1 /* $OpenBSD: read.c,v 1.172 2018/08/25 16:43:52 schwarze Exp $ */ 2 /* 3 * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv> 4 * Copyright (c) 2010-2018 Ingo Schwarze <schwarze@openbsd.org> 5 * Copyright (c) 2010, 2012 Joerg Sonnenberger <joerg@netbsd.org> 6 * 7 * Permission to use, copy, modify, and distribute this software for any 8 * purpose with or without fee is hereby granted, provided that the above 9 * copyright notice and this permission notice appear in all copies. 10 * 11 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES 12 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF 13 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR 14 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES 15 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN 16 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF 17 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. 18 */ 19 #include <sys/types.h> 20 #include <sys/mman.h> 21 #include <sys/stat.h> 22 23 #include <assert.h> 24 #include <ctype.h> 25 #include <errno.h> 26 #include <fcntl.h> 27 #include <stdarg.h> 28 #include <stdio.h> 29 #include <stdlib.h> 30 #include <string.h> 31 #include <unistd.h> 32 #include <zlib.h> 33 34 #include "mandoc_aux.h" 35 #include "mandoc.h" 36 #include "roff.h" 37 #include "mdoc.h" 38 #include "man.h" 39 #include "libmandoc.h" 40 41 #define REPARSE_LIMIT 1000 42 43 struct mparse { 44 struct roff *roff; /* roff parser (!NULL) */ 45 struct roff_man *man; /* man parser */ 46 char *sodest; /* filename pointed to by .so */ 47 const char *file; /* filename of current input file */ 48 struct buf *primary; /* buffer currently being parsed */ 49 struct buf *secondary; /* copy of top level input */ 50 struct buf *loop; /* open .while request line */ 51 const char *os_s; /* default operating system */ 52 mandocmsg mmsg; /* warning/error message handler */ 53 enum mandoclevel file_status; /* status of current parse */ 54 enum mandocerr mmin; /* ignore messages below this */ 55 int options; /* parser options */ 56 int gzip; /* current input file is gzipped */ 57 int filenc; /* encoding of the current file */ 58 int reparse_count; /* finite interp. stack */ 59 int line; /* line number in the file */ 60 }; 61 62 static void choose_parser(struct mparse *); 63 static void free_buf_list(struct buf *); 64 static void resize_buf(struct buf *, size_t); 65 static int mparse_buf_r(struct mparse *, struct buf, size_t, int); 66 static int read_whole_file(struct mparse *, const char *, int, 67 struct buf *, int *); 68 static void mparse_end(struct mparse *); 69 static void mparse_parse_buffer(struct mparse *, struct buf, 70 const char *); 71 72 static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = { 73 MANDOCERR_OK, 74 MANDOCERR_OK, 75 MANDOCERR_WARNING, 76 MANDOCERR_ERROR, 77 MANDOCERR_UNSUPP, 78 MANDOCERR_MAX, 79 MANDOCERR_MAX 80 }; 81 82 static const char * const mandocerrs[MANDOCERR_MAX] = { 83 "ok", 84 85 "base system convention", 86 87 "Mdocdate found", 88 "Mdocdate missing", 89 "unknown architecture", 90 "operating system explicitly specified", 91 "RCS id missing", 92 "referenced manual not found", 93 94 "generic style suggestion", 95 96 "legacy man(7) date format", 97 "normalizing date format to", 98 "lower case character in document title", 99 "duplicate RCS id", 100 "possible typo in section name", 101 "unterminated quoted argument", 102 "useless macro", 103 "consider using OS macro", 104 "errnos out of order", 105 "duplicate errno", 106 "trailing delimiter", 107 "no blank before trailing delimiter", 108 "fill mode already enabled, skipping", 109 "fill mode already disabled, skipping", 110 "verbatim \"--\", maybe consider using \\(em", 111 "function name without markup", 112 "whitespace at end of input line", 113 "bad comment style", 114 115 "generic warning", 116 117 /* related to the prologue */ 118 "missing manual title, using UNTITLED", 119 "missing manual title, using \"\"", 120 "missing manual section, using \"\"", 121 "unknown manual section", 122 "missing date, using today's date", 123 "cannot parse date, using it verbatim", 124 "date in the future, using it anyway", 125 "missing Os macro, using \"\"", 126 "late prologue macro", 127 "prologue macros out of order", 128 129 /* related to document structure */ 130 ".so is fragile, better use ln(1)", 131 "no document body", 132 "content before first section header", 133 "first section is not \"NAME\"", 134 "NAME section without Nm before Nd", 135 "NAME section without description", 136 "description not at the end of NAME", 137 "bad NAME section content", 138 "missing comma before name", 139 "missing description line, using \"\"", 140 "description line outside NAME section", 141 "sections out of conventional order", 142 "duplicate section title", 143 "unexpected section", 144 "cross reference to self", 145 "unusual Xr order", 146 "unusual Xr punctuation", 147 "AUTHORS section without An macro", 148 149 /* related to macros and nesting */ 150 "obsolete macro", 151 "macro neither callable nor escaped", 152 "skipping paragraph macro", 153 "moving paragraph macro out of list", 154 "skipping no-space macro", 155 "blocks badly nested", 156 "nested displays are not portable", 157 "moving content out of list", 158 "first macro on line", 159 "line scope broken", 160 "skipping blank line in line scope", 161 162 /* related to missing macro arguments */ 163 "skipping empty request", 164 "conditional request controls empty scope", 165 "skipping empty macro", 166 "empty block", 167 "empty argument, using 0n", 168 "missing display type, using -ragged", 169 "list type is not the first argument", 170 "missing -width in -tag list, using 6n", 171 "missing utility name, using \"\"", 172 "missing function name, using \"\"", 173 "empty head in list item", 174 "empty list item", 175 "missing argument, using next line", 176 "missing font type, using \\fR", 177 "unknown font type, using \\fR", 178 "nothing follows prefix", 179 "empty reference block", 180 "missing section argument", 181 "missing -std argument, adding it", 182 "missing option string, using \"\"", 183 "missing resource identifier, using \"\"", 184 "missing eqn box, using \"\"", 185 186 /* related to bad macro arguments */ 187 "duplicate argument", 188 "skipping duplicate argument", 189 "skipping duplicate display type", 190 "skipping duplicate list type", 191 "skipping -width argument", 192 "wrong number of cells", 193 "unknown AT&T UNIX version", 194 "comma in function argument", 195 "parenthesis in function name", 196 "unknown library name", 197 "invalid content in Rs block", 198 "invalid Boolean argument", 199 "argument contains two font escapes", 200 "unknown font, skipping request", 201 "odd number of characters in request", 202 203 /* related to plain text */ 204 "blank line in fill mode, using .sp", 205 "tab in filled text", 206 "new sentence, new line", 207 "invalid escape sequence", 208 "undefined string, using \"\"", 209 210 /* related to tables */ 211 "tbl line starts with span", 212 "tbl column starts with span", 213 "skipping vertical bar in tbl layout", 214 215 "generic error", 216 217 /* related to tables */ 218 "non-alphabetic character in tbl options", 219 "skipping unknown tbl option", 220 "missing tbl option argument", 221 "wrong tbl option argument size", 222 "empty tbl layout", 223 "invalid character in tbl layout", 224 "unmatched parenthesis in tbl layout", 225 "tbl without any data cells", 226 "ignoring data in spanned tbl cell", 227 "ignoring extra tbl data cells", 228 "data block open at end of tbl", 229 230 /* related to document structure and macros */ 231 NULL, 232 "duplicate prologue macro", 233 "skipping late title macro", 234 "input stack limit exceeded, infinite loop?", 235 "skipping bad character", 236 "skipping unknown macro", 237 "ignoring request outside macro", 238 "skipping insecure request", 239 "skipping item outside list", 240 "skipping column outside column list", 241 "skipping end of block that is not open", 242 "fewer RS blocks open, skipping", 243 "inserting missing end of block", 244 "appending missing end of block", 245 246 /* related to request and macro arguments */ 247 "escaped character not allowed in a name", 248 "using macro argument outside macro", 249 "argument number is not numeric", 250 "NOT IMPLEMENTED: Bd -file", 251 "skipping display without arguments", 252 "missing list type, using -item", 253 "argument is not numeric, using 1", 254 "argument is not a character", 255 "missing manual name, using \"\"", 256 "uname(3) system call failed, using UNKNOWN", 257 "unknown standard specifier", 258 "skipping request without numeric argument", 259 "excessive shift", 260 "NOT IMPLEMENTED: .so with absolute path or \"..\"", 261 ".so request failed", 262 "skipping all arguments", 263 "skipping excess arguments", 264 "divide by zero", 265 266 "unsupported feature", 267 "input too large", 268 "unsupported control character", 269 "unsupported roff request", 270 "nested .while loops", 271 "end of scope with open .while loop", 272 "end of .while loop in inner scope", 273 "cannot continue this .while loop", 274 "eqn delim option in tbl", 275 "unsupported tbl layout modifier", 276 "ignoring macro in table", 277 }; 278 279 static const char * const mandoclevels[MANDOCLEVEL_MAX] = { 280 "SUCCESS", 281 "STYLE", 282 "WARNING", 283 "ERROR", 284 "UNSUPP", 285 "BADARG", 286 "SYSERR" 287 }; 288 289 290 static void 291 resize_buf(struct buf *buf, size_t initial) 292 { 293 294 buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial; 295 buf->buf = mandoc_realloc(buf->buf, buf->sz); 296 } 297 298 static void 299 free_buf_list(struct buf *buf) 300 { 301 struct buf *tmp; 302 303 while (buf != NULL) { 304 tmp = buf; 305 buf = tmp->next; 306 free(tmp->buf); 307 free(tmp); 308 } 309 } 310 311 static void 312 choose_parser(struct mparse *curp) 313 { 314 char *cp, *ep; 315 int format; 316 317 /* 318 * If neither command line arguments -mdoc or -man select 319 * a parser nor the roff parser found a .Dd or .TH macro 320 * yet, look ahead in the main input buffer. 321 */ 322 323 if ((format = roff_getformat(curp->roff)) == 0) { 324 cp = curp->primary->buf; 325 ep = cp + curp->primary->sz; 326 while (cp < ep) { 327 if (*cp == '.' || *cp == '\'') { 328 cp++; 329 if (cp[0] == 'D' && cp[1] == 'd') { 330 format = MPARSE_MDOC; 331 break; 332 } 333 if (cp[0] == 'T' && cp[1] == 'H') { 334 format = MPARSE_MAN; 335 break; 336 } 337 } 338 cp = memchr(cp, '\n', ep - cp); 339 if (cp == NULL) 340 break; 341 cp++; 342 } 343 } 344 345 if (format == MPARSE_MDOC) { 346 curp->man->macroset = MACROSET_MDOC; 347 if (curp->man->mdocmac == NULL) 348 curp->man->mdocmac = roffhash_alloc(MDOC_Dd, MDOC_MAX); 349 } else { 350 curp->man->macroset = MACROSET_MAN; 351 if (curp->man->manmac == NULL) 352 curp->man->manmac = roffhash_alloc(MAN_TH, MAN_MAX); 353 } 354 curp->man->first->tok = TOKEN_NONE; 355 } 356 357 /* 358 * Main parse routine for a buffer. 359 * It assumes encoding and line numbering are already set up. 360 * It can recurse directly (for invocations of user-defined 361 * macros, inline equations, and input line traps) 362 * and indirectly (for .so file inclusion). 363 */ 364 static int 365 mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start) 366 { 367 struct buf ln; 368 struct buf *firstln, *lastln, *thisln, *loop; 369 const char *save_file; 370 char *cp; 371 size_t pos; /* byte number in the ln buffer */ 372 int line_result, result; 373 int of; 374 int lnn; /* line number in the real file */ 375 int fd; 376 int inloop; /* Saw .while on this level. */ 377 unsigned char c; 378 379 ln.sz = 256; 380 ln.buf = mandoc_malloc(ln.sz); 381 ln.next = NULL; 382 firstln = loop = NULL; 383 lnn = curp->line; 384 pos = 0; 385 inloop = 0; 386 result = ROFF_CONT; 387 388 while (i < blk.sz && (blk.buf[i] != '\0' || pos != 0)) { 389 if (start) { 390 curp->line = lnn; 391 curp->reparse_count = 0; 392 393 if (lnn < 3 && 394 curp->filenc & MPARSE_UTF8 && 395 curp->filenc & MPARSE_LATIN1) 396 curp->filenc = preconv_cue(&blk, i); 397 } 398 399 while (i < blk.sz && (start || blk.buf[i] != '\0')) { 400 401 /* 402 * When finding an unescaped newline character, 403 * leave the character loop to process the line. 404 * Skip a preceding carriage return, if any. 405 */ 406 407 if ('\r' == blk.buf[i] && i + 1 < blk.sz && 408 '\n' == blk.buf[i + 1]) 409 ++i; 410 if ('\n' == blk.buf[i]) { 411 ++i; 412 ++lnn; 413 break; 414 } 415 416 /* 417 * Make sure we have space for the worst 418 * case of 12 bytes: "\\[u10ffff]\n\0" 419 */ 420 421 if (pos + 12 > ln.sz) 422 resize_buf(&ln, 256); 423 424 /* 425 * Encode 8-bit input. 426 */ 427 428 c = blk.buf[i]; 429 if (c & 0x80) { 430 if ( ! (curp->filenc && preconv_encode( 431 &blk, &i, &ln, &pos, &curp->filenc))) { 432 mandoc_vmsg(MANDOCERR_CHAR_BAD, curp, 433 curp->line, pos, "0x%x", c); 434 ln.buf[pos++] = '?'; 435 i++; 436 } 437 continue; 438 } 439 440 /* 441 * Exclude control characters. 442 */ 443 444 if (c == 0x7f || (c < 0x20 && c != 0x09)) { 445 mandoc_vmsg(c == 0x00 || c == 0x04 || 446 c > 0x0a ? MANDOCERR_CHAR_BAD : 447 MANDOCERR_CHAR_UNSUPP, 448 curp, curp->line, pos, "0x%x", c); 449 i++; 450 if (c != '\r') 451 ln.buf[pos++] = '?'; 452 continue; 453 } 454 455 ln.buf[pos++] = blk.buf[i++]; 456 } 457 ln.buf[pos] = '\0'; 458 459 /* 460 * Maintain a lookaside buffer of all lines. 461 * parsed from this input source. 462 */ 463 464 thisln = mandoc_malloc(sizeof(*thisln)); 465 thisln->buf = mandoc_strdup(ln.buf); 466 thisln->sz = strlen(ln.buf) + 1; 467 thisln->next = NULL; 468 if (firstln == NULL) { 469 firstln = lastln = thisln; 470 if (curp->secondary == NULL) 471 curp->secondary = firstln; 472 } else { 473 lastln->next = thisln; 474 lastln = thisln; 475 } 476 477 /* XXX Ugly hack to mark the end of the input. */ 478 479 if (i == blk.sz || blk.buf[i] == '\0') { 480 ln.buf[pos++] = '\n'; 481 ln.buf[pos] = '\0'; 482 } 483 484 /* 485 * A significant amount of complexity is contained by 486 * the roff preprocessor. It's line-oriented but can be 487 * expressed on one line, so we need at times to 488 * readjust our starting point and re-run it. The roff 489 * preprocessor can also readjust the buffers with new 490 * data, so we pass them in wholesale. 491 */ 492 493 of = 0; 494 rerun: 495 line_result = roff_parseln(curp->roff, curp->line, &ln, &of); 496 497 /* Process options. */ 498 499 if (line_result & ROFF_APPEND) 500 assert(line_result == (ROFF_IGN | ROFF_APPEND)); 501 502 if (line_result & ROFF_USERCALL) 503 assert((line_result & ROFF_MASK) == ROFF_REPARSE); 504 505 if (line_result & ROFF_USERRET) { 506 assert(line_result == (ROFF_IGN | ROFF_USERRET)); 507 if (start == 0) { 508 /* Return from the current macro. */ 509 result = ROFF_USERRET; 510 goto out; 511 } 512 } 513 514 switch (line_result & ROFF_LOOPMASK) { 515 case ROFF_IGN: 516 break; 517 case ROFF_WHILE: 518 if (curp->loop != NULL) { 519 if (loop == curp->loop) 520 break; 521 mandoc_msg(MANDOCERR_WHILE_NEST, 522 curp, curp->line, pos, NULL); 523 } 524 curp->loop = thisln; 525 loop = NULL; 526 inloop = 1; 527 break; 528 case ROFF_LOOPCONT: 529 case ROFF_LOOPEXIT: 530 if (curp->loop == NULL) { 531 mandoc_msg(MANDOCERR_WHILE_FAIL, 532 curp, curp->line, pos, NULL); 533 break; 534 } 535 if (inloop == 0) { 536 mandoc_msg(MANDOCERR_WHILE_INTO, 537 curp, curp->line, pos, NULL); 538 curp->loop = loop = NULL; 539 break; 540 } 541 if (line_result & ROFF_LOOPCONT) 542 loop = curp->loop; 543 else { 544 curp->loop = loop = NULL; 545 inloop = 0; 546 } 547 break; 548 default: 549 abort(); 550 } 551 552 /* Process the main instruction from the roff parser. */ 553 554 switch (line_result & ROFF_MASK) { 555 case ROFF_IGN: 556 break; 557 case ROFF_CONT: 558 if (curp->man->macroset == MACROSET_NONE) 559 choose_parser(curp); 560 if ((curp->man->macroset == MACROSET_MDOC ? 561 mdoc_parseln(curp->man, curp->line, ln.buf, of) : 562 man_parseln(curp->man, curp->line, ln.buf, of) 563 ) == 2) 564 goto out; 565 break; 566 case ROFF_RERUN: 567 goto rerun; 568 case ROFF_REPARSE: 569 if (++curp->reparse_count > REPARSE_LIMIT) { 570 /* Abort and return to the top level. */ 571 result = ROFF_IGN; 572 mandoc_msg(MANDOCERR_ROFFLOOP, curp, 573 curp->line, pos, NULL); 574 goto out; 575 } 576 result = mparse_buf_r(curp, ln, of, 0); 577 if (line_result & ROFF_USERCALL) { 578 roff_userret(curp->roff); 579 /* Continue normally. */ 580 if (result & ROFF_USERRET) 581 result = ROFF_CONT; 582 } 583 if (start == 0 && result != ROFF_CONT) 584 goto out; 585 break; 586 case ROFF_SO: 587 if ( ! (curp->options & MPARSE_SO) && 588 (i >= blk.sz || blk.buf[i] == '\0')) { 589 curp->sodest = mandoc_strdup(ln.buf + of); 590 goto out; 591 } 592 save_file = curp->file; 593 if ((fd = mparse_open(curp, ln.buf + of)) != -1) { 594 mparse_readfd(curp, fd, ln.buf + of); 595 close(fd); 596 curp->file = save_file; 597 } else { 598 curp->file = save_file; 599 mandoc_vmsg(MANDOCERR_SO_FAIL, 600 curp, curp->line, pos, 601 ".so %s", ln.buf + of); 602 ln.sz = mandoc_asprintf(&cp, 603 ".sp\nSee the file %s.\n.sp", 604 ln.buf + of); 605 free(ln.buf); 606 ln.buf = cp; 607 of = 0; 608 mparse_buf_r(curp, ln, of, 0); 609 } 610 break; 611 default: 612 abort(); 613 } 614 615 /* Start the next input line. */ 616 617 if (loop != NULL && 618 (line_result & ROFF_LOOPMASK) == ROFF_IGN) 619 loop = loop->next; 620 621 if (loop != NULL) { 622 if ((line_result & ROFF_APPEND) == 0) 623 *ln.buf = '\0'; 624 if (ln.sz < loop->sz) 625 resize_buf(&ln, loop->sz); 626 (void)strlcat(ln.buf, loop->buf, ln.sz); 627 of = 0; 628 goto rerun; 629 } 630 631 pos = (line_result & ROFF_APPEND) ? strlen(ln.buf) : 0; 632 } 633 out: 634 if (inloop) { 635 if (result != ROFF_USERRET) 636 mandoc_msg(MANDOCERR_WHILE_OUTOF, curp, 637 curp->line, pos, NULL); 638 curp->loop = NULL; 639 } 640 free(ln.buf); 641 if (firstln != curp->secondary) 642 free_buf_list(firstln); 643 return result; 644 } 645 646 static int 647 read_whole_file(struct mparse *curp, const char *file, int fd, 648 struct buf *fb, int *with_mmap) 649 { 650 struct stat st; 651 gzFile gz; 652 size_t off; 653 ssize_t ssz; 654 int gzerrnum, retval; 655 656 if (fstat(fd, &st) == -1) { 657 mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0, 658 "fstat: %s", strerror(errno)); 659 return 0; 660 } 661 662 /* 663 * If we're a regular file, try just reading in the whole entry 664 * via mmap(). This is faster than reading it into blocks, and 665 * since each file is only a few bytes to begin with, I'm not 666 * concerned that this is going to tank any machines. 667 */ 668 669 if (curp->gzip == 0 && S_ISREG(st.st_mode)) { 670 if (st.st_size > 0x7fffffff) { 671 mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL); 672 return 0; 673 } 674 *with_mmap = 1; 675 fb->sz = (size_t)st.st_size; 676 fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0); 677 if (fb->buf != MAP_FAILED) 678 return 1; 679 } 680 681 if (curp->gzip) { 682 /* 683 * Duplicating the file descriptor is required 684 * because we will have to call gzclose(3) 685 * to free memory used internally by zlib, 686 * but that will also close the file descriptor, 687 * which this function must not do. 688 */ 689 if ((fd = dup(fd)) == -1) { 690 mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0, 691 "dup: %s", strerror(errno)); 692 return 0; 693 } 694 if ((gz = gzdopen(fd, "rb")) == NULL) { 695 mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0, 696 "gzdopen: %s", strerror(errno)); 697 close(fd); 698 return 0; 699 } 700 } else 701 gz = NULL; 702 703 /* 704 * If this isn't a regular file (like, say, stdin), then we must 705 * go the old way and just read things in bit by bit. 706 */ 707 708 *with_mmap = 0; 709 off = 0; 710 retval = 0; 711 fb->sz = 0; 712 fb->buf = NULL; 713 for (;;) { 714 if (off == fb->sz) { 715 if (fb->sz == (1U << 31)) { 716 mandoc_msg(MANDOCERR_TOOLARGE, curp, 717 0, 0, NULL); 718 break; 719 } 720 resize_buf(fb, 65536); 721 } 722 ssz = curp->gzip ? 723 gzread(gz, fb->buf + (int)off, fb->sz - off) : 724 read(fd, fb->buf + (int)off, fb->sz - off); 725 if (ssz == 0) { 726 fb->sz = off; 727 retval = 1; 728 break; 729 } 730 if (ssz == -1) { 731 if (curp->gzip) 732 (void)gzerror(gz, &gzerrnum); 733 mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0, "read: %s", 734 curp->gzip && gzerrnum != Z_ERRNO ? 735 zError(gzerrnum) : strerror(errno)); 736 break; 737 } 738 off += (size_t)ssz; 739 } 740 741 if (curp->gzip && (gzerrnum = gzclose(gz)) != Z_OK) 742 mandoc_vmsg(MANDOCERR_FILE, curp, 0, 0, "gzclose: %s", 743 gzerrnum == Z_ERRNO ? strerror(errno) : 744 zError(gzerrnum)); 745 if (retval == 0) { 746 free(fb->buf); 747 fb->buf = NULL; 748 } 749 return retval; 750 } 751 752 static void 753 mparse_end(struct mparse *curp) 754 { 755 if (curp->man->macroset == MACROSET_NONE) 756 curp->man->macroset = MACROSET_MAN; 757 if (curp->man->macroset == MACROSET_MDOC) 758 mdoc_endparse(curp->man); 759 else 760 man_endparse(curp->man); 761 roff_endparse(curp->roff); 762 } 763 764 static void 765 mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file) 766 { 767 struct buf *svprimary; 768 const char *svfile; 769 size_t offset; 770 static int recursion_depth; 771 772 if (64 < recursion_depth) { 773 mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL); 774 return; 775 } 776 777 /* Line number is per-file. */ 778 svfile = curp->file; 779 curp->file = file; 780 svprimary = curp->primary; 781 curp->primary = &blk; 782 curp->line = 1; 783 recursion_depth++; 784 785 /* Skip an UTF-8 byte order mark. */ 786 if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 && 787 (unsigned char)blk.buf[0] == 0xef && 788 (unsigned char)blk.buf[1] == 0xbb && 789 (unsigned char)blk.buf[2] == 0xbf) { 790 offset = 3; 791 curp->filenc &= ~MPARSE_LATIN1; 792 } else 793 offset = 0; 794 795 mparse_buf_r(curp, blk, offset, 1); 796 797 if (--recursion_depth == 0) 798 mparse_end(curp); 799 800 curp->primary = svprimary; 801 curp->file = svfile; 802 } 803 804 /* 805 * Read the whole file into memory and call the parsers. 806 * Called recursively when an .so request is encountered. 807 */ 808 enum mandoclevel 809 mparse_readfd(struct mparse *curp, int fd, const char *file) 810 { 811 struct buf blk; 812 int with_mmap; 813 int save_filenc; 814 815 if (read_whole_file(curp, file, fd, &blk, &with_mmap)) { 816 save_filenc = curp->filenc; 817 curp->filenc = curp->options & 818 (MPARSE_UTF8 | MPARSE_LATIN1); 819 mparse_parse_buffer(curp, blk, file); 820 curp->filenc = save_filenc; 821 if (with_mmap) 822 munmap(blk.buf, blk.sz); 823 else 824 free(blk.buf); 825 } 826 return curp->file_status; 827 } 828 829 int 830 mparse_open(struct mparse *curp, const char *file) 831 { 832 char *cp; 833 int fd; 834 835 curp->file = file; 836 cp = strrchr(file, '.'); 837 curp->gzip = (cp != NULL && ! strcmp(cp + 1, "gz")); 838 839 /* First try to use the filename as it is. */ 840 841 if ((fd = open(file, O_RDONLY)) != -1) 842 return fd; 843 844 /* 845 * If that doesn't work and the filename doesn't 846 * already end in .gz, try appending .gz. 847 */ 848 849 if ( ! curp->gzip) { 850 mandoc_asprintf(&cp, "%s.gz", file); 851 fd = open(cp, O_RDONLY); 852 free(cp); 853 if (fd != -1) { 854 curp->gzip = 1; 855 return fd; 856 } 857 } 858 859 /* Neither worked, give up. */ 860 861 mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno)); 862 return -1; 863 } 864 865 struct mparse * 866 mparse_alloc(int options, enum mandocerr mmin, mandocmsg mmsg, 867 enum mandoc_os os_e, const char *os_s) 868 { 869 struct mparse *curp; 870 871 curp = mandoc_calloc(1, sizeof(struct mparse)); 872 873 curp->options = options; 874 curp->mmin = mmin; 875 curp->mmsg = mmsg; 876 curp->os_s = os_s; 877 878 curp->roff = roff_alloc(curp, options); 879 curp->man = roff_man_alloc(curp->roff, curp, curp->os_s, 880 curp->options & MPARSE_QUICK ? 1 : 0); 881 if (curp->options & MPARSE_MDOC) { 882 curp->man->macroset = MACROSET_MDOC; 883 if (curp->man->mdocmac == NULL) 884 curp->man->mdocmac = roffhash_alloc(MDOC_Dd, MDOC_MAX); 885 } else if (curp->options & MPARSE_MAN) { 886 curp->man->macroset = MACROSET_MAN; 887 if (curp->man->manmac == NULL) 888 curp->man->manmac = roffhash_alloc(MAN_TH, MAN_MAX); 889 } 890 curp->man->first->tok = TOKEN_NONE; 891 curp->man->meta.os_e = os_e; 892 return curp; 893 } 894 895 void 896 mparse_reset(struct mparse *curp) 897 { 898 roff_reset(curp->roff); 899 roff_man_reset(curp->man); 900 free_buf_list(curp->secondary); 901 curp->secondary = NULL; 902 903 free(curp->sodest); 904 curp->sodest = NULL; 905 906 curp->file_status = MANDOCLEVEL_OK; 907 curp->gzip = 0; 908 } 909 910 void 911 mparse_free(struct mparse *curp) 912 { 913 roffhash_free(curp->man->mdocmac); 914 roffhash_free(curp->man->manmac); 915 roff_man_free(curp->man); 916 roff_free(curp->roff); 917 free_buf_list(curp->secondary); 918 free(curp->sodest); 919 free(curp); 920 } 921 922 void 923 mparse_result(struct mparse *curp, struct roff_man **man, 924 char **sodest) 925 { 926 927 if (sodest && NULL != (*sodest = curp->sodest)) { 928 *man = NULL; 929 return; 930 } 931 if (man) 932 *man = curp->man; 933 } 934 935 void 936 mparse_updaterc(struct mparse *curp, enum mandoclevel *rc) 937 { 938 if (curp->file_status > *rc) 939 *rc = curp->file_status; 940 } 941 942 void 943 mandoc_vmsg(enum mandocerr t, struct mparse *m, 944 int ln, int pos, const char *fmt, ...) 945 { 946 char buf[256]; 947 va_list ap; 948 949 va_start(ap, fmt); 950 (void)vsnprintf(buf, sizeof(buf), fmt, ap); 951 va_end(ap); 952 953 mandoc_msg(t, m, ln, pos, buf); 954 } 955 956 void 957 mandoc_msg(enum mandocerr er, struct mparse *m, 958 int ln, int col, const char *msg) 959 { 960 enum mandoclevel level; 961 962 if (er < m->mmin && er != MANDOCERR_FILE) 963 return; 964 965 level = MANDOCLEVEL_UNSUPP; 966 while (er < mandoclimits[level]) 967 level--; 968 969 if (m->mmsg) 970 (*m->mmsg)(er, level, m->file, ln, col, msg); 971 972 if (m->file_status < level) 973 m->file_status = level; 974 } 975 976 const char * 977 mparse_strerror(enum mandocerr er) 978 { 979 980 return mandocerrs[er]; 981 } 982 983 const char * 984 mparse_strlevel(enum mandoclevel lvl) 985 { 986 return mandoclevels[lvl]; 987 } 988 989 void 990 mparse_copy(const struct mparse *p) 991 { 992 struct buf *buf; 993 994 for (buf = p->secondary; buf != NULL; buf = buf->next) 995 puts(buf->buf); 996 } 997