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 const char *metainfo_charset;
  14 static struct strbuf line = STRBUF_INIT;
  15 static struct strbuf name = STRBUF_INIT;
  16 static struct strbuf email = STRBUF_INIT;
  17
  18 static enum  {
  19         TE_DONTCARE, TE_QP, TE_BASE64,
  20 } transfer_encoding;
  21 static enum  {
  22         TYPE_TEXT, TYPE_OTHER,
  23 } message_type;
  24
  25 static struct strbuf charset = STRBUF_INIT;
  26 static int patch_lines;
  27 static struct strbuf **p_hdr_data, **s_hdr_data;
  28 static int use_scissors;
  29
  30 #define MAX_HDR_PARSED 10
  31 #define MAX_BOUNDARIES 5
  32
  33 static void cleanup_space(struct strbuf *sb);
  34
  35
  36 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  37 {
  38         struct strbuf *src = name;
  39         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  40                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  41                 src = email;
  42         else if (name == out)
  43                 return;
  44         strbuf_reset(out);
  45         strbuf_addbuf(out, src);
  46 }
  47
  48 static void parse_bogus_from(const struct strbuf *line)
  49 {
  50         /* John Doe <johndoe> */
  51
  52         char *bra, *ket;
  53         /* This is fallback, so do not bother if we already have an
  54          * e-mail address.
  55          */
  56         if (email.len)
  57                 return;
  58
  59         bra = strchr(line->buf, '<');
  60         if (!bra)
  61                 return;
  62         ket = strchr(bra, '>');
  63         if (!ket)
  64                 return;
  65
  66         strbuf_reset(&email);
  67         strbuf_add(&email, bra + 1, ket - bra - 1);
  68
  69         strbuf_reset(&name);
  70         strbuf_add(&name, line->buf, bra - line->buf);
  71         strbuf_trim(&name);
  72         get_sane_name(&name, &name, &email);
  73 }
  74
  75 static void handle_from(const struct strbuf *from)
  76 {
  77         char *at;
  78         size_t el;
  79         struct strbuf f;
  80
  81         strbuf_init(&f, from->len);
  82         strbuf_addbuf(&f, from);
  83
  84         at = strchr(f.buf, '@');
  85         if (!at) {
  86                 parse_bogus_from(from);
  87                 return;
  88         }
  89
  90         /*
  91          * If we already have one email, don't take any confusing lines
  92          */
  93         if (email.len && strchr(at + 1, '@')) {
  94                 strbuf_release(&f);
  95                 return;
  96         }
  97
  98         /* Pick up the string around '@', possibly delimited with <>
  99          * pair; that is the email part.
 100          */
 101         while (at > f.buf) {
 102                 char c = at[-1];
 103                 if (isspace(c))
 104                         break;
 105                 if (c == '<') {
 106                         at[-1] = ' ';
 107                         break;
 108                 }
 109                 at--;
 110         }
 111         el = strcspn(at, " \n\t\r\v\f>");
 112         strbuf_reset(&email);
 113         strbuf_add(&email, at, el);
 114         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 115
 116         /* The remainder is name.  It could be
 117          *
 118          * - "John Doe <john.doe@xz>"                   (a), or
 119          * - "john.doe@xz (John Doe)"                   (b), or
 120          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 121          *
 122          * but we have removed the email part, so
 123          *
 124          * - remove extra spaces which could stay after email (case 'c'), and
 125          * - trim from both ends, possibly removing the () pair at the end
 126          *   (cases 'a' and 'b').
 127          */
 128         cleanup_space(&f);
 129         strbuf_trim(&f);
 130         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 131                 strbuf_remove(&f, 0, 1);
 132                 strbuf_setlen(&f, f.len - 1);
 133         }
 134
 135         get_sane_name(&name, &f, &email);
 136         strbuf_release(&f);
 137 }
 138
 139 static void handle_header(struct strbuf **out, const struct strbuf *line)
 140 {
 141         if (!*out) {
 142                 *out = xmalloc(sizeof(struct strbuf));
 143                 strbuf_init(*out, line->len);
 144         } else
 145                 strbuf_reset(*out);
 146
 147         strbuf_addbuf(*out, line);
 148 }
 149
 150 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 151  * to have enough heuristics to grok MIME encoded patches often found
 152  * on our mailing lists.  For example, we do not even treat header lines
 153  * case insensitively.
 154  */
 155
 156 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 157 {
 158         const char *ends, *ap = strcasestr(line, name);
 159         size_t sz;
 160
 161         if (!ap) {
 162                 strbuf_setlen(attr, 0);
 163                 return 0;
 164         }
 165         ap += strlen(name);
 166         if (*ap == '"') {
 167                 ap++;
 168                 ends = "\"";
 169         }
 170         else
 171                 ends = "; \t";
 172         sz = strcspn(ap, ends);
 173         strbuf_add(attr, ap, sz);
 174         return 1;
 175 }
 176
 177 static struct strbuf *content[MAX_BOUNDARIES];
 178
 179 static struct strbuf **content_top = content;
 180
 181 static void handle_content_type(struct strbuf *line)
 182 {
 183         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 184         strbuf_init(boundary, line->len);
 185
 186         if (!strcasestr(line->buf, "text/"))
 187                  message_type = TYPE_OTHER;
 188         if (slurp_attr(line->buf, "boundary=", boundary)) {
 189                 strbuf_insert(boundary, 0, "--", 2);
 190                 if (++content_top > &content[MAX_BOUNDARIES]) {
 191                         fprintf(stderr, "Too many boundaries to handle\n");
 192                         exit(1);
 193                 }
 194                 *content_top = boundary;
 195                 boundary = NULL;
 196         }
 197         slurp_attr(line->buf, "charset=", &charset);
 198
 199         if (boundary) {
 200                 strbuf_release(boundary);
 201                 free(boundary);
 202         }
 203 }
 204
 205 static void handle_content_transfer_encoding(const struct strbuf *line)
 206 {
 207         if (strcasestr(line->buf, "base64"))
 208                 transfer_encoding = TE_BASE64;
 209         else if (strcasestr(line->buf, "quoted-printable"))
 210                 transfer_encoding = TE_QP;
 211         else
 212                 transfer_encoding = TE_DONTCARE;
 213 }
 214
 215 static int is_multipart_boundary(const struct strbuf *line)
 216 {
 217         return (((*content_top)->len <= line->len) &&
 218                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
 219 }
 220
 221 static void cleanup_subject(struct strbuf *subject)
 222 {
 223         char *pos;
 224         size_t remove;
 225         while (subject->len) {
 226                 switch (*subject->buf) {
 227                 case 'r': case 'R':
 228                         if (subject->len <= 3)
 229                                 break;
 230                         if (!memcmp(subject->buf + 1, "e:", 2)) {
 231                                 strbuf_remove(subject, 0, 3);
 232                                 continue;
 233                         }
 234                         break;
 235                 case ' ': case '\t': case ':':
 236                         strbuf_remove(subject, 0, 1);
 237                         continue;
 238                 case '[':
 239                         if ((pos = strchr(subject->buf, ']'))) {
 240                                 remove = pos - subject->buf;
 241                                 if (remove <= (subject->len - remove) * 2) {
 242                                         strbuf_remove(subject, 0, remove + 1);
 243                                         continue;
 244                                 }
 245                         } else
 246                                 strbuf_remove(subject, 0, 1);
 247                         break;
 248                 }
 249                 strbuf_trim(subject);
 250                 return;
 251         }
 252 }
 253
 254 static void cleanup_space(struct strbuf *sb)
 255 {
 256         size_t pos, cnt;
 257         for (pos = 0; pos < sb->len; pos++) {
 258                 if (isspace(sb->buf[pos])) {
 259                         sb->buf[pos] = ' ';
 260                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 261                         strbuf_remove(sb, pos + 1, cnt);
 262                 }
 263         }
 264 }
 265
 266 static void decode_header(struct strbuf *line);
 267 static const char *header[MAX_HDR_PARSED] = {
 268         "From","Subject","Date",
 269 };
 270
 271 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 272 {
 273         int len = strlen(hdr);
 274         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 275                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 276 }
 277
 278 static int check_header(const struct strbuf *line,
 279                                 struct strbuf *hdr_data[], int overwrite)
 280 {
 281         int i, ret = 0, len;
 282         struct strbuf sb = STRBUF_INIT;
 283         /* search for the interesting parts */
 284         for (i = 0; header[i]; i++) {
 285                 int len = strlen(header[i]);
 286                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 287                         /* Unwrap inline B and Q encoding, and optionally
 288                          * normalize the meta information to utf8.
 289                          */
 290                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 291                         decode_header(&sb);
 292                         handle_header(&hdr_data[i], &sb);
 293                         ret = 1;
 294                         goto check_header_out;
 295                 }
 296         }
 297
 298         /* Content stuff */
 299         if (cmp_header(line, "Content-Type")) {
 300                 len = strlen("Content-Type: ");
 301                 strbuf_add(&sb, line->buf + len, line->len - len);
 302                 decode_header(&sb);
 303                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 304                 handle_content_type(&sb);
 305                 ret = 1;
 306                 goto check_header_out;
 307         }
 308         if (cmp_header(line, "Content-Transfer-Encoding")) {
 309                 len = strlen("Content-Transfer-Encoding: ");
 310                 strbuf_add(&sb, line->buf + len, line->len - len);
 311                 decode_header(&sb);
 312                 handle_content_transfer_encoding(&sb);
 313                 ret = 1;
 314                 goto check_header_out;
 315         }
 316
 317         /* for inbody stuff */
 318         if (!prefixcmp(line->buf, ">From") && isspace(line->buf[5])) {
 319                 ret = 1; /* Should this return 0? */
 320                 goto check_header_out;
 321         }
 322         if (!prefixcmp(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 323                 for (i = 0; header[i]; i++) {
 324                         if (!memcmp("Subject", header[i], 7)) {
 325                                 handle_header(&hdr_data[i], line);
 326                                 ret = 1;
 327                                 goto check_header_out;
 328                         }
 329                 }
 330         }
 331
 332 check_header_out:
 333         strbuf_release(&sb);
 334         return ret;
 335 }
 336
 337 static int is_rfc2822_header(const struct strbuf *line)
 338 {
 339         /*
 340          * The section that defines the loosest possible
 341          * field name is "3.6.8 Optional fields".
 342          *
 343          * optional-field = field-name ":" unstructured CRLF
 344          * field-name = 1*ftext
 345          * ftext = %d33-57 / %59-126
 346          */
 347         int ch;
 348         char *cp = line->buf;
 349
 350         /* Count mbox From headers as headers */
 351         if (!prefixcmp(cp, "From ") || !prefixcmp(cp, ">From "))
 352                 return 1;
 353
 354         while ((ch = *cp++)) {
 355                 if (ch == ':')
 356                         return 1;
 357                 if ((33 <= ch && ch <= 57) ||
 358                     (59 <= ch && ch <= 126))
 359                         continue;
 360                 break;
 361         }
 362         return 0;
 363 }
 364
 365 static int read_one_header_line(struct strbuf *line, FILE *in)
 366 {
 367         /* Get the first part of the line. */
 368         if (strbuf_getline(line, in, '\n'))
 369                 return 0;
 370
 371         /*
 372          * Is it an empty line or not a valid rfc2822 header?
 373          * If so, stop here, and return false ("not a header")
 374          */
 375         strbuf_rtrim(line);
 376         if (!line->len || !is_rfc2822_header(line)) {
 377                 /* Re-add the newline */
 378                 strbuf_addch(line, '\n');
 379                 return 0;
 380         }
 381
 382         /*
 383          * Now we need to eat all the continuation lines..
 384          * Yuck, 2822 header "folding"
 385          */
 386         for (;;) {
 387                 int peek;
 388                 struct strbuf continuation = STRBUF_INIT;
 389
 390                 peek = fgetc(in); ungetc(peek, in);
 391                 if (peek != ' ' && peek != '\t')
 392                         break;
 393                 if (strbuf_getline(&continuation, in, '\n'))
 394                         break;
 395                 continuation.buf[0] = '\n';
 396                 strbuf_rtrim(&continuation);
 397                 strbuf_addbuf(line, &continuation);
 398         }
 399
 400         return 1;
 401 }
 402
 403 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 404 {
 405         const char *in = q_seg->buf;
 406         int c;
 407         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 408         strbuf_init(out, q_seg->len);
 409
 410         while ((c = *in++) != 0) {
 411                 if (c == '=') {
 412                         int d = *in++;
 413                         if (d == '\n' || !d)
 414                                 break; /* drop trailing newline */
 415                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 416                         continue;
 417                 }
 418                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 419                         c = 0x20;
 420                 strbuf_addch(out, c);
 421         }
 422         return out;
 423 }
 424
 425 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 426 {
 427         /* Decode in..ep, possibly in-place to ot */
 428         int c, pos = 0, acc = 0;
 429         const char *in = b_seg->buf;
 430         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 431         strbuf_init(out, b_seg->len);
 432
 433         while ((c = *in++) != 0) {
 434                 if (c == '+')
 435                         c = 62;
 436                 else if (c == '/')
 437                         c = 63;
 438                 else if ('A' <= c && c <= 'Z')
 439                         c -= 'A';
 440                 else if ('a' <= c && c <= 'z')
 441                         c -= 'a' - 26;
 442                 else if ('0' <= c && c <= '9')
 443                         c -= '0' - 52;
 444                 else
 445                         continue; /* garbage */
 446                 switch (pos++) {
 447                 case 0:
 448                         acc = (c << 2);
 449                         break;
 450                 case 1:
 451                         strbuf_addch(out, (acc | (c >> 4)));
 452                         acc = (c & 15) << 4;
 453                         break;
 454                 case 2:
 455                         strbuf_addch(out, (acc | (c >> 2)));
 456                         acc = (c & 3) << 6;
 457                         break;
 458                 case 3:
 459                         strbuf_addch(out, (acc | c));
 460                         acc = pos = 0;
 461                         break;
 462                 }
 463         }
 464         return out;
 465 }
 466
 467 /*
 468  * When there is no known charset, guess.
 469  *
 470  * Right now we assume that if the target is UTF-8 (the default),
 471  * and it already looks like UTF-8 (which includes US-ASCII as its
 472  * subset, of course) then that is what it is and there is nothing
 473  * to do.
 474  *
 475  * Otherwise, we default to assuming it is Latin1 for historical
 476  * reasons.
 477  */
 478 static const char *guess_charset(const struct strbuf *line, const char *target_charset)
 479 {
 480         if (is_encoding_utf8(target_charset)) {
 481                 if (is_utf8(line->buf))
 482                         return NULL;
 483         }
 484         return "ISO8859-1";
 485 }
 486
 487 static void convert_to_utf8(struct strbuf *line, const char *charset)
 488 {
 489         char *out;
 490
 491         if (!charset || !*charset) {
 492                 charset = guess_charset(line, metainfo_charset);
 493                 if (!charset)
 494                         return;
 495         }
 496
 497         if (!strcasecmp(metainfo_charset, charset))
 498                 return;
 499         out = reencode_string(line->buf, metainfo_charset, charset);
 500         if (!out)
 501                 die("cannot convert from %s to %s",
 502                     charset, metainfo_charset);
 503         strbuf_attach(line, out, strlen(out), strlen(out));
 504 }
 505
 506 static int decode_header_bq(struct strbuf *it)
 507 {
 508         char *in, *ep, *cp;
 509         struct strbuf outbuf = STRBUF_INIT, *dec;
 510         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 511         int rfc2047 = 0;
 512
 513         in = it->buf;
 514         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 515                 int encoding;
 516                 strbuf_reset(&charset_q);
 517                 strbuf_reset(&piecebuf);
 518                 rfc2047 = 1;
 519
 520                 if (in != ep) {
 521                         /*
 522                          * We are about to process an encoded-word
 523                          * that begins at ep, but there is something
 524                          * before the encoded word.
 525                          */
 526                         char *scan;
 527                         for (scan = in; scan < ep; scan++)
 528                                 if (!isspace(*scan))
 529                                         break;
 530
 531                         if (scan != ep || in == it->buf) {
 532                                 /*
 533                                  * We should not lose that "something",
 534                                  * unless we have just processed an
 535                                  * encoded-word, and there is only LWS
 536                                  * before the one we are about to process.
 537                                  */
 538                                 strbuf_add(&outbuf, in, ep - in);
 539                         }
 540                 }
 541                 /* E.g.
 542                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 543                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 544                  */
 545                 ep += 2;
 546
 547                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 548                         goto decode_header_bq_out;
 549
 550                 if (cp + 3 - it->buf > it->len)
 551                         goto decode_header_bq_out;
 552                 strbuf_add(&charset_q, ep, cp - ep);
 553
 554                 encoding = cp[1];
 555                 if (!encoding || cp[2] != '?')
 556                         goto decode_header_bq_out;
 557                 ep = strstr(cp + 3, "?=");
 558                 if (!ep)
 559                         goto decode_header_bq_out;
 560                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 561                 switch (tolower(encoding)) {
 562                 default:
 563                         goto decode_header_bq_out;
 564                 case 'b':
 565                         dec = decode_b_segment(&piecebuf);
 566                         break;
 567                 case 'q':
 568                         dec = decode_q_segment(&piecebuf, 1);
 569                         break;
 570                 }
 571                 if (metainfo_charset)
 572                         convert_to_utf8(dec, charset_q.buf);
 573
 574                 strbuf_addbuf(&outbuf, dec);
 575                 strbuf_release(dec);
 576                 free(dec);
 577                 in = ep + 2;
 578         }
 579         strbuf_addstr(&outbuf, in);
 580         strbuf_reset(it);
 581         strbuf_addbuf(it, &outbuf);
 582 decode_header_bq_out:
 583         strbuf_release(&outbuf);
 584         strbuf_release(&charset_q);
 585         strbuf_release(&piecebuf);
 586         return rfc2047;
 587 }
 588
 589 static void decode_header(struct strbuf *it)
 590 {
 591         if (decode_header_bq(it))
 592                 return;
 593         /* otherwise "it" is a straight copy of the input.
 594          * This can be binary guck but there is no charset specified.
 595          */
 596         if (metainfo_charset)
 597                 convert_to_utf8(it, "");
 598 }
 599
 600 static void decode_transfer_encoding(struct strbuf *line)
 601 {
 602         struct strbuf *ret;
 603
 604         switch (transfer_encoding) {
 605         case TE_QP:
 606                 ret = decode_q_segment(line, 0);
 607                 break;
 608         case TE_BASE64:
 609                 ret = decode_b_segment(line);
 610                 break;
 611         case TE_DONTCARE:
 612         default:
 613                 return;
 614         }
 615         strbuf_reset(line);
 616         strbuf_addbuf(line, ret);
 617         strbuf_release(ret);
 618         free(ret);
 619 }
 620
 621 static void handle_filter(struct strbuf *line);
 622
 623 static int find_boundary(void)
 624 {
 625         while (!strbuf_getline(&line, fin, '\n')) {
 626                 if (*content_top && is_multipart_boundary(&line))
 627                         return 1;
 628         }
 629         return 0;
 630 }
 631
 632 static int handle_boundary(void)
 633 {
 634         struct strbuf newline = STRBUF_INIT;
 635
 636         strbuf_addch(&newline, '\n');
 637 again:
 638         if (line.len >= (*content_top)->len + 2 &&
 639             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 640                 /* we hit an end boundary */
 641                 /* pop the current boundary off the stack */
 642                 strbuf_release(*content_top);
 643                 free(*content_top);
 644                 *content_top = NULL;
 645
 646                 /* technically won't happen as is_multipart_boundary()
 647                    will fail first.  But just in case..
 648                  */
 649                 if (--content_top < content) {
 650                         fprintf(stderr, "Detected mismatched boundaries, "
 651                                         "can't recover\n");
 652                         exit(1);
 653                 }
 654                 handle_filter(&newline);
 655                 strbuf_release(&newline);
 656
 657                 /* skip to the next boundary */
 658                 if (!find_boundary())
 659                         return 0;
 660                 goto again;
 661         }
 662
 663         /* set some defaults */
 664         transfer_encoding = TE_DONTCARE;
 665         strbuf_reset(&charset);
 666         message_type = TYPE_TEXT;
 667
 668         /* slurp in this section's info */
 669         while (read_one_header_line(&line, fin))
 670                 check_header(&line, p_hdr_data, 0);
 671
 672         strbuf_release(&newline);
 673         /* replenish line */
 674         if (strbuf_getline(&line, fin, '\n'))
 675                 return 0;
 676         strbuf_addch(&line, '\n');
 677         return 1;
 678 }
 679
 680 static inline int patchbreak(const struct strbuf *line)
 681 {
 682         size_t i;
 683
 684         /* Beginning of a "diff -" header? */
 685         if (!prefixcmp(line->buf, "diff -"))
 686                 return 1;
 687
 688         /* CVS "Index: " line? */
 689         if (!prefixcmp(line->buf, "Index: "))
 690                 return 1;
 691
 692         /*
 693          * "--- <filename>" starts patches without headers
 694          * "---<sp>*" is a manual separator
 695          */
 696         if (line->len < 4)
 697                 return 0;
 698
 699         if (!prefixcmp(line->buf, "---")) {
 700                 /* space followed by a filename? */
 701                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 702                         return 1;
 703                 /* Just whitespace? */
 704                 for (i = 3; i < line->len; i++) {
 705                         unsigned char c = line->buf[i];
 706                         if (c == '\n')
 707                                 return 1;
 708                         if (!isspace(c))
 709                                 break;
 710                 }
 711                 return 0;
 712         }
 713         return 0;
 714 }
 715
 716 static int is_scissors_line(const struct strbuf *line)
 717 {
 718         size_t i, len = line->len;
 719         int scissors = 0, gap = 0;
 720         int first_nonblank = -1;
 721         int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
 722         const char *buf = line->buf;
 723
 724         for (i = 0; i < len; i++) {
 725                 if (isspace(buf[i])) {
 726                         if (in_perforation) {
 727                                 perforation++;
 728                                 gap++;
 729                         }
 730                         continue;
 731                 }
 732                 last_nonblank = i;
 733                 if (first_nonblank < 0)
 734                         first_nonblank = i;
 735                 if (buf[i] == '-') {
 736                         in_perforation = 1;
 737                         perforation++;
 738                         continue;
 739                 }
 740                 if (i + 1 < len &&
 741                     (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2))) {
 742                         in_perforation = 1;
 743                         perforation += 2;
 744                         scissors += 2;
 745                         i++;
 746                         continue;
 747                 }
 748                 in_perforation = 0;
 749         }
 750
 751         /*
 752          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 753          * Even though there can be arbitrary cruft on the same line
 754          * (e.g. "cut here"), in order to avoid misidentification, the
 755          * perforation must occupy more than a third of the visible
 756          * width of the line, and dashes and scissors must occupy more
 757          * than half of the perforation.
 758          */
 759
 760         visible = last_nonblank - first_nonblank + 1;
 761         return (scissors && 8 <= visible &&
 762                 visible < perforation * 3 &&
 763                 gap * 2 < perforation);
 764 }
 765
 766 static int handle_commit_msg(struct strbuf *line)
 767 {
 768         static int still_looking = 1;
 769
 770         if (!cmitmsg)
 771                 return 0;
 772
 773         if (still_looking) {
 774                 strbuf_ltrim(line);
 775                 if (!line->len)
 776                         return 0;
 777                 still_looking = check_header(line, s_hdr_data, 0);
 778                 if (still_looking)
 779                         return 0;
 780         }
 781
 782         /* normalize the log message to UTF-8. */
 783         if (metainfo_charset)
 784                 convert_to_utf8(line, charset.buf);
 785
 786         if (use_scissors && is_scissors_line(line)) {
 787                 int i;
 788                 if (fseek(cmitmsg, 0L, SEEK_SET))
 789                         die_errno("Could not rewind output message file");
 790                 if (ftruncate(fileno(cmitmsg), 0))
 791                         die_errno("Could not truncate output message file at scissors");
 792                 still_looking = 1;
 793
 794                 /*
 795                  * We may have already read "secondary headers"; purge
 796                  * them to give ourselves a clean restart.
 797                  */
 798                 for (i = 0; header[i]; i++) {
 799                         if (s_hdr_data[i])
 800                                 strbuf_release(s_hdr_data[i]);
 801                         s_hdr_data[i] = NULL;
 802                 }
 803                 return 0;
 804         }
 805
 806         if (patchbreak(line)) {
 807                 fclose(cmitmsg);
 808                 cmitmsg = NULL;
 809                 return 1;
 810         }
 811
 812         fputs(line->buf, cmitmsg);
 813         return 0;
 814 }
 815
 816 static void handle_patch(const struct strbuf *line)
 817 {
 818         fwrite(line->buf, 1, line->len, patchfile);
 819         patch_lines++;
 820 }
 821
 822 static void handle_filter(struct strbuf *line)
 823 {
 824         static int filter = 0;
 825
 826         /* filter tells us which part we left off on */
 827         switch (filter) {
 828         case 0:
 829                 if (!handle_commit_msg(line))
 830                         break;
 831                 filter++;
 832         case 1:
 833                 handle_patch(line);
 834                 break;
 835         }
 836 }
 837
 838 static void handle_body(void)
 839 {
 840         struct strbuf prev = STRBUF_INIT;
 841
 842         /* Skip up to the first boundary */
 843         if (*content_top) {
 844                 if (!find_boundary())
 845                         goto handle_body_out;
 846         }
 847
 848         do {
 849                 /* process any boundary lines */
 850                 if (*content_top && is_multipart_boundary(&line)) {
 851                         /* flush any leftover */
 852                         if (prev.len) {
 853                                 handle_filter(&prev);
 854                                 strbuf_reset(&prev);
 855                         }
 856                         if (!handle_boundary())
 857                                 goto handle_body_out;
 858                 }
 859
 860                 /* Unwrap transfer encoding */
 861                 decode_transfer_encoding(&line);
 862
 863                 switch (transfer_encoding) {
 864                 case TE_BASE64:
 865                 case TE_QP:
 866                 {
 867                         struct strbuf **lines, **it, *sb;
 868
 869                         /* Prepend any previous partial lines */
 870                         strbuf_insert(&line, 0, prev.buf, prev.len);
 871                         strbuf_reset(&prev);
 872
 873                         /* binary data most likely doesn't have newlines */
 874                         if (message_type != TYPE_TEXT) {
 875                                 handle_filter(&line);
 876                                 break;
 877                         }
 878                         /*
 879                          * This is a decoded line that may contain
 880                          * multiple new lines.  Pass only one chunk
 881                          * at a time to handle_filter()
 882                          */
 883                         lines = strbuf_split(&line, '\n');
 884                         for (it = lines; (sb = *it); it++) {
 885                                 if (*(it + 1) == NULL) /* The last line */
 886                                         if (sb->buf[sb->len - 1] != '\n') {
 887                                                 /* Partial line, save it for later. */
 888                                                 strbuf_addbuf(&prev, sb);
 889                                                 break;
 890                                         }
 891                                 handle_filter(sb);
 892                         }
 893                         /*
 894                          * The partial chunk is saved in "prev" and will be
 895                          * appended by the next iteration of read_line_with_nul().
 896                          */
 897                         strbuf_list_free(lines);
 898                         break;
 899                 }
 900                 default:
 901                         handle_filter(&line);
 902                 }
 903
 904         } while (!strbuf_getwholeline(&line, fin, '\n'));
 905
 906 handle_body_out:
 907         strbuf_release(&prev);
 908 }
 909
 910 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 911 {
 912         const char *sp = data->buf;
 913         while (1) {
 914                 char *ep = strchr(sp, '\n');
 915                 int len;
 916                 if (!ep)
 917                         len = strlen(sp);
 918                 else
 919                         len = ep - sp;
 920                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 921                 if (!ep)
 922                         break;
 923                 sp = ep + 1;
 924         }
 925 }
 926
 927 static void handle_info(void)
 928 {
 929         struct strbuf *hdr;
 930         int i;
 931
 932         for (i = 0; header[i]; i++) {
 933                 /* only print inbody headers if we output a patch file */
 934                 if (patch_lines && s_hdr_data[i])
 935                         hdr = s_hdr_data[i];
 936                 else if (p_hdr_data[i])
 937                         hdr = p_hdr_data[i];
 938                 else
 939                         continue;
 940
 941                 if (!memcmp(header[i], "Subject", 7)) {
 942                         if (!keep_subject) {
 943                                 cleanup_subject(hdr);
 944                                 cleanup_space(hdr);
 945                         }
 946                         output_header_lines(fout, "Subject", hdr);
 947                 } else if (!memcmp(header[i], "From", 4)) {
 948                         cleanup_space(hdr);
 949                         handle_from(hdr);
 950                         fprintf(fout, "Author: %s\n", name.buf);
 951                         fprintf(fout, "Email: %s\n", email.buf);
 952                 } else {
 953                         cleanup_space(hdr);
 954                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 955                 }
 956         }
 957         fprintf(fout, "\n");
 958 }
 959
 960 static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
 961 {
 962         int peek;
 963         fin = in;
 964         fout = out;
 965
 966         cmitmsg = fopen(msg, "w");
 967         if (!cmitmsg) {
 968                 perror(msg);
 969                 return -1;
 970         }
 971         patchfile = fopen(patch, "w");
 972         if (!patchfile) {
 973                 perror(patch);
 974                 fclose(cmitmsg);
 975                 return -1;
 976         }
 977
 978         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
 979         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
 980
 981         do {
 982                 peek = fgetc(in);
 983         } while (isspace(peek));
 984         ungetc(peek, in);
 985
 986         /* process the email header */
 987         while (read_one_header_line(&line, fin))
 988                 check_header(&line, p_hdr_data, 1);
 989
 990         handle_body();
 991         handle_info();
 992
 993         return 0;
 994 }
 995
 996 static int git_mailinfo_config(const char *var, const char *value, void *unused)
 997 {
 998         if (prefixcmp(var, "mailinfo."))
 999                 return git_default_config(var, value, unused);
1000         if (!strcmp(var, "mailinfo.scissors")) {
1001                 use_scissors = git_config_bool(var, value);
1002                 return 0;
1003         }
1004         /* perhaps others here */
1005         return 0;
1006 }
1007
1008 static const char mailinfo_usage[] =
1009         "git mailinfo [-k] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] msg patch < mail >info";
1010
1011 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1012 {
1013         const char *def_charset;
1014
1015         /* NEEDSWORK: might want to do the optional .git/ directory
1016          * discovery
1017          */
1018         git_config(git_mailinfo_config, NULL);
1019
1020         def_charset = (git_commit_encoding ? git_commit_encoding : "UTF-8");
1021         metainfo_charset = def_charset;
1022
1023         while (1 < argc && argv[1][0] == '-') {
1024                 if (!strcmp(argv[1], "-k"))
1025                         keep_subject = 1;
1026                 else if (!strcmp(argv[1], "-u"))
1027                         metainfo_charset = def_charset;
1028                 else if (!strcmp(argv[1], "-n"))
1029                         metainfo_charset = NULL;
1030                 else if (!prefixcmp(argv[1], "--encoding="))
1031                         metainfo_charset = argv[1] + 11;
1032                 else if (!strcmp(argv[1], "--scissors"))
1033                         use_scissors = 1;
1034                 else if (!strcmp(argv[1], "--no-scissors"))
1035                         use_scissors = 0;
1036                 else
1037                         usage(mailinfo_usage);
1038                 argc--; argv++;
1039         }
1040
1041         if (argc != 3)
1042                 usage(mailinfo_usage);
1043
1044         return !!mailinfo(stdin, stdout, argv[1], argv[2]);
1045 }