xref: /freebsd/contrib/llvm-project/llvm/lib/Support/regcomp.c (revision 43a5ec4eb41567cc92586503212743d89686d78f)
1 /*-
2  * This code is derived from OpenBSD's libc/regex, original license follows:
3  *
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  * 3. 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  *	@(#)regcomp.c	8.5 (Berkeley) 3/20/94
36  */
37 
38 #include <sys/types.h>
39 #include <stdint.h>
40 #include <stdio.h>
41 #include <string.h>
42 #include <ctype.h>
43 #include <limits.h>
44 #include <stdlib.h>
45 #include "regex_impl.h"
46 
47 #include "regutils.h"
48 #include "regex2.h"
49 
50 #include "llvm/Config/config.h"
51 #include "llvm/Support/Compiler.h"
52 
53 /* character-class table */
54 static struct cclass {
55 	const char *name;
56 	const char *chars;
57 	const char *multis;
58 } cclasses[] = {
59 	{ "alnum",	"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\
60 0123456789",				""} ,
61 	{ "alpha",	"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz",
62 					""} ,
63 	{ "blank",	" \t",		""} ,
64 	{ "cntrl",	"\007\b\t\n\v\f\r\1\2\3\4\5\6\16\17\20\21\22\23\24\
65 \25\26\27\30\31\32\33\34\35\36\37\177",	""} ,
66 	{ "digit",	"0123456789",	""} ,
67 	{ "graph",	"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\
68 0123456789!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~",
69 					""} ,
70 	{ "lower",	"abcdefghijklmnopqrstuvwxyz",
71 					""} ,
72 	{ "print",	"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\
73 0123456789!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~ ",
74 					""} ,
75 	{ "punct",	"!\"#$%&'()*+,-./:;<=>?@[\\]^_`{|}~",
76 					""} ,
77 	{ "space",	"\t\n\v\f\r ",	""} ,
78 	{ "upper",	"ABCDEFGHIJKLMNOPQRSTUVWXYZ",
79 					""} ,
80 	{ "xdigit",	"0123456789ABCDEFabcdef",
81 					""} ,
82 	{ NULL,		0,		"" }
83 };
84 
85 /* character-name table */
86 static struct cname {
87 	const char *name;
88 	char code;
89 } cnames[] = {
90 	{ "NUL",			'\0' },
91 	{ "SOH",			'\001' },
92 	{ "STX",			'\002' },
93 	{ "ETX",			'\003' },
94 	{ "EOT",			'\004' },
95 	{ "ENQ",			'\005' },
96 	{ "ACK",			'\006' },
97 	{ "BEL",			'\007' },
98 	{ "alert",			'\007' },
99 	{ "BS",				'\010' },
100 	{ "backspace",			'\b' },
101 	{ "HT",				'\011' },
102 	{ "tab",			'\t' },
103 	{ "LF",				'\012' },
104 	{ "newline",			'\n' },
105 	{ "VT",				'\013' },
106 	{ "vertical-tab",		'\v' },
107 	{ "FF",				'\014' },
108 	{ "form-feed",			'\f' },
109 	{ "CR",				'\015' },
110 	{ "carriage-return",		'\r' },
111 	{ "SO",				'\016' },
112 	{ "SI",				'\017' },
113 	{ "DLE",			'\020' },
114 	{ "DC1",			'\021' },
115 	{ "DC2",			'\022' },
116 	{ "DC3",			'\023' },
117 	{ "DC4",			'\024' },
118 	{ "NAK",			'\025' },
119 	{ "SYN",			'\026' },
120 	{ "ETB",			'\027' },
121 	{ "CAN",			'\030' },
122 	{ "EM",				'\031' },
123 	{ "SUB",			'\032' },
124 	{ "ESC",			'\033' },
125 	{ "IS4",			'\034' },
126 	{ "FS",				'\034' },
127 	{ "IS3",			'\035' },
128 	{ "GS",				'\035' },
129 	{ "IS2",			'\036' },
130 	{ "RS",				'\036' },
131 	{ "IS1",			'\037' },
132 	{ "US",				'\037' },
133 	{ "space",			' ' },
134 	{ "exclamation-mark",		'!' },
135 	{ "quotation-mark",		'"' },
136 	{ "number-sign",		'#' },
137 	{ "dollar-sign",		'$' },
138 	{ "percent-sign",		'%' },
139 	{ "ampersand",			'&' },
140 	{ "apostrophe",			'\'' },
141 	{ "left-parenthesis",		'(' },
142 	{ "right-parenthesis",		')' },
143 	{ "asterisk",			'*' },
144 	{ "plus-sign",			'+' },
145 	{ "comma",			',' },
146 	{ "hyphen",			'-' },
147 	{ "hyphen-minus",		'-' },
148 	{ "period",			'.' },
149 	{ "full-stop",			'.' },
150 	{ "slash",			'/' },
151 	{ "solidus",			'/' },
152 	{ "zero",			'0' },
153 	{ "one",			'1' },
154 	{ "two",			'2' },
155 	{ "three",			'3' },
156 	{ "four",			'4' },
157 	{ "five",			'5' },
158 	{ "six",			'6' },
159 	{ "seven",			'7' },
160 	{ "eight",			'8' },
161 	{ "nine",			'9' },
162 	{ "colon",			':' },
163 	{ "semicolon",			';' },
164 	{ "less-than-sign",		'<' },
165 	{ "equals-sign",		'=' },
166 	{ "greater-than-sign",		'>' },
167 	{ "question-mark",		'?' },
168 	{ "commercial-at",		'@' },
169 	{ "left-square-bracket",	'[' },
170 	{ "backslash",			'\\' },
171 	{ "reverse-solidus",		'\\' },
172 	{ "right-square-bracket",	']' },
173 	{ "circumflex",			'^' },
174 	{ "circumflex-accent",		'^' },
175 	{ "underscore",			'_' },
176 	{ "low-line",			'_' },
177 	{ "grave-accent",		'`' },
178 	{ "left-brace",			'{' },
179 	{ "left-curly-bracket",		'{' },
180 	{ "vertical-line",		'|' },
181 	{ "right-brace",		'}' },
182 	{ "right-curly-bracket",	'}' },
183 	{ "tilde",			'~' },
184 	{ "DEL",			'\177' },
185 	{ NULL,				0 }
186 };
187 
188 /*
189  * parse structure, passed up and down to avoid global variables and
190  * other clumsinesses
191  */
192 struct parse {
193 	char *next;		/* next character in RE */
194 	char *end;		/* end of string (-> NUL normally) */
195 	int error;		/* has an error been seen? */
196 	sop *strip;		/* malloced strip */
197 	sopno ssize;		/* malloced strip size (allocated) */
198 	sopno slen;		/* malloced strip length (used) */
199 	int ncsalloc;		/* number of csets allocated */
200 	struct re_guts *g;
201 #	define	NPAREN	10	/* we need to remember () 1-9 for back refs */
202 	sopno pbegin[NPAREN];	/* -> ( ([0] unused) */
203 	sopno pend[NPAREN];	/* -> ) ([0] unused) */
204 };
205 
206 static void p_ere(struct parse *, int);
207 static void p_ere_exp(struct parse *);
208 static void p_str(struct parse *);
209 static void p_bre(struct parse *, int, int);
210 static int p_simp_re(struct parse *, int);
211 static int p_count(struct parse *);
212 static void p_bracket(struct parse *);
213 static void p_b_term(struct parse *, cset *);
214 static void p_b_cclass(struct parse *, cset *);
215 static void p_b_eclass(struct parse *, cset *);
216 static char p_b_symbol(struct parse *);
217 static char p_b_coll_elem(struct parse *, int);
218 static char othercase(int);
219 static void bothcases(struct parse *, int);
220 static void ordinary(struct parse *, int);
221 static void nonnewline(struct parse *);
222 static void repeat(struct parse *, sopno, int, int);
223 static int seterr(struct parse *, int);
224 static cset *allocset(struct parse *);
225 static void freeset(struct parse *, cset *);
226 static int freezeset(struct parse *, cset *);
227 static int firstch(struct parse *, cset *);
228 static int nch(struct parse *, cset *);
229 static void mcadd(struct parse *, cset *, const char *);
230 static void mcinvert(struct parse *, cset *);
231 static void mccase(struct parse *, cset *);
232 static int isinsets(struct re_guts *, int);
233 static int samesets(struct re_guts *, int, int);
234 static void categorize(struct parse *, struct re_guts *);
235 static sopno dupl(struct parse *, sopno, sopno);
236 static void doemit(struct parse *, sop, size_t);
237 static void doinsert(struct parse *, sop, size_t, sopno);
238 static void dofwd(struct parse *, sopno, sop);
239 static void enlarge(struct parse *, sopno);
240 static void stripsnug(struct parse *, struct re_guts *);
241 static void findmust(struct parse *, struct re_guts *);
242 static sopno pluscount(struct parse *, struct re_guts *);
243 
244 static char nuls[10];		/* place to point scanner in event of error */
245 
246 /*
247  * macros for use with parse structure
248  * BEWARE:  these know that the parse structure is named `p' !!!
249  */
250 #define	PEEK()	(*p->next)
251 #define	PEEK2()	(*(p->next+1))
252 #define	MORE()	(p->next < p->end)
253 #define	MORE2()	(p->next+1 < p->end)
254 #define	SEE(c)	(MORE() && PEEK() == (c))
255 #define	SEETWO(a, b)	(MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
256 #define	EAT(c)	((SEE(c)) ? (NEXT(), 1) : 0)
257 #define	EATTWO(a, b)	((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
258 #define	NEXT()	(p->next++)
259 #define	NEXT2()	(p->next += 2)
260 #define	NEXTn(n)	(p->next += (n))
261 #define	GETNEXT()	(*p->next++)
262 #define	SETERROR(e)	seterr(p, (e))
263 #define	REQUIRE(co, e)	(void)((co) || SETERROR(e))
264 #define	MUSTSEE(c, e)	(REQUIRE(MORE() && PEEK() == (c), e))
265 #define	MUSTEAT(c, e)	(REQUIRE(MORE() && GETNEXT() == (c), e))
266 #define	MUSTNOTSEE(c, e)	(REQUIRE(!MORE() || PEEK() != (c), e))
267 #define	EMIT(op, sopnd)	doemit(p, (sop)(op), (size_t)(sopnd))
268 #define	INSERT(op, pos)	doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
269 #define	AHEAD(pos)		dofwd(p, pos, HERE()-(pos))
270 #define	ASTERN(sop, pos)	EMIT(sop, HERE()-pos)
271 #define	HERE()		(p->slen)
272 #define	THERE()		(p->slen - 1)
273 #define	THERETHERE()	(p->slen - 2)
274 #define	DROP(n)	(p->slen -= (n))
275 
276 #ifdef	_POSIX2_RE_DUP_MAX
277 #define	DUPMAX	_POSIX2_RE_DUP_MAX
278 #else
279 #define	DUPMAX	255
280 #endif
281 #define	INFINITY	(DUPMAX + 1)
282 
283 #ifndef NDEBUG
284 static int never = 0;		/* for use in asserts; shuts lint up */
285 #else
286 #define	never	0		/* some <assert.h>s have bugs too */
287 #endif
288 
289 /*
290  - llvm_regcomp - interface for parser and compilation
291  */
292 int				/* 0 success, otherwise REG_something */
293 llvm_regcomp(llvm_regex_t *preg, const char *pattern, int cflags)
294 {
295 	struct parse pa;
296 	struct re_guts *g;
297 	struct parse *p = &pa;
298 	int i;
299 	size_t len;
300 #ifdef REDEBUG
301 #	define	GOODFLAGS(f)	(f)
302 #else
303 #	define	GOODFLAGS(f)	((f)&~REG_DUMP)
304 #endif
305 
306 	cflags = GOODFLAGS(cflags);
307 	if ((cflags&REG_EXTENDED) && (cflags&REG_NOSPEC))
308 		return(REG_INVARG);
309 
310 	if (cflags&REG_PEND) {
311 		if (preg->re_endp < pattern)
312 			return(REG_INVARG);
313 		len = preg->re_endp - pattern;
314 	} else
315 		len = strlen((const char *)pattern);
316 
317 	/* do the mallocs early so failure handling is easy */
318 	g = (struct re_guts *)malloc(sizeof(struct re_guts) +
319 							(NC-1)*sizeof(cat_t));
320 	if (g == NULL)
321 		return(REG_ESPACE);
322 	p->ssize = len/(size_t)2*(size_t)3 + (size_t)1;	/* ugh */
323 	p->strip = (sop *)calloc(p->ssize, sizeof(sop));
324 	p->slen = 0;
325 	if (p->strip == NULL) {
326 		free((char *)g);
327 		return(REG_ESPACE);
328 	}
329 
330 	/* set things up */
331 	p->g = g;
332 	p->next = (char *)pattern;	/* convenience; we do not modify it */
333 	p->end = p->next + len;
334 	p->error = 0;
335 	p->ncsalloc = 0;
336 	for (i = 0; i < NPAREN; i++) {
337 		p->pbegin[i] = 0;
338 		p->pend[i] = 0;
339 	}
340 	g->csetsize = NC;
341 	g->sets = NULL;
342 	g->setbits = NULL;
343 	g->ncsets = 0;
344 	g->cflags = cflags;
345 	g->iflags = 0;
346 	g->nbol = 0;
347 	g->neol = 0;
348 	g->must = NULL;
349 	g->mlen = 0;
350 	g->nsub = 0;
351 	g->ncategories = 1;	/* category 0 is "everything else" */
352 	g->categories = &g->catspace[-(CHAR_MIN)];
353 	(void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
354 	g->backrefs = 0;
355 
356 	/* do it */
357 	EMIT(OEND, 0);
358 	g->firststate = THERE();
359 	if (cflags&REG_EXTENDED)
360 		p_ere(p, OUT);
361 	else if (cflags&REG_NOSPEC)
362 		p_str(p);
363 	else
364 		p_bre(p, OUT, OUT);
365 	EMIT(OEND, 0);
366 	g->laststate = THERE();
367 
368 	/* tidy up loose ends and fill things in */
369 	categorize(p, g);
370 	stripsnug(p, g);
371 	findmust(p, g);
372 	g->nplus = pluscount(p, g);
373 	g->magic = MAGIC2;
374 	preg->re_nsub = g->nsub;
375 	preg->re_g = g;
376 	preg->re_magic = MAGIC1;
377 #ifndef REDEBUG
378 	/* not debugging, so can't rely on the assert() in llvm_regexec() */
379 	if (g->iflags&REGEX_BAD)
380 		SETERROR(REG_ASSERT);
381 #endif
382 
383 	/* win or lose, we're done */
384 	if (p->error != 0)	/* lose */
385 		llvm_regfree(preg);
386 	return(p->error);
387 }
388 
389 /*
390  - p_ere - ERE parser top level, concatenation and alternation
391  */
392 static void
393 p_ere(struct parse *p, int stop)	/* character this ERE should end at */
394 {
395 	char c;
396 	sopno prevback = 0;
397 	sopno prevfwd = 0;
398 	sopno conc;
399 	int first = 1;		/* is this the first alternative? */
400 
401 	for (;;) {
402 		/* do a bunch of concatenated expressions */
403 		conc = HERE();
404 		while (MORE() && (c = PEEK()) != '|' && c != stop)
405 			p_ere_exp(p);
406 		REQUIRE(HERE() != conc, REG_EMPTY);	/* require nonempty */
407 
408 		if (!EAT('|'))
409 			break;		/* NOTE BREAK OUT */
410 
411 		if (first) {
412 			INSERT(OCH_, conc);	/* offset is wrong */
413 			prevfwd = conc;
414 			prevback = conc;
415 			first = 0;
416 		}
417 		ASTERN(OOR1, prevback);
418 		prevback = THERE();
419 		AHEAD(prevfwd);			/* fix previous offset */
420 		prevfwd = HERE();
421 		EMIT(OOR2, 0);			/* offset is very wrong */
422 	}
423 
424 	if (!first) {		/* tail-end fixups */
425 		AHEAD(prevfwd);
426 		ASTERN(O_CH, prevback);
427 	}
428 
429 	assert(!MORE() || SEE(stop));
430 }
431 
432 /*
433  - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
434  */
435 static void
436 p_ere_exp(struct parse *p)
437 {
438 	char c;
439 	sopno pos;
440 	int count;
441 	int count2;
442 	int backrefnum;
443 	sopno subno;
444 	int wascaret = 0;
445 
446 	assert(MORE());		/* caller should have ensured this */
447 	c = GETNEXT();
448 
449 	pos = HERE();
450 	switch (c) {
451 	case '(':
452 		REQUIRE(MORE(), REG_EPAREN);
453 		p->g->nsub++;
454 		subno = p->g->nsub;
455 		if (subno < NPAREN)
456 			p->pbegin[subno] = HERE();
457 		EMIT(OLPAREN, subno);
458 		if (!SEE(')'))
459 			p_ere(p, ')');
460 		if (subno < NPAREN) {
461 			p->pend[subno] = HERE();
462 			assert(p->pend[subno] != 0);
463 		}
464 		EMIT(ORPAREN, subno);
465 		MUSTEAT(')', REG_EPAREN);
466 		break;
467 #ifndef POSIX_MISTAKE
468 	case ')':		/* happens only if no current unmatched ( */
469 		/*
470 		 * You may ask, why the ifndef?  Because I didn't notice
471 		 * this until slightly too late for 1003.2, and none of the
472 		 * other 1003.2 regular-expression reviewers noticed it at
473 		 * all.  So an unmatched ) is legal POSIX, at least until
474 		 * we can get it fixed.
475 		 */
476 		SETERROR(REG_EPAREN);
477 		break;
478 #endif
479 	case '^':
480 		EMIT(OBOL, 0);
481 		p->g->iflags |= USEBOL;
482 		p->g->nbol++;
483 		wascaret = 1;
484 		break;
485 	case '$':
486 		EMIT(OEOL, 0);
487 		p->g->iflags |= USEEOL;
488 		p->g->neol++;
489 		break;
490 	case '|':
491 		SETERROR(REG_EMPTY);
492 		break;
493 	case '*':
494 	case '+':
495 	case '?':
496 		SETERROR(REG_BADRPT);
497 		break;
498 	case '.':
499 		if (p->g->cflags&REG_NEWLINE)
500 			nonnewline(p);
501 		else
502 			EMIT(OANY, 0);
503 		break;
504 	case '[':
505 		p_bracket(p);
506 		break;
507 	case '\\':
508 		REQUIRE(MORE(), REG_EESCAPE);
509 		c = GETNEXT();
510 		if (c >= '1' && c <= '9') {
511 			/* \[0-9] is taken to be a back-reference to a previously specified
512 			 * matching group. backrefnum will hold the number. The matching
513 			 * group must exist (i.e. if \4 is found there must have been at
514 			 * least 4 matching groups specified in the pattern previously).
515 			 */
516 			backrefnum = c - '0';
517 			if (p->pend[backrefnum] == 0) {
518 				SETERROR(REG_ESUBREG);
519 				break;
520 			}
521 
522 			/* Make sure everything checks out and emit the sequence
523 			 * that marks a back-reference to the parse structure.
524 			 */
525 			assert(backrefnum <= p->g->nsub);
526 			EMIT(OBACK_, backrefnum);
527 			assert(p->pbegin[backrefnum] != 0);
528 			assert(OP(p->strip[p->pbegin[backrefnum]]) != OLPAREN);
529 			assert(OP(p->strip[p->pend[backrefnum]]) != ORPAREN);
530 			(void) dupl(p, p->pbegin[backrefnum]+1, p->pend[backrefnum]);
531 			EMIT(O_BACK, backrefnum);
532 			p->g->backrefs = 1;
533 		} else {
534 			/* Other chars are simply themselves when escaped with a backslash.
535 			 */
536 			ordinary(p, c);
537 		}
538 		break;
539 	case '{':		/* okay as ordinary except if digit follows */
540 		REQUIRE(!MORE() || !isdigit((uch)PEEK()), REG_BADRPT);
541 		LLVM_FALLTHROUGH;
542 	default:
543 		ordinary(p, c);
544 		break;
545 	}
546 
547 	if (!MORE())
548 		return;
549 	c = PEEK();
550 	/* we call { a repetition if followed by a digit */
551 	if (!( c == '*' || c == '+' || c == '?' ||
552 				(c == '{' && MORE2() && isdigit((uch)PEEK2())) ))
553 		return;		/* no repetition, we're done */
554 	NEXT();
555 
556 	REQUIRE(!wascaret, REG_BADRPT);
557 	switch (c) {
558 	case '*':	/* implemented as +? */
559 		/* this case does not require the (y|) trick, noKLUDGE */
560 		INSERT(OPLUS_, pos);
561 		ASTERN(O_PLUS, pos);
562 		INSERT(OQUEST_, pos);
563 		ASTERN(O_QUEST, pos);
564 		break;
565 	case '+':
566 		INSERT(OPLUS_, pos);
567 		ASTERN(O_PLUS, pos);
568 		break;
569 	case '?':
570 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
571 		INSERT(OCH_, pos);		/* offset slightly wrong */
572 		ASTERN(OOR1, pos);		/* this one's right */
573 		AHEAD(pos);			/* fix the OCH_ */
574 		EMIT(OOR2, 0);			/* offset very wrong... */
575 		AHEAD(THERE());			/* ...so fix it */
576 		ASTERN(O_CH, THERETHERE());
577 		break;
578 	case '{':
579 		count = p_count(p);
580 		if (EAT(',')) {
581 			if (isdigit((uch)PEEK())) {
582 				count2 = p_count(p);
583 				REQUIRE(count <= count2, REG_BADBR);
584 			} else		/* single number with comma */
585 				count2 = INFINITY;
586 		} else		/* just a single number */
587 			count2 = count;
588 		repeat(p, pos, count, count2);
589 		if (!EAT('}')) {	/* error heuristics */
590 			while (MORE() && PEEK() != '}')
591 				NEXT();
592 			REQUIRE(MORE(), REG_EBRACE);
593 			SETERROR(REG_BADBR);
594 		}
595 		break;
596 	}
597 
598 	if (!MORE())
599 		return;
600 	c = PEEK();
601 	if (!( c == '*' || c == '+' || c == '?' ||
602 				(c == '{' && MORE2() && isdigit((uch)PEEK2())) ) )
603 		return;
604 	SETERROR(REG_BADRPT);
605 }
606 
607 /*
608  - p_str - string (no metacharacters) "parser"
609  */
610 static void
611 p_str(struct parse *p)
612 {
613 	REQUIRE(MORE(), REG_EMPTY);
614 	while (MORE())
615 		ordinary(p, GETNEXT());
616 }
617 
618 /*
619  - p_bre - BRE parser top level, anchoring and concatenation
620  * Giving end1 as OUT essentially eliminates the end1/end2 check.
621  *
622  * This implementation is a bit of a kludge, in that a trailing $ is first
623  * taken as an ordinary character and then revised to be an anchor.  The
624  * only undesirable side effect is that '$' gets included as a character
625  * category in such cases.  This is fairly harmless; not worth fixing.
626  * The amount of lookahead needed to avoid this kludge is excessive.
627  */
628 static void
629 p_bre(struct parse *p,
630     int end1,		/* first terminating character */
631     int end2)		/* second terminating character */
632 {
633 	sopno start = HERE();
634 	int first = 1;			/* first subexpression? */
635 	int wasdollar = 0;
636 
637 	if (EAT('^')) {
638 		EMIT(OBOL, 0);
639 		p->g->iflags |= USEBOL;
640 		p->g->nbol++;
641 	}
642 	while (MORE() && !SEETWO(end1, end2)) {
643 		wasdollar = p_simp_re(p, first);
644 		first = 0;
645 	}
646 	if (wasdollar) {	/* oops, that was a trailing anchor */
647 		DROP(1);
648 		EMIT(OEOL, 0);
649 		p->g->iflags |= USEEOL;
650 		p->g->neol++;
651 	}
652 
653 	REQUIRE(HERE() != start, REG_EMPTY);	/* require nonempty */
654 }
655 
656 /*
657  - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
658  */
659 static int			/* was the simple RE an unbackslashed $? */
660 p_simp_re(struct parse *p,
661     int starordinary)		/* is a leading * an ordinary character? */
662 {
663 	int c;
664 	int count;
665 	int count2;
666 	sopno pos;
667 	int i;
668 	sopno subno;
669 #	define	BACKSL	(1<<CHAR_BIT)
670 
671         pos = HERE(); /* repetition op, if any, covers from here */
672 
673         assert(MORE()); /* caller should have ensured this */
674         c = GETNEXT();
675 	if (c == '\\') {
676 		REQUIRE(MORE(), REG_EESCAPE);
677 		c = BACKSL | GETNEXT();
678 	}
679 	switch (c) {
680 	case '.':
681 		if (p->g->cflags&REG_NEWLINE)
682 			nonnewline(p);
683 		else
684 			EMIT(OANY, 0);
685 		break;
686 	case '[':
687 		p_bracket(p);
688 		break;
689 	case BACKSL|'{':
690 		SETERROR(REG_BADRPT);
691 		break;
692 	case BACKSL|'(':
693 		p->g->nsub++;
694 		subno = p->g->nsub;
695 		if (subno < NPAREN)
696 			p->pbegin[subno] = HERE();
697 		EMIT(OLPAREN, subno);
698 		/* the MORE here is an error heuristic */
699 		if (MORE() && !SEETWO('\\', ')'))
700 			p_bre(p, '\\', ')');
701 		if (subno < NPAREN) {
702 			p->pend[subno] = HERE();
703 			assert(p->pend[subno] != 0);
704 		}
705 		EMIT(ORPAREN, subno);
706 		REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
707 		break;
708 	case BACKSL|')':	/* should not get here -- must be user */
709 	case BACKSL|'}':
710 		SETERROR(REG_EPAREN);
711 		break;
712 	case BACKSL|'1':
713 	case BACKSL|'2':
714 	case BACKSL|'3':
715 	case BACKSL|'4':
716 	case BACKSL|'5':
717 	case BACKSL|'6':
718 	case BACKSL|'7':
719 	case BACKSL|'8':
720 	case BACKSL|'9':
721 		i = (c&~BACKSL) - '0';
722 		assert(i < NPAREN);
723 		if (p->pend[i] != 0) {
724 			assert(i <= p->g->nsub);
725 			EMIT(OBACK_, i);
726 			assert(p->pbegin[i] != 0);
727 			assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
728 			assert(OP(p->strip[p->pend[i]]) == ORPAREN);
729 			(void) dupl(p, p->pbegin[i]+1, p->pend[i]);
730 			EMIT(O_BACK, i);
731 		} else
732 			SETERROR(REG_ESUBREG);
733 		p->g->backrefs = 1;
734 		break;
735 	case '*':
736 		REQUIRE(starordinary, REG_BADRPT);
737 		LLVM_FALLTHROUGH;
738 	default:
739 		ordinary(p, (char)c);
740 		break;
741 	}
742 
743 	if (EAT('*')) {		/* implemented as +? */
744 		/* this case does not require the (y|) trick, noKLUDGE */
745 		INSERT(OPLUS_, pos);
746 		ASTERN(O_PLUS, pos);
747 		INSERT(OQUEST_, pos);
748 		ASTERN(O_QUEST, pos);
749 	} else if (EATTWO('\\', '{')) {
750 		count = p_count(p);
751 		if (EAT(',')) {
752 			if (MORE() && isdigit((uch)PEEK())) {
753 				count2 = p_count(p);
754 				REQUIRE(count <= count2, REG_BADBR);
755 			} else		/* single number with comma */
756 				count2 = INFINITY;
757 		} else		/* just a single number */
758 			count2 = count;
759 		repeat(p, pos, count, count2);
760 		if (!EATTWO('\\', '}')) {	/* error heuristics */
761 			while (MORE() && !SEETWO('\\', '}'))
762 				NEXT();
763 			REQUIRE(MORE(), REG_EBRACE);
764 			SETERROR(REG_BADBR);
765 		}
766 	} else if (c == '$')	/* $ (but not \$) ends it */
767 		return(1);
768 
769 	return(0);
770 }
771 
772 /*
773  - p_count - parse a repetition count
774  */
775 static int			/* the value */
776 p_count(struct parse *p)
777 {
778 	int count = 0;
779 	int ndigits = 0;
780 
781 	while (MORE() && isdigit((uch)PEEK()) && count <= DUPMAX) {
782 		count = count*10 + (GETNEXT() - '0');
783 		ndigits++;
784 	}
785 
786 	REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
787 	return(count);
788 }
789 
790 /*
791  - p_bracket - parse a bracketed character list
792  *
793  * Note a significant property of this code:  if the allocset() did SETERROR,
794  * no set operations are done.
795  */
796 static void
797 p_bracket(struct parse *p)
798 {
799 	cset *cs;
800 	int invert = 0;
801 
802 	/* Dept of Truly Sickening Special-Case Kludges */
803 	if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
804 		EMIT(OBOW, 0);
805 		NEXTn(6);
806 		return;
807 	}
808 	if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
809 		EMIT(OEOW, 0);
810 		NEXTn(6);
811 		return;
812 	}
813 
814 	if ((cs = allocset(p)) == NULL) {
815 		/* allocset did set error status in p */
816 		return;
817 	}
818 
819 	if (EAT('^'))
820 		invert++;	/* make note to invert set at end */
821 	if (EAT(']'))
822 		CHadd(cs, ']');
823 	else if (EAT('-'))
824 		CHadd(cs, '-');
825 	while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
826 		p_b_term(p, cs);
827 	if (EAT('-'))
828 		CHadd(cs, '-');
829 	MUSTEAT(']', REG_EBRACK);
830 
831 	if (p->error != 0) {	/* don't mess things up further */
832 		freeset(p, cs);
833 		return;
834 	}
835 
836 	if (p->g->cflags&REG_ICASE) {
837 		int i;
838 		int ci;
839 
840 		for (i = p->g->csetsize - 1; i >= 0; i--)
841 			if (CHIN(cs, i) && isalpha(i)) {
842 				ci = othercase(i);
843 				if (ci != i)
844 					CHadd(cs, ci);
845 			}
846 		if (cs->multis != NULL)
847 			mccase(p, cs);
848 	}
849 	if (invert) {
850 		int i;
851 
852 		for (i = p->g->csetsize - 1; i >= 0; i--)
853 			if (CHIN(cs, i))
854 				CHsub(cs, i);
855 			else
856 				CHadd(cs, i);
857 		if (p->g->cflags&REG_NEWLINE)
858 			CHsub(cs, '\n');
859 		if (cs->multis != NULL)
860 			mcinvert(p, cs);
861 	}
862 
863 	assert(cs->multis == NULL);		/* xxx */
864 
865 	if (nch(p, cs) == 1) {		/* optimize singleton sets */
866 		ordinary(p, firstch(p, cs));
867 		freeset(p, cs);
868 	} else
869 		EMIT(OANYOF, freezeset(p, cs));
870 }
871 
872 /*
873  - p_b_term - parse one term of a bracketed character list
874  */
875 static void
876 p_b_term(struct parse *p, cset *cs)
877 {
878 	char c;
879 	char start, finish;
880 	int i;
881 
882 	/* classify what we've got */
883 	switch ((MORE()) ? PEEK() : '\0') {
884 	case '[':
885 		c = (MORE2()) ? PEEK2() : '\0';
886 		break;
887 	case '-':
888 		SETERROR(REG_ERANGE);
889 		return;			/* NOTE RETURN */
890 		break;
891 	default:
892 		c = '\0';
893 		break;
894 	}
895 
896 	switch (c) {
897 	case ':':		/* character class */
898 		NEXT2();
899 		REQUIRE(MORE(), REG_EBRACK);
900 		c = PEEK();
901 		REQUIRE(c != '-' && c != ']', REG_ECTYPE);
902 		p_b_cclass(p, cs);
903 		REQUIRE(MORE(), REG_EBRACK);
904 		REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
905 		break;
906 	case '=':		/* equivalence class */
907 		NEXT2();
908 		REQUIRE(MORE(), REG_EBRACK);
909 		c = PEEK();
910 		REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
911 		p_b_eclass(p, cs);
912 		REQUIRE(MORE(), REG_EBRACK);
913 		REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
914 		break;
915 	default:		/* symbol, ordinary character, or range */
916 /* xxx revision needed for multichar stuff */
917 		start = p_b_symbol(p);
918 		if (SEE('-') && MORE2() && PEEK2() != ']') {
919 			/* range */
920 			NEXT();
921 			if (EAT('-'))
922 				finish = '-';
923 			else
924 				finish = p_b_symbol(p);
925 		} else
926 			finish = start;
927 /* xxx what about signed chars here... */
928 		REQUIRE(start <= finish, REG_ERANGE);
929 		for (i = start; i <= finish; i++)
930 			CHadd(cs, i);
931 		break;
932 	}
933 }
934 
935 /*
936  - p_b_cclass - parse a character-class name and deal with it
937  */
938 static void
939 p_b_cclass(struct parse *p, cset *cs)
940 {
941 	char *sp = p->next;
942 	struct cclass *cp;
943 	size_t len;
944 	const char *u;
945 	char c;
946 
947 	while (MORE() && isalpha((uch)PEEK()))
948 		NEXT();
949 	len = p->next - sp;
950 	for (cp = cclasses; cp->name != NULL; cp++)
951 		if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
952 			break;
953 	if (cp->name == NULL) {
954 		/* oops, didn't find it */
955 		SETERROR(REG_ECTYPE);
956 		return;
957 	}
958 
959 	u = cp->chars;
960 	while ((c = *u++) != '\0')
961 		CHadd(cs, c);
962 	for (u = cp->multis; *u != '\0'; u += strlen(u) + 1)
963 		MCadd(p, cs, u);
964 }
965 
966 /*
967  - p_b_eclass - parse an equivalence-class name and deal with it
968  *
969  * This implementation is incomplete. xxx
970  */
971 static void
972 p_b_eclass(struct parse *p, cset *cs)
973 {
974 	char c;
975 
976 	c = p_b_coll_elem(p, '=');
977 	CHadd(cs, c);
978 }
979 
980 /*
981  - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
982  */
983 static char			/* value of symbol */
984 p_b_symbol(struct parse *p)
985 {
986 	char value;
987 
988 	REQUIRE(MORE(), REG_EBRACK);
989 	if (!EATTWO('[', '.'))
990 		return(GETNEXT());
991 
992 	/* collating symbol */
993 	value = p_b_coll_elem(p, '.');
994 	REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
995 	return(value);
996 }
997 
998 /*
999  - p_b_coll_elem - parse a collating-element name and look it up
1000  */
1001 static char			/* value of collating element */
1002 p_b_coll_elem(struct parse *p,
1003     int endc)			/* name ended by endc,']' */
1004 {
1005 	char *sp = p->next;
1006 	struct cname *cp;
1007 	size_t len;
1008 
1009 	while (MORE() && !SEETWO(endc, ']'))
1010 		NEXT();
1011 	if (!MORE()) {
1012 		SETERROR(REG_EBRACK);
1013 		return(0);
1014 	}
1015 	len = p->next - sp;
1016 	for (cp = cnames; cp->name != NULL; cp++)
1017 		if (strncmp(cp->name, sp, len) == 0 && strlen(cp->name) == len)
1018 			return(cp->code);	/* known name */
1019 	if (len == 1)
1020 		return(*sp);	/* single character */
1021 	SETERROR(REG_ECOLLATE);			/* neither */
1022 	return(0);
1023 }
1024 
1025 /*
1026  - othercase - return the case counterpart of an alphabetic
1027  */
1028 static char			/* if no counterpart, return ch */
1029 othercase(int ch)
1030 {
1031 	ch = (uch)ch;
1032 	assert(isalpha(ch));
1033 	if (isupper(ch))
1034 		return ((uch)tolower(ch));
1035 	else if (islower(ch))
1036 		return ((uch)toupper(ch));
1037 	else			/* peculiar, but could happen */
1038 		return(ch);
1039 }
1040 
1041 /*
1042  - bothcases - emit a dualcase version of a two-case character
1043  *
1044  * Boy, is this implementation ever a kludge...
1045  */
1046 static void
1047 bothcases(struct parse *p, int ch)
1048 {
1049 	char *oldnext = p->next;
1050 	char *oldend = p->end;
1051 	char bracket[3];
1052 
1053 	ch = (uch)ch;
1054 	assert(othercase(ch) != ch);	/* p_bracket() would recurse */
1055 	p->next = bracket;
1056 	p->end = bracket+2;
1057 	bracket[0] = ch;
1058 	bracket[1] = ']';
1059 	bracket[2] = '\0';
1060 	p_bracket(p);
1061 	assert(p->next == bracket+2);
1062 	p->next = oldnext;
1063 	p->end = oldend;
1064 }
1065 
1066 /*
1067  - ordinary - emit an ordinary character
1068  */
1069 static void
1070 ordinary(struct parse *p, int ch)
1071 {
1072 	cat_t *cap = p->g->categories;
1073 
1074 	if ((p->g->cflags&REG_ICASE) && isalpha((uch)ch) && othercase(ch) != ch)
1075 		bothcases(p, ch);
1076 	else {
1077 		EMIT(OCHAR, (uch)ch);
1078 		if (cap[ch] == 0)
1079 			cap[ch] = p->g->ncategories++;
1080 	}
1081 }
1082 
1083 /*
1084  - nonnewline - emit REG_NEWLINE version of OANY
1085  *
1086  * Boy, is this implementation ever a kludge...
1087  */
1088 static void
1089 nonnewline(struct parse *p)
1090 {
1091 	char *oldnext = p->next;
1092 	char *oldend = p->end;
1093 	char bracket[4];
1094 
1095 	p->next = bracket;
1096 	p->end = bracket+3;
1097 	bracket[0] = '^';
1098 	bracket[1] = '\n';
1099 	bracket[2] = ']';
1100 	bracket[3] = '\0';
1101 	p_bracket(p);
1102 	assert(p->next == bracket+3);
1103 	p->next = oldnext;
1104 	p->end = oldend;
1105 }
1106 
1107 /*
1108  - repeat - generate code for a bounded repetition, recursively if needed
1109  */
1110 static void
1111 repeat(struct parse *p,
1112     sopno start,		/* operand from here to end of strip */
1113     int from,			/* repeated from this number */
1114     int to)			/* to this number of times (maybe INFINITY) */
1115 {
1116 	sopno finish = HERE();
1117 #	define	N	2
1118 #	define	INF	3
1119 #	define	REP(f, t)	((f)*8 + (t))
1120 #	define	MAP(n)	(((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1121 	sopno copy;
1122 
1123 	if (p->error != 0)	/* head off possible runaway recursion */
1124 		return;
1125 
1126 	assert(from <= to);
1127 
1128 	switch (REP(MAP(from), MAP(to))) {
1129 	case REP(0, 0):			/* must be user doing this */
1130 		DROP(finish-start);	/* drop the operand */
1131 		break;
1132 	case REP(0, 1):			/* as x{1,1}? */
1133 	case REP(0, N):			/* as x{1,n}? */
1134 	case REP(0, INF):		/* as x{1,}? */
1135 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1136 		INSERT(OCH_, start);		/* offset is wrong... */
1137 		repeat(p, start+1, 1, to);
1138 		ASTERN(OOR1, start);
1139 		AHEAD(start);			/* ... fix it */
1140 		EMIT(OOR2, 0);
1141 		AHEAD(THERE());
1142 		ASTERN(O_CH, THERETHERE());
1143 		break;
1144 	case REP(1, 1):			/* trivial case */
1145 		/* done */
1146 		break;
1147 	case REP(1, N):			/* as x?x{1,n-1} */
1148 		/* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1149 		INSERT(OCH_, start);
1150 		ASTERN(OOR1, start);
1151 		AHEAD(start);
1152 		EMIT(OOR2, 0);			/* offset very wrong... */
1153 		AHEAD(THERE());			/* ...so fix it */
1154 		ASTERN(O_CH, THERETHERE());
1155 		copy = dupl(p, start+1, finish+1);
1156 		assert(copy == finish+4);
1157 		repeat(p, copy, 1, to-1);
1158 		break;
1159 	case REP(1, INF):		/* as x+ */
1160 		INSERT(OPLUS_, start);
1161 		ASTERN(O_PLUS, start);
1162 		break;
1163 	case REP(N, N):			/* as xx{m-1,n-1} */
1164 		copy = dupl(p, start, finish);
1165 		repeat(p, copy, from-1, to-1);
1166 		break;
1167 	case REP(N, INF):		/* as xx{n-1,INF} */
1168 		copy = dupl(p, start, finish);
1169 		repeat(p, copy, from-1, to);
1170 		break;
1171 	default:			/* "can't happen" */
1172 		SETERROR(REG_ASSERT);	/* just in case */
1173 		break;
1174 	}
1175 }
1176 
1177 /*
1178  - seterr - set an error condition
1179  */
1180 static int			/* useless but makes type checking happy */
1181 seterr(struct parse *p, int e)
1182 {
1183 	if (p->error == 0)	/* keep earliest error condition */
1184 		p->error = e;
1185 	p->next = nuls;		/* try to bring things to a halt */
1186 	p->end = nuls;
1187 	return(0);		/* make the return value well-defined */
1188 }
1189 
1190 /*
1191  - allocset - allocate a set of characters for []
1192  */
1193 static cset *
1194 allocset(struct parse *p)
1195 {
1196 	int no = p->g->ncsets++;
1197 	size_t nc;
1198 	size_t nbytes;
1199 	cset *cs;
1200 	size_t css = (size_t)p->g->csetsize;
1201 	int i;
1202 
1203 	if (no >= p->ncsalloc) {	/* need another column of space */
1204 		void *ptr;
1205 
1206 		p->ncsalloc += CHAR_BIT;
1207 		nc = p->ncsalloc;
1208 		if (nc > SIZE_MAX / sizeof(cset))
1209 			goto nomem;
1210 		assert(nc % CHAR_BIT == 0);
1211 		nbytes = nc / CHAR_BIT * css;
1212 
1213 		ptr = (cset *)realloc((char *)p->g->sets, nc * sizeof(cset));
1214 		if (ptr == NULL)
1215 			goto nomem;
1216 		p->g->sets = ptr;
1217 
1218 		ptr = (uch *)realloc((char *)p->g->setbits, nbytes);
1219 		if (ptr == NULL)
1220 			goto nomem;
1221 		p->g->setbits = ptr;
1222 
1223 		for (i = 0; i < no; i++)
1224 			p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
1225 
1226 		(void) memset((char *)p->g->setbits + (nbytes - css), 0, css);
1227 	}
1228 	/* XXX should not happen */
1229 	if (p->g->sets == NULL || p->g->setbits == NULL)
1230 		goto nomem;
1231 
1232 	cs = &p->g->sets[no];
1233 	cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
1234 	cs->mask = 1 << ((no) % CHAR_BIT);
1235 	cs->hash = 0;
1236 	cs->smultis = 0;
1237 	cs->multis = NULL;
1238 
1239 	return(cs);
1240 nomem:
1241 	free(p->g->sets);
1242 	p->g->sets = NULL;
1243 	free(p->g->setbits);
1244 	p->g->setbits = NULL;
1245 
1246 	SETERROR(REG_ESPACE);
1247 	/* caller's responsibility not to do set ops */
1248 	return(NULL);
1249 }
1250 
1251 /*
1252  - freeset - free a now-unused set
1253  */
1254 static void
1255 freeset(struct parse *p, cset *cs)
1256 {
1257 	size_t i;
1258 	cset *top = &p->g->sets[p->g->ncsets];
1259 	size_t css = (size_t)p->g->csetsize;
1260 
1261 	for (i = 0; i < css; i++)
1262 		CHsub(cs, i);
1263 	if (cs == top-1)	/* recover only the easy case */
1264 		p->g->ncsets--;
1265 }
1266 
1267 /*
1268  - freezeset - final processing on a set of characters
1269  *
1270  * The main task here is merging identical sets.  This is usually a waste
1271  * of time (although the hash code minimizes the overhead), but can win
1272  * big if REG_ICASE is being used.  REG_ICASE, by the way, is why the hash
1273  * is done using addition rather than xor -- all ASCII [aA] sets xor to
1274  * the same value!
1275  */
1276 static int			/* set number */
1277 freezeset(struct parse *p, cset *cs)
1278 {
1279 	uch h = cs->hash;
1280 	size_t i;
1281 	cset *top = &p->g->sets[p->g->ncsets];
1282 	cset *cs2;
1283 	size_t css = (size_t)p->g->csetsize;
1284 
1285 	/* look for an earlier one which is the same */
1286 	for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
1287 		if (cs2->hash == h && cs2 != cs) {
1288 			/* maybe */
1289 			for (i = 0; i < css; i++)
1290 				if (!!CHIN(cs2, i) != !!CHIN(cs, i))
1291 					break;		/* no */
1292 			if (i == css)
1293 				break;			/* yes */
1294 		}
1295 
1296 	if (cs2 < top) {	/* found one */
1297 		freeset(p, cs);
1298 		cs = cs2;
1299 	}
1300 
1301 	return((int)(cs - p->g->sets));
1302 }
1303 
1304 /*
1305  - firstch - return first character in a set (which must have at least one)
1306  */
1307 static int			/* character; there is no "none" value */
1308 firstch(struct parse *p, cset *cs)
1309 {
1310 	size_t i;
1311 	size_t css = (size_t)p->g->csetsize;
1312 
1313 	for (i = 0; i < css; i++)
1314 		if (CHIN(cs, i))
1315 			return((char)i);
1316 	assert(never);
1317 	return(0);		/* arbitrary */
1318 }
1319 
1320 /*
1321  - nch - number of characters in a set
1322  */
1323 static int
1324 nch(struct parse *p, cset *cs)
1325 {
1326 	size_t i;
1327 	size_t css = (size_t)p->g->csetsize;
1328 	int n = 0;
1329 
1330 	for (i = 0; i < css; i++)
1331 		if (CHIN(cs, i))
1332 			n++;
1333 	return(n);
1334 }
1335 
1336 /*
1337  - mcadd - add a collating element to a cset
1338  */
1339 static void
1340 mcadd( struct parse *p, cset *cs, const char *cp)
1341 {
1342 	size_t oldend = cs->smultis;
1343 	void *np;
1344 
1345 	cs->smultis += strlen(cp) + 1;
1346 	np = realloc(cs->multis, cs->smultis);
1347 	if (np == NULL) {
1348 		if (cs->multis)
1349 			free(cs->multis);
1350 		cs->multis = NULL;
1351 		SETERROR(REG_ESPACE);
1352 		return;
1353 	}
1354 	cs->multis = np;
1355 
1356 	llvm_strlcpy(cs->multis + oldend - 1, cp, cs->smultis - oldend + 1);
1357 }
1358 
1359 /*
1360  - mcinvert - invert the list of collating elements in a cset
1361  *
1362  * This would have to know the set of possibilities.  Implementation
1363  * is deferred.
1364  */
1365 /* ARGSUSED */
1366 static void
1367 mcinvert(struct parse *p, cset *cs)
1368 {
1369 	assert(cs->multis == NULL);	/* xxx */
1370 }
1371 
1372 /*
1373  - mccase - add case counterparts of the list of collating elements in a cset
1374  *
1375  * This would have to know the set of possibilities.  Implementation
1376  * is deferred.
1377  */
1378 /* ARGSUSED */
1379 static void
1380 mccase(struct parse *p, cset *cs)
1381 {
1382 	assert(cs->multis == NULL);	/* xxx */
1383 }
1384 
1385 /*
1386  - isinsets - is this character in any sets?
1387  */
1388 static int			/* predicate */
1389 isinsets(struct re_guts *g, int c)
1390 {
1391 	uch *col;
1392 	int i;
1393 	int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1394 	unsigned uc = (uch)c;
1395 
1396 	for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1397 		if (col[uc] != 0)
1398 			return(1);
1399 	return(0);
1400 }
1401 
1402 /*
1403  - samesets - are these two characters in exactly the same sets?
1404  */
1405 static int			/* predicate */
1406 samesets(struct re_guts *g, int c1, int c2)
1407 {
1408 	uch *col;
1409 	int i;
1410 	int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1411 	unsigned uc1 = (uch)c1;
1412 	unsigned uc2 = (uch)c2;
1413 
1414 	for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1415 		if (col[uc1] != col[uc2])
1416 			return(0);
1417 	return(1);
1418 }
1419 
1420 /*
1421  - categorize - sort out character categories
1422  */
1423 static void
1424 categorize(struct parse *p, struct re_guts *g)
1425 {
1426 	cat_t *cats = g->categories;
1427 	int c;
1428 	int c2;
1429 	cat_t cat;
1430 
1431 	/* avoid making error situations worse */
1432 	if (p->error != 0)
1433 		return;
1434 
1435 	for (c = CHAR_MIN; c <= CHAR_MAX; c++)
1436 		if (cats[c] == 0 && isinsets(g, c)) {
1437 			cat = g->ncategories++;
1438 			cats[c] = cat;
1439 			for (c2 = c+1; c2 <= CHAR_MAX; c2++)
1440 				if (cats[c2] == 0 && samesets(g, c, c2))
1441 					cats[c2] = cat;
1442 		}
1443 }
1444 
1445 /*
1446  - dupl - emit a duplicate of a bunch of sops
1447  */
1448 static sopno			/* start of duplicate */
1449 dupl(struct parse *p,
1450     sopno start,		/* from here */
1451     sopno finish)		/* to this less one */
1452 {
1453 	sopno ret = HERE();
1454 	sopno len = finish - start;
1455 
1456 	assert(finish >= start);
1457 	if (len == 0)
1458 		return(ret);
1459 	enlarge(p, p->ssize + len);	/* this many unexpected additions */
1460 	assert(p->ssize >= p->slen + len);
1461 	(void) memmove((char *)(p->strip + p->slen),
1462 		(char *)(p->strip + start), (size_t)len*sizeof(sop));
1463 	p->slen += len;
1464 	return(ret);
1465 }
1466 
1467 /*
1468  - doemit - emit a strip operator
1469  *
1470  * It might seem better to implement this as a macro with a function as
1471  * hard-case backup, but it's just too big and messy unless there are
1472  * some changes to the data structures.  Maybe later.
1473  */
1474 static void
1475 doemit(struct parse *p, sop op, size_t opnd)
1476 {
1477 	/* avoid making error situations worse */
1478 	if (p->error != 0)
1479 		return;
1480 
1481 	/* deal with oversize operands ("can't happen", more or less) */
1482 	assert(opnd < 1<<OPSHIFT);
1483 
1484 	/* deal with undersized strip */
1485 	if (p->slen >= p->ssize)
1486 		enlarge(p, (p->ssize+1) / 2 * 3);	/* +50% */
1487 	assert(p->slen < p->ssize);
1488 
1489 	/* finally, it's all reduced to the easy case */
1490 	p->strip[p->slen++] = SOP(op, opnd);
1491 }
1492 
1493 /*
1494  - doinsert - insert a sop into the strip
1495  */
1496 static void
1497 doinsert(struct parse *p, sop op, size_t opnd, sopno pos)
1498 {
1499 	sopno sn;
1500 	sop s;
1501 	int i;
1502 
1503 	/* avoid making error situations worse */
1504 	if (p->error != 0)
1505 		return;
1506 
1507 	sn = HERE();
1508 	EMIT(op, opnd);		/* do checks, ensure space */
1509 	assert(HERE() == sn+1);
1510 	s = p->strip[sn];
1511 
1512 	/* adjust paren pointers */
1513 	assert(pos > 0);
1514 	for (i = 1; i < NPAREN; i++) {
1515 		if (p->pbegin[i] >= pos) {
1516 			p->pbegin[i]++;
1517 		}
1518 		if (p->pend[i] >= pos) {
1519 			p->pend[i]++;
1520 		}
1521 	}
1522 
1523 	memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1524 						(HERE()-pos-1)*sizeof(sop));
1525 	p->strip[pos] = s;
1526 }
1527 
1528 /*
1529  - dofwd - complete a forward reference
1530  */
1531 static void
1532 dofwd(struct parse *p, sopno pos, sop value)
1533 {
1534 	/* avoid making error situations worse */
1535 	if (p->error != 0)
1536 		return;
1537 
1538 	assert(value < 1<<OPSHIFT);
1539 	p->strip[pos] = OP(p->strip[pos]) | value;
1540 }
1541 
1542 /*
1543  - enlarge - enlarge the strip
1544  */
1545 static void
1546 enlarge(struct parse *p, sopno size)
1547 {
1548 	sop *sp;
1549 
1550 	if (p->ssize >= size)
1551 		return;
1552 
1553 	if ((uintptr_t)size > SIZE_MAX / sizeof(sop)) {
1554 		SETERROR(REG_ESPACE);
1555 		return;
1556 	}
1557 
1558 	sp = (sop *)realloc(p->strip, size*sizeof(sop));
1559 	if (sp == NULL) {
1560 		SETERROR(REG_ESPACE);
1561 		return;
1562 	}
1563 	p->strip = sp;
1564 	p->ssize = size;
1565 }
1566 
1567 /*
1568  - stripsnug - compact the strip
1569  */
1570 static void
1571 stripsnug(struct parse *p, struct re_guts *g)
1572 {
1573 	g->nstates = p->slen;
1574 	if ((uintptr_t)p->slen > SIZE_MAX / sizeof(sop)) {
1575 		g->strip = p->strip;
1576 		SETERROR(REG_ESPACE);
1577 		return;
1578 	}
1579 
1580 	g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1581 	if (g->strip == NULL) {
1582 		SETERROR(REG_ESPACE);
1583 		g->strip = p->strip;
1584 	}
1585 }
1586 
1587 /*
1588  - findmust - fill in must and mlen with longest mandatory literal string
1589  *
1590  * This algorithm could do fancy things like analyzing the operands of |
1591  * for common subsequences.  Someday.  This code is simple and finds most
1592  * of the interesting cases.
1593  *
1594  * Note that must and mlen got initialized during setup.
1595  */
1596 static void
1597 findmust(struct parse *p, struct re_guts *g)
1598 {
1599 	sop *scan;
1600 	sop *start = 0; /* start initialized in the default case, after that */
1601 	sop *newstart = 0; /* newstart was initialized in the OCHAR case */
1602 	sopno newlen;
1603 	sop s;
1604 	char *cp;
1605 	sopno i;
1606 
1607 	/* avoid making error situations worse */
1608 	if (p->error != 0)
1609 		return;
1610 
1611 	/* find the longest OCHAR sequence in strip */
1612 	newlen = 0;
1613 	scan = g->strip + 1;
1614 	do {
1615 		s = *scan++;
1616 		switch (OP(s)) {
1617 		case OCHAR:		/* sequence member */
1618 			if (newlen == 0)		/* new sequence */
1619 				newstart = scan - 1;
1620 			newlen++;
1621 			break;
1622 		case OPLUS_:		/* things that don't break one */
1623 		case OLPAREN:
1624 		case ORPAREN:
1625 			break;
1626 		case OQUEST_:		/* things that must be skipped */
1627 		case OCH_:
1628 			scan--;
1629 			do {
1630 				scan += OPND(s);
1631 				s = *scan;
1632 				/* assert() interferes w debug printouts */
1633 				if (OP(s) != O_QUEST && OP(s) != O_CH &&
1634 							OP(s) != OOR2) {
1635 					g->iflags |= REGEX_BAD;
1636 					return;
1637 				}
1638 			} while (OP(s) != O_QUEST && OP(s) != O_CH);
1639 			LLVM_FALLTHROUGH;
1640 		default:		/* things that break a sequence */
1641 			if (newlen > g->mlen) {		/* ends one */
1642 				start = newstart;
1643 				g->mlen = newlen;
1644 			}
1645 			newlen = 0;
1646 			break;
1647 		}
1648 	} while (OP(s) != OEND);
1649 
1650 	if (g->mlen == 0)		/* there isn't one */
1651 		return;
1652 
1653 	/* turn it into a character string */
1654 	g->must = malloc((size_t)g->mlen + 1);
1655 	if (g->must == NULL) {		/* argh; just forget it */
1656 		g->mlen = 0;
1657 		return;
1658 	}
1659 	cp = g->must;
1660 	scan = start;
1661 	for (i = g->mlen; i > 0; i--) {
1662 		while (OP(s = *scan++) != OCHAR)
1663 			continue;
1664 		assert(cp < g->must + g->mlen);
1665 		*cp++ = (char)OPND(s);
1666 	}
1667 	assert(cp == g->must + g->mlen);
1668 	*cp++ = '\0';		/* just on general principles */
1669 }
1670 
1671 /*
1672  - pluscount - count + nesting
1673  */
1674 static sopno			/* nesting depth */
1675 pluscount(struct parse *p, struct re_guts *g)
1676 {
1677 	sop *scan;
1678 	sop s;
1679 	sopno plusnest = 0;
1680 	sopno maxnest = 0;
1681 
1682 	if (p->error != 0)
1683 		return(0);	/* there may not be an OEND */
1684 
1685 	scan = g->strip + 1;
1686 	do {
1687 		s = *scan++;
1688 		switch (OP(s)) {
1689 		case OPLUS_:
1690 			plusnest++;
1691 			break;
1692 		case O_PLUS:
1693 			if (plusnest > maxnest)
1694 				maxnest = plusnest;
1695 			plusnest--;
1696 			break;
1697 		}
1698 	} while (OP(s) != OEND);
1699 	if (plusnest != 0)
1700 		g->iflags |= REGEX_BAD;
1701 	return(maxnest);
1702 }
1703