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         slurp_attr(line->buf, "charset=", &charset);
 197
 198         if (boundary) {
 199                 strbuf_release(boundary);
 200                 free(boundary);
 201         }
 202 }
 203
 204 static void handle_content_transfer_encoding(const struct strbuf *line)
 205 {
 206         if (strcasestr(line->buf, "base64"))
 207                 transfer_encoding = TE_BASE64;
 208         else if (strcasestr(line->buf, "quoted-printable"))
 209                 transfer_encoding = TE_QP;
 210         else
 211                 transfer_encoding = TE_DONTCARE;
 212 }
 213
 214 static int is_multipart_boundary(const struct strbuf *line)
 215 {
 216         return (((*content_top)->len <= line->len) &&
 217                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
 218 }
 219
 220 static void cleanup_subject(struct strbuf *subject)
 221 {
 222         char *pos;
 223         size_t remove;
 224         int brackets_removed = 0;
 225
 226         while (subject->len) {
 227                 switch (*subject->buf) {
 228                 case 'r': case 'R':
 229                         if (subject->len <= 3)
 230                                 break;
 231                         if (!memcmp(subject->buf + 1, "e:", 2)) {
 232                                 strbuf_remove(subject, 0, 3);
 233                                 continue;
 234                         }
 235                         break;
 236                 case ' ': case '\t': case ':':
 237                         strbuf_remove(subject, 0, 1);
 238                         continue;
 239                 case '[':
 240                         /* remove only one set of square brackets */
 241                         if (brackets_removed)
 242                                 break;
 243
 244                         if ((pos = strchr(subject->buf, ']'))) {
 245                                 remove = pos - subject->buf;
 246                                 if (remove <= (subject->len - remove) * 2) {
 247                                         strbuf_remove(subject, 0, remove + 1);
 248                                         brackets_removed = 1;
 249                                         continue;
 250                                 }
 251                         } else
 252                                 strbuf_remove(subject, 0, 1);
 253                         break;
 254                 }
 255                 strbuf_trim(subject);
 256                 return;
 257         }
 258 }
 259
 260 static void cleanup_space(struct strbuf *sb)
 261 {
 262         size_t pos, cnt;
 263         for (pos = 0; pos < sb->len; pos++) {
 264                 if (isspace(sb->buf[pos])) {
 265                         sb->buf[pos] = ' ';
 266                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 267                         strbuf_remove(sb, pos + 1, cnt);
 268                 }
 269         }
 270 }
 271
 272 static void decode_header(struct strbuf *line);
 273 static const char *header[MAX_HDR_PARSED] = {
 274         "From","Subject","Date",
 275 };
 276
 277 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 278 {
 279         int len = strlen(hdr);
 280         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 281                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 282 }
 283
 284 static int check_header(const struct strbuf *line,
 285                                 struct strbuf *hdr_data[], int overwrite)
 286 {
 287         int i, ret = 0, len;
 288         struct strbuf sb = STRBUF_INIT;
 289         /* search for the interesting parts */
 290         for (i = 0; header[i]; i++) {
 291                 int len = strlen(header[i]);
 292                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 293                         /* Unwrap inline B and Q encoding, and optionally
 294                          * normalize the meta information to utf8.
 295                          */
 296                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 297                         decode_header(&sb);
 298                         handle_header(&hdr_data[i], &sb);
 299                         ret = 1;
 300                         goto check_header_out;
 301                 }
 302         }
 303
 304         /* Content stuff */
 305         if (cmp_header(line, "Content-Type")) {
 306                 len = strlen("Content-Type: ");
 307                 strbuf_add(&sb, line->buf + len, line->len - len);
 308                 decode_header(&sb);
 309                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 310                 handle_content_type(&sb);
 311                 ret = 1;
 312                 goto check_header_out;
 313         }
 314         if (cmp_header(line, "Content-Transfer-Encoding")) {
 315                 len = strlen("Content-Transfer-Encoding: ");
 316                 strbuf_add(&sb, line->buf + len, line->len - len);
 317                 decode_header(&sb);
 318                 handle_content_transfer_encoding(&sb);
 319                 ret = 1;
 320                 goto check_header_out;
 321         }
 322
 323         /* for inbody stuff */
 324         if (!prefixcmp(line->buf, ">From") && isspace(line->buf[5])) {
 325                 ret = 1; /* Should this return 0? */
 326                 goto check_header_out;
 327         }
 328         if (!prefixcmp(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 329                 for (i = 0; header[i]; i++) {
 330                         if (!memcmp("Subject", header[i], 7)) {
 331                                 handle_header(&hdr_data[i], line);
 332                                 ret = 1;
 333                                 goto check_header_out;
 334                         }
 335                 }
 336         }
 337
 338 check_header_out:
 339         strbuf_release(&sb);
 340         return ret;
 341 }
 342
 343 static int is_rfc2822_header(const struct strbuf *line)
 344 {
 345         /*
 346          * The section that defines the loosest possible
 347          * field name is "3.6.8 Optional fields".
 348          *
 349          * optional-field = field-name ":" unstructured CRLF
 350          * field-name = 1*ftext
 351          * ftext = %d33-57 / %59-126
 352          */
 353         int ch;
 354         char *cp = line->buf;
 355
 356         /* Count mbox From headers as headers */
 357         if (!prefixcmp(cp, "From ") || !prefixcmp(cp, ">From "))
 358                 return 1;
 359
 360         while ((ch = *cp++)) {
 361                 if (ch == ':')
 362                         return 1;
 363                 if ((33 <= ch && ch <= 57) ||
 364                     (59 <= ch && ch <= 126))
 365                         continue;
 366                 break;
 367         }
 368         return 0;
 369 }
 370
 371 static int read_one_header_line(struct strbuf *line, FILE *in)
 372 {
 373         /* Get the first part of the line. */
 374         if (strbuf_getline(line, in, '\n'))
 375                 return 0;
 376
 377         /*
 378          * Is it an empty line or not a valid rfc2822 header?
 379          * If so, stop here, and return false ("not a header")
 380          */
 381         strbuf_rtrim(line);
 382         if (!line->len || !is_rfc2822_header(line)) {
 383                 /* Re-add the newline */
 384                 strbuf_addch(line, '\n');
 385                 return 0;
 386         }
 387
 388         /*
 389          * Now we need to eat all the continuation lines..
 390          * Yuck, 2822 header "folding"
 391          */
 392         for (;;) {
 393                 int peek;
 394                 struct strbuf continuation = STRBUF_INIT;
 395
 396                 peek = fgetc(in); ungetc(peek, in);
 397                 if (peek != ' ' && peek != '\t')
 398                         break;
 399                 if (strbuf_getline(&continuation, in, '\n'))
 400                         break;
 401                 continuation.buf[0] = '\n';
 402                 strbuf_rtrim(&continuation);
 403                 strbuf_addbuf(line, &continuation);
 404         }
 405
 406         return 1;
 407 }
 408
 409 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 410 {
 411         const char *in = q_seg->buf;
 412         int c;
 413         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 414         strbuf_init(out, q_seg->len);
 415
 416         while ((c = *in++) != 0) {
 417                 if (c == '=') {
 418                         int d = *in++;
 419                         if (d == '\n' || !d)
 420                                 break; /* drop trailing newline */
 421                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 422                         continue;
 423                 }
 424                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 425                         c = 0x20;
 426                 strbuf_addch(out, c);
 427         }
 428         return out;
 429 }
 430
 431 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 432 {
 433         /* Decode in..ep, possibly in-place to ot */
 434         int c, pos = 0, acc = 0;
 435         const char *in = b_seg->buf;
 436         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 437         strbuf_init(out, b_seg->len);
 438
 439         while ((c = *in++) != 0) {
 440                 if (c == '+')
 441                         c = 62;
 442                 else if (c == '/')
 443                         c = 63;
 444                 else if ('A' <= c && c <= 'Z')
 445                         c -= 'A';
 446                 else if ('a' <= c && c <= 'z')
 447                         c -= 'a' - 26;
 448                 else if ('0' <= c && c <= '9')
 449                         c -= '0' - 52;
 450                 else
 451                         continue; /* garbage */
 452                 switch (pos++) {
 453                 case 0:
 454                         acc = (c << 2);
 455                         break;
 456                 case 1:
 457                         strbuf_addch(out, (acc | (c >> 4)));
 458                         acc = (c & 15) << 4;
 459                         break;
 460                 case 2:
 461                         strbuf_addch(out, (acc | (c >> 2)));
 462                         acc = (c & 3) << 6;
 463                         break;
 464                 case 3:
 465                         strbuf_addch(out, (acc | c));
 466                         acc = pos = 0;
 467                         break;
 468                 }
 469         }
 470         return out;
 471 }
 472
 473 /*
 474  * When there is no known charset, guess.
 475  *
 476  * Right now we assume that if the target is UTF-8 (the default),
 477  * and it already looks like UTF-8 (which includes US-ASCII as its
 478  * subset, of course) then that is what it is and there is nothing
 479  * to do.
 480  *
 481  * Otherwise, we default to assuming it is Latin1 for historical
 482  * reasons.
 483  */
 484 static const char *guess_charset(const struct strbuf *line, const char *target_charset)
 485 {
 486         if (is_encoding_utf8(target_charset)) {
 487                 if (is_utf8(line->buf))
 488                         return NULL;
 489         }
 490         return "ISO8859-1";
 491 }
 492
 493 static void convert_to_utf8(struct strbuf *line, const char *charset)
 494 {
 495         char *out;
 496
 497         if (!charset || !*charset) {
 498                 charset = guess_charset(line, metainfo_charset);
 499                 if (!charset)
 500                         return;
 501         }
 502
 503         if (!strcasecmp(metainfo_charset, charset))
 504                 return;
 505         out = reencode_string(line->buf, metainfo_charset, charset);
 506         if (!out)
 507                 die("cannot convert from %s to %s",
 508                     charset, metainfo_charset);
 509         strbuf_attach(line, out, strlen(out), strlen(out));
 510 }
 511
 512 static int decode_header_bq(struct strbuf *it)
 513 {
 514         char *in, *ep, *cp;
 515         struct strbuf outbuf = STRBUF_INIT, *dec;
 516         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 517         int rfc2047 = 0;
 518
 519         in = it->buf;
 520         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 521                 int encoding;
 522                 strbuf_reset(&charset_q);
 523                 strbuf_reset(&piecebuf);
 524                 rfc2047 = 1;
 525
 526                 if (in != ep) {
 527                         /*
 528                          * We are about to process an encoded-word
 529                          * that begins at ep, but there is something
 530                          * before the encoded word.
 531                          */
 532                         char *scan;
 533                         for (scan = in; scan < ep; scan++)
 534                                 if (!isspace(*scan))
 535                                         break;
 536
 537                         if (scan != ep || in == it->buf) {
 538                                 /*
 539                                  * We should not lose that "something",
 540                                  * unless we have just processed an
 541                                  * encoded-word, and there is only LWS
 542                                  * before the one we are about to process.
 543                                  */
 544                                 strbuf_add(&outbuf, in, ep - in);
 545                         }
 546                 }
 547                 /* E.g.
 548                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 549                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 550                  */
 551                 ep += 2;
 552
 553                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 554                         goto decode_header_bq_out;
 555
 556                 if (cp + 3 - it->buf > it->len)
 557                         goto decode_header_bq_out;
 558                 strbuf_add(&charset_q, ep, cp - ep);
 559
 560                 encoding = cp[1];
 561                 if (!encoding || cp[2] != '?')
 562                         goto decode_header_bq_out;
 563                 ep = strstr(cp + 3, "?=");
 564                 if (!ep)
 565                         goto decode_header_bq_out;
 566                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 567                 switch (tolower(encoding)) {
 568                 default:
 569                         goto decode_header_bq_out;
 570                 case 'b':
 571                         dec = decode_b_segment(&piecebuf);
 572                         break;
 573                 case 'q':
 574                         dec = decode_q_segment(&piecebuf, 1);
 575                         break;
 576                 }
 577                 if (metainfo_charset)
 578                         convert_to_utf8(dec, charset_q.buf);
 579
 580                 strbuf_addbuf(&outbuf, dec);
 581                 strbuf_release(dec);
 582                 free(dec);
 583                 in = ep + 2;
 584         }
 585         strbuf_addstr(&outbuf, in);
 586         strbuf_reset(it);
 587         strbuf_addbuf(it, &outbuf);
 588 decode_header_bq_out:
 589         strbuf_release(&outbuf);
 590         strbuf_release(&charset_q);
 591         strbuf_release(&piecebuf);
 592         return rfc2047;
 593 }
 594
 595 static void decode_header(struct strbuf *it)
 596 {
 597         if (decode_header_bq(it))
 598                 return;
 599         /* otherwise "it" is a straight copy of the input.
 600          * This can be binary guck but there is no charset specified.
 601          */
 602         if (metainfo_charset)
 603                 convert_to_utf8(it, "");
 604 }
 605
 606 static void decode_transfer_encoding(struct strbuf *line)
 607 {
 608         struct strbuf *ret;
 609
 610         switch (transfer_encoding) {
 611         case TE_QP:
 612                 ret = decode_q_segment(line, 0);
 613                 break;
 614         case TE_BASE64:
 615                 ret = decode_b_segment(line);
 616                 break;
 617         case TE_DONTCARE:
 618         default:
 619                 return;
 620         }
 621         strbuf_reset(line);
 622         strbuf_addbuf(line, ret);
 623         strbuf_release(ret);
 624         free(ret);
 625 }
 626
 627 static void handle_filter(struct strbuf *line);
 628
 629 static int find_boundary(void)
 630 {
 631         while (!strbuf_getline(&line, fin, '\n')) {
 632                 if (*content_top && is_multipart_boundary(&line))
 633                         return 1;
 634         }
 635         return 0;
 636 }
 637
 638 static int handle_boundary(void)
 639 {
 640         struct strbuf newline = STRBUF_INIT;
 641
 642         strbuf_addch(&newline, '\n');
 643 again:
 644         if (line.len >= (*content_top)->len + 2 &&
 645             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 646                 /* we hit an end boundary */
 647                 /* pop the current boundary off the stack */
 648                 strbuf_release(*content_top);
 649                 free(*content_top);
 650                 *content_top = NULL;
 651
 652                 /* technically won't happen as is_multipart_boundary()
 653                    will fail first.  But just in case..
 654                  */
 655                 if (--content_top < content) {
 656                         fprintf(stderr, "Detected mismatched boundaries, "
 657                                         "can't recover\n");
 658                         exit(1);
 659                 }
 660                 handle_filter(&newline);
 661                 strbuf_release(&newline);
 662
 663                 /* skip to the next boundary */
 664                 if (!find_boundary())
 665                         return 0;
 666                 goto again;
 667         }
 668
 669         /* set some defaults */
 670         transfer_encoding = TE_DONTCARE;
 671         strbuf_reset(&charset);
 672         message_type = TYPE_TEXT;
 673
 674         /* slurp in this section's info */
 675         while (read_one_header_line(&line, fin))
 676                 check_header(&line, p_hdr_data, 0);
 677
 678         strbuf_release(&newline);
 679         /* replenish line */
 680         if (strbuf_getline(&line, fin, '\n'))
 681                 return 0;
 682         strbuf_addch(&line, '\n');
 683         return 1;
 684 }
 685
 686 static inline int patchbreak(const struct strbuf *line)
 687 {
 688         size_t i;
 689
 690         /* Beginning of a "diff -" header? */
 691         if (!prefixcmp(line->buf, "diff -"))
 692                 return 1;
 693
 694         /* CVS "Index: " line? */
 695         if (!prefixcmp(line->buf, "Index: "))
 696                 return 1;
 697
 698         /*
 699          * "--- <filename>" starts patches without headers
 700          * "---<sp>*" is a manual separator
 701          */
 702         if (line->len < 4)
 703                 return 0;
 704
 705         if (!prefixcmp(line->buf, "---")) {
 706                 /* space followed by a filename? */
 707                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 708                         return 1;
 709                 /* Just whitespace? */
 710                 for (i = 3; i < line->len; i++) {
 711                         unsigned char c = line->buf[i];
 712                         if (c == '\n')
 713                                 return 1;
 714                         if (!isspace(c))
 715                                 break;
 716                 }
 717                 return 0;
 718         }
 719         return 0;
 720 }
 721
 722 static int handle_commit_msg(struct strbuf *line)
 723 {
 724         static int still_looking = 1;
 725
 726         if (!cmitmsg)
 727                 return 0;
 728
 729         if (still_looking) {
 730                 strbuf_ltrim(line);
 731                 if (!line->len)
 732                         return 0;
 733                 if ((still_looking = check_header(line, s_hdr_data, 0)) != 0)
 734                         return 0;
 735         }
 736
 737         /* normalize the log message to UTF-8. */
 738         if (metainfo_charset)
 739                 convert_to_utf8(line, charset.buf);
 740
 741         if (patchbreak(line)) {
 742                 fclose(cmitmsg);
 743                 cmitmsg = NULL;
 744                 return 1;
 745         }
 746
 747         fputs(line->buf, cmitmsg);
 748         return 0;
 749 }
 750
 751 static void handle_patch(const struct strbuf *line)
 752 {
 753         fwrite(line->buf, 1, line->len, patchfile);
 754         patch_lines++;
 755 }
 756
 757 static void handle_filter(struct strbuf *line)
 758 {
 759         static int filter = 0;
 760
 761         /* filter tells us which part we left off on */
 762         switch (filter) {
 763         case 0:
 764                 if (!handle_commit_msg(line))
 765                         break;
 766                 filter++;
 767         case 1:
 768                 handle_patch(line);
 769                 break;
 770         }
 771 }
 772
 773 static void handle_body(void)
 774 {
 775         int len = 0;
 776         struct strbuf prev = STRBUF_INIT;
 777
 778         /* Skip up to the first boundary */
 779         if (*content_top) {
 780                 if (!find_boundary())
 781                         goto handle_body_out;
 782         }
 783
 784         do {
 785                 strbuf_setlen(&line, line.len + len);
 786
 787                 /* process any boundary lines */
 788                 if (*content_top && is_multipart_boundary(&line)) {
 789                         /* flush any leftover */
 790                         if (prev.len) {
 791                                 handle_filter(&prev);
 792                                 strbuf_reset(&prev);
 793                         }
 794                         if (!handle_boundary())
 795                                 goto handle_body_out;
 796                 }
 797
 798                 /* Unwrap transfer encoding */
 799                 decode_transfer_encoding(&line);
 800
 801                 switch (transfer_encoding) {
 802                 case TE_BASE64:
 803                 case TE_QP:
 804                 {
 805                         struct strbuf **lines, **it, *sb;
 806
 807                         /* Prepend any previous partial lines */
 808                         strbuf_insert(&line, 0, prev.buf, prev.len);
 809                         strbuf_reset(&prev);
 810
 811                         /* binary data most likely doesn't have newlines */
 812                         if (message_type != TYPE_TEXT) {
 813                                 handle_filter(&line);
 814                                 break;
 815                         }
 816                         /*
 817                          * This is a decoded line that may contain
 818                          * multiple new lines.  Pass only one chunk
 819                          * at a time to handle_filter()
 820                          */
 821                         lines = strbuf_split(&line, '\n');
 822                         for (it = lines; (sb = *it); it++) {
 823                                 if (*(it + 1) == NULL) /* The last line */
 824                                         if (sb->buf[sb->len - 1] != '\n') {
 825                                                 /* Partial line, save it for later. */
 826                                                 strbuf_addbuf(&prev, sb);
 827                                                 break;
 828                                         }
 829                                 handle_filter(sb);
 830                         }
 831                         /*
 832                          * The partial chunk is saved in "prev" and will be
 833                          * appended by the next iteration of read_line_with_nul().
 834                          */
 835                         strbuf_list_free(lines);
 836                         break;
 837                 }
 838                 default:
 839                         handle_filter(&line);
 840                 }
 841
 842                 strbuf_reset(&line);
 843                 if (strbuf_avail(&line) < 100)
 844                         strbuf_grow(&line, 100);
 845         } while ((len = read_line_with_nul(line.buf, strbuf_avail(&line), fin)));
 846
 847 handle_body_out:
 848         strbuf_release(&prev);
 849 }
 850
 851 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 852 {
 853         const char *sp = data->buf;
 854         while (1) {
 855                 char *ep = strchr(sp, '\n');
 856                 int len;
 857                 if (!ep)
 858                         len = strlen(sp);
 859                 else
 860                         len = ep - sp;
 861                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 862                 if (!ep)
 863                         break;
 864                 sp = ep + 1;
 865         }
 866 }
 867
 868 static void handle_info(void)
 869 {
 870         struct strbuf *hdr;
 871         int i;
 872
 873         for (i = 0; header[i]; i++) {
 874                 /* only print inbody headers if we output a patch file */
 875                 if (patch_lines && s_hdr_data[i])
 876                         hdr = s_hdr_data[i];
 877                 else if (p_hdr_data[i])
 878                         hdr = p_hdr_data[i];
 879                 else
 880                         continue;
 881
 882                 if (!memcmp(header[i], "Subject", 7)) {
 883                         if (!keep_subject) {
 884                                 cleanup_subject(hdr);
 885                                 cleanup_space(hdr);
 886                         }
 887                         output_header_lines(fout, "Subject", hdr);
 888                 } else if (!memcmp(header[i], "From", 4)) {
 889                         cleanup_space(hdr);
 890                         handle_from(hdr);
 891                         fprintf(fout, "Author: %s\n", name.buf);
 892                         fprintf(fout, "Email: %s\n", email.buf);
 893                 } else {
 894                         cleanup_space(hdr);
 895                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 896                 }
 897         }
 898         fprintf(fout, "\n");
 899 }
 900
 901 static int mailinfo(FILE *in, FILE *out, int ks, const char *encoding,
 902                     const char *msg, const char *patch)
 903 {
 904         int peek;
 905         keep_subject = ks;
 906         metainfo_charset = encoding;
 907         fin = in;
 908         fout = out;
 909
 910         cmitmsg = fopen(msg, "w");
 911         if (!cmitmsg) {
 912                 perror(msg);
 913                 return -1;
 914         }
 915         patchfile = fopen(patch, "w");
 916         if (!patchfile) {
 917                 perror(patch);
 918                 fclose(cmitmsg);
 919                 return -1;
 920         }
 921
 922         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
 923         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
 924
 925         do {
 926                 peek = fgetc(in);
 927         } while (isspace(peek));
 928         ungetc(peek, in);
 929
 930         /* process the email header */
 931         while (read_one_header_line(&line, fin))
 932                 check_header(&line, p_hdr_data, 1);
 933
 934         handle_body();
 935         handle_info();
 936
 937         return 0;
 938 }
 939
 940 static const char mailinfo_usage[] =
 941         "git mailinfo [-k] [-u | --encoding=<encoding> | -n] msg patch <mail >info";
 942
 943 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
 944 {
 945         const char *def_charset;
 946
 947         /* NEEDSWORK: might want to do the optional .git/ directory
 948          * discovery
 949          */
 950         git_config(git_default_config, NULL);
 951
 952         def_charset = (git_commit_encoding ? git_commit_encoding : "UTF-8");
 953         metainfo_charset = def_charset;
 954
 955         while (1 < argc && argv[1][0] == '-') {
 956                 if (!strcmp(argv[1], "-k"))
 957                         keep_subject = 1;
 958                 else if (!strcmp(argv[1], "-u"))
 959                         metainfo_charset = def_charset;
 960                 else if (!strcmp(argv[1], "-n"))
 961                         metainfo_charset = NULL;
 962                 else if (!prefixcmp(argv[1], "--encoding="))
 963                         metainfo_charset = argv[1] + 11;
 964                 else
 965                         usage(mailinfo_usage);
 966                 argc--; argv++;
 967         }
 968
 969         if (argc != 3)
 970                 usage(mailinfo_usage);
 971
 972         return !!mailinfo(stdin, stdout, keep_subject, metainfo_charset, argv[1], argv[2]);
 973 }