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