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         /* for inbody stuff */
 499         if (starts_with(line->buf, ">From") && isspace(line->buf[5])) {
 500                 ret = is_format_patch_separator(line->buf + 1, line->len - 1);
 501                 goto check_header_out;
 502         }
 503         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 504                 for (i = 0; header[i]; i++) {
 505                         if (!strcmp("Subject", header[i])) {
 506                                 handle_header(&hdr_data[i], line);
 507                                 ret = 1;
 508                                 goto check_header_out;
 509                         }
 510                 }
 511         }
 512
 513 check_header_out:
 514         strbuf_release(&sb);
 515         return ret;
 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 struct strbuf *line)
 576 {
 577         size_t i, len = line->len;
 578         int scissors = 0, gap = 0;
 579         int first_nonblank = -1;
 580         int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
 581         const char *buf = line->buf;
 582
 583         for (i = 0; i < len; i++) {
 584                 if (isspace(buf[i])) {
 585                         if (in_perforation) {
 586                                 perforation++;
 587                                 gap++;
 588                         }
 589                         continue;
 590                 }
 591                 last_nonblank = i;
 592                 if (first_nonblank < 0)
 593                         first_nonblank = i;
 594                 if (buf[i] == '-') {
 595                         in_perforation = 1;
 596                         perforation++;
 597                         continue;
 598                 }
 599                 if (i + 1 < len &&
 600                     (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
 601                      !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
 602                         in_perforation = 1;
 603                         perforation += 2;
 604                         scissors += 2;
 605                         i++;
 606                         continue;
 607                 }
 608                 in_perforation = 0;
 609         }
 610
 611         /*
 612          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 613          * Even though there can be arbitrary cruft on the same line
 614          * (e.g. "cut here"), in order to avoid misidentification, the
 615          * perforation must occupy more than a third of the visible
 616          * width of the line, and dashes and scissors must occupy more
 617          * than half of the perforation.
 618          */
 619
 620         visible = last_nonblank - first_nonblank + 1;
 621         return (scissors && 8 <= visible &&
 622                 visible < perforation * 3 &&
 623                 gap * 2 < perforation);
 624 }
 625
 626 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 627 {
 628         assert(!mi->filter_stage);
 629
 630         if (mi->header_stage) {
 631                 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
 632                         return 0;
 633         }
 634
 635         if (mi->use_inbody_headers && mi->header_stage) {
 636                 mi->header_stage = check_header(mi, line, mi->s_hdr_data, 0);
 637                 if (mi->header_stage)
 638                         return 0;
 639         } else
 640                 /* Only trim the first (blank) line of the commit message
 641                  * when ignoring in-body headers.
 642                  */
 643                 mi->header_stage = 0;
 644
 645         /* normalize the log message to UTF-8. */
 646         if (convert_to_utf8(mi, line, mi->charset.buf))
 647                 return 0; /* mi->input_error already set */
 648
 649         if (mi->use_scissors && is_scissors_line(line)) {
 650                 int i;
 651
 652                 strbuf_setlen(&mi->log_message, 0);
 653                 mi->header_stage = 1;
 654
 655                 /*
 656                  * We may have already read "secondary headers"; purge
 657                  * them to give ourselves a clean restart.
 658                  */
 659                 for (i = 0; header[i]; i++) {
 660                         if (mi->s_hdr_data[i])
 661                                 strbuf_release(mi->s_hdr_data[i]);
 662                         mi->s_hdr_data[i] = NULL;
 663                 }
 664                 return 0;
 665         }
 666
 667         if (patchbreak(line)) {
 668                 if (mi->message_id)
 669                         strbuf_addf(&mi->log_message,
 670                                     "Message-Id: %s\n", mi->message_id);
 671                 return 1;
 672         }
 673
 674         strbuf_addbuf(&mi->log_message, line);
 675         return 0;
 676 }
 677
 678 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 679 {
 680         fwrite(line->buf, 1, line->len, mi->patchfile);
 681         mi->patch_lines++;
 682 }
 683
 684 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 685 {
 686         switch (mi->filter_stage) {
 687         case 0:
 688                 if (!handle_commit_msg(mi, line))
 689                         break;
 690                 mi->filter_stage++;
 691         case 1:
 692                 handle_patch(mi, line);
 693                 break;
 694         }
 695 }
 696
 697 static int is_rfc2822_header(const struct strbuf *line)
 698 {
 699         /*
 700          * The section that defines the loosest possible
 701          * field name is "3.6.8 Optional fields".
 702          *
 703          * optional-field = field-name ":" unstructured CRLF
 704          * field-name = 1*ftext
 705          * ftext = %d33-57 / %59-126
 706          */
 707         int ch;
 708         char *cp = line->buf;
 709
 710         /* Count mbox From headers as headers */
 711         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 712                 return 1;
 713
 714         while ((ch = *cp++)) {
 715                 if (ch == ':')
 716                         return 1;
 717                 if ((33 <= ch && ch <= 57) ||
 718                     (59 <= ch && ch <= 126))
 719                         continue;
 720                 break;
 721         }
 722         return 0;
 723 }
 724
 725 static int read_one_header_line(struct strbuf *line, FILE *in)
 726 {
 727         struct strbuf continuation = STRBUF_INIT;
 728
 729         /* Get the first part of the line. */
 730         if (strbuf_getline_lf(line, in))
 731                 return 0;
 732
 733         /*
 734          * Is it an empty line or not a valid rfc2822 header?
 735          * If so, stop here, and return false ("not a header")
 736          */
 737         strbuf_rtrim(line);
 738         if (!line->len || !is_rfc2822_header(line)) {
 739                 /* Re-add the newline */
 740                 strbuf_addch(line, '\n');
 741                 return 0;
 742         }
 743
 744         /*
 745          * Now we need to eat all the continuation lines..
 746          * Yuck, 2822 header "folding"
 747          */
 748         for (;;) {
 749                 int peek;
 750
 751                 peek = fgetc(in); ungetc(peek, in);
 752                 if (peek != ' ' && peek != '\t')
 753                         break;
 754                 if (strbuf_getline_lf(&continuation, in))
 755                         break;
 756                 continuation.buf[0] = ' ';
 757                 strbuf_rtrim(&continuation);
 758                 strbuf_addbuf(line, &continuation);
 759         }
 760         strbuf_release(&continuation);
 761
 762         return 1;
 763 }
 764
 765 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 766 {
 767         while (!strbuf_getline_lf(line, mi->input)) {
 768                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 769                         return 1;
 770         }
 771         return 0;
 772 }
 773
 774 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 775 {
 776         struct strbuf newline = STRBUF_INIT;
 777
 778         strbuf_addch(&newline, '\n');
 779 again:
 780         if (line->len >= (*(mi->content_top))->len + 2 &&
 781             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 782                 /* we hit an end boundary */
 783                 /* pop the current boundary off the stack */
 784                 strbuf_release(*(mi->content_top));
 785                 free(*(mi->content_top));
 786                 *(mi->content_top) = NULL;
 787
 788                 /* technically won't happen as is_multipart_boundary()
 789                    will fail first.  But just in case..
 790                  */
 791                 if (--mi->content_top < mi->content) {
 792                         error("Detected mismatched boundaries, can't recover");
 793                         mi->input_error = -1;
 794                         mi->content_top = mi->content;
 795                         return 0;
 796                 }
 797                 handle_filter(mi, &newline);
 798                 strbuf_release(&newline);
 799                 if (mi->input_error)
 800                         return 0;
 801
 802                 /* skip to the next boundary */
 803                 if (!find_boundary(mi, line))
 804                         return 0;
 805                 goto again;
 806         }
 807
 808         /* set some defaults */
 809         mi->transfer_encoding = TE_DONTCARE;
 810         strbuf_reset(&mi->charset);
 811
 812         /* slurp in this section's info */
 813         while (read_one_header_line(line, mi->input))
 814                 check_header(mi, line, mi->p_hdr_data, 0);
 815
 816         strbuf_release(&newline);
 817         /* replenish line */
 818         if (strbuf_getline_lf(line, mi->input))
 819                 return 0;
 820         strbuf_addch(line, '\n');
 821         return 1;
 822 }
 823
 824 static void handle_body(struct mailinfo *mi, struct strbuf *line)
 825 {
 826         struct strbuf prev = STRBUF_INIT;
 827
 828         /* Skip up to the first boundary */
 829         if (*(mi->content_top)) {
 830                 if (!find_boundary(mi, line))
 831                         goto handle_body_out;
 832         }
 833
 834         do {
 835                 /* process any boundary lines */
 836                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
 837                         /* flush any leftover */
 838                         if (prev.len) {
 839                                 handle_filter(mi, &prev);
 840                                 strbuf_reset(&prev);
 841                         }
 842                         if (!handle_boundary(mi, line))
 843                                 goto handle_body_out;
 844                 }
 845
 846                 /* Unwrap transfer encoding */
 847                 decode_transfer_encoding(mi, line);
 848
 849                 switch (mi->transfer_encoding) {
 850                 case TE_BASE64:
 851                 case TE_QP:
 852                 {
 853                         struct strbuf **lines, **it, *sb;
 854
 855                         /* Prepend any previous partial lines */
 856                         strbuf_insert(line, 0, prev.buf, prev.len);
 857                         strbuf_reset(&prev);
 858
 859                         /*
 860                          * This is a decoded line that may contain
 861                          * multiple new lines.  Pass only one chunk
 862                          * at a time to handle_filter()
 863                          */
 864                         lines = strbuf_split(line, '\n');
 865                         for (it = lines; (sb = *it); it++) {
 866                                 if (*(it + 1) == NULL) /* The last line */
 867                                         if (sb->buf[sb->len - 1] != '\n') {
 868                                                 /* Partial line, save it for later. */
 869                                                 strbuf_addbuf(&prev, sb);
 870                                                 break;
 871                                         }
 872                                 handle_filter(mi, sb);
 873                         }
 874                         /*
 875                          * The partial chunk is saved in "prev" and will be
 876                          * appended by the next iteration of read_line_with_nul().
 877                          */
 878                         strbuf_list_free(lines);
 879                         break;
 880                 }
 881                 default:
 882                         handle_filter(mi, line);
 883                 }
 884
 885                 if (mi->input_error)
 886                         break;
 887         } while (!strbuf_getwholeline(line, mi->input, '\n'));
 888
 889 handle_body_out:
 890         strbuf_release(&prev);
 891 }
 892
 893 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 894 {
 895         const char *sp = data->buf;
 896         while (1) {
 897                 char *ep = strchr(sp, '\n');
 898                 int len;
 899                 if (!ep)
 900                         len = strlen(sp);
 901                 else
 902                         len = ep - sp;
 903                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 904                 if (!ep)
 905                         break;
 906                 sp = ep + 1;
 907         }
 908 }
 909
 910 static void handle_info(struct mailinfo *mi)
 911 {
 912         struct strbuf *hdr;
 913         int i;
 914
 915         for (i = 0; header[i]; i++) {
 916                 /* only print inbody headers if we output a patch file */
 917                 if (mi->patch_lines && mi->s_hdr_data[i])
 918                         hdr = mi->s_hdr_data[i];
 919                 else if (mi->p_hdr_data[i])
 920                         hdr = mi->p_hdr_data[i];
 921                 else
 922                         continue;
 923
 924                 if (!strcmp(header[i], "Subject")) {
 925                         if (!mi->keep_subject) {
 926                                 cleanup_subject(mi, hdr);
 927                                 cleanup_space(hdr);
 928                         }
 929                         output_header_lines(mi->output, "Subject", hdr);
 930                 } else if (!strcmp(header[i], "From")) {
 931                         cleanup_space(hdr);
 932                         handle_from(mi, hdr);
 933                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
 934                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
 935                 } else {
 936                         cleanup_space(hdr);
 937                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
 938                 }
 939         }
 940         fprintf(mi->output, "\n");
 941 }
 942
 943 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
 944 {
 945         FILE *cmitmsg;
 946         int peek;
 947         struct strbuf line = STRBUF_INIT;
 948
 949         cmitmsg = fopen(msg, "w");
 950         if (!cmitmsg) {
 951                 perror(msg);
 952                 return -1;
 953         }
 954         mi->patchfile = fopen(patch, "w");
 955         if (!mi->patchfile) {
 956                 perror(patch);
 957                 fclose(cmitmsg);
 958                 return -1;
 959         }
 960
 961         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
 962         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
 963
 964         do {
 965                 peek = fgetc(mi->input);
 966         } while (isspace(peek));
 967         ungetc(peek, mi->input);
 968
 969         /* process the email header */
 970         while (read_one_header_line(&line, mi->input))
 971                 check_header(mi, &line, mi->p_hdr_data, 1);
 972
 973         handle_body(mi, &line);
 974         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
 975         fclose(cmitmsg);
 976         fclose(mi->patchfile);
 977
 978         handle_info(mi);
 979         strbuf_release(&line);
 980         return mi->input_error;
 981 }
 982
 983 static int git_mailinfo_config(const char *var, const char *value, void *mi_)
 984 {
 985         struct mailinfo *mi = mi_;
 986
 987         if (!starts_with(var, "mailinfo."))
 988                 return git_default_config(var, value, NULL);
 989         if (!strcmp(var, "mailinfo.scissors")) {
 990                 mi->use_scissors = git_config_bool(var, value);
 991                 return 0;
 992         }
 993         /* perhaps others here */
 994         return 0;
 995 }
 996
 997 void setup_mailinfo(struct mailinfo *mi)
 998 {
 999         memset(mi, 0, sizeof(*mi));
1000         strbuf_init(&mi->name, 0);
1001         strbuf_init(&mi->email, 0);
1002         strbuf_init(&mi->charset, 0);
1003         strbuf_init(&mi->log_message, 0);
1004         mi->header_stage = 1;
1005         mi->use_inbody_headers = 1;
1006         mi->content_top = mi->content;
1007         git_config(git_mailinfo_config, mi);
1008 }
1009
1010 void clear_mailinfo(struct mailinfo *mi)
1011 {
1012         int i;
1013
1014         strbuf_release(&mi->name);
1015         strbuf_release(&mi->email);
1016         strbuf_release(&mi->charset);
1017         free(mi->message_id);
1018
1019         for (i = 0; mi->p_hdr_data[i]; i++)
1020                 strbuf_release(mi->p_hdr_data[i]);
1021         free(mi->p_hdr_data);
1022         for (i = 0; mi->s_hdr_data[i]; i++)
1023                 strbuf_release(mi->s_hdr_data[i]);
1024         free(mi->s_hdr_data);
1025
1026         while (mi->content < mi->content_top) {
1027                 free(*(mi->content_top));
1028                 mi->content_top--;
1029         }
1030
1031         strbuf_release(&mi->log_message);
1032 }