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                 goto out;
 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                 goto out;
 160
 161         /* Pick up the string around '@', possibly delimited with <>
 162          * pair; that is the email part.
 163          */
 164         while (at > f.buf) {
 165                 char c = at[-1];
 166                 if (isspace(c))
 167                         break;
 168                 if (c == '<') {
 169                         at[-1] = ' ';
 170                         break;
 171                 }
 172                 at--;
 173         }
 174         el = strcspn(at, " \n\t\r\v\f>");
 175         strbuf_reset(&mi->email);
 176         strbuf_add(&mi->email, at, el);
 177         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 178
 179         /* The remainder is name.  It could be
 180          *
 181          * - "John Doe <john.doe@xz>"                   (a), or
 182          * - "john.doe@xz (John Doe)"                   (b), or
 183          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 184          *
 185          * but we have removed the email part, so
 186          *
 187          * - remove extra spaces which could stay after email (case 'c'), and
 188          * - trim from both ends, possibly removing the () pair at the end
 189          *   (cases 'a' and 'b').
 190          */
 191         cleanup_space(&f);
 192         strbuf_trim(&f);
 193         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 194                 strbuf_remove(&f, 0, 1);
 195                 strbuf_setlen(&f, f.len - 1);
 196         }
 197
 198         get_sane_name(&mi->name, &f, &mi->email);
 199 out:
 200         strbuf_release(&f);
 201 }
 202
 203 static void handle_header(struct strbuf **out, const struct strbuf *line)
 204 {
 205         if (!*out) {
 206                 *out = xmalloc(sizeof(struct strbuf));
 207                 strbuf_init(*out, line->len);
 208         } else
 209                 strbuf_reset(*out);
 210
 211         strbuf_addbuf(*out, line);
 212 }
 213
 214 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 215  * to have enough heuristics to grok MIME encoded patches often found
 216  * on our mailing lists.  For example, we do not even treat header lines
 217  * case insensitively.
 218  */
 219
 220 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 221 {
 222         const char *ends, *ap = strcasestr(line, name);
 223         size_t sz;
 224
 225         strbuf_setlen(attr, 0);
 226         if (!ap)
 227                 return 0;
 228         ap += strlen(name);
 229         if (*ap == '"') {
 230                 ap++;
 231                 ends = "\"";
 232         }
 233         else
 234                 ends = "; \t";
 235         sz = strcspn(ap, ends);
 236         strbuf_add(attr, ap, sz);
 237         return 1;
 238 }
 239
 240 static void handle_content_type(struct mailinfo *mi, struct strbuf *line)
 241 {
 242         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 243         strbuf_init(boundary, line->len);
 244
 245         if (slurp_attr(line->buf, "boundary=", boundary)) {
 246                 strbuf_insert(boundary, 0, "--", 2);
 247                 if (++mi->content_top >= &mi->content[MAX_BOUNDARIES]) {
 248                         error("Too many boundaries to handle");
 249                         mi->input_error = -1;
 250                         mi->content_top = &mi->content[MAX_BOUNDARIES] - 1;
 251                         return;
 252                 }
 253                 *(mi->content_top) = boundary;
 254                 boundary = NULL;
 255         }
 256         slurp_attr(line->buf, "charset=", &mi->charset);
 257
 258         if (boundary) {
 259                 strbuf_release(boundary);
 260                 free(boundary);
 261         }
 262 }
 263
 264 static void handle_content_transfer_encoding(struct mailinfo *mi,
 265                                              const struct strbuf *line)
 266 {
 267         if (strcasestr(line->buf, "base64"))
 268                 mi->transfer_encoding = TE_BASE64;
 269         else if (strcasestr(line->buf, "quoted-printable"))
 270                 mi->transfer_encoding = TE_QP;
 271         else
 272                 mi->transfer_encoding = TE_DONTCARE;
 273 }
 274
 275 static int is_multipart_boundary(struct mailinfo *mi, const struct strbuf *line)
 276 {
 277         struct strbuf *content_top = *(mi->content_top);
 278
 279         return ((content_top->len <= line->len) &&
 280                 !memcmp(line->buf, content_top->buf, content_top->len));
 281 }
 282
 283 static void cleanup_subject(struct mailinfo *mi, struct strbuf *subject)
 284 {
 285         size_t at = 0;
 286
 287         while (at < subject->len) {
 288                 char *pos;
 289                 size_t remove;
 290
 291                 switch (subject->buf[at]) {
 292                 case 'r': case 'R':
 293                         if (subject->len <= at + 3)
 294                                 break;
 295                         if ((subject->buf[at + 1] == 'e' ||
 296                              subject->buf[at + 1] == 'E') &&
 297                             subject->buf[at + 2] == ':') {
 298                                 strbuf_remove(subject, at, 3);
 299                                 continue;
 300                         }
 301                         at++;
 302                         break;
 303                 case ' ': case '\t': case ':':
 304                         strbuf_remove(subject, at, 1);
 305                         continue;
 306                 case '[':
 307                         pos = strchr(subject->buf + at, ']');
 308                         if (!pos)
 309                                 break;
 310                         remove = pos - subject->buf + at + 1;
 311                         if (!mi->keep_non_patch_brackets_in_subject ||
 312                             (7 <= remove &&
 313                              memmem(subject->buf + at, remove, "PATCH", 5)))
 314                                 strbuf_remove(subject, at, remove);
 315                         else {
 316                                 at += remove;
 317                                 /*
 318                                  * If the input had a space after the ], keep
 319                                  * it.  We don't bother with finding the end of
 320                                  * the space, since we later normalize it
 321                                  * anyway.
 322                                  */
 323                                 if (isspace(subject->buf[at]))
 324                                         at += 1;
 325                         }
 326                         continue;
 327                 }
 328                 break;
 329         }
 330         strbuf_trim(subject);
 331 }
 332
 333 #define MAX_HDR_PARSED 10
 334 static const char *header[MAX_HDR_PARSED] = {
 335         "From","Subject","Date",
 336 };
 337
 338 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 339 {
 340         int len = strlen(hdr);
 341         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 342                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 343 }
 344
 345 static int is_format_patch_separator(const char *line, int len)
 346 {
 347         static const char SAMPLE[] =
 348                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 349         const char *cp;
 350
 351         if (len != strlen(SAMPLE))
 352                 return 0;
 353         if (!skip_prefix(line, "From ", &cp))
 354                 return 0;
 355         if (strspn(cp, "0123456789abcdef") != 40)
 356                 return 0;
 357         cp += 40;
 358         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 359 }
 360
 361 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 362 {
 363         const char *in = q_seg->buf;
 364         int c;
 365         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 366         strbuf_init(out, q_seg->len);
 367
 368         while ((c = *in++) != 0) {
 369                 if (c == '=') {
 370                         int ch, d = *in;
 371                         if (d == '\n' || !d)
 372                                 break; /* drop trailing newline */
 373                         ch = hex2chr(in);
 374                         if (ch >= 0) {
 375                                 strbuf_addch(out, ch);
 376                                 in += 2;
 377                                 continue;
 378                         }
 379                         /* garbage -- fall through */
 380                 }
 381                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 382                         c = 0x20;
 383                 strbuf_addch(out, c);
 384         }
 385         return out;
 386 }
 387
 388 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 389 {
 390         /* Decode in..ep, possibly in-place to ot */
 391         int c, pos = 0, acc = 0;
 392         const char *in = b_seg->buf;
 393         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 394         strbuf_init(out, b_seg->len);
 395
 396         while ((c = *in++) != 0) {
 397                 if (c == '+')
 398                         c = 62;
 399                 else if (c == '/')
 400                         c = 63;
 401                 else if ('A' <= c && c <= 'Z')
 402                         c -= 'A';
 403                 else if ('a' <= c && c <= 'z')
 404                         c -= 'a' - 26;
 405                 else if ('0' <= c && c <= '9')
 406                         c -= '0' - 52;
 407                 else
 408                         continue; /* garbage */
 409                 switch (pos++) {
 410                 case 0:
 411                         acc = (c << 2);
 412                         break;
 413                 case 1:
 414                         strbuf_addch(out, (acc | (c >> 4)));
 415                         acc = (c & 15) << 4;
 416                         break;
 417                 case 2:
 418                         strbuf_addch(out, (acc | (c >> 2)));
 419                         acc = (c & 3) << 6;
 420                         break;
 421                 case 3:
 422                         strbuf_addch(out, (acc | c));
 423                         acc = pos = 0;
 424                         break;
 425                 }
 426         }
 427         return out;
 428 }
 429
 430 static int convert_to_utf8(struct mailinfo *mi,
 431                            struct strbuf *line, const char *charset)
 432 {
 433         char *out;
 434
 435         if (!mi->metainfo_charset || !charset || !*charset)
 436                 return 0;
 437
 438         if (same_encoding(mi->metainfo_charset, charset))
 439                 return 0;
 440         out = reencode_string(line->buf, mi->metainfo_charset, charset);
 441         if (!out) {
 442                 mi->input_error = -1;
 443                 return error("cannot convert from %s to %s",
 444                              charset, mi->metainfo_charset);
 445         }
 446         strbuf_attach(line, out, strlen(out), strlen(out));
 447         return 0;
 448 }
 449
 450 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 451 {
 452         char *in, *ep, *cp;
 453         struct strbuf outbuf = STRBUF_INIT, *dec;
 454         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 455         int found_error = 1; /* pessimism */
 456
 457         in = it->buf;
 458         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 459                 int encoding;
 460                 strbuf_reset(&charset_q);
 461                 strbuf_reset(&piecebuf);
 462
 463                 if (in != ep) {
 464                         /*
 465                          * We are about to process an encoded-word
 466                          * that begins at ep, but there is something
 467                          * before the encoded word.
 468                          */
 469                         char *scan;
 470                         for (scan = in; scan < ep; scan++)
 471                                 if (!isspace(*scan))
 472                                         break;
 473
 474                         if (scan != ep || in == it->buf) {
 475                                 /*
 476                                  * We should not lose that "something",
 477                                  * unless we have just processed an
 478                                  * encoded-word, and there is only LWS
 479                                  * before the one we are about to process.
 480                                  */
 481                                 strbuf_add(&outbuf, in, ep - in);
 482                         }
 483                 }
 484                 /* E.g.
 485                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 486                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 487                  */
 488                 ep += 2;
 489
 490                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 491                         goto release_return;
 492
 493                 if (cp + 3 - it->buf > it->len)
 494                         goto release_return;
 495                 strbuf_add(&charset_q, ep, cp - ep);
 496
 497                 encoding = cp[1];
 498                 if (!encoding || cp[2] != '?')
 499                         goto release_return;
 500                 ep = strstr(cp + 3, "?=");
 501                 if (!ep)
 502                         goto release_return;
 503                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 504                 switch (tolower(encoding)) {
 505                 default:
 506                         goto release_return;
 507                 case 'b':
 508                         dec = decode_b_segment(&piecebuf);
 509                         break;
 510                 case 'q':
 511                         dec = decode_q_segment(&piecebuf, 1);
 512                         break;
 513                 }
 514                 if (convert_to_utf8(mi, dec, charset_q.buf))
 515                         goto release_return;
 516
 517                 strbuf_addbuf(&outbuf, dec);
 518                 strbuf_release(dec);
 519                 free(dec);
 520                 in = ep + 2;
 521         }
 522         strbuf_addstr(&outbuf, in);
 523         strbuf_reset(it);
 524         strbuf_addbuf(it, &outbuf);
 525         found_error = 0;
 526 release_return:
 527         strbuf_release(&outbuf);
 528         strbuf_release(&charset_q);
 529         strbuf_release(&piecebuf);
 530
 531         if (found_error)
 532                 mi->input_error = -1;
 533 }
 534
 535 static int check_header(struct mailinfo *mi,
 536                         const struct strbuf *line,
 537                         struct strbuf *hdr_data[], int overwrite)
 538 {
 539         int i, ret = 0, len;
 540         struct strbuf sb = STRBUF_INIT;
 541
 542         /* search for the interesting parts */
 543         for (i = 0; header[i]; i++) {
 544                 int len = strlen(header[i]);
 545                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 546                         /* Unwrap inline B and Q encoding, and optionally
 547                          * normalize the meta information to utf8.
 548                          */
 549                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 550                         decode_header(mi, &sb);
 551                         handle_header(&hdr_data[i], &sb);
 552                         ret = 1;
 553                         goto check_header_out;
 554                 }
 555         }
 556
 557         /* Content stuff */
 558         if (cmp_header(line, "Content-Type")) {
 559                 len = strlen("Content-Type: ");
 560                 strbuf_add(&sb, line->buf + len, line->len - len);
 561                 decode_header(mi, &sb);
 562                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 563                 handle_content_type(mi, &sb);
 564                 ret = 1;
 565                 goto check_header_out;
 566         }
 567         if (cmp_header(line, "Content-Transfer-Encoding")) {
 568                 len = strlen("Content-Transfer-Encoding: ");
 569                 strbuf_add(&sb, line->buf + len, line->len - len);
 570                 decode_header(mi, &sb);
 571                 handle_content_transfer_encoding(mi, &sb);
 572                 ret = 1;
 573                 goto check_header_out;
 574         }
 575         if (cmp_header(line, "Message-Id")) {
 576                 len = strlen("Message-Id: ");
 577                 strbuf_add(&sb, line->buf + len, line->len - len);
 578                 decode_header(mi, &sb);
 579                 if (mi->add_message_id)
 580                         mi->message_id = strbuf_detach(&sb, NULL);
 581                 ret = 1;
 582                 goto check_header_out;
 583         }
 584
 585 check_header_out:
 586         strbuf_release(&sb);
 587         return ret;
 588 }
 589
 590 /*
 591  * Returns 1 if the given line or any line beginning with the given line is an
 592  * in-body header (that is, check_header will succeed when passed
 593  * mi->s_hdr_data).
 594  */
 595 static int is_inbody_header(const struct mailinfo *mi,
 596                             const struct strbuf *line)
 597 {
 598         int i;
 599         for (i = 0; header[i]; i++)
 600                 if (!mi->s_hdr_data[i] && cmp_header(line, header[i]))
 601                         return 1;
 602         return 0;
 603 }
 604
 605 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 606 {
 607         struct strbuf *ret;
 608
 609         switch (mi->transfer_encoding) {
 610         case TE_QP:
 611                 ret = decode_q_segment(line, 0);
 612                 break;
 613         case TE_BASE64:
 614                 ret = decode_b_segment(line);
 615                 break;
 616         case TE_DONTCARE:
 617         default:
 618                 return;
 619         }
 620         strbuf_reset(line);
 621         strbuf_addbuf(line, ret);
 622         strbuf_release(ret);
 623         free(ret);
 624 }
 625
 626 static inline int patchbreak(const struct strbuf *line)
 627 {
 628         size_t i;
 629
 630         /* Beginning of a "diff -" header? */
 631         if (starts_with(line->buf, "diff -"))
 632                 return 1;
 633
 634         /* CVS "Index: " line? */
 635         if (starts_with(line->buf, "Index: "))
 636                 return 1;
 637
 638         /*
 639          * "--- <filename>" starts patches without headers
 640          * "---<sp>*" is a manual separator
 641          */
 642         if (line->len < 4)
 643                 return 0;
 644
 645         if (starts_with(line->buf, "---")) {
 646                 /* space followed by a filename? */
 647                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 648                         return 1;
 649                 /* Just whitespace? */
 650                 for (i = 3; i < line->len; i++) {
 651                         unsigned char c = line->buf[i];
 652                         if (c == '\n')
 653                                 return 1;
 654                         if (!isspace(c))
 655                                 break;
 656                 }
 657                 return 0;
 658         }
 659         return 0;
 660 }
 661
 662 static int is_scissors_line(const char *line)
 663 {
 664         const char *c;
 665         int scissors = 0, gap = 0;
 666         const char *first_nonblank = NULL, *last_nonblank = NULL;
 667         int visible, perforation = 0, in_perforation = 0;
 668
 669         for (c = line; *c; c++) {
 670                 if (isspace(*c)) {
 671                         if (in_perforation) {
 672                                 perforation++;
 673                                 gap++;
 674                         }
 675                         continue;
 676                 }
 677                 last_nonblank = c;
 678                 if (first_nonblank == NULL)
 679                         first_nonblank = c;
 680                 if (*c == '-') {
 681                         in_perforation = 1;
 682                         perforation++;
 683                         continue;
 684                 }
 685                 if ((!memcmp(c, ">8", 2) || !memcmp(c, "8<", 2) ||
 686                      !memcmp(c, ">%", 2) || !memcmp(c, "%<", 2))) {
 687                         in_perforation = 1;
 688                         perforation += 2;
 689                         scissors += 2;
 690                         c++;
 691                         continue;
 692                 }
 693                 in_perforation = 0;
 694         }
 695
 696         /*
 697          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 698          * Even though there can be arbitrary cruft on the same line
 699          * (e.g. "cut here"), in order to avoid misidentification, the
 700          * perforation must occupy more than a third of the visible
 701          * width of the line, and dashes and scissors must occupy more
 702          * than half of the perforation.
 703          */
 704
 705         if (first_nonblank && last_nonblank)
 706                 visible = last_nonblank - first_nonblank + 1;
 707         else
 708                 visible = 0;
 709         return (scissors && 8 <= visible &&
 710                 visible < perforation * 3 &&
 711                 gap * 2 < perforation);
 712 }
 713
 714 static void flush_inbody_header_accum(struct mailinfo *mi)
 715 {
 716         if (!mi->inbody_header_accum.len)
 717                 return;
 718         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 719                 die("BUG: inbody_header_accum, if not empty, must always contain a valid in-body header");
 720         strbuf_reset(&mi->inbody_header_accum);
 721 }
 722
 723 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 724 {
 725         if (mi->inbody_header_accum.len &&
 726             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 727                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 728                         /*
 729                          * This is a scissors line; do not consider this line
 730                          * as a header continuation line.
 731                          */
 732                         flush_inbody_header_accum(mi);
 733                         return 0;
 734                 }
 735                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 736                 strbuf_addbuf(&mi->inbody_header_accum, line);
 737                 return 1;
 738         }
 739
 740         flush_inbody_header_accum(mi);
 741
 742         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 743                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 744         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 745                 int i;
 746                 for (i = 0; header[i]; i++)
 747                         if (!strcmp("Subject", header[i])) {
 748                                 handle_header(&mi->s_hdr_data[i], line);
 749                                 return 1;
 750                         }
 751                 return 0;
 752         }
 753         if (is_inbody_header(mi, line)) {
 754                 strbuf_addbuf(&mi->inbody_header_accum, line);
 755                 return 1;
 756         }
 757         return 0;
 758 }
 759
 760 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 761 {
 762         assert(!mi->filter_stage);
 763
 764         if (mi->header_stage) {
 765                 if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 766                         if (mi->inbody_header_accum.len) {
 767                                 flush_inbody_header_accum(mi);
 768                                 mi->header_stage = 0;
 769                         }
 770                         return 0;
 771                 }
 772         }
 773
 774         if (mi->use_inbody_headers && mi->header_stage) {
 775                 mi->header_stage = check_inbody_header(mi, line);
 776                 if (mi->header_stage)
 777                         return 0;
 778         } else
 779                 /* Only trim the first (blank) line of the commit message
 780                  * when ignoring in-body headers.
 781                  */
 782                 mi->header_stage = 0;
 783
 784         /* normalize the log message to UTF-8. */
 785         if (convert_to_utf8(mi, line, mi->charset.buf))
 786                 return 0; /* mi->input_error already set */
 787
 788         if (mi->use_scissors && is_scissors_line(line->buf)) {
 789                 int i;
 790
 791                 strbuf_setlen(&mi->log_message, 0);
 792                 mi->header_stage = 1;
 793
 794                 /*
 795                  * We may have already read "secondary headers"; purge
 796                  * them to give ourselves a clean restart.
 797                  */
 798                 for (i = 0; header[i]; i++) {
 799                         if (mi->s_hdr_data[i])
 800                                 strbuf_release(mi->s_hdr_data[i]);
 801                         mi->s_hdr_data[i] = NULL;
 802                 }
 803                 return 0;
 804         }
 805
 806         if (patchbreak(line)) {
 807                 if (mi->message_id)
 808                         strbuf_addf(&mi->log_message,
 809                                     "Message-Id: %s\n", mi->message_id);
 810                 return 1;
 811         }
 812
 813         strbuf_addbuf(&mi->log_message, line);
 814         return 0;
 815 }
 816
 817 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 818 {
 819         fwrite(line->buf, 1, line->len, mi->patchfile);
 820         mi->patch_lines++;
 821 }
 822
 823 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 824 {
 825         switch (mi->filter_stage) {
 826         case 0:
 827                 if (!handle_commit_msg(mi, line))
 828                         break;
 829                 mi->filter_stage++;
 830                 /* fallthrough */
 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                         strbuf_release(&newline);
 938                         return 0;
 939                 }
 940                 handle_filter(mi, &newline);
 941                 strbuf_release(&newline);
 942                 if (mi->input_error)
 943                         return 0;
 944
 945                 /* skip to the next boundary */
 946                 if (!find_boundary(mi, line))
 947                         return 0;
 948                 goto again;
 949         }
 950
 951         /* set some defaults */
 952         mi->transfer_encoding = TE_DONTCARE;
 953         strbuf_reset(&mi->charset);
 954
 955         /* slurp in this section's info */
 956         while (read_one_header_line(line, mi->input))
 957                 check_header(mi, line, mi->p_hdr_data, 0);
 958
 959         strbuf_release(&newline);
 960         /* replenish line */
 961         if (strbuf_getline_lf(line, mi->input))
 962                 return 0;
 963         strbuf_addch(line, '\n');
 964         return 1;
 965 }
 966
 967 static void handle_body(struct mailinfo *mi, struct strbuf *line)
 968 {
 969         struct strbuf prev = STRBUF_INIT;
 970
 971         /* Skip up to the first boundary */
 972         if (*(mi->content_top)) {
 973                 if (!find_boundary(mi, line))
 974                         goto handle_body_out;
 975         }
 976
 977         do {
 978                 /* process any boundary lines */
 979                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
 980                         /* flush any leftover */
 981                         if (prev.len) {
 982                                 handle_filter(mi, &prev);
 983                                 strbuf_reset(&prev);
 984                         }
 985                         if (!handle_boundary(mi, line))
 986                                 goto handle_body_out;
 987                 }
 988
 989                 /* Unwrap transfer encoding */
 990                 decode_transfer_encoding(mi, line);
 991
 992                 switch (mi->transfer_encoding) {
 993                 case TE_BASE64:
 994                 case TE_QP:
 995                 {
 996                         struct strbuf **lines, **it, *sb;
 997
 998                         /* Prepend any previous partial lines */
 999                         strbuf_insert(line, 0, prev.buf, prev.len);
1000                         strbuf_reset(&prev);
1001
1002                         /*
1003                          * This is a decoded line that may contain
1004                          * multiple new lines.  Pass only one chunk
1005                          * at a time to handle_filter()
1006                          */
1007                         lines = strbuf_split(line, '\n');
1008                         for (it = lines; (sb = *it); it++) {
1009                                 if (*(it + 1) == NULL) /* The last line */
1010                                         if (sb->buf[sb->len - 1] != '\n') {
1011                                                 /* Partial line, save it for later. */
1012                                                 strbuf_addbuf(&prev, sb);
1013                                                 break;
1014                                         }
1015                                 handle_filter(mi, sb);
1016                         }
1017                         /*
1018                          * The partial chunk is saved in "prev" and will be
1019                          * appended by the next iteration of read_line_with_nul().
1020                          */
1021                         strbuf_list_free(lines);
1022                         break;
1023                 }
1024                 default:
1025                         handle_filter(mi, line);
1026                 }
1027
1028                 if (mi->input_error)
1029                         break;
1030         } while (!strbuf_getwholeline(line, mi->input, '\n'));
1031
1032         flush_inbody_header_accum(mi);
1033
1034 handle_body_out:
1035         strbuf_release(&prev);
1036 }
1037
1038 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
1039 {
1040         const char *sp = data->buf;
1041         while (1) {
1042                 char *ep = strchr(sp, '\n');
1043                 int len;
1044                 if (!ep)
1045                         len = strlen(sp);
1046                 else
1047                         len = ep - sp;
1048                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
1049                 if (!ep)
1050                         break;
1051                 sp = ep + 1;
1052         }
1053 }
1054
1055 static void handle_info(struct mailinfo *mi)
1056 {
1057         struct strbuf *hdr;
1058         int i;
1059
1060         for (i = 0; header[i]; i++) {
1061                 /* only print inbody headers if we output a patch file */
1062                 if (mi->patch_lines && mi->s_hdr_data[i])
1063                         hdr = mi->s_hdr_data[i];
1064                 else if (mi->p_hdr_data[i])
1065                         hdr = mi->p_hdr_data[i];
1066                 else
1067                         continue;
1068
1069                 if (!strcmp(header[i], "Subject")) {
1070                         if (!mi->keep_subject) {
1071                                 cleanup_subject(mi, hdr);
1072                                 cleanup_space(hdr);
1073                         }
1074                         output_header_lines(mi->output, "Subject", hdr);
1075                 } else if (!strcmp(header[i], "From")) {
1076                         cleanup_space(hdr);
1077                         handle_from(mi, hdr);
1078                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
1079                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
1080                 } else {
1081                         cleanup_space(hdr);
1082                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
1083                 }
1084         }
1085         fprintf(mi->output, "\n");
1086 }
1087
1088 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
1089 {
1090         FILE *cmitmsg;
1091         int peek;
1092         struct strbuf line = STRBUF_INIT;
1093
1094         cmitmsg = fopen(msg, "w");
1095         if (!cmitmsg) {
1096                 perror(msg);
1097                 return -1;
1098         }
1099         mi->patchfile = fopen(patch, "w");
1100         if (!mi->patchfile) {
1101                 perror(patch);
1102                 fclose(cmitmsg);
1103                 return -1;
1104         }
1105
1106         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
1107         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
1108
1109         do {
1110                 peek = fgetc(mi->input);
1111                 if (peek == EOF) {
1112                         fclose(cmitmsg);
1113                         return error("empty patch: '%s'", patch);
1114                 }
1115         } while (isspace(peek));
1116         ungetc(peek, mi->input);
1117
1118         /* process the email header */
1119         while (read_one_header_line(&line, mi->input))
1120                 check_header(mi, &line, mi->p_hdr_data, 1);
1121
1122         handle_body(mi, &line);
1123         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1124         fclose(cmitmsg);
1125         fclose(mi->patchfile);
1126
1127         handle_info(mi);
1128         strbuf_release(&line);
1129         return mi->input_error;
1130 }
1131
1132 static int git_mailinfo_config(const char *var, const char *value, void *mi_)
1133 {
1134         struct mailinfo *mi = mi_;
1135
1136         if (!starts_with(var, "mailinfo."))
1137                 return git_default_config(var, value, NULL);
1138         if (!strcmp(var, "mailinfo.scissors")) {
1139                 mi->use_scissors = git_config_bool(var, value);
1140                 return 0;
1141         }
1142         /* perhaps others here */
1143         return 0;
1144 }
1145
1146 void setup_mailinfo(struct mailinfo *mi)
1147 {
1148         memset(mi, 0, sizeof(*mi));
1149         strbuf_init(&mi->name, 0);
1150         strbuf_init(&mi->email, 0);
1151         strbuf_init(&mi->charset, 0);
1152         strbuf_init(&mi->log_message, 0);
1153         strbuf_init(&mi->inbody_header_accum, 0);
1154         mi->header_stage = 1;
1155         mi->use_inbody_headers = 1;
1156         mi->content_top = mi->content;
1157         git_config(git_mailinfo_config, mi);
1158 }
1159
1160 void clear_mailinfo(struct mailinfo *mi)
1161 {
1162         int i;
1163
1164         strbuf_release(&mi->name);
1165         strbuf_release(&mi->email);
1166         strbuf_release(&mi->charset);
1167         strbuf_release(&mi->inbody_header_accum);
1168         free(mi->message_id);
1169
1170         for (i = 0; mi->p_hdr_data[i]; i++)
1171                 strbuf_release(mi->p_hdr_data[i]);
1172         free(mi->p_hdr_data);
1173         for (i = 0; mi->s_hdr_data[i]; i++)
1174                 strbuf_release(mi->s_hdr_data[i]);
1175         free(mi->s_hdr_data);
1176
1177         while (mi->content < mi->content_top) {
1178                 free(*(mi->content_top));
1179                 mi->content_top--;
1180         }
1181
1182         strbuf_release(&mi->log_message);
1183 }