xref: /titanic_41/usr/src/cmd/mandoc/read.c (revision ffb8ebfab4941f959e7caea93ecfb348cfa3515e)
1 /*	$Id: read.c,v 1.39 2013/09/16 00:25:07 schwarze Exp $ */
2 /*
3  * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4  * Copyright (c) 2010, 2011, 2012, 2013 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 #define	REPARSE_LIMIT	1000
44 
45 struct	buf {
46 	char	 	 *buf; /* binary input buffer */
47 	size_t		  sz; /* size of binary buffer */
48 };
49 
50 struct	mparse {
51 	enum mandoclevel  file_status; /* status of current parse */
52 	enum mandoclevel  wlevel; /* ignore messages below this */
53 	int		  line; /* line number in the file */
54 	enum mparset	  inttype; /* which parser to use */
55 	struct man	 *pman; /* persistent man parser */
56 	struct mdoc	 *pmdoc; /* persistent mdoc parser */
57 	struct man	 *man; /* man parser */
58 	struct mdoc	 *mdoc; /* mdoc parser */
59 	struct roff	 *roff; /* roff parser (!NULL) */
60 	int		  reparse_count; /* finite interp. stack */
61 	mandocmsg	  mmsg; /* warning/error message handler */
62 	void		 *arg; /* argument to mmsg */
63 	const char	 *file;
64 	struct buf	 *secondary;
65 	char		 *defos; /* default operating system */
66 };
67 
68 static	void	  resize_buf(struct buf *, size_t);
69 static	void	  mparse_buf_r(struct mparse *, struct buf, int);
70 static	void	  pset(const char *, int, struct mparse *);
71 static	int	  read_whole_file(const char *, int, struct buf *, int *);
72 static	void	  mparse_end(struct mparse *);
73 static	void	  mparse_parse_buffer(struct mparse *, struct buf,
74 			const char *);
75 
76 static	const enum mandocerr	mandoclimits[MANDOCLEVEL_MAX] = {
77 	MANDOCERR_OK,
78 	MANDOCERR_WARNING,
79 	MANDOCERR_WARNING,
80 	MANDOCERR_ERROR,
81 	MANDOCERR_FATAL,
82 	MANDOCERR_MAX,
83 	MANDOCERR_MAX
84 };
85 
86 static	const char * const	mandocerrs[MANDOCERR_MAX] = {
87 	"ok",
88 
89 	"generic warning",
90 
91 	/* related to the prologue */
92 	"no title in document",
93 	"document title should be all caps",
94 	"unknown manual section",
95 	"unknown manual volume or arch",
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 	"sections out of conventional order",
108 	"duplicate section name",
109 	"section header suited to sections 2, 3, and 9 only",
110 
111 	/* related to macros and nesting */
112 	"skipping obsolete macro",
113 	"skipping paragraph macro",
114 	"moving paragraph macro out of list",
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 	"manual name not yet set",
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 column outside column list",
181 	"skipping end of block that is not open",
182 	"missing end of block",
183 	"scope open on exit",
184 	"uname(3) system call failed",
185 	"macro requires line argument(s)",
186 	"macro requires body argument(s)",
187 	"macro requires argument(s)",
188 	"request requires a numeric argument",
189 	"missing list type",
190 	"line argument(s) will be lost",
191 	"body argument(s) will be lost",
192 
193 	"generic fatal error",
194 
195 	"not a manual",
196 	"column syntax is inconsistent",
197 	"NOT IMPLEMENTED: .Bd -file",
198 	"argument count wrong, violates syntax",
199 	"child violates parent syntax",
200 	"argument count wrong, violates syntax",
201 	"NOT IMPLEMENTED: .so with absolute path or \"..\"",
202 	"no document body",
203 	"no document prologue",
204 	"static buffer exhausted",
205 };
206 
207 static	const char * const	mandoclevels[MANDOCLEVEL_MAX] = {
208 	"SUCCESS",
209 	"RESERVED",
210 	"WARNING",
211 	"ERROR",
212 	"FATAL",
213 	"BADARG",
214 	"SYSERR"
215 };
216 
217 static void
resize_buf(struct buf * buf,size_t initial)218 resize_buf(struct buf *buf, size_t initial)
219 {
220 
221 	buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
222 	buf->buf = mandoc_realloc(buf->buf, buf->sz);
223 }
224 
225 static void
pset(const char * buf,int pos,struct mparse * curp)226 pset(const char *buf, int pos, struct mparse *curp)
227 {
228 	int		 i;
229 
230 	/*
231 	 * Try to intuit which kind of manual parser should be used.  If
232 	 * passed in by command-line (-man, -mdoc), then use that
233 	 * explicitly.  If passed as -mandoc, then try to guess from the
234 	 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
235 	 * default to -man, which is more lenient.
236 	 *
237 	 * Separate out pmdoc/pman from mdoc/man: the first persists
238 	 * through all parsers, while the latter is used per-parse.
239 	 */
240 
241 	if ('.' == buf[0] || '\'' == buf[0]) {
242 		for (i = 1; buf[i]; i++)
243 			if (' ' != buf[i] && '\t' != buf[i])
244 				break;
245 		if ('\0' == buf[i])
246 			return;
247 	}
248 
249 	switch (curp->inttype) {
250 	case (MPARSE_MDOC):
251 		if (NULL == curp->pmdoc)
252 			curp->pmdoc = mdoc_alloc(curp->roff, curp,
253 					curp->defos);
254 		assert(curp->pmdoc);
255 		curp->mdoc = curp->pmdoc;
256 		return;
257 	case (MPARSE_MAN):
258 		if (NULL == curp->pman)
259 			curp->pman = man_alloc(curp->roff, curp);
260 		assert(curp->pman);
261 		curp->man = curp->pman;
262 		return;
263 	default:
264 		break;
265 	}
266 
267 	if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3))  {
268 		if (NULL == curp->pmdoc)
269 			curp->pmdoc = mdoc_alloc(curp->roff, curp,
270 					curp->defos);
271 		assert(curp->pmdoc);
272 		curp->mdoc = curp->pmdoc;
273 		return;
274 	}
275 
276 	if (NULL == curp->pman)
277 		curp->pman = man_alloc(curp->roff, curp);
278 	assert(curp->pman);
279 	curp->man = curp->pman;
280 }
281 
282 /*
283  * Main parse routine for an opened file.  This is called for each
284  * opened file and simply loops around the full input file, possibly
285  * nesting (i.e., with `so').
286  */
287 static void
mparse_buf_r(struct mparse * curp,struct buf blk,int start)288 mparse_buf_r(struct mparse *curp, struct buf blk, int start)
289 {
290 	const struct tbl_span	*span;
291 	struct buf	 ln;
292 	enum rofferr	 rr;
293 	int		 i, of, rc;
294 	int		 pos; /* byte number in the ln buffer */
295 	int		 lnn; /* line number in the real file */
296 	unsigned char	 c;
297 
298 	memset(&ln, 0, sizeof(struct buf));
299 
300 	lnn = curp->line;
301 	pos = 0;
302 
303 	for (i = 0; i < (int)blk.sz; ) {
304 		if (0 == pos && '\0' == blk.buf[i])
305 			break;
306 
307 		if (start) {
308 			curp->line = lnn;
309 			curp->reparse_count = 0;
310 		}
311 
312 		while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
313 
314 			/*
315 			 * When finding an unescaped newline character,
316 			 * leave the character loop to process the line.
317 			 * Skip a preceding carriage return, if any.
318 			 */
319 
320 			if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
321 			    '\n' == blk.buf[i + 1])
322 				++i;
323 			if ('\n' == blk.buf[i]) {
324 				++i;
325 				++lnn;
326 				break;
327 			}
328 
329 			/*
330 			 * Make sure we have space for at least
331 			 * one backslash and one other character
332 			 * and the trailing NUL byte.
333 			 */
334 
335 			if (pos + 2 >= (int)ln.sz)
336 				resize_buf(&ln, 256);
337 
338 			/*
339 			 * Warn about bogus characters.  If you're using
340 			 * non-ASCII encoding, you're screwing your
341 			 * readers.  Since I'd rather this not happen,
342 			 * I'll be helpful and replace these characters
343 			 * with "?", so we don't display gibberish.
344 			 * Note to manual writers: use special characters.
345 			 */
346 
347 			c = (unsigned char) blk.buf[i];
348 
349 			if ( ! (isascii(c) &&
350 					(isgraph(c) || isblank(c)))) {
351 				mandoc_msg(MANDOCERR_BADCHAR, curp,
352 						curp->line, pos, NULL);
353 				i++;
354 				ln.buf[pos++] = '?';
355 				continue;
356 			}
357 
358 			/* Trailing backslash = a plain char. */
359 
360 			if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
361 				ln.buf[pos++] = blk.buf[i++];
362 				continue;
363 			}
364 
365 			/*
366 			 * Found escape and at least one other character.
367 			 * When it's a newline character, skip it.
368 			 * When there is a carriage return in between,
369 			 * skip that one as well.
370 			 */
371 
372 			if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
373 			    '\n' == blk.buf[i + 2])
374 				++i;
375 			if ('\n' == blk.buf[i + 1]) {
376 				i += 2;
377 				++lnn;
378 				continue;
379 			}
380 
381 			if ('"' == blk.buf[i + 1] || '#' == blk.buf[i + 1]) {
382 				i += 2;
383 				/* Comment, skip to end of line */
384 				for (; i < (int)blk.sz; ++i) {
385 					if ('\n' == blk.buf[i]) {
386 						++i;
387 						++lnn;
388 						break;
389 					}
390 				}
391 
392 				/* Backout trailing whitespaces */
393 				for (; pos > 0; --pos) {
394 					if (ln.buf[pos - 1] != ' ')
395 						break;
396 					if (pos > 2 && ln.buf[pos - 2] == '\\')
397 						break;
398 				}
399 				break;
400 			}
401 
402 			/* Catch escaped bogus characters. */
403 
404 			c = (unsigned char) blk.buf[i+1];
405 
406 			if ( ! (isascii(c) &&
407 					(isgraph(c) || isblank(c)))) {
408 				mandoc_msg(MANDOCERR_BADCHAR, curp,
409 						curp->line, pos, NULL);
410 				i += 2;
411 				ln.buf[pos++] = '?';
412 				continue;
413 			}
414 
415 			/* Some other escape sequence, copy & cont. */
416 
417 			ln.buf[pos++] = blk.buf[i++];
418 			ln.buf[pos++] = blk.buf[i++];
419 		}
420 
421  		if (pos >= (int)ln.sz)
422 			resize_buf(&ln, 256);
423 
424 		ln.buf[pos] = '\0';
425 
426 		/*
427 		 * A significant amount of complexity is contained by
428 		 * the roff preprocessor.  It's line-oriented but can be
429 		 * expressed on one line, so we need at times to
430 		 * readjust our starting point and re-run it.  The roff
431 		 * preprocessor can also readjust the buffers with new
432 		 * data, so we pass them in wholesale.
433 		 */
434 
435 		of = 0;
436 
437 		/*
438 		 * Maintain a lookaside buffer of all parsed lines.  We
439 		 * only do this if mparse_keep() has been invoked (the
440 		 * buffer may be accessed with mparse_getkeep()).
441 		 */
442 
443 		if (curp->secondary) {
444 			curp->secondary->buf =
445 				mandoc_realloc
446 				(curp->secondary->buf,
447 				 curp->secondary->sz + pos + 2);
448 			memcpy(curp->secondary->buf +
449 					curp->secondary->sz,
450 					ln.buf, pos);
451 			curp->secondary->sz += pos;
452 			curp->secondary->buf
453 				[curp->secondary->sz] = '\n';
454 			curp->secondary->sz++;
455 			curp->secondary->buf
456 				[curp->secondary->sz] = '\0';
457 		}
458 rerun:
459 		rr = roff_parseln
460 			(curp->roff, curp->line,
461 			 &ln.buf, &ln.sz, of, &of);
462 
463 		switch (rr) {
464 		case (ROFF_REPARSE):
465 			if (REPARSE_LIMIT >= ++curp->reparse_count)
466 				mparse_buf_r(curp, ln, 0);
467 			else
468 				mandoc_msg(MANDOCERR_ROFFLOOP, curp,
469 					curp->line, pos, NULL);
470 			pos = 0;
471 			continue;
472 		case (ROFF_APPEND):
473 			pos = (int)strlen(ln.buf);
474 			continue;
475 		case (ROFF_RERUN):
476 			goto rerun;
477 		case (ROFF_IGN):
478 			pos = 0;
479 			continue;
480 		case (ROFF_ERR):
481 			assert(MANDOCLEVEL_FATAL <= curp->file_status);
482 			break;
483 		case (ROFF_SO):
484 			/*
485 			 * We remove `so' clauses from our lookaside
486 			 * buffer because we're going to descend into
487 			 * the file recursively.
488 			 */
489 			if (curp->secondary)
490 				curp->secondary->sz -= pos + 1;
491 			mparse_readfd(curp, -1, ln.buf + of);
492 			if (MANDOCLEVEL_FATAL <= curp->file_status)
493 				break;
494 			pos = 0;
495 			continue;
496 		default:
497 			break;
498 		}
499 
500 		/*
501 		 * If we encounter errors in the recursive parse, make
502 		 * sure we don't continue parsing.
503 		 */
504 
505 		if (MANDOCLEVEL_FATAL <= curp->file_status)
506 			break;
507 
508 		/*
509 		 * If input parsers have not been allocated, do so now.
510 		 * We keep these instanced between parsers, but set them
511 		 * locally per parse routine since we can use different
512 		 * parsers with each one.
513 		 */
514 
515 		if ( ! (curp->man || curp->mdoc))
516 			pset(ln.buf + of, pos - of, curp);
517 
518 		/*
519 		 * Lastly, push down into the parsers themselves.  One
520 		 * of these will have already been set in the pset()
521 		 * routine.
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 
529 		rc = -1;
530 
531 		if (ROFF_TBL == rr)
532 			while (NULL != (span = roff_span(curp->roff))) {
533 				rc = curp->man ?
534 					man_addspan(curp->man, span) :
535 					mdoc_addspan(curp->mdoc, span);
536 				if (0 == rc)
537 					break;
538 			}
539 		else if (ROFF_EQN == rr)
540 			rc = curp->mdoc ?
541 				mdoc_addeqn(curp->mdoc,
542 					roff_eqn(curp->roff)) :
543 				man_addeqn(curp->man,
544 					roff_eqn(curp->roff));
545 		else if (curp->man || curp->mdoc)
546 			rc = curp->man ?
547 				man_parseln(curp->man,
548 					curp->line, ln.buf, of) :
549 				mdoc_parseln(curp->mdoc,
550 					curp->line, ln.buf, of);
551 
552 		if (0 == rc) {
553 			assert(MANDOCLEVEL_FATAL <= curp->file_status);
554 			break;
555 		}
556 
557 		/* Temporary buffers typically are not full. */
558 
559 		if (0 == start && '\0' == blk.buf[i])
560 			break;
561 
562 		/* Start the next input line. */
563 
564 		pos = 0;
565 	}
566 
567 	free(ln.buf);
568 }
569 
570 static int
read_whole_file(const char * file,int fd,struct buf * fb,int * with_mmap)571 read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap)
572 {
573 	size_t		 off;
574 	ssize_t		 ssz;
575 
576 #ifdef	HAVE_MMAP
577 	struct stat	 st;
578 	if (-1 == fstat(fd, &st)) {
579 		perror(file);
580 		return(0);
581 	}
582 
583 	/*
584 	 * If we're a regular file, try just reading in the whole entry
585 	 * via mmap().  This is faster than reading it into blocks, and
586 	 * since each file is only a few bytes to begin with, I'm not
587 	 * concerned that this is going to tank any machines.
588 	 */
589 
590 	if (S_ISREG(st.st_mode)) {
591 		if (st.st_size >= (1U << 31)) {
592 			fprintf(stderr, "%s: input too large\n", file);
593 			return(0);
594 		}
595 		*with_mmap = 1;
596 		fb->sz = (size_t)st.st_size;
597 		fb->buf = mmap(NULL, fb->sz, PROT_READ, MAP_SHARED, fd, 0);
598 		if (fb->buf != MAP_FAILED)
599 			return(1);
600 	}
601 #endif
602 
603 	/*
604 	 * If this isn't a regular file (like, say, stdin), then we must
605 	 * go the old way and just read things in bit by bit.
606 	 */
607 
608 	*with_mmap = 0;
609 	off = 0;
610 	fb->sz = 0;
611 	fb->buf = NULL;
612 	for (;;) {
613 		if (off == fb->sz) {
614 			if (fb->sz == (1U << 31)) {
615 				fprintf(stderr, "%s: input too large\n", file);
616 				break;
617 			}
618 			resize_buf(fb, 65536);
619 		}
620 		ssz = read(fd, fb->buf + (int)off, fb->sz - off);
621 		if (ssz == 0) {
622 			fb->sz = off;
623 			return(1);
624 		}
625 		if (ssz == -1) {
626 			perror(file);
627 			break;
628 		}
629 		off += (size_t)ssz;
630 	}
631 
632 	free(fb->buf);
633 	fb->buf = NULL;
634 	return(0);
635 }
636 
637 static void
mparse_end(struct mparse * curp)638 mparse_end(struct mparse *curp)
639 {
640 
641 	if (MANDOCLEVEL_FATAL <= curp->file_status)
642 		return;
643 
644 	if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
645 		assert(MANDOCLEVEL_FATAL <= curp->file_status);
646 		return;
647 	}
648 
649 	if (curp->man && ! man_endparse(curp->man)) {
650 		assert(MANDOCLEVEL_FATAL <= curp->file_status);
651 		return;
652 	}
653 
654 	if ( ! (curp->man || curp->mdoc)) {
655 		mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
656 		curp->file_status = MANDOCLEVEL_FATAL;
657 		return;
658 	}
659 
660 	roff_endparse(curp->roff);
661 }
662 
663 static void
mparse_parse_buffer(struct mparse * curp,struct buf blk,const char * file)664 mparse_parse_buffer(struct mparse *curp, struct buf blk, const char *file)
665 {
666 	const char	*svfile;
667 	static int	 recursion_depth;
668 
669 	if (64 < recursion_depth) {
670 		mandoc_msg(MANDOCERR_ROFFLOOP, curp, curp->line, 0, NULL);
671 		return;
672 	}
673 
674 	/* Line number is per-file. */
675 	svfile = curp->file;
676 	curp->file = file;
677 	curp->line = 1;
678 	recursion_depth++;
679 
680 	mparse_buf_r(curp, blk, 1);
681 
682 	if (0 == --recursion_depth && MANDOCLEVEL_FATAL > curp->file_status)
683 		mparse_end(curp);
684 
685 	curp->file = svfile;
686 }
687 
688 enum mandoclevel
mparse_readmem(struct mparse * curp,const void * buf,size_t len,const char * file)689 mparse_readmem(struct mparse *curp, const void *buf, size_t len,
690 		const char *file)
691 {
692 	struct buf blk;
693 
694 	blk.buf = UNCONST(buf);
695 	blk.sz = len;
696 
697 	mparse_parse_buffer(curp, blk, file);
698 	return(curp->file_status);
699 }
700 
701 enum mandoclevel
mparse_readfd(struct mparse * curp,int fd,const char * file)702 mparse_readfd(struct mparse *curp, int fd, const char *file)
703 {
704 	struct buf	 blk;
705 	int		 with_mmap;
706 
707 	if (-1 == fd)
708 		if (-1 == (fd = open(file, O_RDONLY, 0))) {
709 			perror(file);
710 			curp->file_status = MANDOCLEVEL_SYSERR;
711 			goto out;
712 		}
713 	/*
714 	 * Run for each opened file; may be called more than once for
715 	 * each full parse sequence if the opened file is nested (i.e.,
716 	 * from `so').  Simply sucks in the whole file and moves into
717 	 * the parse phase for the file.
718 	 */
719 
720 	if ( ! read_whole_file(file, fd, &blk, &with_mmap)) {
721 		curp->file_status = MANDOCLEVEL_SYSERR;
722 		goto out;
723 	}
724 
725 	mparse_parse_buffer(curp, blk, file);
726 
727 #ifdef	HAVE_MMAP
728 	if (with_mmap)
729 		munmap(blk.buf, blk.sz);
730 	else
731 #endif
732 		free(blk.buf);
733 
734 	if (STDIN_FILENO != fd && -1 == close(fd))
735 		perror(file);
736 out:
737 	return(curp->file_status);
738 }
739 
740 struct mparse *
mparse_alloc(enum mparset inttype,enum mandoclevel wlevel,mandocmsg mmsg,void * arg,char * defos)741 mparse_alloc(enum mparset inttype, enum mandoclevel wlevel,
742 		mandocmsg mmsg, void *arg, char *defos)
743 {
744 	struct mparse	*curp;
745 
746 	assert(wlevel <= MANDOCLEVEL_FATAL);
747 
748 	curp = mandoc_calloc(1, sizeof(struct mparse));
749 
750 	curp->wlevel = wlevel;
751 	curp->mmsg = mmsg;
752 	curp->arg = arg;
753 	curp->inttype = inttype;
754 	curp->defos = defos;
755 
756 	curp->roff = roff_alloc(inttype, curp);
757 	return(curp);
758 }
759 
760 void
mparse_reset(struct mparse * curp)761 mparse_reset(struct mparse *curp)
762 {
763 
764 	roff_reset(curp->roff);
765 
766 	if (curp->mdoc)
767 		mdoc_reset(curp->mdoc);
768 	if (curp->man)
769 		man_reset(curp->man);
770 	if (curp->secondary)
771 		curp->secondary->sz = 0;
772 
773 	curp->file_status = MANDOCLEVEL_OK;
774 	curp->mdoc = NULL;
775 	curp->man = NULL;
776 }
777 
778 void
mparse_free(struct mparse * curp)779 mparse_free(struct mparse *curp)
780 {
781 
782 	if (curp->pmdoc)
783 		mdoc_free(curp->pmdoc);
784 	if (curp->pman)
785 		man_free(curp->pman);
786 	if (curp->roff)
787 		roff_free(curp->roff);
788 	if (curp->secondary)
789 		free(curp->secondary->buf);
790 
791 	free(curp->secondary);
792 	free(curp);
793 }
794 
795 void
mparse_result(struct mparse * curp,struct mdoc ** mdoc,struct man ** man)796 mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
797 {
798 
799 	if (mdoc)
800 		*mdoc = curp->mdoc;
801 	if (man)
802 		*man = curp->man;
803 }
804 
805 void
mandoc_vmsg(enum mandocerr t,struct mparse * m,int ln,int pos,const char * fmt,...)806 mandoc_vmsg(enum mandocerr t, struct mparse *m,
807 		int ln, int pos, const char *fmt, ...)
808 {
809 	char		 buf[256];
810 	va_list		 ap;
811 
812 	va_start(ap, fmt);
813 	vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
814 	va_end(ap);
815 
816 	mandoc_msg(t, m, ln, pos, buf);
817 }
818 
819 void
mandoc_msg(enum mandocerr er,struct mparse * m,int ln,int col,const char * msg)820 mandoc_msg(enum mandocerr er, struct mparse *m,
821 		int ln, int col, const char *msg)
822 {
823 	enum mandoclevel level;
824 
825 	level = MANDOCLEVEL_FATAL;
826 	while (er < mandoclimits[level])
827 		level--;
828 
829 	if (level < m->wlevel)
830 		return;
831 
832 	if (m->mmsg)
833 		(*m->mmsg)(er, level, m->file, ln, col, msg);
834 
835 	if (m->file_status < level)
836 		m->file_status = level;
837 }
838 
839 const char *
mparse_strerror(enum mandocerr er)840 mparse_strerror(enum mandocerr er)
841 {
842 
843 	return(mandocerrs[er]);
844 }
845 
846 const char *
mparse_strlevel(enum mandoclevel lvl)847 mparse_strlevel(enum mandoclevel lvl)
848 {
849 	return(mandoclevels[lvl]);
850 }
851 
852 void
mparse_keep(struct mparse * p)853 mparse_keep(struct mparse *p)
854 {
855 
856 	assert(NULL == p->secondary);
857 	p->secondary = mandoc_calloc(1, sizeof(struct buf));
858 }
859 
860 const char *
mparse_getkeep(const struct mparse * p)861 mparse_getkeep(const struct mparse *p)
862 {
863 
864 	assert(p->secondary);
865 	return(p->secondary->sz ? p->secondary->buf : NULL);
866 }
867