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_BOUNDARIES 5
  32
  33 static void cleanup_space(struct strbuf *sb)
  34 {
  35         size_t pos, cnt;
  36         for (pos = 0; pos < sb->len; pos++) {
  37                 if (isspace(sb->buf[pos])) {
  38                         sb->buf[pos] = ' ';
  39                         for (cnt = 0; isspace(sb->buf[pos + cnt + 1]); cnt++);
  40                         strbuf_remove(sb, pos + 1, cnt);
  41                 }
  42         }
  43 }
  44
  45 static void get_sane_name(struct strbuf *out, struct strbuf *name, struct strbuf *email)
  46 {
  47         struct strbuf *src = name;
  48         if (name->len < 3 || 60 < name->len || strchr(name->buf, '@') ||
  49                 strchr(name->buf, '<') || strchr(name->buf, '>'))
  50                 src = email;
  51         else if (name == out)
  52                 return;
  53         strbuf_reset(out);
  54         strbuf_addbuf(out, src);
  55 }
  56
  57 static void parse_bogus_from(const struct strbuf *line)
  58 {
  59         /* John Doe <johndoe> */
  60
  61         char *bra, *ket;
  62         /* This is fallback, so do not bother if we already have an
  63          * e-mail address.
  64          */
  65         if (email.len)
  66                 return;
  67
  68         bra = strchr(line->buf, '<');
  69         if (!bra)
  70                 return;
  71         ket = strchr(bra, '>');
  72         if (!ket)
  73                 return;
  74
  75         strbuf_reset(&email);
  76         strbuf_add(&email, bra + 1, ket - bra - 1);
  77
  78         strbuf_reset(&name);
  79         strbuf_add(&name, line->buf, bra - line->buf);
  80         strbuf_trim(&name);
  81         get_sane_name(&name, &name, &email);
  82 }
  83
  84 static void handle_from(const struct strbuf *from)
  85 {
  86         char *at;
  87         size_t el;
  88         struct strbuf f;
  89
  90         strbuf_init(&f, from->len);
  91         strbuf_addbuf(&f, from);
  92
  93         at = strchr(f.buf, '@');
  94         if (!at) {
  95                 parse_bogus_from(from);
  96                 return;
  97         }
  98
  99         /*
 100          * If we already have one email, don't take any confusing lines
 101          */
 102         if (email.len && strchr(at + 1, '@')) {
 103                 strbuf_release(&f);
 104                 return;
 105         }
 106
 107         /* Pick up the string around '@', possibly delimited with <>
 108          * pair; that is the email part.
 109          */
 110         while (at > f.buf) {
 111                 char c = at[-1];
 112                 if (isspace(c))
 113                         break;
 114                 if (c == '<') {
 115                         at[-1] = ' ';
 116                         break;
 117                 }
 118                 at--;
 119         }
 120         el = strcspn(at, " \n\t\r\v\f>");
 121         strbuf_reset(&email);
 122         strbuf_add(&email, at, el);
 123         strbuf_remove(&f, at - f.buf, el + (at[el] ? 1 : 0));
 124
 125         /* The remainder is name.  It could be
 126          *
 127          * - "John Doe <john.doe@xz>"                   (a), or
 128          * - "john.doe@xz (John Doe)"                   (b), or
 129          * - "John (zzz) Doe <john.doe@xz> (Comment)"   (c)
 130          *
 131          * but we have removed the email part, so
 132          *
 133          * - remove extra spaces which could stay after email (case 'c'), and
 134          * - trim from both ends, possibly removing the () pair at the end
 135          *   (cases 'a' and 'b').
 136          */
 137         cleanup_space(&f);
 138         strbuf_trim(&f);
 139         if (f.buf[0] == '(' && f.len && f.buf[f.len - 1] == ')') {
 140                 strbuf_remove(&f, 0, 1);
 141                 strbuf_setlen(&f, f.len - 1);
 142         }
 143
 144         get_sane_name(&name, &f, &email);
 145         strbuf_release(&f);
 146 }
 147
 148 static void handle_header(struct strbuf **out, const struct strbuf *line)
 149 {
 150         if (!*out) {
 151                 *out = xmalloc(sizeof(struct strbuf));
 152                 strbuf_init(*out, line->len);
 153         } else
 154                 strbuf_reset(*out);
 155
 156         strbuf_addbuf(*out, line);
 157 }
 158
 159 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 160  * to have enough heuristics to grok MIME encoded patches often found
 161  * on our mailing lists.  For example, we do not even treat header lines
 162  * case insensitively.
 163  */
 164
 165 static int slurp_attr(const char *line, const char *name, struct strbuf *attr)
 166 {
 167         const char *ends, *ap = strcasestr(line, name);
 168         size_t sz;
 169
 170         strbuf_setlen(attr, 0);
 171         if (!ap)
 172                 return 0;
 173         ap += strlen(name);
 174         if (*ap == '"') {
 175                 ap++;
 176                 ends = "\"";
 177         }
 178         else
 179                 ends = "; \t";
 180         sz = strcspn(ap, ends);
 181         strbuf_add(attr, ap, sz);
 182         return 1;
 183 }
 184
 185 static struct strbuf *content[MAX_BOUNDARIES];
 186
 187 static struct strbuf **content_top = content;
 188
 189 static void handle_content_type(struct strbuf *line)
 190 {
 191         struct strbuf *boundary = xmalloc(sizeof(struct strbuf));
 192         strbuf_init(boundary, line->len);
 193
 194         if (slurp_attr(line->buf, "boundary=", boundary)) {
 195                 strbuf_insert(boundary, 0, "--", 2);
 196                 if (++content_top >= &content[MAX_BOUNDARIES]) {
 197                         fprintf(stderr, "Too many boundaries to handle\n");
 198                         exit(1);
 199                 }
 200                 *content_top = boundary;
 201                 boundary = NULL;
 202         }
 203         slurp_attr(line->buf, "charset=", &charset);
 204
 205         if (boundary) {
 206                 strbuf_release(boundary);
 207                 free(boundary);
 208         }
 209 }
 210
 211 static void handle_message_id(const struct strbuf *line)
 212 {
 213         if (add_message_id)
 214                 message_id = strdup(line->buf);
 215 }
 216
 217 static void handle_content_transfer_encoding(const struct strbuf *line)
 218 {
 219         if (strcasestr(line->buf, "base64"))
 220                 transfer_encoding = TE_BASE64;
 221         else if (strcasestr(line->buf, "quoted-printable"))
 222                 transfer_encoding = TE_QP;
 223         else
 224                 transfer_encoding = TE_DONTCARE;
 225 }
 226
 227 static int is_multipart_boundary(const struct strbuf *line)
 228 {
 229         return (((*content_top)->len <= line->len) &&
 230                 !memcmp(line->buf, (*content_top)->buf, (*content_top)->len));
 231 }
 232
 233 static void cleanup_subject(struct strbuf *subject)
 234 {
 235         size_t at = 0;
 236
 237         while (at < subject->len) {
 238                 char *pos;
 239                 size_t remove;
 240
 241                 switch (subject->buf[at]) {
 242                 case 'r': case 'R':
 243                         if (subject->len <= at + 3)
 244                                 break;
 245                         if ((subject->buf[at + 1] == 'e' ||
 246                              subject->buf[at + 1] == 'E') &&
 247                             subject->buf[at + 2] == ':') {
 248                                 strbuf_remove(subject, at, 3);
 249                                 continue;
 250                         }
 251                         at++;
 252                         break;
 253                 case ' ': case '\t': case ':':
 254                         strbuf_remove(subject, at, 1);
 255                         continue;
 256                 case '[':
 257                         pos = strchr(subject->buf + at, ']');
 258                         if (!pos)
 259                                 break;
 260                         remove = pos - subject->buf + at + 1;
 261                         if (!keep_non_patch_brackets_in_subject ||
 262                             (7 <= remove &&
 263                              memmem(subject->buf + at, remove, "PATCH", 5)))
 264                                 strbuf_remove(subject, at, remove);
 265                         else {
 266                                 at += remove;
 267                                 /*
 268                                  * If the input had a space after the ], keep
 269                                  * it.  We don't bother with finding the end of
 270                                  * the space, since we later normalize it
 271                                  * anyway.
 272                                  */
 273                                 if (isspace(subject->buf[at]))
 274                                         at += 1;
 275                         }
 276                         continue;
 277                 }
 278                 break;
 279         }
 280         strbuf_trim(subject);
 281 }
 282
 283 #define MAX_HDR_PARSED 10
 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, int *still_looking)
 647 {
 648         if (!cmitmsg)
 649                 return 0;
 650
 651         if (*still_looking) {
 652                 if (!line->len || (line->len == 1 && line->buf[0] == '\n'))
 653                         return 0;
 654         }
 655
 656         if (use_inbody_headers && *still_looking) {
 657                 *still_looking = check_header(line, s_hdr_data, 0);
 658                 if (*still_looking)
 659                         return 0;
 660         } else
 661                 /* Only trim the first (blank) line of the commit message
 662                  * when ignoring in-body headers.
 663                  */
 664                 *still_looking = 0;
 665
 666         /* normalize the log message to UTF-8. */
 667         if (metainfo_charset)
 668                 convert_to_utf8(line, charset.buf);
 669
 670         if (use_scissors && is_scissors_line(line)) {
 671                 int i;
 672                 if (fseek(cmitmsg, 0L, SEEK_SET))
 673                         die_errno("Could not rewind output message file");
 674                 if (ftruncate(fileno(cmitmsg), 0))
 675                         die_errno("Could not truncate output message file at scissors");
 676                 *still_looking = 1;
 677
 678                 /*
 679                  * We may have already read "secondary headers"; purge
 680                  * them to give ourselves a clean restart.
 681                  */
 682                 for (i = 0; header[i]; i++) {
 683                         if (s_hdr_data[i])
 684                                 strbuf_release(s_hdr_data[i]);
 685                         s_hdr_data[i] = NULL;
 686                 }
 687                 return 0;
 688         }
 689
 690         if (patchbreak(line)) {
 691                 if (message_id)
 692                         fprintf(cmitmsg, "Message-Id: %s\n", message_id);
 693                 fclose(cmitmsg);
 694                 cmitmsg = NULL;
 695                 return 1;
 696         }
 697
 698         fputs(line->buf, cmitmsg);
 699         return 0;
 700 }
 701
 702 static void handle_patch(const struct strbuf *line)
 703 {
 704         fwrite(line->buf, 1, line->len, patchfile);
 705         patch_lines++;
 706 }
 707
 708 static void handle_filter(struct strbuf *line, int *filter_stage, int *header_stage)
 709 {
 710         switch (*filter_stage) {
 711         case 0:
 712                 if (!handle_commit_msg(line, header_stage))
 713                         break;
 714                 (*filter_stage)++;
 715         case 1:
 716                 handle_patch(line);
 717                 break;
 718         }
 719 }
 720
 721 static int is_rfc2822_header(const struct strbuf *line)
 722 {
 723         /*
 724          * The section that defines the loosest possible
 725          * field name is "3.6.8 Optional fields".
 726          *
 727          * optional-field = field-name ":" unstructured CRLF
 728          * field-name = 1*ftext
 729          * ftext = %d33-57 / %59-126
 730          */
 731         int ch;
 732         char *cp = line->buf;
 733
 734         /* Count mbox From headers as headers */
 735         if (starts_with(cp, "From ") || starts_with(cp, ">From "))
 736                 return 1;
 737
 738         while ((ch = *cp++)) {
 739                 if (ch == ':')
 740                         return 1;
 741                 if ((33 <= ch && ch <= 57) ||
 742                     (59 <= ch && ch <= 126))
 743                         continue;
 744                 break;
 745         }
 746         return 0;
 747 }
 748
 749 static int read_one_header_line(struct strbuf *line, FILE *in)
 750 {
 751         struct strbuf continuation = STRBUF_INIT;
 752
 753         /* Get the first part of the line. */
 754         if (strbuf_getline(line, in, '\n'))
 755                 return 0;
 756
 757         /*
 758          * Is it an empty line or not a valid rfc2822 header?
 759          * If so, stop here, and return false ("not a header")
 760          */
 761         strbuf_rtrim(line);
 762         if (!line->len || !is_rfc2822_header(line)) {
 763                 /* Re-add the newline */
 764                 strbuf_addch(line, '\n');
 765                 return 0;
 766         }
 767
 768         /*
 769          * Now we need to eat all the continuation lines..
 770          * Yuck, 2822 header "folding"
 771          */
 772         for (;;) {
 773                 int peek;
 774
 775                 peek = fgetc(in); ungetc(peek, in);
 776                 if (peek != ' ' && peek != '\t')
 777                         break;
 778                 if (strbuf_getline(&continuation, in, '\n'))
 779                         break;
 780                 continuation.buf[0] = ' ';
 781                 strbuf_rtrim(&continuation);
 782                 strbuf_addbuf(line, &continuation);
 783         }
 784         strbuf_release(&continuation);
 785
 786         return 1;
 787 }
 788
 789 static int find_boundary(void)
 790 {
 791         while (!strbuf_getline(&line, fin, '\n')) {
 792                 if (*content_top && is_multipart_boundary(&line))
 793                         return 1;
 794         }
 795         return 0;
 796 }
 797
 798 static int handle_boundary(struct strbuf *line, int *filter_stage, int *header_stage)
 799 {
 800         struct strbuf newline = STRBUF_INIT;
 801
 802         strbuf_addch(&newline, '\n');
 803 again:
 804         if (line->len >= (*content_top)->len + 2 &&
 805             !memcmp(line->buf + (*content_top)->len, "--", 2)) {
 806                 /* we hit an end boundary */
 807                 /* pop the current boundary off the stack */
 808                 strbuf_release(*content_top);
 809                 free(*content_top);
 810                 *content_top = NULL;
 811
 812                 /* technically won't happen as is_multipart_boundary()
 813                    will fail first.  But just in case..
 814                  */
 815                 if (--content_top < content) {
 816                         fprintf(stderr, "Detected mismatched boundaries, "
 817                                         "can't recover\n");
 818                         exit(1);
 819                 }
 820                 handle_filter(&newline, filter_stage, header_stage);
 821                 strbuf_release(&newline);
 822
 823                 /* skip to the next boundary */
 824                 if (!find_boundary())
 825                         return 0;
 826                 goto again;
 827         }
 828
 829         /* set some defaults */
 830         transfer_encoding = TE_DONTCARE;
 831         strbuf_reset(&charset);
 832
 833         /* slurp in this section's info */
 834         while (read_one_header_line(line, fin))
 835                 check_header(line, p_hdr_data, 0);
 836
 837         strbuf_release(&newline);
 838         /* replenish line */
 839         if (strbuf_getline(line, fin, '\n'))
 840                 return 0;
 841         strbuf_addch(line, '\n');
 842         return 1;
 843 }
 844
 845 static void handle_body(struct strbuf *line)
 846 {
 847         struct strbuf prev = STRBUF_INIT;
 848         int filter_stage = 0;
 849         int header_stage = 1;
 850
 851         /* Skip up to the first boundary */
 852         if (*content_top) {
 853                 if (!find_boundary())
 854                         goto handle_body_out;
 855         }
 856
 857         do {
 858                 /* process any boundary lines */
 859                 if (*content_top && is_multipart_boundary(line)) {
 860                         /* flush any leftover */
 861                         if (prev.len) {
 862                                 handle_filter(&prev, &filter_stage, &header_stage);
 863                                 strbuf_reset(&prev);
 864                         }
 865                         if (!handle_boundary(line, &filter_stage, &header_stage))
 866                                 goto handle_body_out;
 867                 }
 868
 869                 /* Unwrap transfer encoding */
 870                 decode_transfer_encoding(line);
 871
 872                 switch (transfer_encoding) {
 873                 case TE_BASE64:
 874                 case TE_QP:
 875                 {
 876                         struct strbuf **lines, **it, *sb;
 877
 878                         /* Prepend any previous partial lines */
 879                         strbuf_insert(line, 0, prev.buf, prev.len);
 880                         strbuf_reset(&prev);
 881
 882                         /*
 883                          * This is a decoded line that may contain
 884                          * multiple new lines.  Pass only one chunk
 885                          * at a time to handle_filter()
 886                          */
 887                         lines = strbuf_split(line, '\n');
 888                         for (it = lines; (sb = *it); it++) {
 889                                 if (*(it + 1) == NULL) /* The last line */
 890                                         if (sb->buf[sb->len - 1] != '\n') {
 891                                                 /* Partial line, save it for later. */
 892                                                 strbuf_addbuf(&prev, sb);
 893                                                 break;
 894                                         }
 895                                 handle_filter(sb, &filter_stage, &header_stage);
 896                         }
 897                         /*
 898                          * The partial chunk is saved in "prev" and will be
 899                          * appended by the next iteration of read_line_with_nul().
 900                          */
 901                         strbuf_list_free(lines);
 902                         break;
 903                 }
 904                 default:
 905                         handle_filter(line, &filter_stage, &header_stage);
 906                 }
 907
 908         } while (!strbuf_getwholeline(line, fin, '\n'));
 909
 910 handle_body_out:
 911         strbuf_release(&prev);
 912 }
 913
 914 static void output_header_lines(FILE *fout, const char *hdr, const struct strbuf *data)
 915 {
 916         const char *sp = data->buf;
 917         while (1) {
 918                 char *ep = strchr(sp, '\n');
 919                 int len;
 920                 if (!ep)
 921                         len = strlen(sp);
 922                 else
 923                         len = ep - sp;
 924                 fprintf(fout, "%s: %.*s\n", hdr, len, sp);
 925                 if (!ep)
 926                         break;
 927                 sp = ep + 1;
 928         }
 929 }
 930
 931 static void handle_info(void)
 932 {
 933         struct strbuf *hdr;
 934         int i;
 935
 936         for (i = 0; header[i]; i++) {
 937                 /* only print inbody headers if we output a patch file */
 938                 if (patch_lines && s_hdr_data[i])
 939                         hdr = s_hdr_data[i];
 940                 else if (p_hdr_data[i])
 941                         hdr = p_hdr_data[i];
 942                 else
 943                         continue;
 944
 945                 if (!strcmp(header[i], "Subject")) {
 946                         if (!keep_subject) {
 947                                 cleanup_subject(hdr);
 948                                 cleanup_space(hdr);
 949                         }
 950                         output_header_lines(fout, "Subject", hdr);
 951                 } else if (!strcmp(header[i], "From")) {
 952                         cleanup_space(hdr);
 953                         handle_from(hdr);
 954                         fprintf(fout, "Author: %s\n", name.buf);
 955                         fprintf(fout, "Email: %s\n", email.buf);
 956                 } else {
 957                         cleanup_space(hdr);
 958                         fprintf(fout, "%s: %s\n", header[i], hdr->buf);
 959                 }
 960         }
 961         fprintf(fout, "\n");
 962 }
 963
 964 static int mailinfo(FILE *in, FILE *out, const char *msg, const char *patch)
 965 {
 966         int peek;
 967         fin = in;
 968         fout = out;
 969
 970         cmitmsg = fopen(msg, "w");
 971         if (!cmitmsg) {
 972                 perror(msg);
 973                 return -1;
 974         }
 975         patchfile = fopen(patch, "w");
 976         if (!patchfile) {
 977                 perror(patch);
 978                 fclose(cmitmsg);
 979                 return -1;
 980         }
 981
 982         p_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*p_hdr_data));
 983         s_hdr_data = xcalloc(MAX_HDR_PARSED, sizeof(*s_hdr_data));
 984
 985         do {
 986                 peek = fgetc(in);
 987         } while (isspace(peek));
 988         ungetc(peek, in);
 989
 990         /* process the email header */
 991         while (read_one_header_line(&line, fin))
 992                 check_header(&line, p_hdr_data, 1);
 993
 994         handle_body(&line);
 995         fclose(patchfile);
 996
 997         handle_info();
 998
 999         return 0;
1000 }
1001
1002 static int git_mailinfo_config(const char *var, const char *value, void *unused)
1003 {
1004         if (!starts_with(var, "mailinfo."))
1005                 return git_default_config(var, value, unused);
1006         if (!strcmp(var, "mailinfo.scissors")) {
1007                 use_scissors = git_config_bool(var, value);
1008                 return 0;
1009         }
1010         /* perhaps others here */
1011         return 0;
1012 }
1013
1014 static const char mailinfo_usage[] =
1015         "git mailinfo [-k | -b] [-m | --message-id] [-u | --encoding=<encoding> | -n] [--scissors | --no-scissors] <msg> <patch> < mail >info";
1016
1017 int cmd_mailinfo(int argc, const char **argv, const char *prefix)
1018 {
1019         const char *def_charset;
1020
1021         /* NEEDSWORK: might want to do the optional .git/ directory
1022          * discovery
1023          */
1024         git_config(git_mailinfo_config, NULL);
1025
1026         def_charset = get_commit_output_encoding();
1027         metainfo_charset = def_charset;
1028
1029         while (1 < argc && argv[1][0] == '-') {
1030                 if (!strcmp(argv[1], "-k"))
1031                         keep_subject = 1;
1032                 else if (!strcmp(argv[1], "-b"))
1033                         keep_non_patch_brackets_in_subject = 1;
1034                 else if (!strcmp(argv[1], "-m") || !strcmp(argv[1], "--message-id"))
1035                         add_message_id = 1;
1036                 else if (!strcmp(argv[1], "-u"))
1037                         metainfo_charset = def_charset;
1038                 else if (!strcmp(argv[1], "-n"))
1039                         metainfo_charset = NULL;
1040                 else if (starts_with(argv[1], "--encoding="))
1041                         metainfo_charset = argv[1] + 11;
1042                 else if (!strcmp(argv[1], "--scissors"))
1043                         use_scissors = 1;
1044                 else if (!strcmp(argv[1], "--no-scissors"))
1045                         use_scissors = 0;
1046                 else if (!strcmp(argv[1], "--no-inbody-headers"))
1047                         use_inbody_headers = 0;
1048                 else
1049                         usage(mailinfo_usage);
1050                 argc--; argv++;
1051         }
1052
1053         if (argc != 3)
1054                 usage(mailinfo_usage);
1055
1056         return !!mailinfo(stdin, stdout, argv[1], argv[2]);
1057 }