xref: /openbsd-src/usr.bin/mandoc/mandoc.c (revision d1df930ffab53da22f3324c32bed7ac5709915e6)
1 /*	$OpenBSD: mandoc.c,v 1.75 2018/08/20 18:06:42 schwarze Exp $ */
2 /*
3  * Copyright (c) 2008-2011, 2014 Kristaps Dzonsons <kristaps@bsd.lv>
4  * Copyright (c) 2011-2015, 2017, 2018 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 #include <sys/types.h>
19 
20 #include <assert.h>
21 #include <ctype.h>
22 #include <errno.h>
23 #include <limits.h>
24 #include <stdlib.h>
25 #include <stdio.h>
26 #include <string.h>
27 #include <time.h>
28 
29 #include "mandoc_aux.h"
30 #include "mandoc.h"
31 #include "roff.h"
32 #include "libmandoc.h"
33 
34 static	int	 a2time(time_t *, const char *, const char *);
35 static	char	*time2a(time_t);
36 
37 
38 enum mandoc_esc
39 mandoc_escape(const char **end, const char **start, int *sz)
40 {
41 	const char	*local_start;
42 	int		 local_sz, c, i;
43 	char		 term;
44 	enum mandoc_esc	 gly;
45 
46 	/*
47 	 * When the caller doesn't provide return storage,
48 	 * use local storage.
49 	 */
50 
51 	if (NULL == start)
52 		start = &local_start;
53 	if (NULL == sz)
54 		sz = &local_sz;
55 
56 	/*
57 	 * Beyond the backslash, at least one input character
58 	 * is part of the escape sequence.  With one exception
59 	 * (see below), that character won't be returned.
60 	 */
61 
62 	gly = ESCAPE_ERROR;
63 	*start = ++*end;
64 	*sz = 0;
65 	term = '\0';
66 
67 	switch ((*start)[-1]) {
68 	/*
69 	 * First the glyphs.  There are several different forms of
70 	 * these, but each eventually returns a substring of the glyph
71 	 * name.
72 	 */
73 	case '(':
74 		gly = ESCAPE_SPECIAL;
75 		*sz = 2;
76 		break;
77 	case '[':
78 		gly = ESCAPE_SPECIAL;
79 		term = ']';
80 		break;
81 	case 'C':
82 		if ('\'' != **start)
83 			return ESCAPE_ERROR;
84 		*start = ++*end;
85 		gly = ESCAPE_SPECIAL;
86 		term = '\'';
87 		break;
88 
89 	/*
90 	 * Escapes taking no arguments at all.
91 	 */
92 	case 'd':
93 	case 'u':
94 	case ',':
95 	case '/':
96 		return ESCAPE_IGNORE;
97 	case 'p':
98 		return ESCAPE_BREAK;
99 
100 	/*
101 	 * The \z escape is supposed to output the following
102 	 * character without advancing the cursor position.
103 	 * Since we are mostly dealing with terminal mode,
104 	 * let us just skip the next character.
105 	 */
106 	case 'z':
107 		return ESCAPE_SKIPCHAR;
108 
109 	/*
110 	 * Handle all triggers matching \X(xy, \Xx, and \X[xxxx], where
111 	 * 'X' is the trigger.  These have opaque sub-strings.
112 	 */
113 	case 'F':
114 	case 'g':
115 	case 'k':
116 	case 'M':
117 	case 'm':
118 	case 'n':
119 	case 'V':
120 	case 'Y':
121 		gly = ESCAPE_IGNORE;
122 		/* FALLTHROUGH */
123 	case 'f':
124 		if (ESCAPE_ERROR == gly)
125 			gly = ESCAPE_FONT;
126 		switch (**start) {
127 		case '(':
128 			*start = ++*end;
129 			*sz = 2;
130 			break;
131 		case '[':
132 			*start = ++*end;
133 			term = ']';
134 			break;
135 		default:
136 			*sz = 1;
137 			break;
138 		}
139 		break;
140 	case '*':
141 		if (strncmp(*start, "(.T", 3) != 0)
142 			abort();
143 		gly = ESCAPE_DEVICE;
144 		*start = ++*end;
145 		*sz = 2;
146 		break;
147 
148 	/*
149 	 * These escapes are of the form \X'Y', where 'X' is the trigger
150 	 * and 'Y' is any string.  These have opaque sub-strings.
151 	 * The \B and \w escapes are handled in roff.c, roff_res().
152 	 */
153 	case 'A':
154 	case 'b':
155 	case 'D':
156 	case 'R':
157 	case 'X':
158 	case 'Z':
159 		gly = ESCAPE_IGNORE;
160 		/* FALLTHROUGH */
161 	case 'o':
162 		if (**start == '\0')
163 			return ESCAPE_ERROR;
164 		if (gly == ESCAPE_ERROR)
165 			gly = ESCAPE_OVERSTRIKE;
166 		term = **start;
167 		*start = ++*end;
168 		break;
169 
170 	/*
171 	 * These escapes are of the form \X'N', where 'X' is the trigger
172 	 * and 'N' resolves to a numerical expression.
173 	 */
174 	case 'h':
175 	case 'H':
176 	case 'L':
177 	case 'l':
178 	case 'S':
179 	case 'v':
180 	case 'x':
181 		if (strchr(" %&()*+-./0123456789:<=>", **start)) {
182 			if ('\0' != **start)
183 				++*end;
184 			return ESCAPE_ERROR;
185 		}
186 		switch ((*start)[-1]) {
187 		case 'h':
188 			gly = ESCAPE_HORIZ;
189 			break;
190 		case 'l':
191 			gly = ESCAPE_HLINE;
192 			break;
193 		default:
194 			gly = ESCAPE_IGNORE;
195 			break;
196 		}
197 		term = **start;
198 		*start = ++*end;
199 		break;
200 
201 	/*
202 	 * Special handling for the numbered character escape.
203 	 * XXX Do any other escapes need similar handling?
204 	 */
205 	case 'N':
206 		if ('\0' == **start)
207 			return ESCAPE_ERROR;
208 		(*end)++;
209 		if (isdigit((unsigned char)**start)) {
210 			*sz = 1;
211 			return ESCAPE_IGNORE;
212 		}
213 		(*start)++;
214 		while (isdigit((unsigned char)**end))
215 			(*end)++;
216 		*sz = *end - *start;
217 		if ('\0' != **end)
218 			(*end)++;
219 		return ESCAPE_NUMBERED;
220 
221 	/*
222 	 * Sizes get a special category of their own.
223 	 */
224 	case 's':
225 		gly = ESCAPE_IGNORE;
226 
227 		/* See +/- counts as a sign. */
228 		if ('+' == **end || '-' == **end || ASCII_HYPH == **end)
229 			*start = ++*end;
230 
231 		switch (**end) {
232 		case '(':
233 			*start = ++*end;
234 			*sz = 2;
235 			break;
236 		case '[':
237 			*start = ++*end;
238 			term = ']';
239 			break;
240 		case '\'':
241 			*start = ++*end;
242 			term = '\'';
243 			break;
244 		case '3':
245 		case '2':
246 		case '1':
247 			*sz = (*end)[-1] == 's' &&
248 			    isdigit((unsigned char)(*end)[1]) ? 2 : 1;
249 			break;
250 		default:
251 			*sz = 1;
252 			break;
253 		}
254 
255 		break;
256 
257 	/*
258 	 * Anything else is assumed to be a glyph.
259 	 * In this case, pass back the character after the backslash.
260 	 */
261 	default:
262 		gly = ESCAPE_SPECIAL;
263 		*start = --*end;
264 		*sz = 1;
265 		break;
266 	}
267 
268 	assert(ESCAPE_ERROR != gly);
269 
270 	/*
271 	 * Read up to the terminating character,
272 	 * paying attention to nested escapes.
273 	 */
274 
275 	if ('\0' != term) {
276 		while (**end != term) {
277 			switch (**end) {
278 			case '\0':
279 				return ESCAPE_ERROR;
280 			case '\\':
281 				(*end)++;
282 				if (ESCAPE_ERROR ==
283 				    mandoc_escape(end, NULL, NULL))
284 					return ESCAPE_ERROR;
285 				break;
286 			default:
287 				(*end)++;
288 				break;
289 			}
290 		}
291 		*sz = (*end)++ - *start;
292 	} else {
293 		assert(*sz > 0);
294 		if ((size_t)*sz > strlen(*start))
295 			return ESCAPE_ERROR;
296 		*end += *sz;
297 	}
298 
299 	/* Run post-processors. */
300 
301 	switch (gly) {
302 	case ESCAPE_FONT:
303 		if (*sz == 2) {
304 			if (**start == 'C') {
305 				/*
306 				 * Treat constant-width font modes
307 				 * just like regular font modes.
308 				 */
309 				(*start)++;
310 				(*sz)--;
311 			} else {
312 				if ((*start)[0] == 'B' && (*start)[1] == 'I')
313 					gly = ESCAPE_FONTBI;
314 				break;
315 			}
316 		} else if (*sz != 1) {
317 			if (*sz == 0)
318 				gly = ESCAPE_FONTPREV;
319 			break;
320 		}
321 
322 		switch (**start) {
323 		case '3':
324 		case 'B':
325 			gly = ESCAPE_FONTBOLD;
326 			break;
327 		case '2':
328 		case 'I':
329 			gly = ESCAPE_FONTITALIC;
330 			break;
331 		case 'P':
332 			gly = ESCAPE_FONTPREV;
333 			break;
334 		case '1':
335 		case 'R':
336 			gly = ESCAPE_FONTROMAN;
337 			break;
338 		}
339 		break;
340 	case ESCAPE_SPECIAL:
341 		if (**start == 'c') {
342 			if (*sz == 1) {
343 				gly = ESCAPE_NOSPACE;
344 				break;
345 			}
346 			if (*sz < 6 || *sz > 7 ||
347 			    strncmp(*start, "char", 4) != 0 ||
348 			    (int)strspn(*start + 4, "0123456789") + 4 < *sz)
349 				break;
350 			c = 0;
351 			for (i = 4; i < *sz; i++)
352 				c = 10 * c + ((*start)[i] - '0');
353 			if (c < 0x21 || (c > 0x7e && c < 0xa0) || c > 0xff)
354 				break;
355 			*start += 4;
356 			*sz -= 4;
357 			gly = ESCAPE_NUMBERED;
358 			break;
359 		}
360 
361 		/*
362 		 * Unicode escapes are defined in groff as \[u0000]
363 		 * to \[u10FFFF], where the contained value must be
364 		 * a valid Unicode codepoint.  Here, however, only
365 		 * check the length and range.
366 		 */
367 		if (**start != 'u' || *sz < 5 || *sz > 7)
368 			break;
369 		if (*sz == 7 && ((*start)[1] != '1' || (*start)[2] != '0'))
370 			break;
371 		if (*sz == 6 && (*start)[1] == '0')
372 			break;
373 		if (*sz == 5 && (*start)[1] == 'D' &&
374 		    strchr("89ABCDEF", (*start)[2]) != NULL)
375 			break;
376 		if ((int)strspn(*start + 1, "0123456789ABCDEFabcdef")
377 		    + 1 == *sz)
378 			gly = ESCAPE_UNICODE;
379 		break;
380 	default:
381 		break;
382 	}
383 
384 	return gly;
385 }
386 
387 /*
388  * Parse a quoted or unquoted roff-style request or macro argument.
389  * Return a pointer to the parsed argument, which is either the original
390  * pointer or advanced by one byte in case the argument is quoted.
391  * NUL-terminate the argument in place.
392  * Collapse pairs of quotes inside quoted arguments.
393  * Advance the argument pointer to the next argument,
394  * or to the NUL byte terminating the argument line.
395  */
396 char *
397 mandoc_getarg(struct mparse *parse, char **cpp, int ln, int *pos)
398 {
399 	char	 *start, *cp;
400 	int	  quoted, pairs, white;
401 
402 	/* Quoting can only start with a new word. */
403 	start = *cpp;
404 	quoted = 0;
405 	if ('"' == *start) {
406 		quoted = 1;
407 		start++;
408 	}
409 
410 	pairs = 0;
411 	white = 0;
412 	for (cp = start; '\0' != *cp; cp++) {
413 
414 		/*
415 		 * Move the following text left
416 		 * after quoted quotes and after "\\" and "\t".
417 		 */
418 		if (pairs)
419 			cp[-pairs] = cp[0];
420 
421 		if ('\\' == cp[0]) {
422 			/*
423 			 * In copy mode, translate double to single
424 			 * backslashes and backslash-t to literal tabs.
425 			 */
426 			switch (cp[1]) {
427 			case 't':
428 				cp[0] = '\t';
429 				/* FALLTHROUGH */
430 			case '\\':
431 				pairs++;
432 				cp++;
433 				break;
434 			case ' ':
435 				/* Skip escaped blanks. */
436 				if (0 == quoted)
437 					cp++;
438 				break;
439 			default:
440 				break;
441 			}
442 		} else if (0 == quoted) {
443 			if (' ' == cp[0]) {
444 				/* Unescaped blanks end unquoted args. */
445 				white = 1;
446 				break;
447 			}
448 		} else if ('"' == cp[0]) {
449 			if ('"' == cp[1]) {
450 				/* Quoted quotes collapse. */
451 				pairs++;
452 				cp++;
453 			} else {
454 				/* Unquoted quotes end quoted args. */
455 				quoted = 2;
456 				break;
457 			}
458 		}
459 	}
460 
461 	/* Quoted argument without a closing quote. */
462 	if (1 == quoted)
463 		mandoc_msg(MANDOCERR_ARG_QUOTE, parse, ln, *pos, NULL);
464 
465 	/* NUL-terminate this argument and move to the next one. */
466 	if (pairs)
467 		cp[-pairs] = '\0';
468 	if ('\0' != *cp) {
469 		*cp++ = '\0';
470 		while (' ' == *cp)
471 			cp++;
472 	}
473 	*pos += (int)(cp - start) + (quoted ? 1 : 0);
474 	*cpp = cp;
475 
476 	if ('\0' == *cp && (white || ' ' == cp[-1]))
477 		mandoc_msg(MANDOCERR_SPACE_EOL, parse, ln, *pos, NULL);
478 
479 	return start;
480 }
481 
482 static int
483 a2time(time_t *t, const char *fmt, const char *p)
484 {
485 	struct tm	 tm;
486 	char		*pp;
487 
488 	memset(&tm, 0, sizeof(struct tm));
489 
490 	pp = strptime(p, fmt, &tm);
491 	if (NULL != pp && '\0' == *pp) {
492 		*t = mktime(&tm);
493 		return 1;
494 	}
495 
496 	return 0;
497 }
498 
499 static char *
500 time2a(time_t t)
501 {
502 	struct tm	*tm;
503 	char		*buf, *p;
504 	size_t		 ssz;
505 	int		 isz;
506 
507 	tm = localtime(&t);
508 	if (tm == NULL)
509 		return NULL;
510 
511 	/*
512 	 * Reserve space:
513 	 * up to 9 characters for the month (September) + blank
514 	 * up to 2 characters for the day + comma + blank
515 	 * 4 characters for the year and a terminating '\0'
516 	 */
517 
518 	p = buf = mandoc_malloc(10 + 4 + 4 + 1);
519 
520 	if ((ssz = strftime(p, 10 + 1, "%B ", tm)) == 0)
521 		goto fail;
522 	p += (int)ssz;
523 
524 	/*
525 	 * The output format is just "%d" here, not "%2d" or "%02d".
526 	 * That's also the reason why we can't just format the
527 	 * date as a whole with "%B %e, %Y" or "%B %d, %Y".
528 	 * Besides, the present approach is less prone to buffer
529 	 * overflows, in case anybody should ever introduce the bug
530 	 * of looking at LC_TIME.
531 	 */
532 
533 	if ((isz = snprintf(p, 4 + 1, "%d, ", tm->tm_mday)) == -1)
534 		goto fail;
535 	p += isz;
536 
537 	if (strftime(p, 4 + 1, "%Y", tm) == 0)
538 		goto fail;
539 	return buf;
540 
541 fail:
542 	free(buf);
543 	return NULL;
544 }
545 
546 char *
547 mandoc_normdate(struct roff_man *man, char *in, int ln, int pos)
548 {
549 	char		*cp;
550 	time_t		 t;
551 
552 	/* No date specified: use today's date. */
553 
554 	if (in == NULL || *in == '\0' || strcmp(in, "$" "Mdocdate$") == 0) {
555 		mandoc_msg(MANDOCERR_DATE_MISSING, man->parse, ln, pos, NULL);
556 		return time2a(time(NULL));
557 	}
558 
559 	/* Valid mdoc(7) date format. */
560 
561 	if (a2time(&t, "$" "Mdocdate: %b %d %Y $", in) ||
562 	    a2time(&t, "%b %d, %Y", in)) {
563 		cp = time2a(t);
564 		if (t > time(NULL) + 86400)
565 			mandoc_msg(MANDOCERR_DATE_FUTURE, man->parse,
566 			    ln, pos, cp);
567 		else if (*in != '$' && strcmp(in, cp) != 0)
568 			mandoc_msg(MANDOCERR_DATE_NORM, man->parse,
569 			    ln, pos, cp);
570 		return cp;
571 	}
572 
573 	/* In man(7), do not warn about the legacy format. */
574 
575 	if (a2time(&t, "%Y-%m-%d", in) == 0)
576 		mandoc_msg(MANDOCERR_DATE_BAD, man->parse, ln, pos, in);
577 	else if (t > time(NULL) + 86400)
578 		mandoc_msg(MANDOCERR_DATE_FUTURE, man->parse, ln, pos, in);
579 	else if (man->macroset == MACROSET_MDOC)
580 		mandoc_vmsg(MANDOCERR_DATE_LEGACY, man->parse,
581 		    ln, pos, "Dd %s", in);
582 
583 	/* Use any non-mdoc(7) date verbatim. */
584 
585 	return mandoc_strdup(in);
586 }
587 
588 int
589 mandoc_eos(const char *p, size_t sz)
590 {
591 	const char	*q;
592 	int		 enclosed, found;
593 
594 	if (0 == sz)
595 		return 0;
596 
597 	/*
598 	 * End-of-sentence recognition must include situations where
599 	 * some symbols, such as `)', allow prior EOS punctuation to
600 	 * propagate outward.
601 	 */
602 
603 	enclosed = found = 0;
604 	for (q = p + (int)sz - 1; q >= p; q--) {
605 		switch (*q) {
606 		case '\"':
607 		case '\'':
608 		case ']':
609 		case ')':
610 			if (0 == found)
611 				enclosed = 1;
612 			break;
613 		case '.':
614 		case '!':
615 		case '?':
616 			found = 1;
617 			break;
618 		default:
619 			return found &&
620 			    (!enclosed || isalnum((unsigned char)*q));
621 		}
622 	}
623 
624 	return found && !enclosed;
625 }
626 
627 /*
628  * Convert a string to a long that may not be <0.
629  * If the string is invalid, or is less than 0, return -1.
630  */
631 int
632 mandoc_strntoi(const char *p, size_t sz, int base)
633 {
634 	char		 buf[32];
635 	char		*ep;
636 	long		 v;
637 
638 	if (sz > 31)
639 		return -1;
640 
641 	memcpy(buf, p, sz);
642 	buf[(int)sz] = '\0';
643 
644 	errno = 0;
645 	v = strtol(buf, &ep, base);
646 
647 	if (buf[0] == '\0' || *ep != '\0')
648 		return -1;
649 
650 	if (v > INT_MAX)
651 		v = INT_MAX;
652 	if (v < INT_MIN)
653 		v = INT_MIN;
654 
655 	return (int)v;
656 }
657