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