xref: /titanic_44/usr/src/lib/libc/port/locale/engine.c (revision 5fd03bc0f2e00e7ba02316c2e08f45d52aab15db)
1 /*
2  * Copyright 2010 Nexenta Systems, Inc.  All rights reserved.
3  * Copyright 2012 Milan Jurik. All rights reserved.
4  * Copyright (c) 1992, 1993, 1994 Henry Spencer.
5  * Copyright (c) 1992, 1993, 1994
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Henry Spencer.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 4. 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 
36 /*
37  * The matching engine and friends.  This file is #included by regexec.c
38  * after suitable #defines of a variety of macros used herein, so that
39  * different state representations can be used without duplicating masses
40  * of code.
41  */
42 
43 #ifdef SNAMES
44 #define	matcher	smatcher
45 #define	fast	sfast
46 #define	slow	sslow
47 #define	dissect	sdissect
48 #define	backref	sbackref
49 #define	step	sstep
50 #define	print	sprint
51 #define	at	sat
52 #define	match	smat
53 #endif
54 #ifdef LNAMES
55 #define	matcher	lmatcher
56 #define	fast	lfast
57 #define	slow	lslow
58 #define	dissect	ldissect
59 #define	backref	lbackref
60 #define	step	lstep
61 #define	print	lprint
62 #define	at	lat
63 #define	match	lmat
64 #endif
65 #ifdef MNAMES
66 #define	matcher	mmatcher
67 #define	fast	mfast
68 #define	slow	mslow
69 #define	dissect	mdissect
70 #define	backref	mbackref
71 #define	step	mstep
72 #define	print	mprint
73 #define	at	mat
74 #define	match	mmat
75 #endif
76 
77 /* another structure passed up and down to avoid zillions of parameters */
78 struct match {
79 	struct re_guts *g;
80 	int eflags;
81 	regmatch_t *pmatch;	/* [nsub+1] (0 element unused) */
82 	const char *offp;	/* offsets work from here */
83 	const char *beginp;	/* start of string -- virtual NUL precedes */
84 	const char *endp;	/* end of string -- virtual NUL here */
85 	const char *coldp;	/* can be no match starting before here */
86 	const char **lastpos;	/* [nplus+1] */
87 	STATEVARS;
88 	states st;		/* current states */
89 	states fresh;		/* states for a fresh start */
90 	states tmp;		/* temporary */
91 	states empty;		/* empty set of states */
92 	mbstate_t mbs;		/* multibyte conversion state */
93 };
94 
95 /* ========= begin header generated by ./mkh ========= */
96 #ifdef __cplusplus
97 extern "C" {
98 #endif
99 
100 /* === engine.c === */
101 static int matcher(struct re_guts *, const char *, size_t, regmatch_t[], int);
102 static const char *dissect(struct match *, const char *, const char *,
103     sopno, sopno);
104 static const char *backref(struct match *, const char *, const char *, sopno,
105     sopno, sopno, int);
106 static const char *fast(struct match *, const char *, const char *, sopno,
107     sopno);
108 static const char *slow(struct match *, const char *, const char *, sopno,
109     sopno);
110 static states step(struct re_guts *, sopno, sopno, states, wint_t, states);
111 #define	MAX_RECURSION	100
112 #define	BOL	(OUT-1)
113 #define	EOL	(BOL-1)
114 #define	BOLEOL	(BOL-2)
115 #define	NOTHING	(BOL-3)
116 #define	BOW	(BOL-4)
117 #define	EOW	(BOL-5)
118 #define	BADCHAR	(BOL-6)
119 #define	NONCHAR(c)	((c) <= OUT)
120 #ifdef REDEBUG
121 static void print(struct match *, const char *, states, int, FILE *);
122 #endif
123 #ifdef REDEBUG
124 static void at(struct match *, const char *, const char *, const char *,
125     sopno, sopno);
126 #endif
127 #ifdef REDEBUG
128 static const char *pchar(int ch);
129 #endif
130 
131 #ifdef __cplusplus
132 }
133 #endif
134 /* ========= end header generated by ./mkh ========= */
135 
136 #ifdef REDEBUG
137 #define	SP(t, s, c)	print(m, t, s, c, stdout)
138 #define	AT(t, p1, p2, s1, s2)	at(m, t, p1, p2, s1, s2)
139 #define	NOTE(str)	{ if (m->eflags&REG_TRACE) printf("=%s\n", (str)); }
140 #else
141 #define	SP(t, s, c)	/* nothing */
142 #define	AT(t, p1, p2, s1, s2)	/* nothing */
143 #define	NOTE(s)	/* nothing */
144 #endif
145 
146 /*
147  * matcher - the actual matching engine
148  */
149 static int			/* 0 success, REG_NOMATCH failure */
150 matcher(struct re_guts *g,
151 	const char *string,
152 	size_t nmatch,
153 	regmatch_t pmatch[],
154 	int eflags)
155 {
156 	const char *endp;
157 	int i;
158 	struct match mv;
159 	struct match *m = &mv;
160 	const char *dp;
161 	const sopno gf = g->firststate+1;	/* +1 for OEND */
162 	const sopno gl = g->laststate;
163 	const char *start;
164 	const char *stop;
165 	/* Boyer-Moore algorithms variables */
166 	const char *pp;
167 	int cj, mj;
168 	const char *mustfirst;
169 	const char *mustlast;
170 	int *matchjump;
171 	int *charjump;
172 
173 	/* simplify the situation where possible */
174 	if (g->cflags&REG_NOSUB)
175 		nmatch = 0;
176 
177 	if (eflags&REG_STARTEND) {
178 		start = string + pmatch[0].rm_so;
179 		stop = string + pmatch[0].rm_eo;
180 	} else {
181 		start = string;
182 		stop = start + strlen(start);
183 	}
184 
185 	if (stop < start)
186 		return (REG_EFATAL);
187 
188 	/* prescreening; this does wonders for this rather slow code */
189 	if (g->must != NULL) {
190 		if (g->charjump != NULL && g->matchjump != NULL) {
191 			mustfirst = g->must;
192 			mustlast = g->must + g->mlen - 1;
193 			charjump = g->charjump;
194 			matchjump = g->matchjump;
195 			pp = mustlast;
196 			for (dp = start+g->mlen-1; dp < stop; ) {
197 				/* Fast skip non-matches */
198 				while (dp < stop && charjump[(int)*dp])
199 					dp += charjump[(int)*dp];
200 
201 				if (dp >= stop)
202 					break;
203 
204 				/* Greedy matcher */
205 				/*
206 				 * We depend on not being used for
207 				 * for strings of length 1
208 				 */
209 				while (*--dp == *--pp && pp != mustfirst)
210 					;
211 
212 				if (*dp == *pp)
213 					break;
214 
215 				/* Jump to next possible match */
216 				mj = matchjump[pp - mustfirst];
217 				cj = charjump[(int)*dp];
218 				dp += (cj < mj ? mj : cj);
219 				pp = mustlast;
220 			}
221 			if (pp != mustfirst)
222 				return (REG_NOMATCH);
223 		} else {
224 			for (dp = start; dp < stop; dp++)
225 				if (*dp == g->must[0] &&
226 				    stop - dp >= g->mlen &&
227 				    memcmp(dp, g->must, (size_t)g->mlen) == 0)
228 					break;
229 			if (dp == stop)		/* we didn't find g->must */
230 				return (REG_NOMATCH);
231 		}
232 	}
233 
234 	/* match struct setup */
235 	m->g = g;
236 	m->eflags = eflags;
237 	m->pmatch = NULL;
238 	m->lastpos = NULL;
239 	m->offp = string;
240 	m->beginp = start;
241 	m->endp = stop;
242 	STATESETUP(m, 4);
243 	SETUP(m->st);
244 	SETUP(m->fresh);
245 	SETUP(m->tmp);
246 	SETUP(m->empty);
247 	CLEAR(m->empty);
248 	ZAPSTATE(&m->mbs);
249 
250 	/* Adjust start according to moffset, to speed things up */
251 	if (g->moffset > -1)
252 		start = ((dp - g->moffset) < start) ? start : dp - g->moffset;
253 
254 	SP("mloop", m->st, *start);
255 
256 	/* this loop does only one repetition except for backrefs */
257 	for (;;) {
258 		endp = fast(m, start, stop, gf, gl);
259 		if (endp == NULL) {		/* a miss */
260 			if (m->pmatch != NULL)
261 				free((char *)m->pmatch);
262 			if (m->lastpos != NULL)
263 				free((char *)m->lastpos);
264 			STATETEARDOWN(m);
265 			return (REG_NOMATCH);
266 		}
267 		if (nmatch == 0 && !g->backrefs)
268 			break;		/* no further info needed */
269 
270 		/* where? */
271 		assert(m->coldp != NULL);
272 		for (;;) {
273 			NOTE("finding start");
274 			endp = slow(m, m->coldp, stop, gf, gl);
275 			if (endp != NULL)
276 				break;
277 			assert(m->coldp < m->endp);
278 			m->coldp += XMBRTOWC(NULL, m->coldp,
279 			    m->endp - m->coldp, &m->mbs, 0);
280 		}
281 		if (nmatch == 1 && !g->backrefs)
282 			break;		/* no further info needed */
283 
284 		/* oh my, he wants the subexpressions... */
285 		if (m->pmatch == NULL)
286 			m->pmatch = (regmatch_t *)malloc((m->g->nsub + 1) *
287 			    sizeof (regmatch_t));
288 		if (m->pmatch == NULL) {
289 			STATETEARDOWN(m);
290 			return (REG_ESPACE);
291 		}
292 		for (i = 1; i <= m->g->nsub; i++)
293 			m->pmatch[i].rm_so = m->pmatch[i].rm_eo = -1;
294 		/* NB: FreeBSD has REG_BACKR, we do not */
295 		if (!g->backrefs /* && !(m->eflags&REG_BACKR) */) {
296 			NOTE("dissecting");
297 			dp = dissect(m, m->coldp, endp, gf, gl);
298 		} else {
299 			if (g->nplus > 0 && m->lastpos == NULL)
300 				m->lastpos = malloc((g->nplus+1) *
301 				    sizeof (const char *));
302 			if (g->nplus > 0 && m->lastpos == NULL) {
303 				free(m->pmatch);
304 				STATETEARDOWN(m);
305 				return (REG_ESPACE);
306 			}
307 			NOTE("backref dissect");
308 			dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
309 		}
310 		if (dp != NULL)
311 			break;
312 
313 		/* uh-oh... we couldn't find a subexpression-level match */
314 		assert(g->backrefs);	/* must be back references doing it */
315 		assert(g->nplus == 0 || m->lastpos != NULL);
316 		for (;;) {
317 			if (dp != NULL || endp <= m->coldp)
318 				break;		/* defeat */
319 			NOTE("backoff");
320 			endp = slow(m, m->coldp, endp-1, gf, gl);
321 			if (endp == NULL)
322 				break;		/* defeat */
323 			/* try it on a shorter possibility */
324 #ifndef NDEBUG
325 			for (i = 1; i <= m->g->nsub; i++) {
326 				assert(m->pmatch[i].rm_so == -1);
327 				assert(m->pmatch[i].rm_eo == -1);
328 			}
329 #endif
330 			NOTE("backoff dissect");
331 			dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
332 		}
333 		assert(dp == NULL || dp == endp);
334 		if (dp != NULL)		/* found a shorter one */
335 			break;
336 
337 		/* despite initial appearances, there is no match here */
338 		NOTE("false alarm");
339 		/* recycle starting later */
340 		start = m->coldp + XMBRTOWC(NULL, m->coldp,
341 		    stop - m->coldp, &m->mbs, 0);
342 		assert(start <= stop);
343 	}
344 
345 	/* fill in the details if requested */
346 	if (nmatch > 0) {
347 		pmatch[0].rm_so = m->coldp - m->offp;
348 		pmatch[0].rm_eo = endp - m->offp;
349 	}
350 	if (nmatch > 1) {
351 		assert(m->pmatch != NULL);
352 		for (i = 1; i < nmatch; i++)
353 			if (i <= m->g->nsub)
354 				pmatch[i] = m->pmatch[i];
355 			else {
356 				pmatch[i].rm_so = -1;
357 				pmatch[i].rm_eo = -1;
358 			}
359 	}
360 
361 	if (m->pmatch != NULL)
362 		free((char *)m->pmatch);
363 	if (m->lastpos != NULL)
364 		free((char *)m->lastpos);
365 	STATETEARDOWN(m);
366 	return (0);
367 }
368 
369 /*
370  * dissect - figure out what matched what, no back references
371  */
372 static const char *
373 dissect(struct match *m, const char *start, const char *stop, sopno startst,
374     sopno stopst)
375 {
376 	int i;
377 	sopno ss;		/* start sop of current subRE */
378 	sopno es;		/* end sop of current subRE */
379 	const char *sp;		/* start of string matched by it */
380 	const char *stp;	/* string matched by it cannot pass here */
381 	const char *rest;	/* start of rest of string */
382 	const char *tail;	/* string unmatched by rest of RE */
383 	sopno ssub;		/* start sop of subsubRE */
384 	sopno esub;		/* end sop of subsubRE */
385 	const char *ssp;	/* start of string matched by subsubRE */
386 	const char *sep;	/* end of string matched by subsubRE */
387 	const char *oldssp;	/* previous ssp */
388 	const char *dp;
389 
390 	AT("diss", start, stop, startst, stopst);
391 	sp = start;
392 	for (ss = startst; ss < stopst; ss = es) {
393 		/* identify end of subRE */
394 		es = ss;
395 		switch (OP(m->g->strip[es])) {
396 		case OPLUS_:
397 		case OQUEST_:
398 			es += OPND(m->g->strip[es]);
399 			break;
400 		case OCH_:
401 			while (OP(m->g->strip[es]) != O_CH)
402 				es += OPND(m->g->strip[es]);
403 			break;
404 		}
405 		es++;
406 
407 		/* figure out what it matched */
408 		switch (OP(m->g->strip[ss])) {
409 		case OEND:
410 			assert(0);
411 			break;
412 		case OCHAR:
413 			sp += XMBRTOWC(NULL, sp, stop - start, &m->mbs, 0);
414 			break;
415 		case OBOL:
416 		case OEOL:
417 		case OBOW:
418 		case OEOW:
419 			break;
420 		case OANY:
421 		case OANYOF:
422 			sp += XMBRTOWC(NULL, sp, stop - start, &m->mbs, 0);
423 			break;
424 		case OBACK_:
425 		case O_BACK:
426 			assert(0);
427 			break;
428 		/* cases where length of match is hard to find */
429 		case OQUEST_:
430 			stp = stop;
431 			for (;;) {
432 				/* how long could this one be? */
433 				rest = slow(m, sp, stp, ss, es);
434 				assert(rest != NULL);	/* it did match */
435 				/* could the rest match the rest? */
436 				tail = slow(m, rest, stop, es, stopst);
437 				if (tail == stop)
438 					break;		/* yes! */
439 				/* no -- try a shorter match for this one */
440 				stp = rest - 1;
441 				assert(stp >= sp);	/* it did work */
442 			}
443 			ssub = ss + 1;
444 			esub = es - 1;
445 			/* did innards match? */
446 			if (slow(m, sp, rest, ssub, esub) != NULL) {
447 				dp = dissect(m, sp, rest, ssub, esub);
448 				assert(dp == rest);
449 #if defined(__lint)
450 				(void) dp;
451 #endif
452 			} else		/* no */
453 				assert(sp == rest);
454 			sp = rest;
455 			break;
456 		case OPLUS_:
457 			stp = stop;
458 			for (;;) {
459 				/* how long could this one be? */
460 				rest = slow(m, sp, stp, ss, es);
461 				assert(rest != NULL);	/* it did match */
462 				/* could the rest match the rest? */
463 				tail = slow(m, rest, stop, es, stopst);
464 				if (tail == stop)
465 					break;		/* yes! */
466 				/* no -- try a shorter match for this one */
467 				stp = rest - 1;
468 				assert(stp >= sp);	/* it did work */
469 			}
470 			ssub = ss + 1;
471 			esub = es - 1;
472 			ssp = sp;
473 			oldssp = ssp;
474 			for (;;) {	/* find last match of innards */
475 				sep = slow(m, ssp, rest, ssub, esub);
476 				if (sep == NULL || sep == ssp)
477 					break;	/* failed or matched null */
478 				oldssp = ssp;	/* on to next try */
479 				ssp = sep;
480 			}
481 			if (sep == NULL) {
482 				/* last successful match */
483 				sep = ssp;
484 				ssp = oldssp;
485 			}
486 			assert(sep == rest);	/* must exhaust substring */
487 			assert(slow(m, ssp, sep, ssub, esub) == rest);
488 			dp = dissect(m, ssp, sep, ssub, esub);
489 			assert(dp == sep);
490 			sp = rest;
491 			break;
492 		case OCH_:
493 			stp = stop;
494 			for (;;) {
495 				/* how long could this one be? */
496 				rest = slow(m, sp, stp, ss, es);
497 				assert(rest != NULL);	/* it did match */
498 				/* could the rest match the rest? */
499 				tail = slow(m, rest, stop, es, stopst);
500 				if (tail == stop)
501 					break;		/* yes! */
502 				/* no -- try a shorter match for this one */
503 				stp = rest - 1;
504 				assert(stp >= sp);	/* it did work */
505 			}
506 			ssub = ss + 1;
507 			esub = ss + OPND(m->g->strip[ss]) - 1;
508 			assert(OP(m->g->strip[esub]) == OOR1);
509 			for (;;) {	/* find first matching branch */
510 				if (slow(m, sp, rest, ssub, esub) == rest)
511 					break;	/* it matched all of it */
512 				/* that one missed, try next one */
513 				assert(OP(m->g->strip[esub]) == OOR1);
514 				esub++;
515 				assert(OP(m->g->strip[esub]) == OOR2);
516 				ssub = esub + 1;
517 				esub += OPND(m->g->strip[esub]);
518 				if (OP(m->g->strip[esub]) == OOR2)
519 					esub--;
520 				else
521 					assert(OP(m->g->strip[esub]) == O_CH);
522 			}
523 			dp = dissect(m, sp, rest, ssub, esub);
524 			assert(dp == rest);
525 			sp = rest;
526 			break;
527 		case O_PLUS:
528 		case O_QUEST:
529 		case OOR1:
530 		case OOR2:
531 		case O_CH:
532 			assert(0);
533 			break;
534 		case OLPAREN:
535 			i = OPND(m->g->strip[ss]);
536 			assert(0 < i && i <= m->g->nsub);
537 			m->pmatch[i].rm_so = sp - m->offp;
538 			break;
539 		case ORPAREN:
540 			i = OPND(m->g->strip[ss]);
541 			assert(0 < i && i <= m->g->nsub);
542 			m->pmatch[i].rm_eo = sp - m->offp;
543 			break;
544 		default:		/* uh oh */
545 			assert(0);
546 			break;
547 		}
548 	}
549 
550 	assert(sp == stop);
551 	return (sp);
552 }
553 
554 /*
555  * backref - figure out what matched what, figuring in back references
556  */
557 static const char *
558 backref(struct match *m, const char *start, const char *stop, sopno startst,
559     sopno stopst, sopno lev,		/* PLUS nesting level */
560     int rec)
561 {
562 	int i;
563 	sopno ss;		/* start sop of current subRE */
564 	const char *sp;		/* start of string matched by it */
565 	sopno ssub;		/* start sop of subsubRE */
566 	sopno esub;		/* end sop of subsubRE */
567 	const char *ssp;	/* start of string matched by subsubRE */
568 	const char *dp;
569 	size_t len;
570 	int hard;
571 	sop s;
572 	regoff_t offsave;
573 	cset *cs;
574 	wint_t wc;
575 
576 	AT("back", start, stop, startst, stopst);
577 	sp = start;
578 
579 	/* get as far as we can with easy stuff */
580 	hard = 0;
581 	for (ss = startst; !hard && ss < stopst; ss++)
582 		switch (OP(s = m->g->strip[ss])) {
583 		case OCHAR:
584 			if (sp == stop)
585 				return (NULL);
586 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
587 			if (wc != OPND(s))
588 				return (NULL);
589 			break;
590 		case OANY:
591 			if (sp == stop)
592 				return (NULL);
593 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
594 			if (wc == BADCHAR)
595 				return (NULL);
596 			break;
597 		case OANYOF:
598 			if (sp == stop)
599 				return (NULL);
600 			cs = &m->g->sets[OPND(s)];
601 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
602 			if (wc == BADCHAR || !CHIN(cs, wc))
603 				return (NULL);
604 			break;
605 		case OBOL:
606 			if ((sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
607 			    (sp < m->endp && *(sp-1) == '\n' &&
608 			    (m->g->cflags&REG_NEWLINE))) {
609 				break;
610 			}
611 			return (NULL);
612 		case OEOL:
613 			if ((sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
614 			    (sp < m->endp && *sp == '\n' &&
615 			    (m->g->cflags&REG_NEWLINE))) {
616 				break;
617 			}
618 			return (NULL);
619 		case OBOW:
620 			if (((sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
621 			    (sp < m->endp && *(sp-1) == '\n' &&
622 			    (m->g->cflags&REG_NEWLINE)) ||
623 			    (sp > m->beginp && !ISWORD(*(sp-1)))) &&
624 			    (sp < m->endp && ISWORD(*sp))) {
625 				break;
626 			}
627 			return (NULL);
628 		case OEOW:
629 			if (((sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
630 			    (sp < m->endp && *sp == '\n' &&
631 			    (m->g->cflags&REG_NEWLINE)) ||
632 			    (sp < m->endp && !ISWORD(*sp))) &&
633 			    (sp > m->beginp && ISWORD(*(sp-1)))) {
634 				break;
635 			}
636 			return (NULL);
637 		case O_QUEST:
638 			break;
639 		case OOR1:	/* matches null but needs to skip */
640 			ss++;
641 			s = m->g->strip[ss];
642 			do {
643 				assert(OP(s) == OOR2);
644 				ss += OPND(s);
645 			} while (OP(s = m->g->strip[ss]) != O_CH);
646 			/* note that the ss++ gets us past the O_CH */
647 			break;
648 		default:	/* have to make a choice */
649 			hard = 1;
650 			break;
651 		}
652 	if (!hard) {		/* that was it! */
653 		if (sp != stop)
654 			return (NULL);
655 		return (sp);
656 	}
657 	ss--;			/* adjust for the for's final increment */
658 
659 	/* the hard stuff */
660 	AT("hard", sp, stop, ss, stopst);
661 	s = m->g->strip[ss];
662 	switch (OP(s)) {
663 	case OBACK_:		/* the vilest depths */
664 		i = OPND(s);
665 		assert(0 < i && i <= m->g->nsub);
666 		if (m->pmatch[i].rm_eo == -1)
667 			return (NULL);
668 		assert(m->pmatch[i].rm_so != -1);
669 		len = m->pmatch[i].rm_eo - m->pmatch[i].rm_so;
670 		if (len == 0 && rec++ > MAX_RECURSION)
671 			return (NULL);
672 		assert(stop - m->beginp >= len);
673 		if (sp > stop - len)
674 			return (NULL);	/* not enough left to match */
675 		ssp = m->offp + m->pmatch[i].rm_so;
676 		if (memcmp(sp, ssp, len) != 0)
677 			return (NULL);
678 		while (m->g->strip[ss] != SOP(O_BACK, i))
679 			ss++;
680 		return (backref(m, sp+len, stop, ss+1, stopst, lev, rec));
681 	case OQUEST_:		/* to null or not */
682 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
683 		if (dp != NULL)
684 			return (dp);	/* not */
685 		return (backref(m, sp, stop, ss+OPND(s)+1, stopst, lev, rec));
686 	case OPLUS_:
687 		assert(m->lastpos != NULL);
688 		assert(lev+1 <= m->g->nplus);
689 		m->lastpos[lev+1] = sp;
690 		return (backref(m, sp, stop, ss+1, stopst, lev+1, rec));
691 	case O_PLUS:
692 		if (sp == m->lastpos[lev])	/* last pass matched null */
693 			return (backref(m, sp, stop, ss+1, stopst, lev-1, rec));
694 		/* try another pass */
695 		m->lastpos[lev] = sp;
696 		dp = backref(m, sp, stop, ss-OPND(s)+1, stopst, lev, rec);
697 		if (dp == NULL)
698 			return (backref(m, sp, stop, ss+1, stopst, lev-1, rec));
699 		return (dp);
700 	case OCH_:		/* find the right one, if any */
701 		ssub = ss + 1;
702 		esub = ss + OPND(s) - 1;
703 		assert(OP(m->g->strip[esub]) == OOR1);
704 		for (;;) {	/* find first matching branch */
705 			dp = backref(m, sp, stop, ssub, esub, lev, rec);
706 			if (dp != NULL)
707 				return (dp);
708 			/* that one missed, try next one */
709 			if (OP(m->g->strip[esub]) == O_CH)
710 				return (NULL);	/* there is none */
711 			esub++;
712 			assert(OP(m->g->strip[esub]) == OOR2);
713 			ssub = esub + 1;
714 			esub += OPND(m->g->strip[esub]);
715 			if (OP(m->g->strip[esub]) == OOR2)
716 				esub--;
717 			else
718 				assert(OP(m->g->strip[esub]) == O_CH);
719 		}
720 		/* NOTREACHED */
721 		break;
722 	case OLPAREN:		/* must undo assignment if rest fails */
723 		i = OPND(s);
724 		assert(0 < i && i <= m->g->nsub);
725 		offsave = m->pmatch[i].rm_so;
726 		m->pmatch[i].rm_so = sp - m->offp;
727 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
728 		if (dp != NULL)
729 			return (dp);
730 		m->pmatch[i].rm_so = offsave;
731 		return (NULL);
732 	case ORPAREN:		/* must undo assignment if rest fails */
733 		i = OPND(s);
734 		assert(0 < i && i <= m->g->nsub);
735 		offsave = m->pmatch[i].rm_eo;
736 		m->pmatch[i].rm_eo = sp - m->offp;
737 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
738 		if (dp != NULL)
739 			return (dp);
740 		m->pmatch[i].rm_eo = offsave;
741 		return (NULL);
742 	default:		/* uh oh */
743 		assert(0);
744 		break;
745 	}
746 
747 	/* "can't happen" */
748 	assert(0);
749 	return (NULL);
750 }
751 
752 /*
753  * fast - step through the string at top speed
754  */
755 static const char *
756 fast(struct match *m, const char *start, const char *stop, sopno startst,
757     sopno stopst)
758 {
759 	states st = m->st;
760 	states fresh = m->fresh;
761 	states tmp = m->tmp;
762 	const char *p = start;
763 	wint_t c;
764 	wint_t lastc;		/* previous c */
765 	wint_t flagch;
766 	int i;
767 	const char *coldp;	/* last p after which no match was underway */
768 	size_t clen;
769 
770 	CLEAR(st);
771 	SET1(st, startst);
772 	SP("fast", st, *p);
773 	st = step(m->g, startst, stopst, st, NOTHING, st);
774 	ASSIGN(fresh, st);
775 	SP("start", st, *p);
776 	coldp = NULL;
777 	if (start == m->beginp)
778 		c = OUT;
779 	else {
780 		/*
781 		 * XXX Wrong if the previous character was multi-byte.
782 		 * Newline never is (in encodings supported by FreeBSD),
783 		 * so this only breaks the ISWORD tests below.
784 		 */
785 		c = (uch)*(start - 1);
786 	}
787 	for (;;) {
788 		/* next character */
789 		lastc = c;
790 		if (p == m->endp) {
791 			clen = 0;
792 			c = OUT;
793 		} else
794 			clen = XMBRTOWC(&c, p, m->endp - p, &m->mbs, BADCHAR);
795 		if (EQ(st, fresh))
796 			coldp = p;
797 
798 		/* is there an EOL and/or BOL between lastc and c? */
799 		flagch = '\0';
800 		i = 0;
801 		if ((lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
802 		    (lastc == OUT && !(m->eflags&REG_NOTBOL))) {
803 			flagch = BOL;
804 			i = m->g->nbol;
805 		}
806 		if ((c == '\n' && m->g->cflags&REG_NEWLINE) ||
807 		    (c == OUT && !(m->eflags&REG_NOTEOL))) {
808 			flagch = (flagch == BOL) ? BOLEOL : EOL;
809 			i += m->g->neol;
810 		}
811 		if (i != 0) {
812 			for (; i > 0; i--)
813 				st = step(m->g, startst, stopst, st,
814 				    flagch, st);
815 			SP("boleol", st, c);
816 		}
817 
818 		/* how about a word boundary? */
819 		if ((flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
820 		    (c != OUT && ISWORD(c))) {
821 			flagch = BOW;
822 		}
823 		if ((lastc != OUT && ISWORD(lastc)) &&
824 		    (flagch == EOL || (c != OUT && !ISWORD(c)))) {
825 			flagch = EOW;
826 		}
827 		if (flagch == BOW || flagch == EOW) {
828 			st = step(m->g, startst, stopst, st, flagch, st);
829 			SP("boweow", st, c);
830 		}
831 
832 		/* are we done? */
833 		if (ISSET(st, stopst) || p == stop || clen > stop - p)
834 			break;		/* NOTE BREAK OUT */
835 
836 		/* no, we must deal with this character */
837 		ASSIGN(tmp, st);
838 		ASSIGN(st, fresh);
839 		assert(c != OUT);
840 		st = step(m->g, startst, stopst, tmp, c, st);
841 		SP("aft", st, c);
842 		assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
843 		p += clen;
844 	}
845 
846 	assert(coldp != NULL);
847 	m->coldp = coldp;
848 	if (ISSET(st, stopst))
849 		return (p+XMBRTOWC(NULL, p, stop - p, &m->mbs, 0));
850 	else
851 		return (NULL);
852 }
853 
854 /*
855  * slow - step through the string more deliberately
856  */
857 static const char *
858 slow(struct match *m, const char *start, const char *stop, sopno startst,
859     sopno stopst)
860 {
861 	states st = m->st;
862 	states empty = m->empty;
863 	states tmp = m->tmp;
864 	const char *p = start;
865 	wint_t c;
866 	wint_t lastc;		/* previous c */
867 	wint_t flagch;
868 	int i;
869 	const char *matchp;	/* last p at which a match ended */
870 	size_t clen;
871 
872 	AT("slow", start, stop, startst, stopst);
873 	CLEAR(st);
874 	SET1(st, startst);
875 	SP("sstart", st, *p);
876 	st = step(m->g, startst, stopst, st, NOTHING, st);
877 	matchp = NULL;
878 	if (start == m->beginp)
879 		c = OUT;
880 	else {
881 		/*
882 		 * XXX Wrong if the previous character was multi-byte.
883 		 * Newline never is (in encodings supported by FreeBSD),
884 		 * so this only breaks the ISWORD tests below.
885 		 */
886 		c = (uch)*(start - 1);
887 	}
888 	for (;;) {
889 		/* next character */
890 		lastc = c;
891 		if (p == m->endp) {
892 			c = OUT;
893 			clen = 0;
894 		} else
895 			clen = XMBRTOWC(&c, p, m->endp - p, &m->mbs, BADCHAR);
896 
897 		/* is there an EOL and/or BOL between lastc and c? */
898 		flagch = '\0';
899 		i = 0;
900 		if ((lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
901 		    (lastc == OUT && !(m->eflags&REG_NOTBOL))) {
902 			flagch = BOL;
903 			i = m->g->nbol;
904 		}
905 		if ((c == '\n' && m->g->cflags&REG_NEWLINE) ||
906 		    (c == OUT && !(m->eflags&REG_NOTEOL))) {
907 			flagch = (flagch == BOL) ? BOLEOL : EOL;
908 			i += m->g->neol;
909 		}
910 		if (i != 0) {
911 			for (; i > 0; i--)
912 				st = step(m->g, startst, stopst, st,
913 				    flagch, st);
914 			SP("sboleol", st, c);
915 		}
916 
917 		/* how about a word boundary? */
918 		if ((flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
919 		    (c != OUT && ISWORD(c))) {
920 			flagch = BOW;
921 		}
922 		if ((lastc != OUT && ISWORD(lastc)) &&
923 		    (flagch == EOL || (c != OUT && !ISWORD(c)))) {
924 			flagch = EOW;
925 		}
926 		if (flagch == BOW || flagch == EOW) {
927 			st = step(m->g, startst, stopst, st, flagch, st);
928 			SP("sboweow", st, c);
929 		}
930 
931 		/* are we done? */
932 		if (ISSET(st, stopst))
933 			matchp = p;
934 		if (EQ(st, empty) || p == stop || clen > stop - p)
935 			break;		/* NOTE BREAK OUT */
936 
937 		/* no, we must deal with this character */
938 		ASSIGN(tmp, st);
939 		ASSIGN(st, empty);
940 		assert(c != OUT);
941 		st = step(m->g, startst, stopst, tmp, c, st);
942 		SP("saft", st, c);
943 		assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
944 		p += clen;
945 	}
946 
947 	return (matchp);
948 }
949 
950 
951 /*
952  * step - map set of states reachable before char to set reachable after
953  */
954 static states
955 step(struct re_guts *g,
956 	sopno start,		/* start state within strip */
957 	sopno stop,		/* state after stop state within strip */
958 	states bef,		/* states reachable before */
959 	wint_t ch,		/* character or NONCHAR code */
960 	states aft)		/* states already known reachable after */
961 {
962 	cset *cs;
963 	sop s;
964 	sopno pc;
965 	onestate here;		/* note, macros know this name */
966 	sopno look;
967 	int i;
968 
969 	for (pc = start, INIT(here, pc); pc != stop; pc++, INC(here)) {
970 		s = g->strip[pc];
971 		switch (OP(s)) {
972 		case OEND:
973 			assert(pc == stop-1);
974 			break;
975 		case OCHAR:
976 			/* only characters can match */
977 			assert(!NONCHAR(ch) || ch != OPND(s));
978 			if (ch == OPND(s))
979 				FWD(aft, bef, 1);
980 			break;
981 		case OBOL:
982 			if (ch == BOL || ch == BOLEOL)
983 				FWD(aft, bef, 1);
984 			break;
985 		case OEOL:
986 			if (ch == EOL || ch == BOLEOL)
987 				FWD(aft, bef, 1);
988 			break;
989 		case OBOW:
990 			if (ch == BOW)
991 				FWD(aft, bef, 1);
992 			break;
993 		case OEOW:
994 			if (ch == EOW)
995 				FWD(aft, bef, 1);
996 			break;
997 		case OANY:
998 			if (!NONCHAR(ch))
999 				FWD(aft, bef, 1);
1000 			break;
1001 		case OANYOF:
1002 			cs = &g->sets[OPND(s)];
1003 			if (!NONCHAR(ch) && CHIN(cs, ch))
1004 				FWD(aft, bef, 1);
1005 			break;
1006 		case OBACK_:		/* ignored here */
1007 		case O_BACK:
1008 			FWD(aft, aft, 1);
1009 			break;
1010 		case OPLUS_:		/* forward, this is just an empty */
1011 			FWD(aft, aft, 1);
1012 			break;
1013 		case O_PLUS:		/* both forward and back */
1014 			FWD(aft, aft, 1);
1015 			i = ISSETBACK(aft, OPND(s));
1016 			BACK(aft, aft, OPND(s));
1017 			if (!i && ISSETBACK(aft, OPND(s))) {
1018 				/* oho, must reconsider loop body */
1019 				pc -= OPND(s) + 1;
1020 				INIT(here, pc);
1021 			}
1022 			break;
1023 		case OQUEST_:		/* two branches, both forward */
1024 			FWD(aft, aft, 1);
1025 			FWD(aft, aft, OPND(s));
1026 			break;
1027 		case O_QUEST:		/* just an empty */
1028 			FWD(aft, aft, 1);
1029 			break;
1030 		case OLPAREN:		/* not significant here */
1031 		case ORPAREN:
1032 			FWD(aft, aft, 1);
1033 			break;
1034 		case OCH_:		/* mark the first two branches */
1035 			FWD(aft, aft, 1);
1036 			assert(OP(g->strip[pc+OPND(s)]) == OOR2);
1037 			FWD(aft, aft, OPND(s));
1038 			break;
1039 		case OOR1:		/* done a branch, find the O_CH */
1040 			if (ISSTATEIN(aft, here)) {
1041 				for (look = 1;
1042 				    OP(s = g->strip[pc+look]) != O_CH;
1043 				    look += OPND(s))
1044 					assert(OP(s) == OOR2);
1045 				FWD(aft, aft, look + 1);
1046 			}
1047 			break;
1048 		case OOR2:		/* propagate OCH_'s marking */
1049 			FWD(aft, aft, 1);
1050 			if (OP(g->strip[pc+OPND(s)]) != O_CH) {
1051 				assert(OP(g->strip[pc+OPND(s)]) == OOR2);
1052 				FWD(aft, aft, OPND(s));
1053 			}
1054 			break;
1055 		case O_CH:		/* just empty */
1056 			FWD(aft, aft, 1);
1057 			break;
1058 		default:		/* ooooops... */
1059 			assert(0);
1060 			break;
1061 		}
1062 	}
1063 
1064 	return (aft);
1065 }
1066 
1067 #ifdef REDEBUG
1068 /*
1069  * print - print a set of states
1070  */
1071 static void
1072 print(struct match *m, const char *caption, states st, int ch, FILE *d)
1073 {
1074 	struct re_guts *g = m->g;
1075 	int i;
1076 	int first = 1;
1077 
1078 	if (!(m->eflags&REG_TRACE))
1079 		return;
1080 
1081 	(void) fprintf(d, "%s", caption);
1082 	if (ch != '\0')
1083 		(void) fprintf(d, " %s", pchar(ch));
1084 	for (i = 0; i < g->nstates; i++)
1085 		if (ISSET(st, i)) {
1086 			(void) fprintf(d, "%s%d", (first) ? "\t" : ", ", i);
1087 			first = 0;
1088 		}
1089 	(void) fprintf(d, "\n");
1090 }
1091 
1092 /*
1093  * at - print current situation
1094  */
1095 static void
1096 at(struct match *m, const char *title, const char *start, const char *stop,
1097     sopno startst, sopno stopst)
1098 {
1099 	if (!(m->eflags&REG_TRACE))
1100 		return;
1101 
1102 	(void) printf("%s %s-", title, pchar(*start));
1103 	(void) printf("%s ", pchar(*stop));
1104 	(void) printf("%ld-%ld\n", (long)startst, (long)stopst);
1105 }
1106 
1107 #ifndef PCHARDONE
1108 #define	PCHARDONE	/* never again */
1109 /*
1110  * pchar - make a character printable
1111  *
1112  * Is this identical to regchar() over in debug.c?  Well, yes.  But a
1113  * duplicate here avoids having a debugging-capable regexec.o tied to
1114  * a matching debug.o, and this is convenient.  It all disappears in
1115  * the non-debug compilation anyway, so it doesn't matter much.
1116  */
1117 static const char *
1118 pchar(int ch)
1119 {
1120 	static char pbuf[10];
1121 
1122 	if (isprint((uch)ch) || ch == ' ')
1123 		(void) sprintf(pbuf, "%c", ch);
1124 	else
1125 		(void) sprintf(pbuf, "\\%o", ch);
1126 	return (pbuf);
1127 }
1128 #endif
1129 #endif
1130 
1131 #undef	matcher
1132 #undef	fast
1133 #undef	slow
1134 #undef	dissect
1135 #undef	backref
1136 #undef	step
1137 #undef	print
1138 #undef	at
1139 #undef	match
1140