xref: /illumos-gate/usr/src/cmd/mandoc/mandoc.c (revision 95c635efb7c3b86efc493e0447eaec7aecca3f0f)
1 /*	$Id: mandoc.c,v 1.62 2011/12/03 16:08:51 schwarze Exp $ */
2 /*
3  * Copyright (c) 2008, 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4  * Copyright (c) 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 AUTHORS DISCLAIM ALL WARRANTIES
11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHORS 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/types.h>
23 
24 #include <assert.h>
25 #include <ctype.h>
26 #include <errno.h>
27 #include <limits.h>
28 #include <stdlib.h>
29 #include <stdio.h>
30 #include <string.h>
31 #include <time.h>
32 
33 #include "mandoc.h"
34 #include "libmandoc.h"
35 
36 #define DATESIZE 32
37 
38 static	int	 a2time(time_t *, const char *, const char *);
39 static	char	*time2a(time_t);
40 static	int	 numescape(const char *);
41 
42 /*
43  * Pass over recursive numerical expressions.  This context of this
44  * function is important: it's only called within character-terminating
45  * escapes (e.g., \s[xxxyyy]), so all we need to do is handle initial
46  * recursion: we don't care about what's in these blocks.
47  * This returns the number of characters skipped or -1 if an error
48  * occurs (the caller should bail).
49  */
50 static int
51 numescape(const char *start)
52 {
53 	int		 i;
54 	size_t		 sz;
55 	const char	*cp;
56 
57 	i = 0;
58 
59 	/* The expression consists of a subexpression. */
60 
61 	if ('\\' == start[i]) {
62 		cp = &start[++i];
63 		/*
64 		 * Read past the end of the subexpression.
65 		 * Bail immediately on errors.
66 		 */
67 		if (ESCAPE_ERROR == mandoc_escape(&cp, NULL, NULL))
68 			return(-1);
69 		return(i + cp - &start[i]);
70 	}
71 
72 	if ('(' != start[i++])
73 		return(0);
74 
75 	/*
76 	 * A parenthesised subexpression.  Read until the closing
77 	 * parenthesis, making sure to handle any nested subexpressions
78 	 * that might ruin our parse.
79 	 */
80 
81 	while (')' != start[i]) {
82 		sz = strcspn(&start[i], ")\\");
83 		i += (int)sz;
84 
85 		if ('\0' == start[i])
86 			return(-1);
87 		else if ('\\' != start[i])
88 			continue;
89 
90 		cp = &start[++i];
91 		if (ESCAPE_ERROR == mandoc_escape(&cp, NULL, NULL))
92 			return(-1);
93 		i += cp - &start[i];
94 	}
95 
96 	/* Read past the terminating ')'. */
97 	return(++i);
98 }
99 
100 enum mandoc_esc
101 mandoc_escape(const char **end, const char **start, int *sz)
102 {
103 	char		 c, term, numeric;
104 	int		 i, lim, ssz, rlim;
105 	const char	*cp, *rstart;
106 	enum mandoc_esc	 gly;
107 
108 	cp = *end;
109 	rstart = cp;
110 	if (start)
111 		*start = rstart;
112 	i = lim = 0;
113 	gly = ESCAPE_ERROR;
114 	term = numeric = '\0';
115 
116 	switch ((c = cp[i++])) {
117 	/*
118 	 * First the glyphs.  There are several different forms of
119 	 * these, but each eventually returns a substring of the glyph
120 	 * name.
121 	 */
122 	case ('('):
123 		gly = ESCAPE_SPECIAL;
124 		lim = 2;
125 		break;
126 	case ('['):
127 		gly = ESCAPE_SPECIAL;
128 		/*
129 		 * Unicode escapes are defined in groff as \[uXXXX] to
130 		 * \[u10FFFF], where the contained value must be a valid
131 		 * Unicode codepoint.  Here, however, only check whether
132 		 * it's not a zero-width escape.
133 		 */
134 		if ('u' == cp[i] && ']' != cp[i + 1])
135 			gly = ESCAPE_UNICODE;
136 		term = ']';
137 		break;
138 	case ('C'):
139 		if ('\'' != cp[i])
140 			return(ESCAPE_ERROR);
141 		gly = ESCAPE_SPECIAL;
142 		term = '\'';
143 		break;
144 
145 	/*
146 	 * Handle all triggers matching \X(xy, \Xx, and \X[xxxx], where
147 	 * 'X' is the trigger.  These have opaque sub-strings.
148 	 */
149 	case ('F'):
150 		/* FALLTHROUGH */
151 	case ('g'):
152 		/* FALLTHROUGH */
153 	case ('k'):
154 		/* FALLTHROUGH */
155 	case ('M'):
156 		/* FALLTHROUGH */
157 	case ('m'):
158 		/* FALLTHROUGH */
159 	case ('n'):
160 		/* FALLTHROUGH */
161 	case ('V'):
162 		/* FALLTHROUGH */
163 	case ('Y'):
164 		gly = ESCAPE_IGNORE;
165 		/* FALLTHROUGH */
166 	case ('f'):
167 		if (ESCAPE_ERROR == gly)
168 			gly = ESCAPE_FONT;
169 
170 		rstart= &cp[i];
171 		if (start)
172 			*start = rstart;
173 
174 		switch (cp[i++]) {
175 		case ('('):
176 			lim = 2;
177 			break;
178 		case ('['):
179 			term = ']';
180 			break;
181 		default:
182 			lim = 1;
183 			i--;
184 			break;
185 		}
186 		break;
187 
188 	/*
189 	 * These escapes are of the form \X'Y', where 'X' is the trigger
190 	 * and 'Y' is any string.  These have opaque sub-strings.
191 	 */
192 	case ('A'):
193 		/* FALLTHROUGH */
194 	case ('b'):
195 		/* FALLTHROUGH */
196 	case ('D'):
197 		/* FALLTHROUGH */
198 	case ('o'):
199 		/* FALLTHROUGH */
200 	case ('R'):
201 		/* FALLTHROUGH */
202 	case ('X'):
203 		/* FALLTHROUGH */
204 	case ('Z'):
205 		if ('\'' != cp[i++])
206 			return(ESCAPE_ERROR);
207 		gly = ESCAPE_IGNORE;
208 		term = '\'';
209 		break;
210 
211 	/*
212 	 * These escapes are of the form \X'N', where 'X' is the trigger
213 	 * and 'N' resolves to a numerical expression.
214 	 */
215 	case ('B'):
216 		/* FALLTHROUGH */
217 	case ('h'):
218 		/* FALLTHROUGH */
219 	case ('H'):
220 		/* FALLTHROUGH */
221 	case ('L'):
222 		/* FALLTHROUGH */
223 	case ('l'):
224 		gly = ESCAPE_NUMBERED;
225 		/* FALLTHROUGH */
226 	case ('S'):
227 		/* FALLTHROUGH */
228 	case ('v'):
229 		/* FALLTHROUGH */
230 	case ('w'):
231 		/* FALLTHROUGH */
232 	case ('x'):
233 		if (ESCAPE_ERROR == gly)
234 			gly = ESCAPE_IGNORE;
235 		if ('\'' != cp[i++])
236 			return(ESCAPE_ERROR);
237 		term = numeric = '\'';
238 		break;
239 
240 	/*
241 	 * Special handling for the numbered character escape.
242 	 * XXX Do any other escapes need similar handling?
243 	 */
244 	case ('N'):
245 		if ('\0' == cp[i])
246 			return(ESCAPE_ERROR);
247 		*end = &cp[++i];
248 		if (isdigit((unsigned char)cp[i-1]))
249 			return(ESCAPE_IGNORE);
250 		while (isdigit((unsigned char)**end))
251 			(*end)++;
252 		if (start)
253 			*start = &cp[i];
254 		if (sz)
255 			*sz = *end - &cp[i];
256 		if ('\0' != **end)
257 			(*end)++;
258 		return(ESCAPE_NUMBERED);
259 
260 	/*
261 	 * Sizes get a special category of their own.
262 	 */
263 	case ('s'):
264 		gly = ESCAPE_IGNORE;
265 
266 		rstart = &cp[i];
267 		if (start)
268 			*start = rstart;
269 
270 		/* See +/- counts as a sign. */
271 		c = cp[i];
272 		if ('+' == c || '-' == c || ASCII_HYPH == c)
273 			++i;
274 
275 		switch (cp[i++]) {
276 		case ('('):
277 			lim = 2;
278 			break;
279 		case ('['):
280 			term = numeric = ']';
281 			break;
282 		case ('\''):
283 			term = numeric = '\'';
284 			break;
285 		default:
286 			lim = 1;
287 			i--;
288 			break;
289 		}
290 
291 		/* See +/- counts as a sign. */
292 		c = cp[i];
293 		if ('+' == c || '-' == c || ASCII_HYPH == c)
294 			++i;
295 
296 		break;
297 
298 	/*
299 	 * Anything else is assumed to be a glyph.
300 	 */
301 	default:
302 		gly = ESCAPE_SPECIAL;
303 		lim = 1;
304 		i--;
305 		break;
306 	}
307 
308 	assert(ESCAPE_ERROR != gly);
309 
310 	rstart = &cp[i];
311 	if (start)
312 		*start = rstart;
313 
314 	/*
315 	 * If a terminating block has been specified, we need to
316 	 * handle the case of recursion, which could have their
317 	 * own terminating blocks that mess up our parse.  This, by the
318 	 * way, means that the "start" and "size" values will be
319 	 * effectively meaningless.
320 	 */
321 
322 	ssz = 0;
323 	if (numeric && -1 == (ssz = numescape(&cp[i])))
324 		return(ESCAPE_ERROR);
325 
326 	i += ssz;
327 	rlim = -1;
328 
329 	/*
330 	 * We have a character terminator.  Try to read up to that
331 	 * character.  If we can't (i.e., we hit the nil), then return
332 	 * an error; if we can, calculate our length, read past the
333 	 * terminating character, and exit.
334 	 */
335 
336 	if ('\0' != term) {
337 		*end = strchr(&cp[i], term);
338 		if ('\0' == *end)
339 			return(ESCAPE_ERROR);
340 
341 		rlim = *end - &cp[i];
342 		if (sz)
343 			*sz = rlim;
344 		(*end)++;
345 		goto out;
346 	}
347 
348 	assert(lim > 0);
349 
350 	/*
351 	 * We have a numeric limit.  If the string is shorter than that,
352 	 * stop and return an error.  Else adjust our endpoint, length,
353 	 * and return the current glyph.
354 	 */
355 
356 	if ((size_t)lim > strlen(&cp[i]))
357 		return(ESCAPE_ERROR);
358 
359 	rlim = lim;
360 	if (sz)
361 		*sz = rlim;
362 
363 	*end = &cp[i] + lim;
364 
365 out:
366 	assert(rlim >= 0 && rstart);
367 
368 	/* Run post-processors. */
369 
370 	switch (gly) {
371 	case (ESCAPE_FONT):
372 		/*
373 		 * Pretend that the constant-width font modes are the
374 		 * same as the regular font modes.
375 		 */
376 		if (2 == rlim && 'C' == *rstart)
377 			rstart++;
378 		else if (1 != rlim)
379 			break;
380 
381 		switch (*rstart) {
382 		case ('3'):
383 			/* FALLTHROUGH */
384 		case ('B'):
385 			gly = ESCAPE_FONTBOLD;
386 			break;
387 		case ('2'):
388 			/* FALLTHROUGH */
389 		case ('I'):
390 			gly = ESCAPE_FONTITALIC;
391 			break;
392 		case ('P'):
393 			gly = ESCAPE_FONTPREV;
394 			break;
395 		case ('1'):
396 			/* FALLTHROUGH */
397 		case ('R'):
398 			gly = ESCAPE_FONTROMAN;
399 			break;
400 		}
401 		break;
402 	case (ESCAPE_SPECIAL):
403 		if (1 != rlim)
404 			break;
405 		if ('c' == *rstart)
406 			gly = ESCAPE_NOSPACE;
407 		break;
408 	default:
409 		break;
410 	}
411 
412 	return(gly);
413 }
414 
415 void *
416 mandoc_calloc(size_t num, size_t size)
417 {
418 	void		*ptr;
419 
420 	ptr = calloc(num, size);
421 	if (NULL == ptr) {
422 		perror(NULL);
423 		exit((int)MANDOCLEVEL_SYSERR);
424 	}
425 
426 	return(ptr);
427 }
428 
429 
430 void *
431 mandoc_malloc(size_t size)
432 {
433 	void		*ptr;
434 
435 	ptr = malloc(size);
436 	if (NULL == ptr) {
437 		perror(NULL);
438 		exit((int)MANDOCLEVEL_SYSERR);
439 	}
440 
441 	return(ptr);
442 }
443 
444 
445 void *
446 mandoc_realloc(void *ptr, size_t size)
447 {
448 
449 	ptr = realloc(ptr, size);
450 	if (NULL == ptr) {
451 		perror(NULL);
452 		exit((int)MANDOCLEVEL_SYSERR);
453 	}
454 
455 	return(ptr);
456 }
457 
458 char *
459 mandoc_strndup(const char *ptr, size_t sz)
460 {
461 	char		*p;
462 
463 	p = mandoc_malloc(sz + 1);
464 	memcpy(p, ptr, sz);
465 	p[(int)sz] = '\0';
466 	return(p);
467 }
468 
469 char *
470 mandoc_strdup(const char *ptr)
471 {
472 	char		*p;
473 
474 	p = strdup(ptr);
475 	if (NULL == p) {
476 		perror(NULL);
477 		exit((int)MANDOCLEVEL_SYSERR);
478 	}
479 
480 	return(p);
481 }
482 
483 /*
484  * Parse a quoted or unquoted roff-style request or macro argument.
485  * Return a pointer to the parsed argument, which is either the original
486  * pointer or advanced by one byte in case the argument is quoted.
487  * Null-terminate the argument in place.
488  * Collapse pairs of quotes inside quoted arguments.
489  * Advance the argument pointer to the next argument,
490  * or to the null byte terminating the argument line.
491  */
492 char *
493 mandoc_getarg(struct mparse *parse, char **cpp, int ln, int *pos)
494 {
495 	char	 *start, *cp;
496 	int	  quoted, pairs, white;
497 
498 	/* Quoting can only start with a new word. */
499 	start = *cpp;
500 	quoted = 0;
501 	if ('"' == *start) {
502 		quoted = 1;
503 		start++;
504 	}
505 
506 	pairs = 0;
507 	white = 0;
508 	for (cp = start; '\0' != *cp; cp++) {
509 		/* Move left after quoted quotes and escaped backslashes. */
510 		if (pairs)
511 			cp[-pairs] = cp[0];
512 		if ('\\' == cp[0]) {
513 			if ('\\' == cp[1]) {
514 				/* Poor man's copy mode. */
515 				pairs++;
516 				cp++;
517 			} else if (0 == quoted && ' ' == cp[1])
518 				/* Skip escaped blanks. */
519 				cp++;
520 		} else if (0 == quoted) {
521 			if (' ' == cp[0]) {
522 				/* Unescaped blanks end unquoted args. */
523 				white = 1;
524 				break;
525 			}
526 		} else if ('"' == cp[0]) {
527 			if ('"' == cp[1]) {
528 				/* Quoted quotes collapse. */
529 				pairs++;
530 				cp++;
531 			} else {
532 				/* Unquoted quotes end quoted args. */
533 				quoted = 2;
534 				break;
535 			}
536 		}
537 	}
538 
539 	/* Quoted argument without a closing quote. */
540 	if (1 == quoted)
541 		mandoc_msg(MANDOCERR_BADQUOTE, parse, ln, *pos, NULL);
542 
543 	/* Null-terminate this argument and move to the next one. */
544 	if (pairs)
545 		cp[-pairs] = '\0';
546 	if ('\0' != *cp) {
547 		*cp++ = '\0';
548 		while (' ' == *cp)
549 			cp++;
550 	}
551 	*pos += (int)(cp - start) + (quoted ? 1 : 0);
552 	*cpp = cp;
553 
554 	if ('\0' == *cp && (white || ' ' == cp[-1]))
555 		mandoc_msg(MANDOCERR_EOLNSPACE, parse, ln, *pos, NULL);
556 
557 	return(start);
558 }
559 
560 static int
561 a2time(time_t *t, const char *fmt, const char *p)
562 {
563 	struct tm	 tm;
564 	char		*pp;
565 
566 	memset(&tm, 0, sizeof(struct tm));
567 
568 	pp = NULL;
569 #ifdef	HAVE_STRPTIME
570 	pp = strptime(p, fmt, &tm);
571 #endif
572 	if (NULL != pp && '\0' == *pp) {
573 		*t = mktime(&tm);
574 		return(1);
575 	}
576 
577 	return(0);
578 }
579 
580 static char *
581 time2a(time_t t)
582 {
583 	struct tm	*tm;
584 	char		*buf, *p;
585 	size_t		 ssz;
586 	int		 isz;
587 
588 	tm = localtime(&t);
589 
590 	/*
591 	 * Reserve space:
592 	 * up to 9 characters for the month (September) + blank
593 	 * up to 2 characters for the day + comma + blank
594 	 * 4 characters for the year and a terminating '\0'
595 	 */
596 	p = buf = mandoc_malloc(10 + 4 + 4 + 1);
597 
598 	if (0 == (ssz = strftime(p, 10 + 1, "%B ", tm)))
599 		goto fail;
600 	p += (int)ssz;
601 
602 	if (-1 == (isz = snprintf(p, 4 + 1, "%d, ", tm->tm_mday)))
603 		goto fail;
604 	p += isz;
605 
606 	if (0 == strftime(p, 4 + 1, "%Y", tm))
607 		goto fail;
608 	return(buf);
609 
610 fail:
611 	free(buf);
612 	return(NULL);
613 }
614 
615 char *
616 mandoc_normdate(struct mparse *parse, char *in, int ln, int pos)
617 {
618 	char		*out;
619 	time_t		 t;
620 
621 	if (NULL == in || '\0' == *in ||
622 	    0 == strcmp(in, "$" "Mdocdate$")) {
623 		mandoc_msg(MANDOCERR_NODATE, parse, ln, pos, NULL);
624 		time(&t);
625 	}
626 	else if (a2time(&t, "%Y-%m-%d", in))
627 		t = 0;
628 	else if (!a2time(&t, "$" "Mdocdate: %b %d %Y $", in) &&
629 	    !a2time(&t, "%b %d, %Y", in)) {
630 		mandoc_msg(MANDOCERR_BADDATE, parse, ln, pos, NULL);
631 		t = 0;
632 	}
633 	out = t ? time2a(t) : NULL;
634 	return(out ? out : mandoc_strdup(in));
635 }
636 
637 int
638 mandoc_eos(const char *p, size_t sz, int enclosed)
639 {
640 	const char *q;
641 	int found;
642 
643 	if (0 == sz)
644 		return(0);
645 
646 	/*
647 	 * End-of-sentence recognition must include situations where
648 	 * some symbols, such as `)', allow prior EOS punctuation to
649 	 * propagate outward.
650 	 */
651 
652 	found = 0;
653 	for (q = p + (int)sz - 1; q >= p; q--) {
654 		switch (*q) {
655 		case ('\"'):
656 			/* FALLTHROUGH */
657 		case ('\''):
658 			/* FALLTHROUGH */
659 		case (']'):
660 			/* FALLTHROUGH */
661 		case (')'):
662 			if (0 == found)
663 				enclosed = 1;
664 			break;
665 		case ('.'):
666 			/* FALLTHROUGH */
667 		case ('!'):
668 			/* FALLTHROUGH */
669 		case ('?'):
670 			found = 1;
671 			break;
672 		default:
673 			return(found && (!enclosed || isalnum((unsigned char)*q)));
674 		}
675 	}
676 
677 	return(found && !enclosed);
678 }
679 
680 /*
681  * Find out whether a line is a macro line or not.  If it is, adjust the
682  * current position and return one; if it isn't, return zero and don't
683  * change the current position.
684  */
685 int
686 mandoc_getcontrol(const char *cp, int *ppos)
687 {
688 	int		pos;
689 
690 	pos = *ppos;
691 
692 	if ('\\' == cp[pos] && '.' == cp[pos + 1])
693 		pos += 2;
694 	else if ('.' == cp[pos] || '\'' == cp[pos])
695 		pos++;
696 	else
697 		return(0);
698 
699 	while (' ' == cp[pos] || '\t' == cp[pos])
700 		pos++;
701 
702 	*ppos = pos;
703 	return(1);
704 }
705 
706 /*
707  * Convert a string to a long that may not be <0.
708  * If the string is invalid, or is less than 0, return -1.
709  */
710 int
711 mandoc_strntoi(const char *p, size_t sz, int base)
712 {
713 	char		 buf[32];
714 	char		*ep;
715 	long		 v;
716 
717 	if (sz > 31)
718 		return(-1);
719 
720 	memcpy(buf, p, sz);
721 	buf[(int)sz] = '\0';
722 
723 	errno = 0;
724 	v = strtol(buf, &ep, base);
725 
726 	if (buf[0] == '\0' || *ep != '\0')
727 		return(-1);
728 
729 	if (v > INT_MAX)
730 		v = INT_MAX;
731 	if (v < INT_MIN)
732 		v = INT_MIN;
733 
734 	return((int)v);
735 }
736