Annotation of src/usr.bin/mandoc/read.c, Revision 1.148
1.148 ! schwarze 1: /* $OpenBSD: read.c,v 1.147 2017/06/11 17:16:36 schwarze Exp $ */
1.1 schwarze 2: /*
3: * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
1.129 schwarze 4: * Copyright (c) 2010-2017 Ingo Schwarze <schwarze@openbsd.org>
1.20 schwarze 5: * Copyright (c) 2010, 2012 Joerg Sonnenberger <joerg@netbsd.org>
1.1 schwarze 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: *
1.108 schwarze 11: * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHORS DISCLAIM ALL WARRANTIES
1.1 schwarze 12: * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
1.108 schwarze 13: * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS BE LIABLE FOR
1.1 schwarze 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: */
1.58 schwarze 19: #include <sys/types.h>
20: #include <sys/mman.h>
1.1 schwarze 21: #include <sys/stat.h>
22:
23: #include <assert.h>
24: #include <ctype.h>
1.118 schwarze 25: #include <err.h>
1.18 schwarze 26: #include <errno.h>
1.1 schwarze 27: #include <fcntl.h>
28: #include <stdarg.h>
29: #include <stdio.h>
30: #include <stdlib.h>
31: #include <string.h>
32: #include <unistd.h>
1.115 schwarze 33: #include <zlib.h>
1.1 schwarze 34:
1.108 schwarze 35: #include "mandoc_aux.h"
1.1 schwarze 36: #include "mandoc.h"
1.108 schwarze 37: #include "roff.h"
1.1 schwarze 38: #include "mdoc.h"
39: #include "man.h"
1.108 schwarze 40: #include "libmandoc.h"
1.114 schwarze 41: #include "roff_int.h"
1.1 schwarze 42:
43: #define REPARSE_LIMIT 1000
44:
45: struct mparse {
1.133 schwarze 46: struct roff *roff; /* roff parser (!NULL) */
1.109 schwarze 47: struct roff_man *man; /* man parser */
1.23 schwarze 48: char *sodest; /* filename pointed to by .so */
1.59 schwarze 49: const char *file; /* filename of current input file */
50: struct buf *primary; /* buffer currently being parsed */
51: struct buf *secondary; /* preprocessed copy of input */
52: const char *defos; /* default operating system */
53: mandocmsg mmsg; /* warning/error message handler */
54: enum mandoclevel file_status; /* status of current parse */
55: enum mandoclevel wlevel; /* ignore messages below this */
56: int options; /* parser options */
1.115 schwarze 57: int gzip; /* current input file is gzipped */
1.70 schwarze 58: int filenc; /* encoding of the current file */
1.1 schwarze 59: int reparse_count; /* finite interp. stack */
1.59 schwarze 60: int line; /* line number in the file */
1.1 schwarze 61: };
62:
1.60 schwarze 63: static void choose_parser(struct mparse *);
1.1 schwarze 64: static void resize_buf(struct buf *, size_t);
1.135 schwarze 65: static int mparse_buf_r(struct mparse *, struct buf, size_t, int);
1.18 schwarze 66: static int read_whole_file(struct mparse *, const char *, int,
67: struct buf *, int *);
1.1 schwarze 68: static void mparse_end(struct mparse *);
1.15 schwarze 69: static void mparse_parse_buffer(struct mparse *, struct buf,
70: const char *);
1.1 schwarze 71:
72: static const enum mandocerr mandoclimits[MANDOCLEVEL_MAX] = {
73: MANDOCERR_OK,
1.137 schwarze 74: MANDOCERR_STYLE,
1.1 schwarze 75: MANDOCERR_WARNING,
76: MANDOCERR_ERROR,
1.87 schwarze 77: MANDOCERR_UNSUPP,
1.1 schwarze 78: MANDOCERR_MAX,
79: MANDOCERR_MAX
80: };
81:
82: static const char * const mandocerrs[MANDOCERR_MAX] = {
83: "ok",
84:
1.137 schwarze 85: "generic style suggestion",
1.138 schwarze 86:
1.148 ! schwarze 87: "Mdocdate found",
! 88: "Mdocdate missing",
! 89: "legacy man(7) date format",
1.138 schwarze 90: "useless macro",
1.139 schwarze 91: "consider using OS macro",
1.144 schwarze 92: "errnos out of order",
93: "duplicate errno",
1.140 schwarze 94: "description line ends with a full stop",
1.146 schwarze 95: "no blank before trailing delimiter",
1.147 schwarze 96: "function name without markup",
1.138 schwarze 97:
1.1 schwarze 98: "generic warning",
99:
100: /* related to the prologue */
1.57 schwarze 101: "missing manual title, using UNTITLED",
102: "missing manual title, using \"\"",
1.32 schwarze 103: "lower case character in document title",
1.57 schwarze 104: "missing manual section, using \"\"",
1.1 schwarze 105: "unknown manual section",
1.32 schwarze 106: "missing date, using today's date",
1.1 schwarze 107: "cannot parse date, using it verbatim",
1.57 schwarze 108: "missing Os macro, using \"\"",
109: "duplicate prologue macro",
110: "late prologue macro",
111: "skipping late title macro",
1.1 schwarze 112: "prologue macros out of order",
113:
114: /* related to document structure */
115: ".so is fragile, better use ln(1)",
1.28 schwarze 116: "no document body",
1.32 schwarze 117: "content before first section header",
118: "first section is not \"NAME\"",
1.129 schwarze 119: "NAME section without Nm before Nd",
1.103 schwarze 120: "NAME section without description",
121: "description not at the end of NAME",
122: "bad NAME section content",
1.129 schwarze 123: "missing comma before name",
1.96 schwarze 124: "missing description line, using \"\"",
1.134 schwarze 125: "description line outside NAME section",
1.1 schwarze 126: "sections out of conventional order",
1.32 schwarze 127: "duplicate section title",
128: "unexpected section",
1.63 schwarze 129: "unusual Xr order",
130: "unusual Xr punctuation",
1.62 schwarze 131: "AUTHORS section without An macro",
1.1 schwarze 132:
133: /* related to macros and nesting */
1.33 schwarze 134: "obsolete macro",
1.78 schwarze 135: "macro neither callable nor escaped",
1.1 schwarze 136: "skipping paragraph macro",
1.10 schwarze 137: "moving paragraph macro out of list",
1.1 schwarze 138: "skipping no-space macro",
139: "blocks badly nested",
140: "nested displays are not portable",
1.35 schwarze 141: "moving content out of list",
1.56 schwarze 142: "fill mode already enabled, skipping",
143: "fill mode already disabled, skipping",
1.1 schwarze 144: "line scope broken",
1.141 schwarze 145: "skipping blank line in line scope",
1.1 schwarze 146:
147: /* related to missing macro arguments */
1.36 schwarze 148: "skipping empty request",
149: "conditional request controls empty scope",
1.1 schwarze 150: "skipping empty macro",
1.99 schwarze 151: "empty block",
1.40 schwarze 152: "empty argument, using 0n",
1.38 schwarze 153: "missing display type, using -ragged",
154: "list type is not the first argument",
1.125 schwarze 155: "missing -width in -tag list, using 6n",
1.56 schwarze 156: "missing utility name, using \"\"",
1.98 schwarze 157: "missing function name, using \"\"",
1.38 schwarze 158: "empty head in list item",
159: "empty list item",
1.39 schwarze 160: "missing font type, using \\fR",
161: "unknown font type, using \\fR",
1.79 schwarze 162: "nothing follows prefix",
1.97 schwarze 163: "empty reference block",
1.128 schwarze 164: "missing section argument",
1.38 schwarze 165: "missing -std argument, adding it",
1.100 schwarze 166: "missing option string, using \"\"",
167: "missing resource identifier, using \"\"",
1.66 schwarze 168: "missing eqn box, using \"\"",
1.1 schwarze 169:
170: /* related to bad macro arguments */
1.42 schwarze 171: "unterminated quoted argument",
1.1 schwarze 172: "duplicate argument",
1.54 schwarze 173: "skipping duplicate argument",
1.41 schwarze 174: "skipping duplicate display type",
175: "skipping duplicate list type",
1.54 schwarze 176: "skipping -width argument",
1.101 schwarze 177: "wrong number of cells",
1.1 schwarze 178: "unknown AT&T UNIX version",
1.64 schwarze 179: "comma in function argument",
1.65 schwarze 180: "parenthesis in function name",
1.145 schwarze 181: "unknown library name",
1.45 schwarze 182: "invalid content in Rs block",
1.41 schwarze 183: "invalid Boolean argument",
184: "unknown font, skipping request",
1.101 schwarze 185: "odd number of characters in request",
1.1 schwarze 186:
187: /* related to plain text */
1.42 schwarze 188: "blank line in fill mode, using .sp",
189: "tab in filled text",
190: "whitespace at end of input line",
1.131 schwarze 191: "new sentence, new line",
1.1 schwarze 192: "bad comment style",
1.42 schwarze 193: "invalid escape sequence",
194: "undefined string, using \"\"",
1.3 schwarze 195:
1.93 schwarze 196: /* related to tables */
197: "tbl line starts with span",
198: "tbl column starts with span",
199: "skipping vertical bar in tbl layout",
200:
1.1 schwarze 201: "generic error",
1.3 schwarze 202:
1.1 schwarze 203: /* related to tables */
1.91 schwarze 204: "non-alphabetic character in tbl options",
205: "skipping unknown tbl option",
206: "missing tbl option argument",
207: "wrong tbl option argument size",
1.93 schwarze 208: "empty tbl layout",
209: "invalid character in tbl layout",
210: "unmatched parenthesis in tbl layout",
1.94 schwarze 211: "tbl without any data cells",
212: "ignoring data in spanned tbl cell",
213: "ignoring extra tbl data cells",
214: "data block open at end of tbl",
1.1 schwarze 215:
1.46 schwarze 216: /* related to document structure and macros */
1.83 schwarze 217: NULL,
1.1 schwarze 218: "input stack limit exceeded, infinite loop?",
219: "skipping bad character",
1.46 schwarze 220: "skipping unknown macro",
1.87 schwarze 221: "skipping insecure request",
1.51 schwarze 222: "skipping item outside list",
1.46 schwarze 223: "skipping column outside column list",
224: "skipping end of block that is not open",
1.90 schwarze 225: "fewer RS blocks open, skipping",
1.46 schwarze 226: "inserting missing end of block",
227: "appending missing end of block",
228:
229: /* related to request and macro arguments */
1.1 schwarze 230: "escaped character not allowed in a name",
1.75 schwarze 231: "NOT IMPLEMENTED: Bd -file",
1.120 schwarze 232: "skipping display without arguments",
1.49 schwarze 233: "missing list type, using -item",
1.143 schwarze 234: "argument is not numeric, using 1",
1.48 schwarze 235: "missing manual name, using \"\"",
1.49 schwarze 236: "uname(3) system call failed, using UNKNOWN",
1.41 schwarze 237: "unknown standard specifier",
1.49 schwarze 238: "skipping request without numeric argument",
1.84 schwarze 239: "NOT IMPLEMENTED: .so with absolute path or \"..\"",
1.85 schwarze 240: ".so request failed",
1.39 schwarze 241: "skipping all arguments",
242: "skipping excess arguments",
1.68 schwarze 243: "divide by zero",
1.87 schwarze 244:
245: "unsupported feature",
246: "input too large",
1.89 schwarze 247: "unsupported control character",
1.87 schwarze 248: "unsupported roff request",
1.94 schwarze 249: "eqn delim option in tbl",
1.93 schwarze 250: "unsupported tbl layout modifier",
1.87 schwarze 251: "ignoring macro in table",
1.1 schwarze 252: };
253:
254: static const char * const mandoclevels[MANDOCLEVEL_MAX] = {
255: "SUCCESS",
1.137 schwarze 256: "STYLE",
1.1 schwarze 257: "WARNING",
258: "ERROR",
1.87 schwarze 259: "UNSUPP",
1.1 schwarze 260: "BADARG",
261: "SYSERR"
262: };
263:
1.25 schwarze 264:
1.1 schwarze 265: static void
266: resize_buf(struct buf *buf, size_t initial)
267: {
268:
269: buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
270: buf->buf = mandoc_realloc(buf->buf, buf->sz);
271: }
272:
273: static void
1.60 schwarze 274: choose_parser(struct mparse *curp)
1.1 schwarze 275: {
1.59 schwarze 276: char *cp, *ep;
277: int format;
1.1 schwarze 278:
1.59 schwarze 279: /*
280: * If neither command line arguments -mdoc or -man select
281: * a parser nor the roff parser found a .Dd or .TH macro
282: * yet, look ahead in the main input buffer.
283: */
284:
285: if ((format = roff_getformat(curp->roff)) == 0) {
286: cp = curp->primary->buf;
287: ep = cp + curp->primary->sz;
288: while (cp < ep) {
1.61 schwarze 289: if (*cp == '.' || *cp == '\'') {
1.59 schwarze 290: cp++;
291: if (cp[0] == 'D' && cp[1] == 'd') {
292: format = MPARSE_MDOC;
293: break;
294: }
295: if (cp[0] == 'T' && cp[1] == 'H') {
296: format = MPARSE_MAN;
297: break;
298: }
299: }
300: cp = memchr(cp, '\n', ep - cp);
301: if (cp == NULL)
302: break;
303: cp++;
304: }
1.1 schwarze 305: }
306:
1.59 schwarze 307: if (format == MPARSE_MDOC) {
1.112 schwarze 308: curp->man->macroset = MACROSET_MDOC;
1.136 schwarze 309: if (curp->man->mdocmac == NULL)
310: curp->man->mdocmac = roffhash_alloc(MDOC_Dd, MDOC_MAX);
1.112 schwarze 311: } else {
312: curp->man->macroset = MACROSET_MAN;
1.136 schwarze 313: if (curp->man->manmac == NULL)
314: curp->man->manmac = roffhash_alloc(MAN_TH, MAN_MAX);
1.25 schwarze 315: }
1.136 schwarze 316: curp->man->first->tok = TOKEN_NONE;
1.1 schwarze 317: }
318:
319: /*
1.71 schwarze 320: * Main parse routine for a buffer.
321: * It assumes encoding and line numbering are already set up.
322: * It can recurse directly (for invocations of user-defined
323: * macros, inline equations, and input line traps)
324: * and indirectly (for .so file inclusion).
1.1 schwarze 325: */
1.135 schwarze 326: static int
1.71 schwarze 327: mparse_buf_r(struct mparse *curp, struct buf blk, size_t i, int start)
1.1 schwarze 328: {
329: const struct tbl_span *span;
330: struct buf ln;
1.88 schwarze 331: const char *save_file;
1.85 schwarze 332: char *cp;
1.71 schwarze 333: size_t pos; /* byte number in the ln buffer */
1.1 schwarze 334: enum rofferr rr;
1.76 schwarze 335: int of;
1.1 schwarze 336: int lnn; /* line number in the real file */
1.85 schwarze 337: int fd;
1.1 schwarze 338: unsigned char c;
339:
1.71 schwarze 340: memset(&ln, 0, sizeof(ln));
1.1 schwarze 341:
1.25 schwarze 342: lnn = curp->line;
343: pos = 0;
1.1 schwarze 344:
1.71 schwarze 345: while (i < blk.sz) {
1.1 schwarze 346: if (0 == pos && '\0' == blk.buf[i])
347: break;
348:
349: if (start) {
350: curp->line = lnn;
351: curp->reparse_count = 0;
1.70 schwarze 352:
353: if (lnn < 3 &&
354: curp->filenc & MPARSE_UTF8 &&
1.71 schwarze 355: curp->filenc & MPARSE_LATIN1)
356: curp->filenc = preconv_cue(&blk, i);
1.1 schwarze 357: }
358:
1.71 schwarze 359: while (i < blk.sz && (start || blk.buf[i] != '\0')) {
1.1 schwarze 360:
361: /*
362: * When finding an unescaped newline character,
363: * leave the character loop to process the line.
364: * Skip a preceding carriage return, if any.
365: */
366:
1.71 schwarze 367: if ('\r' == blk.buf[i] && i + 1 < blk.sz &&
1.1 schwarze 368: '\n' == blk.buf[i + 1])
369: ++i;
370: if ('\n' == blk.buf[i]) {
371: ++i;
372: ++lnn;
373: break;
374: }
375:
1.13 schwarze 376: /*
1.70 schwarze 377: * Make sure we have space for the worst
378: * case of 11 bytes: "\\[u10ffff]\0"
1.13 schwarze 379: */
380:
1.71 schwarze 381: if (pos + 11 > ln.sz)
1.13 schwarze 382: resize_buf(&ln, 256);
383:
1.25 schwarze 384: /*
1.70 schwarze 385: * Encode 8-bit input.
1.1 schwarze 386: */
387:
1.70 schwarze 388: c = blk.buf[i];
389: if (c & 0x80) {
1.71 schwarze 390: if ( ! (curp->filenc && preconv_encode(
391: &blk, &i, &ln, &pos, &curp->filenc))) {
1.89 schwarze 392: mandoc_vmsg(MANDOCERR_CHAR_BAD, curp,
393: curp->line, pos, "0x%x", c);
1.70 schwarze 394: ln.buf[pos++] = '?';
395: i++;
396: }
397: continue;
398: }
399:
400: /*
401: * Exclude control characters.
402: */
1.1 schwarze 403:
1.70 schwarze 404: if (c == 0x7f || (c < 0x20 && c != 0x09)) {
1.89 schwarze 405: mandoc_vmsg(c == 0x00 || c == 0x04 ||
406: c > 0x0a ? MANDOCERR_CHAR_BAD :
407: MANDOCERR_CHAR_UNSUPP,
408: curp, curp->line, pos, "0x%x", c);
1.1 schwarze 409: i++;
1.102 schwarze 410: if (c != '\r')
411: ln.buf[pos++] = '?';
1.1 schwarze 412: continue;
413: }
414:
415: ln.buf[pos++] = blk.buf[i++];
416: }
417:
1.142 schwarze 418: if (pos + 1 >= ln.sz)
1.1 schwarze 419: resize_buf(&ln, 256);
420:
1.142 schwarze 421: if (i == blk.sz || blk.buf[i] == '\0')
422: ln.buf[pos++] = '\n';
1.1 schwarze 423: ln.buf[pos] = '\0';
424:
425: /*
426: * A significant amount of complexity is contained by
427: * the roff preprocessor. It's line-oriented but can be
428: * expressed on one line, so we need at times to
429: * readjust our starting point and re-run it. The roff
430: * preprocessor can also readjust the buffers with new
431: * data, so we pass them in wholesale.
432: */
433:
434: of = 0;
435:
1.4 schwarze 436: /*
437: * Maintain a lookaside buffer of all parsed lines. We
438: * only do this if mparse_keep() has been invoked (the
439: * buffer may be accessed with mparse_getkeep()).
440: */
441:
442: if (curp->secondary) {
1.25 schwarze 443: curp->secondary->buf = mandoc_realloc(
444: curp->secondary->buf,
445: curp->secondary->sz + pos + 2);
446: memcpy(curp->secondary->buf +
447: curp->secondary->sz,
448: ln.buf, pos);
1.4 schwarze 449: curp->secondary->sz += pos;
450: curp->secondary->buf
451: [curp->secondary->sz] = '\n';
452: curp->secondary->sz++;
453: curp->secondary->buf
454: [curp->secondary->sz] = '\0';
455: }
1.1 schwarze 456: rerun:
1.72 schwarze 457: rr = roff_parseln(curp->roff, curp->line, &ln, &of);
1.1 schwarze 458:
459: switch (rr) {
1.25 schwarze 460: case ROFF_REPARSE:
1.135 schwarze 461: if (++curp->reparse_count > REPARSE_LIMIT)
1.1 schwarze 462: mandoc_msg(MANDOCERR_ROFFLOOP, curp,
1.25 schwarze 463: curp->line, pos, NULL);
1.135 schwarze 464: else if (mparse_buf_r(curp, ln, of, 0) == 1 ||
465: start == 1) {
466: pos = 0;
467: continue;
468: }
469: free(ln.buf);
470: return 0;
1.25 schwarze 471: case ROFF_APPEND:
1.71 schwarze 472: pos = strlen(ln.buf);
1.1 schwarze 473: continue;
1.25 schwarze 474: case ROFF_RERUN:
1.1 schwarze 475: goto rerun;
1.25 schwarze 476: case ROFF_IGN:
1.1 schwarze 477: pos = 0;
478: continue;
1.25 schwarze 479: case ROFF_SO:
1.71 schwarze 480: if ( ! (curp->options & MPARSE_SO) &&
481: (i >= blk.sz || blk.buf[i] == '\0')) {
1.23 schwarze 482: curp->sodest = mandoc_strdup(ln.buf + of);
483: free(ln.buf);
1.135 schwarze 484: return 1;
1.23 schwarze 485: }
1.4 schwarze 486: /*
487: * We remove `so' clauses from our lookaside
488: * buffer because we're going to descend into
489: * the file recursively.
490: */
1.25 schwarze 491: if (curp->secondary)
1.4 schwarze 492: curp->secondary->sz -= pos + 1;
1.88 schwarze 493: save_file = curp->file;
1.122 schwarze 494: if ((fd = mparse_open(curp, ln.buf + of)) != -1) {
1.85 schwarze 495: mparse_readfd(curp, fd, ln.buf + of);
1.121 schwarze 496: close(fd);
1.88 schwarze 497: curp->file = save_file;
498: } else {
499: curp->file = save_file;
1.30 schwarze 500: mandoc_vmsg(MANDOCERR_SO_FAIL,
501: curp, curp->line, pos,
502: ".so %s", ln.buf + of);
1.85 schwarze 503: ln.sz = mandoc_asprintf(&cp,
504: ".sp\nSee the file %s.\n.sp",
505: ln.buf + of);
506: free(ln.buf);
507: ln.buf = cp;
508: of = 0;
509: mparse_buf_r(curp, ln, of, 0);
1.30 schwarze 510: }
1.1 schwarze 511: pos = 0;
512: continue;
513: default:
514: break;
515: }
516:
1.124 schwarze 517: if (curp->man->macroset == MACROSET_NONE)
1.60 schwarze 518: choose_parser(curp);
1.1 schwarze 519:
1.25 schwarze 520: /*
1.60 schwarze 521: * Lastly, push down into the parsers themselves.
1.1 schwarze 522: * If libroff returns ROFF_TBL, then add it to the
523: * currently open parse. Since we only get here if
524: * there does exist data (see tbl_data.c), we're
525: * guaranteed that something's been allocated.
526: * Do the same for ROFF_EQN.
527: */
528:
1.114 schwarze 529: if (rr == ROFF_TBL)
1.76 schwarze 530: while ((span = roff_span(curp->roff)) != NULL)
1.114 schwarze 531: roff_addtbl(curp->man, span);
532: else if (rr == ROFF_EQN)
533: roff_addeqn(curp->man, roff_eqn(curp->roff));
534: else if ((curp->man->macroset == MACROSET_MDOC ?
1.110 schwarze 535: mdoc_parseln(curp->man, curp->line, ln.buf, of) :
1.76 schwarze 536: man_parseln(curp->man, curp->line, ln.buf, of)) == 2)
537: break;
1.1 schwarze 538:
539: /* Temporary buffers typically are not full. */
540:
541: if (0 == start && '\0' == blk.buf[i])
542: break;
543:
544: /* Start the next input line. */
545:
546: pos = 0;
547: }
548:
549: free(ln.buf);
1.135 schwarze 550: return 1;
1.1 schwarze 551: }
552:
553: static int
1.18 schwarze 554: read_whole_file(struct mparse *curp, const char *file, int fd,
555: struct buf *fb, int *with_mmap)
1.1 schwarze 556: {
557: struct stat st;
1.115 schwarze 558: gzFile gz;
1.1 schwarze 559: size_t off;
560: ssize_t ssz;
561:
1.118 schwarze 562: if (fstat(fd, &st) == -1)
563: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 564:
565: /*
566: * If we're a regular file, try just reading in the whole entry
567: * via mmap(). This is faster than reading it into blocks, and
568: * since each file is only a few bytes to begin with, I'm not
569: * concerned that this is going to tank any machines.
570: */
571:
1.115 schwarze 572: if (curp->gzip == 0 && S_ISREG(st.st_mode)) {
1.106 schwarze 573: if (st.st_size > 0x7fffffff) {
1.86 schwarze 574: mandoc_msg(MANDOCERR_TOOLARGE, curp, 0, 0, NULL);
1.117 schwarze 575: return 0;
1.1 schwarze 576: }
577: *with_mmap = 1;
578: fb->sz = (size_t)st.st_size;
1.15 schwarze 579: fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
1.1 schwarze 580: if (fb->buf != MAP_FAILED)
1.117 schwarze 581: return 1;
1.1 schwarze 582: }
583:
1.115 schwarze 584: if (curp->gzip) {
1.118 schwarze 585: if ((gz = gzdopen(fd, "rb")) == NULL)
586: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.115 schwarze 587: } else
588: gz = NULL;
589:
1.1 schwarze 590: /*
591: * If this isn't a regular file (like, say, stdin), then we must
592: * go the old way and just read things in bit by bit.
593: */
594:
595: *with_mmap = 0;
596: off = 0;
597: fb->sz = 0;
598: fb->buf = NULL;
599: for (;;) {
600: if (off == fb->sz) {
601: if (fb->sz == (1U << 31)) {
1.86 schwarze 602: mandoc_msg(MANDOCERR_TOOLARGE, curp,
603: 0, 0, NULL);
1.1 schwarze 604: break;
605: }
606: resize_buf(fb, 65536);
607: }
1.115 schwarze 608: ssz = curp->gzip ?
609: gzread(gz, fb->buf + (int)off, fb->sz - off) :
610: read(fd, fb->buf + (int)off, fb->sz - off);
1.1 schwarze 611: if (ssz == 0) {
612: fb->sz = off;
1.117 schwarze 613: return 1;
1.1 schwarze 614: }
1.118 schwarze 615: if (ssz == -1)
616: err((int)MANDOCLEVEL_SYSERR, "%s", file);
1.1 schwarze 617: off += (size_t)ssz;
618: }
619:
620: free(fb->buf);
621: fb->buf = NULL;
1.117 schwarze 622: return 0;
1.1 schwarze 623: }
624:
625: static void
626: mparse_end(struct mparse *curp)
627: {
1.110 schwarze 628: if (curp->man->macroset == MACROSET_NONE)
629: curp->man->macroset = MACROSET_MAN;
630: if (curp->man->macroset == MACROSET_MDOC)
631: mdoc_endparse(curp->man);
632: else
1.86 schwarze 633: man_endparse(curp->man);
1.1 schwarze 634: roff_endparse(curp->roff);
635: }
636:
1.15 schwarze 637: static void
638: mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
1.1 schwarze 639: {
1.61 schwarze 640: struct buf *svprimary;
1.1 schwarze 641: const char *svfile;
1.71 schwarze 642: size_t offset;
1.14 schwarze 643: static int recursion_depth;
644:
645: if (64 < recursion_depth) {
646: mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
1.15 schwarze 647: return;
1.14 schwarze 648: }
1.1 schwarze 649:
1.15 schwarze 650: /* Line number is per-file. */
651: svfile = curp->file;
652: curp->file = file;
1.61 schwarze 653: svprimary = curp->primary;
1.59 schwarze 654: curp->primary = &blk;
1.15 schwarze 655: curp->line = 1;
656: recursion_depth++;
657:
1.70 schwarze 658: /* Skip an UTF-8 byte order mark. */
659: if (curp->filenc & MPARSE_UTF8 && blk.sz > 2 &&
660: (unsigned char)blk.buf[0] == 0xef &&
661: (unsigned char)blk.buf[1] == 0xbb &&
662: (unsigned char)blk.buf[2] == 0xbf) {
1.71 schwarze 663: offset = 3;
1.70 schwarze 664: curp->filenc &= ~MPARSE_LATIN1;
1.71 schwarze 665: } else
666: offset = 0;
1.70 schwarze 667:
1.71 schwarze 668: mparse_buf_r(curp, blk, offset, 1);
1.15 schwarze 669:
1.86 schwarze 670: if (--recursion_depth == 0)
1.15 schwarze 671: mparse_end(curp);
672:
1.61 schwarze 673: curp->primary = svprimary;
1.15 schwarze 674: curp->file = svfile;
675: }
676:
1.74 schwarze 677: /*
678: * Read the whole file into memory and call the parsers.
679: * Called recursively when an .so request is encountered.
680: */
1.15 schwarze 681: enum mandoclevel
682: mparse_readfd(struct mparse *curp, int fd, const char *file)
683: {
684: struct buf blk;
685: int with_mmap;
1.70 schwarze 686: int save_filenc;
1.1 schwarze 687:
1.67 schwarze 688: if (read_whole_file(curp, file, fd, &blk, &with_mmap)) {
1.70 schwarze 689: save_filenc = curp->filenc;
690: curp->filenc = curp->options &
691: (MPARSE_UTF8 | MPARSE_LATIN1);
1.67 schwarze 692: mparse_parse_buffer(curp, blk, file);
1.70 schwarze 693: curp->filenc = save_filenc;
1.67 schwarze 694: if (with_mmap)
695: munmap(blk.buf, blk.sz);
696: else
697: free(blk.buf);
698: }
1.117 schwarze 699: return curp->file_status;
1.58 schwarze 700: }
701:
1.122 schwarze 702: int
703: mparse_open(struct mparse *curp, const char *file)
1.58 schwarze 704: {
705: char *cp;
1.122 schwarze 706: int fd;
1.58 schwarze 707:
708: curp->file = file;
1.115 schwarze 709: cp = strrchr(file, '.');
710: curp->gzip = (cp != NULL && ! strcmp(cp + 1, "gz"));
1.74 schwarze 711:
1.115 schwarze 712: /* First try to use the filename as it is. */
1.74 schwarze 713:
1.122 schwarze 714: if ((fd = open(file, O_RDONLY)) != -1)
715: return fd;
1.74 schwarze 716:
1.115 schwarze 717: /*
718: * If that doesn't work and the filename doesn't
719: * already end in .gz, try appending .gz.
720: */
1.74 schwarze 721:
1.115 schwarze 722: if ( ! curp->gzip) {
1.74 schwarze 723: mandoc_asprintf(&cp, "%s.gz", file);
1.123 schwarze 724: fd = open(cp, O_RDONLY);
1.83 schwarze 725: free(cp);
1.122 schwarze 726: if (fd != -1) {
1.115 schwarze 727: curp->gzip = 1;
1.122 schwarze 728: return fd;
1.58 schwarze 729: }
730: }
731:
1.115 schwarze 732: /* Neither worked, give up. */
1.73 schwarze 733:
1.115 schwarze 734: mandoc_msg(MANDOCERR_FILE, curp, 0, 0, strerror(errno));
1.122 schwarze 735: return -1;
1.1 schwarze 736: }
737:
738: struct mparse *
1.69 schwarze 739: mparse_alloc(int options, enum mandoclevel wlevel, mandocmsg mmsg,
1.119 schwarze 740: const char *defos)
1.1 schwarze 741: {
742: struct mparse *curp;
743:
744: curp = mandoc_calloc(1, sizeof(struct mparse));
745:
1.22 schwarze 746: curp->options = options;
1.1 schwarze 747: curp->wlevel = wlevel;
748: curp->mmsg = mmsg;
1.7 schwarze 749: curp->defos = defos;
1.1 schwarze 750:
1.119 schwarze 751: curp->roff = roff_alloc(curp, options);
1.112 schwarze 752: curp->man = roff_man_alloc( curp->roff, curp, curp->defos,
753: curp->options & MPARSE_QUICK ? 1 : 0);
1.111 schwarze 754: if (curp->options & MPARSE_MDOC) {
1.112 schwarze 755: curp->man->macroset = MACROSET_MDOC;
1.136 schwarze 756: if (curp->man->mdocmac == NULL)
757: curp->man->mdocmac = roffhash_alloc(MDOC_Dd, MDOC_MAX);
1.112 schwarze 758: } else if (curp->options & MPARSE_MAN) {
759: curp->man->macroset = MACROSET_MAN;
1.136 schwarze 760: if (curp->man->manmac == NULL)
761: curp->man->manmac = roffhash_alloc(MAN_TH, MAN_MAX);
1.111 schwarze 762: }
1.113 schwarze 763: curp->man->first->tok = TOKEN_NONE;
1.117 schwarze 764: return curp;
1.1 schwarze 765: }
766:
767: void
768: mparse_reset(struct mparse *curp)
769: {
770: roff_reset(curp->roff);
1.124 schwarze 771: roff_man_reset(curp->man);
1.133 schwarze 772:
773: free(curp->sodest);
774: curp->sodest = NULL;
775:
1.4 schwarze 776: if (curp->secondary)
777: curp->secondary->sz = 0;
1.1 schwarze 778:
779: curp->file_status = MANDOCLEVEL_OK;
1.132 schwarze 780: curp->gzip = 0;
1.1 schwarze 781: }
782:
783: void
784: mparse_free(struct mparse *curp)
785: {
786:
1.136 schwarze 787: roffhash_free(curp->man->mdocmac);
788: roffhash_free(curp->man->manmac);
1.112 schwarze 789: roff_man_free(curp->man);
1.133 schwarze 790: roff_free(curp->roff);
1.4 schwarze 791: if (curp->secondary)
792: free(curp->secondary->buf);
1.1 schwarze 793:
1.4 schwarze 794: free(curp->secondary);
1.23 schwarze 795: free(curp->sodest);
1.1 schwarze 796: free(curp);
797: }
798:
799: void
1.110 schwarze 800: mparse_result(struct mparse *curp, struct roff_man **man,
801: char **sodest)
1.1 schwarze 802: {
803:
1.23 schwarze 804: if (sodest && NULL != (*sodest = curp->sodest)) {
805: *man = NULL;
806: return;
807: }
1.1 schwarze 808: if (man)
809: *man = curp->man;
1.130 schwarze 810: }
811:
812: void
813: mparse_updaterc(struct mparse *curp, enum mandoclevel *rc)
814: {
815: if (curp->file_status > *rc)
816: *rc = curp->file_status;
1.1 schwarze 817: }
818:
819: void
820: mandoc_vmsg(enum mandocerr t, struct mparse *m,
821: int ln, int pos, const char *fmt, ...)
822: {
823: char buf[256];
824: va_list ap;
825:
826: va_start(ap, fmt);
1.26 schwarze 827: (void)vsnprintf(buf, sizeof(buf), fmt, ap);
1.1 schwarze 828: va_end(ap);
829:
830: mandoc_msg(t, m, ln, pos, buf);
831: }
832:
833: void
1.25 schwarze 834: mandoc_msg(enum mandocerr er, struct mparse *m,
1.1 schwarze 835: int ln, int col, const char *msg)
836: {
837: enum mandoclevel level;
838:
1.87 schwarze 839: level = MANDOCLEVEL_UNSUPP;
1.1 schwarze 840: while (er < mandoclimits[level])
841: level--;
842:
1.83 schwarze 843: if (level < m->wlevel && er != MANDOCERR_FILE)
1.1 schwarze 844: return;
845:
846: if (m->mmsg)
847: (*m->mmsg)(er, level, m->file, ln, col, msg);
848:
849: if (m->file_status < level)
850: m->file_status = level;
851: }
852:
853: const char *
854: mparse_strerror(enum mandocerr er)
855: {
856:
1.117 schwarze 857: return mandocerrs[er];
1.1 schwarze 858: }
859:
860: const char *
861: mparse_strlevel(enum mandoclevel lvl)
862: {
1.117 schwarze 863: return mandoclevels[lvl];
1.4 schwarze 864: }
865:
866: void
867: mparse_keep(struct mparse *p)
868: {
869:
870: assert(NULL == p->secondary);
871: p->secondary = mandoc_calloc(1, sizeof(struct buf));
872: }
873:
874: const char *
875: mparse_getkeep(const struct mparse *p)
876: {
877:
878: assert(p->secondary);
1.117 schwarze 879: return p->secondary->sz ? p->secondary->buf : NULL;
1.1 schwarze 880: }