xref: /titanic_41/usr/src/cmd/mandoc/read.c (revision 0b7a972444a5f07ee3f5ae816cd71b2cbbea9bac)
1 /*	$Id: read.c,v 1.28 2012/02/16 20:51:31 joerg Exp $ */
2 /*
3  * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4  * Copyright (c) 2010, 2011 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 AUTHOR DISCLAIMS ALL WARRANTIES
11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR 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 #ifdef HAVE_CONFIG_H
19 #include "config.h"
20 #endif
21 
22 #ifdef HAVE_MMAP
23 # include <sys/stat.h>
24 # include <sys/mman.h>
25 #endif
26 
27 #include <assert.h>
28 #include <ctype.h>
29 #include <fcntl.h>
30 #include <stdarg.h>
31 #include <stdint.h>
32 #include <stdio.h>
33 #include <stdlib.h>
34 #include <string.h>
35 #include <unistd.h>
36 
37 #include "mandoc.h"
38 #include "libmandoc.h"
39 #include "mdoc.h"
40 #include "man.h"
41 #include "main.h"
42 
43 #ifndef MAP_FILE
44 #define	MAP_FILE	0
45 #endif
46 
47 #define	REPARSE_LIMIT	1000
48 
49 struct	buf {
50 	char	 	 *buf; /* binary input buffer */
51 	size_t		  sz; /* size of binary buffer */
52 };
53 
54 struct	mparse {
55 	enum mandoclevel  file_status; /* status of current parse */
56 	enum mandoclevel  wlevel; /* ignore messages below this */
57 	int		  line; /* line number in the file */
58 	enum mparset	  inttype; /* which parser to use */
59 	struct man	 *pman; /* persistent man parser */
60 	struct mdoc	 *pmdoc; /* persistent mdoc parser */
61 	struct man	 *man; /* man parser */
62 	struct mdoc	 *mdoc; /* mdoc parser */
63 	struct roff	 *roff; /* roff parser (!NULL) */
64 	int		  reparse_count; /* finite interp. stack */
65 	mandocmsg	  mmsg; /* warning/error message handler */
66 	void		 *arg; /* argument to mmsg */
67 	const char	 *file;
68 	struct buf	 *secondary;
69 };
70 
71 static	void	  resize_buf(struct buf *, size_t);
72 static	void	  mparse_buf_r(struct mparse *, struct buf, int);
73 static	void	  mparse_readfd_r(struct mparse *, int, const char *, int);
74 static	void	  pset(const char *, int, struct mparse *);
75 static	int	  read_whole_file(const char *, int, struct buf *, int *);
76 static	void	  mparse_end(struct mparse *);
77 
78 static	const enum mandocerr	mandoclimits[MANDOCLEVEL_MAX] = {
79 	MANDOCERR_OK,
80 	MANDOCERR_WARNING,
81 	MANDOCERR_WARNING,
82 	MANDOCERR_ERROR,
83 	MANDOCERR_FATAL,
84 	MANDOCERR_MAX,
85 	MANDOCERR_MAX
86 };
87 
88 static	const char * const	mandocerrs[MANDOCERR_MAX] = {
89 	"ok",
90 
91 	"generic warning",
92 
93 	/* related to the prologue */
94 	"no title in document",
95 	"document title should be all caps",
96 	"unknown manual section",
97 	"date missing, using today's date",
98 	"cannot parse date, using it verbatim",
99 	"prologue macros out of order",
100 	"duplicate prologue macro",
101 	"macro not allowed in prologue",
102 	"macro not allowed in body",
103 
104 	/* related to document structure */
105 	".so is fragile, better use ln(1)",
106 	"NAME section must come first",
107 	"bad NAME section contents",
108 	"manual name not yet set",
109 	"sections out of conventional order",
110 	"duplicate section name",
111 	"section not in conventional manual section",
112 
113 	/* related to macros and nesting */
114 	"skipping obsolete macro",
115 	"skipping paragraph macro",
116 	"skipping no-space macro",
117 	"blocks badly nested",
118 	"child violates parent syntax",
119 	"nested displays are not portable",
120 	"already in literal mode",
121 	"line scope broken",
122 
123 	/* related to missing macro arguments */
124 	"skipping empty macro",
125 	"argument count wrong",
126 	"missing display type",
127 	"list type must come first",
128 	"tag lists require a width argument",
129 	"missing font type",
130 	"skipping end of block that is not open",
131 
132 	/* related to bad macro arguments */
133 	"skipping argument",
134 	"duplicate argument",
135 	"duplicate display type",
136 	"duplicate list type",
137 	"unknown AT&T UNIX version",
138 	"bad Boolean value",
139 	"unknown font",
140 	"unknown standard specifier",
141 	"bad width argument",
142 
143 	/* related to plain text */
144 	"blank line in non-literal context",
145 	"tab in non-literal context",
146 	"end of line whitespace",
147 	"bad comment style",
148 	"bad escape sequence",
149 	"unterminated quoted string",
150 
151 	/* related to equations */
152 	"unexpected literal in equation",
153 
154 	"generic error",
155 
156 	/* related to equations */
157 	"unexpected equation scope closure",
158 	"equation scope open on exit",
159 	"overlapping equation scopes",
160 	"unexpected end of equation",
161 	"equation syntax error",
162 
163 	/* related to tables */
164 	"bad table syntax",
165 	"bad table option",
166 	"bad table layout",
167 	"no table layout cells specified",
168 	"no table data cells specified",
169 	"ignore data in cell",
170 	"data block still open",
171 	"ignoring extra data cells",
172 
173 	"input stack limit exceeded, infinite loop?",
174 	"skipping bad character",
175 	"escaped character not allowed in a name",
176 	"skipping text before the first section header",
177 	"skipping unknown macro",
178 	"NOT IMPLEMENTED, please use groff: skipping request",
179 	"argument count wrong",
180 	"skipping end of block that is not open",
181 	"missing end of block",
182 	"scope open on exit",
183 	"uname(3) system call failed",
184 	"macro requires line argument(s)",
185 	"macro requires body argument(s)",
186 	"macro requires argument(s)",
187 	"missing list type",
188 	"line argument(s) will be lost",
189 	"body argument(s) will be lost",
190 
191 	"generic fatal error",
192 
193 	"not a manual",
194 	"column syntax is inconsistent",
195 	"NOT IMPLEMENTED: .Bd -file",
196 	"argument count wrong, violates syntax",
197 	"child violates parent syntax",
198 	"argument count wrong, violates syntax",
199 	"NOT IMPLEMENTED: .so with absolute path or \"..\"",
200 	"no document body",
201 	"no document prologue",
202 	"static buffer exhausted",
203 };
204 
205 static	const char * const	mandoclevels[MANDOCLEVEL_MAX] = {
206 	"SUCCESS",
207 	"RESERVED",
208 	"WARNING",
209 	"ERROR",
210 	"FATAL",
211 	"BADARG",
212 	"SYSERR"
213 };
214 
215 static void
216 resize_buf(struct buf *buf, size_t initial)
217 {
218 
219 	buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
220 	buf->buf = mandoc_realloc(buf->buf, buf->sz);
221 }
222 
223 static void
224 pset(const char *buf, int pos, struct mparse *curp)
225 {
226 	int		 i;
227 
228 	/*
229 	 * Try to intuit which kind of manual parser should be used.  If
230 	 * passed in by command-line (-man, -mdoc), then use that
231 	 * explicitly.  If passed as -mandoc, then try to guess from the
232 	 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
233 	 * default to -man, which is more lenient.
234 	 *
235 	 * Separate out pmdoc/pman from mdoc/man: the first persists
236 	 * through all parsers, while the latter is used per-parse.
237 	 */
238 
239 	if ('.' == buf[0] || '\'' == buf[0]) {
240 		for (i = 1; buf[i]; i++)
241 			if (' ' != buf[i] && '\t' != buf[i])
242 				break;
243 		if ('\0' == buf[i])
244 			return;
245 	}
246 
247 	switch (curp->inttype) {
248 	case (MPARSE_MDOC):
249 		if (NULL == curp->pmdoc)
250 			curp->pmdoc = mdoc_alloc(curp->roff, curp);
251 		assert(curp->pmdoc);
252 		curp->mdoc = curp->pmdoc;
253 		return;
254 	case (MPARSE_MAN):
255 		if (NULL == curp->pman)
256 			curp->pman = man_alloc(curp->roff, curp);
257 		assert(curp->pman);
258 		curp->man = curp->pman;
259 		return;
260 	default:
261 		break;
262 	}
263 
264 	if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3))  {
265 		if (NULL == curp->pmdoc)
266 			curp->pmdoc = mdoc_alloc(curp->roff, curp);
267 		assert(curp->pmdoc);
268 		curp->mdoc = curp->pmdoc;
269 		return;
270 	}
271 
272 	if (NULL == curp->pman)
273 		curp->pman = man_alloc(curp->roff, curp);
274 	assert(curp->pman);
275 	curp->man = curp->pman;
276 }
277 
278 /*
279  * Main parse routine for an opened file.  This is called for each
280  * opened file and simply loops around the full input file, possibly
281  * nesting (i.e., with `so').
282  */
283 static void
284 mparse_buf_r(struct mparse *curp, struct buf blk, int start)
285 {
286 	const struct tbl_span	*span;
287 	struct buf	 ln;
288 	enum rofferr	 rr;
289 	int		 i, of, rc;
290 	int		 pos; /* byte number in the ln buffer */
291 	int		 lnn; /* line number in the real file */
292 	unsigned char	 c;
293 
294 	memset(&ln, 0, sizeof(struct buf));
295 
296 	lnn = curp->line;
297 	pos = 0;
298 
299 	for (i = 0; i < (int)blk.sz; ) {
300 		if (0 == pos && '\0' == blk.buf[i])
301 			break;
302 
303 		if (start) {
304 			curp->line = lnn;
305 			curp->reparse_count = 0;
306 		}
307 
308 		while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
309 
310 			/*
311 			 * When finding an unescaped newline character,
312 			 * leave the character loop to process the line.
313 			 * Skip a preceding carriage return, if any.
314 			 */
315 
316 			if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
317 			    '\n' == blk.buf[i + 1])
318 				++i;
319 			if ('\n' == blk.buf[i]) {
320 				++i;
321 				++lnn;
322 				break;
323 			}
324 
325 			/*
326 			 * Warn about bogus characters.  If you're using
327 			 * non-ASCII encoding, you're screwing your
328 			 * readers.  Since I'd rather this not happen,
329 			 * I'll be helpful and replace these characters
330 			 * with "?", so we don't display gibberish.
331 			 * Note to manual writers: use special characters.
332 			 */
333 
334 			c = (unsigned char) blk.buf[i];
335 
336 			if ( ! (isascii(c) &&
337 					(isgraph(c) || isblank(c)))) {
338 				mandoc_msg(MANDOCERR_BADCHAR, curp,
339 						curp->line, pos, NULL);
340 				i++;
341 				if (pos >= (int)ln.sz)
342 					resize_buf(&ln, 256);
343 				ln.buf[pos++] = '?';
344 				continue;
345 			}
346 
347 			/* Trailing backslash = a plain char. */
348 
349 			if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
350 				if (pos >= (int)ln.sz)
351 					resize_buf(&ln, 256);
352 				ln.buf[pos++] = blk.buf[i++];
353 				continue;
354 			}
355 
356 			/*
357 			 * Found escape and at least one other character.
358 			 * When it's a newline character, skip it.
359 			 * When there is a carriage return in between,
360 			 * skip that one as well.
361 			 */
362 
363 			if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
364 			    '\n' == blk.buf[i + 2])
365 				++i;
366 			if ('\n' == blk.buf[i + 1]) {
367 				i += 2;
368 				++lnn;
369 				continue;
370 			}
371 
372 			if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
373 				i += 2;
374 				/* Comment, skip to end of line */
375 				for (; i < (int)blk.sz; ++i) {
376 					if ('\n' == blk.buf[i]) {
377 						++i;
378 						++lnn;
379 						break;
380 					}
381 				}
382 
383 				/* Backout trailing whitespaces */
384 				for (; pos > 0; --pos) {
385 					if (ln.buf[pos - 1] != ' ')
386 						break;
387 					if (pos > 2 && ln.buf[pos - 2] == '\\')
388 						break;
389 				}
390 				break;
391 			}
392 
393 			/* Some other escape sequence, copy & cont. */
394 
395 			if (pos + 1 >= (int)ln.sz)
396 				resize_buf(&ln, 256);
397 
398 			ln.buf[pos++] = blk.buf[i++];
399 			ln.buf[pos++] = blk.buf[i++];
400 		}
401 
402  		if (pos >= (int)ln.sz)
403 			resize_buf(&ln, 256);
404 
405 		ln.buf[pos] = '\0';
406 
407 		/*
408 		 * A significant amount of complexity is contained by
409 		 * the roff preprocessor.  It's line-oriented but can be
410 		 * expressed on one line, so we need at times to
411 		 * readjust our starting point and re-run it.  The roff
412 		 * preprocessor can also readjust the buffers with new
413 		 * data, so we pass them in wholesale.
414 		 */
415 
416 		of = 0;
417 
418 		/*
419 		 * Maintain a lookaside buffer of all parsed lines.  We
420 		 * only do this if mparse_keep() has been invoked (the
421 		 * buffer may be accessed with mparse_getkeep()).
422 		 */
423 
424 		if (curp->secondary) {
425 			curp->secondary->buf =
426 				mandoc_realloc
427 				(curp->secondary->buf,
428 				 curp->secondary->sz + pos + 2);
429 			memcpy(curp->secondary->buf +
430 					curp->secondary->sz,
431 					ln.buf, pos);
432 			curp->secondary->sz += pos;
433 			curp->secondary->buf
434 				[curp->secondary->sz] = '\n';
435 			curp->secondary->sz++;
436 			curp->secondary->buf
437 				[curp->secondary->sz] = '\0';
438 		}
439 rerun:
440 		rr = roff_parseln
441 			(curp->roff, curp->line,
442 			 &ln.buf, &ln.sz, of, &of);
443 
444 		switch (rr) {
445 		case (ROFF_REPARSE):
446 			if (REPARSE_LIMIT >= ++curp->reparse_count)
447 				mparse_buf_r(curp, ln, 0);
448 			else
449 				mandoc_msg(MANDOCERR_ROFFLOOP, curp,
450 					curp->line, pos, NULL);
451 			pos = 0;
452 			continue;
453 		case (ROFF_APPEND):
454 			pos = (int)strlen(ln.buf);
455 			continue;
456 		case (ROFF_RERUN):
457 			goto rerun;
458 		case (ROFF_IGN):
459 			pos = 0;
460 			continue;
461 		case (ROFF_ERR):
462 			assert(MANDOCLEVEL_FATAL <= curp->file_status);
463 			break;
464 		case (ROFF_SO):
465 			/*
466 			 * We remove `so' clauses from our lookaside
467 			 * buffer because we're going to descend into
468 			 * the file recursively.
469 			 */
470 			if (curp->secondary)
471 				curp->secondary->sz -= pos + 1;
472 			mparse_readfd_r(curp, -1, ln.buf + of, 1);
473 			if (MANDOCLEVEL_FATAL <= curp->file_status)
474 				break;
475 			pos = 0;
476 			continue;
477 		default:
478 			break;
479 		}
480 
481 		/*
482 		 * If we encounter errors in the recursive parse, make
483 		 * sure we don't continue parsing.
484 		 */
485 
486 		if (MANDOCLEVEL_FATAL <= curp->file_status)
487 			break;
488 
489 		/*
490 		 * If input parsers have not been allocated, do so now.
491 		 * We keep these instanced between parsers, but set them
492 		 * locally per parse routine since we can use different
493 		 * parsers with each one.
494 		 */
495 
496 		if ( ! (curp->man || curp->mdoc))
497 			pset(ln.buf + of, pos - of, curp);
498 
499 		/*
500 		 * Lastly, push down into the parsers themselves.  One
501 		 * of these will have already been set in the pset()
502 		 * routine.
503 		 * If libroff returns ROFF_TBL, then add it to the
504 		 * currently open parse.  Since we only get here if
505 		 * there does exist data (see tbl_data.c), we're
506 		 * guaranteed that something's been allocated.
507 		 * Do the same for ROFF_EQN.
508 		 */
509 
510 		rc = -1;
511 
512 		if (ROFF_TBL == rr)
513 			while (NULL != (span = roff_span(curp->roff))) {
514 				rc = curp->man ?
515 					man_addspan(curp->man, span) :
516 					mdoc_addspan(curp->mdoc, span);
517 				if (0 == rc)
518 					break;
519 			}
520 		else if (ROFF_EQN == rr)
521 			rc = curp->mdoc ?
522 				mdoc_addeqn(curp->mdoc,
523 					roff_eqn(curp->roff)) :
524 				man_addeqn(curp->man,
525 					roff_eqn(curp->roff));
526 		else if (curp->man || curp->mdoc)
527 			rc = curp->man ?
528 				man_parseln(curp->man,
529 					curp->line, ln.buf, of) :
530 				mdoc_parseln(curp->mdoc,
531 					curp->line, ln.buf, of);
532 
533 		if (0 == rc) {
534 			assert(MANDOCLEVEL_FATAL <= curp->file_status);
535 			break;
536 		}
537 
538 		/* Temporary buffers typically are not full. */
539 
540 		if (0 == start && '\0' == blk.buf[i])
541 			break;
542 
543 		/* Start the next input line. */
544 
545 		pos = 0;
546 	}
547 
548 	free(ln.buf);
549 }
550 
551 static int
552 read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap)
553 {
554 	size_t		 off;
555 	ssize_t		 ssz;
556 
557 #ifdef	HAVE_MMAP
558 	struct stat	 st;
559 	if (-1 == fstat(fd, &st)) {
560 		perror(file);
561 		return(0);
562 	}
563 
564 	/*
565 	 * If we're a regular file, try just reading in the whole entry
566 	 * via mmap().  This is faster than reading it into blocks, and
567 	 * since each file is only a few bytes to begin with, I'm not
568 	 * concerned that this is going to tank any machines.
569 	 */
570 
571 	if (S_ISREG(st.st_mode)) {
572 		if (st.st_size >= (1U << 31)) {
573 			fprintf(stderr, "%s: input too large\n", file);
574 			return(0);
575 		}
576 		*with_mmap = 1;
577 		fb->sz = (size_t)st.st_size;
578 		fb->buf = mmap(NULL, fb->sz, PROT_READ,
579 				MAP_FILE|MAP_SHARED, fd, 0);
580 		if (fb->buf != MAP_FAILED)
581 			return(1);
582 	}
583 #endif
584 
585 	/*
586 	 * If this isn't a regular file (like, say, stdin), then we must
587 	 * go the old way and just read things in bit by bit.
588 	 */
589 
590 	*with_mmap = 0;
591 	off = 0;
592 	fb->sz = 0;
593 	fb->buf = NULL;
594 	for (;;) {
595 		if (off == fb->sz) {
596 			if (fb->sz == (1U << 31)) {
597 				fprintf(stderr, "%s: input too large\n", file);
598 				break;
599 			}
600 			resize_buf(fb, 65536);
601 		}
602 		ssz = read(fd, fb->buf + (int)off, fb->sz - off);
603 		if (ssz == 0) {
604 			fb->sz = off;
605 			return(1);
606 		}
607 		if (ssz == -1) {
608 			perror(file);
609 			break;
610 		}
611 		off += (size_t)ssz;
612 	}
613 
614 	free(fb->buf);
615 	fb->buf = NULL;
616 	return(0);
617 }
618 
619 static void
620 mparse_end(struct mparse *curp)
621 {
622 
623 	if (MANDOCLEVEL_FATAL <= curp->file_status)
624 		return;
625 
626 	if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
627 		assert(MANDOCLEVEL_FATAL <= curp->file_status);
628 		return;
629 	}
630 
631 	if (curp->man && ! man_endparse(curp->man)) {
632 		assert(MANDOCLEVEL_FATAL <= curp->file_status);
633 		return;
634 	}
635 
636 	if ( ! (curp->man || curp->mdoc)) {
637 		mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
638 		curp->file_status = MANDOCLEVEL_FATAL;
639 		return;
640 	}
641 
642 	roff_endparse(curp->roff);
643 }
644 
645 static void
646 mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file,
647 		int re)
648 {
649 	const char	*svfile;
650 
651 	/* Line number is per-file. */
652 	svfile = curp->file;
653 	curp->file = file;
654 	curp->line = 1;
655 
656 	mparse_buf_r(curp, blk, 1);
657 
658 	if (0 == re && MANDOCLEVEL_FATAL > curp->file_status)
659 		mparse_end(curp);
660 
661 	curp->file = svfile;
662 }
663 
664 enum mandoclevel
665 mparse_readmem(struct mparse *curp, const void *buf, size_t len,
666 		const char *file)
667 {
668 	struct buf blk;
669 
670 	blk.buf = UNCONST(buf);
671 	blk.sz = len;
672 
673 	mparse_parse_buffer(curp, blk, file, 0);
674 	return(curp->file_status);
675 }
676 
677 static void
678 mparse_readfd_r(struct mparse *curp, int fd, const char *file, int re)
679 {
680 	struct buf	 blk;
681 	int		 with_mmap;
682 
683 	if (-1 == fd)
684 		if (-1 == (fd = open(file, O_RDONLY, 0))) {
685 			perror(file);
686 			curp->file_status = MANDOCLEVEL_SYSERR;
687 			return;
688 		}
689 	/*
690 	 * Run for each opened file; may be called more than once for
691 	 * each full parse sequence if the opened file is nested (i.e.,
692 	 * from `so').  Simply sucks in the whole file and moves into
693 	 * the parse phase for the file.
694 	 */
695 
696 	if ( ! read_whole_file(file, fd, &blk, &with_mmap)) {
697 		curp->file_status = MANDOCLEVEL_SYSERR;
698 		return;
699 	}
700 
701 	mparse_parse_buffer(curp, blk, file, re);
702 
703 #ifdef	HAVE_MMAP
704 	if (with_mmap)
705 		munmap(blk.buf, blk.sz);
706 	else
707 #endif
708 		free(blk.buf);
709 
710 	if (STDIN_FILENO != fd && -1 == close(fd))
711 		perror(file);
712 }
713 
714 enum mandoclevel
715 mparse_readfd(struct mparse *curp, int fd, const char *file)
716 {
717 
718 	mparse_readfd_r(curp, fd, file, 0);
719 	return(curp->file_status);
720 }
721 
722 struct mparse *
723 mparse_alloc(enum mparset inttype, enum mandoclevel wlevel, mandocmsg mmsg, void *arg)
724 {
725 	struct mparse	*curp;
726 
727 	assert(wlevel <= MANDOCLEVEL_FATAL);
728 
729 	curp = mandoc_calloc(1, sizeof(struct mparse));
730 
731 	curp->wlevel = wlevel;
732 	curp->mmsg = mmsg;
733 	curp->arg = arg;
734 	curp->inttype = inttype;
735 
736 	curp->roff = roff_alloc(curp);
737 	return(curp);
738 }
739 
740 void
741 mparse_reset(struct mparse *curp)
742 {
743 
744 	roff_reset(curp->roff);
745 
746 	if (curp->mdoc)
747 		mdoc_reset(curp->mdoc);
748 	if (curp->man)
749 		man_reset(curp->man);
750 	if (curp->secondary)
751 		curp->secondary->sz = 0;
752 
753 	curp->file_status = MANDOCLEVEL_OK;
754 	curp->mdoc = NULL;
755 	curp->man = NULL;
756 }
757 
758 void
759 mparse_free(struct mparse *curp)
760 {
761 
762 	if (curp->pmdoc)
763 		mdoc_free(curp->pmdoc);
764 	if (curp->pman)
765 		man_free(curp->pman);
766 	if (curp->roff)
767 		roff_free(curp->roff);
768 	if (curp->secondary)
769 		free(curp->secondary->buf);
770 
771 	free(curp->secondary);
772 	free(curp);
773 }
774 
775 void
776 mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
777 {
778 
779 	if (mdoc)
780 		*mdoc = curp->mdoc;
781 	if (man)
782 		*man = curp->man;
783 }
784 
785 void
786 mandoc_vmsg(enum mandocerr t, struct mparse *m,
787 		int ln, int pos, const char *fmt, ...)
788 {
789 	char		 buf[256];
790 	va_list		 ap;
791 
792 	va_start(ap, fmt);
793 	vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
794 	va_end(ap);
795 
796 	mandoc_msg(t, m, ln, pos, buf);
797 }
798 
799 void
800 mandoc_msg(enum mandocerr er, struct mparse *m,
801 		int ln, int col, const char *msg)
802 {
803 	enum mandoclevel level;
804 
805 	level = MANDOCLEVEL_FATAL;
806 	while (er < mandoclimits[level])
807 		level--;
808 
809 	if (level < m->wlevel)
810 		return;
811 
812 	if (m->mmsg)
813 		(*m->mmsg)(er, level, m->file, ln, col, msg);
814 
815 	if (m->file_status < level)
816 		m->file_status = level;
817 }
818 
819 const char *
820 mparse_strerror(enum mandocerr er)
821 {
822 
823 	return(mandocerrs[er]);
824 }
825 
826 const char *
827 mparse_strlevel(enum mandoclevel lvl)
828 {
829 	return(mandoclevels[lvl]);
830 }
831 
832 void
833 mparse_keep(struct mparse *p)
834 {
835 
836 	assert(NULL == p->secondary);
837 	p->secondary = mandoc_calloc(1, sizeof(struct buf));
838 }
839 
840 const char *
841 mparse_getkeep(const struct mparse *p)
842 {
843 
844 	assert(p->secondary);
845 	return(p->secondary->sz ? p->secondary->buf : NULL);
846 }
847