xref: /illumos-gate/usr/src/lib/libc/port/regex/engine.c (revision 695dd8d1)
1 /*
2  * Copyright 2019 Nexenta by DDN, Inc. All rights reserved.
3  * Copyright 2012 Milan Jurik. All rights reserved.
4  * Copyright (c) 2016 by Delphix. All rights reserved.
5  * Copyright (c) 1992, 1993, 1994 Henry Spencer.
6  * Copyright (c) 1992, 1993, 1994
7  *	The Regents of the University of California.  All rights reserved.
8  *
9  * This code is derived from software contributed to Berkeley by
10  * Henry Spencer.
11  *
12  * Redistribution and use in source and binary forms, with or without
13  * modification, are permitted provided that the following conditions
14  * are met:
15  * 1. Redistributions of source code must retain the above copyright
16  *    notice, this list of conditions and the following disclaimer.
17  * 2. Redistributions in binary form must reproduce the above copyright
18  *    notice, this list of conditions and the following disclaimer in the
19  *    documentation and/or other materials provided with the distribution.
20  * 3. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  */
36 
37 #include <stdbool.h>
38 
39 /*
40  * The matching engine and friends.  This file is #included by regexec.c
41  * after suitable #defines of a variety of macros used herein, so that
42  * different state representations can be used without duplicating masses
43  * of code.
44  */
45 
46 #ifdef SNAMES
47 #define	stepback sstepback
48 #define	matcher	smatcher
49 #define	walk	swalk
50 #define	dissect	sdissect
51 #define	backref	sbackref
52 #define	step	sstep
53 #define	print	sprint
54 #define	at	sat
55 #define	match	smat
56 #endif
57 #ifdef LNAMES
58 #define	stepback lstepback
59 #define	matcher	lmatcher
60 #define	walk	lwalk
61 #define	dissect	ldissect
62 #define	backref	lbackref
63 #define	step	lstep
64 #define	print	lprint
65 #define	at	lat
66 #define	match	lmat
67 #endif
68 #ifdef MNAMES
69 #define	stepback mstepback
70 #define	matcher	mmatcher
71 #define	walk	mwalk
72 #define	dissect	mdissect
73 #define	backref	mbackref
74 #define	step	mstep
75 #define	print	mprint
76 #define	at	mat
77 #define	match	mmat
78 #endif
79 
80 /* another structure passed up and down to avoid zillions of parameters */
81 struct match {
82 	struct re_guts *g;
83 	int eflags;
84 	regmatch_t *pmatch;	/* [nsub+1] (0 element unused) */
85 	const char *offp;	/* offsets work from here */
86 	const char *beginp;	/* start of string -- virtual NUL precedes */
87 	const char *endp;	/* end of string -- virtual NUL here */
88 	const char *coldp;	/* can be no match starting before here */
89 	const char **lastpos;	/* [nplus+1] */
90 	STATEVARS;
91 	states st;		/* current states */
92 	states fresh;		/* states for a fresh start */
93 	states tmp;		/* temporary */
94 	states empty;		/* empty set of states */
95 	mbstate_t mbs;		/* multibyte conversion state */
96 };
97 
98 /* ========= begin header generated by ./mkh ========= */
99 #ifdef __cplusplus
100 extern "C" {
101 #endif
102 
103 /* === engine.c === */
104 static int matcher(struct re_guts *, const char *, size_t, regmatch_t[], int);
105 static const char *dissect(struct match *, const char *, const char *,
106     sopno, sopno);
107 static const char *backref(struct match *, const char *, const char *, sopno,
108     sopno, sopno, int);
109 static const char *walk(struct match *m, const char *start, const char *stop,
110     sopno startst, sopno stopst, bool fast);
111 static states step(struct re_guts *, sopno, sopno, states, wint_t, states);
112 #define	MAX_RECURSION	100
113 #define	BOL	(OUT-1)
114 #define	EOL	(BOL-1)
115 #define	BOLEOL	(BOL-2)
116 #define	NOTHING	(BOL-3)
117 #define	BOW	(BOL-4)
118 #define	EOW	(BOL-5)
119 #define	BADCHAR	(BOL-6)
120 #define	NONCHAR(c)	((c) <= OUT)
121 #ifdef REDEBUG
122 static void print(struct match *, const char *, states, int, FILE *);
123 #endif
124 #ifdef REDEBUG
125 static void at(struct match *, const char *, const char *, const char *,
126     sopno, sopno);
127 #endif
128 #ifdef REDEBUG
129 static const char *pchar(int ch);
130 #endif
131 
132 #ifdef __cplusplus
133 }
134 #endif
135 /* ========= end header generated by ./mkh ========= */
136 
137 #ifdef REDEBUG
138 #define	SP(t, s, c)	print(m, t, s, c, stdout)
139 #define	AT(t, p1, p2, s1, s2)	at(m, t, p1, p2, s1, s2)
140 #define	NOTE(str)	{ if (m->eflags&REG_TRACE) printf("=%s\n", (str)); }
141 #else
142 #define	SP(t, s, c)	/* nothing */
143 #define	AT(t, p1, p2, s1, s2)	/* nothing */
144 #define	NOTE(s)	/* nothing */
145 #endif
146 
147 /*
148  * Given a multibyte string pointed to by start, step back nchar characters
149  * from current position pointed to by cur.
150  */
151 static const char *
152 stepback(const char *start, const char *cur, int nchar)
153 {
154 	const char *ret;
155 	int wc, mbc;
156 	mbstate_t mbs;
157 	size_t clen;
158 
159 	if (MB_CUR_MAX == 1)
160 		return ((cur - nchar) > start ? cur - nchar : NULL);
161 
162 	ret = cur;
163 	for (wc = nchar; wc > 0; wc--) {
164 		for (mbc = 1; mbc <= MB_CUR_MAX; mbc++) {
165 			if ((ret - mbc) < start)
166 				return (NULL);
167 			memset(&mbs, 0, sizeof (mbs));
168 			clen = mbrtowc(NULL, ret - mbc, mbc, &mbs);
169 			if (clen != (size_t)-1 && clen != (size_t)-2)
170 				break;
171 		}
172 		if (mbc > MB_CUR_MAX)
173 			return (NULL);
174 		ret -= mbc;
175 	}
176 
177 	return (ret);
178 }
179 
180 /*
181  * matcher - the actual matching engine
182  */
183 static int			/* 0 success, REG_NOMATCH failure */
184 matcher(struct re_guts *g, const char *string, size_t nmatch,
185     regmatch_t pmatch[], int eflags)
186 {
187 	const char *endp;
188 	size_t i;
189 	struct match mv;
190 	struct match *m = &mv;
191 	const char *dp = NULL;
192 	const sopno gf = g->firststate+1;	/* +1 for OEND */
193 	const sopno gl = g->laststate;
194 	const char *start;
195 	const char *stop;
196 	/* Boyer-Moore algorithms variables */
197 	const char *pp;
198 	int cj, mj;
199 	const char *mustfirst;
200 	const char *mustlast;
201 	int *matchjump;
202 	int *charjump;
203 
204 	/* simplify the situation where possible */
205 	if (g->cflags&REG_NOSUB)
206 		nmatch = 0;
207 
208 	if (eflags&REG_STARTEND) {
209 		start = string + pmatch[0].rm_so;
210 		stop = string + pmatch[0].rm_eo;
211 	} else {
212 		start = string;
213 		stop = start + strlen(start);
214 	}
215 
216 	if (stop < start)
217 		return (REG_EFATAL);
218 
219 	/* prescreening; this does wonders for this rather slow code */
220 	if (g->must != NULL) {
221 		if (g->charjump != NULL && g->matchjump != NULL) {
222 			mustfirst = g->must;
223 			mustlast = g->must + g->mlen - 1;
224 			charjump = g->charjump;
225 			matchjump = g->matchjump;
226 			pp = mustlast;
227 			for (dp = start+g->mlen-1; dp < stop; ) {
228 				/* Fast skip non-matches */
229 				while (dp < stop && charjump[(int)*dp])
230 					dp += charjump[(int)*dp];
231 
232 				if (dp >= stop)
233 					break;
234 
235 				/* Greedy matcher */
236 				/*
237 				 * We depend on not being used for
238 				 * for strings of length 1
239 				 */
240 				while (*--dp == *--pp && pp != mustfirst)
241 					;
242 
243 				if (*dp == *pp)
244 					break;
245 
246 				/* Jump to next possible match */
247 				mj = matchjump[pp - mustfirst];
248 				cj = charjump[(int)*dp];
249 				dp += (cj < mj ? mj : cj);
250 				pp = mustlast;
251 			}
252 			if (pp != mustfirst)
253 				return (REG_NOMATCH);
254 		} else {
255 			for (dp = start; dp < stop; dp++)
256 				if (*dp == g->must[0] &&
257 				    stop - dp >= g->mlen &&
258 				    memcmp(dp, g->must, (size_t)g->mlen) == 0)
259 					break;
260 			if (dp == stop)		/* we didn't find g->must */
261 				return (REG_NOMATCH);
262 		}
263 	}
264 
265 	/* match struct setup */
266 	m->g = g;
267 	m->eflags = eflags;
268 	m->pmatch = NULL;
269 	m->lastpos = NULL;
270 	m->offp = string;
271 	m->beginp = start;
272 	m->endp = stop;
273 	STATESETUP(m, 4);
274 	SETUP(m->st);
275 	SETUP(m->fresh);
276 	SETUP(m->tmp);
277 	SETUP(m->empty);
278 	CLEAR(m->empty);
279 	ZAPSTATE(&m->mbs);
280 
281 	/* Adjust start according to moffset, to speed things up */
282 	if (dp != NULL && g->moffset > -1) {
283 		const char *nstart;
284 
285 		nstart = stepback(start, dp, g->moffset);
286 		if (nstart != NULL)
287 			start = nstart;
288 	}
289 
290 	SP("mloop", m->st, *start);
291 
292 	/* this loop does only one repetition except for backrefs */
293 	for (;;) {
294 		endp = walk(m, start, stop, gf, gl, true);
295 		if (endp == NULL) {		/* a miss */
296 			if (m->pmatch != NULL)
297 				free((char *)m->pmatch);
298 			if (m->lastpos != NULL)
299 				free((char *)m->lastpos);
300 			STATETEARDOWN(m);
301 			return (REG_NOMATCH);
302 		}
303 		if (nmatch == 0 && !g->backrefs)
304 			break;		/* no further info needed */
305 
306 		/* where? */
307 		assert(m->coldp != NULL);
308 		for (;;) {
309 			NOTE("finding start");
310 			endp = walk(m, m->coldp, stop, gf, gl, false);
311 			if (endp != NULL)
312 				break;
313 			assert(m->coldp < m->endp);
314 			m->coldp += XMBRTOWC(NULL, m->coldp,
315 			    m->endp - m->coldp, &m->mbs, 0);
316 		}
317 		if (nmatch == 1 && !g->backrefs)
318 			break;		/* no further info needed */
319 
320 		/* oh my, it wants the subexpressions... */
321 		if (m->pmatch == NULL)
322 			m->pmatch = (regmatch_t *)malloc((m->g->nsub + 1) *
323 			    sizeof (regmatch_t));
324 		if (m->pmatch == NULL) {
325 			STATETEARDOWN(m);
326 			return (REG_ESPACE);
327 		}
328 		for (i = 1; i <= m->g->nsub; i++)
329 			m->pmatch[i].rm_so = m->pmatch[i].rm_eo = -1;
330 		if (!g->backrefs && !(m->eflags&REG_BACKR)) {
331 			NOTE("dissecting");
332 			dp = dissect(m, m->coldp, endp, gf, gl);
333 		} else {
334 			if (g->nplus > 0 && m->lastpos == NULL)
335 				m->lastpos = malloc((g->nplus+1) *
336 				    sizeof (const char *));
337 			if (g->nplus > 0 && m->lastpos == NULL) {
338 				free(m->pmatch);
339 				STATETEARDOWN(m);
340 				return (REG_ESPACE);
341 			}
342 			NOTE("backref dissect");
343 			dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
344 		}
345 		if (dp != NULL)
346 			break;
347 
348 		/* uh-oh... we couldn't find a subexpression-level match */
349 		assert(g->backrefs);	/* must be back references doing it */
350 		assert(g->nplus == 0 || m->lastpos != NULL);
351 		for (;;) {
352 			if (dp != NULL || endp <= m->coldp)
353 				break;		/* defeat */
354 			NOTE("backoff");
355 			endp = walk(m, m->coldp, endp-1, gf, gl, false);
356 			if (endp == NULL)
357 				break;		/* defeat */
358 			/* try it on a shorter possibility */
359 #ifndef NDEBUG
360 			for (i = 1; i <= m->g->nsub; i++) {
361 				assert(m->pmatch[i].rm_so == -1);
362 				assert(m->pmatch[i].rm_eo == -1);
363 			}
364 #endif
365 			NOTE("backoff dissect");
366 			dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
367 		}
368 		assert(dp == NULL || dp == endp);
369 		if (dp != NULL)		/* found a shorter one */
370 			break;
371 
372 		/* despite initial appearances, there is no match here */
373 		NOTE("false alarm");
374 		/* recycle starting later */
375 		start = m->coldp + XMBRTOWC(NULL, m->coldp,
376 		    stop - m->coldp, &m->mbs, 0);
377 		assert(start <= stop);
378 	}
379 
380 	/* fill in the details if requested */
381 	if (nmatch > 0) {
382 		pmatch[0].rm_so = m->coldp - m->offp;
383 		pmatch[0].rm_eo = endp - m->offp;
384 	}
385 	if (nmatch > 1) {
386 		assert(m->pmatch != NULL);
387 		for (i = 1; i < nmatch; i++)
388 			if (i <= m->g->nsub)
389 				pmatch[i] = m->pmatch[i];
390 			else {
391 				pmatch[i].rm_so = -1;
392 				pmatch[i].rm_eo = -1;
393 			}
394 	}
395 
396 	if (m->pmatch != NULL)
397 		free((char *)m->pmatch);
398 	if (m->lastpos != NULL)
399 		free((char *)m->lastpos);
400 	STATETEARDOWN(m);
401 	return (0);
402 }
403 
404 /*
405  * dissect - figure out what matched what, no back references
406  */
407 static const char *
408 dissect(struct match *m, const char *start, const char *stop, sopno startst,
409     sopno stopst)
410 {
411 	int i;
412 	sopno ss;		/* start sop of current subRE */
413 	sopno es;		/* end sop of current subRE */
414 	const char *sp;		/* start of string matched by it */
415 	const char *stp;	/* string matched by it cannot pass here */
416 	const char *rest;	/* start of rest of string */
417 	const char *tail;	/* string unmatched by rest of RE */
418 	sopno ssub;		/* start sop of subsubRE */
419 	sopno esub;		/* end sop of subsubRE */
420 	const char *ssp;	/* start of string matched by subsubRE */
421 	const char *sep;	/* end of string matched by subsubRE */
422 	const char *oldssp;	/* previous ssp */
423 	const char *dp __unused;
424 
425 	AT("diss", start, stop, startst, stopst);
426 	sp = start;
427 	for (ss = startst; ss < stopst; ss = es) {
428 		/* identify end of subRE */
429 		es = ss;
430 		switch (OP(m->g->strip[es])) {
431 		case OPLUS_:
432 		case OQUEST_:
433 			es += OPND(m->g->strip[es]);
434 			break;
435 		case OCH_:
436 			while (OP(m->g->strip[es]) != (sop)O_CH)
437 				es += OPND(m->g->strip[es]);
438 			break;
439 		}
440 		es++;
441 
442 		/* figure out what it matched */
443 		switch (OP(m->g->strip[ss])) {
444 		case OEND:
445 			assert(0);
446 			break;
447 		case OCHAR:
448 			sp += XMBRTOWC(NULL, sp, stop - start, &m->mbs, 0);
449 			break;
450 		case OBOL:
451 		case OEOL:
452 		case OBOW:
453 		case OEOW:
454 			break;
455 		case OANY:
456 		case OANYOF:
457 			sp += XMBRTOWC(NULL, sp, stop - start, &m->mbs, 0);
458 			break;
459 		case OBACK_:
460 		case O_BACK:
461 			assert(0);
462 			break;
463 		/* cases where length of match is hard to find */
464 		case OQUEST_:
465 			stp = stop;
466 			for (;;) {
467 				/* how long could this one be? */
468 				rest = walk(m, sp, stp, ss, es, false);
469 				assert(rest != NULL);	/* it did match */
470 				/* could the rest match the rest? */
471 				tail = walk(m, rest, stop, es, stopst, false);
472 				if (tail == stop)
473 					break;		/* yes! */
474 				/* no -- try a shorter match for this one */
475 				stp = rest - 1;
476 				assert(stp >= sp);	/* it did work */
477 			}
478 			ssub = ss + 1;
479 			esub = es - 1;
480 			/* did innards match? */
481 			if (walk(m, sp, rest, ssub, esub, false) != NULL) {
482 				dp = dissect(m, sp, rest, ssub, esub);
483 				assert(dp == rest);
484 			} else		/* no */
485 				assert(sp == rest);
486 			sp = rest;
487 			break;
488 		case OPLUS_:
489 			stp = stop;
490 			for (;;) {
491 				/* how long could this one be? */
492 				rest = walk(m, sp, stp, ss, es, false);
493 				assert(rest != NULL);	/* it did match */
494 				/* could the rest match the rest? */
495 				tail = walk(m, rest, stop, es, stopst, false);
496 				if (tail == stop)
497 					break;		/* yes! */
498 				/* no -- try a shorter match for this one */
499 				stp = rest - 1;
500 				assert(stp >= sp);	/* it did work */
501 			}
502 			ssub = ss + 1;
503 			esub = es - 1;
504 			ssp = sp;
505 			oldssp = ssp;
506 			for (;;) {	/* find last match of innards */
507 				sep = walk(m, ssp, rest, ssub, esub, false);
508 				if (sep == NULL || sep == ssp)
509 					break;	/* failed or matched null */
510 				oldssp = ssp;	/* on to next try */
511 				ssp = sep;
512 			}
513 			if (sep == NULL) {
514 				/* last successful match */
515 				sep = ssp;
516 				ssp = oldssp;
517 			}
518 			assert(sep == rest);	/* must exhaust substring */
519 			assert(walk(m, ssp, sep, ssub, esub, false) == rest);
520 			dp = dissect(m, ssp, sep, ssub, esub);
521 			assert(dp == sep);
522 			sp = rest;
523 			break;
524 		case OCH_:
525 			stp = stop;
526 			for (;;) {
527 				/* how long could this one be? */
528 				rest = walk(m, sp, stp, ss, es, false);
529 				assert(rest != NULL);	/* it did match */
530 				/* could the rest match the rest? */
531 				tail = walk(m, rest, stop, es, stopst, false);
532 				if (tail == stop)
533 					break;		/* yes! */
534 				/* no -- try a shorter match for this one */
535 				stp = rest - 1;
536 				assert(stp >= sp);	/* it did work */
537 			}
538 			ssub = ss + 1;
539 			esub = ss + OPND(m->g->strip[ss]) - 1;
540 			assert(OP(m->g->strip[esub]) == OOR1);
541 			for (;;) {	/* find first matching branch */
542 				if (walk(m, sp, rest, ssub, esub,
543 				    false) == rest)
544 					break;	/* it matched all of it */
545 				/* that one missed, try next one */
546 				assert(OP(m->g->strip[esub]) == OOR1);
547 				esub++;
548 				assert(OP(m->g->strip[esub]) == OOR2);
549 				ssub = esub + 1;
550 				esub += OPND(m->g->strip[esub]);
551 				if (OP(m->g->strip[esub]) == (sop)OOR2)
552 					esub--;
553 				else
554 					assert(OP(m->g->strip[esub]) == O_CH);
555 			}
556 			dp = dissect(m, sp, rest, ssub, esub);
557 			assert(dp == rest);
558 			sp = rest;
559 			break;
560 		case O_PLUS:
561 		case O_QUEST:
562 		case OOR1:
563 		case OOR2:
564 		case O_CH:
565 			assert(0);
566 			break;
567 		case OLPAREN:
568 			i = OPND(m->g->strip[ss]);
569 			assert(0 < i && i <= m->g->nsub);
570 			m->pmatch[i].rm_so = sp - m->offp;
571 			break;
572 		case ORPAREN:
573 			i = OPND(m->g->strip[ss]);
574 			assert(0 < i && i <= m->g->nsub);
575 			m->pmatch[i].rm_eo = sp - m->offp;
576 			break;
577 		default:		/* uh oh */
578 			assert(0);
579 			break;
580 		}
581 	}
582 
583 	assert(sp == stop);
584 	return (sp);
585 }
586 
587 /*
588  * backref - figure out what matched what, figuring in back references
589  */
590 static const char *
591 backref(struct match *m, const char *start, const char *stop, sopno startst,
592     sopno stopst, sopno lev,		/* PLUS nesting level */
593     int rec)
594 {
595 	int i;
596 	sopno ss;		/* start sop of current subRE */
597 	const char *sp;		/* start of string matched by it */
598 	sopno ssub;		/* start sop of subsubRE */
599 	sopno esub;		/* end sop of subsubRE */
600 	const char *ssp;	/* start of string matched by subsubRE */
601 	const char *dp;
602 	size_t len;
603 	int hard;
604 	sop s;
605 	regoff_t offsave;
606 	cset *cs;
607 	wint_t wc;
608 
609 	AT("back", start, stop, startst, stopst);
610 	sp = start;
611 
612 	/* get as far as we can with easy stuff */
613 	hard = 0;
614 	for (ss = startst; !hard && ss < stopst; ss++)
615 		switch (OP(s = m->g->strip[ss])) {
616 		case OCHAR:
617 			if (sp == stop)
618 				return (NULL);
619 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
620 			if (wc != OPND(s))
621 				return (NULL);
622 			break;
623 		case OANY:
624 			if (sp == stop)
625 				return (NULL);
626 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
627 			if (wc == BADCHAR)
628 				return (NULL);
629 			break;
630 		case OANYOF:
631 			if (sp == stop)
632 				return (NULL);
633 			cs = &m->g->sets[OPND(s)];
634 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
635 			if (wc == BADCHAR || !CHIN(cs, wc))
636 				return (NULL);
637 			break;
638 		case OBOL:
639 			if ((sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
640 			    (sp > m->offp && sp < m->endp &&
641 			    *(sp-1) == '\n' && (m->g->cflags&REG_NEWLINE))) {
642 				break;
643 			}
644 			return (NULL);
645 		case OEOL:
646 			if ((sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
647 			    (sp < m->endp && *sp == '\n' &&
648 			    (m->g->cflags&REG_NEWLINE))) {
649 				break;
650 			}
651 			return (NULL);
652 		case OBOW:
653 			if (sp < m->endp && ISWORD(*sp) &&
654 			    ((sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
655 			    (sp > m->offp && !ISWORD(*(sp-1))))) {
656 				break;
657 			}
658 			return (NULL);
659 		case OEOW:
660 			if (((sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
661 			    (sp < m->endp && *sp == '\n' &&
662 			    (m->g->cflags&REG_NEWLINE)) ||
663 			    (sp < m->endp && !ISWORD(*sp))) &&
664 			    (sp > m->beginp && ISWORD(*(sp-1)))) {
665 				break;
666 			}
667 			return (NULL);
668 		case O_QUEST:
669 			break;
670 		case OOR1:	/* matches null but needs to skip */
671 			ss++;
672 			s = m->g->strip[ss];
673 			do {
674 				assert(OP(s) == OOR2);
675 				ss += OPND(s);
676 			} while (OP(s = m->g->strip[ss]) != (sop)O_CH);
677 			/* note that the ss++ gets us past the O_CH */
678 			break;
679 		default:	/* have to make a choice */
680 			hard = 1;
681 			break;
682 		}
683 	if (!hard) {		/* that was it! */
684 		if (sp != stop)
685 			return (NULL);
686 		return (sp);
687 	}
688 	ss--;			/* adjust for the for's final increment */
689 
690 	/* the hard stuff */
691 	AT("hard", sp, stop, ss, stopst);
692 	s = m->g->strip[ss];
693 	switch (OP(s)) {
694 	case OBACK_:		/* the vilest depths */
695 		i = OPND(s);
696 		assert(0 < i && i <= m->g->nsub);
697 		if (m->pmatch[i].rm_eo == -1)
698 			return (NULL);
699 		assert(m->pmatch[i].rm_so != -1);
700 		len = m->pmatch[i].rm_eo - m->pmatch[i].rm_so;
701 		if (len == 0 && rec++ > MAX_RECURSION)
702 			return (NULL);
703 		assert(stop - m->beginp >= len);
704 		if (sp > stop - len)
705 			return (NULL);	/* not enough left to match */
706 		ssp = m->offp + m->pmatch[i].rm_so;
707 		if (memcmp(sp, ssp, len) != 0)
708 			return (NULL);
709 		while (m->g->strip[ss] != (sop)SOP(O_BACK, i))
710 			ss++;
711 		return (backref(m, sp+len, stop, ss+1, stopst, lev, rec));
712 	case OQUEST_:		/* to null or not */
713 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
714 		if (dp != NULL)
715 			return (dp);	/* not */
716 		return (backref(m, sp, stop, ss+OPND(s)+1, stopst, lev, rec));
717 	case OPLUS_:
718 		assert(m->lastpos != NULL);
719 		assert(lev+1 <= m->g->nplus);
720 		m->lastpos[lev+1] = sp;
721 		return (backref(m, sp, stop, ss+1, stopst, lev+1, rec));
722 	case O_PLUS:
723 		if (sp == m->lastpos[lev])	/* last pass matched null */
724 			return (backref(m, sp, stop, ss+1, stopst, lev-1, rec));
725 		/* try another pass */
726 		m->lastpos[lev] = sp;
727 		dp = backref(m, sp, stop, ss-OPND(s)+1, stopst, lev, rec);
728 		if (dp == NULL)
729 			return (backref(m, sp, stop, ss+1, stopst, lev-1, rec));
730 		return (dp);
731 	case OCH_:		/* find the right one, if any */
732 		ssub = ss + 1;
733 		esub = ss + OPND(s) - 1;
734 		assert(OP(m->g->strip[esub]) == OOR1);
735 		for (;;) {	/* find first matching branch */
736 			dp = backref(m, sp, stop, ssub, esub, lev, rec);
737 			if (dp != NULL)
738 				return (dp);
739 			/* that one missed, try next one */
740 			if (OP(m->g->strip[esub]) == (sop)O_CH)
741 				return (NULL);	/* there is none */
742 			esub++;
743 			assert(OP(m->g->strip[esub]) == (sop)OOR2);
744 			ssub = esub + 1;
745 			esub += OPND(m->g->strip[esub]);
746 			if (OP(m->g->strip[esub]) == (sop)OOR2)
747 				esub--;
748 			else
749 				assert(OP(m->g->strip[esub]) == O_CH);
750 		}
751 		/* NOTREACHED */
752 		break;
753 	case OLPAREN:		/* must undo assignment if rest fails */
754 		i = OPND(s);
755 		assert(0 < i && i <= m->g->nsub);
756 		offsave = m->pmatch[i].rm_so;
757 		m->pmatch[i].rm_so = sp - m->offp;
758 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
759 		if (dp != NULL)
760 			return (dp);
761 		m->pmatch[i].rm_so = offsave;
762 		return (NULL);
763 	case ORPAREN:		/* must undo assignment if rest fails */
764 		i = OPND(s);
765 		assert(0 < i && i <= m->g->nsub);
766 		offsave = m->pmatch[i].rm_eo;
767 		m->pmatch[i].rm_eo = sp - m->offp;
768 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
769 		if (dp != NULL)
770 			return (dp);
771 		m->pmatch[i].rm_eo = offsave;
772 		return (NULL);
773 	default:		/* uh oh */
774 		assert(0);
775 		break;
776 	}
777 
778 	/* "can't happen" */
779 	assert(0);
780 	return (NULL);
781 }
782 
783 /*
784  * Step through the string either quickly or slowly.  Returns where it ended
785  * or NULL.
786  */
787 static const char *
788 walk(struct match *m, const char *start, const char *stop, sopno startst,
789     sopno stopst, bool fast)
790 {
791 	states st = m->st;
792 	states fresh = m->fresh;
793 	states empty = m->empty;
794 	states tmp = m->tmp;
795 	const char *p = start;
796 	wint_t c;
797 	wint_t lastc;		/* previous c */
798 	wint_t flagch;
799 	int i;
800 	const char *matchp;	/* last p at which a match ended */
801 	size_t clen;
802 
803 	AT("slow", start, stop, startst, stopst);
804 	CLEAR(st);
805 	SET1(st, startst);
806 	SP("sstart", st, *p);
807 	st = step(m->g, startst, stopst, st, NOTHING, st);
808 	if (fast)
809 		ASSIGN(fresh, st);
810 	matchp = NULL;
811 	if (start == m->offp || (start == m->beginp && !(m->eflags&REG_NOTBOL)))
812 		c = OUT;
813 	else {
814 		/*
815 		 * XXX Wrong if the previous character was multi-byte.
816 		 * Newline never is (in encodings supported by FreeBSD),
817 		 * so this only breaks the ISWORD tests below.
818 		 */
819 		c = (uch)*(start - 1);
820 	}
821 	for (;;) {
822 		/* next character */
823 		lastc = c;
824 		if (p == m->endp) {
825 			c = OUT;
826 			clen = 0;
827 		} else
828 			clen = XMBRTOWC(&c, p, m->endp - p, &m->mbs, BADCHAR);
829 
830 		if (fast && EQ(st, fresh))
831 			matchp = p;
832 
833 		/* is there an EOL and/or BOL between lastc and c? */
834 		flagch = '\0';
835 		i = 0;
836 		if ((lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
837 		    (lastc == OUT && !(m->eflags&REG_NOTBOL))) {
838 			flagch = BOL;
839 			i = m->g->nbol;
840 		}
841 		if ((c == '\n' && m->g->cflags&REG_NEWLINE) ||
842 		    (c == OUT && !(m->eflags&REG_NOTEOL))) {
843 			flagch = (flagch == BOL) ? BOLEOL : EOL;
844 			i += m->g->neol;
845 		}
846 		if (i != 0) {
847 			for (; i > 0; i--)
848 				st = step(m->g, startst, stopst, st,
849 				    flagch, st);
850 			SP("sboleol", st, c);
851 		}
852 
853 		/* how about a word boundary? */
854 		if ((flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
855 		    (c != OUT && ISWORD(c))) {
856 			flagch = BOW;
857 		}
858 		if ((lastc != OUT && ISWORD(lastc)) &&
859 		    (flagch == EOL || (c != OUT && !ISWORD(c)))) {
860 			flagch = EOW;
861 		}
862 		if (flagch == BOW || flagch == EOW) {
863 			st = step(m->g, startst, stopst, st, flagch, st);
864 			SP("sboweow", st, c);
865 		}
866 
867 		/* are we done? */
868 		if (ISSET(st, stopst)) {
869 			if (fast)
870 				break;
871 			else
872 				matchp = p;
873 		}
874 		if (EQ(st, empty) || p == stop || clen > (size_t)(stop - p))
875 			break;		/* NOTE BREAK OUT */
876 
877 		/* no, we must deal with this character */
878 		ASSIGN(tmp, st);
879 		if (fast)
880 			ASSIGN(st, fresh);
881 		else
882 			ASSIGN(st, empty);
883 		assert(c != OUT);
884 		st = step(m->g, startst, stopst, tmp, c, st);
885 		SP("saft", st, c);
886 		assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
887 		p += clen;
888 	}
889 
890 	if (fast) {
891 		assert(matchp != NULL);
892 		m->coldp = matchp;
893 		if (ISSET(st, stopst))
894 			return (p + XMBRTOWC(NULL, p, stop - p, &m->mbs, 0));
895 		else
896 			return (NULL);
897 	} else
898 		return (matchp);
899 }
900 
901 /*
902  * step - map set of states reachable before char to set reachable after
903  */
904 static states
905 step(struct re_guts *g,
906     sopno start,	/* start state within strip */
907     sopno stop,		/* state after stop state within strip */
908     states bef,		/* states reachable before */
909     wint_t ch,		/* character or NONCHAR code */
910     states aft)		/* states already known reachable after */
911 {
912 	cset *cs;
913 	sop s;
914 	sopno pc;
915 	onestate here;		/* note, macros know this name */
916 	sopno look;
917 	int i;
918 
919 	for (pc = start, INIT(here, pc); pc != stop; pc++, INC(here)) {
920 		s = g->strip[pc];
921 		switch (OP(s)) {
922 		case OEND:
923 			assert(pc == stop-1);
924 			break;
925 		case OCHAR:
926 			/* only characters can match */
927 			assert(!NONCHAR(ch) || ch != OPND(s));
928 			if (ch == OPND(s))
929 				FWD(aft, bef, 1);
930 			break;
931 		case OBOL:
932 			if (ch == BOL || ch == BOLEOL)
933 				FWD(aft, bef, 1);
934 			break;
935 		case OEOL:
936 			if (ch == EOL || ch == BOLEOL)
937 				FWD(aft, bef, 1);
938 			break;
939 		case OBOW:
940 			if (ch == BOW)
941 				FWD(aft, bef, 1);
942 			break;
943 		case OEOW:
944 			if (ch == EOW)
945 				FWD(aft, bef, 1);
946 			break;
947 		case OANY:
948 			if (!NONCHAR(ch))
949 				FWD(aft, bef, 1);
950 			break;
951 		case OANYOF:
952 			cs = &g->sets[OPND(s)];
953 			if (!NONCHAR(ch) && CHIN(cs, ch))
954 				FWD(aft, bef, 1);
955 			break;
956 		case OBACK_:		/* ignored here */
957 		case O_BACK:
958 			FWD(aft, aft, 1);
959 			break;
960 		case OPLUS_:		/* forward, this is just an empty */
961 			FWD(aft, aft, 1);
962 			break;
963 		case O_PLUS:		/* both forward and back */
964 			FWD(aft, aft, 1);
965 			i = ISSETBACK(aft, OPND(s));
966 			BACK(aft, aft, OPND(s));
967 			if (!i && ISSETBACK(aft, OPND(s))) {
968 				/* oho, must reconsider loop body */
969 				pc -= OPND(s) + 1;
970 				INIT(here, pc);
971 			}
972 			break;
973 		case OQUEST_:		/* two branches, both forward */
974 			FWD(aft, aft, 1);
975 			FWD(aft, aft, OPND(s));
976 			break;
977 		case O_QUEST:		/* just an empty */
978 			FWD(aft, aft, 1);
979 			break;
980 		case OLPAREN:		/* not significant here */
981 		case ORPAREN:
982 			FWD(aft, aft, 1);
983 			break;
984 		case OCH_:		/* mark the first two branches */
985 			FWD(aft, aft, 1);
986 			assert(OP(g->strip[pc+OPND(s)]) == (sop)OOR2);
987 			FWD(aft, aft, OPND(s));
988 			break;
989 		case OOR1:		/* done a branch, find the O_CH */
990 			if (ISSTATEIN(aft, here)) {
991 				for (look = 1;
992 				    OP(s = g->strip[pc+look]) != (sop)O_CH;
993 				    look += OPND(s))
994 					assert(OP(s) == (sop)OOR2);
995 				FWD(aft, aft, look + 1);
996 			}
997 			break;
998 		case OOR2:		/* propagate OCH_'s marking */
999 			FWD(aft, aft, 1);
1000 			if (OP(g->strip[pc+OPND(s)]) != (sop)O_CH) {
1001 				assert(OP(g->strip[pc+OPND(s)]) == (sop)OOR2);
1002 				FWD(aft, aft, OPND(s));
1003 			}
1004 			break;
1005 		case O_CH:		/* just empty */
1006 			FWD(aft, aft, 1);
1007 			break;
1008 		default:		/* ooooops... */
1009 			assert(0);
1010 			break;
1011 		}
1012 	}
1013 
1014 	return (aft);
1015 }
1016 
1017 #ifdef REDEBUG
1018 /*
1019  * print - print a set of states
1020  */
1021 static void
1022 print(struct match *m, const char *caption, states st, int ch, FILE *d)
1023 {
1024 	struct re_guts *g = m->g;
1025 	sopno i;
1026 	int first = 1;
1027 
1028 	if (!(m->eflags&REG_TRACE))
1029 		return;
1030 
1031 	(void) fprintf(d, "%s", caption);
1032 	if (ch != '\0')
1033 		(void) fprintf(d, " %s", pchar(ch));
1034 	for (i = 0; i < g->nstates; i++)
1035 		if (ISSET(st, i)) {
1036 			(void) fprintf(d, "%s%d", (first) ? "\t" : ", ", i);
1037 			first = 0;
1038 		}
1039 	(void) fprintf(d, "\n");
1040 }
1041 
1042 /*
1043  * at - print current situation
1044  */
1045 static void
1046 at(struct match *m, const char *title, const char *start, const char *stop,
1047     sopno startst, sopno stopst)
1048 {
1049 	if (!(m->eflags&REG_TRACE))
1050 		return;
1051 
1052 	(void) printf("%s %s-", title, pchar(*start));
1053 	(void) printf("%s ", pchar(*stop));
1054 	(void) printf("%ld-%ld\n", (long)startst, (long)stopst);
1055 }
1056 
1057 #ifndef PCHARDONE
1058 #define	PCHARDONE	/* never again */
1059 /*
1060  * pchar - make a character printable
1061  *
1062  * Is this identical to regchar() over in debug.c?  Well, yes.  But a
1063  * duplicate here avoids having a debugging-capable regexec.o tied to
1064  * a matching debug.o, and this is convenient.  It all disappears in
1065  * the non-debug compilation anyway, so it doesn't matter much.
1066  */
1067 static const char *
1068 pchar(int ch)
1069 {
1070 	static char pbuf[10];
1071 
1072 	if (isprint((uch)ch) || ch == ' ')
1073 		(void) sprintf(pbuf, "%c", ch);
1074 	else
1075 		(void) sprintf(pbuf, "\\%o", ch);
1076 	return (pbuf);
1077 }
1078 #endif
1079 #endif
1080 
1081 #undef	stepback
1082 #undef	matcher
1083 #undef	walk
1084 #undef	dissect
1085 #undef	backref
1086 #undef	step
1087 #undef	print
1088 #undef	at
1089 #undef	match
1090