mailinfo.c

   1 #include "cache.h"
   2 #include "utf8.h"
   3 #include "strbuf.h"
   4 #include "mailinfo.h"
   5
   6 static void cleanup_space(struct strbuf *sb)
   7 {
   8         size_t pos, cnt;
   9         for (pos = 0; pos < sb->len; pos++) {
  10                 if (isspace(sb->buf[pos])) {
  11                         sb->buf[pos] = ' ';
  12                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
  13                         strbuf_remove(sb, pos + 1, cnt);
  14                 }
  15         }
  16 }
  17
  18 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  19 {
  20         struct strbuf *src = name;
  21         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  22                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  23                 src = email;
  24         else if (name == out)
  25                 return;
  26         strbuf_reset(out);
  27         strbuf_addbuf(out, src);
  28 }
  29
  30 static void parse_bogus_from(struct mailinfo *mi, const struct strbuf *line)
  31 {
  32         /* John Doe <johndoe> */
  33
  34         char *bra, *ket;
  35         /* This is fallback, so do not bother if we already have an
  36          * e-mail address.
  37          */
  38         if (mi->email.len)
  39                 return;
  40
  41         bra = strchr(line->buf, '<');
  42         if (!bra)
  43                 return;
  44         ket = strchr(bra, '>');
  45         if (!ket)
  46                 return;
  47
  48         strbuf_reset(&mi->email);
  49         strbuf_add(&mi->email, bra + 1, ket - bra - 1);
  50
  51         strbuf_reset(&mi->name);
  52         strbuf_add(&mi->name, line->buf, bra - line->buf);
  53         strbuf_trim(&mi->name);
  54         get_sane_name(&mi->name, &mi->name, &mi->email);
  55 }
  56
  57 static void handle_from(struct mailinfo *mi, const struct strbuf *from)
  58 {
  59         char *at;
  60         size_t el;
  61         struct strbuf f;
  62
  63         strbuf_init(&f, from->len);
  64         strbuf_addbuf(&f, from);
  65
  66         at = strchr(f.buf, '@');
  67         if (!at) {
  68                 parse_bogus_from(mi, from);
  69                 return;
  70         }
  71
  72         /*
  73          * If we already have one email, don't take any confusing lines
  74          */
  75         if (mi->email.len && strchr(at + 1, '@')) {
  76                 strbuf_release(&f);
  77                 return;
  78         }
  79
  80         /* Pick up the string around '@', possibly delimited with <>
  81          * pair; that is the email part.
  82          */
  83         while (at > f.buf) {
  84                 char c = at[-1];
  85                 if (isspace(c))
  86                         break;
  87                 if (c == '<') {
  88                         at[-1] = ' ';
  89                         break;
  90                 }
  91                 at--;
  92         }
  93         el = strcspn(at, " \n\t\r\v\f>");
  94         strbuf_reset(&mi->email);
  95         strbuf_add(&mi->email, at, el);
  96         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
  97
  98         /* The remainder is name.  It could be
  99          *
 100          * - "John Doe <john.doe@xz>"                   (a), or
 101          * - "john.doe@xz (John Doe)"                   (b), or
 102          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 103          *
 104          * but we have removed the email part, so
 105          *
 106          * - remove extra spaces which could stay after email (case 'c'), and
 107          * - trim from both ends, possibly removing the () pair at the end
 108          *   (cases 'a' and 'b').
 109          */
 110         cleanup_space(&f);
 111         strbuf_trim(&f);
 112         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 113                 strbuf_remove(&f, 0, 1);
 114                 strbuf_setlen(&f, f.len - 1);
 115         }
 116
 117         get_sane_name(&mi->name, &f, &mi->email);
 118         strbuf_release(&f);
 119 }
 120
 121 static void handle_header(struct strbuf **out, const struct strbuf *line)
 122 {
 123         if (!*out) {
 124                 *out = xmalloc(sizeof(struct strbuf));
 125                 strbuf_init(*out, line->len);
 126         } else
 127                 strbuf_reset(*out);
 128
 129         strbuf_addbuf(*out, line);
 130 }
 131
 132 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 133  * to have enough heuristics to grok MIME encoded patches often found
 134  * on our mailing lists.  For example, we do not even treat header lines
 135  * case insensitively.
 136  */
 137
 138 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 139 {
 140         const char *ends, *ap = strcasestr(line, name);
 141         size_t sz;
 142
 143         strbuf_setlen(attr, 0);
 144         if (!ap)
 145                 return 0;
 146         ap += strlen(name);
 147         if (*ap == '"') {
 148                 ap++;
 149                 ends = "\"";
 150         }
 151         else
 152                 ends = "; \t";
 153         sz = strcspn(ap, ends);
 154         strbuf_add(attr, ap, sz);
 155         return 1;
 156 }
 157
 158 static void handle_content_type(struct mailinfo *mi, struct strbuf *line)
 159 {
 160         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 161         strbuf_init(boundary, line->len);
 162
 163         if (slurp_attr(line->buf, "boundary=", boundary)) {
 164                 strbuf_insert(boundary, 0, "--", 2);
 165                 if (++mi->content_top >= &mi->content[MAX_BOUNDARIES]) {
 166                         error("Too many boundaries to handle");
 167                         mi->input_error = -1;
 168                         mi->content_top = &mi->content[MAX_BOUNDARIES] - 1;
 169                         return;
 170                 }
 171                 *(mi->content_top) = boundary;
 172                 boundary = NULL;
 173         }
 174         slurp_attr(line->buf, "charset=", &mi->charset);
 175
 176         if (boundary) {
 177                 strbuf_release(boundary);
 178                 free(boundary);
 179         }
 180 }
 181
 182 static void handle_content_transfer_encoding(struct mailinfo *mi,
 183                                              const struct strbuf *line)
 184 {
 185         if (strcasestr(line->buf, "base64"))
 186                 mi->transfer_encoding = TE_BASE64;
 187         else if (strcasestr(line->buf, "quoted-printable"))
 188                 mi->transfer_encoding = TE_QP;
 189         else
 190                 mi->transfer_encoding = TE_DONTCARE;
 191 }
 192
 193 static int is_multipart_boundary(struct mailinfo *mi, const struct strbuf *line)
 194 {
 195         struct strbuf *content_top = *(mi->content_top);
 196
 197         return ((content_top->len <= line->len) &&
 198                 !memcmp(line->buf, content_top->buf, content_top->len));
 199 }
 200
 201 static void cleanup_subject(struct mailinfo *mi, struct strbuf *subject)
 202 {
 203         size_t at = 0;
 204
 205         while (at < subject->len) {
 206                 char *pos;
 207                 size_t remove;
 208
 209                 switch (subject->buf[at]) {
 210                 case 'r': case 'R':
 211                         if (subject->len <= at + 3)
 212                                 break;
 213                         if ((subject->buf[at + 1] == 'e' ||
 214                              subject->buf[at + 1] == 'E') &&
 215                             subject->buf[at + 2] == ':') {
 216                                 strbuf_remove(subject, at, 3);
 217                                 continue;
 218                         }
 219                         at++;
 220                         break;
 221                 case ' ': case '\t': case ':':
 222                         strbuf_remove(subject, at, 1);
 223                         continue;
 224                 case '[':
 225                         pos = strchr(subject->buf + at, ']');
 226                         if (!pos)
 227                                 break;
 228                         remove = pos - subject->buf + at + 1;
 229                         if (!mi->keep_non_patch_brackets_in_subject ||
 230                             (7 <= remove &&
 231                              memmem(subject->buf + at, remove, "PATCH", 5)))
 232                                 strbuf_remove(subject, at, remove);
 233                         else {
 234                                 at += remove;
 235                                 /*
 236                                  * If the input had a space after the ], keep
 237                                  * it.  We don't bother with finding the end of
 238                                  * the space, since we later normalize it
 239                                  * anyway.
 240                                  */
 241                                 if (isspace(subject->buf[at]))
 242                                         at += 1;
 243                         }
 244                         continue;
 245                 }
 246                 break;
 247         }
 248         strbuf_trim(subject);
 249 }
 250
 251 #define MAX_HDR_PARSED 10
 252 static const char *header[MAX_HDR_PARSED] = {
 253         "From","Subject","Date",
 254 };
 255
 256 static inline int cmp_header(const struct strbuf *line, const char *hdr)
 257 {
 258         int len = strlen(hdr);
 259         return !strncasecmp(line->buf, hdr, len) && line->len > len &&
 260                         line->buf[len] == ':' && isspace(line->buf[len + 1]);
 261 }
 262
 263 static int is_format_patch_separator(const char *line, int len)
 264 {
 265         static const char SAMPLE[] =
 266                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 267         const char *cp;
 268
 269         if (len != strlen(SAMPLE))
 270                 return 0;
 271         if (!skip_prefix(line, "From ", &cp))
 272                 return 0;
 273         if (strspn(cp, "0123456789abcdef") != 40)
 274                 return 0;
 275         cp += 40;
 276         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 277 }
 278
 279 static struct strbuf *decode_q_segment(const struct strbuf *q_seg, int rfc2047)
 280 {
 281         const char *in = q_seg->buf;
 282         int c;
 283         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 284         strbuf_init(out, q_seg->len);
 285
 286         while ((c = *in++) != 0) {
 287                 if (c == '=') {
 288                         int d = *in++;
 289                         if (d == '\n' || !d)
 290                                 break; /* drop trailing newline */
 291                         strbuf_addch(out, (hexval(d) << 4) | hexval(*in++));
 292                         continue;
 293                 }
 294                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 295                         c = 0x20;
 296                 strbuf_addch(out, c);
 297         }
 298         return out;
 299 }
 300
 301 static struct strbuf *decode_b_segment(const struct strbuf *b_seg)
 302 {
 303         /* Decode in..ep, possibly in-place to ot */
 304         int c, pos = 0, acc = 0;
 305         const char *in = b_seg->buf;
 306         struct strbuf *out = xmalloc(sizeof(struct strbuf));
 307         strbuf_init(out, b_seg->len);
 308
 309         while ((c = *in++) != 0) {
 310                 if (c == '+')
 311                         c = 62;
 312                 else if (c == '/')
 313                         c = 63;
 314                 else if ('A' <= c && c <= 'Z')
 315                         c -= 'A';
 316                 else if ('a' <= c && c <= 'z')
 317                         c -= 'a' - 26;
 318                 else if ('0' <= c && c <= '9')
 319                         c -= '0' - 52;
 320                 else
 321                         continue; /* garbage */
 322                 switch (pos++) {
 323                 case 0:
 324                         acc = (c << 2);
 325                         break;
 326                 case 1:
 327                         strbuf_addch(out, (acc | (c >> 4)));
 328                         acc = (c & 15) << 4;
 329                         break;
 330                 case 2:
 331                         strbuf_addch(out, (acc | (c >> 2)));
 332                         acc = (c & 3) << 6;
 333                         break;
 334                 case 3:
 335                         strbuf_addch(out, (acc | c));
 336                         acc = pos = 0;
 337                         break;
 338                 }
 339         }
 340         return out;
 341 }
 342
 343 static int convert_to_utf8(struct mailinfo *mi,
 344                            struct strbuf *line, const char *charset)
 345 {
 346         char *out;
 347
 348         if (!mi->metainfo_charset || !charset || !*charset)
 349                 return 0;
 350
 351         if (same_encoding(mi->metainfo_charset, charset))
 352                 return 0;
 353         out = reencode_string(line->buf, mi->metainfo_charset, charset);
 354         if (!out) {
 355                 mi->input_error = -1;
 356                 return error("cannot convert from %s to %s",
 357                              charset, mi->metainfo_charset);
 358         }
 359         strbuf_attach(line, out, strlen(out), strlen(out));
 360         return 0;
 361 }
 362
 363 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 364 {
 365         char *in, *ep, *cp;
 366         struct strbuf outbuf = STRBUF_INIT, *dec;
 367         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 368         int found_error = 1; /* pessimism */
 369
 370         in = it->buf;
 371         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 372                 int encoding;
 373                 strbuf_reset(&charset_q);
 374                 strbuf_reset(&piecebuf);
 375
 376                 if (in != ep) {
 377                         /*
 378                          * We are about to process an encoded-word
 379                          * that begins at ep, but there is something
 380                          * before the encoded word.
 381                          */
 382                         char *scan;
 383                         for (scan = in; scan < ep; scan++)
 384                                 if (!isspace(*scan))
 385                                         break;
 386
 387                         if (scan != ep || in == it->buf) {
 388                                 /*
 389                                  * We should not lose that "something",
 390                                  * unless we have just processed an
 391                                  * encoded-word, and there is only LWS
 392                                  * before the one we are about to process.
 393                                  */
 394                                 strbuf_add(&outbuf, in, ep - in);
 395                         }
 396                 }
 397                 /* E.g.
 398                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 399                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 400                  */
 401                 ep += 2;
 402
 403                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 404                         goto release_return;
 405
 406                 if (cp + 3 - it->buf > it->len)
 407                         goto release_return;
 408                 strbuf_add(&charset_q, ep, cp - ep);
 409
 410                 encoding = cp[1];
 411                 if (!encoding || cp[2] != '?')
 412                         goto release_return;
 413                 ep = strstr(cp + 3, "?=");
 414                 if (!ep)
 415                         goto release_return;
 416                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 417                 switch (tolower(encoding)) {
 418                 default:
 419                         goto release_return;
 420                 case 'b':
 421                         dec = decode_b_segment(&piecebuf);
 422                         break;
 423                 case 'q':
 424                         dec = decode_q_segment(&piecebuf, 1);
 425                         break;
 426                 }
 427                 if (convert_to_utf8(mi, dec, charset_q.buf))
 428                         goto release_return;
 429
 430                 strbuf_addbuf(&outbuf, dec);
 431                 strbuf_release(dec);
 432                 free(dec);
 433                 in = ep + 2;
 434         }
 435         strbuf_addstr(&outbuf, in);
 436         strbuf_reset(it);
 437         strbuf_addbuf(it, &outbuf);
 438         found_error = 0;
 439 release_return:
 440         strbuf_release(&outbuf);
 441         strbuf_release(&charset_q);
 442         strbuf_release(&piecebuf);
 443
 444         if (found_error)
 445                 mi->input_error = -1;
 446 }
 447
 448 static int check_header(struct mailinfo *mi,
 449                         const struct strbuf *line,
 450                         struct strbuf *hdr_data[], int overwrite)
 451 {
 452         int i, ret = 0, len;
 453         struct strbuf sb = STRBUF_INIT;
 454
 455         /* search for the interesting parts */
 456         for (i = 0; header[i]; i++) {
 457                 int len = strlen(header[i]);
 458                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 459                         /* Unwrap inline B and Q encoding, and optionally
 460                          * normalize the meta information to utf8.
 461                          */
 462                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 463                         decode_header(mi, &sb);
 464                         handle_header(&hdr_data[i], &sb);
 465                         ret = 1;
 466                         goto check_header_out;
 467                 }
 468         }
 469
 470         /* Content stuff */
 471         if (cmp_header(line, "Content-Type")) {
 472                 len = strlen("Content-Type: ");
 473                 strbuf_add(&sb, line->buf + len, line->len - len);
 474                 decode_header(mi, &sb);
 475                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 476                 handle_content_type(mi, &sb);
 477                 ret = 1;
 478                 goto check_header_out;
 479         }
 480         if (cmp_header(line, "Content-Transfer-Encoding")) {
 481                 len = strlen("Content-Transfer-Encoding: ");
 482                 strbuf_add(&sb, line->buf + len, line->len - len);
 483                 decode_header(mi, &sb);
 484                 handle_content_transfer_encoding(mi, &sb);
 485                 ret = 1;
 486                 goto check_header_out;
 487         }
 488         if (cmp_header(line, "Message-Id")) {
 489                 len = strlen("Message-Id: ");
 490                 strbuf_add(&sb, line->buf + len, line->len - len);
 491                 decode_header(mi, &sb);
 492                 if (mi->add_message_id)
 493                         mi->message_id = strbuf_detach(&sb, NULL);
 494                 ret = 1;
 495                 goto check_header_out;
 496         }
 497
 498 check_header_out:
 499         strbuf_release(&sb);
 500         return ret;
 501 }
 502
 503 /*
 504  * Returns 1 if the given line or any line beginning with the given line is an
 505  * in-body header (that is, check_header will succeed when passed
 506  * mi->s_hdr_data).
 507  */
 508 static int is_inbody_header(const struct mailinfo *mi,
 509                             const struct strbuf *line)
 510 {
 511         int i;
 512         for (i = 0; header[i]; i++)
 513                 if (!mi->s_hdr_data[i] && cmp_header(line, header[i]))
 514                         return 1;
 515         return 0;
 516 }
 517
 518 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 519 {
 520         struct strbuf *ret;
 521
 522         switch (mi->transfer_encoding) {
 523         case TE_QP:
 524                 ret = decode_q_segment(line, 0);
 525                 break;
 526         case TE_BASE64:
 527                 ret = decode_b_segment(line);
 528                 break;
 529         case TE_DONTCARE:
 530         default:
 531                 return;
 532         }
 533         strbuf_reset(line);
 534         strbuf_addbuf(line, ret);
 535         strbuf_release(ret);
 536         free(ret);
 537 }
 538
 539 static inline int patchbreak(const struct strbuf *line)
 540 {
 541         size_t i;
 542
 543         /* Beginning of a "diff -" header? */
 544         if (starts_with(line->buf, "diff -"))
 545                 return 1;
 546
 547         /* CVS "Index: " line? */
 548         if (starts_with(line->buf, "Index: "))
 549                 return 1;
 550
 551         /*
 552          * "--- <filename>" starts patches without headers
 553          * "---<sp>*" is a manual separator
 554          */
 555         if (line->len < 4)
 556                 return 0;
 557
 558         if (starts_with(line->buf, "---")) {
 559                 /* space followed by a filename? */
 560                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 561                         return 1;
 562                 /* Just whitespace? */
 563                 for (i = 3; i < line->len; i++) {
 564                         unsigned char c = line->buf[i];
 565                         if (c == '\n')
 566                                 return 1;
 567                         if (!isspace(c))
 568                                 break;
 569                 }
 570                 return 0;
 571         }
 572         return 0;
 573 }
 574
 575 static int is_scissors_line(const char *line)
 576 {
 577         const char *c;
 578         int scissors = 0, gap = 0;
 579         const char *first_nonblank = NULL, *last_nonblank = NULL;
 580         int visible, perforation = 0, in_perforation = 0;
 581
 582         for (c = line; *c; c++) {
 583                 if (isspace(*c)) {
 584                         if (in_perforation) {
 585                                 perforation++;
 586                                 gap++;
 587                         }
 588                         continue;
 589                 }
 590                 last_nonblank = c;
 591                 if (first_nonblank == NULL)
 592                         first_nonblank = c;
 593                 if (*c == '-') {
 594                         in_perforation = 1;
 595                         perforation++;
 596                         continue;
 597                 }
 598                 if ((!memcmp(c, ">8", 2) || !memcmp(c, "8<", 2) ||
 599                      !memcmp(c, ">%", 2) || !memcmp(c, "%<", 2))) {
 600                         in_perforation = 1;
 601                         perforation += 2;
 602                         scissors += 2;
 603                         c++;
 604                         continue;
 605                 }
 606                 in_perforation = 0;
 607         }
 608
 609         /*
 610          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 611          * Even though there can be arbitrary cruft on the same line
 612          * (e.g. "cut here"), in order to avoid misidentification, the
 613          * perforation must occupy more than a third of the visible
 614          * width of the line, and dashes and scissors must occupy more
 615          * than half of the perforation.
 616          */
 617
 618         if (first_nonblank && last_nonblank)
 619                 visible = last_nonblank - first_nonblank + 1;
 620         else
 621                 visible = 0;
 622         return (scissors && 8 <= visible &&
 623                 visible < perforation * 3 &&
 624                 gap * 2 < perforation);
 625 }
 626
 627 static void flush_inbody_header_accum(struct mailinfo *mi)
 628 {
 629         if (!mi->inbody_header_accum.len)
 630                 return;
 631         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 632                 die("BUG: inbody_header_accum, if not empty, must always contain a valid in-body header");
 633         strbuf_reset(&mi->inbody_header_accum);
 634 }
 635
 636 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 637 {
 638         if (mi->inbody_header_accum.len &&
 639             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 640                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 641                         /*
 642                          * This is a scissors line; do not consider this line
 643                          * as a header continuation line.
 644                          */
 645                         flush_inbody_header_accum(mi);
 646                         return 0;
 647                 }
 648                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 649                 strbuf_addbuf(&mi->inbody_header_accum, line);
 650                 return 1;
 651         }
 652
 653         flush_inbody_header_accum(mi);
 654
 655         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 656                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 657         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 658                 int i;
 659                 for (i = 0; header[i]; i++)
 660                         if (!strcmp("Subject", header[i])) {
 661                                 handle_header(&mi->s_hdr_data[i], line);
 662                                 return 1;
 663                         }
 664                 return 0;
 665         }
 666         if (is_inbody_header(mi, line)) {
 667                 strbuf_addbuf(&mi->inbody_header_accum, line);
 668                 return 1;
 669         }
 670         return 0;
 671 }
 672
 673 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 674 {
 675         assert(!mi->filter_stage);
 676
 677         if (mi->header_stage) {
 678                 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
 679                         return 0;
 680         }
 681
 682         if (mi->use_inbody_headers && mi->header_stage) {
 683                 mi->header_stage = check_inbody_header(mi, line);
 684                 if (mi->header_stage)
 685                         return 0;
 686         } else
 687                 /* Only trim the first (blank) line of the commit message
 688                  * when ignoring in-body headers.
 689                  */
 690                 mi->header_stage = 0;
 691
 692         /* normalize the log message to UTF-8. */
 693         if (convert_to_utf8(mi, line, mi->charset.buf))
 694                 return 0; /* mi->input_error already set */
 695
 696         if (mi->use_scissors && is_scissors_line(line->buf)) {
 697                 int i;
 698
 699                 strbuf_setlen(&mi->log_message, 0);
 700                 mi->header_stage = 1;
 701
 702                 /*
 703                  * We may have already read "secondary headers"; purge
 704                  * them to give ourselves a clean restart.
 705                  */
 706                 for (i = 0; header[i]; i++) {
 707                         if (mi->s_hdr_data[i])
 708                                 strbuf_release(mi->s_hdr_data[i]);
 709                         mi->s_hdr_data[i] = NULL;
 710                 }
 711                 return 0;
 712         }
 713
 714         if (patchbreak(line)) {
 715                 if (mi->message_id)
 716                         strbuf_addf(&mi->log_message,
 717                                     "Message-Id: %s\n", mi->message_id);
 718                 return 1;
 719         }
 720
 721         strbuf_addbuf(&mi->log_message, line);
 722         return 0;
 723 }
 724
 725 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 726 {
 727         fwrite(line->buf, 1, line->len, mi->patchfile);
 728         mi->patch_lines++;
 729 }
 730
 731 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 732 {
 733         switch (mi->filter_stage) {
 734         case 0:
 735                 if (!handle_commit_msg(mi, line))
 736                         break;
 737                 mi->filter_stage++;
 738         case 1:
 739                 handle_patch(mi, line);
 740                 break;
 741         }
 742 }
 743
 744 static int is_rfc2822_header(const struct strbuf *line)
 745 {
 746         /*
 747          * The section that defines the loosest possible
 748          * field name is "3.6.8 Optional fields".
 749          *
 750          * optional-field = field-name ":" unstructured CRLF
 751          * field-name = 1*ftext
 752          * ftext = %d33-57 / %59-126
 753          */
 754         int ch;
 755         char *cp = line->buf;
 756
 757         /* Count mbox From headers as headers */
 758         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 759                 return 1;
 760
 761         while ((ch = *cp++)) {
 762                 if (ch == ':')
 763                         return 1;
 764                 if ((33 <= ch && ch <= 57) ||
 765                     (59 <= ch && ch <= 126))
 766                         continue;
 767                 break;
 768         }
 769         return 0;
 770 }
 771
 772 static int read_one_header_line(struct strbuf *line, FILE *in)
 773 {
 774         struct strbuf continuation = STRBUF_INIT;
 775
 776         /* Get the first part of the line. */
 777         if (strbuf_getline_lf(line, in))
 778                 return 0;
 779
 780         /*
 781          * Is it an empty line or not a valid rfc2822 header?
 782          * If so, stop here, and return false ("not a header")
 783          */
 784         strbuf_rtrim(line);
 785         if (!line->len || !is_rfc2822_header(line)) {
 786                 /* Re-add the newline */
 787                 strbuf_addch(line, '\n');
 788                 return 0;
 789         }
 790
 791         /*
 792          * Now we need to eat all the continuation lines..
 793          * Yuck, 2822 header "folding"
 794          */
 795         for (;;) {
 796                 int peek;
 797
 798                 peek = fgetc(in); ungetc(peek, in);
 799                 if (peek != ' ' && peek != '\t')
 800                         break;
 801                 if (strbuf_getline_lf(&continuation, in))
 802                         break;
 803                 continuation.buf[0] = ' ';
 804                 strbuf_rtrim(&continuation);
 805                 strbuf_addbuf(line, &continuation);
 806         }
 807         strbuf_release(&continuation);
 808
 809         return 1;
 810 }
 811
 812 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 813 {
 814         while (!strbuf_getline_lf(line, mi->input)) {
 815                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 816                         return 1;
 817         }
 818         return 0;
 819 }
 820
 821 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 822 {
 823         struct strbuf newline = STRBUF_INIT;
 824
 825         strbuf_addch(&newline, '\n');
 826 again:
 827         if (line->len >= (*(mi->content_top))->len + 2 &&
 828             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 829                 /* we hit an end boundary */
 830                 /* pop the current boundary off the stack */
 831                 strbuf_release(*(mi->content_top));
 832                 free(*(mi->content_top));
 833                 *(mi->content_top) = NULL;
 834
 835                 /* technically won't happen as is_multipart_boundary()
 836                    will fail first.  But just in case..
 837                  */
 838                 if (--mi->content_top < mi->content) {
 839                         error("Detected mismatched boundaries, can't recover");
 840                         mi->input_error = -1;
 841                         mi->content_top = mi->content;
 842                         return 0;
 843                 }
 844                 handle_filter(mi, &newline);
 845                 strbuf_release(&newline);
 846                 if (mi->input_error)
 847                         return 0;
 848
 849                 /* skip to the next boundary */
 850                 if (!find_boundary(mi, line))
 851                         return 0;
 852                 goto again;
 853         }
 854
 855         /* set some defaults */
 856         mi->transfer_encoding = TE_DONTCARE;
 857         strbuf_reset(&mi->charset);
 858
 859         /* slurp in this section's info */
 860         while (read_one_header_line(line, mi->input))
 861                 check_header(mi, line, mi->p_hdr_data, 0);
 862
 863         strbuf_release(&newline);
 864         /* replenish line */
 865         if (strbuf_getline_lf(line, mi->input))
 866                 return 0;
 867         strbuf_addch(line, '\n');
 868         return 1;
 869 }
 870
 871 static void handle_body(struct mailinfo *mi, struct strbuf *line)
 872 {
 873         struct strbuf prev = STRBUF_INIT;
 874
 875         /* Skip up to the first boundary */
 876         if (*(mi->content_top)) {
 877                 if (!find_boundary(mi, line))
 878                         goto handle_body_out;
 879         }
 880
 881         do {
 882                 /* process any boundary lines */
 883                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
 884                         /* flush any leftover */
 885                         if (prev.len) {
 886                                 handle_filter(mi, &prev);
 887                                 strbuf_reset(&prev);
 888                         }
 889                         if (!handle_boundary(mi, line))
 890                                 goto handle_body_out;
 891                 }
 892
 893                 /* Unwrap transfer encoding */
 894                 decode_transfer_encoding(mi, line);
 895
 896                 switch (mi->transfer_encoding) {
 897                 case TE_BASE64:
 898                 case TE_QP:
 899                 {
 900                         struct strbuf **lines, **it, *sb;
 901
 902                         /* Prepend any previous partial lines */
 903                         strbuf_insert(line, 0, prev.buf, prev.len);
 904                         strbuf_reset(&prev);
 905
 906                         /*
 907                          * This is a decoded line that may contain
 908                          * multiple new lines.  Pass only one chunk
 909                          * at a time to handle_filter()
 910                          */
 911                         lines = strbuf_split(line, '\n');
 912                         for (it = lines; (sb = *it); it++) {
 913                                 if (*(it + 1) == NULL) /* The last line */
 914                                         if (sb->buf[sb->len - 1] != '\n') {
 915                                                 /* Partial line, save it for later. */
 916                                                 strbuf_addbuf(&prev, sb);
 917                                                 break;
 918                                         }
 919                                 handle_filter(mi, sb);
 920                         }
 921                         /*
 922                          * The partial chunk is saved in "prev" and will be
 923                          * appended by the next iteration of read_line_with_nul().
 924                          */
 925                         strbuf_list_free(lines);
 926                         break;
 927                 }
 928                 default:
 929                         handle_filter(mi, line);
 930                 }
 931
 932                 if (mi->input_error)
 933                         break;
 934         } while (!strbuf_getwholeline(line, mi->input, '\n'));
 935
 936         flush_inbody_header_accum(mi);
 937
 938 handle_body_out:
 939         strbuf_release(&prev);
 940 }
 941
 942 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 943 {
 944         const char *sp = data->buf;
 945         while (1) {
 946                 char *ep = strchr(sp, '\n');
 947                 int len;
 948                 if (!ep)
 949                         len = strlen(sp);
 950                 else
 951                         len = ep - sp;
 952                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 953                 if (!ep)
 954                         break;
 955                 sp = ep + 1;
 956         }
 957 }
 958
 959 static void handle_info(struct mailinfo *mi)
 960 {
 961         struct strbuf *hdr;
 962         int i;
 963
 964         for (i = 0; header[i]; i++) {
 965                 /* only print inbody headers if we output a patch file */
 966                 if (mi->patch_lines && mi->s_hdr_data[i])
 967                         hdr = mi->s_hdr_data[i];
 968                 else if (mi->p_hdr_data[i])
 969                         hdr = mi->p_hdr_data[i];
 970                 else
 971                         continue;
 972
 973                 if (!strcmp(header[i], "Subject")) {
 974                         if (!mi->keep_subject) {
 975                                 cleanup_subject(mi, hdr);
 976                                 cleanup_space(hdr);
 977                         }
 978                         output_header_lines(mi->output, "Subject", hdr);
 979                 } else if (!strcmp(header[i], "From")) {
 980                         cleanup_space(hdr);
 981                         handle_from(mi, hdr);
 982                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
 983                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
 984                 } else {
 985                         cleanup_space(hdr);
 986                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
 987                 }
 988         }
 989         fprintf(mi->output, "\n");
 990 }
 991
 992 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
 993 {
 994         FILE *cmitmsg;
 995         int peek;
 996         struct strbuf line = STRBUF_INIT;
 997
 998         cmitmsg = fopen(msg, "w");
 999         if (!cmitmsg) {
1000                 perror(msg);
1001                 return -1;
1002         }
1003         mi->patchfile = fopen(patch, "w");
1004         if (!mi->patchfile) {
1005                 perror(patch);
1006                 fclose(cmitmsg);
1007                 return -1;
1008         }
1009
1010         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
1011         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
1012
1013         do {
1014                 peek = fgetc(mi->input);
1015         } while (isspace(peek));
1016         ungetc(peek, mi->input);
1017
1018         /* process the email header */
1019         while (read_one_header_line(&line, mi->input))
1020                 check_header(mi, &line, mi->p_hdr_data, 1);
1021
1022         handle_body(mi, &line);
1023         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1024         fclose(cmitmsg);
1025         fclose(mi->patchfile);
1026
1027         handle_info(mi);
1028         strbuf_release(&line);
1029         return mi->input_error;
1030 }
1031
1032 static int git_mailinfo_config(const char *var, const char *value, void *mi_)
1033 {
1034         struct mailinfo *mi = mi_;
1035
1036         if (!starts_with(var, "mailinfo."))
1037                 return git_default_config(var, value, NULL);
1038         if (!strcmp(var, "mailinfo.scissors")) {
1039                 mi->use_scissors = git_config_bool(var, value);
1040                 return 0;
1041         }
1042         /* perhaps others here */
1043         return 0;
1044 }
1045
1046 void setup_mailinfo(struct mailinfo *mi)
1047 {
1048         memset(mi, 0, sizeof(*mi));
1049         strbuf_init(&mi->name, 0);
1050         strbuf_init(&mi->email, 0);
1051         strbuf_init(&mi->charset, 0);
1052         strbuf_init(&mi->log_message, 0);
1053         strbuf_init(&mi->inbody_header_accum, 0);
1054         mi->header_stage = 1;
1055         mi->use_inbody_headers = 1;
1056         mi->content_top = mi->content;
1057         git_config(git_mailinfo_config, mi);
1058 }
1059
1060 void clear_mailinfo(struct mailinfo *mi)
1061 {
1062         int i;
1063
1064         strbuf_release(&mi->name);
1065         strbuf_release(&mi->email);
1066         strbuf_release(&mi->charset);
1067         strbuf_release(&mi->inbody_header_accum);
1068         free(mi->message_id);
1069
1070         for (i = 0; mi->p_hdr_data[i]; i++)
1071                 strbuf_release(mi->p_hdr_data[i]);
1072         free(mi->p_hdr_data);
1073         for (i = 0; mi->s_hdr_data[i]; i++)
1074                 strbuf_release(mi->s_hdr_data[i]);
1075         free(mi->s_hdr_data);
1076
1077         while (mi->content < mi->content_top) {
1078                 free(*(mi->content_top));
1079                 mi->content_top--;
1080         }
1081
1082         strbuf_release(&mi->log_message);
1083 }