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