xref: /titanic_52/usr/src/lib/libc/port/locale/engine.c (revision 6a5408e613dc0b372f722907d6c0b18e99c182dd)
1 /*
2  * Copyright 2010 Nexenta Systems, Inc.  All rights reserved.
3  * Copyright (c) 1992, 1993, 1994 Henry Spencer.
4  * Copyright (c) 1992, 1993, 1994
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * This code is derived from software contributed to Berkeley by
8  * Henry Spencer.
9  *
10  * Redistribution and use in source and binary forms, with or without
11  * modification, are permitted provided that the following conditions
12  * are met:
13  * 1. Redistributions of source code must retain the above copyright
14  *    notice, this list of conditions and the following disclaimer.
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in the
17  *    documentation and/or other materials provided with the distribution.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  */
34 
35 /*
36  * The matching engine and friends.  This file is #included by regexec.c
37  * after suitable #defines of a variety of macros used herein, so that
38  * different state representations can be used without duplicating masses
39  * of code.
40  */
41 
42 #ifdef SNAMES
43 #define	matcher	smatcher
44 #define	fast	sfast
45 #define	slow	sslow
46 #define	dissect	sdissect
47 #define	backref	sbackref
48 #define	step	sstep
49 #define	print	sprint
50 #define	at	sat
51 #define	match	smat
52 #endif
53 #ifdef LNAMES
54 #define	matcher	lmatcher
55 #define	fast	lfast
56 #define	slow	lslow
57 #define	dissect	ldissect
58 #define	backref	lbackref
59 #define	step	lstep
60 #define	print	lprint
61 #define	at	lat
62 #define	match	lmat
63 #endif
64 #ifdef MNAMES
65 #define	matcher	mmatcher
66 #define	fast	mfast
67 #define	slow	mslow
68 #define	dissect	mdissect
69 #define	backref	mbackref
70 #define	step	mstep
71 #define	print	mprint
72 #define	at	mat
73 #define	match	mmat
74 #endif
75 
76 /* another structure passed up and down to avoid zillions of parameters */
77 struct match {
78 	struct re_guts *g;
79 	int eflags;
80 	regmatch_t *pmatch;	/* [nsub+1] (0 element unused) */
81 	const char *offp;	/* offsets work from here */
82 	const char *beginp;	/* start of string -- virtual NUL precedes */
83 	const char *endp;	/* end of string -- virtual NUL here */
84 	const char *coldp;	/* can be no match starting before here */
85 	const char **lastpos;	/* [nplus+1] */
86 	STATEVARS;
87 	states st;		/* current states */
88 	states fresh;		/* states for a fresh start */
89 	states tmp;		/* temporary */
90 	states empty;		/* empty set of states */
91 	mbstate_t mbs;		/* multibyte conversion state */
92 };
93 
94 /* ========= begin header generated by ./mkh ========= */
95 #ifdef __cplusplus
96 extern "C" {
97 #endif
98 
99 /* === engine.c === */
100 static int matcher(struct re_guts *, const char *, size_t, regmatch_t[], int);
101 static const char *dissect(struct match *, const char *, const char *,
102     sopno, sopno);
103 static const char *backref(struct match *, const char *, const char *, sopno,
104     sopno, sopno, int);
105 static const char *fast(struct match *, const char *, const char *, sopno,
106     sopno);
107 static const char *slow(struct match *, const char *, const char *, sopno,
108     sopno);
109 static states step(struct re_guts *, sopno, sopno, states, wint_t, states);
110 #define	MAX_RECURSION	100
111 #define	BOL	(OUT-1)
112 #define	EOL	(BOL-1)
113 #define	BOLEOL	(BOL-2)
114 #define	NOTHING	(BOL-3)
115 #define	BOW	(BOL-4)
116 #define	EOW	(BOL-5)
117 #define	BADCHAR	(BOL-6)
118 #define	NONCHAR(c)	((c) <= OUT)
119 #ifdef REDEBUG
120 static void print(struct match *, const char *, states, int, FILE *);
121 #endif
122 #ifdef REDEBUG
123 static void at(struct match *, const char *, const char *, const char *,
124     sopno, sopno);
125 #endif
126 #ifdef REDEBUG
127 static const char *pchar(int ch);
128 #endif
129 
130 #ifdef __cplusplus
131 }
132 #endif
133 /* ========= end header generated by ./mkh ========= */
134 
135 #ifdef REDEBUG
136 #define	SP(t, s, c)	print(m, t, s, c, stdout)
137 #define	AT(t, p1, p2, s1, s2)	at(m, t, p1, p2, s1, s2)
138 #define	NOTE(str)	{ if (m->eflags&REG_TRACE) printf("=%s\n", (str)); }
139 #else
140 #define	SP(t, s, c)	/* nothing */
141 #define	AT(t, p1, p2, s1, s2)	/* nothing */
142 #define	NOTE(s)	/* nothing */
143 #endif
144 
145 /*
146  * matcher - the actual matching engine
147  */
148 static int			/* 0 success, REG_NOMATCH failure */
149 matcher(struct re_guts *g,
150 	const char *string,
151 	size_t nmatch,
152 	regmatch_t pmatch[],
153 	int eflags)
154 {
155 	const char *endp;
156 	int i;
157 	struct match mv;
158 	struct match *m = &mv;
159 	const char *dp;
160 	const sopno gf = g->firststate+1;	/* +1 for OEND */
161 	const sopno gl = g->laststate;
162 	const char *start;
163 	const char *stop;
164 	/* Boyer-Moore algorithms variables */
165 	const char *pp;
166 	int cj, mj;
167 	const char *mustfirst;
168 	const char *mustlast;
169 	int *matchjump;
170 	int *charjump;
171 
172 	/* simplify the situation where possible */
173 	if (g->cflags&REG_NOSUB)
174 		nmatch = 0;
175 
176 	if (eflags&REG_STARTEND) {
177 		start = string + pmatch[0].rm_so;
178 		stop = string + pmatch[0].rm_eo;
179 	} else {
180 		start = string;
181 		stop = start + strlen(start);
182 	}
183 
184 	if (stop < start)
185 		return (REG_EFATAL);
186 
187 	/* prescreening; this does wonders for this rather slow code */
188 	if (g->must != NULL) {
189 		if (g->charjump != NULL && g->matchjump != NULL) {
190 			mustfirst = g->must;
191 			mustlast = g->must + g->mlen - 1;
192 			charjump = g->charjump;
193 			matchjump = g->matchjump;
194 			pp = mustlast;
195 			for (dp = start+g->mlen-1; dp < stop; ) {
196 				/* Fast skip non-matches */
197 				while (dp < stop && charjump[(int)*dp])
198 					dp += charjump[(int)*dp];
199 
200 				if (dp >= stop)
201 					break;
202 
203 				/* Greedy matcher */
204 				/*
205 				 * We depend on not being used for
206 				 * for strings of length 1
207 				 */
208 				while (*--dp == *--pp && pp != mustfirst)
209 					;
210 
211 				if (*dp == *pp)
212 					break;
213 
214 				/* Jump to next possible match */
215 				mj = matchjump[pp - mustfirst];
216 				cj = charjump[(int)*dp];
217 				dp += (cj < mj ? mj : cj);
218 				pp = mustlast;
219 			}
220 			if (pp != mustfirst)
221 				return (REG_NOMATCH);
222 		} else {
223 			for (dp = start; dp < stop; dp++)
224 				if (*dp == g->must[0] &&
225 				    stop - dp >= g->mlen &&
226 				    memcmp(dp, g->must, (size_t)g->mlen) == 0)
227 					break;
228 			if (dp == stop)		/* we didn't find g->must */
229 				return (REG_NOMATCH);
230 		}
231 	}
232 
233 	/* match struct setup */
234 	m->g = g;
235 	m->eflags = eflags;
236 	m->pmatch = NULL;
237 	m->lastpos = NULL;
238 	m->offp = string;
239 	m->beginp = start;
240 	m->endp = stop;
241 	STATESETUP(m, 4);
242 	SETUP(m->st);
243 	SETUP(m->fresh);
244 	SETUP(m->tmp);
245 	SETUP(m->empty);
246 	CLEAR(m->empty);
247 	ZAPSTATE(&m->mbs);
248 
249 	/* Adjust start according to moffset, to speed things up */
250 	if (g->moffset > -1)
251 		start = ((dp - g->moffset) < start) ? start : dp - g->moffset;
252 
253 	SP("mloop", m->st, *start);
254 
255 	/* this loop does only one repetition except for backrefs */
256 	for (;;) {
257 		endp = fast(m, start, stop, gf, gl);
258 		if (endp == NULL) {		/* a miss */
259 			if (m->pmatch != NULL)
260 				free((char *)m->pmatch);
261 			if (m->lastpos != NULL)
262 				free((char *)m->lastpos);
263 			STATETEARDOWN(m);
264 			return (REG_NOMATCH);
265 		}
266 		if (nmatch == 0 && !g->backrefs)
267 			break;		/* no further info needed */
268 
269 		/* where? */
270 		assert(m->coldp != NULL);
271 		for (;;) {
272 			NOTE("finding start");
273 			endp = slow(m, m->coldp, stop, gf, gl);
274 			if (endp != NULL)
275 				break;
276 			assert(m->coldp < m->endp);
277 			m->coldp += XMBRTOWC(NULL, m->coldp,
278 			    m->endp - m->coldp, &m->mbs, 0);
279 		}
280 		if (nmatch == 1 && !g->backrefs)
281 			break;		/* no further info needed */
282 
283 		/* oh my, he wants the subexpressions... */
284 		if (m->pmatch == NULL)
285 			m->pmatch = (regmatch_t *)malloc((m->g->nsub + 1) *
286 			    sizeof (regmatch_t));
287 		if (m->pmatch == NULL) {
288 			STATETEARDOWN(m);
289 			return (REG_ESPACE);
290 		}
291 		for (i = 1; i <= m->g->nsub; i++)
292 			m->pmatch[i].rm_so = m->pmatch[i].rm_eo = -1;
293 		/* NB: FreeBSD has REG_BACKR, we do not */
294 		if (!g->backrefs /* && !(m->eflags&REG_BACKR) */) {
295 			NOTE("dissecting");
296 			dp = dissect(m, m->coldp, endp, gf, gl);
297 		} else {
298 			if (g->nplus > 0 && m->lastpos == NULL)
299 				m->lastpos = malloc((g->nplus+1) *
300 				    sizeof (const char *));
301 			if (g->nplus > 0 && m->lastpos == NULL) {
302 				free(m->pmatch);
303 				STATETEARDOWN(m);
304 				return (REG_ESPACE);
305 			}
306 			NOTE("backref dissect");
307 			dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
308 		}
309 		if (dp != NULL)
310 			break;
311 
312 		/* uh-oh... we couldn't find a subexpression-level match */
313 		assert(g->backrefs);	/* must be back references doing it */
314 		assert(g->nplus == 0 || m->lastpos != NULL);
315 		for (;;) {
316 			if (dp != NULL || endp <= m->coldp)
317 				break;		/* defeat */
318 			NOTE("backoff");
319 			endp = slow(m, m->coldp, endp-1, gf, gl);
320 			if (endp == NULL)
321 				break;		/* defeat */
322 			/* try it on a shorter possibility */
323 #ifndef NDEBUG
324 			for (i = 1; i <= m->g->nsub; i++) {
325 				assert(m->pmatch[i].rm_so == -1);
326 				assert(m->pmatch[i].rm_eo == -1);
327 			}
328 #endif
329 			NOTE("backoff dissect");
330 			dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
331 		}
332 		assert(dp == NULL || dp == endp);
333 		if (dp != NULL)		/* found a shorter one */
334 			break;
335 
336 		/* despite initial appearances, there is no match here */
337 		NOTE("false alarm");
338 		/* recycle starting later */
339 		start = m->coldp + XMBRTOWC(NULL, m->coldp,
340 		    stop - m->coldp, &m->mbs, 0);
341 		assert(start <= stop);
342 	}
343 
344 	/* fill in the details if requested */
345 	if (nmatch > 0) {
346 		pmatch[0].rm_so = m->coldp - m->offp;
347 		pmatch[0].rm_eo = endp - m->offp;
348 	}
349 	if (nmatch > 1) {
350 		assert(m->pmatch != NULL);
351 		for (i = 1; i < nmatch; i++)
352 			if (i <= m->g->nsub)
353 				pmatch[i] = m->pmatch[i];
354 			else {
355 				pmatch[i].rm_so = -1;
356 				pmatch[i].rm_eo = -1;
357 			}
358 	}
359 
360 	if (m->pmatch != NULL)
361 		free((char *)m->pmatch);
362 	if (m->lastpos != NULL)
363 		free((char *)m->lastpos);
364 	STATETEARDOWN(m);
365 	return (0);
366 }
367 
368 /*
369  * dissect - figure out what matched what, no back references
370  */
371 static const char *
372 dissect(struct match *m, const char *start, const char *stop, sopno startst,
373     sopno stopst)
374 {
375 	int i;
376 	sopno ss;		/* start sop of current subRE */
377 	sopno es;		/* end sop of current subRE */
378 	const char *sp;		/* start of string matched by it */
379 	const char *stp;	/* string matched by it cannot pass here */
380 	const char *rest;	/* start of rest of string */
381 	const char *tail;	/* string unmatched by rest of RE */
382 	sopno ssub;		/* start sop of subsubRE */
383 	sopno esub;		/* end sop of subsubRE */
384 	const char *ssp;	/* start of string matched by subsubRE */
385 	const char *sep;	/* end of string matched by subsubRE */
386 	const char *oldssp;	/* previous ssp */
387 	const char *dp;
388 
389 	AT("diss", start, stop, startst, stopst);
390 	sp = start;
391 	for (ss = startst; ss < stopst; ss = es) {
392 		/* identify end of subRE */
393 		es = ss;
394 		switch (OP(m->g->strip[es])) {
395 		case OPLUS_:
396 		case OQUEST_:
397 			es += OPND(m->g->strip[es]);
398 			break;
399 		case OCH_:
400 			while (OP(m->g->strip[es]) != O_CH)
401 				es += OPND(m->g->strip[es]);
402 			break;
403 		}
404 		es++;
405 
406 		/* figure out what it matched */
407 		switch (OP(m->g->strip[ss])) {
408 		case OEND:
409 			assert(0);
410 			break;
411 		case OCHAR:
412 			sp += XMBRTOWC(NULL, sp, stop - start, &m->mbs, 0);
413 			break;
414 		case OBOL:
415 		case OEOL:
416 		case OBOW:
417 		case OEOW:
418 			break;
419 		case OANY:
420 		case OANYOF:
421 			sp += XMBRTOWC(NULL, sp, stop - start, &m->mbs, 0);
422 			break;
423 		case OBACK_:
424 		case O_BACK:
425 			assert(0);
426 			break;
427 		/* cases where length of match is hard to find */
428 		case OQUEST_:
429 			stp = stop;
430 			for (;;) {
431 				/* how long could this one be? */
432 				rest = slow(m, sp, stp, ss, es);
433 				assert(rest != NULL);	/* it did match */
434 				/* could the rest match the rest? */
435 				tail = slow(m, rest, stop, es, stopst);
436 				if (tail == stop)
437 					break;		/* yes! */
438 				/* no -- try a shorter match for this one */
439 				stp = rest - 1;
440 				assert(stp >= sp);	/* it did work */
441 			}
442 			ssub = ss + 1;
443 			esub = es - 1;
444 			/* did innards match? */
445 			if (slow(m, sp, rest, ssub, esub) != NULL) {
446 				dp = dissect(m, sp, rest, ssub, esub);
447 				assert(dp == rest);
448 #if defined(__lint)
449 				(void) dp;
450 #endif
451 			} else		/* no */
452 				assert(sp == rest);
453 			sp = rest;
454 			break;
455 		case OPLUS_:
456 			stp = stop;
457 			for (;;) {
458 				/* how long could this one be? */
459 				rest = slow(m, sp, stp, ss, es);
460 				assert(rest != NULL);	/* it did match */
461 				/* could the rest match the rest? */
462 				tail = slow(m, rest, stop, es, stopst);
463 				if (tail == stop)
464 					break;		/* yes! */
465 				/* no -- try a shorter match for this one */
466 				stp = rest - 1;
467 				assert(stp >= sp);	/* it did work */
468 			}
469 			ssub = ss + 1;
470 			esub = es - 1;
471 			ssp = sp;
472 			oldssp = ssp;
473 			for (;;) {	/* find last match of innards */
474 				sep = slow(m, ssp, rest, ssub, esub);
475 				if (sep == NULL || sep == ssp)
476 					break;	/* failed or matched null */
477 				oldssp = ssp;	/* on to next try */
478 				ssp = sep;
479 			}
480 			if (sep == NULL) {
481 				/* last successful match */
482 				sep = ssp;
483 				ssp = oldssp;
484 			}
485 			assert(sep == rest);	/* must exhaust substring */
486 			assert(slow(m, ssp, sep, ssub, esub) == rest);
487 			dp = dissect(m, ssp, sep, ssub, esub);
488 			assert(dp == sep);
489 			sp = rest;
490 			break;
491 		case OCH_:
492 			stp = stop;
493 			for (;;) {
494 				/* how long could this one be? */
495 				rest = slow(m, sp, stp, ss, es);
496 				assert(rest != NULL);	/* it did match */
497 				/* could the rest match the rest? */
498 				tail = slow(m, rest, stop, es, stopst);
499 				if (tail == stop)
500 					break;		/* yes! */
501 				/* no -- try a shorter match for this one */
502 				stp = rest - 1;
503 				assert(stp >= sp);	/* it did work */
504 			}
505 			ssub = ss + 1;
506 			esub = ss + OPND(m->g->strip[ss]) - 1;
507 			assert(OP(m->g->strip[esub]) == OOR1);
508 			for (;;) {	/* find first matching branch */
509 				if (slow(m, sp, rest, ssub, esub) == rest)
510 					break;	/* it matched all of it */
511 				/* that one missed, try next one */
512 				assert(OP(m->g->strip[esub]) == OOR1);
513 				esub++;
514 				assert(OP(m->g->strip[esub]) == OOR2);
515 				ssub = esub + 1;
516 				esub += OPND(m->g->strip[esub]);
517 				if (OP(m->g->strip[esub]) == OOR2)
518 					esub--;
519 				else
520 					assert(OP(m->g->strip[esub]) == O_CH);
521 			}
522 			dp = dissect(m, sp, rest, ssub, esub);
523 			assert(dp == rest);
524 			sp = rest;
525 			break;
526 		case O_PLUS:
527 		case O_QUEST:
528 		case OOR1:
529 		case OOR2:
530 		case O_CH:
531 			assert(0);
532 			break;
533 		case OLPAREN:
534 			i = OPND(m->g->strip[ss]);
535 			assert(0 < i && i <= m->g->nsub);
536 			m->pmatch[i].rm_so = sp - m->offp;
537 			break;
538 		case ORPAREN:
539 			i = OPND(m->g->strip[ss]);
540 			assert(0 < i && i <= m->g->nsub);
541 			m->pmatch[i].rm_eo = sp - m->offp;
542 			break;
543 		default:		/* uh oh */
544 			assert(0);
545 			break;
546 		}
547 	}
548 
549 	assert(sp == stop);
550 	return (sp);
551 }
552 
553 /*
554  * backref - figure out what matched what, figuring in back references
555  */
556 static const char *
557 backref(struct match *m, const char *start, const char *stop, sopno startst,
558     sopno stopst, sopno lev,		/* PLUS nesting level */
559     int rec)
560 {
561 	int i;
562 	sopno ss;		/* start sop of current subRE */
563 	const char *sp;		/* start of string matched by it */
564 	sopno ssub;		/* start sop of subsubRE */
565 	sopno esub;		/* end sop of subsubRE */
566 	const char *ssp;	/* start of string matched by subsubRE */
567 	const char *dp;
568 	size_t len;
569 	int hard;
570 	sop s;
571 	regoff_t offsave;
572 	cset *cs;
573 	wint_t wc;
574 
575 	AT("back", start, stop, startst, stopst);
576 	sp = start;
577 
578 	/* get as far as we can with easy stuff */
579 	hard = 0;
580 	for (ss = startst; !hard && ss < stopst; ss++)
581 		switch (OP(s = m->g->strip[ss])) {
582 		case OCHAR:
583 			if (sp == stop)
584 				return (NULL);
585 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
586 			if (wc != OPND(s))
587 				return (NULL);
588 			break;
589 		case OANY:
590 			if (sp == stop)
591 				return (NULL);
592 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
593 			if (wc == BADCHAR)
594 				return (NULL);
595 			break;
596 		case OANYOF:
597 			if (sp == stop)
598 				return (NULL);
599 			cs = &m->g->sets[OPND(s)];
600 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
601 			if (wc == BADCHAR || !CHIN(cs, wc))
602 				return (NULL);
603 			break;
604 		case OBOL:
605 			if ((sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
606 			    (sp < m->endp && *(sp-1) == '\n' &&
607 			    (m->g->cflags&REG_NEWLINE))) {
608 				break;
609 			}
610 			else
611 				return (NULL);
612 			break;
613 		case OEOL:
614 			if ((sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
615 			    (sp < m->endp && *sp == '\n' &&
616 			    (m->g->cflags&REG_NEWLINE))) {
617 				break;
618 			}
619 			else
620 				return (NULL);
621 			break;
622 		case OBOW:
623 			if (((sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
624 			    (sp < m->endp && *(sp-1) == '\n' &&
625 			    (m->g->cflags&REG_NEWLINE)) ||
626 			    (sp > m->beginp && !ISWORD(*(sp-1)))) &&
627 			    (sp < m->endp && ISWORD(*sp))) {
628 				break;
629 			} else
630 				return (NULL);
631 			break;
632 		case OEOW:
633 			if (((sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
634 			    (sp < m->endp && *sp == '\n' &&
635 			    (m->g->cflags&REG_NEWLINE)) ||
636 			    (sp < m->endp && !ISWORD(*sp))) &&
637 			    (sp > m->beginp && ISWORD(*(sp-1)))) {
638 				break;
639 			} else
640 				return (NULL);
641 			break;
642 		case O_QUEST:
643 			break;
644 		case OOR1:	/* matches null but needs to skip */
645 			ss++;
646 			s = m->g->strip[ss];
647 			do {
648 				assert(OP(s) == OOR2);
649 				ss += OPND(s);
650 			} while (OP(s = m->g->strip[ss]) != O_CH);
651 			/* note that the ss++ gets us past the O_CH */
652 			break;
653 		default:	/* have to make a choice */
654 			hard = 1;
655 			break;
656 		}
657 	if (!hard) {		/* that was it! */
658 		if (sp != stop)
659 			return (NULL);
660 		return (sp);
661 	}
662 	ss--;			/* adjust for the for's final increment */
663 
664 	/* the hard stuff */
665 	AT("hard", sp, stop, ss, stopst);
666 	s = m->g->strip[ss];
667 	switch (OP(s)) {
668 	case OBACK_:		/* the vilest depths */
669 		i = OPND(s);
670 		assert(0 < i && i <= m->g->nsub);
671 		if (m->pmatch[i].rm_eo == -1)
672 			return (NULL);
673 		assert(m->pmatch[i].rm_so != -1);
674 		len = m->pmatch[i].rm_eo - m->pmatch[i].rm_so;
675 		if (len == 0 && rec++ > MAX_RECURSION)
676 			return (NULL);
677 		assert(stop - m->beginp >= len);
678 		if (sp > stop - len)
679 			return (NULL);	/* not enough left to match */
680 		ssp = m->offp + m->pmatch[i].rm_so;
681 		if (memcmp(sp, ssp, len) != 0)
682 			return (NULL);
683 		while (m->g->strip[ss] != SOP(O_BACK, i))
684 			ss++;
685 		return (backref(m, sp+len, stop, ss+1, stopst, lev, rec));
686 		break;
687 	case OQUEST_:		/* to null or not */
688 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
689 		if (dp != NULL)
690 			return (dp);	/* not */
691 		return (backref(m, sp, stop, ss+OPND(s)+1, stopst, lev, rec));
692 		break;
693 	case OPLUS_:
694 		assert(m->lastpos != NULL);
695 		assert(lev+1 <= m->g->nplus);
696 		m->lastpos[lev+1] = sp;
697 		return (backref(m, sp, stop, ss+1, stopst, lev+1, rec));
698 		break;
699 	case O_PLUS:
700 		if (sp == m->lastpos[lev])	/* last pass matched null */
701 			return (backref(m, sp, stop, ss+1, stopst, lev-1, rec));
702 		/* try another pass */
703 		m->lastpos[lev] = sp;
704 		dp = backref(m, sp, stop, ss-OPND(s)+1, stopst, lev, rec);
705 		if (dp == NULL)
706 			return (backref(m, sp, stop, ss+1, stopst, lev-1, rec));
707 		else
708 			return (dp);
709 		break;
710 	case OCH_:		/* find the right one, if any */
711 		ssub = ss + 1;
712 		esub = ss + OPND(s) - 1;
713 		assert(OP(m->g->strip[esub]) == OOR1);
714 		for (;;) {	/* find first matching branch */
715 			dp = backref(m, sp, stop, ssub, esub, lev, rec);
716 			if (dp != NULL)
717 				return (dp);
718 			/* that one missed, try next one */
719 			if (OP(m->g->strip[esub]) == O_CH)
720 				return (NULL);	/* there is none */
721 			esub++;
722 			assert(OP(m->g->strip[esub]) == OOR2);
723 			ssub = esub + 1;
724 			esub += OPND(m->g->strip[esub]);
725 			if (OP(m->g->strip[esub]) == OOR2)
726 				esub--;
727 			else
728 				assert(OP(m->g->strip[esub]) == O_CH);
729 		}
730 		break;
731 	case OLPAREN:		/* must undo assignment if rest fails */
732 		i = OPND(s);
733 		assert(0 < i && i <= m->g->nsub);
734 		offsave = m->pmatch[i].rm_so;
735 		m->pmatch[i].rm_so = sp - m->offp;
736 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
737 		if (dp != NULL)
738 			return (dp);
739 		m->pmatch[i].rm_so = offsave;
740 		return (NULL);
741 		break;
742 	case ORPAREN:		/* must undo assignment if rest fails */
743 		i = OPND(s);
744 		assert(0 < i && i <= m->g->nsub);
745 		offsave = m->pmatch[i].rm_eo;
746 		m->pmatch[i].rm_eo = sp - m->offp;
747 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
748 		if (dp != NULL)
749 			return (dp);
750 		m->pmatch[i].rm_eo = offsave;
751 		return (NULL);
752 		break;
753 	default:		/* uh oh */
754 		assert(0);
755 		break;
756 	}
757 
758 	/* "can't happen" */
759 	assert(0);
760 	return (NULL);
761 }
762 
763 /*
764  * fast - step through the string at top speed
765  */
766 static const char *
767 fast(struct match *m, const char *start, const char *stop, sopno startst,
768     sopno stopst)
769 {
770 	states st = m->st;
771 	states fresh = m->fresh;
772 	states tmp = m->tmp;
773 	const char *p = start;
774 	wint_t c;
775 	wint_t lastc;		/* previous c */
776 	wint_t flagch;
777 	int i;
778 	const char *coldp;	/* last p after which no match was underway */
779 	size_t clen;
780 
781 	CLEAR(st);
782 	SET1(st, startst);
783 	SP("fast", st, *p);
784 	st = step(m->g, startst, stopst, st, NOTHING, st);
785 	ASSIGN(fresh, st);
786 	SP("start", st, *p);
787 	coldp = NULL;
788 	if (start == m->beginp)
789 		c = OUT;
790 	else {
791 		/*
792 		 * XXX Wrong if the previous character was multi-byte.
793 		 * Newline never is (in encodings supported by FreeBSD),
794 		 * so this only breaks the ISWORD tests below.
795 		 */
796 		c = (uch)*(start - 1);
797 	}
798 	for (;;) {
799 		/* next character */
800 		lastc = c;
801 		if (p == m->endp) {
802 			clen = 0;
803 			c = OUT;
804 		} else
805 			clen = XMBRTOWC(&c, p, m->endp - p, &m->mbs, BADCHAR);
806 		if (EQ(st, fresh))
807 			coldp = p;
808 
809 		/* is there an EOL and/or BOL between lastc and c? */
810 		flagch = '\0';
811 		i = 0;
812 		if ((lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
813 		    (lastc == OUT && !(m->eflags&REG_NOTBOL))) {
814 			flagch = BOL;
815 			i = m->g->nbol;
816 		}
817 		if ((c == '\n' && m->g->cflags&REG_NEWLINE) ||
818 		    (c == OUT && !(m->eflags&REG_NOTEOL))) {
819 			flagch = (flagch == BOL) ? BOLEOL : EOL;
820 			i += m->g->neol;
821 		}
822 		if (i != 0) {
823 			for (; i > 0; i--)
824 				st = step(m->g, startst, stopst, st,
825 				    flagch, st);
826 			SP("boleol", st, c);
827 		}
828 
829 		/* how about a word boundary? */
830 		if ((flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
831 		    (c != OUT && ISWORD(c))) {
832 			flagch = BOW;
833 		}
834 		if ((lastc != OUT && ISWORD(lastc)) &&
835 		    (flagch == EOL || (c != OUT && !ISWORD(c)))) {
836 			flagch = EOW;
837 		}
838 		if (flagch == BOW || flagch == EOW) {
839 			st = step(m->g, startst, stopst, st, flagch, st);
840 			SP("boweow", st, c);
841 		}
842 
843 		/* are we done? */
844 		if (ISSET(st, stopst) || p == stop || clen > stop - p)
845 			break;		/* NOTE BREAK OUT */
846 
847 		/* no, we must deal with this character */
848 		ASSIGN(tmp, st);
849 		ASSIGN(st, fresh);
850 		assert(c != OUT);
851 		st = step(m->g, startst, stopst, tmp, c, st);
852 		SP("aft", st, c);
853 		assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
854 		p += clen;
855 	}
856 
857 	assert(coldp != NULL);
858 	m->coldp = coldp;
859 	if (ISSET(st, stopst))
860 		return (p+XMBRTOWC(NULL, p, stop - p, &m->mbs, 0));
861 	else
862 		return (NULL);
863 }
864 
865 /*
866  * slow - step through the string more deliberately
867  */
868 static const char *
869 slow(struct match *m, const char *start, const char *stop, sopno startst,
870     sopno stopst)
871 {
872 	states st = m->st;
873 	states empty = m->empty;
874 	states tmp = m->tmp;
875 	const char *p = start;
876 	wint_t c;
877 	wint_t lastc;		/* previous c */
878 	wint_t flagch;
879 	int i;
880 	const char *matchp;	/* last p at which a match ended */
881 	size_t clen;
882 
883 	AT("slow", start, stop, startst, stopst);
884 	CLEAR(st);
885 	SET1(st, startst);
886 	SP("sstart", st, *p);
887 	st = step(m->g, startst, stopst, st, NOTHING, st);
888 	matchp = NULL;
889 	if (start == m->beginp)
890 		c = OUT;
891 	else {
892 		/*
893 		 * XXX Wrong if the previous character was multi-byte.
894 		 * Newline never is (in encodings supported by FreeBSD),
895 		 * so this only breaks the ISWORD tests below.
896 		 */
897 		c = (uch)*(start - 1);
898 	}
899 	for (;;) {
900 		/* next character */
901 		lastc = c;
902 		if (p == m->endp) {
903 			c = OUT;
904 			clen = 0;
905 		} else
906 			clen = XMBRTOWC(&c, p, m->endp - p, &m->mbs, BADCHAR);
907 
908 		/* is there an EOL and/or BOL between lastc and c? */
909 		flagch = '\0';
910 		i = 0;
911 		if ((lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
912 		    (lastc == OUT && !(m->eflags&REG_NOTBOL))) {
913 			flagch = BOL;
914 			i = m->g->nbol;
915 		}
916 		if ((c == '\n' && m->g->cflags&REG_NEWLINE) ||
917 		    (c == OUT && !(m->eflags&REG_NOTEOL))) {
918 			flagch = (flagch == BOL) ? BOLEOL : EOL;
919 			i += m->g->neol;
920 		}
921 		if (i != 0) {
922 			for (; i > 0; i--)
923 				st = step(m->g, startst, stopst, st,
924 				    flagch, st);
925 			SP("sboleol", st, c);
926 		}
927 
928 		/* how about a word boundary? */
929 		if ((flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
930 		    (c != OUT && ISWORD(c))) {
931 			flagch = BOW;
932 		}
933 		if ((lastc != OUT && ISWORD(lastc)) &&
934 		    (flagch == EOL || (c != OUT && !ISWORD(c)))) {
935 			flagch = EOW;
936 		}
937 		if (flagch == BOW || flagch == EOW) {
938 			st = step(m->g, startst, stopst, st, flagch, st);
939 			SP("sboweow", st, c);
940 		}
941 
942 		/* are we done? */
943 		if (ISSET(st, stopst))
944 			matchp = p;
945 		if (EQ(st, empty) || p == stop || clen > stop - p)
946 			break;		/* NOTE BREAK OUT */
947 
948 		/* no, we must deal with this character */
949 		ASSIGN(tmp, st);
950 		ASSIGN(st, empty);
951 		assert(c != OUT);
952 		st = step(m->g, startst, stopst, tmp, c, st);
953 		SP("saft", st, c);
954 		assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
955 		p += clen;
956 	}
957 
958 	return (matchp);
959 }
960 
961 
962 /*
963  * step - map set of states reachable before char to set reachable after
964  */
965 static states
966 step(struct re_guts *g,
967 	sopno start,		/* start state within strip */
968 	sopno stop,		/* state after stop state within strip */
969 	states bef,		/* states reachable before */
970 	wint_t ch,		/* character or NONCHAR code */
971 	states aft)		/* states already known reachable after */
972 {
973 	cset *cs;
974 	sop s;
975 	sopno pc;
976 	onestate here;		/* note, macros know this name */
977 	sopno look;
978 	int i;
979 
980 	for (pc = start, INIT(here, pc); pc != stop; pc++, INC(here)) {
981 		s = g->strip[pc];
982 		switch (OP(s)) {
983 		case OEND:
984 			assert(pc == stop-1);
985 			break;
986 		case OCHAR:
987 			/* only characters can match */
988 			assert(!NONCHAR(ch) || ch != OPND(s));
989 			if (ch == OPND(s))
990 				FWD(aft, bef, 1);
991 			break;
992 		case OBOL:
993 			if (ch == BOL || ch == BOLEOL)
994 				FWD(aft, bef, 1);
995 			break;
996 		case OEOL:
997 			if (ch == EOL || ch == BOLEOL)
998 				FWD(aft, bef, 1);
999 			break;
1000 		case OBOW:
1001 			if (ch == BOW)
1002 				FWD(aft, bef, 1);
1003 			break;
1004 		case OEOW:
1005 			if (ch == EOW)
1006 				FWD(aft, bef, 1);
1007 			break;
1008 		case OANY:
1009 			if (!NONCHAR(ch))
1010 				FWD(aft, bef, 1);
1011 			break;
1012 		case OANYOF:
1013 			cs = &g->sets[OPND(s)];
1014 			if (!NONCHAR(ch) && CHIN(cs, ch))
1015 				FWD(aft, bef, 1);
1016 			break;
1017 		case OBACK_:		/* ignored here */
1018 		case O_BACK:
1019 			FWD(aft, aft, 1);
1020 			break;
1021 		case OPLUS_:		/* forward, this is just an empty */
1022 			FWD(aft, aft, 1);
1023 			break;
1024 		case O_PLUS:		/* both forward and back */
1025 			FWD(aft, aft, 1);
1026 			i = ISSETBACK(aft, OPND(s));
1027 			BACK(aft, aft, OPND(s));
1028 			if (!i && ISSETBACK(aft, OPND(s))) {
1029 				/* oho, must reconsider loop body */
1030 				pc -= OPND(s) + 1;
1031 				INIT(here, pc);
1032 			}
1033 			break;
1034 		case OQUEST_:		/* two branches, both forward */
1035 			FWD(aft, aft, 1);
1036 			FWD(aft, aft, OPND(s));
1037 			break;
1038 		case O_QUEST:		/* just an empty */
1039 			FWD(aft, aft, 1);
1040 			break;
1041 		case OLPAREN:		/* not significant here */
1042 		case ORPAREN:
1043 			FWD(aft, aft, 1);
1044 			break;
1045 		case OCH_:		/* mark the first two branches */
1046 			FWD(aft, aft, 1);
1047 			assert(OP(g->strip[pc+OPND(s)]) == OOR2);
1048 			FWD(aft, aft, OPND(s));
1049 			break;
1050 		case OOR1:		/* done a branch, find the O_CH */
1051 			if (ISSTATEIN(aft, here)) {
1052 				for (look = 1;
1053 				    OP(s = g->strip[pc+look]) != O_CH;
1054 				    look += OPND(s))
1055 					assert(OP(s) == OOR2);
1056 				FWD(aft, aft, look + 1);
1057 			}
1058 			break;
1059 		case OOR2:		/* propagate OCH_'s marking */
1060 			FWD(aft, aft, 1);
1061 			if (OP(g->strip[pc+OPND(s)]) != O_CH) {
1062 				assert(OP(g->strip[pc+OPND(s)]) == OOR2);
1063 				FWD(aft, aft, OPND(s));
1064 			}
1065 			break;
1066 		case O_CH:		/* just empty */
1067 			FWD(aft, aft, 1);
1068 			break;
1069 		default:		/* ooooops... */
1070 			assert(0);
1071 			break;
1072 		}
1073 	}
1074 
1075 	return (aft);
1076 }
1077 
1078 #ifdef REDEBUG
1079 /*
1080  * print - print a set of states
1081  */
1082 static void
1083 print(struct match *m, const char *caption, states st, int ch, FILE *d)
1084 {
1085 	struct re_guts *g = m->g;
1086 	int i;
1087 	int first = 1;
1088 
1089 	if (!(m->eflags&REG_TRACE))
1090 		return;
1091 
1092 	(void) fprintf(d, "%s", caption);
1093 	if (ch != '\0')
1094 		(void) fprintf(d, " %s", pchar(ch));
1095 	for (i = 0; i < g->nstates; i++)
1096 		if (ISSET(st, i)) {
1097 			(void) fprintf(d, "%s%d", (first) ? "\t" : ", ", i);
1098 			first = 0;
1099 		}
1100 	(void) fprintf(d, "\n");
1101 }
1102 
1103 /*
1104  * at - print current situation
1105  */
1106 static void
1107 at(struct match *m, const char *title, const char *start, const char *stop,
1108     sopno startst, sopno stopst)
1109 {
1110 	if (!(m->eflags&REG_TRACE))
1111 		return;
1112 
1113 	(void) printf("%s %s-", title, pchar(*start));
1114 	(void) printf("%s ", pchar(*stop));
1115 	(void) printf("%ld-%ld\n", (long)startst, (long)stopst);
1116 }
1117 
1118 #ifndef PCHARDONE
1119 #define	PCHARDONE	/* never again */
1120 /*
1121  * pchar - make a character printable
1122  *
1123  * Is this identical to regchar() over in debug.c?  Well, yes.  But a
1124  * duplicate here avoids having a debugging-capable regexec.o tied to
1125  * a matching debug.o, and this is convenient.  It all disappears in
1126  * the non-debug compilation anyway, so it doesn't matter much.
1127  */
1128 static const char *
1129 pchar(int ch)
1130 {
1131 	static char pbuf[10];
1132 
1133 	if (isprint((uch)ch) || ch == ' ')
1134 		(void) sprintf(pbuf, "%c", ch);
1135 	else
1136 		(void) sprintf(pbuf, "\\%o", ch);
1137 	return (pbuf);
1138 }
1139 #endif
1140 #endif
1141 
1142 #undef	matcher
1143 #undef	fast
1144 #undef	slow
1145 #undef	dissect
1146 #undef	backref
1147 #undef	step
1148 #undef	print
1149 #undef	at
1150 #undef	match
1151