1 /*	$NetBSD: regcomp.c,v 1.36 2015/09/12 19:08:47 christos Exp $	*/
2 
3 /*-
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  * 3. 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  *	@(#)regcomp.c	8.5 (Berkeley) 3/20/94
35  */
36 
37 /*-
38  * Copyright (c) 1992, 1993, 1994 Henry Spencer.
39  *
40  * This code is derived from software contributed to Berkeley by
41  * Henry Spencer.
42  *
43  * Redistribution and use in source and binary forms, with or without
44  * modification, are permitted provided that the following conditions
45  * are met:
46  * 1. Redistributions of source code must retain the above copyright
47  *    notice, this list of conditions and the following disclaimer.
48  * 2. Redistributions in binary form must reproduce the above copyright
49  *    notice, this list of conditions and the following disclaimer in the
50  *    documentation and/or other materials provided with the distribution.
51  * 3. All advertising materials mentioning features or use of this software
52  *    must display the following acknowledgement:
53  *	This product includes software developed by the University of
54  *	California, Berkeley and its contributors.
55  * 4. Neither the name of the University nor the names of its contributors
56  *    may be used to endorse or promote products derived from this software
57  *    without specific prior written permission.
58  *
59  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
60  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
61  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
62  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
63  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
64  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
65  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
66  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
67  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
68  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
69  * SUCH DAMAGE.
70  *
71  *	@(#)regcomp.c	8.5 (Berkeley) 3/20/94
72  */
73 
74 #include <sys/cdefs.h>
75 #if defined(LIBC_SCCS) && !defined(lint)
76 #if 0
77 static char sccsid[] = "@(#)regcomp.c	8.5 (Berkeley) 3/20/94";
78 #else
79 __RCSID("$NetBSD: regcomp.c,v 1.36 2015/09/12 19:08:47 christos Exp $");
80 #endif
81 #endif /* LIBC_SCCS and not lint */
82 
83 #include "namespace.h"
84 #include <sys/types.h>
85 
86 #include <assert.h>
87 #include <ctype.h>
88 #include <limits.h>
89 #include <stdio.h>
90 #include <stdlib.h>
91 #include <string.h>
92 #include <regex.h>
93 
94 #ifdef __weak_alias
95 __weak_alias(regcomp,_regcomp)
96 #endif
97 
98 #include "utils.h"
99 #include "regex2.h"
100 
101 #include "cclass.h"
102 #include "cname.h"
103 
104 /*
105  * parse structure, passed up and down to avoid global variables and
106  * other clumsinesses
107  */
108 struct parse {
109 	const char *next;	/* next character in RE */
110 	const char *end;	/* end of string (-> NUL normally) */
111 	int error;		/* has an error been seen? */
112 	sop *strip;		/* malloced strip */
113 	sopno ssize;		/* malloced strip size (allocated) */
114 	sopno slen;		/* malloced strip length (used) */
115 	size_t ncsalloc;	/* number of csets allocated */
116 	struct re_guts *g;
117 #	define	NPAREN	10	/* we need to remember () 1-9 for back refs */
118 	sopno pbegin[NPAREN];	/* -> ( ([0] unused) */
119 	sopno pend[NPAREN];	/* -> ) ([0] unused) */
120 };
121 
122 /* ========= begin header generated by ./mkh ========= */
123 #ifdef __cplusplus
124 extern "C" {
125 #endif
126 
127 /* === regcomp.c === */
128 static void p_ere(struct parse *p, int stop, size_t reclimit);
129 static void p_ere_exp(struct parse *p, size_t reclimit);
130 static void p_str(struct parse *p);
131 static void p_bre(struct parse *p, int end1, int end2, size_t reclimit);
132 static int p_simp_re(struct parse *p, int starordinary, size_t reclimit);
133 static int p_count(struct parse *p);
134 static void p_bracket(struct parse *p);
135 static void p_b_term(struct parse *p, cset *cs);
136 static void p_b_cclass(struct parse *p, cset *cs);
137 static void p_b_eclass(struct parse *p, cset *cs);
138 static char p_b_symbol(struct parse *p);
139 static char p_b_coll_elem(struct parse *p, int endc);
140 static int othercase(int ch);
141 static void bothcases(struct parse *p, int ch);
142 static void ordinary(struct parse *p, int ch);
143 static void nonnewline(struct parse *p);
144 static void repeat(struct parse *p, sopno start, int from, int to, size_t reclimit);
145 static int seterr(struct parse *p, int e);
146 static cset *allocset(struct parse *p);
147 static void freeset(struct parse *p, cset *cs);
148 static sopno freezeset(struct parse *p, cset *cs);
149 static int firstch(struct parse *p, cset *cs);
150 static int nch(struct parse *p, cset *cs);
151 static void mcadd(struct parse *p, cset *cs, const char *cp);
152 #if 0
153 static void mcsub(cset *cs, char *cp);
154 static int mcin(cset *cs, char *cp);
155 static char *mcfind(cset *cs, char *cp);
156 #endif
157 static void mcinvert(struct parse *p, cset *cs);
158 static void mccase(struct parse *p, cset *cs);
159 static int isinsets(struct re_guts *g, int c);
160 static int samesets(struct re_guts *g, int c1, int c2);
161 static void categorize(struct parse *p, struct re_guts *g);
162 static sopno dupl(struct parse *p, sopno start, sopno finish);
163 static void doemit(struct parse *p, sop op, sopno opnd);
164 static void doinsert(struct parse *p, sop op, sopno opnd, sopno pos);
165 static void dofwd(struct parse *p, sopno pos, sopno value);
166 static int enlarge(struct parse *p, sopno size);
167 static void stripsnug(struct parse *p, struct re_guts *g);
168 static void findmust(struct parse *p, struct re_guts *g);
169 static sopno pluscount(struct parse *p, struct re_guts *g);
170 
171 #ifdef __cplusplus
172 }
173 #endif
174 /* ========= end header generated by ./mkh ========= */
175 
176 static char nuls[10];		/* place to point scanner in event of error */
177 
178 /*
179  * macros for use with parse structure
180  * BEWARE:  these know that the parse structure is named `p' !!!
181  */
182 #define	PEEK()	(*p->next)
183 #define	PEEK2()	(*(p->next+1))
184 #define	MORE()	(p->next < p->end)
185 #define	MORE2()	(p->next+1 < p->end)
186 #define	SEE(c)	(MORE() && PEEK() == (c))
187 #define	SEETWO(a, b)	(MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
188 #define	EAT(c)	((SEE(c)) ? (NEXT(), 1) : 0)
189 #define	EATTWO(a, b)	((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
190 #define	NEXT()	(p->next++)
191 #define	NEXT2()	(p->next += 2)
192 #define	NEXTn(n)	(p->next += (n))
193 #define	GETNEXT()	(*p->next++)
194 #define	SETERROR(e)	seterr(p, (e))
195 #define	REQUIRE(co, e)	(void) ((co) || SETERROR(e))
196 #define	MUSTSEE(c, e)	(REQUIRE(MORE() && PEEK() == (c), e))
197 #define	MUSTEAT(c, e)	(void) (REQUIRE(MORE() && GETNEXT() == (c), e))
198 #define	MUSTNOTSEE(c, e)	(REQUIRE(!MORE() || PEEK() != (c), e))
199 #define	EMIT(op, sopnd)	doemit(p, (sop)(op), sopnd)
200 #define	INSERT(op, pos)	doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
201 #define	AHEAD(pos)		dofwd(p, pos, HERE()-(pos))
202 #define	ASTERN(sop, pos)	EMIT(sop, HERE()-pos)
203 #define	HERE()		(p->slen)
204 #define	THERE()		(p->slen - 1)
205 #define	THERETHERE()	(p->slen - 2)
206 #define	DROP(n)	(p->slen -= (n))
207 
208 #ifndef NDEBUG
209 static int never = 0;		/* for use in asserts; shuts lint up */
210 #else
211 #define	never	0		/* some <assert.h>s have bugs too */
212 #endif
213 
214 #define	MEMLIMIT	0x8000000
215 #define MEMSIZE(p) \
216 	((p)->ncsalloc / CHAR_BIT * (p)->g->csetsize + \
217 	(p)->ncsalloc * sizeof(cset) + \
218 	(p)->ssize * sizeof(sop))
219 #define	RECLIMIT	256
220 
221 /*
222  - regcomp - interface for parser and compilation
223  = extern int regcomp(regex_t *, const char *, int);
224  = #define	REG_BASIC	0000
225  = #define	REG_EXTENDED	0001
226  = #define	REG_ICASE	0002
227  = #define	REG_NOSUB	0004
228  = #define	REG_NEWLINE	0010
229  = #define	REG_NOSPEC	0020
230  = #define	REG_PEND	0040
231  = #define	REG_DUMP	0200
232  */
233 int				/* 0 success, otherwise REG_something */
regcomp(regex_t * preg,const char * pattern,int cflags)234 regcomp(
235     regex_t *preg,
236     const char *pattern,
237     int cflags)
238 {
239 	struct parse pa;
240 	struct re_guts *g;
241 	struct parse *p = &pa;
242 	int i;
243 	size_t len;
244 #ifdef REDEBUG
245 #	define	GOODFLAGS(f)	(f)
246 #else
247 #	define	GOODFLAGS(f)	((f)&~REG_DUMP)
248 #endif
249 
250 	_DIAGASSERT(preg != NULL);
251 	_DIAGASSERT(pattern != NULL);
252 
253 	cflags = GOODFLAGS(cflags);
254 	if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
255 		return(REG_INVARG);
256 
257 	if (cflags&REG_PEND) {
258 		if (preg->re_endp < pattern)
259 			return(REG_INVARG);
260 		len = preg->re_endp - pattern;
261 	} else
262 		len = strlen(pattern);
263 
264 	/* do the mallocs early so failure handling is easy */
265 	g = malloc(sizeof(struct re_guts) + (NC - 1) * sizeof(cat_t));
266 	if (g == NULL)
267 		return(REG_ESPACE);
268 	p->ssize = len/(size_t)2*(size_t)3 + (size_t)1;	/* ugh */
269 	p->strip = calloc(p->ssize, sizeof(sop));
270 	p->slen = 0;
271 	if (p->strip == NULL) {
272 		free(g);
273 		return(REG_ESPACE);
274 	}
275 
276 	/* set things up */
277 	p->g = g;
278 	p->next = pattern;
279 	p->end = p->next + len;
280 	p->error = 0;
281 	p->ncsalloc = 0;
282 	for (i = 0; i < NPAREN; i++) {
283 		p->pbegin[i] = 0;
284 		p->pend[i] = 0;
285 	}
286 	g->csetsize = NC;
287 	g->sets = NULL;
288 	g->setbits = NULL;
289 	g->ncsets = 0;
290 	g->cflags = cflags;
291 	g->iflags = 0;
292 	g->nbol = 0;
293 	g->neol = 0;
294 	g->must = NULL;
295 	g->mlen = 0;
296 	g->nsub = 0;
297 	g->ncategories = 1;	/* category 0 is "everything else" */
298 	g->categories = &g->catspace[-(CHAR_MIN)];
299 	(void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
300 	g->backrefs = 0;
301 
302 	/* do it */
303 	EMIT(OEND, 0);
304 	g->firststate = THERE();
305 	if (cflags&REG_EXTENDED)
306 		p_ere(p, OUT, 0);
307 	else if (cflags&REG_NOSPEC)
308 		p_str(p);
309 	else
310 		p_bre(p, OUT, OUT, 0);
311 	EMIT(OEND, 0);
312 	g->laststate = THERE();
313 
314 	/* tidy up loose ends and fill things in */
315 	categorize(p, g);
316 	stripsnug(p, g);
317 	findmust(p, g);
318 	g->nplus = pluscount(p, g);
319 	g->magic = MAGIC2;
320 	preg->re_nsub = g->nsub;
321 	preg->re_g = g;
322 	preg->re_magic = MAGIC1;
323 #ifndef REDEBUG
324 	/* not debugging, so can't rely on the assert() in regexec() */
325 	if (g->iflags&BAD)
326 		SETERROR(REG_ASSERT);
327 #endif
328 
329 	/* win or lose, we're done */
330 	if (p->error != 0)	/* lose */
331 		regfree(preg);
332 	return(p->error);
333 }
334 
335 /*
336  - p_ere - ERE parser top level, concatenation and alternation
337  == static void p_ere(struct parse *p, int stop, size_t reclimit);
338  */
339 static void
p_ere(struct parse * p,int stop,size_t reclimit)340 p_ere(
341     struct parse *p,
342     int stop,			/* character this ERE should end at */
343     size_t reclimit)
344 {
345 	char c;
346 	sopno prevback = 0;	/* pacify gcc */
347 	sopno prevfwd = 0; 	/* pacify gcc */
348 	sopno conc;
349 	int first = 1;		/* is this the first alternative? */
350 
351 	_DIAGASSERT(p != NULL);
352 
353 	if (reclimit++ > RECLIMIT || p->error == REG_ESPACE) {
354 		p->error = REG_ESPACE;
355 		return;
356 	}
357 
358 	for (;;) {
359 		/* do a bunch of concatenated expressions */
360 		conc = HERE();
361 		while (MORE() && (c = PEEK()) != '|' && c != stop)
362 			p_ere_exp(p, reclimit);
363 		REQUIRE(HERE() != conc, REG_EMPTY);	/* require nonempty */
364 
365 		if (!EAT('|'))
366 			break;		/* NOTE BREAK OUT */
367 
368 		if (first) {
369 			INSERT(OCH_, conc);	/* offset is wrong */
370 			prevfwd = conc;
371 			prevback = conc;
372 			first = 0;
373 		}
374 		ASTERN(OOR1, prevback);
375 		prevback = THERE();
376 		AHEAD(prevfwd);			/* fix previous offset */
377 		prevfwd = HERE();
378 		EMIT(OOR2, 0);			/* offset is very wrong */
379 	}
380 
381 	if (!first) {		/* tail-end fixups */
382 		AHEAD(prevfwd);
383 		ASTERN(O_CH, prevback);
384 	}
385 
386 	assert(!MORE() || SEE(stop));
387 }
388 
389 /*
390  - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
391  == static void p_ere_exp(struct parse *p, size_t reclimit);
392  */
393 static void
p_ere_exp(struct parse * p,size_t reclimit)394 p_ere_exp(
395     struct parse *p,
396     size_t reclimit)
397 {
398 	char c;
399 	sopno pos;
400 	int count;
401 	int count2;
402 	sopno subno;
403 	int wascaret = 0;
404 
405 	_DIAGASSERT(p != NULL);
406 
407 	assert(MORE());		/* caller should have ensured this */
408 	c = GETNEXT();
409 
410 	pos = HERE();
411 	switch (c) {
412 	case '(':
413 		REQUIRE(MORE(), REG_EPAREN);
414 		p->g->nsub++;
415 		subno = p->g->nsub;
416 		if (subno < NPAREN)
417 			p->pbegin[subno] = HERE();
418 		EMIT(OLPAREN, subno);
419 		if (!SEE(')'))
420 			p_ere(p, ')', reclimit);
421 		if (subno < NPAREN) {
422 			p->pend[subno] = HERE();
423 			assert(p->pend[subno] != 0);
424 		}
425 		EMIT(ORPAREN, subno);
426 		MUSTEAT(')', REG_EPAREN);
427 		break;
428 #ifndef POSIX_MISTAKE
429 	case ')':		/* happens only if no current unmatched ( */
430 		/*
431 		 * You may ask, why the ifndef?  Because I didn't notice
432 		 * this until slightly too late for 1003.2, and none of the
433 		 * other 1003.2 regular-expression reviewers noticed it at
434 		 * all.  So an unmatched ) is legal POSIX, at least until
435 		 * we can get it fixed.
436 		 */
437 		SETERROR(REG_EPAREN);
438 		break;
439 #endif
440 	case '^':
441 		EMIT(OBOL, 0);
442 		p->g->iflags |= USEBOL;
443 		p->g->nbol++;
444 		wascaret = 1;
445 		break;
446 	case '$':
447 		EMIT(OEOL, 0);
448 		p->g->iflags |= USEEOL;
449 		p->g->neol++;
450 		break;
451 	case '|':
452 		SETERROR(REG_EMPTY);
453 		break;
454 	case '*':
455 	case '+':
456 	case '?':
457 		SETERROR(REG_BADRPT);
458 		break;
459 	case '.':
460 		if (p->g->cflags&REG_NEWLINE)
461 			nonnewline(p);
462 		else
463 			EMIT(OANY, 0);
464 		break;
465 	case '[':
466 		p_bracket(p);
467 		break;
468 	case '\\':
469 		REQUIRE(MORE(), REG_EESCAPE);
470 		c = GETNEXT();
471 		ordinary(p, c);
472 		break;
473 	case '{':		/* okay as ordinary except if digit follows */
474 		REQUIRE(!MORE() || !isdigit((unsigned char)PEEK()), REG_BADRPT);
475 		/* FALLTHROUGH */
476 	default:
477 		ordinary(p, c);
478 		break;
479 	}
480 
481 	if (!MORE())
482 		return;
483 	c = PEEK();
484 	/* we call { a repetition if followed by a digit */
485 	if (!( c == '*' || c == '+' || c == '?' ||
486 	    (c == '{' && MORE2() && isdigit((unsigned char)PEEK2())) ))
487 		return;		/* no repetition, we're done */
488 	NEXT();
489 
490 	REQUIRE(!wascaret, REG_BADRPT);
491 	switch (c) {
492 	case '*':	/* implemented as +? */
493 		/* this case does not require the (y|) trick, noKLUDGE */
494 		INSERT(OPLUS_, pos);
495 		ASTERN(O_PLUS, pos);
496 		INSERT(OQUEST_, pos);
497 		ASTERN(O_QUEST, pos);
498 		break;
499 	case '+':
500 		INSERT(OPLUS_, pos);
501 		ASTERN(O_PLUS, pos);
502 		break;
503 	case '?':
504 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
505 		INSERT(OCH_, pos);		/* offset slightly wrong */
506 		ASTERN(OOR1, pos);		/* this one's right */
507 		AHEAD(pos);			/* fix the OCH_ */
508 		EMIT(OOR2, 0);			/* offset very wrong... */
509 		AHEAD(THERE());			/* ...so fix it */
510 		ASTERN(O_CH, THERETHERE());
511 		break;
512 	case '{':
513 		count = p_count(p);
514 		if (EAT(',')) {
515 			if (isdigit((unsigned char)PEEK())) {
516 				count2 = p_count(p);
517 				REQUIRE(count <= count2, REG_BADBR);
518 			} else		/* single number with comma */
519 				count2 = INFINITY;
520 		} else		/* just a single number */
521 			count2 = count;
522 		repeat(p, pos, count, count2, 0);
523 		if (!EAT('}')) {	/* error heuristics */
524 			while (MORE() && PEEK() != '}')
525 				NEXT();
526 			REQUIRE(MORE(), REG_EBRACE);
527 			SETERROR(REG_BADBR);
528 		}
529 		break;
530 	}
531 
532 	if (!MORE())
533 		return;
534 	c = PEEK();
535 	if (!( c == '*' || c == '+' || c == '?' ||
536 	    (c == '{' && MORE2() && isdigit((unsigned char)PEEK2())) ) )
537 		return;
538 	SETERROR(REG_BADRPT);
539 }
540 
541 /*
542  - p_str - string (no metacharacters) "parser"
543  == static void p_str(struct parse *p);
544  */
545 static void
p_str(struct parse * p)546 p_str(
547     struct parse *p)
548 {
549 
550 	_DIAGASSERT(p != NULL);
551 
552 	REQUIRE(MORE(), REG_EMPTY);
553 	while (MORE())
554 		ordinary(p, GETNEXT());
555 }
556 
557 /*
558  - p_bre - BRE parser top level, anchoring and concatenation
559  == static void p_bre(struct parse *p, int end1, \
560  ==	int end2, size_t reclimit);
561  * Giving end1 as OUT essentially eliminates the end1/end2 check.
562  *
563  * This implementation is a bit of a kludge, in that a trailing $ is first
564  * taken as an ordinary character and then revised to be an anchor.  The
565  * only undesirable side effect is that '$' gets included as a character
566  * category in such cases.  This is fairly harmless; not worth fixing.
567  * The amount of lookahead needed to avoid this kludge is excessive.
568  */
569 static void
p_bre(struct parse * p,int end1,int end2,size_t reclimit)570 p_bre(
571     struct parse *p,
572     int end1,		/* first terminating character */
573     int end2,		/* second terminating character */
574     size_t reclimit)
575 {
576 	sopno start;
577 	int first = 1;			/* first subexpression? */
578 	int wasdollar = 0;
579 
580 	_DIAGASSERT(p != NULL);
581 
582 	if (reclimit++ > RECLIMIT || p->error == REG_ESPACE) {
583 		p->error = REG_ESPACE;
584 		return;
585 	}
586 
587 	start = HERE();
588 
589 	if (EAT('^')) {
590 		EMIT(OBOL, 0);
591 		p->g->iflags |= USEBOL;
592 		p->g->nbol++;
593 	}
594 	while (MORE() && !SEETWO(end1, end2)) {
595 		wasdollar = p_simp_re(p, first, reclimit);
596 		first = 0;
597 	}
598 	if (wasdollar) {	/* oops, that was a trailing anchor */
599 		DROP(1);
600 		EMIT(OEOL, 0);
601 		p->g->iflags |= USEEOL;
602 		p->g->neol++;
603 	}
604 
605 	REQUIRE(HERE() != start, REG_EMPTY);	/* require nonempty */
606 }
607 
608 /*
609  - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
610  == static int p_simp_re(struct parse *p, int starordinary, size_t reclimit);
611  */
612 static int			/* was the simple RE an unbackslashed $? */
p_simp_re(struct parse * p,int starordinary,size_t reclimit)613 p_simp_re(
614     struct parse *p,
615     int starordinary,		/* is a leading * an ordinary character? */
616     size_t reclimit)
617 {
618 	int c;
619 	int count;
620 	int count2;
621 	sopno pos, i;
622 	sopno subno;
623 #	define	BACKSL	(1<<CHAR_BIT)
624 
625 	_DIAGASSERT(p != NULL);
626 
627 	pos = HERE();		/* repetion op, if any, covers from here */
628 
629 	assert(MORE());		/* caller should have ensured this */
630 	c = GETNEXT();
631 	if (c == '\\') {
632 		REQUIRE(MORE(), REG_EESCAPE);
633 		c = BACKSL | (unsigned char)GETNEXT();
634 	}
635 	switch (c) {
636 	case '.':
637 		if (p->g->cflags&REG_NEWLINE)
638 			nonnewline(p);
639 		else
640 			EMIT(OANY, 0);
641 		break;
642 	case '[':
643 		p_bracket(p);
644 		break;
645 	case BACKSL|'{':
646 		SETERROR(REG_BADRPT);
647 		break;
648 	case BACKSL|'(':
649 		p->g->nsub++;
650 		subno = p->g->nsub;
651 		if (subno < NPAREN)
652 			p->pbegin[subno] = HERE();
653 		EMIT(OLPAREN, subno);
654 		/* the MORE here is an error heuristic */
655 		if (MORE() && !SEETWO('\\', ')'))
656 			p_bre(p, '\\', ')', reclimit);
657 		if (subno < NPAREN) {
658 			p->pend[subno] = HERE();
659 			assert(p->pend[subno] != 0);
660 		}
661 		EMIT(ORPAREN, subno);
662 		REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
663 		break;
664 	case BACKSL|')':	/* should not get here -- must be user */
665 	case BACKSL|'}':
666 		SETERROR(REG_EPAREN);
667 		break;
668 	case BACKSL|'1':
669 	case BACKSL|'2':
670 	case BACKSL|'3':
671 	case BACKSL|'4':
672 	case BACKSL|'5':
673 	case BACKSL|'6':
674 	case BACKSL|'7':
675 	case BACKSL|'8':
676 	case BACKSL|'9':
677 		i = (c&~BACKSL) - '0';
678 		assert(i < NPAREN);
679 		if (p->pend[i] != 0) {
680 			assert(i <= p->g->nsub);
681 			EMIT(OBACK_, i);
682 			assert(p->pbegin[i] != 0);
683 			assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
684 			assert(OP(p->strip[p->pend[i]]) == ORPAREN);
685 			(void) dupl(p, p->pbegin[i]+1, p->pend[i]);
686 			EMIT(O_BACK, i);
687 		} else
688 			SETERROR(REG_ESUBREG);
689 		p->g->backrefs = 1;
690 		break;
691 	case '*':
692 		REQUIRE(starordinary, REG_BADRPT);
693 		/* FALLTHROUGH */
694 	default:
695 		ordinary(p, c &~ BACKSL);
696 		break;
697 	}
698 
699 	if (EAT('*')) {		/* implemented as +? */
700 		/* this case does not require the (y|) trick, noKLUDGE */
701 		INSERT(OPLUS_, pos);
702 		ASTERN(O_PLUS, pos);
703 		INSERT(OQUEST_, pos);
704 		ASTERN(O_QUEST, pos);
705 	} else if (EATTWO('\\', '{')) {
706 		count = p_count(p);
707 		if (EAT(',')) {
708 			if (MORE() && isdigit((unsigned char)PEEK())) {
709 				count2 = p_count(p);
710 				REQUIRE(count <= count2, REG_BADBR);
711 			} else		/* single number with comma */
712 				count2 = INFINITY;
713 		} else		/* just a single number */
714 			count2 = count;
715 		repeat(p, pos, count, count2, 0);
716 		if (!EATTWO('\\', '}')) {	/* error heuristics */
717 			while (MORE() && !SEETWO('\\', '}'))
718 				NEXT();
719 			REQUIRE(MORE(), REG_EBRACE);
720 			SETERROR(REG_BADBR);
721 		}
722 	} else if (c == (unsigned char)'$')	/* $ (but not \$) ends it */
723 		return(1);
724 
725 	return(0);
726 }
727 
728 /*
729  - p_count - parse a repetition count
730  == static int p_count(struct parse *p);
731  */
732 static int			/* the value */
p_count(struct parse * p)733 p_count(
734     struct parse *p)
735 {
736 	int count = 0;
737 	int ndigits = 0;
738 
739 	_DIAGASSERT(p != NULL);
740 
741 	while (MORE() && isdigit((unsigned char)PEEK()) && count <= DUPMAX) {
742 		count = count*10 + (GETNEXT() - '0');
743 		ndigits++;
744 	}
745 
746 	REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
747 	return(count);
748 }
749 
750 /*
751  - p_bracket - parse a bracketed character list
752  == static void p_bracket(struct parse *p);
753  *
754  * Note a significant property of this code:  if the allocset() did SETERROR,
755  * no set operations are done.
756  */
757 static void
p_bracket(struct parse * p)758 p_bracket(
759     struct parse *p)
760 {
761 	cset *cs;
762 	int invert = 0;
763 	_DIAGASSERT(p != NULL);
764 
765 	cs = allocset(p);
766 	if (cs == NULL)
767 		return;
768 
769 	/* Dept of Truly Sickening Special-Case Kludges */
770 	if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]",
771 					    (size_t)6) == 0) {
772 		EMIT(OBOW, 0);
773 		NEXTn(6);
774 		return;
775 	}
776 	if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]",
777 					    (size_t)6) == 0) {
778 		EMIT(OEOW, 0);
779 		NEXTn(6);
780 		return;
781 	}
782 
783 	if (EAT('^'))
784 		invert++;	/* make note to invert set at end */
785 	if (EAT(']'))
786 		CHadd(cs, ']');
787 	else if (EAT('-'))
788 		CHadd(cs, '-');
789 	while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
790 		p_b_term(p, cs);
791 	if (EAT('-'))
792 		CHadd(cs, '-');
793 	MUSTEAT(']', REG_EBRACK);
794 
795 	if (p->error != 0)	/* don't mess things up further */
796 		return;
797 
798 	if (p->g->cflags&REG_ICASE) {
799 		ssize_t i;
800 		int ci;
801 
802 		for (i = p->g->csetsize - 1; i >= 0; i--)
803 			if (CHIN(cs, i) && isalpha(i)) {
804 				ci = othercase((int)i);
805 				if (ci != i)
806 					CHadd(cs, ci);
807 			}
808 		if (cs->multis != NULL)
809 			mccase(p, cs);
810 	}
811 	if (invert) {
812 		ssize_t i;
813 
814 		for (i = p->g->csetsize - 1; i >= 0; i--)
815 			if (CHIN(cs, i))
816 				CHsub(cs, (int)i);
817 			else
818 				CHadd(cs, (int)i);
819 		if (p->g->cflags&REG_NEWLINE)
820 			CHsub(cs, '\n');
821 		if (cs->multis != NULL)
822 			mcinvert(p, cs);
823 	}
824 
825 	assert(cs->multis == NULL);		/* xxx */
826 
827 	if (nch(p, cs) == 1) {		/* optimize singleton sets */
828 		ordinary(p, firstch(p, cs));
829 		freeset(p, cs);
830 	} else
831 		EMIT(OANYOF, freezeset(p, cs));
832 }
833 
834 /*
835  - p_b_term - parse one term of a bracketed character list
836  == static void p_b_term(struct parse *p, cset *cs);
837  */
838 static void
p_b_term(struct parse * p,cset * cs)839 p_b_term(
840     struct parse *p,
841     cset *cs)
842 {
843 	char c;
844 	char start, finish;
845 	int i;
846 
847 	_DIAGASSERT(p != NULL);
848 	_DIAGASSERT(cs != NULL);
849 
850 	/* classify what we've got */
851 	switch ((MORE()) ? PEEK() : '\0') {
852 	case '[':
853 		c = (MORE2()) ? PEEK2() : '\0';
854 		break;
855 
856 	case '-':
857 		SETERROR(REG_ERANGE);
858 		return;			/* NOTE RETURN */
859 
860 	default:
861 		c = '\0';
862 		break;
863 	}
864 
865 	switch (c) {
866 	case ':':		/* character class */
867 		NEXT2();
868 		REQUIRE(MORE(), REG_EBRACK);
869 		c = PEEK();
870 		REQUIRE(c != '-' && c != ']', REG_ECTYPE);
871 		p_b_cclass(p, cs);
872 		REQUIRE(MORE(), REG_EBRACK);
873 		REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
874 		break;
875 	case '=':		/* equivalence class */
876 		NEXT2();
877 		REQUIRE(MORE(), REG_EBRACK);
878 		c = PEEK();
879 		REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
880 		p_b_eclass(p, cs);
881 		REQUIRE(MORE(), REG_EBRACK);
882 		REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
883 		break;
884 	default:		/* symbol, ordinary character, or range */
885 /* xxx revision needed for multichar stuff */
886 		start = p_b_symbol(p);
887 		if (SEE('-') && MORE2() && PEEK2() != ']') {
888 			/* range */
889 			NEXT();
890 			if (EAT('-'))
891 				finish = '-';
892 			else
893 				finish = p_b_symbol(p);
894 		} else
895 			finish = start;
896 /* xxx what about signed chars here... */
897 		REQUIRE(start <= finish, REG_ERANGE);
898 		for (i = start; i <= finish; i++)
899 			CHadd(cs, i);
900 		break;
901 	}
902 }
903 
904 /*
905  - p_b_cclass - parse a character-class name and deal with it
906  == static void p_b_cclass(struct parse *p, cset *cs);
907  */
908 static void
p_b_cclass(struct parse * p,cset * cs)909 p_b_cclass(
910     struct parse *p,
911     cset *cs)
912 {
913 	const char *sp;
914 	const struct cclass *cp;
915 	size_t len;
916 	const char *u;
917 	char c;
918 
919 	_DIAGASSERT(p != NULL);
920 	_DIAGASSERT(cs != NULL);
921 
922 	sp = p->next;
923 
924 	while (MORE() && isalpha((unsigned char)PEEK()))
925 		NEXT();
926 	len = p->next - sp;
927 	for (cp = cclasses; cp->name != NULL; cp++)
928 		if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
929 			break;
930 	if (cp->name == NULL) {
931 		/* oops, didn't find it */
932 		SETERROR(REG_ECTYPE);
933 		return;
934 	}
935 
936 	u = cp->chars;
937 	while ((c = *u++) != '\0')
938 		CHadd(cs, c);
939 	for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
940 		MCadd(p, cs, u);
941 }
942 
943 /*
944  - p_b_eclass - parse an equivalence-class name and deal with it
945  == static void p_b_eclass(struct parse *p, cset *cs);
946  *
947  * This implementation is incomplete. xxx
948  */
949 static void
p_b_eclass(struct parse * p,cset * cs)950 p_b_eclass(
951     struct parse *p,
952     cset *cs)
953 {
954 	char c;
955 
956 	_DIAGASSERT(p != NULL);
957 	_DIAGASSERT(cs != NULL);
958 
959 	c = p_b_coll_elem(p, '=');
960 	CHadd(cs, c);
961 }
962 
963 /*
964  - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
965  == static char p_b_symbol(struct parse *p);
966  */
967 static char			/* value of symbol */
p_b_symbol(struct parse * p)968 p_b_symbol(
969     struct parse *p)
970 {
971 	char value;
972 
973 	_DIAGASSERT(p != NULL);
974 
975 	REQUIRE(MORE(), REG_EBRACK);
976 	if (!EATTWO('[', '.'))
977 		return(GETNEXT());
978 
979 	/* collating symbol */
980 	value = p_b_coll_elem(p, '.');
981 	REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
982 	return(value);
983 }
984 
985 /*
986  - p_b_coll_elem - parse a collating-element name and look it up
987  == static char p_b_coll_elem(struct parse *p, int endc);
988  */
989 static char			/* value of collating element */
p_b_coll_elem(struct parse * p,int endc)990 p_b_coll_elem(
991     struct parse *p,
992     int endc)			/* name ended by endc,']' */
993 {
994 	const char *sp;
995 	const struct cname *cp;
996 	size_t len;
997 
998 	_DIAGASSERT(p != NULL);
999 
1000 	sp = p->next;
1001 
1002 	while (MORE() && !SEETWO(endc, ']'))
1003 		NEXT();
1004 	if (!MORE()) {
1005 		SETERROR(REG_EBRACK);
1006 		return(0);
1007 	}
1008 	len = p->next - sp;
1009 	for (cp = cnames; cp->name != NULL; cp++)
1010 		if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
1011 			return(cp->code);	/* known name */
1012 	if (len == 1)
1013 		return(*sp);	/* single character */
1014 	SETERROR(REG_ECOLLATE);			/* neither */
1015 	return(0);
1016 }
1017 
1018 /*
1019  - othercase - return the case counterpart of an alphabetic
1020  == static int othercase(int ch);
1021  */
1022 static int			/* if no counterpart, return ch */
othercase(int ch)1023 othercase(
1024     int ch)
1025 {
1026 	assert(isalpha(ch));
1027 	if (isupper(ch))
1028 		return(tolower(ch));
1029 	else if (islower(ch))
1030 		return(toupper(ch));
1031 	else			/* peculiar, but could happen */
1032 		return(ch);
1033 }
1034 
1035 /*
1036  - bothcases - emit a dualcase version of a two-case character
1037  == static void bothcases(struct parse *p, int ch);
1038  *
1039  * Boy, is this implementation ever a kludge...
1040  */
1041 static void
bothcases(struct parse * p,int ch)1042 bothcases(
1043     struct parse *p,
1044     int ch)
1045 {
1046 	const char *oldnext;
1047 	const char *oldend;
1048 	char bracket[3];
1049 
1050 	_DIAGASSERT(p != NULL);
1051 
1052 	oldnext = p->next;
1053 	oldend = p->end;
1054 
1055 	assert(othercase(ch) != ch);	/* p_bracket() would recurse */
1056 	p->next = bracket;
1057 	p->end = bracket+2;
1058 	bracket[0] = ch;
1059 	bracket[1] = ']';
1060 	bracket[2] = '\0';
1061 	p_bracket(p);
1062 	assert(p->next == bracket+2);
1063 	p->next = oldnext;
1064 	p->end = oldend;
1065 }
1066 
1067 /*
1068  - ordinary - emit an ordinary character
1069  == static void ordinary(struct parse *p, int ch);
1070  */
1071 static void
ordinary(struct parse * p,int ch)1072 ordinary(
1073     struct parse *p,
1074     int ch)
1075 {
1076 	cat_t *cap;
1077 	unsigned char uc = (unsigned char)ch;
1078 
1079 	_DIAGASSERT(p != NULL);
1080 
1081 	cap = p->g->categories;
1082 	if ((p->g->cflags & REG_ICASE) && isalpha(uc) && othercase(uc) != uc)
1083 		bothcases(p, uc);
1084 	else {
1085 		EMIT(OCHAR, (sopno)uc);
1086 		if (cap[uc] == 0) {
1087 			_DIAGASSERT(__type_fit(unsigned char,
1088 			    p->g->ncategories + 1));
1089 			cap[uc] = (unsigned char)p->g->ncategories++;
1090 		}
1091 	}
1092 }
1093 
1094 /*
1095  - nonnewline - emit REG_NEWLINE version of OANY
1096  == static void nonnewline(struct parse *p);
1097  *
1098  * Boy, is this implementation ever a kludge...
1099  */
1100 static void
nonnewline(struct parse * p)1101 nonnewline(
1102     struct parse *p)
1103 {
1104 	const char *oldnext;
1105 	const char *oldend;
1106 	char bracket[4];
1107 
1108 	_DIAGASSERT(p != NULL);
1109 
1110 	oldnext = p->next;
1111 	oldend = p->end;
1112 
1113 	p->next = bracket;
1114 	p->end = bracket+3;
1115 	bracket[0] = '^';
1116 	bracket[1] = '\n';
1117 	bracket[2] = ']';
1118 	bracket[3] = '\0';
1119 	p_bracket(p);
1120 	assert(p->next == bracket+3);
1121 	p->next = oldnext;
1122 	p->end = oldend;
1123 }
1124 
1125 /*
1126  - repeat - generate code for a bounded repetition, recursively if needed
1127  == static void repeat(struct parse *p, sopno start, int from, int to,
1128  == size_t reclimit);
1129  */
1130 static void
repeat(struct parse * p,sopno start,int from,int to,size_t reclimit)1131 repeat(
1132     struct parse *p,
1133     sopno start,		/* operand from here to end of strip */
1134     int from,			/* repeated from this number */
1135     int to,			/* to this number of times (maybe INFINITY) */
1136     size_t reclimit)
1137 {
1138 	sopno finish;
1139 #	define	N	2
1140 #	define	INF	3
1141 #	define	REP(f, t)	((f)*8 + (t))
1142 #	define	MAP(n)	(((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1143 	sopno copy;
1144 
1145 	_DIAGASSERT(p != NULL);
1146 
1147 	if (reclimit++ > RECLIMIT)
1148 		p->error = REG_ESPACE;
1149 	if (p->error)
1150 		return;
1151 
1152 	finish = HERE();
1153 
1154 	assert(from <= to);
1155 
1156 	switch (REP(MAP(from), MAP(to))) {
1157 	case REP(0, 0):			/* must be user doing this */
1158 		DROP(finish-start);	/* drop the operand */
1159 		break;
1160 	case REP(0, 1):			/* as x{1,1}? */
1161 	case REP(0, N):			/* as x{1,n}? */
1162 	case REP(0, INF):		/* as x{1,}? */
1163 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1164 		INSERT(OCH_, start);		/* offset is wrong... */
1165 		repeat(p, start+1, 1, to, reclimit);
1166 		ASTERN(OOR1, start);
1167 		AHEAD(start);			/* ... fix it */
1168 		EMIT(OOR2, 0);
1169 		AHEAD(THERE());
1170 		ASTERN(O_CH, THERETHERE());
1171 		break;
1172 	case REP(1, 1):			/* trivial case */
1173 		/* done */
1174 		break;
1175 	case REP(1, N):			/* as x?x{1,n-1} */
1176 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1177 		INSERT(OCH_, start);
1178 		ASTERN(OOR1, start);
1179 		AHEAD(start);
1180 		EMIT(OOR2, 0);			/* offset very wrong... */
1181 		AHEAD(THERE());			/* ...so fix it */
1182 		ASTERN(O_CH, THERETHERE());
1183 		copy = dupl(p, start+1, finish+1);
1184 		assert(copy == finish+4);
1185 		repeat(p, copy, 1, to-1, reclimit);
1186 		break;
1187 	case REP(1, INF):		/* as x+ */
1188 		INSERT(OPLUS_, start);
1189 		ASTERN(O_PLUS, start);
1190 		break;
1191 	case REP(N, N):			/* as xx{m-1,n-1} */
1192 		copy = dupl(p, start, finish);
1193 		repeat(p, copy, from-1, to-1, reclimit);
1194 		break;
1195 	case REP(N, INF):		/* as xx{n-1,INF} */
1196 		copy = dupl(p, start, finish);
1197 		repeat(p, copy, from-1, to, reclimit);
1198 		break;
1199 	default:			/* "can't happen" */
1200 		SETERROR(REG_ASSERT);	/* just in case */
1201 		break;
1202 	}
1203 }
1204 
1205 /*
1206  - seterr - set an error condition
1207  == static int seterr(struct parse *p, int e);
1208  */
1209 static int			/* useless but makes type checking happy */
seterr(struct parse * p,int e)1210 seterr(
1211     struct parse *p,
1212     int e)
1213 {
1214 
1215 	_DIAGASSERT(p != NULL);
1216 
1217 	if (p->error == 0)	/* keep earliest error condition */
1218 		p->error = e;
1219 	p->next = nuls;		/* try to bring things to a halt */
1220 	p->end = nuls;
1221 	return(0);		/* make the return value well-defined */
1222 }
1223 
1224 /*
1225  - allocset - allocate a set of characters for []
1226  == static cset *allocset(struct parse *p);
1227  */
1228 static cset *
allocset(struct parse * p)1229 allocset(
1230     struct parse *p)
1231 {
1232 	size_t no;
1233 	size_t nc;
1234 	size_t nbytes;
1235 	cset *cs;
1236 	size_t css;
1237 	size_t i;
1238 	void *old_ptr;
1239 
1240 	_DIAGASSERT(p != NULL);
1241 
1242 	no = p->g->ncsets++;
1243 	css = (size_t)p->g->csetsize;
1244 	if (no >= p->ncsalloc) {	/* need another column of space */
1245 		p->ncsalloc += CHAR_BIT;
1246 		nc = p->ncsalloc;
1247 		assert(nc % CHAR_BIT == 0);
1248 		nbytes = nc / CHAR_BIT * css;
1249 		if (MEMSIZE(p) > MEMLIMIT)
1250 			goto oomem;
1251 		if (reallocarr(&p->g->sets, nc, sizeof(cset)))
1252 			goto oomem;
1253 		old_ptr = p->g->setbits;
1254 		if (reallocarr(&p->g->setbits, nc / CHAR_BIT, css)) {
1255 			free(old_ptr);
1256 			goto oomem;
1257 		}
1258 		if (old_ptr != p->g->setbits) {
1259 			for (i = 0; i < no; i++)
1260 				p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
1261 		}
1262 		(void) memset((char *)p->g->setbits + (nbytes - css), 0, css);
1263 	}
1264 
1265 	cs = &p->g->sets[no];
1266 	cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
1267 	cs->mask = 1 << (unsigned int)((no) % CHAR_BIT);
1268 	cs->hash = 0;
1269 	cs->smultis = 0;
1270 	cs->multis = NULL;
1271 
1272 	return(cs);
1273 
1274 oomem:
1275 	SETERROR(REG_ESPACE);
1276 	/* caller's responsibility not to do set ops */
1277 	return NULL;
1278 }
1279 
1280 /*
1281  - freeset - free a now-unused set
1282  == static void freeset(struct parse *p, cset *cs);
1283  */
1284 static void
freeset(struct parse * p,cset * cs)1285 freeset(
1286     struct parse *p,
1287     cset *cs)
1288 {
1289 	size_t i;
1290 	cset *top;
1291 	size_t css;
1292 
1293 	_DIAGASSERT(p != NULL);
1294 	_DIAGASSERT(cs != NULL);
1295 
1296 	top = &p->g->sets[p->g->ncsets];
1297 	css = (size_t)p->g->csetsize;
1298 
1299 	for (i = 0; i < css; i++)
1300 		CHsub(cs, (int)i);
1301 	if (cs == top-1)	/* recover only the easy case */
1302 		p->g->ncsets--;
1303 }
1304 
1305 /*
1306  - freezeset - final processing on a set of characters
1307  == static int freezeset(struct parse *p, cset *cs);
1308  *
1309  * The main task here is merging identical sets.  This is usually a waste
1310  * of time (although the hash code minimizes the overhead), but can win
1311  * big if REG_ICASE is being used.  REG_ICASE, by the way, is why the hash
1312  * is done using addition rather than xor -- all ASCII [aA] sets xor to
1313  * the same value!
1314  */
1315 static sopno			/* set number */
freezeset(struct parse * p,cset * cs)1316 freezeset(
1317     struct parse *p,
1318     cset *cs)
1319 {
1320 	uch h;
1321 	size_t i;
1322 	cset *top;
1323 	cset *cs2;
1324 	size_t css;
1325 
1326 	_DIAGASSERT(p != NULL);
1327 	_DIAGASSERT(cs != NULL);
1328 
1329 	h = cs->hash;
1330 	top = &p->g->sets[p->g->ncsets];
1331 	css = (size_t)p->g->csetsize;
1332 
1333 	/* look for an earlier one which is the same */
1334 	for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
1335 		if (cs2->hash == h && cs2 != cs) {
1336 			/* maybe */
1337 			for (i = 0; i < css; i++)
1338 				if (!!CHIN(cs2, i) != !!CHIN(cs, i))
1339 					break;		/* no */
1340 			if (i == css)
1341 				break;			/* yes */
1342 		}
1343 
1344 	if (cs2 < top) {	/* found one */
1345 		freeset(p, cs);
1346 		cs = cs2;
1347 	}
1348 
1349 	return (sopno)(cs - p->g->sets);
1350 }
1351 
1352 /*
1353  - firstch - return first character in a set (which must have at least one)
1354  == static int firstch(struct parse *p, cset *cs);
1355  */
1356 static int			/* character; there is no "none" value */
firstch(struct parse * p,cset * cs)1357 firstch(
1358     struct parse *p,
1359     cset *cs)
1360 {
1361 	size_t i;
1362 	size_t css;
1363 
1364 	_DIAGASSERT(p != NULL);
1365 	_DIAGASSERT(cs != NULL);
1366 
1367 	css = (size_t)p->g->csetsize;
1368 
1369 	for (i = 0; i < css; i++)
1370 		if (CHIN(cs, i))
1371 			return((char)i);
1372 	assert(never);
1373 	return(0);		/* arbitrary */
1374 }
1375 
1376 /*
1377  - nch - number of characters in a set
1378  == static int nch(struct parse *p, cset *cs);
1379  */
1380 static int
nch(struct parse * p,cset * cs)1381 nch(
1382     struct parse *p,
1383     cset *cs)
1384 {
1385 	size_t i;
1386 	size_t css;
1387 	int n = 0;
1388 
1389 	_DIAGASSERT(p != NULL);
1390 	_DIAGASSERT(cs != NULL);
1391 
1392 	css = (size_t)p->g->csetsize;
1393 
1394 	for (i = 0; i < css; i++)
1395 		if (CHIN(cs, i))
1396 			n++;
1397 	return(n);
1398 }
1399 
1400 /*
1401  - mcadd - add a collating element to a cset
1402  == static void mcadd(struct parse *p, cset *cs, \
1403  ==	char *cp);
1404  */
1405 static void
mcadd(struct parse * p,cset * cs,const char * cp)1406 mcadd(
1407     struct parse *p,
1408     cset *cs,
1409     const char *cp)
1410 {
1411 	size_t oldend;
1412 
1413 	_DIAGASSERT(p != NULL);
1414 	_DIAGASSERT(cs != NULL);
1415 	_DIAGASSERT(cp != NULL);
1416 
1417 	oldend = cs->smultis;
1418 
1419 	cs->smultis += strlen(cp) + 1;
1420 	if (cs->multis == NULL)
1421 		cs->multis = malloc(cs->smultis);
1422 	else
1423 		cs->multis = realloc(cs->multis, cs->smultis);
1424 	if (cs->multis == NULL) {
1425 		SETERROR(REG_ESPACE);
1426 		return;
1427 	}
1428 
1429 	(void) strcpy(cs->multis + oldend - 1, cp);
1430 	cs->multis[cs->smultis - 1] = '\0';
1431 }
1432 
1433 #if 0
1434 /*
1435  - mcsub - subtract a collating element from a cset
1436  == static void mcsub(cset *cs, char *cp);
1437  */
1438 static void
1439 mcsub(
1440     cset *cs,
1441     char *cp)
1442 {
1443 	char *fp;
1444 	size_t len;
1445 
1446 	_DIAGASSERT(cs != NULL);
1447 	_DIAGASSERT(cp != NULL);
1448 
1449 	fp = mcfind(cs, cp);
1450 	len = strlen(fp);
1451 
1452 	assert(fp != NULL);
1453 	(void) memmove(fp, fp + len + 1,
1454 				cs->smultis - (fp + len + 1 - cs->multis));
1455 	cs->smultis -= len;
1456 
1457 	if (cs->smultis == 0) {
1458 		free(cs->multis);
1459 		cs->multis = NULL;
1460 		return;
1461 	}
1462 
1463 	cs->multis = realloc(cs->multis, cs->smultis);
1464 	assert(cs->multis != NULL);
1465 }
1466 
1467 /*
1468  - mcin - is a collating element in a cset?
1469  == static int mcin(cset *cs, char *cp);
1470  */
1471 static int
1472 mcin(
1473     cset *cs,
1474     char *cp)
1475 {
1476 
1477 	_DIAGASSERT(cs != NULL);
1478 	_DIAGASSERT(cp != NULL);
1479 
1480 	return(mcfind(cs, cp) != NULL);
1481 }
1482 
1483 /*
1484  - mcfind - find a collating element in a cset
1485  == static char *mcfind(cset *cs, char *cp);
1486  */
1487 static char *
1488 mcfind(
1489     cset *cs,
1490     char *cp)
1491 {
1492 	char *p;
1493 
1494 	_DIAGASSERT(cs != NULL);
1495 	_DIAGASSERT(cp != NULL);
1496 
1497 	if (cs->multis == NULL)
1498 		return(NULL);
1499 	for (p = cs->multis; *p != '\0'; p += strlen(p) + 1)
1500 		if (strcmp(cp, p) == 0)
1501 			return(p);
1502 	return(NULL);
1503 }
1504 #endif
1505 
1506 /*
1507  - mcinvert - invert the list of collating elements in a cset
1508  == static void mcinvert(struct parse *p, cset *cs);
1509  *
1510  * This would have to know the set of possibilities.  Implementation
1511  * is deferred.
1512  */
1513 /* ARGSUSED */
1514 static void
mcinvert(struct parse * p,cset * cs)1515 mcinvert(
1516     struct parse *p,
1517     cset *cs)
1518 {
1519 
1520 	_DIAGASSERT(p != NULL);
1521 	_DIAGASSERT(cs != NULL);
1522 
1523 	assert(cs->multis == NULL);	/* xxx */
1524 }
1525 
1526 /*
1527  - mccase - add case counterparts of the list of collating elements in a cset
1528  == static void mccase(struct parse *p, cset *cs);
1529  *
1530  * This would have to know the set of possibilities.  Implementation
1531  * is deferred.
1532  */
1533 /* ARGSUSED */
1534 static void
mccase(struct parse * p,cset * cs)1535 mccase(
1536     struct parse *p,
1537     cset *cs)
1538 {
1539 
1540 	_DIAGASSERT(p != NULL);
1541 	_DIAGASSERT(cs != NULL);
1542 
1543 	assert(cs->multis == NULL);	/* xxx */
1544 }
1545 
1546 /*
1547  - isinsets - is this character in any sets?
1548  == static int isinsets(struct re_guts *g, int c);
1549  */
1550 static int			/* predicate */
isinsets(struct re_guts * g,int c)1551 isinsets(
1552     struct re_guts *g,
1553     int c)
1554 {
1555 	uch *col;
1556 	size_t i;
1557 	size_t ncols;
1558 	unsigned uc = (unsigned char)c;
1559 
1560 	_DIAGASSERT(g != NULL);
1561 
1562 	if (g->setbits == NULL)
1563 		return 0;
1564 
1565 	ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1566 
1567 	for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1568 		if (col[uc] != 0)
1569 			return(1);
1570 	return(0);
1571 }
1572 
1573 /*
1574  - samesets - are these two characters in exactly the same sets?
1575  == static int samesets(struct re_guts *g, int c1, int c2);
1576  */
1577 static int			/* predicate */
samesets(struct re_guts * g,int c1,int c2)1578 samesets(
1579     struct re_guts *g,
1580     int c1,
1581     int c2)
1582 {
1583 	uch *col;
1584 	size_t i;
1585 	size_t ncols;
1586 	unsigned uc1 = (unsigned char)c1;
1587 	unsigned uc2 = (unsigned char)c2;
1588 
1589 	_DIAGASSERT(g != NULL);
1590 
1591 	ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1592 
1593 	for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1594 		if (col[uc1] != col[uc2])
1595 			return(0);
1596 	return(1);
1597 }
1598 
1599 /*
1600  - categorize - sort out character categories
1601  == static void categorize(struct parse *p, struct re_guts *g);
1602  */
1603 static void
categorize(struct parse * p,struct re_guts * g)1604 categorize(
1605     struct parse *p,
1606     struct re_guts *g)
1607 {
1608 	cat_t *cats;
1609 	int c;
1610 	int c2;
1611 	cat_t cat;
1612 
1613 	_DIAGASSERT(p != NULL);
1614 	_DIAGASSERT(g != NULL);
1615 
1616 	cats = g->categories;
1617 
1618 	/* avoid making error situations worse */
1619 	if (p->error != 0)
1620 		return;
1621 
1622 	for (c = CHAR_MIN; c <= CHAR_MAX; c++)
1623 		if (cats[c] == 0 && isinsets(g, c)) {
1624 			_DIAGASSERT(__type_fit(unsigned char,
1625 			    g->ncategories + 1));
1626 			cat = g->ncategories++;
1627 			cats[c] = cat;
1628 			for (c2 = c+1; c2 <= CHAR_MAX; c2++)
1629 				if (cats[c2] == 0 && samesets(g, c, c2))
1630 					cats[c2] = cat;
1631 		}
1632 }
1633 
1634 /*
1635  - dupl - emit a duplicate of a bunch of sops
1636  == static sopno dupl(struct parse *p, sopno start, sopno finish);
1637  */
1638 static sopno			/* start of duplicate */
dupl(struct parse * p,sopno start,sopno finish)1639 dupl(
1640     struct parse *p,
1641     sopno start,			/* from here */
1642     sopno finish)			/* to this less one */
1643 {
1644 	sopno ret;
1645 	sopno len = finish - start;
1646 
1647 	_DIAGASSERT(p != NULL);
1648 
1649 	ret = HERE();
1650 
1651 	assert(finish >= start);
1652 	if (len == 0)
1653 		return(ret);
1654 	if (!enlarge(p, p->ssize + len))/* this many unexpected additions */
1655 		return ret;
1656 	(void)memcpy(p->strip + p->slen, p->strip + start,
1657 	    (size_t)len * sizeof(sop));
1658 	p->slen += len;
1659 	return(ret);
1660 }
1661 
1662 /*
1663  - doemit - emit a strip operator
1664  == static void doemit(struct parse *p, sop op, size_t opnd);
1665  *
1666  * It might seem better to implement this as a macro with a function as
1667  * hard-case backup, but it's just too big and messy unless there are
1668  * some changes to the data structures.  Maybe later.
1669  */
1670 static void
doemit(struct parse * p,sop op,sopno opnd)1671 doemit(
1672     struct parse *p,
1673     sop op,
1674     sopno opnd)
1675 {
1676 	_DIAGASSERT(p != NULL);
1677 
1678 	/* avoid making error situations worse */
1679 	if (p->error != 0)
1680 		return;
1681 
1682 	/* deal with oversize operands ("can't happen", more or less) */
1683 	assert(opnd < 1<<OPSHIFT);
1684 
1685 	/* deal with undersized strip */
1686 	if (p->slen >= p->ssize)
1687 		if (!enlarge(p, (p->ssize+1) / 2 * 3))	/* +50% */
1688 			return;
1689 
1690 	/* finally, it's all reduced to the easy case */
1691 	p->strip[p->slen++] = (sop)SOP(op, opnd);
1692 }
1693 
1694 /*
1695  - doinsert - insert a sop into the strip
1696  == static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
1697  */
1698 static void
doinsert(struct parse * p,sop op,sopno opnd,sopno pos)1699 doinsert(
1700     struct parse *p,
1701     sop op,
1702     sopno opnd,
1703     sopno pos)
1704 {
1705 	sopno sn;
1706 	sop s;
1707 	int i;
1708 
1709 	_DIAGASSERT(p != NULL);
1710 
1711 	/* avoid making error situations worse */
1712 	if (p->error != 0)
1713 		return;
1714 
1715 	sn = HERE();
1716 	EMIT(op, opnd);		/* do checks, ensure space */
1717 	assert(HERE() == sn+1);
1718 	s = p->strip[sn];
1719 
1720 	/* adjust paren pointers */
1721 	assert(pos > 0);
1722 	for (i = 1; i < NPAREN; i++) {
1723 		if (p->pbegin[i] >= pos) {
1724 			p->pbegin[i]++;
1725 		}
1726 		if (p->pend[i] >= pos) {
1727 			p->pend[i]++;
1728 		}
1729 	}
1730 
1731 	memmove(&p->strip[pos+1], &p->strip[pos], (HERE()-pos-1)*sizeof(sop));
1732 	p->strip[pos] = s;
1733 }
1734 
1735 /*
1736  - dofwd - complete a forward reference
1737  == static void dofwd(struct parse *p, sopno pos, sop value);
1738  */
1739 static void
dofwd(struct parse * p,sopno pos,sopno value)1740 dofwd(
1741     struct parse *p,
1742     sopno pos,
1743     sopno value)
1744 {
1745 
1746 	_DIAGASSERT(p != NULL);
1747 
1748 	/* avoid making error situations worse */
1749 	if (p->error != 0)
1750 		return;
1751 
1752 	assert(value < 1<<OPSHIFT);
1753 	p->strip[pos] = (sop)(OP(p->strip[pos]) | value);
1754 }
1755 
1756 /*
1757  - enlarge - enlarge the strip
1758  == static void enlarge(struct parse *p, sopno size);
1759  */
1760 static int
enlarge(struct parse * p,sopno size)1761 enlarge(struct parse *p, sopno size)
1762 {
1763 	_DIAGASSERT(p != NULL);
1764 
1765 	if (p->ssize >= size)
1766 		return 1;
1767 
1768 	if (MEMSIZE(p) > MEMLIMIT || reallocarr(&p->strip, size, sizeof(sop))) {
1769 		SETERROR(REG_ESPACE);
1770 		return 0;
1771 	}
1772 	p->ssize = size;
1773 	return 1;
1774 }
1775 
1776 /*
1777  - stripsnug - compact the strip
1778  == static void stripsnug(struct parse *p, struct re_guts *g);
1779  */
1780 static void
stripsnug(struct parse * p,struct re_guts * g)1781 stripsnug(
1782     struct parse *p,
1783     struct re_guts *g)
1784 {
1785 
1786 	_DIAGASSERT(p != NULL);
1787 	_DIAGASSERT(g != NULL);
1788 
1789 	g->nstates = p->slen;
1790 	g->strip = p->strip;
1791 	reallocarr(&g->strip, p->slen, sizeof(sop));
1792 	/* Ignore error as tries to free memory only. */
1793 }
1794 
1795 /*
1796  - findmust - fill in must and mlen with longest mandatory literal string
1797  == static void findmust(struct parse *p, struct re_guts *g);
1798  *
1799  * This algorithm could do fancy things like analyzing the operands of |
1800  * for common subsequences.  Someday.  This code is simple and finds most
1801  * of the interesting cases.
1802  *
1803  * Note that must and mlen got initialized during setup.
1804  */
1805 static void
findmust(struct parse * p,struct re_guts * g)1806 findmust(
1807     struct parse *p,
1808     struct re_guts *g)
1809 {
1810 	sop *scan;
1811 	sop *start = NULL;
1812 	sop *newstart = NULL;
1813 	sopno newlen;
1814 	sop s;
1815 	char *cp;
1816 	sopno i;
1817 
1818 	_DIAGASSERT(p != NULL);
1819 	_DIAGASSERT(g != NULL);
1820 
1821 	/* avoid making error situations worse */
1822 	if (p->error != 0)
1823 		return;
1824 
1825 	/* find the longest OCHAR sequence in strip */
1826 	newlen = 0;
1827 	scan = g->strip + 1;
1828 	do {
1829 		s = *scan++;
1830 		switch (OP(s)) {
1831 		case OCHAR:		/* sequence member */
1832 			if (newlen == 0)		/* new sequence */
1833 				newstart = scan - 1;
1834 			newlen++;
1835 			break;
1836 		case OPLUS_:		/* things that don't break one */
1837 		case OLPAREN:
1838 		case ORPAREN:
1839 			break;
1840 		case OQUEST_:		/* things that must be skipped */
1841 		case OCH_:
1842 			scan--;
1843 			do {
1844 				scan += OPND(s);
1845 				s = *scan;
1846 				/* assert() interferes w debug printouts */
1847 				if (OP(s) != O_QUEST && OP(s) != O_CH &&
1848 							OP(s) != OOR2) {
1849 					g->iflags |= BAD;
1850 					return;
1851 				}
1852 			} while (OP(s) != O_QUEST && OP(s) != O_CH);
1853 			/* FALLTHROUGH */
1854 		default:		/* things that break a sequence */
1855 			if (newlen > g->mlen) {		/* ends one */
1856 				start = newstart;
1857 				g->mlen = newlen;
1858 			}
1859 			newlen = 0;
1860 			break;
1861 		}
1862 	} while (OP(s) != OEND);
1863 
1864 	if (start == NULL)
1865 		g->mlen = 0;
1866 
1867 	if (g->mlen == 0)	/* there isn't one */
1868 		return;
1869 
1870 	/* turn it into a character string */
1871 	g->must = malloc((size_t)g->mlen + 1);
1872 	if (g->must == NULL) {		/* argh; just forget it */
1873 		g->mlen = 0;
1874 		return;
1875 	}
1876 	cp = g->must;
1877 	scan = start;
1878 	for (i = g->mlen; i > 0; i--) {
1879 		while (OP(s = *scan++) != OCHAR)
1880 			continue;
1881 		assert(cp < g->must + g->mlen);
1882 		*cp++ = (char)OPND(s);
1883 	}
1884 	assert(cp == g->must + g->mlen);
1885 	*cp++ = '\0';		/* just on general principles */
1886 }
1887 
1888 /*
1889  - pluscount - count + nesting
1890  == static sopno pluscount(struct parse *p, struct re_guts *g);
1891  */
1892 static sopno			/* nesting depth */
pluscount(struct parse * p,struct re_guts * g)1893 pluscount(
1894     struct parse *p,
1895     struct re_guts *g)
1896 {
1897 	sop *scan;
1898 	sop s;
1899 	sopno plusnest = 0;
1900 	sopno maxnest = 0;
1901 
1902 	_DIAGASSERT(p != NULL);
1903 	_DIAGASSERT(g != NULL);
1904 
1905 	if (p->error != 0)
1906 		return(0);	/* there may not be an OEND */
1907 
1908 	scan = g->strip + 1;
1909 	do {
1910 		s = *scan++;
1911 		switch (OP(s)) {
1912 		case OPLUS_:
1913 			plusnest++;
1914 			break;
1915 		case O_PLUS:
1916 			if (plusnest > maxnest)
1917 				maxnest = plusnest;
1918 			plusnest--;
1919 			break;
1920 		}
1921 	} while (OP(s) != OEND);
1922 	if (plusnest != 0)
1923 		g->iflags |= BAD;
1924 	return(maxnest);
1925 }
1926