builtin/mailinfo.c

   1 /*
   2  * Another stupid program, this one parsing the headers of an
   3  * email to figure out authorship and subject
   4  */
   5 #include "cache.h"
   6 #include "builtin.h"
   7 #include "utf8.h"
   8 #include "strbuf.h"
   9
  10 static FILE *cmitmsg, *patchfile, *fin, *fout;
  11
  12 static int keep_subject;
  13 static int keep_non_patch_brackets_in_subject;
  14 static const char *metainfo_charset;
  15 static struct strbuf line = STRBUF_INIT;
  16 static struct strbuf name = STRBUF_INIT;
  17 static struct strbuf email = STRBUF_INIT;
  18
  19 static enum  {
  20         TE_DONTCARE, TE_QP, TE_BASE64
  21 } transfer_encoding;
  22 static enum  {
  23         TYPE_TEXT, TYPE_OTHER
  24 } message_type;
  25
  26 static struct strbuf charset = STRBUF_INIT;
  27 static int patch_lines;
  28 static struct strbuf **p_hdr_data, **s_hdr_data;
  29 static int use_scissors;
  30 static int use_inbody_headers = 1;
  31
  32 #define MAX_HDR_PARSED 10
  33 #define MAX_BOUNDARIES 5
  34
  35 static void cleanup_space(struct strbuf *sb);
  36
  37
  38 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  39 {
  40         struct strbuf *src = name;
  41         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  42                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  43                 src = email;
  44         else if (name == out)
  45                 return;
  46         strbuf_reset(out);
  47         strbuf_addbuf(out, src);
  48 }
  49
  50 static void parse_bogus_from(const struct strbuf *line)
  51 {
  52         /* John Doe <johndoe> */
  53
  54         char *bra, *ket;
  55         /* This is fallback, so do not bother if we already have an
  56          * e-mail address.
  57          */
  58         if (email.len)
  59                 return;
  60
  61         bra = strchr(line->buf, '<');
  62         if (!bra)
  63                 return;
  64         ket = strchr(bra, '>');
  65         if (!ket)
  66                 return;
  67
  68         strbuf_reset(&email);
  69         strbuf_add(&email, bra + 1, ket - bra - 1);
  70
  71         strbuf_reset(&name);
  72         strbuf_add(&name, line->buf, bra - line->buf);
  73         strbuf_trim(&name);
  74         get_sane_name(&name, &name, &email);
  75 }
  76
  77 static void handle_from(const struct strbuf *from)
  78 {
  79         char *at;
  80         size_t el;
  81         struct strbuf f;
  82
  83         strbuf_init(&f, from->len);
  84         strbuf_addbuf(&f, from);
  85
  86         at = strchr(f.buf, '@');
  87         if (!at) {
  88                 parse_bogus_from(from);
  89                 return;
  90         }
  91
  92         /*
  93          * If we already have one email, don't take any confusing lines
  94          */
  95         if (email.len && strchr(at + 1, '@')) {
  96                 strbuf_release(&f);
  97                 return;
  98         }
  99
 100         /* Pick up the string around '@', possibly delimited with <>
 101          * pair; that is the email part.
 102          */
 103         while (at > f.buf) {
 104                 char c = at[-1];
 105                 if (isspace(c))
 106                         break;
 107                 if (c == '<') {
 108                         at[-1] = ' ';
 109                         break;
 110                 }
 111                 at--;
 112         }
 113         el = strcspn(at, " \n\t\r\v\f>");
 114         strbuf_reset(&email);
 115         strbuf_add(&email, at, el);
 116         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 117
 118         /* The remainder is name.  It could be
 119          *
 120          * - "John Doe <john.doe@xz>"                   (a), or
 121          * - "john.doe@xz (John Doe)"                   (b), or
 122          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 123          *
 124          * but we have removed the email part, so
 125          *
 126          * - remove extra spaces which could stay after email (case 'c'), and
 127          * - trim from both ends, possibly removing the () pair at the end
 128          *   (cases 'a' and 'b').
 129          */
 130         cleanup_space(&f);
 131         strbuf_trim(&f);
 132         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 133                 strbuf_remove(&f, 0, 1);
 134                 strbuf_setlen(&f, f.len - 1);
 135         }
 136
 137         get_sane_name(&name, &f, &email);
 138         strbuf_release(&f);
 139 }
 140
 141 static void handle_header(struct strbuf **out, const struct strbuf *line)
 142 {
 143         if (!*out) {
 144                 *out = xmalloc(sizeof(struct strbuf));
 145                 strbuf_init(*out, line->len);
 146         } else
 147                 strbuf_reset(*out);
 148
 149         strbuf_addbuf(*out, line);
 150 }
 151
 152 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 153  * to have enough heuristics to grok MIME encoded patches often found
 154  * on our mailing lists.  For example, we do not even treat header lines
 155  * case insensitively.
 156  */
 157
 158 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 159 {
 160         const char *ends, *ap = strcasestr(line, name);
 161         size_t sz;
 162
 163         if (!ap) {
 164                 strbuf_setlen(attr, 0);
 165                 return 0;
 166         }
 167         ap += strlen(name);
 168         if (*ap == '"') {
 169                 ap++;
 170                 ends = "\"";
 171         }
 172         else
 173                 ends = "; \t";
 174         sz = strcspn(ap, ends);
 175         strbuf_add(attr, ap, sz);
 176         return 1;
 177 }
 178
 179 static struct strbuf *content[MAX_BOUNDARIES];
 180
 181 static struct strbuf **content_top = content;
 182
 183 static void handle_content_type(struct strbuf *line)
 184 {
 185         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 186         strbuf_init(boundary, line->len);
 187
 188         if (!strcasestr(line->buf, "text/"))
 189                  message_type = TYPE_OTHER;
 190         if (slurp_attr(line->buf, "boundary=", boundary)) {
 191                 strbuf_insert(boundary, 0, "--", 2);
 192                 if (++content_top > &content[MAX_BOUNDARIES]) {
 193                         fprintf(stderr, "Too many boundaries to handle\n");
 194                         exit(1);
 195                 }
 196                 *content_top = boundary;
 197                 boundary = NULL;
 198         }
 199         slurp_attr(line->buf, "charset=", &charset);
 200
 201         if (boundary) {
 202                 strbuf_release(boundary);
 203                 free(boundary);
 204         }
 205 }
 206
 207 static void handle_content_transfer_encoding(const struct strbuf *line)
 208 {
 209         if (strcasestr(line->buf, "base64"))
 210                 transfer_encoding = TE_BASE64;
 211         else if (strcasestr(line->buf, "quoted-printable"))
 212                 transfer_encoding = TE_QP;
 213         else
 214                 transfer_encoding = TE_DONTCARE;
 215 }
 216
 217 static int is_multipart_boundary(const struct strbuf *line)
 218 {
 219         return (((*content_top)->len <= line->len) &&
 220                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
 221 }
 222
 223 static void cleanup_subject(struct strbuf *subject)
 224 {
 225         size_t at = 0;
 226
 227         while (at < subject->len) {
 228                 char *pos;
 229                 size_t remove;
 230
 231                 switch (subject->buf[at]) {
 232                 case 'r': case 'R':
 233                         if (subject->len <= at + 3)
 234                                 break;
 235                         if ((subject->buf[at + 1] == 'e' ||
 236                              subject->buf[at + 1] == 'E') &&
 237                             subject->buf[at + 2] == ':') {
 238                                 strbuf_remove(subject, at, 3);
 239                                 continue;
 240                         }
 241                         at++;
 242                         break;
 243                 case ' ': case '\t': case ':':
 244                         strbuf_remove(subject, at, 1);
 245                         continue;
 246                 case '[':
 247                         pos = strchr(subject->buf + at, ']');
 248                         if (!pos)
 249                                 break;
 250                         remove = pos - subject->buf + at + 1;
 251                         if (!keep_non_patch_brackets_in_subject ||
 252                             (7 <= remove &&
 253                              memmem(subject->buf + at, remove, "PATCH", 5)))
 254                                 strbuf_remove(subject, at, remove);
 255                         else {
 256                                 at += remove;
 257                                 /*
 258                                  * If the input had a space after the ], keep
 259                                  * it.  We don't bother with finding the end of
 260                                  * the space, since we later normalize it
 261                                  * anyway.
 262                                  */
 263                                 if (isspace(subject->buf[at]))
 264                                         at += 1;
 265                         }
 266                         continue;
 267                 }
 268                 break;
 269         }
 270         strbuf_trim(subject);
 271 }
 272
 273 static void cleanup_space(struct strbuf *sb)
 274 {
 275         size_t pos, cnt;
 276         for (pos = 0; pos < sb->len; pos++) {
 277                 if (isspace(sb->buf[pos])) {
 278                         sb->buf[pos] = ' ';
 279                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 280                         strbuf_remove(sb, pos + 1, cnt);
 281                 }
 282         }
 283 }
 284
 285 static void decode_header(struct strbuf *line);
 286 static const char *header[MAX_HDR_PARSED] = {
 287         "From","Subject","Date",
 288 };
 289
 290 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 291 {
 292         int len = strlen(hdr);
 293         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 294                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 295 }
 296
 297 static int check_header(const struct strbuf *line,
 298                                 struct strbuf *hdr_data[], int overwrite)
 299 {
 300         int i, ret = 0, len;
 301         struct strbuf sb = STRBUF_INIT;
 302         /* search for the interesting parts */
 303         for (i = 0; header[i]; i++) {
 304                 int len = strlen(header[i]);
 305                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 306                         /* Unwrap inline B and Q encoding, and optionally
 307                          * normalize the meta information to utf8.
 308                          */
 309                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 310                         decode_header(&sb);
 311                         handle_header(&hdr_data[i], &sb);
 312                         ret = 1;
 313                         goto check_header_out;
 314                 }
 315         }
 316
 317         /* Content stuff */
 318         if (cmp_header(line, "Content-Type")) {
 319                 len = strlen("Content-Type: ");
 320                 strbuf_add(&sb, line->buf + len, line->len - len);
 321                 decode_header(&sb);
 322                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 323                 handle_content_type(&sb);
 324                 ret = 1;
 325                 goto check_header_out;
 326         }
 327         if (cmp_header(line, "Content-Transfer-Encoding")) {
 328                 len = strlen("Content-Transfer-Encoding: ");
 329                 strbuf_add(&sb, line->buf + len, line->len - len);
 330                 decode_header(&sb);
 331                 handle_content_transfer_encoding(&sb);
 332                 ret = 1;
 333                 goto check_header_out;
 334         }
 335
 336         /* for inbody stuff */
 337         if (!prefixcmp(line->buf, ">From") && isspace(line->buf[5])) {
 338                 ret = 1; /* Should this return 0? */
 339                 goto check_header_out;
 340         }
 341         if (!prefixcmp(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 342                 for (i = 0; header[i]; i++) {
 343                         if (!memcmp("Subject", header[i], 7)) {
 344                                 handle_header(&hdr_data[i], line);
 345                                 ret = 1;
 346                                 goto check_header_out;
 347                         }
 348                 }
 349         }
 350
 351 check_header_out:
 352         strbuf_release(&sb);
 353         return ret;
 354 }
 355
 356 static int is_rfc2822_header(const struct strbuf *line)
 357 {
 358         /*
 359          * The section that defines the loosest possible
 360          * field name is "3.6.8 Optional fields".
 361          *
 362          * optional-field = field-name ":" unstructured CRLF
 363          * field-name = 1*ftext
 364          * ftext = %d33-57 / %59-126
 365          */
 366         int ch;
 367         char *cp = line->buf;
 368
 369         /* Count mbox From headers as headers */
 370         if (!prefixcmp(cp, "From ") || !prefixcmp(cp, ">From "))
 371                 return 1;
 372
 373         while ((ch = *cp++)) {
 374                 if (ch == ':')
 375                         return 1;
 376                 if ((33 <= ch && ch <= 57) ||
 377                     (59 <= ch && ch <= 126))
 378                         continue;
 379                 break;
 380         }
 381         return 0;
 382 }
 383
 384 static int read_one_header_line(struct strbuf *line, FILE *in)
 385 {
 386         /* Get the first part of the line. */
 387         if (strbuf_getline(line, in, '\n'))
 388                 return 0;
 389
 390         /*
 391          * Is it an empty line or not a valid rfc2822 header?
 392          * If so, stop here, and return false ("not a header")
 393          */
 394         strbuf_rtrim(line);
 395         if (!line->len || !is_rfc2822_header(line)) {
 396                 /* Re-add the newline */
 397                 strbuf_addch(line, '\n');
 398                 return 0;
 399         }
 400
 401         /*
 402          * Now we need to eat all the continuation lines..
 403          * Yuck, 2822 header "folding"
 404          */
 405         for (;;) {
 406                 int peek;
 407                 struct strbuf continuation = STRBUF_INIT;
 408
 409                 peek = fgetc(in); ungetc(peek, in);
 410                 if (peek != ' ' && peek != '\t')
 411                         break;
 412                 if (strbuf_getline(&continuation, in, '\n'))
 413                         break;
 414                 continuation.buf[0] = ' ';
 415                 strbuf_rtrim(&continuation);
 416                 strbuf_addbuf(line, &continuation);
 417         }
 418
 419         return 1;
 420 }
 421
 422 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 423 {
 424         const char *in = q_seg->buf;
 425         int c;
 426         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 427         strbuf_init(out, q_seg->len);
 428
 429         while ((c = *in++) != 0) {
 430                 if (c == '=') {
 431                         int d = *in++;
 432                         if (d == '\n' || !d)
 433                                 break; /* drop trailing newline */
 434                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 435                         continue;
 436                 }
 437                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 438                         c = 0x20;
 439                 strbuf_addch(out, c);
 440         }
 441         return out;
 442 }
 443
 444 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 445 {
 446         /* Decode in..ep, possibly in-place to ot */
 447         int c, pos = 0, acc = 0;
 448         const char *in = b_seg->buf;
 449         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 450         strbuf_init(out, b_seg->len);
 451
 452         while ((c = *in++) != 0) {
 453                 if (c == '+')
 454                         c = 62;
 455                 else if (c == '/')
 456                         c = 63;
 457                 else if ('A' <= c && c <= 'Z')
 458                         c -= 'A';
 459                 else if ('a' <= c && c <= 'z')
 460                         c -= 'a' - 26;
 461                 else if ('0' <= c && c <= '9')
 462                         c -= '0' - 52;
 463                 else
 464                         continue; /* garbage */
 465                 switch (pos++) {
 466                 case 0:
 467                         acc = (c << 2);
 468                         break;
 469                 case 1:
 470                         strbuf_addch(out, (acc | (c >> 4)));
 471                         acc = (c & 15) << 4;
 472                         break;
 473                 case 2:
 474                         strbuf_addch(out, (acc | (c >> 2)));
 475                         acc = (c & 3) << 6;
 476                         break;
 477                 case 3:
 478                         strbuf_addch(out, (acc | c));
 479                         acc = pos = 0;
 480                         break;
 481                 }
 482         }
 483         return out;
 484 }
 485
 486 /*
 487  * When there is no known charset, guess.
 488  *
 489  * Right now we assume that if the target is UTF-8 (the default),
 490  * and it already looks like UTF-8 (which includes US-ASCII as its
 491  * subset, of course) then that is what it is and there is nothing
 492  * to do.
 493  *
 494  * Otherwise, we default to assuming it is Latin1 for historical
 495  * reasons.
 496  */
 497 static const char *guess_charset(const struct strbuf *line, const char *target_charset)
 498 {
 499         if (is_encoding_utf8(target_charset)) {
 500                 if (is_utf8(line->buf))
 501                         return NULL;
 502         }
 503         return "ISO8859-1";
 504 }
 505
 506 static void convert_to_utf8(struct strbuf *line, const char *charset)
 507 {
 508         char *out;
 509
 510         if (!charset || !*charset) {
 511                 charset = guess_charset(line, metainfo_charset);
 512                 if (!charset)
 513                         return;
 514         }
 515
 516         if (!strcasecmp(metainfo_charset, charset))
 517                 return;
 518         out = reencode_string(line->buf, metainfo_charset, charset);
 519         if (!out)
 520                 die("cannot convert from %s to %s",
 521                     charset, metainfo_charset);
 522         strbuf_attach(line, out, strlen(out), strlen(out));
 523 }
 524
 525 static int decode_header_bq(struct strbuf *it)
 526 {
 527         char *in, *ep, *cp;
 528         struct strbuf outbuf = STRBUF_INIT, *dec;
 529         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 530         int rfc2047 = 0;
 531
 532         in = it->buf;
 533         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 534                 int encoding;
 535                 strbuf_reset(&charset_q);
 536                 strbuf_reset(&piecebuf);
 537                 rfc2047 = 1;
 538
 539                 if (in != ep) {
 540                         /*
 541                          * We are about to process an encoded-word
 542                          * that begins at ep, but there is something
 543                          * before the encoded word.
 544                          */
 545                         char *scan;
 546                         for (scan = in; scan < ep; scan++)
 547                                 if (!isspace(*scan))
 548                                         break;
 549
 550                         if (scan != ep || in == it->buf) {
 551                                 /*
 552                                  * We should not lose that "something",
 553                                  * unless we have just processed an
 554                                  * encoded-word, and there is only LWS
 555                                  * before the one we are about to process.
 556                                  */
 557                                 strbuf_add(&outbuf, in, ep - in);
 558                         }
 559                 }
 560                 /* E.g.
 561                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 562                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 563                  */
 564                 ep += 2;
 565
 566                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 567                         goto decode_header_bq_out;
 568
 569                 if (cp + 3 - it->buf > it->len)
 570                         goto decode_header_bq_out;
 571                 strbuf_add(&charset_q, ep, cp - ep);
 572
 573                 encoding = cp[1];
 574                 if (!encoding || cp[2] != '?')
 575                         goto decode_header_bq_out;
 576                 ep = strstr(cp + 3, "?=");
 577                 if (!ep)
 578                         goto decode_header_bq_out;
 579                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 580                 switch (tolower(encoding)) {
 581                 default:
 582                         goto decode_header_bq_out;
 583                 case 'b':
 584                         dec = decode_b_segment(&piecebuf);
 585                         break;
 586                 case 'q':
 587                         dec = decode_q_segment(&piecebuf, 1);
 588                         break;
 589                 }
 590                 if (metainfo_charset)
 591                         convert_to_utf8(dec, charset_q.buf);
 592
 593                 strbuf_addbuf(&outbuf, dec);
 594                 strbuf_release(dec);
 595                 free(dec);
 596                 in = ep + 2;
 597         }
 598         strbuf_addstr(&outbuf, in);
 599         strbuf_reset(it);
 600         strbuf_addbuf(it, &outbuf);
 601 decode_header_bq_out:
 602         strbuf_release(&outbuf);
 603         strbuf_release(&charset_q);
 604         strbuf_release(&piecebuf);
 605         return rfc2047;
 606 }
 607
 608 static void decode_header(struct strbuf *it)
 609 {
 610         if (decode_header_bq(it))
 611                 return;
 612         /* otherwise "it" is a straight copy of the input.
 613          * This can be binary guck but there is no charset specified.
 614          */
 615         if (metainfo_charset)
 616                 convert_to_utf8(it, "");
 617 }
 618
 619 static void decode_transfer_encoding(struct strbuf *line)
 620 {
 621         struct strbuf *ret;
 622
 623         switch (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 void handle_filter(struct strbuf *line);
 641
 642 static int find_boundary(void)
 643 {
 644         while (!strbuf_getline(&line, fin, '\n')) {
 645                 if (*content_top && is_multipart_boundary(&line))
 646                         return 1;
 647         }
 648         return 0;
 649 }
 650
 651 static int handle_boundary(void)
 652 {
 653         struct strbuf newline = STRBUF_INIT;
 654
 655         strbuf_addch(&newline, '\n');
 656 again:
 657         if (line.len >= (*content_top)->len + 2 &&
 658             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 659                 /* we hit an end boundary */
 660                 /* pop the current boundary off the stack */
 661                 strbuf_release(*content_top);
 662                 free(*content_top);
 663                 *content_top = NULL;
 664
 665                 /* technically won't happen as is_multipart_boundary()
 666                    will fail first.  But just in case..
 667                  */
 668                 if (--content_top < content) {
 669                         fprintf(stderr, "Detected mismatched boundaries, "
 670                                         "can't recover\n");
 671                         exit(1);
 672                 }
 673                 handle_filter(&newline);
 674                 strbuf_release(&newline);
 675
 676                 /* skip to the next boundary */
 677                 if (!find_boundary())
 678                         return 0;
 679                 goto again;
 680         }
 681
 682         /* set some defaults */
 683         transfer_encoding = TE_DONTCARE;
 684         strbuf_reset(&charset);
 685         message_type = TYPE_TEXT;
 686
 687         /* slurp in this section's info */
 688         while (read_one_header_line(&line, fin))
 689                 check_header(&line, p_hdr_data, 0);
 690
 691         strbuf_release(&newline);
 692         /* replenish line */
 693         if (strbuf_getline(&line, fin, '\n'))
 694                 return 0;
 695         strbuf_addch(&line, '\n');
 696         return 1;
 697 }
 698
 699 static inline int patchbreak(const struct strbuf *line)
 700 {
 701         size_t i;
 702
 703         /* Beginning of a "diff -" header? */
 704         if (!prefixcmp(line->buf, "diff -"))
 705                 return 1;
 706
 707         /* CVS "Index: " line? */
 708         if (!prefixcmp(line->buf, "Index: "))
 709                 return 1;
 710
 711         /*
 712          * "--- <filename>" starts patches without headers
 713          * "---<sp>*" is a manual separator
 714          */
 715         if (line->len < 4)
 716                 return 0;
 717
 718         if (!prefixcmp(line->buf, "---")) {
 719                 /* space followed by a filename? */
 720                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 721                         return 1;
 722                 /* Just whitespace? */
 723                 for (i = 3; i < line->len; i++) {
 724                         unsigned char c = line->buf[i];
 725                         if (c == '\n')
 726                                 return 1;
 727                         if (!isspace(c))
 728                                 break;
 729                 }
 730                 return 0;
 731         }
 732         return 0;
 733 }
 734
 735 static int is_scissors_line(const struct strbuf *line)
 736 {
 737         size_t i, len = line->len;
 738         int scissors = 0, gap = 0;
 739         int first_nonblank = -1;
 740         int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
 741         const char *buf = line->buf;
 742
 743         for (i = 0; i < len; i++) {
 744                 if (isspace(buf[i])) {
 745                         if (in_perforation) {
 746                                 perforation++;
 747                                 gap++;
 748                         }
 749                         continue;
 750                 }
 751                 last_nonblank = i;
 752                 if (first_nonblank < 0)
 753                         first_nonblank = i;
 754                 if (buf[i] == '-') {
 755                         in_perforation = 1;
 756                         perforation++;
 757                         continue;
 758                 }
 759                 if (i + 1 < len &&
 760                     (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
 761                      !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
 762                         in_perforation = 1;
 763                         perforation += 2;
 764                         scissors += 2;
 765                         i++;
 766                         continue;
 767                 }
 768                 in_perforation = 0;
 769         }
 770
 771         /*
 772          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 773          * Even though there can be arbitrary cruft on the same line
 774          * (e.g. "cut here"), in order to avoid misidentification, the
 775          * perforation must occupy more than a third of the visible
 776          * width of the line, and dashes and scissors must occupy more
 777          * than half of the perforation.
 778          */
 779
 780         visible = last_nonblank - first_nonblank + 1;
 781         return (scissors && 8 <= visible &&
 782                 visible < perforation * 3 &&
 783                 gap * 2 < perforation);
 784 }
 785
 786 static int handle_commit_msg(struct strbuf *line)
 787 {
 788         static int still_looking = 1;
 789
 790         if (!cmitmsg)
 791                 return 0;
 792
 793         if (still_looking) {
 794                 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
 795                         return 0;
 796         }
 797
 798         if (use_inbody_headers && still_looking) {
 799                 still_looking = check_header(line, s_hdr_data, 0);
 800                 if (still_looking)
 801                         return 0;
 802         } else
 803                 /* Only trim the first (blank) line of the commit message
 804                  * when ignoring in-body headers.
 805                  */
 806                 still_looking = 0;
 807
 808         /* normalize the log message to UTF-8. */
 809         if (metainfo_charset)
 810                 convert_to_utf8(line, charset.buf);
 811
 812         if (use_scissors && is_scissors_line(line)) {
 813                 int i;
 814                 if (fseek(cmitmsg, 0L, SEEK_SET))
 815                         die_errno("Could not rewind output message file");
 816                 if (ftruncate(fileno(cmitmsg), 0))
 817                         die_errno("Could not truncate output message file at scissors");
 818                 still_looking = 1;
 819
 820                 /*
 821                  * We may have already read "secondary headers"; purge
 822                  * them to give ourselves a clean restart.
 823                  */
 824                 for (i = 0; header[i]; i++) {
 825                         if (s_hdr_data[i])
 826                                 strbuf_release(s_hdr_data[i]);
 827                         s_hdr_data[i] = NULL;
 828                 }
 829                 return 0;
 830         }
 831
 832         if (patchbreak(line)) {
 833                 fclose(cmitmsg);
 834                 cmitmsg = NULL;
 835                 return 1;
 836         }
 837
 838         fputs(line->buf, cmitmsg);
 839         return 0;
 840 }
 841
 842 static void handle_patch(const struct strbuf *line)
 843 {
 844         fwrite(line->buf, 1, line->len, patchfile);
 845         patch_lines++;
 846 }
 847
 848 static void handle_filter(struct strbuf *line)
 849 {
 850         static int filter = 0;
 851
 852         /* filter tells us which part we left off on */
 853         switch (filter) {
 854         case 0:
 855                 if (!handle_commit_msg(line))
 856                         break;
 857                 filter++;
 858         case 1:
 859                 handle_patch(line);
 860                 break;
 861         }
 862 }
 863
 864 static void handle_body(void)
 865 {
 866         struct strbuf prev = STRBUF_INIT;
 867
 868         /* Skip up to the first boundary */
 869         if (*content_top) {
 870                 if (!find_boundary())
 871                         goto handle_body_out;
 872         }
 873
 874         do {
 875                 /* process any boundary lines */
 876                 if (*content_top && is_multipart_boundary(&line)) {
 877                         /* flush any leftover */
 878                         if (prev.len) {
 879                                 handle_filter(&prev);
 880                                 strbuf_reset(&prev);
 881                         }
 882                         if (!handle_boundary())
 883                                 goto handle_body_out;
 884                 }
 885
 886                 /* Unwrap transfer encoding */
 887                 decode_transfer_encoding(&line);
 888
 889                 switch (transfer_encoding) {
 890                 case TE_BASE64:
 891                 case TE_QP:
 892                 {
 893                         struct strbuf **lines, **it, *sb;
 894
 895                         /* Prepend any previous partial lines */
 896                         strbuf_insert(&line, 0, prev.buf, prev.len);
 897                         strbuf_reset(&prev);
 898
 899                         /* binary data most likely doesn't have newlines */
 900                         if (message_type != TYPE_TEXT) {
 901                                 handle_filter(&line);
 902                                 break;
 903                         }
 904                         /*
 905                          * This is a decoded line that may contain
 906                          * multiple new lines.  Pass only one chunk
 907                          * at a time to handle_filter()
 908                          */
 909                         lines = strbuf_split(&line, '\n');
 910                         for (it = lines; (sb = *it); it++) {
 911                                 if (*(it + 1) == NULL) /* The last line */
 912                                         if (sb->buf[sb->len - 1] != '\n') {
 913                                                 /* Partial line, save it for later. */
 914                                                 strbuf_addbuf(&prev, sb);
 915                                                 break;
 916                                         }
 917                                 handle_filter(sb);
 918                         }
 919                         /*
 920                          * The partial chunk is saved in "prev" and will be
 921                          * appended by the next iteration of read_line_with_nul().
 922                          */
 923                         strbuf_list_free(lines);
 924                         break;
 925                 }
 926                 default:
 927                         handle_filter(&line);
 928                 }
 929
 930         } while (!strbuf_getwholeline(&line, fin, '\n'));
 931
 932 handle_body_out:
 933         strbuf_release(&prev);
 934 }
 935
 936 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 937 {
 938         const char *sp = data->buf;
 939         while (1) {
 940                 char *ep = strchr(sp, '\n');
 941                 int len;
 942                 if (!ep)
 943                         len = strlen(sp);
 944                 else
 945                         len = ep - sp;
 946                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 947                 if (!ep)
 948                         break;
 949                 sp = ep + 1;
 950         }
 951 }
 952
 953 static void handle_info(void)
 954 {
 955         struct strbuf *hdr;
 956         int i;
 957
 958         for (i = 0; header[i]; i++) {
 959                 /* only print inbody headers if we output a patch file */
 960                 if (patch_lines && s_hdr_data[i])
 961                         hdr = s_hdr_data[i];
 962                 else if (p_hdr_data[i])
 963                         hdr = p_hdr_data[i];
 964                 else
 965                         continue;
 966
 967                 if (!memcmp(header[i], "Subject", 7)) {
 968                         if (!keep_subject) {
 969                                 cleanup_subject(hdr);
 970                                 cleanup_space(hdr);
 971                         }
 972                         output_header_lines(fout, "Subject", hdr);
 973                 } else if (!memcmp(header[i], "From", 4)) {
 974                         cleanup_space(hdr);
 975                         handle_from(hdr);
 976                         fprintf(fout, "Author: %s\n", name.buf);
 977                         fprintf(fout, "Email: %s\n", email.buf);
 978                 } else {
 979                         cleanup_space(hdr);
 980                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 981                 }
 982         }
 983         fprintf(fout, "\n");
 984 }
 985
 986 static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
 987 {
 988         int peek;
 989         fin = in;
 990         fout = out;
 991
 992         cmitmsg = fopen(msg, "w");
 993         if (!cmitmsg) {
 994                 perror(msg);
 995                 return -1;
 996         }
 997         patchfile = fopen(patch, "w");
 998         if (!patchfile) {
 999                 perror(patch);
1000                 fclose(cmitmsg);
1001                 return -1;
1002         }
1003
1004         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
1005         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
1006
1007         do {
1008                 peek = fgetc(in);
1009         } while (isspace(peek));
1010         ungetc(peek, in);
1011
1012         /* process the email header */
1013         while (read_one_header_line(&line, fin))
1014                 check_header(&line, p_hdr_data, 1);
1015
1016         handle_body();
1017         handle_info();
1018
1019         return 0;
1020 }
1021
1022 static int git_mailinfo_config(const char *var, const char *value, void *unused)
1023 {
1024         if (prefixcmp(var, "mailinfo."))
1025                 return git_default_config(var, value, unused);
1026         if (!strcmp(var, "mailinfo.scissors")) {
1027                 use_scissors = git_config_bool(var, value);
1028                 return 0;
1029         }
1030         /* perhaps others here */
1031         return 0;
1032 }
1033
1034 static const char mailinfo_usage[] =
1035         "git mailinfo [-k|-b] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] msg patch < mail >info";
1036
1037 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1038 {
1039         const char *def_charset;
1040
1041         /* NEEDSWORK: might want to do the optional .git/ directory
1042          * discovery
1043          */
1044         git_config(git_mailinfo_config, NULL);
1045
1046         def_charset = get_commit_output_encoding();
1047         metainfo_charset = def_charset;
1048
1049         while (1 < argc && argv[1][0] == '-') {
1050                 if (!strcmp(argv[1], "-k"))
1051                         keep_subject = 1;
1052                 else if (!strcmp(argv[1], "-b"))
1053                         keep_non_patch_brackets_in_subject = 1;
1054                 else if (!strcmp(argv[1], "-u"))
1055                         metainfo_charset = def_charset;
1056                 else if (!strcmp(argv[1], "-n"))
1057                         metainfo_charset = NULL;
1058                 else if (!prefixcmp(argv[1], "--encoding="))
1059                         metainfo_charset = argv[1] + 11;
1060                 else if (!strcmp(argv[1], "--scissors"))
1061                         use_scissors = 1;
1062                 else if (!strcmp(argv[1], "--no-scissors"))
1063                         use_scissors = 0;
1064                 else if (!strcmp(argv[1], "--no-inbody-headers"))
1065                         use_inbody_headers = 0;
1066                 else
1067                         usage(mailinfo_usage);
1068                 argc--; argv++;
1069         }
1070
1071         if (argc != 3)
1072                 usage(mailinfo_usage);
1073
1074         return !!mailinfo(stdin, stdout, argv[1], argv[2]);
1075 }