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 skip_header(const struct strbuf *line, const char *hdr,
 350                               const char **outval)
 351 {
 352         const char *val;
 353         if (!skip_iprefix(line->buf, hdr, &val) ||
 354             *val++ != ':' ||
 355             !isspace(*val++))
 356                 return 0;
 357         *outval = val;
 358         return 1;
 359 }
 360
 361 static int is_format_patch_separator(const char *line, int len)
 362 {
 363         static const char SAMPLE[] =
 364                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 365         const char *cp;
 366
 367         if (len != strlen(SAMPLE))
 368                 return 0;
 369         if (!skip_prefix(line, "From ", &cp))
 370                 return 0;
 371         if (strspn(cp, "0123456789abcdef") != 40)
 372                 return 0;
 373         cp += 40;
 374         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 375 }
 376
 377 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 378 {
 379         const char *in = q_seg->buf;
 380         int c;
 381         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 382         strbuf_init(out, q_seg->len);
 383
 384         while ((c = *in++) != 0) {
 385                 if (c == '=') {
 386                         int ch, d = *in;
 387                         if (d == '\n' || !d)
 388                                 break; /* drop trailing newline */
 389                         ch = hex2chr(in);
 390                         if (ch >= 0) {
 391                                 strbuf_addch(out, ch);
 392                                 in += 2;
 393                                 continue;
 394                         }
 395                         /* garbage -- fall through */
 396                 }
 397                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 398                         c = 0x20;
 399                 strbuf_addch(out, c);
 400         }
 401         return out;
 402 }
 403
 404 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 405 {
 406         /* Decode in..ep, possibly in-place to ot */
 407         int c, pos = 0, acc = 0;
 408         const char *in = b_seg->buf;
 409         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 410         strbuf_init(out, b_seg->len);
 411
 412         while ((c = *in++) != 0) {
 413                 if (c == '+')
 414                         c = 62;
 415                 else if (c == '/')
 416                         c = 63;
 417                 else if ('A' <= c && c <= 'Z')
 418                         c -= 'A';
 419                 else if ('a' <= c && c <= 'z')
 420                         c -= 'a' - 26;
 421                 else if ('0' <= c && c <= '9')
 422                         c -= '0' - 52;
 423                 else
 424                         continue; /* garbage */
 425                 switch (pos++) {
 426                 case 0:
 427                         acc = (c << 2);
 428                         break;
 429                 case 1:
 430                         strbuf_addch(out, (acc | (c >> 4)));
 431                         acc = (c & 15) << 4;
 432                         break;
 433                 case 2:
 434                         strbuf_addch(out, (acc | (c >> 2)));
 435                         acc = (c & 3) << 6;
 436                         break;
 437                 case 3:
 438                         strbuf_addch(out, (acc | c));
 439                         acc = pos = 0;
 440                         break;
 441                 }
 442         }
 443         return out;
 444 }
 445
 446 static int convert_to_utf8(struct mailinfo *mi,
 447                            struct strbuf *line, const char *charset)
 448 {
 449         char *out;
 450
 451         if (!mi->metainfo_charset || !charset || !*charset)
 452                 return 0;
 453
 454         if (same_encoding(mi->metainfo_charset, charset))
 455                 return 0;
 456         out = reencode_string(line->buf, mi->metainfo_charset, charset);
 457         if (!out) {
 458                 mi->input_error = -1;
 459                 return error("cannot convert from %s to %s",
 460                              charset, mi->metainfo_charset);
 461         }
 462         strbuf_attach(line, out, strlen(out), strlen(out));
 463         return 0;
 464 }
 465
 466 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 467 {
 468         char *in, *ep, *cp;
 469         struct strbuf outbuf = STRBUF_INIT, *dec;
 470         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 471         int found_error = 1; /* pessimism */
 472
 473         in = it->buf;
 474         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 475                 int encoding;
 476                 strbuf_reset(&charset_q);
 477                 strbuf_reset(&piecebuf);
 478
 479                 if (in != ep) {
 480                         /*
 481                          * We are about to process an encoded-word
 482                          * that begins at ep, but there is something
 483                          * before the encoded word.
 484                          */
 485                         char *scan;
 486                         for (scan = in; scan < ep; scan++)
 487                                 if (!isspace(*scan))
 488                                         break;
 489
 490                         if (scan != ep || in == it->buf) {
 491                                 /*
 492                                  * We should not lose that "something",
 493                                  * unless we have just processed an
 494                                  * encoded-word, and there is only LWS
 495                                  * before the one we are about to process.
 496                                  */
 497                                 strbuf_add(&outbuf, in, ep - in);
 498                         }
 499                 }
 500                 /* E.g.
 501                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 502                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 503                  */
 504                 ep += 2;
 505
 506                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 507                         goto release_return;
 508
 509                 if (cp + 3 - it->buf > it->len)
 510                         goto release_return;
 511                 strbuf_add(&charset_q, ep, cp - ep);
 512
 513                 encoding = cp[1];
 514                 if (!encoding || cp[2] != '?')
 515                         goto release_return;
 516                 ep = strstr(cp + 3, "?=");
 517                 if (!ep)
 518                         goto release_return;
 519                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 520                 switch (tolower(encoding)) {
 521                 default:
 522                         goto release_return;
 523                 case 'b':
 524                         dec = decode_b_segment(&piecebuf);
 525                         break;
 526                 case 'q':
 527                         dec = decode_q_segment(&piecebuf, 1);
 528                         break;
 529                 }
 530                 if (convert_to_utf8(mi, dec, charset_q.buf))
 531                         goto release_return;
 532
 533                 strbuf_addbuf(&outbuf, dec);
 534                 strbuf_release(dec);
 535                 free(dec);
 536                 in = ep + 2;
 537         }
 538         strbuf_addstr(&outbuf, in);
 539         strbuf_reset(it);
 540         strbuf_addbuf(it, &outbuf);
 541         found_error = 0;
 542 release_return:
 543         strbuf_release(&outbuf);
 544         strbuf_release(&charset_q);
 545         strbuf_release(&piecebuf);
 546
 547         if (found_error)
 548                 mi->input_error = -1;
 549 }
 550
 551 static int check_header(struct mailinfo *mi,
 552                         const struct strbuf *line,
 553                         struct strbuf *hdr_data[], int overwrite)
 554 {
 555         int i, ret = 0;
 556         struct strbuf sb = STRBUF_INIT;
 557         const char *val;
 558
 559         /* search for the interesting parts */
 560         for (i = 0; header[i]; i++) {
 561                 if ((!hdr_data[i] || overwrite) &&
 562                     skip_header(line, header[i], &val)) {
 563                         /* Unwrap inline B and Q encoding, and optionally
 564                          * normalize the meta information to utf8.
 565                          */
 566                         strbuf_addstr(&sb, val);
 567                         decode_header(mi, &sb);
 568                         handle_header(&hdr_data[i], &sb);
 569                         ret = 1;
 570                         goto check_header_out;
 571                 }
 572         }
 573
 574         /* Content stuff */
 575         if (skip_header(line, "Content-Type", &val)) {
 576                 strbuf_addstr(&sb, val);
 577                 decode_header(mi, &sb);
 578                 handle_content_type(mi, &sb);
 579                 ret = 1;
 580                 goto check_header_out;
 581         }
 582         if (skip_header(line, "Content-Transfer-Encoding", &val)) {
 583                 strbuf_addstr(&sb, val);
 584                 decode_header(mi, &sb);
 585                 handle_content_transfer_encoding(mi, &sb);
 586                 ret = 1;
 587                 goto check_header_out;
 588         }
 589         if (skip_header(line, "Message-Id", &val)) {
 590                 strbuf_addstr(&sb, val);
 591                 decode_header(mi, &sb);
 592                 if (mi->add_message_id)
 593                         mi->message_id = strbuf_detach(&sb, NULL);
 594                 ret = 1;
 595                 goto check_header_out;
 596         }
 597
 598 check_header_out:
 599         strbuf_release(&sb);
 600         return ret;
 601 }
 602
 603 /*
 604  * Returns 1 if the given line or any line beginning with the given line is an
 605  * in-body header (that is, check_header will succeed when passed
 606  * mi->s_hdr_data).
 607  */
 608 static int is_inbody_header(const struct mailinfo *mi,
 609                             const struct strbuf *line)
 610 {
 611         int i;
 612         const char *val;
 613         for (i = 0; header[i]; i++)
 614                 if (!mi->s_hdr_data[i] && skip_header(line, header[i], &val))
 615                         return 1;
 616         return 0;
 617 }
 618
 619 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 620 {
 621         struct strbuf *ret;
 622
 623         switch (mi->transfer_encoding) {
 624         case TE_QP:
 625                 ret = decode_q_segment(line, 0);
 626                 break;
 627         case TE_BASE64:
 628                 ret = decode_b_segment(line);
 629                 break;
 630         case TE_DONTCARE:
 631         default:
 632                 return;
 633         }
 634         strbuf_reset(line);
 635         strbuf_addbuf(line, ret);
 636         strbuf_release(ret);
 637         free(ret);
 638 }
 639
 640 static inline int patchbreak(const struct strbuf *line)
 641 {
 642         size_t i;
 643
 644         /* Beginning of a "diff -" header? */
 645         if (starts_with(line->buf, "diff -"))
 646                 return 1;
 647
 648         /* CVS "Index: " line? */
 649         if (starts_with(line->buf, "Index: "))
 650                 return 1;
 651
 652         /*
 653          * "--- <filename>" starts patches without headers
 654          * "---<sp>*" is a manual separator
 655          */
 656         if (line->len < 4)
 657                 return 0;
 658
 659         if (starts_with(line->buf, "---")) {
 660                 /* space followed by a filename? */
 661                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 662                         return 1;
 663                 /* Just whitespace? */
 664                 for (i = 3; i < line->len; i++) {
 665                         unsigned char c = line->buf[i];
 666                         if (c == '\n')
 667                                 return 1;
 668                         if (!isspace(c))
 669                                 break;
 670                 }
 671                 return 0;
 672         }
 673         return 0;
 674 }
 675
 676 static int is_scissors_line(const char *line)
 677 {
 678         const char *c;
 679         int scissors = 0, gap = 0;
 680         const char *first_nonblank = NULL, *last_nonblank = NULL;
 681         int visible, perforation = 0, in_perforation = 0;
 682
 683         for (c = line; *c; c++) {
 684                 if (isspace(*c)) {
 685                         if (in_perforation) {
 686                                 perforation++;
 687                                 gap++;
 688                         }
 689                         continue;
 690                 }
 691                 last_nonblank = c;
 692                 if (first_nonblank == NULL)
 693                         first_nonblank = c;
 694                 if (*c == '-') {
 695                         in_perforation = 1;
 696                         perforation++;
 697                         continue;
 698                 }
 699                 if ((!memcmp(c, ">8", 2) || !memcmp(c, "8<", 2) ||
 700                      !memcmp(c, ">%", 2) || !memcmp(c, "%<", 2))) {
 701                         in_perforation = 1;
 702                         perforation += 2;
 703                         scissors += 2;
 704                         c++;
 705                         continue;
 706                 }
 707                 in_perforation = 0;
 708         }
 709
 710         /*
 711          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 712          * Even though there can be arbitrary cruft on the same line
 713          * (e.g. "cut here"), in order to avoid misidentification, the
 714          * perforation must occupy more than a third of the visible
 715          * width of the line, and dashes and scissors must occupy more
 716          * than half of the perforation.
 717          */
 718
 719         if (first_nonblank && last_nonblank)
 720                 visible = last_nonblank - first_nonblank + 1;
 721         else
 722                 visible = 0;
 723         return (scissors && 8 <= visible &&
 724                 visible < perforation * 3 &&
 725                 gap * 2 < perforation);
 726 }
 727
 728 static void flush_inbody_header_accum(struct mailinfo *mi)
 729 {
 730         if (!mi->inbody_header_accum.len)
 731                 return;
 732         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 733                 BUG("inbody_header_accum, if not empty, must always contain a valid in-body header");
 734         strbuf_reset(&mi->inbody_header_accum);
 735 }
 736
 737 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 738 {
 739         if (mi->inbody_header_accum.len &&
 740             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 741                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 742                         /*
 743                          * This is a scissors line; do not consider this line
 744                          * as a header continuation line.
 745                          */
 746                         flush_inbody_header_accum(mi);
 747                         return 0;
 748                 }
 749                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 750                 strbuf_addbuf(&mi->inbody_header_accum, line);
 751                 return 1;
 752         }
 753
 754         flush_inbody_header_accum(mi);
 755
 756         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 757                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 758         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 759                 int i;
 760                 for (i = 0; header[i]; i++)
 761                         if (!strcmp("Subject", header[i])) {
 762                                 handle_header(&mi->s_hdr_data[i], line);
 763                                 return 1;
 764                         }
 765                 return 0;
 766         }
 767         if (is_inbody_header(mi, line)) {
 768                 strbuf_addbuf(&mi->inbody_header_accum, line);
 769                 return 1;
 770         }
 771         return 0;
 772 }
 773
 774 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 775 {
 776         assert(!mi->filter_stage);
 777
 778         if (mi->header_stage) {
 779                 if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 780                         if (mi->inbody_header_accum.len) {
 781                                 flush_inbody_header_accum(mi);
 782                                 mi->header_stage = 0;
 783                         }
 784                         return 0;
 785                 }
 786         }
 787
 788         if (mi->use_inbody_headers && mi->header_stage) {
 789                 mi->header_stage = check_inbody_header(mi, line);
 790                 if (mi->header_stage)
 791                         return 0;
 792         } else
 793                 /* Only trim the first (blank) line of the commit message
 794                  * when ignoring in-body headers.
 795                  */
 796                 mi->header_stage = 0;
 797
 798         /* normalize the log message to UTF-8. */
 799         if (convert_to_utf8(mi, line, mi->charset.buf))
 800                 return 0; /* mi->input_error already set */
 801
 802         if (mi->use_scissors && is_scissors_line(line->buf)) {
 803                 int i;
 804
 805                 strbuf_setlen(&mi->log_message, 0);
 806                 mi->header_stage = 1;
 807
 808                 /*
 809                  * We may have already read "secondary headers"; purge
 810                  * them to give ourselves a clean restart.
 811                  */
 812                 for (i = 0; header[i]; i++) {
 813                         if (mi->s_hdr_data[i])
 814                                 strbuf_release(mi->s_hdr_data[i]);
 815                         mi->s_hdr_data[i] = NULL;
 816                 }
 817                 return 0;
 818         }
 819
 820         if (patchbreak(line)) {
 821                 if (mi->message_id)
 822                         strbuf_addf(&mi->log_message,
 823                                     "Message-Id: %s\n", mi->message_id);
 824                 return 1;
 825         }
 826
 827         strbuf_addbuf(&mi->log_message, line);
 828         return 0;
 829 }
 830
 831 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 832 {
 833         fwrite(line->buf, 1, line->len, mi->patchfile);
 834         mi->patch_lines++;
 835 }
 836
 837 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 838 {
 839         switch (mi->filter_stage) {
 840         case 0:
 841                 if (!handle_commit_msg(mi, line))
 842                         break;
 843                 mi->filter_stage++;
 844                 /* fallthrough */
 845         case 1:
 846                 handle_patch(mi, line);
 847                 break;
 848         }
 849 }
 850
 851 static int is_rfc2822_header(const struct strbuf *line)
 852 {
 853         /*
 854          * The section that defines the loosest possible
 855          * field name is "3.6.8 Optional fields".
 856          *
 857          * optional-field = field-name ":" unstructured CRLF
 858          * field-name = 1*ftext
 859          * ftext = %d33-57 / %59-126
 860          */
 861         int ch;
 862         char *cp = line->buf;
 863
 864         /* Count mbox From headers as headers */
 865         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 866                 return 1;
 867
 868         while ((ch = *cp++)) {
 869                 if (ch == ':')
 870                         return 1;
 871                 if ((33 <= ch && ch <= 57) ||
 872                     (59 <= ch && ch <= 126))
 873                         continue;
 874                 break;
 875         }
 876         return 0;
 877 }
 878
 879 static int read_one_header_line(struct strbuf *line, FILE *in)
 880 {
 881         struct strbuf continuation = STRBUF_INIT;
 882
 883         /* Get the first part of the line. */
 884         if (strbuf_getline_lf(line, in))
 885                 return 0;
 886
 887         /*
 888          * Is it an empty line or not a valid rfc2822 header?
 889          * If so, stop here, and return false ("not a header")
 890          */
 891         strbuf_rtrim(line);
 892         if (!line->len || !is_rfc2822_header(line)) {
 893                 /* Re-add the newline */
 894                 strbuf_addch(line, '\n');
 895                 return 0;
 896         }
 897
 898         /*
 899          * Now we need to eat all the continuation lines..
 900          * Yuck, 2822 header "folding"
 901          */
 902         for (;;) {
 903                 int peek;
 904
 905                 peek = fgetc(in);
 906                 if (peek == EOF)
 907                         break;
 908                 ungetc(peek, in);
 909                 if (peek != ' ' && peek != '\t')
 910                         break;
 911                 if (strbuf_getline_lf(&continuation, in))
 912                         break;
 913                 continuation.buf[0] = ' ';
 914                 strbuf_rtrim(&continuation);
 915                 strbuf_addbuf(line, &continuation);
 916         }
 917         strbuf_release(&continuation);
 918
 919         return 1;
 920 }
 921
 922 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 923 {
 924         while (!strbuf_getline_lf(line, mi->input)) {
 925                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 926                         return 1;
 927         }
 928         return 0;
 929 }
 930
 931 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 932 {
 933         struct strbuf newline = STRBUF_INIT;
 934
 935         strbuf_addch(&newline, '\n');
 936 again:
 937         if (line->len >= (*(mi->content_top))->len + 2 &&
 938             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 939                 /* we hit an end boundary */
 940                 /* pop the current boundary off the stack */
 941                 strbuf_release(*(mi->content_top));
 942                 FREE_AND_NULL(*(mi->content_top));
 943
 944                 /* technically won't happen as is_multipart_boundary()
 945                    will fail first.  But just in case..
 946                  */
 947                 if (--mi->content_top < mi->content) {
 948                         error("Detected mismatched boundaries, can't recover");
 949                         mi->input_error = -1;
 950                         mi->content_top = mi->content;
 951                         strbuf_release(&newline);
 952                         return 0;
 953                 }
 954                 handle_filter(mi, &newline);
 955                 strbuf_release(&newline);
 956                 if (mi->input_error)
 957                         return 0;
 958
 959                 /* skip to the next boundary */
 960                 if (!find_boundary(mi, line))
 961                         return 0;
 962                 goto again;
 963         }
 964
 965         /* set some defaults */
 966         mi->transfer_encoding = TE_DONTCARE;
 967         strbuf_reset(&mi->charset);
 968
 969         /* slurp in this section's info */
 970         while (read_one_header_line(line, mi->input))
 971                 check_header(mi, line, mi->p_hdr_data, 0);
 972
 973         strbuf_release(&newline);
 974         /* replenish line */
 975         if (strbuf_getline_lf(line, mi->input))
 976                 return 0;
 977         strbuf_addch(line, '\n');
 978         return 1;
 979 }
 980
 981 static void handle_filter_flowed(struct mailinfo *mi, struct strbuf *line,
 982                                  struct strbuf *prev)
 983 {
 984         size_t len = line->len;
 985         const char *rest;
 986
 987         if (!mi->format_flowed) {
 988                 handle_filter(mi, line);
 989                 return;
 990         }
 991
 992         if (line->buf[len - 1] == '\n') {
 993                 len--;
 994                 if (len && line->buf[len - 1] == '\r')
 995                         len--;
 996         }
 997
 998         /* Keep signature separator as-is. */
 999         if (skip_prefix(line->buf, "-- ", &rest) && rest - line->buf == len) {
1000                 if (prev->len) {
1001                         handle_filter(mi, prev);
1002                         strbuf_reset(prev);
1003                 }
1004                 handle_filter(mi, line);
1005                 return;
1006         }
1007
1008         /* Unstuff space-stuffed line. */
1009         if (len && line->buf[0] == ' ') {
1010                 strbuf_remove(line, 0, 1);
1011                 len--;
1012         }
1013
1014         /* Save flowed line for later, but without the soft line break. */
1015         if (len && line->buf[len - 1] == ' ') {
1016                 strbuf_add(prev, line->buf, len - !!mi->delsp);
1017                 return;
1018         }
1019
1020         /* Prepend any previous partial lines */
1021         strbuf_insert(line, 0, prev->buf, prev->len);
1022         strbuf_reset(prev);
1023
1024         handle_filter(mi, line);
1025 }
1026
1027 static void handle_body(struct mailinfo *mi, struct strbuf *line)
1028 {
1029         struct strbuf prev = STRBUF_INIT;
1030
1031         /* Skip up to the first boundary */
1032         if (*(mi->content_top)) {
1033                 if (!find_boundary(mi, line))
1034                         goto handle_body_out;
1035         }
1036
1037         do {
1038                 /* process any boundary lines */
1039                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
1040                         /* flush any leftover */
1041                         if (prev.len) {
1042                                 handle_filter(mi, &prev);
1043                                 strbuf_reset(&prev);
1044                         }
1045                         if (!handle_boundary(mi, line))
1046                                 goto handle_body_out;
1047                 }
1048
1049                 /* Unwrap transfer encoding */
1050                 decode_transfer_encoding(mi, line);
1051
1052                 switch (mi->transfer_encoding) {
1053                 case TE_BASE64:
1054                 case TE_QP:
1055                 {
1056                         struct strbuf **lines, **it, *sb;
1057
1058                         /* Prepend any previous partial lines */
1059                         strbuf_insert(line, 0, prev.buf, prev.len);
1060                         strbuf_reset(&prev);
1061
1062                         /*
1063                          * This is a decoded line that may contain
1064                          * multiple new lines.  Pass only one chunk
1065                          * at a time to handle_filter()
1066                          */
1067                         lines = strbuf_split(line, '\n');
1068                         for (it = lines; (sb = *it); it++) {
1069                                 if (*(it + 1) == NULL) /* The last line */
1070                                         if (sb->buf[sb->len - 1] != '\n') {
1071                                                 /* Partial line, save it for later. */
1072                                                 strbuf_addbuf(&prev, sb);
1073                                                 break;
1074                                         }
1075                                 handle_filter_flowed(mi, sb, &prev);
1076                         }
1077                         /*
1078                          * The partial chunk is saved in "prev" and will be
1079                          * appended by the next iteration of read_line_with_nul().
1080                          */
1081                         strbuf_list_free(lines);
1082                         break;
1083                 }
1084                 default:
1085                         handle_filter_flowed(mi, line, &prev);
1086                 }
1087
1088                 if (mi->input_error)
1089                         break;
1090         } while (!strbuf_getwholeline(line, mi->input, '\n'));
1091
1092         if (prev.len)
1093                 handle_filter(mi, &prev);
1094
1095         flush_inbody_header_accum(mi);
1096
1097 handle_body_out:
1098         strbuf_release(&prev);
1099 }
1100
1101 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
1102 {
1103         const char *sp = data->buf;
1104         while (1) {
1105                 char *ep = strchr(sp, '\n');
1106                 int len;
1107                 if (!ep)
1108                         len = strlen(sp);
1109                 else
1110                         len = ep - sp;
1111                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
1112                 if (!ep)
1113                         break;
1114                 sp = ep + 1;
1115         }
1116 }
1117
1118 static void handle_info(struct mailinfo *mi)
1119 {
1120         struct strbuf *hdr;
1121         int i;
1122
1123         for (i = 0; header[i]; i++) {
1124                 /* only print inbody headers if we output a patch file */
1125                 if (mi->patch_lines && mi->s_hdr_data[i])
1126                         hdr = mi->s_hdr_data[i];
1127                 else if (mi->p_hdr_data[i])
1128                         hdr = mi->p_hdr_data[i];
1129                 else
1130                         continue;
1131
1132                 if (!strcmp(header[i], "Subject")) {
1133                         if (!mi->keep_subject) {
1134                                 cleanup_subject(mi, hdr);
1135                                 cleanup_space(hdr);
1136                         }
1137                         output_header_lines(mi->output, "Subject", hdr);
1138                 } else if (!strcmp(header[i], "From")) {
1139                         cleanup_space(hdr);
1140                         handle_from(mi, hdr);
1141                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
1142                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
1143                 } else {
1144                         cleanup_space(hdr);
1145                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
1146                 }
1147         }
1148         fprintf(mi->output, "\n");
1149 }
1150
1151 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
1152 {
1153         FILE *cmitmsg;
1154         int peek;
1155         struct strbuf line = STRBUF_INIT;
1156
1157         cmitmsg = fopen(msg, "w");
1158         if (!cmitmsg) {
1159                 perror(msg);
1160                 return -1;
1161         }
1162         mi->patchfile = fopen(patch, "w");
1163         if (!mi->patchfile) {
1164                 perror(patch);
1165                 fclose(cmitmsg);
1166                 return -1;
1167         }
1168
1169         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
1170         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
1171
1172         do {
1173                 peek = fgetc(mi->input);
1174                 if (peek == EOF) {
1175                         fclose(cmitmsg);
1176                         return error("empty patch: '%s'", patch);
1177                 }
1178         } while (isspace(peek));
1179         ungetc(peek, mi->input);
1180
1181         /* process the email header */
1182         while (read_one_header_line(&line, mi->input))
1183                 check_header(mi, &line, mi->p_hdr_data, 1);
1184
1185         handle_body(mi, &line);
1186         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1187         fclose(cmitmsg);
1188         fclose(mi->patchfile);
1189
1190         handle_info(mi);
1191         strbuf_release(&line);
1192         return mi->input_error;
1193 }
1194
1195 static int git_mailinfo_config(const char *var, const char *value, void *mi_)
1196 {
1197         struct mailinfo *mi = mi_;
1198
1199         if (!starts_with(var, "mailinfo."))
1200                 return git_default_config(var, value, NULL);
1201         if (!strcmp(var, "mailinfo.scissors")) {
1202                 mi->use_scissors = git_config_bool(var, value);
1203                 return 0;
1204         }
1205         /* perhaps others here */
1206         return 0;
1207 }
1208
1209 void setup_mailinfo(struct mailinfo *mi)
1210 {
1211         memset(mi, 0, sizeof(*mi));
1212         strbuf_init(&mi->name, 0);
1213         strbuf_init(&mi->email, 0);
1214         strbuf_init(&mi->charset, 0);
1215         strbuf_init(&mi->log_message, 0);
1216         strbuf_init(&mi->inbody_header_accum, 0);
1217         mi->header_stage = 1;
1218         mi->use_inbody_headers = 1;
1219         mi->content_top = mi->content;
1220         git_config(git_mailinfo_config, mi);
1221 }
1222
1223 void clear_mailinfo(struct mailinfo *mi)
1224 {
1225         int i;
1226
1227         strbuf_release(&mi->name);
1228         strbuf_release(&mi->email);
1229         strbuf_release(&mi->charset);
1230         strbuf_release(&mi->inbody_header_accum);
1231         free(mi->message_id);
1232
1233         if (mi->p_hdr_data)
1234                 for (i = 0; mi->p_hdr_data[i]; i++)
1235                         strbuf_release(mi->p_hdr_data[i]);
1236         free(mi->p_hdr_data);
1237         if (mi->s_hdr_data)
1238                 for (i = 0; mi->s_hdr_data[i]; i++)
1239                         strbuf_release(mi->s_hdr_data[i]);
1240         free(mi->s_hdr_data);
1241
1242         while (mi->content < mi->content_top) {
1243                 free(*(mi->content_top));
1244                 mi->content_top--;
1245         }
1246
1247         strbuf_release(&mi->log_message);
1248 }