if (c == '>')
 106                                 *at = ' ';
 107                         break;
 108                 }
 109                 *at++ = ' ';
 110                 *dst++ = c;
 111         }
 112         *dst++ = 0;
 113
 114         /* The remainder is name.  It could be "John Doe <john.doe@xz>"
 115          * or "john.doe@xz (John Doe)", but we have whited out the
 116          * email part, so trim from both ends, possibly removing
 117          * the () pair at the end.
 118          */
 119         at = line + strlen(line);
 120         while (at > line) {
 121                 unsigned char c = *--at;
 122                 if (!isspace(c)) {
 123                         at[(c == ')') ? 0 : 1] = 0;
 124                         break;
 125                 }
 126         }
 127
 128         at = line;
 129         for (;;) {
 130                 unsigned char c = *at;
 131                 if (!c || !isspace(c)) {
 132                         if (c == '(')
 133                                 at++;
 134                         break;
 135                 }
 136                 at++;
 137         }
 138         at = sanity_check(at, email);
 139         strcpy(name, at);
 140         return 1;
 141 }
 142
 143 static int handle_header(char *line, char *data, int ofs)
 144 {
 145         if (!line || !data)
 146                 return 1;
 147
 148         strcpy(data, line+ofs);
 149
 150         return 0;
 151 }
 152
 153 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 154  * to have enough heuristics to grok MIME encoded patches often found
 155  * on our mailing lists.  For example, we do not even treat header lines
 156  * case insensitively.
 157  */
 158
 159 static int slurp_attr(const char *line, const char *name, char *attr)
 160 {
 161         const char *ends, *ap = strcasestr(line, name);
 162         size_t sz;
 163
 164         if (!ap) {
 165                 *attr = 0;
 166                 return 0;
 167         }
 168         ap += strlen(name);
 169         if (*ap == '"') {
 170                 ap++;
 171                 ends = "\"";
 172         }
 173         else
 174                 ends = "; \t";
 175         sz = strcspn(ap, ends);
 176         memcpy(attr, ap, sz);
 177         attr[sz] = 0;
 178         return 1;
 179 }
 180
 181 struct content_type {
 182         char *boundary;
 183         int boundary_len;
 184 };
 185
 186 static struct content_type content[MAX_BOUNDARIES];
 187
 188 static struct content_type *content_top = content;
 189
 190 static int handle_content_type(char *line)
 191 {
 192         char boundary[256];
 193
 194         if (strcasestr(line, "text/") == NULL)
 195                  message_type = TYPE_OTHER;
 196         if (slurp_attr(line, "boundary=", boundary + 2)) {
 197                 memcpy(boundary, "--", 2);
 198                 if (content_top++ >= &content[MAX_BOUNDARIES]) {
 199                         fprintf(stderr, "Too many boundaries to handle\n");
 200                         exit(1);
 201                 }
 202                 content_top->boundary_len = strlen(boundary);
 203                 content_top->boundary = xmalloc(content_top->boundary_len+1);
 204                 strcpy(content_top->boundary, boundary);
 205         }
 206         if (slurp_attr(line, "charset=", charset)) {
 207                 int i, c;
 208                 for (i = 0; (c = charset[i]) != 0; i++)
 209                         charset[i] = tolower(c);
 210         }
 211         return 0;
 212 }
 213
 214 static int handle_content_transfer_encoding(char *line)
 215 {
 216         if (strcasestr(line, "base64"))
 217                 transfer_encoding = TE_BASE64;
 218         else if (strcasestr(line, "quoted-printable"))
 219                 transfer_encoding = TE_QP;
 220         else
 221                 transfer_encoding = TE_DONTCARE;
 222         return 0;
 223 }
 224
 225 static int is_multipart_boundary(const char *line)
 226 {
 227         return (!memcmp(line, content_top->boundary, content_top->boundary_len));
 228 }
 229
 230 static int eatspace(char *line)
 231 {
 232         int len = strlen(line);
 233         while (len > 0 && isspace(line[len-1]))
 234                 line[--len] = 0;
 235         return len;
 236 }
 237
 238 static char *cleanup_subject(char *subject)
 239 {
 240         for (;;) {
 241                 char *p;
 242                 int len, remove;
 243                 switch (*subject) {
 244                 case 'r': case 'R':
 245                         if (!memcmp("e:", subject+1, 2)) {
 246                                 subject += 3;
 247                                 continue;
 248                         }
 249                         break;
 250                 case ' ': case '\t': case ':':
 251                         subject++;
 252                         continue;
 253
 254                 case '[':
 255                         p = strchr(subject, ']');
 256                         if (!p) {
 257                                 subject++;
 258                                 continue;
 259                         }
 260                         len = strlen(p);
 261                         remove = p - subject;
 262                         if (remove <= len *2) {
 263                                 subject = p+1;
 264                                 continue;
 265                         }
 266                         break;
 267                 }
 268                 eatspace(subject);
 269                 return subject;
 270         }
 271 }
 272
 273 static void cleanup_space(char *buf)
 274 {
 275         unsigned char c;
 276         while ((c = *buf) != 0) {
 277                 buf++;
 278                 if (isspace(c)) {
 279                         buf[-1] = ' ';
 280                         c = *buf;
 281                         while (isspace(c)) {
 282                                 int len = strlen(buf);
 283                                 memmove(buf, buf+1, len);
 284                                 c = *buf;
 285                         }
 286                 }
 287         }
 288 }
 289
 290 static void decode_header(char *it, unsigned itsize);
 291 static const char *header[MAX_HDR_PARSED] = {
 292         "From","Subject","Date",
 293 };
 294
 295 static int check_header(char *line, unsigned linesize, char **hdr_data, int overwrite)
 296 {
 297         int i;
 298
 299         /* search for the interesting parts */
 300         for (i = 0; header[i]; i++) {
 301                 int len = strlen(header[i]);
 302                 if ((!hdr_data[i] || overwrite) &&
 303                     !strncasecmp(line, header[i], len) &&
 304                     line[len] == ':' && isspace(line[len + 1])) {
 305                         /* Unwrap inline B and Q encoding, and optionally
 306                          * normalize the meta information to utf8.
 307                          */
 308                         decode_header(line + len + 2, linesize - len - 2);
 309                         hdr_data[i] = xmalloc(1000 * sizeof(char));
 310                         if (! handle_header(line, hdr_data[i], len + 2)) {
 311                                 return 1;
 312                         }
 313                 }
 314         }
 315
 316         /* Content stuff */
 317         if (!strncasecmp(line, "Content-Type", 12) &&
 318                 line[12] == ':' && isspace(line[12 + 1])) {
 319                 decode_header(line + 12 + 2, linesize - 12 - 2);
 320                 if (! handle_content_type(line)) {
 321                         return 1;
 322                 }
 323         }
 324         if (!strncasecmp(line, "Content-Transfer-Encoding", 25) &&
 325                 line[25] == ':' && isspace(line[25 + 1])) {
 326                 decode_header(line + 25 + 2, linesize - 25 - 2);
 327                 if (! handle_content_transfer_encoding(line)) {
 328                         return 1;
 329                 }
 330         }
 331
 332         /* for inbody stuff */
 333         if (!memcmp(">From", line, 5) && isspace(line[5]))
 334                 return 1;
 335         if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
 336                 for (i = 0; header[i]; i++) {
 337                         if (!memcmp("Subject: ", header[i], 9)) {
 338                                 if (! handle_header(line, hdr_data[i], 0)) {
 339                                         return 1;
 340                                 }
 341                         }
 342                 }
 343         }
 344
 345         /* no match */
 346         return 0;
 347 }
 348
 349 static int is_rfc2822_header(char *line)
 350 {
 351         /*
 352          * The section that defines the loosest possible
 353          * field name is "3.6.8 Optional fields".
 354          *
 355          * optional-field = field-name ":" unstructured CRLF
 356          * field-name = 1*ftext
 357          * ftext = %d33-57 / %59-126
 358          */
 359         int ch;
 360         char *cp = line;
 361
 362         /* Count mbox From headers as headers */
 363         if (!memcmp(line, "From ", 5) || !memcmp(line, ">From ", 6))
 364                 return 1;
 365
 366         while ((ch = *cp++)) {
 367                 if (ch == ':')
 368                         return cp != line;
 369                 if ((33 <= ch && ch <= 57) ||
 370                     (59 <= ch && ch <= 126))
 371                         continue;
 372                 break;
 373         }
 374         return 0;
 375 }
 376
 377 /*
 378  * sz is size of 'line' buffer in bytes.  Must be reasonably
 379  * long enough to hold one physical real-world e-mail line.
 380  */
 381 static int read_one_header_line(char *line, int sz, FILE *in)
 382 {
 383         int len;
 384
 385         /*
 386          * We will read at most (sz-1) bytes and then potentially
 387          * re-add NUL after it.  Accessing line[sz] after this is safe
 388          * and we can allow len to grow up to and including sz.
 389          */
 390         sz--;
 391
 392         /* Get the first part of the line. */
 393         if (!fgets(line, sz, in))
 394                 return 0;
 395
 396         /*
 397          * Is it an empty line or not a valid rfc2822 header?
 398          * If so, stop here, and return false ("not a header")
 399          */
 400         len = eatspace(line);
 401         if (!len || !is_rfc2822_header(line)) {
 402                 /* Re-add the newline */
 403                 line[len] = '\n';
 404                 line[len + 1] = '\0';
 405                 return 0;
 406         }
 407
 408         /*
 409          * Now we need to eat all the continuation lines..
 410          * Yuck, 2822 header "folding"
 411          */
 412         for (;;) {
 413                 int peek, addlen;
 414                 static char continuation[1000];
 415
 416                 peek = fgetc(in); ungetc(peek, in);
 417                 if (peek != ' ' && peek != '\t')
 418                         break;
 419                 if (!fgets(continuation, sizeof(continuation), in))
 420                         break;
 421                 addlen = eatspace(continuation);
 422                 if (len < sz - 1) {
 423                         if (addlen >= sz - len)
 424                                 addlen = sz - len - 1;
 425                         memcpy(line + len, continuation, addlen);
 426                         line[len] = '\n';
 427                         len += addlen;
 428                 }
 429         }
 430         line[len] = 0;
 431
 432         return 1;
 433 }
 434
 435 static int decode_q_segment(char *in, char *ot, unsigned otsize, char *ep, int rfc2047)
 436 {
 437         char *otbegin = ot;
 438         char *otend = ot + otsize;
 439         int c;
 440         while ((c = *in++) != 0 && (in <= ep)) {
 441                 if (ot == otend) {
 442                         *--ot = '\0';
 443                         return -1;
 444                 }
 445                 if (c == '=') {
 446                         int d = *in++;
 447                         if (d == '\n' || !d)
 448                                 break; /* drop trailing newline */
 449                         *ot++ = ((hexval(d) << 4) | hexval(*in++));
 450                         continue;
 451                 }
 452                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 453                         c = 0x20;
 454                 *ot++ = c;
 455         }
 456         *ot = 0;
 457         return (ot - otbegin);
 458 }
 459
 460 static int decode_b_segment(char *in, char *ot, unsigned otsize, char *ep)
 461 {
 462         /* Decode in..ep, possibly in-place to ot */
 463         int c, pos = 0, acc = 0;
 464         char *otbegin = ot;
 465         char *otend = ot + otsize;
 466
 467         while ((c = *in++) != 0 && (in <= ep)) {
 468                 if (ot == otend) {
 469                         *--ot = '\0';
 470                         return -1;
 471                 }
 472                 if (c == '+')
 473                         c = 62;
 474                 else if (c == '/')
 475                         c = 63;
 476                 else if ('A' <= c && c <= 'Z')
 477                         c -= 'A';
 478                 else if ('a' <= c && c <= 'z')
 479                         c -= 'a' - 26;
 480                 else if ('0' <= c && c <= '9')
 481                         c -= '0' - 52;
 482                 else if (c == '=') {
 483                         /* padding is almost like (c == 0), except we do
 484                          * not output NUL resulting only from it;
 485                          * for now we just trust the data.
 486                          */
 487                         c = 0;
 488                 }
 489                 else
 490                         continue; /* garbage */
 491                 switch (pos++) {
 492                 case 0:
 493                         acc = (c << 2);
 494                         break;
 495                 case 1:
 496                         *ot++ = (acc | (c >> 4));
 497                         acc = (c & 15) << 4;
 498                         break;
 499                 case 2:
 500                         *ot++ = (acc | (c >> 2));
 501                         acc = (c & 3) << 6;
 502                         break;
 503                 case 3:
 504                         *ot++ = (acc | c);
 505                         acc = pos = 0;
 506                         break;
 507                 }
 508         }
 509         *ot = 0;
 510         return (ot - otbegin);
 511 }
 512
 513 /*
 514  * When there is no known charset, guess.
 515  *
 516  * Right now we assume that if the target is UTF-8 (the default),
 517  * and it already looks like UTF-8 (which includes US-ASCII as its
 518  * subset, of course) then that is what it is and there is nothing
 519  * to do.
 520  *
 521  * Otherwise, we default to assuming it is Latin1 for historical
 522  * reasons.
 523  */
 524 static const char *guess_charset(const char *line, const char *target_charset)
 525 {
 526         if (is_encoding_utf8(target_charset)) {
 527                 if (is_utf8(line))
 528                         return NULL;
 529         }
 530         return "latin1";
 531 }
 532
 533 static void convert_to_utf8(char *line, unsigned linesize, const char *charset)
 534 {
 535         char *out;
 536
 537         if (!charset || !*charset) {
 538                 charset = guess_charset(line, metainfo_charset);
 539                 if (!charset)
 540                         return;
 541         }
 542
 543         if (!strcmp(metainfo_charset, charset))
 544                 return;
 545         out = reencode_string(line, metainfo_charset, charset);
 546         if (!out)
 547                 die("cannot convert from %s to %s\n",
 548                     charset, metainfo_charset);
 549         strlcpy(line, out, linesize);
 550         free(out);
 551 }
 552
 553 static int decode_header_bq(char *it, unsigned itsize)
 554 {
 555         char *in, *out, *ep, *cp, *sp;
 556         char outbuf[1000];
 557         int rfc2047 = 0;
 558
 559         in = it;
 560         out = outbuf;
 561         while ((ep = strstr(in, "=?")) != NULL) {
 562                 int sz, encoding;
 563                 char charset_q[256], piecebuf[256];
 564                 rfc2047 = 1;
 565
 566                 if (in != ep) {
 567                         sz = ep - in;
 568                         memcpy(out, in, sz);
 569                         out += sz;
 570                         in += sz;
 571                 }
 572                 /* E.g.
 573                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 574                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 575                  */
 576                 ep += 2;
 577                 cp = strchr(ep, '?');
 578                 if (!cp)
 579                         return rfc2047; /* no munging */
 580                 for (sp = ep; sp < cp; sp++)
 581                         charset_q[sp - ep] = tolower(*sp);
 582                 charset_q[cp - ep] = 0;
 583                 encoding = cp[1];
 584                 if (!encoding || cp[2] != '?')
 585                         return rfc2047; /* no munging */
 586                 ep = strstr(cp + 3, "?=");
 587                 if (!ep)
 588                         return rfc2047; /* no munging */
 589                 switch (tolower(encoding)) {
 590                 default:
 591                         return rfc2047; /* no munging */
 592                 case 'b':
 593                         sz = decode_b_segment(cp + 3, piecebuf, sizeof(piecebuf), ep);
 594                         break;
 595                 case 'q':
 596                         sz = decode_q_segment(cp + 3, piecebuf, sizeof(piecebuf), ep, 1);
 597                         break;
 598                 }
 599                 if (sz < 0)
 600                         return rfc2047;
 601                 if (metainfo_charset)
 602                         convert_to_utf8(piecebuf, sizeof(piecebuf), charset_q);
 603
 604                 sz = strlen(piecebuf);
 605                 if (outbuf + sizeof(outbuf) <= out + sz)
 606                         return rfc2047; /* no munging */
 607                 strcpy(out, piecebuf);
 608                 out += sz;
 609                 in = ep + 2;
 610         }
 611         strcpy(out, in);
 612         strlcpy(it, outbuf, itsize);
 613         return rfc2047;
 614 }
 615
 616 static void decode_header(char *it, unsigned itsize)
 617 {
 618
 619         if (decode_header_bq(it, itsize))
 620                 return;
 621         /* otherwise "it" is a straight copy of the input.
 622          * This can be binary guck but there is no charset specified.
 623          */
 624         if (metainfo_charset)
 625                 convert_to_utf8(it, itsize, "");
 626 }
 627
 628 static int decode_transfer_encoding(char *line, unsigned linesize, int inputlen)
 629 {
 630         char *ep;
 631
 632         switch (transfer_encoding) {
 633         case TE_QP:
 634                 ep = line + inputlen;
 635                 return decode_q_segment(line, line, linesize, ep, 0);
 636         case TE_BASE64:
 637                 ep = line + inputlen;
 638                 return decode_b_segment(line, line, linesize, ep);
 639         case TE_DONTCARE:
 640         default:
 641                 return inputlen;
 642         }
 643 }
 644
 645 static int handle_filter(char *line, unsigned linesize, int linelen);
 646
 647 static int find_boundary(void)
 648 {
 649         while(fgets(line, sizeof(line), fin) != NULL) {
 650                 if (is_multipart_boundary(line))
 651                         return 1;
 652         }
 653         return 0;
 654 }
 655
 656 static int handle_boundary(void)
 657 {
 658         char newline[]="\n";
 659 again:
 660         if (!memcmp(line+content_top->boundary_len, "--", 2)) {
 661                 /* we hit an end boundary */
 662                 /* pop the current boundary off the stack */
 663                 free(content_top->boundary);
 664
 665                 /* technically won't happen as is_multipart_boundary()
 666                    will fail first.  But just in case..
 667                  */
 668                 if (content_top-- < content) {
 669                         fprintf(stderr, "Detected mismatched boundaries, "
 670                                         "can't recover\n");
 671                         exit(1);
 672                 }
 673                 handle_filter(newline, sizeof(newline), strlen(newline));
 674
 675                 /* skip to the next boundary */
 676                 if (!find_boundary())
 677                         return 0;
 678                 goto again;
 679         }
 680
 681         /* set some defaults */
 682         transfer_encoding = TE_DONTCARE;
 683         charset[0] = 0;
 684         message_type = TYPE_TEXT;
 685
 686         /* slurp in this section's info */
 687         while (read_one_header_line(line, sizeof(line), fin))
 688                 check_header(line, sizeof(line), p_hdr_data, 0);
 689
 690         /* eat the blank line after section info */
 691         return (fgets(line, sizeof(line), fin) != NULL);
 692 }
 693
 694 static inline int patchbreak(const char *line)
 695 {
 696         /* Beginning of a "diff -" header? */
 697         if (!memcmp("diff -", line, 6))
 698                 return 1;
 699
 700         /* CVS "Index: " line? */
 701         if (!memcmp("Index: ", line, 7))
 702                 return 1;
 703
 704         /*
 705          * "--- <filename>" starts patches without headers
 706          * "---<sp>*" is a manual separator
 707          */
 708         if (!memcmp("---", line, 3)) {
 709                 line += 3;
 710                 /* space followed by a filename? */
 711                 if (line[0] == ' ' && !isspace(line[1]))
 712                         return 1;
 713                 /* Just whitespace? */
 714                 for (;;) {
 715                         unsigned char c = *line++;
 716                         if (c == '\n')
 717                                 return 1;
 718                         if (!isspace(c))
 719                                 break;
 720                 }
 721                 return 0;
 722         }
 723         return 0;
 724 }
 725
 726
 727 static int handle_commit_msg(char *line, unsigned linesize)
 728 {
 729         static int still_looking = 1;
 730         char *endline = line + linesize;
 731
 732         if (!cmitmsg)
 733                 return 0;
 734
 735         if (still_looking) {
 736                 char *cp = line;
 737                 if (isspace(*line)) {
 738                         for (cp = line + 1; *cp; cp++) {
 739                                 if (!isspace(*cp))
 740                                         break;
 741                         }
 742                         if (!*cp)
 743                                 return 0;
 744                 }
 745                 if ((still_looking = check_header(cp, endline - cp, s_hdr_data, 0)) != 0)
 746                         return 0;
 747         }
 748
 749         /* normalize the log message to UTF-8. */
 750         if (metainfo_charset)
 751                 convert_to_utf8(line, endline - line, charset);
 752
 753         if (patchbreak(line)) {
 754                 fclose(cmitmsg);
 755                 cmitmsg = NULL;
 756                 return 1;
 757         }
 758
 759         fputs(line, cmitmsg);
 760         return 0;
 761 }
 762
 763 static int handle_patch(char *line, int len)
 764 {
 765         fwrite(line, 1, len, patchfile);
 766         patch_lines++;
 767         return 0;
 768 }
 769
 770 static int handle_filter(char *line, unsigned linesize, int linelen)
 771 {
 772         static int filter = 0;
 773
 774         /* filter tells us which part we left off on
 775          * a non-zero return indicates we hit a filter point
 776          */
 777         switch (filter) {
 778         case 0:
 779                 if (!handle_commit_msg(line, linesize))
 780                         break;
 781                 filter++;
 782         case 1:
 783                 if (!handle_patch(line, linelen))
 784                         break;
 785                 filter++;
 786         default:
 787                 return 1;
 788         }
 789
 790         return 0;
 791 }
 792
 793 static void handle_body(void)
 794 {
 795         int rc = 0;
 796         static char newline[2000];
 797         static char *np = newline;
 798         int len = strlen(line);
 799
 800         /* Skip up to the first boundary */
 801         if (content_top->boundary) {
 802                 if (!find_boundary())
 803                         return;
 804         }
 805
 806         do {
 807                 /* process any boundary lines */
 808                 if (content_top->boundary && is_multipart_boundary(line)) {
 809                         /* flush any leftover */
 810                         if (np != newline)
 811                                 handle_filter(newline, sizeof(newline),
 812                                               np - newline);
 813                         if (!handle_boundary())
 814                                 return;
 815                         len = strlen(line);
 816                 }
 817
 818                 /* Unwrap transfer encoding */
 819                 len = decode_transfer_encoding(line, sizeof(line), len);
 820                 if (len < 0) {
 821                         error("Malformed input line");
 822                         return;
 823                 }
 824
 825                 switch (transfer_encoding) {
 826                 case TE_BASE64:
 827                 case TE_QP:
 828                 {
 829                         char *op = line;
 830
 831                         /* binary data most likely doesn't have newlines */
 832                         if (message_type != TYPE_TEXT) {
 833                                 rc = handle_filter(line, sizeof(line), len);
 834                                 break;
 835                         }
 836
 837                         /*
 838                          * This is a decoded line that may contain
 839                          * multiple new lines.  Pass only one chunk
 840                          * at a time to handle_filter()
 841                          */
 842                         do {
 843                                 while (op < line + len && *op != '\n')
 844                                         *np++ = *op++;
 845                                 *np = *op;
 846                                 if (*np != 0) {
 847                                         /* should be sitting on a new line */
 848                                         *(++np) = 0;
 849                                         op++;
 850                                         rc = handle_filter(newline, sizeof(newline), np - newline);
 851                                         np = newline;
 852                                 }
 853                         } while (op < line + len);
 854                         /*
 855                          * The partial chunk is saved in newline and will be
 856                          * appended by the next iteration of read_line_with_nul().
 857                          */
 858                         break;
 859                 }
 860                 default:
 861                         rc = handle_filter(line, sizeof(line), len);
 862                 }
 863                 if (rc)
 864                         /* nothing left to filter */
 865                         break;
 866         } while ((len = read_line_with_nul(line, sizeof(line), fin)));
 867
 868         return;
 869 }
 870
 871 static void output_header_lines(FILE *fout, const char *hdr, char *data)
 872 {
 873         while (1) {
 874                 char *ep = strchr(data, '\n');
 875                 int len;
 876                 if (!ep)
 877                         len = strlen(data);
 878                 else
 879                         len = ep - data;
 880                 fprintf(fout, "%s: %.*s\n", hdr, len, data);
 881                 if (!ep)
 882                         break;
 883                 data = ep + 1;
 884         }
 885 }
 886
 887 static void handle_info(void)
 888 {
 889         char *sub;
 890         char *hdr;
 891         int i;
 892
 893         for (i = 0; header[i]; i++) {
 894
 895                 /* only print inbody headers if we output a patch file */
 896                 if (patch_lines && s_hdr_data[i])
 897                         hdr = s_hdr_data[i];
 898                 else if (p_hdr_data[i])
 899                         hdr = p_hdr_data[i];
 900                 else
 901                         continue;
 902
 903                 if (!memcmp(header[i], "Subject", 7)) {
 904                         if (keep_subject)
 905                                 sub = hdr;
 906                         else {
 907                                 sub = cleanup_subject(hdr);
 908                                 cleanup_space(sub);
 909                         }
 910                         output_header_lines(fout, "Subject", sub);
 911                 } else if (!memcmp(header[i], "From", 4)) {
 912                         handle_from(hdr);
 913                         fprintf(fout, "Author: %s\n", name);
 914                         fprintf(fout, "Email: %s\n", email);
 915                 } else {
 916                         cleanup_space(hdr);
 917                         fprintf(fout, "%s: %s\n", header[i], hdr);
 918                 }
 919         }
 920         fprintf(fout, "\n");
 921 }
 922
 923 static int mailinfo(FILE *in, FILE *out, int ks, const char *encoding,
 924                     const char *msg, const char *patch)
 925 {
 926         int peek;
 927         keep_subject = ks;
 928         metainfo_charset = encoding;
 929         fin = in;
 930         fout = out;
 931
 932         cmitmsg = fopen(msg, "w");
 933         if (!cmitmsg) {
 934                 perror(msg);
 935                 return -1;
 936         }
 937         patchfile = fopen(patch, "w");
 938         if (!patchfile) {
 939                 perror(patch);
 940                 fclose(cmitmsg);
 941                 return -1;
 942         }
 943
 944         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(char *));
 945         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(char *));
 946
 947         do {
 948                 peek = fgetc(in);
 949         } while (isspace(peek));
 950         ungetc(peek, in);
 951
 952         /* process the email header */
 953         while (read_one_header_line(line, sizeof(line), fin))
 954                 check_header(line, sizeof(line), p_hdr_data, 1);
 955
 956         handle_body();
 957         handle_info();
 958
 959         return 0;
 960 }
 961
 962 static const char mailinfo_usage[] =
 963         "git-mailinfo [-k] [-u | --encoding=<encoding>] msg patch <mail >info";
 964
 965 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
 966 {
 967         const char *def_charset;
 968
 969         /* NEEDSWORK: might want to do the optional .git/ directory
 970          * discovery
 971          */
 972         git_config(git_default_config, NULL);
 973
 974         def_charset = (git_commit_encoding ? git_commit_encoding : "utf-8");
 975         metainfo_charset = def_charset;
 976
 977         while (1 < argc && argv[1][0] == '-') {
 978                 if (!strcmp(argv[1], "-k"))
 979                         keep_subject = 1;
 980                 else if (!strcmp(argv[1], "-u"))
 981                         metainfo_charset = def_charset;
 982                 else if (!strcmp(argv[1], "-n"))
 983                         metainfo_charset = NULL;
 984                 else if (!prefixcmp(argv[1], "--encoding="))
 985                         metainfo_charset = argv[1] + 11;
 986                 else
 987                         usage(mailinfo_usage);
 988                 argc--; argv++;
 989         }
 990
 991         if (argc != 3)
 992                 usage(mailinfo_usage);
 993
 994         return !!mailinfo(stdin, stdout, keep_subject, metainfo_charset, argv[1], argv[2]);
 995 }