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
  23 static struct strbuf charset = STRBUF_INIT;
  24 static int patch_lines;
  25 static struct strbuf **p_hdr_data, **s_hdr_data;
  26 static int use_scissors;
  27 static int use_inbody_headers = 1;
  28
  29 #define MAX_HDR_PARSED 10
  30 #define MAX_BOUNDARIES 5
  31
  32 static void cleanup_space(struct strbuf *sb);
  33
  34
  35 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  36 {
  37         struct strbuf *src = name;
  38         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  39                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  40                 src = email;
  41         else if (name == out)
  42                 return;
  43         strbuf_reset(out);
  44         strbuf_addbuf(out, src);
  45 }
  46
  47 static void parse_bogus_from(const struct strbuf *line)
  48 {
  49         /* John Doe <johndoe> */
  50
  51         char *bra, *ket;
  52         /* This is fallback, so do not bother if we already have an
  53          * e-mail address.
  54          */
  55         if (email.len)
  56                 return;
  57
  58         bra = strchr(line->buf, '<');
  59         if (!bra)
  60                 return;
  61         ket = strchr(bra, '>');
  62         if (!ket)
  63                 return;
  64
  65         strbuf_reset(&email);
  66         strbuf_add(&email, bra + 1, ket - bra - 1);
  67
  68         strbuf_reset(&name);
  69         strbuf_add(&name, line->buf, bra - line->buf);
  70         strbuf_trim(&name);
  71         get_sane_name(&name, &name, &email);
  72 }
  73
  74 static void handle_from(const struct strbuf *from)
  75 {
  76         char *at;
  77         size_t el;
  78         struct strbuf f;
  79
  80         strbuf_init(&f, from->len);
  81         strbuf_addbuf(&f, from);
  82
  83         at = strchr(f.buf, '@');
  84         if (!at) {
  85                 parse_bogus_from(from);
  86                 return;
  87         }
  88
  89         /*
  90          * If we already have one email, don't take any confusing lines
  91          */
  92         if (email.len && strchr(at + 1, '@')) {
  93                 strbuf_release(&f);
  94                 return;
  95         }
  96
  97         /* Pick up the string around '@', possibly delimited with <>
  98          * pair; that is the email part.
  99          */
 100         while (at > f.buf) {
 101                 char c = at[-1];
 102                 if (isspace(c))
 103                         break;
 104                 if (c == '<') {
 105                         at[-1] = ' ';
 106                         break;
 107                 }
 108                 at--;
 109         }
 110         el = strcspn(at, " \n\t\r\v\f>");
 111         strbuf_reset(&email);
 112         strbuf_add(&email, at, el);
 113         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 114
 115         /* The remainder is name.  It could be
 116          *
 117          * - "John Doe <john.doe@xz>"                   (a), or
 118          * - "john.doe@xz (John Doe)"                   (b), or
 119          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 120          *
 121          * but we have removed the email part, so
 122          *
 123          * - remove extra spaces which could stay after email (case 'c'), and
 124          * - trim from both ends, possibly removing the () pair at the end
 125          *   (cases 'a' and 'b').
 126          */
 127         cleanup_space(&f);
 128         strbuf_trim(&f);
 129         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 130                 strbuf_remove(&f, 0, 1);
 131                 strbuf_setlen(&f, f.len - 1);
 132         }
 133
 134         get_sane_name(&name, &f, &email);
 135         strbuf_release(&f);
 136 }
 137
 138 static void handle_header(struct strbuf **out, const struct strbuf *line)
 139 {
 140         if (!*out) {
 141                 *out = xmalloc(sizeof(struct strbuf));
 142                 strbuf_init(*out, line->len);
 143         } else
 144                 strbuf_reset(*out);
 145
 146         strbuf_addbuf(*out, line);
 147 }
 148
 149 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 150  * to have enough heuristics to grok MIME encoded patches often found
 151  * on our mailing lists.  For example, we do not even treat header lines
 152  * case insensitively.
 153  */
 154
 155 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 156 {
 157         const char *ends, *ap = strcasestr(line, name);
 158         size_t sz;
 159
 160         if (!ap) {
 161                 strbuf_setlen(attr, 0);
 162                 return 0;
 163         }
 164         ap += strlen(name);
 165         if (*ap == '"') {
 166                 ap++;
 167                 ends = "\"";
 168         }
 169         else
 170                 ends = "; \t";
 171         sz = strcspn(ap, ends);
 172         strbuf_add(attr, ap, sz);
 173         return 1;
 174 }
 175
 176 static struct strbuf *content[MAX_BOUNDARIES];
 177
 178 static struct strbuf **content_top = content;
 179
 180 static void handle_content_type(struct strbuf *line)
 181 {
 182         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 183         strbuf_init(boundary, line->len);
 184
 185         if (slurp_attr(line->buf, "boundary=", boundary)) {
 186                 strbuf_insert(boundary, 0, "--", 2);
 187                 if (++content_top > &content[MAX_BOUNDARIES]) {
 188                         fprintf(stderr, "Too many boundaries to handle\n");
 189                         exit(1);
 190                 }
 191                 *content_top = boundary;
 192                 boundary = NULL;
 193         }
 194         slurp_attr(line->buf, "charset=", &charset);
 195
 196         if (boundary) {
 197                 strbuf_release(boundary);
 198                 free(boundary);
 199         }
 200 }
 201
 202 static void handle_content_transfer_encoding(const struct strbuf *line)
 203 {
 204         if (strcasestr(line->buf, "base64"))
 205                 transfer_encoding = TE_BASE64;
 206         else if (strcasestr(line->buf, "quoted-printable"))
 207                 transfer_encoding = TE_QP;
 208         else
 209                 transfer_encoding = TE_DONTCARE;
 210 }
 211
 212 static int is_multipart_boundary(const struct strbuf *line)
 213 {
 214         return (((*content_top)->len <= line->len) &&
 215                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
 216 }
 217
 218 static void cleanup_subject(struct strbuf *subject)
 219 {
 220         size_t at = 0;
 221
 222         while (at < subject->len) {
 223                 char *pos;
 224                 size_t remove;
 225
 226                 switch (subject->buf[at]) {
 227                 case 'r': case 'R':
 228                         if (subject->len <= at + 3)
 229                                 break;
 230                         if (!memcmp(subject->buf + at + 1, "e:", 2)) {
 231                                 strbuf_remove(subject, at, 3);
 232                                 continue;
 233                         }
 234                         at++;
 235                         break;
 236                 case ' ': case '\t': case ':':
 237                         strbuf_remove(subject, at, 1);
 238                         continue;
 239                 case '[':
 240                         pos = strchr(subject->buf + at, ']');
 241                         if (!pos)
 242                                 break;
 243                         remove = pos - subject->buf + at + 1;
 244                         if (!keep_non_patch_brackets_in_subject ||
 245                             (7 <= remove &&
 246                              memmem(subject->buf + at, remove, "PATCH", 5)))
 247                                 strbuf_remove(subject, at, remove);
 248                         else
 249                                 at += remove;
 250                         continue;
 251                 }
 252                 break;
 253         }
 254         strbuf_trim(subject);
 255 }
 256
 257 static void cleanup_space(struct strbuf *sb)
 258 {
 259         size_t pos, cnt;
 260         for (pos = 0; pos < sb->len; pos++) {
 261                 if (isspace(sb->buf[pos])) {
 262                         sb->buf[pos] = ' ';
 263                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 264                         strbuf_remove(sb, pos + 1, cnt);
 265                 }
 266         }
 267 }
 268
 269 static void decode_header(struct strbuf *line);
 270 static const char *header[MAX_HDR_PARSED] = {
 271         "From","Subject","Date",
 272 };
 273
 274 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 275 {
 276         int len = strlen(hdr);
 277         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 278                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 279 }
 280
 281 static int check_header(const struct strbuf *line,
 282                                 struct strbuf *hdr_data[], int overwrite)
 283 {
 284         int i, ret = 0, len;
 285         struct strbuf sb = STRBUF_INIT;
 286         /* search for the interesting parts */
 287         for (i = 0; header[i]; i++) {
 288                 int len = strlen(header[i]);
 289                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 290                         /* Unwrap inline B and Q encoding, and optionally
 291                          * normalize the meta information to utf8.
 292                          */
 293                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 294                         decode_header(&sb);
 295                         handle_header(&hdr_data[i], &sb);
 296                         ret = 1;
 297                         goto check_header_out;
 298                 }
 299         }
 300
 301         /* Content stuff */
 302         if (cmp_header(line, "Content-Type")) {
 303                 len = strlen("Content-Type: ");
 304                 strbuf_add(&sb, line->buf + len, line->len - len);
 305                 decode_header(&sb);
 306                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 307                 handle_content_type(&sb);
 308                 ret = 1;
 309                 goto check_header_out;
 310         }
 311         if (cmp_header(line, "Content-Transfer-Encoding")) {
 312                 len = strlen("Content-Transfer-Encoding: ");
 313                 strbuf_add(&sb, line->buf + len, line->len - len);
 314                 decode_header(&sb);
 315                 handle_content_transfer_encoding(&sb);
 316                 ret = 1;
 317                 goto check_header_out;
 318         }
 319
 320         /* for inbody stuff */
 321         if (!prefixcmp(line->buf, ">From") && isspace(line->buf[5])) {
 322                 ret = 1; /* Should this return 0? */
 323                 goto check_header_out;
 324         }
 325         if (!prefixcmp(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 326                 for (i = 0; header[i]; i++) {
 327                         if (!memcmp("Subject", header[i], 7)) {
 328                                 handle_header(&hdr_data[i], line);
 329                                 ret = 1;
 330                                 goto check_header_out;
 331                         }
 332                 }
 333         }
 334
 335 check_header_out:
 336         strbuf_release(&sb);
 337         return ret;
 338 }
 339
 340 static int is_rfc2822_header(const struct strbuf *line)
 341 {
 342         /*
 343          * The section that defines the loosest possible
 344          * field name is "3.6.8 Optional fields".
 345          *
 346          * optional-field = field-name ":" unstructured CRLF
 347          * field-name = 1*ftext
 348          * ftext = %d33-57 / %59-126
 349          */
 350         int ch;
 351         char *cp = line->buf;
 352
 353         /* Count mbox From headers as headers */
 354         if (!prefixcmp(cp, "From ") || !prefixcmp(cp, ">From "))
 355                 return 1;
 356
 357         while ((ch = *cp++)) {
 358                 if (ch == ':')
 359                         return 1;
 360                 if ((33 <= ch && ch <= 57) ||
 361                     (59 <= ch && ch <= 126))
 362                         continue;
 363                 break;
 364         }
 365         return 0;
 366 }
 367
 368 static int read_one_header_line(struct strbuf *line, FILE *in)
 369 {
 370         /* Get the first part of the line. */
 371         if (strbuf_getline(line, in, '\n'))
 372                 return 0;
 373
 374         /*
 375          * Is it an empty line or not a valid rfc2822 header?
 376          * If so, stop here, and return false ("not a header")
 377          */
 378         strbuf_rtrim(line);
 379         if (!line->len || !is_rfc2822_header(line)) {
 380                 /* Re-add the newline */
 381                 strbuf_addch(line, '\n');
 382                 return 0;
 383         }
 384
 385         /*
 386          * Now we need to eat all the continuation lines..
 387          * Yuck, 2822 header "folding"
 388          */
 389         for (;;) {
 390                 int peek;
 391                 struct strbuf continuation = STRBUF_INIT;
 392
 393                 peek = fgetc(in); ungetc(peek, in);
 394                 if (peek != ' ' && peek != '\t')
 395                         break;
 396                 if (strbuf_getline(&continuation, in, '\n'))
 397                         break;
 398                 continuation.buf[0] = ' ';
 399                 strbuf_rtrim(&continuation);
 400                 strbuf_addbuf(line, &continuation);
 401         }
 402
 403         return 1;
 404 }
 405
 406 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 407 {
 408         const char *in = q_seg->buf;
 409         int c;
 410         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 411         strbuf_init(out, q_seg->len);
 412
 413         while ((c = *in++) != 0) {
 414                 if (c == '=') {
 415                         int d = *in++;
 416                         if (d == '\n' || !d)
 417                                 break; /* drop trailing newline */
 418                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 419                         continue;
 420                 }
 421                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 422                         c = 0x20;
 423                 strbuf_addch(out, c);
 424         }
 425         return out;
 426 }
 427
 428 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 429 {
 430         /* Decode in..ep, possibly in-place to ot */
 431         int c, pos = 0, acc = 0;
 432         const char *in = b_seg->buf;
 433         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 434         strbuf_init(out, b_seg->len);
 435
 436         while ((c = *in++) != 0) {
 437                 if (c == '+')
 438                         c = 62;
 439                 else if (c == '/')
 440                         c = 63;
 441                 else if ('A' <= c && c <= 'Z')
 442                         c -= 'A';
 443                 else if ('a' <= c && c <= 'z')
 444                         c -= 'a' - 26;
 445                 else if ('0' <= c && c <= '9')
 446                         c -= '0' - 52;
 447                 else
 448                         continue; /* garbage */
 449                 switch (pos++) {
 450                 case 0:
 451                         acc = (c << 2);
 452                         break;
 453                 case 1:
 454                         strbuf_addch(out, (acc | (c >> 4)));
 455                         acc = (c & 15) << 4;
 456                         break;
 457                 case 2:
 458                         strbuf_addch(out, (acc | (c >> 2)));
 459                         acc = (c & 3) << 6;
 460                         break;
 461                 case 3:
 462                         strbuf_addch(out, (acc | c));
 463                         acc = pos = 0;
 464                         break;
 465                 }
 466         }
 467         return out;
 468 }
 469
 470 /*
 471  * When there is no known charset, guess.
 472  *
 473  * Right now we assume that if the target is UTF-8 (the default),
 474  * and it already looks like UTF-8 (which includes US-ASCII as its
 475  * subset, of course) then that is what it is and there is nothing
 476  * to do.
 477  *
 478  * Otherwise, we default to assuming it is Latin1 for historical
 479  * reasons.
 480  */
 481 static const char *guess_charset(const struct strbuf *line, const char *target_charset)
 482 {
 483         if (is_encoding_utf8(target_charset)) {
 484                 if (is_utf8(line->buf))
 485                         return NULL;
 486         }
 487         return "ISO8859-1";
 488 }
 489
 490 static void convert_to_utf8(struct strbuf *line, const char *charset)
 491 {
 492         char *out;
 493
 494         if (!charset || !*charset) {
 495                 charset = guess_charset(line, metainfo_charset);
 496                 if (!charset)
 497                         return;
 498         }
 499
 500         if (!strcasecmp(metainfo_charset, charset))
 501                 return;
 502         out = reencode_string(line->buf, metainfo_charset, charset);
 503         if (!out)
 504                 die("cannot convert from %s to %s",
 505                     charset, metainfo_charset);
 506         strbuf_attach(line, out, strlen(out), strlen(out));
 507 }
 508
 509 static int decode_header_bq(struct strbuf *it)
 510 {
 511         char *in, *ep, *cp;
 512         struct strbuf outbuf = STRBUF_INIT, *dec;
 513         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 514         int rfc2047 = 0;
 515
 516         in = it->buf;
 517         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 518                 int encoding;
 519                 strbuf_reset(&charset_q);
 520                 strbuf_reset(&piecebuf);
 521                 rfc2047 = 1;
 522
 523                 if (in != ep) {
 524                         /*
 525                          * We are about to process an encoded-word
 526                          * that begins at ep, but there is something
 527                          * before the encoded word.
 528                          */
 529                         char *scan;
 530                         for (scan = in; scan < ep; scan++)
 531                                 if (!isspace(*scan))
 532                                         break;
 533
 534                         if (scan != ep || in == it->buf) {
 535                                 /*
 536                                  * We should not lose that "something",
 537                                  * unless we have just processed an
 538                                  * encoded-word, and there is only LWS
 539                                  * before the one we are about to process.
 540                                  */
 541                                 strbuf_add(&outbuf, in, ep - in);
 542                         }
 543                 }
 544                 /* E.g.
 545                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 546                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 547                  */
 548                 ep += 2;
 549
 550                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 551                         goto decode_header_bq_out;
 552
 553                 if (cp + 3 - it->buf > it->len)
 554                         goto decode_header_bq_out;
 555                 strbuf_add(&charset_q, ep, cp - ep);
 556
 557                 encoding = cp[1];
 558                 if (!encoding || cp[2] != '?')
 559                         goto decode_header_bq_out;
 560                 ep = strstr(cp + 3, "?=");
 561                 if (!ep)
 562                         goto decode_header_bq_out;
 563                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 564                 switch (tolower(encoding)) {
 565                 default:
 566                         goto decode_header_bq_out;
 567                 case 'b':
 568                         dec = decode_b_segment(&piecebuf);
 569                         break;
 570                 case 'q':
 571                         dec = decode_q_segment(&piecebuf, 1);
 572                         break;
 573                 }
 574                 if (metainfo_charset)
 575                         convert_to_utf8(dec, charset_q.buf);
 576
 577                 strbuf_addbuf(&outbuf, dec);
 578                 strbuf_release(dec);
 579                 free(dec);
 580                 in = ep + 2;
 581         }
 582         strbuf_addstr(&outbuf, in);
 583         strbuf_reset(it);
 584         strbuf_addbuf(it, &outbuf);
 585 decode_header_bq_out:
 586         strbuf_release(&outbuf);
 587         strbuf_release(&charset_q);
 588         strbuf_release(&piecebuf);
 589         return rfc2047;
 590 }
 591
 592 static void decode_header(struct strbuf *it)
 593 {
 594         if (decode_header_bq(it))
 595                 return;
 596         /* otherwise "it" is a straight copy of the input.
 597          * This can be binary guck but there is no charset specified.
 598          */
 599         if (metainfo_charset)
 600                 convert_to_utf8(it, "");
 601 }
 602
 603 static void decode_transfer_encoding(struct strbuf *line)
 604 {
 605         struct strbuf *ret;
 606
 607         switch (transfer_encoding) {
 608         case TE_QP:
 609                 ret = decode_q_segment(line, 0);
 610                 break;
 611         case TE_BASE64:
 612                 ret = decode_b_segment(line);
 613                 break;
 614         case TE_DONTCARE:
 615         default:
 616                 return;
 617         }
 618         strbuf_reset(line);
 619         strbuf_addbuf(line, ret);
 620         strbuf_release(ret);
 621         free(ret);
 622 }
 623
 624 static void handle_filter(struct strbuf *line);
 625
 626 static int find_boundary(void)
 627 {
 628         while (!strbuf_getline(&line, fin, '\n')) {
 629                 if (*content_top && is_multipart_boundary(&line))
 630                         return 1;
 631         }
 632         return 0;
 633 }
 634
 635 static int handle_boundary(void)
 636 {
 637         struct strbuf newline = STRBUF_INIT;
 638
 639         strbuf_addch(&newline, '\n');
 640 again:
 641         if (line.len >= (*content_top)->len + 2 &&
 642             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 643                 /* we hit an end boundary */
 644                 /* pop the current boundary off the stack */
 645                 strbuf_release(*content_top);
 646                 free(*content_top);
 647                 *content_top = NULL;
 648
 649                 /* technically won't happen as is_multipart_boundary()
 650                    will fail first.  But just in case..
 651                  */
 652                 if (--content_top < content) {
 653                         fprintf(stderr, "Detected mismatched boundaries, "
 654                                         "can't recover\n");
 655                         exit(1);
 656                 }
 657                 handle_filter(&newline);
 658                 strbuf_release(&newline);
 659
 660                 /* skip to the next boundary */
 661                 if (!find_boundary())
 662                         return 0;
 663                 goto again;
 664         }
 665
 666         /* set some defaults */
 667         transfer_encoding = TE_DONTCARE;
 668         strbuf_reset(&charset);
 669
 670         /* slurp in this section's info */
 671         while (read_one_header_line(&line, fin))
 672                 check_header(&line, p_hdr_data, 0);
 673
 674         strbuf_release(&newline);
 675         /* replenish line */
 676         if (strbuf_getline(&line, fin, '\n'))
 677                 return 0;
 678         strbuf_addch(&line, '\n');
 679         return 1;
 680 }
 681
 682 static inline int patchbreak(const struct strbuf *line)
 683 {
 684         size_t i;
 685
 686         /* Beginning of a "diff -" header? */
 687         if (!prefixcmp(line->buf, "diff -"))
 688                 return 1;
 689
 690         /* CVS "Index: " line? */
 691         if (!prefixcmp(line->buf, "Index: "))
 692                 return 1;
 693
 694         /*
 695          * "--- <filename>" starts patches without headers
 696          * "---<sp>*" is a manual separator
 697          */
 698         if (line->len < 4)
 699                 return 0;
 700
 701         if (!prefixcmp(line->buf, "---")) {
 702                 /* space followed by a filename? */
 703                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 704                         return 1;
 705                 /* Just whitespace? */
 706                 for (i = 3; i < line->len; i++) {
 707                         unsigned char c = line->buf[i];
 708                         if (c == '\n')
 709                                 return 1;
 710                         if (!isspace(c))
 711                                 break;
 712                 }
 713                 return 0;
 714         }
 715         return 0;
 716 }
 717
 718 static int is_scissors_line(const struct strbuf *line)
 719 {
 720         size_t i, len = line->len;
 721         int scissors = 0, gap = 0;
 722         int first_nonblank = -1;
 723         int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
 724         const char *buf = line->buf;
 725
 726         for (i = 0; i < len; i++) {
 727                 if (isspace(buf[i])) {
 728                         if (in_perforation) {
 729                                 perforation++;
 730                                 gap++;
 731                         }
 732                         continue;
 733                 }
 734                 last_nonblank = i;
 735                 if (first_nonblank < 0)
 736                         first_nonblank = i;
 737                 if (buf[i] == '-') {
 738                         in_perforation = 1;
 739                         perforation++;
 740                         continue;
 741                 }
 742                 if (i + 1 < len &&
 743                     (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
 744                      !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
 745                         in_perforation = 1;
 746                         perforation += 2;
 747                         scissors += 2;
 748                         i++;
 749                         continue;
 750                 }
 751                 in_perforation = 0;
 752         }
 753
 754         /*
 755          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 756          * Even though there can be arbitrary cruft on the same line
 757          * (e.g. "cut here"), in order to avoid misidentification, the
 758          * perforation must occupy more than a third of the visible
 759          * width of the line, and dashes and scissors must occupy more
 760          * than half of the perforation.
 761          */
 762
 763         visible = last_nonblank - first_nonblank + 1;
 764         return (scissors && 8 <= visible &&
 765                 visible < perforation * 3 &&
 766                 gap * 2 < perforation);
 767 }
 768
 769 static int handle_commit_msg(struct strbuf *line)
 770 {
 771         static int still_looking = 1;
 772
 773         if (!cmitmsg)
 774                 return 0;
 775
 776         if (still_looking) {
 777                 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
 778                         return 0;
 779         }
 780
 781         if (use_inbody_headers && still_looking) {
 782                 still_looking = check_header(line, s_hdr_data, 0);
 783                 if (still_looking)
 784                         return 0;
 785         } else
 786                 /* Only trim the first (blank) line of the commit message
 787                  * when ignoring in-body headers.
 788                  */
 789                 still_looking = 0;
 790
 791         /* normalize the log message to UTF-8. */
 792         if (metainfo_charset)
 793                 convert_to_utf8(line, charset.buf);
 794
 795         if (use_scissors && is_scissors_line(line)) {
 796                 int i;
 797                 if (fseek(cmitmsg, 0L, SEEK_SET))
 798                         die_errno("Could not rewind output message file");
 799                 if (ftruncate(fileno(cmitmsg), 0))
 800                         die_errno("Could not truncate output message file at scissors");
 801                 still_looking = 1;
 802
 803                 /*
 804                  * We may have already read "secondary headers"; purge
 805                  * them to give ourselves a clean restart.
 806                  */
 807                 for (i = 0; header[i]; i++) {
 808                         if (s_hdr_data[i])
 809                                 strbuf_release(s_hdr_data[i]);
 810                         s_hdr_data[i] = NULL;
 811                 }
 812                 return 0;
 813         }
 814
 815         if (patchbreak(line)) {
 816                 fclose(cmitmsg);
 817                 cmitmsg = NULL;
 818                 return 1;
 819         }
 820
 821         fputs(line->buf, cmitmsg);
 822         return 0;
 823 }
 824
 825 static void handle_patch(const struct strbuf *line)
 826 {
 827         fwrite(line->buf, 1, line->len, patchfile);
 828         patch_lines++;
 829 }
 830
 831 static void handle_filter(struct strbuf *line)
 832 {
 833         static int filter = 0;
 834
 835         /* filter tells us which part we left off on */
 836         switch (filter) {
 837         case 0:
 838                 if (!handle_commit_msg(line))
 839                         break;
 840                 filter++;
 841         case 1:
 842                 handle_patch(line);
 843                 break;
 844         }
 845 }
 846
 847 static void handle_body(void)
 848 {
 849         struct strbuf prev = STRBUF_INIT;
 850
 851         /* Skip up to the first boundary */
 852         if (*content_top) {
 853                 if (!find_boundary())
 854                         goto handle_body_out;
 855         }
 856
 857         do {
 858                 /* process any boundary lines */
 859                 if (*content_top && is_multipart_boundary(&line)) {
 860                         /* flush any leftover */
 861                         if (prev.len) {
 862                                 handle_filter(&prev);
 863                                 strbuf_reset(&prev);
 864                         }
 865                         if (!handle_boundary())
 866                                 goto handle_body_out;
 867                 }
 868
 869                 /* Unwrap transfer encoding */
 870                 decode_transfer_encoding(&line);
 871
 872                 switch (transfer_encoding) {
 873                 case TE_BASE64:
 874                 case TE_QP:
 875                 {
 876                         struct strbuf **lines, **it, *sb;
 877
 878                         /* Prepend any previous partial lines */
 879                         strbuf_insert(&line, 0, prev.buf, prev.len);
 880                         strbuf_reset(&prev);
 881
 882                         /*
 883                          * This is a decoded line that may contain
 884                          * multiple new lines.  Pass only one chunk
 885                          * at a time to handle_filter()
 886                          */
 887                         lines = strbuf_split(&line, '\n');
 888                         for (it = lines; (sb = *it); it++) {
 889                                 if (*(it + 1) == NULL) /* The last line */
 890                                         if (sb->buf[sb->len - 1] != '\n') {
 891                                                 /* Partial line, save it for later. */
 892                                                 strbuf_addbuf(&prev, sb);
 893                                                 break;
 894                                         }
 895                                 handle_filter(sb);
 896                         }
 897                         /*
 898                          * The partial chunk is saved in "prev" and will be
 899                          * appended by the next iteration of read_line_with_nul().
 900                          */
 901                         strbuf_list_free(lines);
 902                         break;
 903                 }
 904                 default:
 905                         handle_filter(&line);
 906                 }
 907
 908         } while (!strbuf_getwholeline(&line, fin, '\n'));
 909
 910 handle_body_out:
 911         strbuf_release(&prev);
 912 }
 913
 914 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 915 {
 916         const char *sp = data->buf;
 917         while (1) {
 918                 char *ep = strchr(sp, '\n');
 919                 int len;
 920                 if (!ep)
 921                         len = strlen(sp);
 922                 else
 923                         len = ep - sp;
 924                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 925                 if (!ep)
 926                         break;
 927                 sp = ep + 1;
 928         }
 929 }
 930
 931 static void handle_info(void)
 932 {
 933         struct strbuf *hdr;
 934         int i;
 935
 936         for (i = 0; header[i]; i++) {
 937                 /* only print inbody headers if we output a patch file */
 938                 if (patch_lines && s_hdr_data[i])
 939                         hdr = s_hdr_data[i];
 940                 else if (p_hdr_data[i])
 941                         hdr = p_hdr_data[i];
 942                 else
 943                         continue;
 944
 945                 if (!memcmp(header[i], "Subject", 7)) {
 946                         if (!keep_subject) {
 947                                 cleanup_subject(hdr);
 948                                 cleanup_space(hdr);
 949                         }
 950                         output_header_lines(fout, "Subject", hdr);
 951                 } else if (!memcmp(header[i], "From", 4)) {
 952                         cleanup_space(hdr);
 953                         handle_from(hdr);
 954                         fprintf(fout, "Author: %s\n", name.buf);
 955                         fprintf(fout, "Email: %s\n", email.buf);
 956                 } else {
 957                         cleanup_space(hdr);
 958                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 959                 }
 960         }
 961         fprintf(fout, "\n");
 962 }
 963
 964 static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
 965 {
 966         int peek;
 967         fin = in;
 968         fout = out;
 969
 970         cmitmsg = fopen(msg, "w");
 971         if (!cmitmsg) {
 972                 perror(msg);
 973                 return -1;
 974         }
 975         patchfile = fopen(patch, "w");
 976         if (!patchfile) {
 977                 perror(patch);
 978                 fclose(cmitmsg);
 979                 return -1;
 980         }
 981
 982         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
 983         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
 984
 985         do {
 986                 peek = fgetc(in);
 987         } while (isspace(peek));
 988         ungetc(peek, in);
 989
 990         /* process the email header */
 991         while (read_one_header_line(&line, fin))
 992                 check_header(&line, p_hdr_data, 1);
 993
 994         handle_body();
 995         handle_info();
 996
 997         return 0;
 998 }
 999
1000 static int git_mailinfo_config(const char *var, const char *value, void *unused)
1001 {
1002         if (prefixcmp(var, "mailinfo."))
1003                 return git_default_config(var, value, unused);
1004         if (!strcmp(var, "mailinfo.scissors")) {
1005                 use_scissors = git_config_bool(var, value);
1006                 return 0;
1007         }
1008         /* perhaps others here */
1009         return 0;
1010 }
1011
1012 static const char mailinfo_usage[] =
1013         "git mailinfo [-k|-b] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] msg patch < mail >info";
1014
1015 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1016 {
1017         const char *def_charset;
1018
1019         /* NEEDSWORK: might want to do the optional .git/ directory
1020          * discovery
1021          */
1022         git_config(git_mailinfo_config, NULL);
1023
1024         def_charset = get_commit_output_encoding();
1025         metainfo_charset = def_charset;
1026
1027         while (1 < argc && argv[1][0] == '-') {
1028                 if (!strcmp(argv[1], "-k"))
1029                         keep_subject = 1;
1030                 else if (!strcmp(argv[1], "-b"))
1031                         keep_non_patch_brackets_in_subject = 1;
1032                 else if (!strcmp(argv[1], "-u"))
1033                         metainfo_charset = def_charset;
1034                 else if (!strcmp(argv[1], "-n"))
1035                         metainfo_charset = NULL;
1036                 else if (!prefixcmp(argv[1], "--encoding="))
1037                         metainfo_charset = argv[1] + 11;
1038                 else if (!strcmp(argv[1], "--scissors"))
1039                         use_scissors = 1;
1040                 else if (!strcmp(argv[1], "--no-scissors"))
1041                         use_scissors = 0;
1042                 else if (!strcmp(argv[1], "--no-inbody-headers"))
1043                         use_inbody_headers = 0;
1044                 else
1045                         usage(mailinfo_usage);
1046                 argc--; argv++;
1047         }
1048
1049         if (argc != 3)
1050                 usage(mailinfo_usage);
1051
1052         return !!mailinfo(stdin, stdout, argv[1], argv[2]);
1053 }