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 static char *message_id;
  19
  20 static enum  {
  21         TE_DONTCARE, TE_QP, TE_BASE64
  22 } transfer_encoding;
  23
  24 static struct strbuf charset = STRBUF_INIT;
  25 static int patch_lines;
  26 static struct strbuf **p_hdr_data, **s_hdr_data;
  27 static int use_scissors;
  28 static int add_message_id;
  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         strbuf_setlen(attr, 0);
 163         if (!ap)
 164                 return 0;
 165         ap += strlen(name);
 166         if (*ap == '"') {
 167                 ap++;
 168                 ends = "\"";
 169         }
 170         else
 171                 ends = "; \t";
 172         sz = strcspn(ap, ends);
 173         strbuf_add(attr, ap, sz);
 174         return 1;
 175 }
 176
 177 static struct strbuf *content[MAX_BOUNDARIES];
 178
 179 static struct strbuf **content_top = content;
 180
 181 static void handle_content_type(struct strbuf *line)
 182 {
 183         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 184         strbuf_init(boundary, line->len);
 185
 186         if (slurp_attr(line->buf, "boundary=", boundary)) {
 187                 strbuf_insert(boundary, 0, "--", 2);
 188                 if (++content_top > &content[MAX_BOUNDARIES]) {
 189                         fprintf(stderr, "Too many boundaries to handle\n");
 190                         exit(1);
 191                 }
 192                 *content_top = boundary;
 193                 boundary = NULL;
 194         }
 195         slurp_attr(line->buf, "charset=", &charset);
 196
 197         if (boundary) {
 198                 strbuf_release(boundary);
 199                 free(boundary);
 200         }
 201 }
 202
 203 static void handle_message_id(const struct strbuf *line)
 204 {
 205         if (add_message_id)
 206                 message_id = strdup(line->buf);
 207 }
 208
 209 static void handle_content_transfer_encoding(const struct strbuf *line)
 210 {
 211         if (strcasestr(line->buf, "base64"))
 212                 transfer_encoding = TE_BASE64;
 213         else if (strcasestr(line->buf, "quoted-printable"))
 214                 transfer_encoding = TE_QP;
 215         else
 216                 transfer_encoding = TE_DONTCARE;
 217 }
 218
 219 static int is_multipart_boundary(const struct strbuf *line)
 220 {
 221         return (((*content_top)->len <= line->len) &&
 222                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
 223 }
 224
 225 static void cleanup_subject(struct strbuf *subject)
 226 {
 227         size_t at = 0;
 228
 229         while (at < subject->len) {
 230                 char *pos;
 231                 size_t remove;
 232
 233                 switch (subject->buf[at]) {
 234                 case 'r': case 'R':
 235                         if (subject->len <= at + 3)
 236                                 break;
 237                         if ((subject->buf[at + 1] == 'e' ||
 238                              subject->buf[at + 1] == 'E') &&
 239                             subject->buf[at + 2] == ':') {
 240                                 strbuf_remove(subject, at, 3);
 241                                 continue;
 242                         }
 243                         at++;
 244                         break;
 245                 case ' ': case '\t': case ':':
 246                         strbuf_remove(subject, at, 1);
 247                         continue;
 248                 case '[':
 249                         pos = strchr(subject->buf + at, ']');
 250                         if (!pos)
 251                                 break;
 252                         remove = pos - subject->buf + at + 1;
 253                         if (!keep_non_patch_brackets_in_subject ||
 254                             (7 <= remove &&
 255                              memmem(subject->buf + at, remove, "PATCH", 5)))
 256                                 strbuf_remove(subject, at, remove);
 257                         else {
 258                                 at += remove;
 259                                 /*
 260                                  * If the input had a space after the ], keep
 261                                  * it.  We don't bother with finding the end of
 262                                  * the space, since we later normalize it
 263                                  * anyway.
 264                                  */
 265                                 if (isspace(subject->buf[at]))
 266                                         at += 1;
 267                         }
 268                         continue;
 269                 }
 270                 break;
 271         }
 272         strbuf_trim(subject);
 273 }
 274
 275 static void cleanup_space(struct strbuf *sb)
 276 {
 277         size_t pos, cnt;
 278         for (pos = 0; pos < sb->len; pos++) {
 279                 if (isspace(sb->buf[pos])) {
 280                         sb->buf[pos] = ' ';
 281                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 282                         strbuf_remove(sb, pos + 1, cnt);
 283                 }
 284         }
 285 }
 286
 287 static void decode_header(struct strbuf *line);
 288 static const char *header[MAX_HDR_PARSED] = {
 289         "From","Subject","Date",
 290 };
 291
 292 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 293 {
 294         int len = strlen(hdr);
 295         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 296                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 297 }
 298
 299 static int is_format_patch_separator(const char *line, int len)
 300 {
 301         static const char SAMPLE[] =
 302                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 303         const char *cp;
 304
 305         if (len != strlen(SAMPLE))
 306                 return 0;
 307         if (!skip_prefix(line, "From ", &cp))
 308                 return 0;
 309         if (strspn(cp, "0123456789abcdef") != 40)
 310                 return 0;
 311         cp += 40;
 312         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 313 }
 314
 315 static int check_header(const struct strbuf *line,
 316                                 struct strbuf *hdr_data[], int overwrite)
 317 {
 318         int i, ret = 0, len;
 319         struct strbuf sb = STRBUF_INIT;
 320         /* search for the interesting parts */
 321         for (i = 0; header[i]; i++) {
 322                 int len = strlen(header[i]);
 323                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 324                         /* Unwrap inline B and Q encoding, and optionally
 325                          * normalize the meta information to utf8.
 326                          */
 327                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 328                         decode_header(&sb);
 329                         handle_header(&hdr_data[i], &sb);
 330                         ret = 1;
 331                         goto check_header_out;
 332                 }
 333         }
 334
 335         /* Content stuff */
 336         if (cmp_header(line, "Content-Type")) {
 337                 len = strlen("Content-Type: ");
 338                 strbuf_add(&sb, line->buf + len, line->len - len);
 339                 decode_header(&sb);
 340                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 341                 handle_content_type(&sb);
 342                 ret = 1;
 343                 goto check_header_out;
 344         }
 345         if (cmp_header(line, "Content-Transfer-Encoding")) {
 346                 len = strlen("Content-Transfer-Encoding: ");
 347                 strbuf_add(&sb, line->buf + len, line->len - len);
 348                 decode_header(&sb);
 349                 handle_content_transfer_encoding(&sb);
 350                 ret = 1;
 351                 goto check_header_out;
 352         }
 353         if (cmp_header(line, "Message-Id")) {
 354                 len = strlen("Message-Id: ");
 355                 strbuf_add(&sb, line->buf + len, line->len - len);
 356                 decode_header(&sb);
 357                 handle_message_id(&sb);
 358                 ret = 1;
 359                 goto check_header_out;
 360         }
 361
 362         /* for inbody stuff */
 363         if (starts_with(line->buf, ">From") && isspace(line->buf[5])) {
 364                 ret = is_format_patch_separator(line->buf + 1, line->len - 1);
 365                 goto check_header_out;
 366         }
 367         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 368                 for (i = 0; header[i]; i++) {
 369                         if (!strcmp("Subject", header[i])) {
 370                                 handle_header(&hdr_data[i], line);
 371                                 ret = 1;
 372                                 goto check_header_out;
 373                         }
 374                 }
 375         }
 376
 377 check_header_out:
 378         strbuf_release(&sb);
 379         return ret;
 380 }
 381
 382 static int is_rfc2822_header(const struct strbuf *line)
 383 {
 384         /*
 385          * The section that defines the loosest possible
 386          * field name is "3.6.8 Optional fields".
 387          *
 388          * optional-field = field-name ":" unstructured CRLF
 389          * field-name = 1*ftext
 390          * ftext = %d33-57 / %59-126
 391          */
 392         int ch;
 393         char *cp = line->buf;
 394
 395         /* Count mbox From headers as headers */
 396         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 397                 return 1;
 398
 399         while ((ch = *cp++)) {
 400                 if (ch == ':')
 401                         return 1;
 402                 if ((33 <= ch && ch <= 57) ||
 403                     (59 <= ch && ch <= 126))
 404                         continue;
 405                 break;
 406         }
 407         return 0;
 408 }
 409
 410 static int read_one_header_line(struct strbuf *line, FILE *in)
 411 {
 412         /* Get the first part of the line. */
 413         if (strbuf_getline(line, in, '\n'))
 414                 return 0;
 415
 416         /*
 417          * Is it an empty line or not a valid rfc2822 header?
 418          * If so, stop here, and return false ("not a header")
 419          */
 420         strbuf_rtrim(line);
 421         if (!line->len || !is_rfc2822_header(line)) {
 422                 /* Re-add the newline */
 423                 strbuf_addch(line, '\n');
 424                 return 0;
 425         }
 426
 427         /*
 428          * Now we need to eat all the continuation lines..
 429          * Yuck, 2822 header "folding"
 430          */
 431         for (;;) {
 432                 int peek;
 433                 struct strbuf continuation = STRBUF_INIT;
 434
 435                 peek = fgetc(in); ungetc(peek, in);
 436                 if (peek != ' ' && peek != '\t')
 437                         break;
 438                 if (strbuf_getline(&continuation, in, '\n'))
 439                         break;
 440                 continuation.buf[0] = ' ';
 441                 strbuf_rtrim(&continuation);
 442                 strbuf_addbuf(line, &continuation);
 443         }
 444
 445         return 1;
 446 }
 447
 448 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 449 {
 450         const char *in = q_seg->buf;
 451         int c;
 452         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 453         strbuf_init(out, q_seg->len);
 454
 455         while ((c = *in++) != 0) {
 456                 if (c == '=') {
 457                         int d = *in++;
 458                         if (d == '\n' || !d)
 459                                 break; /* drop trailing newline */
 460                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 461                         continue;
 462                 }
 463                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 464                         c = 0x20;
 465                 strbuf_addch(out, c);
 466         }
 467         return out;
 468 }
 469
 470 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 471 {
 472         /* Decode in..ep, possibly in-place to ot */
 473         int c, pos = 0, acc = 0;
 474         const char *in = b_seg->buf;
 475         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 476         strbuf_init(out, b_seg->len);
 477
 478         while ((c = *in++) != 0) {
 479                 if (c == '+')
 480                         c = 62;
 481                 else if (c == '/')
 482                         c = 63;
 483                 else if ('A' <= c && c <= 'Z')
 484                         c -= 'A';
 485                 else if ('a' <= c && c <= 'z')
 486                         c -= 'a' - 26;
 487                 else if ('0' <= c && c <= '9')
 488                         c -= '0' - 52;
 489                 else
 490                         continue; /* garbage */
 491                 switch (pos++) {
 492                 case 0:
 493                         acc = (c << 2);
 494                         break;
 495                 case 1:
 496                         strbuf_addch(out, (acc | (c >> 4)));
 497                         acc = (c & 15) << 4;
 498                         break;
 499                 case 2:
 500                         strbuf_addch(out, (acc | (c >> 2)));
 501                         acc = (c & 3) << 6;
 502                         break;
 503                 case 3:
 504                         strbuf_addch(out, (acc | c));
 505                         acc = pos = 0;
 506                         break;
 507                 }
 508         }
 509         return out;
 510 }
 511
 512 static void convert_to_utf8(struct strbuf *line, const char *charset)
 513 {
 514         char *out;
 515
 516         if (!charset || !*charset)
 517                 return;
 518
 519         if (same_encoding(metainfo_charset, charset))
 520                 return;
 521         out = reencode_string(line->buf, metainfo_charset, charset);
 522         if (!out)
 523                 die("cannot convert from %s to %s",
 524                     charset, metainfo_charset);
 525         strbuf_attach(line, out, strlen(out), strlen(out));
 526 }
 527
 528 static int decode_header_bq(struct strbuf *it)
 529 {
 530         char *in, *ep, *cp;
 531         struct strbuf outbuf = STRBUF_INIT, *dec;
 532         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 533         int rfc2047 = 0;
 534
 535         in = it->buf;
 536         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 537                 int encoding;
 538                 strbuf_reset(&charset_q);
 539                 strbuf_reset(&piecebuf);
 540                 rfc2047 = 1;
 541
 542                 if (in != ep) {
 543                         /*
 544                          * We are about to process an encoded-word
 545                          * that begins at ep, but there is something
 546                          * before the encoded word.
 547                          */
 548                         char *scan;
 549                         for (scan = in; scan < ep; scan++)
 550                                 if (!isspace(*scan))
 551                                         break;
 552
 553                         if (scan != ep || in == it->buf) {
 554                                 /*
 555                                  * We should not lose that "something",
 556                                  * unless we have just processed an
 557                                  * encoded-word, and there is only LWS
 558                                  * before the one we are about to process.
 559                                  */
 560                                 strbuf_add(&outbuf, in, ep - in);
 561                         }
 562                 }
 563                 /* E.g.
 564                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 565                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 566                  */
 567                 ep += 2;
 568
 569                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 570                         goto decode_header_bq_out;
 571
 572                 if (cp + 3 - it->buf > it->len)
 573                         goto decode_header_bq_out;
 574                 strbuf_add(&charset_q, ep, cp - ep);
 575
 576                 encoding = cp[1];
 577                 if (!encoding || cp[2] != '?')
 578                         goto decode_header_bq_out;
 579                 ep = strstr(cp + 3, "?=");
 580                 if (!ep)
 581                         goto decode_header_bq_out;
 582                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 583                 switch (tolower(encoding)) {
 584                 default:
 585                         goto decode_header_bq_out;
 586                 case 'b':
 587                         dec = decode_b_segment(&piecebuf);
 588                         break;
 589                 case 'q':
 590                         dec = decode_q_segment(&piecebuf, 1);
 591                         break;
 592                 }
 593                 if (metainfo_charset)
 594                         convert_to_utf8(dec, charset_q.buf);
 595
 596                 strbuf_addbuf(&outbuf, dec);
 597                 strbuf_release(dec);
 598                 free(dec);
 599                 in = ep + 2;
 600         }
 601         strbuf_addstr(&outbuf, in);
 602         strbuf_reset(it);
 603         strbuf_addbuf(it, &outbuf);
 604 decode_header_bq_out:
 605         strbuf_release(&outbuf);
 606         strbuf_release(&charset_q);
 607         strbuf_release(&piecebuf);
 608         return rfc2047;
 609 }
 610
 611 static void decode_header(struct strbuf *it)
 612 {
 613         if (decode_header_bq(it))
 614                 return;
 615         /* otherwise "it" is a straight copy of the input.
 616          * This can be binary guck but there is no charset specified.
 617          */
 618         if (metainfo_charset)
 619                 convert_to_utf8(it, "");
 620 }
 621
 622 static void decode_transfer_encoding(struct strbuf *line)
 623 {
 624         struct strbuf *ret;
 625
 626         switch (transfer_encoding) {
 627         case TE_QP:
 628                 ret = decode_q_segment(line, 0);
 629                 break;
 630         case TE_BASE64:
 631                 ret = decode_b_segment(line);
 632                 break;
 633         case TE_DONTCARE:
 634         default:
 635                 return;
 636         }
 637         strbuf_reset(line);
 638         strbuf_addbuf(line, ret);
 639         strbuf_release(ret);
 640         free(ret);
 641 }
 642
 643 static void handle_filter(struct strbuf *line);
 644
 645 static int find_boundary(void)
 646 {
 647         while (!strbuf_getline(&line, fin, '\n')) {
 648                 if (*content_top && is_multipart_boundary(&line))
 649                         return 1;
 650         }
 651         return 0;
 652 }
 653
 654 static int handle_boundary(void)
 655 {
 656         struct strbuf newline = STRBUF_INIT;
 657
 658         strbuf_addch(&newline, '\n');
 659 again:
 660         if (line.len >= (*content_top)->len + 2 &&
 661             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 662                 /* we hit an end boundary */
 663                 /* pop the current boundary off the stack */
 664                 strbuf_release(*content_top);
 665                 free(*content_top);
 666                 *content_top = NULL;
 667
 668                 /* technically won't happen as is_multipart_boundary()
 669                    will fail first.  But just in case..
 670                  */
 671                 if (--content_top < content) {
 672                         fprintf(stderr, "Detected mismatched boundaries, "
 673                                         "can't recover\n");
 674                         exit(1);
 675                 }
 676                 handle_filter(&newline);
 677                 strbuf_release(&newline);
 678
 679                 /* skip to the next boundary */
 680                 if (!find_boundary())
 681                         return 0;
 682                 goto again;
 683         }
 684
 685         /* set some defaults */
 686         transfer_encoding = TE_DONTCARE;
 687         strbuf_reset(&charset);
 688
 689         /* slurp in this section's info */
 690         while (read_one_header_line(&line, fin))
 691                 check_header(&line, p_hdr_data, 0);
 692
 693         strbuf_release(&newline);
 694         /* replenish line */
 695         if (strbuf_getline(&line, fin, '\n'))
 696                 return 0;
 697         strbuf_addch(&line, '\n');
 698         return 1;
 699 }
 700
 701 static inline int patchbreak(const struct strbuf *line)
 702 {
 703         size_t i;
 704
 705         /* Beginning of a "diff -" header? */
 706         if (starts_with(line->buf, "diff -"))
 707                 return 1;
 708
 709         /* CVS "Index: " line? */
 710         if (starts_with(line->buf, "Index: "))
 711                 return 1;
 712
 713         /*
 714          * "--- <filename>" starts patches without headers
 715          * "---<sp>*" is a manual separator
 716          */
 717         if (line->len < 4)
 718                 return 0;
 719
 720         if (starts_with(line->buf, "---")) {
 721                 /* space followed by a filename? */
 722                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 723                         return 1;
 724                 /* Just whitespace? */
 725                 for (i = 3; i < line->len; i++) {
 726                         unsigned char c = line->buf[i];
 727                         if (c == '\n')
 728                                 return 1;
 729                         if (!isspace(c))
 730                                 break;
 731                 }
 732                 return 0;
 733         }
 734         return 0;
 735 }
 736
 737 static int is_scissors_line(const struct strbuf *line)
 738 {
 739         size_t i, len = line->len;
 740         int scissors = 0, gap = 0;
 741         int first_nonblank = -1;
 742         int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
 743         const char *buf = line->buf;
 744
 745         for (i = 0; i < len; i++) {
 746                 if (isspace(buf[i])) {
 747                         if (in_perforation) {
 748                                 perforation++;
 749                                 gap++;
 750                         }
 751                         continue;
 752                 }
 753                 last_nonblank = i;
 754                 if (first_nonblank < 0)
 755                         first_nonblank = i;
 756                 if (buf[i] == '-') {
 757                         in_perforation = 1;
 758                         perforation++;
 759                         continue;
 760                 }
 761                 if (i + 1 < len &&
 762                     (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
 763                      !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
 764                         in_perforation = 1;
 765                         perforation += 2;
 766                         scissors += 2;
 767                         i++;
 768                         continue;
 769                 }
 770                 in_perforation = 0;
 771         }
 772
 773         /*
 774          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 775          * Even though there can be arbitrary cruft on the same line
 776          * (e.g. "cut here"), in order to avoid misidentification, the
 777          * perforation must occupy more than a third of the visible
 778          * width of the line, and dashes and scissors must occupy more
 779          * than half of the perforation.
 780          */
 781
 782         visible = last_nonblank - first_nonblank + 1;
 783         return (scissors && 8 <= visible &&
 784                 visible < perforation * 3 &&
 785                 gap * 2 < perforation);
 786 }
 787
 788 static int handle_commit_msg(struct strbuf *line)
 789 {
 790         static int still_looking = 1;
 791
 792         if (!cmitmsg)
 793                 return 0;
 794
 795         if (still_looking) {
 796                 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
 797                         return 0;
 798         }
 799
 800         if (use_inbody_headers && still_looking) {
 801                 still_looking = check_header(line, s_hdr_data, 0);
 802                 if (still_looking)
 803                         return 0;
 804         } else
 805                 /* Only trim the first (blank) line of the commit message
 806                  * when ignoring in-body headers.
 807                  */
 808                 still_looking = 0;
 809
 810         /* normalize the log message to UTF-8. */
 811         if (metainfo_charset)
 812                 convert_to_utf8(line, charset.buf);
 813
 814         if (use_scissors && is_scissors_line(line)) {
 815                 int i;
 816                 if (fseek(cmitmsg, 0L, SEEK_SET))
 817                         die_errno("Could not rewind output message file");
 818                 if (ftruncate(fileno(cmitmsg), 0))
 819                         die_errno("Could not truncate output message file at scissors");
 820                 still_looking = 1;
 821
 822                 /*
 823                  * We may have already read "secondary headers"; purge
 824                  * them to give ourselves a clean restart.
 825                  */
 826                 for (i = 0; header[i]; i++) {
 827                         if (s_hdr_data[i])
 828                                 strbuf_release(s_hdr_data[i]);
 829                         s_hdr_data[i] = NULL;
 830                 }
 831                 return 0;
 832         }
 833
 834         if (patchbreak(line)) {
 835                 if (message_id)
 836                         fprintf(cmitmsg, "Message-Id: %s\n", message_id);
 837                 fclose(cmitmsg);
 838                 cmitmsg = NULL;
 839                 return 1;
 840         }
 841
 842         fputs(line->buf, cmitmsg);
 843         return 0;
 844 }
 845
 846 static void handle_patch(const struct strbuf *line)
 847 {
 848         fwrite(line->buf, 1, line->len, patchfile);
 849         patch_lines++;
 850 }
 851
 852 static void handle_filter(struct strbuf *line)
 853 {
 854         static int filter = 0;
 855
 856         /* filter tells us which part we left off on */
 857         switch (filter) {
 858         case 0:
 859                 if (!handle_commit_msg(line))
 860                         break;
 861                 filter++;
 862         case 1:
 863                 handle_patch(line);
 864                 break;
 865         }
 866 }
 867
 868 static void handle_body(void)
 869 {
 870         struct strbuf prev = STRBUF_INIT;
 871
 872         /* Skip up to the first boundary */
 873         if (*content_top) {
 874                 if (!find_boundary())
 875                         goto handle_body_out;
 876         }
 877
 878         do {
 879                 /* process any boundary lines */
 880                 if (*content_top && is_multipart_boundary(&line)) {
 881                         /* flush any leftover */
 882                         if (prev.len) {
 883                                 handle_filter(&prev);
 884                                 strbuf_reset(&prev);
 885                         }
 886                         if (!handle_boundary())
 887                                 goto handle_body_out;
 888                 }
 889
 890                 /* Unwrap transfer encoding */
 891                 decode_transfer_encoding(&line);
 892
 893                 switch (transfer_encoding) {
 894                 case TE_BASE64:
 895                 case TE_QP:
 896                 {
 897                         struct strbuf **lines, **it, *sb;
 898
 899                         /* Prepend any previous partial lines */
 900                         strbuf_insert(&line, 0, prev.buf, prev.len);
 901                         strbuf_reset(&prev);
 902
 903                         /*
 904                          * This is a decoded line that may contain
 905                          * multiple new lines.  Pass only one chunk
 906                          * at a time to handle_filter()
 907                          */
 908                         lines = strbuf_split(&line, '\n');
 909                         for (it = lines; (sb = *it); it++) {
 910                                 if (*(it + 1) == NULL) /* The last line */
 911                                         if (sb->buf[sb->len - 1] != '\n') {
 912                                                 /* Partial line, save it for later. */
 913                                                 strbuf_addbuf(&prev, sb);
 914                                                 break;
 915                                         }
 916                                 handle_filter(sb);
 917                         }
 918                         /*
 919                          * The partial chunk is saved in "prev" and will be
 920                          * appended by the next iteration of read_line_with_nul().
 921                          */
 922                         strbuf_list_free(lines);
 923                         break;
 924                 }
 925                 default:
 926                         handle_filter(&line);
 927                 }
 928
 929         } while (!strbuf_getwholeline(&line, fin, '\n'));
 930
 931 handle_body_out:
 932         strbuf_release(&prev);
 933 }
 934
 935 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 936 {
 937         const char *sp = data->buf;
 938         while (1) {
 939                 char *ep = strchr(sp, '\n');
 940                 int len;
 941                 if (!ep)
 942                         len = strlen(sp);
 943                 else
 944                         len = ep - sp;
 945                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 946                 if (!ep)
 947                         break;
 948                 sp = ep + 1;
 949         }
 950 }
 951
 952 static void handle_info(void)
 953 {
 954         struct strbuf *hdr;
 955         int i;
 956
 957         for (i = 0; header[i]; i++) {
 958                 /* only print inbody headers if we output a patch file */
 959                 if (patch_lines && s_hdr_data[i])
 960                         hdr = s_hdr_data[i];
 961                 else if (p_hdr_data[i])
 962                         hdr = p_hdr_data[i];
 963                 else
 964                         continue;
 965
 966                 if (!strcmp(header[i], "Subject")) {
 967                         if (!keep_subject) {
 968                                 cleanup_subject(hdr);
 969                                 cleanup_space(hdr);
 970                         }
 971                         output_header_lines(fout, "Subject", hdr);
 972                 } else if (!strcmp(header[i], "From")) {
 973                         cleanup_space(hdr);
 974                         handle_from(hdr);
 975                         fprintf(fout, "Author: %s\n", name.buf);
 976                         fprintf(fout, "Email: %s\n", email.buf);
 977                 } else {
 978                         cleanup_space(hdr);
 979                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 980                 }
 981         }
 982         fprintf(fout, "\n");
 983 }
 984
 985 static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
 986 {
 987         int peek;
 988         fin = in;
 989         fout = out;
 990
 991         cmitmsg = fopen(msg, "w");
 992         if (!cmitmsg) {
 993                 perror(msg);
 994                 return -1;
 995         }
 996         patchfile = fopen(patch, "w");
 997         if (!patchfile) {
 998                 perror(patch);
 999                 fclose(cmitmsg);
1000                 return -1;
1001         }
1002
1003         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
1004         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
1005
1006         do {
1007                 peek = fgetc(in);
1008         } while (isspace(peek));
1009         ungetc(peek, in);
1010
1011         /* process the email header */
1012         while (read_one_header_line(&line, fin))
1013                 check_header(&line, p_hdr_data, 1);
1014
1015         handle_body();
1016         handle_info();
1017
1018         return 0;
1019 }
1020
1021 static int git_mailinfo_config(const char *var, const char *value, void *unused)
1022 {
1023         if (!starts_with(var, "mailinfo."))
1024                 return git_default_config(var, value, unused);
1025         if (!strcmp(var, "mailinfo.scissors")) {
1026                 use_scissors = git_config_bool(var, value);
1027                 return 0;
1028         }
1029         /* perhaps others here */
1030         return 0;
1031 }
1032
1033 static const char mailinfo_usage[] =
1034         "git mailinfo [-k|-b] [-m | --message-id] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] msg patch < mail >info";
1035
1036 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1037 {
1038         const char *def_charset;
1039
1040         /* NEEDSWORK: might want to do the optional .git/ directory
1041          * discovery
1042          */
1043         git_config(git_mailinfo_config, NULL);
1044
1045         def_charset = get_commit_output_encoding();
1046         metainfo_charset = def_charset;
1047
1048         while (1 < argc && argv[1][0] == '-') {
1049                 if (!strcmp(argv[1], "-k"))
1050                         keep_subject = 1;
1051                 else if (!strcmp(argv[1], "-b"))
1052                         keep_non_patch_brackets_in_subject = 1;
1053                 else if (!strcmp(argv[1], "-m") || !strcmp(argv[1], "--message-id"))
1054                         add_message_id = 1;
1055                 else if (!strcmp(argv[1], "-u"))
1056                         metainfo_charset = def_charset;
1057                 else if (!strcmp(argv[1], "-n"))
1058                         metainfo_charset = NULL;
1059                 else if (starts_with(argv[1], "--encoding="))
1060                         metainfo_charset = argv[1] + 11;
1061                 else if (!strcmp(argv[1], "--scissors"))
1062                         use_scissors = 1;
1063                 else if (!strcmp(argv[1], "--no-scissors"))
1064                         use_scissors = 0;
1065                 else if (!strcmp(argv[1], "--no-inbody-headers"))
1066                         use_inbody_headers = 0;
1067                 else
1068                         usage(mailinfo_usage);
1069                 argc--; argv++;
1070         }
1071
1072         if (argc != 3)
1073                 usage(mailinfo_usage);
1074
1075         return !!mailinfo(stdin, stdout, argv[1], argv[2]);
1076 }