mandoc.c

   1 /*      $Id: mandoc.c,v 1.111 2018/12/15 19:30:26 schwarze Exp $ */
   2 /*
   3  * Copyright (c) 2008-2011, 2014 Kristaps Dzonsons <kristaps@bsd.lv>
   4  * Copyright (c) 2011-2015, 2017, 2018 Ingo Schwarze <schwarze@openbsd.org>
   5  *
   6  * Permission to use, copy, modify, and distribute this software for any
   7  * purpose with or without fee is hereby granted, provided that the above
   8  * copyright notice and this permission notice appear in all copies.
   9  *
  10  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
  11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
  13  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  14  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  15  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  16  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  17  */
  18 #include "config.h"
  19
  20 #include <sys/types.h>
  21
  22 #include <assert.h>
  23 #include <ctype.h>
  24 #include <errno.h>
  25 #include <limits.h>
  26 #include <stdlib.h>
  27 #include <stdio.h>
  28 #include <string.h>
  29 #include <time.h>
  30
  31 #include "mandoc_aux.h"
  32 #include "mandoc.h"
  33 #include "roff.h"
  34 #include "libmandoc.h"
  35
  36 static  int      a2time(time_t *, const char *, const char *);
  37 static  char    *time2a(time_t);
  38
  39
  40 enum mandoc_esc
  41 mandoc_escape(const char **end, const char **start, int *sz)
  42 {
  43         const char      *local_start;
  44         int              local_sz, c, i;
  45         char             term;
  46         enum mandoc_esc  gly;
  47
  48         /*
  49          * When the caller doesn't provide return storage,
  50          * use local storage.
  51          */
  52
  53         if (NULL == start)
  54                 start = &local_start;
  55         if (NULL == sz)
  56                 sz = &local_sz;
  57
  58         /*
  59          * Treat "\E" just like "\";
  60          * it only makes a difference in copy mode.
  61          */
  62
  63         if (**end == 'E')
  64                 ++*end;
  65
  66         /*
  67          * Beyond the backslash, at least one input character
  68          * is part of the escape sequence.  With one exception
  69          * (see below), that character won't be returned.
  70          */
  71
  72         gly = ESCAPE_ERROR;
  73         *start = ++*end;
  74         *sz = 0;
  75         term = '\0';
  76
  77         switch ((*start)[-1]) {
  78         /*
  79          * First the glyphs.  There are several different forms of
  80          * these, but each eventually returns a substring of the glyph
  81          * name.
  82          */
  83         case '(':
  84                 gly = ESCAPE_SPECIAL;
  85                 *sz = 2;
  86                 break;
  87         case '[':
  88                 if (**start == ' ') {
  89                         ++*end;
  90                         return ESCAPE_ERROR;
  91                 }
  92                 gly = ESCAPE_SPECIAL;
  93                 term = ']';
  94                 break;
  95         case 'C':
  96                 if ('\'' != **start)
  97                         return ESCAPE_ERROR;
  98                 *start = ++*end;
  99                 gly = ESCAPE_SPECIAL;
 100                 term = '\'';
 101                 break;
 102
 103         /*
 104          * Escapes taking no arguments at all.
 105          */
 106         case '!':
 107         case '?':
 108                 return ESCAPE_UNSUPP;
 109         case '%':
 110         case '&':
 111         case ')':
 112         case ',':
 113         case '/':
 114         case '^':
 115         case 'a':
 116         case 'd':
 117         case 'r':
 118         case 't':
 119         case 'u':
 120         case '{':
 121         case '|':
 122         case '}':
 123                 return ESCAPE_IGNORE;
 124         case 'c':
 125                 return ESCAPE_NOSPACE;
 126         case 'p':
 127                 return ESCAPE_BREAK;
 128
 129         /*
 130          * The \z escape is supposed to output the following
 131          * character without advancing the cursor position.
 132          * Since we are mostly dealing with terminal mode,
 133          * let us just skip the next character.
 134          */
 135         case 'z':
 136                 return ESCAPE_SKIPCHAR;
 137
 138         /*
 139          * Handle all triggers matching \X(xy, \Xx, and \X[xxxx], where
 140          * 'X' is the trigger.  These have opaque sub-strings.
 141          */
 142         case 'F':
 143         case 'f':
 144         case 'g':
 145         case 'k':
 146         case 'M':
 147         case 'm':
 148         case 'n':
 149         case 'O':
 150         case 'V':
 151         case 'Y':
 152                 gly = (*start)[-1] == 'f' ? ESCAPE_FONT : ESCAPE_IGNORE;
 153                 switch (**start) {
 154                 case '(':
 155                         if ((*start)[-1] == 'O')
 156                                 gly = ESCAPE_ERROR;
 157                         *start = ++*end;
 158                         *sz = 2;
 159                         break;
 160                 case '[':
 161                         if ((*start)[-1] == 'O')
 162                                 gly = (*start)[1] == '5' ?
 163                                     ESCAPE_UNSUPP : ESCAPE_ERROR;
 164                         *start = ++*end;
 165                         term = ']';
 166                         break;
 167                 default:
 168                         if ((*start)[-1] == 'O') {
 169                                 switch (**start) {
 170                                 case '0':
 171                                         gly = ESCAPE_UNSUPP;
 172                                         break;
 173                                 case '1':
 174                                 case '2':
 175                                 case '3':
 176                                 case '4':
 177                                         break;
 178                                 default:
 179                                         gly = ESCAPE_ERROR;
 180                                         break;
 181                                 }
 182                         }
 183                         *sz = 1;
 184                         break;
 185                 }
 186                 break;
 187         case '*':
 188                 if (strncmp(*start, "(.T", 3) != 0)
 189                         abort();
 190                 gly = ESCAPE_DEVICE;
 191                 *start = ++*end;
 192                 *sz = 2;
 193                 break;
 194
 195         /*
 196          * These escapes are of the form \X'Y', where 'X' is the trigger
 197          * and 'Y' is any string.  These have opaque sub-strings.
 198          * The \B and \w escapes are handled in roff.c, roff_res().
 199          */
 200         case 'A':
 201         case 'b':
 202         case 'D':
 203         case 'R':
 204         case 'X':
 205         case 'Z':
 206                 gly = ESCAPE_IGNORE;
 207                 /* FALLTHROUGH */
 208         case 'o':
 209                 if (**start == '\0')
 210                         return ESCAPE_ERROR;
 211                 if (gly == ESCAPE_ERROR)
 212                         gly = ESCAPE_OVERSTRIKE;
 213                 term = **start;
 214                 *start = ++*end;
 215                 break;
 216
 217         /*
 218          * These escapes are of the form \X'N', where 'X' is the trigger
 219          * and 'N' resolves to a numerical expression.
 220          */
 221         case 'h':
 222         case 'H':
 223         case 'L':
 224         case 'l':
 225         case 'S':
 226         case 'v':
 227         case 'x':
 228                 if (strchr(" %&()*+-./0123456789:<=>", **start)) {
 229                         if ('\0' != **start)
 230                                 ++*end;
 231                         return ESCAPE_ERROR;
 232                 }
 233                 switch ((*start)[-1]) {
 234                 case 'h':
 235                         gly = ESCAPE_HORIZ;
 236                         break;
 237                 case 'l':
 238                         gly = ESCAPE_HLINE;
 239                         break;
 240                 default:
 241                         gly = ESCAPE_IGNORE;
 242                         break;
 243                 }
 244                 term = **start;
 245                 *start = ++*end;
 246                 break;
 247
 248         /*
 249          * Special handling for the numbered character escape.
 250          * XXX Do any other escapes need similar handling?
 251          */
 252         case 'N':
 253                 if ('\0' == **start)
 254                         return ESCAPE_ERROR;
 255                 (*end)++;
 256                 if (isdigit((unsigned char)**start)) {
 257                         *sz = 1;
 258                         return ESCAPE_IGNORE;
 259                 }
 260                 (*start)++;
 261                 while (isdigit((unsigned char)**end))
 262                         (*end)++;
 263                 *sz = *end - *start;
 264                 if ('\0' != **end)
 265                         (*end)++;
 266                 return ESCAPE_NUMBERED;
 267
 268         /*
 269          * Sizes get a special category of their own.
 270          */
 271         case 's':
 272                 gly = ESCAPE_IGNORE;
 273
 274                 /* See +/- counts as a sign. */
 275                 if ('+' == **end || '-' == **end || ASCII_HYPH == **end)
 276                         *start = ++*end;
 277
 278                 switch (**end) {
 279                 case '(':
 280                         *start = ++*end;
 281                         *sz = 2;
 282                         break;
 283                 case '[':
 284                         *start = ++*end;
 285                         term = ']';
 286                         break;
 287                 case '\'':
 288                         *start = ++*end;
 289                         term = '\'';
 290                         break;
 291                 case '3':
 292                 case '2':
 293                 case '1':
 294                         *sz = (*end)[-1] == 's' &&
 295                             isdigit((unsigned char)(*end)[1]) ? 2 : 1;
 296                         break;
 297                 default:
 298                         *sz = 1;
 299                         break;
 300                 }
 301
 302                 break;
 303
 304         /*
 305          * Several special characters can be encoded as
 306          * one-byte escape sequences without using \[].
 307          */
 308         case ' ':
 309         case '\'':
 310         case '-':
 311         case '.':
 312         case '0':
 313         case ':':
 314         case '_':
 315         case '`':
 316         case 'e':
 317         case '~':
 318                 gly = ESCAPE_SPECIAL;
 319                 /* FALLTHROUGH */
 320         default:
 321                 if (gly == ESCAPE_ERROR)
 322                         gly = ESCAPE_UNDEF;
 323                 *start = --*end;
 324                 *sz = 1;
 325                 break;
 326         }
 327
 328         /*
 329          * Read up to the terminating character,
 330          * paying attention to nested escapes.
 331          */
 332
 333         if ('\0' != term) {
 334                 while (**end != term) {
 335                         switch (**end) {
 336                         case '\0':
 337                                 return ESCAPE_ERROR;
 338                         case '\\':
 339                                 (*end)++;
 340                                 if (ESCAPE_ERROR ==
 341                                     mandoc_escape(end, NULL, NULL))
 342                                         return ESCAPE_ERROR;
 343                                 break;
 344                         default:
 345                                 (*end)++;
 346                                 break;
 347                         }
 348                 }
 349                 *sz = (*end)++ - *start;
 350
 351                 /*
 352                  * The file chars.c only provides one common list
 353                  * of character names, but \[-] == \- is the only
 354                  * one of the characters with one-byte names that
 355                  * allows enclosing the name in brackets.
 356                  */
 357                 if (gly == ESCAPE_SPECIAL && *sz == 1 && **start != '-')
 358                         return ESCAPE_ERROR;
 359         } else {
 360                 assert(*sz > 0);
 361                 if ((size_t)*sz > strlen(*start))
 362                         return ESCAPE_ERROR;
 363                 *end += *sz;
 364         }
 365
 366         /* Run post-processors. */
 367
 368         switch (gly) {
 369         case ESCAPE_FONT:
 370                 if (*sz == 2) {
 371                         if (**start == 'C') {
 372                                 if ((*start)[1] == 'W' ||
 373                                     (*start)[1] == 'R') {
 374                                         gly = ESCAPE_FONTCW;
 375                                         break;
 376                                 }
 377                                 /*
 378                                  * Treat other constant-width font modes
 379                                  * just like regular font modes.
 380                                  */
 381                                 (*start)++;
 382                                 (*sz)--;
 383                         } else {
 384                                 if ((*start)[0] == 'B' && (*start)[1] == 'I')
 385                                         gly = ESCAPE_FONTBI;
 386                                 break;
 387                         }
 388                 } else if (*sz != 1) {
 389                         if (*sz == 0)
 390                                 gly = ESCAPE_FONTPREV;
 391                         break;
 392                 }
 393
 394                 switch (**start) {
 395                 case '3':
 396                 case 'B':
 397                         gly = ESCAPE_FONTBOLD;
 398                         break;
 399                 case '2':
 400                 case 'I':
 401                         gly = ESCAPE_FONTITALIC;
 402                         break;
 403                 case 'P':
 404                         gly = ESCAPE_FONTPREV;
 405                         break;
 406                 case '1':
 407                 case 'R':
 408                         gly = ESCAPE_FONTROMAN;
 409                         break;
 410                 }
 411                 break;
 412         case ESCAPE_SPECIAL:
 413                 if (**start == 'c') {
 414                         if (*sz < 6 || *sz > 7 ||
 415                             strncmp(*start, "char", 4) != 0 ||
 416                             (int)strspn(*start + 4, "0123456789") + 4 < *sz)
 417                                 break;
 418                         c = 0;
 419                         for (i = 4; i < *sz; i++)
 420                                 c = 10 * c + ((*start)[i] - '0');
 421                         if (c < 0x21 || (c > 0x7e && c < 0xa0) || c > 0xff)
 422                                 break;
 423                         *start += 4;
 424                         *sz -= 4;
 425                         gly = ESCAPE_NUMBERED;
 426                         break;
 427                 }
 428
 429                 /*
 430                  * Unicode escapes are defined in groff as \[u0000]
 431                  * to \[u10FFFF], where the contained value must be
 432                  * a valid Unicode codepoint.  Here, however, only
 433                  * check the length and range.
 434                  */
 435                 if (**start != 'u' || *sz < 5 || *sz > 7)
 436                         break;
 437                 if (*sz == 7 && ((*start)[1] != '1' || (*start)[2] != '0'))
 438                         break;
 439                 if (*sz == 6 && (*start)[1] == '0')
 440                         break;
 441                 if (*sz == 5 && (*start)[1] == 'D' &&
 442                     strchr("89ABCDEF", (*start)[2]) != NULL)
 443                         break;
 444                 if ((int)strspn(*start + 1, "0123456789ABCDEFabcdef")
 445                     + 1 == *sz)
 446                         gly = ESCAPE_UNICODE;
 447                 break;
 448         default:
 449                 break;
 450         }
 451
 452         return gly;
 453 }
 454
 455 /*
 456  * Parse a quoted or unquoted roff-style request or macro argument.
 457  * Return a pointer to the parsed argument, which is either the original
 458  * pointer or advanced by one byte in case the argument is quoted.
 459  * NUL-terminate the argument in place.
 460  * Collapse pairs of quotes inside quoted arguments.
 461  * Advance the argument pointer to the next argument,
 462  * or to the NUL byte terminating the argument line.
 463  */
 464 char *
 465 mandoc_getarg(char **cpp, int ln, int *pos)
 466 {
 467         char     *start, *cp;
 468         int       quoted, pairs, white;
 469
 470         /* Quoting can only start with a new word. */
 471         start = *cpp;
 472         quoted = 0;
 473         if ('"' == *start) {
 474                 quoted = 1;
 475                 start++;
 476         }
 477
 478         pairs = 0;
 479         white = 0;
 480         for (cp = start; '\0' != *cp; cp++) {
 481
 482                 /*
 483                  * Move the following text left
 484                  * after quoted quotes and after "\\" and "\t".
 485                  */
 486                 if (pairs)
 487                         cp[-pairs] = cp[0];
 488
 489                 if ('\\' == cp[0]) {
 490                         /*
 491                          * In copy mode, translate double to single
 492                          * backslashes and backslash-t to literal tabs.
 493                          */
 494                         switch (cp[1]) {
 495                         case 'a':
 496                         case 't':
 497                                 cp[0] = '\t';
 498                                 /* FALLTHROUGH */
 499                         case '\\':
 500                                 pairs++;
 501                                 cp++;
 502                                 break;
 503                         case ' ':
 504                                 /* Skip escaped blanks. */
 505                                 if (0 == quoted)
 506                                         cp++;
 507                                 break;
 508                         default:
 509                                 break;
 510                         }
 511                 } else if (0 == quoted) {
 512                         if (' ' == cp[0]) {
 513                                 /* Unescaped blanks end unquoted args. */
 514                                 white = 1;
 515                                 break;
 516                         }
 517                 } else if ('"' == cp[0]) {
 518                         if ('"' == cp[1]) {
 519                                 /* Quoted quotes collapse. */
 520                                 pairs++;
 521                                 cp++;
 522                         } else {
 523                                 /* Unquoted quotes end quoted args. */
 524                                 quoted = 2;
 525                                 break;
 526                         }
 527                 }
 528         }
 529
 530         /* Quoted argument without a closing quote. */
 531         if (1 == quoted)
 532                 mandoc_msg(MANDOCERR_ARG_QUOTE, ln, *pos, NULL);
 533
 534         /* NUL-terminate this argument and move to the next one. */
 535         if (pairs)
 536                 cp[-pairs] = '\0';
 537         if ('\0' != *cp) {
 538                 *cp++ = '\0';
 539                 while (' ' == *cp)
 540                         cp++;
 541         }
 542         *pos += (int)(cp - start) + (quoted ? 1 : 0);
 543         *cpp = cp;
 544
 545         if ('\0' == *cp && (white || ' ' == cp[-1]))
 546                 mandoc_msg(MANDOCERR_SPACE_EOL, ln, *pos, NULL);
 547
 548         return start;
 549 }
 550
 551 static int
 552 a2time(time_t *t, const char *fmt, const char *p)
 553 {
 554         struct tm        tm;
 555         char            *pp;
 556
 557         memset(&tm, 0, sizeof(struct tm));
 558
 559         pp = NULL;
 560 #if HAVE_STRPTIME
 561         pp = strptime(p, fmt, &tm);
 562 #endif
 563         if (NULL != pp && '\0' == *pp) {
 564                 *t = mktime(&tm);
 565                 return 1;
 566         }
 567
 568         return 0;
 569 }
 570
 571 static char *
 572 time2a(time_t t)
 573 {
 574         struct tm       *tm;
 575         char            *buf, *p;
 576         size_t           ssz;
 577         int              isz;
 578
 579         tm = localtime(&t);
 580         if (tm == NULL)
 581                 return NULL;
 582
 583         /*
 584          * Reserve space:
 585          * up to 9 characters for the month (September) + blank
 586          * up to 2 characters for the day + comma + blank
 587          * 4 characters for the year and a terminating '\0'
 588          */
 589
 590         p = buf = mandoc_malloc(10 + 4 + 4 + 1);
 591
 592         if ((ssz = strftime(p, 10 + 1, "%B ", tm)) == 0)
 593                 goto fail;
 594         p += (int)ssz;
 595
 596         /*
 597          * The output format is just "%d" here, not "%2d" or "%02d".
 598          * That's also the reason why we can't just format the
 599          * date as a whole with "%B %e, %Y" or "%B %d, %Y".
 600          * Besides, the present approach is less prone to buffer
 601          * overflows, in case anybody should ever introduce the bug
 602          * of looking at LC_TIME.
 603          */
 604
 605         if ((isz = snprintf(p, 4 + 1, "%d, ", tm->tm_mday)) == -1)
 606                 goto fail;
 607         p += isz;
 608
 609         if (strftime(p, 4 + 1, "%Y", tm) == 0)
 610                 goto fail;
 611         return buf;
 612
 613 fail:
 614         free(buf);
 615         return NULL;
 616 }
 617
 618 char *
 619 mandoc_normdate(struct roff_man *man, char *in, int ln, int pos)
 620 {
 621         char            *cp;
 622         time_t           t;
 623
 624         /* No date specified: use today's date. */
 625
 626         if (in == NULL || *in == '\0' || strcmp(in, "$" "Mdocdate$") == 0) {
 627                 mandoc_msg(MANDOCERR_DATE_MISSING, ln, pos, NULL);
 628                 return time2a(time(NULL));
 629         }
 630
 631         /* Valid mdoc(7) date format. */
 632
 633         if (a2time(&t, "$" "Mdocdate: %b %d %Y $", in) ||
 634             a2time(&t, "%b %d, %Y", in)) {
 635                 cp = time2a(t);
 636                 if (t > time(NULL) + 86400)
 637                         mandoc_msg(MANDOCERR_DATE_FUTURE, ln, pos, "%s", cp);
 638                 else if (*in != '$' && strcmp(in, cp) != 0)
 639                         mandoc_msg(MANDOCERR_DATE_NORM, ln, pos, "%s", cp);
 640                 return cp;
 641         }
 642
 643         /* In man(7), do not warn about the legacy format. */
 644
 645         if (a2time(&t, "%Y-%m-%d", in) == 0)
 646                 mandoc_msg(MANDOCERR_DATE_BAD, ln, pos, "%s", in);
 647         else if (t > time(NULL) + 86400)
 648                 mandoc_msg(MANDOCERR_DATE_FUTURE, ln, pos, "%s", in);
 649         else if (man->macroset == MACROSET_MDOC)
 650                 mandoc_msg(MANDOCERR_DATE_LEGACY, ln, pos, "Dd %s", in);
 651
 652         /* Use any non-mdoc(7) date verbatim. */
 653
 654         return mandoc_strdup(in);
 655 }
 656
 657 int
 658 mandoc_eos(const char *p, size_t sz)
 659 {
 660         const char      *q;
 661         int              enclosed, found;
 662
 663         if (0 == sz)
 664                 return 0;
 665
 666         /*
 667          * End-of-sentence recognition must include situations where
 668          * some symbols, such as `)', allow prior EOS punctuation to
 669          * propagate outward.
 670          */
 671
 672         enclosed = found = 0;
 673         for (q = p + (int)sz - 1; q >= p; q--) {
 674                 switch (*q) {
 675                 case '\"':
 676                 case '\'':
 677                 case ']':
 678                 case ')':
 679                         if (0 == found)
 680                                 enclosed = 1;
 681                         break;
 682                 case '.':
 683                 case '!':
 684                 case '?':
 685                         found = 1;
 686                         break;
 687                 default:
 688                         return found &&
 689                             (!enclosed || isalnum((unsigned char)*q));
 690                 }
 691         }
 692
 693         return found && !enclosed;
 694 }
 695
 696 /*
 697  * Convert a string to a long that may not be <0.
 698  * If the string is invalid, or is less than 0, return -1.
 699  */
 700 int
 701 mandoc_strntoi(const char *p, size_t sz, int base)
 702 {
 703         char             buf[32];
 704         char            *ep;
 705         long             v;
 706
 707         if (sz > 31)
 708                 return -1;
 709
 710         memcpy(buf, p, sz);
 711         buf[(int)sz] = '\0';
 712
 713         errno = 0;
 714         v = strtol(buf, &ep, base);
 715
 716         if (buf[0] == '\0' || *ep != '\0')
 717                 return -1;
 718
 719         if (v > INT_MAX)
 720                 v = INT_MAX;
 721         if (v < INT_MIN)
 722                 v = INT_MIN;
 723
 724         return (int)v;
 725 }