contrib/mdocml/preconv.c

   1 /*      $Id: preconv.c,v 1.6 2013/06/02 03:52:21 schwarze Exp $ */
   2 /*
   3  * Copyright (c) 2011 Kristaps Dzonsons <kristaps@bsd.lv>
   4  *
   5  * Permission to use, copy, modify, and distribute this software for any
   6  * purpose with or without fee is hereby granted, provided that the above
   7  * copyright notice and this permission notice appear in all copies.
   8  *
   9  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  10  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  11  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
  12  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  13  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  14  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  15  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  16  */
  17 #ifdef HAVE_CONFIG_H
  18 #include "config.h"
  19 #endif
  20
  21 #ifdef HAVE_MMAP
  22 #include <sys/stat.h>
  23 #include <sys/mman.h>
  24 #endif
  25
  26 #include <assert.h>
  27 #include <fcntl.h>
  28 #include <stdio.h>
  29 #include <stdlib.h>
  30 #include <string.h>
  31 #include <unistd.h>
  32
  33 /*
  34  * The read_whole_file() and resize_buf() functions are copied from
  35  * read.c, including all dependency code.
  36  */
  37
  38 enum    enc {
  39         ENC_UTF_8, /* UTF-8 */
  40         ENC_US_ASCII, /* US-ASCII */
  41         ENC_LATIN_1, /* Latin-1 */
  42         ENC__MAX
  43 };
  44
  45 struct  buf {
  46         char             *buf; /* binary input buffer */
  47         size_t            sz; /* size of binary buffer */
  48         size_t            offs; /* starting buffer offset */
  49 };
  50
  51 struct  encode {
  52         const char       *name;
  53         int             (*conv)(const struct buf *);
  54 };
  55
  56 static  int      cue_enc(const struct buf *, size_t *, enum enc *);
  57 static  int      conv_latin_1(const struct buf *);
  58 static  int      conv_us_ascii(const struct buf *);
  59 static  int      conv_utf_8(const struct buf *);
  60 static  int      read_whole_file(const char *, int,
  61                         struct buf *, int *);
  62 static  void     resize_buf(struct buf *, size_t);
  63 static  void     usage(void);
  64
  65 static  const struct encode encs[ENC__MAX] = {
  66         { "utf-8", conv_utf_8 }, /* ENC_UTF_8 */
  67         { "us-ascii", conv_us_ascii }, /* ENC_US_ASCII */
  68         { "latin-1", conv_latin_1 }, /* ENC_LATIN_1 */
  69 };
  70
  71 static  const char       *progname;
  72
  73 static void
  74 usage(void)
  75 {
  76
  77         fprintf(stderr, "usage: %s "
  78                         "[-D enc] "
  79                         "[-e ENC] "
  80                         "[file]\n", progname);
  81 }
  82
  83 static int
  84 conv_latin_1(const struct buf *b)
  85 {
  86         size_t           i;
  87         unsigned char    cu;
  88         const char      *cp;
  89
  90         cp = b->buf + (int)b->offs;
  91
  92         /*
  93          * Latin-1 falls into the first 256 code-points of Unicode, so
  94          * there's no need for any sort of translation.  Just make the
  95          * 8-bit characters use the Unicode escape.
  96          * Note that binary values 128 < v < 160 are passed through
  97          * unmodified to mandoc.
  98          */
  99
 100         for (i = b->offs; i < b->sz; i++) {
 101                 cu = (unsigned char)*cp++;
 102                 cu < 160U ? putchar(cu) : printf("\\[u%.4X]", cu);
 103         }
 104
 105         return(1);
 106 }
 107
 108 static int
 109 conv_us_ascii(const struct buf *b)
 110 {
 111
 112         /*
 113          * US-ASCII has no conversion since it falls into the first 128
 114          * bytes of Unicode.
 115          */
 116
 117         fwrite(b->buf, 1, b->sz, stdout);
 118         return(1);
 119 }
 120
 121 static int
 122 conv_utf_8(const struct buf *b)
 123 {
 124         int              state, be;
 125         unsigned int     accum;
 126         size_t           i;
 127         unsigned char    cu;
 128         const char      *cp;
 129         const long       one = 1L;
 130
 131         cp = b->buf + (int)b->offs;
 132         state = 0;
 133         accum = 0U;
 134         be = 0;
 135
 136         /* Quick test for big-endian value. */
 137
 138         if ( ! (*((const char *)(&one))))
 139                 be = 1;
 140
 141         for (i = b->offs; i < b->sz; i++) {
 142                 cu = (unsigned char)*cp++;
 143                 if (state) {
 144                         if ( ! (cu & 128) || (cu & 64)) {
 145                                 /* Bad sequence header. */
 146                                 return(0);
 147                         }
 148
 149                         /* Accept only legitimate bit patterns. */
 150
 151                         if (cu > 191 || cu < 128) {
 152                                 /* Bad in-sequence bits. */
 153                                 return(0);
 154                         }
 155
 156                         accum |= (cu & 63) << --state * 6;
 157
 158                         /*
 159                          * Accum is held in little-endian order as
 160                          * stipulated by the UTF-8 sequence coding.  We
 161                          * need to convert to a native big-endian if our
 162                          * architecture requires it.
 163                          */
 164
 165                         if (0 == state && be)
 166                                 accum = (accum >> 24) |
 167                                         ((accum << 8) & 0x00FF0000) |
 168                                         ((accum >> 8) & 0x0000FF00) |
 169                                         (accum << 24);
 170
 171                         if (0 == state) {
 172                                 accum < 128U ? putchar(accum) :
 173                                         printf("\\[u%.4X]", accum);
 174                                 accum = 0U;
 175                         }
 176                 } else if (cu & (1 << 7)) {
 177                         /*
 178                          * Entering a UTF-8 state:  if we encounter a
 179                          * UTF-8 bitmask, calculate the expected UTF-8
 180                          * state from it.
 181                          */
 182                         for (state = 0; state < 7; state++)
 183                                 if ( ! (cu & (1 << (7 - state))))
 184                                         break;
 185
 186                         /* Accept only legitimate bit patterns. */
 187
 188                         switch (state) {
 189                         case (4):
 190                                 if (cu <= 244 && cu >= 240) {
 191                                         accum = (cu & 7) << 18;
 192                                         break;
 193                                 }
 194                                 /* Bad 4-sequence start bits. */
 195                                 return(0);
 196                         case (3):
 197                                 if (cu <= 239 && cu >= 224) {
 198                                         accum = (cu & 15) << 12;
 199                                         break;
 200                                 }
 201                                 /* Bad 3-sequence start bits. */
 202                                 return(0);
 203                         case (2):
 204                                 if (cu <= 223 && cu >= 194) {
 205                                         accum = (cu & 31) << 6;
 206                                         break;
 207                                 }
 208                                 /* Bad 2-sequence start bits. */
 209                                 return(0);
 210                         default:
 211                                 /* Bad sequence bit mask. */
 212                                 return(0);
 213                         }
 214                         state--;
 215                 } else
 216                         putchar(cu);
 217         }
 218
 219         if (0 != state) {
 220                 /* Bad trailing bits. */
 221                 return(0);
 222         }
 223
 224         return(1);
 225 }
 226
 227 static void
 228 resize_buf(struct buf *buf, size_t initial)
 229 {
 230
 231         buf->sz = buf->sz > initial / 2 ?
 232                 2 * buf->sz : initial;
 233
 234         buf->buf = realloc(buf->buf, buf->sz);
 235         if (NULL == buf->buf) {
 236                 perror(NULL);
 237                 exit(EXIT_FAILURE);
 238         }
 239 }
 240
 241 static int
 242 read_whole_file(const char *f, int fd,
 243                 struct buf *fb, int *with_mmap)
 244 {
 245         size_t           off;
 246         ssize_t          ssz;
 247
 248 #ifdef  HAVE_MMAP
 249         struct stat      st;
 250         if (-1 == fstat(fd, &st)) {
 251                 perror(f);
 252                 return(0);
 253         }
 254
 255         /*
 256          * If we're a regular file, try just reading in the whole entry
 257          * via mmap().  This is faster than reading it into blocks, and
 258          * since each file is only a few bytes to begin with, I'm not
 259          * concerned that this is going to tank any machines.
 260          */
 261
 262         if (S_ISREG(st.st_mode) && st.st_size >= (1U << 31)) {
 263                 fprintf(stderr, "%s: input too large\n", f);
 264                 return(0);
 265         }
 266
 267         if (S_ISREG(st.st_mode)) {
 268                 *with_mmap = 1;
 269                 fb->sz = (size_t)st.st_size;
 270                 fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
 271                 if (fb->buf != MAP_FAILED)
 272                         return(1);
 273         }
 274 #endif
 275
 276         /*
 277          * If this isn't a regular file (like, say, stdin), then we must
 278          * go the old way and just read things in bit by bit.
 279          */
 280
 281         *with_mmap = 0;
 282         off = 0;
 283         fb->sz = 0;
 284         fb->buf = NULL;
 285         for (;;) {
 286                 if (off == fb->sz && fb->sz == (1U << 31)) {
 287                         fprintf(stderr, "%s: input too large\n", f);
 288                         break;
 289                 }
 290
 291                 if (off == fb->sz)
 292                         resize_buf(fb, 65536);
 293
 294                 ssz = read(fd, fb->buf + (int)off, fb->sz - off);
 295                 if (ssz == 0) {
 296                         fb->sz = off;
 297                         return(1);
 298                 }
 299                 if (ssz == -1) {
 300                         perror(f);
 301                         break;
 302                 }
 303                 off += (size_t)ssz;
 304         }
 305
 306         free(fb->buf);
 307         fb->buf = NULL;
 308         return(0);
 309 }
 310
 311 static int
 312 cue_enc(const struct buf *b, size_t *offs, enum enc *enc)
 313 {
 314         const char      *ln, *eoln, *eoph;
 315         size_t           sz, phsz, nsz;
 316         int              i;
 317
 318         ln = b->buf + (int)*offs;
 319         sz = b->sz - *offs;
 320
 321         /* Look for the end-of-line. */
 322
 323         if (NULL == (eoln = memchr(ln, '\n', sz)))
 324                 return(-1);
 325
 326         /* Set next-line marker. */
 327
 328         *offs = (size_t)((eoln + 1) - b->buf);
 329
 330         /* Check if we have the correct header/trailer. */
 331
 332         if ((sz = (size_t)(eoln - ln)) < 10 ||
 333                         memcmp(ln, ".\\\" -*-", 7) ||
 334                         memcmp(eoln - 3, "-*-", 3))
 335                 return(0);
 336
 337         /* Move after the header and adjust for the trailer. */
 338
 339         ln += 7;
 340         sz -= 10;
 341
 342         while (sz > 0) {
 343                 while (sz > 0 && ' ' == *ln) {
 344                         ln++;
 345                         sz--;
 346                 }
 347                 if (0 == sz)
 348                         break;
 349
 350                 /* Find the end-of-phrase marker (or eoln). */
 351
 352                 if (NULL == (eoph = memchr(ln, ';', sz)))
 353                         eoph = eoln - 3;
 354                 else
 355                         eoph++;
 356
 357                 /* Only account for the "coding" phrase. */
 358
 359                 if ((phsz = (size_t)(eoph - ln)) < 7 ||
 360                                 strncasecmp(ln, "coding:", 7)) {
 361                         sz -= phsz;
 362                         ln += phsz;
 363                         continue;
 364                 }
 365
 366                 sz -= 7;
 367                 ln += 7;
 368
 369                 while (sz > 0 && ' ' == *ln) {
 370                         ln++;
 371                         sz--;
 372                 }
 373                 if (0 == sz)
 374                         break;
 375
 376                 /* Check us against known encodings. */
 377
 378                 for (i = 0; i < (int)ENC__MAX; i++) {
 379                         nsz = strlen(encs[i].name);
 380                         if (phsz < nsz)
 381                                 continue;
 382                         if (strncasecmp(ln, encs[i].name, nsz))
 383                                 continue;
 384
 385                         *enc = (enum enc)i;
 386                         return(1);
 387                 }
 388
 389                 /* Unknown encoding. */
 390
 391                 *enc = ENC__MAX;
 392                 return(1);
 393         }
 394
 395         return(0);
 396 }
 397
 398 int
 399 main(int argc, char *argv[])
 400 {
 401         int              i, ch, map, fd, rc;
 402         struct buf       b;
 403         const char      *fn;
 404         enum enc         enc, def;
 405         unsigned char    bom[3] = { 0xEF, 0xBB, 0xBF };
 406         size_t           offs;
 407         extern int       optind;
 408         extern char     *optarg;
 409
 410         progname = strrchr(argv[0], '/');
 411         if (progname == NULL)
 412                 progname = argv[0];
 413         else
 414                 ++progname;
 415
 416         fn = "<stdin>";
 417         fd = STDIN_FILENO;
 418         rc = EXIT_FAILURE;
 419         enc = def = ENC__MAX;
 420         map = 0;
 421
 422         memset(&b, 0, sizeof(struct buf));
 423
 424         while (-1 != (ch = getopt(argc, argv, "D:e:rdvh")))
 425                 switch (ch) {
 426                 case ('D'):
 427                         /* FALLTHROUGH */
 428                 case ('e'):
 429                         for (i = 0; i < (int)ENC__MAX; i++) {
 430                                 if (strcasecmp(optarg, encs[i].name))
 431                                         continue;
 432                                 break;
 433                         }
 434                         if (i < (int)ENC__MAX) {
 435                                 if ('D' == ch)
 436                                         def = (enum enc)i;
 437                                 else
 438                                         enc = (enum enc)i;
 439                                 break;
 440                         }
 441
 442                         fprintf(stderr, "%s: Bad encoding\n", optarg);
 443                         return(EXIT_FAILURE);
 444                 case ('r'):
 445                         /* FALLTHROUGH */
 446                 case ('d'):
 447                         /* FALLTHROUGH */
 448                 case ('v'):
 449                         /* Compatibility with GNU preconv. */
 450                         break;
 451                 case ('h'):
 452                         /* Compatibility with GNU preconv. */
 453                         /* FALLTHROUGH */
 454                 default:
 455                         usage();
 456                         return(EXIT_FAILURE);
 457                 }
 458
 459         argc -= optind;
 460         argv += optind;
 461
 462         /*
 463          * Open and read the first argument on the command-line.
 464          * If we don't have one, we default to stdin.
 465          */
 466
 467         if (argc > 0) {
 468                 fn = *argv;
 469                 fd = open(fn, O_RDONLY, 0);
 470                 if (-1 == fd) {
 471                         perror(fn);
 472                         return(EXIT_FAILURE);
 473                 }
 474         }
 475
 476         if ( ! read_whole_file(fn, fd, &b, &map))
 477                 goto out;
 478
 479         /* Try to read the UTF-8 BOM. */
 480
 481         if (ENC__MAX == enc)
 482                 if (b.sz > 3 && 0 == memcmp(b.buf, bom, 3)) {
 483                         b.offs = 3;
 484                         enc = ENC_UTF_8;
 485                 }
 486
 487         /* Try reading from the "-*-" cue. */
 488
 489         if (ENC__MAX == enc) {
 490                 offs = b.offs;
 491                 ch = cue_enc(&b, &offs, &enc);
 492                 if (0 == ch)
 493                         ch = cue_enc(&b, &offs, &enc);
 494         }
 495
 496         /*
 497          * No encoding has been detected.
 498          * Thus, we either fall into our default encoder, if specified,
 499          * or use Latin-1 if all else fails.
 500          */
 501
 502         if (ENC__MAX == enc)
 503                 enc = ENC__MAX == def ? ENC_LATIN_1 : def;
 504
 505         if ( ! (*encs[(int)enc].conv)(&b)) {
 506                 fprintf(stderr, "%s: Bad encoding\n", fn);
 507                 goto out;
 508         }
 509
 510         rc = EXIT_SUCCESS;
 511 out:
 512 #ifdef  HAVE_MMAP
 513         if (map)
 514                 munmap(b.buf, b.sz);
 515         else
 516 #endif
 517                 free(b.buf);
 518
 519         if (fd > STDIN_FILENO)
 520                 close(fd);
 521
 522         return(rc);
 523 }