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