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