mailinfo.c

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