Andrew's git / gitweb.git / mailinfo.c
?
mailinfo.con commit mailinfo: skip bogus UNIX From line inside body (81c5cf7)
   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
  16static FILE *cmitmsg, *patchfile;
  17
  18static int keep_subject = 0;
  19static char *metainfo_charset = NULL;
  20static char line[1000];
  21static char date[1000];
  22static char name[1000];
  23static char email[1000];
  24static char subject[1000];
  25
  26static enum  {
  27        TE_DONTCARE, TE_QP, TE_BASE64,
  28} transfer_encoding;
  29static char charset[256];
  30
  31static char multipart_boundary[1000];
  32static int multipart_boundary_len;
  33static int patch_lines = 0;
  34
  35static 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
  45static 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
  75static 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
 144static int handle_date(char *line)
 145{
 146        strcpy(date, line);
 147        return 0;
 148}
 149
 150static 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
 162static 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
 184static 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
 202static 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
 213static 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
 224static int is_multipart_boundary(const char *line)
 225{
 226        return (!memcmp(line, multipart_boundary, multipart_boundary_len));
 227}
 228
 229static 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_BOGUS_UNIX_FROM 010
 241
 242/* First lines of body can have From:, Date:, and Subject: */
 243static int handle_inbody_header(int *seen, char *line)
 244{
 245        if (!memcmp(">From", line, 5) && isspace(line[5])) {
 246                if (!(*seen & SEEN_BOGUS_UNIX_FROM)) {
 247                        *seen |= SEEN_BOGUS_UNIX_FROM;
 248                        return 1;
 249                }
 250        }
 251        if (!memcmp("From:", line, 5) && isspace(line[5])) {
 252                if (!(*seen & SEEN_FROM) && handle_from(line+6)) {
 253                        *seen |= SEEN_FROM;
 254                        return 1;
 255                }
 256        }
 257        if (!memcmp("Date:", line, 5) && isspace(line[5])) {
 258                if (!(*seen & SEEN_DATE)) {
 259                        handle_date(line+6);
 260                        *seen |= SEEN_DATE;
 261                        return 1;
 262                }
 263        }
 264        if (!memcmp("Subject:", line, 8) && isspace(line[8])) {
 265                if (!(*seen & SEEN_SUBJECT)) {
 266                        handle_subject(line+9);
 267                        *seen |= SEEN_SUBJECT;
 268                        return 1;
 269                }
 270        }
 271        if (!memcmp("[PATCH]", line, 7) && isspace(line[7])) {
 272                if (!(*seen & SEEN_SUBJECT)) {
 273                        handle_subject(line);
 274                        *seen |= SEEN_SUBJECT;
 275                        return 1;
 276                }
 277        }
 278        return 0;
 279}
 280
 281static char *cleanup_subject(char *subject)
 282{
 283        if (keep_subject)
 284                return subject;
 285        for (;;) {
 286                char *p;
 287                int len, remove;
 288                switch (*subject) {
 289                case 'r': case 'R':
 290                        if (!memcmp("e:", subject+1, 2)) {
 291                                subject +=3;
 292                                continue;
 293                        }
 294                        break;
 295                case ' ': case '\t': case ':':
 296                        subject++;
 297                        continue;
 298
 299                case '[':
 300                        p = strchr(subject, ']');
 301                        if (!p) {
 302                                subject++;
 303                                continue;
 304                        }
 305                        len = strlen(p);
 306                        remove = p - subject;
 307                        if (remove <= len *2) {
 308                                subject = p+1;
 309                                continue;
 310                        }       
 311                        break;
 312                }
 313                return subject;
 314        }
 315}                       
 316
 317static void cleanup_space(char *buf)
 318{
 319        unsigned char c;
 320        while ((c = *buf) != 0) {
 321                buf++;
 322                if (isspace(c)) {
 323                        buf[-1] = ' ';
 324                        c = *buf;
 325                        while (isspace(c)) {
 326                                int len = strlen(buf);
 327                                memmove(buf, buf+1, len);
 328                                c = *buf;
 329                        }
 330                }
 331        }
 332}
 333
 334typedef int (*header_fn_t)(char *);
 335struct header_def {
 336        const char *name;
 337        header_fn_t func;
 338        int namelen;
 339};
 340
 341static void check_header(char *line, int len, struct header_def *header)
 342{
 343        int i;
 344
 345        if (header[0].namelen <= 0) {
 346                for (i = 0; header[i].name; i++)
 347                        header[i].namelen = strlen(header[i].name);
 348        }
 349        for (i = 0; header[i].name; i++) {
 350                int len = header[i].namelen;
 351                if (!strncasecmp(line, header[i].name, len) &&
 352                    line[len] == ':' && isspace(line[len + 1])) {
 353                        header[i].func(line + len + 2);
 354                        break;
 355                }
 356        }
 357}
 358
 359static void check_subheader_line(char *line, int len)
 360{
 361        static struct header_def header[] = {
 362                { "Content-Type", handle_subcontent_type },
 363                { "Content-Transfer-Encoding",
 364                  handle_content_transfer_encoding },
 365                { NULL },
 366        };
 367        check_header(line, len, header);
 368}
 369static void check_header_line(char *line, int len)
 370{
 371        static struct header_def header[] = {
 372                { "From", handle_from },
 373                { "Date", handle_date },
 374                { "Subject", handle_subject },
 375                { "Content-Type", handle_content_type },
 376                { "Content-Transfer-Encoding",
 377                  handle_content_transfer_encoding },
 378                { NULL },
 379        };
 380        check_header(line, len, header);
 381}
 382
 383static int read_one_header_line(char *line, int sz, FILE *in)
 384{
 385        int ofs = 0;
 386        while (ofs < sz) {
 387                int peek, len;
 388                if (fgets(line + ofs, sz - ofs, in) == NULL)
 389                        return ofs;
 390                len = eatspace(line + ofs);
 391                if (len == 0)
 392                        return ofs;
 393                peek = fgetc(in); ungetc(peek, in);
 394                if (peek == ' ' || peek == '\t') {
 395                        /* Yuck, 2822 header "folding" */
 396                        ofs += len;
 397                        continue;
 398                }
 399                return ofs + len;
 400        }
 401        return ofs;
 402}
 403
 404static unsigned hexval(int c)
 405{
 406        if (c >= '0' && c <= '9')
 407                return c - '0';
 408        if (c >= 'a' && c <= 'f')
 409                return c - 'a' + 10;
 410        if (c >= 'A' && c <= 'F')
 411                return c - 'A' + 10;
 412        return ~0;
 413}
 414
 415static int decode_q_segment(char *in, char *ot, char *ep, int rfc2047)
 416{
 417        int c;
 418        while ((c = *in++) != 0 && (in <= ep)) {
 419                if (c == '=') {
 420                        int d = *in++;
 421                        if (d == '\n' || !d)
 422                                break; /* drop trailing newline */
 423                        *ot++ = ((hexval(d) << 4) | hexval(*in++));
 424                        continue;
 425                }
 426                if (rfc2047 && c == '_') /* rfc2047 4.2 (2) */
 427                        c = 0x20;
 428                *ot++ = c;
 429        }
 430        *ot = 0;
 431        return 0;
 432}
 433
 434static int decode_b_segment(char *in, char *ot, char *ep)
 435{
 436        /* Decode in..ep, possibly in-place to ot */
 437        int c, pos = 0, acc = 0;
 438
 439        while ((c = *in++) != 0 && (in <= ep)) {
 440                if (c == '+')
 441                        c = 62;
 442                else if (c == '/')
 443                        c = 63;
 444                else if ('A' <= c && c <= 'Z')
 445                        c -= 'A';
 446                else if ('a' <= c && c <= 'z')
 447                        c -= 'a' - 26;
 448                else if ('0' <= c && c <= '9')
 449                        c -= '0' - 52;
 450                else if (c == '=') {
 451                        /* padding is almost like (c == 0), except we do
 452                         * not output NUL resulting only from it;
 453                         * for now we just trust the data.
 454                         */
 455                        c = 0;
 456                }
 457                else
 458                        continue; /* garbage */
 459                switch (pos++) {
 460                case 0:
 461                        acc = (c << 2);
 462                        break;
 463                case 1:
 464                        *ot++ = (acc | (c >> 4));
 465                        acc = (c & 15) << 4;
 466                        break;
 467                case 2:
 468                        *ot++ = (acc | (c >> 2));
 469                        acc = (c & 3) << 6;
 470                        break;
 471                case 3:
 472                        *ot++ = (acc | c);
 473                        acc = pos = 0;
 474                        break;
 475                }
 476        }
 477        *ot = 0;
 478        return 0;
 479}
 480
 481static void convert_to_utf8(char *line, char *charset)
 482{
 483#ifndef NO_ICONV
 484        char *in, *out;
 485        size_t insize, outsize, nrc;
 486        char outbuf[4096]; /* cheat */
 487        static char latin_one[] = "latin1";
 488        char *input_charset = *charset ? charset : latin_one;
 489        iconv_t conv = iconv_open(metainfo_charset, input_charset);
 490
 491        if (conv == (iconv_t) -1) {
 492                static int warned_latin1_once = 0;
 493                if (input_charset != latin_one) {
 494                        fprintf(stderr, "cannot convert from %s to %s\n",
 495                                input_charset, metainfo_charset);
 496                        *charset = 0;
 497                }
 498                else if (!warned_latin1_once) {
 499                        warned_latin1_once = 1;
 500                        fprintf(stderr, "tried to convert from %s to %s, "
 501                                "but your iconv does not work with it.\n",
 502                                input_charset, metainfo_charset);
 503                }
 504                return;
 505        }
 506        in = line;
 507        insize = strlen(in);
 508        out = outbuf;
 509        outsize = sizeof(outbuf);
 510        nrc = iconv(conv, &in, &insize, &out, &outsize);
 511        iconv_close(conv);
 512        if (nrc == (size_t) -1)
 513                return;
 514        *out = 0;
 515        strcpy(line, outbuf);
 516#endif
 517}
 518
 519static void decode_header_bq(char *it)
 520{
 521        char *in, *out, *ep, *cp, *sp;
 522        char outbuf[1000];
 523
 524        in = it;
 525        out = outbuf;
 526        while ((ep = strstr(in, "=?")) != NULL) {
 527                int sz, encoding;
 528                char charset_q[256], piecebuf[256];
 529                if (in != ep) {
 530                        sz = ep - in;
 531                        memcpy(out, in, sz);
 532                        out += sz;
 533                        in += sz;
 534                }
 535                /* E.g.
 536                 * ep : "=?iso-2022-jp?B?GyR...?= foo"
 537                 * ep : "=?ISO-8859-1?Q?Foo=FCbar?= baz"
 538                 */
 539                ep += 2;
 540                cp = strchr(ep, '?');
 541                if (!cp)
 542                        return; /* no munging */
 543                for (sp = ep; sp < cp; sp++)
 544                        charset_q[sp - ep] = tolower(*sp);
 545                charset_q[cp - ep] = 0;
 546                encoding = cp[1];
 547                if (!encoding || cp[2] != '?')
 548                        return; /* no munging */
 549                ep = strstr(cp + 3, "?=");
 550                if (!ep)
 551                        return; /* no munging */
 552                switch (tolower(encoding)) {
 553                default:
 554                        return; /* no munging */
 555                case 'b':
 556                        sz = decode_b_segment(cp + 3, piecebuf, ep);
 557                        break;
 558                case 'q':
 559                        sz = decode_q_segment(cp + 3, piecebuf, ep, 1);
 560                        break;
 561                }
 562                if (sz < 0)
 563                        return;
 564                if (metainfo_charset)
 565                        convert_to_utf8(piecebuf, charset_q);
 566                strcpy(out, piecebuf);
 567                out += strlen(out);
 568                in = ep + 2;
 569        }
 570        strcpy(out, in);
 571        strcpy(it, outbuf);
 572}
 573
 574static void decode_transfer_encoding(char *line)
 575{
 576        char *ep;
 577
 578        switch (transfer_encoding) {
 579        case TE_QP:
 580                ep = line + strlen(line);
 581                decode_q_segment(line, line, ep, 0);
 582                break;
 583        case TE_BASE64:
 584                ep = line + strlen(line);
 585                decode_b_segment(line, line, ep);
 586                break;
 587        case TE_DONTCARE:
 588                break;
 589        }
 590}
 591
 592static void handle_info(void)
 593{
 594        char *sub;
 595        static int done_info = 0;
 596
 597        if (done_info)
 598                return;
 599
 600        done_info = 1;
 601        sub = cleanup_subject(subject);
 602        cleanup_space(name);
 603        cleanup_space(date);
 604        cleanup_space(email);
 605        cleanup_space(sub);
 606
 607        /* Unwrap inline B and Q encoding, and optionally
 608         * normalize the meta information to utf8.
 609         */
 610        decode_header_bq(name);
 611        decode_header_bq(date);
 612        decode_header_bq(email);
 613        decode_header_bq(sub);
 614        printf("Author: %s\nEmail: %s\nSubject: %s\nDate: %s\n\n",
 615               name, email, sub, date);
 616}
 617
 618/* We are inside message body and have read line[] already.
 619 * Spit out the commit log.
 620 */
 621static int handle_commit_msg(void)
 622{
 623        if (!cmitmsg)
 624                return 0;
 625        do {
 626                if (!memcmp("diff -", line, 6) ||
 627                    !memcmp("---", line, 3) ||
 628                    !memcmp("Index: ", line, 7))
 629                        break;
 630                if ((multipart_boundary[0] && is_multipart_boundary(line))) {
 631                        /* We come here when the first part had only
 632                         * the commit message without any patch.  We
 633                         * pretend we have not seen this line yet, and
 634                         * go back to the loop.
 635                         */
 636                        return 1;
 637                }
 638
 639                /* Unwrap transfer encoding and optionally
 640                 * normalize the log message to UTF-8.
 641                 */
 642                decode_transfer_encoding(line);
 643                if (metainfo_charset)
 644                        convert_to_utf8(line, charset);
 645                fputs(line, cmitmsg);
 646        } while (fgets(line, sizeof(line), stdin) != NULL);
 647        fclose(cmitmsg);
 648        cmitmsg = NULL;
 649        return 0;
 650}
 651
 652/* We have done the commit message and have the first
 653 * line of the patch in line[].
 654 */
 655static void handle_patch(void)
 656{
 657        do {
 658                if (multipart_boundary[0] && is_multipart_boundary(line))
 659                        break;
 660                /* Only unwrap transfer encoding but otherwise do not
 661                 * do anything.  We do *NOT* want UTF-8 conversion
 662                 * here; we are dealing with the user payload.
 663                 */
 664                decode_transfer_encoding(line);
 665                fputs(line, patchfile);
 666                patch_lines++;
 667        } while (fgets(line, sizeof(line), stdin) != NULL);
 668}
 669
 670/* multipart boundary and transfer encoding are set up for us, and we
 671 * are at the end of the sub header.  do equivalent of handle_body up
 672 * to the next boundary without closing patchfile --- we will expect
 673 * that the first part to contain commit message and a patch, and
 674 * handle other parts as pure patches.
 675 */
 676static int handle_multipart_one_part(void)
 677{
 678        int seen = 0;
 679        int n = 0;
 680        int len;
 681
 682        while (fgets(line, sizeof(line), stdin) != NULL) {
 683        again:
 684                len = eatspace(line);
 685                n++;
 686                if (!len)
 687                        continue;
 688                if (is_multipart_boundary(line))
 689                        break;
 690                if (0 <= seen && handle_inbody_header(&seen, line))
 691                        continue;
 692                seen = -1; /* no more inbody headers */
 693                line[len] = '\n';
 694                handle_info();
 695                if (handle_commit_msg())
 696                        goto again;
 697                handle_patch();
 698                break;
 699        }
 700        if (n == 0)
 701                return -1;
 702        return 0;
 703}
 704
 705static void handle_multipart_body(void)
 706{
 707        int part_num = 0;
 708
 709        /* Skip up to the first boundary */
 710        while (fgets(line, sizeof(line), stdin) != NULL)
 711                if (is_multipart_boundary(line)) {
 712                        part_num = 1;
 713                        break;
 714                }
 715        if (!part_num)
 716                return;
 717        /* We are on boundary line.  Start slurping the subhead. */
 718        while (1) {
 719                int len = read_one_header_line(line, sizeof(line), stdin);
 720                if (!len) {
 721                        if (handle_multipart_one_part() < 0)
 722                                return;
 723                        /* Reset per part headers */
 724                        transfer_encoding = TE_DONTCARE;
 725                        charset[0] = 0;
 726                }
 727                else
 728                        check_subheader_line(line, len);
 729        }
 730        fclose(patchfile);
 731        if (!patch_lines) {
 732                fprintf(stderr, "No patch found\n");
 733                exit(1);
 734        }
 735}
 736
 737/* Non multipart message */
 738static void handle_body(void)
 739{
 740        int seen = 0;
 741
 742        while (fgets(line, sizeof(line), stdin) != NULL) {
 743                int len = eatspace(line);
 744                if (!len)
 745                        continue;
 746                if (0 <= seen && handle_inbody_header(&seen, line))
 747                        continue;
 748                seen = -1; /* no more inbody headers */
 749                line[len] = '\n';
 750                handle_info();
 751                handle_commit_msg();
 752                handle_patch();
 753                break;
 754        }
 755        fclose(patchfile);
 756        if (!patch_lines) {
 757                fprintf(stderr, "No patch found\n");
 758                exit(1);
 759        }
 760}
 761
 762static const char mailinfo_usage[] =
 763        "git-mailinfo [-k] [-u | --encoding=<encoding>] msg patch <mail >info";
 764
 765int main(int argc, char **argv)
 766{
 767        /* NEEDSWORK: might want to do the optional .git/ directory
 768         * discovery
 769         */
 770        git_config(git_default_config);
 771
 772        while (1 < argc && argv[1][0] == '-') {
 773                if (!strcmp(argv[1], "-k"))
 774                        keep_subject = 1;
 775                else if (!strcmp(argv[1], "-u"))
 776                        metainfo_charset = git_commit_encoding;
 777                else if (!strncmp(argv[1], "--encoding=", 11))
 778                        metainfo_charset = argv[1] + 11;
 779                else
 780                        usage(mailinfo_usage);
 781                argc--; argv++;
 782        }
 783
 784        if (argc != 3)
 785                usage(mailinfo_usage);
 786        cmitmsg = fopen(argv[1], "w");
 787        if (!cmitmsg) {
 788                perror(argv[1]);
 789                exit(1);
 790        }
 791        patchfile = fopen(argv[2], "w");
 792        if (!patchfile) {
 793                perror(argv[2]);
 794                exit(1);
 795        }
 796        while (1) {
 797                int len = read_one_header_line(line, sizeof(line), stdin);
 798                if (!len) {
 799                        if (multipart_boundary[0])
 800                                handle_multipart_body();
 801                        else
 802                                handle_body();
 803                        break;
 804                }
 805                check_header_line(line, len);
 806        }
 807        return 0;
 808}