git.oblomov.eu Git - git/blob - 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 struct mailinfo {
  11         FILE *input;
  12         FILE *output;
  13         FILE *cmitmsg;
  14         FILE *patchfile;
  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         struct strbuf charset;
  26         char *message_id;
  27         enum  {
  28                 TE_DONTCARE, TE_QP, TE_BASE64
  29         } transfer_encoding;
  30         int patch_lines;
  31         int filter_stage; /* still reading log or are we copying patch? */
  32         int header_stage; /* still checking in-body headers? */
  33         struct strbuf **p_hdr_data;
  34         struct strbuf **s_hdr_data;
  35 };
  36
  37 #define MAX_BOUNDARIES 5
  38
  39 static void cleanup_space(struct strbuf *sb)
  40 {
  41         size_t pos, cnt;
  42         for (pos = 0; pos < sb->len; pos++) {
  43                 if (isspace(sb->buf[pos])) {
  44                         sb->buf[pos] = ' ';
  45                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
  46                         strbuf_remove(sb, pos + 1, cnt);
  47                 }
  48         }
  49 }
  50
  51 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  52 {
  53         struct strbuf *src = name;
  54         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  55                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  56                 src = email;
  57         else if (name == out)
  58                 return;
  59         strbuf_reset(out);
  60         strbuf_addbuf(out, src);
  61 }
  62
  63 static void parse_bogus_from(struct mailinfo *mi, const struct strbuf *line)
  64 {
  65         /* John Doe <johndoe> */
  66
  67         char *bra, *ket;
  68         /* This is fallback, so do not bother if we already have an
  69          * e-mail address.
  70          */
  71         if (mi->email.len)
  72                 return;
  73
  74         bra = strchr(line->buf, '<');
  75         if (!bra)
  76                 return;
  77         ket = strchr(bra, '>');
  78         if (!ket)
  79                 return;
  80
  81         strbuf_reset(&mi->email);
  82         strbuf_add(&mi->email, bra + 1, ket - bra - 1);
  83
  84         strbuf_reset(&mi->name);
  85         strbuf_add(&mi->name, line->buf, bra - line->buf);
  86         strbuf_trim(&mi->name);
  87         get_sane_name(&mi->name, &mi->name, &mi->email);
  88 }
  89
  90 static void handle_from(struct mailinfo *mi, const struct strbuf *from)
  91 {
  92         char *at;
  93         size_t el;
  94         struct strbuf f;
  95
  96         strbuf_init(&f, from->len);
  97         strbuf_addbuf(&f, from);
  98
  99         at = strchr(f.buf, '@');
 100         if (!at) {
 101                 parse_bogus_from(mi, from);
 102                 return;
 103         }
 104
 105         /*
 106          * If we already have one email, don't take any confusing lines
 107          */
 108         if (mi->email.len && strchr(at + 1, '@')) {
 109                 strbuf_release(&f);
 110                 return;
 111         }
 112
 113         /* Pick up the string around '@', possibly delimited with <>
 114          * pair; that is the email part.
 115          */
 116         while (at > f.buf) {
 117                 char c = at[-1];
 118                 if (isspace(c))
 119                         break;
 120                 if (c == '<') {
 121                         at[-1] = ' ';
 122                         break;
 123                 }
 124                 at--;
 125         }
 126         el = strcspn(at, " \n\t\r\v\f>");
 127         strbuf_reset(&mi->email);
 128         strbuf_add(&mi->email, at, el);
 129         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 130
 131         /* The remainder is name.  It could be
 132          *
 133          * - "John Doe <john.doe@xz>"                   (a), or
 134          * - "john.doe@xz (John Doe)"                   (b), or
 135          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 136          *
 137          * but we have removed the email part, so
 138          *
 139          * - remove extra spaces which could stay after email (case 'c'), and
 140          * - trim from both ends, possibly removing the () pair at the end
 141          *   (cases 'a' and 'b').
 142          */
 143         cleanup_space(&f);
 144         strbuf_trim(&f);
 145         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 146                 strbuf_remove(&f, 0, 1);
 147                 strbuf_setlen(&f, f.len - 1);
 148         }
 149
 150         get_sane_name(&mi->name, &f, &mi->email);
 151         strbuf_release(&f);
 152 }
 153
 154 static void handle_header(struct strbuf **out, const struct strbuf *line)
 155 {
 156         if (!*out) {
 157                 *out = xmalloc(sizeof(struct strbuf));
 158                 strbuf_init(*out, line->len);
 159         } else
 160                 strbuf_reset(*out);
 161
 162         strbuf_addbuf(*out, line);
 163 }
 164
 165 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 166  * to have enough heuristics to grok MIME encoded patches often found
 167  * on our mailing lists.  For example, we do not even treat header lines
 168  * case insensitively.
 169  */
 170
 171 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 172 {
 173         const char *ends, *ap = strcasestr(line, name);
 174         size_t sz;
 175
 176         strbuf_setlen(attr, 0);
 177         if (!ap)
 178                 return 0;
 179         ap += strlen(name);
 180         if (*ap == '"') {
 181                 ap++;
 182                 ends = "\"";
 183         }
 184         else
 185                 ends = "; \t";
 186         sz = strcspn(ap, ends);
 187         strbuf_add(attr, ap, sz);
 188         return 1;
 189 }
 190
 191 static struct strbuf *content[MAX_BOUNDARIES];
 192
 193 static struct strbuf **content_top = content;
 194
 195 static void handle_content_type(struct mailinfo *mi, struct strbuf *line)
 196 {
 197         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 198         strbuf_init(boundary, line->len);
 199
 200         if (slurp_attr(line->buf, "boundary=", boundary)) {
 201                 strbuf_insert(boundary, 0, "--", 2);
 202                 if (++content_top >= &content[MAX_BOUNDARIES]) {
 203                         fprintf(stderr, "Too many boundaries to handle\n");
 204                         exit(1);
 205                 }
 206                 *content_top = boundary;
 207                 boundary = NULL;
 208         }
 209         slurp_attr(line->buf, "charset=", &mi->charset);
 210
 211         if (boundary) {
 212                 strbuf_release(boundary);
 213                 free(boundary);
 214         }
 215 }
 216
 217 static void handle_message_id(struct mailinfo *mi, const struct strbuf *line)
 218 {
 219         if (mi->add_message_id)
 220                 mi->message_id = strdup(line->buf);
 221 }
 222
 223 static void handle_content_transfer_encoding(struct mailinfo *mi,
 224                                              const struct strbuf *line)
 225 {
 226         if (strcasestr(line->buf, "base64"))
 227                 mi->transfer_encoding = TE_BASE64;
 228         else if (strcasestr(line->buf, "quoted-printable"))
 229                 mi->transfer_encoding = TE_QP;
 230         else
 231                 mi->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 (!mi->metainfo_charset || !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                 convert_to_utf8(mi, dec, charset_q.buf);
 463
 464                 strbuf_addbuf(&outbuf, dec);
 465                 strbuf_release(dec);
 466                 free(dec);
 467                 in = ep + 2;
 468         }
 469         strbuf_addstr(&outbuf, in);
 470         strbuf_reset(it);
 471         strbuf_addbuf(it, &outbuf);
 472 release_return:
 473         strbuf_release(&outbuf);
 474         strbuf_release(&charset_q);
 475         strbuf_release(&piecebuf);
 476 }
 477
 478 static int check_header(struct mailinfo *mi,
 479                         const struct strbuf *line,
 480                         struct strbuf *hdr_data[], int overwrite)
 481 {
 482         int i, ret = 0, len;
 483         struct strbuf sb = STRBUF_INIT;
 484
 485         /* search for the interesting parts */
 486         for (i = 0; header[i]; i++) {
 487                 int len = strlen(header[i]);
 488                 if ((!hdr_data[i] || overwrite) && cmp_header(line, header[i])) {
 489                         /* Unwrap inline B and Q encoding, and optionally
 490                          * normalize the meta information to utf8.
 491                          */
 492                         strbuf_add(&sb, line->buf + len + 2, line->len - len - 2);
 493                         decode_header(mi, &sb);
 494                         handle_header(&hdr_data[i], &sb);
 495                         ret = 1;
 496                         goto check_header_out;
 497                 }
 498         }
 499
 500         /* Content stuff */
 501         if (cmp_header(line, "Content-Type")) {
 502                 len = strlen("Content-Type: ");
 503                 strbuf_add(&sb, line->buf + len, line->len - len);
 504                 decode_header(mi, &sb);
 505                 strbuf_insert(&sb, 0, "Content-Type: ", len);
 506                 handle_content_type(mi, &sb);
 507                 ret = 1;
 508                 goto check_header_out;
 509         }
 510         if (cmp_header(line, "Content-Transfer-Encoding")) {
 511                 len = strlen("Content-Transfer-Encoding: ");
 512                 strbuf_add(&sb, line->buf + len, line->len - len);
 513                 decode_header(mi, &sb);
 514                 handle_content_transfer_encoding(mi, &sb);
 515                 ret = 1;
 516                 goto check_header_out;
 517         }
 518         if (cmp_header(line, "Message-Id")) {
 519                 len = strlen("Message-Id: ");
 520                 strbuf_add(&sb, line->buf + len, line->len - len);
 521                 decode_header(mi, &sb);
 522                 handle_message_id(mi, &sb);
 523                 ret = 1;
 524                 goto check_header_out;
 525         }
 526
 527         /* for inbody stuff */
 528         if (starts_with(line->buf, ">From") && isspace(line->buf[5])) {
 529                 ret = is_format_patch_separator(line->buf + 1, line->len - 1);
 530                 goto check_header_out;
 531         }
 532         if (starts_with(line->buf, "[PATCH]") && isspace(line->buf[7])) {
 533                 for (i = 0; header[i]; i++) {
 534                         if (!strcmp("Subject", header[i])) {
 535                                 handle_header(&hdr_data[i], line);
 536                                 ret = 1;
 537                                 goto check_header_out;
 538                         }
 539                 }
 540         }
 541
 542 check_header_out:
 543         strbuf_release(&sb);
 544         return ret;
 545 }
 546
 547 static void decode_transfer_encoding(struct mailinfo *mi, struct strbuf *line)
 548 {
 549         struct strbuf *ret;
 550
 551         switch (mi->transfer_encoding) {
 552         case TE_QP:
 553                 ret = decode_q_segment(line, 0);
 554                 break;
 555         case TE_BASE64:
 556                 ret = decode_b_segment(line);
 557                 break;
 558         case TE_DONTCARE:
 559         default:
 560                 return;
 561         }
 562         strbuf_reset(line);
 563         strbuf_addbuf(line, ret);
 564         strbuf_release(ret);
 565         free(ret);
 566 }
 567
 568 static inline int patchbreak(const struct strbuf *line)
 569 {
 570         size_t i;
 571
 572         /* Beginning of a "diff -" header? */
 573         if (starts_with(line->buf, "diff -"))
 574                 return 1;
 575
 576         /* CVS "Index: " line? */
 577         if (starts_with(line->buf, "Index: "))
 578                 return 1;
 579
 580         /*
 581          * "--- <filename>" starts patches without headers
 582          * "---<sp>*" is a manual separator
 583          */
 584         if (line->len < 4)
 585                 return 0;
 586
 587         if (starts_with(line->buf, "---")) {
 588                 /* space followed by a filename? */
 589                 if (line->buf[3] == ' ' && !isspace(line->buf[4]))
 590                         return 1;
 591                 /* Just whitespace? */
 592                 for (i = 3; i < line->len; i++) {
 593                         unsigned char c = line->buf[i];
 594                         if (c == '\n')
 595                                 return 1;
 596                         if (!isspace(c))
 597                                 break;
 598                 }
 599                 return 0;
 600         }
 601         return 0;
 602 }
 603
 604 static int is_scissors_line(const struct strbuf *line)
 605 {
 606         size_t i, len = line->len;
 607         int scissors = 0, gap = 0;
 608         int first_nonblank = -1;
 609         int last_nonblank = 0, visible, perforation = 0, in_perforation = 0;
 610         const char *buf = line->buf;
 611
 612         for (i = 0; i < len; i++) {
 613                 if (isspace(buf[i])) {
 614                         if (in_perforation) {
 615                                 perforation++;
 616                                 gap++;
 617                         }
 618                         continue;
 619                 }
 620                 last_nonblank = i;
 621                 if (first_nonblank < 0)
 622                         first_nonblank = i;
 623                 if (buf[i] == '-') {
 624                         in_perforation = 1;
 625                         perforation++;
 626                         continue;
 627                 }
 628                 if (i + 1 < len &&
 629                     (!memcmp(buf + i, ">8", 2) || !memcmp(buf + i, "8<", 2) ||
 630                      !memcmp(buf + i, ">%", 2) || !memcmp(buf + i, "%<", 2))) {
 631                         in_perforation = 1;
 632                         perforation += 2;
 633                         scissors += 2;
 634                         i++;
 635                         continue;
 636                 }
 637                 in_perforation = 0;
 638         }
 639
 640         /*
 641          * The mark must be at least 8 bytes long (e.g. "-- >8 --").
 642          * Even though there can be arbitrary cruft on the same line
 643          * (e.g. "cut here"), in order to avoid misidentification, the
 644          * perforation must occupy more than a third of the visible
 645          * width of the line, and dashes and scissors must occupy more
 646          * than half of the perforation.
 647          */
 648
 649         visible = last_nonblank - first_nonblank + 1;
 650         return (scissors && 8 <= visible &&
 651                 visible < perforation * 3 &&
 652                 gap * 2 < perforation);
 653 }
 654
 655 static int handle_commit_msg(struct mailinfo *mi, struct strbuf *line)
 656 {
 657         if (!mi->cmitmsg)
 658                 return 0;
 659
 660         if (mi->header_stage) {
 661                 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
 662                         return 0;
 663         }
 664
 665         if (mi->use_inbody_headers && mi->header_stage) {
 666                 mi->header_stage = check_header(mi, line, mi->s_hdr_data, 0);
 667                 if (mi->header_stage)
 668                         return 0;
 669         } else
 670                 /* Only trim the first (blank) line of the commit message
 671                  * when ignoring in-body headers.
 672                  */
 673                 mi->header_stage = 0;
 674
 675         /* normalize the log message to UTF-8. */
 676         convert_to_utf8(mi, line, mi->charset.buf);
 677
 678         if (mi->use_scissors && is_scissors_line(line)) {
 679                 int i;
 680                 if (fseek(mi->cmitmsg, 0L, SEEK_SET))
 681                         die_errno("Could not rewind output message file");
 682                 if (ftruncate(fileno(mi->cmitmsg), 0))
 683                         die_errno("Could not truncate output message file at scissors");
 684                 mi->header_stage = 1;
 685
 686                 /*
 687                  * We may have already read "secondary headers"; purge
 688                  * them to give ourselves a clean restart.
 689                  */
 690                 for (i = 0; header[i]; i++) {
 691                         if (mi->s_hdr_data[i])
 692                                 strbuf_release(mi->s_hdr_data[i]);
 693                         mi->s_hdr_data[i] = NULL;
 694                 }
 695                 return 0;
 696         }
 697
 698         if (patchbreak(line)) {
 699                 if (mi->message_id)
 700                         fprintf(mi->cmitmsg, "Message-Id: %s\n", mi->message_id);
 701                 fclose(mi->cmitmsg);
 702                 mi->cmitmsg = NULL;
 703                 return 1;
 704         }
 705
 706         fputs(line->buf, mi->cmitmsg);
 707         return 0;
 708 }
 709
 710 static void handle_patch(struct mailinfo *mi, const struct strbuf *line)
 711 {
 712         fwrite(line->buf, 1, line->len, mi->patchfile);
 713         mi->patch_lines++;
 714 }
 715
 716 static void handle_filter(struct mailinfo *mi, struct strbuf *line)
 717 {
 718         switch (mi->filter_stage) {
 719         case 0:
 720                 if (!handle_commit_msg(mi, line))
 721                         break;
 722                 mi->filter_stage++;
 723         case 1:
 724                 handle_patch(mi, line);
 725                 break;
 726         }
 727 }
 728
 729 static int is_rfc2822_header(const struct strbuf *line)
 730 {
 731         /*
 732          * The section that defines the loosest possible
 733          * field name is "3.6.8 Optional fields".
 734          *
 735          * optional-field = field-name ":" unstructured CRLF
 736          * field-name = 1*ftext
 737          * ftext = %d33-57 / %59-126
 738          */
 739         int ch;
 740         char *cp = line->buf;
 741
 742         /* Count mbox From headers as headers */
 743         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 744                 return 1;
 745
 746         while ((ch = *cp++)) {
 747                 if (ch == ':')
 748                         return 1;
 749                 if ((33 <= ch && ch <= 57) ||
 750                     (59 <= ch && ch <= 126))
 751                         continue;
 752                 break;
 753         }
 754         return 0;
 755 }
 756
 757 static int read_one_header_line(struct strbuf *line, FILE *in)
 758 {
 759         struct strbuf continuation = STRBUF_INIT;
 760
 761         /* Get the first part of the line. */
 762         if (strbuf_getline(line, in, '\n'))
 763                 return 0;
 764
 765         /*
 766          * Is it an empty line or not a valid rfc2822 header?
 767          * If so, stop here, and return false ("not a header")
 768          */
 769         strbuf_rtrim(line);
 770         if (!line->len || !is_rfc2822_header(line)) {
 771                 /* Re-add the newline */
 772                 strbuf_addch(line, '\n');
 773                 return 0;
 774         }
 775
 776         /*
 777          * Now we need to eat all the continuation lines..
 778          * Yuck, 2822 header "folding"
 779          */
 780         for (;;) {
 781                 int peek;
 782
 783                 peek = fgetc(in); ungetc(peek, in);
 784                 if (peek != ' ' && peek != '\t')
 785                         break;
 786                 if (strbuf_getline(&continuation, in, '\n'))
 787                         break;
 788                 continuation.buf[0] = ' ';
 789                 strbuf_rtrim(&continuation);
 790                 strbuf_addbuf(line, &continuation);
 791         }
 792         strbuf_release(&continuation);
 793
 794         return 1;
 795 }
 796
 797 static int find_boundary(struct mailinfo *mi, struct strbuf *line)
 798 {
 799         while (!strbuf_getline(line, mi->input, '\n')) {
 800                 if (*content_top && is_multipart_boundary(line))
 801                         return 1;
 802         }
 803         return 0;
 804 }
 805
 806 static int handle_boundary(struct mailinfo *mi, struct strbuf *line)
 807 {
 808         struct strbuf newline = STRBUF_INIT;
 809
 810         strbuf_addch(&newline, '\n');
 811 again:
 812         if (line->len >= (*content_top)->len + 2 &&
 813             !memcmp(line->buf + (*content_top)->len, "--", 2)) {
 814                 /* we hit an end boundary */
 815                 /* pop the current boundary off the stack */
 816                 strbuf_release(*content_top);
 817                 free(*content_top);
 818                 *content_top = NULL;
 819
 820                 /* technically won't happen as is_multipart_boundary()
 821                    will fail first.  But just in case..
 822                  */
 823                 if (--content_top < content) {
 824                         fprintf(stderr, "Detected mismatched boundaries, "
 825                                         "can't recover\n");
 826                         exit(1);
 827                 }
 828                 handle_filter(mi, &newline);
 829                 strbuf_release(&newline);
 830
 831                 /* skip to the next boundary */
 832                 if (!find_boundary(mi, line))
 833                         return 0;
 834                 goto again;
 835         }
 836
 837         /* set some defaults */
 838         mi->transfer_encoding = TE_DONTCARE;
 839         strbuf_reset(&mi->charset);
 840
 841         /* slurp in this section's info */
 842         while (read_one_header_line(line, mi->input))
 843                 check_header(mi, line, mi->p_hdr_data, 0);
 844
 845         strbuf_release(&newline);
 846         /* replenish line */
 847         if (strbuf_getline(line, mi->input, '\n'))
 848                 return 0;
 849         strbuf_addch(line, '\n');
 850         return 1;
 851 }
 852
 853 static void handle_body(struct mailinfo *mi, struct strbuf *line)
 854 {
 855         struct strbuf prev = STRBUF_INIT;
 856
 857         /* Skip up to the first boundary */
 858         if (*content_top) {
 859                 if (!find_boundary(mi, line))
 860                         goto handle_body_out;
 861         }
 862
 863         do {
 864                 /* process any boundary lines */
 865                 if (*content_top && is_multipart_boundary(line)) {
 866                         /* flush any leftover */
 867                         if (prev.len) {
 868                                 handle_filter(mi, &prev);
 869                                 strbuf_reset(&prev);
 870                         }
 871                         if (!handle_boundary(mi, line))
 872                                 goto handle_body_out;
 873                 }
 874
 875                 /* Unwrap transfer encoding */
 876                 decode_transfer_encoding(mi, line);
 877
 878                 switch (mi->transfer_encoding) {
 879                 case TE_BASE64:
 880                 case TE_QP:
 881                 {
 882                         struct strbuf **lines, **it, *sb;
 883
 884                         /* Prepend any previous partial lines */
 885                         strbuf_insert(line, 0, prev.buf, prev.len);
 886                         strbuf_reset(&prev);
 887
 888                         /*
 889                          * This is a decoded line that may contain
 890                          * multiple new lines.  Pass only one chunk
 891                          * at a time to handle_filter()
 892                          */
 893                         lines = strbuf_split(line, '\n');
 894                         for (it = lines; (sb = *it); it++) {
 895                                 if (*(it + 1) == NULL) /* The last line */
 896                                         if (sb->buf[sb->len - 1] != '\n') {
 897                                                 /* Partial line, save it for later. */
 898                                                 strbuf_addbuf(&prev, sb);
 899                                                 break;
 900                                         }
 901                                 handle_filter(mi, sb);
 902                         }
 903                         /*
 904                          * The partial chunk is saved in "prev" and will be
 905                          * appended by the next iteration of read_line_with_nul().
 906                          */
 907                         strbuf_list_free(lines);
 908                         break;
 909                 }
 910                 default:
 911                         handle_filter(mi, line);
 912                 }
 913
 914         } while (!strbuf_getwholeline(line, mi->input, '\n'));
 915
 916 handle_body_out:
 917         strbuf_release(&prev);
 918 }
 919
 920 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 921 {
 922         const char *sp = data->buf;
 923         while (1) {
 924                 char *ep = strchr(sp, '\n');
 925                 int len;
 926                 if (!ep)
 927                         len = strlen(sp);
 928                 else
 929                         len = ep - sp;
 930                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 931                 if (!ep)
 932                         break;
 933                 sp = ep + 1;
 934         }
 935 }
 936
 937 static void handle_info(struct mailinfo *mi)
 938 {
 939         struct strbuf *hdr;
 940         int i;
 941
 942         for (i = 0; header[i]; i++) {
 943                 /* only print inbody headers if we output a patch file */
 944                 if (mi->patch_lines && mi->s_hdr_data[i])
 945                         hdr = mi->s_hdr_data[i];
 946                 else if (mi->p_hdr_data[i])
 947                         hdr = mi->p_hdr_data[i];
 948                 else
 949                         continue;
 950
 951                 if (!strcmp(header[i], "Subject")) {
 952                         if (!mi->keep_subject) {
 953                                 cleanup_subject(mi, hdr);
 954                                 cleanup_space(hdr);
 955                         }
 956                         output_header_lines(mi->output, "Subject", hdr);
 957                 } else if (!strcmp(header[i], "From")) {
 958                         cleanup_space(hdr);
 959                         handle_from(mi, hdr);
 960                         fprintf(mi->output, "Author: %s\n", mi->name.buf);
 961                         fprintf(mi->output, "Email: %s\n", mi->email.buf);
 962                 } else {
 963                         cleanup_space(hdr);
 964                         fprintf(mi->output, "%s: %s\n", header[i], hdr->buf);
 965                 }
 966         }
 967         fprintf(mi->output, "\n");
 968 }
 969
 970 static int mailinfo(struct mailinfo *mi, const char *msg, const char *patch)
 971 {
 972         int peek;
 973         struct strbuf line = STRBUF_INIT;
 974
 975         mi->cmitmsg = fopen(msg, "w");
 976         if (!mi->cmitmsg) {
 977                 perror(msg);
 978                 return -1;
 979         }
 980         mi->patchfile = fopen(patch, "w");
 981         if (!mi->patchfile) {
 982                 perror(patch);
 983                 fclose(mi->cmitmsg);
 984                 return -1;
 985         }
 986
 987         mi->p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->p_hdr_data)));
 988         mi->s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*(mi->s_hdr_data)));
 989
 990         do {
 991                 peek = fgetc(mi->input);
 992         } while (isspace(peek));
 993         ungetc(peek, mi->input);
 994
 995         /* process the email header */
 996         while (read_one_header_line(&line, mi->input))
 997                 check_header(mi, &line, mi->p_hdr_data, 1);
 998
 999         handle_body(mi, &line);
1000         fclose(mi->patchfile);
1001
1002         handle_info(mi);
1003         strbuf_release(&line);
1004         return 0;
1005 }
1006
1007 static int git_mailinfo_config(const char *var, const char *value, void *mi_)
1008 {
1009         struct mailinfo *mi = mi_;
1010
1011         if (!starts_with(var, "mailinfo."))
1012                 return git_default_config(var, value, NULL);
1013         if (!strcmp(var, "mailinfo.scissors")) {
1014                 mi->use_scissors = git_config_bool(var, value);
1015                 return 0;
1016         }
1017         /* perhaps others here */
1018         return 0;
1019 }
1020
1021 static void setup_mailinfo(struct mailinfo *mi)
1022 {
1023         memset(mi, 0, sizeof(*mi));
1024         strbuf_init(&mi->name, 0);
1025         strbuf_init(&mi->email, 0);
1026         strbuf_init(&mi->charset, 0);
1027         mi->header_stage = 1;
1028         mi->use_inbody_headers = 1;
1029         git_config(git_mailinfo_config, &mi);
1030 }
1031
1032 static void clear_mailinfo(struct mailinfo *mi)
1033 {
1034         int i;
1035
1036         strbuf_release(&mi->name);
1037         strbuf_release(&mi->email);
1038         strbuf_release(&mi->charset);
1039         free(mi->message_id);
1040
1041         for (i = 0; mi->p_hdr_data[i]; i++)
1042                 strbuf_release(mi->p_hdr_data[i]);
1043         free(mi->p_hdr_data);
1044         for (i = 0; mi->s_hdr_data[i]; i++)
1045                 strbuf_release(mi->s_hdr_data[i]);
1046         free(mi->s_hdr_data);
1047 }
1048
1049 static const char mailinfo_usage[] =
1050         "git mailinfo [-k | -b] [-m | --message-id] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] <msg> <patch> < mail >info";
1051
1052 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1053 {
1054         const char *def_charset;
1055         struct mailinfo mi;
1056         int status;
1057
1058         /* NEEDSWORK: might want to do the optional .git/ directory
1059          * discovery
1060          */
1061         setup_mailinfo(&mi);
1062
1063         def_charset = get_commit_output_encoding();
1064         mi.metainfo_charset = def_charset;
1065
1066         while (1 < argc && argv[1][0] == '-') {
1067                 if (!strcmp(argv[1], "-k"))
1068                         mi.keep_subject = 1;
1069                 else if (!strcmp(argv[1], "-b"))
1070                         mi.keep_non_patch_brackets_in_subject = 1;
1071                 else if (!strcmp(argv[1], "-m") || !strcmp(argv[1], "--message-id"))
1072                         mi.add_message_id = 1;
1073                 else if (!strcmp(argv[1], "-u"))
1074                         mi.metainfo_charset = def_charset;
1075                 else if (!strcmp(argv[1], "-n"))
1076                         mi.metainfo_charset = NULL;
1077                 else if (starts_with(argv[1], "--encoding="))
1078                         mi.metainfo_charset = argv[1] + 11;
1079                 else if (!strcmp(argv[1], "--scissors"))
1080                         mi.use_scissors = 1;
1081                 else if (!strcmp(argv[1], "--no-scissors"))
1082                         mi.use_scissors = 0;
1083                 else if (!strcmp(argv[1], "--no-inbody-headers"))
1084                         mi.use_inbody_headers = 0;
1085                 else
1086                         usage(mailinfo_usage);
1087                 argc--; argv++;
1088         }
1089
1090         if (argc != 3)
1091                 usage(mailinfo_usage);
1092
1093         mi.input = stdin;
1094         mi.output = stdout;
1095         status = !!mailinfo(&mi, argv[1], argv[2]);
1096         clear_mailinfo(&mi);
1097
1098         return status;
1099 }