mailinfo.c

   1 #include "cache.h"
   2 #include "config.h"
   3 #include "utf8.h"
   4 #include "strbuf.h"
   5 #include "mailinfo.h"
   6
   7 static void cleanup_space(struct strbuf *sb)
   8 {
   9         size_t pos, cnt;
  10         for (pos = 0; pos < sb->len; pos++) {
  11                 if (isspace(sb->buf[pos])) {
  12                         sb->buf[pos] = ' ';
  13                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
  14                         strbuf_remove(sb, pos + 1, cnt);
  15                 }
  16         }
  17 }
  18
  19 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  20 {
  21         struct strbuf *src = name;
  22         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  23                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  24                 src = email;
  25         else if (name == out)
  26                 return;
  27         strbuf_reset(out);
  28         strbuf_addbuf(out, src);
  29 }
  30
  31 static void parse_bogus_from(struct mailinfo *mi, const struct strbuf *line)
  32 {
  33         /* John Doe <johndoe> */
  34
  35         char *bra, *ket;
  36         /* This is fallback, so do not bother if we already have an
  37          * e-mail address.
  38          */
  39         if (mi->email.len)
  40                 return;
  41
  42         bra = strchr(line->buf, '<');
  43         if (!bra)
  44                 return;
  45         ket = strchr(bra, '>');
  46         if (!ket)
  47                 return;
  48
  49         strbuf_reset(&mi->email);
  50         strbuf_add(&mi->email, bra + 1, ket - bra - 1);
  51
  52         strbuf_reset(&mi->name);
  53         strbuf_add(&mi->name, line->buf, bra - line->buf);
  54         strbuf_trim(&mi->name);
  55         get_sane_name(&mi->name, &mi->name, &mi->email);
  56 }
  57
  58 static const char *unquote_comment(struct strbuf *outbuf, const char *in)
  59 {
  60         int c;
  61         int take_next_literally = 0;
  62
  63         strbuf_addch(outbuf, '(');
  64
  65         while ((c = *in++) != 0) {
  66                 if (take_next_literally == 1) {
  67                         take_next_literally = 0;
  68                 } else {
  69                         switch (c) {
  70                         case '\\':
  71                                 take_next_literally = 1;
  72                                 continue;
  73                         case '(':
  74                                 in = unquote_comment(outbuf, in);
  75                                 continue;
  76                         case ')':
  77                                 strbuf_addch(outbuf, ')');
  78                                 return in;
  79                         }
  80                 }
  81
  82                 strbuf_addch(outbuf, c);
  83         }
  84
  85         return in;
  86 }
  87
  88 static const char *unquote_quoted_string(struct strbuf *outbuf, const char *in)
  89 {
  90         int c;
  91         int take_next_literally = 0;
  92
  93         while ((c = *in++) != 0) {
  94                 if (take_next_literally == 1) {
  95                         take_next_literally = 0;
  96                 } else {
  97                         switch (c) {
  98                         case '\\':
  99                                 take_next_literally = 1;
 100                                 continue;
 101                         case '"':
 102                                 return in;
 103                         }
 104                 }
 105
 106                 strbuf_addch(outbuf, c);
 107         }
 108
 109         return in;
 110 }
 111
 112 static void unquote_quoted_pair(struct strbuf *line)
 113 {
 114         struct strbuf outbuf;
 115         const char *in = line->buf;
 116         int c;
 117
 118         strbuf_init(&outbuf, line->len);
 119
 120         while ((c = *in++) != 0) {
 121                 switch (c) {
 122                 case '"':
 123                         in = unquote_quoted_string(&outbuf, in);
 124                         continue;
 125                 case '(':
 126                         in = unquote_comment(&outbuf, in);
 127                         continue;
 128                 }
 129
 130                 strbuf_addch(&outbuf, c);
 131         }
 132
 133         strbuf_swap(&outbuf, line);
 134         strbuf_release(&outbuf);
 135
 136 }
 137
 138 static void handle_from(struct mailinfo *mi, const struct strbuf *from)
 139 {
 140         char *at;
 141         size_t el;
 142         struct strbuf f;
 143
 144         strbuf_init(&f, from->len);
 145         strbuf_addbuf(&f, from);
 146
 147         unquote_quoted_pair(&f);
 148
 149         at = strchr(f.buf, '@');
 150         if (!at) {
 151                 parse_bogus_from(mi, from);
 152                 goto out;
 153         }
 154
 155         /*
 156          * If we already have one email, don't take any confusing lines
 157          */
 158         if (mi->email.len && strchr(at + 1, '@'))
 159                 goto out;
 160
 161         /* Pick up the string around '@', possibly delimited with <>
 162          * pair; that is the email part.
 163          */
 164         while (at > f.buf) {
 165                 char c = at[-1];
 166                 if (isspace(c))
 167                         break;
 168                 if (c == '<') {
 169                         at[-1] = ' ';
 170                         break;
 171                 }
 172                 at--;
 173         }
 174         el = strcspn(at, " \n\t\r\v\f>");
 175         strbuf_reset(&mi->email);
 176         strbuf_add(&mi->email, at, el);
 177         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 178
 179         /* The remainder is name.  It could be
 180          *
 181          * - "John Doe <john.doe@xz>"                   (a), or
 182          * - "john.doe@xz (John Doe)"                   (b), or
 183          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 184          *
 185          * but we have removed the email part, so
 186          *
 187          * - remove extra spaces which could stay after email (case 'c'), and
 188          * - trim from both ends, possibly removing the () pair at the end
 189          *   (cases 'a' and 'b').
 190          */
 191         cleanup_space(&f);
 192         strbuf_trim(&f);
 193         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 194                 strbuf_remove(&f, 0, 1);
 195                 strbuf_setlen(&f, f.len - 1);
 196         }
 197
 198         get_sane_name(&mi->name, &f, &mi->email);
 199 out:
 200         strbuf_release(&f);
 201 }
 202
 203 static void handle_header(struct strbuf **out, const struct strbuf *line)
 204 {
 205         if (!*out) {
 206                 *out = xmalloc(sizeof(struct strbuf));
 207                 strbuf_init(*out, line->len);
 208         } else
 209                 strbuf_reset(*out);
 210
 211         strbuf_addbuf(*out, line);
 212 }
 213
 214 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 215  * to have enough heuristics to grok MIME encoded patches often found
 216  * on our mailing lists.  For example, we do not even treat header lines
 217  * case insensitively.
 218  */
 219
 220 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 221 {
 222         const char *ends, *ap = strcasestr(line, name);
 223         size_t sz;
 224
 225         strbuf_setlen(attr, 0);
 226         if (!ap)
 227                 return 0;
 228         ap += strlen(name);
 229         if (*ap == '"') {
 230                 ap++;
 231                 ends = "\"";
 232         }
 233         else
 234                 ends = "; \t";
 235         sz = strcspn(ap, ends);
 236         strbuf_add(attr, ap, sz);
 237         return 1;
 238 }
 239
 240 static int has_attr_value(const char *line, const char *name, const char *value)
 241 {
 242         struct strbuf sb = STRBUF_INIT;
 243         int rc = slurp_attr(line, name, &sb) && !strcasecmp(sb.buf, value);
 244         strbuf_release(&sb);
 245         return rc;
 246 }
 247
 248 static void handle_content_type(struct mailinfo *mi, struct strbuf *line)
 249 {
 250         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 251         strbuf_init(boundary, line->len);
 252
 253         mi->format_flowed = has_attr_value(line->buf, "format=", "flowed");
 254         mi->delsp = has_attr_value(line->buf, "delsp=", "yes");
 255
 256         if (slurp_attr(line->buf, "boundary=", boundary)) {
 257                 strbuf_insertstr(boundary, 0, "--");
 258                 if (++mi->content_top >= &mi->content[MAX_BOUNDARIES]) {
 259                         error("Too many boundaries to handle");
 260                         mi->input_error = -1;
 261                         mi->content_top = &mi->content[MAX_BOUNDARIES] - 1;
 262                         return;
 263                 }
 264                 *(mi->content_top) = boundary;
 265                 boundary = NULL;
 266         }
 267         slurp_attr(line->buf, "charset=", &mi->charset);
 268
 269         if (boundary) {
 270                 strbuf_release(boundary);
 271                 free(boundary);
 272         }
 273 }
 274
 275 static void handle_content_transfer_encoding(struct mailinfo *mi,
 276                                              const struct strbuf *line)
 277 {
 278         if (strcasestr(line->buf, "base64"))
 279                 mi->transfer_encoding = TE_BASE64;
 280         else if (strcasestr(line->buf, "quoted-printable"))
 281                 mi->transfer_encoding = TE_QP;
 282         else
 283                 mi->transfer_encoding = TE_DONTCARE;
 284 }
 285
 286 static int is_multipart_boundary(struct mailinfo *mi, const struct strbuf *line)
 287 {
 288         struct strbuf *content_top = *(mi->content_top);
 289
 290         return ((content_top->len <= line->len) &&
 291                 !memcmp(line->buf, content_top->buf, content_top->len));
 292 }
 293
 294 static void cleanup_subject(struct mailinfo *mi, struct strbuf *subject)
 295 {
 296         size_t at = 0;
 297
 298         while (at < subject->len) {
 299                 char *pos;
 300                 size_t remove;
 301
 302                 switch (subject->buf[at]) {
 303                 case 'r': case 'R':
 304                         if (subject->len <= at + 3)
 305                                 break;
 306                         if ((subject->buf[at + 1] == 'e' ||
 307                              subject->buf[at + 1] == 'E') &&
 308                             subject->buf[at + 2] == ':') {
 309                                 strbuf_remove(subject, at, 3);
 310                                 continue;
 311                         }
 312                         at++;
 313                         break;
 314                 case ' ': case '\t': case ':':
 315                         strbuf_remove(subject, at, 1);
 316                         continue;
 317                 case '[':
 318                         pos = strchr(subject->buf + at, ']');
 319                         if (!pos)
 320                                 break;
 321                         remove = pos - subject->buf + at + 1;
 322                         if (!mi->keep_non_patch_brackets_in_subject ||
 323                             (7 <= remove &&
 324                              memmem(subject->buf + at, remove, "PATCH", 5)))
 325                                 strbuf_remove(subject, at, remove);
 326                         else {
 327                                 at += remove;
 328                                 /*
 329                                  * If the input had a space after the ], keep
 330                                  * it.  We don't bother with finding the end of
 331                                  * the space, since we later normalize it
 332                                  * anyway.
 333                                  */
 334                                 if (isspace(subject->buf[at]))
 335                                         at += 1;
 336                         }
 337                         continue;
 338                 }
 339                 break;
 340         }
 341         strbuf_trim(subject);
 342 }
 343
 344 #define MAX_HDR_PARSED 10
 345 static const char *header[MAX_HDR_PARSED] = {
 346         "From","Subject","Date",
 347 };
 348
 349 static inline int skip_header(const struct strbuf *line, const char *hdr,
 350                               const char **outval)
 351 {
 352         const char *val;
 353         if (!skip_iprefix(line->buf, hdr, &val) ||
 354             *val++ != ':')
 355                 return 0;
 356         while (isspace(*val))
 357                 val++;
 358         *outval = val;
 359         return 1;
 360 }
 361
 362 static int is_format_patch_separator(const char *line, int len)
 363 {
 364         static const char SAMPLE[] =
 365                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 366         const char *cp;
 367
 368         if (len != strlen(SAMPLE))
 369                 return 0;
 370         if (!skip_prefix(line, "From ", &cp))
 371                 return 0;
 372         if (strspn(cp, "0123456789abcdef") != 40)
 373                 return 0;
 374         cp += 40;
 375         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 376 }
 377
 378 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 379 {
 380         const char *in = q_seg->buf;
 381         int c;
 382         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 383         strbuf_init(out, q_seg->len);
 384
 385         while ((c = *in++) != 0) {
 386                 if (c == '=') {
 387                         int ch, d = *in;
 388                         if (d == '\n' || !d)
 389                                 break; /* drop trailing newline */
 390                         ch = hex2chr(in);
 391                         if (ch >= 0) {
 392                                 strbuf_addch(out, ch);
 393                                 in += 2;
 394                                 continue;
 395                         }
 396                         /* garbage -- fall through */
 397                 }
 398                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 399                         c = 0x20;
 400                 strbuf_addch(out, c);
 401         }
 402         return out;
 403 }
 404
 405 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 406 {
 407         /* Decode in..ep, possibly in-place to ot */
 408         int c, pos = 0, acc = 0;
 409         const char *in = b_seg->buf;
 410         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 411         strbuf_init(out, b_seg->len);
 412
 413         while ((c = *in++) != 0) {
 414                 if (c == '+')
 415                         c = 62;
 416                 else if (c == '/')
 417                         c = 63;
 418                 else if ('A' <= c && c <= 'Z')
 419                         c -= 'A';
 420                 else if ('a' <= c && c <= 'z')
 421                         c -= 'a' - 26;
 422                 else if ('0' <= c && c <= '9')
 423                         c -= '0' - 52;
 424                 else
 425                         continue; /* garbage */
 426                 switch (pos++) {
 427                 case 0:
 428                         acc = (c << 2);
 429                         break;
 430                 case 1:
 431                         strbuf_addch(out, (acc | (c >> 4)));
 432                         acc = (c & 15) << 4;
 433                         break;
 434                 case 2:
 435                         strbuf_addch(out, (acc | (c >> 2)));
 436                         acc = (c & 3) << 6;
 437                         break;
 438                 case 3:
 439                         strbuf_addch(out, (acc | c));
 440                         acc = pos = 0;
 441                         break;
 442                 }
 443         }
 444         return out;
 445 }
 446
 447 static int convert_to_utf8(struct mailinfo *mi,
 448                            struct strbuf *line, const char *charset)
 449 {
 450         char *out;
 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(line->buf, mi->metainfo_charset, charset);
 458         if (!out) {
 459                 mi->input_error = -1;
 460                 return error("cannot convert from %s to %s",
 461                              charset, mi->metainfo_charset);
 462         }
 463         strbuf_attach(line, out, strlen(out), strlen(out));
 464         return 0;
 465 }
 466
 467 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 468 {
 469         char *in, *ep, *cp;
 470         struct strbuf outbuf = STRBUF_INIT, *dec;
 471         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 472         int found_error = 1; /* pessimism */
 473
 474         in = it->buf;
 475         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 476                 int encoding;
 477                 strbuf_reset(&charset_q);
 478                 strbuf_reset(&piecebuf);
 479
 480                 if (in != ep) {
 481                         /*
 482                          * We are about to process an encoded-word
 483                          * that begins at ep, but there is something
 484                          * before the encoded word.
 485                          */
 486                         char *scan;
 487                         for (scan = in; scan < ep; scan++)
 488                                 if (!isspace(*scan))
 489                                         break;
 490
 491                         if (scan != ep || in == it->buf) {
 492                                 /*
 493                                  * We should not lose that "something",
 494                                  * unless we have just processed an
 495                                  * encoded-word, and there is only LWS
 496                                  * before the one we are about to process.
 497                                  */
 498                                 strbuf_add(&outbuf, in, ep - in);
 499                         }
 500                 }
 501                 /* E.g.
 502                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 503                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 504                  */
 505                 ep += 2;
 506
 507                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 508                         goto release_return;
 509
 510                 if (cp + 3 - it->buf > it->len)
 511                         goto release_return;
 512                 strbuf_add(&charset_q, ep, cp - ep);
 513
 514                 encoding = cp[1];
 515                 if (!encoding || cp[2] != '?')
 516                         goto release_return;
 517                 ep = strstr(cp + 3, "?=");
 518                 if (!ep)
 519                         goto release_return;
 520                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 521                 switch (tolower(encoding)) {
 522                 default:
 523                         goto release_return;
 524                 case 'b':
 525                         dec = decode_b_segment(&piecebuf);
 526                         break;
 527                 case 'q':
 528                         dec = decode_q_segment(&piecebuf, 1);
 529                         break;
 530                 }
 531                 if (convert_to_utf8(mi, dec, charset_q.buf))
 532                         goto release_return;
 533
 534                 strbuf_addbuf(&outbuf, dec);
 535                 strbuf_release(dec);
 536                 free(dec);
 537                 in = ep + 2;
 538         }
 539         strbuf_addstr(&outbuf, in);
 540         strbuf_reset(it);
 541         strbuf_addbuf(it, &outbuf);
 542         found_error = 0;
 543 release_return:
 544         strbuf_release(&outbuf);
 545         strbuf_release(&charset_q);
 546         strbuf_release(&piecebuf);
 547
 548         if (found_error)
 549                 mi->input_error = -1;
 550 }
 551
 552 /*
 553  * Returns true if "line" contains a header matching "hdr", in which case "val"
 554  * will contain the value of the header with any RFC2047 B and Q encoding
 555  * unwrapped, and optionally normalize the meta information to utf8.
 556  */
 557 static int parse_header(const struct strbuf *line,
 558                         const char *hdr,
 559                         struct mailinfo *mi,
 560                         struct strbuf *val)
 561 {
 562         const char *val_str;
 563
 564         if (!skip_header(line, hdr, &val_str))
 565                 return 0;
 566         strbuf_addstr(val, val_str);
 567         decode_header(mi, val);
 568         return 1;
 569 }
 570
 571 static int check_header(struct mailinfo *mi,
 572                         const struct strbuf *line,
 573                         struct strbuf *hdr_data[], int overwrite)
 574 {
 575         int i, ret = 0;
 576         struct strbuf sb = STRBUF_INIT;
 577
 578         /* search for the interesting parts */
 579         for (i = 0; header[i]; i++) {
 580                 if ((!hdr_data[i] || overwrite) &&
 581                     parse_header(line, header[i], mi, &sb)) {
 582                         handle_header(&hdr_data[i], &sb);
 583                         ret = 1;
 584                         goto check_header_out;
 585                 }
 586         }
 587
 588         /* Content stuff */
 589         if (parse_header(line, "Content-Type", mi, &sb)) {
 590                 handle_content_type(mi, &sb);
 591                 ret = 1;
 592                 goto check_header_out;
 593         }
 594         if (parse_header(line, "Content-Transfer-Encoding", mi, &sb)) {
 595                 handle_content_transfer_encoding(mi, &sb);
 596                 ret = 1;
 597                 goto check_header_out;
 598         }
 599         if (parse_header(line, "Message-Id", mi, &sb)) {
 600                 if (mi->add_message_id)
 601                         mi->message_id = strbuf_detach(&sb, NULL);
 602                 ret = 1;
 603                 goto check_header_out;
 604         }
 605
 606 check_header_out:
 607         strbuf_release(&sb);
 608         return ret;
 609 }
 610
 611 /*
 612  * Returns 1 if the given line or any line beginning with the given line is an
 613  * in-body header (that is, check_header will succeed when passed
 614  * mi->s_hdr_data).
 615  */
 616 static int is_inbody_header(const struct mailinfo *mi,
 617                             const struct strbuf *line)
 618 {
 619         int i;
 620         const char *val;
 621         for (i = 0; header[i]; i++)
 622                 if (!mi->s_hdr_data[i] && skip_header(line, header[i], &val))
 623                         return 1;
 624         return 0;
 625 }
 626
 627 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 628 {
 629         struct strbuf *ret;
 630
 631         switch (mi->transfer_encoding) {
 632         case TE_QP:
 633                 ret = decode_q_segment(line, 0);
 634                 break;
 635         case TE_BASE64:
 636                 ret = decode_b_segment(line);
 637                 break;
 638         case TE_DONTCARE:
 639         default:
 640                 return;
 641         }
 642         strbuf_reset(line);
 643         strbuf_addbuf(line, ret);
 644         strbuf_release(ret);
 645         free(ret);
 646 }
 647
 648 static inline int patchbreak(const struct strbuf *line)
 649 {
 650         size_t i;
 651
 652         /* Beginning of a "diff -" header? */
 653         if (starts_with(line->buf, "diff -"))
 654                 return 1;
 655
 656         /* CVS "Index: " line? */
 657         if (starts_with(line->buf, "Index: "))
 658                 return 1;
 659
 660         /*
 661          * "--- <filename>" starts patches without headers
 662          * "---<sp>*" is a manual separator
 663          */
 664         if (line->len < 4)
 665                 return 0;
 666
 667         if (starts_with(line->buf, "---")) {
 668                 /* space followed by a filename? */
 669                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 670                         return 1;
 671                 /* Just whitespace? */
 672                 for (i = 3; i < line->len; i++) {
 673                         unsigned char c = line->buf[i];
 674                         if (c == '\n')
 675                                 return 1;
 676                         if (!isspace(c))
 677                                 break;
 678                 }
 679                 return 0;
 680         }
 681         return 0;
 682 }
 683
 684 static int is_scissors_line(const char *line)
 685 {
 686         const char *c;
 687         int scissors = 0, gap = 0;
 688         const char *first_nonblank = NULL, *last_nonblank = NULL;
 689         int visible, perforation = 0, in_perforation = 0;
 690
 691         for (c = line; *c; c++) {
 692                 if (isspace(*c)) {
 693                         if (in_perforation) {
 694                                 perforation++;
 695                                 gap++;
 696                         }
 697                         continue;
 698                 }
 699                 last_nonblank = c;
 700                 if (first_nonblank == NULL)
 701                         first_nonblank = c;
 702                 if (*c == '-') {
 703                         in_perforation = 1;
 704                         perforation++;
 705                         continue;
 706                 }
 707                 if ((!memcmp(c, ">8", 2) || !memcmp(c, "8<", 2) ||
 708                      !memcmp(c, ">%", 2) || !memcmp(c, "%<", 2))) {
 709                         in_perforation = 1;
 710                         perforation += 2;
 711                         scissors += 2;
 712                         c++;
 713                         continue;
 714                 }
 715                 in_perforation = 0;
 716         }
 717
 718         /*
 719          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 720          * Even though there can be arbitrary cruft on the same line
 721          * (e.g. "cut here"), in order to avoid misidentification, the
 722          * perforation must occupy more than a third of the visible
 723          * width of the line, and dashes and scissors must occupy more
 724          * than half of the perforation.
 725          */
 726
 727         if (first_nonblank && last_nonblank)
 728                 visible = last_nonblank - first_nonblank + 1;
 729         else
 730                 visible = 0;
 731         return (scissors && 8 <= visible &&
 732                 visible < perforation * 3 &&
 733                 gap * 2 < perforation);
 734 }
 735
 736 static void flush_inbody_header_accum(struct mailinfo *mi)
 737 {
 738         if (!mi->inbody_header_accum.len)
 739                 return;
 740         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 741                 BUG("inbody_header_accum, if not empty, must always contain a valid in-body header");
 742         strbuf_reset(&mi->inbody_header_accum);
 743 }
 744
 745 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 746 {
 747         if (mi->inbody_header_accum.len &&
 748             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 749                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 750                         /*
 751                          * This is a scissors line; do not consider this line
 752                          * as a header continuation line.
 753                          */
 754                         flush_inbody_header_accum(mi);
 755                         return 0;
 756                 }
 757                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 758                 strbuf_addbuf(&mi->inbody_header_accum, line);
 759                 return 1;
 760         }
 761
 762         flush_inbody_header_accum(mi);
 763
 764         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 765                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 766         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 767                 int i;
 768                 for (i = 0; header[i]; i++)
 769                         if (!strcmp("Subject", header[i])) {
 770                                 handle_header(&mi->s_hdr_data[i], line);
 771                                 return 1;
 772                         }
 773                 return 0;
 774         }
 775         if (is_inbody_header(mi, line)) {
 776                 strbuf_addbuf(&mi->inbody_header_accum, line);
 777                 return 1;
 778         }
 779         return 0;
 780 }
 781
 782 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 783 {
 784         assert(!mi->filter_stage);
 785
 786         if (mi->header_stage) {
 787                 if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 788                         if (mi->inbody_header_accum.len) {
 789                                 flush_inbody_header_accum(mi);
 790                                 mi->header_stage = 0;
 791                         }
 792                         return 0;
 793                 }
 794         }
 795
 796         if (mi->use_inbody_headers && mi->header_stage) {
 797                 mi->header_stage = check_inbody_header(mi, line);
 798                 if (mi->header_stage)
 799                         return 0;
 800         } else
 801                 /* Only trim the first (blank) line of the commit message
 802                  * when ignoring in-body headers.
 803                  */
 804                 mi->header_stage = 0;
 805
 806         /* normalize the log message to UTF-8. */
 807         if (convert_to_utf8(mi, line, mi->charset.buf))
 808                 return 0; /* mi->input_error already set */
 809
 810         if (mi->use_scissors && is_scissors_line(line->buf)) {
 811                 int i;
 812
 813                 strbuf_setlen(&mi->log_message, 0);
 814                 mi->header_stage = 1;
 815
 816                 /*
 817                  * We may have already read "secondary headers"; purge
 818                  * them to give ourselves a clean restart.
 819                  */
 820                 for (i = 0; header[i]; i++) {
 821                         if (mi->s_hdr_data[i])
 822                                 strbuf_release(mi->s_hdr_data[i]);
 823                         mi->s_hdr_data[i] = NULL;
 824                 }
 825                 return 0;
 826         }
 827
 828         if (patchbreak(line)) {
 829                 if (mi->message_id)
 830                         strbuf_addf(&mi->log_message,
 831                                     "Message-Id: %s\n", mi->message_id);
 832                 return 1;
 833         }
 834
 835         strbuf_addbuf(&mi->log_message, line);
 836         return 0;
 837 }
 838
 839 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 840 {
 841         fwrite(line->buf, 1, line->len, mi->patchfile);
 842         mi->patch_lines++;
 843 }
 844
 845 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 846 {
 847         switch (mi->filter_stage) {
 848         case 0:
 849                 if (!handle_commit_msg(mi, line))
 850                         break;
 851                 mi->filter_stage++;
 852                 /* fallthrough */
 853         case 1:
 854                 handle_patch(mi, line);
 855                 break;
 856         }
 857 }
 858
 859 static int is_rfc2822_header(const struct strbuf *line)
 860 {
 861         /*
 862          * The section that defines the loosest possible
 863          * field name is "3.6.8 Optional fields".
 864          *
 865          * optional-field = field-name ":" unstructured CRLF
 866          * field-name = 1*ftext
 867          * ftext = %d33-57 / %59-126
 868          */
 869         int ch;
 870         char *cp = line->buf;
 871
 872         /* Count mbox From headers as headers */
 873         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 874                 return 1;
 875
 876         while ((ch = *cp++)) {
 877                 if (ch == ':')
 878                         return 1;
 879                 if ((33 <= ch && ch <= 57) ||
 880                     (59 <= ch && ch <= 126))
 881                         continue;
 882                 break;
 883         }
 884         return 0;
 885 }
 886
 887 static int read_one_header_line(struct strbuf *line, FILE *in)
 888 {
 889         struct strbuf continuation = STRBUF_INIT;
 890
 891         /* Get the first part of the line. */
 892         if (strbuf_getline_lf(line, in))
 893                 return 0;
 894
 895         /*
 896          * Is it an empty line or not a valid rfc2822 header?
 897          * If so, stop here, and return false ("not a header")
 898          */
 899         strbuf_rtrim(line);
 900         if (!line->len || !is_rfc2822_header(line)) {
 901                 /* Re-add the newline */
 902                 strbuf_addch(line, '\n');
 903                 return 0;
 904         }
 905
 906         /*
 907          * Now we need to eat all the continuation lines..
 908          * Yuck, 2822 header "folding"
 909          */
 910         for (;;) {
 911                 int peek;
 912
 913                 peek = fgetc(in);
 914                 if (peek == EOF)
 915                         break;
 916                 ungetc(peek, in);
 917                 if (peek != ' ' && peek != '\t')
 918                         break;
 919                 if (strbuf_getline_lf(&continuation, in))
 920                         break;
 921                 continuation.buf[0] = ' ';
 922                 strbuf_rtrim(&continuation);
 923                 strbuf_addbuf(line, &continuation);
 924         }
 925         strbuf_release(&continuation);
 926
 927         return 1;
 928 }
 929
 930 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 931 {
 932         while (!strbuf_getline_lf(line, mi->input)) {
 933                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 934                         return 1;
 935         }
 936         return 0;
 937 }
 938
 939 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 940 {
 941         struct strbuf newline = STRBUF_INIT;
 942
 943         strbuf_addch(&newline, '\n');
 944 again:
 945         if (line->len >= (*(mi->content_top))->len + 2 &&
 946             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 947                 /* we hit an end boundary */
 948                 /* pop the current boundary off the stack */
 949                 strbuf_release(*(mi->content_top));
 950                 FREE_AND_NULL(*(mi->content_top));
 951
 952                 /* technically won't happen as is_multipart_boundary()
 953                    will fail first.  But just in case..
 954                  */
 955                 if (--mi->content_top < mi->content) {
 956                         error("Detected mismatched boundaries, can't recover");
 957                         mi->input_error = -1;
 958                         mi->content_top = mi->content;
 959                         strbuf_release(&newline);
 960                         return 0;
 961                 }
 962                 handle_filter(mi, &newline);
 963                 strbuf_release(&newline);
 964                 if (mi->input_error)
 965                         return 0;
 966
 967                 /* skip to the next boundary */
 968                 if (!find_boundary(mi, line))
 969                         return 0;
 970                 goto again;
 971         }
 972
 973         /* set some defaults */
 974         mi->transfer_encoding = TE_DONTCARE;
 975         strbuf_reset(&mi->charset);
 976
 977         /* slurp in this section's info */
 978         while (read_one_header_line(line, mi->input))
 979                 check_header(mi, line, mi->p_hdr_data, 0);
 980
 981         strbuf_release(&newline);
 982         /* replenish line */
 983         if (strbuf_getline_lf(line, mi->input))
 984                 return 0;
 985         strbuf_addch(line, '\n');
 986         return 1;
 987 }
 988
 989 static void handle_filter_flowed(struct mailinfo *mi, struct strbuf *line,
 990                                  struct strbuf *prev)
 991 {
 992         size_t len = line->len;
 993         const char *rest;
 994
 995         if (!mi->format_flowed) {
 996                 handle_filter(mi, line);
 997                 return;
 998         }
 999
1000         if (line->buf[len - 1] == '\n') {
1001                 len--;
1002                 if (len && line->buf[len - 1] == '\r')
1003                         len--;
1004         }
1005
1006         /* Keep signature separator as-is. */
1007         if (skip_prefix(line->buf, "-- ", &rest) && rest - line->buf == len) {
1008                 if (prev->len) {
1009                         handle_filter(mi, prev);
1010                         strbuf_reset(prev);
1011                 }
1012                 handle_filter(mi, line);
1013                 return;
1014         }
1015
1016         /* Unstuff space-stuffed line. */
1017         if (len && line->buf[0] == ' ') {
1018                 strbuf_remove(line, 0, 1);
1019                 len--;
1020         }
1021
1022         /* Save flowed line for later, but without the soft line break. */
1023         if (len && line->buf[len - 1] == ' ') {
1024                 strbuf_add(prev, line->buf, len - !!mi->delsp);
1025                 return;
1026         }
1027
1028         /* Prepend any previous partial lines */
1029         strbuf_insert(line, 0, prev->buf, prev->len);
1030         strbuf_reset(prev);
1031
1032         handle_filter(mi, line);
1033 }
1034
1035 static void handle_body(struct mailinfo *mi, struct strbuf *line)
1036 {
1037         struct strbuf prev = STRBUF_INIT;
1038
1039         /* Skip up to the first boundary */
1040         if (*(mi->content_top)) {
1041                 if (!find_boundary(mi, line))
1042                         goto handle_body_out;
1043         }
1044
1045         do {
1046                 /* process any boundary lines */
1047                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
1048                         /* flush any leftover */
1049                         if (prev.len) {
1050                                 handle_filter(mi, &prev);
1051                                 strbuf_reset(&prev);
1052                         }
1053                         if (!handle_boundary(mi, line))
1054                                 goto handle_body_out;
1055                 }
1056
1057                 /* Unwrap transfer encoding */
1058                 decode_transfer_encoding(mi, line);
1059
1060                 switch (mi->transfer_encoding) {
1061                 case TE_BASE64:
1062                 case TE_QP:
1063                 {
1064                         struct strbuf **lines, **it, *sb;
1065
1066                         /* Prepend any previous partial lines */
1067                         strbuf_insert(line, 0, prev.buf, prev.len);
1068                         strbuf_reset(&prev);
1069
1070                         /*
1071                          * This is a decoded line that may contain
1072                          * multiple new lines.  Pass only one chunk
1073                          * at a time to handle_filter()
1074                          */
1075                         lines = strbuf_split(line, '\n');
1076                         for (it = lines; (sb = *it); it++) {
1077                                 if (*(it + 1) == NULL) /* The last line */
1078                                         if (sb->buf[sb->len - 1] != '\n') {
1079                                                 /* Partial line, save it for later. */
1080                                                 strbuf_addbuf(&prev, sb);
1081                                                 break;
1082                                         }
1083                                 handle_filter_flowed(mi, sb, &prev);
1084                         }
1085                         /*
1086                          * The partial chunk is saved in "prev" and will be
1087                          * appended by the next iteration of read_line_with_nul().
1088                          */
1089                         strbuf_list_free(lines);
1090                         break;
1091                 }
1092                 default:
1093                         handle_filter_flowed(mi, line, &prev);
1094                 }
1095
1096                 if (mi->input_error)
1097                         break;
1098         } while (!strbuf_getwholeline(line, mi->input, '\n'));
1099
1100         if (prev.len)
1101                 handle_filter(mi, &prev);
1102
1103         flush_inbody_header_accum(mi);
1104
1105 handle_body_out:
1106         strbuf_release(&prev);
1107 }
1108
1109 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
1110 {
1111         const char *sp = data->buf;
1112         while (1) {
1113                 char *ep = strchr(sp, '\n');
1114                 int len;
1115                 if (!ep)
1116                         len = strlen(sp);
1117                 else
1118                         len = ep - sp;
1119                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
1120                 if (!ep)
1121                         break;
1122                 sp = ep + 1;
1123         }
1124 }
1125
1126 static void handle_info(struct mailinfo *mi)
1127 {
1128         struct strbuf *hdr;
1129         int i;
1130
1131         for (i = 0; header[i]; i++) {
1132                 /* only print inbody headers if we output a patch file */
1133                 if (mi->patch_lines && mi->s_hdr_data[i])
1134                         hdr = mi->s_hdr_data[i];
1135                 else if (mi->p_hdr_data[i])
1136                         hdr = mi->p_hdr_data[i];
1137                 else
1138                         continue;
1139
1140                 if (!strcmp(header[i], "Subject")) {
1141                         if (!mi->keep_subject) {
1142                                 cleanup_subject(mi, hdr);
1143                                 cleanup_space(hdr);
1144                         }
1145                         output_header_lines(mi->output, "Subject", hdr);
1146                 } else if (!strcmp(header[i], "From")) {
1147                         cleanup_space(hdr);
1148                         handle_from(mi, hdr);
1149                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
1150                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
1151                 } else {
1152                         cleanup_space(hdr);
1153                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
1154                 }
1155         }
1156         fprintf(mi->output, "\n");
1157 }
1158
1159 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
1160 {
1161         FILE *cmitmsg;
1162         int peek;
1163         struct strbuf line = STRBUF_INIT;
1164
1165         cmitmsg = fopen(msg, "w");
1166         if (!cmitmsg) {
1167                 perror(msg);
1168                 return -1;
1169         }
1170         mi->patchfile = fopen(patch, "w");
1171         if (!mi->patchfile) {
1172                 perror(patch);
1173                 fclose(cmitmsg);
1174                 return -1;
1175         }
1176
1177         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
1178         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
1179
1180         do {
1181                 peek = fgetc(mi->input);
1182                 if (peek == EOF) {
1183                         fclose(cmitmsg);
1184                         return error("empty patch: '%s'", patch);
1185                 }
1186         } while (isspace(peek));
1187         ungetc(peek, mi->input);
1188
1189         /* process the email header */
1190         while (read_one_header_line(&line, mi->input))
1191                 check_header(mi, &line, mi->p_hdr_data, 1);
1192
1193         handle_body(mi, &line);
1194         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1195         fclose(cmitmsg);
1196         fclose(mi->patchfile);
1197
1198         handle_info(mi);
1199         strbuf_release(&line);
1200         return mi->input_error;
1201 }
1202
1203 static int git_mailinfo_config(const char *var, const char *value, void *mi_)
1204 {
1205         struct mailinfo *mi = mi_;
1206
1207         if (!starts_with(var, "mailinfo."))
1208                 return git_default_config(var, value, NULL);
1209         if (!strcmp(var, "mailinfo.scissors")) {
1210                 mi->use_scissors = git_config_bool(var, value);
1211                 return 0;
1212         }
1213         /* perhaps others here */
1214         return 0;
1215 }
1216
1217 void setup_mailinfo(struct mailinfo *mi)
1218 {
1219         memset(mi, 0, sizeof(*mi));
1220         strbuf_init(&mi->name, 0);
1221         strbuf_init(&mi->email, 0);
1222         strbuf_init(&mi->charset, 0);
1223         strbuf_init(&mi->log_message, 0);
1224         strbuf_init(&mi->inbody_header_accum, 0);
1225         mi->header_stage = 1;
1226         mi->use_inbody_headers = 1;
1227         mi->content_top = mi->content;
1228         git_config(git_mailinfo_config, mi);
1229 }
1230
1231 void clear_mailinfo(struct mailinfo *mi)
1232 {
1233         int i;
1234
1235         strbuf_release(&mi->name);
1236         strbuf_release(&mi->email);
1237         strbuf_release(&mi->charset);
1238         strbuf_release(&mi->inbody_header_accum);
1239         free(mi->message_id);
1240
1241         if (mi->p_hdr_data)
1242                 for (i = 0; mi->p_hdr_data[i]; i++)
1243                         strbuf_release(mi->p_hdr_data[i]);
1244         free(mi->p_hdr_data);
1245         if (mi->s_hdr_data)
1246                 for (i = 0; mi->s_hdr_data[i]; i++)
1247                         strbuf_release(mi->s_hdr_data[i]);
1248         free(mi->s_hdr_data);
1249
1250         while (mi->content < mi->content_top) {
1251                 free(*(mi->content_top));
1252                 mi->content_top--;
1253         }
1254
1255         strbuf_release(&mi->log_message);
1256 }