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
 451         if (!mi->metainfo_charset || !charset || !*charset)
 452                 return 0;
 453
 454         if (same_encoding(mi->metainfo_charset, charset))
 455                 return 0;
 456         out = reencode_string(line->buf, mi->metainfo_charset, charset);
 457         if (!out) {
 458                 mi->input_error = -1;
 459                 return error("cannot convert from %s to %s",
 460                              charset, mi->metainfo_charset);
 461         }
 462         strbuf_attach(line, out, strlen(out), strlen(out));
 463         return 0;
 464 }
 465
 466 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 467 {
 468         char *in, *ep, *cp;
 469         struct strbuf outbuf = STRBUF_INIT, *dec;
 470         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 471         int found_error = 1; /* pessimism */
 472
 473         in = it->buf;
 474         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 475                 int encoding;
 476                 strbuf_reset(&charset_q);
 477                 strbuf_reset(&piecebuf);
 478
 479                 if (in != ep) {
 480                         /*
 481                          * We are about to process an encoded-word
 482                          * that begins at ep, but there is something
 483                          * before the encoded word.
 484                          */
 485                         char *scan;
 486                         for (scan = in; scan < ep; scan++)
 487                                 if (!isspace(*scan))
 488                                         break;
 489
 490                         if (scan != ep || in == it->buf) {
 491                                 /*
 492                                  * We should not lose that "something",
 493                                  * unless we have just processed an
 494                                  * encoded-word, and there is only LWS
 495                                  * before the one we are about to process.
 496                                  */
 497                                 strbuf_add(&outbuf, in, ep - in);
 498                         }
 499                 }
 500                 /* E.g.
 501                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 502                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 503                  */
 504                 ep += 2;
 505
 506                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 507                         goto release_return;
 508
 509                 if (cp + 3 - it->buf > it->len)
 510                         goto release_return;
 511                 strbuf_add(&charset_q, ep, cp - ep);
 512
 513                 encoding = cp[1];
 514                 if (!encoding || cp[2] != '?')
 515                         goto release_return;
 516                 ep = strstr(cp + 3, "?=");
 517                 if (!ep)
 518                         goto release_return;
 519                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 520                 switch (tolower(encoding)) {
 521                 default:
 522                         goto release_return;
 523                 case 'b':
 524                         dec = decode_b_segment(&piecebuf);
 525                         break;
 526                 case 'q':
 527                         dec = decode_q_segment(&piecebuf, 1);
 528                         break;
 529                 }
 530                 if (convert_to_utf8(mi, dec, charset_q.buf))
 531                         goto release_return;
 532
 533                 strbuf_addbuf(&outbuf, dec);
 534                 strbuf_release(dec);
 535                 free(dec);
 536                 in = ep + 2;
 537         }
 538         strbuf_addstr(&outbuf, in);
 539         strbuf_reset(it);
 540         strbuf_addbuf(it, &outbuf);
 541         found_error = 0;
 542 release_return:
 543         strbuf_release(&outbuf);
 544         strbuf_release(&charset_q);
 545         strbuf_release(&piecebuf);
 546
 547         if (found_error)
 548                 mi->input_error = -1;
 549 }
 550
 551 /*
 552  * Returns true if "line" contains a header matching "hdr", in which case "val"
 553  * will contain the value of the header with any RFC2047 B and Q encoding
 554  * unwrapped, and optionally normalize the meta information to utf8.
 555  */
 556 static int parse_header(const struct strbuf *line,
 557                         const char *hdr,
 558                         struct mailinfo *mi,
 559                         struct strbuf *val)
 560 {
 561         const char *val_str;
 562
 563         if (!skip_header(line, hdr, &val_str))
 564                 return 0;
 565         strbuf_addstr(val, val_str);
 566         decode_header(mi, val);
 567         return 1;
 568 }
 569
 570 static int check_header(struct mailinfo *mi,
 571                         const struct strbuf *line,
 572                         struct strbuf *hdr_data[], int overwrite)
 573 {
 574         int i, ret = 0;
 575         struct strbuf sb = STRBUF_INIT;
 576
 577         /* search for the interesting parts */
 578         for (i = 0; header[i]; i++) {
 579                 if ((!hdr_data[i] || overwrite) &&
 580                     parse_header(line, header[i], mi, &sb)) {
 581                         handle_header(&hdr_data[i], &sb);
 582                         ret = 1;
 583                         goto check_header_out;
 584                 }
 585         }
 586
 587         /* Content stuff */
 588         if (parse_header(line, "Content-Type", mi, &sb)) {
 589                 handle_content_type(mi, &sb);
 590                 ret = 1;
 591                 goto check_header_out;
 592         }
 593         if (parse_header(line, "Content-Transfer-Encoding", mi, &sb)) {
 594                 handle_content_transfer_encoding(mi, &sb);
 595                 ret = 1;
 596                 goto check_header_out;
 597         }
 598         if (parse_header(line, "Message-Id", mi, &sb)) {
 599                 if (mi->add_message_id)
 600                         mi->message_id = strbuf_detach(&sb, NULL);
 601                 ret = 1;
 602                 goto check_header_out;
 603         }
 604
 605 check_header_out:
 606         strbuf_release(&sb);
 607         return ret;
 608 }
 609
 610 /*
 611  * Returns 1 if the given line or any line beginning with the given line is an
 612  * in-body header (that is, check_header will succeed when passed
 613  * mi->s_hdr_data).
 614  */
 615 static int is_inbody_header(const struct mailinfo *mi,
 616                             const struct strbuf *line)
 617 {
 618         int i;
 619         const char *val;
 620         for (i = 0; header[i]; i++)
 621                 if (!mi->s_hdr_data[i] && skip_header(line, header[i], &val))
 622                         return 1;
 623         return 0;
 624 }
 625
 626 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 627 {
 628         struct strbuf *ret;
 629
 630         switch (mi->transfer_encoding) {
 631         case TE_QP:
 632                 ret = decode_q_segment(line, 0);
 633                 break;
 634         case TE_BASE64:
 635                 ret = decode_b_segment(line);
 636                 break;
 637         case TE_DONTCARE:
 638         default:
 639                 return;
 640         }
 641         strbuf_reset(line);
 642         strbuf_addbuf(line, ret);
 643         strbuf_release(ret);
 644         free(ret);
 645 }
 646
 647 static inline int patchbreak(const struct strbuf *line)
 648 {
 649         size_t i;
 650
 651         /* Beginning of a "diff -" header? */
 652         if (starts_with(line->buf, "diff -"))
 653                 return 1;
 654
 655         /* CVS "Index: " line? */
 656         if (starts_with(line->buf, "Index: "))
 657                 return 1;
 658
 659         /*
 660          * "--- <filename>" starts patches without headers
 661          * "---<sp>*" is a manual separator
 662          */
 663         if (line->len < 4)
 664                 return 0;
 665
 666         if (starts_with(line->buf, "---")) {
 667                 /* space followed by a filename? */
 668                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 669                         return 1;
 670                 /* Just whitespace? */
 671                 for (i = 3; i < line->len; i++) {
 672                         unsigned char c = line->buf[i];
 673                         if (c == '\n')
 674                                 return 1;
 675                         if (!isspace(c))
 676                                 break;
 677                 }
 678                 return 0;
 679         }
 680         return 0;
 681 }
 682
 683 static int is_scissors_line(const char *line)
 684 {
 685         const char *c;
 686         int scissors = 0, gap = 0;
 687         const char *first_nonblank = NULL, *last_nonblank = NULL;
 688         int visible, perforation = 0, in_perforation = 0;
 689
 690         for (c = line; *c; c++) {
 691                 if (isspace(*c)) {
 692                         if (in_perforation) {
 693                                 perforation++;
 694                                 gap++;
 695                         }
 696                         continue;
 697                 }
 698                 last_nonblank = c;
 699                 if (first_nonblank == NULL)
 700                         first_nonblank = c;
 701                 if (*c == '-') {
 702                         in_perforation = 1;
 703                         perforation++;
 704                         continue;
 705                 }
 706                 if ((!memcmp(c, ">8", 2) || !memcmp(c, "8<", 2) ||
 707                      !memcmp(c, ">%", 2) || !memcmp(c, "%<", 2))) {
 708                         in_perforation = 1;
 709                         perforation += 2;
 710                         scissors += 2;
 711                         c++;
 712                         continue;
 713                 }
 714                 in_perforation = 0;
 715         }
 716
 717         /*
 718          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 719          * Even though there can be arbitrary cruft on the same line
 720          * (e.g. "cut here"), in order to avoid misidentification, the
 721          * perforation must occupy more than a third of the visible
 722          * width of the line, and dashes and scissors must occupy more
 723          * than half of the perforation.
 724          */
 725
 726         if (first_nonblank && last_nonblank)
 727                 visible = last_nonblank - first_nonblank + 1;
 728         else
 729                 visible = 0;
 730         return (scissors && 8 <= visible &&
 731                 visible < perforation * 3 &&
 732                 gap * 2 < perforation);
 733 }
 734
 735 static void flush_inbody_header_accum(struct mailinfo *mi)
 736 {
 737         if (!mi->inbody_header_accum.len)
 738                 return;
 739         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 740                 BUG("inbody_header_accum, if not empty, must always contain a valid in-body header");
 741         strbuf_reset(&mi->inbody_header_accum);
 742 }
 743
 744 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 745 {
 746         if (mi->inbody_header_accum.len &&
 747             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 748                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 749                         /*
 750                          * This is a scissors line; do not consider this line
 751                          * as a header continuation line.
 752                          */
 753                         flush_inbody_header_accum(mi);
 754                         return 0;
 755                 }
 756                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 757                 strbuf_addbuf(&mi->inbody_header_accum, line);
 758                 return 1;
 759         }
 760
 761         flush_inbody_header_accum(mi);
 762
 763         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 764                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 765         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 766                 int i;
 767                 for (i = 0; header[i]; i++)
 768                         if (!strcmp("Subject", header[i])) {
 769                                 handle_header(&mi->s_hdr_data[i], line);
 770                                 return 1;
 771                         }
 772                 return 0;
 773         }
 774         if (is_inbody_header(mi, line)) {
 775                 strbuf_addbuf(&mi->inbody_header_accum, line);
 776                 return 1;
 777         }
 778         return 0;
 779 }
 780
 781 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 782 {
 783         assert(!mi->filter_stage);
 784
 785         if (mi->header_stage) {
 786                 if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 787                         if (mi->inbody_header_accum.len) {
 788                                 flush_inbody_header_accum(mi);
 789                                 mi->header_stage = 0;
 790                         }
 791                         return 0;
 792                 }
 793         }
 794
 795         if (mi->use_inbody_headers && mi->header_stage) {
 796                 mi->header_stage = check_inbody_header(mi, line);
 797                 if (mi->header_stage)
 798                         return 0;
 799         } else
 800                 /* Only trim the first (blank) line of the commit message
 801                  * when ignoring in-body headers.
 802                  */
 803                 mi->header_stage = 0;
 804
 805         /* normalize the log message to UTF-8. */
 806         if (convert_to_utf8(mi, line, mi->charset.buf))
 807                 return 0; /* mi->input_error already set */
 808
 809         if (mi->use_scissors && is_scissors_line(line->buf)) {
 810                 int i;
 811
 812                 strbuf_setlen(&mi->log_message, 0);
 813                 mi->header_stage = 1;
 814
 815                 /*
 816                  * We may have already read "secondary headers"; purge
 817                  * them to give ourselves a clean restart.
 818                  */
 819                 for (i = 0; header[i]; i++) {
 820                         if (mi->s_hdr_data[i])
 821                                 strbuf_release(mi->s_hdr_data[i]);
 822                         mi->s_hdr_data[i] = NULL;
 823                 }
 824                 return 0;
 825         }
 826
 827         if (patchbreak(line)) {
 828                 if (mi->message_id)
 829                         strbuf_addf(&mi->log_message,
 830                                     "Message-Id: %s\n", mi->message_id);
 831                 return 1;
 832         }
 833
 834         strbuf_addbuf(&mi->log_message, line);
 835         return 0;
 836 }
 837
 838 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 839 {
 840         fwrite(line->buf, 1, line->len, mi->patchfile);
 841         mi->patch_lines++;
 842 }
 843
 844 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 845 {
 846         switch (mi->filter_stage) {
 847         case 0:
 848                 if (!handle_commit_msg(mi, line))
 849                         break;
 850                 mi->filter_stage++;
 851                 /* fallthrough */
 852         case 1:
 853                 handle_patch(mi, line);
 854                 break;
 855         }
 856 }
 857
 858 static int is_rfc2822_header(const struct strbuf *line)
 859 {
 860         /*
 861          * The section that defines the loosest possible
 862          * field name is "3.6.8 Optional fields".
 863          *
 864          * optional-field = field-name ":" unstructured CRLF
 865          * field-name = 1*ftext
 866          * ftext = %d33-57 / %59-126
 867          */
 868         int ch;
 869         char *cp = line->buf;
 870
 871         /* Count mbox From headers as headers */
 872         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 873                 return 1;
 874
 875         while ((ch = *cp++)) {
 876                 if (ch == ':')
 877                         return 1;
 878                 if ((33 <= ch && ch <= 57) ||
 879                     (59 <= ch && ch <= 126))
 880                         continue;
 881                 break;
 882         }
 883         return 0;
 884 }
 885
 886 static int read_one_header_line(struct strbuf *line, FILE *in)
 887 {
 888         struct strbuf continuation = STRBUF_INIT;
 889
 890         /* Get the first part of the line. */
 891         if (strbuf_getline_lf(line, in))
 892                 return 0;
 893
 894         /*
 895          * Is it an empty line or not a valid rfc2822 header?
 896          * If so, stop here, and return false ("not a header")
 897          */
 898         strbuf_rtrim(line);
 899         if (!line->len || !is_rfc2822_header(line)) {
 900                 /* Re-add the newline */
 901                 strbuf_addch(line, '\n');
 902                 return 0;
 903         }
 904
 905         /*
 906          * Now we need to eat all the continuation lines..
 907          * Yuck, 2822 header "folding"
 908          */
 909         for (;;) {
 910                 int peek;
 911
 912                 peek = fgetc(in);
 913                 if (peek == EOF)
 914                         break;
 915                 ungetc(peek, in);
 916                 if (peek != ' ' && peek != '\t')
 917                         break;
 918                 if (strbuf_getline_lf(&continuation, in))
 919                         break;
 920                 continuation.buf[0] = ' ';
 921                 strbuf_rtrim(&continuation);
 922                 strbuf_addbuf(line, &continuation);
 923         }
 924         strbuf_release(&continuation);
 925
 926         return 1;
 927 }
 928
 929 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 930 {
 931         while (!strbuf_getline_lf(line, mi->input)) {
 932                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 933                         return 1;
 934         }
 935         return 0;
 936 }
 937
 938 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 939 {
 940         struct strbuf newline = STRBUF_INIT;
 941
 942         strbuf_addch(&newline, '\n');
 943 again:
 944         if (line->len >= (*(mi->content_top))->len + 2 &&
 945             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 946                 /* we hit an end boundary */
 947                 /* pop the current boundary off the stack */
 948                 strbuf_release(*(mi->content_top));
 949                 FREE_AND_NULL(*(mi->content_top));
 950
 951                 /* technically won't happen as is_multipart_boundary()
 952                    will fail first.  But just in case..
 953                  */
 954                 if (--mi->content_top < mi->content) {
 955                         error("Detected mismatched boundaries, can't recover");
 956                         mi->input_error = -1;
 957                         mi->content_top = mi->content;
 958                         strbuf_release(&newline);
 959                         return 0;
 960                 }
 961                 handle_filter(mi, &newline);
 962                 strbuf_release(&newline);
 963                 if (mi->input_error)
 964                         return 0;
 965
 966                 /* skip to the next boundary */
 967                 if (!find_boundary(mi, line))
 968                         return 0;
 969                 goto again;
 970         }
 971
 972         /* set some defaults */
 973         mi->transfer_encoding = TE_DONTCARE;
 974         strbuf_reset(&mi->charset);
 975
 976         /* slurp in this section's info */
 977         while (read_one_header_line(line, mi->input))
 978                 check_header(mi, line, mi->p_hdr_data, 0);
 979
 980         strbuf_release(&newline);
 981         /* replenish line */
 982         if (strbuf_getline_lf(line, mi->input))
 983                 return 0;
 984         strbuf_addch(line, '\n');
 985         return 1;
 986 }
 987
 988 static void handle_filter_flowed(struct mailinfo *mi, struct strbuf *line,
 989                                  struct strbuf *prev)
 990 {
 991         size_t len = line->len;
 992         const char *rest;
 993
 994         if (!mi->format_flowed) {
 995                 handle_filter(mi, line);
 996                 return;
 997         }
 998
 999         if (line->buf[len - 1] == '\n') {
1000                 len--;
1001                 if (len && line->buf[len - 1] == '\r')
1002                         len--;
1003         }
1004
1005         /* Keep signature separator as-is. */
1006         if (skip_prefix(line->buf, "-- ", &rest) && rest - line->buf == len) {
1007                 if (prev->len) {
1008                         handle_filter(mi, prev);
1009                         strbuf_reset(prev);
1010                 }
1011                 handle_filter(mi, line);
1012                 return;
1013         }
1014
1015         /* Unstuff space-stuffed line. */
1016         if (len && line->buf[0] == ' ') {
1017                 strbuf_remove(line, 0, 1);
1018                 len--;
1019         }
1020
1021         /* Save flowed line for later, but without the soft line break. */
1022         if (len && line->buf[len - 1] == ' ') {
1023                 strbuf_add(prev, line->buf, len - !!mi->delsp);
1024                 return;
1025         }
1026
1027         /* Prepend any previous partial lines */
1028         strbuf_insert(line, 0, prev->buf, prev->len);
1029         strbuf_reset(prev);
1030
1031         handle_filter(mi, line);
1032 }
1033
1034 static void handle_body(struct mailinfo *mi, struct strbuf *line)
1035 {
1036         struct strbuf prev = STRBUF_INIT;
1037
1038         /* Skip up to the first boundary */
1039         if (*(mi->content_top)) {
1040                 if (!find_boundary(mi, line))
1041                         goto handle_body_out;
1042         }
1043
1044         do {
1045                 /* process any boundary lines */
1046                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
1047                         /* flush any leftover */
1048                         if (prev.len) {
1049                                 handle_filter(mi, &prev);
1050                                 strbuf_reset(&prev);
1051                         }
1052                         if (!handle_boundary(mi, line))
1053                                 goto handle_body_out;
1054                 }
1055
1056                 /* Unwrap transfer encoding */
1057                 decode_transfer_encoding(mi, line);
1058
1059                 switch (mi->transfer_encoding) {
1060                 case TE_BASE64:
1061                 case TE_QP:
1062                 {
1063                         struct strbuf **lines, **it, *sb;
1064
1065                         /* Prepend any previous partial lines */
1066                         strbuf_insert(line, 0, prev.buf, prev.len);
1067                         strbuf_reset(&prev);
1068
1069                         /*
1070                          * This is a decoded line that may contain
1071                          * multiple new lines.  Pass only one chunk
1072                          * at a time to handle_filter()
1073                          */
1074                         lines = strbuf_split(line, '\n');
1075                         for (it = lines; (sb = *it); it++) {
1076                                 if (*(it + 1) == NULL) /* The last line */
1077                                         if (sb->buf[sb->len - 1] != '\n') {
1078                                                 /* Partial line, save it for later. */
1079                                                 strbuf_addbuf(&prev, sb);
1080                                                 break;
1081                                         }
1082                                 handle_filter_flowed(mi, sb, &prev);
1083                         }
1084                         /*
1085                          * The partial chunk is saved in "prev" and will be
1086                          * appended by the next iteration of read_line_with_nul().
1087                          */
1088                         strbuf_list_free(lines);
1089                         break;
1090                 }
1091                 default:
1092                         handle_filter_flowed(mi, line, &prev);
1093                 }
1094
1095                 if (mi->input_error)
1096                         break;
1097         } while (!strbuf_getwholeline(line, mi->input, '\n'));
1098
1099         if (prev.len)
1100                 handle_filter(mi, &prev);
1101
1102         flush_inbody_header_accum(mi);
1103
1104 handle_body_out:
1105         strbuf_release(&prev);
1106 }
1107
1108 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
1109 {
1110         const char *sp = data->buf;
1111         while (1) {
1112                 char *ep = strchr(sp, '\n');
1113                 int len;
1114                 if (!ep)
1115                         len = strlen(sp);
1116                 else
1117                         len = ep - sp;
1118                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
1119                 if (!ep)
1120                         break;
1121                 sp = ep + 1;
1122         }
1123 }
1124
1125 static void handle_info(struct mailinfo *mi)
1126 {
1127         struct strbuf *hdr;
1128         int i;
1129
1130         for (i = 0; header[i]; i++) {
1131                 /* only print inbody headers if we output a patch file */
1132                 if (mi->patch_lines && mi->s_hdr_data[i])
1133                         hdr = mi->s_hdr_data[i];
1134                 else if (mi->p_hdr_data[i])
1135                         hdr = mi->p_hdr_data[i];
1136                 else
1137                         continue;
1138
1139                 if (!strcmp(header[i], "Subject")) {
1140                         if (!mi->keep_subject) {
1141                                 cleanup_subject(mi, hdr);
1142                                 cleanup_space(hdr);
1143                         }
1144                         output_header_lines(mi->output, "Subject", hdr);
1145                 } else if (!strcmp(header[i], "From")) {
1146                         cleanup_space(hdr);
1147                         handle_from(mi, hdr);
1148                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
1149                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
1150                 } else {
1151                         cleanup_space(hdr);
1152                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
1153                 }
1154         }
1155         fprintf(mi->output, "\n");
1156 }
1157
1158 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
1159 {
1160         FILE *cmitmsg;
1161         int peek;
1162         struct strbuf line = STRBUF_INIT;
1163
1164         cmitmsg = fopen(msg, "w");
1165         if (!cmitmsg) {
1166                 perror(msg);
1167                 return -1;
1168         }
1169         mi->patchfile = fopen(patch, "w");
1170         if (!mi->patchfile) {
1171                 perror(patch);
1172                 fclose(cmitmsg);
1173                 return -1;
1174         }
1175
1176         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
1177         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
1178
1179         do {
1180                 peek = fgetc(mi->input);
1181                 if (peek == EOF) {
1182                         fclose(cmitmsg);
1183                         return error("empty patch: '%s'", patch);
1184                 }
1185         } while (isspace(peek));
1186         ungetc(peek, mi->input);
1187
1188         /* process the email header */
1189         while (read_one_header_line(&line, mi->input))
1190                 check_header(mi, &line, mi->p_hdr_data, 1);
1191
1192         handle_body(mi, &line);
1193         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1194         fclose(cmitmsg);
1195         fclose(mi->patchfile);
1196
1197         handle_info(mi);
1198         strbuf_release(&line);
1199         return mi->input_error;
1200 }
1201
1202 static int git_mailinfo_config(const char *var, const char *value, void *mi_)
1203 {
1204         struct mailinfo *mi = mi_;
1205
1206         if (!starts_with(var, "mailinfo."))
1207                 return git_default_config(var, value, NULL);
1208         if (!strcmp(var, "mailinfo.scissors")) {
1209                 mi->use_scissors = git_config_bool(var, value);
1210                 return 0;
1211         }
1212         /* perhaps others here */
1213         return 0;
1214 }
1215
1216 void setup_mailinfo(struct mailinfo *mi)
1217 {
1218         memset(mi, 0, sizeof(*mi));
1219         strbuf_init(&mi->name, 0);
1220         strbuf_init(&mi->email, 0);
1221         strbuf_init(&mi->charset, 0);
1222         strbuf_init(&mi->log_message, 0);
1223         strbuf_init(&mi->inbody_header_accum, 0);
1224         mi->header_stage = 1;
1225         mi->use_inbody_headers = 1;
1226         mi->content_top = mi->content;
1227         git_config(git_mailinfo_config, mi);
1228 }
1229
1230 void clear_mailinfo(struct mailinfo *mi)
1231 {
1232         int i;
1233
1234         strbuf_release(&mi->name);
1235         strbuf_release(&mi->email);
1236         strbuf_release(&mi->charset);
1237         strbuf_release(&mi->inbody_header_accum);
1238         free(mi->message_id);
1239
1240         if (mi->p_hdr_data)
1241                 for (i = 0; mi->p_hdr_data[i]; i++)
1242                         strbuf_release(mi->p_hdr_data[i]);
1243         free(mi->p_hdr_data);
1244         if (mi->s_hdr_data)
1245                 for (i = 0; mi->s_hdr_data[i]; i++)
1246                         strbuf_release(mi->s_hdr_data[i]);
1247         free(mi->s_hdr_data);
1248
1249         while (mi->content < mi->content_top) {
1250                 free(*(mi->content_top));
1251                 mi->content_top--;
1252         }
1253
1254         strbuf_release(&mi->log_message);
1255 }