xref: /freebsd/lib/libc/regex/regcomp.c (revision f77b5b295da3146c3b601767cbc4e85e6713192c)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1992, 1993, 1994 Henry Spencer.
5  * Copyright (c) 1992, 1993, 1994
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Copyright (c) 2011 The FreeBSD Foundation
9  *
10  * Portions of this software were developed by David Chisnall
11  * under sponsorship from the FreeBSD Foundation.
12  *
13  * This code is derived from software contributed to Berkeley by
14  * Henry Spencer.
15  *
16  * Redistribution and use in source and binary forms, with or without
17  * modification, are permitted provided that the following conditions
18  * are met:
19  * 1. Redistributions of source code must retain the above copyright
20  *    notice, this list of conditions and the following disclaimer.
21  * 2. Redistributions in binary form must reproduce the above copyright
22  *    notice, this list of conditions and the following disclaimer in the
23  *    documentation and/or other materials provided with the distribution.
24  * 3. Neither the name of the University nor the names of its contributors
25  *    may be used to endorse or promote products derived from this software
26  *    without specific prior written permission.
27  *
28  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
29  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
30  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
31  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
32  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
33  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
34  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
35  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
36  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
37  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
38  * SUCH DAMAGE.
39  */
40 
41 #include <sys/types.h>
42 #include <stdio.h>
43 #include <string.h>
44 #include <ctype.h>
45 #include <limits.h>
46 #include <stdlib.h>
47 #include <regex.h>
48 #include <stdbool.h>
49 #include <wchar.h>
50 #include <wctype.h>
51 
52 #ifndef LIBREGEX
53 #include "collate.h"
54 #endif
55 
56 #include "utils.h"
57 #include "regex2.h"
58 
59 #include "cname.h"
60 
61 /*
62  * Branching context, used to keep track of branch state for all of the branch-
63  * aware functions. In addition to keeping track of branch positions for the
64  * p_branch_* functions, we use this to simplify some clumsiness in BREs for
65  * detection of whether ^ is acting as an anchor or being used erroneously and
66  * also for whether we're in a sub-expression or not.
67  */
68 struct branchc {
69 	sopno start;
70 	sopno back;
71 	sopno fwd;
72 
73 	int nbranch;
74 	int nchain;
75 	bool outer;
76 	bool terminate;
77 };
78 
79 /*
80  * parse structure, passed up and down to avoid global variables and
81  * other clumsinesses
82  */
83 struct parse {
84 	const char *next;	/* next character in RE */
85 	const char *end;	/* end of string (-> NUL normally) */
86 	int error;		/* has an error been seen? */
87 	int gnuext;
88 	sop *strip;		/* malloced strip */
89 	sopno ssize;		/* malloced strip size (allocated) */
90 	sopno slen;		/* malloced strip length (used) */
91 	int ncsalloc;		/* number of csets allocated */
92 	struct re_guts *g;
93 #	define	NPAREN	10	/* we need to remember () 1-9 for back refs */
94 	sopno pbegin[NPAREN];	/* -> ( ([0] unused) */
95 	sopno pend[NPAREN];	/* -> ) ([0] unused) */
96 	bool allowbranch;	/* can this expression branch? */
97 	bool bre;		/* convenience; is this a BRE? */
98 	int pflags;		/* other parsing flags -- legacy escapes? */
99 	bool (*parse_expr)(struct parse *, struct branchc *);
100 	void (*pre_parse)(struct parse *, struct branchc *);
101 	void (*post_parse)(struct parse *, struct branchc *);
102 };
103 
104 #define PFLAG_LEGACY_ESC	0x00000001
105 
106 /* ========= begin header generated by ./mkh ========= */
107 #ifdef __cplusplus
108 extern "C" {
109 #endif
110 
111 /* === regcomp.c === */
112 static bool p_ere_exp(struct parse *p, struct branchc *bc);
113 static void p_str(struct parse *p);
114 static int p_branch_eat_delim(struct parse *p, struct branchc *bc);
115 static void p_branch_ins_offset(struct parse *p, struct branchc *bc);
116 static void p_branch_fix_tail(struct parse *p, struct branchc *bc);
117 static bool p_branch_empty(struct parse *p, struct branchc *bc);
118 static bool p_branch_do(struct parse *p, struct branchc *bc);
119 static void p_bre_pre_parse(struct parse *p, struct branchc *bc);
120 static void p_bre_post_parse(struct parse *p, struct branchc *bc);
121 static void p_re(struct parse *p, int end1, int end2);
122 static bool p_simp_re(struct parse *p, struct branchc *bc);
123 static int p_count(struct parse *p);
124 static void p_bracket(struct parse *p);
125 static int p_range_cmp(wchar_t c1, wchar_t c2);
126 static void p_b_term(struct parse *p, cset *cs);
127 static int p_b_pseudoclass(struct parse *p, char c);
128 static void p_b_cclass(struct parse *p, cset *cs);
129 static void p_b_cclass_named(struct parse *p, cset *cs, const char[]);
130 static void p_b_eclass(struct parse *p, cset *cs);
131 static wint_t p_b_symbol(struct parse *p);
132 static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
133 static bool may_escape(struct parse *p, const wint_t ch);
134 static wint_t othercase(wint_t ch);
135 static void bothcases(struct parse *p, wint_t ch);
136 static void ordinary(struct parse *p, wint_t ch);
137 static void nonnewline(struct parse *p);
138 static void repeat(struct parse *p, sopno start, int from, int to);
139 static int seterr(struct parse *p, int e);
140 static cset *allocset(struct parse *p);
141 static void freeset(struct parse *p, cset *cs);
142 static void CHadd(struct parse *p, cset *cs, wint_t ch);
143 static void CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max);
144 static void CHaddtype(struct parse *p, cset *cs, wctype_t wct);
145 static wint_t singleton(cset *cs);
146 static sopno dupl(struct parse *p, sopno start, sopno finish);
147 static void doemit(struct parse *p, sop op, size_t opnd);
148 static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
149 static void dofwd(struct parse *p, sopno pos, sop value);
150 static int enlarge(struct parse *p, sopno size);
151 static void stripsnug(struct parse *p, struct re_guts *g);
152 static void findmust(struct parse *p, struct re_guts *g);
153 static int altoffset(sop *scan, int offset);
154 static void computejumps(struct parse *p, struct re_guts *g);
155 static void computematchjumps(struct parse *p, struct re_guts *g);
156 static sopno pluscount(struct parse *p, struct re_guts *g);
157 static wint_t wgetnext(struct parse *p);
158 
159 #ifdef __cplusplus
160 }
161 #endif
162 /* ========= end header generated by ./mkh ========= */
163 
164 static char nuls[10];		/* place to point scanner in event of error */
165 
166 /*
167  * macros for use with parse structure
168  * BEWARE:  these know that the parse structure is named `p' !!!
169  */
170 #define	PEEK()	(*p->next)
171 #define	PEEK2()	(*(p->next+1))
172 #define	MORE()	(p->end - p->next > 0)
173 #define	MORE2()	(p->end - p->next > 1)
174 #define	SEE(c)	(MORE() && PEEK() == (c))
175 #define	SEETWO(a, b)	(MORE2() && PEEK() == (a) && PEEK2() == (b))
176 #define	SEESPEC(a)	(p->bre ? SEETWO('\\', a) : SEE(a))
177 #define	EAT(c)	((SEE(c)) ? (NEXT(), 1) : 0)
178 #define	EATTWO(a, b)	((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
179 #define	EATSPEC(a)	(p->bre ? EATTWO('\\', a) : EAT(a))
180 #define	NEXT()	(p->next++)
181 #define	NEXT2()	(p->next += 2)
182 #define	NEXTn(n)	(p->next += (n))
183 #define	GETNEXT()	(*p->next++)
184 #define	WGETNEXT()	wgetnext(p)
185 #define	SETERROR(e)	seterr(p, (e))
186 #define	REQUIRE(co, e)	((co) || SETERROR(e))
187 #define	MUSTSEE(c, e)	(REQUIRE(MORE() && PEEK() == (c), e))
188 #define	MUSTEAT(c, e)	(REQUIRE(MORE() && GETNEXT() == (c), e))
189 #define	MUSTNOTSEE(c, e)	(REQUIRE(!MORE() || PEEK() != (c), e))
190 #define	EMIT(op, sopnd)	doemit(p, (sop)(op), (size_t)(sopnd))
191 #define	INSERT(op, pos)	doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
192 #define	AHEAD(pos)		dofwd(p, pos, HERE()-(pos))
193 #define	ASTERN(sop, pos)	EMIT(sop, HERE()-pos)
194 #define	HERE()		(p->slen)
195 #define	THERE()		(p->slen - 1)
196 #define	THERETHERE()	(p->slen - 2)
197 #define	DROP(n)	(p->slen -= (n))
198 
199 /* Macro used by computejump()/computematchjump() */
200 #define MIN(a,b)	((a)<(b)?(a):(b))
201 
202 static int				/* 0 success, otherwise REG_something */
regcomp_internal(regex_t * __restrict preg,const char * __restrict pattern,int cflags,int pflags)203 regcomp_internal(regex_t * __restrict preg,
204 	const char * __restrict pattern,
205 	int cflags, int pflags)
206 {
207 	struct parse pa;
208 	struct re_guts *g;
209 	struct parse *p = &pa;
210 	int i;
211 	size_t len;
212 	size_t maxlen;
213 #ifdef REDEBUG
214 #	define	GOODFLAGS(f)	(f)
215 #else
216 #	define	GOODFLAGS(f)	((f)&~REG_DUMP)
217 #endif
218 
219 	cflags = GOODFLAGS(cflags);
220 	if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
221 		return(REG_INVARG);
222 
223 	if (cflags&REG_PEND) {
224 		if (preg->re_endp < pattern)
225 			return(REG_INVARG);
226 		len = preg->re_endp - pattern;
227 	} else
228 		len = strlen(pattern);
229 
230 	/* do the mallocs early so failure handling is easy */
231 	g = (struct re_guts *)malloc(sizeof(struct re_guts));
232 	if (g == NULL)
233 		return(REG_ESPACE);
234 	/*
235 	 * Limit the pattern space to avoid a 32-bit overflow on buffer
236 	 * extension.  Also avoid any signed overflow in case of conversion
237 	 * so make the real limit based on a 31-bit overflow.
238 	 *
239 	 * Likely not applicable on 64-bit systems but handle the case
240 	 * generically (who are we to stop people from using ~715MB+
241 	 * patterns?).
242 	 */
243 	maxlen = ((size_t)-1 >> 1) / sizeof(sop) * 2 / 3;
244 	if (len >= maxlen) {
245 		free((char *)g);
246 		return(REG_ESPACE);
247 	}
248 	p->ssize = len/(size_t)2*(size_t)3 + (size_t)1;	/* ugh */
249 	assert(p->ssize >= len);
250 
251 	p->strip = (sop *)malloc(p->ssize * sizeof(sop));
252 	p->slen = 0;
253 	if (p->strip == NULL) {
254 		free((char *)g);
255 		return(REG_ESPACE);
256 	}
257 
258 	/* set things up */
259 	p->g = g;
260 	p->next = pattern;	/* convenience; we do not modify it */
261 	p->end = p->next + len;
262 	p->error = 0;
263 	p->ncsalloc = 0;
264 	p->pflags = pflags;
265 	for (i = 0; i < NPAREN; i++) {
266 		p->pbegin[i] = 0;
267 		p->pend[i] = 0;
268 	}
269 #ifdef LIBREGEX
270 	if (cflags&REG_POSIX) {
271 		p->gnuext = false;
272 		p->allowbranch = (cflags & REG_EXTENDED) != 0;
273 	} else
274 		p->gnuext = p->allowbranch = true;
275 #else
276 	p->gnuext = false;
277 	p->allowbranch = (cflags & REG_EXTENDED) != 0;
278 #endif
279 	if (cflags & REG_EXTENDED) {
280 		p->bre = false;
281 		p->parse_expr = p_ere_exp;
282 		p->pre_parse = NULL;
283 		p->post_parse = NULL;
284 	} else {
285 		p->bre = true;
286 		p->parse_expr = p_simp_re;
287 		p->pre_parse = p_bre_pre_parse;
288 		p->post_parse = p_bre_post_parse;
289 	}
290 	g->sets = NULL;
291 	g->ncsets = 0;
292 	g->cflags = cflags;
293 	g->iflags = 0;
294 	g->nbol = 0;
295 	g->neol = 0;
296 	g->must = NULL;
297 	g->moffset = -1;
298 	g->charjump = NULL;
299 	g->matchjump = NULL;
300 	g->mlen = 0;
301 	g->nsub = 0;
302 	g->backrefs = 0;
303 
304 	/* do it */
305 	EMIT(OEND, 0);
306 	g->firststate = THERE();
307 	if (cflags & REG_NOSPEC)
308 		p_str(p);
309 	else
310 		p_re(p, OUT, OUT);
311 	EMIT(OEND, 0);
312 	g->laststate = THERE();
313 
314 	/* tidy up loose ends and fill things in */
315 	stripsnug(p, g);
316 	findmust(p, g);
317 	/* only use Boyer-Moore algorithm if the pattern is bigger
318 	 * than three characters
319 	 */
320 	if(g->mlen > 3) {
321 		computejumps(p, g);
322 		computematchjumps(p, g);
323 		if(g->matchjump == NULL && g->charjump != NULL) {
324 			free(&g->charjump[CHAR_MIN]);
325 			g->charjump = NULL;
326 		}
327 	}
328 	g->nplus = pluscount(p, g);
329 	g->magic = MAGIC2;
330 	preg->re_nsub = g->nsub;
331 	preg->re_g = g;
332 	preg->re_magic = MAGIC1;
333 #ifndef REDEBUG
334 	/* not debugging, so can't rely on the assert() in regexec() */
335 	if (g->iflags&BAD)
336 		SETERROR(REG_ASSERT);
337 #endif
338 
339 	/* win or lose, we're done */
340 	if (p->error != 0)	/* lose */
341 		regfree(preg);
342 	return(p->error);
343 }
344 
345 /*
346  - regcomp - interface for parser and compilation
347  = extern int regcomp(regex_t *, const char *, int);
348  = #define	REG_BASIC	0000
349  = #define	REG_EXTENDED	0001
350  = #define	REG_ICASE	0002
351  = #define	REG_NOSUB	0004
352  = #define	REG_NEWLINE	0010
353  = #define	REG_NOSPEC	0020
354  = #define	REG_PEND	0040
355  = #define	REG_DUMP	0200
356  */
357 int				/* 0 success, otherwise REG_something */
regcomp(regex_t * __restrict preg,const char * __restrict pattern,int cflags)358 regcomp(regex_t * __restrict preg,
359 	const char * __restrict pattern,
360 	int cflags)
361 {
362 
363 	return (regcomp_internal(preg, pattern, cflags, 0));
364 }
365 
366 #ifndef LIBREGEX
367 /*
368  * Legacy interface that requires more lax escaping behavior.
369  */
370 int
freebsd12_regcomp(regex_t * __restrict preg,const char * __restrict pattern,int cflags,int pflags)371 freebsd12_regcomp(regex_t * __restrict preg,
372 	const char * __restrict pattern,
373 	int cflags, int pflags)
374 {
375 
376 	return (regcomp_internal(preg, pattern, cflags, PFLAG_LEGACY_ESC));
377 }
378 
379 __sym_compat(regcomp, freebsd12_regcomp, FBSD_1.0);
380 #endif	/* !LIBREGEX */
381 
382 /*
383  - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op,
384  - return whether we should terminate or not
385  == static bool p_ere_exp(struct parse *p);
386  */
387 static bool
p_ere_exp(struct parse * p,struct branchc * bc)388 p_ere_exp(struct parse *p, struct branchc *bc)
389 {
390 	char c;
391 	wint_t wc;
392 	sopno pos;
393 	int count;
394 	int count2;
395 #ifdef LIBREGEX
396 	int i;
397 	int handled;
398 #endif
399 	sopno subno;
400 	int wascaret = 0;
401 
402 	(void)bc;
403 	assert(MORE());		/* caller should have ensured this */
404 	c = GETNEXT();
405 
406 #ifdef LIBREGEX
407 	handled = 0;
408 #endif
409 	pos = HERE();
410 	switch (c) {
411 	case '(':
412 		(void)REQUIRE(MORE(), REG_EPAREN);
413 		p->g->nsub++;
414 		subno = p->g->nsub;
415 		if (subno < NPAREN)
416 			p->pbegin[subno] = HERE();
417 		EMIT(OLPAREN, subno);
418 		if (!SEE(')'))
419 			p_re(p, ')', IGN);
420 		if (subno < NPAREN) {
421 			p->pend[subno] = HERE();
422 			assert(p->pend[subno] != 0);
423 		}
424 		EMIT(ORPAREN, subno);
425 		(void)MUSTEAT(')', REG_EPAREN);
426 		break;
427 #ifndef POSIX_MISTAKE
428 	case ')':		/* happens only if no current unmatched ( */
429 		/*
430 		 * You may ask, why the ifndef?  Because I didn't notice
431 		 * this until slightly too late for 1003.2, and none of the
432 		 * other 1003.2 regular-expression reviewers noticed it at
433 		 * all.  So an unmatched ) is legal POSIX, at least until
434 		 * we can get it fixed.
435 		 */
436 		SETERROR(REG_EPAREN);
437 		break;
438 #endif
439 	case '^':
440 		EMIT(OBOL, 0);
441 		p->g->iflags |= USEBOL;
442 		p->g->nbol++;
443 		wascaret = 1;
444 		break;
445 	case '$':
446 		EMIT(OEOL, 0);
447 		p->g->iflags |= USEEOL;
448 		p->g->neol++;
449 		break;
450 	case '|':
451 		SETERROR(REG_EMPTY);
452 		break;
453 	case '*':
454 	case '+':
455 	case '?':
456 #ifndef NO_STRICT_REGEX
457 	case '{':
458 #endif
459 		SETERROR(REG_BADRPT);
460 		break;
461 	case '.':
462 		if (p->g->cflags&REG_NEWLINE)
463 			nonnewline(p);
464 		else
465 			EMIT(OANY, 0);
466 		break;
467 	case '[':
468 		p_bracket(p);
469 		break;
470 	case '\\':
471 		(void)REQUIRE(MORE(), REG_EESCAPE);
472 		wc = WGETNEXT();
473 #ifdef LIBREGEX
474 		if (p->gnuext) {
475 			handled = 1;
476 			switch (wc) {
477 			case '`':
478 				EMIT(OBOS, 0);
479 				break;
480 			case '\'':
481 				EMIT(OEOS, 0);
482 				break;
483 			case 'B':
484 				EMIT(ONWBND, 0);
485 				break;
486 			case 'b':
487 				EMIT(OWBND, 0);
488 				break;
489 			case 'W':
490 			case 'w':
491 			case 'S':
492 			case 's':
493 				p_b_pseudoclass(p, wc);
494 				break;
495 			case '1':
496 			case '2':
497 			case '3':
498 			case '4':
499 			case '5':
500 			case '6':
501 			case '7':
502 			case '8':
503 			case '9':
504 				i = wc - '0';
505 				assert(i < NPAREN);
506 				if (p->pend[i] != 0) {
507 					assert(i <= p->g->nsub);
508 					EMIT(OBACK_, i);
509 					assert(p->pbegin[i] != 0);
510 					assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
511 					assert(OP(p->strip[p->pend[i]]) == ORPAREN);
512 					(void) dupl(p, p->pbegin[i]+1, p->pend[i]);
513 					EMIT(O_BACK, i);
514 				} else
515 					SETERROR(REG_ESUBREG);
516 				p->g->backrefs = 1;
517 				break;
518 			default:
519 				handled = 0;
520 			}
521 			/* Don't proceed to the POSIX bits if we've already handled it */
522 			if (handled)
523 				break;
524 		}
525 #endif
526 		switch (wc) {
527 		case '<':
528 			EMIT(OBOW, 0);
529 			break;
530 		case '>':
531 			EMIT(OEOW, 0);
532 			break;
533 		default:
534 			if (may_escape(p, wc))
535 				ordinary(p, wc);
536 			else
537 				SETERROR(REG_EESCAPE);
538 			break;
539 		}
540 		break;
541 #ifdef NO_STRICT_REGEX
542 	case '{':               /* okay as ordinary except if digit follows */
543 	    (void)REQUIRE(!MORE() || !isdigit((uch)PEEK()), REG_BADRPT);
544 	    /* FALLTHROUGH */
545 #endif
546 	default:
547 		if (p->error != 0)
548 			return (false);
549 		p->next--;
550 		wc = WGETNEXT();
551 		ordinary(p, wc);
552 		break;
553 	}
554 
555 	if (!MORE())
556 		return (false);
557 	c = PEEK();
558 	/* we call { a repetition if followed by a digit */
559 	if (!( c == '*' || c == '+' || c == '?' ||
560 #ifdef NO_STRICT_REGEX
561 	       (c == '{' && MORE2() && isdigit((uch)PEEK2()))
562 #else
563 	       c == '{'
564 #endif
565 	       ))
566 		return (false);		/* no repetition, we're done */
567 #ifndef NO_STRICT_REGEX
568 	else if (c == '{')
569 		(void)REQUIRE(MORE2() && \
570 		    (isdigit((uch)PEEK2()) || PEEK2() == ','), REG_BADRPT);
571 #endif
572 	NEXT();
573 
574 	(void)REQUIRE(!wascaret, REG_BADRPT);
575 	switch (c) {
576 	case '*':	/* implemented as +? */
577 		/* this case does not require the (y|) trick, noKLUDGE */
578 		INSERT(OPLUS_, pos);
579 		ASTERN(O_PLUS, pos);
580 		INSERT(OQUEST_, pos);
581 		ASTERN(O_QUEST, pos);
582 		break;
583 	case '+':
584 		INSERT(OPLUS_, pos);
585 		ASTERN(O_PLUS, pos);
586 		break;
587 	case '?':
588 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
589 		INSERT(OCH_, pos);		/* offset slightly wrong */
590 		ASTERN(OOR1, pos);		/* this one's right */
591 		AHEAD(pos);			/* fix the OCH_ */
592 		EMIT(OOR2, 0);			/* offset very wrong... */
593 		AHEAD(THERE());			/* ...so fix it */
594 		ASTERN(O_CH, THERETHERE());
595 		break;
596 	case '{':
597 		count = p_count(p);
598 		if (EAT(',')) {
599 			if (isdigit((uch)PEEK())) {
600 				count2 = p_count(p);
601 				(void)REQUIRE(count <= count2, REG_BADBR);
602 			} else		/* single number with comma */
603 				count2 = INFINITY;
604 		} else		/* just a single number */
605 			count2 = count;
606 		repeat(p, pos, count, count2);
607 		if (!EAT('}')) {	/* error heuristics */
608 			while (MORE() && PEEK() != '}')
609 				NEXT();
610 			(void)REQUIRE(MORE(), REG_EBRACE);
611 			SETERROR(REG_BADBR);
612 		}
613 		break;
614 	}
615 
616 	if (!MORE())
617 		return (false);
618 	c = PEEK();
619 	if (!( c == '*' || c == '+' || c == '?' ||
620 				(c == '{' && MORE2() && isdigit((uch)PEEK2())) ) )
621 		return (false);
622 	SETERROR(REG_BADRPT);
623 	return (false);
624 }
625 
626 /*
627  - p_str - string (no metacharacters) "parser"
628  == static void p_str(struct parse *p);
629  */
630 static void
p_str(struct parse * p)631 p_str(struct parse *p)
632 {
633 	(void)REQUIRE(MORE(), REG_EMPTY);
634 	while (MORE())
635 		ordinary(p, WGETNEXT());
636 }
637 
638 /*
639  * Eat consecutive branch delimiters for the kind of expression that we are
640  * parsing, return the number of delimiters that we ate.
641  */
642 static int
p_branch_eat_delim(struct parse * p,struct branchc * bc)643 p_branch_eat_delim(struct parse *p, struct branchc *bc)
644 {
645 	int nskip;
646 
647 	(void)bc;
648 	nskip = 0;
649 	while (EATSPEC('|'))
650 		++nskip;
651 	return (nskip);
652 }
653 
654 /*
655  * Insert necessary branch book-keeping operations. This emits a
656  * bogus 'next' offset, since we still have more to parse
657  */
658 static void
p_branch_ins_offset(struct parse * p,struct branchc * bc)659 p_branch_ins_offset(struct parse *p, struct branchc *bc)
660 {
661 
662 	if (bc->nbranch == 0) {
663 		INSERT(OCH_, bc->start);	/* offset is wrong */
664 		bc->fwd = bc->start;
665 		bc->back = bc->start;
666 	}
667 
668 	ASTERN(OOR1, bc->back);
669 	bc->back = THERE();
670 	AHEAD(bc->fwd);			/* fix previous offset */
671 	bc->fwd = HERE();
672 	EMIT(OOR2, 0);			/* offset is very wrong */
673 	++bc->nbranch;
674 }
675 
676 /*
677  * Fix the offset of the tail branch, if we actually had any branches.
678  * This is to correct the bogus placeholder offset that we use.
679  */
680 static void
p_branch_fix_tail(struct parse * p,struct branchc * bc)681 p_branch_fix_tail(struct parse *p, struct branchc *bc)
682 {
683 
684 	/* Fix bogus offset at the tail if we actually have branches */
685 	if (bc->nbranch > 0) {
686 		AHEAD(bc->fwd);
687 		ASTERN(O_CH, bc->back);
688 	}
689 }
690 
691 /*
692  * Signal to the parser that an empty branch has been encountered; this will,
693  * in the future, be used to allow for more permissive behavior with empty
694  * branches. The return value should indicate whether parsing may continue
695  * or not.
696  */
697 static bool
p_branch_empty(struct parse * p,struct branchc * bc)698 p_branch_empty(struct parse *p, struct branchc *bc)
699 {
700 
701 	(void)bc;
702 	SETERROR(REG_EMPTY);
703 	return (false);
704 }
705 
706 /*
707  * Take care of any branching requirements. This includes inserting the
708  * appropriate branching instructions as well as eating all of the branch
709  * delimiters until we either run out of pattern or need to parse more pattern.
710  */
711 static bool
p_branch_do(struct parse * p,struct branchc * bc)712 p_branch_do(struct parse *p, struct branchc *bc)
713 {
714 	int ate = 0;
715 
716 	ate = p_branch_eat_delim(p, bc);
717 	if (ate == 0)
718 		return (false);
719 	else if ((ate > 1 || (bc->outer && !MORE())) && !p_branch_empty(p, bc))
720 		/*
721 		 * Halt parsing only if we have an empty branch and p_branch_empty
722 		 * indicates that we must not continue. In the future, this will not
723 		 * necessarily be an error.
724 		 */
725 		return (false);
726 	p_branch_ins_offset(p, bc);
727 
728 	return (true);
729 }
730 
731 static void
p_bre_pre_parse(struct parse * p,struct branchc * bc)732 p_bre_pre_parse(struct parse *p, struct branchc *bc)
733 {
734 
735 	(void) bc;
736 	/*
737 	 * Does not move cleanly into expression parser because of
738 	 * ordinary interpration of * at the beginning position of
739 	 * an expression.
740 	 */
741 	if (EAT('^')) {
742 		EMIT(OBOL, 0);
743 		p->g->iflags |= USEBOL;
744 		p->g->nbol++;
745 	}
746 }
747 
748 static void
p_bre_post_parse(struct parse * p,struct branchc * bc)749 p_bre_post_parse(struct parse *p, struct branchc *bc)
750 {
751 
752 	/* Expression is terminating due to EOL token */
753 	if (bc->terminate) {
754 		DROP(1);
755 		EMIT(OEOL, 0);
756 		p->g->iflags |= USEEOL;
757 		p->g->neol++;
758 	}
759 }
760 
761 /*
762  - p_re - Top level parser, concatenation and BRE anchoring
763  == static void p_re(struct parse *p, int end1, int end2);
764  * Giving end1 as OUT essentially eliminates the end1/end2 check.
765  *
766  * This implementation is a bit of a kludge, in that a trailing $ is first
767  * taken as an ordinary character and then revised to be an anchor.
768  * The amount of lookahead needed to avoid this kludge is excessive.
769  */
770 static void
p_re(struct parse * p,int end1,int end2)771 p_re(struct parse *p,
772 	int end1,	/* first terminating character */
773 	int end2)	/* second terminating character; ignored for EREs */
774 {
775 	struct branchc bc;
776 
777 	bc.nbranch = 0;
778 	if (end1 == OUT && end2 == OUT)
779 		bc.outer = true;
780 	else
781 		bc.outer = false;
782 #define	SEEEND()	(!p->bre ? SEE(end1) : SEETWO(end1, end2))
783 	for (;;) {
784 		bc.start = HERE();
785 		bc.nchain = 0;
786 		bc.terminate = false;
787 		if (p->pre_parse != NULL)
788 			p->pre_parse(p, &bc);
789 		while (MORE() && (!p->allowbranch || !SEESPEC('|')) && !SEEEND()) {
790 			bc.terminate = p->parse_expr(p, &bc);
791 			++bc.nchain;
792 		}
793 		if (p->post_parse != NULL)
794 			p->post_parse(p, &bc);
795 		(void) REQUIRE(p->gnuext || HERE() != bc.start, REG_EMPTY);
796 #ifdef LIBREGEX
797 		if (HERE() == bc.start && !p_branch_empty(p, &bc))
798 			break;
799 #endif
800 		if (!p->allowbranch)
801 			break;
802 		/*
803 		 * p_branch_do's return value indicates whether we should
804 		 * continue parsing or not. This is both for correctness and
805 		 * a slight optimization, because it will check if we've
806 		 * encountered an empty branch or the end of the string
807 		 * immediately following a branch delimiter.
808 		 */
809 		if (!p_branch_do(p, &bc))
810 			break;
811 	}
812 #undef SEE_END
813 	if (p->allowbranch)
814 		p_branch_fix_tail(p, &bc);
815 	assert(!MORE() || SEE(end1));
816 }
817 
818 /*
819  - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
820  == static bool p_simp_re(struct parse *p, struct branchc *bc);
821  */
822 static bool			/* was the simple RE an unbackslashed $? */
p_simp_re(struct parse * p,struct branchc * bc)823 p_simp_re(struct parse *p, struct branchc *bc)
824 {
825 	int c;
826 	int cc;			/* convenient/control character */
827 	int count;
828 	int count2;
829 	sopno pos;
830 	bool handled;
831 	int i;
832 	wint_t wc;
833 	sopno subno;
834 #	define	BACKSL	(1<<CHAR_BIT)
835 
836 	pos = HERE();		/* repetition op, if any, covers from here */
837 	handled = false;
838 
839 	assert(MORE());		/* caller should have ensured this */
840 	c = (uch)GETNEXT();
841 	if (c == '\\') {
842 		(void)REQUIRE(MORE(), REG_EESCAPE);
843 		cc = (uch)GETNEXT();
844 		c = BACKSL | cc;
845 #ifdef LIBREGEX
846 		if (p->gnuext) {
847 			handled = true;
848 			switch (c) {
849 			case BACKSL|'`':
850 				EMIT(OBOS, 0);
851 				break;
852 			case BACKSL|'\'':
853 				EMIT(OEOS, 0);
854 				break;
855 			case BACKSL|'B':
856 				EMIT(ONWBND, 0);
857 				break;
858 			case BACKSL|'b':
859 				EMIT(OWBND, 0);
860 				break;
861 			case BACKSL|'W':
862 			case BACKSL|'w':
863 			case BACKSL|'S':
864 			case BACKSL|'s':
865 				p_b_pseudoclass(p, cc);
866 				break;
867 			default:
868 				handled = false;
869 			}
870 		}
871 #endif
872 	}
873 	if (!handled) {
874 		switch (c) {
875 		case '.':
876 			if (p->g->cflags&REG_NEWLINE)
877 				nonnewline(p);
878 			else
879 				EMIT(OANY, 0);
880 			break;
881 		case '[':
882 			p_bracket(p);
883 			break;
884 		case BACKSL|'<':
885 			EMIT(OBOW, 0);
886 			break;
887 		case BACKSL|'>':
888 			EMIT(OEOW, 0);
889 			break;
890 		case BACKSL|'{':
891 			SETERROR(REG_BADRPT);
892 			break;
893 		case BACKSL|'(':
894 			p->g->nsub++;
895 			subno = p->g->nsub;
896 			if (subno < NPAREN)
897 				p->pbegin[subno] = HERE();
898 			EMIT(OLPAREN, subno);
899 			/* the MORE here is an error heuristic */
900 			if (MORE() && !SEETWO('\\', ')'))
901 				p_re(p, '\\', ')');
902 			if (subno < NPAREN) {
903 				p->pend[subno] = HERE();
904 				assert(p->pend[subno] != 0);
905 			}
906 			EMIT(ORPAREN, subno);
907 			(void)REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
908 			break;
909 		case BACKSL|')':	/* should not get here -- must be user */
910 #ifdef NO_STRICT_REGEX
911 		case BACKSL|'}':
912 #endif
913 			SETERROR(REG_EPAREN);
914 			break;
915 		case BACKSL|'1':
916 		case BACKSL|'2':
917 		case BACKSL|'3':
918 		case BACKSL|'4':
919 		case BACKSL|'5':
920 		case BACKSL|'6':
921 		case BACKSL|'7':
922 		case BACKSL|'8':
923 		case BACKSL|'9':
924 			i = (c&~BACKSL) - '0';
925 			assert(i < NPAREN);
926 			if (p->pend[i] != 0) {
927 				assert(i <= p->g->nsub);
928 				EMIT(OBACK_, i);
929 				assert(p->pbegin[i] != 0);
930 				assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
931 				assert(OP(p->strip[p->pend[i]]) == ORPAREN);
932 				(void) dupl(p, p->pbegin[i]+1, p->pend[i]);
933 				EMIT(O_BACK, i);
934 			} else
935 				SETERROR(REG_ESUBREG);
936 			p->g->backrefs = 1;
937 			break;
938 		case '*':
939 			/*
940 			 * Ordinary if used as the first character beyond BOL anchor of
941 			 * a (sub-)expression, counts as a bad repetition operator if it
942 			 * appears otherwise.
943 			 */
944 			(void)REQUIRE(bc->nchain == 0, REG_BADRPT);
945 			/* FALLTHROUGH */
946 		default:
947 			if (p->error != 0)
948 				return (false);	/* Definitely not $... */
949 			p->next--;
950 			wc = WGETNEXT();
951 			if ((c & BACKSL) == 0 || may_escape(p, wc))
952 				ordinary(p, wc);
953 			else
954 				SETERROR(REG_EESCAPE);
955 			break;
956 		}
957 	}
958 
959 	if (EAT('*')) {		/* implemented as +? */
960 		/* this case does not require the (y|) trick, noKLUDGE */
961 		INSERT(OPLUS_, pos);
962 		ASTERN(O_PLUS, pos);
963 		INSERT(OQUEST_, pos);
964 		ASTERN(O_QUEST, pos);
965 #ifdef LIBREGEX
966 	} else if (p->gnuext && EATTWO('\\', '?')) {
967 		INSERT(OQUEST_, pos);
968 		ASTERN(O_QUEST, pos);
969 	} else if (p->gnuext && EATTWO('\\', '+')) {
970 		INSERT(OPLUS_, pos);
971 		ASTERN(O_PLUS, pos);
972 #endif
973 	} else if (EATTWO('\\', '{')) {
974 		count = p_count(p);
975 		if (EAT(',')) {
976 			if (MORE() && isdigit((uch)PEEK())) {
977 				count2 = p_count(p);
978 				(void)REQUIRE(count <= count2, REG_BADBR);
979 			} else		/* single number with comma */
980 				count2 = INFINITY;
981 		} else		/* just a single number */
982 			count2 = count;
983 		repeat(p, pos, count, count2);
984 		if (!EATTWO('\\', '}')) {	/* error heuristics */
985 			while (MORE() && !SEETWO('\\', '}'))
986 				NEXT();
987 			(void)REQUIRE(MORE(), REG_EBRACE);
988 			SETERROR(REG_BADBR);
989 		}
990 	} else if (c == '$')     /* $ (but not \$) ends it */
991 		return (true);
992 
993 	return (false);
994 }
995 
996 /*
997  - p_count - parse a repetition count
998  == static int p_count(struct parse *p);
999  */
1000 static int			/* the value */
p_count(struct parse * p)1001 p_count(struct parse *p)
1002 {
1003 	int count = 0;
1004 	int ndigits = 0;
1005 
1006 	while (MORE() && isdigit((uch)PEEK()) && count <= DUPMAX) {
1007 		count = count*10 + ((uch)GETNEXT() - '0');
1008 		ndigits++;
1009 	}
1010 
1011 	(void)REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
1012 	return(count);
1013 }
1014 
1015 /*
1016  - p_bracket - parse a bracketed character list
1017  == static void p_bracket(struct parse *p);
1018  */
1019 static void
p_bracket(struct parse * p)1020 p_bracket(struct parse *p)
1021 {
1022 	cset *cs;
1023 	wint_t ch;
1024 
1025 	/* Dept of Truly Sickening Special-Case Kludges */
1026 	if (p->end - p->next > 5) {
1027 		if (strncmp(p->next, "[:<:]]", 6) == 0) {
1028 			EMIT(OBOW, 0);
1029 			NEXTn(6);
1030 			return;
1031 		}
1032 		if (strncmp(p->next, "[:>:]]", 6) == 0) {
1033 			EMIT(OEOW, 0);
1034 			NEXTn(6);
1035 			return;
1036 		}
1037 	}
1038 
1039 	if ((cs = allocset(p)) == NULL)
1040 		return;
1041 
1042 	if (p->g->cflags&REG_ICASE)
1043 		cs->icase = 1;
1044 	if (EAT('^'))
1045 		cs->invert = 1;
1046 	if (EAT(']'))
1047 		CHadd(p, cs, ']');
1048 	else if (EAT('-'))
1049 		CHadd(p, cs, '-');
1050 	while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
1051 		p_b_term(p, cs);
1052 	if (EAT('-'))
1053 		CHadd(p, cs, '-');
1054 	(void)MUSTEAT(']', REG_EBRACK);
1055 
1056 	if (p->error != 0)	/* don't mess things up further */
1057 		return;
1058 
1059 	if (cs->invert && p->g->cflags&REG_NEWLINE)
1060 		cs->bmp['\n' >> 3] |= 1 << ('\n' & 7);
1061 
1062 	if ((ch = singleton(cs)) != OUT) {	/* optimize singleton sets */
1063 		ordinary(p, ch);
1064 		freeset(p, cs);
1065 	} else
1066 		EMIT(OANYOF, (int)(cs - p->g->sets));
1067 }
1068 
1069 static int
p_range_cmp(wchar_t c1,wchar_t c2)1070 p_range_cmp(wchar_t c1, wchar_t c2)
1071 {
1072 #ifndef LIBREGEX
1073 	return __wcollate_range_cmp(c1, c2);
1074 #else
1075 	/* Copied from libc/collate __wcollate_range_cmp */
1076 	wchar_t s1[2], s2[2];
1077 
1078 	s1[0] = c1;
1079 	s1[1] = L'\0';
1080 	s2[0] = c2;
1081 	s2[1] = L'\0';
1082 	return (wcscoll(s1, s2));
1083 #endif
1084 }
1085 
1086 /*
1087  - p_b_term - parse one term of a bracketed character list
1088  == static void p_b_term(struct parse *p, cset *cs);
1089  */
1090 static void
p_b_term(struct parse * p,cset * cs)1091 p_b_term(struct parse *p, cset *cs)
1092 {
1093 	char c;
1094 	wint_t start, finish;
1095 	wint_t i;
1096 #ifndef LIBREGEX
1097 	struct xlocale_collate *table =
1098 		(struct xlocale_collate*)__get_locale()->components[XLC_COLLATE];
1099 #endif
1100 	/* classify what we've got */
1101 	switch ((MORE()) ? PEEK() : '\0') {
1102 	case '[':
1103 		c = (MORE2()) ? PEEK2() : '\0';
1104 		break;
1105 	case '-':
1106 		SETERROR(REG_ERANGE);
1107 		return;			/* NOTE RETURN */
1108 	default:
1109 		c = '\0';
1110 		break;
1111 	}
1112 
1113 	switch (c) {
1114 	case ':':		/* character class */
1115 		NEXT2();
1116 		(void)REQUIRE(MORE(), REG_EBRACK);
1117 		c = PEEK();
1118 		(void)REQUIRE(c != '-' && c != ']', REG_ECTYPE);
1119 		p_b_cclass(p, cs);
1120 		(void)REQUIRE(MORE(), REG_EBRACK);
1121 		(void)REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
1122 		break;
1123 	case '=':		/* equivalence class */
1124 		NEXT2();
1125 		(void)REQUIRE(MORE(), REG_EBRACK);
1126 		c = PEEK();
1127 		(void)REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
1128 		p_b_eclass(p, cs);
1129 		(void)REQUIRE(MORE(), REG_EBRACK);
1130 		(void)REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
1131 		break;
1132 	default:		/* symbol, ordinary character, or range */
1133 		start = p_b_symbol(p);
1134 		if (SEE('-') && MORE2() && PEEK2() != ']') {
1135 			/* range */
1136 			NEXT();
1137 			if (EAT('-'))
1138 				finish = '-';
1139 			else
1140 				finish = p_b_symbol(p);
1141 		} else
1142 			finish = start;
1143 		if (start == finish)
1144 			CHadd(p, cs, start);
1145 		else {
1146 #ifndef LIBREGEX
1147 			if (table->__collate_load_error || MB_CUR_MAX > 1) {
1148 #else
1149 			if (MB_CUR_MAX > 1) {
1150 #endif
1151 				(void)REQUIRE(start <= finish, REG_ERANGE);
1152 				CHaddrange(p, cs, start, finish);
1153 			} else {
1154 				(void)REQUIRE(p_range_cmp(start, finish) <= 0, REG_ERANGE);
1155 				for (i = 0; i <= UCHAR_MAX; i++) {
1156 					if (p_range_cmp(start, i) <= 0 &&
1157 					    p_range_cmp(i, finish) <= 0 )
1158 						CHadd(p, cs, i);
1159 				}
1160 			}
1161 		}
1162 		break;
1163 	}
1164 }
1165 
1166 /*
1167  - p_b_pseudoclass - parse a pseudo-class (\w, \W, \s, \S)
1168  == static int p_b_pseudoclass(struct parse *p, char c)
1169  */
1170 static int
1171 p_b_pseudoclass(struct parse *p, char c) {
1172 	cset *cs;
1173 
1174 	if ((cs = allocset(p)) == NULL)
1175 		return(0);
1176 
1177 	if (p->g->cflags&REG_ICASE)
1178 		cs->icase = 1;
1179 
1180 	switch (c) {
1181 	case 'W':
1182 		cs->invert = 1;
1183 		/* PASSTHROUGH */
1184 	case 'w':
1185 		p_b_cclass_named(p, cs, "alnum");
1186 		break;
1187 	case 'S':
1188 		cs->invert = 1;
1189 		/* PASSTHROUGH */
1190 	case 's':
1191 		p_b_cclass_named(p, cs, "space");
1192 		break;
1193 	default:
1194 		return(0);
1195 	}
1196 
1197 	EMIT(OANYOF, (int)(cs - p->g->sets));
1198 	return(1);
1199 }
1200 
1201 /*
1202  - p_b_cclass - parse a character-class name and deal with it
1203  == static void p_b_cclass(struct parse *p, cset *cs);
1204  */
1205 static void
1206 p_b_cclass(struct parse *p, cset *cs)
1207 {
1208 	const char *sp = p->next;
1209 	size_t len;
1210 	char clname[16];
1211 
1212 	while (MORE() && isalpha((uch)PEEK()))
1213 		NEXT();
1214 	len = p->next - sp;
1215 	if (len >= sizeof(clname) - 1) {
1216 		SETERROR(REG_ECTYPE);
1217 		return;
1218 	}
1219 	memcpy(clname, sp, len);
1220 	clname[len] = '\0';
1221 
1222 	p_b_cclass_named(p, cs, clname);
1223 }
1224 /*
1225  - p_b_cclass_named - deal with a named character class
1226  == static void p_b_cclass_named(struct parse *p, cset *cs, const char []);
1227  */
1228 static void
1229 p_b_cclass_named(struct parse *p, cset *cs, const char clname[]) {
1230 	wctype_t wct;
1231 
1232 	if ((wct = wctype(clname)) == 0) {
1233 		SETERROR(REG_ECTYPE);
1234 		return;
1235 	}
1236 	CHaddtype(p, cs, wct);
1237 }
1238 
1239 /*
1240  - p_b_eclass - parse an equivalence-class name and deal with it
1241  == static void p_b_eclass(struct parse *p, cset *cs);
1242  *
1243  * This implementation is incomplete. xxx
1244  */
1245 static void
1246 p_b_eclass(struct parse *p, cset *cs)
1247 {
1248 	wint_t c;
1249 
1250 	c = p_b_coll_elem(p, '=');
1251 	CHadd(p, cs, c);
1252 }
1253 
1254 /*
1255  - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
1256  == static wint_t p_b_symbol(struct parse *p);
1257  */
1258 static wint_t			/* value of symbol */
1259 p_b_symbol(struct parse *p)
1260 {
1261 	wint_t value;
1262 
1263 	(void)REQUIRE(MORE(), REG_EBRACK);
1264 	if (!EATTWO('[', '.'))
1265 		return(WGETNEXT());
1266 
1267 	/* collating symbol */
1268 	value = p_b_coll_elem(p, '.');
1269 	(void)REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
1270 	return(value);
1271 }
1272 
1273 /*
1274  - p_b_coll_elem - parse a collating-element name and look it up
1275  == static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
1276  */
1277 static wint_t			/* value of collating element */
1278 p_b_coll_elem(struct parse *p,
1279 	wint_t endc)		/* name ended by endc,']' */
1280 {
1281 	const char *sp = p->next;
1282 	struct cname *cp;
1283 	mbstate_t mbs;
1284 	wchar_t wc;
1285 	size_t clen, len;
1286 
1287 	while (MORE() && !SEETWO(endc, ']'))
1288 		NEXT();
1289 	if (!MORE()) {
1290 		SETERROR(REG_EBRACK);
1291 		return(0);
1292 	}
1293 	len = p->next - sp;
1294 	for (cp = cnames; cp->name != NULL; cp++)
1295 		if (strncmp(cp->name, sp, len) == 0 && strlen(cp->name) == len)
1296 			return(cp->code);	/* known name */
1297 	memset(&mbs, 0, sizeof(mbs));
1298 	if ((clen = mbrtowc(&wc, sp, len, &mbs)) == len)
1299 		return (wc);			/* single character */
1300 	else if (clen == (size_t)-1 || clen == (size_t)-2)
1301 		SETERROR(REG_ILLSEQ);
1302 	else
1303 		SETERROR(REG_ECOLLATE);		/* neither */
1304 	return(0);
1305 }
1306 
1307 /*
1308  - may_escape - determine whether 'ch' is escape-able in the current context
1309  == static int may_escape(struct parse *p, const wint_t ch)
1310  */
1311 static bool
1312 may_escape(struct parse *p, const wint_t ch)
1313 {
1314 
1315 	if ((p->pflags & PFLAG_LEGACY_ESC) != 0)
1316 		return (true);
1317 	if (iswalpha(ch) || ch == '\'' || ch == '`')
1318 		return (false);
1319 	return (true);
1320 #ifdef NOTYET
1321 	/*
1322 	 * Build a whitelist of characters that may be escaped to produce an
1323 	 * ordinary in the current context. This assumes that these have not
1324 	 * been otherwise interpreted as a special character. Escaping an
1325 	 * ordinary character yields undefined results according to
1326 	 * IEEE 1003.1-2008. Some extensions (notably, some GNU extensions) take
1327 	 * advantage of this and use escaped ordinary characters to provide
1328 	 * special meaning, e.g. \b, \B, \w, \W, \s, \S.
1329 	 */
1330 	switch(ch) {
1331 	case '|':
1332 	case '+':
1333 	case '?':
1334 		/* The above characters may not be escaped in BREs */
1335 		if (!(p->g->cflags&REG_EXTENDED))
1336 			return (false);
1337 		/* Fallthrough */
1338 	case '(':
1339 	case ')':
1340 	case '{':
1341 	case '}':
1342 	case '.':
1343 	case '[':
1344 	case ']':
1345 	case '\\':
1346 	case '*':
1347 	case '^':
1348 	case '$':
1349 		return (true);
1350 	default:
1351 		return (false);
1352 	}
1353 #endif
1354 }
1355 
1356 /*
1357  - othercase - return the case counterpart of an alphabetic
1358  == static wint_t othercase(wint_t ch);
1359  */
1360 static wint_t			/* if no counterpart, return ch */
1361 othercase(wint_t ch)
1362 {
1363 	assert(iswalpha(ch));
1364 	if (iswupper(ch))
1365 		return(towlower(ch));
1366 	else if (iswlower(ch))
1367 		return(towupper(ch));
1368 	else			/* peculiar, but could happen */
1369 		return(ch);
1370 }
1371 
1372 /*
1373  - bothcases - emit a dualcase version of a two-case character
1374  == static void bothcases(struct parse *p, wint_t ch);
1375  *
1376  * Boy, is this implementation ever a kludge...
1377  */
1378 static void
1379 bothcases(struct parse *p, wint_t ch)
1380 {
1381 	const char *oldnext = p->next;
1382 	const char *oldend = p->end;
1383 	char bracket[3 + MB_LEN_MAX];
1384 	size_t n;
1385 	mbstate_t mbs;
1386 
1387 	assert(othercase(ch) != ch);	/* p_bracket() would recurse */
1388 	p->next = bracket;
1389 	memset(&mbs, 0, sizeof(mbs));
1390 	n = wcrtomb(bracket, ch, &mbs);
1391 	assert(n != (size_t)-1);
1392 	bracket[n] = ']';
1393 	bracket[n + 1] = '\0';
1394 	p->end = bracket+n+1;
1395 	p_bracket(p);
1396 	assert(p->next == p->end);
1397 	p->next = oldnext;
1398 	p->end = oldend;
1399 }
1400 
1401 /*
1402  - ordinary - emit an ordinary character
1403  == static void ordinary(struct parse *p, wint_t ch);
1404  */
1405 static void
1406 ordinary(struct parse *p, wint_t ch)
1407 {
1408 	cset *cs;
1409 
1410 	if ((p->g->cflags&REG_ICASE) && iswalpha(ch) && othercase(ch) != ch)
1411 		bothcases(p, ch);
1412 	else if ((ch & OPDMASK) == ch)
1413 		EMIT(OCHAR, ch);
1414 	else {
1415 		/*
1416 		 * Kludge: character is too big to fit into an OCHAR operand.
1417 		 * Emit a singleton set.
1418 		 */
1419 		if ((cs = allocset(p)) == NULL)
1420 			return;
1421 		CHadd(p, cs, ch);
1422 		EMIT(OANYOF, (int)(cs - p->g->sets));
1423 	}
1424 }
1425 
1426 /*
1427  - nonnewline - emit REG_NEWLINE version of OANY
1428  == static void nonnewline(struct parse *p);
1429  *
1430  * Boy, is this implementation ever a kludge...
1431  */
1432 static void
1433 nonnewline(struct parse *p)
1434 {
1435 	const char *oldnext = p->next;
1436 	const char *oldend = p->end;
1437 	char bracket[4];
1438 
1439 	p->next = bracket;
1440 	p->end = bracket+3;
1441 	bracket[0] = '^';
1442 	bracket[1] = '\n';
1443 	bracket[2] = ']';
1444 	bracket[3] = '\0';
1445 	p_bracket(p);
1446 	assert(p->next == bracket+3);
1447 	p->next = oldnext;
1448 	p->end = oldend;
1449 }
1450 
1451 /*
1452  - repeat - generate code for a bounded repetition, recursively if needed
1453  == static void repeat(struct parse *p, sopno start, int from, int to);
1454  */
1455 static void
1456 repeat(struct parse *p,
1457 	sopno start,		/* operand from here to end of strip */
1458 	int from,		/* repeated from this number */
1459 	int to)			/* to this number of times (maybe INFINITY) */
1460 {
1461 	sopno finish = HERE();
1462 #	define	N	2
1463 #	define	INF	3
1464 #	define	REP(f, t)	((f)*8 + (t))
1465 #	define	MAP(n)	(((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1466 	sopno copy;
1467 
1468 	if (p->error != 0)	/* head off possible runaway recursion */
1469 		return;
1470 
1471 	assert(from <= to);
1472 
1473 	switch (REP(MAP(from), MAP(to))) {
1474 	case REP(0, 0):			/* must be user doing this */
1475 		DROP(finish-start);	/* drop the operand */
1476 		break;
1477 	case REP(0, 1):			/* as x{1,1}? */
1478 	case REP(0, N):			/* as x{1,n}? */
1479 	case REP(0, INF):		/* as x{1,}? */
1480 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1481 		INSERT(OCH_, start);		/* offset is wrong... */
1482 		repeat(p, start+1, 1, to);
1483 		ASTERN(OOR1, start);
1484 		AHEAD(start);			/* ... fix it */
1485 		EMIT(OOR2, 0);
1486 		AHEAD(THERE());
1487 		ASTERN(O_CH, THERETHERE());
1488 		break;
1489 	case REP(1, 1):			/* trivial case */
1490 		/* done */
1491 		break;
1492 	case REP(1, N):			/* as x?x{1,n-1} */
1493 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1494 		INSERT(OCH_, start);
1495 		ASTERN(OOR1, start);
1496 		AHEAD(start);
1497 		EMIT(OOR2, 0);			/* offset very wrong... */
1498 		AHEAD(THERE());			/* ...so fix it */
1499 		ASTERN(O_CH, THERETHERE());
1500 		copy = dupl(p, start+1, finish+1);
1501 		assert(copy == finish+4);
1502 		repeat(p, copy, 1, to-1);
1503 		break;
1504 	case REP(1, INF):		/* as x+ */
1505 		INSERT(OPLUS_, start);
1506 		ASTERN(O_PLUS, start);
1507 		break;
1508 	case REP(N, N):			/* as xx{m-1,n-1} */
1509 		copy = dupl(p, start, finish);
1510 		repeat(p, copy, from-1, to-1);
1511 		break;
1512 	case REP(N, INF):		/* as xx{n-1,INF} */
1513 		copy = dupl(p, start, finish);
1514 		repeat(p, copy, from-1, to);
1515 		break;
1516 	default:			/* "can't happen" */
1517 		SETERROR(REG_ASSERT);	/* just in case */
1518 		break;
1519 	}
1520 }
1521 
1522 /*
1523  - wgetnext - helper function for WGETNEXT() macro. Gets the next wide
1524  - character from the parse struct, signals a REG_ILLSEQ error if the
1525  - character can't be converted. Returns the number of bytes consumed.
1526  */
1527 static wint_t
1528 wgetnext(struct parse *p)
1529 {
1530 	mbstate_t mbs;
1531 	wchar_t wc;
1532 	size_t n;
1533 
1534 	memset(&mbs, 0, sizeof(mbs));
1535 	n = mbrtowc(&wc, p->next, p->end - p->next, &mbs);
1536 	if (n == (size_t)-1 || n == (size_t)-2) {
1537 		SETERROR(REG_ILLSEQ);
1538 		return (0);
1539 	}
1540 	if (n == 0)
1541 		n = 1;
1542 	p->next += n;
1543 	return (wc);
1544 }
1545 
1546 /*
1547  - seterr - set an error condition
1548  == static int seterr(struct parse *p, int e);
1549  */
1550 static int			/* useless but makes type checking happy */
1551 seterr(struct parse *p, int e)
1552 {
1553 	if (p->error == 0)	/* keep earliest error condition */
1554 		p->error = e;
1555 	p->next = nuls;		/* try to bring things to a halt */
1556 	p->end = nuls;
1557 	return(0);		/* make the return value well-defined */
1558 }
1559 
1560 /*
1561  - allocset - allocate a set of characters for []
1562  == static cset *allocset(struct parse *p);
1563  */
1564 static cset *
1565 allocset(struct parse *p)
1566 {
1567 	cset *cs, *ncs;
1568 
1569 	ncs = reallocarray(p->g->sets, p->g->ncsets + 1, sizeof(*ncs));
1570 	if (ncs == NULL) {
1571 		SETERROR(REG_ESPACE);
1572 		return (NULL);
1573 	}
1574 	p->g->sets = ncs;
1575 	cs = &p->g->sets[p->g->ncsets++];
1576 	memset(cs, 0, sizeof(*cs));
1577 
1578 	return(cs);
1579 }
1580 
1581 /*
1582  - freeset - free a now-unused set
1583  == static void freeset(struct parse *p, cset *cs);
1584  */
1585 static void
1586 freeset(struct parse *p, cset *cs)
1587 {
1588 	cset *top = &p->g->sets[p->g->ncsets];
1589 
1590 	free(cs->wides);
1591 	free(cs->ranges);
1592 	free(cs->types);
1593 	memset(cs, 0, sizeof(*cs));
1594 	if (cs == top-1)	/* recover only the easy case */
1595 		p->g->ncsets--;
1596 }
1597 
1598 /*
1599  - singleton - Determine whether a set contains only one character,
1600  - returning it if so, otherwise returning OUT.
1601  */
1602 static wint_t
1603 singleton(cset *cs)
1604 {
1605 	wint_t i, s, n;
1606 
1607 	/* Exclude the complicated cases we don't want to deal with */
1608 	if (cs->nranges != 0 || cs->ntypes != 0 || cs->icase != 0)
1609 		return (OUT);
1610 
1611 	if (cs->nwides > 1)
1612 		return (OUT);
1613 
1614 	/* Count the number of characters present in the bitmap */
1615 	for (i = n = 0; i < NC; i++)
1616 		if (CHIN(cs, i)) {
1617 			n++;
1618 			s = i;
1619 		}
1620 
1621 	if (n > 1)
1622 		return (OUT);
1623 
1624 	if (n == 1) {
1625 		if (cs->nwides == 0)
1626 			return (s);
1627 		else
1628 			return (OUT);
1629 	}
1630 	if (cs->nwides == 1)
1631 		return (cs->wides[0]);
1632 
1633 	return (OUT);
1634 }
1635 
1636 /*
1637  - CHadd - add character to character set.
1638  */
1639 static void
1640 CHadd(struct parse *p, cset *cs, wint_t ch)
1641 {
1642 	wint_t nch, *newwides;
1643 	assert(ch >= 0);
1644 	if (ch < NC)
1645 		cs->bmp[ch >> 3] |= 1 << (ch & 7);
1646 	else {
1647 		newwides = reallocarray(cs->wides, cs->nwides + 1,
1648 		    sizeof(*cs->wides));
1649 		if (newwides == NULL) {
1650 			SETERROR(REG_ESPACE);
1651 			return;
1652 		}
1653 		cs->wides = newwides;
1654 		cs->wides[cs->nwides++] = ch;
1655 	}
1656 	if (cs->icase) {
1657 		if ((nch = towlower(ch)) < NC)
1658 			cs->bmp[nch >> 3] |= 1 << (nch & 7);
1659 		if ((nch = towupper(ch)) < NC)
1660 			cs->bmp[nch >> 3] |= 1 << (nch & 7);
1661 	}
1662 }
1663 
1664 /*
1665  - CHaddrange - add all characters in the range [min,max] to a character set.
1666  */
1667 static void
1668 CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max)
1669 {
1670 	crange *newranges;
1671 
1672 	for (; min < NC && min <= max; min++)
1673 		CHadd(p, cs, min);
1674 	if (min >= max)
1675 		return;
1676 	newranges = reallocarray(cs->ranges, cs->nranges + 1,
1677 	    sizeof(*cs->ranges));
1678 	if (newranges == NULL) {
1679 		SETERROR(REG_ESPACE);
1680 		return;
1681 	}
1682 	cs->ranges = newranges;
1683 	cs->ranges[cs->nranges].min = min;
1684 	cs->ranges[cs->nranges].max = max;
1685 	cs->nranges++;
1686 }
1687 
1688 /*
1689  - CHaddtype - add all characters of a certain type to a character set.
1690  */
1691 static void
1692 CHaddtype(struct parse *p, cset *cs, wctype_t wct)
1693 {
1694 	wint_t i;
1695 	wctype_t *newtypes;
1696 
1697 	for (i = 0; i < NC; i++)
1698 		if (iswctype(i, wct))
1699 			CHadd(p, cs, i);
1700 	newtypes = reallocarray(cs->types, cs->ntypes + 1,
1701 	    sizeof(*cs->types));
1702 	if (newtypes == NULL) {
1703 		SETERROR(REG_ESPACE);
1704 		return;
1705 	}
1706 	cs->types = newtypes;
1707 	cs->types[cs->ntypes++] = wct;
1708 }
1709 
1710 /*
1711  - dupl - emit a duplicate of a bunch of sops
1712  == static sopno dupl(struct parse *p, sopno start, sopno finish);
1713  */
1714 static sopno			/* start of duplicate */
1715 dupl(struct parse *p,
1716 	sopno start,		/* from here */
1717 	sopno finish)		/* to this less one */
1718 {
1719 	sopno ret = HERE();
1720 	sopno len = finish - start;
1721 
1722 	assert(finish >= start);
1723 	if (len == 0)
1724 		return(ret);
1725 	if (!enlarge(p, p->ssize + len)) /* this many unexpected additions */
1726 		return(ret);
1727 	(void) memcpy((char *)(p->strip + p->slen),
1728 		(char *)(p->strip + start), (size_t)len*sizeof(sop));
1729 	p->slen += len;
1730 	return(ret);
1731 }
1732 
1733 /*
1734  - doemit - emit a strip operator
1735  == static void doemit(struct parse *p, sop op, size_t opnd);
1736  *
1737  * It might seem better to implement this as a macro with a function as
1738  * hard-case backup, but it's just too big and messy unless there are
1739  * some changes to the data structures.  Maybe later.
1740  */
1741 static void
1742 doemit(struct parse *p, sop op, size_t opnd)
1743 {
1744 	/* avoid making error situations worse */
1745 	if (p->error != 0)
1746 		return;
1747 
1748 	/* deal with oversize operands ("can't happen", more or less) */
1749 	assert(opnd < 1<<OPSHIFT);
1750 
1751 	/* deal with undersized strip */
1752 	if (p->slen >= p->ssize)
1753 		if (!enlarge(p, (p->ssize+1) / 2 * 3))	/* +50% */
1754 			return;
1755 
1756 	/* finally, it's all reduced to the easy case */
1757 	p->strip[p->slen++] = SOP(op, opnd);
1758 }
1759 
1760 /*
1761  - doinsert - insert a sop into the strip
1762  == static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
1763  */
1764 static void
1765 doinsert(struct parse *p, sop op, size_t opnd, sopno pos)
1766 {
1767 	sopno sn;
1768 	sop s;
1769 	int i;
1770 
1771 	/* avoid making error situations worse */
1772 	if (p->error != 0)
1773 		return;
1774 
1775 	sn = HERE();
1776 	EMIT(op, opnd);		/* do checks, ensure space */
1777 	assert(HERE() == sn+1);
1778 	s = p->strip[sn];
1779 
1780 	/* adjust paren pointers */
1781 	assert(pos > 0);
1782 	for (i = 1; i < NPAREN; i++) {
1783 		if (p->pbegin[i] >= pos) {
1784 			p->pbegin[i]++;
1785 		}
1786 		if (p->pend[i] >= pos) {
1787 			p->pend[i]++;
1788 		}
1789 	}
1790 
1791 	memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1792 						(HERE()-pos-1)*sizeof(sop));
1793 	p->strip[pos] = s;
1794 }
1795 
1796 /*
1797  - dofwd - complete a forward reference
1798  == static void dofwd(struct parse *p, sopno pos, sop value);
1799  */
1800 static void
1801 dofwd(struct parse *p, sopno pos, sop value)
1802 {
1803 	/* avoid making error situations worse */
1804 	if (p->error != 0)
1805 		return;
1806 
1807 	assert(value < 1<<OPSHIFT);
1808 	p->strip[pos] = OP(p->strip[pos]) | value;
1809 }
1810 
1811 /*
1812  - enlarge - enlarge the strip
1813  == static int enlarge(struct parse *p, sopno size);
1814  */
1815 static int
1816 enlarge(struct parse *p, sopno size)
1817 {
1818 	sop *sp;
1819 
1820 	if (p->ssize >= size)
1821 		return 1;
1822 
1823 	sp = reallocarray(p->strip, size, sizeof(sop));
1824 	if (sp == NULL) {
1825 		SETERROR(REG_ESPACE);
1826 		return 0;
1827 	}
1828 	p->strip = sp;
1829 	p->ssize = size;
1830 	return 1;
1831 }
1832 
1833 /*
1834  - stripsnug - compact the strip
1835  == static void stripsnug(struct parse *p, struct re_guts *g);
1836  */
1837 static void
1838 stripsnug(struct parse *p, struct re_guts *g)
1839 {
1840 	g->nstates = p->slen;
1841 	g->strip = reallocarray((char *)p->strip, p->slen, sizeof(sop));
1842 	if (g->strip == NULL) {
1843 		SETERROR(REG_ESPACE);
1844 		g->strip = p->strip;
1845 	}
1846 }
1847 
1848 /*
1849  - findmust - fill in must and mlen with longest mandatory literal string
1850  == static void findmust(struct parse *p, struct re_guts *g);
1851  *
1852  * This algorithm could do fancy things like analyzing the operands of |
1853  * for common subsequences.  Someday.  This code is simple and finds most
1854  * of the interesting cases.
1855  *
1856  * Note that must and mlen got initialized during setup.
1857  */
1858 static void
1859 findmust(struct parse *p, struct re_guts *g)
1860 {
1861 	sop *scan;
1862 	sop *start = NULL;
1863 	sop *newstart = NULL;
1864 	sopno newlen;
1865 	sop s;
1866 	char *cp;
1867 	int offset;
1868 	char buf[MB_LEN_MAX];
1869 	size_t clen;
1870 	mbstate_t mbs;
1871 
1872 	/* avoid making error situations worse */
1873 	if (p->error != 0)
1874 		return;
1875 
1876 	/*
1877 	 * It's not generally safe to do a ``char'' substring search on
1878 	 * multibyte character strings, but it's safe for at least
1879 	 * UTF-8 (see RFC 3629).
1880 	 */
1881 	if (MB_CUR_MAX > 1 &&
1882 	    strcmp(_CurrentRuneLocale->__encoding, "UTF-8") != 0)
1883 		return;
1884 
1885 	/* find the longest OCHAR sequence in strip */
1886 	newlen = 0;
1887 	offset = 0;
1888 	g->moffset = 0;
1889 	scan = g->strip + 1;
1890 	do {
1891 		s = *scan++;
1892 		switch (OP(s)) {
1893 		case OCHAR:		/* sequence member */
1894 			if (newlen == 0) {		/* new sequence */
1895 				memset(&mbs, 0, sizeof(mbs));
1896 				newstart = scan - 1;
1897 			}
1898 			clen = wcrtomb(buf, OPND(s), &mbs);
1899 			if (clen == (size_t)-1)
1900 				goto toohard;
1901 			newlen += clen;
1902 			break;
1903 		case OPLUS_:		/* things that don't break one */
1904 		case OLPAREN:
1905 		case ORPAREN:
1906 			break;
1907 		case OQUEST_:		/* things that must be skipped */
1908 		case OCH_:
1909 			offset = altoffset(scan, offset);
1910 			scan--;
1911 			do {
1912 				scan += OPND(s);
1913 				s = *scan;
1914 				/* assert() interferes w debug printouts */
1915 				if (OP(s) != (sop)O_QUEST &&
1916 				    OP(s) != (sop)O_CH && OP(s) != (sop)OOR2) {
1917 					g->iflags |= BAD;
1918 					return;
1919 				}
1920 			} while (OP(s) != (sop)O_QUEST && OP(s) != (sop)O_CH);
1921 			/* FALLTHROUGH */
1922 		case OBOW:		/* things that break a sequence */
1923 		case OEOW:
1924 		case OBOL:
1925 		case OEOL:
1926 		case OBOS:
1927 		case OEOS:
1928 		case OWBND:
1929 		case ONWBND:
1930 		case O_QUEST:
1931 		case O_CH:
1932 		case OEND:
1933 			if (newlen > (sopno)g->mlen) {		/* ends one */
1934 				start = newstart;
1935 				g->mlen = newlen;
1936 				if (offset > -1) {
1937 					g->moffset += offset;
1938 					offset = newlen;
1939 				} else
1940 					g->moffset = offset;
1941 			} else {
1942 				if (offset > -1)
1943 					offset += newlen;
1944 			}
1945 			newlen = 0;
1946 			break;
1947 		case OANY:
1948 			if (newlen > (sopno)g->mlen) {		/* ends one */
1949 				start = newstart;
1950 				g->mlen = newlen;
1951 				if (offset > -1) {
1952 					g->moffset += offset;
1953 					offset = newlen;
1954 				} else
1955 					g->moffset = offset;
1956 			} else {
1957 				if (offset > -1)
1958 					offset += newlen;
1959 			}
1960 			if (offset > -1)
1961 				offset++;
1962 			newlen = 0;
1963 			break;
1964 		case OANYOF:		/* may or may not invalidate offset */
1965 			/* First, everything as OANY */
1966 			if (newlen > (sopno)g->mlen) {		/* ends one */
1967 				start = newstart;
1968 				g->mlen = newlen;
1969 				if (offset > -1) {
1970 					g->moffset += offset;
1971 					offset = newlen;
1972 				} else
1973 					g->moffset = offset;
1974 			} else {
1975 				if (offset > -1)
1976 					offset += newlen;
1977 			}
1978 			if (offset > -1)
1979 				offset++;
1980 			newlen = 0;
1981 			break;
1982 		toohard:
1983 		default:
1984 			/* Anything here makes it impossible or too hard
1985 			 * to calculate the offset -- so we give up;
1986 			 * save the last known good offset, in case the
1987 			 * must sequence doesn't occur later.
1988 			 */
1989 			if (newlen > (sopno)g->mlen) {		/* ends one */
1990 				start = newstart;
1991 				g->mlen = newlen;
1992 				if (offset > -1)
1993 					g->moffset += offset;
1994 				else
1995 					g->moffset = offset;
1996 			}
1997 			offset = -1;
1998 			newlen = 0;
1999 			break;
2000 		}
2001 	} while (OP(s) != OEND);
2002 
2003 	if (g->mlen == 0) {		/* there isn't one */
2004 		g->moffset = -1;
2005 		return;
2006 	}
2007 
2008 	/* turn it into a character string */
2009 	g->must = malloc((size_t)g->mlen + 1);
2010 	if (g->must == NULL) {		/* argh; just forget it */
2011 		g->mlen = 0;
2012 		g->moffset = -1;
2013 		return;
2014 	}
2015 	cp = g->must;
2016 	scan = start;
2017 	memset(&mbs, 0, sizeof(mbs));
2018 	while (cp < g->must + g->mlen) {
2019 		while (OP(s = *scan++) != OCHAR)
2020 			continue;
2021 		clen = wcrtomb(cp, OPND(s), &mbs);
2022 		assert(clen != (size_t)-1);
2023 		cp += clen;
2024 	}
2025 	assert(cp == g->must + g->mlen);
2026 	*cp++ = '\0';		/* just on general principles */
2027 }
2028 
2029 /*
2030  - altoffset - choose biggest offset among multiple choices
2031  == static int altoffset(sop *scan, int offset);
2032  *
2033  * Compute, recursively if necessary, the largest offset among multiple
2034  * re paths.
2035  */
2036 static int
2037 altoffset(sop *scan, int offset)
2038 {
2039 	int largest;
2040 	int try;
2041 	sop s;
2042 
2043 	/* If we gave up already on offsets, return */
2044 	if (offset == -1)
2045 		return -1;
2046 
2047 	largest = 0;
2048 	try = 0;
2049 	s = *scan++;
2050 	while (OP(s) != (sop)O_QUEST && OP(s) != (sop)O_CH) {
2051 		switch (OP(s)) {
2052 		case OOR1:
2053 			if (try > largest)
2054 				largest = try;
2055 			try = 0;
2056 			break;
2057 		case OQUEST_:
2058 		case OCH_:
2059 			try = altoffset(scan, try);
2060 			if (try == -1)
2061 				return -1;
2062 			scan--;
2063 			do {
2064 				scan += OPND(s);
2065 				s = *scan;
2066 				if (OP(s) != (sop)O_QUEST &&
2067 				    OP(s) != (sop)O_CH && OP(s) != (sop)OOR2)
2068 					return -1;
2069 			} while (OP(s) != (sop)O_QUEST && OP(s) != (sop)O_CH);
2070 			/* We must skip to the next position, or we'll
2071 			 * leave altoffset() too early.
2072 			 */
2073 			scan++;
2074 			break;
2075 		case OANYOF:
2076 		case OCHAR:
2077 		case OANY:
2078 			try++;
2079 		case OBOW:
2080 		case OEOW:
2081 		case OWBND:
2082 		case ONWBND:
2083 		case OLPAREN:
2084 		case ORPAREN:
2085 		case OOR2:
2086 			break;
2087 		default:
2088 			try = -1;
2089 			break;
2090 		}
2091 		if (try == -1)
2092 			return -1;
2093 		s = *scan++;
2094 	}
2095 
2096 	if (try > largest)
2097 		largest = try;
2098 
2099 	return largest+offset;
2100 }
2101 
2102 /*
2103  - computejumps - compute char jumps for BM scan
2104  == static void computejumps(struct parse *p, struct re_guts *g);
2105  *
2106  * This algorithm assumes g->must exists and is has size greater than
2107  * zero. It's based on the algorithm found on Computer Algorithms by
2108  * Sara Baase.
2109  *
2110  * A char jump is the number of characters one needs to jump based on
2111  * the value of the character from the text that was mismatched.
2112  */
2113 static void
2114 computejumps(struct parse *p, struct re_guts *g)
2115 {
2116 	int ch;
2117 	int mindex;
2118 
2119 	/* Avoid making errors worse */
2120 	if (p->error != 0)
2121 		return;
2122 
2123 	g->charjump = (int *)malloc((NC_MAX + 1) * sizeof(int));
2124 	if (g->charjump == NULL)	/* Not a fatal error */
2125 		return;
2126 	/* Adjust for signed chars, if necessary */
2127 	g->charjump = &g->charjump[-(CHAR_MIN)];
2128 
2129 	/* If the character does not exist in the pattern, the jump
2130 	 * is equal to the number of characters in the pattern.
2131 	 */
2132 	for (ch = CHAR_MIN; ch < (CHAR_MAX + 1); ch++)
2133 		g->charjump[ch] = g->mlen;
2134 
2135 	/* If the character does exist, compute the jump that would
2136 	 * take us to the last character in the pattern equal to it
2137 	 * (notice that we match right to left, so that last character
2138 	 * is the first one that would be matched).
2139 	 */
2140 	for (mindex = 0; mindex < g->mlen; mindex++)
2141 		g->charjump[(int)g->must[mindex]] = g->mlen - mindex - 1;
2142 }
2143 
2144 /*
2145  - computematchjumps - compute match jumps for BM scan
2146  == static void computematchjumps(struct parse *p, struct re_guts *g);
2147  *
2148  * This algorithm assumes g->must exists and is has size greater than
2149  * zero. It's based on the algorithm found on Computer Algorithms by
2150  * Sara Baase.
2151  *
2152  * A match jump is the number of characters one needs to advance based
2153  * on the already-matched suffix.
2154  * Notice that all values here are minus (g->mlen-1), because of the way
2155  * the search algorithm works.
2156  */
2157 static void
2158 computematchjumps(struct parse *p, struct re_guts *g)
2159 {
2160 	int mindex;		/* General "must" iterator */
2161 	int suffix;		/* Keeps track of matching suffix */
2162 	int ssuffix;		/* Keeps track of suffixes' suffix */
2163 	int* pmatches;		/* pmatches[k] points to the next i
2164 				 * such that i+1...mlen is a substring
2165 				 * of k+1...k+mlen-i-1
2166 				 */
2167 
2168 	/* Avoid making errors worse */
2169 	if (p->error != 0)
2170 		return;
2171 
2172 	pmatches = (int*) malloc(g->mlen * sizeof(int));
2173 	if (pmatches == NULL) {
2174 		g->matchjump = NULL;
2175 		return;
2176 	}
2177 
2178 	g->matchjump = (int*) malloc(g->mlen * sizeof(int));
2179 	if (g->matchjump == NULL) {	/* Not a fatal error */
2180 		free(pmatches);
2181 		return;
2182 	}
2183 
2184 	/* Set maximum possible jump for each character in the pattern */
2185 	for (mindex = 0; mindex < g->mlen; mindex++)
2186 		g->matchjump[mindex] = 2*g->mlen - mindex - 1;
2187 
2188 	/* Compute pmatches[] */
2189 	for (mindex = g->mlen - 1, suffix = g->mlen; mindex >= 0;
2190 	    mindex--, suffix--) {
2191 		pmatches[mindex] = suffix;
2192 
2193 		/* If a mismatch is found, interrupting the substring,
2194 		 * compute the matchjump for that position. If no
2195 		 * mismatch is found, then a text substring mismatched
2196 		 * against the suffix will also mismatch against the
2197 		 * substring.
2198 		 */
2199 		while (suffix < g->mlen
2200 		    && g->must[mindex] != g->must[suffix]) {
2201 			g->matchjump[suffix] = MIN(g->matchjump[suffix],
2202 			    g->mlen - mindex - 1);
2203 			suffix = pmatches[suffix];
2204 		}
2205 	}
2206 
2207 	/* Compute the matchjump up to the last substring found to jump
2208 	 * to the beginning of the largest must pattern prefix matching
2209 	 * it's own suffix.
2210 	 */
2211 	for (mindex = 0; mindex <= suffix; mindex++)
2212 		g->matchjump[mindex] = MIN(g->matchjump[mindex],
2213 		    g->mlen + suffix - mindex);
2214 
2215         ssuffix = pmatches[suffix];
2216         while (suffix < g->mlen) {
2217                 while (suffix <= ssuffix && suffix < g->mlen) {
2218                         g->matchjump[suffix] = MIN(g->matchjump[suffix],
2219 			    g->mlen + ssuffix - suffix);
2220                         suffix++;
2221                 }
2222 		if (suffix < g->mlen)
2223                 	ssuffix = pmatches[ssuffix];
2224         }
2225 
2226 	free(pmatches);
2227 }
2228 
2229 /*
2230  - pluscount - count + nesting
2231  == static sopno pluscount(struct parse *p, struct re_guts *g);
2232  */
2233 static sopno			/* nesting depth */
2234 pluscount(struct parse *p, struct re_guts *g)
2235 {
2236 	sop *scan;
2237 	sop s;
2238 	sopno plusnest = 0;
2239 	sopno maxnest = 0;
2240 
2241 	if (p->error != 0)
2242 		return(0);	/* there may not be an OEND */
2243 
2244 	scan = g->strip + 1;
2245 	do {
2246 		s = *scan++;
2247 		switch (OP(s)) {
2248 		case OPLUS_:
2249 			plusnest++;
2250 			break;
2251 		case O_PLUS:
2252 			if (plusnest > maxnest)
2253 				maxnest = plusnest;
2254 			plusnest--;
2255 			break;
2256 		}
2257 	} while (OP(s) != OEND);
2258 	if (plusnest != 0)
2259 		g->iflags |= BAD;
2260 	return(maxnest);
2261 }
2262