mailinfo.c

   1 #define DISABLE_SIGN_COMPARE_WARNINGS
   2
   3 #include "git-compat-util.h"
   4 #include "config.h"
   5 #include "gettext.h"
   6 #include "hex-ll.h"
   7 #include "utf8.h"
   8 #include "strbuf.h"
   9 #include "mailinfo.h"
  10
  11 static void cleanup_space(struct strbuf *sb)
  12 {
  13         size_t pos, cnt;
  14         for (pos = 0; pos < sb->len; pos++) {
  15                 if (isspace(sb->buf[pos])) {
  16                         sb->buf[pos] = ' ';
  17                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
  18                         strbuf_remove(sb, pos + 1, cnt);
  19                 }
  20         }
  21 }
  22
  23 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  24 {
  25         struct strbuf *src = name;
  26         if (!name->len || 60 < name->len || strpbrk(name->buf, "@<>"))
  27                 src = email;
  28         else if (name == out)
  29                 return;
  30         strbuf_reset(out);
  31         strbuf_addbuf(out, src);
  32 }
  33
  34 static void parse_bogus_from(struct mailinfo *mi, const struct strbuf *line)
  35 {
  36         /* John Doe <johndoe> */
  37
  38         char *bra, *ket;
  39         /* This is fallback, so do not bother if we already have an
  40          * e-mail address.
  41          */
  42         if (mi->email.len)
  43                 return;
  44
  45         bra = strchr(line->buf, '<');
  46         if (!bra)
  47                 return;
  48         ket = strchr(bra, '>');
  49         if (!ket)
  50                 return;
  51
  52         strbuf_reset(&mi->email);
  53         strbuf_add(&mi->email, bra + 1, ket - bra - 1);
  54
  55         strbuf_reset(&mi->name);
  56         strbuf_add(&mi->name, line->buf, bra - line->buf);
  57         strbuf_trim(&mi->name);
  58         get_sane_name(&mi->name, &mi->name, &mi->email);
  59 }
  60
  61 static const char *unquote_comment(struct strbuf *outbuf, const char *in)
  62 {
  63         int take_next_literally = 0;
  64         int depth = 1;
  65
  66         strbuf_addch(outbuf, '(');
  67
  68         while (*in) {
  69                 int c = *in++;
  70                 if (take_next_literally == 1) {
  71                         take_next_literally = 0;
  72                 } else {
  73                         switch (c) {
  74                         case '\\':
  75                                 take_next_literally = 1;
  76                                 continue;
  77                         case '(':
  78                                 strbuf_addch(outbuf, '(');
  79                                 depth++;
  80                                 continue;
  81                         case ')':
  82                                 strbuf_addch(outbuf, ')');
  83                                 if (!--depth)
  84                                         return in;
  85                                 continue;
  86                         }
  87                 }
  88
  89                 strbuf_addch(outbuf, c);
  90         }
  91
  92         return in;
  93 }
  94
  95 static const char *unquote_quoted_string(struct strbuf *outbuf, const char *in)
  96 {
  97         int take_next_literally = 0;
  98
  99         while (*in) {
 100                 int c = *in++;
 101                 if (take_next_literally == 1) {
 102                         take_next_literally = 0;
 103                 } else {
 104                         switch (c) {
 105                         case '\\':
 106                                 take_next_literally = 1;
 107                                 continue;
 108                         case '"':
 109                                 return in;
 110                         }
 111                 }
 112
 113                 strbuf_addch(outbuf, c);
 114         }
 115
 116         return in;
 117 }
 118
 119 static void unquote_quoted_pair(struct strbuf *line)
 120 {
 121         struct strbuf outbuf;
 122         const char *in = line->buf;
 123         int c;
 124
 125         strbuf_init(&outbuf, line->len);
 126
 127         while ((c = *in++) != 0) {
 128                 switch (c) {
 129                 case '"':
 130                         in = unquote_quoted_string(&outbuf, in);
 131                         continue;
 132                 case '(':
 133                         in = unquote_comment(&outbuf, in);
 134                         continue;
 135                 }
 136
 137                 strbuf_addch(&outbuf, c);
 138         }
 139
 140         strbuf_swap(&outbuf, line);
 141         strbuf_release(&outbuf);
 142
 143 }
 144
 145 static void handle_from(struct mailinfo *mi, const struct strbuf *from)
 146 {
 147         char *at;
 148         size_t el;
 149         struct strbuf f;
 150
 151         strbuf_init(&f, from->len);
 152         strbuf_addbuf(&f, from);
 153
 154         unquote_quoted_pair(&f);
 155
 156         at = strchr(f.buf, '@');
 157         if (!at) {
 158                 parse_bogus_from(mi, from);
 159                 goto out;
 160         }
 161
 162         /*
 163          * If we already have one email, don't take any confusing lines
 164          */
 165         if (mi->email.len && strchr(at + 1, '@'))
 166                 goto out;
 167
 168         /* Pick up the string around '@', possibly delimited with <>
 169          * pair; that is the email part.
 170          */
 171         while (at > f.buf) {
 172                 char c = at[-1];
 173                 if (isspace(c))
 174                         break;
 175                 if (c == '<') {
 176                         at[-1] = ' ';
 177                         break;
 178                 }
 179                 at--;
 180         }
 181         el = strcspn(at, " \n\t\r\v\f>");
 182         strbuf_reset(&mi->email);
 183         strbuf_add(&mi->email, at, el);
 184         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 185
 186         /* The remainder is name.  It could be
 187          *
 188          * - "John Doe <john.doe@xz>"                   (a), or
 189          * - "john.doe@xz (John Doe)"                   (b), or
 190          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 191          *
 192          * but we have removed the email part, so
 193          *
 194          * - remove extra spaces which could stay after email (case 'c'), and
 195          * - trim from both ends, possibly removing the () pair at the end
 196          *   (cases 'a' and 'b').
 197          */
 198         cleanup_space(&f);
 199         strbuf_trim(&f);
 200         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 201                 strbuf_remove(&f, 0, 1);
 202                 strbuf_setlen(&f, f.len - 1);
 203         }
 204
 205         get_sane_name(&mi->name, &f, &mi->email);
 206 out:
 207         strbuf_release(&f);
 208 }
 209
 210 static void handle_header(struct strbuf **out, const struct strbuf *line)
 211 {
 212         if (!*out) {
 213                 *out = xmalloc(sizeof(struct strbuf));
 214                 strbuf_init(*out, line->len);
 215         } else
 216                 strbuf_reset(*out);
 217
 218         strbuf_addbuf(*out, line);
 219 }
 220
 221 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 222  * to have enough heuristics to grok MIME encoded patches often found
 223  * on our mailing lists.  For example, we do not even treat header lines
 224  * case insensitively.
 225  */
 226
 227 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 228 {
 229         const char *ends, *ap = strcasestr(line, name);
 230         size_t sz;
 231
 232         strbuf_setlen(attr, 0);
 233         if (!ap)
 234                 return 0;
 235         ap += strlen(name);
 236         if (*ap == '"') {
 237                 ap++;
 238                 ends = "\"";
 239         }
 240         else
 241                 ends = "; \t";
 242         sz = strcspn(ap, ends);
 243         strbuf_add(attr, ap, sz);
 244         return 1;
 245 }
 246
 247 static int has_attr_value(const char *line, const char *name, const char *value)
 248 {
 249         struct strbuf sb = STRBUF_INIT;
 250         int rc = slurp_attr(line, name, &sb) && !strcasecmp(sb.buf, value);
 251         strbuf_release(&sb);
 252         return rc;
 253 }
 254
 255 static void handle_content_type(struct mailinfo *mi, struct strbuf *line)
 256 {
 257         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 258         strbuf_init(boundary, line->len);
 259
 260         mi->format_flowed = has_attr_value(line->buf, "format=", "flowed");
 261         mi->delsp = has_attr_value(line->buf, "delsp=", "yes");
 262
 263         if (slurp_attr(line->buf, "boundary=", boundary)) {
 264                 strbuf_insertstr(boundary, 0, "--");
 265                 if (++mi->content_top >= &mi->content[MAX_BOUNDARIES]) {
 266                         error("Too many boundaries to handle");
 267                         mi->input_error = -1;
 268                         mi->content_top = &mi->content[MAX_BOUNDARIES] - 1;
 269                         strbuf_release(boundary);
 270                         free(boundary);
 271                         return;
 272                 }
 273                 *(mi->content_top) = boundary;
 274                 boundary = NULL;
 275         }
 276         slurp_attr(line->buf, "charset=", &mi->charset);
 277
 278         if (boundary) {
 279                 strbuf_release(boundary);
 280                 free(boundary);
 281         }
 282 }
 283
 284 static void handle_content_transfer_encoding(struct mailinfo *mi,
 285                                              const struct strbuf *line)
 286 {
 287         if (strcasestr(line->buf, "base64"))
 288                 mi->transfer_encoding = TE_BASE64;
 289         else if (strcasestr(line->buf, "quoted-printable"))
 290                 mi->transfer_encoding = TE_QP;
 291         else
 292                 mi->transfer_encoding = TE_DONTCARE;
 293 }
 294
 295 static int is_multipart_boundary(struct mailinfo *mi, const struct strbuf *line)
 296 {
 297         struct strbuf *content_top = *(mi->content_top);
 298
 299         return ((content_top->len <= line->len) &&
 300                 !memcmp(line->buf, content_top->buf, content_top->len));
 301 }
 302
 303 static void cleanup_subject(struct mailinfo *mi, struct strbuf *subject)
 304 {
 305         size_t at = 0;
 306
 307         while (at < subject->len) {
 308                 char *pos;
 309                 size_t remove;
 310
 311                 switch (subject->buf[at]) {
 312                 case 'r': case 'R':
 313                         if (subject->len <= at + 3)
 314                                 break;
 315                         if ((subject->buf[at + 1] == 'e' ||
 316                              subject->buf[at + 1] == 'E') &&
 317                             subject->buf[at + 2] == ':') {
 318                                 strbuf_remove(subject, at, 3);
 319                                 continue;
 320                         }
 321                         at++;
 322                         break;
 323                 case ' ': case '\t': case ':':
 324                         strbuf_remove(subject, at, 1);
 325                         continue;
 326                 case '[':
 327                         pos = strchr(subject->buf + at, ']');
 328                         if (!pos)
 329                                 break;
 330                         remove = pos - (subject->buf + at) + 1;
 331                         if (!mi->keep_non_patch_brackets_in_subject ||
 332                             (7 <= remove &&
 333                              memmem(subject->buf + at, remove, "PATCH", 5)))
 334                                 strbuf_remove(subject, at, remove);
 335                         else {
 336                                 at += remove;
 337                                 /*
 338                                  * If the input had a space after the ], keep
 339                                  * it.  We don't bother with finding the end of
 340                                  * the space, since we later normalize it
 341                                  * anyway.
 342                                  */
 343                                 if (isspace(subject->buf[at]))
 344                                         at += 1;
 345                         }
 346                         continue;
 347                 }
 348                 break;
 349         }
 350         strbuf_trim(subject);
 351 }
 352
 353 static const char * const header[] = {
 354         "From", "Subject", "Date",
 355 };
 356
 357 static inline int skip_header(const struct strbuf *line, const char *hdr,
 358                               const char **outval)
 359 {
 360         const char *val;
 361         if (!skip_iprefix(line->buf, hdr, &val) ||
 362             *val++ != ':')
 363                 return 0;
 364         while (isspace(*val))
 365                 val++;
 366         *outval = val;
 367         return 1;
 368 }
 369
 370 static int is_format_patch_separator(const char *line, int len)
 371 {
 372         static const char SAMPLE[] =
 373                 "From e6807f3efca28b30decfecb1732a56c7db1137ee Mon Sep 17 00:00:00 2001\n";
 374         const char *cp;
 375
 376         if (len != strlen(SAMPLE))
 377                 return 0;
 378         if (!skip_prefix(line, "From ", &cp))
 379                 return 0;
 380         if (strspn(cp, "0123456789abcdef") != 40)
 381                 return 0;
 382         cp += 40;
 383         return !memcmp(SAMPLE + (cp - line), cp, strlen(SAMPLE) - (cp - line));
 384 }
 385
 386 static int decode_q_segment(struct strbuf *out, const struct strbuf *q_seg,
 387                             int rfc2047)
 388 {
 389         const char *in = q_seg->buf;
 390         int c;
 391         strbuf_grow(out, q_seg->len);
 392
 393         while ((c = *in++) != 0) {
 394                 if (c == '=') {
 395                         int ch, d = *in;
 396                         if (d == '\n' || !d)
 397                                 break; /* drop trailing newline */
 398                         ch = hex2chr(in);
 399                         if (ch >= 0) {
 400                                 strbuf_addch(out, ch);
 401                                 in += 2;
 402                                 continue;
 403                         }
 404                         /* garbage -- fall through */
 405                 }
 406                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 407                         c = 0x20;
 408                 strbuf_addch(out, c);
 409         }
 410         return 0;
 411 }
 412
 413 static int decode_b_segment(struct strbuf *out, const struct strbuf *b_seg)
 414 {
 415         /* Decode in..ep, possibly in-place to ot */
 416         int c, pos = 0, acc = 0;
 417         const char *in = b_seg->buf;
 418         strbuf_grow(out, b_seg->len);
 419
 420         while ((c = *in++) != 0) {
 421                 if (c == '+')
 422                         c = 62;
 423                 else if (c == '/')
 424                         c = 63;
 425                 else if ('A' <= c && c <= 'Z')
 426                         c -= 'A';
 427                 else if ('a' <= c && c <= 'z')
 428                         c -= 'a' - 26;
 429                 else if ('0' <= c && c <= '9')
 430                         c -= '0' - 52;
 431                 else
 432                         continue; /* garbage */
 433                 switch (pos++) {
 434                 case 0:
 435                         acc = (c << 2);
 436                         break;
 437                 case 1:
 438                         strbuf_addch(out, (acc | (c >> 4)));
 439                         acc = (c & 15) << 4;
 440                         break;
 441                 case 2:
 442                         strbuf_addch(out, (acc | (c >> 2)));
 443                         acc = (c & 3) << 6;
 444                         break;
 445                 case 3:
 446                         strbuf_addch(out, (acc | c));
 447                         acc = pos = 0;
 448                         break;
 449                 }
 450         }
 451         return 0;
 452 }
 453
 454 static int convert_to_utf8(struct mailinfo *mi,
 455                            struct strbuf *line, const char *charset)
 456 {
 457         char *out;
 458         size_t out_len;
 459
 460         if (!mi->metainfo_charset || !charset || !*charset)
 461                 return 0;
 462
 463         if (same_encoding(mi->metainfo_charset, charset))
 464                 return 0;
 465         out = reencode_string_len(line->buf, line->len,
 466                                   mi->metainfo_charset, charset, &out_len);
 467         if (!out) {
 468                 mi->input_error = -1;
 469                 return error("cannot convert from %s to %s",
 470                              charset, mi->metainfo_charset);
 471         }
 472         strbuf_attach(line, out, out_len, out_len);
 473         return 0;
 474 }
 475
 476 static void decode_header(struct mailinfo *mi, struct strbuf *it)
 477 {
 478         char *in, *ep, *cp;
 479         struct strbuf outbuf = STRBUF_INIT, dec = STRBUF_INIT;
 480         struct strbuf charset_q = STRBUF_INIT, piecebuf = STRBUF_INIT;
 481         int found_error = 1; /* pessimism */
 482
 483         in = it->buf;
 484         while (in - it->buf <= it->len && (ep = strstr(in, "=?")) != NULL) {
 485                 int encoding;
 486                 strbuf_reset(&charset_q);
 487                 strbuf_reset(&piecebuf);
 488
 489                 if (in != ep) {
 490                         /*
 491                          * We are about to process an encoded-word
 492                          * that begins at ep, but there is something
 493                          * before the encoded word.
 494                          */
 495                         char *scan;
 496                         for (scan = in; scan < ep; scan++)
 497                                 if (!isspace(*scan))
 498                                         break;
 499
 500                         if (scan != ep || in == it->buf) {
 501                                 /*
 502                                  * We should not lose that "something",
 503                                  * unless we have just processed an
 504                                  * encoded-word, and there is only LWS
 505                                  * before the one we are about to process.
 506                                  */
 507                                 strbuf_add(&outbuf, in, ep - in);
 508                         }
 509                 }
 510                 /* E.g.
 511                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 512                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 513                  */
 514                 ep += 2;
 515
 516                 if (ep - it->buf >= it->len || !(cp = strchr(ep, '?')))
 517                         goto release_return;
 518
 519                 if (cp + 3 - it->buf > it->len)
 520                         goto release_return;
 521                 strbuf_add(&charset_q, ep, cp - ep);
 522
 523                 encoding = cp[1];
 524                 if (!encoding || cp[2] != '?')
 525                         goto release_return;
 526                 ep = strstr(cp + 3, "?=");
 527                 if (!ep)
 528                         goto release_return;
 529                 strbuf_add(&piecebuf, cp + 3, ep - cp - 3);
 530                 switch (tolower(encoding)) {
 531                 default:
 532                         goto release_return;
 533                 case 'b':
 534                         if ((found_error = decode_b_segment(&dec, &piecebuf)))
 535                                 goto release_return;
 536                         break;
 537                 case 'q':
 538                         if ((found_error = decode_q_segment(&dec, &piecebuf, 1)))
 539                                 goto release_return;
 540                         break;
 541                 }
 542                 if (convert_to_utf8(mi, &dec, charset_q.buf))
 543                         goto release_return;
 544
 545                 strbuf_addbuf(&outbuf, &dec);
 546                 strbuf_release(&dec);
 547                 in = ep + 2;
 548         }
 549         strbuf_addstr(&outbuf, in);
 550         strbuf_reset(it);
 551         strbuf_addbuf(it, &outbuf);
 552         found_error = 0;
 553 release_return:
 554         strbuf_release(&outbuf);
 555         strbuf_release(&charset_q);
 556         strbuf_release(&piecebuf);
 557         strbuf_release(&dec);
 558
 559         if (found_error)
 560                 mi->input_error = -1;
 561 }
 562
 563 /*
 564  * Returns true if "line" contains a header matching "hdr", in which case "val"
 565  * will contain the value of the header with any RFC2047 B and Q encoding
 566  * unwrapped, and optionally normalize the meta information to utf8.
 567  */
 568 static int parse_header(const struct strbuf *line,
 569                         const char *hdr,
 570                         struct mailinfo *mi,
 571                         struct strbuf *val)
 572 {
 573         const char *val_str;
 574
 575         if (!skip_header(line, hdr, &val_str))
 576                 return 0;
 577         strbuf_addstr(val, val_str);
 578         decode_header(mi, val);
 579         return 1;
 580 }
 581
 582 static int check_header(struct mailinfo *mi,
 583                         const struct strbuf *line,
 584                         struct strbuf *hdr_data[], int overwrite)
 585 {
 586         int i, ret = 0;
 587         struct strbuf sb = STRBUF_INIT;
 588
 589         /* search for the interesting parts */
 590         for (i = 0; i < ARRAY_SIZE(header); i++) {
 591                 if ((!hdr_data[i] || overwrite) &&
 592                     parse_header(line, header[i], mi, &sb)) {
 593                         handle_header(&hdr_data[i], &sb);
 594                         ret = 1;
 595                         goto check_header_out;
 596                 }
 597         }
 598
 599         /* Content stuff */
 600         if (parse_header(line, "Content-Type", mi, &sb)) {
 601                 handle_content_type(mi, &sb);
 602                 ret = 1;
 603                 goto check_header_out;
 604         }
 605         if (parse_header(line, "Content-Transfer-Encoding", mi, &sb)) {
 606                 handle_content_transfer_encoding(mi, &sb);
 607                 ret = 1;
 608                 goto check_header_out;
 609         }
 610         if (parse_header(line, "Message-ID", mi, &sb)) {
 611                 if (mi->add_message_id)
 612                         mi->message_id = strbuf_detach(&sb, NULL);
 613                 ret = 1;
 614                 goto check_header_out;
 615         }
 616
 617 check_header_out:
 618         strbuf_release(&sb);
 619         return ret;
 620 }
 621
 622 /*
 623  * Returns 1 if the given line or any line beginning with the given line is an
 624  * in-body header (that is, check_header will succeed when passed
 625  * mi->s_hdr_data).
 626  */
 627 static int is_inbody_header(const struct mailinfo *mi,
 628                             const struct strbuf *line)
 629 {
 630         int i;
 631         const char *val;
 632         for (i = 0; i < ARRAY_SIZE(header); i++)
 633                 if (!mi->s_hdr_data[i] && skip_header(line, header[i], &val))
 634                         return 1;
 635         return 0;
 636 }
 637
 638 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 639 {
 640         struct strbuf ret = STRBUF_INIT;
 641
 642         switch (mi->transfer_encoding) {
 643         case TE_QP:
 644                 decode_q_segment(&ret, line, 0);
 645                 break;
 646         case TE_BASE64:
 647                 decode_b_segment(&ret, line);
 648                 break;
 649         case TE_DONTCARE:
 650         default:
 651                 return;
 652         }
 653         strbuf_reset(line);
 654         strbuf_addbuf(line, &ret);
 655         strbuf_release(&ret);
 656 }
 657
 658 static inline int patchbreak(const struct strbuf *line)
 659 {
 660         size_t i;
 661
 662         /* Beginning of a "diff -" header? */
 663         if (starts_with(line->buf, "diff -"))
 664                 return 1;
 665
 666         /* CVS "Index: " line? */
 667         if (starts_with(line->buf, "Index: "))
 668                 return 1;
 669
 670         /*
 671          * "--- <filename>" starts patches without headers
 672          * "---<sp>*" is a manual separator
 673          */
 674         if (line->len < 4)
 675                 return 0;
 676
 677         if (starts_with(line->buf, "---")) {
 678                 /* space followed by a filename? */
 679                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 680                         return 1;
 681                 /* Just whitespace? */
 682                 for (i = 3; i < line->len; i++) {
 683                         unsigned char c = line->buf[i];
 684                         if (c == '\n')
 685                                 return 1;
 686                         if (!isspace(c))
 687                                 break;
 688                 }
 689                 return 0;
 690         }
 691         return 0;
 692 }
 693
 694 static int is_scissors_line(const char *line)
 695 {
 696         const char *c;
 697         int scissors = 0, gap = 0;
 698         const char *first_nonblank = NULL, *last_nonblank = NULL;
 699         int visible, perforation = 0, in_perforation = 0;
 700
 701         for (c = line; *c; c++) {
 702                 if (isspace(*c)) {
 703                         if (in_perforation) {
 704                                 perforation++;
 705                                 gap++;
 706                         }
 707                         continue;
 708                 }
 709                 last_nonblank = c;
 710                 if (!first_nonblank)
 711                         first_nonblank = c;
 712                 if (*c == '-') {
 713                         in_perforation = 1;
 714                         perforation++;
 715                         continue;
 716                 }
 717                 if (starts_with(c, ">8") || starts_with(c, "8<") ||
 718                     starts_with(c, ">%") || starts_with(c, "%<")) {
 719                         in_perforation = 1;
 720                         perforation += 2;
 721                         scissors += 2;
 722                         c++;
 723                         continue;
 724                 }
 725                 in_perforation = 0;
 726         }
 727
 728         /*
 729          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 730          * Even though there can be arbitrary cruft on the same line
 731          * (e.g. "cut here"), in order to avoid misidentification, the
 732          * perforation must occupy more than a third of the visible
 733          * width of the line, and dashes and scissors must occupy more
 734          * than half of the perforation.
 735          */
 736
 737         if (first_nonblank && last_nonblank)
 738                 visible = last_nonblank - first_nonblank + 1;
 739         else
 740                 visible = 0;
 741         return (scissors && 8 <= visible &&
 742                 visible < perforation * 3 &&
 743                 gap * 2 < perforation);
 744 }
 745
 746 static void flush_inbody_header_accum(struct mailinfo *mi)
 747 {
 748         if (!mi->inbody_header_accum.len)
 749                 return;
 750         if (!check_header(mi, &mi->inbody_header_accum, mi->s_hdr_data, 0))
 751                 BUG("inbody_header_accum, if not empty, must always contain a valid in-body header");
 752         strbuf_reset(&mi->inbody_header_accum);
 753 }
 754
 755 static int check_inbody_header(struct mailinfo *mi, const struct strbuf *line)
 756 {
 757         if (mi->inbody_header_accum.len &&
 758             (line->buf[0] == ' ' || line->buf[0] == '\t')) {
 759                 if (mi->use_scissors && is_scissors_line(line->buf)) {
 760                         /*
 761                          * This is a scissors line; do not consider this line
 762                          * as a header continuation line.
 763                          */
 764                         flush_inbody_header_accum(mi);
 765                         return 0;
 766                 }
 767                 strbuf_strip_suffix(&mi->inbody_header_accum, "\n");
 768                 strbuf_addbuf(&mi->inbody_header_accum, line);
 769                 return 1;
 770         }
 771
 772         flush_inbody_header_accum(mi);
 773
 774         if (starts_with(line->buf, ">From") && isspace(line->buf[5]))
 775                 return is_format_patch_separator(line->buf + 1, line->len - 1);
 776         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 777                 int i;
 778                 for (i = 0; i < ARRAY_SIZE(header); i++)
 779                         if (!strcmp("Subject", header[i])) {
 780                                 handle_header(&mi->s_hdr_data[i], line);
 781                                 return 1;
 782                         }
 783                 return 0;
 784         }
 785         if (is_inbody_header(mi, line)) {
 786                 strbuf_addbuf(&mi->inbody_header_accum, line);
 787                 return 1;
 788         }
 789         return 0;
 790 }
 791
 792 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 793 {
 794         assert(!mi->filter_stage);
 795
 796         if (mi->header_stage) {
 797                 if (!line->len || (line->len == 1 && line->buf[0] == '\n')) {
 798                         if (mi->inbody_header_accum.len) {
 799                                 flush_inbody_header_accum(mi);
 800                                 mi->header_stage = 0;
 801                         }
 802                         return 0;
 803                 }
 804         }
 805
 806         if (mi->use_inbody_headers && mi->header_stage) {
 807                 mi->header_stage = check_inbody_header(mi, line);
 808                 if (mi->header_stage)
 809                         return 0;
 810         } else
 811                 /* Only trim the first (blank) line of the commit message
 812                  * when ignoring in-body headers.
 813                  */
 814                 mi->header_stage = 0;
 815
 816         /* normalize the log message to UTF-8. */
 817         if (convert_to_utf8(mi, line, mi->charset.buf))
 818                 return 0; /* mi->input_error already set */
 819
 820         if (mi->use_scissors && is_scissors_line(line->buf)) {
 821                 int i;
 822
 823                 strbuf_setlen(&mi->log_message, 0);
 824                 mi->header_stage = 1;
 825
 826                 /*
 827                  * We may have already read "secondary headers"; purge
 828                  * them to give ourselves a clean restart.
 829                  */
 830                 for (i = 0; i < ARRAY_SIZE(header); i++) {
 831                         if (mi->s_hdr_data[i])
 832                                 strbuf_release(mi->s_hdr_data[i]);
 833                         FREE_AND_NULL(mi->s_hdr_data[i]);
 834                 }
 835                 return 0;
 836         }
 837
 838         if (patchbreak(line)) {
 839                 if (mi->message_id)
 840                         strbuf_addf(&mi->log_message,
 841                                     "Message-ID: %s\n", mi->message_id);
 842                 return 1;
 843         }
 844
 845         strbuf_addbuf(&mi->log_message, line);
 846         return 0;
 847 }
 848
 849 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 850 {
 851         fwrite(line->buf, 1, line->len, mi->patchfile);
 852         mi->patch_lines++;
 853 }
 854
 855 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 856 {
 857         switch (mi->filter_stage) {
 858         case 0:
 859                 if (!handle_commit_msg(mi, line))
 860                         break;
 861                 mi->filter_stage++;
 862                 /* fallthrough */
 863         case 1:
 864                 handle_patch(mi, line);
 865                 break;
 866         }
 867 }
 868
 869 static int is_rfc2822_header(const struct strbuf *line)
 870 {
 871         /*
 872          * The section that defines the loosest possible
 873          * field name is "3.6.8 Optional fields".
 874          *
 875          * optional-field = field-name ":" unstructured CRLF
 876          * field-name = 1*ftext
 877          * ftext = %d33-57 / %59-126
 878          */
 879         int ch;
 880         char *cp = line->buf;
 881
 882         /* Count mbox From headers as headers */
 883         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 884                 return 1;
 885
 886         while ((ch = *cp++)) {
 887                 if (ch == ':')
 888                         return 1;
 889                 if ((33 <= ch && ch <= 57) ||
 890                     (59 <= ch && ch <= 126))
 891                         continue;
 892                 break;
 893         }
 894         return 0;
 895 }
 896
 897 static int read_one_header_line(struct strbuf *line, FILE *in)
 898 {
 899         struct strbuf continuation = STRBUF_INIT;
 900
 901         /* Get the first part of the line. */
 902         if (strbuf_getline_lf(line, in))
 903                 return 0;
 904
 905         /*
 906          * Is it an empty line or not a valid rfc2822 header?
 907          * If so, stop here, and return false ("not a header")
 908          */
 909         strbuf_rtrim(line);
 910         if (!line->len || !is_rfc2822_header(line)) {
 911                 /* Re-add the newline */
 912                 strbuf_addch(line, '\n');
 913                 return 0;
 914         }
 915
 916         /*
 917          * Now we need to eat all the continuation lines..
 918          * Yuck, 2822 header "folding"
 919          */
 920         for (;;) {
 921                 int peek;
 922
 923                 peek = fgetc(in);
 924                 if (peek == EOF)
 925                         break;
 926                 ungetc(peek, in);
 927                 if (peek != ' ' && peek != '\t')
 928                         break;
 929                 if (strbuf_getline_lf(&continuation, in))
 930                         break;
 931                 continuation.buf[0] = ' ';
 932                 strbuf_rtrim(&continuation);
 933                 strbuf_addbuf(line, &continuation);
 934         }
 935         strbuf_release(&continuation);
 936
 937         return 1;
 938 }
 939
 940 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 941 {
 942         while (!strbuf_getline_lf(line, mi->input)) {
 943                 if (*(mi->content_top) && is_multipart_boundary(mi, line))
 944                         return 1;
 945         }
 946         return 0;
 947 }
 948
 949 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 950 {
 951         struct strbuf newline = STRBUF_INIT;
 952
 953         strbuf_addch(&newline, '\n');
 954 again:
 955         if (line->len >= (*(mi->content_top))->len + 2 &&
 956             !memcmp(line->buf + (*(mi->content_top))->len, "--", 2)) {
 957                 /* we hit an end boundary */
 958                 /* pop the current boundary off the stack */
 959                 strbuf_release(*(mi->content_top));
 960                 FREE_AND_NULL(*(mi->content_top));
 961
 962                 /* technically won't happen as is_multipart_boundary()
 963                    will fail first.  But just in case..
 964                  */
 965                 if (--mi->content_top < mi->content) {
 966                         error("Detected mismatched boundaries, can't recover");
 967                         mi->input_error = -1;
 968                         mi->content_top = mi->content;
 969                         strbuf_release(&newline);
 970                         return 0;
 971                 }
 972                 handle_filter(mi, &newline);
 973                 strbuf_release(&newline);
 974                 if (mi->input_error)
 975                         return 0;
 976
 977                 /* skip to the next boundary */
 978                 if (!find_boundary(mi, line))
 979                         return 0;
 980                 goto again;
 981         }
 982
 983         /* set some defaults */
 984         mi->transfer_encoding = TE_DONTCARE;
 985         strbuf_reset(&mi->charset);
 986
 987         /* slurp in this section's info */
 988         while (read_one_header_line(line, mi->input))
 989                 check_header(mi, line, mi->p_hdr_data, 0);
 990
 991         strbuf_release(&newline);
 992         /* replenish line */
 993         if (strbuf_getline_lf(line, mi->input))
 994                 return 0;
 995         strbuf_addch(line, '\n');
 996         return 1;
 997 }
 998
 999 static void handle_filter_flowed(struct mailinfo *mi, struct strbuf *line,
1000                                  struct strbuf *prev)
1001 {
1002         size_t len = line->len;
1003         const char *rest;
1004
1005         if (!mi->format_flowed) {
1006                 if (len >= 2 &&
1007                     line->buf[len - 2] == '\r' &&
1008                     line->buf[len - 1] == '\n') {
1009                         mi->have_quoted_cr = 1;
1010                         if (mi->quoted_cr == quoted_cr_strip) {
1011                                 strbuf_setlen(line, len - 2);
1012                                 strbuf_addch(line, '\n');
1013                                 len--;
1014                         }
1015                 }
1016                 handle_filter(mi, line);
1017                 return;
1018         }
1019
1020         if (line->buf[len - 1] == '\n') {
1021                 len--;
1022                 if (len && line->buf[len - 1] == '\r')
1023                         len--;
1024         }
1025
1026         /* Keep signature separator as-is. */
1027         if (skip_prefix(line->buf, "-- ", &rest) && rest - line->buf == len) {
1028                 if (prev->len) {
1029                         handle_filter(mi, prev);
1030                         strbuf_reset(prev);
1031                 }
1032                 handle_filter(mi, line);
1033                 return;
1034         }
1035
1036         /* Unstuff space-stuffed line. */
1037         if (len && line->buf[0] == ' ') {
1038                 strbuf_remove(line, 0, 1);
1039                 len--;
1040         }
1041
1042         /* Save flowed line for later, but without the soft line break. */
1043         if (len && line->buf[len - 1] == ' ') {
1044                 strbuf_add(prev, line->buf, len - !!mi->delsp);
1045                 return;
1046         }
1047
1048         /* Prepend any previous partial lines */
1049         strbuf_insert(line, 0, prev->buf, prev->len);
1050         strbuf_reset(prev);
1051
1052         handle_filter(mi, line);
1053 }
1054
1055 static void summarize_quoted_cr(struct mailinfo *mi)
1056 {
1057         if (mi->have_quoted_cr &&
1058             mi->quoted_cr == quoted_cr_warn)
1059                 warning(_("quoted CRLF detected"));
1060 }
1061
1062 static void handle_body(struct mailinfo *mi, struct strbuf *line)
1063 {
1064         struct strbuf prev = STRBUF_INIT;
1065
1066         /* Skip up to the first boundary */
1067         if (*(mi->content_top)) {
1068                 if (!find_boundary(mi, line))
1069                         goto handle_body_out;
1070         }
1071
1072         do {
1073                 /* process any boundary lines */
1074                 if (*(mi->content_top) && is_multipart_boundary(mi, line)) {
1075                         /* flush any leftover */
1076                         if (prev.len) {
1077                                 handle_filter(mi, &prev);
1078                                 strbuf_reset(&prev);
1079                         }
1080                         summarize_quoted_cr(mi);
1081                         mi->have_quoted_cr = 0;
1082                         if (!handle_boundary(mi, line))
1083                                 goto handle_body_out;
1084                 }
1085
1086                 /* Unwrap transfer encoding */
1087                 decode_transfer_encoding(mi, line);
1088
1089                 switch (mi->transfer_encoding) {
1090                 case TE_BASE64:
1091                 case TE_QP:
1092                 {
1093                         struct strbuf **lines, **it, *sb;
1094
1095                         /* Prepend any previous partial lines */
1096                         strbuf_insert(line, 0, prev.buf, prev.len);
1097                         strbuf_reset(&prev);
1098
1099                         /*
1100                          * This is a decoded line that may contain
1101                          * multiple new lines.  Pass only one chunk
1102                          * at a time to handle_filter()
1103                          */
1104                         lines = strbuf_split(line, '\n');
1105                         for (it = lines; (sb = *it); it++) {
1106                                 if (!*(it + 1)) /* The last line */
1107                                         if (sb->buf[sb->len - 1] != '\n') {
1108                                                 /* Partial line, save it for later. */
1109                                                 strbuf_addbuf(&prev, sb);
1110                                                 break;
1111                                         }
1112                                 handle_filter_flowed(mi, sb, &prev);
1113                         }
1114                         /*
1115                          * The partial chunk is saved in "prev" and will be
1116                          * appended by the next iteration of read_line_with_nul().
1117                          */
1118                         strbuf_list_free(lines);
1119                         break;
1120                 }
1121                 default:
1122                         handle_filter_flowed(mi, line, &prev);
1123                 }
1124
1125                 if (mi->input_error)
1126                         break;
1127         } while (!strbuf_getwholeline(line, mi->input, '\n'));
1128
1129         if (prev.len)
1130                 handle_filter(mi, &prev);
1131         summarize_quoted_cr(mi);
1132
1133         flush_inbody_header_accum(mi);
1134
1135 handle_body_out:
1136         strbuf_release(&prev);
1137 }
1138
1139 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
1140 {
1141         const char *sp = data->buf;
1142         while (1) {
1143                 char *ep = strchr(sp, '\n');
1144                 int len;
1145                 if (!ep)
1146                         len = strlen(sp);
1147                 else
1148                         len = ep - sp;
1149                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
1150                 if (!ep)
1151                         break;
1152                 sp = ep + 1;
1153         }
1154 }
1155
1156 static void handle_info(struct mailinfo *mi)
1157 {
1158         struct strbuf *hdr;
1159         int i;
1160
1161         for (i = 0; i < ARRAY_SIZE(header); i++) {
1162                 /* only print inbody headers if we output a patch file */
1163                 if (mi->patch_lines && mi->s_hdr_data[i])
1164                         hdr = mi->s_hdr_data[i];
1165                 else if (mi->p_hdr_data[i])
1166                         hdr = mi->p_hdr_data[i];
1167                 else
1168                         continue;
1169
1170                 if (memchr(hdr->buf, '\0', hdr->len)) {
1171                         error("a NUL byte in '%s' is not allowed.", header[i]);
1172                         mi->input_error = -1;
1173                 }
1174
1175                 if (!strcmp(header[i], "Subject")) {
1176                         if (!mi->keep_subject) {
1177                                 cleanup_subject(mi, hdr);
1178                                 cleanup_space(hdr);
1179                         }
1180                         output_header_lines(mi->output, "Subject", hdr);
1181                 } else if (!strcmp(header[i], "From")) {
1182                         cleanup_space(hdr);
1183                         handle_from(mi, hdr);
1184                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
1185                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
1186                 } else {
1187                         cleanup_space(hdr);
1188                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
1189                 }
1190         }
1191         fprintf(mi->output, "\n");
1192 }
1193
1194 int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
1195 {
1196         FILE *cmitmsg;
1197         int peek;
1198         struct strbuf line = STRBUF_INIT;
1199
1200         cmitmsg = fopen(msg, "w");
1201         if (!cmitmsg) {
1202                 perror(msg);
1203                 return -1;
1204         }
1205         mi->patchfile = fopen(patch, "w");
1206         if (!mi->patchfile) {
1207                 perror(patch);
1208                 fclose(cmitmsg);
1209                 return -1;
1210         }
1211
1212         mi->p_hdr_data = xcalloc(ARRAY_SIZE(header), sizeof(*(mi->p_hdr_data)));
1213         mi->s_hdr_data = xcalloc(ARRAY_SIZE(header), sizeof(*(mi->s_hdr_data)));
1214
1215         do {
1216                 peek = fgetc(mi->input);
1217                 if (peek == EOF) {
1218                         fclose(cmitmsg);
1219                         return error("empty patch: '%s'", patch);
1220                 }
1221         } while (isspace(peek));
1222         ungetc(peek, mi->input);
1223
1224         /* process the email header */
1225         while (read_one_header_line(&line, mi->input))
1226                 check_header(mi, &line, mi->p_hdr_data, 1);
1227
1228         handle_body(mi, &line);
1229         fwrite(mi->log_message.buf, 1, mi->log_message.len, cmitmsg);
1230         fclose(cmitmsg);
1231         fclose(mi->patchfile);
1232
1233         handle_info(mi);
1234         strbuf_release(&line);
1235         return mi->input_error;
1236 }
1237
1238 int mailinfo_parse_quoted_cr_action(const char *actionstr, int *action)
1239 {
1240         if (!strcmp(actionstr, "nowarn"))
1241                 *action = quoted_cr_nowarn;
1242         else if (!strcmp(actionstr, "warn"))
1243                 *action = quoted_cr_warn;
1244         else if (!strcmp(actionstr, "strip"))
1245                 *action = quoted_cr_strip;
1246         else
1247                 return -1;
1248         return 0;
1249 }
1250
1251 static int git_mailinfo_config(const char *var, const char *value,
1252                                const struct config_context *ctx, void *mi_)
1253 {
1254         struct mailinfo *mi = mi_;
1255
1256         if (!starts_with(var, "mailinfo."))
1257                 return git_default_config(var, value, ctx, NULL);
1258         if (!strcmp(var, "mailinfo.scissors")) {
1259                 mi->use_scissors = git_config_bool(var, value);
1260                 return 0;
1261         }
1262         if (!strcmp(var, "mailinfo.quotedcr")) {
1263                 if (!value)
1264                         return config_error_nonbool(var);
1265                 if (mailinfo_parse_quoted_cr_action(value, &mi->quoted_cr) != 0)
1266                         return error(_("bad action '%s' for '%s'"), value, var);
1267                 return 0;
1268         }
1269         /* perhaps others here */
1270         return 0;
1271 }
1272
1273 void setup_mailinfo(struct repository *r, struct mailinfo *mi)
1274 {
1275         memset(mi, 0, sizeof(*mi));
1276         strbuf_init(&mi->name, 0);
1277         strbuf_init(&mi->email, 0);
1278         strbuf_init(&mi->charset, 0);
1279         strbuf_init(&mi->log_message, 0);
1280         strbuf_init(&mi->inbody_header_accum, 0);
1281         mi->quoted_cr = quoted_cr_warn;
1282         mi->header_stage = 1;
1283         mi->use_inbody_headers = 1;
1284         mi->content_top = mi->content;
1285         repo_config(r, git_mailinfo_config, mi);
1286 }
1287
1288 void clear_mailinfo(struct mailinfo *mi)
1289 {
1290         strbuf_release(&mi->name);
1291         strbuf_release(&mi->email);
1292         strbuf_release(&mi->charset);
1293         strbuf_release(&mi->inbody_header_accum);
1294         free(mi->message_id);
1295
1296         for (size_t i = 0; i < ARRAY_SIZE(header); i++) {
1297                 if (!mi->p_hdr_data[i])
1298                         continue;
1299                 strbuf_release(mi->p_hdr_data[i]);
1300                 free(mi->p_hdr_data[i]);
1301         }
1302         free(mi->p_hdr_data);
1303
1304         for (size_t i = 0; i < ARRAY_SIZE(header); i++) {
1305                 if (!mi->s_hdr_data[i])
1306                         continue;
1307                 strbuf_release(mi->s_hdr_data[i]);
1308                 free(mi->s_hdr_data[i]);
1309         }
1310         free(mi->s_hdr_data);
1311
1312         while (mi->content < mi->content_top) {
1313                 free(*(mi->content_top));
1314                 mi->content_top--;
1315         }
1316
1317         strbuf_release(&mi->log_message);
1318 }