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 get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  33 {
  34         struct strbuf *src = name;
  35         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  36                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  37                 src = email;
  38         else if (name == out)
  39                 return;
  40         strbuf_reset(out);
  41         strbuf_addbuf(out, src);
  42 }
  43
  44 static void parse_bogus_from(const struct strbuf *line)
  45 {
  46         /* John Doe <johndoe> */
  47
  48         char *bra, *ket;
  49         /* This is fallback, so do not bother if we already have an
  50          * e-mail address.
  51          */
  52         if (email.len)
  53                 return;
  54
  55         bra = strchr(line->buf, '<');
  56         if (!bra)
  57                 return;
  58         ket = strchr(bra, '>');
  59         if (!ket)
  60                 return;
  61
  62         strbuf_reset(&email);
  63         strbuf_add(&email, bra + 1, ket - bra - 1);
  64
  65         strbuf_reset(&name);
  66         strbuf_add(&name, line->buf, bra - line->buf);
  67         strbuf_trim(&name);
  68         get_sane_name(&name, &name, &email);
  69 }
  70
  71 static void handle_from(const struct strbuf *from)
  72 {
  73         char *at;
  74         size_t el;
  75         struct strbuf f;
  76
  77         strbuf_init(&f, from->len);
  78         strbuf_addbuf(&f, from);
  79
  80         at = strchr(f.buf, '@');
  81         if (!at) {
  82                 parse_bogus_from(from);
  83                 return;
  84         }
  85
  86         /*
  87          * If we already have one email, don't take any confusing lines
  88          */
  89         if (email.len && strchr(at + 1, '@')) {
  90                 strbuf_release(&f);
  91                 return;
  92         }
  93
  94         /* Pick up the string around '@', possibly delimited with <>
  95          * pair; that is the email part.
  96          */
  97         while (at > f.buf) {
  98                 char c = at[-1];
  99                 if (isspace(c))
 100                         break;
 101                 if (c == '<') {
 102                         at[-1] = ' ';
 103                         break;
 104                 }
 105                 at--;
 106         }
 107         el = strcspn(at, " \n\t\r\v\f>");
 108         strbuf_reset(&email);
 109         strbuf_add(&email, at, el);
 110         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 111
 112         /* The remainder is name.  It could be "John Doe <john.doe@xz>"
 113          * or "john.doe@xz (John Doe)", but we have removed the
 114          * email part, so trim from both ends, possibly removing
 115          * the () pair at the end.
 116          */
 117         strbuf_trim(&f);
 118         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 119                 strbuf_remove(&f, 0, 1);
 120                 strbuf_setlen(&f, f.len - 1);
 121         }
 122
 123         get_sane_name(&name, &f, &email);
 124         strbuf_release(&f);
 125 }
 126
 127 static void handle_header(struct strbuf **out, const struct strbuf *line)
 128 {
 129         if (!*out) {
 130                 *out = xmalloc(sizeof(struct strbuf));
 131                 strbuf_init(*out, line->len);
 132         } else
 133                 strbuf_reset(*out);
 134
 135         strbuf_addbuf(*out, line);
 136 }
 137
 138 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 139  * to have enough heuristics to grok MIME encoded patches often found
 140  * on our mailing lists.  For example, we do not even treat header lines
 141  * case insensitively.
 142  */
 143
 144 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 145 {
 146         const char *ends, *ap = strcasestr(line, name);
 147         size_t sz;
 148
 149         if (!ap) {
 150                 strbuf_setlen(attr, 0);
 151                 return 0;
 152         }
 153         ap += strlen(name);
 154         if (*ap == '"') {
 155                 ap++;
 156                 ends = "\"";
 157         }
 158         else
 159                 ends = "; \t";
 160         sz = strcspn(ap, ends);
 161         strbuf_add(attr, ap, sz);
 162         return 1;
 163 }
 164
 165 static struct strbuf *content[MAX_BOUNDARIES];
 166
 167 static struct strbuf **content_top = content;
 168
 169 static void handle_content_type(struct strbuf *line)
 170 {
 171         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 172         strbuf_init(boundary, line->len);
 173
 174         if (!strcasestr(line->buf, "text/"))
 175                  message_type = TYPE_OTHER;
 176         if (slurp_attr(line->buf, "boundary=", boundary)) {
 177                 strbuf_insert(boundary, 0, "--", 2);
 178                 if (++content_top > &content[MAX_BOUNDARIES]) {
 179                         fprintf(stderr, "Too many boundaries to handle\n");
 180                         exit(1);
 181                 }
 182                 *content_top = boundary;
 183                 boundary = NULL;
 184         }
 185         if (slurp_attr(line->buf, "charset=", &charset))
 186                 strbuf_tolower(&charset);
 187
 188         if (boundary) {
 189                 strbuf_release(boundary);
 190                 free(boundary);
 191         }
 192 }
 193
 194 static void handle_content_transfer_encoding(const struct strbuf *line)
 195 {
 196         if (strcasestr(line->buf, "base64"))
 197                 transfer_encoding = TE_BASE64;
 198         else if (strcasestr(line->buf, "quoted-printable"))
 199                 transfer_encoding = TE_QP;
 200         else
 201                 transfer_encoding = TE_DONTCARE;
 202 }
 203
 204 static int is_multipart_boundary(const struct strbuf *line)
 205 {
 206         return (((*content_top)->len <= line->len) &&
 207                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
 208 }
 209
 210 static void cleanup_subject(struct strbuf *subject)
 211 {
 212         char *pos;
 213         size_t remove;
 214         while (subject->len) {
 215                 switch (*subject->buf) {
 216                 case 'r': case 'R':
 217                         if (subject->len <= 3)
 218                                 break;
 219                         if (!memcmp(subject->buf + 1, "e:", 2)) {
 220                                 strbuf_remove(subject, 0, 3);
 221                                 continue;
 222                         }
 223                         break;
 224                 case ' ': case '\t': case ':':
 225                         strbuf_remove(subject, 0, 1);
 226                         continue;
 227                 case '[':
 228                         if ((pos = strchr(subject->buf, ']'))) {
 229                                 remove = pos - subject->buf;
 230                                 if (remove <= (subject->len - remove) * 2) {
 231                                         strbuf_remove(subject, 0, remove + 1);
 232                                         continue;
 233                                 }
 234                         } else
 235                                 strbuf_remove(subject, 0, 1);
 236                         break;
 237                 }
 238                 strbuf_trim(subject);
 239                 return;
 240         }
 241 }
 242
 243 static void cleanup_space(struct strbuf *sb)
 244 {
 245         size_t pos, cnt;
 246         for (pos = 0; pos < sb->len; pos++) {
 247                 if (isspace(sb->buf[pos])) {
 248                         sb->buf[pos] = ' ';
 249                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
 250                         strbuf_remove(sb, pos + 1, cnt);
 251                 }
 252         }
 253 }
 254
 255 static void decode_header(struct strbuf *line);
 256 static const char *header[MAX_HDR_PARSED] = {
 257         "From","Subject","Date",
 258 };
 259
 260 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 261 {
 262         int len = strlen(hdr);
 263         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 264                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 265 }
 266
 267 static int check_header(const struct strbuf *line,
 268                                 struct strbuf *hdr_data[], int overwrite)
 269 {
 270         int i, ret = 0, len;
 271         struct strbuf sb = STRBUF_INIT;
 272         /* search for the interesting parts */
 273         for (i = 0; header[i]; i++) {
 274                 int len = strlen(header[i]);
 275                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 276                         /* Unwrap inline B and Q encoding, and optionally
 277                          * normalize the meta information to utf8.
 278                          */
 279                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 280                         decode_header(&sb);
 281                         handle_header(&hdr_data[i], &sb);
 282                         ret = 1;
 283                         goto check_header_out;
 284                 }
 285         }
 286
 287         /* Content stuff */
 288         if (cmp_header(line, "Content-Type")) {
 289                 len = strlen("Content-Type: ");
 290                 strbuf_add(&sb, line->buf + len, line->len - len);
 291                 decode_header(&sb);
 292                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 293                 handle_content_type(&sb);
 294                 ret = 1;
 295                 goto check_header_out;
 296         }
 297         if (cmp_header(line, "Content-Transfer-Encoding")) {
 298                 len = strlen("Content-Transfer-Encoding: ");
 299                 strbuf_add(&sb, line->buf + len, line->len - len);
 300                 decode_header(&sb);
 301                 handle_content_transfer_encoding(&sb);
 302                 ret = 1;
 303                 goto check_header_out;
 304         }
 305
 306         /* for inbody stuff */
 307         if (!prefixcmp(line->buf, ">From") && isspace(line->buf[5])) {
 308                 ret = 1; /* Should this return 0? */
 309                 goto check_header_out;
 310         }
 311         if (!prefixcmp(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 312                 for (i = 0; header[i]; i++) {
 313                         if (!memcmp("Subject", header[i], 7)) {
 314                                 handle_header(&hdr_data[i], line);
 315                                 ret = 1;
 316                                 goto check_header_out;
 317                         }
 318                 }
 319         }
 320
 321 check_header_out:
 322         strbuf_release(&sb);
 323         return ret;
 324 }
 325
 326 static int is_rfc2822_header(const struct strbuf *line)
 327 {
 328         /*
 329          * The section that defines the loosest possible
 330          * field name is "3.6.8 Optional fields".
 331          *
 332          * optional-field = field-name ":" unstructured CRLF
 333          * field-name = 1*ftext
 334          * ftext = %d33-57 / %59-126
 335          */
 336         int ch;
 337         char *cp = line->buf;
 338
 339         /* Count mbox From headers as headers */
 340         if (!prefixcmp(cp, "From ") || !prefixcmp(cp, ">From "))
 341                 return 1;
 342
 343         while ((ch = *cp++)) {
 344                 if (ch == ':')
 345                         return 1;
 346                 if ((33 <= ch && ch <= 57) ||
 347                     (59 <= ch && ch <= 126))
 348                         continue;
 349                 break;
 350         }
 351         return 0;
 352 }
 353
 354 static int read_one_header_line(struct strbuf *line, FILE *in)
 355 {
 356         /* Get the first part of the line. */
 357         if (strbuf_getline(line, in, '\n'))
 358                 return 0;
 359
 360         /*
 361          * Is it an empty line or not a valid rfc2822 header?
 362          * If so, stop here, and return false ("not a header")
 363          */
 364         strbuf_rtrim(line);
 365         if (!line->len || !is_rfc2822_header(line)) {
 366                 /* Re-add the newline */
 367                 strbuf_addch(line, '\n');
 368                 return 0;
 369         }
 370
 371         /*
 372          * Now we need to eat all the continuation lines..
 373          * Yuck, 2822 header "folding"
 374          */
 375         for (;;) {
 376                 int peek;
 377                 struct strbuf continuation = STRBUF_INIT;
 378
 379                 peek = fgetc(in); ungetc(peek, in);
 380                 if (peek != ' ' && peek != '\t')
 381                         break;
 382                 if (strbuf_getline(&continuation, in, '\n'))
 383                         break;
 384                 continuation.buf[0] = '\n';
 385                 strbuf_rtrim(&continuation);
 386                 strbuf_addbuf(line, &continuation);
 387         }
 388
 389         return 1;
 390 }
 391
 392 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 393 {
 394         const char *in = q_seg->buf;
 395         int c;
 396         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 397         strbuf_init(out, q_seg->len);
 398
 399         while ((c = *in++) != 0) {
 400                 if (c == '=') {
 401                         int d = *in++;
 402                         if (d == '\n' || !d)
 403                                 break; /* drop trailing newline */
 404                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 405                         continue;
 406                 }
 407                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 408                         c = 0x20;
 409                 strbuf_addch(out, c);
 410         }
 411         return out;
 412 }
 413
 414 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 415 {
 416         /* Decode in..ep, possibly in-place to ot */
 417         int c, pos = 0, acc = 0;
 418         const char *in = b_seg->buf;
 419         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 420         strbuf_init(out, b_seg->len);
 421
 422         while ((c = *in++) != 0) {
 423                 if (c == '+')
 424                         c = 62;
 425                 else if (c == '/')
 426                         c = 63;
 427                 else if ('A' <= c && c <= 'Z')
 428                         c -= 'A';
 429                 else if ('a' <= c && c <= 'z')
 430                         c -= 'a' - 26;
 431                 else if ('0' <= c && c <= '9')
 432                         c -= '0' - 52;
 433                 else if (c == '=') {
 434                         /* padding is almost like (c == 0), except we do
 435                          * not output NUL resulting only from it;
 436                          * for now we just trust the data.
 437                          */
 438                         c = 0;
 439                 }
 440                 else
 441                         continue; /* garbage */
 442                 switch (pos++) {
 443                 case 0:
 444                         acc = (c << 2);
 445                         break;
 446                 case 1:
 447                         strbuf_addch(out, (acc | (c >> 4)));
 448                         acc = (c & 15) << 4;
 449                         break;
 450                 case 2:
 451                         strbuf_addch(out, (acc | (c >> 2)));
 452                         acc = (c & 3) << 6;
 453                         break;
 454                 case 3:
 455                         strbuf_addch(out, (acc | c));
 456                         acc = pos = 0;
 457                         break;
 458                 }
 459         }
 460         return out;
 461 }
 462
 463 /*
 464  * When there is no known charset, guess.
 465  *
 466  * Right now we assume that if the target is UTF-8 (the default),
 467  * and it already looks like UTF-8 (which includes US-ASCII as its
 468  * subset, of course) then that is what it is and there is nothing
 469  * to do.
 470  *
 471  * Otherwise, we default to assuming it is Latin1 for historical
 472  * reasons.
 473  */
 474 static const char *guess_charset(const struct strbuf *line, const char *target_charset)
 475 {
 476         if (is_encoding_utf8(target_charset)) {
 477                 if (is_utf8(line->buf))
 478                         return NULL;
 479         }
 480         return "latin1";
 481 }
 482
 483 static void convert_to_utf8(struct strbuf *line, const char *charset)
 484 {
 485         char *out;
 486
 487         if (!charset || !*charset) {
 488                 charset = guess_charset(line, metainfo_charset);
 489                 if (!charset)
 490                         return;
 491         }
 492
 493         if (!strcmp(metainfo_charset, charset))
 494                 return;
 495         out = reencode_string(line->buf, metainfo_charset, charset);
 496         if (!out)
 497                 die("cannot convert from %s to %s\n",
 498                     charset, metainfo_charset);
 499         strbuf_attach(line, out, strlen(out), strlen(out));
 500 }
 501
 502 static int decode_header_bq(struct strbuf *it)
 503 {
 504         char *in, *ep, *cp;
 505         struct strbuf outbuf = STRBUF_INIT, *dec;
 506         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 507         int rfc2047 = 0;
 508
 509         in = it->buf;
 510         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 511                 int encoding;
 512                 strbuf_reset(&charset_q);
 513                 strbuf_reset(&piecebuf);
 514                 rfc2047 = 1;
 515
 516                 if (in != ep) {
 517                         strbuf_add(&outbuf, in, ep - in);
 518                         in = ep;
 519                 }
 520                 /* E.g.
 521                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 522                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 523                  */
 524                 ep += 2;
 525
 526                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 527                         goto decode_header_bq_out;
 528
 529                 if (cp + 3 - it->buf > it->len)
 530                         goto decode_header_bq_out;
 531                 strbuf_add(&charset_q, ep, cp - ep);
 532                 strbuf_tolower(&charset_q);
 533
 534                 encoding = cp[1];
 535                 if (!encoding || cp[2] != '?')
 536                         goto decode_header_bq_out;
 537                 ep = strstr(cp + 3, "?=");
 538                 if (!ep)
 539                         goto decode_header_bq_out;
 540                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 541                 switch (tolower(encoding)) {
 542                 default:
 543                         goto decode_header_bq_out;
 544                 case 'b':
 545                         dec = decode_b_segment(&piecebuf);
 546                         break;
 547                 case 'q':
 548                         dec = decode_q_segment(&piecebuf, 1);
 549                         break;
 550                 }
 551                 if (metainfo_charset)
 552                         convert_to_utf8(dec, charset_q.buf);
 553
 554                 strbuf_addbuf(&outbuf, dec);
 555                 strbuf_release(dec);
 556                 free(dec);
 557                 in = ep + 2;
 558         }
 559         strbuf_addstr(&outbuf, in);
 560         strbuf_reset(it);
 561         strbuf_addbuf(it, &outbuf);
 562 decode_header_bq_out:
 563         strbuf_release(&outbuf);
 564         strbuf_release(&charset_q);
 565         strbuf_release(&piecebuf);
 566         return rfc2047;
 567 }
 568
 569 static void decode_header(struct strbuf *it)
 570 {
 571         if (decode_header_bq(it))
 572                 return;
 573         /* otherwise "it" is a straight copy of the input.
 574          * This can be binary guck but there is no charset specified.
 575          */
 576         if (metainfo_charset)
 577                 convert_to_utf8(it, "");
 578 }
 579
 580 static void decode_transfer_encoding(struct strbuf *line)
 581 {
 582         struct strbuf *ret;
 583
 584         switch (transfer_encoding) {
 585         case TE_QP:
 586                 ret = decode_q_segment(line, 0);
 587                 break;
 588         case TE_BASE64:
 589                 ret = decode_b_segment(line);
 590                 break;
 591         case TE_DONTCARE:
 592         default:
 593                 return;
 594         }
 595         strbuf_reset(line);
 596         strbuf_addbuf(line, ret);
 597         strbuf_release(ret);
 598         free(ret);
 599 }
 600
 601 static void handle_filter(struct strbuf *line);
 602
 603 static int find_boundary(void)
 604 {
 605         while (!strbuf_getline(&line, fin, '\n')) {
 606                 if (*content_top && is_multipart_boundary(&line))
 607                         return 1;
 608         }
 609         return 0;
 610 }
 611
 612 static int handle_boundary(void)
 613 {
 614         struct strbuf newline = STRBUF_INIT;
 615
 616         strbuf_addch(&newline, '\n');
 617 again:
 618         if (line.len >= (*content_top)->len + 2 &&
 619             !memcmp(line.buf + (*content_top)->len, "--", 2)) {
 620                 /* we hit an end boundary */
 621                 /* pop the current boundary off the stack */
 622                 strbuf_release(*content_top);
 623                 free(*content_top);
 624                 *content_top = NULL;
 625
 626                 /* technically won't happen as is_multipart_boundary()
 627                    will fail first.  But just in case..
 628                  */
 629                 if (--content_top < content) {
 630                         fprintf(stderr, "Detected mismatched boundaries, "
 631                                         "can't recover\n");
 632                         exit(1);
 633                 }
 634                 handle_filter(&newline);
 635                 strbuf_release(&newline);
 636
 637                 /* skip to the next boundary */
 638                 if (!find_boundary())
 639                         return 0;
 640                 goto again;
 641         }
 642
 643         /* set some defaults */
 644         transfer_encoding = TE_DONTCARE;
 645         strbuf_reset(&charset);
 646         message_type = TYPE_TEXT;
 647
 648         /* slurp in this section's info */
 649         while (read_one_header_line(&line, fin))
 650                 check_header(&line, p_hdr_data, 0);
 651
 652         strbuf_release(&newline);
 653         /* replenish line */
 654         if (strbuf_getline(&line, fin, '\n'))
 655                 return 0;
 656         strbuf_addch(&line, '\n');
 657         return 1;
 658 }
 659
 660 static inline int patchbreak(const struct strbuf *line)
 661 {
 662         size_t i;
 663
 664         /* Beginning of a "diff -" header? */
 665         if (!prefixcmp(line->buf, "diff -"))
 666                 return 1;
 667
 668         /* CVS "Index: " line? */
 669         if (!prefixcmp(line->buf, "Index: "))
 670                 return 1;
 671
 672         /*
 673          * "--- <filename>" starts patches without headers
 674          * "---<sp>*" is a manual separator
 675          */
 676         if (line->len < 4)
 677                 return 0;
 678
 679         if (!prefixcmp(line->buf, "---")) {
 680                 /* space followed by a filename? */
 681                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 682                         return 1;
 683                 /* Just whitespace? */
 684                 for (i = 3; i < line->len; i++) {
 685                         unsigned char c = line->buf[i];
 686                         if (c == '\n')
 687                                 return 1;
 688                         if (!isspace(c))
 689                                 break;
 690                 }
 691                 return 0;
 692         }
 693         return 0;
 694 }
 695
 696 static int handle_commit_msg(struct strbuf *line)
 697 {
 698         static int still_looking = 1;
 699
 700         if (!cmitmsg)
 701                 return 0;
 702
 703         if (still_looking) {
 704                 strbuf_ltrim(line);
 705                 if (!line->len)
 706                         return 0;
 707                 if ((still_looking = check_header(line, s_hdr_data, 0)) != 0)
 708                         return 0;
 709         }
 710
 711         /* normalize the log message to UTF-8. */
 712         if (metainfo_charset)
 713                 convert_to_utf8(line, charset.buf);
 714
 715         if (patchbreak(line)) {
 716                 fclose(cmitmsg);
 717                 cmitmsg = NULL;
 718                 return 1;
 719         }
 720
 721         fputs(line->buf, cmitmsg);
 722         return 0;
 723 }
 724
 725 static void handle_patch(const struct strbuf *line)
 726 {
 727         fwrite(line->buf, 1, line->len, patchfile);
 728         patch_lines++;
 729 }
 730
 731 static void handle_filter(struct strbuf *line)
 732 {
 733         static int filter = 0;
 734
 735         /* filter tells us which part we left off on */
 736         switch (filter) {
 737         case 0:
 738                 if (!handle_commit_msg(line))
 739                         break;
 740                 filter++;
 741         case 1:
 742                 handle_patch(line);
 743                 break;
 744         }
 745 }
 746
 747 static void handle_body(void)
 748 {
 749         int len = 0;
 750         struct strbuf prev = STRBUF_INIT;
 751
 752         /* Skip up to the first boundary */
 753         if (*content_top) {
 754                 if (!find_boundary())
 755                         goto handle_body_out;
 756         }
 757
 758         do {
 759                 strbuf_setlen(&line, line.len + len);
 760
 761                 /* process any boundary lines */
 762                 if (*content_top && is_multipart_boundary(&line)) {
 763                         /* flush any leftover */
 764                         if (prev.len) {
 765                                 handle_filter(&prev);
 766                                 strbuf_reset(&prev);
 767                         }
 768                         if (!handle_boundary())
 769                                 goto handle_body_out;
 770                 }
 771
 772                 /* Unwrap transfer encoding */
 773                 decode_transfer_encoding(&line);
 774
 775                 switch (transfer_encoding) {
 776                 case TE_BASE64:
 777                 case TE_QP:
 778                 {
 779                         struct strbuf **lines, **it, *sb;
 780
 781                         /* Prepend any previous partial lines */
 782                         strbuf_insert(&line, 0, prev.buf, prev.len);
 783                         strbuf_reset(&prev);
 784
 785                         /* binary data most likely doesn't have newlines */
 786                         if (message_type != TYPE_TEXT) {
 787                                 handle_filter(&line);
 788                                 break;
 789                         }
 790                         /*
 791                          * This is a decoded line that may contain
 792                          * multiple new lines.  Pass only one chunk
 793                          * at a time to handle_filter()
 794                          */
 795                         lines = strbuf_split(&line, '\n');
 796                         for (it = lines; (sb = *it); it++) {
 797                                 if (*(it + 1) == NULL) /* The last line */
 798                                         if (sb->buf[sb->len - 1] != '\n') {
 799                                                 /* Partial line, save it for later. */
 800                                                 strbuf_addbuf(&prev, sb);
 801                                                 break;
 802                                         }
 803                                 handle_filter(sb);
 804                         }
 805                         /*
 806                          * The partial chunk is saved in "prev" and will be
 807                          * appended by the next iteration of read_line_with_nul().
 808                          */
 809                         strbuf_list_free(lines);
 810                         break;
 811                 }
 812                 default:
 813                         handle_filter(&line);
 814                 }
 815
 816                 strbuf_reset(&line);
 817                 if (strbuf_avail(&line) < 100)
 818                         strbuf_grow(&line, 100);
 819         } while ((len = read_line_with_nul(line.buf, strbuf_avail(&line), fin)));
 820
 821 handle_body_out:
 822         strbuf_release(&prev);
 823 }
 824
 825 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 826 {
 827         const char *sp = data->buf;
 828         while (1) {
 829                 char *ep = strchr(sp, '\n');
 830                 int len;
 831                 if (!ep)
 832                         len = strlen(sp);
 833                 else
 834                         len = ep - sp;
 835                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 836                 if (!ep)
 837                         break;
 838                 sp = ep + 1;
 839         }
 840 }
 841
 842 static void handle_info(void)
 843 {
 844         struct strbuf *hdr;
 845         int i;
 846
 847         for (i = 0; header[i]; i++) {
 848                 /* only print inbody headers if we output a patch file */
 849                 if (patch_lines && s_hdr_data[i])
 850                         hdr = s_hdr_data[i];
 851                 else if (p_hdr_data[i])
 852                         hdr = p_hdr_data[i];
 853                 else
 854                         continue;
 855
 856                 if (!memcmp(header[i], "Subject", 7)) {
 857                         if (!keep_subject) {
 858                                 cleanup_subject(hdr);
 859                                 cleanup_space(hdr);
 860                         }
 861                         output_header_lines(fout, "Subject", hdr);
 862                 } else if (!memcmp(header[i], "From", 4)) {
 863                         handle_from(hdr);
 864                         fprintf(fout, "Author: %s\n", name.buf);
 865                         fprintf(fout, "Email: %s\n", email.buf);
 866                 } else {
 867                         cleanup_space(hdr);
 868                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 869                 }
 870         }
 871         fprintf(fout, "\n");
 872 }
 873
 874 static int mailinfo(FILE *in, FILE *out, int ks, const char *encoding,
 875                     const char *msg, const char *patch)
 876 {
 877         int peek;
 878         keep_subject = ks;
 879         metainfo_charset = encoding;
 880         fin = in;
 881         fout = out;
 882
 883         cmitmsg = fopen(msg, "w");
 884         if (!cmitmsg) {
 885                 perror(msg);
 886                 return -1;
 887         }
 888         patchfile = fopen(patch, "w");
 889         if (!patchfile) {
 890                 perror(patch);
 891                 fclose(cmitmsg);
 892                 return -1;
 893         }
 894
 895         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
 896         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
 897
 898         do {
 899                 peek = fgetc(in);
 900         } while (isspace(peek));
 901         ungetc(peek, in);
 902
 903         /* process the email header */
 904         while (read_one_header_line(&line, fin))
 905                 check_header(&line, p_hdr_data, 1);
 906
 907         handle_body();
 908         handle_info();
 909
 910         return 0;
 911 }
 912
 913 static const char mailinfo_usage[] =
 914         "git mailinfo [-k] [-u | --encoding=<encoding> | -n] msg patch <mail >info";
 915
 916 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
 917 {
 918         const char *def_charset;
 919
 920         /* NEEDSWORK: might want to do the optional .git/ directory
 921          * discovery
 922          */
 923         git_config(git_default_config, NULL);
 924
 925         def_charset = (git_commit_encoding ? git_commit_encoding : "utf-8");
 926         metainfo_charset = def_charset;
 927
 928         while (1 < argc && argv[1][0] == '-') {
 929                 if (!strcmp(argv[1], "-k"))
 930                         keep_subject = 1;
 931                 else if (!strcmp(argv[1], "-u"))
 932                         metainfo_charset = def_charset;
 933                 else if (!strcmp(argv[1], "-n"))
 934                         metainfo_charset = NULL;
 935                 else if (!prefixcmp(argv[1], "--encoding="))
 936                         metainfo_charset = argv[1] + 11;
 937                 else
 938                         usage(mailinfo_usage);
 939                 argc--; argv++;
 940         }
 941
 942         if (argc != 3)
 943                 usage(mailinfo_usage);
 944
 945         return !!mailinfo(stdin, stdout, keep_subject, metainfo_charset, argv[1], argv[2]);
 946 }