Home | History | Annotate | Line # | Download | only in regex
regcomp.c revision 1.9
      1 /*	$NetBSD: regcomp.c,v 1.9 1998/02/03 18:38:13 perry Exp $	*/
      2 
      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  * This code is derived from software contributed to Berkeley by
      9  * Henry Spencer.
     10  *
     11  * Redistribution and use in source and binary forms, with or without
     12  * modification, are permitted provided that the following conditions
     13  * are met:
     14  * 1. Redistributions of source code must retain the above copyright
     15  *    notice, this list of conditions and the following disclaimer.
     16  * 2. Redistributions in binary form must reproduce the above copyright
     17  *    notice, this list of conditions and the following disclaimer in the
     18  *    documentation and/or other materials provided with the distribution.
     19  * 3. All advertising materials mentioning features or use of this software
     20  *    must display the following acknowledgement:
     21  *	This product includes software developed by the University of
     22  *	California, Berkeley and its contributors.
     23  * 4. Neither the name of the University nor the names of its contributors
     24  *    may be used to endorse or promote products derived from this software
     25  *    without specific prior written permission.
     26  *
     27  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     29  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     30  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     31  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     32  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     33  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     34  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     35  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     36  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     37  * SUCH DAMAGE.
     38  *
     39  *	@(#)regcomp.c	8.5 (Berkeley) 3/20/94
     40  */
     41 
     42 #include <sys/cdefs.h>
     43 #if defined(LIBC_SCCS) && !defined(lint)
     44 #if 0
     45 static char sccsid[] = "@(#)regcomp.c	8.5 (Berkeley) 3/20/94";
     46 #else
     47 __RCSID("$NetBSD: regcomp.c,v 1.9 1998/02/03 18:38:13 perry Exp $");
     48 #endif
     49 #endif /* LIBC_SCCS and not lint */
     50 
     51 #include "namespace.h"
     52 #include <sys/types.h>
     53 #include <stdio.h>
     54 #include <string.h>
     55 #include <ctype.h>
     56 #include <limits.h>
     57 #include <stdlib.h>
     58 #include <regex.h>
     59 
     60 #ifdef __weak_alias
     61 __weak_alias(regcomp,_regcomp);
     62 #endif
     63 
     64 #include "utils.h"
     65 #include "regex2.h"
     66 
     67 #include "cclass.h"
     68 #include "cname.h"
     69 
     70 /*
     71  * parse structure, passed up and down to avoid global variables and
     72  * other clumsinesses
     73  */
     74 struct parse {
     75 	char *next;		/* next character in RE */
     76 	char *end;		/* end of string (-> NUL normally) */
     77 	int error;		/* has an error been seen? */
     78 	sop *strip;		/* malloced strip */
     79 	sopno ssize;		/* malloced strip size (allocated) */
     80 	sopno slen;		/* malloced strip length (used) */
     81 	int ncsalloc;		/* number of csets allocated */
     82 	struct re_guts *g;
     83 #	define	NPAREN	10	/* we need to remember () 1-9 for back refs */
     84 	sopno pbegin[NPAREN];	/* -> ( ([0] unused) */
     85 	sopno pend[NPAREN];	/* -> ) ([0] unused) */
     86 };
     87 
     88 /* ========= begin header generated by ./mkh ========= */
     89 #ifdef __cplusplus
     90 extern "C" {
     91 #endif
     92 
     93 /* === regcomp.c === */
     94 static void p_ere __P((struct parse *p, int stop));
     95 static void p_ere_exp __P((struct parse *p));
     96 static void p_str __P((struct parse *p));
     97 static void p_bre __P((struct parse *p, int end1, int end2));
     98 static int p_simp_re __P((struct parse *p, int starordinary));
     99 static int p_count __P((struct parse *p));
    100 static void p_bracket __P((struct parse *p));
    101 static void p_b_term __P((struct parse *p, cset *cs));
    102 static void p_b_cclass __P((struct parse *p, cset *cs));
    103 static void p_b_eclass __P((struct parse *p, cset *cs));
    104 static char p_b_symbol __P((struct parse *p));
    105 static char p_b_coll_elem __P((struct parse *p, int endc));
    106 static char othercase __P((int ch));
    107 static void bothcases __P((struct parse *p, int ch));
    108 static void ordinary __P((struct parse *p, int ch));
    109 static void nonnewline __P((struct parse *p));
    110 static void repeat __P((struct parse *p, sopno start, int from, int to));
    111 static int seterr __P((struct parse *p, int e));
    112 static cset *allocset __P((struct parse *p));
    113 static void freeset __P((struct parse *p, cset *cs));
    114 static int freezeset __P((struct parse *p, cset *cs));
    115 static int firstch __P((struct parse *p, cset *cs));
    116 static int nch __P((struct parse *p, cset *cs));
    117 static void mcadd __P((struct parse *p, cset *cs, char *cp));
    118 #if 0
    119 static void mcsub __P((cset *cs, char *cp));
    120 static int mcin __P((cset *cs, char *cp));
    121 static char *mcfind __P((cset *cs, char *cp));
    122 #endif
    123 static void mcinvert __P((struct parse *p, cset *cs));
    124 static void mccase __P((struct parse *p, cset *cs));
    125 static int isinsets __P((struct re_guts *g, int c));
    126 static int samesets __P((struct re_guts *g, int c1, int c2));
    127 static void categorize __P((struct parse *p, struct re_guts *g));
    128 static sopno dupl __P((struct parse *p, sopno start, sopno finish));
    129 static void doemit __P((struct parse *p, sop op, size_t opnd));
    130 static void doinsert __P((struct parse *p, sop op, size_t opnd, sopno pos));
    131 static void dofwd __P((struct parse *p, sopno pos, sop value));
    132 static void enlarge __P((struct parse *p, sopno size));
    133 static void stripsnug __P((struct parse *p, struct re_guts *g));
    134 static void findmust __P((struct parse *p, struct re_guts *g));
    135 static sopno pluscount __P((struct parse *p, struct re_guts *g));
    136 
    137 #ifdef __cplusplus
    138 }
    139 #endif
    140 /* ========= end header generated by ./mkh ========= */
    141 
    142 static char nuls[10];		/* place to point scanner in event of error */
    143 
    144 /*
    145  * macros for use with parse structure
    146  * BEWARE:  these know that the parse structure is named `p' !!!
    147  */
    148 #define	PEEK()	(*p->next)
    149 #define	PEEK2()	(*(p->next+1))
    150 #define	MORE()	(p->next < p->end)
    151 #define	MORE2()	(p->next+1 < p->end)
    152 #define	SEE(c)	(MORE() && PEEK() == (c))
    153 #define	SEETWO(a, b)	(MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
    154 #define	EAT(c)	((SEE(c)) ? (NEXT(), 1) : 0)
    155 #define	EATTWO(a, b)	((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
    156 #define	NEXT()	(p->next++)
    157 #define	NEXT2()	(p->next += 2)
    158 #define	NEXTn(n)	(p->next += (n))
    159 #define	GETNEXT()	(*p->next++)
    160 #define	SETERROR(e)	seterr(p, (e))
    161 #define	REQUIRE(co, e)	(void) ((co) || SETERROR(e))
    162 #define	MUSTSEE(c, e)	(REQUIRE(MORE() && PEEK() == (c), e))
    163 #define	MUSTEAT(c, e)	(void) (REQUIRE(MORE() && GETNEXT() == (c), e))
    164 #define	MUSTNOTSEE(c, e)	(REQUIRE(!MORE() || PEEK() != (c), e))
    165 #define	EMIT(op, sopnd)	doemit(p, (sop)(op), (size_t)(sopnd))
    166 #define	INSERT(op, pos)	doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
    167 #define	AHEAD(pos)		dofwd(p, pos, HERE()-(pos))
    168 #define	ASTERN(sop, pos)	EMIT(sop, HERE()-pos)
    169 #define	HERE()		(p->slen)
    170 #define	THERE()		(p->slen - 1)
    171 #define	THERETHERE()	(p->slen - 2)
    172 #define	DROP(n)	(p->slen -= (n))
    173 
    174 #ifndef NDEBUG
    175 static int never = 0;		/* for use in asserts; shuts lint up */
    176 #else
    177 #define	never	0		/* some <assert.h>s have bugs too */
    178 #endif
    179 
    180 /*
    181  - regcomp - interface for parser and compilation
    182  = extern int regcomp(regex_t *, const char *, int);
    183  = #define	REG_BASIC	0000
    184  = #define	REG_EXTENDED	0001
    185  = #define	REG_ICASE	0002
    186  = #define	REG_NOSUB	0004
    187  = #define	REG_NEWLINE	0010
    188  = #define	REG_NOSPEC	0020
    189  = #define	REG_PEND	0040
    190  = #define	REG_DUMP	0200
    191  */
    192 int				/* 0 success, otherwise REG_something */
    193 regcomp(preg, pattern, cflags)
    194 regex_t *preg;
    195 const char *pattern;
    196 int cflags;
    197 {
    198 	struct parse pa;
    199 	struct re_guts *g;
    200 	struct parse *p = &pa;
    201 	int i;
    202 	size_t len;
    203 #ifdef REDEBUG
    204 #	define	GOODFLAGS(f)	(f)
    205 #else
    206 #	define	GOODFLAGS(f)	((f)&~REG_DUMP)
    207 #endif
    208 
    209 	cflags = GOODFLAGS(cflags);
    210 	if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
    211 		return(REG_INVARG);
    212 
    213 	if (cflags&REG_PEND) {
    214 		if (preg->re_endp < pattern)
    215 			return(REG_INVARG);
    216 		len = preg->re_endp - pattern;
    217 	} else
    218 		len = strlen((char *)pattern);
    219 
    220 	/* do the mallocs early so failure handling is easy */
    221 	g = (struct re_guts *)malloc(sizeof(struct re_guts) +
    222 							(NC-1)*sizeof(cat_t));
    223 	if (g == NULL)
    224 		return(REG_ESPACE);
    225 	p->ssize = len/(size_t)2*(size_t)3 + (size_t)1;	/* ugh */
    226 	p->strip = (sop *)malloc(p->ssize * sizeof(sop));
    227 	p->slen = 0;
    228 	if (p->strip == NULL) {
    229 		free((char *)g);
    230 		return(REG_ESPACE);
    231 	}
    232 
    233 	/* set things up */
    234 	p->g = g;
    235 	p->next = (char *)pattern;	/* convenience; we do not modify it */
    236 	p->end = p->next + len;
    237 	p->error = 0;
    238 	p->ncsalloc = 0;
    239 	for (i = 0; i < NPAREN; i++) {
    240 		p->pbegin[i] = 0;
    241 		p->pend[i] = 0;
    242 	}
    243 	g->csetsize = NC;
    244 	g->sets = NULL;
    245 	g->setbits = NULL;
    246 	g->ncsets = 0;
    247 	g->cflags = cflags;
    248 	g->iflags = 0;
    249 	g->nbol = 0;
    250 	g->neol = 0;
    251 	g->must = NULL;
    252 	g->mlen = 0;
    253 	g->nsub = 0;
    254 	g->ncategories = 1;	/* category 0 is "everything else" */
    255 	g->categories = &g->catspace[-(CHAR_MIN)];
    256 	(void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
    257 	g->backrefs = 0;
    258 
    259 	/* do it */
    260 	EMIT(OEND, 0);
    261 	g->firststate = THERE();
    262 	if (cflags&REG_EXTENDED)
    263 		p_ere(p, OUT);
    264 	else if (cflags&REG_NOSPEC)
    265 		p_str(p);
    266 	else
    267 		p_bre(p, OUT, OUT);
    268 	EMIT(OEND, 0);
    269 	g->laststate = THERE();
    270 
    271 	/* tidy up loose ends and fill things in */
    272 	categorize(p, g);
    273 	stripsnug(p, g);
    274 	findmust(p, g);
    275 	g->nplus = pluscount(p, g);
    276 	g->magic = MAGIC2;
    277 	preg->re_nsub = g->nsub;
    278 	preg->re_g = g;
    279 	preg->re_magic = MAGIC1;
    280 #ifndef REDEBUG
    281 	/* not debugging, so can't rely on the assert() in regexec() */
    282 	if (g->iflags&BAD)
    283 		SETERROR(REG_ASSERT);
    284 #endif
    285 
    286 	/* win or lose, we're done */
    287 	if (p->error != 0)	/* lose */
    288 		regfree(preg);
    289 	return(p->error);
    290 }
    291 
    292 /*
    293  - p_ere - ERE parser top level, concatenation and alternation
    294  == static void p_ere(struct parse *p, int stop);
    295  */
    296 static void
    297 p_ere(p, stop)
    298 struct parse *p;
    299 int stop;			/* character this ERE should end at */
    300 {
    301 	char c;
    302 	sopno prevback = 0;	/* pacify gcc */
    303 	sopno prevfwd = 0; 	/* pacify gcc */
    304 	sopno conc;
    305 	int first = 1;		/* is this the first alternative? */
    306 
    307 	for (;;) {
    308 		/* do a bunch of concatenated expressions */
    309 		conc = HERE();
    310 		while (MORE() && (c = PEEK()) != '|' && c != stop)
    311 			p_ere_exp(p);
    312 		REQUIRE(HERE() != conc, REG_EMPTY);	/* require nonempty */
    313 
    314 		if (!EAT('|'))
    315 			break;		/* NOTE BREAK OUT */
    316 
    317 		if (first) {
    318 			INSERT(OCH_, conc);	/* offset is wrong */
    319 			prevfwd = conc;
    320 			prevback = conc;
    321 			first = 0;
    322 		}
    323 		ASTERN(OOR1, prevback);
    324 		prevback = THERE();
    325 		AHEAD(prevfwd);			/* fix previous offset */
    326 		prevfwd = HERE();
    327 		EMIT(OOR2, 0);			/* offset is very wrong */
    328 	}
    329 
    330 	if (!first) {		/* tail-end fixups */
    331 		AHEAD(prevfwd);
    332 		ASTERN(O_CH, prevback);
    333 	}
    334 
    335 	assert(!MORE() || SEE(stop));
    336 }
    337 
    338 /*
    339  - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
    340  == static void p_ere_exp(struct parse *p);
    341  */
    342 static void
    343 p_ere_exp(p)
    344 struct parse *p;
    345 {
    346 	char c;
    347 	sopno pos;
    348 	int count;
    349 	int count2;
    350 	sopno subno;
    351 	int wascaret = 0;
    352 
    353 	assert(MORE());		/* caller should have ensured this */
    354 	c = GETNEXT();
    355 
    356 	pos = HERE();
    357 	switch (c) {
    358 	case '(':
    359 		REQUIRE(MORE(), REG_EPAREN);
    360 		p->g->nsub++;
    361 		subno = p->g->nsub;
    362 		if (subno < NPAREN)
    363 			p->pbegin[subno] = HERE();
    364 		EMIT(OLPAREN, subno);
    365 		if (!SEE(')'))
    366 			p_ere(p, ')');
    367 		if (subno < NPAREN) {
    368 			p->pend[subno] = HERE();
    369 			assert(p->pend[subno] != 0);
    370 		}
    371 		EMIT(ORPAREN, subno);
    372 		MUSTEAT(')', REG_EPAREN);
    373 		break;
    374 #ifndef POSIX_MISTAKE
    375 	case ')':		/* happens only if no current unmatched ( */
    376 		/*
    377 		 * You may ask, why the ifndef?  Because I didn't notice
    378 		 * this until slightly too late for 1003.2, and none of the
    379 		 * other 1003.2 regular-expression reviewers noticed it at
    380 		 * all.  So an unmatched ) is legal POSIX, at least until
    381 		 * we can get it fixed.
    382 		 */
    383 		SETERROR(REG_EPAREN);
    384 		break;
    385 #endif
    386 	case '^':
    387 		EMIT(OBOL, 0);
    388 		p->g->iflags |= USEBOL;
    389 		p->g->nbol++;
    390 		wascaret = 1;
    391 		break;
    392 	case '$':
    393 		EMIT(OEOL, 0);
    394 		p->g->iflags |= USEEOL;
    395 		p->g->neol++;
    396 		break;
    397 	case '|':
    398 		SETERROR(REG_EMPTY);
    399 		break;
    400 	case '*':
    401 	case '+':
    402 	case '?':
    403 		SETERROR(REG_BADRPT);
    404 		break;
    405 	case '.':
    406 		if (p->g->cflags&REG_NEWLINE)
    407 			nonnewline(p);
    408 		else
    409 			EMIT(OANY, 0);
    410 		break;
    411 	case '[':
    412 		p_bracket(p);
    413 		break;
    414 	case '\\':
    415 		REQUIRE(MORE(), REG_EESCAPE);
    416 		c = GETNEXT();
    417 		ordinary(p, c);
    418 		break;
    419 	case '{':		/* okay as ordinary except if digit follows */
    420 		REQUIRE(!MORE() || !isdigit(PEEK()), REG_BADRPT);
    421 		/* FALLTHROUGH */
    422 	default:
    423 		ordinary(p, c);
    424 		break;
    425 	}
    426 
    427 	if (!MORE())
    428 		return;
    429 	c = PEEK();
    430 	/* we call { a repetition if followed by a digit */
    431 	if (!( c == '*' || c == '+' || c == '?' ||
    432 				(c == '{' && MORE2() && isdigit(PEEK2())) ))
    433 		return;		/* no repetition, we're done */
    434 	NEXT();
    435 
    436 	REQUIRE(!wascaret, REG_BADRPT);
    437 	switch (c) {
    438 	case '*':	/* implemented as +? */
    439 		/* this case does not require the (y|) trick, noKLUDGE */
    440 		INSERT(OPLUS_, pos);
    441 		ASTERN(O_PLUS, pos);
    442 		INSERT(OQUEST_, pos);
    443 		ASTERN(O_QUEST, pos);
    444 		break;
    445 	case '+':
    446 		INSERT(OPLUS_, pos);
    447 		ASTERN(O_PLUS, pos);
    448 		break;
    449 	case '?':
    450 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
    451 		INSERT(OCH_, pos);		/* offset slightly wrong */
    452 		ASTERN(OOR1, pos);		/* this one's right */
    453 		AHEAD(pos);			/* fix the OCH_ */
    454 		EMIT(OOR2, 0);			/* offset very wrong... */
    455 		AHEAD(THERE());			/* ...so fix it */
    456 		ASTERN(O_CH, THERETHERE());
    457 		break;
    458 	case '{':
    459 		count = p_count(p);
    460 		if (EAT(',')) {
    461 			if (isdigit(PEEK())) {
    462 				count2 = p_count(p);
    463 				REQUIRE(count <= count2, REG_BADBR);
    464 			} else		/* single number with comma */
    465 				count2 = INFINITY;
    466 		} else		/* just a single number */
    467 			count2 = count;
    468 		repeat(p, pos, count, count2);
    469 		if (!EAT('}')) {	/* error heuristics */
    470 			while (MORE() && PEEK() != '}')
    471 				NEXT();
    472 			REQUIRE(MORE(), REG_EBRACE);
    473 			SETERROR(REG_BADBR);
    474 		}
    475 		break;
    476 	}
    477 
    478 	if (!MORE())
    479 		return;
    480 	c = PEEK();
    481 	if (!( c == '*' || c == '+' || c == '?' ||
    482 				(c == '{' && MORE2() && isdigit(PEEK2())) ) )
    483 		return;
    484 	SETERROR(REG_BADRPT);
    485 }
    486 
    487 /*
    488  - p_str - string (no metacharacters) "parser"
    489  == static void p_str(struct parse *p);
    490  */
    491 static void
    492 p_str(p)
    493 struct parse *p;
    494 {
    495 	REQUIRE(MORE(), REG_EMPTY);
    496 	while (MORE())
    497 		ordinary(p, GETNEXT());
    498 }
    499 
    500 /*
    501  - p_bre - BRE parser top level, anchoring and concatenation
    502  == static void p_bre(struct parse *p, int end1, \
    503  ==	int end2);
    504  * Giving end1 as OUT essentially eliminates the end1/end2 check.
    505  *
    506  * This implementation is a bit of a kludge, in that a trailing $ is first
    507  * taken as an ordinary character and then revised to be an anchor.  The
    508  * only undesirable side effect is that '$' gets included as a character
    509  * category in such cases.  This is fairly harmless; not worth fixing.
    510  * The amount of lookahead needed to avoid this kludge is excessive.
    511  */
    512 static void
    513 p_bre(p, end1, end2)
    514 struct parse *p;
    515 int end1;		/* first terminating character */
    516 int end2;		/* second terminating character */
    517 {
    518 	sopno start = HERE();
    519 	int first = 1;			/* first subexpression? */
    520 	int wasdollar = 0;
    521 
    522 	if (EAT('^')) {
    523 		EMIT(OBOL, 0);
    524 		p->g->iflags |= USEBOL;
    525 		p->g->nbol++;
    526 	}
    527 	while (MORE() && !SEETWO(end1, end2)) {
    528 		wasdollar = p_simp_re(p, first);
    529 		first = 0;
    530 	}
    531 	if (wasdollar) {	/* oops, that was a trailing anchor */
    532 		DROP(1);
    533 		EMIT(OEOL, 0);
    534 		p->g->iflags |= USEEOL;
    535 		p->g->neol++;
    536 	}
    537 
    538 	REQUIRE(HERE() != start, REG_EMPTY);	/* require nonempty */
    539 }
    540 
    541 /*
    542  - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
    543  == static int p_simp_re(struct parse *p, int starordinary);
    544  */
    545 static int			/* was the simple RE an unbackslashed $? */
    546 p_simp_re(p, starordinary)
    547 struct parse *p;
    548 int starordinary;		/* is a leading * an ordinary character? */
    549 {
    550 	int c;
    551 	int count;
    552 	int count2;
    553 	sopno pos;
    554 	int i;
    555 	sopno subno;
    556 #	define	BACKSL	(1<<CHAR_BIT)
    557 
    558 	pos = HERE();		/* repetion op, if any, covers from here */
    559 
    560 	assert(MORE());		/* caller should have ensured this */
    561 	c = GETNEXT();
    562 	if (c == '\\') {
    563 		REQUIRE(MORE(), REG_EESCAPE);
    564 		c = BACKSL | (unsigned char)GETNEXT();
    565 	}
    566 	switch (c) {
    567 	case '.':
    568 		if (p->g->cflags&REG_NEWLINE)
    569 			nonnewline(p);
    570 		else
    571 			EMIT(OANY, 0);
    572 		break;
    573 	case '[':
    574 		p_bracket(p);
    575 		break;
    576 	case BACKSL|'{':
    577 		SETERROR(REG_BADRPT);
    578 		break;
    579 	case BACKSL|'(':
    580 		p->g->nsub++;
    581 		subno = p->g->nsub;
    582 		if (subno < NPAREN)
    583 			p->pbegin[subno] = HERE();
    584 		EMIT(OLPAREN, subno);
    585 		/* the MORE here is an error heuristic */
    586 		if (MORE() && !SEETWO('\\', ')'))
    587 			p_bre(p, '\\', ')');
    588 		if (subno < NPAREN) {
    589 			p->pend[subno] = HERE();
    590 			assert(p->pend[subno] != 0);
    591 		}
    592 		EMIT(ORPAREN, subno);
    593 		REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
    594 		break;
    595 	case BACKSL|')':	/* should not get here -- must be user */
    596 	case BACKSL|'}':
    597 		SETERROR(REG_EPAREN);
    598 		break;
    599 	case BACKSL|'1':
    600 	case BACKSL|'2':
    601 	case BACKSL|'3':
    602 	case BACKSL|'4':
    603 	case BACKSL|'5':
    604 	case BACKSL|'6':
    605 	case BACKSL|'7':
    606 	case BACKSL|'8':
    607 	case BACKSL|'9':
    608 		i = (c&~BACKSL) - '0';
    609 		assert(i < NPAREN);
    610 		if (p->pend[i] != 0) {
    611 			assert(i <= p->g->nsub);
    612 			EMIT(OBACK_, i);
    613 			assert(p->pbegin[i] != 0);
    614 			assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
    615 			assert(OP(p->strip[p->pend[i]]) == ORPAREN);
    616 			(void) dupl(p, p->pbegin[i]+1, p->pend[i]);
    617 			EMIT(O_BACK, i);
    618 		} else
    619 			SETERROR(REG_ESUBREG);
    620 		p->g->backrefs = 1;
    621 		break;
    622 	case '*':
    623 		REQUIRE(starordinary, REG_BADRPT);
    624 		/* FALLTHROUGH */
    625 	default:
    626 		ordinary(p, c &~ BACKSL);
    627 		break;
    628 	}
    629 
    630 	if (EAT('*')) {		/* implemented as +? */
    631 		/* this case does not require the (y|) trick, noKLUDGE */
    632 		INSERT(OPLUS_, pos);
    633 		ASTERN(O_PLUS, pos);
    634 		INSERT(OQUEST_, pos);
    635 		ASTERN(O_QUEST, pos);
    636 	} else if (EATTWO('\\', '{')) {
    637 		count = p_count(p);
    638 		if (EAT(',')) {
    639 			if (MORE() && isdigit(PEEK())) {
    640 				count2 = p_count(p);
    641 				REQUIRE(count <= count2, REG_BADBR);
    642 			} else		/* single number with comma */
    643 				count2 = INFINITY;
    644 		} else		/* just a single number */
    645 			count2 = count;
    646 		repeat(p, pos, count, count2);
    647 		if (!EATTWO('\\', '}')) {	/* error heuristics */
    648 			while (MORE() && !SEETWO('\\', '}'))
    649 				NEXT();
    650 			REQUIRE(MORE(), REG_EBRACE);
    651 			SETERROR(REG_BADBR);
    652 		}
    653 	} else if (c == (unsigned char)'$')	/* $ (but not \$) ends it */
    654 		return(1);
    655 
    656 	return(0);
    657 }
    658 
    659 /*
    660  - p_count - parse a repetition count
    661  == static int p_count(struct parse *p);
    662  */
    663 static int			/* the value */
    664 p_count(p)
    665 struct parse *p;
    666 {
    667 	int count = 0;
    668 	int ndigits = 0;
    669 
    670 	while (MORE() && isdigit(PEEK()) && count <= DUPMAX) {
    671 		count = count*10 + (GETNEXT() - '0');
    672 		ndigits++;
    673 	}
    674 
    675 	REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
    676 	return(count);
    677 }
    678 
    679 /*
    680  - p_bracket - parse a bracketed character list
    681  == static void p_bracket(struct parse *p);
    682  *
    683  * Note a significant property of this code:  if the allocset() did SETERROR,
    684  * no set operations are done.
    685  */
    686 static void
    687 p_bracket(p)
    688 struct parse *p;
    689 {
    690 	cset *cs = allocset(p);
    691 	int invert = 0;
    692 
    693 	/* Dept of Truly Sickening Special-Case Kludges */
    694 	if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
    695 		EMIT(OBOW, 0);
    696 		NEXTn(6);
    697 		return;
    698 	}
    699 	if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
    700 		EMIT(OEOW, 0);
    701 		NEXTn(6);
    702 		return;
    703 	}
    704 
    705 	if (EAT('^'))
    706 		invert++;	/* make note to invert set at end */
    707 	if (EAT(']'))
    708 		CHadd(cs, ']');
    709 	else if (EAT('-'))
    710 		CHadd(cs, '-');
    711 	while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
    712 		p_b_term(p, cs);
    713 	if (EAT('-'))
    714 		CHadd(cs, '-');
    715 	MUSTEAT(']', REG_EBRACK);
    716 
    717 	if (p->error != 0)	/* don't mess things up further */
    718 		return;
    719 
    720 	if (p->g->cflags&REG_ICASE) {
    721 		int i;
    722 		int ci;
    723 
    724 		for (i = p->g->csetsize - 1; i >= 0; i--)
    725 			if (CHIN(cs, i) && isalpha(i)) {
    726 				ci = othercase(i);
    727 				if (ci != i)
    728 					CHadd(cs, ci);
    729 			}
    730 		if (cs->multis != NULL)
    731 			mccase(p, cs);
    732 	}
    733 	if (invert) {
    734 		int i;
    735 
    736 		for (i = p->g->csetsize - 1; i >= 0; i--)
    737 			if (CHIN(cs, i))
    738 				CHsub(cs, i);
    739 			else
    740 				CHadd(cs, i);
    741 		if (p->g->cflags&REG_NEWLINE)
    742 			CHsub(cs, '\n');
    743 		if (cs->multis != NULL)
    744 			mcinvert(p, cs);
    745 	}
    746 
    747 	assert(cs->multis == NULL);		/* xxx */
    748 
    749 	if (nch(p, cs) == 1) {		/* optimize singleton sets */
    750 		ordinary(p, firstch(p, cs));
    751 		freeset(p, cs);
    752 	} else
    753 		EMIT(OANYOF, freezeset(p, cs));
    754 }
    755 
    756 /*
    757  - p_b_term - parse one term of a bracketed character list
    758  == static void p_b_term(struct parse *p, cset *cs);
    759  */
    760 static void
    761 p_b_term(p, cs)
    762 struct parse *p;
    763 cset *cs;
    764 {
    765 	char c;
    766 	char start, finish;
    767 	int i;
    768 
    769 	/* classify what we've got */
    770 	switch ((MORE()) ? PEEK() : '\0') {
    771 	case '[':
    772 		c = (MORE2()) ? PEEK2() : '\0';
    773 		break;
    774 	case '-':
    775 		SETERROR(REG_ERANGE);
    776 		return;			/* NOTE RETURN */
    777 		break;
    778 	default:
    779 		c = '\0';
    780 		break;
    781 	}
    782 
    783 	switch (c) {
    784 	case ':':		/* character class */
    785 		NEXT2();
    786 		REQUIRE(MORE(), REG_EBRACK);
    787 		c = PEEK();
    788 		REQUIRE(c != '-' && c != ']', REG_ECTYPE);
    789 		p_b_cclass(p, cs);
    790 		REQUIRE(MORE(), REG_EBRACK);
    791 		REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
    792 		break;
    793 	case '=':		/* equivalence class */
    794 		NEXT2();
    795 		REQUIRE(MORE(), REG_EBRACK);
    796 		c = PEEK();
    797 		REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
    798 		p_b_eclass(p, cs);
    799 		REQUIRE(MORE(), REG_EBRACK);
    800 		REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
    801 		break;
    802 	default:		/* symbol, ordinary character, or range */
    803 /* xxx revision needed for multichar stuff */
    804 		start = p_b_symbol(p);
    805 		if (SEE('-') && MORE2() && PEEK2() != ']') {
    806 			/* range */
    807 			NEXT();
    808 			if (EAT('-'))
    809 				finish = '-';
    810 			else
    811 				finish = p_b_symbol(p);
    812 		} else
    813 			finish = start;
    814 /* xxx what about signed chars here... */
    815 		REQUIRE(start <= finish, REG_ERANGE);
    816 		for (i = start; i <= finish; i++)
    817 			CHadd(cs, i);
    818 		break;
    819 	}
    820 }
    821 
    822 /*
    823  - p_b_cclass - parse a character-class name and deal with it
    824  == static void p_b_cclass(struct parse *p, cset *cs);
    825  */
    826 static void
    827 p_b_cclass(p, cs)
    828 struct parse *p;
    829 cset *cs;
    830 {
    831 	char *sp = p->next;
    832 	struct cclass *cp;
    833 	size_t len;
    834 	char *u;
    835 	char c;
    836 
    837 	while (MORE() && isalpha(PEEK()))
    838 		NEXT();
    839 	len = p->next - sp;
    840 	for (cp = cclasses; cp->name != NULL; cp++)
    841 		if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
    842 			break;
    843 	if (cp->name == NULL) {
    844 		/* oops, didn't find it */
    845 		SETERROR(REG_ECTYPE);
    846 		return;
    847 	}
    848 
    849 	u = cp->chars;
    850 	while ((c = *u++) != '\0')
    851 		CHadd(cs, c);
    852 	for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
    853 		MCadd(p, cs, u);
    854 }
    855 
    856 /*
    857  - p_b_eclass - parse an equivalence-class name and deal with it
    858  == static void p_b_eclass(struct parse *p, cset *cs);
    859  *
    860  * This implementation is incomplete. xxx
    861  */
    862 static void
    863 p_b_eclass(p, cs)
    864 struct parse *p;
    865 cset *cs;
    866 {
    867 	char c;
    868 
    869 	c = p_b_coll_elem(p, '=');
    870 	CHadd(cs, c);
    871 }
    872 
    873 /*
    874  - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
    875  == static char p_b_symbol(struct parse *p);
    876  */
    877 static char			/* value of symbol */
    878 p_b_symbol(p)
    879 struct parse *p;
    880 {
    881 	char value;
    882 
    883 	REQUIRE(MORE(), REG_EBRACK);
    884 	if (!EATTWO('[', '.'))
    885 		return(GETNEXT());
    886 
    887 	/* collating symbol */
    888 	value = p_b_coll_elem(p, '.');
    889 	REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
    890 	return(value);
    891 }
    892 
    893 /*
    894  - p_b_coll_elem - parse a collating-element name and look it up
    895  == static char p_b_coll_elem(struct parse *p, int endc);
    896  */
    897 static char			/* value of collating element */
    898 p_b_coll_elem(p, endc)
    899 struct parse *p;
    900 int endc;			/* name ended by endc,']' */
    901 {
    902 	char *sp = p->next;
    903 	struct cname *cp;
    904 	int len;
    905 
    906 	while (MORE() && !SEETWO(endc, ']'))
    907 		NEXT();
    908 	if (!MORE()) {
    909 		SETERROR(REG_EBRACK);
    910 		return(0);
    911 	}
    912 	len = p->next - sp;
    913 	for (cp = cnames; cp->name != NULL; cp++)
    914 		if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
    915 			return(cp->code);	/* known name */
    916 	if (len == 1)
    917 		return(*sp);	/* single character */
    918 	SETERROR(REG_ECOLLATE);			/* neither */
    919 	return(0);
    920 }
    921 
    922 /*
    923  - othercase - return the case counterpart of an alphabetic
    924  == static char othercase(int ch);
    925  */
    926 static char			/* if no counterpart, return ch */
    927 othercase(ch)
    928 int ch;
    929 {
    930 	assert(isalpha(ch));
    931 	if (isupper(ch))
    932 		return(tolower(ch));
    933 	else if (islower(ch))
    934 		return(toupper(ch));
    935 	else			/* peculiar, but could happen */
    936 		return(ch);
    937 }
    938 
    939 /*
    940  - bothcases - emit a dualcase version of a two-case character
    941  == static void bothcases(struct parse *p, int ch);
    942  *
    943  * Boy, is this implementation ever a kludge...
    944  */
    945 static void
    946 bothcases(p, ch)
    947 struct parse *p;
    948 int ch;
    949 {
    950 	char *oldnext = p->next;
    951 	char *oldend = p->end;
    952 	char bracket[3];
    953 
    954 	assert(othercase(ch) != ch);	/* p_bracket() would recurse */
    955 	p->next = bracket;
    956 	p->end = bracket+2;
    957 	bracket[0] = ch;
    958 	bracket[1] = ']';
    959 	bracket[2] = '\0';
    960 	p_bracket(p);
    961 	assert(p->next == bracket+2);
    962 	p->next = oldnext;
    963 	p->end = oldend;
    964 }
    965 
    966 /*
    967  - ordinary - emit an ordinary character
    968  == static void ordinary(struct parse *p, int ch);
    969  */
    970 static void
    971 ordinary(p, ch)
    972 struct parse *p;
    973 int ch;
    974 {
    975 	cat_t *cap = p->g->categories;
    976 
    977 	if ((p->g->cflags&REG_ICASE) && isalpha(ch) && othercase(ch) != ch)
    978 		bothcases(p, ch);
    979 	else {
    980 		EMIT(OCHAR, (unsigned char)ch);
    981 		if (cap[ch] == 0)
    982 			cap[ch] = p->g->ncategories++;
    983 	}
    984 }
    985 
    986 /*
    987  - nonnewline - emit REG_NEWLINE version of OANY
    988  == static void nonnewline(struct parse *p);
    989  *
    990  * Boy, is this implementation ever a kludge...
    991  */
    992 static void
    993 nonnewline(p)
    994 struct parse *p;
    995 {
    996 	char *oldnext = p->next;
    997 	char *oldend = p->end;
    998 	char bracket[4];
    999 
   1000 	p->next = bracket;
   1001 	p->end = bracket+3;
   1002 	bracket[0] = '^';
   1003 	bracket[1] = '\n';
   1004 	bracket[2] = ']';
   1005 	bracket[3] = '\0';
   1006 	p_bracket(p);
   1007 	assert(p->next == bracket+3);
   1008 	p->next = oldnext;
   1009 	p->end = oldend;
   1010 }
   1011 
   1012 /*
   1013  - repeat - generate code for a bounded repetition, recursively if needed
   1014  == static void repeat(struct parse *p, sopno start, int from, int to);
   1015  */
   1016 static void
   1017 repeat(p, start, from, to)
   1018 struct parse *p;
   1019 sopno start;			/* operand from here to end of strip */
   1020 int from;			/* repeated from this number */
   1021 int to;				/* to this number of times (maybe INFINITY) */
   1022 {
   1023 	sopno finish = HERE();
   1024 #	define	N	2
   1025 #	define	INF	3
   1026 #	define	REP(f, t)	((f)*8 + (t))
   1027 #	define	MAP(n)	(((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
   1028 	sopno copy;
   1029 
   1030 	if (p->error != 0)	/* head off possible runaway recursion */
   1031 		return;
   1032 
   1033 	assert(from <= to);
   1034 
   1035 	switch (REP(MAP(from), MAP(to))) {
   1036 	case REP(0, 0):			/* must be user doing this */
   1037 		DROP(finish-start);	/* drop the operand */
   1038 		break;
   1039 	case REP(0, 1):			/* as x{1,1}? */
   1040 	case REP(0, N):			/* as x{1,n}? */
   1041 	case REP(0, INF):		/* as x{1,}? */
   1042 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
   1043 		INSERT(OCH_, start);		/* offset is wrong... */
   1044 		repeat(p, start+1, 1, to);
   1045 		ASTERN(OOR1, start);
   1046 		AHEAD(start);			/* ... fix it */
   1047 		EMIT(OOR2, 0);
   1048 		AHEAD(THERE());
   1049 		ASTERN(O_CH, THERETHERE());
   1050 		break;
   1051 	case REP(1, 1):			/* trivial case */
   1052 		/* done */
   1053 		break;
   1054 	case REP(1, N):			/* as x?x{1,n-1} */
   1055 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
   1056 		INSERT(OCH_, start);
   1057 		ASTERN(OOR1, start);
   1058 		AHEAD(start);
   1059 		EMIT(OOR2, 0);			/* offset very wrong... */
   1060 		AHEAD(THERE());			/* ...so fix it */
   1061 		ASTERN(O_CH, THERETHERE());
   1062 		copy = dupl(p, start+1, finish+1);
   1063 		assert(copy == finish+4);
   1064 		repeat(p, copy, 1, to-1);
   1065 		break;
   1066 	case REP(1, INF):		/* as x+ */
   1067 		INSERT(OPLUS_, start);
   1068 		ASTERN(O_PLUS, start);
   1069 		break;
   1070 	case REP(N, N):			/* as xx{m-1,n-1} */
   1071 		copy = dupl(p, start, finish);
   1072 		repeat(p, copy, from-1, to-1);
   1073 		break;
   1074 	case REP(N, INF):		/* as xx{n-1,INF} */
   1075 		copy = dupl(p, start, finish);
   1076 		repeat(p, copy, from-1, to);
   1077 		break;
   1078 	default:			/* "can't happen" */
   1079 		SETERROR(REG_ASSERT);	/* just in case */
   1080 		break;
   1081 	}
   1082 }
   1083 
   1084 /*
   1085  - seterr - set an error condition
   1086  == static int seterr(struct parse *p, int e);
   1087  */
   1088 static int			/* useless but makes type checking happy */
   1089 seterr(p, e)
   1090 struct parse *p;
   1091 int e;
   1092 {
   1093 	if (p->error == 0)	/* keep earliest error condition */
   1094 		p->error = e;
   1095 	p->next = nuls;		/* try to bring things to a halt */
   1096 	p->end = nuls;
   1097 	return(0);		/* make the return value well-defined */
   1098 }
   1099 
   1100 /*
   1101  - allocset - allocate a set of characters for []
   1102  == static cset *allocset(struct parse *p);
   1103  */
   1104 static cset *
   1105 allocset(p)
   1106 struct parse *p;
   1107 {
   1108 	int no = p->g->ncsets++;
   1109 	size_t nc;
   1110 	size_t nbytes;
   1111 	cset *cs;
   1112 	size_t css = (size_t)p->g->csetsize;
   1113 	int i;
   1114 
   1115 	if (no >= p->ncsalloc) {	/* need another column of space */
   1116 		p->ncsalloc += CHAR_BIT;
   1117 		nc = p->ncsalloc;
   1118 		assert(nc % CHAR_BIT == 0);
   1119 		nbytes = nc / CHAR_BIT * css;
   1120 		if (p->g->sets == NULL)
   1121 			p->g->sets = (cset *)malloc(nc * sizeof(cset));
   1122 		else
   1123 			p->g->sets = (cset *)realloc((char *)p->g->sets,
   1124 							nc * sizeof(cset));
   1125 		if (p->g->setbits == NULL)
   1126 			p->g->setbits = (uch *)malloc(nbytes);
   1127 		else {
   1128 			p->g->setbits = (uch *)realloc((char *)p->g->setbits,
   1129 								nbytes);
   1130 			/* xxx this isn't right if setbits is now NULL */
   1131 			for (i = 0; i < no; i++)
   1132 				p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
   1133 		}
   1134 		if (p->g->sets != NULL && p->g->setbits != NULL)
   1135 			(void) memset((char *)p->g->setbits + (nbytes - css),
   1136 								0, css);
   1137 		else {
   1138 			no = 0;
   1139 			SETERROR(REG_ESPACE);
   1140 			/* caller's responsibility not to do set ops */
   1141 		}
   1142 	}
   1143 
   1144 	assert(p->g->sets != NULL);	/* xxx */
   1145 	cs = &p->g->sets[no];
   1146 	cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
   1147 	cs->mask = 1 << ((no) % CHAR_BIT);
   1148 	cs->hash = 0;
   1149 	cs->smultis = 0;
   1150 	cs->multis = NULL;
   1151 
   1152 	return(cs);
   1153 }
   1154 
   1155 /*
   1156  - freeset - free a now-unused set
   1157  == static void freeset(struct parse *p, cset *cs);
   1158  */
   1159 static void
   1160 freeset(p, cs)
   1161 struct parse *p;
   1162 cset *cs;
   1163 {
   1164 	int i;
   1165 	cset *top = &p->g->sets[p->g->ncsets];
   1166 	size_t css = (size_t)p->g->csetsize;
   1167 
   1168 	for (i = 0; i < css; i++)
   1169 		CHsub(cs, i);
   1170 	if (cs == top-1)	/* recover only the easy case */
   1171 		p->g->ncsets--;
   1172 }
   1173 
   1174 /*
   1175  - freezeset - final processing on a set of characters
   1176  == static int freezeset(struct parse *p, cset *cs);
   1177  *
   1178  * The main task here is merging identical sets.  This is usually a waste
   1179  * of time (although the hash code minimizes the overhead), but can win
   1180  * big if REG_ICASE is being used.  REG_ICASE, by the way, is why the hash
   1181  * is done using addition rather than xor -- all ASCII [aA] sets xor to
   1182  * the same value!
   1183  */
   1184 static int			/* set number */
   1185 freezeset(p, cs)
   1186 struct parse *p;
   1187 cset *cs;
   1188 {
   1189 	uch h = cs->hash;
   1190 	int i;
   1191 	cset *top = &p->g->sets[p->g->ncsets];
   1192 	cset *cs2;
   1193 	size_t css = (size_t)p->g->csetsize;
   1194 
   1195 	/* look for an earlier one which is the same */
   1196 	for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
   1197 		if (cs2->hash == h && cs2 != cs) {
   1198 			/* maybe */
   1199 			for (i = 0; i < css; i++)
   1200 				if (!!CHIN(cs2, i) != !!CHIN(cs, i))
   1201 					break;		/* no */
   1202 			if (i == css)
   1203 				break;			/* yes */
   1204 		}
   1205 
   1206 	if (cs2 < top) {	/* found one */
   1207 		freeset(p, cs);
   1208 		cs = cs2;
   1209 	}
   1210 
   1211 	return((int)(cs - p->g->sets));
   1212 }
   1213 
   1214 /*
   1215  - firstch - return first character in a set (which must have at least one)
   1216  == static int firstch(struct parse *p, cset *cs);
   1217  */
   1218 static int			/* character; there is no "none" value */
   1219 firstch(p, cs)
   1220 struct parse *p;
   1221 cset *cs;
   1222 {
   1223 	int i;
   1224 	size_t css = (size_t)p->g->csetsize;
   1225 
   1226 	for (i = 0; i < css; i++)
   1227 		if (CHIN(cs, i))
   1228 			return((char)i);
   1229 	assert(never);
   1230 	return(0);		/* arbitrary */
   1231 }
   1232 
   1233 /*
   1234  - nch - number of characters in a set
   1235  == static int nch(struct parse *p, cset *cs);
   1236  */
   1237 static int
   1238 nch(p, cs)
   1239 struct parse *p;
   1240 cset *cs;
   1241 {
   1242 	int i;
   1243 	size_t css = (size_t)p->g->csetsize;
   1244 	int n = 0;
   1245 
   1246 	for (i = 0; i < css; i++)
   1247 		if (CHIN(cs, i))
   1248 			n++;
   1249 	return(n);
   1250 }
   1251 
   1252 /*
   1253  - mcadd - add a collating element to a cset
   1254  == static void mcadd(struct parse *p, cset *cs, \
   1255  ==	char *cp);
   1256  */
   1257 static void
   1258 mcadd(p, cs, cp)
   1259 struct parse *p;
   1260 cset *cs;
   1261 char *cp;
   1262 {
   1263 	size_t oldend = cs->smultis;
   1264 
   1265 	cs->smultis += strlen(cp) + 1;
   1266 	if (cs->multis == NULL)
   1267 		cs->multis = malloc(cs->smultis);
   1268 	else
   1269 		cs->multis = realloc(cs->multis, cs->smultis);
   1270 	if (cs->multis == NULL) {
   1271 		SETERROR(REG_ESPACE);
   1272 		return;
   1273 	}
   1274 
   1275 	(void) strcpy(cs->multis + oldend - 1, cp);
   1276 	cs->multis[cs->smultis - 1] = '\0';
   1277 }
   1278 
   1279 #if 0
   1280 /*
   1281  - mcsub - subtract a collating element from a cset
   1282  == static void mcsub(cset *cs, char *cp);
   1283  */
   1284 static void
   1285 mcsub(cs, cp)
   1286 cset *cs;
   1287 char *cp;
   1288 {
   1289 	char *fp = mcfind(cs, cp);
   1290 	size_t len = strlen(fp);
   1291 
   1292 	assert(fp != NULL);
   1293 	(void) memmove(fp, fp + len + 1,
   1294 				cs->smultis - (fp + len + 1 - cs->multis));
   1295 	cs->smultis -= len;
   1296 
   1297 	if (cs->smultis == 0) {
   1298 		free(cs->multis);
   1299 		cs->multis = NULL;
   1300 		return;
   1301 	}
   1302 
   1303 	cs->multis = realloc(cs->multis, cs->smultis);
   1304 	assert(cs->multis != NULL);
   1305 }
   1306 
   1307 /*
   1308  - mcin - is a collating element in a cset?
   1309  == static int mcin(cset *cs, char *cp);
   1310  */
   1311 static int
   1312 mcin(cs, cp)
   1313 cset *cs;
   1314 char *cp;
   1315 {
   1316 	return(mcfind(cs, cp) != NULL);
   1317 }
   1318 
   1319 /*
   1320  - mcfind - find a collating element in a cset
   1321  == static char *mcfind(cset *cs, char *cp);
   1322  */
   1323 static char *
   1324 mcfind(cs, cp)
   1325 cset *cs;
   1326 char *cp;
   1327 {
   1328 	char *p;
   1329 
   1330 	if (cs->multis == NULL)
   1331 		return(NULL);
   1332 	for (p = cs->multis; *p != '\0'; p += strlen(p) + 1)
   1333 		if (strcmp(cp, p) == 0)
   1334 			return(p);
   1335 	return(NULL);
   1336 }
   1337 #endif
   1338 
   1339 /*
   1340  - mcinvert - invert the list of collating elements in a cset
   1341  == static void mcinvert(struct parse *p, cset *cs);
   1342  *
   1343  * This would have to know the set of possibilities.  Implementation
   1344  * is deferred.
   1345  */
   1346 static void
   1347 mcinvert(p, cs)
   1348 struct parse *p;
   1349 cset *cs;
   1350 {
   1351 	assert(cs->multis == NULL);	/* xxx */
   1352 }
   1353 
   1354 /*
   1355  - mccase - add case counterparts of the list of collating elements in a cset
   1356  == static void mccase(struct parse *p, cset *cs);
   1357  *
   1358  * This would have to know the set of possibilities.  Implementation
   1359  * is deferred.
   1360  */
   1361 static void
   1362 mccase(p, cs)
   1363 struct parse *p;
   1364 cset *cs;
   1365 {
   1366 	assert(cs->multis == NULL);	/* xxx */
   1367 }
   1368 
   1369 /*
   1370  - isinsets - is this character in any sets?
   1371  == static int isinsets(struct re_guts *g, int c);
   1372  */
   1373 static int			/* predicate */
   1374 isinsets(g, c)
   1375 struct re_guts *g;
   1376 int c;
   1377 {
   1378 	uch *col;
   1379 	int i;
   1380 	int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
   1381 	unsigned uc = (unsigned char)c;
   1382 
   1383 	for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
   1384 		if (col[uc] != 0)
   1385 			return(1);
   1386 	return(0);
   1387 }
   1388 
   1389 /*
   1390  - samesets - are these two characters in exactly the same sets?
   1391  == static int samesets(struct re_guts *g, int c1, int c2);
   1392  */
   1393 static int			/* predicate */
   1394 samesets(g, c1, c2)
   1395 struct re_guts *g;
   1396 int c1;
   1397 int c2;
   1398 {
   1399 	uch *col;
   1400 	int i;
   1401 	int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
   1402 	unsigned uc1 = (unsigned char)c1;
   1403 	unsigned uc2 = (unsigned char)c2;
   1404 
   1405 	for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
   1406 		if (col[uc1] != col[uc2])
   1407 			return(0);
   1408 	return(1);
   1409 }
   1410 
   1411 /*
   1412  - categorize - sort out character categories
   1413  == static void categorize(struct parse *p, struct re_guts *g);
   1414  */
   1415 static void
   1416 categorize(p, g)
   1417 struct parse *p;
   1418 struct re_guts *g;
   1419 {
   1420 	cat_t *cats = g->categories;
   1421 	int c;
   1422 	int c2;
   1423 	cat_t cat;
   1424 
   1425 	/* avoid making error situations worse */
   1426 	if (p->error != 0)
   1427 		return;
   1428 
   1429 	for (c = CHAR_MIN; c <= CHAR_MAX; c++)
   1430 		if (cats[c] == 0 && isinsets(g, c)) {
   1431 			cat = g->ncategories++;
   1432 			cats[c] = cat;
   1433 			for (c2 = c+1; c2 <= CHAR_MAX; c2++)
   1434 				if (cats[c2] == 0 && samesets(g, c, c2))
   1435 					cats[c2] = cat;
   1436 		}
   1437 }
   1438 
   1439 /*
   1440  - dupl - emit a duplicate of a bunch of sops
   1441  == static sopno dupl(struct parse *p, sopno start, sopno finish);
   1442  */
   1443 static sopno			/* start of duplicate */
   1444 dupl(p, start, finish)
   1445 struct parse *p;
   1446 sopno start;			/* from here */
   1447 sopno finish;			/* to this less one */
   1448 {
   1449 	sopno ret = HERE();
   1450 	sopno len = finish - start;
   1451 
   1452 	assert(finish >= start);
   1453 	if (len == 0)
   1454 		return(ret);
   1455 	enlarge(p, p->ssize + len);	/* this many unexpected additions */
   1456 	assert(p->ssize >= p->slen + len);
   1457 	(void) memcpy((char *)(p->strip + p->slen),
   1458 		(char *)(p->strip + start), (size_t)len*sizeof(sop));
   1459 	p->slen += len;
   1460 	return(ret);
   1461 }
   1462 
   1463 /*
   1464  - doemit - emit a strip operator
   1465  == static void doemit(struct parse *p, sop op, size_t opnd);
   1466  *
   1467  * It might seem better to implement this as a macro with a function as
   1468  * hard-case backup, but it's just too big and messy unless there are
   1469  * some changes to the data structures.  Maybe later.
   1470  */
   1471 static void
   1472 doemit(p, op, opnd)
   1473 struct parse *p;
   1474 sop op;
   1475 size_t opnd;
   1476 {
   1477 	/* avoid making error situations worse */
   1478 	if (p->error != 0)
   1479 		return;
   1480 
   1481 	/* deal with oversize operands ("can't happen", more or less) */
   1482 	assert(opnd < 1<<OPSHIFT);
   1483 
   1484 	/* deal with undersized strip */
   1485 	if (p->slen >= p->ssize)
   1486 		enlarge(p, (p->ssize+1) / 2 * 3);	/* +50% */
   1487 	assert(p->slen < p->ssize);
   1488 
   1489 	/* finally, it's all reduced to the easy case */
   1490 	p->strip[p->slen++] = SOP(op, opnd);
   1491 }
   1492 
   1493 /*
   1494  - doinsert - insert a sop into the strip
   1495  == static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
   1496  */
   1497 static void
   1498 doinsert(p, op, opnd, pos)
   1499 struct parse *p;
   1500 sop op;
   1501 size_t opnd;
   1502 sopno pos;
   1503 {
   1504 	sopno sn;
   1505 	sop s;
   1506 	int i;
   1507 
   1508 	/* avoid making error situations worse */
   1509 	if (p->error != 0)
   1510 		return;
   1511 
   1512 	sn = HERE();
   1513 	EMIT(op, opnd);		/* do checks, ensure space */
   1514 	assert(HERE() == sn+1);
   1515 	s = p->strip[sn];
   1516 
   1517 	/* adjust paren pointers */
   1518 	assert(pos > 0);
   1519 	for (i = 1; i < NPAREN; i++) {
   1520 		if (p->pbegin[i] >= pos) {
   1521 			p->pbegin[i]++;
   1522 		}
   1523 		if (p->pend[i] >= pos) {
   1524 			p->pend[i]++;
   1525 		}
   1526 	}
   1527 
   1528 	memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
   1529 						(HERE()-pos-1)*sizeof(sop));
   1530 	p->strip[pos] = s;
   1531 }
   1532 
   1533 /*
   1534  - dofwd - complete a forward reference
   1535  == static void dofwd(struct parse *p, sopno pos, sop value);
   1536  */
   1537 static void
   1538 dofwd(p, pos, value)
   1539 struct parse *p;
   1540 sopno pos;
   1541 sop value;
   1542 {
   1543 	/* avoid making error situations worse */
   1544 	if (p->error != 0)
   1545 		return;
   1546 
   1547 	assert(value < 1<<OPSHIFT);
   1548 	p->strip[pos] = OP(p->strip[pos]) | value;
   1549 }
   1550 
   1551 /*
   1552  - enlarge - enlarge the strip
   1553  == static void enlarge(struct parse *p, sopno size);
   1554  */
   1555 static void
   1556 enlarge(p, size)
   1557 struct parse *p;
   1558 sopno size;
   1559 {
   1560 	sop *sp;
   1561 
   1562 	if (p->ssize >= size)
   1563 		return;
   1564 
   1565 	sp = (sop *)realloc(p->strip, size*sizeof(sop));
   1566 	if (sp == NULL) {
   1567 		SETERROR(REG_ESPACE);
   1568 		return;
   1569 	}
   1570 	p->strip = sp;
   1571 	p->ssize = size;
   1572 }
   1573 
   1574 /*
   1575  - stripsnug - compact the strip
   1576  == static void stripsnug(struct parse *p, struct re_guts *g);
   1577  */
   1578 static void
   1579 stripsnug(p, g)
   1580 struct parse *p;
   1581 struct re_guts *g;
   1582 {
   1583 	g->nstates = p->slen;
   1584 	g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
   1585 	if (g->strip == NULL) {
   1586 		SETERROR(REG_ESPACE);
   1587 		g->strip = p->strip;
   1588 	}
   1589 }
   1590 
   1591 /*
   1592  - findmust - fill in must and mlen with longest mandatory literal string
   1593  == static void findmust(struct parse *p, struct re_guts *g);
   1594  *
   1595  * This algorithm could do fancy things like analyzing the operands of |
   1596  * for common subsequences.  Someday.  This code is simple and finds most
   1597  * of the interesting cases.
   1598  *
   1599  * Note that must and mlen got initialized during setup.
   1600  */
   1601 static void
   1602 findmust(p, g)
   1603 struct parse *p;
   1604 struct re_guts *g;
   1605 {
   1606 	sop *scan;
   1607 	sop *start = NULL;
   1608 	sop *newstart = NULL;
   1609 	sopno newlen;
   1610 	sop s;
   1611 	char *cp;
   1612 	sopno i;
   1613 
   1614 	/* avoid making error situations worse */
   1615 	if (p->error != 0)
   1616 		return;
   1617 
   1618 	/* find the longest OCHAR sequence in strip */
   1619 	newlen = 0;
   1620 	scan = g->strip + 1;
   1621 	do {
   1622 		s = *scan++;
   1623 		switch (OP(s)) {
   1624 		case OCHAR:		/* sequence member */
   1625 			if (newlen == 0)		/* new sequence */
   1626 				newstart = scan - 1;
   1627 			newlen++;
   1628 			break;
   1629 		case OPLUS_:		/* things that don't break one */
   1630 		case OLPAREN:
   1631 		case ORPAREN:
   1632 			break;
   1633 		case OQUEST_:		/* things that must be skipped */
   1634 		case OCH_:
   1635 			scan--;
   1636 			do {
   1637 				scan += OPND(s);
   1638 				s = *scan;
   1639 				/* assert() interferes w debug printouts */
   1640 				if (OP(s) != O_QUEST && OP(s) != O_CH &&
   1641 							OP(s) != OOR2) {
   1642 					g->iflags |= BAD;
   1643 					return;
   1644 				}
   1645 			} while (OP(s) != O_QUEST && OP(s) != O_CH);
   1646 			/* fallthrough */
   1647 		default:		/* things that break a sequence */
   1648 			if (newlen > g->mlen) {		/* ends one */
   1649 				start = newstart;
   1650 				g->mlen = newlen;
   1651 			}
   1652 			newlen = 0;
   1653 			break;
   1654 		}
   1655 	} while (OP(s) != OEND);
   1656 
   1657 	if (g->mlen == 0)		/* there isn't one */
   1658 		return;
   1659 
   1660 	/* turn it into a character string */
   1661 	g->must = malloc((size_t)g->mlen + 1);
   1662 	if (g->must == NULL) {		/* argh; just forget it */
   1663 		g->mlen = 0;
   1664 		return;
   1665 	}
   1666 	cp = g->must;
   1667 	scan = start;
   1668 	for (i = g->mlen; i > 0; i--) {
   1669 		while (OP(s = *scan++) != OCHAR)
   1670 			continue;
   1671 		assert(cp < g->must + g->mlen);
   1672 		*cp++ = (char)OPND(s);
   1673 	}
   1674 	assert(cp == g->must + g->mlen);
   1675 	*cp++ = '\0';		/* just on general principles */
   1676 }
   1677 
   1678 /*
   1679  - pluscount - count + nesting
   1680  == static sopno pluscount(struct parse *p, struct re_guts *g);
   1681  */
   1682 static sopno			/* nesting depth */
   1683 pluscount(p, g)
   1684 struct parse *p;
   1685 struct re_guts *g;
   1686 {
   1687 	sop *scan;
   1688 	sop s;
   1689 	sopno plusnest = 0;
   1690 	sopno maxnest = 0;
   1691 
   1692 	if (p->error != 0)
   1693 		return(0);	/* there may not be an OEND */
   1694 
   1695 	scan = g->strip + 1;
   1696 	do {
   1697 		s = *scan++;
   1698 		switch (OP(s)) {
   1699 		case OPLUS_:
   1700 			plusnest++;
   1701 			break;
   1702 		case O_PLUS:
   1703 			if (plusnest > maxnest)
   1704 				maxnest = plusnest;
   1705 			plusnest--;
   1706 			break;
   1707 		}
   1708 	} while (OP(s) != OEND);
   1709 	if (plusnest != 0)
   1710 		g->iflags |= BAD;
   1711 	return(maxnest);
   1712 }
   1713