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 const char *metainfo_charset;
  14 static struct strbuf line = STRBUF_INIT;
  15 static struct strbuf name = STRBUF_INIT;
  16 static struct strbuf email = STRBUF_INIT;
  17
  18 static enum  {
  19         TE_DONTCARE, TE_QP, TE_BASE64,
  20 } transfer_encoding;
  21 static enum  {
  22         TYPE_TEXT, TYPE_OTHER,
  23 } message_type;
  24
  25 static struct strbuf charset = STRBUF_INIT;
  26 static int patch_lines;
  27 static struct strbuf **p_hdr_data, **s_hdr_data;
  28 static int use_scissors;
  29 static int use_inbody_headers = 1;
  30
  31 #define MAX_HDR_PARSED 10
  32 #define MAX_BOUNDARIES 5
  33
  34 static void cleanup_space(struct strbuf *sb);
  35
  36
  37 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  38 {
  39         struct strbuf *src = name;
  40         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  41                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  42                 src = email;
  43         else if (name == out)
  44                 return;
  45         strbuf_reset(out);
  46         strbuf_addbuf(out, src);
  47 }
  48
  49 static void parse_bogus_from(const struct strbuf *line)
  50 {
  51         /* John Doe <johndoe> */
  52
  53         char *bra, *ket;
  54         /* This is fallback, so do not bother if we already have an
  55          * e-mail address.
  56          */
  57         if (email.len)
  58                 return;
  59
  60         bra = strchr(line->buf, '<');
  61         if (!bra)
  62                 return;
  63         ket = strchr(bra, '>');
  64         if (!ket)
  65                 return;
  66
  67         strbuf_reset(&email);
  68         strbuf_add(&email, bra + 1, ket - bra - 1);
  69
  70         strbuf_reset(&name);
  71         strbuf_add(&name, line->buf, bra - line->buf);
  72         strbuf_trim(&name);
  73         get_sane_name(&name, &name, &email);
  74 }
  75
  76 static void handle_from(const struct strbuf *from)
  77 {
  78         char *at;
  79         size_t el;
  80         struct strbuf f;
  81
  82         strbuf_init(&f, from->len);
  83         strbuf_addbuf(&f, from);
  84
  85         at = strchr(f.buf, '@');
  86         if (!at) {
  87                 parse_bogus_from(from);
  88                 return;
  89         }
  90
  91         /*
  92          * If we already have one email, don't take any confusing lines
  93          */
  94         if (email.len && strchr(at + 1, '@')) {
  95                 strbuf_release(&f);
  96                 return;
  97         }
  98
  99         /* Pick up the string around '@', possibly delimited with <>
 100          * pair; that is the email part.
 101          */
 102         while (at > f.buf) {
 103                 char c = at[-1];
 104                 if (isspace(c))
 105                         break;
 106                 if (c == '<') {
 107                         at[-1] = ' ';
 108                         break;
 109                 }
 110                 at--;
 111         }
 112         el = strcspn(at, " \n\t\r\v\f>");
 113         strbuf_reset(&email);
 114         strbuf_add(&email, at, el);
 115         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 116
 117         /* The remainder is name.  It could be
 118          *
 119          * - "John Doe <john.doe@xz>"                   (a), or
 120          * - "john.doe@xz (John Doe)"                   (b), or
 121          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 122          *
 123          * but we have removed the email part, so
 124          *
 125          * - remove extra spaces which could stay after email (case 'c'), and
 126          * - trim from both ends, possibly removing the () pair at the end
 127          *   (cases 'a' and 'b').
 128          */
 129         cleanup_space(&f);
 130         strbuf_trim(&f);
 131         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 132                 strbuf_remove(&f, 0, 1);
 133                 strbuf_setlen(&f, f.len - 1);
 134         }
 135
 136         get_sane_name(&name, &f, &email);
 137         strbuf_release(&f);
 138 }
 139
 140 static void handle_header(struct strbuf **out, const struct strbuf *line)
 141 {
 142         if (!*out) {
 143                 *out = xmalloc(sizeof(struct strbuf));
 144                 strbuf_init(*out, line->len);
 145         } else
 146                 strbuf_reset(*out);
 147
 148         strbuf_addbuf(*out, line);
 149 }
 150
 151 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 152  * to have enough heuristics to grok MIME encoded patches often found
 153  * on our mailing lists.  For example, we do not even treat header lines
 154  * case insensitively.
 155  */
 156
 157 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 158 {
 159         const char *ends, *ap = strcasestr(line, name);
 160         size_t sz;
 161
 162         if (!ap) {
 163                 strbuf_setlen(attr, 0);
 164                 return 0;
 165         }
 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         char *pos;
 225         size_t remove;
 226         while (subject->len) {
 227                 switch (*subject->buf) {
 228                 case 'r': case 'R':
 229                         if (subject->len <= 3)
 230                                 break;
 231                         if (!memcmp(subject->buf + 1, "e:", 2)) {
 232                                 strbuf_remove(subject, 0, 3);
 233                                 continue;
 234                         }
 235                         break;
 236                 case ' ': case '\t': case ':':
 237                         strbuf_remove(subject, 0, 1);
 238                         continue;
 239                 case '[':
 240                         if ((pos = strchr(subject->buf, ']'))) {
 241                                 remove = pos - subject->buf;
 242                                 if (remove <= (subject->len - remove) * 2) {
 243                                         strbuf_remove(subject, 0, remove + 1);
 244                                         continue;
 245                                 }
 246                         } else
 247                                 strbuf_remove(subject, 0, 1);
 248                         break;
 249                 }
 250                 strbuf_trim(subject);
 251                 return;
 252         }
 253 }
 254
 255 static void cleanup_space(struct strbuf *sb)
 256 {
 257         size_t pos, cnt;
 258         for (pos = 0; pos < sb->len; pos++) {
 259                 if (isspace(sb->buf[pos])) {
 260                         sb->buf[pos] = ' ';
 261                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 262                         strbuf_remove(sb, pos + 1, cnt);
 263                 }
 264         }
 265 }
 266
 267 static void decode_header(struct strbuf *line);
 268 static const char *header[MAX_HDR_PARSED] = {
 269         "From","Subject","Date",
 270 };
 271
 272 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 273 {
 274         int len = strlen(hdr);
 275         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 276                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 277 }
 278
 279 static int check_header(const struct strbuf *line,
 280                                 struct strbuf *hdr_data[], int overwrite)
 281 {
 282         int i, ret = 0, len;
 283         struct strbuf sb = STRBUF_INIT;
 284         /* search for the interesting parts */
 285         for (i = 0; header[i]; i++) {
 286                 int len = strlen(header[i]);
 287                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 288                         /* Unwrap inline B and Q encoding, and optionally
 289                          * normalize the meta information to utf8.
 290                          */
 291                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 292                         decode_header(&sb);
 293                         handle_header(&hdr_data[i], &sb);
 294                         ret = 1;
 295                         goto check_header_out;
 296                 }
 297         }
 298
 299         /* Content stuff */
 300         if (cmp_header(line, "Content-Type")) {
 301                 len = strlen("Content-Type: ");
 302                 strbuf_add(&sb, line->buf + len, line->len - len);
 303                 decode_header(&sb);
 304                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 305                 handle_content_type(&sb);
 306                 ret = 1;
 307                 goto check_header_out;
 308         }
 309         if (cmp_header(line, "Content-Transfer-Encoding")) {
 310                 len = strlen("Content-Transfer-Encoding: ");
 311                 strbuf_add(&sb, line->buf + len, line->len - len);
 312                 decode_header(&sb);
 313                 handle_content_transfer_encoding(&sb);
 314                 ret = 1;
 315                 goto check_header_out;
 316         }
 317
 318         /* for inbody stuff */
 319         if (!prefixcmp(line->buf, ">From") && isspace(line->buf[5])) {
 320                 ret = 1; /* Should this return 0? */
 321                 goto check_header_out;
 322         }
 323         if (!prefixcmp(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 324                 for (i = 0; header[i]; i++) {
 325                         if (!memcmp("Subject", header[i], 7)) {
 326                                 handle_header(&hdr_data[i], line);
 327                                 ret = 1;
 328                                 goto check_header_out;
 329                         }
 330                 }
 331         }
 332
 333 check_header_out:
 334         strbuf_release(&sb);
 335         return ret;
 336 }
 337
 338 static int is_rfc2822_header(const struct strbuf *line)
 339 {
 340         /*
 341          * The section that defines the loosest possible
 342          * field name is "3.6.8 Optional fields".
 343          *
 344          * optional-field = field-name ":" unstructured CRLF
 345          * field-name = 1*ftext
 346          * ftext = %d33-57 / %59-126
 347          */
 348         int ch;
 349         char *cp = line->buf;
 350
 351         /* Count mbox From headers as headers */
 352         if (!prefixcmp(cp, "From ") || !prefixcmp(cp, ">From "))
 353                 return 1;
 354
 355         while ((ch = *cp++)) {
 356                 if (ch == ':')
 357                         return 1;
 358                 if ((33 <= ch && ch <= 57) ||
 359                     (59 <= ch && ch <= 126))
 360                         continue;
 361                 break;
 362         }
 363         return 0;
 364 }
 365
 366 static int read_one_header_line(struct strbuf *line, FILE *in)
 367 {
 368         /* Get the first part of the line. */
 369         if (strbuf_getline(line, in, '\n'))
 370                 return 0;
 371
 372         /*
 373          * Is it an empty line or not a valid rfc2822 header?
 374          * If so, stop here, and return false ("not a header")
 375          */
 376         strbuf_rtrim(line);
 377         if (!line->len || !is_rfc2822_header(line)) {
 378                 /* Re-add the newline */
 379                 strbuf_addch(line, '\n');
 380                 return 0;
 381         }
 382
 383         /*
 384          * Now we need to eat all the continuation lines..
 385          * Yuck, 2822 header "folding"
 386          */
 387         for (;;) {
 388                 int peek;
 389                 struct strbuf continuation = STRBUF_INIT;
 390
 391                 peek = fgetc(in); ungetc(peek, in);
 392                 if (peek != ' ' && peek != '\t')
 393                         break;
 394                 if (strbuf_getline(&continuation, in, '\n'))
 395                         break;
 396                 continuation.buf[0] = '\n';
 397                 strbuf_rtrim(&continuation);
 398                 strbuf_addbuf(line, &continuation);
 399         }
 400
 401         return 1;
 402 }
 403
 404 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 405 {
 406         const char *in = q_seg->buf;
 407         int c;
 408         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 409         strbuf_init(out, q_seg->len);
 410
 411         while ((c = *in++) != 0) {
 412                 if (c == '=') {
 413                         int d = *in++;
 414                         if (d == '\n' || !d)
 415                                 break; /* drop trailing newline */
 416                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 417                         continue;
 418                 }
 419                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 420                         c = 0x20;
 421                 strbuf_addch(out, c);
 422         }
 423         return out;
 424 }
 425
 426 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 427 {
 428         /* Decode in..ep, possibly in-place to ot */
 429         int c, pos = 0, acc = 0;
 430         const char *in = b_seg->buf;
 431         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 432         strbuf_init(out, b_seg->len);
 433
 434         while ((c = *in++) != 0) {
 435                 if (c == '+')
 436                         c = 62;
 437                 else if (c == '/')
 438                         c = 63;
 439                 else if ('A' <= c && c <= 'Z')
 440                         c -= 'A';
 441                 else if ('a' <= c && c <= 'z')
 442                         c -= 'a' - 26;
 443                 else if ('0' <= c && c <= '9')
 444                         c -= '0' - 52;
 445                 else
 446                         continue; /* garbage */
 447                 switch (pos++) {
 448                 case 0:
 449                         acc = (c << 2);
 450                         break;
 451                 case 1:
 452                         strbuf_addch(out, (acc | (c >> 4)));
 453                         acc = (c & 15) << 4;
 454                         break;
 455                 case 2:
 456                         strbuf_addch(out, (acc | (c >> 2)));
 457                         acc = (c & 3) << 6;
 458                         break;
 459                 case 3:
 460                         strbuf_addch(out, (acc | c));
 461                         acc = pos = 0;
 462                         break;
 463                 }
 464         }
 465         return out;
 466 }
 467
 468 /*
 469  * When there is no known charset, guess.
 470  *
 471  * Right now we assume that if the target is UTF-8 (the default),
 472  * and it already looks like UTF-8 (which includes US-ASCII as its
 473  * subset, of course) then that is what it is and there is nothing
 474  * to do.
 475  *
 476  * Otherwise, we default to assuming it is Latin1 for historical
 477  * reasons.
 478  */
 479 static const char *guess_charset(const struct strbuf *line, const char *target_charset)
 480 {
 481         if (is_encoding_utf8(target_charset)) {
 482                 if (is_utf8(line->buf))
 483                         return NULL;
 484         }
 485         return "ISO8859-1";
 486 }
 487
 488 static void convert_to_utf8(struct strbuf *line, const char *charset)
 489 {
 490         char *out;
 491
 492         if (!charset || !*charset) {
 493                 charset = guess_charset(line, metainfo_charset);
 494                 if (!charset)
 495                         return;
 496         }
 497
 498         if (!strcasecmp(metainfo_charset, charset))
 499                 return;
 500         out = reencode_string(line->buf, metainfo_charset, charset);
 501         if (!out)
 502                 die("cannot convert from %s to %s",
 503                     charset, metainfo_charset);
 504         strbuf_attach(line, out, strlen(out), strlen(out));
 505 }
 506
 507 static int decode_header_bq(struct strbuf *it)
 508 {
 509         char *in, *ep, *cp;
 510         struct strbuf outbuf = STRBUF_INIT, *dec;
 511         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 512         int rfc2047 = 0;
 513
 514         in = it->buf;
 515         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 516                 int encoding;
 517                 strbuf_reset(&charset_q);
 518                 strbuf_reset(&piecebuf);
 519                 rfc2047 = 1;
 520
 521                 if (in != ep) {
 522                         /*
 523                          * We are about to process an encoded-word
 524                          * that begins at ep, but there is something
 525                          * before the encoded word.
 526                          */
 527                         char *scan;
 528                         for (scan = in; scan < ep; scan++)
 529                                 if (!isspace(*scan))
 530                                         break;
 531
 532                         if (scan != ep || in == it->buf) {
 533                                 /*
 534                                  * We should not lose that "something",
 535                                  * unless we have just processed an
 536                                  * encoded-word, and there is only LWS
 537                                  * before the one we are about to process.
 538                                  */
 539                                 strbuf_add(&outbuf, in, ep - in);
 540                         }
 541                 }
 542                 /* E.g.
 543                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 544                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 545                  */
 546                 ep += 2;
 547
 548                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 549                         goto decode_header_bq_out;
 550
 551                 if (cp + 3 - it->buf > it->len)
 552                         goto decode_header_bq_out;
 553                 strbuf_add(&charset_q, ep, cp - ep);
 554
 555                 encoding = cp[1];
 556                 if (!encoding || cp[2] != '?')
 557                         goto decode_header_bq_out;
 558                 ep = strstr(cp + 3, "?=");
 559                 if (!ep)
 560                         goto decode_header_bq_out;
 561                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 562                 switch (tolower(encoding)) {
 563                 default:
 564                         goto decode_header_bq_out;
 565                 case 'b':
 566                         dec = decode_b_segment(&piecebuf);
 567                         break;
 568                 case 'q':
 569                         dec = decode_q_segment(&piecebuf, 1);
 570                         break;
 571                 }
 572                 if (metainfo_charset)
 573                         convert_to_utf8(dec, charset_q.buf);
 574
 575                 strbuf_addbuf(&outbuf, dec);
 576                 strbuf_release(dec);
 577                 free(dec);
 578                 in = ep + 2;
 579         }
 580         strbuf_addstr(&outbuf, in);
 581         strbuf_reset(it);
 582         strbuf_addbuf(it, &outbuf);
 583 decode_header_bq_out:
 584         strbuf_release(&outbuf);
 585         strbuf_release(&charset_q);
 586         strbuf_release(&piecebuf);
 587         return rfc2047;
 588 }
 589
 590 static void decode_header(struct strbuf *it)
 591 {
 592         if (decode_header_bq(it))
 593                 return;
 594         /* otherwise "it" is a straight copy of the input.
 595          * This can be binary guck but there is no charset specified.
 596          */
 597         if (metainfo_charset)
 598                 convert_to_utf8(it, "");
 599 }
 600
 601 static void decode_transfer_encoding(struct strbuf *line)
 602 {
 603         struct strbuf *ret;
 604
 605         switch (transfer_encoding) {
 606         case TE_QP:
 607                 ret = decode_q_segment(line, 0);
 608                 break;
 609         case TE_BASE64:
 610                 ret = decode_b_segment(line);
 611                 break;
 612         case TE_DONTCARE:
 613         default:
 614                 return;
 615         }
 616         strbuf_reset(line);
 617         strbuf_addbuf(line, ret);
 618         strbuf_release(ret);
 619         free(ret);
 620 }
 621
 622 static void handle_filter(struct strbuf *line);
 623
 624 static int find_boundary(void)
 625 {
 626         while (!strbuf_getline(&line, fin, '\n')) {
 627                 if (*content_top && is_multipart_boundary(&line))
 628                         return 1;
 629         }
 630         return 0;
 631 }
 632
 633 static int handle_boundary(void)
 634 {
 635         struct strbuf newline = STRBUF_INIT;
 636
 637         strbuf_addch(&newline, '\n');
 638 again:
 639         if (line.len >= (*content_top)->len + 2 &&
 640             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 641                 /* we hit an end boundary */
 642                 /* pop the current boundary off the stack */
 643                 strbuf_release(*content_top);
 644                 free(*content_top);
 645                 *content_top = NULL;
 646
 647                 /* technically won't happen as is_multipart_boundary()
 648                    will fail first.  But just in case..
 649                  */
 650                 if (--content_top < content) {
 651                         fprintf(stderr, "Detected mismatched boundaries, "
 652                                         "can't recover\n");
 653                         exit(1);
 654                 }
 655                 handle_filter(&newline);
 656                 strbuf_release(&newline);
 657
 658                 /* skip to the next boundary */
 659                 if (!find_boundary())
 660                         return 0;
 661                 goto again;
 662         }
 663
 664         /* set some defaults */
 665         transfer_encoding = TE_DONTCARE;
 666         strbuf_reset(&charset);
 667         message_type = TYPE_TEXT;
 668
 669         /* slurp in this section's info */
 670         while (read_one_header_line(&line, fin))
 671                 check_header(&line, p_hdr_data, 0);
 672
 673         strbuf_release(&newline);
 674         /* replenish line */
 675         if (strbuf_getline(&line, fin, '\n'))
 676                 return 0;
 677         strbuf_addch(&line, '\n');
 678         return 1;
 679 }
 680
 681 static inline int patchbreak(const struct strbuf *line)
 682 {
 683         size_t i;
 684
 685         /* Beginning of a "diff -" header? */
 686         if (!prefixcmp(line->buf, "diff -"))
 687                 return 1;
 688
 689         /* CVS "Index: " line? */
 690         if (!prefixcmp(line->buf, "Index: "))
 691                 return 1;
 692
 693         /*
 694          * "--- <filename>" starts patches without headers
 695          * "---<sp>*" is a manual separator
 696          */
 697         if (line->len < 4)
 698                 return 0;
 699
 700         if (!prefixcmp(line->buf, "---")) {
 701                 /* space followed by a filename? */
 702                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 703                         return 1;
 704                 /* Just whitespace? */
 705                 for (i = 3; i < line->len; i++) {
 706                         unsigned char c = line->buf[i];
 707                         if (c == '\n')
 708                                 return 1;
 709                         if (!isspace(c))
 710                                 break;
 711                 }
 712                 return 0;
 713         }
 714         return 0;
 715 }
 716
 717 static int is_scissors_line(const struct strbuf *line)
 718 {
 719         size_t i, len = line->len;
 720         int scissors = 0, gap = 0;
 721         int first_nonblank = -1;
 722         int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
 723         const char *buf = line->buf;
 724
 725         for (i = 0; i < len; i++) {
 726                 if (isspace(buf[i])) {
 727                         if (in_perforation) {
 728                                 perforation++;
 729                                 gap++;
 730                         }
 731                         continue;
 732                 }
 733                 last_nonblank = i;
 734                 if (first_nonblank < 0)
 735                         first_nonblank = i;
 736                 if (buf[i] == '-') {
 737                         in_perforation = 1;
 738                         perforation++;
 739                         continue;
 740                 }
 741                 if (i + 1 < len &&
 742                     (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2))) {
 743                         in_perforation = 1;
 744                         perforation += 2;
 745                         scissors += 2;
 746                         i++;
 747                         continue;
 748                 }
 749                 in_perforation = 0;
 750         }
 751
 752         /*
 753          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 754          * Even though there can be arbitrary cruft on the same line
 755          * (e.g. "cut here"), in order to avoid misidentification, the
 756          * perforation must occupy more than a third of the visible
 757          * width of the line, and dashes and scissors must occupy more
 758          * than half of the perforation.
 759          */
 760
 761         visible = last_nonblank - first_nonblank + 1;
 762         return (scissors && 8 <= visible &&
 763                 visible < perforation * 3 &&
 764                 gap * 2 < perforation);
 765 }
 766
 767 static int handle_commit_msg(struct strbuf *line)
 768 {
 769         static int still_looking = 1;
 770
 771         if (!cmitmsg)
 772                 return 0;
 773
 774         if (still_looking) {
 775                 strbuf_ltrim(line);
 776                 if (!line->len)
 777                         return 0;
 778         }
 779
 780         if (use_inbody_headers && still_looking) {
 781                 still_looking = check_header(line, s_hdr_data, 0);
 782                 if (still_looking)
 783                         return 0;
 784         } else
 785                 /* Only trim the first (blank) line of the commit message
 786                  * when ignoring in-body headers.
 787                  */
 788                 still_looking = 0;
 789
 790         /* normalize the log message to UTF-8. */
 791         if (metainfo_charset)
 792                 convert_to_utf8(line, charset.buf);
 793
 794         if (use_scissors && is_scissors_line(line)) {
 795                 int i;
 796                 if (fseek(cmitmsg, 0L, SEEK_SET))
 797                         die_errno("Could not rewind output message file");
 798                 if (ftruncate(fileno(cmitmsg), 0))
 799                         die_errno("Could not truncate output message file at scissors");
 800                 still_looking = 1;
 801
 802                 /*
 803                  * We may have already read "secondary headers"; purge
 804                  * them to give ourselves a clean restart.
 805                  */
 806                 for (i = 0; header[i]; i++) {
 807                         if (s_hdr_data[i])
 808                                 strbuf_release(s_hdr_data[i]);
 809                         s_hdr_data[i] = NULL;
 810                 }
 811                 return 0;
 812         }
 813
 814         if (patchbreak(line)) {
 815                 fclose(cmitmsg);
 816                 cmitmsg = NULL;
 817                 return 1;
 818         }
 819
 820         fputs(line->buf, cmitmsg);
 821         return 0;
 822 }
 823
 824 static void handle_patch(const struct strbuf *line)
 825 {
 826         fwrite(line->buf, 1, line->len, patchfile);
 827         patch_lines++;
 828 }
 829
 830 static void handle_filter(struct strbuf *line)
 831 {
 832         static int filter = 0;
 833
 834         /* filter tells us which part we left off on */
 835         switch (filter) {
 836         case 0:
 837                 if (!handle_commit_msg(line))
 838                         break;
 839                 filter++;
 840         case 1:
 841                 handle_patch(line);
 842                 break;
 843         }
 844 }
 845
 846 static void handle_body(void)
 847 {
 848         struct strbuf prev = STRBUF_INIT;
 849
 850         /* Skip up to the first boundary */
 851         if (*content_top) {
 852                 if (!find_boundary())
 853                         goto handle_body_out;
 854         }
 855
 856         do {
 857                 /* process any boundary lines */
 858                 if (*content_top && is_multipart_boundary(&line)) {
 859                         /* flush any leftover */
 860                         if (prev.len) {
 861                                 handle_filter(&prev);
 862                                 strbuf_reset(&prev);
 863                         }
 864                         if (!handle_boundary())
 865                                 goto handle_body_out;
 866                 }
 867
 868                 /* Unwrap transfer encoding */
 869                 decode_transfer_encoding(&line);
 870
 871                 switch (transfer_encoding) {
 872                 case TE_BASE64:
 873                 case TE_QP:
 874                 {
 875                         struct strbuf **lines, **it, *sb;
 876
 877                         /* Prepend any previous partial lines */
 878                         strbuf_insert(&line, 0, prev.buf, prev.len);
 879                         strbuf_reset(&prev);
 880
 881                         /* binary data most likely doesn't have newlines */
 882                         if (message_type != TYPE_TEXT) {
 883                                 handle_filter(&line);
 884                                 break;
 885                         }
 886                         /*
 887                          * This is a decoded line that may contain
 888                          * multiple new lines.  Pass only one chunk
 889                          * at a time to handle_filter()
 890                          */
 891                         lines = strbuf_split(&line, '\n');
 892                         for (it = lines; (sb = *it); it++) {
 893                                 if (*(it + 1) == NULL) /* The last line */
 894                                         if (sb->buf[sb->len - 1] != '\n') {
 895                                                 /* Partial line, save it for later. */
 896                                                 strbuf_addbuf(&prev, sb);
 897                                                 break;
 898                                         }
 899                                 handle_filter(sb);
 900                         }
 901                         /*
 902                          * The partial chunk is saved in "prev" and will be
 903                          * appended by the next iteration of read_line_with_nul().
 904                          */
 905                         strbuf_list_free(lines);
 906                         break;
 907                 }
 908                 default:
 909                         handle_filter(&line);
 910                 }
 911
 912         } while (!strbuf_getwholeline(&line, fin, '\n'));
 913
 914 handle_body_out:
 915         strbuf_release(&prev);
 916 }
 917
 918 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 919 {
 920         const char *sp = data->buf;
 921         while (1) {
 922                 char *ep = strchr(sp, '\n');
 923                 int len;
 924                 if (!ep)
 925                         len = strlen(sp);
 926                 else
 927                         len = ep - sp;
 928                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 929                 if (!ep)
 930                         break;
 931                 sp = ep + 1;
 932         }
 933 }
 934
 935 static void handle_info(void)
 936 {
 937         struct strbuf *hdr;
 938         int i;
 939
 940         for (i = 0; header[i]; i++) {
 941                 /* only print inbody headers if we output a patch file */
 942                 if (patch_lines && s_hdr_data[i])
 943                         hdr = s_hdr_data[i];
 944                 else if (p_hdr_data[i])
 945                         hdr = p_hdr_data[i];
 946                 else
 947                         continue;
 948
 949                 if (!memcmp(header[i], "Subject", 7)) {
 950                         if (!keep_subject) {
 951                                 cleanup_subject(hdr);
 952                                 cleanup_space(hdr);
 953                         }
 954                         output_header_lines(fout, "Subject", hdr);
 955                 } else if (!memcmp(header[i], "From", 4)) {
 956                         cleanup_space(hdr);
 957                         handle_from(hdr);
 958                         fprintf(fout, "Author: %s\n", name.buf);
 959                         fprintf(fout, "Email: %s\n", email.buf);
 960                 } else {
 961                         cleanup_space(hdr);
 962                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 963                 }
 964         }
 965         fprintf(fout, "\n");
 966 }
 967
 968 static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
 969 {
 970         int peek;
 971         fin = in;
 972         fout = out;
 973
 974         cmitmsg = fopen(msg, "w");
 975         if (!cmitmsg) {
 976                 perror(msg);
 977                 return -1;
 978         }
 979         patchfile = fopen(patch, "w");
 980         if (!patchfile) {
 981                 perror(patch);
 982                 fclose(cmitmsg);
 983                 return -1;
 984         }
 985
 986         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
 987         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
 988
 989         do {
 990                 peek = fgetc(in);
 991         } while (isspace(peek));
 992         ungetc(peek, in);
 993
 994         /* process the email header */
 995         while (read_one_header_line(&line, fin))
 996                 check_header(&line, p_hdr_data, 1);
 997
 998         handle_body();
 999         handle_info();
1000
1001         return 0;
1002 }
1003
1004 static int git_mailinfo_config(const char *var, const char *value, void *unused)
1005 {
1006         if (prefixcmp(var, "mailinfo."))
1007                 return git_default_config(var, value, unused);
1008         if (!strcmp(var, "mailinfo.scissors")) {
1009                 use_scissors = git_config_bool(var, value);
1010                 return 0;
1011         }
1012         /* perhaps others here */
1013         return 0;
1014 }
1015
1016 static const char mailinfo_usage[] =
1017         "git mailinfo [-k] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] msg patch < mail >info";
1018
1019 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1020 {
1021         const char *def_charset;
1022
1023         /* NEEDSWORK: might want to do the optional .git/ directory
1024          * discovery
1025          */
1026         git_config(git_mailinfo_config, NULL);
1027
1028         def_charset = (git_commit_encoding ? git_commit_encoding : "UTF-8");
1029         metainfo_charset = def_charset;
1030
1031         while (1 < argc && argv[1][0] == '-') {
1032                 if (!strcmp(argv[1], "-k"))
1033                         keep_subject = 1;
1034                 else if (!strcmp(argv[1], "-u"))
1035                         metainfo_charset = def_charset;
1036                 else if (!strcmp(argv[1], "-n"))
1037                         metainfo_charset = NULL;
1038                 else if (!prefixcmp(argv[1], "--encoding="))
1039                         metainfo_charset = argv[1] + 11;
1040                 else if (!strcmp(argv[1], "--scissors"))
1041                         use_scissors = 1;
1042                 else if (!strcmp(argv[1], "--no-scissors"))
1043                         use_scissors = 0;
1044                 else if (!strcmp(argv[1], "--no-inbody-headers"))
1045                         use_inbody_headers = 0;
1046                 else
1047                         usage(mailinfo_usage);
1048                 argc--; argv++;
1049         }
1050
1051         if (argc != 3)
1052                 usage(mailinfo_usage);
1053
1054         return !!mailinfo(stdin, stdout, argv[1], argv[2]);
1055 }