mailinfo.c

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