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], 7)) {
 338                                 if (!hdr_data[i])
 339                                         hdr_data[i] = xmalloc(linesize + 20);
 340                                 if (! handle_header(line, hdr_data[i], 0)) {
 341                                         return 1;
 342                                 }
 343                         }
 344                 }
 345         }
 346
 347         /* no match */
 348         return 0;
 349 }
 350
 351 static int is_rfc2822_header(char *line)
 352 {
 353         /*
 354          * The section that defines the loosest possible
 355          * field name is "3.6.8 Optional fields".
 356          *
 357          * optional-field = field-name ":" unstructured CRLF
 358          * field-name = 1*ftext
 359          * ftext = %d33-57 / %59-126
 360          */
 361         int ch;
 362         char *cp = line;
 363
 364         /* Count mbox From headers as headers */
 365         if (!memcmp(line, "From ", 5) || !memcmp(line, ">From ", 6))
 366                 return 1;
 367
 368         while ((ch = *cp++)) {
 369                 if (ch == ':')
 370                         return cp != line;
 371                 if ((33 <= ch && ch <= 57) ||
 372                     (59 <= ch && ch <= 126))
 373                         continue;
 374                 break;
 375         }
 376         return 0;
 377 }
 378
 379 /*
 380  * sz is size of 'line' buffer in bytes.  Must be reasonably
 381  * long enough to hold one physical real-world e-mail line.
 382  */
 383 static int read_one_header_line(char *line, int sz, FILE *in)
 384 {
 385         int len;
 386
 387         /*
 388          * We will read at most (sz-1) bytes and then potentially
 389          * re-add NUL after it.  Accessing line[sz] after this is safe
 390          * and we can allow len to grow up to and including sz.
 391          */
 392         sz--;
 393
 394         /* Get the first part of the line. */
 395         if (!fgets(line, sz, in))
 396                 return 0;
 397
 398         /*
 399          * Is it an empty line or not a valid rfc2822 header?
 400          * If so, stop here, and return false ("not a header")
 401          */
 402         len = eatspace(line);
 403         if (!len || !is_rfc2822_header(line)) {
 404                 /* Re-add the newline */
 405                 line[len] = '\n';
 406                 line[len + 1] = '\0';
 407                 return 0;
 408         }
 409
 410         /*
 411          * Now we need to eat all the continuation lines..
 412          * Yuck, 2822 header "folding"
 413          */
 414         for (;;) {
 415                 int peek, addlen;
 416                 static char continuation[1000];
 417
 418                 peek = fgetc(in); ungetc(peek, in);
 419                 if (peek != ' ' && peek != '\t')
 420                         break;
 421                 if (!fgets(continuation, sizeof(continuation), in))
 422                         break;
 423                 addlen = eatspace(continuation);
 424                 if (len < sz - 1) {
 425                         if (addlen >= sz - len)
 426                                 addlen = sz - len - 1;
 427                         memcpy(line + len, continuation, addlen);
 428                         line[len] = '\n';
 429                         len += addlen;
 430                 }
 431         }
 432         line[len] = 0;
 433
 434         return 1;
 435 }
 436
 437 static int decode_q_segment(char *in, char *ot, unsigned otsize, char *ep, int rfc2047)
 438 {
 439         char *otbegin = ot;
 440         char *otend = ot + otsize;
 441         int c;
 442         while ((c = *in++) != 0 && (in <= ep)) {
 443                 if (ot == otend) {
 444                         *--ot = '\0';
 445                         return -1;
 446                 }
 447                 if (c == '=') {
 448                         int d = *in++;
 449                         if (d == '\n' || !d)
 450                                 break; /* drop trailing newline */
 451                         *ot++ = ((hexval(d) << 4) | hexval(*in++));
 452                         continue;
 453                 }
 454                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 455                         c = 0x20;
 456                 *ot++ = c;
 457         }
 458         *ot = 0;
 459         return (ot - otbegin);
 460 }
 461
 462 static int decode_b_segment(char *in, char *ot, unsigned otsize, char *ep)
 463 {
 464         /* Decode in..ep, possibly in-place to ot */
 465         int c, pos = 0, acc = 0;
 466         char *otbegin = ot;
 467         char *otend = ot + otsize;
 468
 469         while ((c = *in++) != 0 && (in <= ep)) {
 470                 if (ot == otend) {
 471                         *--ot = '\0';
 472                         return -1;
 473                 }
 474                 if (c == '+')
 475                         c = 62;
 476                 else if (c == '/')
 477                         c = 63;
 478                 else if ('A' <= c && c <= 'Z')
 479                         c -= 'A';
 480                 else if ('a' <= c && c <= 'z')
 481                         c -= 'a' - 26;
 482                 else if ('0' <= c && c <= '9')
 483                         c -= '0' - 52;
 484                 else if (c == '=') {
 485                         /* padding is almost like (c == 0), except we do
 486                          * not output NUL resulting only from it;
 487                          * for now we just trust the data.
 488                          */
 489                         c = 0;
 490                 }
 491                 else
 492                         continue; /* garbage */
 493                 switch (pos++) {
 494                 case 0:
 495                         acc = (c << 2);
 496                         break;
 497                 case 1:
 498                         *ot++ = (acc | (c >> 4));
 499                         acc = (c & 15) << 4;
 500                         break;
 501                 case 2:
 502                         *ot++ = (acc | (c >> 2));
 503                         acc = (c & 3) << 6;
 504                         break;
 505                 case 3:
 506                         *ot++ = (acc | c);
 507                         acc = pos = 0;
 508                         break;
 509                 }
 510         }
 511         *ot = 0;
 512         return (ot - otbegin);
 513 }
 514
 515 /*
 516  * When there is no known charset, guess.
 517  *
 518  * Right now we assume that if the target is UTF-8 (the default),
 519  * and it already looks like UTF-8 (which includes US-ASCII as its
 520  * subset, of course) then that is what it is and there is nothing
 521  * to do.
 522  *
 523  * Otherwise, we default to assuming it is Latin1 for historical
 524  * reasons.
 525  */
 526 static const char *guess_charset(const char *line, const char *target_charset)
 527 {
 528         if (is_encoding_utf8(target_charset)) {
 529                 if (is_utf8(line))
 530                         return NULL;
 531         }
 532         return "latin1";
 533 }
 534
 535 static void convert_to_utf8(char *line, unsigned linesize, const char *charset)
 536 {
 537         char *out;
 538
 539         if (!charset || !*charset) {
 540                 charset = guess_charset(line, metainfo_charset);
 541                 if (!charset)
 542                         return;
 543         }
 544
 545         if (!strcmp(metainfo_charset, charset))
 546                 return;
 547         out = reencode_string(line, metainfo_charset, charset);
 548         if (!out)
 549                 die("cannot convert from %s to %s\n",
 550                     charset, metainfo_charset);
 551         strlcpy(line, out, linesize);
 552         free(out);
 553 }
 554
 555 static int decode_header_bq(char *it, unsigned itsize)
 556 {
 557         char *in, *out, *ep, *cp, *sp;
 558         char outbuf[1000];
 559         int rfc2047 = 0;
 560
 561         in = it;
 562         out = outbuf;
 563         while ((ep = strstr(in, "=?")) != NULL) {
 564                 int sz, encoding;
 565                 char charset_q[256], piecebuf[256];
 566                 rfc2047 = 1;
 567
 568                 if (in != ep) {
 569                         sz = ep - in;
 570                         memcpy(out, in, sz);
 571                         out += sz;
 572                         in += sz;
 573                 }
 574                 /* E.g.
 575                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 576                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 577                  */
 578                 ep += 2;
 579                 cp = strchr(ep, '?');
 580                 if (!cp)
 581                         return rfc2047; /* no munging */
 582                 for (sp = ep; sp < cp; sp++)
 583                         charset_q[sp - ep] = tolower(*sp);
 584                 charset_q[cp - ep] = 0;
 585                 encoding = cp[1];
 586                 if (!encoding || cp[2] != '?')
 587                         return rfc2047; /* no munging */
 588                 ep = strstr(cp + 3, "?=");
 589                 if (!ep)
 590                         return rfc2047; /* no munging */
 591                 switch (tolower(encoding)) {
 592                 default:
 593                         return rfc2047; /* no munging */
 594                 case 'b':
 595                         sz = decode_b_segment(cp + 3, piecebuf, sizeof(piecebuf), ep);
 596                         break;
 597                 case 'q':
 598                         sz = decode_q_segment(cp + 3, piecebuf, sizeof(piecebuf), ep, 1);
 599                         break;
 600                 }
 601                 if (sz < 0)
 602                         return rfc2047;
 603                 if (metainfo_charset)
 604                         convert_to_utf8(piecebuf, sizeof(piecebuf), charset_q);
 605
 606                 sz = strlen(piecebuf);
 607                 if (outbuf + sizeof(outbuf) <= out + sz)
 608                         return rfc2047; /* no munging */
 609                 strcpy(out, piecebuf);
 610                 out += sz;
 611                 in = ep + 2;
 612         }
 613         strcpy(out, in);
 614         strlcpy(it, outbuf, itsize);
 615         return rfc2047;
 616 }
 617
 618 static void decode_header(char *it, unsigned itsize)
 619 {
 620
 621         if (decode_header_bq(it, itsize))
 622                 return;
 623         /* otherwise "it" is a straight copy of the input.
 624          * This can be binary guck but there is no charset specified.
 625          */
 626         if (metainfo_charset)
 627                 convert_to_utf8(it, itsize, "");
 628 }
 629
 630 static int decode_transfer_encoding(char *line, unsigned linesize, int inputlen)
 631 {
 632         char *ep;
 633
 634         switch (transfer_encoding) {
 635         case TE_QP:
 636                 ep = line + inputlen;
 637                 return decode_q_segment(line, line, linesize, ep, 0);
 638         case TE_BASE64:
 639                 ep = line + inputlen;
 640                 return decode_b_segment(line, line, linesize, ep);
 641         case TE_DONTCARE:
 642         default:
 643                 return inputlen;
 644         }
 645 }
 646
 647 static int handle_filter(char *line, unsigned linesize, int linelen);
 648
 649 static int find_boundary(void)
 650 {
 651         while(fgets(line, sizeof(line), fin) != NULL) {
 652                 if (is_multipart_boundary(line))
 653                         return 1;
 654         }
 655         return 0;
 656 }
 657
 658 static int handle_boundary(void)
 659 {
 660         char newline[]="\n";
 661 again:
 662         if (!memcmp(line+content_top->boundary_len, "--", 2)) {
 663                 /* we hit an end boundary */
 664                 /* pop the current boundary off the stack */
 665                 free(content_top->boundary);
 666
 667                 /* technically won't happen as is_multipart_boundary()
 668                    will fail first.  But just in case..
 669                  */
 670                 if (content_top-- < content) {
 671                         fprintf(stderr, "Detected mismatched boundaries, "
 672                                         "can't recover\n");
 673                         exit(1);
 674                 }
 675                 handle_filter(newline, sizeof(newline), strlen(newline));
 676
 677                 /* skip to the next boundary */
 678                 if (!find_boundary())
 679                         return 0;
 680                 goto again;
 681         }
 682
 683         /* set some defaults */
 684         transfer_encoding = TE_DONTCARE;
 685         charset[0] = 0;
 686         message_type = TYPE_TEXT;
 687
 688         /* slurp in this section's info */
 689         while (read_one_header_line(line, sizeof(line), fin))
 690                 check_header(line, sizeof(line), p_hdr_data, 0);
 691
 692         /* eat the blank line after section info */
 693         return (fgets(line, sizeof(line), fin) != NULL);
 694 }
 695
 696 static inline int patchbreak(const char *line)
 697 {
 698         /* Beginning of a "diff -" header? */
 699         if (!memcmp("diff -", line, 6))
 700                 return 1;
 701
 702         /* CVS "Index: " line? */
 703         if (!memcmp("Index: ", line, 7))
 704                 return 1;
 705
 706         /*
 707          * "--- <filename>" starts patches without headers
 708          * "---<sp>*" is a manual separator
 709          */
 710         if (!memcmp("---", line, 3)) {
 711                 line += 3;
 712                 /* space followed by a filename? */
 713                 if (line[0] == ' ' && !isspace(line[1]))
 714                         return 1;
 715                 /* Just whitespace? */
 716                 for (;;) {
 717                         unsigned char c = *line++;
 718                         if (c == '\n')
 719                                 return 1;
 720                         if (!isspace(c))
 721                                 break;
 722                 }
 723                 return 0;
 724         }
 725         return 0;
 726 }
 727
 728
 729 static int handle_commit_msg(char *line, unsigned linesize)
 730 {
 731         static int still_looking = 1;
 732         char *endline = line + linesize;
 733
 734         if (!cmitmsg)
 735                 return 0;
 736
 737         if (still_looking) {
 738                 char *cp = line;
 739                 if (isspace(*line)) {
 740                         for (cp = line + 1; *cp; cp++) {
 741                                 if (!isspace(*cp))
 742                                         break;
 743                         }
 744                         if (!*cp)
 745                                 return 0;
 746                 }
 747                 if ((still_looking = check_header(cp, endline - cp, s_hdr_data, 0)) != 0)
 748                         return 0;
 749         }
 750
 751         /* normalize the log message to UTF-8. */
 752         if (metainfo_charset)
 753                 convert_to_utf8(line, endline - line, charset);
 754
 755         if (patchbreak(line)) {
 756                 fclose(cmitmsg);
 757                 cmitmsg = NULL;
 758                 return 1;
 759         }
 760
 761         fputs(line, cmitmsg);
 762         return 0;
 763 }
 764
 765 static int handle_patch(char *line, int len)
 766 {
 767         fwrite(line, 1, len, patchfile);
 768         patch_lines++;
 769         return 0;
 770 }
 771
 772 static int handle_filter(char *line, unsigned linesize, int linelen)
 773 {
 774         static int filter = 0;
 775
 776         /* filter tells us which part we left off on
 777          * a non-zero return indicates we hit a filter point
 778          */
 779         switch (filter) {
 780         case 0:
 781                 if (!handle_commit_msg(line, linesize))
 782                         break;
 783                 filter++;
 784         case 1:
 785                 if (!handle_patch(line, linelen))
 786                         break;
 787                 filter++;
 788         default:
 789                 return 1;
 790         }
 791
 792         return 0;
 793 }
 794
 795 static void handle_body(void)
 796 {
 797         int rc = 0;
 798         static char newline[2000];
 799         static char *np = newline;
 800         int len = strlen(line);
 801
 802         /* Skip up to the first boundary */
 803         if (content_top->boundary) {
 804                 if (!find_boundary())
 805                         return;
 806         }
 807
 808         do {
 809                 /* process any boundary lines */
 810                 if (content_top->boundary && is_multipart_boundary(line)) {
 811                         /* flush any leftover */
 812                         if (np != newline)
 813                                 handle_filter(newline, sizeof(newline),
 814                                               np - newline);
 815                         if (!handle_boundary())
 816                                 return;
 817                         len = strlen(line);
 818                 }
 819
 820                 /* Unwrap transfer encoding */
 821                 len = decode_transfer_encoding(line, sizeof(line), len);
 822                 if (len < 0) {
 823                         error("Malformed input line");
 824                         return;
 825                 }
 826
 827                 switch (transfer_encoding) {
 828                 case TE_BASE64:
 829                 case TE_QP:
 830                 {
 831                         char *op = line;
 832
 833                         /* binary data most likely doesn't have newlines */
 834                         if (message_type != TYPE_TEXT) {
 835                                 rc = handle_filter(line, sizeof(line), len);
 836                                 break;
 837                         }
 838
 839                         /*
 840                          * This is a decoded line that may contain
 841                          * multiple new lines.  Pass only one chunk
 842                          * at a time to handle_filter()
 843                          */
 844                         do {
 845                                 while (op < line + len && *op != '\n')
 846                                         *np++ = *op++;
 847                                 *np = *op;
 848                                 if (*np != 0) {
 849                                         /* should be sitting on a new line */
 850                                         *(++np) = 0;
 851                                         op++;
 852                                         rc = handle_filter(newline, sizeof(newline), np - newline);
 853                                         np = newline;
 854                                 }
 855                         } while (op < line + len);
 856                         /*
 857                          * The partial chunk is saved in newline and will be
 858                          * appended by the next iteration of read_line_with_nul().
 859                          */
 860                         break;
 861                 }
 862                 default:
 863                         rc = handle_filter(line, sizeof(line), len);
 864                 }
 865                 if (rc)
 866                         /* nothing left to filter */
 867                         break;
 868         } while ((len = read_line_with_nul(line, sizeof(line), fin)));
 869
 870         return;
 871 }
 872
 873 static void output_header_lines(FILE *fout, const char *hdr, char *data)
 874 {
 875         while (1) {
 876                 char *ep = strchr(data, '\n');
 877                 int len;
 878                 if (!ep)
 879                         len = strlen(data);
 880                 else
 881                         len = ep - data;
 882                 fprintf(fout, "%s: %.*s\n", hdr, len, data);
 883                 if (!ep)
 884                         break;
 885                 data = ep + 1;
 886         }
 887 }
 888
 889 static void handle_info(void)
 890 {
 891         char *sub;
 892         char *hdr;
 893         int i;
 894
 895         for (i = 0; header[i]; i++) {
 896
 897                 /* only print inbody headers if we output a patch file */
 898                 if (patch_lines && s_hdr_data[i])
 899                         hdr = s_hdr_data[i];
 900                 else if (p_hdr_data[i])
 901                         hdr = p_hdr_data[i];
 902                 else
 903                         continue;
 904
 905                 if (!memcmp(header[i], "Subject", 7)) {
 906                         if (keep_subject)
 907                                 sub = hdr;
 908                         else {
 909                                 sub = cleanup_subject(hdr);
 910                                 cleanup_space(sub);
 911                         }
 912                         output_header_lines(fout, "Subject", sub);
 913                 } else if (!memcmp(header[i], "From", 4)) {
 914                         handle_from(hdr);
 915                         fprintf(fout, "Author: %s\n", name);
 916                         fprintf(fout, "Email: %s\n", email);
 917                 } else {
 918                         cleanup_space(hdr);
 919                         fprintf(fout, "%s: %s\n", header[i], hdr);
 920                 }
 921         }
 922         fprintf(fout, "\n");
 923 }
 924
 925 static int mailinfo(FILE *in, FILE *out, int ks, const char *encoding,
 926                     const char *msg, const char *patch)
 927 {
 928         int peek;
 929         keep_subject = ks;
 930         metainfo_charset = encoding;
 931         fin = in;
 932         fout = out;
 933
 934         cmitmsg = fopen(msg, "w");
 935         if (!cmitmsg) {
 936                 perror(msg);
 937                 return -1;
 938         }
 939         patchfile = fopen(patch, "w");
 940         if (!patchfile) {
 941                 perror(patch);
 942                 fclose(cmitmsg);
 943                 return -1;
 944         }
 945
 946         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(char *));
 947         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(char *));
 948
 949         do {
 950                 peek = fgetc(in);
 951         } while (isspace(peek));
 952         ungetc(peek, in);
 953
 954         /* process the email header */
 955         while (read_one_header_line(line, sizeof(line), fin))
 956                 check_header(line, sizeof(line), p_hdr_data, 1);
 957
 958         handle_body();
 959         handle_info();
 960
 961         return 0;
 962 }
 963
 964 static const char mailinfo_usage[] =
 965         "git-mailinfo [-k] [-u | --encoding=<encoding> | -n] msg patch <mail >info";
 966
 967 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
 968 {
 969         const char *def_charset;
 970
 971         /* NEEDSWORK: might want to do the optional .git/ directory
 972          * discovery
 973          */
 974         git_config(git_default_config, NULL);
 975
 976         def_charset = (git_commit_encoding ? git_commit_encoding : "utf-8");
 977         metainfo_charset = def_charset;
 978
 979         while (1 < argc && argv[1][0] == '-') {
 980                 if (!strcmp(argv[1], "-k"))
 981                         keep_subject = 1;
 982                 else if (!strcmp(argv[1], "-u"))
 983                         metainfo_charset = def_charset;
 984                 else if (!strcmp(argv[1], "-n"))
 985                         metainfo_charset = NULL;
 986                 else if (!prefixcmp(argv[1], "--encoding="))
 987                         metainfo_charset = argv[1] + 11;
 988                 else
 989                         usage(mailinfo_usage);
 990                 argc--; argv++;
 991         }
 992
 993         if (argc != 3)
 994                 usage(mailinfo_usage);
 995
 996         return !!mailinfo(stdin, stdout, keep_subject, metainfo_charset, argv[1], argv[2]);
 997 }