builtin/mailinfo.c

   1 /*
   2  * Another stupid program, this one parsing the headers of an
   3  * email to figure out authorship and subject
   4  */
   5 #include "cache.h"
   6 #include "builtin.h"
   7 #include "utf8.h"
   8 #include "strbuf.h"
   9
  10 static FILE *cmitmsg, *patchfile, *fin, *fout;
  11
  12 static int keep_subject;
  13 static int keep_non_patch_brackets_in_subject;
  14 static const char *metainfo_charset;
  15 static struct strbuf line = STRBUF_INIT;
  16 static struct strbuf name = STRBUF_INIT;
  17 static struct strbuf email = STRBUF_INIT;
  18
  19 static enum  {
  20         TE_DONTCARE, TE_QP, TE_BASE64
  21 } transfer_encoding;
  22 static enum  {
  23         TYPE_TEXT, TYPE_OTHER
  24 } message_type;
  25
  26 static struct strbuf charset = STRBUF_INIT;
  27 static int patch_lines;
  28 static struct strbuf **p_hdr_data, **s_hdr_data;
  29 static int use_scissors;
  30 static int use_inbody_headers = 1;
  31
  32 #define MAX_HDR_PARSED 10
  33 #define MAX_BOUNDARIES 5
  34
  35 static void cleanup_space(struct strbuf *sb);
  36
  37
  38 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  39 {
  40         struct strbuf *src = name;
  41         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  42                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  43                 src = email;
  44         else if (name == out)
  45                 return;
  46         strbuf_reset(out);
  47         strbuf_addbuf(out, src);
  48 }
  49
  50 static void parse_bogus_from(const struct strbuf *line)
  51 {
  52         /* John Doe <johndoe> */
  53
  54         char *bra, *ket;
  55         /* This is fallback, so do not bother if we already have an
  56          * e-mail address.
  57          */
  58         if (email.len)
  59                 return;
  60
  61         bra = strchr(line->buf, '<');
  62         if (!bra)
  63                 return;
  64         ket = strchr(bra, '>');
  65         if (!ket)
  66                 return;
  67
  68         strbuf_reset(&email);
  69         strbuf_add(&email, bra + 1, ket - bra - 1);
  70
  71         strbuf_reset(&name);
  72         strbuf_add(&name, line->buf, bra - line->buf);
  73         strbuf_trim(&name);
  74         get_sane_name(&name, &name, &email);
  75 }
  76
  77 static void handle_from(const struct strbuf *from)
  78 {
  79         char *at;
  80         size_t el;
  81         struct strbuf f;
  82
  83         strbuf_init(&f, from->len);
  84         strbuf_addbuf(&f, from);
  85
  86         at = strchr(f.buf, '@');
  87         if (!at) {
  88                 parse_bogus_from(from);
  89                 return;
  90         }
  91
  92         /*
  93          * If we already have one email, don't take any confusing lines
  94          */
  95         if (email.len && strchr(at + 1, '@')) {
  96                 strbuf_release(&f);
  97                 return;
  98         }
  99
 100         /* Pick up the string around '@', possibly delimited with <>
 101          * pair; that is the email part.
 102          */
 103         while (at > f.buf) {
 104                 char c = at[-1];
 105                 if (isspace(c))
 106                         break;
 107                 if (c == '<') {
 108                         at[-1] = ' ';
 109                         break;
 110                 }
 111                 at--;
 112         }
 113         el = strcspn(at, " \n\t\r\v\f>");
 114         strbuf_reset(&email);
 115         strbuf_add(&email, at, el);
 116         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 117
 118         /* The remainder is name.  It could be
 119          *
 120          * - "John Doe <john.doe@xz>"                   (a), or
 121          * - "john.doe@xz (John Doe)"                   (b), or
 122          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 123          *
 124          * but we have removed the email part, so
 125          *
 126          * - remove extra spaces which could stay after email (case 'c'), and
 127          * - trim from both ends, possibly removing the () pair at the end
 128          *   (cases 'a' and 'b').
 129          */
 130         cleanup_space(&f);
 131         strbuf_trim(&f);
 132         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 133                 strbuf_remove(&f, 0, 1);
 134                 strbuf_setlen(&f, f.len - 1);
 135         }
 136
 137         get_sane_name(&name, &f, &email);
 138         strbuf_release(&f);
 139 }
 140
 141 static void handle_header(struct strbuf **out, const struct strbuf *line)
 142 {
 143         if (!*out) {
 144                 *out = xmalloc(sizeof(struct strbuf));
 145                 strbuf_init(*out, line->len);
 146         } else
 147                 strbuf_reset(*out);
 148
 149         strbuf_addbuf(*out, line);
 150 }
 151
 152 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 153  * to have enough heuristics to grok MIME encoded patches often found
 154  * on our mailing lists.  For example, we do not even treat header lines
 155  * case insensitively.
 156  */
 157
 158 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 159 {
 160         const char *ends, *ap = strcasestr(line, name);
 161         size_t sz;
 162
 163         if (!ap) {
 164                 strbuf_setlen(attr, 0);
 165                 return 0;
 166         }
 167         ap += strlen(name);
 168         if (*ap == '"') {
 169                 ap++;
 170                 ends = "\"";
 171         }
 172         else
 173                 ends = "; \t";
 174         sz = strcspn(ap, ends);
 175         strbuf_add(attr, ap, sz);
 176         return 1;
 177 }
 178
 179 static struct strbuf *content[MAX_BOUNDARIES];
 180
 181 static struct strbuf **content_top = content;
 182
 183 static void handle_content_type(struct strbuf *line)
 184 {
 185         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 186         strbuf_init(boundary, line->len);
 187
 188         if (!strcasestr(line->buf, "text/"))
 189                  message_type = TYPE_OTHER;
 190         if (slurp_attr(line->buf, "boundary=", boundary)) {
 191                 strbuf_insert(boundary, 0, "--", 2);
 192                 if (++content_top > &content[MAX_BOUNDARIES]) {
 193                         fprintf(stderr, "Too many boundaries to handle\n");
 194                         exit(1);
 195                 }
 196                 *content_top = boundary;
 197                 boundary = NULL;
 198         }
 199         slurp_attr(line->buf, "charset=", &charset);
 200
 201         if (boundary) {
 202                 strbuf_release(boundary);
 203                 free(boundary);
 204         }
 205 }
 206
 207 static void handle_content_transfer_encoding(const struct strbuf *line)
 208 {
 209         if (strcasestr(line->buf, "base64"))
 210                 transfer_encoding = TE_BASE64;
 211         else if (strcasestr(line->buf, "quoted-printable"))
 212                 transfer_encoding = TE_QP;
 213         else
 214                 transfer_encoding = TE_DONTCARE;
 215 }
 216
 217 static int is_multipart_boundary(const struct strbuf *line)
 218 {
 219         return (((*content_top)->len <= line->len) &&
 220                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
 221 }
 222
 223 static void cleanup_subject(struct strbuf *subject)
 224 {
 225         size_t at = 0;
 226
 227         while (at < subject->len) {
 228                 char *pos;
 229                 size_t remove;
 230
 231                 switch (subject->buf[at]) {
 232                 case 'r': case 'R':
 233                         if (subject->len <= at + 3)
 234                                 break;
 235                         if ((subject->buf[at + 1] == 'e' ||
 236                              subject->buf[at + 1] == 'E') &&
 237                             subject->buf[at + 2] == ':') {
 238                                 strbuf_remove(subject, at, 3);
 239                                 continue;
 240                         }
 241                         at++;
 242                         break;
 243                 case ' ': case '\t': case ':':
 244                         strbuf_remove(subject, at, 1);
 245                         continue;
 246                 case '[':
 247                         pos = strchr(subject->buf + at, ']');
 248                         if (!pos)
 249                                 break;
 250                         remove = pos - subject->buf + at + 1;
 251                         if (!keep_non_patch_brackets_in_subject ||
 252                             (7 <= remove &&
 253                              memmem(subject->buf + at, remove, "PATCH", 5)))
 254                                 strbuf_remove(subject, at, remove);
 255                         else {
 256                                 at += remove;
 257                                 /*
 258                                  * If the input had a space after the ], keep
 259                                  * it.  We don't bother with finding the end of
 260                                  * the space, since we later normalize it
 261                                  * anyway.
 262                                  */
 263                                 if (isspace(subject->buf[at]))
 264                                         at += 1;
 265                         }
 266                         continue;
 267                 }
 268                 break;
 269         }
 270         strbuf_trim(subject);
 271 }
 272
 273 static void cleanup_space(struct strbuf *sb)
 274 {
 275         size_t pos, cnt;
 276         for (pos = 0; pos < sb->len; pos++) {
 277                 if (isspace(sb->buf[pos])) {
 278                         sb->buf[pos] = ' ';
 279                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 280                         strbuf_remove(sb, pos + 1, cnt);
 281                 }
 282         }
 283 }
 284
 285 static void decode_header(struct strbuf *line);
 286 static const char *header[MAX_HDR_PARSED] = {
 287         "From","Subject","Date",
 288 };
 289
 290 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 291 {
 292         int len = strlen(hdr);
 293         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 294                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 295 }
 296
 297 static int check_header(const struct strbuf *line,
 298                                 struct strbuf *hdr_data[], int overwrite)
 299 {
 300         int i, ret = 0, len;
 301         struct strbuf sb = STRBUF_INIT;
 302         /* search for the interesting parts */
 303         for (i = 0; header[i]; i++) {
 304                 int len = strlen(header[i]);
 305                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 306                         /* Unwrap inline B and Q encoding, and optionally
 307                          * normalize the meta information to utf8.
 308                          */
 309                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 310                         decode_header(&sb);
 311                         handle_header(&hdr_data[i], &sb);
 312                         ret = 1;
 313                         goto check_header_out;
 314                 }
 315         }
 316
 317         /* Content stuff */
 318         if (cmp_header(line, "Content-Type")) {
 319                 len = strlen("Content-Type: ");
 320                 strbuf_add(&sb, line->buf + len, line->len - len);
 321                 decode_header(&sb);
 322                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 323                 handle_content_type(&sb);
 324                 ret = 1;
 325                 goto check_header_out;
 326         }
 327         if (cmp_header(line, "Content-Transfer-Encoding")) {
 328                 len = strlen("Content-Transfer-Encoding: ");
 329                 strbuf_add(&sb, line->buf + len, line->len - len);
 330                 decode_header(&sb);
 331                 handle_content_transfer_encoding(&sb);
 332                 ret = 1;
 333                 goto check_header_out;
 334         }
 335
 336         /* for inbody stuff */
 337         if (!prefixcmp(line->buf, ">From") && isspace(line->buf[5])) {
 338                 ret = 1; /* Should this return 0? */
 339                 goto check_header_out;
 340         }
 341         if (!prefixcmp(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 342                 for (i = 0; header[i]; i++) {
 343                         if (!memcmp("Subject", header[i], 7)) {
 344                                 handle_header(&hdr_data[i], line);
 345                                 ret = 1;
 346                                 goto check_header_out;
 347                         }
 348                 }
 349         }
 350
 351 check_header_out:
 352         strbuf_release(&sb);
 353         return ret;
 354 }
 355
 356 static int is_rfc2822_header(const struct strbuf *line)
 357 {
 358         /*
 359          * The section that defines the loosest possible
 360          * field name is "3.6.8 Optional fields".
 361          *
 362          * optional-field = field-name ":" unstructured CRLF
 363          * field-name = 1*ftext
 364          * ftext = %d33-57 / %59-126
 365          */
 366         int ch;
 367         char *cp = line->buf;
 368
 369         /* Count mbox From headers as headers */
 370         if (!prefixcmp(cp, "From ") || !prefixcmp(cp, ">From "))
 371                 return 1;
 372
 373         while ((ch = *cp++)) {
 374                 if (ch == ':')
 375                         return 1;
 376                 if ((33 <= ch && ch <= 57) ||
 377                     (59 <= ch && ch <= 126))
 378                         continue;
 379                 break;
 380         }
 381         return 0;
 382 }
 383
 384 static int read_one_header_line(struct strbuf *line, FILE *in)
 385 {
 386         /* Get the first part of the line. */
 387         if (strbuf_getline(line, in, '\n'))
 388                 return 0;
 389
 390         /*
 391          * Is it an empty line or not a valid rfc2822 header?
 392          * If so, stop here, and return false ("not a header")
 393          */
 394         strbuf_rtrim(line);
 395         if (!line->len || !is_rfc2822_header(line)) {
 396                 /* Re-add the newline */
 397                 strbuf_addch(line, '\n');
 398                 return 0;
 399         }
 400
 401         /*
 402          * Now we need to eat all the continuation lines..
 403          * Yuck, 2822 header "folding"
 404          */
 405         for (;;) {
 406                 int peek;
 407                 struct strbuf continuation = STRBUF_INIT;
 408
 409                 peek = fgetc(in); ungetc(peek, in);
 410                 if (peek != ' ' && peek != '\t')
 411                         break;
 412                 if (strbuf_getline(&continuation, in, '\n'))
 413                         break;
 414                 continuation.buf[0] = ' ';
 415                 strbuf_rtrim(&continuation);
 416                 strbuf_addbuf(line, &continuation);
 417         }
 418
 419         return 1;
 420 }
 421
 422 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 423 {
 424         const char *in = q_seg->buf;
 425         int c;
 426         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 427         strbuf_init(out, q_seg->len);
 428
 429         while ((c = *in++) != 0) {
 430                 if (c == '=') {
 431                         int d = *in++;
 432                         if (d == '\n' || !d)
 433                                 break; /* drop trailing newline */
 434                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 435                         continue;
 436                 }
 437                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 438                         c = 0x20;
 439                 strbuf_addch(out, c);
 440         }
 441         return out;
 442 }
 443
 444 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 445 {
 446         /* Decode in..ep, possibly in-place to ot */
 447         int c, pos = 0, acc = 0;
 448         const char *in = b_seg->buf;
 449         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 450         strbuf_init(out, b_seg->len);
 451
 452         while ((c = *in++) != 0) {
 453                 if (c == '+')
 454                         c = 62;
 455                 else if (c == '/')
 456                         c = 63;
 457                 else if ('A' <= c && c <= 'Z')
 458                         c -= 'A';
 459                 else if ('a' <= c && c <= 'z')
 460                         c -= 'a' - 26;
 461                 else if ('0' <= c && c <= '9')
 462                         c -= '0' - 52;
 463                 else
 464                         continue; /* garbage */
 465                 switch (pos++) {
 466                 case 0:
 467                         acc = (c << 2);
 468                         break;
 469                 case 1:
 470                         strbuf_addch(out, (acc | (c >> 4)));
 471                         acc = (c & 15) << 4;
 472                         break;
 473                 case 2:
 474                         strbuf_addch(out, (acc | (c >> 2)));
 475                         acc = (c & 3) << 6;
 476                         break;
 477                 case 3:
 478                         strbuf_addch(out, (acc | c));
 479                         acc = pos = 0;
 480                         break;
 481                 }
 482         }
 483         return out;
 484 }
 485
 486 static void convert_to_utf8(struct strbuf *line, const char *charset)
 487 {
 488         char *out;
 489
 490         if (!charset || !*charset)
 491                 return;
 492         if (!strcasecmp(metainfo_charset, charset))
 493                 return;
 494         out = reencode_string(line->buf, metainfo_charset, charset);
 495         if (!out)
 496                 die("cannot convert from %s to %s",
 497                     charset, metainfo_charset);
 498         strbuf_attach(line, out, strlen(out), strlen(out));
 499 }
 500
 501 static int decode_header_bq(struct strbuf *it)
 502 {
 503         char *in, *ep, *cp;
 504         struct strbuf outbuf = STRBUF_INIT, *dec;
 505         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 506         int rfc2047 = 0;
 507
 508         in = it->buf;
 509         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 510                 int encoding;
 511                 strbuf_reset(&charset_q);
 512                 strbuf_reset(&piecebuf);
 513                 rfc2047 = 1;
 514
 515                 if (in != ep) {
 516                         /*
 517                          * We are about to process an encoded-word
 518                          * that begins at ep, but there is something
 519                          * before the encoded word.
 520                          */
 521                         char *scan;
 522                         for (scan = in; scan < ep; scan++)
 523                                 if (!isspace(*scan))
 524                                         break;
 525
 526                         if (scan != ep || in == it->buf) {
 527                                 /*
 528                                  * We should not lose that "something",
 529                                  * unless we have just processed an
 530                                  * encoded-word, and there is only LWS
 531                                  * before the one we are about to process.
 532                                  */
 533                                 strbuf_add(&outbuf, in, ep - in);
 534                         }
 535                 }
 536                 /* E.g.
 537                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 538                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 539                  */
 540                 ep += 2;
 541
 542                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 543                         goto decode_header_bq_out;
 544
 545                 if (cp + 3 - it->buf > it->len)
 546                         goto decode_header_bq_out;
 547                 strbuf_add(&charset_q, ep, cp - ep);
 548
 549                 encoding = cp[1];
 550                 if (!encoding || cp[2] != '?')
 551                         goto decode_header_bq_out;
 552                 ep = strstr(cp + 3, "?=");
 553                 if (!ep)
 554                         goto decode_header_bq_out;
 555                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 556                 switch (tolower(encoding)) {
 557                 default:
 558                         goto decode_header_bq_out;
 559                 case 'b':
 560                         dec = decode_b_segment(&piecebuf);
 561                         break;
 562                 case 'q':
 563                         dec = decode_q_segment(&piecebuf, 1);
 564                         break;
 565                 }
 566                 if (metainfo_charset)
 567                         convert_to_utf8(dec, charset_q.buf);
 568
 569                 strbuf_addbuf(&outbuf, dec);
 570                 strbuf_release(dec);
 571                 free(dec);
 572                 in = ep + 2;
 573         }
 574         strbuf_addstr(&outbuf, in);
 575         strbuf_reset(it);
 576         strbuf_addbuf(it, &outbuf);
 577 decode_header_bq_out:
 578         strbuf_release(&outbuf);
 579         strbuf_release(&charset_q);
 580         strbuf_release(&piecebuf);
 581         return rfc2047;
 582 }
 583
 584 static void decode_header(struct strbuf *it)
 585 {
 586         if (decode_header_bq(it))
 587                 return;
 588         /* otherwise "it" is a straight copy of the input.
 589          * This can be binary guck but there is no charset specified.
 590          */
 591         if (metainfo_charset)
 592                 convert_to_utf8(it, "");
 593 }
 594
 595 static void decode_transfer_encoding(struct strbuf *line)
 596 {
 597         struct strbuf *ret;
 598
 599         switch (transfer_encoding) {
 600         case TE_QP:
 601                 ret = decode_q_segment(line, 0);
 602                 break;
 603         case TE_BASE64:
 604                 ret = decode_b_segment(line);
 605                 break;
 606         case TE_DONTCARE:
 607         default:
 608                 return;
 609         }
 610         strbuf_reset(line);
 611         strbuf_addbuf(line, ret);
 612         strbuf_release(ret);
 613         free(ret);
 614 }
 615
 616 static void handle_filter(struct strbuf *line);
 617
 618 static int find_boundary(void)
 619 {
 620         while (!strbuf_getline(&line, fin, '\n')) {
 621                 if (*content_top && is_multipart_boundary(&line))
 622                         return 1;
 623         }
 624         return 0;
 625 }
 626
 627 static int handle_boundary(void)
 628 {
 629         struct strbuf newline = STRBUF_INIT;
 630
 631         strbuf_addch(&newline, '\n');
 632 again:
 633         if (line.len >= (*content_top)->len + 2 &&
 634             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 635                 /* we hit an end boundary */
 636                 /* pop the current boundary off the stack */
 637                 strbuf_release(*content_top);
 638                 free(*content_top);
 639                 *content_top = NULL;
 640
 641                 /* technically won't happen as is_multipart_boundary()
 642                    will fail first.  But just in case..
 643                  */
 644                 if (--content_top < content) {
 645                         fprintf(stderr, "Detected mismatched boundaries, "
 646                                         "can't recover\n");
 647                         exit(1);
 648                 }
 649                 handle_filter(&newline);
 650                 strbuf_release(&newline);
 651
 652                 /* skip to the next boundary */
 653                 if (!find_boundary())
 654                         return 0;
 655                 goto again;
 656         }
 657
 658         /* set some defaults */
 659         transfer_encoding = TE_DONTCARE;
 660         strbuf_reset(&charset);
 661         message_type = TYPE_TEXT;
 662
 663         /* slurp in this section's info */
 664         while (read_one_header_line(&line, fin))
 665                 check_header(&line, p_hdr_data, 0);
 666
 667         strbuf_release(&newline);
 668         /* replenish line */
 669         if (strbuf_getline(&line, fin, '\n'))
 670                 return 0;
 671         strbuf_addch(&line, '\n');
 672         return 1;
 673 }
 674
 675 static inline int patchbreak(const struct strbuf *line)
 676 {
 677         size_t i;
 678
 679         /* Beginning of a "diff -" header? */
 680         if (!prefixcmp(line->buf, "diff -"))
 681                 return 1;
 682
 683         /* CVS "Index: " line? */
 684         if (!prefixcmp(line->buf, "Index: "))
 685                 return 1;
 686
 687         /*
 688          * "--- <filename>" starts patches without headers
 689          * "---<sp>*" is a manual separator
 690          */
 691         if (line->len < 4)
 692                 return 0;
 693
 694         if (!prefixcmp(line->buf, "---")) {
 695                 /* space followed by a filename? */
 696                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 697                         return 1;
 698                 /* Just whitespace? */
 699                 for (i = 3; i < line->len; i++) {
 700                         unsigned char c = line->buf[i];
 701                         if (c == '\n')
 702                                 return 1;
 703                         if (!isspace(c))
 704                                 break;
 705                 }
 706                 return 0;
 707         }
 708         return 0;
 709 }
 710
 711 static int is_scissors_line(const struct strbuf *line)
 712 {
 713         size_t i, len = line->len;
 714         int scissors = 0, gap = 0;
 715         int first_nonblank = -1;
 716         int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
 717         const char *buf = line->buf;
 718
 719         for (i = 0; i < len; i++) {
 720                 if (isspace(buf[i])) {
 721                         if (in_perforation) {
 722                                 perforation++;
 723                                 gap++;
 724                         }
 725                         continue;
 726                 }
 727                 last_nonblank = i;
 728                 if (first_nonblank < 0)
 729                         first_nonblank = i;
 730                 if (buf[i] == '-') {
 731                         in_perforation = 1;
 732                         perforation++;
 733                         continue;
 734                 }
 735                 if (i + 1 < len &&
 736                     (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
 737                      !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
 738                         in_perforation = 1;
 739                         perforation += 2;
 740                         scissors += 2;
 741                         i++;
 742                         continue;
 743                 }
 744                 in_perforation = 0;
 745         }
 746
 747         /*
 748          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 749          * Even though there can be arbitrary cruft on the same line
 750          * (e.g. "cut here"), in order to avoid misidentification, the
 751          * perforation must occupy more than a third of the visible
 752          * width of the line, and dashes and scissors must occupy more
 753          * than half of the perforation.
 754          */
 755
 756         visible = last_nonblank - first_nonblank + 1;
 757         return (scissors && 8 <= visible &&
 758                 visible < perforation * 3 &&
 759                 gap * 2 < perforation);
 760 }
 761
 762 static int handle_commit_msg(struct strbuf *line)
 763 {
 764         static int still_looking = 1;
 765
 766         if (!cmitmsg)
 767                 return 0;
 768
 769         if (still_looking) {
 770                 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
 771                         return 0;
 772         }
 773
 774         if (use_inbody_headers && still_looking) {
 775                 still_looking = check_header(line, s_hdr_data, 0);
 776                 if (still_looking)
 777                         return 0;
 778         } else
 779                 /* Only trim the first (blank) line of the commit message
 780                  * when ignoring in-body headers.
 781                  */
 782                 still_looking = 0;
 783
 784         /* normalize the log message to UTF-8. */
 785         if (metainfo_charset)
 786                 convert_to_utf8(line, charset.buf);
 787
 788         if (use_scissors && is_scissors_line(line)) {
 789                 int i;
 790                 if (fseek(cmitmsg, 0L, SEEK_SET))
 791                         die_errno("Could not rewind output message file");
 792                 if (ftruncate(fileno(cmitmsg), 0))
 793                         die_errno("Could not truncate output message file at scissors");
 794                 still_looking = 1;
 795
 796                 /*
 797                  * We may have already read "secondary headers"; purge
 798                  * them to give ourselves a clean restart.
 799                  */
 800                 for (i = 0; header[i]; i++) {
 801                         if (s_hdr_data[i])
 802                                 strbuf_release(s_hdr_data[i]);
 803                         s_hdr_data[i] = NULL;
 804                 }
 805                 return 0;
 806         }
 807
 808         if (patchbreak(line)) {
 809                 fclose(cmitmsg);
 810                 cmitmsg = NULL;
 811                 return 1;
 812         }
 813
 814         fputs(line->buf, cmitmsg);
 815         return 0;
 816 }
 817
 818 static void handle_patch(const struct strbuf *line)
 819 {
 820         fwrite(line->buf, 1, line->len, patchfile);
 821         patch_lines++;
 822 }
 823
 824 static void handle_filter(struct strbuf *line)
 825 {
 826         static int filter = 0;
 827
 828         /* filter tells us which part we left off on */
 829         switch (filter) {
 830         case 0:
 831                 if (!handle_commit_msg(line))
 832                         break;
 833                 filter++;
 834         case 1:
 835                 handle_patch(line);
 836                 break;
 837         }
 838 }
 839
 840 static void handle_body(void)
 841 {
 842         struct strbuf prev = STRBUF_INIT;
 843
 844         /* Skip up to the first boundary */
 845         if (*content_top) {
 846                 if (!find_boundary())
 847                         goto handle_body_out;
 848         }
 849
 850         do {
 851                 /* process any boundary lines */
 852                 if (*content_top && is_multipart_boundary(&line)) {
 853                         /* flush any leftover */
 854                         if (prev.len) {
 855                                 handle_filter(&prev);
 856                                 strbuf_reset(&prev);
 857                         }
 858                         if (!handle_boundary())
 859                                 goto handle_body_out;
 860                 }
 861
 862                 /* Unwrap transfer encoding */
 863                 decode_transfer_encoding(&line);
 864
 865                 switch (transfer_encoding) {
 866                 case TE_BASE64:
 867                 case TE_QP:
 868                 {
 869                         struct strbuf **lines, **it, *sb;
 870
 871                         /* Prepend any previous partial lines */
 872                         strbuf_insert(&line, 0, prev.buf, prev.len);
 873                         strbuf_reset(&prev);
 874
 875                         /* binary data most likely doesn't have newlines */
 876                         if (message_type != TYPE_TEXT) {
 877                                 handle_filter(&line);
 878                                 break;
 879                         }
 880                         /*
 881                          * This is a decoded line that may contain
 882                          * multiple new lines.  Pass only one chunk
 883                          * at a time to handle_filter()
 884                          */
 885                         lines = strbuf_split(&line, '\n');
 886                         for (it = lines; (sb = *it); it++) {
 887                                 if (*(it + 1) == NULL) /* The last line */
 888                                         if (sb->buf[sb->len - 1] != '\n') {
 889                                                 /* Partial line, save it for later. */
 890                                                 strbuf_addbuf(&prev, sb);
 891                                                 break;
 892                                         }
 893                                 handle_filter(sb);
 894                         }
 895                         /*
 896                          * The partial chunk is saved in "prev" and will be
 897                          * appended by the next iteration of read_line_with_nul().
 898                          */
 899                         strbuf_list_free(lines);
 900                         break;
 901                 }
 902                 default:
 903                         handle_filter(&line);
 904                 }
 905
 906         } while (!strbuf_getwholeline(&line, fin, '\n'));
 907
 908 handle_body_out:
 909         strbuf_release(&prev);
 910 }
 911
 912 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 913 {
 914         const char *sp = data->buf;
 915         while (1) {
 916                 char *ep = strchr(sp, '\n');
 917                 int len;
 918                 if (!ep)
 919                         len = strlen(sp);
 920                 else
 921                         len = ep - sp;
 922                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 923                 if (!ep)
 924                         break;
 925                 sp = ep + 1;
 926         }
 927 }
 928
 929 static void handle_info(void)
 930 {
 931         struct strbuf *hdr;
 932         int i;
 933
 934         for (i = 0; header[i]; i++) {
 935                 /* only print inbody headers if we output a patch file */
 936                 if (patch_lines && s_hdr_data[i])
 937                         hdr = s_hdr_data[i];
 938                 else if (p_hdr_data[i])
 939                         hdr = p_hdr_data[i];
 940                 else
 941                         continue;
 942
 943                 if (!memcmp(header[i], "Subject", 7)) {
 944                         if (!keep_subject) {
 945                                 cleanup_subject(hdr);
 946                                 cleanup_space(hdr);
 947                         }
 948                         output_header_lines(fout, "Subject", hdr);
 949                 } else if (!memcmp(header[i], "From", 4)) {
 950                         cleanup_space(hdr);
 951                         handle_from(hdr);
 952                         fprintf(fout, "Author: %s\n", name.buf);
 953                         fprintf(fout, "Email: %s\n", email.buf);
 954                 } else {
 955                         cleanup_space(hdr);
 956                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 957                 }
 958         }
 959         fprintf(fout, "\n");
 960 }
 961
 962 static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
 963 {
 964         int peek;
 965         fin = in;
 966         fout = out;
 967
 968         cmitmsg = fopen(msg, "w");
 969         if (!cmitmsg) {
 970                 perror(msg);
 971                 return -1;
 972         }
 973         patchfile = fopen(patch, "w");
 974         if (!patchfile) {
 975                 perror(patch);
 976                 fclose(cmitmsg);
 977                 return -1;
 978         }
 979
 980         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
 981         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
 982
 983         do {
 984                 peek = fgetc(in);
 985         } while (isspace(peek));
 986         ungetc(peek, in);
 987
 988         /* process the email header */
 989         while (read_one_header_line(&line, fin))
 990                 check_header(&line, p_hdr_data, 1);
 991
 992         handle_body();
 993         handle_info();
 994
 995         return 0;
 996 }
 997
 998 static int git_mailinfo_config(const char *var, const char *value, void *unused)
 999 {
1000         if (prefixcmp(var, "mailinfo."))
1001                 return git_default_config(var, value, unused);
1002         if (!strcmp(var, "mailinfo.scissors")) {
1003                 use_scissors = git_config_bool(var, value);
1004                 return 0;
1005         }
1006         /* perhaps others here */
1007         return 0;
1008 }
1009
1010 static const char mailinfo_usage[] =
1011         "git mailinfo [-k|-b] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] msg patch < mail >info";
1012
1013 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1014 {
1015         const char *def_charset;
1016
1017         /* NEEDSWORK: might want to do the optional .git/ directory
1018          * discovery
1019          */
1020         git_config(git_mailinfo_config, NULL);
1021
1022         def_charset = get_commit_output_encoding();
1023         metainfo_charset = def_charset;
1024
1025         while (1 < argc && argv[1][0] == '-') {
1026                 if (!strcmp(argv[1], "-k"))
1027                         keep_subject = 1;
1028                 else if (!strcmp(argv[1], "-b"))
1029                         keep_non_patch_brackets_in_subject = 1;
1030                 else if (!strcmp(argv[1], "-u"))
1031                         metainfo_charset = def_charset;
1032                 else if (!strcmp(argv[1], "-n"))
1033                         metainfo_charset = NULL;
1034                 else if (!prefixcmp(argv[1], "--encoding="))
1035                         metainfo_charset = argv[1] + 11;
1036                 else if (!strcmp(argv[1], "--scissors"))
1037                         use_scissors = 1;
1038                 else if (!strcmp(argv[1], "--no-scissors"))
1039                         use_scissors = 0;
1040                 else if (!strcmp(argv[1], "--no-inbody-headers"))
1041                         use_inbody_headers = 0;
1042                 else
1043                         usage(mailinfo_usage);
1044                 argc--; argv++;
1045         }
1046
1047         if (argc != 3)
1048                 usage(mailinfo_usage);
1049
1050         return !!mailinfo(stdin, stdout, argv[1], argv[2]);
1051 }