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