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         int brackets_removed = 0;
 226
 227         while (subject->len) {
 228                 switch (*subject->buf) {
 229                 case 'r': case 'R':
 230                         if (subject->len <= 3)
 231                                 break;
 232                         if (!memcmp(subject->buf + 1, "e:", 2)) {
 233                                 strbuf_remove(subject, 0, 3);
 234                                 continue;
 235                         }
 236                         break;
 237                 case ' ': case '\t': case ':':
 238                         strbuf_remove(subject, 0, 1);
 239                         continue;
 240                 case '[':
 241                         /* remove only one set of square brackets */
 242                         if (brackets_removed)
 243                                 break;
 244
 245                         if ((pos = strchr(subject->buf, ']'))) {
 246                                 remove = pos - subject->buf;
 247                                 if (remove <= (subject->len - remove) * 2) {
 248                                         strbuf_remove(subject, 0, remove + 1);
 249                                         brackets_removed = 1;
 250                                         continue;
 251                                 }
 252                         } else
 253                                 strbuf_remove(subject, 0, 1);
 254                         break;
 255                 }
 256                 strbuf_trim(subject);
 257                 return;
 258         }
 259 }
 260
 261 static void cleanup_space(struct strbuf *sb)
 262 {
 263         size_t pos, cnt;
 264         for (pos = 0; pos < sb->len; pos++) {
 265                 if (isspace(sb->buf[pos])) {
 266                         sb->buf[pos] = ' ';
 267                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 268                         strbuf_remove(sb, pos + 1, cnt);
 269                 }
 270         }
 271 }
 272
 273 static void decode_header(struct strbuf *line);
 274 static const char *header[MAX_HDR_PARSED] = {
 275         "From","Subject","Date",
 276 };
 277
 278 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 279 {
 280         int len = strlen(hdr);
 281         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 282                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 283 }
 284
 285 static int check_header(const struct strbuf *line,
 286                                 struct strbuf *hdr_data[], int overwrite)
 287 {
 288         int i, ret = 0, len;
 289         struct strbuf sb = STRBUF_INIT;
 290         /* search for the interesting parts */
 291         for (i = 0; header[i]; i++) {
 292                 int len = strlen(header[i]);
 293                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 294                         /* Unwrap inline B and Q encoding, and optionally
 295                          * normalize the meta information to utf8.
 296                          */
 297                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 298                         decode_header(&sb);
 299                         handle_header(&hdr_data[i], &sb);
 300                         ret = 1;
 301                         goto check_header_out;
 302                 }
 303         }
 304
 305         /* Content stuff */
 306         if (cmp_header(line, "Content-Type")) {
 307                 len = strlen("Content-Type: ");
 308                 strbuf_add(&sb, line->buf + len, line->len - len);
 309                 decode_header(&sb);
 310                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 311                 handle_content_type(&sb);
 312                 ret = 1;
 313                 goto check_header_out;
 314         }
 315         if (cmp_header(line, "Content-Transfer-Encoding")) {
 316                 len = strlen("Content-Transfer-Encoding: ");
 317                 strbuf_add(&sb, line->buf + len, line->len - len);
 318                 decode_header(&sb);
 319                 handle_content_transfer_encoding(&sb);
 320                 ret = 1;
 321                 goto check_header_out;
 322         }
 323
 324         /* for inbody stuff */
 325         if (!prefixcmp(line->buf, ">From") && isspace(line->buf[5])) {
 326                 ret = 1; /* Should this return 0? */
 327                 goto check_header_out;
 328         }
 329         if (!prefixcmp(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 330                 for (i = 0; header[i]; i++) {
 331                         if (!memcmp("Subject", header[i], 7)) {
 332                                 handle_header(&hdr_data[i], line);
 333                                 ret = 1;
 334                                 goto check_header_out;
 335                         }
 336                 }
 337         }
 338
 339 check_header_out:
 340         strbuf_release(&sb);
 341         return ret;
 342 }
 343
 344 static int is_rfc2822_header(const struct strbuf *line)
 345 {
 346         /*
 347          * The section that defines the loosest possible
 348          * field name is "3.6.8 Optional fields".
 349          *
 350          * optional-field = field-name ":" unstructured CRLF
 351          * field-name = 1*ftext
 352          * ftext = %d33-57 / %59-126
 353          */
 354         int ch;
 355         char *cp = line->buf;
 356
 357         /* Count mbox From headers as headers */
 358         if (!prefixcmp(cp, "From ") || !prefixcmp(cp, ">From "))
 359                 return 1;
 360
 361         while ((ch = *cp++)) {
 362                 if (ch == ':')
 363                         return 1;
 364                 if ((33 <= ch && ch <= 57) ||
 365                     (59 <= ch && ch <= 126))
 366                         continue;
 367                 break;
 368         }
 369         return 0;
 370 }
 371
 372 static int read_one_header_line(struct strbuf *line, FILE *in)
 373 {
 374         /* Get the first part of the line. */
 375         if (strbuf_getline(line, in, '\n'))
 376                 return 0;
 377
 378         /*
 379          * Is it an empty line or not a valid rfc2822 header?
 380          * If so, stop here, and return false ("not a header")
 381          */
 382         strbuf_rtrim(line);
 383         if (!line->len || !is_rfc2822_header(line)) {
 384                 /* Re-add the newline */
 385                 strbuf_addch(line, '\n');
 386                 return 0;
 387         }
 388
 389         /*
 390          * Now we need to eat all the continuation lines..
 391          * Yuck, 2822 header "folding"
 392          */
 393         for (;;) {
 394                 int peek;
 395                 struct strbuf continuation = STRBUF_INIT;
 396
 397                 peek = fgetc(in); ungetc(peek, in);
 398                 if (peek != ' ' && peek != '\t')
 399                         break;
 400                 if (strbuf_getline(&continuation, in, '\n'))
 401                         break;
 402                 continuation.buf[0] = '\n';
 403                 strbuf_rtrim(&continuation);
 404                 strbuf_addbuf(line, &continuation);
 405         }
 406
 407         return 1;
 408 }
 409
 410 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 411 {
 412         const char *in = q_seg->buf;
 413         int c;
 414         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 415         strbuf_init(out, q_seg->len);
 416
 417         while ((c = *in++) != 0) {
 418                 if (c == '=') {
 419                         int d = *in++;
 420                         if (d == '\n' || !d)
 421                                 break; /* drop trailing newline */
 422                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 423                         continue;
 424                 }
 425                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 426                         c = 0x20;
 427                 strbuf_addch(out, c);
 428         }
 429         return out;
 430 }
 431
 432 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 433 {
 434         /* Decode in..ep, possibly in-place to ot */
 435         int c, pos = 0, acc = 0;
 436         const char *in = b_seg->buf;
 437         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 438         strbuf_init(out, b_seg->len);
 439
 440         while ((c = *in++) != 0) {
 441                 if (c == '+')
 442                         c = 62;
 443                 else if (c == '/')
 444                         c = 63;
 445                 else if ('A' <= c && c <= 'Z')
 446                         c -= 'A';
 447                 else if ('a' <= c && c <= 'z')
 448                         c -= 'a' - 26;
 449                 else if ('0' <= c && c <= '9')
 450                         c -= '0' - 52;
 451                 else
 452                         continue; /* garbage */
 453                 switch (pos++) {
 454                 case 0:
 455                         acc = (c << 2);
 456                         break;
 457                 case 1:
 458                         strbuf_addch(out, (acc | (c >> 4)));
 459                         acc = (c & 15) << 4;
 460                         break;
 461                 case 2:
 462                         strbuf_addch(out, (acc | (c >> 2)));
 463                         acc = (c & 3) << 6;
 464                         break;
 465                 case 3:
 466                         strbuf_addch(out, (acc | c));
 467                         acc = pos = 0;
 468                         break;
 469                 }
 470         }
 471         return out;
 472 }
 473
 474 /*
 475  * When there is no known charset, guess.
 476  *
 477  * Right now we assume that if the target is UTF-8 (the default),
 478  * and it already looks like UTF-8 (which includes US-ASCII as its
 479  * subset, of course) then that is what it is and there is nothing
 480  * to do.
 481  *
 482  * Otherwise, we default to assuming it is Latin1 for historical
 483  * reasons.
 484  */
 485 static const char *guess_charset(const struct strbuf *line, const char *target_charset)
 486 {
 487         if (is_encoding_utf8(target_charset)) {
 488                 if (is_utf8(line->buf))
 489                         return NULL;
 490         }
 491         return "latin1";
 492 }
 493
 494 static void convert_to_utf8(struct strbuf *line, const char *charset)
 495 {
 496         char *out;
 497
 498         if (!charset || !*charset) {
 499                 charset = guess_charset(line, metainfo_charset);
 500                 if (!charset)
 501                         return;
 502         }
 503
 504         if (!strcmp(metainfo_charset, charset))
 505                 return;
 506         out = reencode_string(line->buf, metainfo_charset, charset);
 507         if (!out)
 508                 die("cannot convert from %s to %s",
 509                     charset, metainfo_charset);
 510         strbuf_attach(line, out, strlen(out), strlen(out));
 511 }
 512
 513 static int decode_header_bq(struct strbuf *it)
 514 {
 515         char *in, *ep, *cp;
 516         struct strbuf outbuf = STRBUF_INIT, *dec;
 517         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 518         int rfc2047 = 0;
 519
 520         in = it->buf;
 521         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 522                 int encoding;
 523                 strbuf_reset(&charset_q);
 524                 strbuf_reset(&piecebuf);
 525                 rfc2047 = 1;
 526
 527                 if (in != ep) {
 528                         /*
 529                          * We are about to process an encoded-word
 530                          * that begins at ep, but there is something
 531                          * before the encoded word.
 532                          */
 533                         char *scan;
 534                         for (scan = in; scan < ep; scan++)
 535                                 if (!isspace(*scan))
 536                                         break;
 537
 538                         if (scan != ep || in == it->buf) {
 539                                 /*
 540                                  * We should not lose that "something",
 541                                  * unless we have just processed an
 542                                  * encoded-word, and there is only LWS
 543                                  * before the one we are about to process.
 544                                  */
 545                                 strbuf_add(&outbuf, in, ep - in);
 546                         }
 547                 }
 548                 /* E.g.
 549                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 550                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 551                  */
 552                 ep += 2;
 553
 554                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 555                         goto decode_header_bq_out;
 556
 557                 if (cp + 3 - it->buf > it->len)
 558                         goto decode_header_bq_out;
 559                 strbuf_add(&charset_q, ep, cp - ep);
 560                 strbuf_tolower(&charset_q);
 561
 562                 encoding = cp[1];
 563                 if (!encoding || cp[2] != '?')
 564                         goto decode_header_bq_out;
 565                 ep = strstr(cp + 3, "?=");
 566                 if (!ep)
 567                         goto decode_header_bq_out;
 568                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 569                 switch (tolower(encoding)) {
 570                 default:
 571                         goto decode_header_bq_out;
 572                 case 'b':
 573                         dec = decode_b_segment(&piecebuf);
 574                         break;
 575                 case 'q':
 576                         dec = decode_q_segment(&piecebuf, 1);
 577                         break;
 578                 }
 579                 if (metainfo_charset)
 580                         convert_to_utf8(dec, charset_q.buf);
 581
 582                 strbuf_addbuf(&outbuf, dec);
 583                 strbuf_release(dec);
 584                 free(dec);
 585                 in = ep + 2;
 586         }
 587         strbuf_addstr(&outbuf, in);
 588         strbuf_reset(it);
 589         strbuf_addbuf(it, &outbuf);
 590 decode_header_bq_out:
 591         strbuf_release(&outbuf);
 592         strbuf_release(&charset_q);
 593         strbuf_release(&piecebuf);
 594         return rfc2047;
 595 }
 596
 597 static void decode_header(struct strbuf *it)
 598 {
 599         if (decode_header_bq(it))
 600                 return;
 601         /* otherwise "it" is a straight copy of the input.
 602          * This can be binary guck but there is no charset specified.
 603          */
 604         if (metainfo_charset)
 605                 convert_to_utf8(it, "");
 606 }
 607
 608 static void decode_transfer_encoding(struct strbuf *line)
 609 {
 610         struct strbuf *ret;
 611
 612         switch (transfer_encoding) {
 613         case TE_QP:
 614                 ret = decode_q_segment(line, 0);
 615                 break;
 616         case TE_BASE64:
 617                 ret = decode_b_segment(line);
 618                 break;
 619         case TE_DONTCARE:
 620         default:
 621                 return;
 622         }
 623         strbuf_reset(line);
 624         strbuf_addbuf(line, ret);
 625         strbuf_release(ret);
 626         free(ret);
 627 }
 628
 629 static void handle_filter(struct strbuf *line);
 630
 631 static int find_boundary(void)
 632 {
 633         while (!strbuf_getline(&line, fin, '\n')) {
 634                 if (*content_top && is_multipart_boundary(&line))
 635                         return 1;
 636         }
 637         return 0;
 638 }
 639
 640 static int handle_boundary(void)
 641 {
 642         struct strbuf newline = STRBUF_INIT;
 643
 644         strbuf_addch(&newline, '\n');
 645 again:
 646         if (line.len >= (*content_top)->len + 2 &&
 647             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 648                 /* we hit an end boundary */
 649                 /* pop the current boundary off the stack */
 650                 strbuf_release(*content_top);
 651                 free(*content_top);
 652                 *content_top = NULL;
 653
 654                 /* technically won't happen as is_multipart_boundary()
 655                    will fail first.  But just in case..
 656                  */
 657                 if (--content_top < content) {
 658                         fprintf(stderr, "Detected mismatched boundaries, "
 659                                         "can't recover\n");
 660                         exit(1);
 661                 }
 662                 handle_filter(&newline);
 663                 strbuf_release(&newline);
 664
 665                 /* skip to the next boundary */
 666                 if (!find_boundary())
 667                         return 0;
 668                 goto again;
 669         }
 670
 671         /* set some defaults */
 672         transfer_encoding = TE_DONTCARE;
 673         strbuf_reset(&charset);
 674         message_type = TYPE_TEXT;
 675
 676         /* slurp in this section's info */
 677         while (read_one_header_line(&line, fin))
 678                 check_header(&line, p_hdr_data, 0);
 679
 680         strbuf_release(&newline);
 681         /* replenish line */
 682         if (strbuf_getline(&line, fin, '\n'))
 683                 return 0;
 684         strbuf_addch(&line, '\n');
 685         return 1;
 686 }
 687
 688 static inline int patchbreak(const struct strbuf *line)
 689 {
 690         size_t i;
 691
 692         /* Beginning of a "diff -" header? */
 693         if (!prefixcmp(line->buf, "diff -"))
 694                 return 1;
 695
 696         /* CVS "Index: " line? */
 697         if (!prefixcmp(line->buf, "Index: "))
 698                 return 1;
 699
 700         /*
 701          * "--- <filename>" starts patches without headers
 702          * "---<sp>*" is a manual separator
 703          */
 704         if (line->len < 4)
 705                 return 0;
 706
 707         if (!prefixcmp(line->buf, "---")) {
 708                 /* space followed by a filename? */
 709                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 710                         return 1;
 711                 /* Just whitespace? */
 712                 for (i = 3; i < line->len; i++) {
 713                         unsigned char c = line->buf[i];
 714                         if (c == '\n')
 715                                 return 1;
 716                         if (!isspace(c))
 717                                 break;
 718                 }
 719                 return 0;
 720         }
 721         return 0;
 722 }
 723
 724 static int handle_commit_msg(struct strbuf *line)
 725 {
 726         static int still_looking = 1;
 727
 728         if (!cmitmsg)
 729                 return 0;
 730
 731         if (still_looking) {
 732                 strbuf_ltrim(line);
 733                 if (!line->len)
 734                         return 0;
 735                 if ((still_looking = check_header(line, s_hdr_data, 0)) != 0)
 736                         return 0;
 737         }
 738
 739         /* normalize the log message to UTF-8. */
 740         if (metainfo_charset)
 741                 convert_to_utf8(line, charset.buf);
 742
 743         if (patchbreak(line)) {
 744                 fclose(cmitmsg);
 745                 cmitmsg = NULL;
 746                 return 1;
 747         }
 748
 749         fputs(line->buf, cmitmsg);
 750         return 0;
 751 }
 752
 753 static void handle_patch(const struct strbuf *line)
 754 {
 755         fwrite(line->buf, 1, line->len, patchfile);
 756         patch_lines++;
 757 }
 758
 759 static void handle_filter(struct strbuf *line)
 760 {
 761         static int filter = 0;
 762
 763         /* filter tells us which part we left off on */
 764         switch (filter) {
 765         case 0:
 766                 if (!handle_commit_msg(line))
 767                         break;
 768                 filter++;
 769         case 1:
 770                 handle_patch(line);
 771                 break;
 772         }
 773 }
 774
 775 static void handle_body(void)
 776 {
 777         int len = 0;
 778         struct strbuf prev = STRBUF_INIT;
 779
 780         /* Skip up to the first boundary */
 781         if (*content_top) {
 782                 if (!find_boundary())
 783                         goto handle_body_out;
 784         }
 785
 786         do {
 787                 strbuf_setlen(&line, line.len + len);
 788
 789                 /* process any boundary lines */
 790                 if (*content_top && is_multipart_boundary(&line)) {
 791                         /* flush any leftover */
 792                         if (prev.len) {
 793                                 handle_filter(&prev);
 794                                 strbuf_reset(&prev);
 795                         }
 796                         if (!handle_boundary())
 797                                 goto handle_body_out;
 798                 }
 799
 800                 /* Unwrap transfer encoding */
 801                 decode_transfer_encoding(&line);
 802
 803                 switch (transfer_encoding) {
 804                 case TE_BASE64:
 805                 case TE_QP:
 806                 {
 807                         struct strbuf **lines, **it, *sb;
 808
 809                         /* Prepend any previous partial lines */
 810                         strbuf_insert(&line, 0, prev.buf, prev.len);
 811                         strbuf_reset(&prev);
 812
 813                         /* binary data most likely doesn't have newlines */
 814                         if (message_type != TYPE_TEXT) {
 815                                 handle_filter(&line);
 816                                 break;
 817                         }
 818                         /*
 819                          * This is a decoded line that may contain
 820                          * multiple new lines.  Pass only one chunk
 821                          * at a time to handle_filter()
 822                          */
 823                         lines = strbuf_split(&line, '\n');
 824                         for (it = lines; (sb = *it); it++) {
 825                                 if (*(it + 1) == NULL) /* The last line */
 826                                         if (sb->buf[sb->len - 1] != '\n') {
 827                                                 /* Partial line, save it for later. */
 828                                                 strbuf_addbuf(&prev, sb);
 829                                                 break;
 830                                         }
 831                                 handle_filter(sb);
 832                         }
 833                         /*
 834                          * The partial chunk is saved in "prev" and will be
 835                          * appended by the next iteration of read_line_with_nul().
 836                          */
 837                         strbuf_list_free(lines);
 838                         break;
 839                 }
 840                 default:
 841                         handle_filter(&line);
 842                 }
 843
 844                 strbuf_reset(&line);
 845                 if (strbuf_avail(&line) < 100)
 846                         strbuf_grow(&line, 100);
 847         } while ((len = read_line_with_nul(line.buf, strbuf_avail(&line), fin)));
 848
 849 handle_body_out:
 850         strbuf_release(&prev);
 851 }
 852
 853 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 854 {
 855         const char *sp = data->buf;
 856         while (1) {
 857                 char *ep = strchr(sp, '\n');
 858                 int len;
 859                 if (!ep)
 860                         len = strlen(sp);
 861                 else
 862                         len = ep - sp;
 863                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 864                 if (!ep)
 865                         break;
 866                 sp = ep + 1;
 867         }
 868 }
 869
 870 static void handle_info(void)
 871 {
 872         struct strbuf *hdr;
 873         int i;
 874
 875         for (i = 0; header[i]; i++) {
 876                 /* only print inbody headers if we output a patch file */
 877                 if (patch_lines && s_hdr_data[i])
 878                         hdr = s_hdr_data[i];
 879                 else if (p_hdr_data[i])
 880                         hdr = p_hdr_data[i];
 881                 else
 882                         continue;
 883
 884                 if (!memcmp(header[i], "Subject", 7)) {
 885                         if (!keep_subject) {
 886                                 cleanup_subject(hdr);
 887                                 cleanup_space(hdr);
 888                         }
 889                         output_header_lines(fout, "Subject", hdr);
 890                 } else if (!memcmp(header[i], "From", 4)) {
 891                         cleanup_space(hdr);
 892                         handle_from(hdr);
 893                         fprintf(fout, "Author: %s\n", name.buf);
 894                         fprintf(fout, "Email: %s\n", email.buf);
 895                 } else {
 896                         cleanup_space(hdr);
 897                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 898                 }
 899         }
 900         fprintf(fout, "\n");
 901 }
 902
 903 static int mailinfo(FILE *in, FILE *out, int ks, const char *encoding,
 904                     const char *msg, const char *patch)
 905 {
 906         int peek;
 907         keep_subject = ks;
 908         metainfo_charset = encoding;
 909         fin = in;
 910         fout = out;
 911
 912         cmitmsg = fopen(msg, "w");
 913         if (!cmitmsg) {
 914                 perror(msg);
 915                 return -1;
 916         }
 917         patchfile = fopen(patch, "w");
 918         if (!patchfile) {
 919                 perror(patch);
 920                 fclose(cmitmsg);
 921                 return -1;
 922         }
 923
 924         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
 925         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
 926
 927         do {
 928                 peek = fgetc(in);
 929         } while (isspace(peek));
 930         ungetc(peek, in);
 931
 932         /* process the email header */
 933         while (read_one_header_line(&line, fin))
 934                 check_header(&line, p_hdr_data, 1);
 935
 936         handle_body();
 937         handle_info();
 938
 939         return 0;
 940 }
 941
 942 static const char mailinfo_usage[] =
 943         "git mailinfo [-k] [-u | --encoding=<encoding> | -n] msg patch <mail >info";
 944
 945 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
 946 {
 947         const char *def_charset;
 948
 949         /* NEEDSWORK: might want to do the optional .git/ directory
 950          * discovery
 951          */
 952         git_config(git_default_config, NULL);
 953
 954         def_charset = (git_commit_encoding ? git_commit_encoding : "utf-8");
 955         metainfo_charset = def_charset;
 956
 957         while (1 < argc && argv[1][0] == '-') {
 958                 if (!strcmp(argv[1], "-k"))
 959                         keep_subject = 1;
 960                 else if (!strcmp(argv[1], "-u"))
 961                         metainfo_charset = def_charset;
 962                 else if (!strcmp(argv[1], "-n"))
 963                         metainfo_charset = NULL;
 964                 else if (!prefixcmp(argv[1], "--encoding="))
 965                         metainfo_charset = argv[1] + 11;
 966                 else
 967                         usage(mailinfo_usage);
 968                 argc--; argv++;
 969         }
 970
 971         if (argc != 3)
 972                 usage(mailinfo_usage);
 973
 974         return !!mailinfo(stdin, stdout, keep_subject, metainfo_charset, argv[1], argv[2]);
 975 }