xref: /titanic_41/usr/src/lib/libc/port/locale/engine.c (revision 4297a3b0d0a35d80f86fff155e288e885a100e6d)
1 /*
2  * Copyright (c) 1992, 1993, 1994 Henry Spencer.
3  * Copyright (c) 1992, 1993, 1994
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * This code is derived from software contributed to Berkeley by
7  * Henry Spencer.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright 2010 Nexenta Systems, Inc.  All rights reserved.
36  * Use is subject to license terms.
37  */
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	matcher	smatcher
48 #define	fast	sfast
49 #define	slow	sslow
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	matcher	lmatcher
59 #define	fast	lfast
60 #define	slow	lslow
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	matcher	mmatcher
70 #define	fast	mfast
71 #define	slow	mslow
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 *fast(struct match *, const char *, const char *, sopno,
110     sopno);
111 static const char *slow(struct match *, const char *, const char *, sopno,
112     sopno);
113 static states step(struct re_guts *, sopno, sopno, states, wint_t, states);
114 #define	MAX_RECURSION	100
115 #define	BOL	(OUT-1)
116 #define	EOL	(BOL-1)
117 #define	BOLEOL	(BOL-2)
118 #define	NOTHING	(BOL-3)
119 #define	BOW	(BOL-4)
120 #define	EOW	(BOL-5)
121 #define	BADCHAR	(BOL-6)
122 #define	NONCHAR(c)	((c) <= OUT)
123 #ifdef REDEBUG
124 static void print(struct match *, const char *, states, int, FILE *);
125 #endif
126 #ifdef REDEBUG
127 static void at(struct match *, const char *, const char *, const char *,
128     sopno, sopno);
129 #endif
130 #ifdef REDEBUG
131 static const char *pchar(int ch);
132 #endif
133 
134 #ifdef __cplusplus
135 }
136 #endif
137 /* ========= end header generated by ./mkh ========= */
138 
139 #ifdef REDEBUG
140 #define	SP(t, s, c)	print(m, t, s, c, stdout)
141 #define	AT(t, p1, p2, s1, s2)	at(m, t, p1, p2, s1, s2)
142 #define	NOTE(str)	{ if (m->eflags&REG_TRACE) printf("=%s\n", (str)); }
143 #else
144 #define	SP(t, s, c)	/* nothing */
145 #define	AT(t, p1, p2, s1, s2)	/* nothing */
146 #define	NOTE(s)	/* nothing */
147 #endif
148 
149 /*
150  * matcher - the actual matching engine
151  */
152 static int			/* 0 success, REG_NOMATCH failure */
153 matcher(struct re_guts *g,
154 	const char *string,
155 	size_t nmatch,
156 	regmatch_t pmatch[],
157 	int eflags)
158 {
159 	const char *endp;
160 	int i;
161 	struct match mv;
162 	struct match *m = &mv;
163 	const char *dp;
164 	const sopno gf = g->firststate+1;	/* +1 for OEND */
165 	const sopno gl = g->laststate;
166 	const char *start;
167 	const char *stop;
168 	/* Boyer-Moore algorithms variables */
169 	const char *pp;
170 	int cj, mj;
171 	const char *mustfirst;
172 	const char *mustlast;
173 	int *matchjump;
174 	int *charjump;
175 
176 	/* simplify the situation where possible */
177 	if (g->cflags&REG_NOSUB)
178 		nmatch = 0;
179 #ifdef	REG_STARTEND
180 	if (eflags&REG_STARTEND) {
181 		start = string + pmatch[0].rm_so;
182 		stop = string + pmatch[0].rm_eo;
183 	} else {
184 		start = string;
185 		stop = start + strlen(start);
186 	}
187 #else
188 	start = string;
189 	stop = start + strlen(start);
190 #endif
191 	if (stop < start)
192 		return (REG_EFATAL);
193 
194 	/* prescreening; this does wonders for this rather slow code */
195 	if (g->must != NULL) {
196 		if (g->charjump != NULL && g->matchjump != NULL) {
197 			mustfirst = g->must;
198 			mustlast = g->must + g->mlen - 1;
199 			charjump = g->charjump;
200 			matchjump = g->matchjump;
201 			pp = mustlast;
202 			for (dp = start+g->mlen-1; dp < stop; ) {
203 				/* Fast skip non-matches */
204 				while (dp < stop && charjump[(int)*dp])
205 					dp += charjump[(int)*dp];
206 
207 				if (dp >= stop)
208 					break;
209 
210 				/* Greedy matcher */
211 				/*
212 				 * We depend on not being used for
213 				 * for strings of length 1
214 				 */
215 				while (*--dp == *--pp && pp != mustfirst)
216 					;
217 
218 				if (*dp == *pp)
219 					break;
220 
221 				/* Jump to next possible match */
222 				mj = matchjump[pp - mustfirst];
223 				cj = charjump[(int)*dp];
224 				dp += (cj < mj ? mj : cj);
225 				pp = mustlast;
226 			}
227 			if (pp != mustfirst)
228 				return (REG_NOMATCH);
229 		} else {
230 			for (dp = start; dp < stop; dp++)
231 				if (*dp == g->must[0] &&
232 				    stop - dp >= g->mlen &&
233 				    memcmp(dp, g->must, (size_t)g->mlen) == 0)
234 					break;
235 			if (dp == stop)		/* we didn't find g->must */
236 				return (REG_NOMATCH);
237 		}
238 	}
239 
240 	/* match struct setup */
241 	m->g = g;
242 	m->eflags = eflags;
243 	m->pmatch = NULL;
244 	m->lastpos = NULL;
245 	m->offp = string;
246 	m->beginp = start;
247 	m->endp = stop;
248 	STATESETUP(m, 4);
249 	SETUP(m->st);
250 	SETUP(m->fresh);
251 	SETUP(m->tmp);
252 	SETUP(m->empty);
253 	CLEAR(m->empty);
254 	ZAPSTATE(&m->mbs);
255 
256 	/* Adjust start according to moffset, to speed things up */
257 	if (g->moffset > -1)
258 		start = ((dp - g->moffset) < start) ? start : dp - g->moffset;
259 
260 	SP("mloop", m->st, *start);
261 
262 	/* this loop does only one repetition except for backrefs */
263 	for (;;) {
264 		endp = fast(m, start, stop, gf, gl);
265 		if (endp == NULL) {		/* a miss */
266 			if (m->pmatch != NULL)
267 				free((char *)m->pmatch);
268 			if (m->lastpos != NULL)
269 				free((char *)m->lastpos);
270 			STATETEARDOWN(m);
271 			return (REG_NOMATCH);
272 		}
273 		if (nmatch == 0 && !g->backrefs)
274 			break;		/* no further info needed */
275 
276 		/* where? */
277 		assert(m->coldp != NULL);
278 		for (;;) {
279 			NOTE("finding start");
280 			endp = slow(m, m->coldp, stop, gf, gl);
281 			if (endp != NULL)
282 				break;
283 			assert(m->coldp < m->endp);
284 			m->coldp += XMBRTOWC(NULL, m->coldp,
285 			    m->endp - m->coldp, &m->mbs, 0);
286 		}
287 		if (nmatch == 1 && !g->backrefs)
288 			break;		/* no further info needed */
289 
290 		/* oh my, he wants the subexpressions... */
291 		if (m->pmatch == NULL)
292 			m->pmatch = (regmatch_t *)malloc((m->g->nsub + 1) *
293 			    sizeof (regmatch_t));
294 		if (m->pmatch == NULL) {
295 			STATETEARDOWN(m);
296 			return (REG_ESPACE);
297 		}
298 		for (i = 1; i <= m->g->nsub; i++)
299 			m->pmatch[i].rm_so = m->pmatch[i].rm_eo = -1;
300 		/* NB: FreeBSD has REG_BACKR, we do not */
301 		if (!g->backrefs /* && !(m->eflags&REG_BACKR) */) {
302 			NOTE("dissecting");
303 			dp = dissect(m, m->coldp, endp, gf, gl);
304 		} else {
305 			if (g->nplus > 0 && m->lastpos == NULL)
306 				m->lastpos = malloc((g->nplus+1) *
307 				    sizeof (const char *));
308 			if (g->nplus > 0 && m->lastpos == NULL) {
309 				free(m->pmatch);
310 				STATETEARDOWN(m);
311 				return (REG_ESPACE);
312 			}
313 			NOTE("backref dissect");
314 			dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
315 		}
316 		if (dp != NULL)
317 			break;
318 
319 		/* uh-oh... we couldn't find a subexpression-level match */
320 		assert(g->backrefs);	/* must be back references doing it */
321 		assert(g->nplus == 0 || m->lastpos != NULL);
322 		for (;;) {
323 			if (dp != NULL || endp <= m->coldp)
324 				break;		/* defeat */
325 			NOTE("backoff");
326 			endp = slow(m, m->coldp, endp-1, gf, gl);
327 			if (endp == NULL)
328 				break;		/* defeat */
329 			/* try it on a shorter possibility */
330 #ifndef NDEBUG
331 			for (i = 1; i <= m->g->nsub; i++) {
332 				assert(m->pmatch[i].rm_so == -1);
333 				assert(m->pmatch[i].rm_eo == -1);
334 			}
335 #endif
336 			NOTE("backoff dissect");
337 			dp = backref(m, m->coldp, endp, gf, gl, (sopno)0, 0);
338 		}
339 		assert(dp == NULL || dp == endp);
340 		if (dp != NULL)		/* found a shorter one */
341 			break;
342 
343 		/* despite initial appearances, there is no match here */
344 		NOTE("false alarm");
345 		/* recycle starting later */
346 		start = m->coldp + XMBRTOWC(NULL, m->coldp,
347 		    stop - m->coldp, &m->mbs, 0);
348 		assert(start <= stop);
349 	}
350 
351 	/* fill in the details if requested */
352 	if (nmatch > 0) {
353 		pmatch[0].rm_so = m->coldp - m->offp;
354 		pmatch[0].rm_eo = endp - m->offp;
355 	}
356 	if (nmatch > 1) {
357 		assert(m->pmatch != NULL);
358 		for (i = 1; i < nmatch; i++)
359 			if (i <= m->g->nsub)
360 				pmatch[i] = m->pmatch[i];
361 			else {
362 				pmatch[i].rm_so = -1;
363 				pmatch[i].rm_eo = -1;
364 			}
365 	}
366 
367 	if (m->pmatch != NULL)
368 		free((char *)m->pmatch);
369 	if (m->lastpos != NULL)
370 		free((char *)m->lastpos);
371 	STATETEARDOWN(m);
372 	return (0);
373 }
374 
375 /*
376  * dissect - figure out what matched what, no back references
377  */
378 static const char *
379 dissect(struct match *m, const char *start, const char *stop, sopno startst,
380     sopno stopst)
381 {
382 	int i;
383 	sopno ss;		/* start sop of current subRE */
384 	sopno es;		/* end sop of current subRE */
385 	const char *sp;		/* start of string matched by it */
386 	const char *stp;	/* string matched by it cannot pass here */
387 	const char *rest;	/* start of rest of string */
388 	const char *tail;	/* string unmatched by rest of RE */
389 	sopno ssub;		/* start sop of subsubRE */
390 	sopno esub;		/* end sop of subsubRE */
391 	const char *ssp;	/* start of string matched by subsubRE */
392 	const char *sep;	/* end of string matched by subsubRE */
393 	const char *oldssp;	/* previous ssp */
394 	const char *dp;
395 
396 	AT("diss", start, stop, startst, stopst);
397 	sp = start;
398 	for (ss = startst; ss < stopst; ss = es) {
399 		/* identify end of subRE */
400 		es = ss;
401 		switch (OP(m->g->strip[es])) {
402 		case OPLUS_:
403 		case OQUEST_:
404 			es += OPND(m->g->strip[es]);
405 			break;
406 		case OCH_:
407 			while (OP(m->g->strip[es]) != O_CH)
408 				es += OPND(m->g->strip[es]);
409 			break;
410 		}
411 		es++;
412 
413 		/* figure out what it matched */
414 		switch (OP(m->g->strip[ss])) {
415 		case OEND:
416 			assert(0);
417 			break;
418 		case OCHAR:
419 			sp += XMBRTOWC(NULL, sp, stop - start, &m->mbs, 0);
420 			break;
421 		case OBOL:
422 		case OEOL:
423 		case OBOW:
424 		case OEOW:
425 			break;
426 		case OANY:
427 		case OANYOF:
428 			sp += XMBRTOWC(NULL, sp, stop - start, &m->mbs, 0);
429 			break;
430 		case OBACK_:
431 		case O_BACK:
432 			assert(0);
433 			break;
434 		/* cases where length of match is hard to find */
435 		case OQUEST_:
436 			stp = stop;
437 			for (;;) {
438 				/* how long could this one be? */
439 				rest = slow(m, sp, stp, ss, es);
440 				assert(rest != NULL);	/* it did match */
441 				/* could the rest match the rest? */
442 				tail = slow(m, rest, stop, es, stopst);
443 				if (tail == stop)
444 					break;		/* yes! */
445 				/* no -- try a shorter match for this one */
446 				stp = rest - 1;
447 				assert(stp >= sp);	/* it did work */
448 			}
449 			ssub = ss + 1;
450 			esub = es - 1;
451 			/* did innards match? */
452 			if (slow(m, sp, rest, ssub, esub) != NULL) {
453 				dp = dissect(m, sp, rest, ssub, esub);
454 				assert(dp == rest);
455 #if defined(__lint)
456 				(void) dp;
457 #endif
458 			} else		/* no */
459 				assert(sp == rest);
460 			sp = rest;
461 			break;
462 		case OPLUS_:
463 			stp = stop;
464 			for (;;) {
465 				/* how long could this one be? */
466 				rest = slow(m, sp, stp, ss, es);
467 				assert(rest != NULL);	/* it did match */
468 				/* could the rest match the rest? */
469 				tail = slow(m, rest, stop, es, stopst);
470 				if (tail == stop)
471 					break;		/* yes! */
472 				/* no -- try a shorter match for this one */
473 				stp = rest - 1;
474 				assert(stp >= sp);	/* it did work */
475 			}
476 			ssub = ss + 1;
477 			esub = es - 1;
478 			ssp = sp;
479 			oldssp = ssp;
480 			for (;;) {	/* find last match of innards */
481 				sep = slow(m, ssp, rest, ssub, esub);
482 				if (sep == NULL || sep == ssp)
483 					break;	/* failed or matched null */
484 				oldssp = ssp;	/* on to next try */
485 				ssp = sep;
486 			}
487 			if (sep == NULL) {
488 				/* last successful match */
489 				sep = ssp;
490 				ssp = oldssp;
491 			}
492 			assert(sep == rest);	/* must exhaust substring */
493 			assert(slow(m, ssp, sep, ssub, esub) == rest);
494 			dp = dissect(m, ssp, sep, ssub, esub);
495 			assert(dp == sep);
496 			sp = rest;
497 			break;
498 		case OCH_:
499 			stp = stop;
500 			for (;;) {
501 				/* how long could this one be? */
502 				rest = slow(m, sp, stp, ss, es);
503 				assert(rest != NULL);	/* it did match */
504 				/* could the rest match the rest? */
505 				tail = slow(m, rest, stop, es, stopst);
506 				if (tail == stop)
507 					break;		/* yes! */
508 				/* no -- try a shorter match for this one */
509 				stp = rest - 1;
510 				assert(stp >= sp);	/* it did work */
511 			}
512 			ssub = ss + 1;
513 			esub = ss + OPND(m->g->strip[ss]) - 1;
514 			assert(OP(m->g->strip[esub]) == OOR1);
515 			for (;;) {	/* find first matching branch */
516 				if (slow(m, sp, rest, ssub, esub) == rest)
517 					break;	/* it matched all of it */
518 				/* that one missed, try next one */
519 				assert(OP(m->g->strip[esub]) == OOR1);
520 				esub++;
521 				assert(OP(m->g->strip[esub]) == OOR2);
522 				ssub = esub + 1;
523 				esub += OPND(m->g->strip[esub]);
524 				if (OP(m->g->strip[esub]) == OOR2)
525 					esub--;
526 				else
527 					assert(OP(m->g->strip[esub]) == O_CH);
528 			}
529 			dp = dissect(m, sp, rest, ssub, esub);
530 			assert(dp == rest);
531 			sp = rest;
532 			break;
533 		case O_PLUS:
534 		case O_QUEST:
535 		case OOR1:
536 		case OOR2:
537 		case O_CH:
538 			assert(0);
539 			break;
540 		case OLPAREN:
541 			i = OPND(m->g->strip[ss]);
542 			assert(0 < i && i <= m->g->nsub);
543 			m->pmatch[i].rm_so = sp - m->offp;
544 			break;
545 		case ORPAREN:
546 			i = OPND(m->g->strip[ss]);
547 			assert(0 < i && i <= m->g->nsub);
548 			m->pmatch[i].rm_eo = sp - m->offp;
549 			break;
550 		default:		/* uh oh */
551 			assert(0);
552 			break;
553 		}
554 	}
555 
556 	assert(sp == stop);
557 	return (sp);
558 }
559 
560 /*
561  * backref - figure out what matched what, figuring in back references
562  */
563 static const char *
564 backref(struct match *m, const char *start, const char *stop, sopno startst,
565     sopno stopst, sopno lev,		/* PLUS nesting level */
566     int rec)
567 {
568 	int i;
569 	sopno ss;		/* start sop of current subRE */
570 	const char *sp;		/* start of string matched by it */
571 	sopno ssub;		/* start sop of subsubRE */
572 	sopno esub;		/* end sop of subsubRE */
573 	const char *ssp;	/* start of string matched by subsubRE */
574 	const char *dp;
575 	size_t len;
576 	int hard;
577 	sop s;
578 	regoff_t offsave;
579 	cset *cs;
580 	wint_t wc;
581 
582 	AT("back", start, stop, startst, stopst);
583 	sp = start;
584 
585 	/* get as far as we can with easy stuff */
586 	hard = 0;
587 	for (ss = startst; !hard && ss < stopst; ss++)
588 		switch (OP(s = m->g->strip[ss])) {
589 		case OCHAR:
590 			if (sp == stop)
591 				return (NULL);
592 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
593 			if (wc != OPND(s))
594 				return (NULL);
595 			break;
596 		case OANY:
597 			if (sp == stop)
598 				return (NULL);
599 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
600 			if (wc == BADCHAR)
601 				return (NULL);
602 			break;
603 		case OANYOF:
604 			if (sp == stop)
605 				return (NULL);
606 			cs = &m->g->sets[OPND(s)];
607 			sp += XMBRTOWC(&wc, sp, stop - sp, &m->mbs, BADCHAR);
608 			if (wc == BADCHAR || !CHIN(cs, wc))
609 				return (NULL);
610 			break;
611 		case OBOL:
612 			if ((sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
613 			    (sp < m->endp && *(sp-1) == '\n' &&
614 			    (m->g->cflags&REG_NEWLINE))) {
615 				break;
616 			}
617 			else
618 				return (NULL);
619 			break;
620 		case OEOL:
621 			if ((sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
622 			    (sp < m->endp && *sp == '\n' &&
623 			    (m->g->cflags&REG_NEWLINE))) {
624 				break;
625 			}
626 			else
627 				return (NULL);
628 			break;
629 		case OBOW:
630 			if (((sp == m->beginp && !(m->eflags&REG_NOTBOL)) ||
631 			    (sp < m->endp && *(sp-1) == '\n' &&
632 			    (m->g->cflags&REG_NEWLINE)) ||
633 			    (sp > m->beginp && !ISWORD(*(sp-1)))) &&
634 			    (sp < m->endp && ISWORD(*sp))) {
635 				break;
636 			} else
637 				return (NULL);
638 			break;
639 		case OEOW:
640 			if (((sp == m->endp && !(m->eflags&REG_NOTEOL)) ||
641 			    (sp < m->endp && *sp == '\n' &&
642 			    (m->g->cflags&REG_NEWLINE)) ||
643 			    (sp < m->endp && !ISWORD(*sp))) &&
644 			    (sp > m->beginp && ISWORD(*(sp-1)))) {
645 				break;
646 			} else
647 				return (NULL);
648 			break;
649 		case O_QUEST:
650 			break;
651 		case OOR1:	/* matches null but needs to skip */
652 			ss++;
653 			s = m->g->strip[ss];
654 			do {
655 				assert(OP(s) == OOR2);
656 				ss += OPND(s);
657 			} while (OP(s = m->g->strip[ss]) != O_CH);
658 			/* note that the ss++ gets us past the O_CH */
659 			break;
660 		default:	/* have to make a choice */
661 			hard = 1;
662 			break;
663 		}
664 	if (!hard) {		/* that was it! */
665 		if (sp != stop)
666 			return (NULL);
667 		return (sp);
668 	}
669 	ss--;			/* adjust for the for's final increment */
670 
671 	/* the hard stuff */
672 	AT("hard", sp, stop, ss, stopst);
673 	s = m->g->strip[ss];
674 	switch (OP(s)) {
675 	case OBACK_:		/* the vilest depths */
676 		i = OPND(s);
677 		assert(0 < i && i <= m->g->nsub);
678 		if (m->pmatch[i].rm_eo == -1)
679 			return (NULL);
680 		assert(m->pmatch[i].rm_so != -1);
681 		len = m->pmatch[i].rm_eo - m->pmatch[i].rm_so;
682 		if (len == 0 && rec++ > MAX_RECURSION)
683 			return (NULL);
684 		assert(stop - m->beginp >= len);
685 		if (sp > stop - len)
686 			return (NULL);	/* not enough left to match */
687 		ssp = m->offp + m->pmatch[i].rm_so;
688 		if (memcmp(sp, ssp, len) != 0)
689 			return (NULL);
690 		while (m->g->strip[ss] != SOP(O_BACK, i))
691 			ss++;
692 		return (backref(m, sp+len, stop, ss+1, stopst, lev, rec));
693 		break;
694 	case OQUEST_:		/* to null or not */
695 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
696 		if (dp != NULL)
697 			return (dp);	/* not */
698 		return (backref(m, sp, stop, ss+OPND(s)+1, stopst, lev, rec));
699 		break;
700 	case OPLUS_:
701 		assert(m->lastpos != NULL);
702 		assert(lev+1 <= m->g->nplus);
703 		m->lastpos[lev+1] = sp;
704 		return (backref(m, sp, stop, ss+1, stopst, lev+1, rec));
705 		break;
706 	case O_PLUS:
707 		if (sp == m->lastpos[lev])	/* last pass matched null */
708 			return (backref(m, sp, stop, ss+1, stopst, lev-1, rec));
709 		/* try another pass */
710 		m->lastpos[lev] = sp;
711 		dp = backref(m, sp, stop, ss-OPND(s)+1, stopst, lev, rec);
712 		if (dp == NULL)
713 			return (backref(m, sp, stop, ss+1, stopst, lev-1, rec));
714 		else
715 			return (dp);
716 		break;
717 	case OCH_:		/* find the right one, if any */
718 		ssub = ss + 1;
719 		esub = ss + OPND(s) - 1;
720 		assert(OP(m->g->strip[esub]) == OOR1);
721 		for (;;) {	/* find first matching branch */
722 			dp = backref(m, sp, stop, ssub, esub, lev, rec);
723 			if (dp != NULL)
724 				return (dp);
725 			/* that one missed, try next one */
726 			if (OP(m->g->strip[esub]) == O_CH)
727 				return (NULL);	/* there is none */
728 			esub++;
729 			assert(OP(m->g->strip[esub]) == OOR2);
730 			ssub = esub + 1;
731 			esub += OPND(m->g->strip[esub]);
732 			if (OP(m->g->strip[esub]) == OOR2)
733 				esub--;
734 			else
735 				assert(OP(m->g->strip[esub]) == O_CH);
736 		}
737 		break;
738 	case OLPAREN:		/* must undo assignment if rest fails */
739 		i = OPND(s);
740 		assert(0 < i && i <= m->g->nsub);
741 		offsave = m->pmatch[i].rm_so;
742 		m->pmatch[i].rm_so = sp - m->offp;
743 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
744 		if (dp != NULL)
745 			return (dp);
746 		m->pmatch[i].rm_so = offsave;
747 		return (NULL);
748 		break;
749 	case ORPAREN:		/* must undo assignment if rest fails */
750 		i = OPND(s);
751 		assert(0 < i && i <= m->g->nsub);
752 		offsave = m->pmatch[i].rm_eo;
753 		m->pmatch[i].rm_eo = sp - m->offp;
754 		dp = backref(m, sp, stop, ss+1, stopst, lev, rec);
755 		if (dp != NULL)
756 			return (dp);
757 		m->pmatch[i].rm_eo = offsave;
758 		return (NULL);
759 		break;
760 	default:		/* uh oh */
761 		assert(0);
762 		break;
763 	}
764 
765 	/* "can't happen" */
766 	assert(0);
767 	return (NULL);
768 }
769 
770 /*
771  * fast - step through the string at top speed
772  */
773 static const char *
774 fast(struct match *m, const char *start, const char *stop, sopno startst,
775     sopno stopst)
776 {
777 	states st = m->st;
778 	states fresh = m->fresh;
779 	states tmp = m->tmp;
780 	const char *p = start;
781 	wint_t c;
782 	wint_t lastc;		/* previous c */
783 	wint_t flagch;
784 	int i;
785 	const char *coldp;	/* last p after which no match was underway */
786 	size_t clen;
787 
788 	CLEAR(st);
789 	SET1(st, startst);
790 	SP("fast", st, *p);
791 	st = step(m->g, startst, stopst, st, NOTHING, st);
792 	ASSIGN(fresh, st);
793 	SP("start", st, *p);
794 	coldp = NULL;
795 	if (start == m->beginp)
796 		c = OUT;
797 	else {
798 		/*
799 		 * XXX Wrong if the previous character was multi-byte.
800 		 * Newline never is (in encodings supported by FreeBSD),
801 		 * so this only breaks the ISWORD tests below.
802 		 */
803 		c = (uch)*(start - 1);
804 	}
805 	for (;;) {
806 		/* next character */
807 		lastc = c;
808 		if (p == m->endp) {
809 			clen = 0;
810 			c = OUT;
811 		} else
812 			clen = XMBRTOWC(&c, p, m->endp - p, &m->mbs, BADCHAR);
813 		if (EQ(st, fresh))
814 			coldp = p;
815 
816 		/* is there an EOL and/or BOL between lastc and c? */
817 		flagch = '\0';
818 		i = 0;
819 		if ((lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
820 		    (lastc == OUT && !(m->eflags&REG_NOTBOL))) {
821 			flagch = BOL;
822 			i = m->g->nbol;
823 		}
824 		if ((c == '\n' && m->g->cflags&REG_NEWLINE) ||
825 		    (c == OUT && !(m->eflags&REG_NOTEOL))) {
826 			flagch = (flagch == BOL) ? BOLEOL : EOL;
827 			i += m->g->neol;
828 		}
829 		if (i != 0) {
830 			for (; i > 0; i--)
831 				st = step(m->g, startst, stopst, st,
832 				    flagch, st);
833 			SP("boleol", st, c);
834 		}
835 
836 		/* how about a word boundary? */
837 		if ((flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
838 		    (c != OUT && ISWORD(c))) {
839 			flagch = BOW;
840 		}
841 		if ((lastc != OUT && ISWORD(lastc)) &&
842 		    (flagch == EOL || (c != OUT && !ISWORD(c)))) {
843 			flagch = EOW;
844 		}
845 		if (flagch == BOW || flagch == EOW) {
846 			st = step(m->g, startst, stopst, st, flagch, st);
847 			SP("boweow", st, c);
848 		}
849 
850 		/* are we done? */
851 		if (ISSET(st, stopst) || p == stop || clen > stop - p)
852 			break;		/* NOTE BREAK OUT */
853 
854 		/* no, we must deal with this character */
855 		ASSIGN(tmp, st);
856 		ASSIGN(st, fresh);
857 		assert(c != OUT);
858 		st = step(m->g, startst, stopst, tmp, c, st);
859 		SP("aft", st, c);
860 		assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
861 		p += clen;
862 	}
863 
864 	assert(coldp != NULL);
865 	m->coldp = coldp;
866 	if (ISSET(st, stopst))
867 		return (p+XMBRTOWC(NULL, p, stop - p, &m->mbs, 0));
868 	else
869 		return (NULL);
870 }
871 
872 /*
873  * slow - step through the string more deliberately
874  */
875 static const char *
876 slow(struct match *m, const char *start, const char *stop, sopno startst,
877     sopno stopst)
878 {
879 	states st = m->st;
880 	states empty = m->empty;
881 	states tmp = m->tmp;
882 	const char *p = start;
883 	wint_t c;
884 	wint_t lastc;		/* previous c */
885 	wint_t flagch;
886 	int i;
887 	const char *matchp;	/* last p at which a match ended */
888 	size_t clen;
889 
890 	AT("slow", start, stop, startst, stopst);
891 	CLEAR(st);
892 	SET1(st, startst);
893 	SP("sstart", st, *p);
894 	st = step(m->g, startst, stopst, st, NOTHING, st);
895 	matchp = NULL;
896 	if (start == m->beginp)
897 		c = OUT;
898 	else {
899 		/*
900 		 * XXX Wrong if the previous character was multi-byte.
901 		 * Newline never is (in encodings supported by FreeBSD),
902 		 * so this only breaks the ISWORD tests below.
903 		 */
904 		c = (uch)*(start - 1);
905 	}
906 	for (;;) {
907 		/* next character */
908 		lastc = c;
909 		if (p == m->endp) {
910 			c = OUT;
911 			clen = 0;
912 		} else
913 			clen = XMBRTOWC(&c, p, m->endp - p, &m->mbs, BADCHAR);
914 
915 		/* is there an EOL and/or BOL between lastc and c? */
916 		flagch = '\0';
917 		i = 0;
918 		if ((lastc == '\n' && m->g->cflags&REG_NEWLINE) ||
919 		    (lastc == OUT && !(m->eflags&REG_NOTBOL))) {
920 			flagch = BOL;
921 			i = m->g->nbol;
922 		}
923 		if ((c == '\n' && m->g->cflags&REG_NEWLINE) ||
924 		    (c == OUT && !(m->eflags&REG_NOTEOL))) {
925 			flagch = (flagch == BOL) ? BOLEOL : EOL;
926 			i += m->g->neol;
927 		}
928 		if (i != 0) {
929 			for (; i > 0; i--)
930 				st = step(m->g, startst, stopst, st,
931 				    flagch, st);
932 			SP("sboleol", st, c);
933 		}
934 
935 		/* how about a word boundary? */
936 		if ((flagch == BOL || (lastc != OUT && !ISWORD(lastc))) &&
937 		    (c != OUT && ISWORD(c))) {
938 			flagch = BOW;
939 		}
940 		if ((lastc != OUT && ISWORD(lastc)) &&
941 		    (flagch == EOL || (c != OUT && !ISWORD(c)))) {
942 			flagch = EOW;
943 		}
944 		if (flagch == BOW || flagch == EOW) {
945 			st = step(m->g, startst, stopst, st, flagch, st);
946 			SP("sboweow", st, c);
947 		}
948 
949 		/* are we done? */
950 		if (ISSET(st, stopst))
951 			matchp = p;
952 		if (EQ(st, empty) || p == stop || clen > stop - p)
953 			break;		/* NOTE BREAK OUT */
954 
955 		/* no, we must deal with this character */
956 		ASSIGN(tmp, st);
957 		ASSIGN(st, empty);
958 		assert(c != OUT);
959 		st = step(m->g, startst, stopst, tmp, c, st);
960 		SP("saft", st, c);
961 		assert(EQ(step(m->g, startst, stopst, st, NOTHING, st), st));
962 		p += clen;
963 	}
964 
965 	return (matchp);
966 }
967 
968 
969 /*
970  * step - map set of states reachable before char to set reachable after
971  */
972 static states
973 step(struct re_guts *g,
974 	sopno start,		/* start state within strip */
975 	sopno stop,		/* state after stop state within strip */
976 	states bef,		/* states reachable before */
977 	wint_t ch,		/* character or NONCHAR code */
978 	states aft)		/* states already known reachable after */
979 {
980 	cset *cs;
981 	sop s;
982 	sopno pc;
983 	onestate here;		/* note, macros know this name */
984 	sopno look;
985 	int i;
986 
987 	for (pc = start, INIT(here, pc); pc != stop; pc++, INC(here)) {
988 		s = g->strip[pc];
989 		switch (OP(s)) {
990 		case OEND:
991 			assert(pc == stop-1);
992 			break;
993 		case OCHAR:
994 			/* only characters can match */
995 			assert(!NONCHAR(ch) || ch != OPND(s));
996 			if (ch == OPND(s))
997 				FWD(aft, bef, 1);
998 			break;
999 		case OBOL:
1000 			if (ch == BOL || ch == BOLEOL)
1001 				FWD(aft, bef, 1);
1002 			break;
1003 		case OEOL:
1004 			if (ch == EOL || ch == BOLEOL)
1005 				FWD(aft, bef, 1);
1006 			break;
1007 		case OBOW:
1008 			if (ch == BOW)
1009 				FWD(aft, bef, 1);
1010 			break;
1011 		case OEOW:
1012 			if (ch == EOW)
1013 				FWD(aft, bef, 1);
1014 			break;
1015 		case OANY:
1016 			if (!NONCHAR(ch))
1017 				FWD(aft, bef, 1);
1018 			break;
1019 		case OANYOF:
1020 			cs = &g->sets[OPND(s)];
1021 			if (!NONCHAR(ch) && CHIN(cs, ch))
1022 				FWD(aft, bef, 1);
1023 			break;
1024 		case OBACK_:		/* ignored here */
1025 		case O_BACK:
1026 			FWD(aft, aft, 1);
1027 			break;
1028 		case OPLUS_:		/* forward, this is just an empty */
1029 			FWD(aft, aft, 1);
1030 			break;
1031 		case O_PLUS:		/* both forward and back */
1032 			FWD(aft, aft, 1);
1033 			i = ISSETBACK(aft, OPND(s));
1034 			BACK(aft, aft, OPND(s));
1035 			if (!i && ISSETBACK(aft, OPND(s))) {
1036 				/* oho, must reconsider loop body */
1037 				pc -= OPND(s) + 1;
1038 				INIT(here, pc);
1039 			}
1040 			break;
1041 		case OQUEST_:		/* two branches, both forward */
1042 			FWD(aft, aft, 1);
1043 			FWD(aft, aft, OPND(s));
1044 			break;
1045 		case O_QUEST:		/* just an empty */
1046 			FWD(aft, aft, 1);
1047 			break;
1048 		case OLPAREN:		/* not significant here */
1049 		case ORPAREN:
1050 			FWD(aft, aft, 1);
1051 			break;
1052 		case OCH_:		/* mark the first two branches */
1053 			FWD(aft, aft, 1);
1054 			assert(OP(g->strip[pc+OPND(s)]) == OOR2);
1055 			FWD(aft, aft, OPND(s));
1056 			break;
1057 		case OOR1:		/* done a branch, find the O_CH */
1058 			if (ISSTATEIN(aft, here)) {
1059 				for (look = 1;
1060 				    OP(s = g->strip[pc+look]) != O_CH;
1061 				    look += OPND(s))
1062 					assert(OP(s) == OOR2);
1063 				FWD(aft, aft, look + 1);
1064 			}
1065 			break;
1066 		case OOR2:		/* propagate OCH_'s marking */
1067 			FWD(aft, aft, 1);
1068 			if (OP(g->strip[pc+OPND(s)]) != O_CH) {
1069 				assert(OP(g->strip[pc+OPND(s)]) == OOR2);
1070 				FWD(aft, aft, OPND(s));
1071 			}
1072 			break;
1073 		case O_CH:		/* just empty */
1074 			FWD(aft, aft, 1);
1075 			break;
1076 		default:		/* ooooops... */
1077 			assert(0);
1078 			break;
1079 		}
1080 	}
1081 
1082 	return (aft);
1083 }
1084 
1085 #ifdef REDEBUG
1086 /*
1087  * print - print a set of states
1088  */
1089 static void
1090 print(struct match *m, const char *caption, states st, int ch, FILE *d)
1091 {
1092 	struct re_guts *g = m->g;
1093 	int i;
1094 	int first = 1;
1095 
1096 	if (!(m->eflags&REG_TRACE))
1097 		return;
1098 
1099 	(void) fprintf(d, "%s", caption);
1100 	if (ch != '\0')
1101 		(void) fprintf(d, " %s", pchar(ch));
1102 	for (i = 0; i < g->nstates; i++)
1103 		if (ISSET(st, i)) {
1104 			(void) fprintf(d, "%s%d", (first) ? "\t" : ", ", i);
1105 			first = 0;
1106 		}
1107 	(void) fprintf(d, "\n");
1108 }
1109 
1110 /*
1111  * at - print current situation
1112  */
1113 static void
1114 at(struct match *m, const char *title, const char *start, const char *stop,
1115     sopno startst, sopno stopst)
1116 {
1117 	if (!(m->eflags&REG_TRACE))
1118 		return;
1119 
1120 	(void) printf("%s %s-", title, pchar(*start));
1121 	(void) printf("%s ", pchar(*stop));
1122 	(void) printf("%ld-%ld\n", (long)startst, (long)stopst);
1123 }
1124 
1125 #ifndef PCHARDONE
1126 #define	PCHARDONE	/* never again */
1127 /*
1128  * pchar - make a character printable
1129  *
1130  * Is this identical to regchar() over in debug.c?  Well, yes.  But a
1131  * duplicate here avoids having a debugging-capable regexec.o tied to
1132  * a matching debug.o, and this is convenient.  It all disappears in
1133  * the non-debug compilation anyway, so it doesn't matter much.
1134  */
1135 static const char *
1136 pchar(int ch)
1137 {
1138 	static char pbuf[10];
1139 
1140 	if (isprint((uch)ch) || ch == ' ')
1141 		(void) sprintf(pbuf, "%c", ch);
1142 	else
1143 		(void) sprintf(pbuf, "\\%o", ch);
1144 	return (pbuf);
1145 }
1146 #endif
1147 #endif
1148 
1149 #undef	matcher
1150 #undef	fast
1151 #undef	slow
1152 #undef	dissect
1153 #undef	backref
1154 #undef	step
1155 #undef	print
1156 #undef	at
1157 #undef	match
1158