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