mailinfo.c

   1 /*
   2  * Another stupid program, this one parsing the headers of an
   3  * email to figure out authorship and subject
   4  */
   5 #define _GNU_SOURCE
   6 #include <stdio.h>
   7 #include <stdlib.h>
   8 #include <string.h>
   9 #include <ctype.h>
  10 #include <iconv.h>
  11
  12 #ifdef NO_STRCASESTR
  13 extern char *gitstrcasestr(const char *haystack, const char *needle);
  14 #endif
  15
  16 static FILE *cmitmsg, *patchfile;
  17
  18 static int keep_subject = 0;
  19 static int metainfo_utf8 = 0;
  20 static char line[1000];
  21 static char date[1000];
  22 static char name[1000];
  23 static char email[1000];
  24 static char subject[1000];
  25
  26 static enum  {
  27         TE_DONTCARE, TE_QP, TE_BASE64,
  28 } transfer_encoding;
  29 static char charset[256];
  30
  31 static char multipart_boundary[1000];
  32 static int multipart_boundary_len;
  33 static int patch_lines = 0;
  34
  35 static char *sanity_check(char *name, char *email)
  36 {
  37         int len = strlen(name);
  38         if (len < 3 || len > 60)
  39                 return email;
  40         if (strchr(name, '@') || strchr(name, '<') || strchr(name, '>'))
  41                 return email;
  42         return name;
  43 }
  44
  45 static int handle_from(char *line)
  46 {
  47         char *at = strchr(line, '@');
  48         char *dst;
  49
  50         if (!at)
  51                 return 0;
  52
  53         /*
  54          * If we already have one email, don't take any confusing lines
  55          */
  56         if (*email && strchr(at+1, '@'))
  57                 return 0;
  58
  59         /* Pick up the string around '@', possibly delimited with <>
  60          * pair; that is the email part.  White them out while copying.
  61          */
  62         while (at > line) {
  63                 char c = at[-1];
  64                 if (isspace(c))
  65                         break;
  66                 if (c == '<') {
  67                         at[-1] = ' ';
  68                         break;
  69                 }
  70                 at--;
  71         }
  72         dst = email;
  73         for (;;) {
  74                 unsigned char c = *at;
  75                 if (!c || c == '>' || isspace(c)) {
  76                         if (c == '>')
  77                                 *at = ' ';
  78                         break;
  79                 }
  80                 *at++ = ' ';
  81                 *dst++ = c;
  82         }
  83         *dst++ = 0;
  84
  85         /* The remainder is name.  It could be "John Doe <john.doe@xz>"
  86          * or "john.doe@xz (John Doe)", but we have whited out the
  87          * email part, so trim from both ends, possibly removing
  88          * the () pair at the end.
  89          */
  90         at = line + strlen(line);
  91         while (at > line) {
  92                 unsigned char c = *--at;
  93                 if (!isspace(c)) {
  94                         at[(c == ')') ? 0 : 1] = 0;
  95                         break;
  96                 }
  97         }
  98
  99         at = line;
 100         for (;;) {
 101                 unsigned char c = *at;
 102                 if (!c || !isspace(c)) {
 103                         if (c == '(')
 104                                 at++;
 105                         break;
 106                 }
 107                 at++;
 108         }
 109         at = sanity_check(at, email);
 110         strcpy(name, at);
 111         return 1;
 112 }
 113
 114 static int handle_date(char *line)
 115 {
 116         strcpy(date, line);
 117         return 0;
 118 }
 119
 120 static int handle_subject(char *line)
 121 {
 122         strcpy(subject, line);
 123         return 0;
 124 }
 125
 126 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 127  * to have enough heuristics to grok MIME encoded patches often found
 128  * on our mailing lists.  For example, we do not even treat header lines
 129  * case insensitively.
 130  */
 131
 132 static int slurp_attr(const char *line, const char *name, char *attr)
 133 {
 134         char *ends, *ap = strcasestr(line, name);
 135         size_t sz;
 136
 137         if (!ap) {
 138                 *attr = 0;
 139                 return 0;
 140         }
 141         ap += strlen(name);
 142         if (*ap == '"') {
 143                 ap++;
 144                 ends = "\"";
 145         }
 146         else
 147                 ends = "; \t";
 148         sz = strcspn(ap, ends);
 149         memcpy(attr, ap, sz);
 150         attr[sz] = 0;
 151         return 1;
 152 }
 153
 154 static int handle_subcontent_type(char *line)
 155 {
 156         /* We do not want to mess with boundary.  Note that we do not
 157          * handle nested multipart.
 158          */
 159         if (strcasestr(line, "boundary=")) {
 160                 fprintf(stderr, "Not handling nested multipart message.\n");
 161                 exit(1);
 162         }
 163         slurp_attr(line, "charset=", charset);
 164         if (*charset) {
 165                 int i, c;
 166                 for (i = 0; (c = charset[i]) != 0; i++)
 167                         charset[i] = tolower(c);
 168         }
 169         return 0;
 170 }
 171
 172 static int handle_content_type(char *line)
 173 {
 174         *multipart_boundary = 0;
 175         if (slurp_attr(line, "boundary=", multipart_boundary + 2)) {
 176                 memcpy(multipart_boundary, "--", 2);
 177                 multipart_boundary_len = strlen(multipart_boundary);
 178         }
 179         slurp_attr(line, "charset=", charset);
 180         return 0;
 181 }
 182
 183 static int handle_content_transfer_encoding(char *line)
 184 {
 185         if (strcasestr(line, "base64"))
 186                 transfer_encoding = TE_BASE64;
 187         else if (strcasestr(line, "quoted-printable"))
 188                 transfer_encoding = TE_QP;
 189         else
 190                 transfer_encoding = TE_DONTCARE;
 191         return 0;
 192 }
 193
 194 static int is_multipart_boundary(const char *line)
 195 {
 196         return (!memcmp(line, multipart_boundary, multipart_boundary_len));
 197 }
 198
 199 static int eatspace(char *line)
 200 {
 201         int len = strlen(line);
 202         while (len > 0 && isspace(line[len-1]))
 203                 line[--len] = 0;
 204         return len;
 205 }
 206
 207 #define SEEN_FROM 01
 208 #define SEEN_DATE 02
 209 #define SEEN_SUBJECT 04
 210
 211 /* First lines of body can have From:, Date:, and Subject: */
 212 static int handle_inbody_header(int *seen, char *line)
 213 {
 214         if (!memcmp("From:", line, 5) && isspace(line[5])) {
 215                 if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
 216                         *seen |= SEEN_FROM;
 217                         return 1;
 218                 }
 219         }
 220         if (!memcmp("Date:", line, 5) && isspace(line[5])) {
 221                 if (!(*seen & SEEN_DATE)) {
 222                         handle_date(line+6);
 223                         *seen |= SEEN_DATE;
 224                         return 1;
 225                 }
 226         }
 227         if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
 228                 if (!(*seen & SEEN_SUBJECT)) {
 229                         handle_subject(line+9);
 230                         *seen |= SEEN_SUBJECT;
 231                         return 1;
 232                 }
 233         }
 234         if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
 235                 if (!(*seen & SEEN_SUBJECT)) {
 236                         handle_subject(line);
 237                         *seen |= SEEN_SUBJECT;
 238                         return 1;
 239                 }
 240         }
 241         return 0;
 242 }
 243
 244 static char *cleanup_subject(char *subject)
 245 {
 246         if (keep_subject)
 247                 return subject;
 248         for (;;) {
 249                 char *p;
 250                 int len, remove;
 251                 switch (*subject) {
 252                 case 'r': case 'R':
 253                         if (!memcmp("e:", subject+1, 2)) {
 254                                 subject +=3;
 255                                 continue;
 256                         }
 257                         break;
 258                 case ' ': case '\t': case ':':
 259                         subject++;
 260                         continue;
 261
 262                 case '[':
 263                         p = strchr(subject, ']');
 264                         if (!p) {
 265                                 subject++;
 266                                 continue;
 267                         }
 268                         len = strlen(p);
 269                         remove = p - subject;
 270                         if (remove <= len *2) {
 271                                 subject = p+1;
 272                                 continue;
 273                         }
 274                         break;
 275                 }
 276                 return subject;
 277         }
 278 }
 279
 280 static void cleanup_space(char *buf)
 281 {
 282         unsigned char c;
 283         while ((c = *buf) != 0) {
 284                 buf++;
 285                 if (isspace(c)) {
 286                         buf[-1] = ' ';
 287                         c = *buf;
 288                         while (isspace(c)) {
 289                                 int len = strlen(buf);
 290                                 memmove(buf, buf+1, len);
 291                                 c = *buf;
 292                         }
 293                 }
 294         }
 295 }
 296
 297 typedef int (*header_fn_t)(char *);
 298 struct header_def {
 299         const char *name;
 300         header_fn_t func;
 301         int namelen;
 302 };
 303
 304 static void check_header(char *line, int len, struct header_def *header)
 305 {
 306         int i;
 307
 308         if (header[0].namelen <= 0) {
 309                 for (i = 0; header[i].name; i++)
 310                         header[i].namelen = strlen(header[i].name);
 311         }
 312         for (i = 0; header[i].name; i++) {
 313                 int len = header[i].namelen;
 314                 if (!strncasecmp(line, header[i].name, len) &&
 315                     line[len] == ':' && isspace(line[len + 1])) {
 316                         header[i].func(line + len + 2);
 317                         break;
 318                 }
 319         }
 320 }
 321
 322 static void check_subheader_line(char *line, int len)
 323 {
 324         static struct header_def header[] = {
 325                 { "Content-Type", handle_subcontent_type },
 326                 { "Content-Transfer-Encoding",
 327                   handle_content_transfer_encoding },
 328                 { NULL },
 329         };
 330         check_header(line, len, header);
 331 }
 332 static void check_header_line(char *line, int len)
 333 {
 334         static struct header_def header[] = {
 335                 { "From", handle_from },
 336                 { "Date", handle_date },
 337                 { "Subject", handle_subject },
 338                 { "Content-Type", handle_content_type },
 339                 { "Content-Transfer-Encoding",
 340                   handle_content_transfer_encoding },
 341                 { NULL },
 342         };
 343         check_header(line, len, header);
 344 }
 345
 346 static int read_one_header_line(char *line, int sz, FILE *in)
 347 {
 348         int ofs = 0;
 349         while (ofs < sz) {
 350                 int peek, len;
 351                 if (fgets(line + ofs, sz - ofs, in) == NULL)
 352                         return ofs;
 353                 len = eatspace(line + ofs);
 354                 if (len == 0)
 355                         return ofs;
 356                 peek = fgetc(in); ungetc(peek, in);
 357                 if (peek == ' ' || peek == '\t') {
 358                         /* Yuck, 2822 header "folding" */
 359                         ofs += len;
 360                         continue;
 361                 }
 362                 return ofs + len;
 363         }
 364         return ofs;
 365 }
 366
 367 static unsigned hexval(int c)
 368 {
 369         if (c >= '0' && c <= '9')
 370                 return c - '0';
 371         if (c >= 'a' && c <= 'f')
 372                 return c - 'a' + 10;
 373         if (c >= 'A' && c <= 'F')
 374                 return c - 'A' + 10;
 375         return ~0;
 376 }
 377
 378 static int decode_q_segment(char *in, char *ot, char *ep)
 379 {
 380         int c;
 381         while ((c = *in++) != 0 && (in <= ep)) {
 382                 if (c == '=') {
 383                         int d = *in++;
 384                         if (d == '\n' || !d)
 385                                 break; /* drop trailing newline */
 386                         *ot++ = ((hexval(d) << 4) | hexval(*in++));
 387                 }
 388                 else
 389                         *ot++ = c;
 390         }
 391         *ot = 0;
 392         return 0;
 393 }
 394
 395 static int decode_b_segment(char *in, char *ot, char *ep)
 396 {
 397         /* Decode in..ep, possibly in-place to ot */
 398         int c, pos = 0, acc = 0;
 399
 400         while ((c = *in++) != 0 && (in <= ep)) {
 401                 if (c == '+')
 402                         c = 62;
 403                 else if (c == '/')
 404                         c = 63;
 405                 else if ('A' <= c && c <= 'Z')
 406                         c -= 'A';
 407                 else if ('a' <= c && c <= 'z')
 408                         c -= 'a' - 26;
 409                 else if ('0' <= c && c <= '9')
 410                         c -= '0' - 52;
 411                 else if (c == '=') {
 412                         /* padding is almost like (c == 0), except we do
 413                          * not output NUL resulting only from it;
 414                          * for now we just trust the data.
 415                          */
 416                         c = 0;
 417                 }
 418                 else
 419                         continue; /* garbage */
 420                 switch (pos++) {
 421                 case 0:
 422                         acc = (c << 2);
 423                         break;
 424                 case 1:
 425                         *ot++ = (acc | (c >> 4));
 426                         acc = (c & 15) << 4;
 427                         break;
 428                 case 2:
 429                         *ot++ = (acc | (c >> 2));
 430                         acc = (c & 3) << 6;
 431                         break;
 432                 case 3:
 433                         *ot++ = (acc | c);
 434                         acc = pos = 0;
 435                         break;
 436                 }
 437         }
 438         *ot = 0;
 439         return 0;
 440 }
 441
 442 static void convert_to_utf8(char *line, char *charset)
 443 {
 444         if (*charset) {
 445                 char *in, *out;
 446                 size_t insize, outsize, nrc;
 447                 char outbuf[4096]; /* cheat */
 448                 iconv_t conv = iconv_open("utf-8", charset);
 449
 450                 if (conv == (iconv_t) -1) {
 451                         fprintf(stderr, "cannot convert from %s to utf-8\n",
 452                                 charset);
 453                         *charset = 0;
 454                         return;
 455                 }
 456                 in = line;
 457                 insize = strlen(in);
 458                 out = outbuf;
 459                 outsize = sizeof(outbuf);
 460                 nrc = iconv(conv, &in, &insize, &out, &outsize);
 461                 iconv_close(conv);
 462                 if (nrc == (size_t) -1)
 463                         return;
 464                 *out = 0;
 465                 strcpy(line, outbuf);
 466         }
 467 }
 468
 469 static void decode_header_bq(char *it)
 470 {
 471         char *in, *out, *ep, *cp, *sp;
 472         char outbuf[1000];
 473
 474         in = it;
 475         out = outbuf;
 476         while ((ep = strstr(in, "=?")) != NULL) {
 477                 int sz, encoding;
 478                 char charset_q[256], piecebuf[256];
 479                 if (in != ep) {
 480                         sz = ep - in;
 481                         memcpy(out, in, sz);
 482                         out += sz;
 483                         in += sz;
 484                 }
 485                 /* E.g.
 486                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 487                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 488                  */
 489                 ep += 2;
 490                 cp = strchr(ep, '?');
 491                 if (!cp)
 492                         return; /* no munging */
 493                 for (sp = ep; sp < cp; sp++)
 494                         charset_q[sp - ep] = tolower(*sp);
 495                 charset_q[cp - ep] = 0;
 496                 encoding = cp[1];
 497                 if (!encoding || cp[2] != '?')
 498                         return; /* no munging */
 499                 ep = strstr(cp + 3, "?=");
 500                 if (!ep)
 501                         return; /* no munging */
 502                 switch (tolower(encoding)) {
 503                 default:
 504                         return; /* no munging */
 505                 case 'b':
 506                         sz = decode_b_segment(cp + 3, piecebuf, ep);
 507                         break;
 508                 case 'q':
 509                         sz = decode_q_segment(cp + 3, piecebuf, ep);
 510                         break;
 511                 }
 512                 if (sz < 0)
 513                         return;
 514                 if (metainfo_utf8)
 515                         convert_to_utf8(piecebuf, charset_q);
 516                 strcpy(out, piecebuf);
 517                 out += strlen(out);
 518                 in = ep + 2;
 519         }
 520         strcpy(out, in);
 521         strcpy(it, outbuf);
 522 }
 523
 524 static void decode_transfer_encoding(char *line)
 525 {
 526         char *ep;
 527
 528         switch (transfer_encoding) {
 529         case TE_QP:
 530                 ep = line + strlen(line);
 531                 decode_q_segment(line, line, ep);
 532                 break;
 533         case TE_BASE64:
 534                 ep = line + strlen(line);
 535                 decode_b_segment(line, line, ep);
 536                 break;
 537         case TE_DONTCARE:
 538                 break;
 539         }
 540 }
 541
 542 static void handle_info(void)
 543 {
 544         char *sub;
 545         static int done_info = 0;
 546
 547         if (done_info)
 548                 return;
 549
 550         done_info = 1;
 551         sub = cleanup_subject(subject);
 552         cleanup_space(name);
 553         cleanup_space(date);
 554         cleanup_space(email);
 555         cleanup_space(sub);
 556
 557         /* Unwrap inline B and Q encoding, and optionally
 558          * normalize the meta information to utf8.
 559          */
 560         decode_header_bq(name);
 561         decode_header_bq(date);
 562         decode_header_bq(email);
 563         decode_header_bq(sub);
 564         printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
 565                name, email, sub, date);
 566 }
 567
 568 /* We are inside message body and have read line[] already.
 569  * Spit out the commit log.
 570  */
 571 static int handle_commit_msg(void)
 572 {
 573         if (!cmitmsg)
 574                 return 0;
 575         do {
 576                 if (!memcmp("diff -", line, 6) ||
 577                     !memcmp("---", line, 3) ||
 578                     !memcmp("Index: ", line, 7))
 579                         break;
 580                 if ((multipart_boundary[0] && is_multipart_boundary(line))) {
 581                         /* We come here when the first part had only
 582                          * the commit message without any patch.  We
 583                          * pretend we have not seen this line yet, and
 584                          * go back to the loop.
 585                          */
 586                         return 1;
 587                 }
 588
 589                 /* Unwrap transfer encoding and optionally
 590                  * normalize the log message to UTF-8.
 591                  */
 592                 decode_transfer_encoding(line);
 593                 if (metainfo_utf8)
 594                         convert_to_utf8(line, charset);
 595                 fputs(line, cmitmsg);
 596         } while (fgets(line, sizeof(line), stdin) != NULL);
 597         fclose(cmitmsg);
 598         cmitmsg = NULL;
 599         return 0;
 600 }
 601
 602 /* We have done the commit message and have the first
 603  * line of the patch in line[].
 604  */
 605 static void handle_patch(void)
 606 {
 607         do {
 608                 if (multipart_boundary[0] && is_multipart_boundary(line))
 609                         break;
 610                 /* Only unwrap transfer encoding but otherwise do not
 611                  * do anything.  We do *NOT* want UTF-8 conversion
 612                  * here; we are dealing with the user payload.
 613                  */
 614                 decode_transfer_encoding(line);
 615                 fputs(line, patchfile);
 616                 patch_lines++;
 617         } while (fgets(line, sizeof(line), stdin) != NULL);
 618 }
 619
 620 /* multipart boundary and transfer encoding are set up for us, and we
 621  * are at the end of the sub header.  do equivalent of handle_body up
 622  * to the next boundary without closing patchfile --- we will expect
 623  * that the first part to contain commit message and a patch, and
 624  * handle other parts as pure patches.
 625  */
 626 static int handle_multipart_one_part(void)
 627 {
 628         int seen = 0;
 629         int n = 0;
 630         int len;
 631
 632         while (fgets(line, sizeof(line), stdin) != NULL) {
 633         again:
 634                 len = eatspace(line);
 635                 n++;
 636                 if (!len)
 637                         continue;
 638                 if (is_multipart_boundary(line))
 639                         break;
 640                 if (0 <= seen && handle_inbody_header(&seen, line))
 641                         continue;
 642                 seen = -1; /* no more inbody headers */
 643                 line[len] = '\n';
 644                 handle_info();
 645                 if (handle_commit_msg())
 646                         goto again;
 647                 handle_patch();
 648                 break;
 649         }
 650         if (n == 0)
 651                 return -1;
 652         return 0;
 653 }
 654
 655 static void handle_multipart_body(void)
 656 {
 657         int part_num = 0;
 658
 659         /* Skip up to the first boundary */
 660         while (fgets(line, sizeof(line), stdin) != NULL)
 661                 if (is_multipart_boundary(line)) {
 662                         part_num = 1;
 663                         break;
 664                 }
 665         if (!part_num)
 666                 return;
 667         /* We are on boundary line.  Start slurping the subhead. */
 668         while (1) {
 669                 int len = read_one_header_line(line, sizeof(line), stdin);
 670                 if (!len) {
 671                         if (handle_multipart_one_part() < 0)
 672                                 return;
 673                 }
 674                 else
 675                         check_subheader_line(line, len);
 676         }
 677         fclose(patchfile);
 678         if (!patch_lines) {
 679                 fprintf(stderr, "No patch found\n");
 680                 exit(1);
 681         }
 682 }
 683
 684 /* Non multipart message */
 685 static void handle_body(void)
 686 {
 687         int seen = 0;
 688
 689         while (fgets(line, sizeof(line), stdin) != NULL) {
 690                 int len = eatspace(line);
 691                 if (!len)
 692                         continue;
 693                 if (0 <= seen && handle_inbody_header(&seen, line))
 694                         continue;
 695                 seen = -1; /* no more inbody headers */
 696                 line[len] = '\n';
 697                 handle_info();
 698                 handle_commit_msg();
 699                 handle_patch();
 700                 break;
 701         }
 702         fclose(patchfile);
 703         if (!patch_lines) {
 704                 fprintf(stderr, "No patch found\n");
 705                 exit(1);
 706         }
 707 }
 708
 709 static const char mailinfo_usage[] =
 710         "git-mailinfo [-k] [-u] msg patch <mail >info";
 711
 712 static void usage(void) {
 713         fprintf(stderr, "%s\n", mailinfo_usage);
 714         exit(1);
 715 }
 716
 717 int main(int argc, char **argv)
 718 {
 719         while (1 < argc && argv[1][0] == '-') {
 720                 if (!strcmp(argv[1], "-k"))
 721                         keep_subject = 1;
 722                 else if (!strcmp(argv[1], "-u"))
 723                         metainfo_utf8 = 1;
 724                 else
 725                         usage();
 726                 argc--; argv++;
 727         }
 728
 729         if (argc != 3)
 730                 usage();
 731         cmitmsg = fopen(argv[1], "w");
 732         if (!cmitmsg) {
 733                 perror(argv[1]);
 734                 exit(1);
 735         }
 736         patchfile = fopen(argv[2], "w");
 737         if (!patchfile) {
 738                 perror(argv[2]);
 739                 exit(1);
 740         }
 741         while (1) {
 742                 int len = read_one_header_line(line, sizeof(line), stdin);
 743                 if (!len) {
 744                         if (multipart_boundary[0])
 745                                 handle_multipart_body();
 746                         else
 747                                 handle_body();
 748                         break;
 749                 }
 750                 check_header_line(line, len);
 751         }
 752         return 0;
 753 }