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 #ifndef NO_ICONV
  11 #include <iconv.h>
  12 #endif
  13 #include "git-compat-util.h"
  14 #include "cache.h"
  15
  16 static FILE *cmitmsg, *patchfile;
  17
  18 static int keep_subject = 0;
  19 static char *metainfo_charset = NULL;
  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 bogus_from(char *line)
  46 {
  47         /* John Doe <johndoe> */
  48         char *bra, *ket, *dst, *cp;
  49
  50         /* This is fallback, so do not bother if we already have an
  51          * e-mail address.
  52          */
  53         if (*email)
  54                 return 0;
  55
  56         bra = strchr(line, '<');
  57         if (!bra)
  58                 return 0;
  59         ket = strchr(bra, '>');
  60         if (!ket)
  61                 return 0;
  62
  63         for (dst = email, cp = bra+1; cp < ket; )
  64                 *dst++ = *cp++;
  65         *dst = 0;
  66         for (cp = line; isspace(*cp); cp++)
  67                 ;
  68         for (bra--; isspace(*bra); bra--)
  69                 *bra = 0;
  70         cp = sanity_check(cp, email);
  71         strcpy(name, cp);
  72         return 1;
  73 }
  74
  75 static int handle_from(char *line)
  76 {
  77         char *at = strchr(line, '@');
  78         char *dst;
  79
  80         if (!at)
  81                 return bogus_from(line);
  82
  83         /*
  84          * If we already have one email, don't take any confusing lines
  85          */
  86         if (*email && strchr(at+1, '@'))
  87                 return 0;
  88
  89         /* Pick up the string around '@', possibly delimited with <>
  90          * pair; that is the email part.  White them out while copying.
  91          */
  92         while (at > line) {
  93                 char c = at[-1];
  94                 if (isspace(c))
  95                         break;
  96                 if (c == '<') {
  97                         at[-1] = ' ';
  98                         break;
  99                 }
 100                 at--;
 101         }
 102         dst = email;
 103         for (;;) {
 104                 unsigned char c = *at;
 105                 if (!c || c == '>' || isspace(c)) {
 106                         if (c == '>')
 107                                 *at = ' ';
 108                         break;
 109                 }
 110                 *at++ = ' ';
 111                 *dst++ = c;
 112         }
 113         *dst++ = 0;
 114
 115         /* The remainder is name.  It could be "John Doe <john.doe@xz>"
 116          * or "john.doe@xz (John Doe)", but we have whited out the
 117          * email part, so trim from both ends, possibly removing
 118          * the () pair at the end.
 119          */
 120         at = line + strlen(line);
 121         while (at > line) {
 122                 unsigned char c = *--at;
 123                 if (!isspace(c)) {
 124                         at[(c == ')') ? 0 : 1] = 0;
 125                         break;
 126                 }
 127         }
 128
 129         at = line;
 130         for (;;) {
 131                 unsigned char c = *at;
 132                 if (!c || !isspace(c)) {
 133                         if (c == '(')
 134                                 at++;
 135                         break;
 136                 }
 137                 at++;
 138         }
 139         at = sanity_check(at, email);
 140         strcpy(name, at);
 141         return 1;
 142 }
 143
 144 static int handle_date(char *line)
 145 {
 146         strcpy(date, line);
 147         return 0;
 148 }
 149
 150 static int handle_subject(char *line)
 151 {
 152         strcpy(subject, line);
 153         return 0;
 154 }
 155
 156 /* NOTE NOTE NOTE.  We do not claim we do full MIME.  We just attempt
 157  * to have enough heuristics to grok MIME encoded patches often found
 158  * on our mailing lists.  For example, we do not even treat header lines
 159  * case insensitively.
 160  */
 161
 162 static int slurp_attr(const char *line, const char *name, char *attr)
 163 {
 164         char *ends, *ap = strcasestr(line, name);
 165         size_t sz;
 166
 167         if (!ap) {
 168                 *attr = 0;
 169                 return 0;
 170         }
 171         ap += strlen(name);
 172         if (*ap == '"') {
 173                 ap++;
 174                 ends = "\"";
 175         }
 176         else
 177                 ends = "; \t";
 178         sz = strcspn(ap, ends);
 179         memcpy(attr, ap, sz);
 180         attr[sz] = 0;
 181         return 1;
 182 }
 183
 184 static int handle_subcontent_type(char *line)
 185 {
 186         /* We do not want to mess with boundary.  Note that we do not
 187          * handle nested multipart.
 188          */
 189         if (strcasestr(line, "boundary=")) {
 190                 fprintf(stderr, "Not handling nested multipart message.\n");
 191                 exit(1);
 192         }
 193         slurp_attr(line, "charset=", charset);
 194         if (*charset) {
 195                 int i, c;
 196                 for (i = 0; (c = charset[i]) != 0; i++)
 197                         charset[i] = tolower(c);
 198         }
 199         return 0;
 200 }
 201
 202 static int handle_content_type(char *line)
 203 {
 204         *multipart_boundary = 0;
 205         if (slurp_attr(line, "boundary=", multipart_boundary + 2)) {
 206                 memcpy(multipart_boundary, "--", 2);
 207                 multipart_boundary_len = strlen(multipart_boundary);
 208         }
 209         slurp_attr(line, "charset=", charset);
 210         return 0;
 211 }
 212
 213 static int handle_content_transfer_encoding(char *line)
 214 {
 215         if (strcasestr(line, "base64"))
 216                 transfer_encoding = TE_BASE64;
 217         else if (strcasestr(line, "quoted-printable"))
 218                 transfer_encoding = TE_QP;
 219         else
 220                 transfer_encoding = TE_DONTCARE;
 221         return 0;
 222 }
 223
 224 static int is_multipart_boundary(const char *line)
 225 {
 226         return (!memcmp(line, multipart_boundary, multipart_boundary_len));
 227 }
 228
 229 static int eatspace(char *line)
 230 {
 231         int len = strlen(line);
 232         while (len > 0 && isspace(line[len-1]))
 233                 line[--len] = 0;
 234         return len;
 235 }
 236
 237 #define SEEN_FROM 01
 238 #define SEEN_DATE 02
 239 #define SEEN_SUBJECT 04
 240 #define SEEN_PREFIX  0x08
 241
 242 /* First lines of body can have From:, Date:, and Subject: */
 243 static void handle_inbody_header(int *seen, char *line)
 244 {
 245         if (*seen & SEEN_PREFIX)
 246                 return;
 247         if (!memcmp("From:", line, 5) && isspace(line[5])) {
 248                 if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
 249                         *seen |= SEEN_FROM;
 250                         return;
 251                 }
 252         }
 253         if (!memcmp("Date:", line, 5) && isspace(line[5])) {
 254                 if (!(*seen & SEEN_DATE)) {
 255                         handle_date(line+6);
 256                         *seen |= SEEN_DATE;
 257                         return;
 258                 }
 259         }
 260         if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
 261                 if (!(*seen & SEEN_SUBJECT)) {
 262                         handle_subject(line+9);
 263                         *seen |= SEEN_SUBJECT;
 264                         return;
 265                 }
 266         }
 267         if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
 268                 if (!(*seen & SEEN_SUBJECT)) {
 269                         handle_subject(line);
 270                         *seen |= SEEN_SUBJECT;
 271                         return;
 272                 }
 273         }
 274         *seen |= SEEN_PREFIX;
 275 }
 276
 277 static char *cleanup_subject(char *subject)
 278 {
 279         if (keep_subject)
 280                 return subject;
 281         for (;;) {
 282                 char *p;
 283                 int len, remove;
 284                 switch (*subject) {
 285                 case 'r': case 'R':
 286                         if (!memcmp("e:", subject+1, 2)) {
 287                                 subject +=3;
 288                                 continue;
 289                         }
 290                         break;
 291                 case ' ': case '\t': case ':':
 292                         subject++;
 293                         continue;
 294
 295                 case '[':
 296                         p = strchr(subject, ']');
 297                         if (!p) {
 298                                 subject++;
 299                                 continue;
 300                         }
 301                         len = strlen(p);
 302                         remove = p - subject;
 303                         if (remove <= len *2) {
 304                                 subject = p+1;
 305                                 continue;
 306                         }
 307                         break;
 308                 }
 309                 return subject;
 310         }
 311 }
 312
 313 static void cleanup_space(char *buf)
 314 {
 315         unsigned char c;
 316         while ((c = *buf) != 0) {
 317                 buf++;
 318                 if (isspace(c)) {
 319                         buf[-1] = ' ';
 320                         c = *buf;
 321                         while (isspace(c)) {
 322                                 int len = strlen(buf);
 323                                 memmove(buf, buf+1, len);
 324                                 c = *buf;
 325                         }
 326                 }
 327         }
 328 }
 329
 330 static void decode_header_bq(char *it);
 331 typedef int (*header_fn_t)(char *);
 332 struct header_def {
 333         const char *name;
 334         header_fn_t func;
 335         int namelen;
 336 };
 337
 338 static void check_header(char *line, struct header_def *header)
 339 {
 340         int i;
 341
 342         if (header[0].namelen <= 0) {
 343                 for (i = 0; header[i].name; i++)
 344                         header[i].namelen = strlen(header[i].name);
 345         }
 346         for (i = 0; header[i].name; i++) {
 347                 int len = header[i].namelen;
 348                 if (!strncasecmp(line, header[i].name, len) &&
 349                     line[len] == ':' && isspace(line[len + 1])) {
 350                         /* Unwrap inline B and Q encoding, and optionally
 351                          * normalize the meta information to utf8.
 352                          */
 353                         decode_header_bq(line + len + 2);
 354                         header[i].func(line + len + 2);
 355                         break;
 356                 }
 357         }
 358 }
 359
 360 static void check_subheader_line(char *line)
 361 {
 362         static struct header_def header[] = {
 363                 { "Content-Type", handle_subcontent_type },
 364                 { "Content-Transfer-Encoding",
 365                   handle_content_transfer_encoding },
 366                 { NULL },
 367         };
 368         check_header(line, header);
 369 }
 370 static void check_header_line(char *line)
 371 {
 372         static struct header_def header[] = {
 373                 { "From", handle_from },
 374                 { "Date", handle_date },
 375                 { "Subject", handle_subject },
 376                 { "Content-Type", handle_content_type },
 377                 { "Content-Transfer-Encoding",
 378                   handle_content_transfer_encoding },
 379                 { NULL },
 380         };
 381         check_header(line, header);
 382 }
 383
 384 static int read_one_header_line(char *line, int sz, FILE *in)
 385 {
 386         int ofs = 0;
 387         while (ofs < sz) {
 388                 int peek, len;
 389                 if (fgets(line + ofs, sz - ofs, in) == NULL)
 390                         return ofs;
 391                 len = eatspace(line + ofs);
 392                 if (len == 0)
 393                         return ofs;
 394                 peek = fgetc(in); ungetc(peek, in);
 395                 if (peek == ' ' || peek == '\t') {
 396                         /* Yuck, 2822 header "folding" */
 397                         ofs += len;
 398                         continue;
 399                 }
 400                 return ofs + len;
 401         }
 402         return ofs;
 403 }
 404
 405 static unsigned hexval(int c)
 406 {
 407         if (c >= '0' && c <= '9')
 408                 return c - '0';
 409         if (c >= 'a' && c <= 'f')
 410                 return c - 'a' + 10;
 411         if (c >= 'A' && c <= 'F')
 412                 return c - 'A' + 10;
 413         return ~0;
 414 }
 415
 416 static int decode_q_segment(char *in, char *ot, char *ep, int rfc2047)
 417 {
 418         int c;
 419         while ((c = *in++) != 0 && (in <= ep)) {
 420                 if (c == '=') {
 421                         int d = *in++;
 422                         if (d == '\n' || !d)
 423                                 break; /* drop trailing newline */
 424                         *ot++ = ((hexval(d) << 4) | hexval(*in++));
 425                         continue;
 426                 }
 427                 if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 428                         c = 0x20;
 429                 *ot++ = c;
 430         }
 431         *ot = 0;
 432         return 0;
 433 }
 434
 435 static int decode_b_segment(char *in, char *ot, char *ep)
 436 {
 437         /* Decode in..ep, possibly in-place to ot */
 438         int c, pos = 0, acc = 0;
 439
 440         while ((c = *in++) != 0 && (in <= ep)) {
 441                 if (c == '+')
 442                         c = 62;
 443                 else if (c == '/')
 444                         c = 63;
 445                 else if ('A' <= c && c <= 'Z')
 446                         c -= 'A';
 447                 else if ('a' <= c && c <= 'z')
 448                         c -= 'a' - 26;
 449                 else if ('0' <= c && c <= '9')
 450                         c -= '0' - 52;
 451                 else if (c == '=') {
 452                         /* padding is almost like (c == 0), except we do
 453                          * not output NUL resulting only from it;
 454                          * for now we just trust the data.
 455                          */
 456                         c = 0;
 457                 }
 458                 else
 459                         continue; /* garbage */
 460                 switch (pos++) {
 461                 case 0:
 462                         acc = (c << 2);
 463                         break;
 464                 case 1:
 465                         *ot++ = (acc | (c >> 4));
 466                         acc = (c & 15) << 4;
 467                         break;
 468                 case 2:
 469                         *ot++ = (acc | (c >> 2));
 470                         acc = (c & 3) << 6;
 471                         break;
 472                 case 3:
 473                         *ot++ = (acc | c);
 474                         acc = pos = 0;
 475                         break;
 476                 }
 477         }
 478         *ot = 0;
 479         return 0;
 480 }
 481
 482 static void convert_to_utf8(char *line, char *charset)
 483 {
 484 #ifndef NO_ICONV
 485         char *in, *out;
 486         size_t insize, outsize, nrc;
 487         char outbuf[4096]; /* cheat */
 488         static char latin_one[] = "latin1";
 489         char *input_charset = *charset ? charset : latin_one;
 490         iconv_t conv = iconv_open(metainfo_charset, input_charset);
 491
 492         if (conv == (iconv_t) -1) {
 493                 static int warned_latin1_once = 0;
 494                 if (input_charset != latin_one) {
 495                         fprintf(stderr, "cannot convert from %s to %s\n",
 496                                 input_charset, metainfo_charset);
 497                         *charset = 0;
 498                 }
 499                 else if (!warned_latin1_once) {
 500                         warned_latin1_once = 1;
 501                         fprintf(stderr, "tried to convert from %s to %s, "
 502                                 "but your iconv does not work with it.\n",
 503                                 input_charset, metainfo_charset);
 504                 }
 505                 return;
 506         }
 507         in = line;
 508         insize = strlen(in);
 509         out = outbuf;
 510         outsize = sizeof(outbuf);
 511         nrc = iconv(conv, &in, &insize, &out, &outsize);
 512         iconv_close(conv);
 513         if (nrc == (size_t) -1)
 514                 return;
 515         *out = 0;
 516         strcpy(line, outbuf);
 517 #endif
 518 }
 519
 520 static void decode_header_bq(char *it)
 521 {
 522         char *in, *out, *ep, *cp, *sp;
 523         char outbuf[1000];
 524
 525         in = it;
 526         out = outbuf;
 527         while ((ep = strstr(in, "=?")) != NULL) {
 528                 int sz, encoding;
 529                 char charset_q[256], piecebuf[256];
 530                 if (in != ep) {
 531                         sz = ep - in;
 532                         memcpy(out, in, sz);
 533                         out += sz;
 534                         in += sz;
 535                 }
 536                 /* E.g.
 537                  * ep : "=?iso-2022-jp?B?GyR...?= foo"
 538                  * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 539                  */
 540                 ep += 2;
 541                 cp = strchr(ep, '?');
 542                 if (!cp)
 543                         return; /* no munging */
 544                 for (sp = ep; sp < cp; sp++)
 545                         charset_q[sp - ep] = tolower(*sp);
 546                 charset_q[cp - ep] = 0;
 547                 encoding = cp[1];
 548                 if (!encoding || cp[2] != '?')
 549                         return; /* no munging */
 550                 ep = strstr(cp + 3, "?=");
 551                 if (!ep)
 552                         return; /* no munging */
 553                 switch (tolower(encoding)) {
 554                 default:
 555                         return; /* no munging */
 556                 case 'b':
 557                         sz = decode_b_segment(cp + 3, piecebuf, ep);
 558                         break;
 559                 case 'q':
 560                         sz = decode_q_segment(cp + 3, piecebuf, ep, 1);
 561                         break;
 562                 }
 563                 if (sz < 0)
 564                         return;
 565                 if (metainfo_charset)
 566                         convert_to_utf8(piecebuf, charset_q);
 567                 strcpy(out, piecebuf);
 568                 out += strlen(out);
 569                 in = ep + 2;
 570         }
 571         strcpy(out, in);
 572         strcpy(it, outbuf);
 573 }
 574
 575 static void decode_transfer_encoding(char *line)
 576 {
 577         char *ep;
 578
 579         switch (transfer_encoding) {
 580         case TE_QP:
 581                 ep = line + strlen(line);
 582                 decode_q_segment(line, line, ep, 0);
 583                 break;
 584         case TE_BASE64:
 585                 ep = line + strlen(line);
 586                 decode_b_segment(line, line, ep);
 587                 break;
 588         case TE_DONTCARE:
 589                 break;
 590         }
 591 }
 592
 593 static void handle_info(void)
 594 {
 595         char *sub;
 596
 597         sub = cleanup_subject(subject);
 598         cleanup_space(name);
 599         cleanup_space(date);
 600         cleanup_space(email);
 601         cleanup_space(sub);
 602
 603         printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
 604                name, email, sub, date);
 605 }
 606
 607 /* We are inside message body and have read line[] already.
 608  * Spit out the commit log.
 609  */
 610 static int handle_commit_msg(int *seen)
 611 {
 612         if (!cmitmsg)
 613                 return 0;
 614         do {
 615                 if (!memcmp("diff -", line, 6) ||
 616                     !memcmp("---", line, 3) ||
 617                     !memcmp("Index: ", line, 7))
 618                         break;
 619                 if ((multipart_boundary[0] && is_multipart_boundary(line))) {
 620                         /* We come here when the first part had only
 621                          * the commit message without any patch.  We
 622                          * pretend we have not seen this line yet, and
 623                          * go back to the loop.
 624                          */
 625                         return 1;
 626                 }
 627
 628                 /* Unwrap transfer encoding and optionally
 629                  * normalize the log message to UTF-8.
 630                  */
 631                 decode_transfer_encoding(line);
 632                 if (metainfo_charset)
 633                         convert_to_utf8(line, charset);
 634
 635                 handle_inbody_header(seen, line);
 636                 if (!(*seen & SEEN_PREFIX))
 637                         continue;
 638
 639                 fputs(line, cmitmsg);
 640         } while (fgets(line, sizeof(line), stdin) != NULL);
 641         fclose(cmitmsg);
 642         cmitmsg = NULL;
 643         return 0;
 644 }
 645
 646 /* We have done the commit message and have the first
 647  * line of the patch in line[].
 648  */
 649 static void handle_patch(void)
 650 {
 651         do {
 652                 if (multipart_boundary[0] && is_multipart_boundary(line))
 653                         break;
 654                 /* Only unwrap transfer encoding but otherwise do not
 655                  * do anything.  We do *NOT* want UTF-8 conversion
 656                  * here; we are dealing with the user payload.
 657                  */
 658                 decode_transfer_encoding(line);
 659                 fputs(line, patchfile);
 660                 patch_lines++;
 661         } while (fgets(line, sizeof(line), stdin) != NULL);
 662 }
 663
 664 /* multipart boundary and transfer encoding are set up for us, and we
 665  * are at the end of the sub header.  do equivalent of handle_body up
 666  * to the next boundary without closing patchfile --- we will expect
 667  * that the first part to contain commit message and a patch, and
 668  * handle other parts as pure patches.
 669  */
 670 static int handle_multipart_one_part(int *seen)
 671 {
 672         int n = 0;
 673
 674         while (fgets(line, sizeof(line), stdin) != NULL) {
 675         again:
 676                 n++;
 677                 if (is_multipart_boundary(line))
 678                         break;
 679                 if (handle_commit_msg(seen))
 680                         goto again;
 681                 handle_patch();
 682                 break;
 683         }
 684         if (n == 0)
 685                 return -1;
 686         return 0;
 687 }
 688
 689 static void handle_multipart_body(void)
 690 {
 691         int seen = 0;
 692         int part_num = 0;
 693
 694         /* Skip up to the first boundary */
 695         while (fgets(line, sizeof(line), stdin) != NULL)
 696                 if (is_multipart_boundary(line)) {
 697                         part_num = 1;
 698                         break;
 699                 }
 700         if (!part_num)
 701                 return;
 702         /* We are on boundary line.  Start slurping the subhead. */
 703         while (1) {
 704                 int hdr = read_one_header_line(line, sizeof(line), stdin);
 705                 if (!hdr) {
 706                         if (handle_multipart_one_part(&seen) < 0)
 707                                 return;
 708                         /* Reset per part headers */
 709                         transfer_encoding = TE_DONTCARE;
 710                         charset[0] = 0;
 711                 }
 712                 else
 713                         check_subheader_line(line);
 714         }
 715         fclose(patchfile);
 716         if (!patch_lines) {
 717                 fprintf(stderr, "No patch found\n");
 718                 exit(1);
 719         }
 720 }
 721
 722 /* Non multipart message */
 723 static void handle_body(void)
 724 {
 725         int seen = 0;
 726
 727         if (line[0] || fgets(line, sizeof(line), stdin) != NULL) {
 728                 handle_commit_msg(&seen);
 729                 handle_patch();
 730         }
 731         fclose(patchfile);
 732         if (!patch_lines) {
 733                 fprintf(stderr, "No patch found\n");
 734                 exit(1);
 735         }
 736 }
 737
 738 static const char mailinfo_usage[] =
 739         "git-mailinfo [-k] [-u | --encoding=<encoding>] msg patch <mail >info";
 740
 741 int main(int argc, char **argv)
 742 {
 743         /* NEEDSWORK: might want to do the optional .git/ directory
 744          * discovery
 745          */
 746         git_config(git_default_config);
 747
 748         while (1 < argc && argv[1][0] == '-') {
 749                 if (!strcmp(argv[1], "-k"))
 750                         keep_subject = 1;
 751                 else if (!strcmp(argv[1], "-u"))
 752                         metainfo_charset = git_commit_encoding;
 753                 else if (!strncmp(argv[1], "--encoding=", 11))
 754                         metainfo_charset = argv[1] + 11;
 755                 else
 756                         usage(mailinfo_usage);
 757                 argc--; argv++;
 758         }
 759
 760         if (argc != 3)
 761                 usage(mailinfo_usage);
 762         cmitmsg = fopen(argv[1], "w");
 763         if (!cmitmsg) {
 764                 perror(argv[1]);
 765                 exit(1);
 766         }
 767         patchfile = fopen(argv[2], "w");
 768         if (!patchfile) {
 769                 perror(argv[2]);
 770                 exit(1);
 771         }
 772         while (1) {
 773                 int hdr = read_one_header_line(line, sizeof(line), stdin);
 774                 if (!hdr) {
 775                         if (multipart_boundary[0])
 776                                 handle_multipart_body();
 777                         else
 778                                 handle_body();
 779                         handle_info();
 780                         break;
 781                 }
 782                 check_header_line(line);
 783         }
 784         return 0;
 785 }