xref: /freebsd/contrib/nvi/regex/regcomp.c (revision b97ee269eae3cbaf35c18f51a459aea581c2a7dc)
1 /*	$NetBSD: regcomp.c,v 1.7 2011/11/19 17:45:11 tnozaki 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 of the University of Toronto.
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	8.4 (Berkeley) 3/19/94
36  */
37 
38 #if defined(LIBC_SCCS) && !defined(lint)
39 static char sccsid[] = "@(#)regcomp.c	8.4 (Berkeley) 3/19/94";
40 #endif /* LIBC_SCCS and not lint */
41 
42 #include <sys/types.h>
43 #include <stdio.h>
44 #include <string.h>
45 #include <ctype.h>
46 #include <limits.h>
47 #include <stdlib.h>
48 #include <regex.h>
49 
50 #include "utils.h"
51 #include "regex2.h"
52 
53 #include "cclass.h"
54 #include "cname.h"
55 
56 /*
57  * parse structure, passed up and down to avoid global variables and
58  * other clumsinesses
59  */
60 struct parse {
61 	RCHAR_T *next;		/* next character in RE */
62 	RCHAR_T *end;		/* end of string (-> NUL normally) */
63 	int error;		/* has an error been seen? */
64 	sop *strip;		/* malloced strip */
65 	RCHAR_T *stripdata;	/* malloced stripdata */
66 	sopno ssize;		/* malloced strip size (allocated) */
67 	sopno slen;		/* malloced strip length (used) */
68 	int ncsalloc;		/* number of csets allocated */
69 	struct re_guts *g;
70 #	define	NPAREN	10	/* we need to remember () 1-9 for back refs */
71 	sopno pbegin[NPAREN];	/* -> ( ([0] unused) */
72 	sopno pend[NPAREN];	/* -> ) ([0] unused) */
73 };
74 
75 /* ========= begin header generated by ./mkh ========= */
76 #ifdef __cplusplus
77 extern "C" {
78 #endif
79 
80 /* === regcomp.c === */
81 static void p_ere(struct parse *p, int stop, size_t reclimit);
82 static void p_ere_exp(struct parse *p, size_t reclimit);
83 static void p_str(struct parse *p);
84 static void p_bre(struct parse *p, int end1, int end2, size_t reclimit);
85 static int p_simp_re(struct parse *p, int starordinary, size_t reclimit);
86 static int p_count(struct parse *p);
87 static void p_bracket(struct parse *p);
88 static void p_b_term(struct parse *p, cset *cs);
89 static void p_b_cclass(struct parse *p, cset *cs);
90 static void p_b_eclass(struct parse *p, cset *cs);
91 static char p_b_symbol(struct parse *p);
92 static char p_b_coll_elem(struct parse *p, int endc);
93 static char othercase(int ch);
94 static void bothcases(struct parse *p, int ch);
95 static void ordinary(struct parse *p, int ch);
96 static void nonnewline(struct parse *p);
97 static void repeat(struct parse *p, sopno start, int from, int to, size_t reclimit);
98 static int seterr(struct parse *p, int e);
99 static cset *allocset(struct parse *p);
100 static void freeset(struct parse *p, cset *cs);
101 static int freezeset(struct parse *p, cset *cs);
102 static int firstch(struct parse *p, cset *cs);
103 static int nch(struct parse *p, cset *cs);
104 static void mcadd(struct parse *p, cset *cs, const char *cp);
105 #ifdef notdef
106 static void mcsub(cset *cs, char *cp);
107 static int mcin(cset *cs, char *cp);
108 static char *mcfind(cset *cs, char *cp);
109 #endif
110 static void mcinvert(struct parse *p, cset *cs);
111 static void mccase(struct parse *p, cset *cs);
112 #ifdef notdef
113 static int isinsets(struct re_guts *g, int c);
114 static int samesets(struct re_guts *g, int c1, int c2);
115 #endif
116 static void categorize(struct parse *p, struct re_guts *g);
117 static sopno dupl(struct parse *p, sopno start, sopno finish);
118 static void doemit(struct parse *p, sop op, size_t opnd);
119 static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
120 static void dofwd(struct parse *p, sopno pos, sop value);
121 static int enlarge(struct parse *p, sopno size);
122 static void stripsnug(struct parse *p, struct re_guts *g);
123 static void findmust(struct parse *p, struct re_guts *g);
124 static sopno pluscount(struct parse *p, struct re_guts *g);
125 
126 #ifdef __cplusplus
127 }
128 #endif
129 /* ========= end header generated by ./mkh ========= */
130 
131 static RCHAR_T nuls[10];		/* place to point scanner in event of error */
132 
133 /*
134  * macros for use with parse structure
135  * BEWARE:  these know that the parse structure is named `p' !!!
136  */
137 #define	PEEK()	(*p->next)
138 #define	PEEK2()	(*(p->next+1))
139 #define	MORE()	(p->next < p->end)
140 #define	MORE2()	(p->next+1 < p->end)
141 #define	SEE(c)	(MORE() && PEEK() == (c))
142 #define	SEETWO(a, b)	(MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
143 #define	EAT(c)	((SEE(c)) ? (NEXT(), 1) : 0)
144 #define	EATTWO(a, b)	((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
145 #define	NEXT()	(p->next++)
146 #define	NEXT2()	(p->next += 2)
147 #define	NEXTn(n)	(p->next += (n))
148 #define	GETNEXT()	(*p->next++)
149 #define	SETERROR(e)	seterr(p, (e))
150 #define	REQUIRE(co, e)	((co) || SETERROR(e))
151 #define	MUSTSEE(c, e)	(REQUIRE(MORE() && PEEK() == (c), e))
152 #define	MUSTEAT(c, e)	(REQUIRE(MORE() && GETNEXT() == (c), e))
153 #define	MUSTNOTSEE(c, e)	(REQUIRE(!MORE() || PEEK() != (c), e))
154 #define	EMIT(op, sopnd)	doemit(p, (sop)(op), (size_t)(sopnd))
155 #define	INSERT(op, pos)	doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
156 #define	AHEAD(pos)		dofwd(p, pos, HERE()-(pos))
157 #define	ASTERN(sop, pos)	EMIT(sop, HERE()-pos)
158 #define	HERE()		(p->slen)
159 #define	THERE()		(p->slen - 1)
160 #define	THERETHERE()	(p->slen - 2)
161 #define	DROP(n)	(p->slen -= (n))
162 
163 #ifndef NDEBUG
164 static int never = 0;		/* for use in asserts; shuts lint up */
165 #else
166 #define	never	0		/* some <assert.h>s have bugs too */
167 #endif
168 
169 #define	MEMLIMIT	0x8000000
170 #define MEMSIZE(p) \
171 	((p)->ncsalloc / CHAR_BIT * (p)->g->csetsize + \
172 	(p)->ncsalloc * sizeof(cset) + \
173 	(p)->ssize * sizeof(sop))
174 #define	RECLIMIT	256
175 
176 /*
177  - regcomp - interface for parser and compilation
178  */
179 int				/* 0 success, otherwise REG_something */
180 regcomp(regex_t *preg, const RCHAR_T *pattern, int cflags)
181 {
182 	struct parse pa;
183 	struct re_guts *g;
184 	struct parse *p = &pa;
185 	int i;
186 	size_t len;
187 #ifdef REDEBUG
188 #	define	GOODFLAGS(f)	(f)
189 #else
190 #	define	GOODFLAGS(f)	((f)&~REG_DUMP)
191 #endif
192 
193 	cflags = GOODFLAGS(cflags);
194 	if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
195 		return(REG_INVARG);
196 
197 	if (cflags&REG_PEND) {
198 		if (preg->re_endp < pattern)
199 			return(REG_INVARG);
200 		len = preg->re_endp - pattern;
201 	} else
202 		len = STRLEN(pattern);
203 
204 	/* do the mallocs early so failure handling is easy */
205 	g = (struct re_guts *)malloc(sizeof(struct re_guts) +
206 							(NC-1)*sizeof(cat_t));
207 	if (g == NULL)
208 		return(REG_ESPACE);
209 	p->ssize = len/(size_t)2*(size_t)3 + (size_t)1;	/* ugh */
210 	p->strip = (sop *)malloc(p->ssize * sizeof(sop));
211 	if (p->strip == NULL) {
212 		free((char *)g);
213 		return(REG_ESPACE);
214 	}
215 	p->stripdata = (RCHAR_T *)malloc(p->ssize * sizeof(RCHAR_T));
216 	if (p->stripdata == NULL) {
217 		free((char *)p->strip);
218 		free((char *)g);
219 		return(REG_ESPACE);
220 	}
221 	p->slen = 0;
222 
223 	/* set things up */
224 	p->g = g;
225 	p->next = (RCHAR_T *)pattern;	/* convenience; we do not modify it */
226 	p->end = p->next + len;
227 	p->error = 0;
228 	p->ncsalloc = 0;
229 	for (i = 0; i < NPAREN; i++) {
230 		p->pbegin[i] = 0;
231 		p->pend[i] = 0;
232 	}
233 	g->csetsize = NC;
234 	g->sets = NULL;
235 	g->setbits = NULL;
236 	g->ncsets = 0;
237 	g->cflags = cflags;
238 	g->iflags = 0;
239 	g->nbol = 0;
240 	g->neol = 0;
241 	g->must = NULL;
242 	g->mlen = 0;
243 	g->nsub = 0;
244 #if 0
245 	g->ncategories = 1;	/* category 0 is "everything else" */
246 	g->categories = &g->catspace[-(CHAR_MIN)];
247 	memset((char *)g->catspace, 0, NC*sizeof(cat_t));
248 #endif
249 	g->backrefs = 0;
250 
251 	/* do it */
252 	EMIT(OEND, 0);
253 	g->firststate = THERE();
254 	if (cflags&REG_EXTENDED)
255 		p_ere(p, OUT, 0);
256 	else if (cflags&REG_NOSPEC)
257 		p_str(p);
258 	else
259 		p_bre(p, OUT, OUT, 0);
260 	EMIT(OEND, 0);
261 	g->laststate = THERE();
262 
263 	/* tidy up loose ends and fill things in */
264 	categorize(p, g);
265 	stripsnug(p, g);
266 	findmust(p, g);
267 	g->nplus = pluscount(p, g);
268 	g->magic = MAGIC2;
269 	preg->re_nsub = g->nsub;
270 	preg->re_g = g;
271 	preg->re_magic = MAGIC1;
272 #ifndef REDEBUG
273 	/* not debugging, so can't rely on the assert() in regexec() */
274 	if (g->iflags&BAD)
275 		SETERROR(REG_ASSERT);
276 #endif
277 
278 	/* win or lose, we're done */
279 	if (p->error != 0)	/* lose */
280 		regfree(preg);
281 	return(p->error);
282 }
283 
284 /*
285  - p_ere - ERE parser top level, concatenation and alternation
286  */
287 static void
288 p_ere(struct parse *p, int stop, size_t reclimit)
289          			/* character this ERE should end at */
290 {
291 	char c;
292 	sopno prevback = 0;
293 	sopno prevfwd = 0;
294 	sopno conc;
295 	int first = 1;		/* is this the first alternative? */
296 
297 	if (reclimit++ > RECLIMIT || p->error == REG_ESPACE) {
298 		p->error = REG_ESPACE;
299 		return;
300 	}
301 
302 	for (;;) {
303 		/* do a bunch of concatenated expressions */
304 		conc = HERE();
305 		while (MORE() && (c = PEEK()) != '|' && c != stop)
306 			p_ere_exp(p, reclimit);
307 		(void)REQUIRE(HERE() != conc, REG_EMPTY);	/* require nonempty */
308 
309 		if (!EAT('|'))
310 			break;		/* NOTE BREAK OUT */
311 
312 		if (first) {
313 			INSERT(OCH_, conc);	/* offset is wrong */
314 			prevfwd = conc;
315 			prevback = conc;
316 			first = 0;
317 		}
318 		ASTERN(OOR1, prevback);
319 		prevback = THERE();
320 		AHEAD(prevfwd);			/* fix previous offset */
321 		prevfwd = HERE();
322 		EMIT(OOR2, 0);			/* offset is very wrong */
323 	}
324 
325 	if (!first) {		/* tail-end fixups */
326 		AHEAD(prevfwd);
327 		ASTERN(O_CH, prevback);
328 	}
329 
330 	assert(!MORE() || SEE(stop));
331 }
332 
333 /*
334  - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
335  */
336 static void
337 p_ere_exp(struct parse *p, size_t reclimit)
338 {
339 	char c;
340 	sopno pos;
341 	int count;
342 	int count2;
343 	sopno subno;
344 	int wascaret = 0;
345 
346 	assert(MORE());		/* caller should have ensured this */
347 	c = GETNEXT();
348 
349 	pos = HERE();
350 	switch (c) {
351 	case '(':
352 		(void)REQUIRE(MORE(), REG_EPAREN);
353 		p->g->nsub++;
354 		subno = p->g->nsub;
355 		if (subno < NPAREN)
356 			p->pbegin[subno] = HERE();
357 		EMIT(OLPAREN, subno);
358 		if (!SEE(')'))
359 			p_ere(p, ')', reclimit);
360 		if (subno < NPAREN) {
361 			p->pend[subno] = HERE();
362 			assert(p->pend[subno] != 0);
363 		}
364 		EMIT(ORPAREN, subno);
365 		(void)MUSTEAT(')', REG_EPAREN);
366 		break;
367 #ifndef POSIX_MISTAKE
368 	case ')':		/* happens only if no current unmatched ( */
369 		/*
370 		 * You may ask, why the ifndef?  Because I didn't notice
371 		 * this until slightly too late for 1003.2, and none of the
372 		 * other 1003.2 regular-expression reviewers noticed it at
373 		 * all.  So an unmatched ) is legal POSIX, at least until
374 		 * we can get it fixed.
375 		 */
376 		SETERROR(REG_EPAREN);
377 		break;
378 #endif
379 	case '^':
380 		EMIT(OBOL, 0);
381 		p->g->iflags |= USEBOL;
382 		p->g->nbol++;
383 		wascaret = 1;
384 		break;
385 	case '$':
386 		EMIT(OEOL, 0);
387 		p->g->iflags |= USEEOL;
388 		p->g->neol++;
389 		break;
390 	case '|':
391 		SETERROR(REG_EMPTY);
392 		break;
393 	case '*':
394 	case '+':
395 	case '?':
396 		SETERROR(REG_BADRPT);
397 		break;
398 	case '.':
399 		if (p->g->cflags&REG_NEWLINE)
400 			nonnewline(p);
401 		else
402 			EMIT(OANY, 0);
403 		break;
404 	case '[':
405 		p_bracket(p);
406 		break;
407 	case '\\':
408 		(void)REQUIRE(MORE(), REG_EESCAPE);
409 		c = GETNEXT();
410 		ordinary(p, c);
411 		break;
412 	case '{':		/* okay as ordinary except if digit follows */
413 		(void)REQUIRE(!MORE() || !ISDIGIT((UCHAR_T)PEEK()), REG_BADRPT);
414 		/* FALLTHROUGH */
415 	default:
416 		ordinary(p, c);
417 		break;
418 	}
419 
420 	if (!MORE())
421 		return;
422 	c = PEEK();
423 	/* we call { a repetition if followed by a digit */
424 	if (!( c == '*' || c == '+' || c == '?' ||
425 				(c == '{' && MORE2() && ISDIGIT((UCHAR_T)PEEK2())) ))
426 		return;		/* no repetition, we're done */
427 	NEXT();
428 
429 	(void)REQUIRE(!wascaret, REG_BADRPT);
430 	switch (c) {
431 	case '*':	/* implemented as +? */
432 		/* this case does not require the (y|) trick, noKLUDGE */
433 		INSERT(OPLUS_, pos);
434 		ASTERN(O_PLUS, pos);
435 		INSERT(OQUEST_, pos);
436 		ASTERN(O_QUEST, pos);
437 		break;
438 	case '+':
439 		INSERT(OPLUS_, pos);
440 		ASTERN(O_PLUS, pos);
441 		break;
442 	case '?':
443 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
444 		INSERT(OCH_, pos);		/* offset slightly wrong */
445 		ASTERN(OOR1, pos);		/* this one's right */
446 		AHEAD(pos);			/* fix the OCH_ */
447 		EMIT(OOR2, 0);			/* offset very wrong... */
448 		AHEAD(THERE());			/* ...so fix it */
449 		ASTERN(O_CH, THERETHERE());
450 		break;
451 	case '{':
452 		count = p_count(p);
453 		if (EAT(',')) {
454 			if (ISDIGIT((UCHAR_T)PEEK())) {
455 				count2 = p_count(p);
456 				(void)REQUIRE(count <= count2, REG_BADBR);
457 			} else		/* single number with comma */
458 				count2 = INFINITY;
459 		} else		/* just a single number */
460 			count2 = count;
461 		repeat(p, pos, count, count2, 0);
462 		if (!EAT('}')) {	/* error heuristics */
463 			while (MORE() && PEEK() != '}')
464 				NEXT();
465 			(void)REQUIRE(MORE(), REG_EBRACE);
466 			SETERROR(REG_BADBR);
467 		}
468 		break;
469 	}
470 
471 	if (!MORE())
472 		return;
473 	c = PEEK();
474 	if (!( c == '*' || c == '+' || c == '?' ||
475 				(c == '{' && MORE2() && ISDIGIT((UCHAR_T)PEEK2())) ) )
476 		return;
477 	SETERROR(REG_BADRPT);
478 }
479 
480 /*
481  - p_str - string (no metacharacters) "parser"
482  */
483 static void
484 p_str(struct parse *p)
485 {
486 	(void)REQUIRE(MORE(), REG_EMPTY);
487 	while (MORE())
488 		ordinary(p, GETNEXT());
489 }
490 
491 /*
492  - p_bre - BRE parser top level, anchoring and concatenation
493  * Giving end1 as OUT essentially eliminates the end1/end2 check.
494  *
495  * This implementation is a bit of a kludge, in that a trailing $ is first
496  * taken as an ordinary character and then revised to be an anchor.  The
497  * only undesirable side effect is that '$' gets included as a character
498  * category in such cases.  This is fairly harmless; not worth fixing.
499  * The amount of lookahead needed to avoid this kludge is excessive.
500  */
501 static void
502 p_bre(struct parse *p,
503     int end1,		/* first terminating character */
504     int end2,		/* second terminating character */
505     size_t reclimit)
506 {
507 	sopno start;
508 	int first = 1;			/* first subexpression? */
509 	int wasdollar = 0;
510 
511 	if (reclimit++ > RECLIMIT || p->error == REG_ESPACE) {
512 		p->error = REG_ESPACE;
513 		return;
514 	}
515 
516 	start = HERE();
517 
518 	if (EAT('^')) {
519 		EMIT(OBOL, 0);
520 		p->g->iflags |= USEBOL;
521 		p->g->nbol++;
522 	}
523 	while (MORE() && !SEETWO(end1, end2)) {
524 		wasdollar = p_simp_re(p, first, reclimit);
525 		first = 0;
526 	}
527 	if (wasdollar) {	/* oops, that was a trailing anchor */
528 		DROP(1);
529 		EMIT(OEOL, 0);
530 		p->g->iflags |= USEEOL;
531 		p->g->neol++;
532 	}
533 
534 	(void)REQUIRE(HERE() != start, REG_EMPTY);	/* require nonempty */
535 }
536 
537 /*
538  - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
539  */
540 static int			/* was the simple RE an unbackslashed $? */
541 p_simp_re(struct parse *p,
542     int starordinary,		/* is a leading * an ordinary character? */
543     size_t reclimit)
544 {
545 	int c;
546 	int count;
547 	int count2;
548 	sopno pos;
549 	int i;
550 	sopno subno;
551 	int backsl;
552 
553 	pos = HERE();		/* repetion op, if any, covers from here */
554 
555 	assert(MORE());		/* caller should have ensured this */
556 	c = GETNEXT();
557 	backsl = c == '\\';
558 	if (backsl) {
559 		(void)REQUIRE(MORE(), REG_EESCAPE);
560 		c = (unsigned char)GETNEXT();
561 		switch (c) {
562 		case '{':
563 			SETERROR(REG_BADRPT);
564 			break;
565 		case '(':
566 			p->g->nsub++;
567 			subno = p->g->nsub;
568 			if (subno < NPAREN)
569 				p->pbegin[subno] = HERE();
570 			EMIT(OLPAREN, subno);
571 			/* the MORE here is an error heuristic */
572 			if (MORE() && !SEETWO('\\', ')'))
573 				p_bre(p, '\\', ')', reclimit);
574 			if (subno < NPAREN) {
575 				p->pend[subno] = HERE();
576 				assert(p->pend[subno] != 0);
577 			}
578 			EMIT(ORPAREN, subno);
579 			(void)REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
580 			break;
581 		case ')':	/* should not get here -- must be user */
582 		case '}':
583 			SETERROR(REG_EPAREN);
584 			break;
585 		case '1':
586 		case '2':
587 		case '3':
588 		case '4':
589 		case '5':
590 		case '6':
591 		case '7':
592 		case '8':
593 		case '9':
594 			i = c - '0';
595 			assert(i < NPAREN);
596 			if (p->pend[i] != 0) {
597 				assert(i <= p->g->nsub);
598 				EMIT(OBACK_, i);
599 				assert(p->pbegin[i] != 0);
600 				assert(p->strip[p->pbegin[i]] == OLPAREN);
601 				assert(p->strip[p->pend[i]] == ORPAREN);
602 				(void) dupl(p, p->pbegin[i]+1, p->pend[i]);
603 				EMIT(O_BACK, i);
604 			} else
605 				SETERROR(REG_ESUBREG);
606 			p->g->backrefs = 1;
607 			break;
608 		default:
609 			ordinary(p, c);
610 			break;
611 		}
612 	} else {
613 		switch (c) {
614 		case '.':
615 			if (p->g->cflags&REG_NEWLINE)
616 				nonnewline(p);
617 			else
618 				EMIT(OANY, 0);
619 			break;
620 		case '[':
621 			p_bracket(p);
622 			break;
623 		case '*':
624 			(void)REQUIRE(starordinary, REG_BADRPT);
625 			/* FALLTHROUGH */
626 		default:
627 			ordinary(p, c);
628 			break;
629 		}
630 	}
631 
632 	if (EAT('*')) {		/* implemented as +? */
633 		/* this case does not require the (y|) trick, noKLUDGE */
634 		INSERT(OPLUS_, pos);
635 		ASTERN(O_PLUS, pos);
636 		INSERT(OQUEST_, pos);
637 		ASTERN(O_QUEST, pos);
638 	} else if (EATTWO('\\', '{')) {
639 		count = p_count(p);
640 		if (EAT(',')) {
641 			if (MORE() && ISDIGIT((UCHAR_T)PEEK())) {
642 				count2 = p_count(p);
643 				(void)REQUIRE(count <= count2, REG_BADBR);
644 			} else		/* single number with comma */
645 				count2 = INFINITY;
646 		} else		/* just a single number */
647 			count2 = count;
648 		repeat(p, pos, count, count2, reclimit);
649 		if (!EATTWO('\\', '}')) {	/* error heuristics */
650 			while (MORE() && !SEETWO('\\', '}'))
651 				NEXT();
652 			(void)REQUIRE(MORE(), REG_EBRACE);
653 			SETERROR(REG_BADBR);
654 		}
655 	} else if (!backsl && c == (unsigned char)'$')	/* $ (but not \$) ends it */
656 		return(1);
657 
658 	return(0);
659 }
660 
661 /*
662  - p_count - parse a repetition count
663  */
664 static int			/* the value */
665 p_count(struct parse *p)
666 {
667 	int count = 0;
668 	int ndigits = 0;
669 
670 	while (MORE() && ISDIGIT((UCHAR_T)PEEK()) && count <= DUPMAX) {
671 		count = count*10 + (GETNEXT() - '0');
672 		ndigits++;
673 	}
674 
675 	(void)REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
676 	return(count);
677 }
678 
679 /*
680  - p_bracket - parse a bracketed character list
681  *
682  * Note a significant property of this code:  if the allocset() did SETERROR,
683  * no set operations are done.
684  */
685 static void
686 p_bracket(struct parse *p)
687 {
688 	cset *cs;
689 	int invert = 0;
690 	static RCHAR_T bow[] = { '[', ':', '<', ':', ']', ']' };
691 	static RCHAR_T eow[] = { '[', ':', '>', ':', ']', ']' };
692 
693 	cs = allocset(p);
694 	if (cs == NULL)
695 		return;
696 
697 	/* Dept of Truly Sickening Special-Case Kludges */
698 	if (p->next + 5 < p->end && MEMCMP(p->next, bow, 6) == 0) {
699 		EMIT(OBOW, 0);
700 		NEXTn(6);
701 		return;
702 	}
703 	if (p->next + 5 < p->end && MEMCMP(p->next, eow, 6) == 0) {
704 		EMIT(OEOW, 0);
705 		NEXTn(6);
706 		return;
707 	}
708 
709 	if (EAT('^'))
710 		invert++;	/* make note to invert set at end */
711 	if (EAT(']'))
712 		CHadd(cs, ']');
713 	else if (EAT('-'))
714 		CHadd(cs, '-');
715 	while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
716 		p_b_term(p, cs);
717 	if (EAT('-'))
718 		CHadd(cs, '-');
719 	(void)MUSTEAT(']', REG_EBRACK);
720 
721 	if (p->error != 0)	/* don't mess things up further */
722 		return;
723 
724 	if (p->g->cflags&REG_ICASE) {
725 		int i;
726 		int ci;
727 
728 		for (i = p->g->csetsize - 1; i >= 0; i--)
729 			if (CHIN(cs, i) && isalpha(i)) {
730 				ci = othercase(i);
731 				if (ci != i)
732 					CHadd(cs, ci);
733 			}
734 		if (cs->multis != NULL)
735 			mccase(p, cs);
736 	}
737 	if (invert) {
738 		int i;
739 
740 		for (i = p->g->csetsize - 1; i >= 0; i--)
741 			if (CHIN(cs, i))
742 				CHsub(cs, i);
743 			else
744 				CHadd(cs, i);
745 		if (p->g->cflags&REG_NEWLINE)
746 			CHsub(cs, '\n');
747 		if (cs->multis != NULL)
748 			mcinvert(p, cs);
749 	}
750 
751 	assert(cs->multis == NULL);		/* xxx */
752 
753 	if (nch(p, cs) == 1) {		/* optimize singleton sets */
754 		ordinary(p, firstch(p, cs));
755 		freeset(p, cs);
756 	} else
757 		EMIT(OANYOF, freezeset(p, cs));
758 }
759 
760 /*
761  - p_b_term - parse one term of a bracketed character list
762  */
763 static void
764 p_b_term(struct parse *p, cset *cs)
765 {
766 	char c;
767 	char start, finish;
768 	int i;
769 
770 	/* classify what we've got */
771 	switch ((MORE()) ? PEEK() : '\0') {
772 	case '[':
773 		c = (MORE2()) ? PEEK2() : '\0';
774 		break;
775 	case '-':
776 		SETERROR(REG_ERANGE);
777 		return;			/* NOTE RETURN */
778 		break;
779 	default:
780 		c = '\0';
781 		break;
782 	}
783 
784 	switch (c) {
785 	case ':':		/* character class */
786 		NEXT2();
787 		(void)REQUIRE(MORE(), REG_EBRACK);
788 		c = PEEK();
789 		(void)REQUIRE(c != '-' && c != ']', REG_ECTYPE);
790 		p_b_cclass(p, cs);
791 		(void)REQUIRE(MORE(), REG_EBRACK);
792 		(void)REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
793 		break;
794 	case '=':		/* equivalence class */
795 		NEXT2();
796 		(void)REQUIRE(MORE(), REG_EBRACK);
797 		c = PEEK();
798 		(void)REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
799 		p_b_eclass(p, cs);
800 		(void)REQUIRE(MORE(), REG_EBRACK);
801 		(void)REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
802 		break;
803 	default:		/* symbol, ordinary character, or range */
804 /* xxx revision needed for multichar stuff */
805 		start = p_b_symbol(p);
806 		if (SEE('-') && MORE2() && PEEK2() != ']') {
807 			/* range */
808 			NEXT();
809 			if (EAT('-'))
810 				finish = '-';
811 			else
812 				finish = p_b_symbol(p);
813 		} else
814 			finish = start;
815 /* xxx what about signed chars here... */
816 		(void)REQUIRE(start <= finish, REG_ERANGE);
817 		for (i = start; i <= finish; i++)
818 			CHadd(cs, i);
819 		break;
820 	}
821 }
822 
823 /*
824  - p_b_cclass - parse a character-class name and deal with it
825  */
826 static void
827 p_b_cclass(struct parse *p, cset *cs)
828 {
829 	RCHAR_T *sp = p->next;
830 	struct cclass *cp;
831 	size_t len;
832 	const char *u;
833 	char c;
834 
835 	while (MORE() && isalpha(PEEK()))
836 		NEXT();
837 	len = p->next - sp;
838 	for (cp = cclasses; cp->name != NULL; cp++)
839 		if (STRLEN(cp->name) == len && !MEMCMP(cp->name, sp, len))
840 			break;
841 	if (cp->name == NULL) {
842 		/* oops, didn't find it */
843 		SETERROR(REG_ECTYPE);
844 		return;
845 	}
846 
847 	u = cp->chars;
848 	while ((c = *u++) != '\0')
849 		CHadd(cs, c);
850 	for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
851 		MCadd(p, cs, u);
852 }
853 
854 /*
855  - p_b_eclass - parse an equivalence-class name and deal with it
856  *
857  * This implementation is incomplete. xxx
858  */
859 static void
860 p_b_eclass(struct parse *p, cset *cs)
861 {
862 	char c;
863 
864 	c = p_b_coll_elem(p, '=');
865 	CHadd(cs, c);
866 }
867 
868 /*
869  - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
870  */
871 static char			/* value of symbol */
872 p_b_symbol(struct parse *p)
873 {
874 	char value;
875 
876 	(void)REQUIRE(MORE(), REG_EBRACK);
877 	if (!EATTWO('[', '.'))
878 		return(GETNEXT());
879 
880 	/* collating symbol */
881 	value = p_b_coll_elem(p, '.');
882 	(void)REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
883 	return(value);
884 }
885 
886 /*
887  - p_b_coll_elem - parse a collating-element name and look it up
888  */
889 static char			/* value of collating element */
890 p_b_coll_elem(struct parse *p, int endc)
891 
892          			/* name ended by endc,']' */
893 {
894 	RCHAR_T *sp = p->next;
895 	struct cname *cp;
896 	size_t len;
897 
898 	while (MORE() && !SEETWO(endc, ']'))
899 		NEXT();
900 	if (!MORE()) {
901 		SETERROR(REG_EBRACK);
902 		return(0);
903 	}
904 	len = p->next - sp;
905 	for (cp = cnames; cp->name != NULL; cp++)
906 		if (STRLEN(cp->name) == len && MEMCMP(cp->name, sp, len))
907 			return(cp->code);	/* known name */
908 	if (len == 1)
909 		return(*sp);	/* single character */
910 	SETERROR(REG_ECOLLATE);			/* neither */
911 	return(0);
912 }
913 
914 /*
915  - othercase - return the case counterpart of an alphabetic
916  */
917 static char			/* if no counterpart, return ch */
918 othercase(int ch)
919 {
920 	assert(isalpha(ch));
921 	if (isupper(ch))
922 		return(tolower(ch));
923 	else if (islower(ch))
924 		return(toupper(ch));
925 	else			/* peculiar, but could happen */
926 		return(ch);
927 }
928 
929 /*
930  - bothcases - emit a dualcase version of a two-case character
931  *
932  * Boy, is this implementation ever a kludge...
933  */
934 static void
935 bothcases(struct parse *p, int ch)
936 {
937 	RCHAR_T *oldnext = p->next;
938 	RCHAR_T *oldend = p->end;
939 	RCHAR_T bracket[3];
940 
941 	assert(othercase(ch) != ch);	/* p_bracket() would recurse */
942 	p->next = bracket;
943 	p->end = bracket+2;
944 	bracket[0] = ch;
945 	bracket[1] = ']';
946 	bracket[2] = '\0';
947 	p_bracket(p);
948 	assert(p->next == bracket+2);
949 	p->next = oldnext;
950 	p->end = oldend;
951 }
952 
953 /*
954  - ordinary - emit an ordinary character
955  */
956 static void
957 ordinary(struct parse *p, int ch)
958 {
959 /*
960 	cat_t *cap = p->g->categories;
961 */
962 
963 	if ((p->g->cflags&REG_ICASE) && isalpha(ch) && othercase(ch) != ch)
964 		bothcases(p, ch);
965 	else {
966 		EMIT(OCHAR, (UCHAR_T)ch);
967 /*
968 		if (cap[ch] == 0)
969 			cap[ch] = p->g->ncategories++;
970 */
971 	}
972 }
973 
974 /*
975  - nonnewline - emit REG_NEWLINE version of OANY
976  *
977  * Boy, is this implementation ever a kludge...
978  */
979 static void
980 nonnewline(struct parse *p)
981 {
982 	RCHAR_T *oldnext = p->next;
983 	RCHAR_T *oldend = p->end;
984 	RCHAR_T bracket[4];
985 
986 	p->next = bracket;
987 	p->end = bracket+3;
988 	bracket[0] = '^';
989 	bracket[1] = '\n';
990 	bracket[2] = ']';
991 	bracket[3] = '\0';
992 	p_bracket(p);
993 	assert(p->next == bracket+3);
994 	p->next = oldnext;
995 	p->end = oldend;
996 }
997 
998 /*
999  - repeat - generate code for a bounded repetition, recursively if needed
1000  */
1001 static void
1002 repeat(struct parse *p,
1003     sopno start,		/* operand from here to end of strip */
1004     int from,			/* repeated from this number */
1005     int to,			/* to this number of times (maybe INFINITY) */
1006     size_t reclimit)
1007 {
1008 	sopno finish;
1009 #	define	N	2
1010 #	define	INF	3
1011 #	define	REP(f, t)	((f)*8 + (t))
1012 #	define	MAP(n)	(((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1013 	sopno copy;
1014 
1015 	if (reclimit++ > RECLIMIT)
1016 		p->error = REG_ESPACE;
1017 	if (p->error)
1018 		return;
1019 
1020 	finish = HERE();
1021 
1022 	assert(from <= to);
1023 
1024 	switch (REP(MAP(from), MAP(to))) {
1025 	case REP(0, 0):			/* must be user doing this */
1026 		DROP(finish-start);	/* drop the operand */
1027 		break;
1028 	case REP(0, 1):			/* as x{1,1}? */
1029 	case REP(0, N):			/* as x{1,n}? */
1030 	case REP(0, INF):		/* as x{1,}? */
1031 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1032 		INSERT(OCH_, start);		/* offset is wrong... */
1033 		repeat(p, start+1, 1, to, reclimit);
1034 		ASTERN(OOR1, start);
1035 		AHEAD(start);			/* ... fix it */
1036 		EMIT(OOR2, 0);
1037 		AHEAD(THERE());
1038 		ASTERN(O_CH, THERETHERE());
1039 		break;
1040 	case REP(1, 1):			/* trivial case */
1041 		/* done */
1042 		break;
1043 	case REP(1, N):			/* as x?x{1,n-1} */
1044 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1045 		INSERT(OCH_, start);
1046 		ASTERN(OOR1, start);
1047 		AHEAD(start);
1048 		EMIT(OOR2, 0);			/* offset very wrong... */
1049 		AHEAD(THERE());			/* ...so fix it */
1050 		ASTERN(O_CH, THERETHERE());
1051 		copy = dupl(p, start+1, finish+1);
1052 		assert(copy == finish+4);
1053 		repeat(p, copy, 1, to-1, reclimit);
1054 		break;
1055 	case REP(1, INF):		/* as x+ */
1056 		INSERT(OPLUS_, start);
1057 		ASTERN(O_PLUS, start);
1058 		break;
1059 	case REP(N, N):			/* as xx{m-1,n-1} */
1060 		copy = dupl(p, start, finish);
1061 		repeat(p, copy, from-1, to-1, reclimit);
1062 		break;
1063 	case REP(N, INF):		/* as xx{n-1,INF} */
1064 		copy = dupl(p, start, finish);
1065 		repeat(p, copy, from-1, to, reclimit);
1066 		break;
1067 	default:			/* "can't happen" */
1068 		SETERROR(REG_ASSERT);	/* just in case */
1069 		break;
1070 	}
1071 }
1072 
1073 /*
1074  - seterr - set an error condition
1075  */
1076 static int			/* useless but makes type checking happy */
1077 seterr(struct parse *p, int e)
1078 {
1079 	if (p->error == 0)	/* keep earliest error condition */
1080 		p->error = e;
1081 	p->next = nuls;		/* try to bring things to a halt */
1082 	p->end = nuls;
1083 	return(0);		/* make the return value well-defined */
1084 }
1085 
1086 /*
1087  - allocset - allocate a set of characters for []
1088  */
1089 static cset *
1090 allocset(struct parse *p)
1091 {
1092 	int no = p->g->ncsets++;
1093 	size_t nc;
1094 	size_t nbytes;
1095 	cset *cs;
1096 	size_t css = (size_t)p->g->csetsize;
1097 	int i;
1098 
1099 	if (no >= p->ncsalloc) {	/* need another column of space */
1100 		p->ncsalloc += CHAR_BIT;
1101 		nc = p->ncsalloc;
1102 		assert(nc % CHAR_BIT == 0);
1103 		nbytes = nc / CHAR_BIT * css;
1104 		if (MEMSIZE(p) > MEMLIMIT)
1105 			goto oomem;
1106 		if (p->g->sets == NULL)
1107 			p->g->sets = (cset *)malloc(nc * sizeof(cset));
1108 		else
1109 			p->g->sets = (cset *)realloc((char *)p->g->sets,
1110 							nc * sizeof(cset));
1111 		if (p->g->setbits == NULL)
1112 			p->g->setbits = (uch *)malloc(nbytes);
1113 		else {
1114 			p->g->setbits = (uch *)realloc((char *)p->g->setbits,
1115 								nbytes);
1116 			/* xxx this isn't right if setbits is now NULL */
1117 			for (i = 0; i < no; i++)
1118 				p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
1119 		}
1120 		if (p->g->sets != NULL && p->g->setbits != NULL)
1121 			memset((char *)p->g->setbits + (nbytes - css),
1122 								0, css);
1123 		else {
1124 oomem:
1125 			no = 0;
1126 			SETERROR(REG_ESPACE);
1127 			/* caller's responsibility not to do set ops */
1128 			return NULL;
1129 		}
1130 	}
1131 
1132 	cs = &p->g->sets[no];
1133 	cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
1134 	cs->mask = 1 << ((no) % CHAR_BIT);
1135 	cs->hash = 0;
1136 	cs->smultis = 0;
1137 	cs->multis = NULL;
1138 
1139 	return(cs);
1140 }
1141 
1142 /*
1143  - freeset - free a now-unused set
1144  */
1145 static void
1146 freeset(struct parse *p, cset *cs)
1147 {
1148 	size_t i;
1149 	cset *top = &p->g->sets[p->g->ncsets];
1150 	size_t css = (size_t)p->g->csetsize;
1151 
1152 	for (i = 0; i < css; i++)
1153 		CHsub(cs, i);
1154 	if (cs == top-1)	/* recover only the easy case */
1155 		p->g->ncsets--;
1156 }
1157 
1158 /*
1159  - freezeset - final processing on a set of characters
1160  *
1161  * The main task here is merging identical sets.  This is usually a waste
1162  * of time (although the hash code minimizes the overhead), but can win
1163  * big if REG_ICASE is being used.  REG_ICASE, by the way, is why the hash
1164  * is done using addition rather than xor -- all ASCII [aA] sets xor to
1165  * the same value!
1166  */
1167 static int			/* set number */
1168 freezeset(struct parse *p, cset *cs)
1169 {
1170 	uch h = cs->hash;
1171 	size_t i;
1172 	cset *top = &p->g->sets[p->g->ncsets];
1173 	cset *cs2;
1174 	size_t css = (size_t)p->g->csetsize;
1175 
1176 	/* look for an earlier one which is the same */
1177 	for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
1178 		if (cs2->hash == h && cs2 != cs) {
1179 			/* maybe */
1180 			for (i = 0; i < css; i++)
1181 				if (!!CHIN(cs2, i) != !!CHIN(cs, i))
1182 					break;		/* no */
1183 			if (i == css)
1184 				break;			/* yes */
1185 		}
1186 
1187 	if (cs2 < top) {	/* found one */
1188 		freeset(p, cs);
1189 		cs = cs2;
1190 	}
1191 
1192 	return((int)(cs - p->g->sets));
1193 }
1194 
1195 /*
1196  - firstch - return first character in a set (which must have at least one)
1197  */
1198 static int			/* character; there is no "none" value */
1199 firstch(struct parse *p, cset *cs)
1200 {
1201 	size_t i;
1202 	size_t css = (size_t)p->g->csetsize;
1203 
1204 	for (i = 0; i < css; i++)
1205 		if (CHIN(cs, i))
1206 			return((char)i);
1207 	assert(never);
1208 	return(0);		/* arbitrary */
1209 }
1210 
1211 /*
1212  - nch - number of characters in a set
1213  */
1214 static int
1215 nch(struct parse *p, cset *cs)
1216 {
1217 	size_t i;
1218 	size_t css = (size_t)p->g->csetsize;
1219 	int n = 0;
1220 
1221 	for (i = 0; i < css; i++)
1222 		if (CHIN(cs, i))
1223 			n++;
1224 	return(n);
1225 }
1226 
1227 /*
1228  - mcadd - add a collating element to a cset
1229  */
1230 static void
1231 mcadd(struct parse *p, cset *cs, const char *cp)
1232 {
1233 	size_t oldend = cs->smultis;
1234 	void *np;
1235 
1236 	cs->smultis += strlen(cp) + 1;
1237 	np = realloc(cs->multis, cs->smultis);
1238 	if (np == NULL) {
1239 		if (cs->multis)
1240 			free(cs->multis);
1241 		cs->multis = NULL;
1242 		SETERROR(REG_ESPACE);
1243 		return;
1244 	}
1245 	cs->multis = np;
1246 
1247 	strlcpy(cs->multis + oldend - 1, cp, cs->smultis - oldend + 1);
1248 }
1249 
1250 /*
1251  - mcinvert - invert the list of collating elements in a cset
1252  *
1253  * This would have to know the set of possibilities.  Implementation
1254  * is deferred.
1255  */
1256 static void
1257 mcinvert(struct parse *p, cset *cs)
1258 {
1259 	assert(cs->multis == NULL);	/* xxx */
1260 }
1261 
1262 /*
1263  - mccase - add case counterparts of the list of collating elements in a cset
1264  *
1265  * This would have to know the set of possibilities.  Implementation
1266  * is deferred.
1267  */
1268 static void
1269 mccase(struct parse *p, cset *cs)
1270 {
1271 	assert(cs->multis == NULL);	/* xxx */
1272 }
1273 
1274 #ifdef notdef
1275 /*
1276  - isinsets - is this character in any sets?
1277  */
1278 static int			/* predicate */
1279 isinsets(struct re_guts *g, int c)
1280 {
1281 	uch *col;
1282 	int i;
1283 	int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1284 	unsigned uc = (unsigned char)c;
1285 
1286 	for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1287 		if (col[uc] != 0)
1288 			return(1);
1289 	return(0);
1290 }
1291 
1292 /*
1293  - samesets - are these two characters in exactly the same sets?
1294  */
1295 static int			/* predicate */
1296 samesets(struct re_guts *g, int c1, int c2)
1297 {
1298 	uch *col;
1299 	int i;
1300 	int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1301 	unsigned uc1 = (unsigned char)c1;
1302 	unsigned uc2 = (unsigned char)c2;
1303 
1304 	for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1305 		if (col[uc1] != col[uc2])
1306 			return(0);
1307 	return(1);
1308 }
1309 #endif
1310 
1311 /*
1312  - categorize - sort out character categories
1313  */
1314 static void
1315 categorize(struct parse *p, struct re_guts *g)
1316 {
1317 #ifdef notdef
1318 	cat_t *cats = g->categories;
1319 	int c;
1320 	int c2;
1321 	cat_t cat;
1322 
1323 	/* avoid making error situations worse */
1324 	if (p->error != 0)
1325 		return;
1326 
1327 	for (c = CHAR_MIN; c <= CHAR_MAX; c++)
1328 		if (cats[c] == 0 && isinsets(g, c)) {
1329 			cat = g->ncategories++;
1330 			cats[c] = cat;
1331 			for (c2 = c+1; c2 <= CHAR_MAX; c2++)
1332 				if (cats[c2] == 0 && samesets(g, c, c2))
1333 					cats[c2] = cat;
1334 		}
1335 #endif
1336 }
1337 
1338 /*
1339  - dupl - emit a duplicate of a bunch of sops
1340  */
1341 static sopno			/* start of duplicate */
1342 dupl(struct parse *p,
1343     sopno start,		/* from here */
1344     sopno finish)		/* to this less one */
1345 {
1346 	sopno ret = HERE();
1347 	sopno len = finish - start;
1348 
1349 	assert(finish >= start);
1350 	if (len == 0)
1351 		return(ret);
1352 	if (!enlarge(p, p->ssize + len))	/* this many unexpected additions */
1353 		return ret;
1354 	assert(p->ssize >= p->slen + len);
1355 	(void) memcpy((char *)(p->strip + p->slen),
1356 		(char *)(p->strip + start), (size_t)len*sizeof(sop));
1357 	(void) memcpy((char *)(p->stripdata + p->slen),
1358 		(char *)(p->stripdata + start), (size_t)len*sizeof(RCHAR_T));
1359 	p->slen += len;
1360 	return(ret);
1361 }
1362 
1363 /*
1364  - doemit - emit a strip operator
1365  *
1366  * It might seem better to implement this as a macro with a function as
1367  * hard-case backup, but it's just too big and messy unless there are
1368  * some changes to the data structures.  Maybe later.
1369  */
1370 static void
1371 doemit(struct parse *p, sop op, size_t opnd)
1372 {
1373 	/* avoid making error situations worse */
1374 	if (p->error != 0)
1375 		return;
1376 
1377 	/* deal with oversize operands ("can't happen", more or less) */
1378 	assert(opnd < 1);
1379 
1380 	/* deal with undersized strip */
1381 	if (p->slen >= p->ssize)
1382 		if (!enlarge(p, (p->ssize+1) / 2 * 3))	/* +50% */
1383 			return;
1384 
1385 	/* finally, it's all reduced to the easy case */
1386 	p->strip[p->slen] = op;
1387 	p->stripdata[p->slen] = opnd;
1388 	p->slen++;
1389 }
1390 
1391 /*
1392  - doinsert - insert a sop into the strip
1393  */
1394 static void
1395 doinsert(struct parse *p, sop op, size_t opnd, sopno pos)
1396 {
1397 	sopno sn;
1398 	sop s;
1399 	RCHAR_T d;
1400 	int i;
1401 
1402 	/* avoid making error situations worse */
1403 	if (p->error != 0)
1404 		return;
1405 
1406 	sn = HERE();
1407 	EMIT(op, opnd);		/* do checks, ensure space */
1408 	assert(HERE() == sn+1);
1409 	s = p->strip[sn];
1410 	d = p->stripdata[sn];
1411 
1412 	/* adjust paren pointers */
1413 	assert(pos > 0);
1414 	for (i = 1; i < NPAREN; i++) {
1415 		if (p->pbegin[i] >= pos) {
1416 			p->pbegin[i]++;
1417 		}
1418 		if (p->pend[i] >= pos) {
1419 			p->pend[i]++;
1420 		}
1421 	}
1422 
1423 	memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1424 						(HERE()-pos-1)*sizeof(sop));
1425 	memmove((char *)&p->stripdata[pos+1], (char *)&p->stripdata[pos],
1426 						(HERE()-pos-1)*sizeof(RCHAR_T));
1427 	p->strip[pos] = s;
1428 	p->stripdata[pos] = d;
1429 }
1430 
1431 /*
1432  - dofwd - complete a forward reference
1433  */
1434 static void
1435 dofwd(struct parse *p, sopno pos, sop value)
1436 {
1437 	/* avoid making error situations worse */
1438 	if (p->error != 0)
1439 		return;
1440 
1441 	assert(value < 1);
1442 	p->stripdata[pos] = value;
1443 }
1444 
1445 /*
1446  - enlarge - enlarge the strip
1447  */
1448 static int
1449 enlarge(struct parse *p, sopno size)
1450 {
1451 	sop *sp;
1452 	RCHAR_T *dp;
1453 	sopno osize;
1454 
1455 	if (p->ssize >= size)
1456 		return 1;
1457 
1458 	osize = p->ssize;
1459 	p->ssize = size;
1460 	if (MEMSIZE(p) > MEMLIMIT)
1461 		goto oomem;
1462 	sp = realloc(p->strip, p->ssize * sizeof(sop));
1463 	if (sp == NULL)
1464 		goto oomem;
1465 	p->strip = sp;
1466 	dp = realloc(p->stripdata, p->ssize * sizeof(RCHAR_T));
1467 	if (dp == NULL) {
1468 oomem:
1469 		p->ssize = osize;
1470 		SETERROR(REG_ESPACE);
1471 		return 0;
1472 	}
1473 	p->stripdata = dp;
1474 	return 1;
1475 }
1476 
1477 /*
1478  - stripsnug - compact the strip
1479  */
1480 static void
1481 stripsnug(struct parse *p, struct re_guts *g)
1482 {
1483 	g->nstates = p->slen;
1484 	g->strip = (sop *)realloc((char *)p->strip,
1485 	    p->slen * sizeof(sop));
1486 	if (g->strip == NULL) {
1487 		SETERROR(REG_ESPACE);
1488 		g->strip = p->strip;
1489 	}
1490 	g->stripdata = (RCHAR_T *)realloc((char *)p->stripdata,
1491 	    p->slen * sizeof(RCHAR_T));
1492 	if (g->stripdata == NULL) {
1493 		SETERROR(REG_ESPACE);
1494 		g->stripdata = p->stripdata;
1495 	}
1496 }
1497 
1498 /*
1499  - findmust - fill in must and mlen with longest mandatory literal string
1500  *
1501  * This algorithm could do fancy things like analyzing the operands of |
1502  * for common subsequences.  Someday.  This code is simple and finds most
1503  * of the interesting cases.
1504  *
1505  * Note that must and mlen got initialized during setup.
1506  */
1507 static void
1508 findmust(struct parse *p, struct re_guts *g)
1509 {
1510 	sop *scans;
1511 	RCHAR_T *scand;
1512 	sop *starts = 0;
1513 	RCHAR_T *startd = NULL;
1514 	sop *newstarts = 0;
1515 	RCHAR_T *newstartd = NULL;
1516 	sopno newlen;
1517 	sop s;
1518 	RCHAR_T d;
1519 	RCHAR_T *cp;
1520 	sopno i;
1521 
1522 	/* avoid making error situations worse */
1523 	if (p->error != 0)
1524 		return;
1525 
1526 	/* find the longest OCHAR sequence in strip */
1527 	newlen = 0;
1528 	scans = g->strip + 1;
1529 	scand = g->stripdata + 1;
1530 	do {
1531 		s = *scans++;
1532 		d = *scand++;
1533 		switch (s) {
1534 		case OCHAR:		/* sequence member */
1535 			if (newlen == 0) {		/* new sequence */
1536 				newstarts = scans - 1;
1537 				newstartd = scand - 1;
1538 			}
1539 			newlen++;
1540 			break;
1541 		case OPLUS_:		/* things that don't break one */
1542 		case OLPAREN:
1543 		case ORPAREN:
1544 			break;
1545 		case OQUEST_:		/* things that must be skipped */
1546 		case OCH_:
1547 			scans--;
1548 			scand--;
1549 			do {
1550 				scans += d;
1551 				scand += d;
1552 				s = *scans;
1553 				d = *scand;
1554 				/* assert() interferes w debug printouts */
1555 				if (s != O_QUEST && s != O_CH && s != OOR2) {
1556 					g->iflags |= BAD;
1557 					return;
1558 				}
1559 			} while (s != O_QUEST && s != O_CH);
1560 			/* fallthrough */
1561 		default:		/* things that break a sequence */
1562 			if (newlen > g->mlen) {		/* ends one */
1563 				starts = newstarts;
1564 				startd = newstartd;
1565 				g->mlen = newlen;
1566 			}
1567 			newlen = 0;
1568 			break;
1569 		}
1570 	} while (s != OEND);
1571 
1572 	if (g->mlen == 0)		/* there isn't one */
1573 		return;
1574 
1575 	/* turn it into a character string */
1576 	g->must = malloc(((size_t)g->mlen + 1) * sizeof(RCHAR_T));
1577 	if (g->must == NULL) {		/* argh; just forget it */
1578 		g->mlen = 0;
1579 		return;
1580 	}
1581 	cp = g->must;
1582 	scans = starts;
1583 	scand = startd;
1584 	for (i = g->mlen; i > 0; i--) {
1585 		for (;;) {
1586 			s = *scans++;
1587 			d = *scand++;
1588 			if (s == OCHAR)
1589 				break;
1590 		}
1591 		assert(cp < g->must + g->mlen);
1592 		*cp++ = d;
1593 	}
1594 	assert(cp == g->must + g->mlen);
1595 	*cp++ = '\0';		/* just on general principles */
1596 }
1597 
1598 /*
1599  - pluscount - count + nesting
1600  */
1601 static sopno			/* nesting depth */
1602 pluscount(struct parse *p, struct re_guts *g)
1603 {
1604 	sop *scan;
1605 	sop s;
1606 	sopno plusnest = 0;
1607 	sopno maxnest = 0;
1608 
1609 	if (p->error != 0)
1610 		return(0);	/* there may not be an OEND */
1611 
1612 	scan = g->strip + 1;
1613 	do {
1614 		s = *scan++;
1615 		switch (s) {
1616 		case OPLUS_:
1617 			plusnest++;
1618 			break;
1619 		case O_PLUS:
1620 			if (plusnest > maxnest)
1621 				maxnest = plusnest;
1622 			plusnest--;
1623 			break;
1624 		}
1625 	} while (s != OEND);
1626 	if (plusnest != 0)
1627 		g->iflags |= BAD;
1628 	return(maxnest);
1629 }
1630