mailinfo.c

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