xref: /titanic_50/usr/src/lib/libc/port/locale/regcomp.c (revision b515258426fed6c7311fd3f1dea697cfbd4085c6)
1 /*
2  * Copyright 2013 Garrett D'Amore <garrett@damore.org>
3  * Copyright 2010 Nexenta Systems, Inc.  All rights reserved.
4  * Copyright 2012 Milan Jurik. All rights reserved.
5  * Copyright (c) 1992, 1993, 1994 Henry Spencer.
6  * Copyright (c) 1992, 1993, 1994
7  *	The Regents of the University of California.  All rights reserved.
8  *
9  * This code is derived from software contributed to Berkeley by
10  * Henry Spencer.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  */
36 
37 #include "lint.h"
38 #include "file64.h"
39 #include <sys/types.h>
40 #include <stdio.h>
41 #include <string.h>
42 #include <ctype.h>
43 #include <limits.h>
44 #include <stdlib.h>
45 #include <regex.h>
46 #include <wchar.h>
47 #include <wctype.h>
48 
49 #include "runetype.h"
50 #include "collate.h"
51 
52 #include "utils.h"
53 #include "regex2.h"
54 
55 #include "cname.h"
56 #include "mblocal.h"
57 
58 /*
59  * parse structure, passed up and down to avoid global variables and
60  * other clumsinesses
61  */
62 struct parse {
63 	char *next;		/* next character in RE */
64 	char *end;		/* end of string (-> NUL normally) */
65 	int error;		/* has an error been seen? */
66 	sop *strip;		/* malloced strip */
67 	sopno ssize;		/* malloced strip size (allocated) */
68 	sopno slen;		/* malloced strip length (used) */
69 	int ncsalloc;		/* number of csets allocated */
70 	struct re_guts *g;
71 #define	NPAREN	10		/* we need to remember () 1-9 for back refs */
72 	sopno pbegin[NPAREN];	/* -> ( ([0] unused) */
73 	sopno pend[NPAREN];	/* -> ) ([0] unused) */
74 };
75 
76 /* ========= begin header generated by ./mkh ========= */
77 #ifdef __cplusplus
78 extern "C" {
79 #endif
80 
81 /* === regcomp.c === */
82 static void p_ere(struct parse *p, wint_t stop);
83 static void p_ere_exp(struct parse *p);
84 static void p_str(struct parse *p);
85 static void p_bre(struct parse *p, wint_t end1, wint_t end2);
86 static int p_simp_re(struct parse *p, int starordinary);
87 static int p_count(struct parse *p);
88 static void p_bracket(struct parse *p);
89 static void p_b_term(struct parse *p, cset *cs);
90 static void p_b_cclass(struct parse *p, cset *cs);
91 static void p_b_eclass(struct parse *p, cset *cs);
92 static wint_t p_b_symbol(struct parse *p);
93 static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
94 static wint_t othercase(wint_t ch);
95 static void bothcases(struct parse *p, wint_t ch);
96 static void ordinary(struct parse *p, wint_t ch);
97 static void nonnewline(struct parse *p);
98 static void repeat(struct parse *p, sopno start, int from, int to);
99 static int seterr(struct parse *p, int e);
100 static cset *allocset(struct parse *p);
101 static void freeset(struct parse *p, cset *cs);
102 static void CHadd(struct parse *p, cset *cs, wint_t ch);
103 static void CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max);
104 static void CHaddtype(struct parse *p, cset *cs, wctype_t wct);
105 static wint_t singleton(cset *cs);
106 static sopno dupl(struct parse *p, sopno start, sopno finish);
107 static void doemit(struct parse *p, sop op, size_t opnd);
108 static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
109 static void dofwd(struct parse *p, sopno pos, sop value);
110 static void enlarge(struct parse *p, sopno size);
111 static void stripsnug(struct parse *p, struct re_guts *g);
112 static void findmust(struct parse *p, struct re_guts *g);
113 static int altoffset(sop *scan, int offset);
114 static void computejumps(struct parse *p, struct re_guts *g);
115 static void computematchjumps(struct parse *p, struct re_guts *g);
116 static sopno pluscount(struct parse *p, struct re_guts *g);
117 static wint_t wgetnext(struct parse *p);
118 
119 #ifdef __cplusplus
120 }
121 #endif
122 /* ========= end header generated by ./mkh ========= */
123 
124 static char nuls[10];		/* place to point scanner in event of error */
125 
126 /*
127  * macros for use with parse structure
128  * BEWARE:  these know that the parse structure is named `p' !!!
129  */
130 #define	PEEK()	(*p->next)
131 #define	PEEK2()	(*(p->next+1))
132 #define	MORE()	(p->next < p->end)
133 #define	MORE2()	(p->next+1 < p->end)
134 #define	SEE(c)	(MORE() && PEEK() == (c))
135 #define	SEETWO(a, b)	(MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
136 #define	EAT(c)	((SEE(c)) ? (NEXT(), 1) : 0)
137 #define	EATTWO(a, b)	((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
138 #define	NEXT()	(p->next++)
139 #define	NEXT2()	(p->next += 2)
140 #define	NEXTn(n)	(p->next += (n))
141 #define	GETNEXT()	(*p->next++)
142 #define	WGETNEXT()	wgetnext(p)
143 #define	SETERROR(e)	((void)seterr(p, (e)))
144 #define	REQUIRE(co, e)	((co) || seterr(p, e))
145 #define	MUSTSEE(c, e)	(REQUIRE(MORE() && PEEK() == (c), e))
146 #define	MUSTEAT(c, e)	(REQUIRE(MORE() && GETNEXT() == (c), e))
147 #define	MUSTNOTSEE(c, e)	(REQUIRE(!MORE() || PEEK() != (c), e))
148 #define	EMIT(op, sopnd)	doemit(p, (sop)(op), (size_t)(sopnd))
149 #define	INSERT(op, pos)	doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
150 #define	AHEAD(pos)		dofwd(p, pos, HERE()-(pos))
151 #define	ASTERN(sop, pos)	EMIT(sop, HERE()-pos)
152 #define	HERE()		(p->slen)
153 #define	THERE()		(p->slen - 1)
154 #define	THERETHERE()	(p->slen - 2)
155 #define	DROP(n)	(p->slen -= (n))
156 
157 #ifndef NDEBUG
158 static int never = 0;		/* for use in asserts; shuts lint up */
159 #else
160 #define	never	0		/* some <assert.h>s have bugs too */
161 #endif
162 
163 /*
164  * regcomp - interface for parser and compilation
165  */
166 int				/* 0 success, otherwise REG_something */
167 regcomp(regex_t *_RESTRICT_KYWD preg,
168 	const char *_RESTRICT_KYWD pattern,
169 	int cflags)
170 {
171 	struct parse pa;
172 	struct re_guts *g;
173 	struct parse *p = &pa;
174 	int i;
175 	size_t len;
176 #ifdef REDEBUG
177 #define	GOODFLAGS(f)	(f)
178 #else
179 #define	GOODFLAGS(f)	((f)&~REG_DUMP)
180 #endif
181 
182 	/* We had REG_INVARG, but we don't have that on Solaris. */
183 	cflags = GOODFLAGS(cflags);
184 	if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
185 		return (REG_EFATAL);
186 
187 	if (cflags&REG_PEND) {
188 		if (preg->re_endp < pattern)
189 			return (REG_EFATAL);
190 		len = preg->re_endp - pattern;
191 	} else
192 		len = strlen((char *)pattern);
193 
194 	/* do the mallocs early so failure handling is easy */
195 	g = (struct re_guts *)malloc(sizeof (struct re_guts));
196 	if (g == NULL)
197 		return (REG_ESPACE);
198 	p->ssize = len/(size_t)2*(size_t)3 + (size_t)1;	/* ugh */
199 	p->strip = (sop *)malloc(p->ssize * sizeof (sop));
200 	p->slen = 0;
201 	if (p->strip == NULL) {
202 		free((char *)g);
203 		return (REG_ESPACE);
204 	}
205 
206 	/* set things up */
207 	p->g = g;
208 	p->next = (char *)pattern;	/* convenience; we do not modify it */
209 	p->end = p->next + len;
210 	p->error = 0;
211 	p->ncsalloc = 0;
212 	for (i = 0; i < NPAREN; i++) {
213 		p->pbegin[i] = 0;
214 		p->pend[i] = 0;
215 	}
216 	g->sets = NULL;
217 	g->ncsets = 0;
218 	g->cflags = cflags;
219 	g->iflags = 0;
220 	g->nbol = 0;
221 	g->neol = 0;
222 	g->must = NULL;
223 	g->moffset = -1;
224 	g->charjump = NULL;
225 	g->matchjump = NULL;
226 	g->mlen = 0;
227 	g->nsub = 0;
228 	g->backrefs = 0;
229 
230 	/* do it */
231 	EMIT(OEND, 0);
232 	g->firststate = THERE();
233 	if (cflags&REG_EXTENDED)
234 		p_ere(p, OUT);
235 	else if (cflags&REG_NOSPEC)
236 		p_str(p);
237 	else
238 		p_bre(p, OUT, OUT);
239 	EMIT(OEND, 0);
240 	g->laststate = THERE();
241 
242 	/* tidy up loose ends and fill things in */
243 	stripsnug(p, g);
244 	findmust(p, g);
245 	/*
246 	 * only use Boyer-Moore algorithm if the pattern is bigger
247 	 * than three characters
248 	 */
249 	if (g->mlen > 3) {
250 		computejumps(p, g);
251 		computematchjumps(p, g);
252 		if (g->matchjump == NULL && g->charjump != NULL) {
253 			free(g->charjump);
254 			g->charjump = NULL;
255 		}
256 	}
257 	g->nplus = pluscount(p, g);
258 	g->magic = MAGIC2;
259 	preg->re_nsub = g->nsub;
260 	preg->re_g = g;
261 	preg->re_magic = MAGIC1;
262 #ifndef REDEBUG
263 	/* not debugging, so can't rely on the assert() in regexec() */
264 	if (g->iflags&BAD)
265 		SETERROR(REG_EFATAL);
266 #endif
267 
268 	/* win or lose, we're done */
269 	if (p->error != 0)	/* lose */
270 		regfree(preg);
271 	return (p->error);
272 }
273 
274 /*
275  * p_ere - ERE parser top level, concatenation and alternation
276  */
277 static void
278 p_ere(struct parse *p,
279     wint_t stop)		/* character this ERE should end at */
280 {
281 	char c;
282 	sopno prevback;
283 	sopno prevfwd;
284 	sopno conc;
285 	int first = 1;		/* is this the first alternative? */
286 
287 	for (;;) {
288 		/* do a bunch of concatenated expressions */
289 		conc = HERE();
290 		while (MORE() && (c = PEEK()) != '|' && c != stop)
291 			p_ere_exp(p);
292 		/* require nonempty */
293 		(void) REQUIRE(HERE() != conc, REG_BADPAT);
294 
295 		if (!EAT('|'))
296 			break;		/* NOTE BREAK OUT */
297 
298 		if (first) {
299 			INSERT(OCH_, conc);	/* offset is wrong */
300 			prevfwd = conc;
301 			prevback = conc;
302 			first = 0;
303 		}
304 		ASTERN(OOR1, prevback);
305 		prevback = THERE();
306 		AHEAD(prevfwd);			/* fix previous offset */
307 		prevfwd = HERE();
308 		EMIT(OOR2, 0);			/* offset is very wrong */
309 	}
310 
311 	if (!first) {		/* tail-end fixups */
312 		AHEAD(prevfwd);
313 		ASTERN(O_CH, prevback);
314 	}
315 
316 	assert(!MORE() || SEE(stop));
317 }
318 
319 /*
320  * p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
321  */
322 static void
323 p_ere_exp(struct parse *p)
324 {
325 	char c;
326 	wint_t wc;
327 	sopno pos;
328 	int count;
329 	int count2;
330 	sopno subno;
331 	int wascaret = 0;
332 
333 	assert(MORE());		/* caller should have ensured this */
334 	c = GETNEXT();
335 
336 	pos = HERE();
337 	switch (c) {
338 	case '(':
339 		(void) REQUIRE(MORE(), REG_EPAREN);
340 		p->g->nsub++;
341 		subno = p->g->nsub;
342 		if (subno < NPAREN)
343 			p->pbegin[subno] = HERE();
344 		EMIT(OLPAREN, subno);
345 		if (!SEE(')'))
346 			p_ere(p, ')');
347 		if (subno < NPAREN) {
348 			p->pend[subno] = HERE();
349 			assert(p->pend[subno] != 0);
350 		}
351 		EMIT(ORPAREN, subno);
352 		(void) MUSTEAT(')', REG_EPAREN);
353 		break;
354 #ifndef POSIX_MISTAKE
355 	case ')':		/* happens only if no current unmatched ( */
356 		/*
357 		 * You may ask, why the ifndef?  Because I didn't notice
358 		 * this until slightly too late for 1003.2, and none of the
359 		 * other 1003.2 regular-expression reviewers noticed it at
360 		 * all.  So an unmatched ) is legal POSIX, at least until
361 		 * we can get it fixed.
362 		 */
363 		SETERROR(REG_EPAREN);
364 		break;
365 #endif
366 	case '^':
367 		EMIT(OBOL, 0);
368 		p->g->iflags |= USEBOL;
369 		p->g->nbol++;
370 		wascaret = 1;
371 		break;
372 	case '$':
373 		EMIT(OEOL, 0);
374 		p->g->iflags |= USEEOL;
375 		p->g->neol++;
376 		break;
377 	case '|':
378 		SETERROR(REG_BADPAT);
379 		break;
380 	case '*':
381 	case '+':
382 	case '?':
383 		SETERROR(REG_BADRPT);
384 		break;
385 	case '.':
386 		if (p->g->cflags&REG_NEWLINE)
387 			nonnewline(p);
388 		else
389 			EMIT(OANY, 0);
390 		break;
391 	case '[':
392 		p_bracket(p);
393 		break;
394 	case '\\':
395 		(void) REQUIRE(MORE(), REG_EESCAPE);
396 		wc = WGETNEXT();
397 		switch (wc) {
398 		case '<':
399 			EMIT(OBOW, 0);
400 			break;
401 		case '>':
402 			EMIT(OEOW, 0);
403 			break;
404 		default:
405 			ordinary(p, wc);
406 			break;
407 		}
408 		break;
409 	case '{':		/* okay as ordinary except if digit follows */
410 		(void) REQUIRE(!MORE() || !isdigit((uch)PEEK()), REG_BADRPT);
411 		/* FALLTHROUGH */
412 	default:
413 		p->next--;
414 		wc = WGETNEXT();
415 		ordinary(p, wc);
416 		break;
417 	}
418 
419 	if (!MORE())
420 		return;
421 	c = PEEK();
422 	/* we call { a repetition if followed by a digit */
423 	if (!(c == '*' || c == '+' || c == '?' ||
424 	    (c == '{' && MORE2() && isdigit((uch)PEEK2()))))
425 		return;		/* no repetition, we're done */
426 	NEXT();
427 
428 	(void) REQUIRE(!wascaret, REG_BADRPT);
429 	switch (c) {
430 	case '*':	/* implemented as +? */
431 		/* this case does not require the (y|) trick, noKLUDGE */
432 		INSERT(OPLUS_, pos);
433 		ASTERN(O_PLUS, pos);
434 		INSERT(OQUEST_, pos);
435 		ASTERN(O_QUEST, pos);
436 		break;
437 	case '+':
438 		INSERT(OPLUS_, pos);
439 		ASTERN(O_PLUS, pos);
440 		break;
441 	case '?':
442 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
443 		INSERT(OCH_, pos);		/* offset slightly wrong */
444 		ASTERN(OOR1, pos);		/* this one's right */
445 		AHEAD(pos);			/* fix the OCH_ */
446 		EMIT(OOR2, 0);			/* offset very wrong... */
447 		AHEAD(THERE());			/* ...so fix it */
448 		ASTERN(O_CH, THERETHERE());
449 		break;
450 	case '{':
451 		count = p_count(p);
452 		if (EAT(',')) {
453 			if (isdigit((uch)PEEK())) {
454 				count2 = p_count(p);
455 				(void) REQUIRE(count <= count2, REG_BADBR);
456 			} else		/* single number with comma */
457 				count2 = INFINITY;
458 		} else		/* just a single number */
459 			count2 = count;
460 		repeat(p, pos, count, count2);
461 		if (!EAT('}')) {	/* error heuristics */
462 			while (MORE() && PEEK() != '}')
463 				NEXT();
464 			(void) REQUIRE(MORE(), REG_EBRACE);
465 			SETERROR(REG_BADBR);
466 		}
467 		break;
468 	}
469 
470 	if (!MORE())
471 		return;
472 	c = PEEK();
473 	if (!(c == '*' || c == '+' || c == '?' ||
474 	    (c == '{' && MORE2() && isdigit((uch)PEEK2()))))
475 		return;
476 	SETERROR(REG_BADRPT);
477 }
478 
479 /*
480  * p_str - string (no metacharacters) "parser"
481  */
482 static void
483 p_str(struct parse *p)
484 {
485 	(void) REQUIRE(MORE(), REG_BADPAT);
486 	while (MORE())
487 		ordinary(p, WGETNEXT());
488 }
489 
490 /*
491  * p_bre - BRE parser top level, anchoring and concatenation
492  * Giving end1 as OUT essentially eliminates the end1/end2 check.
493  *
494  * This implementation is a bit of a kludge, in that a trailing $ is first
495  * taken as an ordinary character and then revised to be an anchor.
496  * The amount of lookahead needed to avoid this kludge is excessive.
497  */
498 static void
499 p_bre(struct parse *p,
500     wint_t end1,		/* first terminating character */
501     wint_t end2)		/* second terminating character */
502 {
503 	sopno start = HERE();
504 	int first = 1;			/* first subexpression? */
505 	int wasdollar = 0;
506 
507 	if (EAT('^')) {
508 		EMIT(OBOL, 0);
509 		p->g->iflags |= USEBOL;
510 		p->g->nbol++;
511 	}
512 	while (MORE() && !SEETWO(end1, end2)) {
513 		wasdollar = p_simp_re(p, first);
514 		first = 0;
515 	}
516 	if (wasdollar) {	/* oops, that was a trailing anchor */
517 		DROP(1);
518 		EMIT(OEOL, 0);
519 		p->g->iflags |= USEEOL;
520 		p->g->neol++;
521 	}
522 
523 	(void) REQUIRE(HERE() != start, REG_BADPAT);	/* require nonempty */
524 }
525 
526 /*
527  * p_simp_re - parse a simple RE, an atom possibly followed by a repetition
528  */
529 static int			/* was the simple RE an unbackslashed $? */
530 p_simp_re(struct parse *p,
531 	int starordinary)	/* is a leading * an ordinary character? */
532 {
533 	int c;
534 	int count;
535 	int count2;
536 	sopno pos;
537 	int i;
538 	wint_t wc;
539 	sopno subno;
540 #define	BACKSL	(1<<CHAR_BIT)
541 
542 	pos = HERE();		/* repetion op, if any, covers from here */
543 
544 	assert(MORE());		/* caller should have ensured this */
545 	c = GETNEXT();
546 	if (c == '\\') {
547 		(void) REQUIRE(MORE(), REG_EESCAPE);
548 		c = BACKSL | GETNEXT();
549 	}
550 	switch (c) {
551 	case '.':
552 		if (p->g->cflags&REG_NEWLINE)
553 			nonnewline(p);
554 		else
555 			EMIT(OANY, 0);
556 		break;
557 	case '[':
558 		p_bracket(p);
559 		break;
560 	case BACKSL|'<':
561 		EMIT(OBOW, 0);
562 		break;
563 	case BACKSL|'>':
564 		EMIT(OEOW, 0);
565 		break;
566 	case BACKSL|'{':
567 		SETERROR(REG_BADRPT);
568 		break;
569 	case BACKSL|'(':
570 		p->g->nsub++;
571 		subno = p->g->nsub;
572 		if (subno < NPAREN)
573 			p->pbegin[subno] = HERE();
574 		EMIT(OLPAREN, subno);
575 		/* the MORE here is an error heuristic */
576 		if (MORE() && !SEETWO('\\', ')'))
577 			p_bre(p, '\\', ')');
578 		if (subno < NPAREN) {
579 			p->pend[subno] = HERE();
580 			assert(p->pend[subno] != 0);
581 		}
582 		EMIT(ORPAREN, subno);
583 		(void) REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
584 		break;
585 	case BACKSL|')':	/* should not get here -- must be user */
586 	case BACKSL|'}':
587 		SETERROR(REG_EPAREN);
588 		break;
589 	case BACKSL|'1':
590 	case BACKSL|'2':
591 	case BACKSL|'3':
592 	case BACKSL|'4':
593 	case BACKSL|'5':
594 	case BACKSL|'6':
595 	case BACKSL|'7':
596 	case BACKSL|'8':
597 	case BACKSL|'9':
598 		i = (c&~BACKSL) - '0';
599 		assert(i < NPAREN);
600 		if (p->pend[i] != 0) {
601 			assert(i <= p->g->nsub);
602 			EMIT(OBACK_, i);
603 			assert(p->pbegin[i] != 0);
604 			assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
605 			assert(OP(p->strip[p->pend[i]]) == ORPAREN);
606 			(void) dupl(p, p->pbegin[i]+1, p->pend[i]);
607 			EMIT(O_BACK, i);
608 		} else
609 			SETERROR(REG_ESUBREG);
610 		p->g->backrefs = 1;
611 		break;
612 	case '*':
613 		(void) REQUIRE(starordinary, REG_BADRPT);
614 		/* FALLTHROUGH */
615 	default:
616 		p->next--;
617 		wc = WGETNEXT();
618 		ordinary(p, wc);
619 		break;
620 	}
621 
622 	if (EAT('*')) {		/* implemented as +? */
623 		/* this case does not require the (y|) trick, noKLUDGE */
624 		INSERT(OPLUS_, pos);
625 		ASTERN(O_PLUS, pos);
626 		INSERT(OQUEST_, pos);
627 		ASTERN(O_QUEST, pos);
628 	} else if (EATTWO('\\', '{')) {
629 		count = p_count(p);
630 		if (EAT(',')) {
631 			if (MORE() && isdigit((uch)PEEK())) {
632 				count2 = p_count(p);
633 				(void) REQUIRE(count <= count2, REG_BADBR);
634 			} else		/* single number with comma */
635 				count2 = INFINITY;
636 		} else		/* just a single number */
637 			count2 = count;
638 		repeat(p, pos, count, count2);
639 		if (!EATTWO('\\', '}')) {	/* error heuristics */
640 			while (MORE() && !SEETWO('\\', '}'))
641 				NEXT();
642 			(void) REQUIRE(MORE(), REG_EBRACE);
643 			SETERROR(REG_BADBR);
644 		}
645 	} else if (c == '$')	/* $ (but not \$) ends it */
646 		return (1);
647 
648 	return (0);
649 }
650 
651 /*
652  * p_count - parse a repetition count
653  */
654 static int			/* the value */
655 p_count(struct parse *p)
656 {
657 	int count = 0;
658 	int ndigits = 0;
659 
660 	while (MORE() && isdigit((uch)PEEK()) && count <= DUPMAX) {
661 		count = count*10 + (GETNEXT() - '0');
662 		ndigits++;
663 	}
664 
665 	(void) REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
666 	return (count);
667 }
668 
669 /*
670  * p_bracket - parse a bracketed character list
671  */
672 static void
673 p_bracket(struct parse *p)
674 {
675 	cset *cs;
676 	wint_t ch;
677 
678 	/* Dept of Truly Sickening Special-Case Kludges */
679 	if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
680 		EMIT(OBOW, 0);
681 		NEXTn(6);
682 		return;
683 	}
684 	if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
685 		EMIT(OEOW, 0);
686 		NEXTn(6);
687 		return;
688 	}
689 
690 	if ((cs = allocset(p)) == NULL)
691 		return;
692 
693 	if (p->g->cflags&REG_ICASE)
694 		cs->icase = 1;
695 	if (EAT('^'))
696 		cs->invert = 1;
697 	if (EAT(']'))
698 		CHadd(p, cs, ']');
699 	else if (EAT('-'))
700 		CHadd(p, cs, '-');
701 	while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
702 		p_b_term(p, cs);
703 	if (EAT('-'))
704 		CHadd(p, cs, '-');
705 	(void) MUSTEAT(']', REG_EBRACK);
706 
707 	if (p->error != 0)	/* don't mess things up further */
708 		return;
709 
710 	if (cs->invert && p->g->cflags&REG_NEWLINE)
711 		cs->bmp['\n' >> 3] |= 1 << ('\n' & 7);
712 
713 	if ((ch = singleton(cs)) != OUT) {	/* optimize singleton sets */
714 		ordinary(p, ch);
715 		freeset(p, cs);
716 	} else
717 		EMIT(OANYOF, (int)(cs - p->g->sets));
718 }
719 
720 /*
721  * p_b_term - parse one term of a bracketed character list
722  */
723 static void
724 p_b_term(struct parse *p, cset *cs)
725 {
726 	char c;
727 	wint_t start, finish;
728 	wint_t i;
729 	locale_t loc = uselocale(NULL);
730 
731 	/* classify what we've got */
732 	switch ((MORE()) ? PEEK() : '\0') {
733 	case '[':
734 		c = (MORE2()) ? PEEK2() : '\0';
735 		break;
736 	case '-':
737 		SETERROR(REG_ERANGE);
738 		return;			/* NOTE RETURN */
739 	default:
740 		c = '\0';
741 		break;
742 	}
743 
744 	switch (c) {
745 	case ':':		/* character class */
746 		NEXT2();
747 		(void) REQUIRE(MORE(), REG_EBRACK);
748 		c = PEEK();
749 		(void) REQUIRE(c != '-' && c != ']', REG_ECTYPE);
750 		p_b_cclass(p, cs);
751 		(void) REQUIRE(MORE(), REG_EBRACK);
752 		(void) REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
753 		break;
754 	case '=':		/* equivalence class */
755 		NEXT2();
756 		(void) REQUIRE(MORE(), REG_EBRACK);
757 		c = PEEK();
758 		(void) REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
759 		p_b_eclass(p, cs);
760 		(void) REQUIRE(MORE(), REG_EBRACK);
761 		(void) REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
762 		break;
763 	default:		/* symbol, ordinary character, or range */
764 		start = p_b_symbol(p);
765 		if (SEE('-') && MORE2() && PEEK2() != ']') {
766 			/* range */
767 			NEXT();
768 			if (EAT('-'))
769 				finish = '-';
770 			else
771 				finish = p_b_symbol(p);
772 		} else
773 			finish = start;
774 		if (start == finish)
775 			CHadd(p, cs, start);
776 		else {
777 			if (loc->collate->lc_is_posix) {
778 				(void) REQUIRE((uch)start <= (uch)finish,
779 				    REG_ERANGE);
780 				CHaddrange(p, cs, start, finish);
781 			} else {
782 				(void) REQUIRE(_collate_range_cmp(start,
783 				    finish, loc) <= 0, REG_ERANGE);
784 				for (i = 0; i <= UCHAR_MAX; i++) {
785 					if (_collate_range_cmp(start, i, loc)
786 					    <= 0 &&
787 					    _collate_range_cmp(i, finish, loc)
788 					    <= 0)
789 						CHadd(p, cs, i);
790 				}
791 			}
792 		}
793 		break;
794 	}
795 }
796 
797 /*
798  * p_b_cclass - parse a character-class name and deal with it
799  */
800 static void
801 p_b_cclass(struct parse *p, cset *cs)
802 {
803 	char *sp = p->next;
804 	size_t len;
805 	wctype_t wct;
806 	char clname[16];
807 
808 	while (MORE() && isalpha((uch)PEEK()))
809 		NEXT();
810 	len = p->next - sp;
811 	if (len >= sizeof (clname) - 1) {
812 		SETERROR(REG_ECTYPE);
813 		return;
814 	}
815 	(void) memcpy(clname, sp, len);
816 	clname[len] = '\0';
817 	if ((wct = wctype(clname)) == 0) {
818 		SETERROR(REG_ECTYPE);
819 		return;
820 	}
821 	CHaddtype(p, cs, wct);
822 }
823 
824 /*
825  * p_b_eclass - parse an equivalence-class name and deal with it
826  *
827  * This implementation is incomplete. xxx
828  */
829 static void
830 p_b_eclass(struct parse *p, cset *cs)
831 {
832 	wint_t c;
833 
834 	c = p_b_coll_elem(p, '=');
835 	CHadd(p, cs, c);
836 }
837 
838 /*
839  * p_b_symbol - parse a character or [..]ed multicharacter collating symbol
840  */
841 static wint_t			/* value of symbol */
842 p_b_symbol(struct parse *p)
843 {
844 	wint_t value;
845 
846 	(void) REQUIRE(MORE(), REG_EBRACK);
847 	if (!EATTWO('[', '.'))
848 		return (WGETNEXT());
849 
850 	/* collating symbol */
851 	value = p_b_coll_elem(p, '.');
852 	(void) REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
853 	return (value);
854 }
855 
856 /*
857  * p_b_coll_elem - parse a collating-element name and look it up
858  */
859 static wint_t			/* value of collating element */
860 p_b_coll_elem(struct parse *p,
861 	wint_t endc)		/* name ended by endc,']' */
862 {
863 	char *sp = p->next;
864 	struct cname *cp;
865 	int len;
866 	mbstate_t mbs;
867 	wchar_t wc;
868 	size_t clen;
869 
870 	while (MORE() && !SEETWO(endc, ']'))
871 		NEXT();
872 	if (!MORE()) {
873 		SETERROR(REG_EBRACK);
874 		return (0);
875 	}
876 	len = p->next - sp;
877 	for (cp = cnames; cp->name != NULL; cp++)
878 		if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
879 			return (cp->code);	/* known name */
880 	(void) memset(&mbs, 0, sizeof (mbs));
881 	if ((clen = mbrtowc(&wc, sp, len, &mbs)) == len)
882 		return (wc);			/* single character */
883 	else if (clen == (size_t)-1 || clen == (size_t)-2)
884 		SETERROR(REG_ECHAR);
885 	else
886 		SETERROR(REG_ECOLLATE);		/* neither */
887 	return (0);
888 }
889 
890 /*
891  * othercase - return the case counterpart of an alphabetic
892  */
893 static wint_t			/* if no counterpart, return ch */
894 othercase(wint_t ch)
895 {
896 	assert(iswalpha(ch));
897 	if (iswupper(ch))
898 		return (towlower(ch));
899 	else if (iswlower(ch))
900 		return (towupper(ch));
901 	else			/* peculiar, but could happen */
902 		return (ch);
903 }
904 
905 /*
906  * bothcases - emit a dualcase version of a two-case character
907  *
908  * Boy, is this implementation ever a kludge...
909  */
910 static void
911 bothcases(struct parse *p, wint_t ch)
912 {
913 	char *oldnext = p->next;
914 	char *oldend = p->end;
915 	char bracket[3 + MB_LEN_MAX];
916 	size_t n;
917 	mbstate_t mbs;
918 
919 	assert(othercase(ch) != ch);	/* p_bracket() would recurse */
920 	p->next = bracket;
921 	(void) memset(&mbs, 0, sizeof (mbs));
922 	n = wcrtomb(bracket, ch, &mbs);
923 	assert(n != (size_t)-1);
924 	bracket[n] = ']';
925 	bracket[n + 1] = '\0';
926 	p->end = bracket+n+1;
927 	p_bracket(p);
928 	assert(p->next == p->end);
929 	p->next = oldnext;
930 	p->end = oldend;
931 }
932 
933 /*
934  * ordinary - emit an ordinary character
935  */
936 static void
937 ordinary(struct parse *p, wint_t ch)
938 {
939 	cset *cs;
940 
941 	if ((p->g->cflags&REG_ICASE) && iswalpha(ch) && othercase(ch) != ch)
942 		bothcases(p, ch);
943 	else if ((ch & OPDMASK) == ch)
944 		EMIT(OCHAR, ch);
945 	else {
946 		/*
947 		 * Kludge: character is too big to fit into an OCHAR operand.
948 		 * Emit a singleton set.
949 		 */
950 		if ((cs = allocset(p)) == NULL)
951 			return;
952 		CHadd(p, cs, ch);
953 		EMIT(OANYOF, (int)(cs - p->g->sets));
954 	}
955 }
956 
957 /*
958  * nonnewline - emit REG_NEWLINE version of OANY
959  *
960  * Boy, is this implementation ever a kludge...
961  */
962 static void
963 nonnewline(struct parse *p)
964 {
965 	char *oldnext = p->next;
966 	char *oldend = p->end;
967 	char bracket[4];
968 
969 	p->next = bracket;
970 	p->end = bracket+3;
971 	bracket[0] = '^';
972 	bracket[1] = '\n';
973 	bracket[2] = ']';
974 	bracket[3] = '\0';
975 	p_bracket(p);
976 	assert(p->next == bracket+3);
977 	p->next = oldnext;
978 	p->end = oldend;
979 }
980 
981 /*
982  * repeat - generate code for a bounded repetition, recursively if needed
983  */
984 static void
985 repeat(struct parse *p,
986     sopno start,		/* operand from here to end of strip */
987     int from,			/* repeated from this number */
988     int to)			/* to this number of times (maybe INFINITY) */
989 {
990 	sopno finish = HERE();
991 #define	N	2
992 #define	INF	3
993 #define	REP(f, t)	((f)*8 + (t))
994 #define	MAP(n)	(((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
995 	sopno copy;
996 
997 	if (p->error != 0)	/* head off possible runaway recursion */
998 		return;
999 
1000 	assert(from <= to);
1001 
1002 	switch (REP(MAP(from), MAP(to))) {
1003 	case REP(0, 0):			/* must be user doing this */
1004 		DROP(finish-start);	/* drop the operand */
1005 		break;
1006 	case REP(0, 1):			/* as x{1,1}? */
1007 	case REP(0, N):			/* as x{1,n}? */
1008 	case REP(0, INF):		/* as x{1,}? */
1009 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1010 		INSERT(OCH_, start);		/* offset is wrong... */
1011 		repeat(p, start+1, 1, to);
1012 		ASTERN(OOR1, start);
1013 		AHEAD(start);			/* ... fix it */
1014 		EMIT(OOR2, 0);
1015 		AHEAD(THERE());
1016 		ASTERN(O_CH, THERETHERE());
1017 		break;
1018 	case REP(1, 1):			/* trivial case */
1019 		/* done */
1020 		break;
1021 	case REP(1, N):			/* as x?x{1,n-1} */
1022 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1023 		INSERT(OCH_, start);
1024 		ASTERN(OOR1, start);
1025 		AHEAD(start);
1026 		EMIT(OOR2, 0);			/* offset very wrong... */
1027 		AHEAD(THERE());			/* ...so fix it */
1028 		ASTERN(O_CH, THERETHERE());
1029 		copy = dupl(p, start+1, finish+1);
1030 		assert(copy == finish+4);
1031 		repeat(p, copy, 1, to-1);
1032 		break;
1033 	case REP(1, INF):		/* as x+ */
1034 		INSERT(OPLUS_, start);
1035 		ASTERN(O_PLUS, start);
1036 		break;
1037 	case REP(N, N):			/* as xx{m-1,n-1} */
1038 		copy = dupl(p, start, finish);
1039 		repeat(p, copy, from-1, to-1);
1040 		break;
1041 	case REP(N, INF):		/* as xx{n-1,INF} */
1042 		copy = dupl(p, start, finish);
1043 		repeat(p, copy, from-1, to);
1044 		break;
1045 	default:			/* "can't happen" */
1046 		SETERROR(REG_EFATAL);	/* just in case */
1047 		break;
1048 	}
1049 }
1050 
1051 /*
1052  * wgetnext - helper function for WGETNEXT() macro. Gets the next wide
1053  * character from the parse struct, signals a REG_ILLSEQ error if the
1054  * character can't be converted. Returns the number of bytes consumed.
1055  */
1056 static wint_t
1057 wgetnext(struct parse *p)
1058 {
1059 	mbstate_t mbs;
1060 	wchar_t wc;
1061 	size_t n;
1062 
1063 	(void) memset(&mbs, 0, sizeof (mbs));
1064 	n = mbrtowc(&wc, p->next, p->end - p->next, &mbs);
1065 	if (n == (size_t)-1 || n == (size_t)-2) {
1066 		SETERROR(REG_ECHAR);
1067 		return (0);
1068 	}
1069 	if (n == 0)
1070 		n = 1;
1071 	p->next += n;
1072 	return (wc);
1073 }
1074 
1075 /*
1076  * seterr - set an error condition
1077  */
1078 static int			/* useless but makes type checking happy */
1079 seterr(struct parse *p, int e)
1080 {
1081 	if (p->error == 0)	/* keep earliest error condition */
1082 		p->error = e;
1083 	p->next = nuls;		/* try to bring things to a halt */
1084 	p->end = nuls;
1085 	return (0);		/* make the return value well-defined */
1086 }
1087 
1088 /*
1089  * allocset - allocate a set of characters for []
1090  */
1091 static cset *
1092 allocset(struct parse *p)
1093 {
1094 	cset *cs, *ncs;
1095 
1096 	ncs = realloc(p->g->sets, (p->g->ncsets + 1) * sizeof (*ncs));
1097 	if (ncs == NULL) {
1098 		SETERROR(REG_ESPACE);
1099 		return (NULL);
1100 	}
1101 	p->g->sets = ncs;
1102 	cs = &p->g->sets[p->g->ncsets++];
1103 	(void) memset(cs, 0, sizeof (*cs));
1104 
1105 	return (cs);
1106 }
1107 
1108 /*
1109  * freeset - free a now-unused set
1110  */
1111 static void
1112 freeset(struct parse *p, cset *cs)
1113 {
1114 	cset *top = &p->g->sets[p->g->ncsets];
1115 
1116 	free(cs->wides);
1117 	free(cs->ranges);
1118 	free(cs->types);
1119 	(void) memset(cs, 0, sizeof (*cs));
1120 	if (cs == top-1)	/* recover only the easy case */
1121 		p->g->ncsets--;
1122 }
1123 
1124 /*
1125  * singleton - Determine whether a set contains only one character,
1126  * returning it if so, otherwise returning OUT.
1127  */
1128 static wint_t
1129 singleton(cset *cs)
1130 {
1131 	wint_t i, s, n;
1132 
1133 	for (i = n = 0; i < NC; i++)
1134 		if (CHIN(cs, i)) {
1135 			n++;
1136 			s = i;
1137 		}
1138 	if (n == 1)
1139 		return (s);
1140 	if (cs->nwides == 1 && cs->nranges == 0 && cs->ntypes == 0 &&
1141 	    cs->icase == 0)
1142 		return (cs->wides[0]);
1143 	/* Don't bother handling the other cases. */
1144 	return (OUT);
1145 }
1146 
1147 /*
1148  * CHadd - add character to character set.
1149  */
1150 static void
1151 CHadd(struct parse *p, cset *cs, wint_t ch)
1152 {
1153 	wint_t nch, *newwides;
1154 	assert(ch >= 0);
1155 	if (ch < NC)
1156 		cs->bmp[ch >> 3] |= 1 << (ch & 7);
1157 	else {
1158 		newwides = realloc(cs->wides, (cs->nwides + 1) *
1159 		    sizeof (*cs->wides));
1160 		if (newwides == NULL) {
1161 			SETERROR(REG_ESPACE);
1162 			return;
1163 		}
1164 		cs->wides = newwides;
1165 		cs->wides[cs->nwides++] = ch;
1166 	}
1167 	if (cs->icase) {
1168 		if ((nch = towlower(ch)) < NC)
1169 			cs->bmp[nch >> 3] |= 1 << (nch & 7);
1170 		if ((nch = towupper(ch)) < NC)
1171 			cs->bmp[nch >> 3] |= 1 << (nch & 7);
1172 	}
1173 }
1174 
1175 /*
1176  * CHaddrange - add all characters in the range [min,max] to a character set.
1177  */
1178 static void
1179 CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max)
1180 {
1181 	crange *newranges;
1182 
1183 	for (; min < NC && min <= max; min++)
1184 		CHadd(p, cs, min);
1185 	if (min >= max)
1186 		return;
1187 	newranges = realloc(cs->ranges, (cs->nranges + 1) *
1188 	    sizeof (*cs->ranges));
1189 	if (newranges == NULL) {
1190 		SETERROR(REG_ESPACE);
1191 		return;
1192 	}
1193 	cs->ranges = newranges;
1194 	cs->ranges[cs->nranges].min = min;
1195 	cs->ranges[cs->nranges].min = max;
1196 	cs->nranges++;
1197 }
1198 
1199 /*
1200  * CHaddtype - add all characters of a certain type to a character set.
1201  */
1202 static void
1203 CHaddtype(struct parse *p, cset *cs, wctype_t wct)
1204 {
1205 	wint_t i;
1206 	wctype_t *newtypes;
1207 
1208 	for (i = 0; i < NC; i++)
1209 		if (iswctype(i, wct))
1210 			CHadd(p, cs, i);
1211 	newtypes = realloc(cs->types, (cs->ntypes + 1) *
1212 	    sizeof (*cs->types));
1213 	if (newtypes == NULL) {
1214 		SETERROR(REG_ESPACE);
1215 		return;
1216 	}
1217 	cs->types = newtypes;
1218 	cs->types[cs->ntypes++] = wct;
1219 }
1220 
1221 /*
1222  * dupl - emit a duplicate of a bunch of sops
1223  */
1224 static sopno			/* start of duplicate */
1225 dupl(struct parse *p,
1226 	sopno start,		/* from here */
1227 	sopno finish)		/* to this less one */
1228 {
1229 	sopno ret = HERE();
1230 	sopno len = finish - start;
1231 
1232 	assert(finish >= start);
1233 	if (len == 0)
1234 		return (ret);
1235 	enlarge(p, p->ssize + len);	/* this many unexpected additions */
1236 	assert(p->ssize >= p->slen + len);
1237 	(void) memcpy((char *)(p->strip + p->slen),
1238 	    (char *)(p->strip + start), (size_t)len*sizeof (sop));
1239 	p->slen += len;
1240 	return (ret);
1241 }
1242 
1243 /*
1244  * doemit - emit a strip operator
1245  *
1246  * It might seem better to implement this as a macro with a function as
1247  * hard-case backup, but it's just too big and messy unless there are
1248  * some changes to the data structures.  Maybe later.
1249  */
1250 static void
1251 doemit(struct parse *p, sop op, size_t opnd)
1252 {
1253 	/* avoid making error situations worse */
1254 	if (p->error != 0)
1255 		return;
1256 
1257 	/* deal with oversize operands ("can't happen", more or less) */
1258 	assert(opnd < 1<<OPSHIFT);
1259 
1260 	/* deal with undersized strip */
1261 	if (p->slen >= p->ssize)
1262 		enlarge(p, (p->ssize+1) / 2 * 3);	/* +50% */
1263 	assert(p->slen < p->ssize);
1264 
1265 	/* finally, it's all reduced to the easy case */
1266 	p->strip[p->slen++] = SOP(op, opnd);
1267 }
1268 
1269 /*
1270  * doinsert - insert a sop into the strip
1271  */
1272 static void
1273 doinsert(struct parse *p, sop op, size_t opnd, sopno pos)
1274 {
1275 	sopno sn;
1276 	sop s;
1277 	int i;
1278 
1279 	/* avoid making error situations worse */
1280 	if (p->error != 0)
1281 		return;
1282 
1283 	sn = HERE();
1284 	EMIT(op, opnd);		/* do checks, ensure space */
1285 	assert(HERE() == sn+1);
1286 	s = p->strip[sn];
1287 
1288 	/* adjust paren pointers */
1289 	assert(pos > 0);
1290 	for (i = 1; i < NPAREN; i++) {
1291 		if (p->pbegin[i] >= pos) {
1292 			p->pbegin[i]++;
1293 		}
1294 		if (p->pend[i] >= pos) {
1295 			p->pend[i]++;
1296 		}
1297 	}
1298 
1299 	(void) memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1300 	    (HERE()-pos-1)*sizeof (sop));
1301 	p->strip[pos] = s;
1302 }
1303 
1304 /*
1305  * dofwd - complete a forward reference
1306  */
1307 static void
1308 dofwd(struct parse *p, sopno pos, sop value)
1309 {
1310 	/* avoid making error situations worse */
1311 	if (p->error != 0)
1312 		return;
1313 
1314 	assert(value < 1<<OPSHIFT);
1315 	p->strip[pos] = OP(p->strip[pos]) | value;
1316 }
1317 
1318 /*
1319  * enlarge - enlarge the strip
1320  */
1321 static void
1322 enlarge(struct parse *p, sopno size)
1323 {
1324 	sop *sp;
1325 
1326 	if (p->ssize >= size)
1327 		return;
1328 
1329 	sp = (sop *)realloc(p->strip, size*sizeof (sop));
1330 	if (sp == NULL) {
1331 		SETERROR(REG_ESPACE);
1332 		return;
1333 	}
1334 	p->strip = sp;
1335 	p->ssize = size;
1336 }
1337 
1338 /*
1339  * stripsnug - compact the strip
1340  */
1341 static void
1342 stripsnug(struct parse *p, struct re_guts *g)
1343 {
1344 	g->nstates = p->slen;
1345 	g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof (sop));
1346 	if (g->strip == NULL) {
1347 		SETERROR(REG_ESPACE);
1348 		g->strip = p->strip;
1349 	}
1350 }
1351 
1352 /*
1353  * findmust - fill in must and mlen with longest mandatory literal string
1354  *
1355  * This algorithm could do fancy things like analyzing the operands of |
1356  * for common subsequences.  Someday.  This code is simple and finds most
1357  * of the interesting cases.
1358  *
1359  * Note that must and mlen got initialized during setup.
1360  */
1361 static void
1362 findmust(struct parse *p, struct re_guts *g)
1363 {
1364 	sop *scan;
1365 	sop *start;
1366 	sop *newstart;
1367 	sopno newlen;
1368 	sop s;
1369 	char *cp;
1370 	int offset;
1371 	char buf[MB_LEN_MAX];
1372 	size_t clen;
1373 	mbstate_t mbs;
1374 	locale_t loc = uselocale(NULL);
1375 
1376 	/* avoid making error situations worse */
1377 	if (p->error != 0)
1378 		return;
1379 
1380 	/*
1381 	 * It's not generally safe to do a ``char'' substring search on
1382 	 * multibyte character strings, but it's safe for at least
1383 	 * UTF-8 (see RFC 3629).
1384 	 */
1385 	if (MB_CUR_MAX > 1 &&
1386 	    strcmp(loc->runelocale->__encoding, "UTF-8") != 0)
1387 		return;
1388 
1389 	/* find the longest OCHAR sequence in strip */
1390 	newlen = 0;
1391 	offset = 0;
1392 	g->moffset = 0;
1393 	scan = g->strip + 1;
1394 	do {
1395 		s = *scan++;
1396 		switch (OP(s)) {
1397 		case OCHAR:		/* sequence member */
1398 			if (newlen == 0) {		/* new sequence */
1399 				(void) memset(&mbs, 0, sizeof (mbs));
1400 				newstart = scan - 1;
1401 			}
1402 			clen = wcrtomb(buf, OPND(s), &mbs);
1403 			if (clen == (size_t)-1)
1404 				goto toohard;
1405 			newlen += clen;
1406 			break;
1407 		case OPLUS_:		/* things that don't break one */
1408 		case OLPAREN:
1409 		case ORPAREN:
1410 			break;
1411 		case OQUEST_:		/* things that must be skipped */
1412 		case OCH_:
1413 			offset = altoffset(scan, offset);
1414 			scan--;
1415 			do {
1416 				scan += OPND(s);
1417 				s = *scan;
1418 				/* assert() interferes w debug printouts */
1419 				if (OP(s) != O_QUEST && OP(s) != O_CH &&
1420 				    OP(s) != OOR2) {
1421 					g->iflags |= BAD;
1422 					return;
1423 				}
1424 			} while (OP(s) != O_QUEST && OP(s) != O_CH);
1425 			/* FALLTHROUGH */
1426 		case OBOW:		/* things that break a sequence */
1427 		case OEOW:
1428 		case OBOL:
1429 		case OEOL:
1430 		case O_QUEST:
1431 		case O_CH:
1432 		case OEND:
1433 			if (newlen > g->mlen) {		/* ends one */
1434 				start = newstart;
1435 				g->mlen = newlen;
1436 				if (offset > -1) {
1437 					g->moffset += offset;
1438 					offset = newlen;
1439 				} else
1440 					g->moffset = offset;
1441 			} else {
1442 				if (offset > -1)
1443 					offset += newlen;
1444 			}
1445 			newlen = 0;
1446 			break;
1447 		case OANY:
1448 			if (newlen > g->mlen) {		/* ends one */
1449 				start = newstart;
1450 				g->mlen = newlen;
1451 				if (offset > -1) {
1452 					g->moffset += offset;
1453 					offset = newlen;
1454 				} else
1455 					g->moffset = offset;
1456 			} else {
1457 				if (offset > -1)
1458 					offset += newlen;
1459 			}
1460 			if (offset > -1)
1461 				offset++;
1462 			newlen = 0;
1463 			break;
1464 		case OANYOF:		/* may or may not invalidate offset */
1465 			/* First, everything as OANY */
1466 			if (newlen > g->mlen) {		/* ends one */
1467 				start = newstart;
1468 				g->mlen = newlen;
1469 				if (offset > -1) {
1470 					g->moffset += offset;
1471 					offset = newlen;
1472 				} else
1473 					g->moffset = offset;
1474 			} else {
1475 				if (offset > -1)
1476 					offset += newlen;
1477 			}
1478 			if (offset > -1)
1479 				offset++;
1480 			newlen = 0;
1481 			break;
1482 		toohard:
1483 		default:
1484 			/*
1485 			 * Anything here makes it impossible or too hard
1486 			 * to calculate the offset -- so we give up;
1487 			 * save the last known good offset, in case the
1488 			 * must sequence doesn't occur later.
1489 			 */
1490 			if (newlen > g->mlen) {		/* ends one */
1491 				start = newstart;
1492 				g->mlen = newlen;
1493 				if (offset > -1)
1494 					g->moffset += offset;
1495 				else
1496 					g->moffset = offset;
1497 			}
1498 			offset = -1;
1499 			newlen = 0;
1500 			break;
1501 		}
1502 	} while (OP(s) != OEND);
1503 
1504 	if (g->mlen == 0) {		/* there isn't one */
1505 		g->moffset = -1;
1506 		return;
1507 	}
1508 
1509 	/* turn it into a character string */
1510 	g->must = malloc((size_t)g->mlen + 1);
1511 	if (g->must == NULL) {		/* argh; just forget it */
1512 		g->mlen = 0;
1513 		g->moffset = -1;
1514 		return;
1515 	}
1516 	cp = g->must;
1517 	scan = start;
1518 	(void) memset(&mbs, 0, sizeof (mbs));
1519 	while (cp < g->must + g->mlen) {
1520 		while (OP(s = *scan++) != OCHAR)
1521 			continue;
1522 		clen = wcrtomb(cp, OPND(s), &mbs);
1523 		assert(clen != (size_t)-1);
1524 		cp += clen;
1525 	}
1526 	assert(cp == g->must + g->mlen);
1527 	*cp++ = '\0';		/* just on general principles */
1528 }
1529 
1530 /*
1531  * altoffset - choose biggest offset among multiple choices
1532  *
1533  * Compute, recursively if necessary, the largest offset among multiple
1534  * re paths.
1535  */
1536 static int
1537 altoffset(sop *scan, int offset)
1538 {
1539 	int largest;
1540 	int try;
1541 	sop s;
1542 
1543 	/* If we gave up already on offsets, return */
1544 	if (offset == -1)
1545 		return (-1);
1546 
1547 	largest = 0;
1548 	try = 0;
1549 	s = *scan++;
1550 	while (OP(s) != O_QUEST && OP(s) != O_CH) {
1551 		switch (OP(s)) {
1552 		case OOR1:
1553 			if (try > largest)
1554 				largest = try;
1555 			try = 0;
1556 			break;
1557 		case OQUEST_:
1558 		case OCH_:
1559 			try = altoffset(scan, try);
1560 			if (try == -1)
1561 				return (-1);
1562 			scan--;
1563 			do {
1564 				scan += OPND(s);
1565 				s = *scan;
1566 				if (OP(s) != O_QUEST && OP(s) != O_CH &&
1567 				    OP(s) != OOR2)
1568 					return (-1);
1569 			} while (OP(s) != O_QUEST && OP(s) != O_CH);
1570 			/*
1571 			 * We must skip to the next position, or we'll
1572 			 * leave altoffset() too early.
1573 			 */
1574 			scan++;
1575 			break;
1576 		case OANYOF:
1577 		case OCHAR:
1578 		case OANY:
1579 			try++;
1580 			/*FALLTHRU*/
1581 		case OBOW:
1582 		case OEOW:
1583 		case OLPAREN:
1584 		case ORPAREN:
1585 		case OOR2:
1586 			break;
1587 		default:
1588 			try = -1;
1589 			break;
1590 		}
1591 		if (try == -1)
1592 			return (-1);
1593 		s = *scan++;
1594 	}
1595 
1596 	if (try > largest)
1597 		largest = try;
1598 
1599 	return (largest+offset);
1600 }
1601 
1602 /*
1603  * computejumps - compute char jumps for BM scan
1604  *
1605  * This algorithm assumes g->must exists and is has size greater than
1606  * zero. It's based on the algorithm found on Computer Algorithms by
1607  * Sara Baase.
1608  *
1609  * A char jump is the number of characters one needs to jump based on
1610  * the value of the character from the text that was mismatched.
1611  */
1612 static void
1613 computejumps(struct parse *p, struct re_guts *g)
1614 {
1615 	int ch;
1616 	int mindex;
1617 
1618 	/* Avoid making errors worse */
1619 	if (p->error != 0)
1620 		return;
1621 
1622 	g->charjump = (int *)malloc((NC + 1) * sizeof (int));
1623 	if (g->charjump == NULL)	/* Not a fatal error */
1624 		return;
1625 	/* Adjust for signed chars, if necessary */
1626 	g->charjump = &g->charjump[-(CHAR_MIN)];
1627 
1628 	/*
1629 	 * If the character does not exist in the pattern, the jump
1630 	 * is equal to the number of characters in the pattern.
1631 	 */
1632 	for (ch = CHAR_MIN; ch < (CHAR_MAX + 1); ch++)
1633 		g->charjump[ch] = g->mlen;
1634 
1635 	/*
1636 	 * If the character does exist, compute the jump that would
1637 	 * take us to the last character in the pattern equal to it
1638 	 * (notice that we match right to left, so that last character
1639 	 * is the first one that would be matched).
1640 	 */
1641 	for (mindex = 0; mindex < g->mlen; mindex++)
1642 		g->charjump[(int)g->must[mindex]] = g->mlen - mindex - 1;
1643 }
1644 
1645 /*
1646  * computematchjumps - compute match jumps for BM scan
1647  *
1648  * This algorithm assumes g->must exists and is has size greater than
1649  * zero. It's based on the algorithm found on Computer Algorithms by
1650  * Sara Baase.
1651  *
1652  * A match jump is the number of characters one needs to advance based
1653  * on the already-matched suffix.
1654  * Notice that all values here are minus (g->mlen-1), because of the way
1655  * the search algorithm works.
1656  */
1657 static void
1658 computematchjumps(struct parse *p, struct re_guts *g)
1659 {
1660 	int mindex;		/* General "must" iterator */
1661 	int suffix;		/* Keeps track of matching suffix */
1662 	int ssuffix;		/* Keeps track of suffixes' suffix */
1663 	int *pmatches;
1664 				/*
1665 				 * pmatches[k] points to the next i
1666 				 * such that i+1...mlen is a substring
1667 				 * of k+1...k+mlen-i-1
1668 				 */
1669 
1670 	/* Avoid making errors worse */
1671 	if (p->error != 0)
1672 		return;
1673 
1674 	pmatches = (int *)malloc(g->mlen * sizeof (unsigned int));
1675 	if (pmatches == NULL) {
1676 		g->matchjump = NULL;
1677 		return;
1678 	}
1679 
1680 	g->matchjump = (int *)malloc(g->mlen * sizeof (unsigned int));
1681 	if (g->matchjump == NULL)	/* Not a fatal error */
1682 		return;
1683 
1684 	/* Set maximum possible jump for each character in the pattern */
1685 	for (mindex = 0; mindex < g->mlen; mindex++)
1686 		g->matchjump[mindex] = 2*g->mlen - mindex - 1;
1687 
1688 	/* Compute pmatches[] */
1689 	for (mindex = g->mlen - 1, suffix = g->mlen; mindex >= 0;
1690 	    mindex--, suffix--) {
1691 		pmatches[mindex] = suffix;
1692 
1693 		/*
1694 		 * If a mismatch is found, interrupting the substring,
1695 		 * compute the matchjump for that position. If no
1696 		 * mismatch is found, then a text substring mismatched
1697 		 * against the suffix will also mismatch against the
1698 		 * substring.
1699 		 */
1700 		while (suffix < g->mlen && g->must[mindex] != g->must[suffix]) {
1701 			g->matchjump[suffix] = MIN(g->matchjump[suffix],
1702 			    g->mlen - mindex - 1);
1703 			suffix = pmatches[suffix];
1704 		}
1705 	}
1706 
1707 	/*
1708 	 * Compute the matchjump up to the last substring found to jump
1709 	 * to the beginning of the largest must pattern prefix matching
1710 	 * it's own suffix.
1711 	 */
1712 	for (mindex = 0; mindex <= suffix; mindex++)
1713 		g->matchjump[mindex] = MIN(g->matchjump[mindex],
1714 		    g->mlen + suffix - mindex);
1715 
1716 	ssuffix = pmatches[suffix];
1717 	while (suffix < g->mlen) {
1718 		while (suffix <= ssuffix && suffix < g->mlen) {
1719 			g->matchjump[suffix] = MIN(g->matchjump[suffix],
1720 			    g->mlen + ssuffix - suffix);
1721 			suffix++;
1722 		}
1723 		if (suffix < g->mlen)
1724 			ssuffix = pmatches[ssuffix];
1725 	}
1726 
1727 	free(pmatches);
1728 }
1729 
1730 /*
1731  * pluscount - count + nesting
1732  */
1733 static sopno			/* nesting depth */
1734 pluscount(struct parse *p, struct re_guts *g)
1735 {
1736 	sop *scan;
1737 	sop s;
1738 	sopno plusnest = 0;
1739 	sopno maxnest = 0;
1740 
1741 	if (p->error != 0)
1742 		return (0);	/* there may not be an OEND */
1743 
1744 	scan = g->strip + 1;
1745 	do {
1746 		s = *scan++;
1747 		switch (OP(s)) {
1748 		case OPLUS_:
1749 			plusnest++;
1750 			break;
1751 		case O_PLUS:
1752 			if (plusnest > maxnest)
1753 				maxnest = plusnest;
1754 			plusnest--;
1755 			break;
1756 		}
1757 	} while (OP(s) != OEND);
1758 	if (plusnest != 0)
1759 		g->iflags |= BAD;
1760 	return (maxnest);
1761 }
1762