builtin/mailinfo.c

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