xref: /netbsd-src/external/bsd/mdocml/dist/read.c (revision a5847cc334d9a7029f6352b847e9e8d71a0f9e0c)
1 /*	$Vendor-Id: read.c,v 1.25 2011/10/08 15:42:29 kristaps 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 <stdio.h>
32 #include <stdlib.h>
33 #include <string.h>
34 #include <unistd.h>
35 
36 #include "mandoc.h"
37 #include "libmandoc.h"
38 #include "mdoc.h"
39 #include "man.h"
40 
41 #ifndef MAP_FILE
42 #define	MAP_FILE	0
43 #endif
44 
45 #define	REPARSE_LIMIT	1000
46 
47 struct	buf {
48 	char	 	 *buf; /* binary input buffer */
49 	size_t		  sz; /* size of binary buffer */
50 };
51 
52 struct	mparse {
53 	enum mandoclevel  file_status; /* status of current parse */
54 	enum mandoclevel  wlevel; /* ignore messages below this */
55 	int		  line; /* line number in the file */
56 	enum mparset	  inttype; /* which parser to use */
57 	struct man	 *pman; /* persistent man parser */
58 	struct mdoc	 *pmdoc; /* persistent mdoc parser */
59 	struct man	 *man; /* man parser */
60 	struct mdoc	 *mdoc; /* mdoc parser */
61 	struct roff	 *roff; /* roff parser (!NULL) */
62 	int		  reparse_count; /* finite interp. stack */
63 	mandocmsg	  mmsg; /* warning/error message handler */
64 	void		 *arg; /* argument to mmsg */
65 	const char	 *file;
66 	struct buf	 *secondary;
67 };
68 
69 static	void	  resize_buf(struct buf *, size_t);
70 static	void	  mparse_buf_r(struct mparse *, struct buf, int);
71 static	void	  mparse_readfd_r(struct mparse *, int, const char *, int);
72 static	void	  pset(const char *, int, struct mparse *);
73 static	void	  pdesc(struct mparse *, const char *, int);
74 static	int	  read_whole_file(const char *, int, struct buf *, int *);
75 static	void	  mparse_end(struct mparse *);
76 
77 static	const enum mandocerr	mandoclimits[MANDOCLEVEL_MAX] = {
78 	MANDOCERR_OK,
79 	MANDOCERR_WARNING,
80 	MANDOCERR_WARNING,
81 	MANDOCERR_ERROR,
82 	MANDOCERR_FATAL,
83 	MANDOCERR_MAX,
84 	MANDOCERR_MAX
85 };
86 
87 static	const char * const	mandocerrs[MANDOCERR_MAX] = {
88 	"ok",
89 
90 	"generic warning",
91 
92 	/* related to the prologue */
93 	"no title in document",
94 	"document title should be all caps",
95 	"unknown manual section",
96 	"date missing, using today's date",
97 	"cannot parse date, using it verbatim",
98 	"prologue macros out of order",
99 	"duplicate prologue macro",
100 	"macro not allowed in prologue",
101 	"macro not allowed in body",
102 
103 	/* related to document structure */
104 	".so is fragile, better use ln(1)",
105 	"NAME section must come first",
106 	"bad NAME section contents",
107 	"manual name not yet set",
108 	"sections out of conventional order",
109 	"duplicate section name",
110 	"section not in conventional manual section",
111 
112 	/* related to macros and nesting */
113 	"skipping obsolete macro",
114 	"skipping paragraph macro",
115 	"skipping no-space macro",
116 	"blocks badly nested",
117 	"child violates parent syntax",
118 	"nested displays are not portable",
119 	"already in literal mode",
120 	"line scope broken",
121 
122 	/* related to missing macro arguments */
123 	"skipping empty macro",
124 	"argument count wrong",
125 	"missing display type",
126 	"list type must come first",
127 	"tag lists require a width argument",
128 	"missing font type",
129 	"skipping end of block that is not open",
130 
131 	/* related to bad macro arguments */
132 	"skipping argument",
133 	"duplicate argument",
134 	"duplicate display type",
135 	"duplicate list type",
136 	"unknown AT&T UNIX version",
137 	"bad Boolean value",
138 	"unknown font",
139 	"unknown standard specifier",
140 	"bad width argument",
141 
142 	/* related to plain text */
143 	"blank line in non-literal context",
144 	"tab in non-literal context",
145 	"end of line whitespace",
146 	"bad comment style",
147 	"bad escape sequence",
148 	"unterminated quoted string",
149 
150 	/* related to equations */
151 	"unexpected literal in equation",
152 
153 	"generic error",
154 
155 	/* related to equations */
156 	"unexpected equation scope closure",
157 	"equation scope open on exit",
158 	"overlapping equation scopes",
159 	"unexpected end of equation",
160 	"equation syntax error",
161 
162 	/* related to tables */
163 	"bad table syntax",
164 	"bad table option",
165 	"bad table layout",
166 	"no table layout cells specified",
167 	"no table data cells specified",
168 	"ignore data in cell",
169 	"data block still open",
170 	"ignoring extra data cells",
171 
172 	"input stack limit exceeded, infinite loop?",
173 	"skipping bad character",
174 	"escaped character not allowed in a name",
175 	"skipping text before the first section header",
176 	"skipping unknown macro",
177 	"NOT IMPLEMENTED, please use groff: skipping request",
178 	"argument count wrong",
179 	"skipping end of block that is not open",
180 	"missing end of block",
181 	"scope open on exit",
182 	"uname(3) system call failed",
183 	"macro requires line argument(s)",
184 	"macro requires body argument(s)",
185 	"macro requires argument(s)",
186 	"missing list type",
187 	"line argument(s) will be lost",
188 	"body argument(s) will be lost",
189 
190 	"generic fatal error",
191 
192 	"not a manual",
193 	"column syntax is inconsistent",
194 	"NOT IMPLEMENTED: .Bd -file",
195 	"line scope broken, syntax violated",
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 drop these characters so
330 			 * we don't display gibberish.  Note to manual
331 			 * 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, "ignoring byte");
340 				i++;
341 				continue;
342 			}
343 
344 			/* Trailing backslash = a plain char. */
345 
346 			if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
347 				if (pos >= (int)ln.sz)
348 					resize_buf(&ln, 256);
349 				ln.buf[pos++] = blk.buf[i++];
350 				continue;
351 			}
352 
353 			/*
354 			 * Found escape and at least one other character.
355 			 * When it's a newline character, skip it.
356 			 * When there is a carriage return in between,
357 			 * skip that one as well.
358 			 */
359 
360 			if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
361 			    '\n' == blk.buf[i + 2])
362 				++i;
363 			if ('\n' == blk.buf[i + 1]) {
364 				i += 2;
365 				++lnn;
366 				continue;
367 			}
368 
369 			if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
370 				i += 2;
371 				/* Comment, skip to end of line */
372 				for (; i < (int)blk.sz; ++i) {
373 					if ('\n' == blk.buf[i]) {
374 						++i;
375 						++lnn;
376 						break;
377 					}
378 				}
379 
380 				/* Backout trailing whitespaces */
381 				for (; pos > 0; --pos) {
382 					if (ln.buf[pos - 1] != ' ')
383 						break;
384 					if (pos > 2 && ln.buf[pos - 2] == '\\')
385 						break;
386 				}
387 				break;
388 			}
389 
390 			/* Some other escape sequence, copy & cont. */
391 
392 			if (pos + 1 >= (int)ln.sz)
393 				resize_buf(&ln, 256);
394 
395 			ln.buf[pos++] = blk.buf[i++];
396 			ln.buf[pos++] = blk.buf[i++];
397 		}
398 
399  		if (pos >= (int)ln.sz)
400 			resize_buf(&ln, 256);
401 
402 		ln.buf[pos] = '\0';
403 
404 		/*
405 		 * A significant amount of complexity is contained by
406 		 * the roff preprocessor.  It's line-oriented but can be
407 		 * expressed on one line, so we need at times to
408 		 * readjust our starting point and re-run it.  The roff
409 		 * preprocessor can also readjust the buffers with new
410 		 * data, so we pass them in wholesale.
411 		 */
412 
413 		of = 0;
414 
415 		/*
416 		 * Maintain a lookaside buffer of all parsed lines.  We
417 		 * only do this if mparse_keep() has been invoked (the
418 		 * buffer may be accessed with mparse_getkeep()).
419 		 */
420 
421 		if (curp->secondary) {
422 			curp->secondary->buf =
423 				mandoc_realloc
424 				(curp->secondary->buf,
425 				 curp->secondary->sz + pos + 2);
426 			memcpy(curp->secondary->buf +
427 					curp->secondary->sz,
428 					ln.buf, pos);
429 			curp->secondary->sz += pos;
430 			curp->secondary->buf
431 				[curp->secondary->sz] = '\n';
432 			curp->secondary->sz++;
433 			curp->secondary->buf
434 				[curp->secondary->sz] = '\0';
435 		}
436 rerun:
437 		rr = roff_parseln
438 			(curp->roff, curp->line,
439 			 &ln.buf, &ln.sz, of, &of);
440 
441 		switch (rr) {
442 		case (ROFF_REPARSE):
443 			if (REPARSE_LIMIT >= ++curp->reparse_count)
444 				mparse_buf_r(curp, ln, 0);
445 			else
446 				mandoc_msg(MANDOCERR_ROFFLOOP, curp,
447 					curp->line, pos, NULL);
448 			pos = 0;
449 			continue;
450 		case (ROFF_APPEND):
451 			pos = (int)strlen(ln.buf);
452 			continue;
453 		case (ROFF_RERUN):
454 			goto rerun;
455 		case (ROFF_IGN):
456 			pos = 0;
457 			continue;
458 		case (ROFF_ERR):
459 			assert(MANDOCLEVEL_FATAL <= curp->file_status);
460 			break;
461 		case (ROFF_SO):
462 			/*
463 			 * We remove `so' clauses from our lookaside
464 			 * buffer because we're going to descend into
465 			 * the file recursively.
466 			 */
467 			if (curp->secondary)
468 				curp->secondary->sz -= pos + 1;
469 			mparse_readfd_r(curp, -1, ln.buf + of, 1);
470 			if (MANDOCLEVEL_FATAL <= curp->file_status)
471 				break;
472 			pos = 0;
473 			continue;
474 		default:
475 			break;
476 		}
477 
478 		/*
479 		 * If we encounter errors in the recursive parse, make
480 		 * sure we don't continue parsing.
481 		 */
482 
483 		if (MANDOCLEVEL_FATAL <= curp->file_status)
484 			break;
485 
486 		/*
487 		 * If input parsers have not been allocated, do so now.
488 		 * We keep these instanced between parsers, but set them
489 		 * locally per parse routine since we can use different
490 		 * parsers with each one.
491 		 */
492 
493 		if ( ! (curp->man || curp->mdoc))
494 			pset(ln.buf + of, pos - of, curp);
495 
496 		/*
497 		 * Lastly, push down into the parsers themselves.  One
498 		 * of these will have already been set in the pset()
499 		 * routine.
500 		 * If libroff returns ROFF_TBL, then add it to the
501 		 * currently open parse.  Since we only get here if
502 		 * there does exist data (see tbl_data.c), we're
503 		 * guaranteed that something's been allocated.
504 		 * Do the same for ROFF_EQN.
505 		 */
506 
507 		rc = -1;
508 
509 		if (ROFF_TBL == rr)
510 			while (NULL != (span = roff_span(curp->roff))) {
511 				rc = curp->man ?
512 					man_addspan(curp->man, span) :
513 					mdoc_addspan(curp->mdoc, span);
514 				if (0 == rc)
515 					break;
516 			}
517 		else if (ROFF_EQN == rr)
518 			rc = curp->mdoc ?
519 				mdoc_addeqn(curp->mdoc,
520 					roff_eqn(curp->roff)) :
521 				man_addeqn(curp->man,
522 					roff_eqn(curp->roff));
523 		else if (curp->man || curp->mdoc)
524 			rc = curp->man ?
525 				man_parseln(curp->man,
526 					curp->line, ln.buf, of) :
527 				mdoc_parseln(curp->mdoc,
528 					curp->line, ln.buf, of);
529 
530 		if (0 == rc) {
531 			assert(MANDOCLEVEL_FATAL <= curp->file_status);
532 			break;
533 		}
534 
535 		/* Temporary buffers typically are not full. */
536 
537 		if (0 == start && '\0' == blk.buf[i])
538 			break;
539 
540 		/* Start the next input line. */
541 
542 		pos = 0;
543 	}
544 
545 	free(ln.buf);
546 }
547 
548 static void
549 pdesc(struct mparse *curp, const char *file, int fd)
550 {
551 	struct buf	 blk;
552 	int		 with_mmap;
553 
554 	/*
555 	 * Run for each opened file; may be called more than once for
556 	 * each full parse sequence if the opened file is nested (i.e.,
557 	 * from `so').  Simply sucks in the whole file and moves into
558 	 * the parse phase for the file.
559 	 */
560 
561 	if ( ! read_whole_file(file, fd, &blk, &with_mmap)) {
562 		curp->file_status = MANDOCLEVEL_SYSERR;
563 		return;
564 	}
565 
566 	/* Line number is per-file. */
567 
568 	curp->line = 1;
569 
570 	mparse_buf_r(curp, blk, 1);
571 
572 #ifdef	HAVE_MMAP
573 	if (with_mmap)
574 		munmap(blk.buf, blk.sz);
575 	else
576 #endif
577 		free(blk.buf);
578 }
579 
580 static int
581 read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap)
582 {
583 	size_t		 off;
584 	ssize_t		 ssz;
585 
586 #ifdef	HAVE_MMAP
587 	struct stat	 st;
588 	if (-1 == fstat(fd, &st)) {
589 		perror(file);
590 		return(0);
591 	}
592 
593 	/*
594 	 * If we're a regular file, try just reading in the whole entry
595 	 * via mmap().  This is faster than reading it into blocks, and
596 	 * since each file is only a few bytes to begin with, I'm not
597 	 * concerned that this is going to tank any machines.
598 	 */
599 
600 	if (S_ISREG(st.st_mode)) {
601 		if (st.st_size >= (1U << 31)) {
602 			fprintf(stderr, "%s: input too large\n", file);
603 			return(0);
604 		}
605 		*with_mmap = 1;
606 		fb->sz = (size_t)st.st_size;
607 		fb->buf = mmap(NULL, fb->sz, PROT_READ,
608 				MAP_FILE|MAP_SHARED, fd, 0);
609 		if (fb->buf != MAP_FAILED)
610 			return(1);
611 	}
612 #endif
613 
614 	/*
615 	 * If this isn't a regular file (like, say, stdin), then we must
616 	 * go the old way and just read things in bit by bit.
617 	 */
618 
619 	*with_mmap = 0;
620 	off = 0;
621 	fb->sz = 0;
622 	fb->buf = NULL;
623 	for (;;) {
624 		if (off == fb->sz) {
625 			if (fb->sz == (1U << 31)) {
626 				fprintf(stderr, "%s: input too large\n", file);
627 				break;
628 			}
629 			resize_buf(fb, 65536);
630 		}
631 		ssz = read(fd, fb->buf + (int)off, fb->sz - off);
632 		if (ssz == 0) {
633 			fb->sz = off;
634 			return(1);
635 		}
636 		if (ssz == -1) {
637 			perror(file);
638 			break;
639 		}
640 		off += (size_t)ssz;
641 	}
642 
643 	free(fb->buf);
644 	fb->buf = NULL;
645 	return(0);
646 }
647 
648 static void
649 mparse_end(struct mparse *curp)
650 {
651 
652 	if (MANDOCLEVEL_FATAL <= curp->file_status)
653 		return;
654 
655 	if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
656 		assert(MANDOCLEVEL_FATAL <= curp->file_status);
657 		return;
658 	}
659 
660 	if (curp->man && ! man_endparse(curp->man)) {
661 		assert(MANDOCLEVEL_FATAL <= curp->file_status);
662 		return;
663 	}
664 
665 	if ( ! (curp->man || curp->mdoc)) {
666 		mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
667 		curp->file_status = MANDOCLEVEL_FATAL;
668 		return;
669 	}
670 
671 	roff_endparse(curp->roff);
672 }
673 
674 static void
675 mparse_readfd_r(struct mparse *curp, int fd, const char *file, int re)
676 {
677 	const char	*svfile;
678 
679 	if (-1 == fd)
680 		if (-1 == (fd = open(file, O_RDONLY, 0))) {
681 			perror(file);
682 			curp->file_status = MANDOCLEVEL_SYSERR;
683 			return;
684 		}
685 
686 	svfile = curp->file;
687 	curp->file = file;
688 
689 	pdesc(curp, file, fd);
690 
691 	if (0 == re && MANDOCLEVEL_FATAL > curp->file_status)
692 		mparse_end(curp);
693 
694 	if (STDIN_FILENO != fd && -1 == close(fd))
695 		perror(file);
696 
697 	curp->file = svfile;
698 }
699 
700 enum mandoclevel
701 mparse_readfd(struct mparse *curp, int fd, const char *file)
702 {
703 
704 	mparse_readfd_r(curp, fd, file, 0);
705 	return(curp->file_status);
706 }
707 
708 struct mparse *
709 mparse_alloc(enum mparset inttype, enum mandoclevel wlevel, mandocmsg mmsg, void *arg)
710 {
711 	struct mparse	*curp;
712 
713 	assert(wlevel <= MANDOCLEVEL_FATAL);
714 
715 	curp = mandoc_calloc(1, sizeof(struct mparse));
716 
717 	curp->wlevel = wlevel;
718 	curp->mmsg = mmsg;
719 	curp->arg = arg;
720 	curp->inttype = inttype;
721 
722 	curp->roff = roff_alloc(curp);
723 	return(curp);
724 }
725 
726 void
727 mparse_reset(struct mparse *curp)
728 {
729 
730 	roff_reset(curp->roff);
731 
732 	if (curp->mdoc)
733 		mdoc_reset(curp->mdoc);
734 	if (curp->man)
735 		man_reset(curp->man);
736 	if (curp->secondary)
737 		curp->secondary->sz = 0;
738 
739 	curp->file_status = MANDOCLEVEL_OK;
740 	curp->mdoc = NULL;
741 	curp->man = NULL;
742 }
743 
744 void
745 mparse_free(struct mparse *curp)
746 {
747 
748 	if (curp->pmdoc)
749 		mdoc_free(curp->pmdoc);
750 	if (curp->pman)
751 		man_free(curp->pman);
752 	if (curp->roff)
753 		roff_free(curp->roff);
754 	if (curp->secondary)
755 		free(curp->secondary->buf);
756 
757 	free(curp->secondary);
758 	free(curp);
759 }
760 
761 void
762 mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
763 {
764 
765 	if (mdoc)
766 		*mdoc = curp->mdoc;
767 	if (man)
768 		*man = curp->man;
769 }
770 
771 void
772 mandoc_vmsg(enum mandocerr t, struct mparse *m,
773 		int ln, int pos, const char *fmt, ...)
774 {
775 	char		 buf[256];
776 	va_list		 ap;
777 
778 	va_start(ap, fmt);
779 	vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
780 	va_end(ap);
781 
782 	mandoc_msg(t, m, ln, pos, buf);
783 }
784 
785 void
786 mandoc_msg(enum mandocerr er, struct mparse *m,
787 		int ln, int col, const char *msg)
788 {
789 	enum mandoclevel level;
790 
791 	level = MANDOCLEVEL_FATAL;
792 	while (er < mandoclimits[level])
793 		level--;
794 
795 	if (level < m->wlevel)
796 		return;
797 
798 	if (m->mmsg)
799 		(*m->mmsg)(er, level, m->file, ln, col, msg);
800 
801 	if (m->file_status < level)
802 		m->file_status = level;
803 }
804 
805 const char *
806 mparse_strerror(enum mandocerr er)
807 {
808 
809 	return(mandocerrs[er]);
810 }
811 
812 const char *
813 mparse_strlevel(enum mandoclevel lvl)
814 {
815 	return(mandoclevels[lvl]);
816 }
817 
818 void
819 mparse_keep(struct mparse *p)
820 {
821 
822 	assert(NULL == p->secondary);
823 	p->secondary = mandoc_calloc(1, sizeof(struct buf));
824 }
825 
826 const char *
827 mparse_getkeep(const struct mparse *p)
828 {
829 
830 	assert(p->secondary);
831 	return(p->secondary->sz ? p->secondary->buf : NULL);
832 }
833