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
  23 static struct strbuf charset = STRBUF_INIT;
  24 static int patch_lines;
  25 static struct strbuf **p_hdr_data, **s_hdr_data;
  26 static int use_scissors;
  27 static int use_inbody_headers = 1;
  28
  29 #define MAX_HDR_PARSED 10
  30 #define MAX_BOUNDARIES 5
  31
  32 static void cleanup_space(struct strbuf *sb);
  33
  34
  35 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  36 {
  37         struct strbuf *src = name;
  38         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  39                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  40                 src = email;
  41         else if (name == out)
  42                 return;
  43         strbuf_reset(out);
  44         strbuf_addbuf(out, src);
  45 }
  46
  47 static void parse_bogus_from(const struct strbuf *line)
  48 {
  49         /* John Doe <johndoe> */
  50
  51         char *bra, *ket;
  52         /* This is fallback, so do not bother if we already have an
  53          * e-mail address.
  54          */
  55         if (email.len)
  56                 return;
  57
  58         bra = strchr(line->buf, '<');
  59         if (!bra)
  60                 return;
  61         ket = strchr(bra, '>');
  62         if (!ket)
  63                 return;
  64
  65         strbuf_reset(&email);
  66         strbuf_add(&email, bra + 1, ket - bra - 1);
  67
  68         strbuf_reset(&name);
  69         strbuf_add(&name, line->buf, bra - line->buf);
  70         strbuf_trim(&name);
  71         get_sane_name(&name, &name, &email);
  72 }
  73
  74 static void handle_from(const struct strbuf *from)
  75 {
  76         char *at;
  77         size_t el;
  78         struct strbuf f;
  79
  80         strbuf_init(&f, from->len);
  81         strbuf_addbuf(&f, from);
  82
  83         at = strchr(f.buf, '@');
  84         if (!at) {
  85                 parse_bogus_from(from);
  86                 return;
  87         }
  88
  89         /*
  90          * If we already have one email, don't take any confusing lines
  91          */
  92         if (email.len && strchr(at + 1, '@')) {
  93                 strbuf_release(&f);
  94                 return;
  95         }
  96
  97         /* Pick up the string around '@', possibly delimited with <>
  98          * pair; that is the email part.
  99          */
 100         while (at > f.buf) {
 101                 char c = at[-1];
 102                 if (isspace(c))
 103                         break;
 104                 if (c == '<') {
 105                         at[-1] = ' ';
 106                         break;
 107                 }
 108                 at--;
 109         }
 110         el = strcspn(at, " \n\t\r\v\f>");
 111         strbuf_reset(&email);
 112         strbuf_add(&email, at, el);
 113         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 114
 115         /* The remainder is name.  It could be
 116          *
 117          * - "John Doe <john.doe@xz>"                   (a), or
 118          * - "john.doe@xz (John Doe)"                   (b), or
 119          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 120          *
 121          * but we have removed the email part, so
 122          *
 123          * - remove extra spaces which could stay after email (case 'c'), and
 124          * - trim from both ends, possibly removing the () pair at the end
 125          *   (cases 'a' and 'b').
 126          */
 127         cleanup_space(&f);
 128         strbuf_trim(&f);
 129         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 130                 strbuf_remove(&f, 0, 1);
 131                 strbuf_setlen(&f, f.len - 1);
 132         }
 133
 134         get_sane_name(&name, &f, &email);
 135         strbuf_release(&f);
 136 }
 137
 138 static void handle_header(struct strbuf **out, const struct strbuf *line)
 139 {
 140         if (!*out) {
 141                 *out = xmalloc(sizeof(struct strbuf));
 142                 strbuf_init(*out, line->len);
 143         } else
 144                 strbuf_reset(*out);
 145
 146         strbuf_addbuf(*out, line);
 147 }
 148
 149 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 150  * to have enough heuristics to grok MIME encoded patches often found
 151  * on our mailing lists.  For example, we do not even treat header lines
 152  * case insensitively.
 153  */
 154
 155 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 156 {
 157         const char *ends, *ap = strcasestr(line, name);
 158         size_t sz;
 159
 160         strbuf_setlen(attr, 0);
 161         if (!ap)
 162                 return 0;
 163         ap += strlen(name);
 164         if (*ap == '"') {
 165                 ap++;
 166                 ends = "\"";
 167         }
 168         else
 169                 ends = "; \t";
 170         sz = strcspn(ap, ends);
 171         strbuf_add(attr, ap, sz);
 172         return 1;
 173 }
 174
 175 static struct strbuf *content[MAX_BOUNDARIES];
 176
 177 static struct strbuf **content_top = content;
 178
 179 static void handle_content_type(struct strbuf *line)
 180 {
 181         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 182         strbuf_init(boundary, line->len);
 183
 184         if (slurp_attr(line->buf, "boundary=", boundary)) {
 185                 strbuf_insert(boundary, 0, "--", 2);
 186                 if (++content_top > &content[MAX_BOUNDARIES]) {
 187                         fprintf(stderr, "Too many boundaries to handle\n");
 188                         exit(1);
 189                 }
 190                 *content_top = boundary;
 191                 boundary = NULL;
 192         }
 193         slurp_attr(line->buf, "charset=", &charset);
 194
 195         if (boundary) {
 196                 strbuf_release(boundary);
 197                 free(boundary);
 198         }
 199 }
 200
 201 static void handle_content_transfer_encoding(const struct strbuf *line)
 202 {
 203         if (strcasestr(line->buf, "base64"))
 204                 transfer_encoding = TE_BASE64;
 205         else if (strcasestr(line->buf, "quoted-printable"))
 206                 transfer_encoding = TE_QP;
 207         else
 208                 transfer_encoding = TE_DONTCARE;
 209 }
 210
 211 static int is_multipart_boundary(const struct strbuf *line)
 212 {
 213         return (((*content_top)->len <= line->len) &&
 214                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
 215 }
 216
 217 static void cleanup_subject(struct strbuf *subject)
 218 {
 219         size_t at = 0;
 220
 221         while (at < subject->len) {
 222                 char *pos;
 223                 size_t remove;
 224
 225                 switch (subject->buf[at]) {
 226                 case 'r': case 'R':
 227                         if (subject->len <= at + 3)
 228                                 break;
 229                         if ((subject->buf[at + 1] == 'e' ||
 230                              subject->buf[at + 1] == 'E') &&
 231                             subject->buf[at + 2] == ':') {
 232                                 strbuf_remove(subject, at, 3);
 233                                 continue;
 234                         }
 235                         at++;
 236                         break;
 237                 case ' ': case '\t': case ':':
 238                         strbuf_remove(subject, at, 1);
 239                         continue;
 240                 case '[':
 241                         pos = strchr(subject->buf + at, ']');
 242                         if (!pos)
 243                                 break;
 244                         remove = pos - subject->buf + at + 1;
 245                         if (!keep_non_patch_brackets_in_subject ||
 246                             (7 <= remove &&
 247                              memmem(subject->buf + at, remove, "PATCH", 5)))
 248                                 strbuf_remove(subject, at, remove);
 249                         else {
 250                                 at += remove;
 251                                 /*
 252                                  * If the input had a space after the ], keep
 253                                  * it.  We don't bother with finding the end of
 254                                  * the space, since we later normalize it
 255                                  * anyway.
 256                                  */
 257                                 if (isspace(subject->buf[at]))
 258                                         at += 1;
 259                         }
 260                         continue;
 261                 }
 262                 break;
 263         }
 264         strbuf_trim(subject);
 265 }
 266
 267 static void cleanup_space(struct strbuf *sb)
 268 {
 269         size_t pos, cnt;
 270         for (pos = 0; pos < sb->len; pos++) {
 271                 if (isspace(sb->buf[pos])) {
 272                         sb->buf[pos] = ' ';
 273                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 274                         strbuf_remove(sb, pos + 1, cnt);
 275                 }
 276         }
 277 }
 278
 279 static void decode_header(struct strbuf *line);
 280 static const char *header[MAX_HDR_PARSED] = {
 281         "From","Subject","Date",
 282 };
 283
 284 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 285 {
 286         int len = strlen(hdr);
 287         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 288                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 289 }
 290
 291 #define SAMPLE "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n"
 292 static int is_format_patch_separator(const char *line, int len)
 293 {
 294         const char *cp;
 295
 296         if (len != strlen(SAMPLE))
 297                 return 0;
 298         if (!skip_prefix(line, "From ", &cp))
 299                 return 0;
 300         if (strspn(cp, "0123456789abcdef") != 40)
 301                 return 0;
 302         cp += 40;
 303         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 304 }
 305
 306 static int check_header(const struct strbuf *line,
 307                                 struct strbuf *hdr_data[], int overwrite)
 308 {
 309         int i, ret = 0, len;
 310         struct strbuf sb = STRBUF_INIT;
 311         /* search for the interesting parts */
 312         for (i = 0; header[i]; i++) {
 313                 int len = strlen(header[i]);
 314                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 315                         /* Unwrap inline B and Q encoding, and optionally
 316                          * normalize the meta information to utf8.
 317                          */
 318                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 319                         decode_header(&sb);
 320                         handle_header(&hdr_data[i], &sb);
 321                         ret = 1;
 322                         goto check_header_out;
 323                 }
 324         }
 325
 326         /* Content stuff */
 327         if (cmp_header(line, "Content-Type")) {
 328                 len = strlen("Content-Type: ");
 329                 strbuf_add(&sb, line->buf + len, line->len - len);
 330                 decode_header(&sb);
 331                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 332                 handle_content_type(&sb);
 333                 ret = 1;
 334                 goto check_header_out;
 335         }
 336         if (cmp_header(line, "Content-Transfer-Encoding")) {
 337                 len = strlen("Content-Transfer-Encoding: ");
 338                 strbuf_add(&sb, line->buf + len, line->len - len);
 339                 decode_header(&sb);
 340                 handle_content_transfer_encoding(&sb);
 341                 ret = 1;
 342                 goto check_header_out;
 343         }
 344
 345         /* for inbody stuff */
 346         if (starts_with(line->buf, ">From") && isspace(line->buf[5])) {
 347                 ret = is_format_patch_separator(line->buf + 1, line->len - 1);
 348                 goto check_header_out;
 349         }
 350         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 351                 for (i = 0; header[i]; i++) {
 352                         if (!strcmp("Subject", header[i])) {
 353                                 handle_header(&hdr_data[i], line);
 354                                 ret = 1;
 355                                 goto check_header_out;
 356                         }
 357                 }
 358         }
 359
 360 check_header_out:
 361         strbuf_release(&sb);
 362         return ret;
 363 }
 364
 365 static int is_rfc2822_header(const struct strbuf *line)
 366 {
 367         /*
 368          * The section that defines the loosest possible
 369          * field name is "3.6.8 Optional fields".
 370          *
 371          * optional-field = field-name ":" unstructured CRLF
 372          * field-name = 1*ftext
 373          * ftext = %d33-57 / %59-126
 374          */
 375         int ch;
 376         char *cp = line->buf;
 377
 378         /* Count mbox From headers as headers */
 379         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 380                 return 1;
 381
 382         while ((ch = *cp++)) {
 383                 if (ch == ':')
 384                         return 1;
 385                 if ((33 <= ch && ch <= 57) ||
 386                     (59 <= ch && ch <= 126))
 387                         continue;
 388                 break;
 389         }
 390         return 0;
 391 }
 392
 393 static int read_one_header_line(struct strbuf *line, FILE *in)
 394 {
 395         /* Get the first part of the line. */
 396         if (strbuf_getline(line, in, '\n'))
 397                 return 0;
 398
 399         /*
 400          * Is it an empty line or not a valid rfc2822 header?
 401          * If so, stop here, and return false ("not a header")
 402          */
 403         strbuf_rtrim(line);
 404         if (!line->len || !is_rfc2822_header(line)) {
 405                 /* Re-add the newline */
 406                 strbuf_addch(line, '\n');
 407                 return 0;
 408         }
 409
 410         /*
 411          * Now we need to eat all the continuation lines..
 412          * Yuck, 2822 header "folding"
 413          */
 414         for (;;) {
 415                 int peek;
 416                 struct strbuf continuation = STRBUF_INIT;
 417
 418                 peek = fgetc(in); ungetc(peek, in);
 419                 if (peek != ' ' && peek != '\t')
 420                         break;
 421                 if (strbuf_getline(&continuation, in, '\n'))
 422                         break;
 423                 continuation.buf[0] = ' ';
 424                 strbuf_rtrim(&continuation);
 425                 strbuf_addbuf(line, &continuation);
 426         }
 427
 428         return 1;
 429 }
 430
 431 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 432 {
 433         const char *in = q_seg->buf;
 434         int c;
 435         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 436         strbuf_init(out, q_seg->len);
 437
 438         while ((c = *in++) != 0) {
 439                 if (c == '=') {
 440                         int d = *in++;
 441                         if (d == '\n' || !d)
 442                                 break; /* drop trailing newline */
 443                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 444                         continue;
 445                 }
 446                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 447                         c = 0x20;
 448                 strbuf_addch(out, c);
 449         }
 450         return out;
 451 }
 452
 453 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 454 {
 455         /* Decode in..ep, possibly in-place to ot */
 456         int c, pos = 0, acc = 0;
 457         const char *in = b_seg->buf;
 458         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 459         strbuf_init(out, b_seg->len);
 460
 461         while ((c = *in++) != 0) {
 462                 if (c == '+')
 463                         c = 62;
 464                 else if (c == '/')
 465                         c = 63;
 466                 else if ('A' <= c && c <= 'Z')
 467                         c -= 'A';
 468                 else if ('a' <= c && c <= 'z')
 469                         c -= 'a' - 26;
 470                 else if ('0' <= c && c <= '9')
 471                         c -= '0' - 52;
 472                 else
 473                         continue; /* garbage */
 474                 switch (pos++) {
 475                 case 0:
 476                         acc = (c << 2);
 477                         break;
 478                 case 1:
 479                         strbuf_addch(out, (acc | (c >> 4)));
 480                         acc = (c & 15) << 4;
 481                         break;
 482                 case 2:
 483                         strbuf_addch(out, (acc | (c >> 2)));
 484                         acc = (c & 3) << 6;
 485                         break;
 486                 case 3:
 487                         strbuf_addch(out, (acc | c));
 488                         acc = pos = 0;
 489                         break;
 490                 }
 491         }
 492         return out;
 493 }
 494
 495 static void convert_to_utf8(struct strbuf *line, const char *charset)
 496 {
 497         char *out;
 498
 499         if (!charset || !*charset)
 500                 return;
 501
 502         if (same_encoding(metainfo_charset, charset))
 503                 return;
 504         out = reencode_string(line->buf, metainfo_charset, charset);
 505         if (!out)
 506                 die("cannot convert from %s to %s",
 507                     charset, metainfo_charset);
 508         strbuf_attach(line, out, strlen(out), strlen(out));
 509 }
 510
 511 static int decode_header_bq(struct strbuf *it)
 512 {
 513         char *in, *ep, *cp;
 514         struct strbuf outbuf = STRBUF_INIT, *dec;
 515         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 516         int rfc2047 = 0;
 517
 518         in = it->buf;
 519         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 520                 int encoding;
 521                 strbuf_reset(&charset_q);
 522                 strbuf_reset(&piecebuf);
 523                 rfc2047 = 1;
 524
 525                 if (in != ep) {
 526                         /*
 527                          * We are about to process an encoded-word
 528                          * that begins at ep, but there is something
 529                          * before the encoded word.
 530                          */
 531                         char *scan;
 532                         for (scan = in; scan < ep; scan++)
 533                                 if (!isspace(*scan))
 534                                         break;
 535
 536                         if (scan != ep || in == it->buf) {
 537                                 /*
 538                                  * We should not lose that "something",
 539                                  * unless we have just processed an
 540                                  * encoded-word, and there is only LWS
 541                                  * before the one we are about to process.
 542                                  */
 543                                 strbuf_add(&outbuf, in, ep - in);
 544                         }
 545                 }
 546                 /* E.g.
 547                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 548                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 549                  */
 550                 ep += 2;
 551
 552                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 553                         goto decode_header_bq_out;
 554
 555                 if (cp + 3 - it->buf > it->len)
 556                         goto decode_header_bq_out;
 557                 strbuf_add(&charset_q, ep, cp - ep);
 558
 559                 encoding = cp[1];
 560                 if (!encoding || cp[2] != '?')
 561                         goto decode_header_bq_out;
 562                 ep = strstr(cp + 3, "?=");
 563                 if (!ep)
 564                         goto decode_header_bq_out;
 565                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 566                 switch (tolower(encoding)) {
 567                 default:
 568                         goto decode_header_bq_out;
 569                 case 'b':
 570                         dec = decode_b_segment(&piecebuf);
 571                         break;
 572                 case 'q':
 573                         dec = decode_q_segment(&piecebuf, 1);
 574                         break;
 575                 }
 576                 if (metainfo_charset)
 577                         convert_to_utf8(dec, charset_q.buf);
 578
 579                 strbuf_addbuf(&outbuf, dec);
 580                 strbuf_release(dec);
 581                 free(dec);
 582                 in = ep + 2;
 583         }
 584         strbuf_addstr(&outbuf, in);
 585         strbuf_reset(it);
 586         strbuf_addbuf(it, &outbuf);
 587 decode_header_bq_out:
 588         strbuf_release(&outbuf);
 589         strbuf_release(&charset_q);
 590         strbuf_release(&piecebuf);
 591         return rfc2047;
 592 }
 593
 594 static void decode_header(struct strbuf *it)
 595 {
 596         if (decode_header_bq(it))
 597                 return;
 598         /* otherwise "it" is a straight copy of the input.
 599          * This can be binary guck but there is no charset specified.
 600          */
 601         if (metainfo_charset)
 602                 convert_to_utf8(it, "");
 603 }
 604
 605 static void decode_transfer_encoding(struct strbuf *line)
 606 {
 607         struct strbuf *ret;
 608
 609         switch (transfer_encoding) {
 610         case TE_QP:
 611                 ret = decode_q_segment(line, 0);
 612                 break;
 613         case TE_BASE64:
 614                 ret = decode_b_segment(line);
 615                 break;
 616         case TE_DONTCARE:
 617         default:
 618                 return;
 619         }
 620         strbuf_reset(line);
 621         strbuf_addbuf(line, ret);
 622         strbuf_release(ret);
 623         free(ret);
 624 }
 625
 626 static void handle_filter(struct strbuf *line);
 627
 628 static int find_boundary(void)
 629 {
 630         while (!strbuf_getline(&line, fin, '\n')) {
 631                 if (*content_top && is_multipart_boundary(&line))
 632                         return 1;
 633         }
 634         return 0;
 635 }
 636
 637 static int handle_boundary(void)
 638 {
 639         struct strbuf newline = STRBUF_INIT;
 640
 641         strbuf_addch(&newline, '\n');
 642 again:
 643         if (line.len >= (*content_top)->len + 2 &&
 644             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 645                 /* we hit an end boundary */
 646                 /* pop the current boundary off the stack */
 647                 strbuf_release(*content_top);
 648                 free(*content_top);
 649                 *content_top = NULL;
 650
 651                 /* technically won't happen as is_multipart_boundary()
 652                    will fail first.  But just in case..
 653                  */
 654                 if (--content_top < content) {
 655                         fprintf(stderr, "Detected mismatched boundaries, "
 656                                         "can't recover\n");
 657                         exit(1);
 658                 }
 659                 handle_filter(&newline);
 660                 strbuf_release(&newline);
 661
 662                 /* skip to the next boundary */
 663                 if (!find_boundary())
 664                         return 0;
 665                 goto again;
 666         }
 667
 668         /* set some defaults */
 669         transfer_encoding = TE_DONTCARE;
 670         strbuf_reset(&charset);
 671
 672         /* slurp in this section's info */
 673         while (read_one_header_line(&line, fin))
 674                 check_header(&line, p_hdr_data, 0);
 675
 676         strbuf_release(&newline);
 677         /* replenish line */
 678         if (strbuf_getline(&line, fin, '\n'))
 679                 return 0;
 680         strbuf_addch(&line, '\n');
 681         return 1;
 682 }
 683
 684 static inline int patchbreak(const struct strbuf *line)
 685 {
 686         size_t i;
 687
 688         /* Beginning of a "diff -" header? */
 689         if (starts_with(line->buf, "diff -"))
 690                 return 1;
 691
 692         /* CVS "Index: " line? */
 693         if (starts_with(line->buf, "Index: "))
 694                 return 1;
 695
 696         /*
 697          * "--- <filename>" starts patches without headers
 698          * "---<sp>*" is a manual separator
 699          */
 700         if (line->len < 4)
 701                 return 0;
 702
 703         if (starts_with(line->buf, "---")) {
 704                 /* space followed by a filename? */
 705                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 706                         return 1;
 707                 /* Just whitespace? */
 708                 for (i = 3; i < line->len; i++) {
 709                         unsigned char c = line->buf[i];
 710                         if (c == '\n')
 711                                 return 1;
 712                         if (!isspace(c))
 713                                 break;
 714                 }
 715                 return 0;
 716         }
 717         return 0;
 718 }
 719
 720 static int is_scissors_line(const struct strbuf *line)
 721 {
 722         size_t i, len = line->len;
 723         int scissors = 0, gap = 0;
 724         int first_nonblank = -1;
 725         int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
 726         const char *buf = line->buf;
 727
 728         for (i = 0; i < len; i++) {
 729                 if (isspace(buf[i])) {
 730                         if (in_perforation) {
 731                                 perforation++;
 732                                 gap++;
 733                         }
 734                         continue;
 735                 }
 736                 last_nonblank = i;
 737                 if (first_nonblank < 0)
 738                         first_nonblank = i;
 739                 if (buf[i] == '-') {
 740                         in_perforation = 1;
 741                         perforation++;
 742                         continue;
 743                 }
 744                 if (i + 1 < len &&
 745                     (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
 746                      !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
 747                         in_perforation = 1;
 748                         perforation += 2;
 749                         scissors += 2;
 750                         i++;
 751                         continue;
 752                 }
 753                 in_perforation = 0;
 754         }
 755
 756         /*
 757          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 758          * Even though there can be arbitrary cruft on the same line
 759          * (e.g. "cut here"), in order to avoid misidentification, the
 760          * perforation must occupy more than a third of the visible
 761          * width of the line, and dashes and scissors must occupy more
 762          * than half of the perforation.
 763          */
 764
 765         visible = last_nonblank - first_nonblank + 1;
 766         return (scissors && 8 <= visible &&
 767                 visible < perforation * 3 &&
 768                 gap * 2 < perforation);
 769 }
 770
 771 static int handle_commit_msg(struct strbuf *line)
 772 {
 773         static int still_looking = 1;
 774
 775         if (!cmitmsg)
 776                 return 0;
 777
 778         if (still_looking) {
 779                 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
 780                         return 0;
 781         }
 782
 783         if (use_inbody_headers && still_looking) {
 784                 still_looking = check_header(line, s_hdr_data, 0);
 785                 if (still_looking)
 786                         return 0;
 787         } else
 788                 /* Only trim the first (blank) line of the commit message
 789                  * when ignoring in-body headers.
 790                  */
 791                 still_looking = 0;
 792
 793         /* normalize the log message to UTF-8. */
 794         if (metainfo_charset)
 795                 convert_to_utf8(line, charset.buf);
 796
 797         if (use_scissors && is_scissors_line(line)) {
 798                 int i;
 799                 if (fseek(cmitmsg, 0L, SEEK_SET))
 800                         die_errno("Could not rewind output message file");
 801                 if (ftruncate(fileno(cmitmsg), 0))
 802                         die_errno("Could not truncate output message file at scissors");
 803                 still_looking = 1;
 804
 805                 /*
 806                  * We may have already read "secondary headers"; purge
 807                  * them to give ourselves a clean restart.
 808                  */
 809                 for (i = 0; header[i]; i++) {
 810                         if (s_hdr_data[i])
 811                                 strbuf_release(s_hdr_data[i]);
 812                         s_hdr_data[i] = NULL;
 813                 }
 814                 return 0;
 815         }
 816
 817         if (patchbreak(line)) {
 818                 fclose(cmitmsg);
 819                 cmitmsg = NULL;
 820                 return 1;
 821         }
 822
 823         fputs(line->buf, cmitmsg);
 824         return 0;
 825 }
 826
 827 static void handle_patch(const struct strbuf *line)
 828 {
 829         fwrite(line->buf, 1, line->len, patchfile);
 830         patch_lines++;
 831 }
 832
 833 static void handle_filter(struct strbuf *line)
 834 {
 835         static int filter = 0;
 836
 837         /* filter tells us which part we left off on */
 838         switch (filter) {
 839         case 0:
 840                 if (!handle_commit_msg(line))
 841                         break;
 842                 filter++;
 843         case 1:
 844                 handle_patch(line);
 845                 break;
 846         }
 847 }
 848
 849 static void handle_body(void)
 850 {
 851         struct strbuf prev = STRBUF_INIT;
 852
 853         /* Skip up to the first boundary */
 854         if (*content_top) {
 855                 if (!find_boundary())
 856                         goto handle_body_out;
 857         }
 858
 859         do {
 860                 /* process any boundary lines */
 861                 if (*content_top && is_multipart_boundary(&line)) {
 862                         /* flush any leftover */
 863                         if (prev.len) {
 864                                 handle_filter(&prev);
 865                                 strbuf_reset(&prev);
 866                         }
 867                         if (!handle_boundary())
 868                                 goto handle_body_out;
 869                 }
 870
 871                 /* Unwrap transfer encoding */
 872                 decode_transfer_encoding(&line);
 873
 874                 switch (transfer_encoding) {
 875                 case TE_BASE64:
 876                 case TE_QP:
 877                 {
 878                         struct strbuf **lines, **it, *sb;
 879
 880                         /* Prepend any previous partial lines */
 881                         strbuf_insert(&line, 0, prev.buf, prev.len);
 882                         strbuf_reset(&prev);
 883
 884                         /*
 885                          * This is a decoded line that may contain
 886                          * multiple new lines.  Pass only one chunk
 887                          * at a time to handle_filter()
 888                          */
 889                         lines = strbuf_split(&line, '\n');
 890                         for (it = lines; (sb = *it); it++) {
 891                                 if (*(it + 1) == NULL) /* The last line */
 892                                         if (sb->buf[sb->len - 1] != '\n') {
 893                                                 /* Partial line, save it for later. */
 894                                                 strbuf_addbuf(&prev, sb);
 895                                                 break;
 896                                         }
 897                                 handle_filter(sb);
 898                         }
 899                         /*
 900                          * The partial chunk is saved in "prev" and will be
 901                          * appended by the next iteration of read_line_with_nul().
 902                          */
 903                         strbuf_list_free(lines);
 904                         break;
 905                 }
 906                 default:
 907                         handle_filter(&line);
 908                 }
 909
 910         } while (!strbuf_getwholeline(&line, fin, '\n'));
 911
 912 handle_body_out:
 913         strbuf_release(&prev);
 914 }
 915
 916 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 917 {
 918         const char *sp = data->buf;
 919         while (1) {
 920                 char *ep = strchr(sp, '\n');
 921                 int len;
 922                 if (!ep)
 923                         len = strlen(sp);
 924                 else
 925                         len = ep - sp;
 926                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 927                 if (!ep)
 928                         break;
 929                 sp = ep + 1;
 930         }
 931 }
 932
 933 static void handle_info(void)
 934 {
 935         struct strbuf *hdr;
 936         int i;
 937
 938         for (i = 0; header[i]; i++) {
 939                 /* only print inbody headers if we output a patch file */
 940                 if (patch_lines && s_hdr_data[i])
 941                         hdr = s_hdr_data[i];
 942                 else if (p_hdr_data[i])
 943                         hdr = p_hdr_data[i];
 944                 else
 945                         continue;
 946
 947                 if (!strcmp(header[i], "Subject")) {
 948                         if (!keep_subject) {
 949                                 cleanup_subject(hdr);
 950                                 cleanup_space(hdr);
 951                         }
 952                         output_header_lines(fout, "Subject", hdr);
 953                 } else if (!strcmp(header[i], "From")) {
 954                         cleanup_space(hdr);
 955                         handle_from(hdr);
 956                         fprintf(fout, "Author: %s\n", name.buf);
 957                         fprintf(fout, "Email: %s\n", email.buf);
 958                 } else {
 959                         cleanup_space(hdr);
 960                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 961                 }
 962         }
 963         fprintf(fout, "\n");
 964 }
 965
 966 static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
 967 {
 968         int peek;
 969         fin = in;
 970         fout = out;
 971
 972         cmitmsg = fopen(msg, "w");
 973         if (!cmitmsg) {
 974                 perror(msg);
 975                 return -1;
 976         }
 977         patchfile = fopen(patch, "w");
 978         if (!patchfile) {
 979                 perror(patch);
 980                 fclose(cmitmsg);
 981                 return -1;
 982         }
 983
 984         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
 985         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
 986
 987         do {
 988                 peek = fgetc(in);
 989         } while (isspace(peek));
 990         ungetc(peek, in);
 991
 992         /* process the email header */
 993         while (read_one_header_line(&line, fin))
 994                 check_header(&line, p_hdr_data, 1);
 995
 996         handle_body();
 997         handle_info();
 998
 999         return 0;
1000 }
1001
1002 static int git_mailinfo_config(const char *var, const char *value, void *unused)
1003 {
1004         if (!starts_with(var, "mailinfo."))
1005                 return git_default_config(var, value, unused);
1006         if (!strcmp(var, "mailinfo.scissors")) {
1007                 use_scissors = git_config_bool(var, value);
1008                 return 0;
1009         }
1010         /* perhaps others here */
1011         return 0;
1012 }
1013
1014 static const char mailinfo_usage[] =
1015         "git mailinfo [-k|-b] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] msg patch < mail >info";
1016
1017 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1018 {
1019         const char *def_charset;
1020
1021         /* NEEDSWORK: might want to do the optional .git/ directory
1022          * discovery
1023          */
1024         git_config(git_mailinfo_config, NULL);
1025
1026         def_charset = get_commit_output_encoding();
1027         metainfo_charset = def_charset;
1028
1029         while (1 < argc && argv[1][0] == '-') {
1030                 if (!strcmp(argv[1], "-k"))
1031                         keep_subject = 1;
1032                 else if (!strcmp(argv[1], "-b"))
1033                         keep_non_patch_brackets_in_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 (starts_with(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 }