xref: /illumos-gate/usr/src/cmd/mandoc/mandoc.c (revision 371584c2eae4cf827fd406ba26c14f021adaaa70)
1 /*	$Id: mandoc.c,v 1.98 2015/11/12 22:44:27 schwarze Exp $ */
2 /*
3  * Copyright (c) 2008-2011, 2014 Kristaps Dzonsons <kristaps@bsd.lv>
4  * Copyright (c) 2011-2015 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 "config.h"
19 
20 #include <sys/types.h>
21 
22 #include <assert.h>
23 #include <ctype.h>
24 #include <errno.h>
25 #include <limits.h>
26 #include <stdlib.h>
27 #include <stdio.h>
28 #include <string.h>
29 #include <time.h>
30 
31 #include "mandoc.h"
32 #include "mandoc_aux.h"
33 #include "libmandoc.h"
34 
35 static	int	 a2time(time_t *, const char *, const char *);
36 static	char	*time2a(time_t);
37 
38 
39 enum mandoc_esc
40 mandoc_escape(const char **end, const char **start, int *sz)
41 {
42 	const char	*local_start;
43 	int		 local_sz;
44 	char		 term;
45 	enum mandoc_esc	 gly;
46 
47 	/*
48 	 * When the caller doesn't provide return storage,
49 	 * use local storage.
50 	 */
51 
52 	if (NULL == start)
53 		start = &local_start;
54 	if (NULL == sz)
55 		sz = &local_sz;
56 
57 	/*
58 	 * Beyond the backslash, at least one input character
59 	 * is part of the escape sequence.  With one exception
60 	 * (see below), that character won't be returned.
61 	 */
62 
63 	gly = ESCAPE_ERROR;
64 	*start = ++*end;
65 	*sz = 0;
66 	term = '\0';
67 
68 	switch ((*start)[-1]) {
69 	/*
70 	 * First the glyphs.  There are several different forms of
71 	 * these, but each eventually returns a substring of the glyph
72 	 * name.
73 	 */
74 	case '(':
75 		gly = ESCAPE_SPECIAL;
76 		*sz = 2;
77 		break;
78 	case '[':
79 		gly = ESCAPE_SPECIAL;
80 		term = ']';
81 		break;
82 	case 'C':
83 		if ('\'' != **start)
84 			return ESCAPE_ERROR;
85 		*start = ++*end;
86 		gly = ESCAPE_SPECIAL;
87 		term = '\'';
88 		break;
89 
90 	/*
91 	 * Escapes taking no arguments at all.
92 	 */
93 	case 'd':
94 	case 'u':
95 	case ',':
96 	case '/':
97 		return ESCAPE_IGNORE;
98 
99 	/*
100 	 * The \z escape is supposed to output the following
101 	 * character without advancing the cursor position.
102 	 * Since we are mostly dealing with terminal mode,
103 	 * let us just skip the next character.
104 	 */
105 	case 'z':
106 		return ESCAPE_SKIPCHAR;
107 
108 	/*
109 	 * Handle all triggers matching \X(xy, \Xx, and \X[xxxx], where
110 	 * 'X' is the trigger.  These have opaque sub-strings.
111 	 */
112 	case 'F':
113 	case 'g':
114 	case 'k':
115 	case 'M':
116 	case 'm':
117 	case 'n':
118 	case 'V':
119 	case 'Y':
120 		gly = ESCAPE_IGNORE;
121 		/* FALLTHROUGH */
122 	case 'f':
123 		if (ESCAPE_ERROR == gly)
124 			gly = ESCAPE_FONT;
125 		switch (**start) {
126 		case '(':
127 			*start = ++*end;
128 			*sz = 2;
129 			break;
130 		case '[':
131 			*start = ++*end;
132 			term = ']';
133 			break;
134 		default:
135 			*sz = 1;
136 			break;
137 		}
138 		break;
139 
140 	/*
141 	 * These escapes are of the form \X'Y', where 'X' is the trigger
142 	 * and 'Y' is any string.  These have opaque sub-strings.
143 	 * The \B and \w escapes are handled in roff.c, roff_res().
144 	 */
145 	case 'A':
146 	case 'b':
147 	case 'D':
148 	case 'R':
149 	case 'X':
150 	case 'Z':
151 		gly = ESCAPE_IGNORE;
152 		/* FALLTHROUGH */
153 	case 'o':
154 		if (**start == '\0')
155 			return ESCAPE_ERROR;
156 		if (gly == ESCAPE_ERROR)
157 			gly = ESCAPE_OVERSTRIKE;
158 		term = **start;
159 		*start = ++*end;
160 		break;
161 
162 	/*
163 	 * These escapes are of the form \X'N', where 'X' is the trigger
164 	 * and 'N' resolves to a numerical expression.
165 	 */
166 	case 'h':
167 	case 'H':
168 	case 'L':
169 	case 'l':
170 	case 'S':
171 	case 'v':
172 	case 'x':
173 		if (strchr(" %&()*+-./0123456789:<=>", **start)) {
174 			if ('\0' != **start)
175 				++*end;
176 			return ESCAPE_ERROR;
177 		}
178 		gly = ESCAPE_IGNORE;
179 		term = **start;
180 		*start = ++*end;
181 		break;
182 
183 	/*
184 	 * Special handling for the numbered character escape.
185 	 * XXX Do any other escapes need similar handling?
186 	 */
187 	case 'N':
188 		if ('\0' == **start)
189 			return ESCAPE_ERROR;
190 		(*end)++;
191 		if (isdigit((unsigned char)**start)) {
192 			*sz = 1;
193 			return ESCAPE_IGNORE;
194 		}
195 		(*start)++;
196 		while (isdigit((unsigned char)**end))
197 			(*end)++;
198 		*sz = *end - *start;
199 		if ('\0' != **end)
200 			(*end)++;
201 		return ESCAPE_NUMBERED;
202 
203 	/*
204 	 * Sizes get a special category of their own.
205 	 */
206 	case 's':
207 		gly = ESCAPE_IGNORE;
208 
209 		/* See +/- counts as a sign. */
210 		if ('+' == **end || '-' == **end || ASCII_HYPH == **end)
211 			*start = ++*end;
212 
213 		switch (**end) {
214 		case '(':
215 			*start = ++*end;
216 			*sz = 2;
217 			break;
218 		case '[':
219 			*start = ++*end;
220 			term = ']';
221 			break;
222 		case '\'':
223 			*start = ++*end;
224 			term = '\'';
225 			break;
226 		case '3':
227 		case '2':
228 		case '1':
229 			*sz = (*end)[-1] == 's' &&
230 			    isdigit((unsigned char)(*end)[1]) ? 2 : 1;
231 			break;
232 		default:
233 			*sz = 1;
234 			break;
235 		}
236 
237 		break;
238 
239 	/*
240 	 * Anything else is assumed to be a glyph.
241 	 * In this case, pass back the character after the backslash.
242 	 */
243 	default:
244 		gly = ESCAPE_SPECIAL;
245 		*start = --*end;
246 		*sz = 1;
247 		break;
248 	}
249 
250 	assert(ESCAPE_ERROR != gly);
251 
252 	/*
253 	 * Read up to the terminating character,
254 	 * paying attention to nested escapes.
255 	 */
256 
257 	if ('\0' != term) {
258 		while (**end != term) {
259 			switch (**end) {
260 			case '\0':
261 				return ESCAPE_ERROR;
262 			case '\\':
263 				(*end)++;
264 				if (ESCAPE_ERROR ==
265 				    mandoc_escape(end, NULL, NULL))
266 					return ESCAPE_ERROR;
267 				break;
268 			default:
269 				(*end)++;
270 				break;
271 			}
272 		}
273 		*sz = (*end)++ - *start;
274 	} else {
275 		assert(*sz > 0);
276 		if ((size_t)*sz > strlen(*start))
277 			return ESCAPE_ERROR;
278 		*end += *sz;
279 	}
280 
281 	/* Run post-processors. */
282 
283 	switch (gly) {
284 	case ESCAPE_FONT:
285 		if (2 == *sz) {
286 			if ('C' == **start) {
287 				/*
288 				 * Treat constant-width font modes
289 				 * just like regular font modes.
290 				 */
291 				(*start)++;
292 				(*sz)--;
293 			} else {
294 				if ('B' == (*start)[0] && 'I' == (*start)[1])
295 					gly = ESCAPE_FONTBI;
296 				break;
297 			}
298 		} else if (1 != *sz)
299 			break;
300 
301 		switch (**start) {
302 		case '3':
303 		case 'B':
304 			gly = ESCAPE_FONTBOLD;
305 			break;
306 		case '2':
307 		case 'I':
308 			gly = ESCAPE_FONTITALIC;
309 			break;
310 		case 'P':
311 			gly = ESCAPE_FONTPREV;
312 			break;
313 		case '1':
314 		case 'R':
315 			gly = ESCAPE_FONTROMAN;
316 			break;
317 		}
318 		break;
319 	case ESCAPE_SPECIAL:
320 		if (1 == *sz && 'c' == **start)
321 			gly = ESCAPE_NOSPACE;
322 		/*
323 		 * Unicode escapes are defined in groff as \[u0000]
324 		 * to \[u10FFFF], where the contained value must be
325 		 * a valid Unicode codepoint.  Here, however, only
326 		 * check the length and range.
327 		 */
328 		if (**start != 'u' || *sz < 5 || *sz > 7)
329 			break;
330 		if (*sz == 7 && ((*start)[1] != '1' || (*start)[2] != '0'))
331 			break;
332 		if (*sz == 6 && (*start)[1] == '0')
333 			break;
334 		if (*sz == 5 && (*start)[1] == 'D' &&
335 		    strchr("89ABCDEF", (*start)[2]) != NULL)
336 			break;
337 		if ((int)strspn(*start + 1, "0123456789ABCDEFabcdef")
338 		    + 1 == *sz)
339 			gly = ESCAPE_UNICODE;
340 		break;
341 	default:
342 		break;
343 	}
344 
345 	return gly;
346 }
347 
348 /*
349  * Parse a quoted or unquoted roff-style request or macro argument.
350  * Return a pointer to the parsed argument, which is either the original
351  * pointer or advanced by one byte in case the argument is quoted.
352  * NUL-terminate the argument in place.
353  * Collapse pairs of quotes inside quoted arguments.
354  * Advance the argument pointer to the next argument,
355  * or to the NUL byte terminating the argument line.
356  */
357 char *
358 mandoc_getarg(struct mparse *parse, char **cpp, int ln, int *pos)
359 {
360 	char	 *start, *cp;
361 	int	  quoted, pairs, white;
362 
363 	/* Quoting can only start with a new word. */
364 	start = *cpp;
365 	quoted = 0;
366 	if ('"' == *start) {
367 		quoted = 1;
368 		start++;
369 	}
370 
371 	pairs = 0;
372 	white = 0;
373 	for (cp = start; '\0' != *cp; cp++) {
374 
375 		/*
376 		 * Move the following text left
377 		 * after quoted quotes and after "\\" and "\t".
378 		 */
379 		if (pairs)
380 			cp[-pairs] = cp[0];
381 
382 		if ('\\' == cp[0]) {
383 			/*
384 			 * In copy mode, translate double to single
385 			 * backslashes and backslash-t to literal tabs.
386 			 */
387 			switch (cp[1]) {
388 			case 't':
389 				cp[0] = '\t';
390 				/* FALLTHROUGH */
391 			case '\\':
392 				pairs++;
393 				cp++;
394 				break;
395 			case ' ':
396 				/* Skip escaped blanks. */
397 				if (0 == quoted)
398 					cp++;
399 				break;
400 			default:
401 				break;
402 			}
403 		} else if (0 == quoted) {
404 			if (' ' == cp[0]) {
405 				/* Unescaped blanks end unquoted args. */
406 				white = 1;
407 				break;
408 			}
409 		} else if ('"' == cp[0]) {
410 			if ('"' == cp[1]) {
411 				/* Quoted quotes collapse. */
412 				pairs++;
413 				cp++;
414 			} else {
415 				/* Unquoted quotes end quoted args. */
416 				quoted = 2;
417 				break;
418 			}
419 		}
420 	}
421 
422 	/* Quoted argument without a closing quote. */
423 	if (1 == quoted)
424 		mandoc_msg(MANDOCERR_ARG_QUOTE, parse, ln, *pos, NULL);
425 
426 	/* NUL-terminate this argument and move to the next one. */
427 	if (pairs)
428 		cp[-pairs] = '\0';
429 	if ('\0' != *cp) {
430 		*cp++ = '\0';
431 		while (' ' == *cp)
432 			cp++;
433 	}
434 	*pos += (int)(cp - start) + (quoted ? 1 : 0);
435 	*cpp = cp;
436 
437 	if ('\0' == *cp && (white || ' ' == cp[-1]))
438 		mandoc_msg(MANDOCERR_SPACE_EOL, parse, ln, *pos, NULL);
439 
440 	return start;
441 }
442 
443 static int
444 a2time(time_t *t, const char *fmt, const char *p)
445 {
446 	struct tm	 tm;
447 	char		*pp;
448 
449 	memset(&tm, 0, sizeof(struct tm));
450 
451 	pp = NULL;
452 #if HAVE_STRPTIME
453 	pp = strptime(p, fmt, &tm);
454 #endif
455 	if (NULL != pp && '\0' == *pp) {
456 		*t = mktime(&tm);
457 		return 1;
458 	}
459 
460 	return 0;
461 }
462 
463 static char *
464 time2a(time_t t)
465 {
466 	struct tm	*tm;
467 	char		*buf, *p;
468 	size_t		 ssz;
469 	int		 isz;
470 
471 	tm = localtime(&t);
472 	if (tm == NULL)
473 		return NULL;
474 
475 	/*
476 	 * Reserve space:
477 	 * up to 9 characters for the month (September) + blank
478 	 * up to 2 characters for the day + comma + blank
479 	 * 4 characters for the year and a terminating '\0'
480 	 */
481 
482 	p = buf = mandoc_malloc(10 + 4 + 4 + 1);
483 
484 	if ((ssz = strftime(p, 10 + 1, "%B ", tm)) == 0)
485 		goto fail;
486 	p += (int)ssz;
487 
488 	/*
489 	 * The output format is just "%d" here, not "%2d" or "%02d".
490 	 * That's also the reason why we can't just format the
491 	 * date as a whole with "%B %e, %Y" or "%B %d, %Y".
492 	 * Besides, the present approach is less prone to buffer
493 	 * overflows, in case anybody should ever introduce the bug
494 	 * of looking at LC_TIME.
495 	 */
496 
497 	if ((isz = snprintf(p, 4 + 1, "%d, ", tm->tm_mday)) == -1)
498 		goto fail;
499 	p += isz;
500 
501 	if (strftime(p, 4 + 1, "%Y", tm) == 0)
502 		goto fail;
503 	return buf;
504 
505 fail:
506 	free(buf);
507 	return NULL;
508 }
509 
510 char *
511 mandoc_normdate(struct mparse *parse, char *in, int ln, int pos)
512 {
513 	time_t		 t;
514 
515 	/* No date specified: use today's date. */
516 
517 	if (in == NULL || *in == '\0' || strcmp(in, "$" "Mdocdate$") == 0) {
518 		mandoc_msg(MANDOCERR_DATE_MISSING, parse, ln, pos, NULL);
519 		return time2a(time(NULL));
520 	}
521 
522 	/* Valid mdoc(7) date format. */
523 
524 	if (a2time(&t, "$" "Mdocdate: %b %d %Y $", in) ||
525 	    a2time(&t, "%b %d, %Y", in))
526 		return time2a(t);
527 
528 	/* Do not warn about the legacy man(7) format. */
529 
530 	if ( ! a2time(&t, "%Y-%m-%d", in))
531 		mandoc_msg(MANDOCERR_DATE_BAD, parse, ln, pos, in);
532 
533 	/* Use any non-mdoc(7) date verbatim. */
534 
535 	return mandoc_strdup(in);
536 }
537 
538 int
539 mandoc_eos(const char *p, size_t sz)
540 {
541 	const char	*q;
542 	int		 enclosed, found;
543 
544 	if (0 == sz)
545 		return 0;
546 
547 	/*
548 	 * End-of-sentence recognition must include situations where
549 	 * some symbols, such as `)', allow prior EOS punctuation to
550 	 * propagate outward.
551 	 */
552 
553 	enclosed = found = 0;
554 	for (q = p + (int)sz - 1; q >= p; q--) {
555 		switch (*q) {
556 		case '\"':
557 		case '\'':
558 		case ']':
559 		case ')':
560 			if (0 == found)
561 				enclosed = 1;
562 			break;
563 		case '.':
564 		case '!':
565 		case '?':
566 			found = 1;
567 			break;
568 		default:
569 			return found &&
570 			    (!enclosed || isalnum((unsigned char)*q));
571 		}
572 	}
573 
574 	return found && !enclosed;
575 }
576 
577 /*
578  * Convert a string to a long that may not be <0.
579  * If the string is invalid, or is less than 0, return -1.
580  */
581 int
582 mandoc_strntoi(const char *p, size_t sz, int base)
583 {
584 	char		 buf[32];
585 	char		*ep;
586 	long		 v;
587 
588 	if (sz > 31)
589 		return -1;
590 
591 	memcpy(buf, p, sz);
592 	buf[(int)sz] = '\0';
593 
594 	errno = 0;
595 	v = strtol(buf, &ep, base);
596 
597 	if (buf[0] == '\0' || *ep != '\0')
598 		return -1;
599 
600 	if (v > INT_MAX)
601 		v = INT_MAX;
602 	if (v < INT_MIN)
603 		v = INT_MIN;
604 
605 	return (int)v;
606 }
607