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