mailinfo.c

   1 #include "cache.h"
   2 #include "config.h"
   3 #include "utf8.h"
   4 #include "strbuf.h"
   5 #include "mailinfo.h"
   6
   7 static void cleanup_space(struct strbuf *sb)
   8 {
   9         size_t pos, cnt;
  10         for (pos = 0; pos < sb->len; pos++) {
  11                 if (isspace(sb->buf[pos])) {
  12                         sb->buf[pos] = ' ';
  13                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
  14                         strbuf_remove(sb, pos + 1, cnt);
  15                 }
  16         }
  17 }
  18
  19 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  20 {
  21         struct strbuf *src = name;
  22         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  23                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  24                 src = email;
  25         else if (name == out)
  26                 return;
  27         strbuf_reset(out);
  28         strbuf_addbuf(out, src);
  29 }
  30
  31 static void parse_bogus_from(struct mailinfo *mi, const struct strbuf *line)
  32 {
  33         /* John Doe <johndoe> */
  34
  35         char *bra, *ket;
  36         /* This is fallback, so do not bother if we already have an
  37          * e-mail address.
  38          */
  39         if (mi->email.len)
  40                 return;
  41
  42         bra = strchr(line->buf, '<');
  43         if (!bra)
  44                 return;
  45         ket = strchr(bra, '>');
  46         if (!ket)
  47                 return;
  48
  49         strbuf_reset(&mi->email);
  50         strbuf_add(&mi->email, bra + 1, ket - bra - 1);
  51
  52         strbuf_reset(&mi->name);
  53         strbuf_add(&mi->name, line->buf, bra - line->buf);
  54         strbuf_trim(&mi->name);
  55         get_sane_name(&mi->name, &mi->name, &mi->email);
  56 }
  57
  58 static const char *unquote_comment(struct strbuf *outbuf, const char *in)
  59 {
  60         int c;
  61         int take_next_literally = 0;
  62
  63         strbuf_addch(outbuf, '(');
  64
  65         while ((c = *in++) != 0) {
  66                 if (take_next_literally == 1) {
  67                         take_next_literally = 0;
  68                 } else {
  69                         switch (c) {
  70                         case '\\':
  71                                 take_next_literally = 1;
  72                                 continue;
  73                         case '(':
  74                                 in = unquote_comment(outbuf, in);
  75                                 continue;
  76                         case ')':
  77                                 strbuf_addch(outbuf, ')');
  78                                 return in;
  79                         }
  80                 }
  81
  82                 strbuf_addch(outbuf, c);
  83         }
  84
  85         return in;
  86 }
  87
  88 static const char *unquote_quoted_string(struct strbuf *outbuf, const char *in)
  89 {
  90         int c;
  91         int take_next_literally = 0;
  92
  93         while ((c = *in++) != 0) {
  94                 if (take_next_literally == 1) {
  95                         take_next_literally = 0;
  96                 } else {
  97                         switch (c) {
  98                         case '\\':
  99                                 take_next_literally = 1;
 100                                 continue;
 101                         case '"':
 102                                 return in;
 103                         }
 104                 }
 105
 106                 strbuf_addch(outbuf, c);
 107         }
 108
 109         return in;
 110 }
 111
 112 static void unquote_quoted_pair(struct strbuf *line)
 113 {
 114         struct strbuf outbuf;
 115         const char *in = line->buf;
 116         int c;
 117
 118         strbuf_init(&outbuf, line->len);
 119
 120         while ((c = *in++) != 0) {
 121                 switch (c) {
 122                 case '"':
 123                         in = unquote_quoted_string(&outbuf, in);
 124                         continue;
 125                 case '(':
 126                         in = unquote_comment(&outbuf, in);
 127                         continue;
 128                 }
 129
 130                 strbuf_addch(&outbuf, c);
 131         }
 132
 133         strbuf_swap(&outbuf, line);
 134         strbuf_release(&outbuf);
 135
 136 }
 137
 138 static void handle_from(struct mailinfo *mi, const struct strbuf *from)
 139 {
 140         char *at;
 141         size_t el;
 142         struct strbuf f;
 143
 144         strbuf_init(&f, from->len);
 145         strbuf_addbuf(&f, from);
 146
 147         unquote_quoted_pair(&f);
 148
 149         at = strchr(f.buf, '@');
 150         if (!at) {
 151                 parse_bogus_from(mi, from);
 152                 return;
 153         }
 154
 155         /*
 156          * If we already have one email, don't take any confusing lines
 157          */
 158         if (mi->email.len && strchr(at + 1, '@')) {
 159                 strbuf_release(&f);
 160                 return;
 161         }
 162
 163         /* Pick up the string around '@', possibly delimited with <>
 164          * pair; that is the email part.
 165          */
 166         while (at > f.buf) {
 167                 char c = at[-1];
 168                 if (isspace(c))
 169                         break;
 170                 if (c == '<') {
 171                         at[-1] = ' ';
 172                         break;
 173                 }
 174                 at--;
 175         }
 176         el = strcspn(at, " \n\t\r\v\f>");
 177         strbuf_reset(&mi->email);
 178         strbuf_add(&mi->email, at, el);
 179         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 180
 181         /* The remainder is name.  It could be
 182          *
 183          * - "John Doe <john.doe@xz>"                   (a), or
 184          * - "john.doe@xz (John Doe)"                   (b), or
 185          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 186          *
 187          * but we have removed the email part, so
 188          *
 189          * - remove extra spaces which could stay after email (case 'c'), and
 190          * - trim from both ends, possibly removing the () pair at the end
 191          *   (cases 'a' and 'b').
 192          */
 193         cleanup_space(&f);
 194         strbuf_trim(&f);
 195         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 196                 strbuf_remove(&f, 0, 1);
 197                 strbuf_setlen(&f, f.len - 1);
 198         }
 199
 200         get_sane_name(&mi->name, &f, &mi->email);
 201         strbuf_release(&f);
 202 }
 203
 204 static void handle_header(struct strbuf **out, const struct strbuf *line)
 205 {
 206         if (!*out) {
 207                 *out = xmalloc(sizeof(struct strbuf));
 208                 strbuf_init(*out, line->len);
 209         } else
 210                 strbuf_reset(*out);
 211
 212         strbuf_addbuf(*out, line);
 213 }
 214
 215 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 216  * to have enough heuristics to grok MIME encoded patches often found
 217  * on our mailing lists.  For example, we do not even treat header lines
 218  * case insensitively.
 219  */
 220
 221 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 222 {
 223         const char *ends, *ap = strcasestr(line, name);
 224         size_t sz;
 225
 226         strbuf_setlen(attr, 0);
 227         if (!ap)
 228                 return 0;
 229         ap += strlen(name);
 230         if (*ap == '"') {
 231                 ap++;
 232                 ends = "\"";
 233         }
 234         else
 235                 ends = "; \t";
 236         sz = strcspn(ap, ends);
 237         strbuf_add(attr, ap, sz);
 238         return 1;
 239 }
 240
 241 static void handle_content_type(struct mailinfo *mi, struct strbuf *line)
 242 {
 243         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 244         strbuf_init(boundary, line->len);
 245
 246         if (slurp_attr(line->buf, "boundary=", boundary)) {
 247                 strbuf_insert(boundary, 0, "--", 2);
 248                 if (++mi->content_top >= &mi->content[MAX_BOUNDARIES]) {
 249                         error("Too many boundaries to handle");
 250                         mi->input_error = -1;
 251                         mi->content_top = &mi->content[MAX_BOUNDARIES] - 1;
 252                         return;
 253                 }
 254                 *(mi->content_top) = boundary;
 255                 boundary = NULL;
 256         }
 257         slurp_attr(line->buf, "charset=", &mi->charset);
 258
 259         if (boundary) {
 260                 strbuf_release(boundary);
 261                 free(boundary);
 262         }
 263 }
 264
 265 static void handle_content_transfer_encoding(struct mailinfo *mi,
 266                                              const struct strbuf *line)
 267 {
 268         if (strcasestr(line->buf, "base64"))
 269                 mi->transfer_encoding = TE_BASE64;
 270         else if (strcasestr(line->buf, "quoted-printable"))
 271                 mi->transfer_encoding = TE_QP;
 272         else
 273                 mi->transfer_encoding = TE_DONTCARE;
 274 }
 275
 276 static int is_multipart_boundary(struct mailinfo *mi, const struct strbuf *line)
 277 {
 278         struct strbuf *content_top = *(mi->content_top);
 279
 280         return ((content_top->len <= line->len) &&
 281                 !memcmp(line->buf, content_top->buf, content_top->len));
 282 }
 283
 284 static void cleanup_subject(struct mailinfo *mi, struct strbuf *subject)
 285 {
 286         size_t at = 0;
 287
 288         while (at < subject->len) {
 289                 char *pos;
 290                 size_t remove;
 291
 292                 switch (subject->buf[at]) {
 293                 case 'r': case 'R':
 294                         if (subject->len <= at + 3)
 295                                 break;
 296                         if ((subject->buf[at + 1] == 'e' ||
 297                              subject->buf[at + 1] == 'E') &&
 298                             subject->buf[at + 2] == ':') {
 299                                 strbuf_remove(subject, at, 3);
 300                                 continue;
 301                         }
 302                         at++;
 303                         break;
 304                 case ' ': case '\t': case ':':
 305                         strbuf_remove(subject, at, 1);
 306                         continue;
 307                 case '[':
 308                         pos = strchr(subject->buf + at, ']');
 309                         if (!pos)
 310                                 break;
 311                         remove = pos - subject->buf + at + 1;
 312                         if (!mi->keep_non_patch_brackets_in_subject ||
 313                             (7 <= remove &&
 314                              memmem(subject->buf + at, remove, "PATCH", 5)))
 315                                 strbuf_remove(subject, at, remove);
 316                         else {
 317                                 at += remove;
 318                                 /*
 319                                  * If the input had a space after the ], keep
 320                                  * it.  We don't bother with finding the end of
 321                                  * the space, since we later normalize it
 322                                  * anyway.
 323                                  */
 324                                 if (isspace(subject->buf[at]))
 325                                         at += 1;
 326                         }
 327                         continue;
 328                 }
 329                 break;
 330         }
 331         strbuf_trim(subject);
 332 }
 333
 334 #define MAX_HDR_PARSED 10
 335 static const char *header[MAX_HDR_PARSED] = {
 336         "From","Subject","Date",
 337 };
 338
 339 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 340 {
 341         int len = strlen(hdr);
 342         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 343                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 344 }
 345
 346 static int is_format_patch_separator(const char *line, int len)
 347 {
 348         static const char SAMPLE[] =
 349                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 350         const char *cp;
 351
 352         if (len != strlen(SAMPLE))
 353                 return 0;
 354         if (!skip_prefix(line, "From ", &cp))
 355                 return 0;
 356         if (strspn(cp, "0123456789abcdef") != 40)
 357                 return 0;
 358         cp += 40;
 359         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 360 }
 361
 362 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 363 {
 364         const char *in = q_seg->buf;
 365         int c;
 366         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 367         strbuf_init(out, q_seg->len);
 368
 369         while ((c = *in++) != 0) {
 370                 if (c == '=') {
 371                         int ch, d = *in;
 372                         if (d == '\n' || !d)
 373                                 break; /* drop trailing newline */
 374                         ch = hex2chr(in);
 375                         if (ch >= 0) {
 376                                 strbuf_addch(out, ch);
 377                                 in += 2;
 378                                 continue;
 379                         }
 380                         /* garbage -- fall through */
 381                 }
 382                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 383                         c = 0x20;
 384                 strbuf_addch(out, c);
 385         }
 386         return out;
 387 }
 388
 389 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 390 {
 391         /* Decode in..ep, possibly in-place to ot */
 392         int c, pos = 0, acc = 0;
 393         const char *in = b_seg->buf;
 394         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 395         strbuf_init(out, b_seg->len);
 396
 397         while ((c = *in++) != 0) {
 398                 if (c == '+')
 399                         c = 62;
 400                 else if (c == '/')
 401                         c = 63;
 402                 else if ('A' <= c && c <= 'Z')
 403                         c -= 'A';
 404                 else if ('a' <= c && c <= 'z')
 405                         c -= 'a' - 26;
 406                 else if ('0' <= c && c <= '9')
 407                         c -= '0' - 52;
 408                 else
 409                         continue; /* garbage */
 410                 switch (pos++) {
 411                 case 0:
 412                         acc = (c << 2);
 413                         break;
 414                 case 1:
 415                         strbuf_addch(out, (acc | (c >> 4)));
 416                         acc = (c & 15) << 4;
 417                         break;
 418                 case 2:
 419                         strbuf_addch(out, (acc | (c >> 2)));
 420                         acc = (c & 3) << 6;
 421                         break;
 422                 case 3:
 423                         strbuf_addch(out, (acc | c));
 424                         acc = pos = 0;
 425                         break;
 426                 }
 427         }
 428         return out;
 429 }
 430
 431 static int convert_to_utf8(struct mailinfo *mi,
 432                            struct strbuf *line, const char *charset)
 433 {
 434         char *out;
 435
 436         if (!mi->metainfo_charset || !charset || !*charset)
 437                 return 0;
 438
 439         if (same_encoding(mi->metainfo_charset, charset))
 440                 return 0;
 441         out = reencode_string(line->buf, mi->metainfo_charset, charset);
 442         if (!out) {
 443                 mi->input_error = -1;
 444                 return error("cannot convert from %s to %s",
 445                              charset, mi->metainfo_charset);
 446         }
 447         strbuf_attach(line, out, strlen(out), strlen(out));
 448         return 0;
 449 }
 450
 451 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 452 {
 453         char *in, *ep, *cp;
 454         struct strbuf outbuf = STRBUF_INIT, *dec;
 455         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 456         int found_error = 1; /* pessimism */
 457
 458         in = it->buf;
 459         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 460                 int encoding;
 461                 strbuf_reset(&charset_q);
 462                 strbuf_reset(&piecebuf);
 463
 464                 if (in != ep) {
 465                         /*
 466                          * We are about to process an encoded-word
 467                          * that begins at ep, but there is something
 468                          * before the encoded word.
 469                          */
 470                         char *scan;
 471                         for (scan = in; scan < ep; scan++)
 472                                 if (!isspace(*scan))
 473                                         break;
 474
 475                         if (scan != ep || in == it->buf) {
 476                                 /*
 477                                  * We should not lose that "something",
 478                                  * unless we have just processed an
 479                                  * encoded-word, and there is only LWS
 480                                  * before the one we are about to process.
 481                                  */
 482                                 strbuf_add(&outbuf, in, ep - in);
 483                         }
 484                 }
 485                 /* E.g.
 486                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 487                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 488                  */
 489                 ep += 2;
 490
 491                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 492                         goto release_return;
 493
 494                 if (cp + 3 - it->buf > it->len)
 495                         goto release_return;
 496                 strbuf_add(&charset_q, ep, cp - ep);
 497
 498                 encoding = cp[1];
 499                 if (!encoding || cp[2] != '?')
 500                         goto release_return;
 501                 ep = strstr(cp + 3, "?=");
 502                 if (!ep)
 503                         goto release_return;
 504                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 505                 switch (tolower(encoding)) {
 506                 default:
 507                         goto release_return;
 508                 case 'b':
 509                         dec = decode_b_segment(&piecebuf);
 510                         break;
 511                 case 'q':
 512                         dec = decode_q_segment(&piecebuf, 1);
 513                         break;
 514                 }
 515                 if (convert_to_utf8(mi, dec, charset_q.buf))
 516                         goto release_return;
 517
 518                 strbuf_addbuf(&outbuf, dec);
 519                 strbuf_release(dec);
 520                 free(dec);
 521                 in = ep + 2;
 522         }
 523         strbuf_addstr(&outbuf, in);
 524         strbuf_reset(it);
 525         strbuf_addbuf(it, &outbuf);
 526         found_error = 0;
 527 release_return:
 528         strbuf_release(&outbuf);
 529         strbuf_release(&charset_q);
 530         strbuf_release(&piecebuf);
 531
 532         if (found_error)
 533                 mi->input_error = -1;
 534 }
 535
 536 static int check_header(struct mailinfo *mi,
 537                         const struct strbuf *line,
 538                         struct strbuf *hdr_data[], int overwrite)
 539 {
 540         int i, ret = 0, len;
 541         struct strbuf sb = STRBUF_INIT;
 542
 543         /* search for the interesting parts */
 544         for (i = 0; header[i]; i++) {
 545                 int len = strlen(header[i]);
 546                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 547                         /* Unwrap inline B and Q encoding, and optionally
 548                          * normalize the meta information to utf8.
 549                          */
 550                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 551                         decode_header(mi, &sb);
 552                         handle_header(&hdr_data[i], &sb);
 553                         ret = 1;
 554                         goto check_header_out;
 555                 }
 556         }
 557
 558         /* Content stuff */
 559         if (cmp_header(line, "Content-Type")) {
 560                 len = strlen("Content-Type: ");
 561                 strbuf_add(&sb, line->buf + len, line->len - len);
 562                 decode_header(mi, &sb);
 563                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 564                 handle_content_type(mi, &sb);
 565                 ret = 1;
 566                 goto check_header_out;
 567         }
 568         if (cmp_header(line, "Content-Transfer-Encoding")) {
 569                 len = strlen("Content-Transfer-Encoding: ");
 570                 strbuf_add(&sb, line->buf + len, line->len - len);
 571                 decode_header(mi, &sb);
 572                 handle_content_transfer_encoding(mi, &sb);
 573                 ret = 1;
 574                 goto check_header_out;
 575         }
 576         if (cmp_header(line, "Message-Id")) {
 577                 len = strlen("Message-Id: ");
 578                 strbuf_add(&sb, line->buf + len, line->len - len);
 579                 decode_header(mi, &sb);
 580                 if (mi->add_message_id)
 581                         mi->message_id = strbuf_detach(&sb, NULL);
 582                 ret = 1;
 583                 goto check_header_out;
 584         }
 585
 586 check_header_out:
 587         strbuf_release(&sb);
 588         return ret;
 589 }
 590
 591 /*
 592  * Returns 1 if the given line or any line beginning with the given line is an
 593  * in-body header (that is, check_header will succeed when passed
 594  * mi->s_hdr_data).
 595  */
 596 static int is_inbody_header(const struct mailinfo *mi,
 597                             const struct strbuf *line)
 598 {
 599         int i;
 600         for (i = 0; header[i]; i++)
 601                 if (!mi->s_hdr_data[i] && cmp_header(line, header[i]))
 602                         return 1;
 603         return 0;
 604 }
 605
 606 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 607 {
 608         struct strbuf *ret;
 609
 610         switch (mi->transfer_encoding) {
 611         case TE_QP:
 612                 ret = decode_q_segment(line, 0);
 613                 break;
 614         case TE_BASE64:
 615                 ret = decode_b_segment(line);
 616                 break;
 617         case TE_DONTCARE:
 618         default:
 619                 return;
 620         }
 621         strbuf_reset(line);
 622         strbuf_addbuf(line, ret);
 623         strbuf_release(ret);
 624         free(ret);
 625 }
 626
 627 static inline int patchbreak(const struct strbuf *line)
 628 {
 629         size_t i;
 630
 631         /* Beginning of a "diff -" header? */
 632         if (starts_with(line->buf, "diff -"))
 633                 return 1;
 634
 635         /* CVS "Index: " line? */
 636         if (starts_with(line->buf, "Index: "))
 637                 return 1;
 638
 639         /*
 640          * "--- <filename>" starts patches without headers
 641          * "---<sp>*" is a manual separator
 642          */
 643         if (line->len < 4)
 644                 return 0;
 645
 646         if (starts_with(line->buf, "---")) {
 647                 /* space followed by a filename? */
 648                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 649                         return 1;
 650                 /* Just whitespace? */
 651                 for (i = 3; i < line->len; i++) {
 652                         unsigned char c = line->buf[i];
 653                         if (c == '\n')
 654                                 return 1;
 655                         if (!isspace(c))
 656                                 break;
 657                 }
 658                 return 0;
 659         }
 660         return 0;
 661 }
 662
 663 static int is_scissors_line(const char *line)
 664 {
 665         const char *c;
 666         int scissors = 0, gap = 0;
 667         const char *first_nonblank = NULL, *last_nonblank = NULL;
 668         int visible, perforation = 0, in_perforation = 0;
 669
 670         for (c = line; *c; c++) {
 671                 if (isspace(*c)) {
 672                         if (in_perforation) {
 673                                 perforation++;
 674                                 gap++;
 675                         }
 676                         continue;
 677                 }
 678                 last_nonblank = c;
 679                 if (first_nonblank == NULL)
 680                         first_nonblank = c;
 681                 if (*c == '-') {
 682                         in_perforation = 1;
 683                         perforation++;
 684                         continue;
 685                 }
 686                 if ((!memcmp(c, ">8", 2) || !memcmp(c, "8<", 2) ||
 687                      !memcmp(c, ">%", 2) || !memcmp(c, "%<", 2))) {
 688                         in_perforation = 1;
 689                         perforation += 2;
 690                         scissors += 2;
 691                         c++;
 692                         continue;
 693                 }
 694                 in_perforation = 0;
 695         }
 696
 697         /*
 698          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 699          * Even though there can be arbitrary cruft on the same line
 700          * (e.g. "cut here"), in order to avoid misidentification, the
 701          * perforation must occupy more than a third of the visible
 702          * width of the line, and dashes and scissors must occupy more
 703          * than half of the perforation.
 704          */
 705
 706         if (first_nonblank && last_nonblank)
 707                 visible = last_nonblank - first_nonblank + 1;
 708         else
 709                 visible = 0;
 710         return (scissors && 8 <= visible &&
 711                 visible < perforation * 3 &&
 712                 gap * 2 < perforation);
 713 }
 714
 715 static void flush_inbody_header_accum(struct mailinfo *mi)
 716 {
 717         if (!mi->inbody_header_accum.len)
 718                 return;
 719         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 720                 die("BUG: inbody_header_accum, if not empty, must always contain a valid in-body header");
 721         strbuf_reset(&mi->inbody_header_accum);
 722 }
 723
 724 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 725 {
 726         if (mi->inbody_header_accum.len &&
 727             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 728                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 729                         /*
 730                          * This is a scissors line; do not consider this line
 731                          * as a header continuation line.
 732                          */
 733                         flush_inbody_header_accum(mi);
 734                         return 0;
 735                 }
 736                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 737                 strbuf_addbuf(&mi->inbody_header_accum, line);
 738                 return 1;
 739         }
 740
 741         flush_inbody_header_accum(mi);
 742
 743         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 744                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 745         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 746                 int i;
 747                 for (i = 0; header[i]; i++)
 748                         if (!strcmp("Subject", header[i])) {
 749                                 handle_header(&mi->s_hdr_data[i], line);
 750                                 return 1;
 751                         }
 752                 return 0;
 753         }
 754         if (is_inbody_header(mi, line)) {
 755                 strbuf_addbuf(&mi->inbody_header_accum, line);
 756                 return 1;
 757         }
 758         return 0;
 759 }
 760
 761 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 762 {
 763         assert(!mi->filter_stage);
 764
 765         if (mi->header_stage) {
 766                 if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 767                         if (mi->inbody_header_accum.len) {
 768                                 flush_inbody_header_accum(mi);
 769                                 mi->header_stage = 0;
 770                         }
 771                         return 0;
 772                 }
 773         }
 774
 775         if (mi->use_inbody_headers && mi->header_stage) {
 776                 mi->header_stage = check_inbody_header(mi, line);
 777                 if (mi->header_stage)
 778                         return 0;
 779         } else
 780                 /* Only trim the first (blank) line of the commit message
 781                  * when ignoring in-body headers.
 782                  */
 783                 mi->header_stage = 0;
 784
 785         /* normalize the log message to UTF-8. */
 786         if (convert_to_utf8(mi, line, mi->charset.buf))
 787                 return 0; /* mi->input_error already set */
 788
 789         if (mi->use_scissors && is_scissors_line(line->buf)) {
 790                 int i;
 791
 792                 strbuf_setlen(&mi->log_message, 0);
 793                 mi->header_stage = 1;
 794
 795                 /*
 796                  * We may have already read "secondary headers"; purge
 797                  * them to give ourselves a clean restart.
 798                  */
 799                 for (i = 0; header[i]; i++) {
 800                         if (mi->s_hdr_data[i])
 801                                 strbuf_release(mi->s_hdr_data[i]);
 802                         mi->s_hdr_data[i] = NULL;
 803                 }
 804                 return 0;
 805         }
 806
 807         if (patchbreak(line)) {
 808                 if (mi->message_id)
 809                         strbuf_addf(&mi->log_message,
 810                                     "Message-Id: %s\n", mi->message_id);
 811                 return 1;
 812         }
 813
 814         strbuf_addbuf(&mi->log_message, line);
 815         return 0;
 816 }
 817
 818 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 819 {
 820         fwrite(line->buf, 1, line->len, mi->patchfile);
 821         mi->patch_lines++;
 822 }
 823
 824 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 825 {
 826         switch (mi->filter_stage) {
 827         case 0:
 828                 if (!handle_commit_msg(mi, line))
 829                         break;
 830                 mi->filter_stage++;
 831         case 1:
 832                 handle_patch(mi, line);
 833                 break;
 834         }
 835 }
 836
 837 static int is_rfc2822_header(const struct strbuf *line)
 838 {
 839         /*
 840          * The section that defines the loosest possible
 841          * field name is "3.6.8 Optional fields".
 842          *
 843          * optional-field = field-name ":" unstructured CRLF
 844          * field-name = 1*ftext
 845          * ftext = %d33-57 / %59-126
 846          */
 847         int ch;
 848         char *cp = line->buf;
 849
 850         /* Count mbox From headers as headers */
 851         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 852                 return 1;
 853
 854         while ((ch = *cp++)) {
 855                 if (ch == ':')
 856                         return 1;
 857                 if ((33 <= ch && ch <= 57) ||
 858                     (59 <= ch && ch <= 126))
 859                         continue;
 860                 break;
 861         }
 862         return 0;
 863 }
 864
 865 static int read_one_header_line(struct strbuf *line, FILE *in)
 866 {
 867         struct strbuf continuation = STRBUF_INIT;
 868
 869         /* Get the first part of the line. */
 870         if (strbuf_getline_lf(line, in))
 871                 return 0;
 872
 873         /*
 874          * Is it an empty line or not a valid rfc2822 header?
 875          * If so, stop here, and return false ("not a header")
 876          */
 877         strbuf_rtrim(line);
 878         if (!line->len || !is_rfc2822_header(line)) {
 879                 /* Re-add the newline */
 880                 strbuf_addch(line, '\n');
 881                 return 0;
 882         }
 883
 884         /*
 885          * Now we need to eat all the continuation lines..
 886          * Yuck, 2822 header "folding"
 887          */
 888         for (;;) {
 889                 int peek;
 890
 891                 peek = fgetc(in);
 892                 if (peek == EOF)
 893                         break;
 894                 ungetc(peek, in);
 895                 if (peek != ' ' && peek != '\t')
 896                         break;
 897                 if (strbuf_getline_lf(&continuation, in))
 898                         break;
 899                 continuation.buf[0] = ' ';
 900                 strbuf_rtrim(&continuation);
 901                 strbuf_addbuf(line, &continuation);
 902         }
 903         strbuf_release(&continuation);
 904
 905         return 1;
 906 }
 907
 908 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 909 {
 910         while (!strbuf_getline_lf(line, mi->input)) {
 911                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 912                         return 1;
 913         }
 914         return 0;
 915 }
 916
 917 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 918 {
 919         struct strbuf newline = STRBUF_INIT;
 920
 921         strbuf_addch(&newline, '\n');
 922 again:
 923         if (line->len >= (*(mi->content_top))->len + 2 &&
 924             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 925                 /* we hit an end boundary */
 926                 /* pop the current boundary off the stack */
 927                 strbuf_release(*(mi->content_top));
 928                 FREE_AND_NULL(*(mi->content_top));
 929
 930                 /* technically won't happen as is_multipart_boundary()
 931                    will fail first.  But just in case..
 932                  */
 933                 if (--mi->content_top < mi->content) {
 934                         error("Detected mismatched boundaries, can't recover");
 935                         mi->input_error = -1;
 936                         mi->content_top = mi->content;
 937                         return 0;
 938                 }
 939                 handle_filter(mi, &newline);
 940                 strbuf_release(&newline);
 941                 if (mi->input_error)
 942                         return 0;
 943
 944                 /* skip to the next boundary */
 945                 if (!find_boundary(mi, line))
 946                         return 0;
 947                 goto again;
 948         }
 949
 950         /* set some defaults */
 951         mi->transfer_encoding = TE_DONTCARE;
 952         strbuf_reset(&mi->charset);
 953
 954         /* slurp in this section's info */
 955         while (read_one_header_line(line, mi->input))
 956                 check_header(mi, line, mi->p_hdr_data, 0);
 957
 958         strbuf_release(&newline);
 959         /* replenish line */
 960         if (strbuf_getline_lf(line, mi->input))
 961                 return 0;
 962         strbuf_addch(line, '\n');
 963         return 1;
 964 }
 965
 966 static void handle_body(struct mailinfo *mi, struct strbuf *line)
 967 {
 968         struct strbuf prev = STRBUF_INIT;
 969
 970         /* Skip up to the first boundary */
 971         if (*(mi->content_top)) {
 972                 if (!find_boundary(mi, line))
 973                         goto handle_body_out;
 974         }
 975
 976         do {
 977                 /* process any boundary lines */
 978                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
 979                         /* flush any leftover */
 980                         if (prev.len) {
 981                                 handle_filter(mi, &prev);
 982                                 strbuf_reset(&prev);
 983                         }
 984                         if (!handle_boundary(mi, line))
 985                                 goto handle_body_out;
 986                 }
 987
 988                 /* Unwrap transfer encoding */
 989                 decode_transfer_encoding(mi, line);
 990
 991                 switch (mi->transfer_encoding) {
 992                 case TE_BASE64:
 993                 case TE_QP:
 994                 {
 995                         struct strbuf **lines, **it, *sb;
 996
 997                         /* Prepend any previous partial lines */
 998                         strbuf_insert(line, 0, prev.buf, prev.len);
 999                         strbuf_reset(&prev);
1000
1001                         /*
1002                          * This is a decoded line that may contain
1003                          * multiple new lines.  Pass only one chunk
1004                          * at a time to handle_filter()
1005                          */
1006                         lines = strbuf_split(line, '\n');
1007                         for (it = lines; (sb = *it); it++) {
1008                                 if (*(it + 1) == NULL) /* The last line */
1009                                         if (sb->buf[sb->len - 1] != '\n') {
1010                                                 /* Partial line, save it for later. */
1011                                                 strbuf_addbuf(&prev, sb);
1012                                                 break;
1013                                         }
1014                                 handle_filter(mi, sb);
1015                         }
1016                         /*
1017                          * The partial chunk is saved in "prev" and will be
1018                          * appended by the next iteration of read_line_with_nul().
1019                          */
1020                         strbuf_list_free(lines);
1021                         break;
1022                 }
1023                 default:
1024                         handle_filter(mi, line);
1025                 }
1026
1027                 if (mi->input_error)
1028                         break;
1029         } while (!strbuf_getwholeline(line, mi->input, '\n'));
1030
1031         flush_inbody_header_accum(mi);
1032
1033 handle_body_out:
1034         strbuf_release(&prev);
1035 }
1036
1037 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
1038 {
1039         const char *sp = data->buf;
1040         while (1) {
1041                 char *ep = strchr(sp, '\n');
1042                 int len;
1043                 if (!ep)
1044                         len = strlen(sp);
1045                 else
1046                         len = ep - sp;
1047                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
1048                 if (!ep)
1049                         break;
1050                 sp = ep + 1;
1051         }
1052 }
1053
1054 static void handle_info(struct mailinfo *mi)
1055 {
1056         struct strbuf *hdr;
1057         int i;
1058
1059         for (i = 0; header[i]; i++) {
1060                 /* only print inbody headers if we output a patch file */
1061                 if (mi->patch_lines && mi->s_hdr_data[i])
1062                         hdr = mi->s_hdr_data[i];
1063                 else if (mi->p_hdr_data[i])
1064                         hdr = mi->p_hdr_data[i];
1065                 else
1066                         continue;
1067
1068                 if (!strcmp(header[i], "Subject")) {
1069                         if (!mi->keep_subject) {
1070                                 cleanup_subject(mi, hdr);
1071                                 cleanup_space(hdr);
1072                         }
1073                         output_header_lines(mi->output, "Subject", hdr);
1074                 } else if (!strcmp(header[i], "From")) {
1075                         cleanup_space(hdr);
1076                         handle_from(mi, hdr);
1077                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
1078                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
1079                 } else {
1080                         cleanup_space(hdr);
1081                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
1082                 }
1083         }
1084         fprintf(mi->output, "\n");
1085 }
1086
1087 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
1088 {
1089         FILE *cmitmsg;
1090         int peek;
1091         struct strbuf line = STRBUF_INIT;
1092
1093         cmitmsg = fopen(msg, "w");
1094         if (!cmitmsg) {
1095                 perror(msg);
1096                 return -1;
1097         }
1098         mi->patchfile = fopen(patch, "w");
1099         if (!mi->patchfile) {
1100                 perror(patch);
1101                 fclose(cmitmsg);
1102                 return -1;
1103         }
1104
1105         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
1106         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
1107
1108         do {
1109                 peek = fgetc(mi->input);
1110                 if (peek == EOF) {
1111                         fclose(cmitmsg);
1112                         return error("empty patch: '%s'", patch);
1113                 }
1114         } while (isspace(peek));
1115         ungetc(peek, mi->input);
1116
1117         /* process the email header */
1118         while (read_one_header_line(&line, mi->input))
1119                 check_header(mi, &line, mi->p_hdr_data, 1);
1120
1121         handle_body(mi, &line);
1122         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1123         fclose(cmitmsg);
1124         fclose(mi->patchfile);
1125
1126         handle_info(mi);
1127         strbuf_release(&line);
1128         return mi->input_error;
1129 }
1130
1131 static int git_mailinfo_config(const char *var, const char *value, void *mi_)
1132 {
1133         struct mailinfo *mi = mi_;
1134
1135         if (!starts_with(var, "mailinfo."))
1136                 return git_default_config(var, value, NULL);
1137         if (!strcmp(var, "mailinfo.scissors")) {
1138                 mi->use_scissors = git_config_bool(var, value);
1139                 return 0;
1140         }
1141         /* perhaps others here */
1142         return 0;
1143 }
1144
1145 void setup_mailinfo(struct mailinfo *mi)
1146 {
1147         memset(mi, 0, sizeof(*mi));
1148         strbuf_init(&mi->name, 0);
1149         strbuf_init(&mi->email, 0);
1150         strbuf_init(&mi->charset, 0);
1151         strbuf_init(&mi->log_message, 0);
1152         strbuf_init(&mi->inbody_header_accum, 0);
1153         mi->header_stage = 1;
1154         mi->use_inbody_headers = 1;
1155         mi->content_top = mi->content;
1156         git_config(git_mailinfo_config, mi);
1157 }
1158
1159 void clear_mailinfo(struct mailinfo *mi)
1160 {
1161         int i;
1162
1163         strbuf_release(&mi->name);
1164         strbuf_release(&mi->email);
1165         strbuf_release(&mi->charset);
1166         strbuf_release(&mi->inbody_header_accum);
1167         free(mi->message_id);
1168
1169         for (i = 0; mi->p_hdr_data[i]; i++)
1170                 strbuf_release(mi->p_hdr_data[i]);
1171         free(mi->p_hdr_data);
1172         for (i = 0; mi->s_hdr_data[i]; i++)
1173                 strbuf_release(mi->s_hdr_data[i]);
1174         free(mi->s_hdr_data);
1175
1176         while (mi->content < mi->content_top) {
1177                 free(*(mi->content_top));
1178                 mi->content_top--;
1179         }
1180
1181         strbuf_release(&mi->log_message);
1182 }