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
  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         if (!ap) {
 161                 strbuf_setlen(attr, 0);
 162                 return 0;
 163         }
 164         ap += strlen(name);
 165         if (*ap == '"') {
 166                 ap++;
 167                 ends = "\"";
 168         }
 169         else
 170                 ends = "; \t";
 171         sz = strcspn(ap, ends);
 172         strbuf_add(attr, ap, sz);
 173         return 1;
 174 }
 175
 176 static struct strbuf *content[MAX_BOUNDARIES];
 177
 178 static struct strbuf **content_top = content;
 179
 180 static void handle_content_type(struct strbuf *line)
 181 {
 182         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 183         strbuf_init(boundary, line->len);
 184
 185         if (!strcasestr(line->buf, "text/"))
 186                  message_type = TYPE_OTHER;
 187         if (slurp_attr(line->buf, "boundary=", boundary)) {
 188                 strbuf_insert(boundary, 0, "--", 2);
 189                 if (++content_top > &content[MAX_BOUNDARIES]) {
 190                         fprintf(stderr, "Too many boundaries to handle\n");
 191                         exit(1);
 192                 }
 193                 *content_top = boundary;
 194                 boundary = NULL;
 195         }
 196         if (slurp_attr(line->buf, "charset=", &charset))
 197                 strbuf_tolower(&charset);
 198
 199         if (boundary) {
 200                 strbuf_release(boundary);
 201                 free(boundary);
 202         }
 203 }
 204
 205 static void handle_content_transfer_encoding(const struct strbuf *line)
 206 {
 207         if (strcasestr(line->buf, "base64"))
 208                 transfer_encoding = TE_BASE64;
 209         else if (strcasestr(line->buf, "quoted-printable"))
 210                 transfer_encoding = TE_QP;
 211         else
 212                 transfer_encoding = TE_DONTCARE;
 213 }
 214
 215 static int is_multipart_boundary(const struct strbuf *line)
 216 {
 217         return (((*content_top)->len <= line->len) &&
 218                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
 219 }
 220
 221 static void cleanup_subject(struct strbuf *subject)
 222 {
 223         char *pos;
 224         size_t remove;
 225         while (subject->len) {
 226                 switch (*subject->buf) {
 227                 case 'r': case 'R':
 228                         if (subject->len <= 3)
 229                                 break;
 230                         if (!memcmp(subject->buf + 1, "e:", 2)) {
 231                                 strbuf_remove(subject, 0, 3);
 232                                 continue;
 233                         }
 234                         break;
 235                 case ' ': case '\t': case ':':
 236                         strbuf_remove(subject, 0, 1);
 237                         continue;
 238                 case '[':
 239                         if ((pos = strchr(subject->buf, ']'))) {
 240                                 remove = pos - subject->buf;
 241                                 if (remove <= (subject->len - remove) * 2) {
 242                                         strbuf_remove(subject, 0, remove + 1);
 243                                         continue;
 244                                 }
 245                         } else
 246                                 strbuf_remove(subject, 0, 1);
 247                         break;
 248                 }
 249                 strbuf_trim(subject);
 250                 return;
 251         }
 252 }
 253
 254 static void cleanup_space(struct strbuf *sb)
 255 {
 256         size_t pos, cnt;
 257         for (pos = 0; pos < sb->len; pos++) {
 258                 if (isspace(sb->buf[pos])) {
 259                         sb->buf[pos] = ' ';
 260                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 261                         strbuf_remove(sb, pos + 1, cnt);
 262                 }
 263         }
 264 }
 265
 266 static void decode_header(struct strbuf *line);
 267 static const char *header[MAX_HDR_PARSED] = {
 268         "From","Subject","Date",
 269 };
 270
 271 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 272 {
 273         int len = strlen(hdr);
 274         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 275                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 276 }
 277
 278 static int check_header(const struct strbuf *line,
 279                                 struct strbuf *hdr_data[], int overwrite)
 280 {
 281         int i, ret = 0, len;
 282         struct strbuf sb = STRBUF_INIT;
 283         /* search for the interesting parts */
 284         for (i = 0; header[i]; i++) {
 285                 int len = strlen(header[i]);
 286                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 287                         /* Unwrap inline B and Q encoding, and optionally
 288                          * normalize the meta information to utf8.
 289                          */
 290                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 291                         decode_header(&sb);
 292                         handle_header(&hdr_data[i], &sb);
 293                         ret = 1;
 294                         goto check_header_out;
 295                 }
 296         }
 297
 298         /* Content stuff */
 299         if (cmp_header(line, "Content-Type")) {
 300                 len = strlen("Content-Type: ");
 301                 strbuf_add(&sb, line->buf + len, line->len - len);
 302                 decode_header(&sb);
 303                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 304                 handle_content_type(&sb);
 305                 ret = 1;
 306                 goto check_header_out;
 307         }
 308         if (cmp_header(line, "Content-Transfer-Encoding")) {
 309                 len = strlen("Content-Transfer-Encoding: ");
 310                 strbuf_add(&sb, line->buf + len, line->len - len);
 311                 decode_header(&sb);
 312                 handle_content_transfer_encoding(&sb);
 313                 ret = 1;
 314                 goto check_header_out;
 315         }
 316
 317         /* for inbody stuff */
 318         if (!prefixcmp(line->buf, ">From") && isspace(line->buf[5])) {
 319                 ret = 1; /* Should this return 0? */
 320                 goto check_header_out;
 321         }
 322         if (!prefixcmp(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 323                 for (i = 0; header[i]; i++) {
 324                         if (!memcmp("Subject", header[i], 7)) {
 325                                 handle_header(&hdr_data[i], line);
 326                                 ret = 1;
 327                                 goto check_header_out;
 328                         }
 329                 }
 330         }
 331
 332 check_header_out:
 333         strbuf_release(&sb);
 334         return ret;
 335 }
 336
 337 static int is_rfc2822_header(const struct strbuf *line)
 338 {
 339         /*
 340          * The section that defines the loosest possible
 341          * field name is "3.6.8 Optional fields".
 342          *
 343          * optional-field = field-name ":" unstructured CRLF
 344          * field-name = 1*ftext
 345          * ftext = %d33-57 / %59-126
 346          */
 347         int ch;
 348         char *cp = line->buf;
 349
 350         /* Count mbox From headers as headers */
 351         if (!prefixcmp(cp, "From ") || !prefixcmp(cp, ">From "))
 352                 return 1;
 353
 354         while ((ch = *cp++)) {
 355                 if (ch == ':')
 356                         return 1;
 357                 if ((33 <= ch && ch <= 57) ||
 358                     (59 <= ch && ch <= 126))
 359                         continue;
 360                 break;
 361         }
 362         return 0;
 363 }
 364
 365 static int read_one_header_line(struct strbuf *line, FILE *in)
 366 {
 367         /* Get the first part of the line. */
 368         if (strbuf_getline(line, in, '\n'))
 369                 return 0;
 370
 371         /*
 372          * Is it an empty line or not a valid rfc2822 header?
 373          * If so, stop here, and return false ("not a header")
 374          */
 375         strbuf_rtrim(line);
 376         if (!line->len || !is_rfc2822_header(line)) {
 377                 /* Re-add the newline */
 378                 strbuf_addch(line, '\n');
 379                 return 0;
 380         }
 381
 382         /*
 383          * Now we need to eat all the continuation lines..
 384          * Yuck, 2822 header "folding"
 385          */
 386         for (;;) {
 387                 int peek;
 388                 struct strbuf continuation = STRBUF_INIT;
 389
 390                 peek = fgetc(in); ungetc(peek, in);
 391                 if (peek != ' ' && peek != '\t')
 392                         break;
 393                 if (strbuf_getline(&continuation, in, '\n'))
 394                         break;
 395                 continuation.buf[0] = '\n';
 396                 strbuf_rtrim(&continuation);
 397                 strbuf_addbuf(line, &continuation);
 398         }
 399
 400         return 1;
 401 }
 402
 403 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 404 {
 405         const char *in = q_seg->buf;
 406         int c;
 407         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 408         strbuf_init(out, q_seg->len);
 409
 410         while ((c = *in++) != 0) {
 411                 if (c == '=') {
 412                         int d = *in++;
 413                         if (d == '\n' || !d)
 414                                 break; /* drop trailing newline */
 415                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 416                         continue;
 417                 }
 418                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 419                         c = 0x20;
 420                 strbuf_addch(out, c);
 421         }
 422         return out;
 423 }
 424
 425 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 426 {
 427         /* Decode in..ep, possibly in-place to ot */
 428         int c, pos = 0, acc = 0;
 429         const char *in = b_seg->buf;
 430         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 431         strbuf_init(out, b_seg->len);
 432
 433         while ((c = *in++) != 0) {
 434                 if (c == '+')
 435                         c = 62;
 436                 else if (c == '/')
 437                         c = 63;
 438                 else if ('A' <= c && c <= 'Z')
 439                         c -= 'A';
 440                 else if ('a' <= c && c <= 'z')
 441                         c -= 'a' - 26;
 442                 else if ('0' <= c && c <= '9')
 443                         c -= '0' - 52;
 444                 else
 445                         continue; /* garbage */
 446                 switch (pos++) {
 447                 case 0:
 448                         acc = (c << 2);
 449                         break;
 450                 case 1:
 451                         strbuf_addch(out, (acc | (c >> 4)));
 452                         acc = (c & 15) << 4;
 453                         break;
 454                 case 2:
 455                         strbuf_addch(out, (acc | (c >> 2)));
 456                         acc = (c & 3) << 6;
 457                         break;
 458                 case 3:
 459                         strbuf_addch(out, (acc | c));
 460                         acc = pos = 0;
 461                         break;
 462                 }
 463         }
 464         return out;
 465 }
 466
 467 /*
 468  * When there is no known charset, guess.
 469  *
 470  * Right now we assume that if the target is UTF-8 (the default),
 471  * and it already looks like UTF-8 (which includes US-ASCII as its
 472  * subset, of course) then that is what it is and there is nothing
 473  * to do.
 474  *
 475  * Otherwise, we default to assuming it is Latin1 for historical
 476  * reasons.
 477  */
 478 static const char *guess_charset(const struct strbuf *line, const char *target_charset)
 479 {
 480         if (is_encoding_utf8(target_charset)) {
 481                 if (is_utf8(line->buf))
 482                         return NULL;
 483         }
 484         return "latin1";
 485 }
 486
 487 static void convert_to_utf8(struct strbuf *line, const char *charset)
 488 {
 489         char *out;
 490
 491         if (!charset || !*charset) {
 492                 charset = guess_charset(line, metainfo_charset);
 493                 if (!charset)
 494                         return;
 495         }
 496
 497         if (!strcmp(metainfo_charset, charset))
 498                 return;
 499         out = reencode_string(line->buf, metainfo_charset, charset);
 500         if (!out)
 501                 die("cannot convert from %s to %s",
 502                     charset, metainfo_charset);
 503         strbuf_attach(line, out, strlen(out), strlen(out));
 504 }
 505
 506 static int decode_header_bq(struct strbuf *it)
 507 {
 508         char *in, *ep, *cp;
 509         struct strbuf outbuf = STRBUF_INIT, *dec;
 510         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 511         int rfc2047 = 0;
 512
 513         in = it->buf;
 514         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 515                 int encoding;
 516                 strbuf_reset(&charset_q);
 517                 strbuf_reset(&piecebuf);
 518                 rfc2047 = 1;
 519
 520                 if (in != ep) {
 521                         /*
 522                          * We are about to process an encoded-word
 523                          * that begins at ep, but there is something
 524                          * before the encoded word.
 525                          */
 526                         char *scan;
 527                         for (scan = in; scan < ep; scan++)
 528                                 if (!isspace(*scan))
 529                                         break;
 530
 531                         if (scan != ep || in == it->buf) {
 532                                 /*
 533                                  * We should not lose that "something",
 534                                  * unless we have just processed an
 535                                  * encoded-word, and there is only LWS
 536                                  * before the one we are about to process.
 537                                  */
 538                                 strbuf_add(&outbuf, in, ep - in);
 539                         }
 540                         in = ep;
 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                 strbuf_tolower(&charset_q);
 555
 556                 encoding = cp[1];
 557                 if (!encoding || cp[2] != '?')
 558                         goto decode_header_bq_out;
 559                 ep = strstr(cp + 3, "?=");
 560                 if (!ep)
 561                         goto decode_header_bq_out;
 562                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 563                 switch (tolower(encoding)) {
 564                 default:
 565                         goto decode_header_bq_out;
 566                 case 'b':
 567                         dec = decode_b_segment(&piecebuf);
 568                         break;
 569                 case 'q':
 570                         dec = decode_q_segment(&piecebuf, 1);
 571                         break;
 572                 }
 573                 if (metainfo_charset)
 574                         convert_to_utf8(dec, charset_q.buf);
 575
 576                 strbuf_addbuf(&outbuf, dec);
 577                 strbuf_release(dec);
 578                 free(dec);
 579                 in = ep + 2;
 580         }
 581         strbuf_addstr(&outbuf, in);
 582         strbuf_reset(it);
 583         strbuf_addbuf(it, &outbuf);
 584 decode_header_bq_out:
 585         strbuf_release(&outbuf);
 586         strbuf_release(&charset_q);
 587         strbuf_release(&piecebuf);
 588         return rfc2047;
 589 }
 590
 591 static void decode_header(struct strbuf *it)
 592 {
 593         if (decode_header_bq(it))
 594                 return;
 595         /* otherwise "it" is a straight copy of the input.
 596          * This can be binary guck but there is no charset specified.
 597          */
 598         if (metainfo_charset)
 599                 convert_to_utf8(it, "");
 600 }
 601
 602 static void decode_transfer_encoding(struct strbuf *line)
 603 {
 604         struct strbuf *ret;
 605
 606         switch (transfer_encoding) {
 607         case TE_QP:
 608                 ret = decode_q_segment(line, 0);
 609                 break;
 610         case TE_BASE64:
 611                 ret = decode_b_segment(line);
 612                 break;
 613         case TE_DONTCARE:
 614         default:
 615                 return;
 616         }
 617         strbuf_reset(line);
 618         strbuf_addbuf(line, ret);
 619         strbuf_release(ret);
 620         free(ret);
 621 }
 622
 623 static void handle_filter(struct strbuf *line);
 624
 625 static int find_boundary(void)
 626 {
 627         while (!strbuf_getline(&line, fin, '\n')) {
 628                 if (*content_top && is_multipart_boundary(&line))
 629                         return 1;
 630         }
 631         return 0;
 632 }
 633
 634 static int handle_boundary(void)
 635 {
 636         struct strbuf newline = STRBUF_INIT;
 637
 638         strbuf_addch(&newline, '\n');
 639 again:
 640         if (line.len >= (*content_top)->len + 2 &&
 641             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 642                 /* we hit an end boundary */
 643                 /* pop the current boundary off the stack */
 644                 strbuf_release(*content_top);
 645                 free(*content_top);
 646                 *content_top = NULL;
 647
 648                 /* technically won't happen as is_multipart_boundary()
 649                    will fail first.  But just in case..
 650                  */
 651                 if (--content_top < content) {
 652                         fprintf(stderr, "Detected mismatched boundaries, "
 653                                         "can't recover\n");
 654                         exit(1);
 655                 }
 656                 handle_filter(&newline);
 657                 strbuf_release(&newline);
 658
 659                 /* skip to the next boundary */
 660                 if (!find_boundary())
 661                         return 0;
 662                 goto again;
 663         }
 664
 665         /* set some defaults */
 666         transfer_encoding = TE_DONTCARE;
 667         strbuf_reset(&charset);
 668         message_type = TYPE_TEXT;
 669
 670         /* slurp in this section's info */
 671         while (read_one_header_line(&line, fin))
 672                 check_header(&line, p_hdr_data, 0);
 673
 674         strbuf_release(&newline);
 675         /* replenish line */
 676         if (strbuf_getline(&line, fin, '\n'))
 677                 return 0;
 678         strbuf_addch(&line, '\n');
 679         return 1;
 680 }
 681
 682 static inline int patchbreak(const struct strbuf *line)
 683 {
 684         size_t i;
 685
 686         /* Beginning of a "diff -" header? */
 687         if (!prefixcmp(line->buf, "diff -"))
 688                 return 1;
 689
 690         /* CVS "Index: " line? */
 691         if (!prefixcmp(line->buf, "Index: "))
 692                 return 1;
 693
 694         /*
 695          * "--- <filename>" starts patches without headers
 696          * "---<sp>*" is a manual separator
 697          */
 698         if (line->len < 4)
 699                 return 0;
 700
 701         if (!prefixcmp(line->buf, "---")) {
 702                 /* space followed by a filename? */
 703                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 704                         return 1;
 705                 /* Just whitespace? */
 706                 for (i = 3; i < line->len; i++) {
 707                         unsigned char c = line->buf[i];
 708                         if (c == '\n')
 709                                 return 1;
 710                         if (!isspace(c))
 711                                 break;
 712                 }
 713                 return 0;
 714         }
 715         return 0;
 716 }
 717
 718 static int handle_commit_msg(struct strbuf *line)
 719 {
 720         static int still_looking = 1;
 721
 722         if (!cmitmsg)
 723                 return 0;
 724
 725         if (still_looking) {
 726                 strbuf_ltrim(line);
 727                 if (!line->len)
 728                         return 0;
 729                 if ((still_looking = check_header(line, s_hdr_data, 0)) != 0)
 730                         return 0;
 731         }
 732
 733         /* normalize the log message to UTF-8. */
 734         if (metainfo_charset)
 735                 convert_to_utf8(line, charset.buf);
 736
 737         if (patchbreak(line)) {
 738                 fclose(cmitmsg);
 739                 cmitmsg = NULL;
 740                 return 1;
 741         }
 742
 743         fputs(line->buf, cmitmsg);
 744         return 0;
 745 }
 746
 747 static void handle_patch(const struct strbuf *line)
 748 {
 749         fwrite(line->buf, 1, line->len, patchfile);
 750         patch_lines++;
 751 }
 752
 753 static void handle_filter(struct strbuf *line)
 754 {
 755         static int filter = 0;
 756
 757         /* filter tells us which part we left off on */
 758         switch (filter) {
 759         case 0:
 760                 if (!handle_commit_msg(line))
 761                         break;
 762                 filter++;
 763         case 1:
 764                 handle_patch(line);
 765                 break;
 766         }
 767 }
 768
 769 static void handle_body(void)
 770 {
 771         int len = 0;
 772         struct strbuf prev = STRBUF_INIT;
 773
 774         /* Skip up to the first boundary */
 775         if (*content_top) {
 776                 if (!find_boundary())
 777                         goto handle_body_out;
 778         }
 779
 780         do {
 781                 strbuf_setlen(&line, line.len + len);
 782
 783                 /* process any boundary lines */
 784                 if (*content_top && is_multipart_boundary(&line)) {
 785                         /* flush any leftover */
 786                         if (prev.len) {
 787                                 handle_filter(&prev);
 788                                 strbuf_reset(&prev);
 789                         }
 790                         if (!handle_boundary())
 791                                 goto handle_body_out;
 792                 }
 793
 794                 /* Unwrap transfer encoding */
 795                 decode_transfer_encoding(&line);
 796
 797                 switch (transfer_encoding) {
 798                 case TE_BASE64:
 799                 case TE_QP:
 800                 {
 801                         struct strbuf **lines, **it, *sb;
 802
 803                         /* Prepend any previous partial lines */
 804                         strbuf_insert(&line, 0, prev.buf, prev.len);
 805                         strbuf_reset(&prev);
 806
 807                         /* binary data most likely doesn't have newlines */
 808                         if (message_type != TYPE_TEXT) {
 809                                 handle_filter(&line);
 810                                 break;
 811                         }
 812                         /*
 813                          * This is a decoded line that may contain
 814                          * multiple new lines.  Pass only one chunk
 815                          * at a time to handle_filter()
 816                          */
 817                         lines = strbuf_split(&line, '\n');
 818                         for (it = lines; (sb = *it); it++) {
 819                                 if (*(it + 1) == NULL) /* The last line */
 820                                         if (sb->buf[sb->len - 1] != '\n') {
 821                                                 /* Partial line, save it for later. */
 822                                                 strbuf_addbuf(&prev, sb);
 823                                                 break;
 824                                         }
 825                                 handle_filter(sb);
 826                         }
 827                         /*
 828                          * The partial chunk is saved in "prev" and will be
 829                          * appended by the next iteration of read_line_with_nul().
 830                          */
 831                         strbuf_list_free(lines);
 832                         break;
 833                 }
 834                 default:
 835                         handle_filter(&line);
 836                 }
 837
 838                 strbuf_reset(&line);
 839                 if (strbuf_avail(&line) < 100)
 840                         strbuf_grow(&line, 100);
 841         } while ((len = read_line_with_nul(line.buf, strbuf_avail(&line), fin)));
 842
 843 handle_body_out:
 844         strbuf_release(&prev);
 845 }
 846
 847 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 848 {
 849         const char *sp = data->buf;
 850         while (1) {
 851                 char *ep = strchr(sp, '\n');
 852                 int len;
 853                 if (!ep)
 854                         len = strlen(sp);
 855                 else
 856                         len = ep - sp;
 857                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 858                 if (!ep)
 859                         break;
 860                 sp = ep + 1;
 861         }
 862 }
 863
 864 static void handle_info(void)
 865 {
 866         struct strbuf *hdr;
 867         int i;
 868
 869         for (i = 0; header[i]; i++) {
 870                 /* only print inbody headers if we output a patch file */
 871                 if (patch_lines && s_hdr_data[i])
 872                         hdr = s_hdr_data[i];
 873                 else if (p_hdr_data[i])
 874                         hdr = p_hdr_data[i];
 875                 else
 876                         continue;
 877
 878                 if (!memcmp(header[i], "Subject", 7)) {
 879                         if (!keep_subject) {
 880                                 cleanup_subject(hdr);
 881                                 cleanup_space(hdr);
 882                         }
 883                         output_header_lines(fout, "Subject", hdr);
 884                 } else if (!memcmp(header[i], "From", 4)) {
 885                         cleanup_space(hdr);
 886                         handle_from(hdr);
 887                         fprintf(fout, "Author: %s\n", name.buf);
 888                         fprintf(fout, "Email: %s\n", email.buf);
 889                 } else {
 890                         cleanup_space(hdr);
 891                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 892                 }
 893         }
 894         fprintf(fout, "\n");
 895 }
 896
 897 static int mailinfo(FILE *in, FILE *out, int ks, const char *encoding,
 898                     const char *msg, const char *patch)
 899 {
 900         int peek;
 901         keep_subject = ks;
 902         metainfo_charset = encoding;
 903         fin = in;
 904         fout = out;
 905
 906         cmitmsg = fopen(msg, "w");
 907         if (!cmitmsg) {
 908                 perror(msg);
 909                 return -1;
 910         }
 911         patchfile = fopen(patch, "w");
 912         if (!patchfile) {
 913                 perror(patch);
 914                 fclose(cmitmsg);
 915                 return -1;
 916         }
 917
 918         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
 919         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
 920
 921         do {
 922                 peek = fgetc(in);
 923         } while (isspace(peek));
 924         ungetc(peek, in);
 925
 926         /* process the email header */
 927         while (read_one_header_line(&line, fin))
 928                 check_header(&line, p_hdr_data, 1);
 929
 930         handle_body();
 931         handle_info();
 932
 933         return 0;
 934 }
 935
 936 static const char mailinfo_usage[] =
 937         "git mailinfo [-k] [-u | --encoding=<encoding> | -n] msg patch <mail >info";
 938
 939 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
 940 {
 941         const char *def_charset;
 942
 943         /* NEEDSWORK: might want to do the optional .git/ directory
 944          * discovery
 945          */
 946         git_config(git_default_config, NULL);
 947
 948         def_charset = (git_commit_encoding ? git_commit_encoding : "utf-8");
 949         metainfo_charset = def_charset;
 950
 951         while (1 < argc && argv[1][0] == '-') {
 952                 if (!strcmp(argv[1], "-k"))
 953                         keep_subject = 1;
 954                 else if (!strcmp(argv[1], "-u"))
 955                         metainfo_charset = def_charset;
 956                 else if (!strcmp(argv[1], "-n"))
 957                         metainfo_charset = NULL;
 958                 else if (!prefixcmp(argv[1], "--encoding="))
 959                         metainfo_charset = argv[1] + 11;
 960                 else
 961                         usage(mailinfo_usage);
 962                 argc--; argv++;
 963         }
 964
 965         if (argc != 3)
 966                 usage(mailinfo_usage);
 967
 968         return !!mailinfo(stdin, stdout, keep_subject, metainfo_charset, argv[1], argv[2]);
 969 }