read.c revision 1.2 
1/*	$Vendor-Id: read.c,v 1.10 2011/04/03 10:11:25 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#include <sys/stat.h>
23#include <sys/mman.h>
24
25#include <assert.h>
26#include <ctype.h>
27#include <fcntl.h>
28#include <stdarg.h>
29#include <stdio.h>
30#include <stdlib.h>
31#include <string.h>
32#include <unistd.h>
33
34#include "mandoc.h"
35#include "libmandoc.h"
36#include "mdoc.h"
37#include "man.h"
38
39#ifndef MAP_FILE
40#define	MAP_FILE	0
41#endif
42
43#define	REPARSE_LIMIT	1000
44
45struct	buf {
46	char	 	 *buf; /* binary input buffer */
47	size_t		  sz; /* size of binary buffer */
48};
49
50struct	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	struct regset	  regs; /* roff registers */
61	int		  reparse_count; /* finite interp. stack */
62	mandocmsg	  mmsg; /* warning/error message handler */
63	void		 *arg; /* argument to mmsg */
64	const char	 *file;
65};
66
67static	void	  resize_buf(struct buf *, size_t);
68static	void	  mparse_buf_r(struct mparse *, struct buf, int);
69static	void	  mparse_readfd_r(struct mparse *, int, const char *, int);
70static	void	  pset(const char *, int, struct mparse *);
71static	void	  pdesc(struct mparse *, const char *, int);
72static	int	  read_whole_file(const char *, int, struct buf *, int *);
73static	void	  mparse_end(struct mparse *);
74
75static	const enum mandocerr	mandoclimits[MANDOCLEVEL_MAX] = {
76	MANDOCERR_OK,
77	MANDOCERR_WARNING,
78	MANDOCERR_WARNING,
79	MANDOCERR_ERROR,
80	MANDOCERR_FATAL,
81	MANDOCERR_MAX,
82	MANDOCERR_MAX
83};
84
85static	const char * const	mandocerrs[MANDOCERR_MAX] = {
86	"ok",
87
88	"generic warning",
89
90	/* related to the prologue */
91	"no title in document",
92	"document title should be all caps",
93	"unknown manual section",
94	"date missing, using today's date",
95	"cannot parse date, using it verbatim",
96	"prologue macros out of order",
97	"duplicate prologue macro",
98	"macro not allowed in prologue",
99	"macro not allowed in body",
100
101	/* related to document structure */
102	".so is fragile, better use ln(1)",
103	"NAME section must come first",
104	"bad NAME section contents",
105	"manual name not yet set",
106	"sections out of conventional order",
107	"duplicate section name",
108	"section not in conventional manual section",
109
110	/* related to macros and nesting */
111	"skipping obsolete macro",
112	"skipping paragraph macro",
113	"skipping no-space macro",
114	"blocks badly nested",
115	"child violates parent syntax",
116	"nested displays are not portable",
117	"already in literal mode",
118	"line scope broken",
119
120	/* related to missing macro arguments */
121	"skipping empty macro",
122	"argument count wrong",
123	"missing display type",
124	"list type must come first",
125	"tag lists require a width argument",
126	"missing font type",
127	"skipping end of block that is not open",
128
129	/* related to bad macro arguments */
130	"skipping argument",
131	"duplicate argument",
132	"duplicate display type",
133	"duplicate list type",
134	"unknown AT&T UNIX version",
135	"bad Boolean value",
136	"unknown font",
137	"unknown standard specifier",
138	"bad width argument",
139
140	/* related to plain text */
141	"blank line in non-literal context",
142	"tab in non-literal context",
143	"end of line whitespace",
144	"bad comment style",
145	"unknown escape sequence",
146	"unterminated quoted string",
147
148	"generic error",
149
150	/* related to tables */
151	"bad table syntax",
152	"bad table option",
153	"bad table layout",
154	"no table layout cells specified",
155	"no table data cells specified",
156	"ignore data in cell",
157	"data block still open",
158	"ignoring extra data cells",
159
160	"input stack limit exceeded, infinite loop?",
161	"skipping bad character",
162	"escaped character not allowed in a name",
163	"skipping text before the first section header",
164	"skipping unknown macro",
165	"NOT IMPLEMENTED, please use groff: skipping request",
166	"argument count wrong",
167	"skipping end of block that is not open",
168	"missing end of block",
169	"scope open on exit",
170	"uname(3) system call failed",
171	"macro requires line argument(s)",
172	"macro requires body argument(s)",
173	"macro requires argument(s)",
174	"missing list type",
175	"line argument(s) will be lost",
176	"body argument(s) will be lost",
177
178	"generic fatal error",
179
180	"not a manual",
181	"column syntax is inconsistent",
182	"NOT IMPLEMENTED: .Bd -file",
183	"line scope broken, syntax violated",
184	"argument count wrong, violates syntax",
185	"child violates parent syntax",
186	"argument count wrong, violates syntax",
187	"NOT IMPLEMENTED: .so with absolute path or \"..\"",
188	"no document body",
189	"no document prologue",
190	"static buffer exhausted",
191};
192
193static	const char * const	mandoclevels[MANDOCLEVEL_MAX] = {
194	"SUCCESS",
195	"RESERVED",
196	"WARNING",
197	"ERROR",
198	"FATAL",
199	"BADARG",
200	"SYSERR"
201};
202
203static void
204resize_buf(struct buf *buf, size_t initial)
205{
206
207	buf->sz = buf->sz > initial/2 ? 2 * buf->sz : initial;
208	buf->buf = mandoc_realloc(buf->buf, buf->sz);
209}
210
211static void
212pset(const char *buf, int pos, struct mparse *curp)
213{
214	int		 i;
215
216	/*
217	 * Try to intuit which kind of manual parser should be used.  If
218	 * passed in by command-line (-man, -mdoc), then use that
219	 * explicitly.  If passed as -mandoc, then try to guess from the
220	 * line: either skip dot-lines, use -mdoc when finding `.Dt', or
221	 * default to -man, which is more lenient.
222	 *
223	 * Separate out pmdoc/pman from mdoc/man: the first persists
224	 * through all parsers, while the latter is used per-parse.
225	 */
226
227	if ('.' == buf[0] || '\'' == buf[0]) {
228		for (i = 1; buf[i]; i++)
229			if (' ' != buf[i] && '\t' != buf[i])
230				break;
231		if ('\0' == buf[i])
232			return;
233	}
234
235	switch (curp->inttype) {
236	case (MPARSE_MDOC):
237		if (NULL == curp->pmdoc)
238			curp->pmdoc = mdoc_alloc(&curp->regs, curp);
239		assert(curp->pmdoc);
240		curp->mdoc = curp->pmdoc;
241		return;
242	case (MPARSE_MAN):
243		if (NULL == curp->pman)
244			curp->pman = man_alloc(&curp->regs, curp);
245		assert(curp->pman);
246		curp->man = curp->pman;
247		return;
248	default:
249		break;
250	}
251
252	if (pos >= 3 && 0 == memcmp(buf, ".Dd", 3))  {
253		if (NULL == curp->pmdoc)
254			curp->pmdoc = mdoc_alloc(&curp->regs, curp);
255		assert(curp->pmdoc);
256		curp->mdoc = curp->pmdoc;
257		return;
258	}
259
260	if (NULL == curp->pman)
261		curp->pman = man_alloc(&curp->regs, curp);
262	assert(curp->pman);
263	curp->man = curp->pman;
264}
265
266/*
267 * Main parse routine for an opened file.  This is called for each
268 * opened file and simply loops around the full input file, possibly
269 * nesting (i.e., with `so').
270 */
271static void
272mparse_buf_r(struct mparse *curp, struct buf blk, int start)
273{
274	const struct tbl_span	*span;
275	struct buf	 ln;
276	enum rofferr	 rr;
277	int		 i, of, rc;
278	int		 pos; /* byte number in the ln buffer */
279	int		 lnn; /* line number in the real file */
280	unsigned char	 c;
281
282	memset(&ln, 0, sizeof(struct buf));
283
284	lnn = curp->line;
285	pos = 0;
286
287	for (i = 0; i < (int)blk.sz; ) {
288		if (0 == pos && '\0' == blk.buf[i])
289			break;
290
291		if (start) {
292			curp->line = lnn;
293			curp->reparse_count = 0;
294		}
295
296		while (i < (int)blk.sz && (start || '\0' != blk.buf[i])) {
297
298			/*
299			 * When finding an unescaped newline character,
300			 * leave the character loop to process the line.
301			 * Skip a preceding carriage return, if any.
302			 */
303
304			if ('\r' == blk.buf[i] && i + 1 < (int)blk.sz &&
305			    '\n' == blk.buf[i + 1])
306				++i;
307			if ('\n' == blk.buf[i]) {
308				++i;
309				++lnn;
310				break;
311			}
312
313			/*
314			 * Warn about bogus characters.  If you're using
315			 * non-ASCII encoding, you're screwing your
316			 * readers.  Since I'd rather this not happen,
317			 * I'll be helpful and drop these characters so
318			 * we don't display gibberish.  Note to manual
319			 * writers: use special characters.
320			 */
321
322			c = (unsigned char) blk.buf[i];
323
324			if ( ! (isascii(c) &&
325					(isgraph(c) || isblank(c)))) {
326				mandoc_msg(MANDOCERR_BADCHAR, curp,
327						curp->line, pos, "ignoring byte");
328				i++;
329				continue;
330			}
331
332			/* Trailing backslash = a plain char. */
333
334			if ('\\' != blk.buf[i] || i + 1 == (int)blk.sz) {
335				if (pos >= (int)ln.sz)
336					resize_buf(&ln, 256);
337				ln.buf[pos++] = blk.buf[i++];
338				continue;
339			}
340
341			/*
342			 * Found escape and at least one other character.
343			 * When it's a newline character, skip it.
344			 * When there is a carriage return in between,
345			 * skip that one as well.
346			 */
347
348			if ('\r' == blk.buf[i + 1] && i + 2 < (int)blk.sz &&
349			    '\n' == blk.buf[i + 2])
350				++i;
351			if ('\n' == blk.buf[i + 1]) {
352				i += 2;
353				++lnn;
354				continue;
355			}
356
357			if ('"' == blk.buf[i + 1]) {
358				i += 2;
359				/* Comment, skip to end of line */
360				for (; i < (int)blk.sz; ++i) {
361					if ('\n' == blk.buf[i]) {
362						++i;
363						++lnn;
364						break;
365					}
366				}
367
368				/* Backout trailing whitespaces */
369				for (; pos > 0; --pos) {
370					if (ln.buf[pos - 1] != ' ')
371						break;
372					if (pos > 2 && ln.buf[pos - 2] == '\\')
373						break;
374				}
375				break;
376			}
377
378			/* Some other escape sequence, copy & cont. */
379
380			if (pos + 1 >= (int)ln.sz)
381				resize_buf(&ln, 256);
382
383			ln.buf[pos++] = blk.buf[i++];
384			ln.buf[pos++] = blk.buf[i++];
385		}
386
387 		if (pos >= (int)ln.sz)
388			resize_buf(&ln, 256);
389
390		ln.buf[pos] = '\0';
391
392		/*
393		 * A significant amount of complexity is contained by
394		 * the roff preprocessor.  It's line-oriented but can be
395		 * expressed on one line, so we need at times to
396		 * readjust our starting point and re-run it.  The roff
397		 * preprocessor can also readjust the buffers with new
398		 * data, so we pass them in wholesale.
399		 */
400
401		of = 0;
402
403rerun:
404		rr = roff_parseln
405			(curp->roff, curp->line,
406			 &ln.buf, &ln.sz, of, &of);
407
408		switch (rr) {
409		case (ROFF_REPARSE):
410			if (REPARSE_LIMIT >= ++curp->reparse_count)
411				mparse_buf_r(curp, ln, 0);
412			else
413				mandoc_msg(MANDOCERR_ROFFLOOP, curp,
414					curp->line, pos, NULL);
415			pos = 0;
416			continue;
417		case (ROFF_APPEND):
418			pos = (int)strlen(ln.buf);
419			continue;
420		case (ROFF_RERUN):
421			goto rerun;
422		case (ROFF_IGN):
423			pos = 0;
424			continue;
425		case (ROFF_ERR):
426			assert(MANDOCLEVEL_FATAL <= curp->file_status);
427			break;
428		case (ROFF_SO):
429			mparse_readfd_r(curp, -1, ln.buf + of, 1);
430			if (MANDOCLEVEL_FATAL <= curp->file_status)
431				break;
432			pos = 0;
433			continue;
434		default:
435			break;
436		}
437
438		/*
439		 * If we encounter errors in the recursive parse, make
440		 * sure we don't continue parsing.
441		 */
442
443		if (MANDOCLEVEL_FATAL <= curp->file_status)
444			break;
445
446		/*
447		 * If input parsers have not been allocated, do so now.
448		 * We keep these instanced betwen parsers, but set them
449		 * locally per parse routine since we can use different
450		 * parsers with each one.
451		 */
452
453		if ( ! (curp->man || curp->mdoc))
454			pset(ln.buf + of, pos - of, curp);
455
456		/*
457		 * Lastly, push down into the parsers themselves.  One
458		 * of these will have already been set in the pset()
459		 * routine.
460		 * If libroff returns ROFF_TBL, then add it to the
461		 * currently open parse.  Since we only get here if
462		 * there does exist data (see tbl_data.c), we're
463		 * guaranteed that something's been allocated.
464		 * Do the same for ROFF_EQN.
465		 */
466
467		rc = -1;
468
469		if (ROFF_TBL == rr)
470			while (NULL != (span = roff_span(curp->roff))) {
471				rc = curp->man ?
472					man_addspan(curp->man, span) :
473					mdoc_addspan(curp->mdoc, span);
474				if (0 == rc)
475					break;
476			}
477		else if (ROFF_EQN == rr)
478			rc = curp->mdoc ?
479				mdoc_addeqn(curp->mdoc,
480					roff_eqn(curp->roff)) :
481				man_addeqn(curp->man,
482					roff_eqn(curp->roff));
483		else if (curp->man || curp->mdoc)
484			rc = curp->man ?
485				man_parseln(curp->man,
486					curp->line, ln.buf, of) :
487				mdoc_parseln(curp->mdoc,
488					curp->line, ln.buf, of);
489
490		if (0 == rc) {
491			assert(MANDOCLEVEL_FATAL <= curp->file_status);
492			break;
493		}
494
495		/* Temporary buffers typically are not full. */
496
497		if (0 == start && '\0' == blk.buf[i])
498			break;
499
500		/* Start the next input line. */
501
502		pos = 0;
503	}
504
505	free(ln.buf);
506}
507
508static void
509pdesc(struct mparse *curp, const char *file, int fd)
510{
511	struct buf	 blk;
512	int		 with_mmap;
513
514	/*
515	 * Run for each opened file; may be called more than once for
516	 * each full parse sequence if the opened file is nested (i.e.,
517	 * from `so').  Simply sucks in the whole file and moves into
518	 * the parse phase for the file.
519	 */
520
521	if ( ! read_whole_file(file, fd, &blk, &with_mmap)) {
522		curp->file_status = MANDOCLEVEL_SYSERR;
523		return;
524	}
525
526	/* Line number is per-file. */
527
528	curp->line = 1;
529
530	mparse_buf_r(curp, blk, 1);
531
532	if (with_mmap)
533		munmap(blk.buf, blk.sz);
534	else
535		free(blk.buf);
536}
537
538static int
539read_whole_file(const char *file, int fd, struct buf *fb, int *with_mmap)
540{
541	struct stat	 st;
542	size_t		 off;
543	ssize_t		 ssz;
544
545	if (-1 == fstat(fd, &st)) {
546		perror(file);
547		return(0);
548	}
549
550	/*
551	 * If we're a regular file, try just reading in the whole entry
552	 * via mmap().  This is faster than reading it into blocks, and
553	 * since each file is only a few bytes to begin with, I'm not
554	 * concerned that this is going to tank any machines.
555	 */
556
557	if (S_ISREG(st.st_mode)) {
558		if (st.st_size >= (1U << 31)) {
559			fprintf(stderr, "%s: input too large\n", file);
560			return(0);
561		}
562		*with_mmap = 1;
563		fb->sz = (size_t)st.st_size;
564		fb->buf = mmap(NULL, fb->sz, PROT_READ,
565				MAP_FILE|MAP_SHARED, fd, 0);
566		if (fb->buf != MAP_FAILED)
567			return(1);
568	}
569
570	/*
571	 * If this isn't a regular file (like, say, stdin), then we must
572	 * go the old way and just read things in bit by bit.
573	 */
574
575	*with_mmap = 0;
576	off = 0;
577	fb->sz = 0;
578	fb->buf = NULL;
579	for (;;) {
580		if (off == fb->sz) {
581			if (fb->sz == (1U << 31)) {
582				fprintf(stderr, "%s: input too large\n", file);
583				break;
584			}
585			resize_buf(fb, 65536);
586		}
587		ssz = read(fd, fb->buf + (int)off, fb->sz - off);
588		if (ssz == 0) {
589			fb->sz = off;
590			return(1);
591		}
592		if (ssz == -1) {
593			perror(file);
594			break;
595		}
596		off += (size_t)ssz;
597	}
598
599	free(fb->buf);
600	fb->buf = NULL;
601	return(0);
602}
603
604static void
605mparse_end(struct mparse *curp)
606{
607
608	if (MANDOCLEVEL_FATAL <= curp->file_status)
609		return;
610
611	if (curp->mdoc && ! mdoc_endparse(curp->mdoc)) {
612		assert(MANDOCLEVEL_FATAL <= curp->file_status);
613		return;
614	}
615
616	if (curp->man && ! man_endparse(curp->man)) {
617		assert(MANDOCLEVEL_FATAL <= curp->file_status);
618		return;
619	}
620
621	if ( ! (curp->man || curp->mdoc)) {
622		mandoc_msg(MANDOCERR_NOTMANUAL, curp, 1, 0, NULL);
623		curp->file_status = MANDOCLEVEL_FATAL;
624		return;
625	}
626
627	roff_endparse(curp->roff);
628}
629
630static void
631mparse_readfd_r(struct mparse *curp, int fd, const char *file, int re)
632{
633	const char	*svfile;
634
635	if (-1 == fd)
636		if (-1 == (fd = open(file, O_RDONLY, 0))) {
637			perror(file);
638			curp->file_status = MANDOCLEVEL_SYSERR;
639			return;
640		}
641
642	svfile = curp->file;
643	curp->file = file;
644
645	pdesc(curp, file, fd);
646
647	if (0 == re && MANDOCLEVEL_FATAL > curp->file_status)
648		mparse_end(curp);
649
650	if (STDIN_FILENO != fd && -1 == close(fd))
651		perror(file);
652
653	curp->file = svfile;
654}
655
656enum mandoclevel
657mparse_readfd(struct mparse *curp, int fd, const char *file)
658{
659
660	mparse_readfd_r(curp, fd, file, 0);
661	return(curp->file_status);
662}
663
664struct mparse *
665mparse_alloc(enum mparset inttype, enum mandoclevel wlevel, mandocmsg mmsg, void *arg)
666{
667	struct mparse	*curp;
668
669	assert(wlevel <= MANDOCLEVEL_FATAL);
670
671	curp = mandoc_calloc(1, sizeof(struct mparse));
672
673	curp->wlevel = wlevel;
674	curp->mmsg = mmsg;
675	curp->arg = arg;
676	curp->inttype = inttype;
677
678	curp->roff = roff_alloc(&curp->regs, curp);
679	return(curp);
680}
681
682void
683mparse_reset(struct mparse *curp)
684{
685
686	memset(&curp->regs, 0, sizeof(struct regset));
687
688	roff_reset(curp->roff);
689
690	if (curp->mdoc)
691		mdoc_reset(curp->mdoc);
692	if (curp->man)
693		man_reset(curp->man);
694
695	curp->file_status = MANDOCLEVEL_OK;
696	curp->mdoc = NULL;
697	curp->man = NULL;
698}
699
700void
701mparse_free(struct mparse *curp)
702{
703
704	if (curp->pmdoc)
705		mdoc_free(curp->pmdoc);
706	if (curp->pman)
707		man_free(curp->pman);
708	if (curp->roff)
709		roff_free(curp->roff);
710
711	free(curp);
712}
713
714void
715mparse_result(struct mparse *curp, struct mdoc **mdoc, struct man **man)
716{
717
718	if (mdoc)
719		*mdoc = curp->mdoc;
720	if (man)
721		*man = curp->man;
722}
723
724void
725mandoc_vmsg(enum mandocerr t, struct mparse *m,
726		int ln, int pos, const char *fmt, ...)
727{
728	char		 buf[256];
729	va_list		 ap;
730
731	va_start(ap, fmt);
732	vsnprintf(buf, sizeof(buf) - 1, fmt, ap);
733	va_end(ap);
734
735	mandoc_msg(t, m, ln, pos, buf);
736}
737
738void
739mandoc_msg(enum mandocerr er, struct mparse *m,
740		int ln, int col, const char *msg)
741{
742	enum mandoclevel level;
743
744	level = MANDOCLEVEL_FATAL;
745	while (er < mandoclimits[level])
746		level--;
747
748	if (level < m->wlevel)
749		return;
750
751	if (m->mmsg)
752		(*m->mmsg)(er, level, m->file, ln, col, msg);
753
754	if (m->file_status < level)
755		m->file_status = level;
756}
757
758const char *
759mparse_strerror(enum mandocerr er)
760{
761
762	return(mandocerrs[er]);
763}
764
765const char *
766mparse_strlevel(enum mandoclevel lvl)
767{
768	return(mandoclevels[lvl]);
769}
770