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 || strpbrk(name->buf, "@<>"))
  23                 src = email;
  24         else if (name == out)
  25                 return;
  26         strbuf_reset(out);
  27         strbuf_addbuf(out, src);
  28 }
  29
  30 static void parse_bogus_from(struct mailinfo *mi, const struct strbuf *line)
  31 {
  32         /* John Doe <johndoe> */
  33
  34         char *bra, *ket;
  35         /* This is fallback, so do not bother if we already have an
  36          * e-mail address.
  37          */
  38         if (mi->email.len)
  39                 return;
  40
  41         bra = strchr(line->buf, '<');
  42         if (!bra)
  43                 return;
  44         ket = strchr(bra, '>');
  45         if (!ket)
  46                 return;
  47
  48         strbuf_reset(&mi->email);
  49         strbuf_add(&mi->email, bra + 1, ket - bra - 1);
  50
  51         strbuf_reset(&mi->name);
  52         strbuf_add(&mi->name, line->buf, bra - line->buf);
  53         strbuf_trim(&mi->name);
  54         get_sane_name(&mi->name, &mi->name, &mi->email);
  55 }
  56
  57 static const char *unquote_comment(struct strbuf *outbuf, const char *in)
  58 {
  59         int c;
  60         int take_next_literally = 0;
  61
  62         strbuf_addch(outbuf, '(');
  63
  64         while ((c = *in++) != 0) {
  65                 if (take_next_literally == 1) {
  66                         take_next_literally = 0;
  67                 } else {
  68                         switch (c) {
  69                         case '\\':
  70                                 take_next_literally = 1;
  71                                 continue;
  72                         case '(':
  73                                 in = unquote_comment(outbuf, in);
  74                                 continue;
  75                         case ')':
  76                                 strbuf_addch(outbuf, ')');
  77                                 return in;
  78                         }
  79                 }
  80
  81                 strbuf_addch(outbuf, c);
  82         }
  83
  84         return in;
  85 }
  86
  87 static const char *unquote_quoted_string(struct strbuf *outbuf, const char *in)
  88 {
  89         int c;
  90         int take_next_literally = 0;
  91
  92         while ((c = *in++) != 0) {
  93                 if (take_next_literally == 1) {
  94                         take_next_literally = 0;
  95                 } else {
  96                         switch (c) {
  97                         case '\\':
  98                                 take_next_literally = 1;
  99                                 continue;
 100                         case '"':
 101                                 return in;
 102                         }
 103                 }
 104
 105                 strbuf_addch(outbuf, c);
 106         }
 107
 108         return in;
 109 }
 110
 111 static void unquote_quoted_pair(struct strbuf *line)
 112 {
 113         struct strbuf outbuf;
 114         const char *in = line->buf;
 115         int c;
 116
 117         strbuf_init(&outbuf, line->len);
 118
 119         while ((c = *in++) != 0) {
 120                 switch (c) {
 121                 case '"':
 122                         in = unquote_quoted_string(&outbuf, in);
 123                         continue;
 124                 case '(':
 125                         in = unquote_comment(&outbuf, in);
 126                         continue;
 127                 }
 128
 129                 strbuf_addch(&outbuf, c);
 130         }
 131
 132         strbuf_swap(&outbuf, line);
 133         strbuf_release(&outbuf);
 134
 135 }
 136
 137 static void handle_from(struct mailinfo *mi, const struct strbuf *from)
 138 {
 139         char *at;
 140         size_t el;
 141         struct strbuf f;
 142
 143         strbuf_init(&f, from->len);
 144         strbuf_addbuf(&f, from);
 145
 146         unquote_quoted_pair(&f);
 147
 148         at = strchr(f.buf, '@');
 149         if (!at) {
 150                 parse_bogus_from(mi, from);
 151                 goto out;
 152         }
 153
 154         /*
 155          * If we already have one email, don't take any confusing lines
 156          */
 157         if (mi->email.len && strchr(at + 1, '@'))
 158                 goto out;
 159
 160         /* Pick up the string around '@', possibly delimited with <>
 161          * pair; that is the email part.
 162          */
 163         while (at > f.buf) {
 164                 char c = at[-1];
 165                 if (isspace(c))
 166                         break;
 167                 if (c == '<') {
 168                         at[-1] = ' ';
 169                         break;
 170                 }
 171                 at--;
 172         }
 173         el = strcspn(at, " \n\t\r\v\f>");
 174         strbuf_reset(&mi->email);
 175         strbuf_add(&mi->email, at, el);
 176         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 177
 178         /* The remainder is name.  It could be
 179          *
 180          * - "John Doe <john.doe@xz>"                   (a), or
 181          * - "john.doe@xz (John Doe)"                   (b), or
 182          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 183          *
 184          * but we have removed the email part, so
 185          *
 186          * - remove extra spaces which could stay after email (case 'c'), and
 187          * - trim from both ends, possibly removing the () pair at the end
 188          *   (cases 'a' and 'b').
 189          */
 190         cleanup_space(&f);
 191         strbuf_trim(&f);
 192         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 193                 strbuf_remove(&f, 0, 1);
 194                 strbuf_setlen(&f, f.len - 1);
 195         }
 196
 197         get_sane_name(&mi->name, &f, &mi->email);
 198 out:
 199         strbuf_release(&f);
 200 }
 201
 202 static void handle_header(struct strbuf **out, const struct strbuf *line)
 203 {
 204         if (!*out) {
 205                 *out = xmalloc(sizeof(struct strbuf));
 206                 strbuf_init(*out, line->len);
 207         } else
 208                 strbuf_reset(*out);
 209
 210         strbuf_addbuf(*out, line);
 211 }
 212
 213 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 214  * to have enough heuristics to grok MIME encoded patches often found
 215  * on our mailing lists.  For example, we do not even treat header lines
 216  * case insensitively.
 217  */
 218
 219 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 220 {
 221         const char *ends, *ap = strcasestr(line, name);
 222         size_t sz;
 223
 224         strbuf_setlen(attr, 0);
 225         if (!ap)
 226                 return 0;
 227         ap += strlen(name);
 228         if (*ap == '"') {
 229                 ap++;
 230                 ends = "\"";
 231         }
 232         else
 233                 ends = "; \t";
 234         sz = strcspn(ap, ends);
 235         strbuf_add(attr, ap, sz);
 236         return 1;
 237 }
 238
 239 static int has_attr_value(const char *line, const char *name, const char *value)
 240 {
 241         struct strbuf sb = STRBUF_INIT;
 242         int rc = slurp_attr(line, name, &sb) && !strcasecmp(sb.buf, value);
 243         strbuf_release(&sb);
 244         return rc;
 245 }
 246
 247 static void handle_content_type(struct mailinfo *mi, struct strbuf *line)
 248 {
 249         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 250         strbuf_init(boundary, line->len);
 251
 252         mi->format_flowed = has_attr_value(line->buf, "format=", "flowed");
 253         mi->delsp = has_attr_value(line->buf, "delsp=", "yes");
 254
 255         if (slurp_attr(line->buf, "boundary=", boundary)) {
 256                 strbuf_insertstr(boundary, 0, "--");
 257                 if (++mi->content_top >= &mi->content[MAX_BOUNDARIES]) {
 258                         error("Too many boundaries to handle");
 259                         mi->input_error = -1;
 260                         mi->content_top = &mi->content[MAX_BOUNDARIES] - 1;
 261                         return;
 262                 }
 263                 *(mi->content_top) = boundary;
 264                 boundary = NULL;
 265         }
 266         slurp_attr(line->buf, "charset=", &mi->charset);
 267
 268         if (boundary) {
 269                 strbuf_release(boundary);
 270                 free(boundary);
 271         }
 272 }
 273
 274 static void handle_content_transfer_encoding(struct mailinfo *mi,
 275                                              const struct strbuf *line)
 276 {
 277         if (strcasestr(line->buf, "base64"))
 278                 mi->transfer_encoding = TE_BASE64;
 279         else if (strcasestr(line->buf, "quoted-printable"))
 280                 mi->transfer_encoding = TE_QP;
 281         else
 282                 mi->transfer_encoding = TE_DONTCARE;
 283 }
 284
 285 static int is_multipart_boundary(struct mailinfo *mi, const struct strbuf *line)
 286 {
 287         struct strbuf *content_top = *(mi->content_top);
 288
 289         return ((content_top->len <= line->len) &&
 290                 !memcmp(line->buf, content_top->buf, content_top->len));
 291 }
 292
 293 static void cleanup_subject(struct mailinfo *mi, struct strbuf *subject)
 294 {
 295         size_t at = 0;
 296
 297         while (at < subject->len) {
 298                 char *pos;
 299                 size_t remove;
 300
 301                 switch (subject->buf[at]) {
 302                 case 'r': case 'R':
 303                         if (subject->len <= at + 3)
 304                                 break;
 305                         if ((subject->buf[at + 1] == 'e' ||
 306                              subject->buf[at + 1] == 'E') &&
 307                             subject->buf[at + 2] == ':') {
 308                                 strbuf_remove(subject, at, 3);
 309                                 continue;
 310                         }
 311                         at++;
 312                         break;
 313                 case ' ': case '\t': case ':':
 314                         strbuf_remove(subject, at, 1);
 315                         continue;
 316                 case '[':
 317                         pos = strchr(subject->buf + at, ']');
 318                         if (!pos)
 319                                 break;
 320                         remove = pos - subject->buf + at + 1;
 321                         if (!mi->keep_non_patch_brackets_in_subject ||
 322                             (7 <= remove &&
 323                              memmem(subject->buf + at, remove, "PATCH", 5)))
 324                                 strbuf_remove(subject, at, remove);
 325                         else {
 326                                 at += remove;
 327                                 /*
 328                                  * If the input had a space after the ], keep
 329                                  * it.  We don't bother with finding the end of
 330                                  * the space, since we later normalize it
 331                                  * anyway.
 332                                  */
 333                                 if (isspace(subject->buf[at]))
 334                                         at += 1;
 335                         }
 336                         continue;
 337                 }
 338                 break;
 339         }
 340         strbuf_trim(subject);
 341 }
 342
 343 #define MAX_HDR_PARSED 10
 344 static const char *header[MAX_HDR_PARSED] = {
 345         "From","Subject","Date",
 346 };
 347
 348 static inline int skip_header(const struct strbuf *line, const char *hdr,
 349                               const char **outval)
 350 {
 351         const char *val;
 352         if (!skip_iprefix(line->buf, hdr, &val) ||
 353             *val++ != ':')
 354                 return 0;
 355         while (isspace(*val))
 356                 val++;
 357         *outval = val;
 358         return 1;
 359 }
 360
 361 static int is_format_patch_separator(const char *line, int len)
 362 {
 363         static const char SAMPLE[] =
 364                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 365         const char *cp;
 366
 367         if (len != strlen(SAMPLE))
 368                 return 0;
 369         if (!skip_prefix(line, "From ", &cp))
 370                 return 0;
 371         if (strspn(cp, "0123456789abcdef") != 40)
 372                 return 0;
 373         cp += 40;
 374         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 375 }
 376
 377 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 378 {
 379         const char *in = q_seg->buf;
 380         int c;
 381         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 382         strbuf_init(out, q_seg->len);
 383
 384         while ((c = *in++) != 0) {
 385                 if (c == '=') {
 386                         int ch, d = *in;
 387                         if (d == '\n' || !d)
 388                                 break; /* drop trailing newline */
 389                         ch = hex2chr(in);
 390                         if (ch >= 0) {
 391                                 strbuf_addch(out, ch);
 392                                 in += 2;
 393                                 continue;
 394                         }
 395                         /* garbage -- fall through */
 396                 }
 397                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 398                         c = 0x20;
 399                 strbuf_addch(out, c);
 400         }
 401         return out;
 402 }
 403
 404 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 405 {
 406         /* Decode in..ep, possibly in-place to ot */
 407         int c, pos = 0, acc = 0;
 408         const char *in = b_seg->buf;
 409         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 410         strbuf_init(out, b_seg->len);
 411
 412         while ((c = *in++) != 0) {
 413                 if (c == '+')
 414                         c = 62;
 415                 else if (c == '/')
 416                         c = 63;
 417                 else if ('A' <= c && c <= 'Z')
 418                         c -= 'A';
 419                 else if ('a' <= c && c <= 'z')
 420                         c -= 'a' - 26;
 421                 else if ('0' <= c && c <= '9')
 422                         c -= '0' - 52;
 423                 else
 424                         continue; /* garbage */
 425                 switch (pos++) {
 426                 case 0:
 427                         acc = (c << 2);
 428                         break;
 429                 case 1:
 430                         strbuf_addch(out, (acc | (c >> 4)));
 431                         acc = (c & 15) << 4;
 432                         break;
 433                 case 2:
 434                         strbuf_addch(out, (acc | (c >> 2)));
 435                         acc = (c & 3) << 6;
 436                         break;
 437                 case 3:
 438                         strbuf_addch(out, (acc | c));
 439                         acc = pos = 0;
 440                         break;
 441                 }
 442         }
 443         return out;
 444 }
 445
 446 static int convert_to_utf8(struct mailinfo *mi,
 447                            struct strbuf *line, const char *charset)
 448 {
 449         char *out;
 450         size_t out_len;
 451
 452         if (!mi->metainfo_charset || !charset || !*charset)
 453                 return 0;
 454
 455         if (same_encoding(mi->metainfo_charset, charset))
 456                 return 0;
 457         out = reencode_string_len(line->buf, line->len,
 458                                   mi->metainfo_charset, charset, &out_len);
 459         if (!out) {
 460                 mi->input_error = -1;
 461                 return error("cannot convert from %s to %s",
 462                              charset, mi->metainfo_charset);
 463         }
 464         strbuf_attach(line, out, out_len, out_len);
 465         return 0;
 466 }
 467
 468 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 469 {
 470         char *in, *ep, *cp;
 471         struct strbuf outbuf = STRBUF_INIT, *dec;
 472         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 473         int found_error = 1; /* pessimism */
 474
 475         in = it->buf;
 476         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 477                 int encoding;
 478                 strbuf_reset(&charset_q);
 479                 strbuf_reset(&piecebuf);
 480
 481                 if (in != ep) {
 482                         /*
 483                          * We are about to process an encoded-word
 484                          * that begins at ep, but there is something
 485                          * before the encoded word.
 486                          */
 487                         char *scan;
 488                         for (scan = in; scan < ep; scan++)
 489                                 if (!isspace(*scan))
 490                                         break;
 491
 492                         if (scan != ep || in == it->buf) {
 493                                 /*
 494                                  * We should not lose that "something",
 495                                  * unless we have just processed an
 496                                  * encoded-word, and there is only LWS
 497                                  * before the one we are about to process.
 498                                  */
 499                                 strbuf_add(&outbuf, in, ep - in);
 500                         }
 501                 }
 502                 /* E.g.
 503                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 504                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 505                  */
 506                 ep += 2;
 507
 508                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 509                         goto release_return;
 510
 511                 if (cp + 3 - it->buf > it->len)
 512                         goto release_return;
 513                 strbuf_add(&charset_q, ep, cp - ep);
 514
 515                 encoding = cp[1];
 516                 if (!encoding || cp[2] != '?')
 517                         goto release_return;
 518                 ep = strstr(cp + 3, "?=");
 519                 if (!ep)
 520                         goto release_return;
 521                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 522                 switch (tolower(encoding)) {
 523                 default:
 524                         goto release_return;
 525                 case 'b':
 526                         dec = decode_b_segment(&piecebuf);
 527                         break;
 528                 case 'q':
 529                         dec = decode_q_segment(&piecebuf, 1);
 530                         break;
 531                 }
 532                 if (convert_to_utf8(mi, dec, charset_q.buf))
 533                         goto release_return;
 534
 535                 strbuf_addbuf(&outbuf, dec);
 536                 strbuf_release(dec);
 537                 free(dec);
 538                 in = ep + 2;
 539         }
 540         strbuf_addstr(&outbuf, in);
 541         strbuf_reset(it);
 542         strbuf_addbuf(it, &outbuf);
 543         found_error = 0;
 544 release_return:
 545         strbuf_release(&outbuf);
 546         strbuf_release(&charset_q);
 547         strbuf_release(&piecebuf);
 548
 549         if (found_error)
 550                 mi->input_error = -1;
 551 }
 552
 553 /*
 554  * Returns true if "line" contains a header matching "hdr", in which case "val"
 555  * will contain the value of the header with any RFC2047 B and Q encoding
 556  * unwrapped, and optionally normalize the meta information to utf8.
 557  */
 558 static int parse_header(const struct strbuf *line,
 559                         const char *hdr,
 560                         struct mailinfo *mi,
 561                         struct strbuf *val)
 562 {
 563         const char *val_str;
 564
 565         if (!skip_header(line, hdr, &val_str))
 566                 return 0;
 567         strbuf_addstr(val, val_str);
 568         decode_header(mi, val);
 569         return 1;
 570 }
 571
 572 static int check_header(struct mailinfo *mi,
 573                         const struct strbuf *line,
 574                         struct strbuf *hdr_data[], int overwrite)
 575 {
 576         int i, ret = 0;
 577         struct strbuf sb = STRBUF_INIT;
 578
 579         /* search for the interesting parts */
 580         for (i = 0; header[i]; i++) {
 581                 if ((!hdr_data[i] || overwrite) &&
 582                     parse_header(line, header[i], mi, &sb)) {
 583                         handle_header(&hdr_data[i], &sb);
 584                         ret = 1;
 585                         goto check_header_out;
 586                 }
 587         }
 588
 589         /* Content stuff */
 590         if (parse_header(line, "Content-Type", mi, &sb)) {
 591                 handle_content_type(mi, &sb);
 592                 ret = 1;
 593                 goto check_header_out;
 594         }
 595         if (parse_header(line, "Content-Transfer-Encoding", mi, &sb)) {
 596                 handle_content_transfer_encoding(mi, &sb);
 597                 ret = 1;
 598                 goto check_header_out;
 599         }
 600         if (parse_header(line, "Message-Id", mi, &sb)) {
 601                 if (mi->add_message_id)
 602                         mi->message_id = strbuf_detach(&sb, NULL);
 603                 ret = 1;
 604                 goto check_header_out;
 605         }
 606
 607 check_header_out:
 608         strbuf_release(&sb);
 609         return ret;
 610 }
 611
 612 /*
 613  * Returns 1 if the given line or any line beginning with the given line is an
 614  * in-body header (that is, check_header will succeed when passed
 615  * mi->s_hdr_data).
 616  */
 617 static int is_inbody_header(const struct mailinfo *mi,
 618                             const struct strbuf *line)
 619 {
 620         int i;
 621         const char *val;
 622         for (i = 0; header[i]; i++)
 623                 if (!mi->s_hdr_data[i] && skip_header(line, header[i], &val))
 624                         return 1;
 625         return 0;
 626 }
 627
 628 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 629 {
 630         struct strbuf *ret;
 631
 632         switch (mi->transfer_encoding) {
 633         case TE_QP:
 634                 ret = decode_q_segment(line, 0);
 635                 break;
 636         case TE_BASE64:
 637                 ret = decode_b_segment(line);
 638                 break;
 639         case TE_DONTCARE:
 640         default:
 641                 return;
 642         }
 643         strbuf_reset(line);
 644         strbuf_addbuf(line, ret);
 645         strbuf_release(ret);
 646         free(ret);
 647 }
 648
 649 static inline int patchbreak(const struct strbuf *line)
 650 {
 651         size_t i;
 652
 653         /* Beginning of a "diff -" header? */
 654         if (starts_with(line->buf, "diff -"))
 655                 return 1;
 656
 657         /* CVS "Index: " line? */
 658         if (starts_with(line->buf, "Index: "))
 659                 return 1;
 660
 661         /*
 662          * "--- <filename>" starts patches without headers
 663          * "---<sp>*" is a manual separator
 664          */
 665         if (line->len < 4)
 666                 return 0;
 667
 668         if (starts_with(line->buf, "---")) {
 669                 /* space followed by a filename? */
 670                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 671                         return 1;
 672                 /* Just whitespace? */
 673                 for (i = 3; i < line->len; i++) {
 674                         unsigned char c = line->buf[i];
 675                         if (c == '\n')
 676                                 return 1;
 677                         if (!isspace(c))
 678                                 break;
 679                 }
 680                 return 0;
 681         }
 682         return 0;
 683 }
 684
 685 static int is_scissors_line(const char *line)
 686 {
 687         const char *c;
 688         int scissors = 0, gap = 0;
 689         const char *first_nonblank = NULL, *last_nonblank = NULL;
 690         int visible, perforation = 0, in_perforation = 0;
 691
 692         for (c = line; *c; c++) {
 693                 if (isspace(*c)) {
 694                         if (in_perforation) {
 695                                 perforation++;
 696                                 gap++;
 697                         }
 698                         continue;
 699                 }
 700                 last_nonblank = c;
 701                 if (first_nonblank == NULL)
 702                         first_nonblank = c;
 703                 if (*c == '-') {
 704                         in_perforation = 1;
 705                         perforation++;
 706                         continue;
 707                 }
 708                 if ((!memcmp(c, ">8", 2) || !memcmp(c, "8<", 2) ||
 709                      !memcmp(c, ">%", 2) || !memcmp(c, "%<", 2))) {
 710                         in_perforation = 1;
 711                         perforation += 2;
 712                         scissors += 2;
 713                         c++;
 714                         continue;
 715                 }
 716                 in_perforation = 0;
 717         }
 718
 719         /*
 720          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 721          * Even though there can be arbitrary cruft on the same line
 722          * (e.g. "cut here"), in order to avoid misidentification, the
 723          * perforation must occupy more than a third of the visible
 724          * width of the line, and dashes and scissors must occupy more
 725          * than half of the perforation.
 726          */
 727
 728         if (first_nonblank && last_nonblank)
 729                 visible = last_nonblank - first_nonblank + 1;
 730         else
 731                 visible = 0;
 732         return (scissors && 8 <= visible &&
 733                 visible < perforation * 3 &&
 734                 gap * 2 < perforation);
 735 }
 736
 737 static void flush_inbody_header_accum(struct mailinfo *mi)
 738 {
 739         if (!mi->inbody_header_accum.len)
 740                 return;
 741         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 742                 BUG("inbody_header_accum, if not empty, must always contain a valid in-body header");
 743         strbuf_reset(&mi->inbody_header_accum);
 744 }
 745
 746 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 747 {
 748         if (mi->inbody_header_accum.len &&
 749             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 750                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 751                         /*
 752                          * This is a scissors line; do not consider this line
 753                          * as a header continuation line.
 754                          */
 755                         flush_inbody_header_accum(mi);
 756                         return 0;
 757                 }
 758                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 759                 strbuf_addbuf(&mi->inbody_header_accum, line);
 760                 return 1;
 761         }
 762
 763         flush_inbody_header_accum(mi);
 764
 765         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 766                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 767         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 768                 int i;
 769                 for (i = 0; header[i]; i++)
 770                         if (!strcmp("Subject", header[i])) {
 771                                 handle_header(&mi->s_hdr_data[i], line);
 772                                 return 1;
 773                         }
 774                 return 0;
 775         }
 776         if (is_inbody_header(mi, line)) {
 777                 strbuf_addbuf(&mi->inbody_header_accum, line);
 778                 return 1;
 779         }
 780         return 0;
 781 }
 782
 783 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 784 {
 785         assert(!mi->filter_stage);
 786
 787         if (mi->header_stage) {
 788                 if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 789                         if (mi->inbody_header_accum.len) {
 790                                 flush_inbody_header_accum(mi);
 791                                 mi->header_stage = 0;
 792                         }
 793                         return 0;
 794                 }
 795         }
 796
 797         if (mi->use_inbody_headers && mi->header_stage) {
 798                 mi->header_stage = check_inbody_header(mi, line);
 799                 if (mi->header_stage)
 800                         return 0;
 801         } else
 802                 /* Only trim the first (blank) line of the commit message
 803                  * when ignoring in-body headers.
 804                  */
 805                 mi->header_stage = 0;
 806
 807         /* normalize the log message to UTF-8. */
 808         if (convert_to_utf8(mi, line, mi->charset.buf))
 809                 return 0; /* mi->input_error already set */
 810
 811         if (mi->use_scissors && is_scissors_line(line->buf)) {
 812                 int i;
 813
 814                 strbuf_setlen(&mi->log_message, 0);
 815                 mi->header_stage = 1;
 816
 817                 /*
 818                  * We may have already read "secondary headers"; purge
 819                  * them to give ourselves a clean restart.
 820                  */
 821                 for (i = 0; header[i]; i++) {
 822                         if (mi->s_hdr_data[i])
 823                                 strbuf_release(mi->s_hdr_data[i]);
 824                         FREE_AND_NULL(mi->s_hdr_data[i]);
 825                 }
 826                 return 0;
 827         }
 828
 829         if (patchbreak(line)) {
 830                 if (mi->message_id)
 831                         strbuf_addf(&mi->log_message,
 832                                     "Message-Id: %s\n", mi->message_id);
 833                 return 1;
 834         }
 835
 836         strbuf_addbuf(&mi->log_message, line);
 837         return 0;
 838 }
 839
 840 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 841 {
 842         fwrite(line->buf, 1, line->len, mi->patchfile);
 843         mi->patch_lines++;
 844 }
 845
 846 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 847 {
 848         switch (mi->filter_stage) {
 849         case 0:
 850                 if (!handle_commit_msg(mi, line))
 851                         break;
 852                 mi->filter_stage++;
 853                 /* fallthrough */
 854         case 1:
 855                 handle_patch(mi, line);
 856                 break;
 857         }
 858 }
 859
 860 static int is_rfc2822_header(const struct strbuf *line)
 861 {
 862         /*
 863          * The section that defines the loosest possible
 864          * field name is "3.6.8 Optional fields".
 865          *
 866          * optional-field = field-name ":" unstructured CRLF
 867          * field-name = 1*ftext
 868          * ftext = %d33-57 / %59-126
 869          */
 870         int ch;
 871         char *cp = line->buf;
 872
 873         /* Count mbox From headers as headers */
 874         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 875                 return 1;
 876
 877         while ((ch = *cp++)) {
 878                 if (ch == ':')
 879                         return 1;
 880                 if ((33 <= ch && ch <= 57) ||
 881                     (59 <= ch && ch <= 126))
 882                         continue;
 883                 break;
 884         }
 885         return 0;
 886 }
 887
 888 static int read_one_header_line(struct strbuf *line, FILE *in)
 889 {
 890         struct strbuf continuation = STRBUF_INIT;
 891
 892         /* Get the first part of the line. */
 893         if (strbuf_getline_lf(line, in))
 894                 return 0;
 895
 896         /*
 897          * Is it an empty line or not a valid rfc2822 header?
 898          * If so, stop here, and return false ("not a header")
 899          */
 900         strbuf_rtrim(line);
 901         if (!line->len || !is_rfc2822_header(line)) {
 902                 /* Re-add the newline */
 903                 strbuf_addch(line, '\n');
 904                 return 0;
 905         }
 906
 907         /*
 908          * Now we need to eat all the continuation lines..
 909          * Yuck, 2822 header "folding"
 910          */
 911         for (;;) {
 912                 int peek;
 913
 914                 peek = fgetc(in);
 915                 if (peek == EOF)
 916                         break;
 917                 ungetc(peek, in);
 918                 if (peek != ' ' && peek != '\t')
 919                         break;
 920                 if (strbuf_getline_lf(&continuation, in))
 921                         break;
 922                 continuation.buf[0] = ' ';
 923                 strbuf_rtrim(&continuation);
 924                 strbuf_addbuf(line, &continuation);
 925         }
 926         strbuf_release(&continuation);
 927
 928         return 1;
 929 }
 930
 931 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 932 {
 933         while (!strbuf_getline_lf(line, mi->input)) {
 934                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 935                         return 1;
 936         }
 937         return 0;
 938 }
 939
 940 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 941 {
 942         struct strbuf newline = STRBUF_INIT;
 943
 944         strbuf_addch(&newline, '\n');
 945 again:
 946         if (line->len >= (*(mi->content_top))->len + 2 &&
 947             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 948                 /* we hit an end boundary */
 949                 /* pop the current boundary off the stack */
 950                 strbuf_release(*(mi->content_top));
 951                 FREE_AND_NULL(*(mi->content_top));
 952
 953                 /* technically won't happen as is_multipart_boundary()
 954                    will fail first.  But just in case..
 955                  */
 956                 if (--mi->content_top < mi->content) {
 957                         error("Detected mismatched boundaries, can't recover");
 958                         mi->input_error = -1;
 959                         mi->content_top = mi->content;
 960                         strbuf_release(&newline);
 961                         return 0;
 962                 }
 963                 handle_filter(mi, &newline);
 964                 strbuf_release(&newline);
 965                 if (mi->input_error)
 966                         return 0;
 967
 968                 /* skip to the next boundary */
 969                 if (!find_boundary(mi, line))
 970                         return 0;
 971                 goto again;
 972         }
 973
 974         /* set some defaults */
 975         mi->transfer_encoding = TE_DONTCARE;
 976         strbuf_reset(&mi->charset);
 977
 978         /* slurp in this section's info */
 979         while (read_one_header_line(line, mi->input))
 980                 check_header(mi, line, mi->p_hdr_data, 0);
 981
 982         strbuf_release(&newline);
 983         /* replenish line */
 984         if (strbuf_getline_lf(line, mi->input))
 985                 return 0;
 986         strbuf_addch(line, '\n');
 987         return 1;
 988 }
 989
 990 static void handle_filter_flowed(struct mailinfo *mi, struct strbuf *line,
 991                                  struct strbuf *prev)
 992 {
 993         size_t len = line->len;
 994         const char *rest;
 995
 996         if (!mi->format_flowed) {
 997                 handle_filter(mi, line);
 998                 return;
 999         }
1000
1001         if (line->buf[len - 1] == '\n') {
1002                 len--;
1003                 if (len && line->buf[len - 1] == '\r')
1004                         len--;
1005         }
1006
1007         /* Keep signature separator as-is. */
1008         if (skip_prefix(line->buf, "-- ", &rest) && rest - line->buf == len) {
1009                 if (prev->len) {
1010                         handle_filter(mi, prev);
1011                         strbuf_reset(prev);
1012                 }
1013                 handle_filter(mi, line);
1014                 return;
1015         }
1016
1017         /* Unstuff space-stuffed line. */
1018         if (len && line->buf[0] == ' ') {
1019                 strbuf_remove(line, 0, 1);
1020                 len--;
1021         }
1022
1023         /* Save flowed line for later, but without the soft line break. */
1024         if (len && line->buf[len - 1] == ' ') {
1025                 strbuf_add(prev, line->buf, len - !!mi->delsp);
1026                 return;
1027         }
1028
1029         /* Prepend any previous partial lines */
1030         strbuf_insert(line, 0, prev->buf, prev->len);
1031         strbuf_reset(prev);
1032
1033         handle_filter(mi, line);
1034 }
1035
1036 static void handle_body(struct mailinfo *mi, struct strbuf *line)
1037 {
1038         struct strbuf prev = STRBUF_INIT;
1039
1040         /* Skip up to the first boundary */
1041         if (*(mi->content_top)) {
1042                 if (!find_boundary(mi, line))
1043                         goto handle_body_out;
1044         }
1045
1046         do {
1047                 /* process any boundary lines */
1048                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
1049                         /* flush any leftover */
1050                         if (prev.len) {
1051                                 handle_filter(mi, &prev);
1052                                 strbuf_reset(&prev);
1053                         }
1054                         if (!handle_boundary(mi, line))
1055                                 goto handle_body_out;
1056                 }
1057
1058                 /* Unwrap transfer encoding */
1059                 decode_transfer_encoding(mi, line);
1060
1061                 switch (mi->transfer_encoding) {
1062                 case TE_BASE64:
1063                 case TE_QP:
1064                 {
1065                         struct strbuf **lines, **it, *sb;
1066
1067                         /* Prepend any previous partial lines */
1068                         strbuf_insert(line, 0, prev.buf, prev.len);
1069                         strbuf_reset(&prev);
1070
1071                         /*
1072                          * This is a decoded line that may contain
1073                          * multiple new lines.  Pass only one chunk
1074                          * at a time to handle_filter()
1075                          */
1076                         lines = strbuf_split(line, '\n');
1077                         for (it = lines; (sb = *it); it++) {
1078                                 if (*(it + 1) == NULL) /* The last line */
1079                                         if (sb->buf[sb->len - 1] != '\n') {
1080                                                 /* Partial line, save it for later. */
1081                                                 strbuf_addbuf(&prev, sb);
1082                                                 break;
1083                                         }
1084                                 handle_filter_flowed(mi, sb, &prev);
1085                         }
1086                         /*
1087                          * The partial chunk is saved in "prev" and will be
1088                          * appended by the next iteration of read_line_with_nul().
1089                          */
1090                         strbuf_list_free(lines);
1091                         break;
1092                 }
1093                 default:
1094                         handle_filter_flowed(mi, line, &prev);
1095                 }
1096
1097                 if (mi->input_error)
1098                         break;
1099         } while (!strbuf_getwholeline(line, mi->input, '\n'));
1100
1101         if (prev.len)
1102                 handle_filter(mi, &prev);
1103
1104         flush_inbody_header_accum(mi);
1105
1106 handle_body_out:
1107         strbuf_release(&prev);
1108 }
1109
1110 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
1111 {
1112         const char *sp = data->buf;
1113         while (1) {
1114                 char *ep = strchr(sp, '\n');
1115                 int len;
1116                 if (!ep)
1117                         len = strlen(sp);
1118                 else
1119                         len = ep - sp;
1120                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
1121                 if (!ep)
1122                         break;
1123                 sp = ep + 1;
1124         }
1125 }
1126
1127 static void handle_info(struct mailinfo *mi)
1128 {
1129         struct strbuf *hdr;
1130         int i;
1131
1132         for (i = 0; header[i]; i++) {
1133                 /* only print inbody headers if we output a patch file */
1134                 if (mi->patch_lines && mi->s_hdr_data[i])
1135                         hdr = mi->s_hdr_data[i];
1136                 else if (mi->p_hdr_data[i])
1137                         hdr = mi->p_hdr_data[i];
1138                 else
1139                         continue;
1140
1141                 if (memchr(hdr->buf, '\0', hdr->len)) {
1142                         error("a NUL byte in '%s' is not allowed.", header[i]);
1143                         mi->input_error = -1;
1144                 }
1145
1146                 if (!strcmp(header[i], "Subject")) {
1147                         if (!mi->keep_subject) {
1148                                 cleanup_subject(mi, hdr);
1149                                 cleanup_space(hdr);
1150                         }
1151                         output_header_lines(mi->output, "Subject", hdr);
1152                 } else if (!strcmp(header[i], "From")) {
1153                         cleanup_space(hdr);
1154                         handle_from(mi, hdr);
1155                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
1156                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
1157                 } else {
1158                         cleanup_space(hdr);
1159                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
1160                 }
1161         }
1162         fprintf(mi->output, "\n");
1163 }
1164
1165 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
1166 {
1167         FILE *cmitmsg;
1168         int peek;
1169         struct strbuf line = STRBUF_INIT;
1170
1171         cmitmsg = fopen(msg, "w");
1172         if (!cmitmsg) {
1173                 perror(msg);
1174                 return -1;
1175         }
1176         mi->patchfile = fopen(patch, "w");
1177         if (!mi->patchfile) {
1178                 perror(patch);
1179                 fclose(cmitmsg);
1180                 return -1;
1181         }
1182
1183         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
1184         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
1185
1186         do {
1187                 peek = fgetc(mi->input);
1188                 if (peek == EOF) {
1189                         fclose(cmitmsg);
1190                         return error("empty patch: '%s'", patch);
1191                 }
1192         } while (isspace(peek));
1193         ungetc(peek, mi->input);
1194
1195         /* process the email header */
1196         while (read_one_header_line(&line, mi->input))
1197                 check_header(mi, &line, mi->p_hdr_data, 1);
1198
1199         handle_body(mi, &line);
1200         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1201         fclose(cmitmsg);
1202         fclose(mi->patchfile);
1203
1204         handle_info(mi);
1205         strbuf_release(&line);
1206         return mi->input_error;
1207 }
1208
1209 static int git_mailinfo_config(const char *var, const char *value, void *mi_)
1210 {
1211         struct mailinfo *mi = mi_;
1212
1213         if (!starts_with(var, "mailinfo."))
1214                 return git_default_config(var, value, NULL);
1215         if (!strcmp(var, "mailinfo.scissors")) {
1216                 mi->use_scissors = git_config_bool(var, value);
1217                 return 0;
1218         }
1219         /* perhaps others here */
1220         return 0;
1221 }
1222
1223 void setup_mailinfo(struct mailinfo *mi)
1224 {
1225         memset(mi, 0, sizeof(*mi));
1226         strbuf_init(&mi->name, 0);
1227         strbuf_init(&mi->email, 0);
1228         strbuf_init(&mi->charset, 0);
1229         strbuf_init(&mi->log_message, 0);
1230         strbuf_init(&mi->inbody_header_accum, 0);
1231         mi->header_stage = 1;
1232         mi->use_inbody_headers = 1;
1233         mi->content_top = mi->content;
1234         git_config(git_mailinfo_config, mi);
1235 }
1236
1237 void clear_mailinfo(struct mailinfo *mi)
1238 {
1239         strbuf_release(&mi->name);
1240         strbuf_release(&mi->email);
1241         strbuf_release(&mi->charset);
1242         strbuf_release(&mi->inbody_header_accum);
1243         free(mi->message_id);
1244
1245         strbuf_list_free(mi->p_hdr_data);
1246         strbuf_list_free(mi->s_hdr_data);
1247
1248         while (mi->content < mi->content_top) {
1249                 free(*(mi->content_top));
1250                 mi->content_top--;
1251         }
1252
1253         strbuf_release(&mi->log_message);
1254 }