mailinfo.c

   1 #define USE_THE_REPOSITORY_VARIABLE
   2
   3 #include "git-compat-util.h"
   4 #include "config.h"
   5 #include "gettext.h"
   6 #include "hex-ll.h"
   7 #include "utf8.h"
   8 #include "strbuf.h"
   9 #include "mailinfo.h"
  10
  11 static void cleanup_space(struct strbuf *sb)
  12 {
  13         size_t pos, cnt;
  14         for (pos = 0; pos < sb->len; pos++) {
  15                 if (isspace(sb->buf[pos])) {
  16                         sb->buf[pos] = ' ';
  17                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
  18                         strbuf_remove(sb, pos + 1, cnt);
  19                 }
  20         }
  21 }
  22
  23 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  24 {
  25         struct strbuf *src = name;
  26         if (!name->len || 60 < name->len || strpbrk(name->buf, "@<>"))
  27                 src = email;
  28         else if (name == out)
  29                 return;
  30         strbuf_reset(out);
  31         strbuf_addbuf(out, src);
  32 }
  33
  34 static void parse_bogus_from(struct mailinfo *mi, const struct strbuf *line)
  35 {
  36         /* John Doe <johndoe> */
  37
  38         char *bra, *ket;
  39         /* This is fallback, so do not bother if we already have an
  40          * e-mail address.
  41          */
  42         if (mi->email.len)
  43                 return;
  44
  45         bra = strchr(line->buf, '<');
  46         if (!bra)
  47                 return;
  48         ket = strchr(bra, '>');
  49         if (!ket)
  50                 return;
  51
  52         strbuf_reset(&mi->email);
  53         strbuf_add(&mi->email, bra + 1, ket - bra - 1);
  54
  55         strbuf_reset(&mi->name);
  56         strbuf_add(&mi->name, line->buf, bra - line->buf);
  57         strbuf_trim(&mi->name);
  58         get_sane_name(&mi->name, &mi->name, &mi->email);
  59 }
  60
  61 static const char *unquote_comment(struct strbuf *outbuf, const char *in)
  62 {
  63         int take_next_literally = 0;
  64         int depth = 1;
  65
  66         strbuf_addch(outbuf, '(');
  67
  68         while (*in) {
  69                 int c = *in++;
  70                 if (take_next_literally == 1) {
  71                         take_next_literally = 0;
  72                 } else {
  73                         switch (c) {
  74                         case '\\':
  75                                 take_next_literally = 1;
  76                                 continue;
  77                         case '(':
  78                                 strbuf_addch(outbuf, '(');
  79                                 depth++;
  80                                 continue;
  81                         case ')':
  82                                 strbuf_addch(outbuf, ')');
  83                                 if (!--depth)
  84                                         return in;
  85                                 continue;
  86                         }
  87                 }
  88
  89                 strbuf_addch(outbuf, c);
  90         }
  91
  92         return in;
  93 }
  94
  95 static const char *unquote_quoted_string(struct strbuf *outbuf, const char *in)
  96 {
  97         int take_next_literally = 0;
  98
  99         while (*in) {
 100                 int c = *in++;
 101                 if (take_next_literally == 1) {
 102                         take_next_literally = 0;
 103                 } else {
 104                         switch (c) {
 105                         case '\\':
 106                                 take_next_literally = 1;
 107                                 continue;
 108                         case '"':
 109                                 return in;
 110                         }
 111                 }
 112
 113                 strbuf_addch(outbuf, c);
 114         }
 115
 116         return in;
 117 }
 118
 119 static void unquote_quoted_pair(struct strbuf *line)
 120 {
 121         struct strbuf outbuf;
 122         const char *in = line->buf;
 123         int c;
 124
 125         strbuf_init(&outbuf, line->len);
 126
 127         while ((c = *in++) != 0) {
 128                 switch (c) {
 129                 case '"':
 130                         in = unquote_quoted_string(&outbuf, in);
 131                         continue;
 132                 case '(':
 133                         in = unquote_comment(&outbuf, in);
 134                         continue;
 135                 }
 136
 137                 strbuf_addch(&outbuf, c);
 138         }
 139
 140         strbuf_swap(&outbuf, line);
 141         strbuf_release(&outbuf);
 142
 143 }
 144
 145 static void handle_from(struct mailinfo *mi, const struct strbuf *from)
 146 {
 147         char *at;
 148         size_t el;
 149         struct strbuf f;
 150
 151         strbuf_init(&f, from->len);
 152         strbuf_addbuf(&f, from);
 153
 154         unquote_quoted_pair(&f);
 155
 156         at = strchr(f.buf, '@');
 157         if (!at) {
 158                 parse_bogus_from(mi, from);
 159                 goto out;
 160         }
 161
 162         /*
 163          * If we already have one email, don't take any confusing lines
 164          */
 165         if (mi->email.len && strchr(at + 1, '@'))
 166                 goto out;
 167
 168         /* Pick up the string around '@', possibly delimited with <>
 169          * pair; that is the email part.
 170          */
 171         while (at > f.buf) {
 172                 char c = at[-1];
 173                 if (isspace(c))
 174                         break;
 175                 if (c == '<') {
 176                         at[-1] = ' ';
 177                         break;
 178                 }
 179                 at--;
 180         }
 181         el = strcspn(at, " \n\t\r\v\f>");
 182         strbuf_reset(&mi->email);
 183         strbuf_add(&mi->email, at, el);
 184         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 185
 186         /* The remainder is name.  It could be
 187          *
 188          * - "John Doe <john.doe@xz>"                   (a), or
 189          * - "john.doe@xz (John Doe)"                   (b), or
 190          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 191          *
 192          * but we have removed the email part, so
 193          *
 194          * - remove extra spaces which could stay after email (case 'c'), and
 195          * - trim from both ends, possibly removing the () pair at the end
 196          *   (cases 'a' and 'b').
 197          */
 198         cleanup_space(&f);
 199         strbuf_trim(&f);
 200         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 201                 strbuf_remove(&f, 0, 1);
 202                 strbuf_setlen(&f, f.len - 1);
 203         }
 204
 205         get_sane_name(&mi->name, &f, &mi->email);
 206 out:
 207         strbuf_release(&f);
 208 }
 209
 210 static void handle_header(struct strbuf **out, const struct strbuf *line)
 211 {
 212         if (!*out) {
 213                 *out = xmalloc(sizeof(struct strbuf));
 214                 strbuf_init(*out, line->len);
 215         } else
 216                 strbuf_reset(*out);
 217
 218         strbuf_addbuf(*out, line);
 219 }
 220
 221 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 222  * to have enough heuristics to grok MIME encoded patches often found
 223  * on our mailing lists.  For example, we do not even treat header lines
 224  * case insensitively.
 225  */
 226
 227 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 228 {
 229         const char *ends, *ap = strcasestr(line, name);
 230         size_t sz;
 231
 232         strbuf_setlen(attr, 0);
 233         if (!ap)
 234                 return 0;
 235         ap += strlen(name);
 236         if (*ap == '"') {
 237                 ap++;
 238                 ends = "\"";
 239         }
 240         else
 241                 ends = "; \t";
 242         sz = strcspn(ap, ends);
 243         strbuf_add(attr, ap, sz);
 244         return 1;
 245 }
 246
 247 static int has_attr_value(const char *line, const char *name, const char *value)
 248 {
 249         struct strbuf sb = STRBUF_INIT;
 250         int rc = slurp_attr(line, name, &sb) && !strcasecmp(sb.buf, value);
 251         strbuf_release(&sb);
 252         return rc;
 253 }
 254
 255 static void handle_content_type(struct mailinfo *mi, struct strbuf *line)
 256 {
 257         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 258         strbuf_init(boundary, line->len);
 259
 260         mi->format_flowed = has_attr_value(line->buf, "format=", "flowed");
 261         mi->delsp = has_attr_value(line->buf, "delsp=", "yes");
 262
 263         if (slurp_attr(line->buf, "boundary=", boundary)) {
 264                 strbuf_insertstr(boundary, 0, "--");
 265                 if (++mi->content_top >= &mi->content[MAX_BOUNDARIES]) {
 266                         error("Too many boundaries to handle");
 267                         mi->input_error = -1;
 268                         mi->content_top = &mi->content[MAX_BOUNDARIES] - 1;
 269                         return;
 270                 }
 271                 *(mi->content_top) = boundary;
 272                 boundary = NULL;
 273         }
 274         slurp_attr(line->buf, "charset=", &mi->charset);
 275
 276         if (boundary) {
 277                 strbuf_release(boundary);
 278                 free(boundary);
 279         }
 280 }
 281
 282 static void handle_content_transfer_encoding(struct mailinfo *mi,
 283                                              const struct strbuf *line)
 284 {
 285         if (strcasestr(line->buf, "base64"))
 286                 mi->transfer_encoding = TE_BASE64;
 287         else if (strcasestr(line->buf, "quoted-printable"))
 288                 mi->transfer_encoding = TE_QP;
 289         else
 290                 mi->transfer_encoding = TE_DONTCARE;
 291 }
 292
 293 static int is_multipart_boundary(struct mailinfo *mi, const struct strbuf *line)
 294 {
 295         struct strbuf *content_top = *(mi->content_top);
 296
 297         return ((content_top->len <= line->len) &&
 298                 !memcmp(line->buf, content_top->buf, content_top->len));
 299 }
 300
 301 static void cleanup_subject(struct mailinfo *mi, struct strbuf *subject)
 302 {
 303         size_t at = 0;
 304
 305         while (at < subject->len) {
 306                 char *pos;
 307                 size_t remove;
 308
 309                 switch (subject->buf[at]) {
 310                 case 'r': case 'R':
 311                         if (subject->len <= at + 3)
 312                                 break;
 313                         if ((subject->buf[at + 1] == 'e' ||
 314                              subject->buf[at + 1] == 'E') &&
 315                             subject->buf[at + 2] == ':') {
 316                                 strbuf_remove(subject, at, 3);
 317                                 continue;
 318                         }
 319                         at++;
 320                         break;
 321                 case ' ': case '\t': case ':':
 322                         strbuf_remove(subject, at, 1);
 323                         continue;
 324                 case '[':
 325                         pos = strchr(subject->buf + at, ']');
 326                         if (!pos)
 327                                 break;
 328                         remove = pos - (subject->buf + at) + 1;
 329                         if (!mi->keep_non_patch_brackets_in_subject ||
 330                             (7 <= remove &&
 331                              memmem(subject->buf + at, remove, "PATCH", 5)))
 332                                 strbuf_remove(subject, at, remove);
 333                         else {
 334                                 at += remove;
 335                                 /*
 336                                  * If the input had a space after the ], keep
 337                                  * it.  We don't bother with finding the end of
 338                                  * the space, since we later normalize it
 339                                  * anyway.
 340                                  */
 341                                 if (isspace(subject->buf[at]))
 342                                         at += 1;
 343                         }
 344                         continue;
 345                 }
 346                 break;
 347         }
 348         strbuf_trim(subject);
 349 }
 350
 351 static const char * const header[] = {
 352         "From", "Subject", "Date",
 353 };
 354
 355 static inline int skip_header(const struct strbuf *line, const char *hdr,
 356                               const char **outval)
 357 {
 358         const char *val;
 359         if (!skip_iprefix(line->buf, hdr, &val) ||
 360             *val++ != ':')
 361                 return 0;
 362         while (isspace(*val))
 363                 val++;
 364         *outval = val;
 365         return 1;
 366 }
 367
 368 static int is_format_patch_separator(const char *line, int len)
 369 {
 370         static const char SAMPLE[] =
 371                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 372         const char *cp;
 373
 374         if (len != strlen(SAMPLE))
 375                 return 0;
 376         if (!skip_prefix(line, "From ", &cp))
 377                 return 0;
 378         if (strspn(cp, "0123456789abcdef") != 40)
 379                 return 0;
 380         cp += 40;
 381         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 382 }
 383
 384 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 385 {
 386         const char *in = q_seg->buf;
 387         int c;
 388         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 389         strbuf_init(out, q_seg->len);
 390
 391         while ((c = *in++) != 0) {
 392                 if (c == '=') {
 393                         int ch, d = *in;
 394                         if (d == '\n' || !d)
 395                                 break; /* drop trailing newline */
 396                         ch = hex2chr(in);
 397                         if (ch >= 0) {
 398                                 strbuf_addch(out, ch);
 399                                 in += 2;
 400                                 continue;
 401                         }
 402                         /* garbage -- fall through */
 403                 }
 404                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 405                         c = 0x20;
 406                 strbuf_addch(out, c);
 407         }
 408         return out;
 409 }
 410
 411 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 412 {
 413         /* Decode in..ep, possibly in-place to ot */
 414         int c, pos = 0, acc = 0;
 415         const char *in = b_seg->buf;
 416         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 417         strbuf_init(out, b_seg->len);
 418
 419         while ((c = *in++) != 0) {
 420                 if (c == '+')
 421                         c = 62;
 422                 else if (c == '/')
 423                         c = 63;
 424                 else if ('A' <= c && c <= 'Z')
 425                         c -= 'A';
 426                 else if ('a' <= c && c <= 'z')
 427                         c -= 'a' - 26;
 428                 else if ('0' <= c && c <= '9')
 429                         c -= '0' - 52;
 430                 else
 431                         continue; /* garbage */
 432                 switch (pos++) {
 433                 case 0:
 434                         acc = (c << 2);
 435                         break;
 436                 case 1:
 437                         strbuf_addch(out, (acc | (c >> 4)));
 438                         acc = (c & 15) << 4;
 439                         break;
 440                 case 2:
 441                         strbuf_addch(out, (acc | (c >> 2)));
 442                         acc = (c & 3) << 6;
 443                         break;
 444                 case 3:
 445                         strbuf_addch(out, (acc | c));
 446                         acc = pos = 0;
 447                         break;
 448                 }
 449         }
 450         return out;
 451 }
 452
 453 static int convert_to_utf8(struct mailinfo *mi,
 454                            struct strbuf *line, const char *charset)
 455 {
 456         char *out;
 457         size_t out_len;
 458
 459         if (!mi->metainfo_charset || !charset || !*charset)
 460                 return 0;
 461
 462         if (same_encoding(mi->metainfo_charset, charset))
 463                 return 0;
 464         out = reencode_string_len(line->buf, line->len,
 465                                   mi->metainfo_charset, charset, &out_len);
 466         if (!out) {
 467                 mi->input_error = -1;
 468                 return error("cannot convert from %s to %s",
 469                              charset, mi->metainfo_charset);
 470         }
 471         strbuf_attach(line, out, out_len, out_len);
 472         return 0;
 473 }
 474
 475 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 476 {
 477         char *in, *ep, *cp;
 478         struct strbuf outbuf = STRBUF_INIT, *dec;
 479         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 480         int found_error = 1; /* pessimism */
 481
 482         in = it->buf;
 483         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 484                 int encoding;
 485                 strbuf_reset(&charset_q);
 486                 strbuf_reset(&piecebuf);
 487
 488                 if (in != ep) {
 489                         /*
 490                          * We are about to process an encoded-word
 491                          * that begins at ep, but there is something
 492                          * before the encoded word.
 493                          */
 494                         char *scan;
 495                         for (scan = in; scan < ep; scan++)
 496                                 if (!isspace(*scan))
 497                                         break;
 498
 499                         if (scan != ep || in == it->buf) {
 500                                 /*
 501                                  * We should not lose that "something",
 502                                  * unless we have just processed an
 503                                  * encoded-word, and there is only LWS
 504                                  * before the one we are about to process.
 505                                  */
 506                                 strbuf_add(&outbuf, in, ep - in);
 507                         }
 508                 }
 509                 /* E.g.
 510                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 511                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 512                  */
 513                 ep += 2;
 514
 515                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 516                         goto release_return;
 517
 518                 if (cp + 3 - it->buf > it->len)
 519                         goto release_return;
 520                 strbuf_add(&charset_q, ep, cp - ep);
 521
 522                 encoding = cp[1];
 523                 if (!encoding || cp[2] != '?')
 524                         goto release_return;
 525                 ep = strstr(cp + 3, "?=");
 526                 if (!ep)
 527                         goto release_return;
 528                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 529                 switch (tolower(encoding)) {
 530                 default:
 531                         goto release_return;
 532                 case 'b':
 533                         dec = decode_b_segment(&piecebuf);
 534                         break;
 535                 case 'q':
 536                         dec = decode_q_segment(&piecebuf, 1);
 537                         break;
 538                 }
 539                 if (convert_to_utf8(mi, dec, charset_q.buf))
 540                         goto release_return;
 541
 542                 strbuf_addbuf(&outbuf, dec);
 543                 strbuf_release(dec);
 544                 free(dec);
 545                 in = ep + 2;
 546         }
 547         strbuf_addstr(&outbuf, in);
 548         strbuf_reset(it);
 549         strbuf_addbuf(it, &outbuf);
 550         found_error = 0;
 551 release_return:
 552         strbuf_release(&outbuf);
 553         strbuf_release(&charset_q);
 554         strbuf_release(&piecebuf);
 555
 556         if (found_error)
 557                 mi->input_error = -1;
 558 }
 559
 560 /*
 561  * Returns true if "line" contains a header matching "hdr", in which case "val"
 562  * will contain the value of the header with any RFC2047 B and Q encoding
 563  * unwrapped, and optionally normalize the meta information to utf8.
 564  */
 565 static int parse_header(const struct strbuf *line,
 566                         const char *hdr,
 567                         struct mailinfo *mi,
 568                         struct strbuf *val)
 569 {
 570         const char *val_str;
 571
 572         if (!skip_header(line, hdr, &val_str))
 573                 return 0;
 574         strbuf_addstr(val, val_str);
 575         decode_header(mi, val);
 576         return 1;
 577 }
 578
 579 static int check_header(struct mailinfo *mi,
 580                         const struct strbuf *line,
 581                         struct strbuf *hdr_data[], int overwrite)
 582 {
 583         int i, ret = 0;
 584         struct strbuf sb = STRBUF_INIT;
 585
 586         /* search for the interesting parts */
 587         for (i = 0; i < ARRAY_SIZE(header); i++) {
 588                 if ((!hdr_data[i] || overwrite) &&
 589                     parse_header(line, header[i], mi, &sb)) {
 590                         handle_header(&hdr_data[i], &sb);
 591                         ret = 1;
 592                         goto check_header_out;
 593                 }
 594         }
 595
 596         /* Content stuff */
 597         if (parse_header(line, "Content-Type", mi, &sb)) {
 598                 handle_content_type(mi, &sb);
 599                 ret = 1;
 600                 goto check_header_out;
 601         }
 602         if (parse_header(line, "Content-Transfer-Encoding", mi, &sb)) {
 603                 handle_content_transfer_encoding(mi, &sb);
 604                 ret = 1;
 605                 goto check_header_out;
 606         }
 607         if (parse_header(line, "Message-ID", mi, &sb)) {
 608                 if (mi->add_message_id)
 609                         mi->message_id = strbuf_detach(&sb, NULL);
 610                 ret = 1;
 611                 goto check_header_out;
 612         }
 613
 614 check_header_out:
 615         strbuf_release(&sb);
 616         return ret;
 617 }
 618
 619 /*
 620  * Returns 1 if the given line or any line beginning with the given line is an
 621  * in-body header (that is, check_header will succeed when passed
 622  * mi->s_hdr_data).
 623  */
 624 static int is_inbody_header(const struct mailinfo *mi,
 625                             const struct strbuf *line)
 626 {
 627         int i;
 628         const char *val;
 629         for (i = 0; i < ARRAY_SIZE(header); i++)
 630                 if (!mi->s_hdr_data[i] && skip_header(line, header[i], &val))
 631                         return 1;
 632         return 0;
 633 }
 634
 635 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 636 {
 637         struct strbuf *ret;
 638
 639         switch (mi->transfer_encoding) {
 640         case TE_QP:
 641                 ret = decode_q_segment(line, 0);
 642                 break;
 643         case TE_BASE64:
 644                 ret = decode_b_segment(line);
 645                 break;
 646         case TE_DONTCARE:
 647         default:
 648                 return;
 649         }
 650         strbuf_reset(line);
 651         strbuf_addbuf(line, ret);
 652         strbuf_release(ret);
 653         free(ret);
 654 }
 655
 656 static inline int patchbreak(const struct strbuf *line)
 657 {
 658         size_t i;
 659
 660         /* Beginning of a "diff -" header? */
 661         if (starts_with(line->buf, "diff -"))
 662                 return 1;
 663
 664         /* CVS "Index: " line? */
 665         if (starts_with(line->buf, "Index: "))
 666                 return 1;
 667
 668         /*
 669          * "--- <filename>" starts patches without headers
 670          * "---<sp>*" is a manual separator
 671          */
 672         if (line->len < 4)
 673                 return 0;
 674
 675         if (starts_with(line->buf, "---")) {
 676                 /* space followed by a filename? */
 677                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 678                         return 1;
 679                 /* Just whitespace? */
 680                 for (i = 3; i < line->len; i++) {
 681                         unsigned char c = line->buf[i];
 682                         if (c == '\n')
 683                                 return 1;
 684                         if (!isspace(c))
 685                                 break;
 686                 }
 687                 return 0;
 688         }
 689         return 0;
 690 }
 691
 692 static int is_scissors_line(const char *line)
 693 {
 694         const char *c;
 695         int scissors = 0, gap = 0;
 696         const char *first_nonblank = NULL, *last_nonblank = NULL;
 697         int visible, perforation = 0, in_perforation = 0;
 698
 699         for (c = line; *c; c++) {
 700                 if (isspace(*c)) {
 701                         if (in_perforation) {
 702                                 perforation++;
 703                                 gap++;
 704                         }
 705                         continue;
 706                 }
 707                 last_nonblank = c;
 708                 if (!first_nonblank)
 709                         first_nonblank = c;
 710                 if (*c == '-') {
 711                         in_perforation = 1;
 712                         perforation++;
 713                         continue;
 714                 }
 715                 if (starts_with(c, ">8") || starts_with(c, "8<") ||
 716                     starts_with(c, ">%") || starts_with(c, "%<")) {
 717                         in_perforation = 1;
 718                         perforation += 2;
 719                         scissors += 2;
 720                         c++;
 721                         continue;
 722                 }
 723                 in_perforation = 0;
 724         }
 725
 726         /*
 727          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 728          * Even though there can be arbitrary cruft on the same line
 729          * (e.g. "cut here"), in order to avoid misidentification, the
 730          * perforation must occupy more than a third of the visible
 731          * width of the line, and dashes and scissors must occupy more
 732          * than half of the perforation.
 733          */
 734
 735         if (first_nonblank && last_nonblank)
 736                 visible = last_nonblank - first_nonblank + 1;
 737         else
 738                 visible = 0;
 739         return (scissors && 8 <= visible &&
 740                 visible < perforation * 3 &&
 741                 gap * 2 < perforation);
 742 }
 743
 744 static void flush_inbody_header_accum(struct mailinfo *mi)
 745 {
 746         if (!mi->inbody_header_accum.len)
 747                 return;
 748         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 749                 BUG("inbody_header_accum, if not empty, must always contain a valid in-body header");
 750         strbuf_reset(&mi->inbody_header_accum);
 751 }
 752
 753 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 754 {
 755         if (mi->inbody_header_accum.len &&
 756             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 757                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 758                         /*
 759                          * This is a scissors line; do not consider this line
 760                          * as a header continuation line.
 761                          */
 762                         flush_inbody_header_accum(mi);
 763                         return 0;
 764                 }
 765                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 766                 strbuf_addbuf(&mi->inbody_header_accum, line);
 767                 return 1;
 768         }
 769
 770         flush_inbody_header_accum(mi);
 771
 772         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 773                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 774         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 775                 int i;
 776                 for (i = 0; i < ARRAY_SIZE(header); i++)
 777                         if (!strcmp("Subject", header[i])) {
 778                                 handle_header(&mi->s_hdr_data[i], line);
 779                                 return 1;
 780                         }
 781                 return 0;
 782         }
 783         if (is_inbody_header(mi, line)) {
 784                 strbuf_addbuf(&mi->inbody_header_accum, line);
 785                 return 1;
 786         }
 787         return 0;
 788 }
 789
 790 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 791 {
 792         assert(!mi->filter_stage);
 793
 794         if (mi->header_stage) {
 795                 if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 796                         if (mi->inbody_header_accum.len) {
 797                                 flush_inbody_header_accum(mi);
 798                                 mi->header_stage = 0;
 799                         }
 800                         return 0;
 801                 }
 802         }
 803
 804         if (mi->use_inbody_headers && mi->header_stage) {
 805                 mi->header_stage = check_inbody_header(mi, line);
 806                 if (mi->header_stage)
 807                         return 0;
 808         } else
 809                 /* Only trim the first (blank) line of the commit message
 810                  * when ignoring in-body headers.
 811                  */
 812                 mi->header_stage = 0;
 813
 814         /* normalize the log message to UTF-8. */
 815         if (convert_to_utf8(mi, line, mi->charset.buf))
 816                 return 0; /* mi->input_error already set */
 817
 818         if (mi->use_scissors && is_scissors_line(line->buf)) {
 819                 int i;
 820
 821                 strbuf_setlen(&mi->log_message, 0);
 822                 mi->header_stage = 1;
 823
 824                 /*
 825                  * We may have already read "secondary headers"; purge
 826                  * them to give ourselves a clean restart.
 827                  */
 828                 for (i = 0; i < ARRAY_SIZE(header); i++) {
 829                         if (mi->s_hdr_data[i])
 830                                 strbuf_release(mi->s_hdr_data[i]);
 831                         FREE_AND_NULL(mi->s_hdr_data[i]);
 832                 }
 833                 return 0;
 834         }
 835
 836         if (patchbreak(line)) {
 837                 if (mi->message_id)
 838                         strbuf_addf(&mi->log_message,
 839                                     "Message-ID: %s\n", mi->message_id);
 840                 return 1;
 841         }
 842
 843         strbuf_addbuf(&mi->log_message, line);
 844         return 0;
 845 }
 846
 847 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 848 {
 849         fwrite(line->buf, 1, line->len, mi->patchfile);
 850         mi->patch_lines++;
 851 }
 852
 853 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 854 {
 855         switch (mi->filter_stage) {
 856         case 0:
 857                 if (!handle_commit_msg(mi, line))
 858                         break;
 859                 mi->filter_stage++;
 860                 /* fallthrough */
 861         case 1:
 862                 handle_patch(mi, line);
 863                 break;
 864         }
 865 }
 866
 867 static int is_rfc2822_header(const struct strbuf *line)
 868 {
 869         /*
 870          * The section that defines the loosest possible
 871          * field name is "3.6.8 Optional fields".
 872          *
 873          * optional-field = field-name ":" unstructured CRLF
 874          * field-name = 1*ftext
 875          * ftext = %d33-57 / %59-126
 876          */
 877         int ch;
 878         char *cp = line->buf;
 879
 880         /* Count mbox From headers as headers */
 881         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 882                 return 1;
 883
 884         while ((ch = *cp++)) {
 885                 if (ch == ':')
 886                         return 1;
 887                 if ((33 <= ch && ch <= 57) ||
 888                     (59 <= ch && ch <= 126))
 889                         continue;
 890                 break;
 891         }
 892         return 0;
 893 }
 894
 895 static int read_one_header_line(struct strbuf *line, FILE *in)
 896 {
 897         struct strbuf continuation = STRBUF_INIT;
 898
 899         /* Get the first part of the line. */
 900         if (strbuf_getline_lf(line, in))
 901                 return 0;
 902
 903         /*
 904          * Is it an empty line or not a valid rfc2822 header?
 905          * If so, stop here, and return false ("not a header")
 906          */
 907         strbuf_rtrim(line);
 908         if (!line->len || !is_rfc2822_header(line)) {
 909                 /* Re-add the newline */
 910                 strbuf_addch(line, '\n');
 911                 return 0;
 912         }
 913
 914         /*
 915          * Now we need to eat all the continuation lines..
 916          * Yuck, 2822 header "folding"
 917          */
 918         for (;;) {
 919                 int peek;
 920
 921                 peek = fgetc(in);
 922                 if (peek == EOF)
 923                         break;
 924                 ungetc(peek, in);
 925                 if (peek != ' ' && peek != '\t')
 926                         break;
 927                 if (strbuf_getline_lf(&continuation, in))
 928                         break;
 929                 continuation.buf[0] = ' ';
 930                 strbuf_rtrim(&continuation);
 931                 strbuf_addbuf(line, &continuation);
 932         }
 933         strbuf_release(&continuation);
 934
 935         return 1;
 936 }
 937
 938 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 939 {
 940         while (!strbuf_getline_lf(line, mi->input)) {
 941                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 942                         return 1;
 943         }
 944         return 0;
 945 }
 946
 947 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 948 {
 949         struct strbuf newline = STRBUF_INIT;
 950
 951         strbuf_addch(&newline, '\n');
 952 again:
 953         if (line->len >= (*(mi->content_top))->len + 2 &&
 954             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 955                 /* we hit an end boundary */
 956                 /* pop the current boundary off the stack */
 957                 strbuf_release(*(mi->content_top));
 958                 FREE_AND_NULL(*(mi->content_top));
 959
 960                 /* technically won't happen as is_multipart_boundary()
 961                    will fail first.  But just in case..
 962                  */
 963                 if (--mi->content_top < mi->content) {
 964                         error("Detected mismatched boundaries, can't recover");
 965                         mi->input_error = -1;
 966                         mi->content_top = mi->content;
 967                         strbuf_release(&newline);
 968                         return 0;
 969                 }
 970                 handle_filter(mi, &newline);
 971                 strbuf_release(&newline);
 972                 if (mi->input_error)
 973                         return 0;
 974
 975                 /* skip to the next boundary */
 976                 if (!find_boundary(mi, line))
 977                         return 0;
 978                 goto again;
 979         }
 980
 981         /* set some defaults */
 982         mi->transfer_encoding = TE_DONTCARE;
 983         strbuf_reset(&mi->charset);
 984
 985         /* slurp in this section's info */
 986         while (read_one_header_line(line, mi->input))
 987                 check_header(mi, line, mi->p_hdr_data, 0);
 988
 989         strbuf_release(&newline);
 990         /* replenish line */
 991         if (strbuf_getline_lf(line, mi->input))
 992                 return 0;
 993         strbuf_addch(line, '\n');
 994         return 1;
 995 }
 996
 997 static void handle_filter_flowed(struct mailinfo *mi, struct strbuf *line,
 998                                  struct strbuf *prev)
 999 {
1000         size_t len = line->len;
1001         const char *rest;
1002
1003         if (!mi->format_flowed) {
1004                 if (len >= 2 &&
1005                     line->buf[len - 2] == '\r' &&
1006                     line->buf[len - 1] == '\n') {
1007                         mi->have_quoted_cr = 1;
1008                         if (mi->quoted_cr == quoted_cr_strip) {
1009                                 strbuf_setlen(line, len - 2);
1010                                 strbuf_addch(line, '\n');
1011                                 len--;
1012                         }
1013                 }
1014                 handle_filter(mi, line);
1015                 return;
1016         }
1017
1018         if (line->buf[len - 1] == '\n') {
1019                 len--;
1020                 if (len && line->buf[len - 1] == '\r')
1021                         len--;
1022         }
1023
1024         /* Keep signature separator as-is. */
1025         if (skip_prefix(line->buf, "-- ", &rest) && rest - line->buf == len) {
1026                 if (prev->len) {
1027                         handle_filter(mi, prev);
1028                         strbuf_reset(prev);
1029                 }
1030                 handle_filter(mi, line);
1031                 return;
1032         }
1033
1034         /* Unstuff space-stuffed line. */
1035         if (len && line->buf[0] == ' ') {
1036                 strbuf_remove(line, 0, 1);
1037                 len--;
1038         }
1039
1040         /* Save flowed line for later, but without the soft line break. */
1041         if (len && line->buf[len - 1] == ' ') {
1042                 strbuf_add(prev, line->buf, len - !!mi->delsp);
1043                 return;
1044         }
1045
1046         /* Prepend any previous partial lines */
1047         strbuf_insert(line, 0, prev->buf, prev->len);
1048         strbuf_reset(prev);
1049
1050         handle_filter(mi, line);
1051 }
1052
1053 static void summarize_quoted_cr(struct mailinfo *mi)
1054 {
1055         if (mi->have_quoted_cr &&
1056             mi->quoted_cr == quoted_cr_warn)
1057                 warning(_("quoted CRLF detected"));
1058 }
1059
1060 static void handle_body(struct mailinfo *mi, struct strbuf *line)
1061 {
1062         struct strbuf prev = STRBUF_INIT;
1063
1064         /* Skip up to the first boundary */
1065         if (*(mi->content_top)) {
1066                 if (!find_boundary(mi, line))
1067                         goto handle_body_out;
1068         }
1069
1070         do {
1071                 /* process any boundary lines */
1072                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
1073                         /* flush any leftover */
1074                         if (prev.len) {
1075                                 handle_filter(mi, &prev);
1076                                 strbuf_reset(&prev);
1077                         }
1078                         summarize_quoted_cr(mi);
1079                         mi->have_quoted_cr = 0;
1080                         if (!handle_boundary(mi, line))
1081                                 goto handle_body_out;
1082                 }
1083
1084                 /* Unwrap transfer encoding */
1085                 decode_transfer_encoding(mi, line);
1086
1087                 switch (mi->transfer_encoding) {
1088                 case TE_BASE64:
1089                 case TE_QP:
1090                 {
1091                         struct strbuf **lines, **it, *sb;
1092
1093                         /* Prepend any previous partial lines */
1094                         strbuf_insert(line, 0, prev.buf, prev.len);
1095                         strbuf_reset(&prev);
1096
1097                         /*
1098                          * This is a decoded line that may contain
1099                          * multiple new lines.  Pass only one chunk
1100                          * at a time to handle_filter()
1101                          */
1102                         lines = strbuf_split(line, '\n');
1103                         for (it = lines; (sb = *it); it++) {
1104                                 if (!*(it + 1)) /* The last line */
1105                                         if (sb->buf[sb->len - 1] != '\n') {
1106                                                 /* Partial line, save it for later. */
1107                                                 strbuf_addbuf(&prev, sb);
1108                                                 break;
1109                                         }
1110                                 handle_filter_flowed(mi, sb, &prev);
1111                         }
1112                         /*
1113                          * The partial chunk is saved in "prev" and will be
1114                          * appended by the next iteration of read_line_with_nul().
1115                          */
1116                         strbuf_list_free(lines);
1117                         break;
1118                 }
1119                 default:
1120                         handle_filter_flowed(mi, line, &prev);
1121                 }
1122
1123                 if (mi->input_error)
1124                         break;
1125         } while (!strbuf_getwholeline(line, mi->input, '\n'));
1126
1127         if (prev.len)
1128                 handle_filter(mi, &prev);
1129         summarize_quoted_cr(mi);
1130
1131         flush_inbody_header_accum(mi);
1132
1133 handle_body_out:
1134         strbuf_release(&prev);
1135 }
1136
1137 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
1138 {
1139         const char *sp = data->buf;
1140         while (1) {
1141                 char *ep = strchr(sp, '\n');
1142                 int len;
1143                 if (!ep)
1144                         len = strlen(sp);
1145                 else
1146                         len = ep - sp;
1147                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
1148                 if (!ep)
1149                         break;
1150                 sp = ep + 1;
1151         }
1152 }
1153
1154 static void handle_info(struct mailinfo *mi)
1155 {
1156         struct strbuf *hdr;
1157         int i;
1158
1159         for (i = 0; i < ARRAY_SIZE(header); i++) {
1160                 /* only print inbody headers if we output a patch file */
1161                 if (mi->patch_lines && mi->s_hdr_data[i])
1162                         hdr = mi->s_hdr_data[i];
1163                 else if (mi->p_hdr_data[i])
1164                         hdr = mi->p_hdr_data[i];
1165                 else
1166                         continue;
1167
1168                 if (memchr(hdr->buf, '\0', hdr->len)) {
1169                         error("a NUL byte in '%s' is not allowed.", header[i]);
1170                         mi->input_error = -1;
1171                 }
1172
1173                 if (!strcmp(header[i], "Subject")) {
1174                         if (!mi->keep_subject) {
1175                                 cleanup_subject(mi, hdr);
1176                                 cleanup_space(hdr);
1177                         }
1178                         output_header_lines(mi->output, "Subject", hdr);
1179                 } else if (!strcmp(header[i], "From")) {
1180                         cleanup_space(hdr);
1181                         handle_from(mi, hdr);
1182                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
1183                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
1184                 } else {
1185                         cleanup_space(hdr);
1186                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
1187                 }
1188         }
1189         fprintf(mi->output, "\n");
1190 }
1191
1192 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
1193 {
1194         FILE *cmitmsg;
1195         int peek;
1196         struct strbuf line = STRBUF_INIT;
1197
1198         cmitmsg = fopen(msg, "w");
1199         if (!cmitmsg) {
1200                 perror(msg);
1201                 return -1;
1202         }
1203         mi->patchfile = fopen(patch, "w");
1204         if (!mi->patchfile) {
1205                 perror(patch);
1206                 fclose(cmitmsg);
1207                 return -1;
1208         }
1209
1210         mi->p_hdr_data = xcalloc(ARRAY_SIZE(header), sizeof(*(mi->p_hdr_data)));
1211         mi->s_hdr_data = xcalloc(ARRAY_SIZE(header), sizeof(*(mi->s_hdr_data)));
1212
1213         do {
1214                 peek = fgetc(mi->input);
1215                 if (peek == EOF) {
1216                         fclose(cmitmsg);
1217                         return error("empty patch: '%s'", patch);
1218                 }
1219         } while (isspace(peek));
1220         ungetc(peek, mi->input);
1221
1222         /* process the email header */
1223         while (read_one_header_line(&line, mi->input))
1224                 check_header(mi, &line, mi->p_hdr_data, 1);
1225
1226         handle_body(mi, &line);
1227         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1228         fclose(cmitmsg);
1229         fclose(mi->patchfile);
1230
1231         handle_info(mi);
1232         strbuf_release(&line);
1233         return mi->input_error;
1234 }
1235
1236 int mailinfo_parse_quoted_cr_action(const char *actionstr, int *action)
1237 {
1238         if (!strcmp(actionstr, "nowarn"))
1239                 *action = quoted_cr_nowarn;
1240         else if (!strcmp(actionstr, "warn"))
1241                 *action = quoted_cr_warn;
1242         else if (!strcmp(actionstr, "strip"))
1243                 *action = quoted_cr_strip;
1244         else
1245                 return -1;
1246         return 0;
1247 }
1248
1249 static int git_mailinfo_config(const char *var, const char *value,
1250                                const struct config_context *ctx, void *mi_)
1251 {
1252         struct mailinfo *mi = mi_;
1253
1254         if (!starts_with(var, "mailinfo."))
1255                 return git_default_config(var, value, ctx, NULL);
1256         if (!strcmp(var, "mailinfo.scissors")) {
1257                 mi->use_scissors = git_config_bool(var, value);
1258                 return 0;
1259         }
1260         if (!strcmp(var, "mailinfo.quotedcr")) {
1261                 if (!value)
1262                         return config_error_nonbool(var);
1263                 if (mailinfo_parse_quoted_cr_action(value, &mi->quoted_cr) != 0)
1264                         return error(_("bad action '%s' for '%s'"), value, var);
1265                 return 0;
1266         }
1267         /* perhaps others here */
1268         return 0;
1269 }
1270
1271 void setup_mailinfo(struct mailinfo *mi)
1272 {
1273         memset(mi, 0, sizeof(*mi));
1274         strbuf_init(&mi->name, 0);
1275         strbuf_init(&mi->email, 0);
1276         strbuf_init(&mi->charset, 0);
1277         strbuf_init(&mi->log_message, 0);
1278         strbuf_init(&mi->inbody_header_accum, 0);
1279         mi->quoted_cr = quoted_cr_warn;
1280         mi->header_stage = 1;
1281         mi->use_inbody_headers = 1;
1282         mi->content_top = mi->content;
1283         git_config(git_mailinfo_config, mi);
1284 }
1285
1286 void clear_mailinfo(struct mailinfo *mi)
1287 {
1288         strbuf_release(&mi->name);
1289         strbuf_release(&mi->email);
1290         strbuf_release(&mi->charset);
1291         strbuf_release(&mi->inbody_header_accum);
1292         free(mi->message_id);
1293
1294         for (size_t i = 0; i < ARRAY_SIZE(header); i++) {
1295                 if (!mi->p_hdr_data[i])
1296                         continue;
1297                 strbuf_release(mi->p_hdr_data[i]);
1298                 free(mi->p_hdr_data[i]);
1299         }
1300         free(mi->p_hdr_data);
1301
1302         for (size_t i = 0; i < ARRAY_SIZE(header); i++) {
1303                 if (!mi->s_hdr_data[i])
1304                         continue;
1305                 strbuf_release(mi->s_hdr_data[i]);
1306                 free(mi->s_hdr_data[i]);
1307         }
1308         free(mi->s_hdr_data);
1309
1310         while (mi->content < mi->content_top) {
1311                 free(*(mi->content_top));
1312                 mi->content_top--;
1313         }
1314
1315         strbuf_release(&mi->log_message);
1316 }