1/*-
2 * This code is derived from OpenBSD's libc/regex, original license follows:
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. Neither the name of the University nor the names of its contributors
20 *    may be used to endorse or promote products derived from this software
21 *    without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 *
35 *	@(#)regcomp.c	8.5 (Berkeley) 3/20/94
36 */
37
38#include <sys/types.h>
39#include <stdio.h>
40#include <string.h>
41#include <ctype.h>
42#include <limits.h>
43#include <stdlib.h>
44#include "regex_impl.h"
45
46#include "regutils.h"
47#include "regex2.h"
48
49#include "regcclass.h"
50#include "regcname.h"
51
52#include "llvm/Config/config.h"
53#if HAVE_STDINT_H
54#include <stdint.h>
55#else
56/* Pessimistically bound memory use */
57#define SIZE_MAX UINT_MAX
58#endif
59
60/*
61 * parse structure, passed up and down to avoid global variables and
62 * other clumsinesses
63 */
64struct parse {
65	char *next;		/* next character in RE */
66	char *end;		/* end of string (-> NUL normally) */
67	int error;		/* has an error been seen? */
68	sop *strip;		/* malloced strip */
69	sopno ssize;		/* malloced strip size (allocated) */
70	sopno slen;		/* malloced strip length (used) */
71	int ncsalloc;		/* number of csets allocated */
72	struct re_guts *g;
73#	define	NPAREN	10	/* we need to remember () 1-9 for back refs */
74	sopno pbegin[NPAREN];	/* -> ( ([0] unused) */
75	sopno pend[NPAREN];	/* -> ) ([0] unused) */
76};
77
78static void p_ere(struct parse *, int);
79static void p_ere_exp(struct parse *);
80static void p_str(struct parse *);
81static void p_bre(struct parse *, int, int);
82static int p_simp_re(struct parse *, int);
83static int p_count(struct parse *);
84static void p_bracket(struct parse *);
85static void p_b_term(struct parse *, cset *);
86static void p_b_cclass(struct parse *, cset *);
87static void p_b_eclass(struct parse *, cset *);
88static char p_b_symbol(struct parse *);
89static char p_b_coll_elem(struct parse *, int);
90static char othercase(int);
91static void bothcases(struct parse *, int);
92static void ordinary(struct parse *, int);
93static void nonnewline(struct parse *);
94static void repeat(struct parse *, sopno, int, int);
95static int seterr(struct parse *, int);
96static cset *allocset(struct parse *);
97static void freeset(struct parse *, cset *);
98static int freezeset(struct parse *, cset *);
99static int firstch(struct parse *, cset *);
100static int nch(struct parse *, cset *);
101static void mcadd(struct parse *, cset *, const char *);
102static void mcinvert(struct parse *, cset *);
103static void mccase(struct parse *, cset *);
104static int isinsets(struct re_guts *, int);
105static int samesets(struct re_guts *, int, int);
106static void categorize(struct parse *, struct re_guts *);
107static sopno dupl(struct parse *, sopno, sopno);
108static void doemit(struct parse *, sop, size_t);
109static void doinsert(struct parse *, sop, size_t, sopno);
110static void dofwd(struct parse *, sopno, sop);
111static void enlarge(struct parse *, sopno);
112static void stripsnug(struct parse *, struct re_guts *);
113static void findmust(struct parse *, struct re_guts *);
114static sopno pluscount(struct parse *, struct re_guts *);
115
116static char nuls[10];		/* place to point scanner in event of error */
117
118/*
119 * macros for use with parse structure
120 * BEWARE:  these know that the parse structure is named `p' !!!
121 */
122#define	PEEK()	(*p->next)
123#define	PEEK2()	(*(p->next+1))
124#define	MORE()	(p->next < p->end)
125#define	MORE2()	(p->next+1 < p->end)
126#define	SEE(c)	(MORE() && PEEK() == (c))
127#define	SEETWO(a, b)	(MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
128#define	EAT(c)	((SEE(c)) ? (NEXT(), 1) : 0)
129#define	EATTWO(a, b)	((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
130#define	NEXT()	(p->next++)
131#define	NEXT2()	(p->next += 2)
132#define	NEXTn(n)	(p->next += (n))
133#define	GETNEXT()	(*p->next++)
134#define	SETERROR(e)	seterr(p, (e))
135#define	REQUIRE(co, e)	(void)((co) || SETERROR(e))
136#define	MUSTSEE(c, e)	(REQUIRE(MORE() && PEEK() == (c), e))
137#define	MUSTEAT(c, e)	(REQUIRE(MORE() && GETNEXT() == (c), e))
138#define	MUSTNOTSEE(c, e)	(REQUIRE(!MORE() || PEEK() != (c), e))
139#define	EMIT(op, sopnd)	doemit(p, (sop)(op), (size_t)(sopnd))
140#define	INSERT(op, pos)	doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
141#define	AHEAD(pos)		dofwd(p, pos, HERE()-(pos))
142#define	ASTERN(sop, pos)	EMIT(sop, HERE()-pos)
143#define	HERE()		(p->slen)
144#define	THERE()		(p->slen - 1)
145#define	THERETHERE()	(p->slen - 2)
146#define	DROP(n)	(p->slen -= (n))
147
148#ifdef	_POSIX2_RE_DUP_MAX
149#define	DUPMAX	_POSIX2_RE_DUP_MAX
150#else
151#define	DUPMAX	255
152#endif
153#define	INFINITY	(DUPMAX + 1)
154
155#ifndef NDEBUG
156static int never = 0;		/* for use in asserts; shuts lint up */
157#else
158#define	never	0		/* some <assert.h>s have bugs too */
159#endif
160
161/*
162 - llvm_regcomp - interface for parser and compilation
163 */
164int				/* 0 success, otherwise REG_something */
165llvm_regcomp(llvm_regex_t *preg, const char *pattern, int cflags)
166{
167	struct parse pa;
168	struct re_guts *g;
169	struct parse *p = &pa;
170	int i;
171	size_t len;
172#ifdef REDEBUG
173#	define	GOODFLAGS(f)	(f)
174#else
175#	define	GOODFLAGS(f)	((f)&~REG_DUMP)
176#endif
177
178	cflags = GOODFLAGS(cflags);
179	if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
180		return(REG_INVARG);
181
182	if (cflags&REG_PEND) {
183		if (preg->re_endp < pattern)
184			return(REG_INVARG);
185		len = preg->re_endp - pattern;
186	} else
187		len = strlen((const char *)pattern);
188
189	/* do the mallocs early so failure handling is easy */
190	g = (struct re_guts *)malloc(sizeof(struct re_guts) +
191							(NC-1)*sizeof(cat_t));
192	if (g == NULL)
193		return(REG_ESPACE);
194	p->ssize = len/(size_t)2*(size_t)3 + (size_t)1;	/* ugh */
195	p->strip = (sop *)calloc(p->ssize, sizeof(sop));
196	p->slen = 0;
197	if (p->strip == NULL) {
198		free((char *)g);
199		return(REG_ESPACE);
200	}
201
202	/* set things up */
203	p->g = g;
204	p->next = (char *)pattern;	/* convenience; we do not modify it */
205	p->end = p->next + len;
206	p->error = 0;
207	p->ncsalloc = 0;
208	for (i = 0; i < NPAREN; i++) {
209		p->pbegin[i] = 0;
210		p->pend[i] = 0;
211	}
212	g->csetsize = NC;
213	g->sets = NULL;
214	g->setbits = NULL;
215	g->ncsets = 0;
216	g->cflags = cflags;
217	g->iflags = 0;
218	g->nbol = 0;
219	g->neol = 0;
220	g->must = NULL;
221	g->mlen = 0;
222	g->nsub = 0;
223	g->ncategories = 1;	/* category 0 is "everything else" */
224	g->categories = &g->catspace[-(CHAR_MIN)];
225	(void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
226	g->backrefs = 0;
227
228	/* do it */
229	EMIT(OEND, 0);
230	g->firststate = THERE();
231	if (cflags&REG_EXTENDED)
232		p_ere(p, OUT);
233	else if (cflags&REG_NOSPEC)
234		p_str(p);
235	else
236		p_bre(p, OUT, OUT);
237	EMIT(OEND, 0);
238	g->laststate = THERE();
239
240	/* tidy up loose ends and fill things in */
241	categorize(p, g);
242	stripsnug(p, g);
243	findmust(p, g);
244	g->nplus = pluscount(p, g);
245	g->magic = MAGIC2;
246	preg->re_nsub = g->nsub;
247	preg->re_g = g;
248	preg->re_magic = MAGIC1;
249#ifndef REDEBUG
250	/* not debugging, so can't rely on the assert() in llvm_regexec() */
251	if (g->iflags&REGEX_BAD)
252		SETERROR(REG_ASSERT);
253#endif
254
255	/* win or lose, we're done */
256	if (p->error != 0)	/* lose */
257		llvm_regfree(preg);
258	return(p->error);
259}
260
261/*
262 - p_ere - ERE parser top level, concatenation and alternation
263 */
264static void
265p_ere(struct parse *p, int stop)	/* character this ERE should end at */
266{
267	char c;
268	sopno prevback = 0;
269	sopno prevfwd = 0;
270	sopno conc;
271	int first = 1;		/* is this the first alternative? */
272
273	for (;;) {
274		/* do a bunch of concatenated expressions */
275		conc = HERE();
276		while (MORE() && (c = PEEK()) != '|' && c != stop)
277			p_ere_exp(p);
278		REQUIRE(HERE() != conc, REG_EMPTY);	/* require nonempty */
279
280		if (!EAT('|'))
281			break;		/* NOTE BREAK OUT */
282
283		if (first) {
284			INSERT(OCH_, conc);	/* offset is wrong */
285			prevfwd = conc;
286			prevback = conc;
287			first = 0;
288		}
289		ASTERN(OOR1, prevback);
290		prevback = THERE();
291		AHEAD(prevfwd);			/* fix previous offset */
292		prevfwd = HERE();
293		EMIT(OOR2, 0);			/* offset is very wrong */
294	}
295
296	if (!first) {		/* tail-end fixups */
297		AHEAD(prevfwd);
298		ASTERN(O_CH, prevback);
299	}
300
301	assert(!MORE() || SEE(stop));
302}
303
304/*
305 - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
306 */
307static void
308p_ere_exp(struct parse *p)
309{
310	char c;
311	sopno pos;
312	int count;
313	int count2;
314	int backrefnum;
315	sopno subno;
316	int wascaret = 0;
317
318	assert(MORE());		/* caller should have ensured this */
319	c = GETNEXT();
320
321	pos = HERE();
322	switch (c) {
323	case '(':
324		REQUIRE(MORE(), REG_EPAREN);
325		p->g->nsub++;
326		subno = p->g->nsub;
327		if (subno < NPAREN)
328			p->pbegin[subno] = HERE();
329		EMIT(OLPAREN, subno);
330		if (!SEE(')'))
331			p_ere(p, ')');
332		if (subno < NPAREN) {
333			p->pend[subno] = HERE();
334			assert(p->pend[subno] != 0);
335		}
336		EMIT(ORPAREN, subno);
337		MUSTEAT(')', REG_EPAREN);
338		break;
339#ifndef POSIX_MISTAKE
340	case ')':		/* happens only if no current unmatched ( */
341		/*
342		 * You may ask, why the ifndef?  Because I didn't notice
343		 * this until slightly too late for 1003.2, and none of the
344		 * other 1003.2 regular-expression reviewers noticed it at
345		 * all.  So an unmatched ) is legal POSIX, at least until
346		 * we can get it fixed.
347		 */
348		SETERROR(REG_EPAREN);
349		break;
350#endif
351	case '^':
352		EMIT(OBOL, 0);
353		p->g->iflags |= USEBOL;
354		p->g->nbol++;
355		wascaret = 1;
356		break;
357	case '$':
358		EMIT(OEOL, 0);
359		p->g->iflags |= USEEOL;
360		p->g->neol++;
361		break;
362	case '|':
363		SETERROR(REG_EMPTY);
364		break;
365	case '*':
366	case '+':
367	case '?':
368		SETERROR(REG_BADRPT);
369		break;
370	case '.':
371		if (p->g->cflags&REG_NEWLINE)
372			nonnewline(p);
373		else
374			EMIT(OANY, 0);
375		break;
376	case '[':
377		p_bracket(p);
378		break;
379	case '\\':
380		REQUIRE(MORE(), REG_EESCAPE);
381		c = GETNEXT();
382		if (c >= '1' && c <= '9') {
383			/* \[0-9] is taken to be a back-reference to a previously specified
384			 * matching group. backrefnum will hold the number. The matching
385			 * group must exist (i.e. if \4 is found there must have been at
386			 * least 4 matching groups specified in the pattern previously).
387			 */
388			backrefnum = c - '0';
389			if (p->pend[backrefnum] == 0) {
390				SETERROR(REG_ESUBREG);
391				break;
392			}
393
394			/* Make sure everything checks out and emit the sequence
395			 * that marks a back-reference to the parse structure.
396			 */
397			assert(backrefnum <= p->g->nsub);
398			EMIT(OBACK_, backrefnum);
399			assert(p->pbegin[backrefnum] != 0);
400			assert(OP(p->strip[p->pbegin[backrefnum]]) != OLPAREN);
401			assert(OP(p->strip[p->pend[backrefnum]]) != ORPAREN);
402			(void) dupl(p, p->pbegin[backrefnum]+1, p->pend[backrefnum]);
403			EMIT(O_BACK, backrefnum);
404			p->g->backrefs = 1;
405		} else {
406			/* Other chars are simply themselves when escaped with a backslash.
407			 */
408			ordinary(p, c);
409		}
410		break;
411	case '{':		/* okay as ordinary except if digit follows */
412		REQUIRE(!MORE() || !isdigit((uch)PEEK()), REG_BADRPT);
413		/* FALLTHROUGH */
414	default:
415		ordinary(p, c);
416		break;
417	}
418
419	if (!MORE())
420		return;
421	c = PEEK();
422	/* we call { a repetition if followed by a digit */
423	if (!( c == '*' || c == '+' || c == '?' ||
424				(c == '{' && MORE2() && isdigit((uch)PEEK2())) ))
425		return;		/* no repetition, we're done */
426	NEXT();
427
428	REQUIRE(!wascaret, REG_BADRPT);
429	switch (c) {
430	case '*':	/* implemented as +? */
431		/* this case does not require the (y|) trick, noKLUDGE */
432		INSERT(OPLUS_, pos);
433		ASTERN(O_PLUS, pos);
434		INSERT(OQUEST_, pos);
435		ASTERN(O_QUEST, pos);
436		break;
437	case '+':
438		INSERT(OPLUS_, pos);
439		ASTERN(O_PLUS, pos);
440		break;
441	case '?':
442		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
443		INSERT(OCH_, pos);		/* offset slightly wrong */
444		ASTERN(OOR1, pos);		/* this one's right */
445		AHEAD(pos);			/* fix the OCH_ */
446		EMIT(OOR2, 0);			/* offset very wrong... */
447		AHEAD(THERE());			/* ...so fix it */
448		ASTERN(O_CH, THERETHERE());
449		break;
450	case '{':
451		count = p_count(p);
452		if (EAT(',')) {
453			if (isdigit((uch)PEEK())) {
454				count2 = p_count(p);
455				REQUIRE(count <= count2, REG_BADBR);
456			} else		/* single number with comma */
457				count2 = INFINITY;
458		} else		/* just a single number */
459			count2 = count;
460		repeat(p, pos, count, count2);
461		if (!EAT('}')) {	/* error heuristics */
462			while (MORE() && PEEK() != '}')
463				NEXT();
464			REQUIRE(MORE(), REG_EBRACE);
465			SETERROR(REG_BADBR);
466		}
467		break;
468	}
469
470	if (!MORE())
471		return;
472	c = PEEK();
473	if (!( c == '*' || c == '+' || c == '?' ||
474				(c == '{' && MORE2() && isdigit((uch)PEEK2())) ) )
475		return;
476	SETERROR(REG_BADRPT);
477}
478
479/*
480 - p_str - string (no metacharacters) "parser"
481 */
482static void
483p_str(struct parse *p)
484{
485	REQUIRE(MORE(), REG_EMPTY);
486	while (MORE())
487		ordinary(p, GETNEXT());
488}
489
490/*
491 - p_bre - BRE parser top level, anchoring and concatenation
492 * Giving end1 as OUT essentially eliminates the end1/end2 check.
493 *
494 * This implementation is a bit of a kludge, in that a trailing $ is first
495 * taken as an ordinary character and then revised to be an anchor.  The
496 * only undesirable side effect is that '$' gets included as a character
497 * category in such cases.  This is fairly harmless; not worth fixing.
498 * The amount of lookahead needed to avoid this kludge is excessive.
499 */
500static void
501p_bre(struct parse *p,
502    int end1,		/* first terminating character */
503    int end2)		/* second terminating character */
504{
505	sopno start = HERE();
506	int first = 1;			/* first subexpression? */
507	int wasdollar = 0;
508
509	if (EAT('^')) {
510		EMIT(OBOL, 0);
511		p->g->iflags |= USEBOL;
512		p->g->nbol++;
513	}
514	while (MORE() && !SEETWO(end1, end2)) {
515		wasdollar = p_simp_re(p, first);
516		first = 0;
517	}
518	if (wasdollar) {	/* oops, that was a trailing anchor */
519		DROP(1);
520		EMIT(OEOL, 0);
521		p->g->iflags |= USEEOL;
522		p->g->neol++;
523	}
524
525	REQUIRE(HERE() != start, REG_EMPTY);	/* require nonempty */
526}
527
528/*
529 - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
530 */
531static int			/* was the simple RE an unbackslashed $? */
532p_simp_re(struct parse *p,
533    int starordinary)		/* is a leading * an ordinary character? */
534{
535	int c;
536	int count;
537	int count2;
538	sopno pos;
539	int i;
540	sopno subno;
541#	define	BACKSL	(1<<CHAR_BIT)
542
543        pos = HERE(); /* repetition op, if any, covers from here */
544
545        assert(MORE()); /* caller should have ensured this */
546        c = GETNEXT();
547	if (c == '\\') {
548		REQUIRE(MORE(), REG_EESCAPE);
549		c = BACKSL | GETNEXT();
550	}
551	switch (c) {
552	case '.':
553		if (p->g->cflags&REG_NEWLINE)
554			nonnewline(p);
555		else
556			EMIT(OANY, 0);
557		break;
558	case '[':
559		p_bracket(p);
560		break;
561	case BACKSL|'{':
562		SETERROR(REG_BADRPT);
563		break;
564	case BACKSL|'(':
565		p->g->nsub++;
566		subno = p->g->nsub;
567		if (subno < NPAREN)
568			p->pbegin[subno] = HERE();
569		EMIT(OLPAREN, subno);
570		/* the MORE here is an error heuristic */
571		if (MORE() && !SEETWO('\\', ')'))
572			p_bre(p, '\\', ')');
573		if (subno < NPAREN) {
574			p->pend[subno] = HERE();
575			assert(p->pend[subno] != 0);
576		}
577		EMIT(ORPAREN, subno);
578		REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
579		break;
580	case BACKSL|')':	/* should not get here -- must be user */
581	case BACKSL|'}':
582		SETERROR(REG_EPAREN);
583		break;
584	case BACKSL|'1':
585	case BACKSL|'2':
586	case BACKSL|'3':
587	case BACKSL|'4':
588	case BACKSL|'5':
589	case BACKSL|'6':
590	case BACKSL|'7':
591	case BACKSL|'8':
592	case BACKSL|'9':
593		i = (c&~BACKSL) - '0';
594		assert(i < NPAREN);
595		if (p->pend[i] != 0) {
596			assert(i <= p->g->nsub);
597			EMIT(OBACK_, i);
598			assert(p->pbegin[i] != 0);
599			assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
600			assert(OP(p->strip[p->pend[i]]) == ORPAREN);
601			(void) dupl(p, p->pbegin[i]+1, p->pend[i]);
602			EMIT(O_BACK, i);
603		} else
604			SETERROR(REG_ESUBREG);
605		p->g->backrefs = 1;
606		break;
607	case '*':
608		REQUIRE(starordinary, REG_BADRPT);
609		/* FALLTHROUGH */
610	default:
611		ordinary(p, (char)c);
612		break;
613	}
614
615	if (EAT('*')) {		/* implemented as +? */
616		/* this case does not require the (y|) trick, noKLUDGE */
617		INSERT(OPLUS_, pos);
618		ASTERN(O_PLUS, pos);
619		INSERT(OQUEST_, pos);
620		ASTERN(O_QUEST, pos);
621	} else if (EATTWO('\\', '{')) {
622		count = p_count(p);
623		if (EAT(',')) {
624			if (MORE() && isdigit((uch)PEEK())) {
625				count2 = p_count(p);
626				REQUIRE(count <= count2, REG_BADBR);
627			} else		/* single number with comma */
628				count2 = INFINITY;
629		} else		/* just a single number */
630			count2 = count;
631		repeat(p, pos, count, count2);
632		if (!EATTWO('\\', '}')) {	/* error heuristics */
633			while (MORE() && !SEETWO('\\', '}'))
634				NEXT();
635			REQUIRE(MORE(), REG_EBRACE);
636			SETERROR(REG_BADBR);
637		}
638	} else if (c == '$')	/* $ (but not \$) ends it */
639		return(1);
640
641	return(0);
642}
643
644/*
645 - p_count - parse a repetition count
646 */
647static int			/* the value */
648p_count(struct parse *p)
649{
650	int count = 0;
651	int ndigits = 0;
652
653	while (MORE() && isdigit((uch)PEEK()) && count <= DUPMAX) {
654		count = count*10 + (GETNEXT() - '0');
655		ndigits++;
656	}
657
658	REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
659	return(count);
660}
661
662/*
663 - p_bracket - parse a bracketed character list
664 *
665 * Note a significant property of this code:  if the allocset() did SETERROR,
666 * no set operations are done.
667 */
668static void
669p_bracket(struct parse *p)
670{
671	cset *cs;
672	int invert = 0;
673
674	/* Dept of Truly Sickening Special-Case Kludges */
675	if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
676		EMIT(OBOW, 0);
677		NEXTn(6);
678		return;
679	}
680	if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
681		EMIT(OEOW, 0);
682		NEXTn(6);
683		return;
684	}
685
686	if ((cs = allocset(p)) == NULL) {
687		/* allocset did set error status in p */
688		return;
689	}
690
691	if (EAT('^'))
692		invert++;	/* make note to invert set at end */
693	if (EAT(']'))
694		CHadd(cs, ']');
695	else if (EAT('-'))
696		CHadd(cs, '-');
697	while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
698		p_b_term(p, cs);
699	if (EAT('-'))
700		CHadd(cs, '-');
701	MUSTEAT(']', REG_EBRACK);
702
703	if (p->error != 0) {	/* don't mess things up further */
704		freeset(p, cs);
705		return;
706	}
707
708	if (p->g->cflags&REG_ICASE) {
709		int i;
710		int ci;
711
712		for (i = p->g->csetsize - 1; i >= 0; i--)
713			if (CHIN(cs, i) && isalpha(i)) {
714				ci = othercase(i);
715				if (ci != i)
716					CHadd(cs, ci);
717			}
718		if (cs->multis != NULL)
719			mccase(p, cs);
720	}
721	if (invert) {
722		int i;
723
724		for (i = p->g->csetsize - 1; i >= 0; i--)
725			if (CHIN(cs, i))
726				CHsub(cs, i);
727			else
728				CHadd(cs, i);
729		if (p->g->cflags&REG_NEWLINE)
730			CHsub(cs, '\n');
731		if (cs->multis != NULL)
732			mcinvert(p, cs);
733	}
734
735	assert(cs->multis == NULL);		/* xxx */
736
737	if (nch(p, cs) == 1) {		/* optimize singleton sets */
738		ordinary(p, firstch(p, cs));
739		freeset(p, cs);
740	} else
741		EMIT(OANYOF, freezeset(p, cs));
742}
743
744/*
745 - p_b_term - parse one term of a bracketed character list
746 */
747static void
748p_b_term(struct parse *p, cset *cs)
749{
750	char c;
751	char start, finish;
752	int i;
753
754	/* classify what we've got */
755	switch ((MORE()) ? PEEK() : '\0') {
756	case '[':
757		c = (MORE2()) ? PEEK2() : '\0';
758		break;
759	case '-':
760		SETERROR(REG_ERANGE);
761		return;			/* NOTE RETURN */
762		break;
763	default:
764		c = '\0';
765		break;
766	}
767
768	switch (c) {
769	case ':':		/* character class */
770		NEXT2();
771		REQUIRE(MORE(), REG_EBRACK);
772		c = PEEK();
773		REQUIRE(c != '-' && c != ']', REG_ECTYPE);
774		p_b_cclass(p, cs);
775		REQUIRE(MORE(), REG_EBRACK);
776		REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
777		break;
778	case '=':		/* equivalence class */
779		NEXT2();
780		REQUIRE(MORE(), REG_EBRACK);
781		c = PEEK();
782		REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
783		p_b_eclass(p, cs);
784		REQUIRE(MORE(), REG_EBRACK);
785		REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
786		break;
787	default:		/* symbol, ordinary character, or range */
788/* xxx revision needed for multichar stuff */
789		start = p_b_symbol(p);
790		if (SEE('-') && MORE2() && PEEK2() != ']') {
791			/* range */
792			NEXT();
793			if (EAT('-'))
794				finish = '-';
795			else
796				finish = p_b_symbol(p);
797		} else
798			finish = start;
799/* xxx what about signed chars here... */
800		REQUIRE(start <= finish, REG_ERANGE);
801		for (i = start; i <= finish; i++)
802			CHadd(cs, i);
803		break;
804	}
805}
806
807/*
808 - p_b_cclass - parse a character-class name and deal with it
809 */
810static void
811p_b_cclass(struct parse *p, cset *cs)
812{
813	char *sp = p->next;
814	struct cclass *cp;
815	size_t len;
816	const char *u;
817	char c;
818
819	while (MORE() && isalpha((uch)PEEK()))
820		NEXT();
821	len = p->next - sp;
822	for (cp = cclasses; cp->name != NULL; cp++)
823		if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
824			break;
825	if (cp->name == NULL) {
826		/* oops, didn't find it */
827		SETERROR(REG_ECTYPE);
828		return;
829	}
830
831	u = cp->chars;
832	while ((c = *u++) != '\0')
833		CHadd(cs, c);
834	for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
835		MCadd(p, cs, u);
836}
837
838/*
839 - p_b_eclass - parse an equivalence-class name and deal with it
840 *
841 * This implementation is incomplete. xxx
842 */
843static void
844p_b_eclass(struct parse *p, cset *cs)
845{
846	char c;
847
848	c = p_b_coll_elem(p, '=');
849	CHadd(cs, c);
850}
851
852/*
853 - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
854 */
855static char			/* value of symbol */
856p_b_symbol(struct parse *p)
857{
858	char value;
859
860	REQUIRE(MORE(), REG_EBRACK);
861	if (!EATTWO('[', '.'))
862		return(GETNEXT());
863
864	/* collating symbol */
865	value = p_b_coll_elem(p, '.');
866	REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
867	return(value);
868}
869
870/*
871 - p_b_coll_elem - parse a collating-element name and look it up
872 */
873static char			/* value of collating element */
874p_b_coll_elem(struct parse *p,
875    int endc)			/* name ended by endc,']' */
876{
877	char *sp = p->next;
878	struct cname *cp;
879	int len;
880
881	while (MORE() && !SEETWO(endc, ']'))
882		NEXT();
883	if (!MORE()) {
884		SETERROR(REG_EBRACK);
885		return(0);
886	}
887	len = p->next - sp;
888	for (cp = cnames; cp->name != NULL; cp++)
889		if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
890			return(cp->code);	/* known name */
891	if (len == 1)
892		return(*sp);	/* single character */
893	SETERROR(REG_ECOLLATE);			/* neither */
894	return(0);
895}
896
897/*
898 - othercase - return the case counterpart of an alphabetic
899 */
900static char			/* if no counterpart, return ch */
901othercase(int ch)
902{
903	ch = (uch)ch;
904	assert(isalpha(ch));
905	if (isupper(ch))
906		return ((uch)tolower(ch));
907	else if (islower(ch))
908		return ((uch)toupper(ch));
909	else			/* peculiar, but could happen */
910		return(ch);
911}
912
913/*
914 - bothcases - emit a dualcase version of a two-case character
915 *
916 * Boy, is this implementation ever a kludge...
917 */
918static void
919bothcases(struct parse *p, int ch)
920{
921	char *oldnext = p->next;
922	char *oldend = p->end;
923	char bracket[3];
924
925	ch = (uch)ch;
926	assert(othercase(ch) != ch);	/* p_bracket() would recurse */
927	p->next = bracket;
928	p->end = bracket+2;
929	bracket[0] = ch;
930	bracket[1] = ']';
931	bracket[2] = '\0';
932	p_bracket(p);
933	assert(p->next == bracket+2);
934	p->next = oldnext;
935	p->end = oldend;
936}
937
938/*
939 - ordinary - emit an ordinary character
940 */
941static void
942ordinary(struct parse *p, int ch)
943{
944	cat_t *cap = p->g->categories;
945
946	if ((p->g->cflags&REG_ICASE) && isalpha((uch)ch) && othercase(ch) != ch)
947		bothcases(p, ch);
948	else {
949		EMIT(OCHAR, (uch)ch);
950		if (cap[ch] == 0)
951			cap[ch] = p->g->ncategories++;
952	}
953}
954
955/*
956 - nonnewline - emit REG_NEWLINE version of OANY
957 *
958 * Boy, is this implementation ever a kludge...
959 */
960static void
961nonnewline(struct parse *p)
962{
963	char *oldnext = p->next;
964	char *oldend = p->end;
965	char bracket[4];
966
967	p->next = bracket;
968	p->end = bracket+3;
969	bracket[0] = '^';
970	bracket[1] = '\n';
971	bracket[2] = ']';
972	bracket[3] = '\0';
973	p_bracket(p);
974	assert(p->next == bracket+3);
975	p->next = oldnext;
976	p->end = oldend;
977}
978
979/*
980 - repeat - generate code for a bounded repetition, recursively if needed
981 */
982static void
983repeat(struct parse *p,
984    sopno start,		/* operand from here to end of strip */
985    int from,			/* repeated from this number */
986    int to)			/* to this number of times (maybe INFINITY) */
987{
988	sopno finish = HERE();
989#	define	N	2
990#	define	INF	3
991#	define	REP(f, t)	((f)*8 + (t))
992#	define	MAP(n)	(((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
993	sopno copy;
994
995	if (p->error != 0)	/* head off possible runaway recursion */
996		return;
997
998	assert(from <= to);
999
1000	switch (REP(MAP(from), MAP(to))) {
1001	case REP(0, 0):			/* must be user doing this */
1002		DROP(finish-start);	/* drop the operand */
1003		break;
1004	case REP(0, 1):			/* as x{1,1}? */
1005	case REP(0, N):			/* as x{1,n}? */
1006	case REP(0, INF):		/* as x{1,}? */
1007		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1008		INSERT(OCH_, start);		/* offset is wrong... */
1009		repeat(p, start+1, 1, to);
1010		ASTERN(OOR1, start);
1011		AHEAD(start);			/* ... fix it */
1012		EMIT(OOR2, 0);
1013		AHEAD(THERE());
1014		ASTERN(O_CH, THERETHERE());
1015		break;
1016	case REP(1, 1):			/* trivial case */
1017		/* done */
1018		break;
1019	case REP(1, N):			/* as x?x{1,n-1} */
1020		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1021		INSERT(OCH_, start);
1022		ASTERN(OOR1, start);
1023		AHEAD(start);
1024		EMIT(OOR2, 0);			/* offset very wrong... */
1025		AHEAD(THERE());			/* ...so fix it */
1026		ASTERN(O_CH, THERETHERE());
1027		copy = dupl(p, start+1, finish+1);
1028		assert(copy == finish+4);
1029		repeat(p, copy, 1, to-1);
1030		break;
1031	case REP(1, INF):		/* as x+ */
1032		INSERT(OPLUS_, start);
1033		ASTERN(O_PLUS, start);
1034		break;
1035	case REP(N, N):			/* as xx{m-1,n-1} */
1036		copy = dupl(p, start, finish);
1037		repeat(p, copy, from-1, to-1);
1038		break;
1039	case REP(N, INF):		/* as xx{n-1,INF} */
1040		copy = dupl(p, start, finish);
1041		repeat(p, copy, from-1, to);
1042		break;
1043	default:			/* "can't happen" */
1044		SETERROR(REG_ASSERT);	/* just in case */
1045		break;
1046	}
1047}
1048
1049/*
1050 - seterr - set an error condition
1051 */
1052static int			/* useless but makes type checking happy */
1053seterr(struct parse *p, int e)
1054{
1055	if (p->error == 0)	/* keep earliest error condition */
1056		p->error = e;
1057	p->next = nuls;		/* try to bring things to a halt */
1058	p->end = nuls;
1059	return(0);		/* make the return value well-defined */
1060}
1061
1062/*
1063 - allocset - allocate a set of characters for []
1064 */
1065static cset *
1066allocset(struct parse *p)
1067{
1068	int no = p->g->ncsets++;
1069	size_t nc;
1070	size_t nbytes;
1071	cset *cs;
1072	size_t css = (size_t)p->g->csetsize;
1073	int i;
1074
1075	if (no >= p->ncsalloc) {	/* need another column of space */
1076		void *ptr;
1077
1078		p->ncsalloc += CHAR_BIT;
1079		nc = p->ncsalloc;
1080		if (nc > SIZE_MAX / sizeof(cset))
1081			goto nomem;
1082		assert(nc % CHAR_BIT == 0);
1083		nbytes = nc / CHAR_BIT * css;
1084
1085		ptr = (cset *)realloc((char *)p->g->sets, nc * sizeof(cset));
1086		if (ptr == NULL)
1087			goto nomem;
1088		p->g->sets = ptr;
1089
1090		ptr = (uch *)realloc((char *)p->g->setbits, nbytes);
1091		if (ptr == NULL)
1092			goto nomem;
1093		p->g->setbits = ptr;
1094
1095		for (i = 0; i < no; i++)
1096			p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
1097
1098		(void) memset((char *)p->g->setbits + (nbytes - css), 0, css);
1099	}
1100	/* XXX should not happen */
1101	if (p->g->sets == NULL || p->g->setbits == NULL)
1102		goto nomem;
1103
1104	cs = &p->g->sets[no];
1105	cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
1106	cs->mask = 1 << ((no) % CHAR_BIT);
1107	cs->hash = 0;
1108	cs->smultis = 0;
1109	cs->multis = NULL;
1110
1111	return(cs);
1112nomem:
1113	free(p->g->sets);
1114	p->g->sets = NULL;
1115	free(p->g->setbits);
1116	p->g->setbits = NULL;
1117
1118	SETERROR(REG_ESPACE);
1119	/* caller's responsibility not to do set ops */
1120	return(NULL);
1121}
1122
1123/*
1124 - freeset - free a now-unused set
1125 */
1126static void
1127freeset(struct parse *p, cset *cs)
1128{
1129	size_t i;
1130	cset *top = &p->g->sets[p->g->ncsets];
1131	size_t css = (size_t)p->g->csetsize;
1132
1133	for (i = 0; i < css; i++)
1134		CHsub(cs, i);
1135	if (cs == top-1)	/* recover only the easy case */
1136		p->g->ncsets--;
1137}
1138
1139/*
1140 - freezeset - final processing on a set of characters
1141 *
1142 * The main task here is merging identical sets.  This is usually a waste
1143 * of time (although the hash code minimizes the overhead), but can win
1144 * big if REG_ICASE is being used.  REG_ICASE, by the way, is why the hash
1145 * is done using addition rather than xor -- all ASCII [aA] sets xor to
1146 * the same value!
1147 */
1148static int			/* set number */
1149freezeset(struct parse *p, cset *cs)
1150{
1151	uch h = cs->hash;
1152	size_t i;
1153	cset *top = &p->g->sets[p->g->ncsets];
1154	cset *cs2;
1155	size_t css = (size_t)p->g->csetsize;
1156
1157	/* look for an earlier one which is the same */
1158	for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
1159		if (cs2->hash == h && cs2 != cs) {
1160			/* maybe */
1161			for (i = 0; i < css; i++)
1162				if (!!CHIN(cs2, i) != !!CHIN(cs, i))
1163					break;		/* no */
1164			if (i == css)
1165				break;			/* yes */
1166		}
1167
1168	if (cs2 < top) {	/* found one */
1169		freeset(p, cs);
1170		cs = cs2;
1171	}
1172
1173	return((int)(cs - p->g->sets));
1174}
1175
1176/*
1177 - firstch - return first character in a set (which must have at least one)
1178 */
1179static int			/* character; there is no "none" value */
1180firstch(struct parse *p, cset *cs)
1181{
1182	size_t i;
1183	size_t css = (size_t)p->g->csetsize;
1184
1185	for (i = 0; i < css; i++)
1186		if (CHIN(cs, i))
1187			return((char)i);
1188	assert(never);
1189	return(0);		/* arbitrary */
1190}
1191
1192/*
1193 - nch - number of characters in a set
1194 */
1195static int
1196nch(struct parse *p, cset *cs)
1197{
1198	size_t i;
1199	size_t css = (size_t)p->g->csetsize;
1200	int n = 0;
1201
1202	for (i = 0; i < css; i++)
1203		if (CHIN(cs, i))
1204			n++;
1205	return(n);
1206}
1207
1208/*
1209 - mcadd - add a collating element to a cset
1210 */
1211static void
1212mcadd( struct parse *p, cset *cs, const char *cp)
1213{
1214	size_t oldend = cs->smultis;
1215	void *np;
1216
1217	cs->smultis += strlen(cp) + 1;
1218	np = realloc(cs->multis, cs->smultis);
1219	if (np == NULL) {
1220		if (cs->multis)
1221			free(cs->multis);
1222		cs->multis = NULL;
1223		SETERROR(REG_ESPACE);
1224		return;
1225	}
1226	cs->multis = np;
1227
1228	llvm_strlcpy(cs->multis + oldend - 1, cp, cs->smultis - oldend + 1);
1229}
1230
1231/*
1232 - mcinvert - invert the list of collating elements in a cset
1233 *
1234 * This would have to know the set of possibilities.  Implementation
1235 * is deferred.
1236 */
1237/* ARGSUSED */
1238static void
1239mcinvert(struct parse *p, cset *cs)
1240{
1241	assert(cs->multis == NULL);	/* xxx */
1242}
1243
1244/*
1245 - mccase - add case counterparts of the list of collating elements in a cset
1246 *
1247 * This would have to know the set of possibilities.  Implementation
1248 * is deferred.
1249 */
1250/* ARGSUSED */
1251static void
1252mccase(struct parse *p, cset *cs)
1253{
1254	assert(cs->multis == NULL);	/* xxx */
1255}
1256
1257/*
1258 - isinsets - is this character in any sets?
1259 */
1260static int			/* predicate */
1261isinsets(struct re_guts *g, int c)
1262{
1263	uch *col;
1264	int i;
1265	int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1266	unsigned uc = (uch)c;
1267
1268	for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1269		if (col[uc] != 0)
1270			return(1);
1271	return(0);
1272}
1273
1274/*
1275 - samesets - are these two characters in exactly the same sets?
1276 */
1277static int			/* predicate */
1278samesets(struct re_guts *g, int c1, int c2)
1279{
1280	uch *col;
1281	int i;
1282	int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1283	unsigned uc1 = (uch)c1;
1284	unsigned uc2 = (uch)c2;
1285
1286	for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1287		if (col[uc1] != col[uc2])
1288			return(0);
1289	return(1);
1290}
1291
1292/*
1293 - categorize - sort out character categories
1294 */
1295static void
1296categorize(struct parse *p, struct re_guts *g)
1297{
1298	cat_t *cats = g->categories;
1299	int c;
1300	int c2;
1301	cat_t cat;
1302
1303	/* avoid making error situations worse */
1304	if (p->error != 0)
1305		return;
1306
1307	for (c = CHAR_MIN; c <= CHAR_MAX; c++)
1308		if (cats[c] == 0 && isinsets(g, c)) {
1309			cat = g->ncategories++;
1310			cats[c] = cat;
1311			for (c2 = c+1; c2 <= CHAR_MAX; c2++)
1312				if (cats[c2] == 0 && samesets(g, c, c2))
1313					cats[c2] = cat;
1314		}
1315}
1316
1317/*
1318 - dupl - emit a duplicate of a bunch of sops
1319 */
1320static sopno			/* start of duplicate */
1321dupl(struct parse *p,
1322    sopno start,		/* from here */
1323    sopno finish)		/* to this less one */
1324{
1325	sopno ret = HERE();
1326	sopno len = finish - start;
1327
1328	assert(finish >= start);
1329	if (len == 0)
1330		return(ret);
1331	enlarge(p, p->ssize + len);	/* this many unexpected additions */
1332	assert(p->ssize >= p->slen + len);
1333	(void) memmove((char *)(p->strip + p->slen),
1334		(char *)(p->strip + start), (size_t)len*sizeof(sop));
1335	p->slen += len;
1336	return(ret);
1337}
1338
1339/*
1340 - doemit - emit a strip operator
1341 *
1342 * It might seem better to implement this as a macro with a function as
1343 * hard-case backup, but it's just too big and messy unless there are
1344 * some changes to the data structures.  Maybe later.
1345 */
1346static void
1347doemit(struct parse *p, sop op, size_t opnd)
1348{
1349	/* avoid making error situations worse */
1350	if (p->error != 0)
1351		return;
1352
1353	/* deal with oversize operands ("can't happen", more or less) */
1354	assert(opnd < 1<<OPSHIFT);
1355
1356	/* deal with undersized strip */
1357	if (p->slen >= p->ssize)
1358		enlarge(p, (p->ssize+1) / 2 * 3);	/* +50% */
1359	assert(p->slen < p->ssize);
1360
1361	/* finally, it's all reduced to the easy case */
1362	p->strip[p->slen++] = SOP(op, opnd);
1363}
1364
1365/*
1366 - doinsert - insert a sop into the strip
1367 */
1368static void
1369doinsert(struct parse *p, sop op, size_t opnd, sopno pos)
1370{
1371	sopno sn;
1372	sop s;
1373	int i;
1374
1375	/* avoid making error situations worse */
1376	if (p->error != 0)
1377		return;
1378
1379	sn = HERE();
1380	EMIT(op, opnd);		/* do checks, ensure space */
1381	assert(HERE() == sn+1);
1382	s = p->strip[sn];
1383
1384	/* adjust paren pointers */
1385	assert(pos > 0);
1386	for (i = 1; i < NPAREN; i++) {
1387		if (p->pbegin[i] >= pos) {
1388			p->pbegin[i]++;
1389		}
1390		if (p->pend[i] >= pos) {
1391			p->pend[i]++;
1392		}
1393	}
1394
1395	memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1396						(HERE()-pos-1)*sizeof(sop));
1397	p->strip[pos] = s;
1398}
1399
1400/*
1401 - dofwd - complete a forward reference
1402 */
1403static void
1404dofwd(struct parse *p, sopno pos, sop value)
1405{
1406	/* avoid making error situations worse */
1407	if (p->error != 0)
1408		return;
1409
1410	assert(value < 1<<OPSHIFT);
1411	p->strip[pos] = OP(p->strip[pos]) | value;
1412}
1413
1414/*
1415 - enlarge - enlarge the strip
1416 */
1417static void
1418enlarge(struct parse *p, sopno size)
1419{
1420	sop *sp;
1421
1422	if (p->ssize >= size)
1423		return;
1424
1425	if ((uintptr_t)size > SIZE_MAX / sizeof(sop)) {
1426		SETERROR(REG_ESPACE);
1427		return;
1428	}
1429
1430	sp = (sop *)realloc(p->strip, size*sizeof(sop));
1431	if (sp == NULL) {
1432		SETERROR(REG_ESPACE);
1433		return;
1434	}
1435	p->strip = sp;
1436	p->ssize = size;
1437}
1438
1439/*
1440 - stripsnug - compact the strip
1441 */
1442static void
1443stripsnug(struct parse *p, struct re_guts *g)
1444{
1445	g->nstates = p->slen;
1446	if ((uintptr_t)p->slen > SIZE_MAX / sizeof(sop)) {
1447		g->strip = p->strip;
1448		SETERROR(REG_ESPACE);
1449		return;
1450	}
1451
1452	g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1453	if (g->strip == NULL) {
1454		SETERROR(REG_ESPACE);
1455		g->strip = p->strip;
1456	}
1457}
1458
1459/*
1460 - findmust - fill in must and mlen with longest mandatory literal string
1461 *
1462 * This algorithm could do fancy things like analyzing the operands of |
1463 * for common subsequences.  Someday.  This code is simple and finds most
1464 * of the interesting cases.
1465 *
1466 * Note that must and mlen got initialized during setup.
1467 */
1468static void
1469findmust(struct parse *p, struct re_guts *g)
1470{
1471	sop *scan;
1472	sop *start = 0; /* start initialized in the default case, after that */
1473	sop *newstart = 0; /* newstart was initialized in the OCHAR case */
1474	sopno newlen;
1475	sop s;
1476	char *cp;
1477	sopno i;
1478
1479	/* avoid making error situations worse */
1480	if (p->error != 0)
1481		return;
1482
1483	/* find the longest OCHAR sequence in strip */
1484	newlen = 0;
1485	scan = g->strip + 1;
1486	do {
1487		s = *scan++;
1488		switch (OP(s)) {
1489		case OCHAR:		/* sequence member */
1490			if (newlen == 0)		/* new sequence */
1491				newstart = scan - 1;
1492			newlen++;
1493			break;
1494		case OPLUS_:		/* things that don't break one */
1495		case OLPAREN:
1496		case ORPAREN:
1497			break;
1498		case OQUEST_:		/* things that must be skipped */
1499		case OCH_:
1500			scan--;
1501			do {
1502				scan += OPND(s);
1503				s = *scan;
1504				/* assert() interferes w debug printouts */
1505				if (OP(s) != O_QUEST && OP(s) != O_CH &&
1506							OP(s) != OOR2) {
1507					g->iflags |= REGEX_BAD;
1508					return;
1509				}
1510			} while (OP(s) != O_QUEST && OP(s) != O_CH);
1511			/* fallthrough */
1512		default:		/* things that break a sequence */
1513			if (newlen > g->mlen) {		/* ends one */
1514				start = newstart;
1515				g->mlen = newlen;
1516			}
1517			newlen = 0;
1518			break;
1519		}
1520	} while (OP(s) != OEND);
1521
1522	if (g->mlen == 0)		/* there isn't one */
1523		return;
1524
1525	/* turn it into a character string */
1526	g->must = malloc((size_t)g->mlen + 1);
1527	if (g->must == NULL) {		/* argh; just forget it */
1528		g->mlen = 0;
1529		return;
1530	}
1531	cp = g->must;
1532	scan = start;
1533	for (i = g->mlen; i > 0; i--) {
1534		while (OP(s = *scan++) != OCHAR)
1535			continue;
1536		assert(cp < g->must + g->mlen);
1537		*cp++ = (char)OPND(s);
1538	}
1539	assert(cp == g->must + g->mlen);
1540	*cp++ = '\0';		/* just on general principles */
1541}
1542
1543/*
1544 - pluscount - count + nesting
1545 */
1546static sopno			/* nesting depth */
1547pluscount(struct parse *p, struct re_guts *g)
1548{
1549	sop *scan;
1550	sop s;
1551	sopno plusnest = 0;
1552	sopno maxnest = 0;
1553
1554	if (p->error != 0)
1555		return(0);	/* there may not be an OEND */
1556
1557	scan = g->strip + 1;
1558	do {
1559		s = *scan++;
1560		switch (OP(s)) {
1561		case OPLUS_:
1562			plusnest++;
1563			break;
1564		case O_PLUS:
1565			if (plusnest > maxnest)
1566				maxnest = plusnest;
1567			plusnest--;
1568			break;
1569		}
1570	} while (OP(s) != OEND);
1571	if (plusnest != 0)
1572		g->iflags |= REGEX_BAD;
1573	return(maxnest);
1574}
1575