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 #define MAX_HDR_PARSED 10
 352 static const char *header[MAX_HDR_PARSED] = {
 353         "From","Subject","Date",
 354 };
 355
 356 static inline int skip_header(const struct strbuf *line, const char *hdr,
 357                               const char **outval)
 358 {
 359         const char *val;
 360         if (!skip_iprefix(line->buf, hdr, &val) ||
 361             *val++ != ':')
 362                 return 0;
 363         while (isspace(*val))
 364                 val++;
 365         *outval = val;
 366         return 1;
 367 }
 368
 369 static int is_format_patch_separator(const char *line, int len)
 370 {
 371         static const char SAMPLE[] =
 372                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 373         const char *cp;
 374
 375         if (len != strlen(SAMPLE))
 376                 return 0;
 377         if (!skip_prefix(line, "From ", &cp))
 378                 return 0;
 379         if (strspn(cp, "0123456789abcdef") != 40)
 380                 return 0;
 381         cp += 40;
 382         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 383 }
 384
 385 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 386 {
 387         const char *in = q_seg->buf;
 388         int c;
 389         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 390         strbuf_init(out, q_seg->len);
 391
 392         while ((c = *in++) != 0) {
 393                 if (c == '=') {
 394                         int ch, d = *in;
 395                         if (d == '\n' || !d)
 396                                 break; /* drop trailing newline */
 397                         ch = hex2chr(in);
 398                         if (ch >= 0) {
 399                                 strbuf_addch(out, ch);
 400                                 in += 2;
 401                                 continue;
 402                         }
 403                         /* garbage -- fall through */
 404                 }
 405                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 406                         c = 0x20;
 407                 strbuf_addch(out, c);
 408         }
 409         return out;
 410 }
 411
 412 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 413 {
 414         /* Decode in..ep, possibly in-place to ot */
 415         int c, pos = 0, acc = 0;
 416         const char *in = b_seg->buf;
 417         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 418         strbuf_init(out, b_seg->len);
 419
 420         while ((c = *in++) != 0) {
 421                 if (c == '+')
 422                         c = 62;
 423                 else if (c == '/')
 424                         c = 63;
 425                 else if ('A' <= c && c <= 'Z')
 426                         c -= 'A';
 427                 else if ('a' <= c && c <= 'z')
 428                         c -= 'a' - 26;
 429                 else if ('0' <= c && c <= '9')
 430                         c -= '0' - 52;
 431                 else
 432                         continue; /* garbage */
 433                 switch (pos++) {
 434                 case 0:
 435                         acc = (c << 2);
 436                         break;
 437                 case 1:
 438                         strbuf_addch(out, (acc | (c >> 4)));
 439                         acc = (c & 15) << 4;
 440                         break;
 441                 case 2:
 442                         strbuf_addch(out, (acc | (c >> 2)));
 443                         acc = (c & 3) << 6;
 444                         break;
 445                 case 3:
 446                         strbuf_addch(out, (acc | c));
 447                         acc = pos = 0;
 448                         break;
 449                 }
 450         }
 451         return out;
 452 }
 453
 454 static int convert_to_utf8(struct mailinfo *mi,
 455                            struct strbuf *line, const char *charset)
 456 {
 457         char *out;
 458         size_t out_len;
 459
 460         if (!mi->metainfo_charset || !charset || !*charset)
 461                 return 0;
 462
 463         if (same_encoding(mi->metainfo_charset, charset))
 464                 return 0;
 465         out = reencode_string_len(line->buf, line->len,
 466                                   mi->metainfo_charset, charset, &out_len);
 467         if (!out) {
 468                 mi->input_error = -1;
 469                 return error("cannot convert from %s to %s",
 470                              charset, mi->metainfo_charset);
 471         }
 472         strbuf_attach(line, out, out_len, out_len);
 473         return 0;
 474 }
 475
 476 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 477 {
 478         char *in, *ep, *cp;
 479         struct strbuf outbuf = STRBUF_INIT, *dec;
 480         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 481         int found_error = 1; /* pessimism */
 482
 483         in = it->buf;
 484         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 485                 int encoding;
 486                 strbuf_reset(&charset_q);
 487                 strbuf_reset(&piecebuf);
 488
 489                 if (in != ep) {
 490                         /*
 491                          * We are about to process an encoded-word
 492                          * that begins at ep, but there is something
 493                          * before the encoded word.
 494                          */
 495                         char *scan;
 496                         for (scan = in; scan < ep; scan++)
 497                                 if (!isspace(*scan))
 498                                         break;
 499
 500                         if (scan != ep || in == it->buf) {
 501                                 /*
 502                                  * We should not lose that "something",
 503                                  * unless we have just processed an
 504                                  * encoded-word, and there is only LWS
 505                                  * before the one we are about to process.
 506                                  */
 507                                 strbuf_add(&outbuf, in, ep - in);
 508                         }
 509                 }
 510                 /* E.g.
 511                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 512                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 513                  */
 514                 ep += 2;
 515
 516                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 517                         goto release_return;
 518
 519                 if (cp + 3 - it->buf > it->len)
 520                         goto release_return;
 521                 strbuf_add(&charset_q, ep, cp - ep);
 522
 523                 encoding = cp[1];
 524                 if (!encoding || cp[2] != '?')
 525                         goto release_return;
 526                 ep = strstr(cp + 3, "?=");
 527                 if (!ep)
 528                         goto release_return;
 529                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 530                 switch (tolower(encoding)) {
 531                 default:
 532                         goto release_return;
 533                 case 'b':
 534                         dec = decode_b_segment(&piecebuf);
 535                         break;
 536                 case 'q':
 537                         dec = decode_q_segment(&piecebuf, 1);
 538                         break;
 539                 }
 540                 if (convert_to_utf8(mi, dec, charset_q.buf))
 541                         goto release_return;
 542
 543                 strbuf_addbuf(&outbuf, dec);
 544                 strbuf_release(dec);
 545                 free(dec);
 546                 in = ep + 2;
 547         }
 548         strbuf_addstr(&outbuf, in);
 549         strbuf_reset(it);
 550         strbuf_addbuf(it, &outbuf);
 551         found_error = 0;
 552 release_return:
 553         strbuf_release(&outbuf);
 554         strbuf_release(&charset_q);
 555         strbuf_release(&piecebuf);
 556
 557         if (found_error)
 558                 mi->input_error = -1;
 559 }
 560
 561 /*
 562  * Returns true if "line" contains a header matching "hdr", in which case "val"
 563  * will contain the value of the header with any RFC2047 B and Q encoding
 564  * unwrapped, and optionally normalize the meta information to utf8.
 565  */
 566 static int parse_header(const struct strbuf *line,
 567                         const char *hdr,
 568                         struct mailinfo *mi,
 569                         struct strbuf *val)
 570 {
 571         const char *val_str;
 572
 573         if (!skip_header(line, hdr, &val_str))
 574                 return 0;
 575         strbuf_addstr(val, val_str);
 576         decode_header(mi, val);
 577         return 1;
 578 }
 579
 580 static int check_header(struct mailinfo *mi,
 581                         const struct strbuf *line,
 582                         struct strbuf *hdr_data[], int overwrite)
 583 {
 584         int i, ret = 0;
 585         struct strbuf sb = STRBUF_INIT;
 586
 587         /* search for the interesting parts */
 588         for (i = 0; header[i]; i++) {
 589                 if ((!hdr_data[i] || overwrite) &&
 590                     parse_header(line, header[i], mi, &sb)) {
 591                         handle_header(&hdr_data[i], &sb);
 592                         ret = 1;
 593                         goto check_header_out;
 594                 }
 595         }
 596
 597         /* Content stuff */
 598         if (parse_header(line, "Content-Type", mi, &sb)) {
 599                 handle_content_type(mi, &sb);
 600                 ret = 1;
 601                 goto check_header_out;
 602         }
 603         if (parse_header(line, "Content-Transfer-Encoding", mi, &sb)) {
 604                 handle_content_transfer_encoding(mi, &sb);
 605                 ret = 1;
 606                 goto check_header_out;
 607         }
 608         if (parse_header(line, "Message-ID", mi, &sb)) {
 609                 if (mi->add_message_id)
 610                         mi->message_id = strbuf_detach(&sb, NULL);
 611                 ret = 1;
 612                 goto check_header_out;
 613         }
 614
 615 check_header_out:
 616         strbuf_release(&sb);
 617         return ret;
 618 }
 619
 620 /*
 621  * Returns 1 if the given line or any line beginning with the given line is an
 622  * in-body header (that is, check_header will succeed when passed
 623  * mi->s_hdr_data).
 624  */
 625 static int is_inbody_header(const struct mailinfo *mi,
 626                             const struct strbuf *line)
 627 {
 628         int i;
 629         const char *val;
 630         for (i = 0; header[i]; i++)
 631                 if (!mi->s_hdr_data[i] && skip_header(line, header[i], &val))
 632                         return 1;
 633         return 0;
 634 }
 635
 636 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 637 {
 638         struct strbuf *ret;
 639
 640         switch (mi->transfer_encoding) {
 641         case TE_QP:
 642                 ret = decode_q_segment(line, 0);
 643                 break;
 644         case TE_BASE64:
 645                 ret = decode_b_segment(line);
 646                 break;
 647         case TE_DONTCARE:
 648         default:
 649                 return;
 650         }
 651         strbuf_reset(line);
 652         strbuf_addbuf(line, ret);
 653         strbuf_release(ret);
 654         free(ret);
 655 }
 656
 657 static inline int patchbreak(const struct strbuf *line)
 658 {
 659         size_t i;
 660
 661         /* Beginning of a "diff -" header? */
 662         if (starts_with(line->buf, "diff -"))
 663                 return 1;
 664
 665         /* CVS "Index: " line? */
 666         if (starts_with(line->buf, "Index: "))
 667                 return 1;
 668
 669         /*
 670          * "--- <filename>" starts patches without headers
 671          * "---<sp>*" is a manual separator
 672          */
 673         if (line->len < 4)
 674                 return 0;
 675
 676         if (starts_with(line->buf, "---")) {
 677                 /* space followed by a filename? */
 678                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 679                         return 1;
 680                 /* Just whitespace? */
 681                 for (i = 3; i < line->len; i++) {
 682                         unsigned char c = line->buf[i];
 683                         if (c == '\n')
 684                                 return 1;
 685                         if (!isspace(c))
 686                                 break;
 687                 }
 688                 return 0;
 689         }
 690         return 0;
 691 }
 692
 693 static int is_scissors_line(const char *line)
 694 {
 695         const char *c;
 696         int scissors = 0, gap = 0;
 697         const char *first_nonblank = NULL, *last_nonblank = NULL;
 698         int visible, perforation = 0, in_perforation = 0;
 699
 700         for (c = line; *c; c++) {
 701                 if (isspace(*c)) {
 702                         if (in_perforation) {
 703                                 perforation++;
 704                                 gap++;
 705                         }
 706                         continue;
 707                 }
 708                 last_nonblank = c;
 709                 if (!first_nonblank)
 710                         first_nonblank = c;
 711                 if (*c == '-') {
 712                         in_perforation = 1;
 713                         perforation++;
 714                         continue;
 715                 }
 716                 if (starts_with(c, ">8") || starts_with(c, "8<") ||
 717                     starts_with(c, ">%") || starts_with(c, "%<")) {
 718                         in_perforation = 1;
 719                         perforation += 2;
 720                         scissors += 2;
 721                         c++;
 722                         continue;
 723                 }
 724                 in_perforation = 0;
 725         }
 726
 727         /*
 728          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 729          * Even though there can be arbitrary cruft on the same line
 730          * (e.g. "cut here"), in order to avoid misidentification, the
 731          * perforation must occupy more than a third of the visible
 732          * width of the line, and dashes and scissors must occupy more
 733          * than half of the perforation.
 734          */
 735
 736         if (first_nonblank && last_nonblank)
 737                 visible = last_nonblank - first_nonblank + 1;
 738         else
 739                 visible = 0;
 740         return (scissors && 8 <= visible &&
 741                 visible < perforation * 3 &&
 742                 gap * 2 < perforation);
 743 }
 744
 745 static void flush_inbody_header_accum(struct mailinfo *mi)
 746 {
 747         if (!mi->inbody_header_accum.len)
 748                 return;
 749         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 750                 BUG("inbody_header_accum, if not empty, must always contain a valid in-body header");
 751         strbuf_reset(&mi->inbody_header_accum);
 752 }
 753
 754 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 755 {
 756         if (mi->inbody_header_accum.len &&
 757             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 758                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 759                         /*
 760                          * This is a scissors line; do not consider this line
 761                          * as a header continuation line.
 762                          */
 763                         flush_inbody_header_accum(mi);
 764                         return 0;
 765                 }
 766                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 767                 strbuf_addbuf(&mi->inbody_header_accum, line);
 768                 return 1;
 769         }
 770
 771         flush_inbody_header_accum(mi);
 772
 773         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 774                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 775         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 776                 int i;
 777                 for (i = 0; header[i]; i++)
 778                         if (!strcmp("Subject", header[i])) {
 779                                 handle_header(&mi->s_hdr_data[i], line);
 780                                 return 1;
 781                         }
 782                 return 0;
 783         }
 784         if (is_inbody_header(mi, line)) {
 785                 strbuf_addbuf(&mi->inbody_header_accum, line);
 786                 return 1;
 787         }
 788         return 0;
 789 }
 790
 791 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 792 {
 793         assert(!mi->filter_stage);
 794
 795         if (mi->header_stage) {
 796                 if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 797                         if (mi->inbody_header_accum.len) {
 798                                 flush_inbody_header_accum(mi);
 799                                 mi->header_stage = 0;
 800                         }
 801                         return 0;
 802                 }
 803         }
 804
 805         if (mi->use_inbody_headers && mi->header_stage) {
 806                 mi->header_stage = check_inbody_header(mi, line);
 807                 if (mi->header_stage)
 808                         return 0;
 809         } else
 810                 /* Only trim the first (blank) line of the commit message
 811                  * when ignoring in-body headers.
 812                  */
 813                 mi->header_stage = 0;
 814
 815         /* normalize the log message to UTF-8. */
 816         if (convert_to_utf8(mi, line, mi->charset.buf))
 817                 return 0; /* mi->input_error already set */
 818
 819         if (mi->use_scissors && is_scissors_line(line->buf)) {
 820                 int i;
 821
 822                 strbuf_setlen(&mi->log_message, 0);
 823                 mi->header_stage = 1;
 824
 825                 /*
 826                  * We may have already read "secondary headers"; purge
 827                  * them to give ourselves a clean restart.
 828                  */
 829                 for (i = 0; header[i]; i++) {
 830                         if (mi->s_hdr_data[i])
 831                                 strbuf_release(mi->s_hdr_data[i]);
 832                         FREE_AND_NULL(mi->s_hdr_data[i]);
 833                 }
 834                 return 0;
 835         }
 836
 837         if (patchbreak(line)) {
 838                 if (mi->message_id)
 839                         strbuf_addf(&mi->log_message,
 840                                     "Message-ID: %s\n", mi->message_id);
 841                 return 1;
 842         }
 843
 844         strbuf_addbuf(&mi->log_message, line);
 845         return 0;
 846 }
 847
 848 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 849 {
 850         fwrite(line->buf, 1, line->len, mi->patchfile);
 851         mi->patch_lines++;
 852 }
 853
 854 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 855 {
 856         switch (mi->filter_stage) {
 857         case 0:
 858                 if (!handle_commit_msg(mi, line))
 859                         break;
 860                 mi->filter_stage++;
 861                 /* fallthrough */
 862         case 1:
 863                 handle_patch(mi, line);
 864                 break;
 865         }
 866 }
 867
 868 static int is_rfc2822_header(const struct strbuf *line)
 869 {
 870         /*
 871          * The section that defines the loosest possible
 872          * field name is "3.6.8 Optional fields".
 873          *
 874          * optional-field = field-name ":" unstructured CRLF
 875          * field-name = 1*ftext
 876          * ftext = %d33-57 / %59-126
 877          */
 878         int ch;
 879         char *cp = line->buf;
 880
 881         /* Count mbox From headers as headers */
 882         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 883                 return 1;
 884
 885         while ((ch = *cp++)) {
 886                 if (ch == ':')
 887                         return 1;
 888                 if ((33 <= ch && ch <= 57) ||
 889                     (59 <= ch && ch <= 126))
 890                         continue;
 891                 break;
 892         }
 893         return 0;
 894 }
 895
 896 static int read_one_header_line(struct strbuf *line, FILE *in)
 897 {
 898         struct strbuf continuation = STRBUF_INIT;
 899
 900         /* Get the first part of the line. */
 901         if (strbuf_getline_lf(line, in))
 902                 return 0;
 903
 904         /*
 905          * Is it an empty line or not a valid rfc2822 header?
 906          * If so, stop here, and return false ("not a header")
 907          */
 908         strbuf_rtrim(line);
 909         if (!line->len || !is_rfc2822_header(line)) {
 910                 /* Re-add the newline */
 911                 strbuf_addch(line, '\n');
 912                 return 0;
 913         }
 914
 915         /*
 916          * Now we need to eat all the continuation lines..
 917          * Yuck, 2822 header "folding"
 918          */
 919         for (;;) {
 920                 int peek;
 921
 922                 peek = fgetc(in);
 923                 if (peek == EOF)
 924                         break;
 925                 ungetc(peek, in);
 926                 if (peek != ' ' && peek != '\t')
 927                         break;
 928                 if (strbuf_getline_lf(&continuation, in))
 929                         break;
 930                 continuation.buf[0] = ' ';
 931                 strbuf_rtrim(&continuation);
 932                 strbuf_addbuf(line, &continuation);
 933         }
 934         strbuf_release(&continuation);
 935
 936         return 1;
 937 }
 938
 939 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 940 {
 941         while (!strbuf_getline_lf(line, mi->input)) {
 942                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 943                         return 1;
 944         }
 945         return 0;
 946 }
 947
 948 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 949 {
 950         struct strbuf newline = STRBUF_INIT;
 951
 952         strbuf_addch(&newline, '\n');
 953 again:
 954         if (line->len >= (*(mi->content_top))->len + 2 &&
 955             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 956                 /* we hit an end boundary */
 957                 /* pop the current boundary off the stack */
 958                 strbuf_release(*(mi->content_top));
 959                 FREE_AND_NULL(*(mi->content_top));
 960
 961                 /* technically won't happen as is_multipart_boundary()
 962                    will fail first.  But just in case..
 963                  */
 964                 if (--mi->content_top < mi->content) {
 965                         error("Detected mismatched boundaries, can't recover");
 966                         mi->input_error = -1;
 967                         mi->content_top = mi->content;
 968                         strbuf_release(&newline);
 969                         return 0;
 970                 }
 971                 handle_filter(mi, &newline);
 972                 strbuf_release(&newline);
 973                 if (mi->input_error)
 974                         return 0;
 975
 976                 /* skip to the next boundary */
 977                 if (!find_boundary(mi, line))
 978                         return 0;
 979                 goto again;
 980         }
 981
 982         /* set some defaults */
 983         mi->transfer_encoding = TE_DONTCARE;
 984         strbuf_reset(&mi->charset);
 985
 986         /* slurp in this section's info */
 987         while (read_one_header_line(line, mi->input))
 988                 check_header(mi, line, mi->p_hdr_data, 0);
 989
 990         strbuf_release(&newline);
 991         /* replenish line */
 992         if (strbuf_getline_lf(line, mi->input))
 993                 return 0;
 994         strbuf_addch(line, '\n');
 995         return 1;
 996 }
 997
 998 static void handle_filter_flowed(struct mailinfo *mi, struct strbuf *line,
 999                                  struct strbuf *prev)
1000 {
1001         size_t len = line->len;
1002         const char *rest;
1003
1004         if (!mi->format_flowed) {
1005                 if (len >= 2 &&
1006                     line->buf[len - 2] == '\r' &&
1007                     line->buf[len - 1] == '\n') {
1008                         mi->have_quoted_cr = 1;
1009                         if (mi->quoted_cr == quoted_cr_strip) {
1010                                 strbuf_setlen(line, len - 2);
1011                                 strbuf_addch(line, '\n');
1012                                 len--;
1013                         }
1014                 }
1015                 handle_filter(mi, line);
1016                 return;
1017         }
1018
1019         if (line->buf[len - 1] == '\n') {
1020                 len--;
1021                 if (len && line->buf[len - 1] == '\r')
1022                         len--;
1023         }
1024
1025         /* Keep signature separator as-is. */
1026         if (skip_prefix(line->buf, "-- ", &rest) && rest - line->buf == len) {
1027                 if (prev->len) {
1028                         handle_filter(mi, prev);
1029                         strbuf_reset(prev);
1030                 }
1031                 handle_filter(mi, line);
1032                 return;
1033         }
1034
1035         /* Unstuff space-stuffed line. */
1036         if (len && line->buf[0] == ' ') {
1037                 strbuf_remove(line, 0, 1);
1038                 len--;
1039         }
1040
1041         /* Save flowed line for later, but without the soft line break. */
1042         if (len && line->buf[len - 1] == ' ') {
1043                 strbuf_add(prev, line->buf, len - !!mi->delsp);
1044                 return;
1045         }
1046
1047         /* Prepend any previous partial lines */
1048         strbuf_insert(line, 0, prev->buf, prev->len);
1049         strbuf_reset(prev);
1050
1051         handle_filter(mi, line);
1052 }
1053
1054 static void summarize_quoted_cr(struct mailinfo *mi)
1055 {
1056         if (mi->have_quoted_cr &&
1057             mi->quoted_cr == quoted_cr_warn)
1058                 warning(_("quoted CRLF detected"));
1059 }
1060
1061 static void handle_body(struct mailinfo *mi, struct strbuf *line)
1062 {
1063         struct strbuf prev = STRBUF_INIT;
1064
1065         /* Skip up to the first boundary */
1066         if (*(mi->content_top)) {
1067                 if (!find_boundary(mi, line))
1068                         goto handle_body_out;
1069         }
1070
1071         do {
1072                 /* process any boundary lines */
1073                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
1074                         /* flush any leftover */
1075                         if (prev.len) {
1076                                 handle_filter(mi, &prev);
1077                                 strbuf_reset(&prev);
1078                         }
1079                         summarize_quoted_cr(mi);
1080                         mi->have_quoted_cr = 0;
1081                         if (!handle_boundary(mi, line))
1082                                 goto handle_body_out;
1083                 }
1084
1085                 /* Unwrap transfer encoding */
1086                 decode_transfer_encoding(mi, line);
1087
1088                 switch (mi->transfer_encoding) {
1089                 case TE_BASE64:
1090                 case TE_QP:
1091                 {
1092                         struct strbuf **lines, **it, *sb;
1093
1094                         /* Prepend any previous partial lines */
1095                         strbuf_insert(line, 0, prev.buf, prev.len);
1096                         strbuf_reset(&prev);
1097
1098                         /*
1099                          * This is a decoded line that may contain
1100                          * multiple new lines.  Pass only one chunk
1101                          * at a time to handle_filter()
1102                          */
1103                         lines = strbuf_split(line, '\n');
1104                         for (it = lines; (sb = *it); it++) {
1105                                 if (!*(it + 1)) /* The last line */
1106                                         if (sb->buf[sb->len - 1] != '\n') {
1107                                                 /* Partial line, save it for later. */
1108                                                 strbuf_addbuf(&prev, sb);
1109                                                 break;
1110                                         }
1111                                 handle_filter_flowed(mi, sb, &prev);
1112                         }
1113                         /*
1114                          * The partial chunk is saved in "prev" and will be
1115                          * appended by the next iteration of read_line_with_nul().
1116                          */
1117                         strbuf_list_free(lines);
1118                         break;
1119                 }
1120                 default:
1121                         handle_filter_flowed(mi, line, &prev);
1122                 }
1123
1124                 if (mi->input_error)
1125                         break;
1126         } while (!strbuf_getwholeline(line, mi->input, '\n'));
1127
1128         if (prev.len)
1129                 handle_filter(mi, &prev);
1130         summarize_quoted_cr(mi);
1131
1132         flush_inbody_header_accum(mi);
1133
1134 handle_body_out:
1135         strbuf_release(&prev);
1136 }
1137
1138 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
1139 {
1140         const char *sp = data->buf;
1141         while (1) {
1142                 char *ep = strchr(sp, '\n');
1143                 int len;
1144                 if (!ep)
1145                         len = strlen(sp);
1146                 else
1147                         len = ep - sp;
1148                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
1149                 if (!ep)
1150                         break;
1151                 sp = ep + 1;
1152         }
1153 }
1154
1155 static void handle_info(struct mailinfo *mi)
1156 {
1157         struct strbuf *hdr;
1158         int i;
1159
1160         for (i = 0; header[i]; i++) {
1161                 /* only print inbody headers if we output a patch file */
1162                 if (mi->patch_lines && mi->s_hdr_data[i])
1163                         hdr = mi->s_hdr_data[i];
1164                 else if (mi->p_hdr_data[i])
1165                         hdr = mi->p_hdr_data[i];
1166                 else
1167                         continue;
1168
1169                 if (memchr(hdr->buf, '\0', hdr->len)) {
1170                         error("a NUL byte in '%s' is not allowed.", header[i]);
1171                         mi->input_error = -1;
1172                 }
1173
1174                 if (!strcmp(header[i], "Subject")) {
1175                         if (!mi->keep_subject) {
1176                                 cleanup_subject(mi, hdr);
1177                                 cleanup_space(hdr);
1178                         }
1179                         output_header_lines(mi->output, "Subject", hdr);
1180                 } else if (!strcmp(header[i], "From")) {
1181                         cleanup_space(hdr);
1182                         handle_from(mi, hdr);
1183                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
1184                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
1185                 } else {
1186                         cleanup_space(hdr);
1187                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
1188                 }
1189         }
1190         fprintf(mi->output, "\n");
1191 }
1192
1193 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
1194 {
1195         FILE *cmitmsg;
1196         int peek;
1197         struct strbuf line = STRBUF_INIT;
1198
1199         cmitmsg = fopen(msg, "w");
1200         if (!cmitmsg) {
1201                 perror(msg);
1202                 return -1;
1203         }
1204         mi->patchfile = fopen(patch, "w");
1205         if (!mi->patchfile) {
1206                 perror(patch);
1207                 fclose(cmitmsg);
1208                 return -1;
1209         }
1210
1211         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
1212         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
1213
1214         do {
1215                 peek = fgetc(mi->input);
1216                 if (peek == EOF) {
1217                         fclose(cmitmsg);
1218                         return error("empty patch: '%s'", patch);
1219                 }
1220         } while (isspace(peek));
1221         ungetc(peek, mi->input);
1222
1223         /* process the email header */
1224         while (read_one_header_line(&line, mi->input))
1225                 check_header(mi, &line, mi->p_hdr_data, 1);
1226
1227         handle_body(mi, &line);
1228         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1229         fclose(cmitmsg);
1230         fclose(mi->patchfile);
1231
1232         handle_info(mi);
1233         strbuf_release(&line);
1234         return mi->input_error;
1235 }
1236
1237 int mailinfo_parse_quoted_cr_action(const char *actionstr, int *action)
1238 {
1239         if (!strcmp(actionstr, "nowarn"))
1240                 *action = quoted_cr_nowarn;
1241         else if (!strcmp(actionstr, "warn"))
1242                 *action = quoted_cr_warn;
1243         else if (!strcmp(actionstr, "strip"))
1244                 *action = quoted_cr_strip;
1245         else
1246                 return -1;
1247         return 0;
1248 }
1249
1250 static int git_mailinfo_config(const char *var, const char *value,
1251                                const struct config_context *ctx, void *mi_)
1252 {
1253         struct mailinfo *mi = mi_;
1254
1255         if (!starts_with(var, "mailinfo."))
1256                 return git_default_config(var, value, ctx, NULL);
1257         if (!strcmp(var, "mailinfo.scissors")) {
1258                 mi->use_scissors = git_config_bool(var, value);
1259                 return 0;
1260         }
1261         if (!strcmp(var, "mailinfo.quotedcr")) {
1262                 if (!value)
1263                         return config_error_nonbool(var);
1264                 if (mailinfo_parse_quoted_cr_action(value, &mi->quoted_cr) != 0)
1265                         return error(_("bad action '%s' for '%s'"), value, var);
1266                 return 0;
1267         }
1268         /* perhaps others here */
1269         return 0;
1270 }
1271
1272 void setup_mailinfo(struct mailinfo *mi)
1273 {
1274         memset(mi, 0, sizeof(*mi));
1275         strbuf_init(&mi->name, 0);
1276         strbuf_init(&mi->email, 0);
1277         strbuf_init(&mi->charset, 0);
1278         strbuf_init(&mi->log_message, 0);
1279         strbuf_init(&mi->inbody_header_accum, 0);
1280         mi->quoted_cr = quoted_cr_warn;
1281         mi->header_stage = 1;
1282         mi->use_inbody_headers = 1;
1283         mi->content_top = mi->content;
1284         git_config(git_mailinfo_config, mi);
1285 }
1286
1287 void clear_mailinfo(struct mailinfo *mi)
1288 {
1289         strbuf_release(&mi->name);
1290         strbuf_release(&mi->email);
1291         strbuf_release(&mi->charset);
1292         strbuf_release(&mi->inbody_header_accum);
1293         free(mi->message_id);
1294
1295         strbuf_list_free(mi->p_hdr_data);
1296         strbuf_list_free(mi->s_hdr_data);
1297
1298         while (mi->content < mi->content_top) {
1299                 free(*(mi->content_top));
1300                 mi->content_top--;
1301         }
1302
1303         strbuf_release(&mi->log_message);
1304 }