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 #include "git-compat-util.h"
  12 #include "cache.h"
  13
  14 static FILE *cmitmsg, *patchfile;
  15
  16 static int keep_subject = 0;
  17 static char *metainfo_charset = NULL;
  18 static char line[1000];
  19 static char date[1000];
  20 static char name[1000];
  21 static char email[1000];
  22 static char subject[1000];
  23
  24 static enum  {
  25         TE_DONTCARE, TE_QP, TE_BASE64,
  26 } transfer_encoding;
  27 static char charset[256];
  28
  29 static char multipart_boundary[1000];
  30 static int multipart_boundary_len;
  31 static int patch_lines = 0;
  32
  33 static char *sanity_check(char *name, char *email)
  34 {
  35         int len = strlen(name);
  36         if (len < 3 || len > 60)
  37                 return email;
  38         if (strchr(name, '@') || strchr(name, '<') || strchr(name, '>'))
  39                 return email;
  40         return name;
  41 }
  42
  43 static int handle_from(char *line)
  44 {
  45         char *at = strchr(line, '@');
  46         char *dst;
  47
  48         if (!at)
  49                 return 0;
  50
  51         /*
  52          * If we already have one email, don't take any confusing lines
  53          */
  54         if (*email && strchr(at+1, '@'))
  55                 return 0;
  56
  57         /* Pick up the string around '@', possibly delimited with <>
  58          * pair; that is the email part.  White them out while copying.
  59          */
  60         while (at > line) {
  61                 char c = at[-1];
  62                 if (isspace(c))
  63                         break;
  64                 if (c == '<') {
  65                         at[-1] = ' ';
  66                         break;
  67                 }
  68                 at--;
  69         }
  70         dst = email;
  71         for (;;) {
  72                 unsigned char c = *at;
  73                 if (!c || c == '>' || isspace(c)) {
  74                         if (c == '>')
  75                                 *at = ' ';
  76                         break;
  77                 }
  78                 *at++ = ' ';
  79                 *dst++ = c;
  80         }
  81         *dst++ = 0;
  82
  83         /* The remainder is name.  It could be "John Doe <john.doe@xz>"
  84          * or "john.doe@xz (John Doe)", but we have whited out the
  85          * email part, so trim from both ends, possibly removing
  86          * the () pair at the end.
  87          */
  88         at = line + strlen(line);
  89         while (at > line) {
  90                 unsigned char c = *--at;
  91                 if (!isspace(c)) {
  92                         at[(c == ')') ? 0 : 1] = 0;
  93                         break;
  94                 }
  95         }
  96
  97         at = line;
  98         for (;;) {
  99                 unsigned char c = *at;
 100                 if (!c || !isspace(c)) {
 101                         if (c == '(')
 102                                 at++;
 103                         break;
 104                 }
 105                 at++;
 106         }
 107         at = sanity_check(at, email);
 108         strcpy(name, at);
 109         return 1;
 110 }
 111
 112 static int handle_date(char *line)
 113 {
 114         strcpy(date, line);
 115         return 0;
 116 }
 117
 118 static int handle_subject(char *line)
 119 {
 120         strcpy(subject, line);
 121         return 0;
 122 }
 123
 124 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 125  * to have enough heuristics to grok MIME encoded patches often found
 126  * on our mailing lists.  For example, we do not even treat header lines
 127  * case insensitively.
 128  */
 129
 130 static int slurp_attr(const char *line, const char *name, char *attr)
 131 {
 132         char *ends, *ap = strcasestr(line, name);
 133         size_t sz;
 134
 135         if (!ap) {
 136                 *attr = 0;
 137                 return 0;
 138         }
 139         ap += strlen(name);
 140         if (*ap == '"') {
 141                 ap++;
 142                 ends = "\"";
 143         }
 144         else
 145                 ends = "; \t";
 146         sz = strcspn(ap, ends);
 147         memcpy(attr, ap, sz);
 148         attr[sz] = 0;
 149         return 1;
 150 }
 151
 152 static int handle_subcontent_type(char *line)
 153 {
 154         /* We do not want to mess with boundary.  Note that we do not
 155          * handle nested multipart.
 156          */
 157         if (strcasestr(line, "boundary=")) {
 158                 fprintf(stderr, "Not handling nested multipart message.\n");
 159                 exit(1);
 160         }
 161         slurp_attr(line, "charset=", charset);
 162         if (*charset) {
 163                 int i, c;
 164                 for (i = 0; (c = charset[i]) != 0; i++)
 165                         charset[i] = tolower(c);
 166         }
 167         return 0;
 168 }
 169
 170 static int handle_content_type(char *line)
 171 {
 172         *multipart_boundary = 0;
 173         if (slurp_attr(line, "boundary=", multipart_boundary + 2)) {
 174                 memcpy(multipart_boundary, "--", 2);
 175                 multipart_boundary_len = strlen(multipart_boundary);
 176         }
 177         slurp_attr(line, "charset=", charset);
 178         return 0;
 179 }
 180
 181 static int handle_content_transfer_encoding(char *line)
 182 {
 183         if (strcasestr(line, "base64"))
 184                 transfer_encoding = TE_BASE64;
 185         else if (strcasestr(line, "quoted-printable"))
 186                 transfer_encoding = TE_QP;
 187         else
 188                 transfer_encoding = TE_DONTCARE;
 189         return 0;
 190 }
 191
 192 static int is_multipart_boundary(const char *line)
 193 {
 194         return (!memcmp(line, multipart_boundary, multipart_boundary_len));
 195 }
 196
 197 static int eatspace(char *line)
 198 {
 199         int len = strlen(line);
 200         while (len > 0 && isspace(line[len-1]))
 201                 line[--len] = 0;
 202         return len;
 203 }
 204
 205 #define SEEN_FROM 01
 206 #define SEEN_DATE 02
 207 #define SEEN_SUBJECT 04
 208
 209 /* First lines of body can have From:, Date:, and Subject: */
 210 static int handle_inbody_header(int *seen, char *line)
 211 {
 212         if (!memcmp("From:", line, 5) && isspace(line[5])) {
 213                 if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
 214                         *seen |= SEEN_FROM;
 215                         return 1;
 216                 }
 217         }
 218         if (!memcmp("Date:", line, 5) && isspace(line[5])) {
 219                 if (!(*seen & SEEN_DATE)) {
 220                         handle_date(line+6);
 221                         *seen |= SEEN_DATE;
 222                         return 1;
 223                 }
 224         }
 225         if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
 226                 if (!(*seen & SEEN_SUBJECT)) {
 227                         handle_subject(line+9);
 228                         *seen |= SEEN_SUBJECT;
 229                         return 1;
 230                 }
 231         }
 232         if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
 233                 if (!(*seen & SEEN_SUBJECT)) {
 234                         handle_subject(line);
 235                         *seen |= SEEN_SUBJECT;
 236                         return 1;
 237                 }
 238         }
 239         return 0;
 240 }
 241
 242 static char *cleanup_subject(char *subject)
 243 {
 244         if (keep_subject)
 245                 return subject;
 246         for (;;) {
 247                 char *p;
 248                 int len, remove;
 249                 switch (*subject) {
 250                 case 'r': case 'R':
 251                         if (!memcmp("e:", subject+1, 2)) {
 252                                 subject +=3;
 253                                 continue;
 254                         }
 255                         break;
 256                 case ' ': case '\t': case ':':
 257                         subject++;
 258                         continue;
 259
 260                 case '[':
 261                         p = strchr(subject, ']');
 262                         if (!p) {
 263                                 subject++;
 264                                 continue;
 265                         }
 266                         len = strlen(p);
 267                         remove = p - subject;
 268                         if (remove <= len *2) {
 269                                 subject = p+1;
 270                                 continue;
 271                         }
 272                         break;
 273                 }
 274                 return subject;
 275         }
 276 }
 277
 278 static void cleanup_space(char *buf)
 279 {
 280         unsigned char c;
 281         while ((c = *buf) != 0) {
 282                 buf++;
 283                 if (isspace(c)) {
 284                         buf[-1] = ' ';
 285                         c = *buf;
 286                         while (isspace(c)) {
 287                                 int len = strlen(buf);
 288                                 memmove(buf, buf+1, len);
 289                                 c = *buf;
 290                         }
 291                 }
 292         }
 293 }
 294
 295 typedef int (*header_fn_t)(char *);
 296 struct header_def {
 297         const char *name;
 298         header_fn_t func;
 299         int namelen;
 300 };
 301
 302 static void check_header(char *line, int len, struct header_def *header)
 303 {
 304         int i;
 305
 306         if (header[0].namelen <= 0) {
 307                 for (i = 0; header[i].name; i++)
 308                         header[i].namelen = strlen(header[i].name);
 309         }
 310         for (i = 0; header[i].name; i++) {
 311                 int len = header[i].namelen;
 312                 if (!strncasecmp(line, header[i].name, len) &&
 313                     line[len] == ':' && isspace(line[len + 1])) {
 314                         header[i].func(line + len + 2);
 315                         break;
 316                 }
 317         }
 318 }
 319
 320 static void check_subheader_line(char *line, int len)
 321 {
 322         static struct header_def header[] = {
 323                 { "Content-Type", handle_subcontent_type },
 324                 { "Content-Transfer-Encoding",
 325                   handle_content_transfer_encoding },
 326                 { NULL },
 327         };
 328         check_header(line, len, header);
 329 }
 330 static void check_header_line(char *line, int len)
 331 {
 332         static struct header_def header[] = {
 333                 { "From", handle_from },
 334                 { "Date", handle_date },
 335                 { "Subject", handle_subject },
 336                 { "Content-Type", handle_content_type },
 337                 { "Content-Transfer-Encoding",
 338                   handle_content_transfer_encoding },
 339                 { NULL },
 340         };
 341         check_header(line, len, header);
 342 }
 343
 344 static int read_one_header_line(char *line, int sz, FILE *in)
 345 {
 346         int ofs = 0;
 347         while (ofs < sz) {
 348                 int peek, len;
 349                 if (fgets(line + ofs, sz - ofs, in) == NULL)
 350                         return ofs;
 351                 len = eatspace(line + ofs);
 352                 if (len == 0)
 353                         return ofs;
 354                 peek = fgetc(in); ungetc(peek, in);
 355                 if (peek == ' ' || peek == '\t') {
 356                         /* Yuck, 2822 header "folding" */
 357                         ofs += len;
 358                         continue;
 359                 }
 360                 return ofs + len;
 361         }
 362         return ofs;
 363 }
 364
 365 static unsigned hexval(int c)
 366 {
 367         if (c >= '0' && c <= '9')
 368                 return c - '0';
 369         if (c >= 'a' && c <= 'f')
 370                 return c - 'a' + 10;
 371         if (c >= 'A' && c <= 'F')
 372                 return c - 'A' + 10;
 373         return ~0;
 374 }
 375
 376 static int decode_q_segment(char *in, char *ot, char *ep)
 377 {
 378         int c;
 379         while ((c = *in++) != 0 && (in <= ep)) {
 380                 if (c == '=') {
 381                         int d = *in++;
 382                         if (d == '\n' || !d)
 383                                 break; /* drop trailing newline */
 384                         *ot++ = ((hexval(d) << 4) | hexval(*in++));
 385                 }
 386                 else
 387                         *ot++ = c;
 388         }
 389         *ot = 0;
 390         return 0;
 391 }
 392
 393 static int decode_b_segment(char *in, char *ot, char *ep)
 394 {
 395         /* Decode in..ep, possibly in-place to ot */
 396         int c, pos = 0, acc = 0;
 397
 398         while ((c = *in++) != 0 && (in <= ep)) {
 399                 if (c == '+')
 400                         c = 62;
 401                 else if (c == '/')
 402                         c = 63;
 403                 else if ('A' <= c && c <= 'Z')
 404                         c -= 'A';
 405                 else if ('a' <= c && c <= 'z')
 406                         c -= 'a' - 26;
 407                 else if ('0' <= c && c <= '9')
 408                         c -= '0' - 52;
 409                 else if (c == '=') {
 410                         /* padding is almost like (c == 0), except we do
 411                          * not output NUL resulting only from it;
 412                          * for now we just trust the data.
 413                          */
 414                         c = 0;
 415                 }
 416                 else
 417                         continue; /* garbage */
 418                 switch (pos++) {
 419                 case 0:
 420                         acc = (c << 2);
 421                         break;
 422                 case 1:
 423                         *ot++ = (acc | (c >> 4));
 424                         acc = (c & 15) << 4;
 425                         break;
 426                 case 2:
 427                         *ot++ = (acc | (c >> 2));
 428                         acc = (c & 3) << 6;
 429                         break;
 430                 case 3:
 431                         *ot++ = (acc | c);
 432                         acc = pos = 0;
 433                         break;
 434                 }
 435         }
 436         *ot = 0;
 437         return 0;
 438 }
 439
 440 static void convert_to_utf8(char *line, char *charset)
 441 {
 442         char *in, *out;
 443         size_t insize, outsize, nrc;
 444         char outbuf[4096]; /* cheat */
 445         static char latin_one[] = "latin-1";
 446         char *input_charset = *charset ? charset : latin_one;
 447         iconv_t conv = iconv_open(metainfo_charset, input_charset);
 448
 449         if (conv == (iconv_t) -1) {
 450                 static int warned_latin1_once = 0;
 451                 if (input_charset != latin_one) {
 452                         fprintf(stderr, "cannot convert from %s to %s\n",
 453                                 input_charset, metainfo_charset);
 454                         *charset = 0;
 455                 }
 456                 else if (!warned_latin1_once) {
 457                         warned_latin1_once = 1;
 458                         fprintf(stderr, "tried to convert from %s to %s, "
 459                                 "but your iconv does not work with it.\n",
 460                                 input_charset, metainfo_charset);
 461                 }
 462                 return;
 463         }
 464         in = line;
 465         insize = strlen(in);
 466         out = outbuf;
 467         outsize = sizeof(outbuf);
 468         nrc = iconv(conv, &in, &insize, &out, &outsize);
 469         iconv_close(conv);
 470         if (nrc == (size_t) -1)
 471                 return;
 472         *out = 0;
 473         strcpy(line, outbuf);
 474 }
 475
 476 static void decode_header_bq(char *it)
 477 {
 478         char *in, *out, *ep, *cp, *sp;
 479         char outbuf[1000];
 480
 481         in = it;
 482         out = outbuf;
 483         while ((ep = strstr(in, "=?")) != NULL) {
 484                 int sz, encoding;
 485                 char charset_q[256], piecebuf[256];
 486                 if (in != ep) {
 487                         sz = ep - in;
 488                         memcpy(out, in, sz);
 489                         out += sz;
 490                         in += sz;
 491                 }
 492                 /* E.g.
 493                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 494                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 495                  */
 496                 ep += 2;
 497                 cp = strchr(ep, '?');
 498                 if (!cp)
 499                         return; /* no munging */
 500                 for (sp = ep; sp < cp; sp++)
 501                         charset_q[sp - ep] = tolower(*sp);
 502                 charset_q[cp - ep] = 0;
 503                 encoding = cp[1];
 504                 if (!encoding || cp[2] != '?')
 505                         return; /* no munging */
 506                 ep = strstr(cp + 3, "?=");
 507                 if (!ep)
 508                         return; /* no munging */
 509                 switch (tolower(encoding)) {
 510                 default:
 511                         return; /* no munging */
 512                 case 'b':
 513                         sz = decode_b_segment(cp + 3, piecebuf, ep);
 514                         break;
 515                 case 'q':
 516                         sz = decode_q_segment(cp + 3, piecebuf, ep);
 517                         break;
 518                 }
 519                 if (sz < 0)
 520                         return;
 521                 if (metainfo_charset)
 522                         convert_to_utf8(piecebuf, charset_q);
 523                 strcpy(out, piecebuf);
 524                 out += strlen(out);
 525                 in = ep + 2;
 526         }
 527         strcpy(out, in);
 528         strcpy(it, outbuf);
 529 }
 530
 531 static void decode_transfer_encoding(char *line)
 532 {
 533         char *ep;
 534
 535         switch (transfer_encoding) {
 536         case TE_QP:
 537                 ep = line + strlen(line);
 538                 decode_q_segment(line, line, ep);
 539                 break;
 540         case TE_BASE64:
 541                 ep = line + strlen(line);
 542                 decode_b_segment(line, line, ep);
 543                 break;
 544         case TE_DONTCARE:
 545                 break;
 546         }
 547 }
 548
 549 static void handle_info(void)
 550 {
 551         char *sub;
 552         static int done_info = 0;
 553
 554         if (done_info)
 555                 return;
 556
 557         done_info = 1;
 558         sub = cleanup_subject(subject);
 559         cleanup_space(name);
 560         cleanup_space(date);
 561         cleanup_space(email);
 562         cleanup_space(sub);
 563
 564         /* Unwrap inline B and Q encoding, and optionally
 565          * normalize the meta information to utf8.
 566          */
 567         decode_header_bq(name);
 568         decode_header_bq(date);
 569         decode_header_bq(email);
 570         decode_header_bq(sub);
 571         printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
 572                name, email, sub, date);
 573 }
 574
 575 /* We are inside message body and have read line[] already.
 576  * Spit out the commit log.
 577  */
 578 static int handle_commit_msg(void)
 579 {
 580         if (!cmitmsg)
 581                 return 0;
 582         do {
 583                 if (!memcmp("diff -", line, 6) ||
 584                     !memcmp("---", line, 3) ||
 585                     !memcmp("Index: ", line, 7))
 586                         break;
 587                 if ((multipart_boundary[0] && is_multipart_boundary(line))) {
 588                         /* We come here when the first part had only
 589                          * the commit message without any patch.  We
 590                          * pretend we have not seen this line yet, and
 591                          * go back to the loop.
 592                          */
 593                         return 1;
 594                 }
 595
 596                 /* Unwrap transfer encoding and optionally
 597                  * normalize the log message to UTF-8.
 598                  */
 599                 decode_transfer_encoding(line);
 600                 if (metainfo_charset)
 601                         convert_to_utf8(line, charset);
 602                 fputs(line, cmitmsg);
 603         } while (fgets(line, sizeof(line), stdin) != NULL);
 604         fclose(cmitmsg);
 605         cmitmsg = NULL;
 606         return 0;
 607 }
 608
 609 /* We have done the commit message and have the first
 610  * line of the patch in line[].
 611  */
 612 static void handle_patch(void)
 613 {
 614         do {
 615                 if (multipart_boundary[0] && is_multipart_boundary(line))
 616                         break;
 617                 /* Only unwrap transfer encoding but otherwise do not
 618                  * do anything.  We do *NOT* want UTF-8 conversion
 619                  * here; we are dealing with the user payload.
 620                  */
 621                 decode_transfer_encoding(line);
 622                 fputs(line, patchfile);
 623                 patch_lines++;
 624         } while (fgets(line, sizeof(line), stdin) != NULL);
 625 }
 626
 627 /* multipart boundary and transfer encoding are set up for us, and we
 628  * are at the end of the sub header.  do equivalent of handle_body up
 629  * to the next boundary without closing patchfile --- we will expect
 630  * that the first part to contain commit message and a patch, and
 631  * handle other parts as pure patches.
 632  */
 633 static int handle_multipart_one_part(void)
 634 {
 635         int seen = 0;
 636         int n = 0;
 637         int len;
 638
 639         while (fgets(line, sizeof(line), stdin) != NULL) {
 640         again:
 641                 len = eatspace(line);
 642                 n++;
 643                 if (!len)
 644                         continue;
 645                 if (is_multipart_boundary(line))
 646                         break;
 647                 if (0 <= seen && handle_inbody_header(&seen, line))
 648                         continue;
 649                 seen = -1; /* no more inbody headers */
 650                 line[len] = '\n';
 651                 handle_info();
 652                 if (handle_commit_msg())
 653                         goto again;
 654                 handle_patch();
 655                 break;
 656         }
 657         if (n == 0)
 658                 return -1;
 659         return 0;
 660 }
 661
 662 static void handle_multipart_body(void)
 663 {
 664         int part_num = 0;
 665
 666         /* Skip up to the first boundary */
 667         while (fgets(line, sizeof(line), stdin) != NULL)
 668                 if (is_multipart_boundary(line)) {
 669                         part_num = 1;
 670                         break;
 671                 }
 672         if (!part_num)
 673                 return;
 674         /* We are on boundary line.  Start slurping the subhead. */
 675         while (1) {
 676                 int len = read_one_header_line(line, sizeof(line), stdin);
 677                 if (!len) {
 678                         if (handle_multipart_one_part() < 0)
 679                                 return;
 680                 }
 681                 else
 682                         check_subheader_line(line, len);
 683         }
 684         fclose(patchfile);
 685         if (!patch_lines) {
 686                 fprintf(stderr, "No patch found\n");
 687                 exit(1);
 688         }
 689 }
 690
 691 /* Non multipart message */
 692 static void handle_body(void)
 693 {
 694         int seen = 0;
 695
 696         while (fgets(line, sizeof(line), stdin) != NULL) {
 697                 int len = eatspace(line);
 698                 if (!len)
 699                         continue;
 700                 if (0 <= seen && handle_inbody_header(&seen, line))
 701                         continue;
 702                 seen = -1; /* no more inbody headers */
 703                 line[len] = '\n';
 704                 handle_info();
 705                 handle_commit_msg();
 706                 handle_patch();
 707                 break;
 708         }
 709         fclose(patchfile);
 710         if (!patch_lines) {
 711                 fprintf(stderr, "No patch found\n");
 712                 exit(1);
 713         }
 714 }
 715
 716 static const char mailinfo_usage[] =
 717         "git-mailinfo [-k] [-u | --encoding=<encoding>] msg patch <mail >info";
 718
 719 int main(int argc, char **argv)
 720 {
 721         /* NEEDSWORK: might want to do the optional .git/ directory
 722          * discovery
 723          */
 724         git_config(git_default_config);
 725
 726         while (1 < argc && argv[1][0] == '-') {
 727                 if (!strcmp(argv[1], "-k"))
 728                         keep_subject = 1;
 729                 else if (!strcmp(argv[1], "-u"))
 730                         metainfo_charset = git_commit_encoding;
 731                 else if (!strncmp(argv[1], "--encoding=", 11))
 732                         metainfo_charset = argv[1] + 11;
 733                 else
 734                         usage(mailinfo_usage);
 735                 argc--; argv++;
 736         }
 737
 738         if (argc != 3)
 739                 usage(mailinfo_usage);
 740         cmitmsg = fopen(argv[1], "w");
 741         if (!cmitmsg) {
 742                 perror(argv[1]);
 743                 exit(1);
 744         }
 745         patchfile = fopen(argv[2], "w");
 746         if (!patchfile) {
 747                 perror(argv[2]);
 748                 exit(1);
 749         }
 750         while (1) {
 751                 int len = read_one_header_line(line, sizeof(line), stdin);
 752                 if (!len) {
 753                         if (multipart_boundary[0])
 754                                 handle_multipart_body();
 755                         else
 756                                 handle_body();
 757                         break;
 758                 }
 759                 check_header_line(line, len);
 760         }
 761         return 0;
 762 }