1/*-
2 * Copyright (c) 1992, 1993, 1994 Henry Spencer.
3 * Copyright (c) 1992, 1993, 1994
4 * The Regents of the University of California. All rights reserved.
5 *
6 * Copyright (c) 2011 The FreeBSD Foundation
7 * All rights reserved.
8 * Portions of this software were developed by David Chisnall
9 * under sponsorship from the FreeBSD Foundation.
10 *
11 * This code is derived from software contributed to Berkeley by
12 * Henry Spencer.
13 *
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
16 * are met:
17 * 1. Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * 2. Redistributions in binary form must reproduce the above copyright
20 * notice, this list of conditions and the following disclaimer in the
21 * documentation and/or other materials provided with the distribution.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 * @(#)regcomp.c 8.5 (Berkeley) 3/20/94
39 */
40
41#if defined(LIBC_SCCS) && !defined(lint)
42static char sccsid[] = "@(#)regcomp.c 8.5 (Berkeley) 3/20/94";
43#endif /* LIBC_SCCS and not lint */
44#include <sys/cdefs.h>
| 1/*-
2 * Copyright (c) 1992, 1993, 1994 Henry Spencer.
3 * Copyright (c) 1992, 1993, 1994
4 * The Regents of the University of California. All rights reserved.
5 *
6 * Copyright (c) 2011 The FreeBSD Foundation
7 * All rights reserved.
8 * Portions of this software were developed by David Chisnall
9 * under sponsorship from the FreeBSD Foundation.
10 *
11 * This code is derived from software contributed to Berkeley by
12 * Henry Spencer.
13 *
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
16 * are met:
17 * 1. Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * 2. Redistributions in binary form must reproduce the above copyright
20 * notice, this list of conditions and the following disclaimer in the
21 * documentation and/or other materials provided with the distribution.
22 * 4. Neither the name of the University nor the names of its contributors
23 * may be used to endorse or promote products derived from this software
24 * without specific prior written permission.
25 *
26 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
27 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
30 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
31 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
32 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
33 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
34 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
35 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
36 * SUCH DAMAGE.
37 *
38 * @(#)regcomp.c 8.5 (Berkeley) 3/20/94
39 */
40
41#if defined(LIBC_SCCS) && !defined(lint)
42static char sccsid[] = "@(#)regcomp.c 8.5 (Berkeley) 3/20/94";
43#endif /* LIBC_SCCS and not lint */
44#include <sys/cdefs.h>
|
45__FBSDID("$FreeBSD: head/lib/libc/regex/regcomp.c 298830 2016-04-30 01:24:24Z pfg $");
| 45__FBSDID("$FreeBSD: head/lib/libc/regex/regcomp.c 301461 2016-06-05 19:12:52Z pfg $");
|
46
47#include <sys/types.h>
48#include <stdio.h>
49#include <string.h>
50#include <ctype.h>
51#include <limits.h>
52#include <stdlib.h>
53#include <regex.h>
54#include <runetype.h>
55#include <wchar.h>
56#include <wctype.h>
57
58#include "collate.h"
59
60#include "utils.h"
61#include "regex2.h"
62
63#include "cname.h"
64
65/*
66 * parse structure, passed up and down to avoid global variables and
67 * other clumsinesses
68 */
69struct parse {
70 char *next; /* next character in RE */
71 char *end; /* end of string (-> NUL normally) */
72 int error; /* has an error been seen? */
73 sop *strip; /* malloced strip */
74 sopno ssize; /* malloced strip size (allocated) */
75 sopno slen; /* malloced strip length (used) */
76 int ncsalloc; /* number of csets allocated */
77 struct re_guts *g;
78# define NPAREN 10 /* we need to remember () 1-9 for back refs */
79 sopno pbegin[NPAREN]; /* -> ( ([0] unused) */
80 sopno pend[NPAREN]; /* -> ) ([0] unused) */
81};
82
83/* ========= begin header generated by ./mkh ========= */
84#ifdef __cplusplus
85extern "C" {
86#endif
87
88/* === regcomp.c === */
89static void p_ere(struct parse *p, int stop);
90static void p_ere_exp(struct parse *p);
91static void p_str(struct parse *p);
92static void p_bre(struct parse *p, int end1, int end2);
93static int p_simp_re(struct parse *p, int starordinary);
94static int p_count(struct parse *p);
95static void p_bracket(struct parse *p);
96static void p_b_term(struct parse *p, cset *cs);
97static void p_b_cclass(struct parse *p, cset *cs);
98static void p_b_eclass(struct parse *p, cset *cs);
99static wint_t p_b_symbol(struct parse *p);
100static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
101static wint_t othercase(wint_t ch);
102static void bothcases(struct parse *p, wint_t ch);
103static void ordinary(struct parse *p, wint_t ch);
104static void nonnewline(struct parse *p);
105static void repeat(struct parse *p, sopno start, int from, int to);
106static int seterr(struct parse *p, int e);
107static cset *allocset(struct parse *p);
108static void freeset(struct parse *p, cset *cs);
109static void CHadd(struct parse *p, cset *cs, wint_t ch);
110static void CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max);
111static void CHaddtype(struct parse *p, cset *cs, wctype_t wct);
112static wint_t singleton(cset *cs);
113static sopno dupl(struct parse *p, sopno start, sopno finish);
114static void doemit(struct parse *p, sop op, size_t opnd);
115static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
116static void dofwd(struct parse *p, sopno pos, sop value);
117static int enlarge(struct parse *p, sopno size);
118static void stripsnug(struct parse *p, struct re_guts *g);
119static void findmust(struct parse *p, struct re_guts *g);
120static int altoffset(sop *scan, int offset);
121static void computejumps(struct parse *p, struct re_guts *g);
122static void computematchjumps(struct parse *p, struct re_guts *g);
123static sopno pluscount(struct parse *p, struct re_guts *g);
124static wint_t wgetnext(struct parse *p);
125
126#ifdef __cplusplus
127}
128#endif
129/* ========= end header generated by ./mkh ========= */
130
131static char nuls[10]; /* place to point scanner in event of error */
132
133/*
134 * macros for use with parse structure
135 * BEWARE: these know that the parse structure is named `p' !!!
136 */
137#define PEEK() (*p->next)
138#define PEEK2() (*(p->next+1))
139#define MORE() (p->next < p->end)
140#define MORE2() (p->next+1 < p->end)
141#define SEE(c) (MORE() && PEEK() == (c))
142#define SEETWO(a, b) (MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
143#define EAT(c) ((SEE(c)) ? (NEXT(), 1) : 0)
144#define EATTWO(a, b) ((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
145#define NEXT() (p->next++)
146#define NEXT2() (p->next += 2)
147#define NEXTn(n) (p->next += (n))
148#define GETNEXT() (*p->next++)
149#define WGETNEXT() wgetnext(p)
150#define SETERROR(e) seterr(p, (e))
151#define REQUIRE(co, e) ((co) || SETERROR(e))
152#define MUSTSEE(c, e) (REQUIRE(MORE() && PEEK() == (c), e))
153#define MUSTEAT(c, e) (REQUIRE(MORE() && GETNEXT() == (c), e))
154#define MUSTNOTSEE(c, e) (REQUIRE(!MORE() || PEEK() != (c), e))
155#define EMIT(op, sopnd) doemit(p, (sop)(op), (size_t)(sopnd))
156#define INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
157#define AHEAD(pos) dofwd(p, pos, HERE()-(pos))
158#define ASTERN(sop, pos) EMIT(sop, HERE()-pos)
159#define HERE() (p->slen)
160#define THERE() (p->slen - 1)
161#define THERETHERE() (p->slen - 2)
162#define DROP(n) (p->slen -= (n))
163
164#ifndef NDEBUG
165static int never = 0; /* for use in asserts; shuts lint up */
166#else
167#define never 0 /* some <assert.h>s have bugs too */
168#endif
169
170/* Macro used by computejump()/computematchjump() */
171#define MIN(a,b) ((a)<(b)?(a):(b))
172
173/*
174 - regcomp - interface for parser and compilation
175 = extern int regcomp(regex_t *, const char *, int);
176 = #define REG_BASIC 0000
177 = #define REG_EXTENDED 0001
178 = #define REG_ICASE 0002
179 = #define REG_NOSUB 0004
180 = #define REG_NEWLINE 0010
181 = #define REG_NOSPEC 0020
182 = #define REG_PEND 0040
183 = #define REG_DUMP 0200
184 */
185int /* 0 success, otherwise REG_something */
186regcomp(regex_t * __restrict preg,
187 const char * __restrict pattern,
188 int cflags)
189{
190 struct parse pa;
191 struct re_guts *g;
192 struct parse *p = &pa;
193 int i;
194 size_t len;
195 size_t maxlen;
196#ifdef REDEBUG
197# define GOODFLAGS(f) (f)
198#else
199# define GOODFLAGS(f) ((f)&~REG_DUMP)
200#endif
201
202 cflags = GOODFLAGS(cflags);
203 if ((cflags®_EXTENDED) && (cflags®_NOSPEC))
204 return(REG_INVARG);
205
206 if (cflags®_PEND) {
207 if (preg->re_endp < pattern)
208 return(REG_INVARG);
209 len = preg->re_endp - pattern;
210 } else
211 len = strlen((char *)pattern);
212
213 /* do the mallocs early so failure handling is easy */
214 g = (struct re_guts *)malloc(sizeof(struct re_guts));
215 if (g == NULL)
216 return(REG_ESPACE);
217 /*
218 * Limit the pattern space to avoid a 32-bit overflow on buffer
219 * extension. Also avoid any signed overflow in case of conversion
220 * so make the real limit based on a 31-bit overflow.
221 *
222 * Likely not applicable on 64-bit systems but handle the case
223 * generically (who are we to stop people from using ~715MB+
224 * patterns?).
225 */
226 maxlen = ((size_t)-1 >> 1) / sizeof(sop) * 2 / 3;
227 if (len >= maxlen) {
228 free((char *)g);
229 return(REG_ESPACE);
230 }
231 p->ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */
232 assert(p->ssize >= len);
233
234 p->strip = (sop *)malloc(p->ssize * sizeof(sop));
235 p->slen = 0;
236 if (p->strip == NULL) {
237 free((char *)g);
238 return(REG_ESPACE);
239 }
240
241 /* set things up */
242 p->g = g;
243 p->next = (char *)pattern; /* convenience; we do not modify it */
244 p->end = p->next + len;
245 p->error = 0;
246 p->ncsalloc = 0;
247 for (i = 0; i < NPAREN; i++) {
248 p->pbegin[i] = 0;
249 p->pend[i] = 0;
250 }
251 g->sets = NULL;
252 g->ncsets = 0;
253 g->cflags = cflags;
254 g->iflags = 0;
255 g->nbol = 0;
256 g->neol = 0;
257 g->must = NULL;
258 g->moffset = -1;
259 g->charjump = NULL;
260 g->matchjump = NULL;
261 g->mlen = 0;
262 g->nsub = 0;
263 g->backrefs = 0;
264
265 /* do it */
266 EMIT(OEND, 0);
267 g->firststate = THERE();
268 if (cflags®_EXTENDED)
269 p_ere(p, OUT);
270 else if (cflags®_NOSPEC)
271 p_str(p);
272 else
273 p_bre(p, OUT, OUT);
274 EMIT(OEND, 0);
275 g->laststate = THERE();
276
277 /* tidy up loose ends and fill things in */
278 stripsnug(p, g);
279 findmust(p, g);
280 /* only use Boyer-Moore algorithm if the pattern is bigger
281 * than three characters
282 */
283 if(g->mlen > 3) {
284 computejumps(p, g);
285 computematchjumps(p, g);
286 if(g->matchjump == NULL && g->charjump != NULL) {
287 free(g->charjump);
288 g->charjump = NULL;
289 }
290 }
291 g->nplus = pluscount(p, g);
292 g->magic = MAGIC2;
293 preg->re_nsub = g->nsub;
294 preg->re_g = g;
295 preg->re_magic = MAGIC1;
296#ifndef REDEBUG
297 /* not debugging, so can't rely on the assert() in regexec() */
298 if (g->iflags&BAD)
299 SETERROR(REG_ASSERT);
300#endif
301
302 /* win or lose, we're done */
303 if (p->error != 0) /* lose */
304 regfree(preg);
305 return(p->error);
306}
307
308/*
309 - p_ere - ERE parser top level, concatenation and alternation
310 == static void p_ere(struct parse *p, int_t stop);
311 */
312static void
313p_ere(struct parse *p,
314 int stop) /* character this ERE should end at */
315{
316 char c;
317 sopno prevback;
318 sopno prevfwd;
319 sopno conc;
320 int first = 1; /* is this the first alternative? */
321
322 for (;;) {
323 /* do a bunch of concatenated expressions */
324 conc = HERE();
325 while (MORE() && (c = PEEK()) != '|' && c != stop)
326 p_ere_exp(p);
327 (void)REQUIRE(HERE() != conc, REG_EMPTY); /* require nonempty */
328
329 if (!EAT('|'))
330 break; /* NOTE BREAK OUT */
331
332 if (first) {
333 INSERT(OCH_, conc); /* offset is wrong */
334 prevfwd = conc;
335 prevback = conc;
336 first = 0;
337 }
338 ASTERN(OOR1, prevback);
339 prevback = THERE();
340 AHEAD(prevfwd); /* fix previous offset */
341 prevfwd = HERE();
342 EMIT(OOR2, 0); /* offset is very wrong */
343 }
344
345 if (!first) { /* tail-end fixups */
346 AHEAD(prevfwd);
347 ASTERN(O_CH, prevback);
348 }
349
350 assert(!MORE() || SEE(stop));
351}
352
353/*
354 - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
355 == static void p_ere_exp(struct parse *p);
356 */
357static void
358p_ere_exp(struct parse *p)
359{
360 char c;
361 wint_t wc;
362 sopno pos;
363 int count;
364 int count2;
365 sopno subno;
366 int wascaret = 0;
367
368 assert(MORE()); /* caller should have ensured this */
369 c = GETNEXT();
370
371 pos = HERE();
372 switch (c) {
373 case '(':
374 (void)REQUIRE(MORE(), REG_EPAREN);
375 p->g->nsub++;
376 subno = p->g->nsub;
377 if (subno < NPAREN)
378 p->pbegin[subno] = HERE();
379 EMIT(OLPAREN, subno);
380 if (!SEE(')'))
381 p_ere(p, ')');
382 if (subno < NPAREN) {
383 p->pend[subno] = HERE();
384 assert(p->pend[subno] != 0);
385 }
386 EMIT(ORPAREN, subno);
387 (void)MUSTEAT(')', REG_EPAREN);
388 break;
389#ifndef POSIX_MISTAKE
390 case ')': /* happens only if no current unmatched ( */
391 /*
392 * You may ask, why the ifndef? Because I didn't notice
393 * this until slightly too late for 1003.2, and none of the
394 * other 1003.2 regular-expression reviewers noticed it at
395 * all. So an unmatched ) is legal POSIX, at least until
396 * we can get it fixed.
397 */
398 SETERROR(REG_EPAREN);
399 break;
400#endif
401 case '^':
402 EMIT(OBOL, 0);
403 p->g->iflags |= USEBOL;
404 p->g->nbol++;
405 wascaret = 1;
406 break;
407 case '$':
408 EMIT(OEOL, 0);
409 p->g->iflags |= USEEOL;
410 p->g->neol++;
411 break;
412 case '|':
413 SETERROR(REG_EMPTY);
414 break;
415 case '*':
416 case '+':
417 case '?':
418 SETERROR(REG_BADRPT);
419 break;
420 case '.':
421 if (p->g->cflags®_NEWLINE)
422 nonnewline(p);
423 else
424 EMIT(OANY, 0);
425 break;
426 case '[':
427 p_bracket(p);
428 break;
429 case '\\':
430 (void)REQUIRE(MORE(), REG_EESCAPE);
431 wc = WGETNEXT();
432 switch (wc) {
433 case '<':
434 EMIT(OBOW, 0);
435 break;
436 case '>':
437 EMIT(OEOW, 0);
438 break;
439 default:
440 ordinary(p, wc);
441 break;
442 }
443 break;
444 case '{': /* okay as ordinary except if digit follows */
445 (void)REQUIRE(!MORE() || !isdigit((uch)PEEK()), REG_BADRPT);
446 /* FALLTHROUGH */
447 default:
448 p->next--;
449 wc = WGETNEXT();
450 ordinary(p, wc);
451 break;
452 }
453
454 if (!MORE())
455 return;
456 c = PEEK();
457 /* we call { a repetition if followed by a digit */
458 if (!( c == '*' || c == '+' || c == '?' ||
459 (c == '{' && MORE2() && isdigit((uch)PEEK2())) ))
460 return; /* no repetition, we're done */
461 NEXT();
462
463 (void)REQUIRE(!wascaret, REG_BADRPT);
464 switch (c) {
465 case '*': /* implemented as +? */
466 /* this case does not require the (y|) trick, noKLUDGE */
467 INSERT(OPLUS_, pos);
468 ASTERN(O_PLUS, pos);
469 INSERT(OQUEST_, pos);
470 ASTERN(O_QUEST, pos);
471 break;
472 case '+':
473 INSERT(OPLUS_, pos);
474 ASTERN(O_PLUS, pos);
475 break;
476 case '?':
477 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
478 INSERT(OCH_, pos); /* offset slightly wrong */
479 ASTERN(OOR1, pos); /* this one's right */
480 AHEAD(pos); /* fix the OCH_ */
481 EMIT(OOR2, 0); /* offset very wrong... */
482 AHEAD(THERE()); /* ...so fix it */
483 ASTERN(O_CH, THERETHERE());
484 break;
485 case '{':
486 count = p_count(p);
487 if (EAT(',')) {
488 if (isdigit((uch)PEEK())) {
489 count2 = p_count(p);
490 (void)REQUIRE(count <= count2, REG_BADBR);
491 } else /* single number with comma */
492 count2 = INFINITY;
493 } else /* just a single number */
494 count2 = count;
495 repeat(p, pos, count, count2);
496 if (!EAT('}')) { /* error heuristics */
497 while (MORE() && PEEK() != '}')
498 NEXT();
499 (void)REQUIRE(MORE(), REG_EBRACE);
500 SETERROR(REG_BADBR);
501 }
502 break;
503 }
504
505 if (!MORE())
506 return;
507 c = PEEK();
508 if (!( c == '*' || c == '+' || c == '?' ||
509 (c == '{' && MORE2() && isdigit((uch)PEEK2())) ) )
510 return;
511 SETERROR(REG_BADRPT);
512}
513
514/*
515 - p_str - string (no metacharacters) "parser"
516 == static void p_str(struct parse *p);
517 */
518static void
519p_str(struct parse *p)
520{
521 (void)REQUIRE(MORE(), REG_EMPTY);
522 while (MORE())
523 ordinary(p, WGETNEXT());
524}
525
526/*
527 - p_bre - BRE parser top level, anchoring and concatenation
528 == static void p_bre(struct parse *p, int end1, \
529 == int end2);
530 * Giving end1 as OUT essentially eliminates the end1/end2 check.
531 *
532 * This implementation is a bit of a kludge, in that a trailing $ is first
533 * taken as an ordinary character and then revised to be an anchor.
534 * The amount of lookahead needed to avoid this kludge is excessive.
535 */
536static void
537p_bre(struct parse *p,
538 int end1, /* first terminating character */
539 int end2) /* second terminating character */
540{
541 sopno start = HERE();
542 int first = 1; /* first subexpression? */
543 int wasdollar = 0;
544
545 if (EAT('^')) {
546 EMIT(OBOL, 0);
547 p->g->iflags |= USEBOL;
548 p->g->nbol++;
549 }
550 while (MORE() && !SEETWO(end1, end2)) {
551 wasdollar = p_simp_re(p, first);
552 first = 0;
553 }
554 if (wasdollar) { /* oops, that was a trailing anchor */
555 DROP(1);
556 EMIT(OEOL, 0);
557 p->g->iflags |= USEEOL;
558 p->g->neol++;
559 }
560
561 (void)REQUIRE(HERE() != start, REG_EMPTY); /* require nonempty */
562}
563
564/*
565 - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
566 == static int p_simp_re(struct parse *p, int starordinary);
567 */
568static int /* was the simple RE an unbackslashed $? */
569p_simp_re(struct parse *p,
570 int starordinary) /* is a leading * an ordinary character? */
571{
572 int c;
573 int count;
574 int count2;
575 sopno pos;
576 int i;
577 wint_t wc;
578 sopno subno;
579# define BACKSL (1<<CHAR_BIT)
580
581 pos = HERE(); /* repetition op, if any, covers from here */
582
583 assert(MORE()); /* caller should have ensured this */
584 c = GETNEXT();
585 if (c == '\\') {
586 (void)REQUIRE(MORE(), REG_EESCAPE);
587 c = BACKSL | GETNEXT();
588 }
589 switch (c) {
590 case '.':
591 if (p->g->cflags®_NEWLINE)
592 nonnewline(p);
593 else
594 EMIT(OANY, 0);
595 break;
596 case '[':
597 p_bracket(p);
598 break;
599 case BACKSL|'<':
600 EMIT(OBOW, 0);
601 break;
602 case BACKSL|'>':
603 EMIT(OEOW, 0);
604 break;
605 case BACKSL|'{':
606 SETERROR(REG_BADRPT);
607 break;
608 case BACKSL|'(':
609 p->g->nsub++;
610 subno = p->g->nsub;
611 if (subno < NPAREN)
612 p->pbegin[subno] = HERE();
613 EMIT(OLPAREN, subno);
614 /* the MORE here is an error heuristic */
615 if (MORE() && !SEETWO('\\', ')'))
616 p_bre(p, '\\', ')');
617 if (subno < NPAREN) {
618 p->pend[subno] = HERE();
619 assert(p->pend[subno] != 0);
620 }
621 EMIT(ORPAREN, subno);
622 (void)REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
623 break;
624 case BACKSL|')': /* should not get here -- must be user */
625 case BACKSL|'}':
626 SETERROR(REG_EPAREN);
627 break;
628 case BACKSL|'1':
629 case BACKSL|'2':
630 case BACKSL|'3':
631 case BACKSL|'4':
632 case BACKSL|'5':
633 case BACKSL|'6':
634 case BACKSL|'7':
635 case BACKSL|'8':
636 case BACKSL|'9':
637 i = (c&~BACKSL) - '0';
638 assert(i < NPAREN);
639 if (p->pend[i] != 0) {
640 assert(i <= p->g->nsub);
641 EMIT(OBACK_, i);
642 assert(p->pbegin[i] != 0);
643 assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
644 assert(OP(p->strip[p->pend[i]]) == ORPAREN);
645 (void) dupl(p, p->pbegin[i]+1, p->pend[i]);
646 EMIT(O_BACK, i);
647 } else
648 SETERROR(REG_ESUBREG);
649 p->g->backrefs = 1;
650 break;
651 case '*':
652 (void)REQUIRE(starordinary, REG_BADRPT);
653 /* FALLTHROUGH */
654 default:
655 p->next--;
656 wc = WGETNEXT();
657 ordinary(p, wc);
658 break;
659 }
660
661 if (EAT('*')) { /* implemented as +? */
662 /* this case does not require the (y|) trick, noKLUDGE */
663 INSERT(OPLUS_, pos);
664 ASTERN(O_PLUS, pos);
665 INSERT(OQUEST_, pos);
666 ASTERN(O_QUEST, pos);
667 } else if (EATTWO('\\', '{')) {
668 count = p_count(p);
669 if (EAT(',')) {
670 if (MORE() && isdigit((uch)PEEK())) {
671 count2 = p_count(p);
672 (void)REQUIRE(count <= count2, REG_BADBR);
673 } else /* single number with comma */
674 count2 = INFINITY;
675 } else /* just a single number */
676 count2 = count;
677 repeat(p, pos, count, count2);
678 if (!EATTWO('\\', '}')) { /* error heuristics */
679 while (MORE() && !SEETWO('\\', '}'))
680 NEXT();
681 (void)REQUIRE(MORE(), REG_EBRACE);
682 SETERROR(REG_BADBR);
683 }
684 } else if (c == '$') /* $ (but not \$) ends it */
685 return(1);
686
687 return(0);
688}
689
690/*
691 - p_count - parse a repetition count
692 == static int p_count(struct parse *p);
693 */
694static int /* the value */
695p_count(struct parse *p)
696{
697 int count = 0;
698 int ndigits = 0;
699
700 while (MORE() && isdigit((uch)PEEK()) && count <= DUPMAX) {
701 count = count*10 + (GETNEXT() - '0');
702 ndigits++;
703 }
704
705 (void)REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
706 return(count);
707}
708
709/*
710 - p_bracket - parse a bracketed character list
711 == static void p_bracket(struct parse *p);
712 */
713static void
714p_bracket(struct parse *p)
715{
716 cset *cs;
717 wint_t ch;
718
719 /* Dept of Truly Sickening Special-Case Kludges */
720 if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
721 EMIT(OBOW, 0);
722 NEXTn(6);
723 return;
724 }
725 if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
726 EMIT(OEOW, 0);
727 NEXTn(6);
728 return;
729 }
730
731 if ((cs = allocset(p)) == NULL)
732 return;
733
734 if (p->g->cflags®_ICASE)
735 cs->icase = 1;
736 if (EAT('^'))
737 cs->invert = 1;
738 if (EAT(']'))
739 CHadd(p, cs, ']');
740 else if (EAT('-'))
741 CHadd(p, cs, '-');
742 while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
743 p_b_term(p, cs);
744 if (EAT('-'))
745 CHadd(p, cs, '-');
746 (void)MUSTEAT(']', REG_EBRACK);
747
748 if (p->error != 0) /* don't mess things up further */
749 return;
750
751 if (cs->invert && p->g->cflags®_NEWLINE)
752 cs->bmp['\n' >> 3] |= 1 << ('\n' & 7);
753
754 if ((ch = singleton(cs)) != OUT) { /* optimize singleton sets */
755 ordinary(p, ch);
756 freeset(p, cs);
757 } else
758 EMIT(OANYOF, (int)(cs - p->g->sets));
759}
760
761/*
762 - p_b_term - parse one term of a bracketed character list
763 == static void p_b_term(struct parse *p, cset *cs);
764 */
765static void
766p_b_term(struct parse *p, cset *cs)
767{
768 char c;
769 wint_t start, finish;
770 wint_t i;
771 struct xlocale_collate *table =
772 (struct xlocale_collate*)__get_locale()->components[XLC_COLLATE];
773
774 /* classify what we've got */
775 switch ((MORE()) ? PEEK() : '\0') {
776 case '[':
777 c = (MORE2()) ? PEEK2() : '\0';
778 break;
779 case '-':
780 SETERROR(REG_ERANGE);
781 return; /* NOTE RETURN */
782 default:
783 c = '\0';
784 break;
785 }
786
787 switch (c) {
788 case ':': /* character class */
789 NEXT2();
790 (void)REQUIRE(MORE(), REG_EBRACK);
791 c = PEEK();
792 (void)REQUIRE(c != '-' && c != ']', REG_ECTYPE);
793 p_b_cclass(p, cs);
794 (void)REQUIRE(MORE(), REG_EBRACK);
795 (void)REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
796 break;
797 case '=': /* equivalence class */
798 NEXT2();
799 (void)REQUIRE(MORE(), REG_EBRACK);
800 c = PEEK();
801 (void)REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
802 p_b_eclass(p, cs);
803 (void)REQUIRE(MORE(), REG_EBRACK);
804 (void)REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
805 break;
806 default: /* symbol, ordinary character, or range */
807 start = p_b_symbol(p);
808 if (SEE('-') && MORE2() && PEEK2() != ']') {
809 /* range */
810 NEXT();
811 if (EAT('-'))
812 finish = '-';
813 else
814 finish = p_b_symbol(p);
815 } else
816 finish = start;
817 if (start == finish)
818 CHadd(p, cs, start);
819 else {
820 if (table->__collate_load_error) {
821 (void)REQUIRE((uch)start <= (uch)finish, REG_ERANGE);
822 CHaddrange(p, cs, start, finish);
823 } else {
| 46
47#include <sys/types.h>
48#include <stdio.h>
49#include <string.h>
50#include <ctype.h>
51#include <limits.h>
52#include <stdlib.h>
53#include <regex.h>
54#include <runetype.h>
55#include <wchar.h>
56#include <wctype.h>
57
58#include "collate.h"
59
60#include "utils.h"
61#include "regex2.h"
62
63#include "cname.h"
64
65/*
66 * parse structure, passed up and down to avoid global variables and
67 * other clumsinesses
68 */
69struct parse {
70 char *next; /* next character in RE */
71 char *end; /* end of string (-> NUL normally) */
72 int error; /* has an error been seen? */
73 sop *strip; /* malloced strip */
74 sopno ssize; /* malloced strip size (allocated) */
75 sopno slen; /* malloced strip length (used) */
76 int ncsalloc; /* number of csets allocated */
77 struct re_guts *g;
78# define NPAREN 10 /* we need to remember () 1-9 for back refs */
79 sopno pbegin[NPAREN]; /* -> ( ([0] unused) */
80 sopno pend[NPAREN]; /* -> ) ([0] unused) */
81};
82
83/* ========= begin header generated by ./mkh ========= */
84#ifdef __cplusplus
85extern "C" {
86#endif
87
88/* === regcomp.c === */
89static void p_ere(struct parse *p, int stop);
90static void p_ere_exp(struct parse *p);
91static void p_str(struct parse *p);
92static void p_bre(struct parse *p, int end1, int end2);
93static int p_simp_re(struct parse *p, int starordinary);
94static int p_count(struct parse *p);
95static void p_bracket(struct parse *p);
96static void p_b_term(struct parse *p, cset *cs);
97static void p_b_cclass(struct parse *p, cset *cs);
98static void p_b_eclass(struct parse *p, cset *cs);
99static wint_t p_b_symbol(struct parse *p);
100static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
101static wint_t othercase(wint_t ch);
102static void bothcases(struct parse *p, wint_t ch);
103static void ordinary(struct parse *p, wint_t ch);
104static void nonnewline(struct parse *p);
105static void repeat(struct parse *p, sopno start, int from, int to);
106static int seterr(struct parse *p, int e);
107static cset *allocset(struct parse *p);
108static void freeset(struct parse *p, cset *cs);
109static void CHadd(struct parse *p, cset *cs, wint_t ch);
110static void CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max);
111static void CHaddtype(struct parse *p, cset *cs, wctype_t wct);
112static wint_t singleton(cset *cs);
113static sopno dupl(struct parse *p, sopno start, sopno finish);
114static void doemit(struct parse *p, sop op, size_t opnd);
115static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
116static void dofwd(struct parse *p, sopno pos, sop value);
117static int enlarge(struct parse *p, sopno size);
118static void stripsnug(struct parse *p, struct re_guts *g);
119static void findmust(struct parse *p, struct re_guts *g);
120static int altoffset(sop *scan, int offset);
121static void computejumps(struct parse *p, struct re_guts *g);
122static void computematchjumps(struct parse *p, struct re_guts *g);
123static sopno pluscount(struct parse *p, struct re_guts *g);
124static wint_t wgetnext(struct parse *p);
125
126#ifdef __cplusplus
127}
128#endif
129/* ========= end header generated by ./mkh ========= */
130
131static char nuls[10]; /* place to point scanner in event of error */
132
133/*
134 * macros for use with parse structure
135 * BEWARE: these know that the parse structure is named `p' !!!
136 */
137#define PEEK() (*p->next)
138#define PEEK2() (*(p->next+1))
139#define MORE() (p->next < p->end)
140#define MORE2() (p->next+1 < p->end)
141#define SEE(c) (MORE() && PEEK() == (c))
142#define SEETWO(a, b) (MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
143#define EAT(c) ((SEE(c)) ? (NEXT(), 1) : 0)
144#define EATTWO(a, b) ((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
145#define NEXT() (p->next++)
146#define NEXT2() (p->next += 2)
147#define NEXTn(n) (p->next += (n))
148#define GETNEXT() (*p->next++)
149#define WGETNEXT() wgetnext(p)
150#define SETERROR(e) seterr(p, (e))
151#define REQUIRE(co, e) ((co) || SETERROR(e))
152#define MUSTSEE(c, e) (REQUIRE(MORE() && PEEK() == (c), e))
153#define MUSTEAT(c, e) (REQUIRE(MORE() && GETNEXT() == (c), e))
154#define MUSTNOTSEE(c, e) (REQUIRE(!MORE() || PEEK() != (c), e))
155#define EMIT(op, sopnd) doemit(p, (sop)(op), (size_t)(sopnd))
156#define INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
157#define AHEAD(pos) dofwd(p, pos, HERE()-(pos))
158#define ASTERN(sop, pos) EMIT(sop, HERE()-pos)
159#define HERE() (p->slen)
160#define THERE() (p->slen - 1)
161#define THERETHERE() (p->slen - 2)
162#define DROP(n) (p->slen -= (n))
163
164#ifndef NDEBUG
165static int never = 0; /* for use in asserts; shuts lint up */
166#else
167#define never 0 /* some <assert.h>s have bugs too */
168#endif
169
170/* Macro used by computejump()/computematchjump() */
171#define MIN(a,b) ((a)<(b)?(a):(b))
172
173/*
174 - regcomp - interface for parser and compilation
175 = extern int regcomp(regex_t *, const char *, int);
176 = #define REG_BASIC 0000
177 = #define REG_EXTENDED 0001
178 = #define REG_ICASE 0002
179 = #define REG_NOSUB 0004
180 = #define REG_NEWLINE 0010
181 = #define REG_NOSPEC 0020
182 = #define REG_PEND 0040
183 = #define REG_DUMP 0200
184 */
185int /* 0 success, otherwise REG_something */
186regcomp(regex_t * __restrict preg,
187 const char * __restrict pattern,
188 int cflags)
189{
190 struct parse pa;
191 struct re_guts *g;
192 struct parse *p = &pa;
193 int i;
194 size_t len;
195 size_t maxlen;
196#ifdef REDEBUG
197# define GOODFLAGS(f) (f)
198#else
199# define GOODFLAGS(f) ((f)&~REG_DUMP)
200#endif
201
202 cflags = GOODFLAGS(cflags);
203 if ((cflags®_EXTENDED) && (cflags®_NOSPEC))
204 return(REG_INVARG);
205
206 if (cflags®_PEND) {
207 if (preg->re_endp < pattern)
208 return(REG_INVARG);
209 len = preg->re_endp - pattern;
210 } else
211 len = strlen((char *)pattern);
212
213 /* do the mallocs early so failure handling is easy */
214 g = (struct re_guts *)malloc(sizeof(struct re_guts));
215 if (g == NULL)
216 return(REG_ESPACE);
217 /*
218 * Limit the pattern space to avoid a 32-bit overflow on buffer
219 * extension. Also avoid any signed overflow in case of conversion
220 * so make the real limit based on a 31-bit overflow.
221 *
222 * Likely not applicable on 64-bit systems but handle the case
223 * generically (who are we to stop people from using ~715MB+
224 * patterns?).
225 */
226 maxlen = ((size_t)-1 >> 1) / sizeof(sop) * 2 / 3;
227 if (len >= maxlen) {
228 free((char *)g);
229 return(REG_ESPACE);
230 }
231 p->ssize = len/(size_t)2*(size_t)3 + (size_t)1; /* ugh */
232 assert(p->ssize >= len);
233
234 p->strip = (sop *)malloc(p->ssize * sizeof(sop));
235 p->slen = 0;
236 if (p->strip == NULL) {
237 free((char *)g);
238 return(REG_ESPACE);
239 }
240
241 /* set things up */
242 p->g = g;
243 p->next = (char *)pattern; /* convenience; we do not modify it */
244 p->end = p->next + len;
245 p->error = 0;
246 p->ncsalloc = 0;
247 for (i = 0; i < NPAREN; i++) {
248 p->pbegin[i] = 0;
249 p->pend[i] = 0;
250 }
251 g->sets = NULL;
252 g->ncsets = 0;
253 g->cflags = cflags;
254 g->iflags = 0;
255 g->nbol = 0;
256 g->neol = 0;
257 g->must = NULL;
258 g->moffset = -1;
259 g->charjump = NULL;
260 g->matchjump = NULL;
261 g->mlen = 0;
262 g->nsub = 0;
263 g->backrefs = 0;
264
265 /* do it */
266 EMIT(OEND, 0);
267 g->firststate = THERE();
268 if (cflags®_EXTENDED)
269 p_ere(p, OUT);
270 else if (cflags®_NOSPEC)
271 p_str(p);
272 else
273 p_bre(p, OUT, OUT);
274 EMIT(OEND, 0);
275 g->laststate = THERE();
276
277 /* tidy up loose ends and fill things in */
278 stripsnug(p, g);
279 findmust(p, g);
280 /* only use Boyer-Moore algorithm if the pattern is bigger
281 * than three characters
282 */
283 if(g->mlen > 3) {
284 computejumps(p, g);
285 computematchjumps(p, g);
286 if(g->matchjump == NULL && g->charjump != NULL) {
287 free(g->charjump);
288 g->charjump = NULL;
289 }
290 }
291 g->nplus = pluscount(p, g);
292 g->magic = MAGIC2;
293 preg->re_nsub = g->nsub;
294 preg->re_g = g;
295 preg->re_magic = MAGIC1;
296#ifndef REDEBUG
297 /* not debugging, so can't rely on the assert() in regexec() */
298 if (g->iflags&BAD)
299 SETERROR(REG_ASSERT);
300#endif
301
302 /* win or lose, we're done */
303 if (p->error != 0) /* lose */
304 regfree(preg);
305 return(p->error);
306}
307
308/*
309 - p_ere - ERE parser top level, concatenation and alternation
310 == static void p_ere(struct parse *p, int_t stop);
311 */
312static void
313p_ere(struct parse *p,
314 int stop) /* character this ERE should end at */
315{
316 char c;
317 sopno prevback;
318 sopno prevfwd;
319 sopno conc;
320 int first = 1; /* is this the first alternative? */
321
322 for (;;) {
323 /* do a bunch of concatenated expressions */
324 conc = HERE();
325 while (MORE() && (c = PEEK()) != '|' && c != stop)
326 p_ere_exp(p);
327 (void)REQUIRE(HERE() != conc, REG_EMPTY); /* require nonempty */
328
329 if (!EAT('|'))
330 break; /* NOTE BREAK OUT */
331
332 if (first) {
333 INSERT(OCH_, conc); /* offset is wrong */
334 prevfwd = conc;
335 prevback = conc;
336 first = 0;
337 }
338 ASTERN(OOR1, prevback);
339 prevback = THERE();
340 AHEAD(prevfwd); /* fix previous offset */
341 prevfwd = HERE();
342 EMIT(OOR2, 0); /* offset is very wrong */
343 }
344
345 if (!first) { /* tail-end fixups */
346 AHEAD(prevfwd);
347 ASTERN(O_CH, prevback);
348 }
349
350 assert(!MORE() || SEE(stop));
351}
352
353/*
354 - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
355 == static void p_ere_exp(struct parse *p);
356 */
357static void
358p_ere_exp(struct parse *p)
359{
360 char c;
361 wint_t wc;
362 sopno pos;
363 int count;
364 int count2;
365 sopno subno;
366 int wascaret = 0;
367
368 assert(MORE()); /* caller should have ensured this */
369 c = GETNEXT();
370
371 pos = HERE();
372 switch (c) {
373 case '(':
374 (void)REQUIRE(MORE(), REG_EPAREN);
375 p->g->nsub++;
376 subno = p->g->nsub;
377 if (subno < NPAREN)
378 p->pbegin[subno] = HERE();
379 EMIT(OLPAREN, subno);
380 if (!SEE(')'))
381 p_ere(p, ')');
382 if (subno < NPAREN) {
383 p->pend[subno] = HERE();
384 assert(p->pend[subno] != 0);
385 }
386 EMIT(ORPAREN, subno);
387 (void)MUSTEAT(')', REG_EPAREN);
388 break;
389#ifndef POSIX_MISTAKE
390 case ')': /* happens only if no current unmatched ( */
391 /*
392 * You may ask, why the ifndef? Because I didn't notice
393 * this until slightly too late for 1003.2, and none of the
394 * other 1003.2 regular-expression reviewers noticed it at
395 * all. So an unmatched ) is legal POSIX, at least until
396 * we can get it fixed.
397 */
398 SETERROR(REG_EPAREN);
399 break;
400#endif
401 case '^':
402 EMIT(OBOL, 0);
403 p->g->iflags |= USEBOL;
404 p->g->nbol++;
405 wascaret = 1;
406 break;
407 case '$':
408 EMIT(OEOL, 0);
409 p->g->iflags |= USEEOL;
410 p->g->neol++;
411 break;
412 case '|':
413 SETERROR(REG_EMPTY);
414 break;
415 case '*':
416 case '+':
417 case '?':
418 SETERROR(REG_BADRPT);
419 break;
420 case '.':
421 if (p->g->cflags®_NEWLINE)
422 nonnewline(p);
423 else
424 EMIT(OANY, 0);
425 break;
426 case '[':
427 p_bracket(p);
428 break;
429 case '\\':
430 (void)REQUIRE(MORE(), REG_EESCAPE);
431 wc = WGETNEXT();
432 switch (wc) {
433 case '<':
434 EMIT(OBOW, 0);
435 break;
436 case '>':
437 EMIT(OEOW, 0);
438 break;
439 default:
440 ordinary(p, wc);
441 break;
442 }
443 break;
444 case '{': /* okay as ordinary except if digit follows */
445 (void)REQUIRE(!MORE() || !isdigit((uch)PEEK()), REG_BADRPT);
446 /* FALLTHROUGH */
447 default:
448 p->next--;
449 wc = WGETNEXT();
450 ordinary(p, wc);
451 break;
452 }
453
454 if (!MORE())
455 return;
456 c = PEEK();
457 /* we call { a repetition if followed by a digit */
458 if (!( c == '*' || c == '+' || c == '?' ||
459 (c == '{' && MORE2() && isdigit((uch)PEEK2())) ))
460 return; /* no repetition, we're done */
461 NEXT();
462
463 (void)REQUIRE(!wascaret, REG_BADRPT);
464 switch (c) {
465 case '*': /* implemented as +? */
466 /* this case does not require the (y|) trick, noKLUDGE */
467 INSERT(OPLUS_, pos);
468 ASTERN(O_PLUS, pos);
469 INSERT(OQUEST_, pos);
470 ASTERN(O_QUEST, pos);
471 break;
472 case '+':
473 INSERT(OPLUS_, pos);
474 ASTERN(O_PLUS, pos);
475 break;
476 case '?':
477 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
478 INSERT(OCH_, pos); /* offset slightly wrong */
479 ASTERN(OOR1, pos); /* this one's right */
480 AHEAD(pos); /* fix the OCH_ */
481 EMIT(OOR2, 0); /* offset very wrong... */
482 AHEAD(THERE()); /* ...so fix it */
483 ASTERN(O_CH, THERETHERE());
484 break;
485 case '{':
486 count = p_count(p);
487 if (EAT(',')) {
488 if (isdigit((uch)PEEK())) {
489 count2 = p_count(p);
490 (void)REQUIRE(count <= count2, REG_BADBR);
491 } else /* single number with comma */
492 count2 = INFINITY;
493 } else /* just a single number */
494 count2 = count;
495 repeat(p, pos, count, count2);
496 if (!EAT('}')) { /* error heuristics */
497 while (MORE() && PEEK() != '}')
498 NEXT();
499 (void)REQUIRE(MORE(), REG_EBRACE);
500 SETERROR(REG_BADBR);
501 }
502 break;
503 }
504
505 if (!MORE())
506 return;
507 c = PEEK();
508 if (!( c == '*' || c == '+' || c == '?' ||
509 (c == '{' && MORE2() && isdigit((uch)PEEK2())) ) )
510 return;
511 SETERROR(REG_BADRPT);
512}
513
514/*
515 - p_str - string (no metacharacters) "parser"
516 == static void p_str(struct parse *p);
517 */
518static void
519p_str(struct parse *p)
520{
521 (void)REQUIRE(MORE(), REG_EMPTY);
522 while (MORE())
523 ordinary(p, WGETNEXT());
524}
525
526/*
527 - p_bre - BRE parser top level, anchoring and concatenation
528 == static void p_bre(struct parse *p, int end1, \
529 == int end2);
530 * Giving end1 as OUT essentially eliminates the end1/end2 check.
531 *
532 * This implementation is a bit of a kludge, in that a trailing $ is first
533 * taken as an ordinary character and then revised to be an anchor.
534 * The amount of lookahead needed to avoid this kludge is excessive.
535 */
536static void
537p_bre(struct parse *p,
538 int end1, /* first terminating character */
539 int end2) /* second terminating character */
540{
541 sopno start = HERE();
542 int first = 1; /* first subexpression? */
543 int wasdollar = 0;
544
545 if (EAT('^')) {
546 EMIT(OBOL, 0);
547 p->g->iflags |= USEBOL;
548 p->g->nbol++;
549 }
550 while (MORE() && !SEETWO(end1, end2)) {
551 wasdollar = p_simp_re(p, first);
552 first = 0;
553 }
554 if (wasdollar) { /* oops, that was a trailing anchor */
555 DROP(1);
556 EMIT(OEOL, 0);
557 p->g->iflags |= USEEOL;
558 p->g->neol++;
559 }
560
561 (void)REQUIRE(HERE() != start, REG_EMPTY); /* require nonempty */
562}
563
564/*
565 - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
566 == static int p_simp_re(struct parse *p, int starordinary);
567 */
568static int /* was the simple RE an unbackslashed $? */
569p_simp_re(struct parse *p,
570 int starordinary) /* is a leading * an ordinary character? */
571{
572 int c;
573 int count;
574 int count2;
575 sopno pos;
576 int i;
577 wint_t wc;
578 sopno subno;
579# define BACKSL (1<<CHAR_BIT)
580
581 pos = HERE(); /* repetition op, if any, covers from here */
582
583 assert(MORE()); /* caller should have ensured this */
584 c = GETNEXT();
585 if (c == '\\') {
586 (void)REQUIRE(MORE(), REG_EESCAPE);
587 c = BACKSL | GETNEXT();
588 }
589 switch (c) {
590 case '.':
591 if (p->g->cflags®_NEWLINE)
592 nonnewline(p);
593 else
594 EMIT(OANY, 0);
595 break;
596 case '[':
597 p_bracket(p);
598 break;
599 case BACKSL|'<':
600 EMIT(OBOW, 0);
601 break;
602 case BACKSL|'>':
603 EMIT(OEOW, 0);
604 break;
605 case BACKSL|'{':
606 SETERROR(REG_BADRPT);
607 break;
608 case BACKSL|'(':
609 p->g->nsub++;
610 subno = p->g->nsub;
611 if (subno < NPAREN)
612 p->pbegin[subno] = HERE();
613 EMIT(OLPAREN, subno);
614 /* the MORE here is an error heuristic */
615 if (MORE() && !SEETWO('\\', ')'))
616 p_bre(p, '\\', ')');
617 if (subno < NPAREN) {
618 p->pend[subno] = HERE();
619 assert(p->pend[subno] != 0);
620 }
621 EMIT(ORPAREN, subno);
622 (void)REQUIRE(EATTWO('\\', ')'), REG_EPAREN);
623 break;
624 case BACKSL|')': /* should not get here -- must be user */
625 case BACKSL|'}':
626 SETERROR(REG_EPAREN);
627 break;
628 case BACKSL|'1':
629 case BACKSL|'2':
630 case BACKSL|'3':
631 case BACKSL|'4':
632 case BACKSL|'5':
633 case BACKSL|'6':
634 case BACKSL|'7':
635 case BACKSL|'8':
636 case BACKSL|'9':
637 i = (c&~BACKSL) - '0';
638 assert(i < NPAREN);
639 if (p->pend[i] != 0) {
640 assert(i <= p->g->nsub);
641 EMIT(OBACK_, i);
642 assert(p->pbegin[i] != 0);
643 assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
644 assert(OP(p->strip[p->pend[i]]) == ORPAREN);
645 (void) dupl(p, p->pbegin[i]+1, p->pend[i]);
646 EMIT(O_BACK, i);
647 } else
648 SETERROR(REG_ESUBREG);
649 p->g->backrefs = 1;
650 break;
651 case '*':
652 (void)REQUIRE(starordinary, REG_BADRPT);
653 /* FALLTHROUGH */
654 default:
655 p->next--;
656 wc = WGETNEXT();
657 ordinary(p, wc);
658 break;
659 }
660
661 if (EAT('*')) { /* implemented as +? */
662 /* this case does not require the (y|) trick, noKLUDGE */
663 INSERT(OPLUS_, pos);
664 ASTERN(O_PLUS, pos);
665 INSERT(OQUEST_, pos);
666 ASTERN(O_QUEST, pos);
667 } else if (EATTWO('\\', '{')) {
668 count = p_count(p);
669 if (EAT(',')) {
670 if (MORE() && isdigit((uch)PEEK())) {
671 count2 = p_count(p);
672 (void)REQUIRE(count <= count2, REG_BADBR);
673 } else /* single number with comma */
674 count2 = INFINITY;
675 } else /* just a single number */
676 count2 = count;
677 repeat(p, pos, count, count2);
678 if (!EATTWO('\\', '}')) { /* error heuristics */
679 while (MORE() && !SEETWO('\\', '}'))
680 NEXT();
681 (void)REQUIRE(MORE(), REG_EBRACE);
682 SETERROR(REG_BADBR);
683 }
684 } else if (c == '$') /* $ (but not \$) ends it */
685 return(1);
686
687 return(0);
688}
689
690/*
691 - p_count - parse a repetition count
692 == static int p_count(struct parse *p);
693 */
694static int /* the value */
695p_count(struct parse *p)
696{
697 int count = 0;
698 int ndigits = 0;
699
700 while (MORE() && isdigit((uch)PEEK()) && count <= DUPMAX) {
701 count = count*10 + (GETNEXT() - '0');
702 ndigits++;
703 }
704
705 (void)REQUIRE(ndigits > 0 && count <= DUPMAX, REG_BADBR);
706 return(count);
707}
708
709/*
710 - p_bracket - parse a bracketed character list
711 == static void p_bracket(struct parse *p);
712 */
713static void
714p_bracket(struct parse *p)
715{
716 cset *cs;
717 wint_t ch;
718
719 /* Dept of Truly Sickening Special-Case Kludges */
720 if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
721 EMIT(OBOW, 0);
722 NEXTn(6);
723 return;
724 }
725 if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
726 EMIT(OEOW, 0);
727 NEXTn(6);
728 return;
729 }
730
731 if ((cs = allocset(p)) == NULL)
732 return;
733
734 if (p->g->cflags®_ICASE)
735 cs->icase = 1;
736 if (EAT('^'))
737 cs->invert = 1;
738 if (EAT(']'))
739 CHadd(p, cs, ']');
740 else if (EAT('-'))
741 CHadd(p, cs, '-');
742 while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
743 p_b_term(p, cs);
744 if (EAT('-'))
745 CHadd(p, cs, '-');
746 (void)MUSTEAT(']', REG_EBRACK);
747
748 if (p->error != 0) /* don't mess things up further */
749 return;
750
751 if (cs->invert && p->g->cflags®_NEWLINE)
752 cs->bmp['\n' >> 3] |= 1 << ('\n' & 7);
753
754 if ((ch = singleton(cs)) != OUT) { /* optimize singleton sets */
755 ordinary(p, ch);
756 freeset(p, cs);
757 } else
758 EMIT(OANYOF, (int)(cs - p->g->sets));
759}
760
761/*
762 - p_b_term - parse one term of a bracketed character list
763 == static void p_b_term(struct parse *p, cset *cs);
764 */
765static void
766p_b_term(struct parse *p, cset *cs)
767{
768 char c;
769 wint_t start, finish;
770 wint_t i;
771 struct xlocale_collate *table =
772 (struct xlocale_collate*)__get_locale()->components[XLC_COLLATE];
773
774 /* classify what we've got */
775 switch ((MORE()) ? PEEK() : '\0') {
776 case '[':
777 c = (MORE2()) ? PEEK2() : '\0';
778 break;
779 case '-':
780 SETERROR(REG_ERANGE);
781 return; /* NOTE RETURN */
782 default:
783 c = '\0';
784 break;
785 }
786
787 switch (c) {
788 case ':': /* character class */
789 NEXT2();
790 (void)REQUIRE(MORE(), REG_EBRACK);
791 c = PEEK();
792 (void)REQUIRE(c != '-' && c != ']', REG_ECTYPE);
793 p_b_cclass(p, cs);
794 (void)REQUIRE(MORE(), REG_EBRACK);
795 (void)REQUIRE(EATTWO(':', ']'), REG_ECTYPE);
796 break;
797 case '=': /* equivalence class */
798 NEXT2();
799 (void)REQUIRE(MORE(), REG_EBRACK);
800 c = PEEK();
801 (void)REQUIRE(c != '-' && c != ']', REG_ECOLLATE);
802 p_b_eclass(p, cs);
803 (void)REQUIRE(MORE(), REG_EBRACK);
804 (void)REQUIRE(EATTWO('=', ']'), REG_ECOLLATE);
805 break;
806 default: /* symbol, ordinary character, or range */
807 start = p_b_symbol(p);
808 if (SEE('-') && MORE2() && PEEK2() != ']') {
809 /* range */
810 NEXT();
811 if (EAT('-'))
812 finish = '-';
813 else
814 finish = p_b_symbol(p);
815 } else
816 finish = start;
817 if (start == finish)
818 CHadd(p, cs, start);
819 else {
820 if (table->__collate_load_error) {
821 (void)REQUIRE((uch)start <= (uch)finish, REG_ERANGE);
822 CHaddrange(p, cs, start, finish);
823 } else {
|
824 (void)REQUIRE(__collate_range_cmp(table, start, finish) <= 0, REG_ERANGE);
| 824 (void)REQUIRE(__wcollate_range_cmp(table, start, finish) <= 0, REG_ERANGE);
|
825 for (i = 0; i <= UCHAR_MAX; i++) {
| 825 for (i = 0; i <= UCHAR_MAX; i++) {
|
826 if ( __collate_range_cmp(table, start, i) <= 0
827 && __collate_range_cmp(table, i, finish) <= 0
| 826 if ( __wcollate_range_cmp(table, start, i) <= 0
827 && __wcollate_range_cmp(table, i, finish) <= 0
|
828 )
829 CHadd(p, cs, i);
830 }
831 }
832 }
833 break;
834 }
835}
836
837/*
838 - p_b_cclass - parse a character-class name and deal with it
839 == static void p_b_cclass(struct parse *p, cset *cs);
840 */
841static void
842p_b_cclass(struct parse *p, cset *cs)
843{
844 char *sp = p->next;
845 size_t len;
846 wctype_t wct;
847 char clname[16];
848
849 while (MORE() && isalpha((uch)PEEK()))
850 NEXT();
851 len = p->next - sp;
852 if (len >= sizeof(clname) - 1) {
853 SETERROR(REG_ECTYPE);
854 return;
855 }
856 memcpy(clname, sp, len);
857 clname[len] = '\0';
858 if ((wct = wctype(clname)) == 0) {
859 SETERROR(REG_ECTYPE);
860 return;
861 }
862 CHaddtype(p, cs, wct);
863}
864
865/*
866 - p_b_eclass - parse an equivalence-class name and deal with it
867 == static void p_b_eclass(struct parse *p, cset *cs);
868 *
869 * This implementation is incomplete. xxx
870 */
871static void
872p_b_eclass(struct parse *p, cset *cs)
873{
874 wint_t c;
875
876 c = p_b_coll_elem(p, '=');
877 CHadd(p, cs, c);
878}
879
880/*
881 - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
882 == static wint_t p_b_symbol(struct parse *p);
883 */
884static wint_t /* value of symbol */
885p_b_symbol(struct parse *p)
886{
887 wint_t value;
888
889 (void)REQUIRE(MORE(), REG_EBRACK);
890 if (!EATTWO('[', '.'))
891 return(WGETNEXT());
892
893 /* collating symbol */
894 value = p_b_coll_elem(p, '.');
895 (void)REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
896 return(value);
897}
898
899/*
900 - p_b_coll_elem - parse a collating-element name and look it up
901 == static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
902 */
903static wint_t /* value of collating element */
904p_b_coll_elem(struct parse *p,
905 wint_t endc) /* name ended by endc,']' */
906{
907 char *sp = p->next;
908 struct cname *cp;
909 int len;
910 mbstate_t mbs;
911 wchar_t wc;
912 size_t clen;
913
914 while (MORE() && !SEETWO(endc, ']'))
915 NEXT();
916 if (!MORE()) {
917 SETERROR(REG_EBRACK);
918 return(0);
919 }
920 len = p->next - sp;
921 for (cp = cnames; cp->name != NULL; cp++)
922 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
923 return(cp->code); /* known name */
924 memset(&mbs, 0, sizeof(mbs));
925 if ((clen = mbrtowc(&wc, sp, len, &mbs)) == len)
926 return (wc); /* single character */
927 else if (clen == (size_t)-1 || clen == (size_t)-2)
928 SETERROR(REG_ILLSEQ);
929 else
930 SETERROR(REG_ECOLLATE); /* neither */
931 return(0);
932}
933
934/*
935 - othercase - return the case counterpart of an alphabetic
936 == static wint_t othercase(wint_t ch);
937 */
938static wint_t /* if no counterpart, return ch */
939othercase(wint_t ch)
940{
941 assert(iswalpha(ch));
942 if (iswupper(ch))
943 return(towlower(ch));
944 else if (iswlower(ch))
945 return(towupper(ch));
946 else /* peculiar, but could happen */
947 return(ch);
948}
949
950/*
951 - bothcases - emit a dualcase version of a two-case character
952 == static void bothcases(struct parse *p, wint_t ch);
953 *
954 * Boy, is this implementation ever a kludge...
955 */
956static void
957bothcases(struct parse *p, wint_t ch)
958{
959 char *oldnext = p->next;
960 char *oldend = p->end;
961 char bracket[3 + MB_LEN_MAX];
962 size_t n;
963 mbstate_t mbs;
964
965 assert(othercase(ch) != ch); /* p_bracket() would recurse */
966 p->next = bracket;
967 memset(&mbs, 0, sizeof(mbs));
968 n = wcrtomb(bracket, ch, &mbs);
969 assert(n != (size_t)-1);
970 bracket[n] = ']';
971 bracket[n + 1] = '\0';
972 p->end = bracket+n+1;
973 p_bracket(p);
974 assert(p->next == p->end);
975 p->next = oldnext;
976 p->end = oldend;
977}
978
979/*
980 - ordinary - emit an ordinary character
981 == static void ordinary(struct parse *p, wint_t ch);
982 */
983static void
984ordinary(struct parse *p, wint_t ch)
985{
986 cset *cs;
987
988 if ((p->g->cflags®_ICASE) && iswalpha(ch) && othercase(ch) != ch)
989 bothcases(p, ch);
990 else if ((ch & OPDMASK) == ch)
991 EMIT(OCHAR, ch);
992 else {
993 /*
994 * Kludge: character is too big to fit into an OCHAR operand.
995 * Emit a singleton set.
996 */
997 if ((cs = allocset(p)) == NULL)
998 return;
999 CHadd(p, cs, ch);
1000 EMIT(OANYOF, (int)(cs - p->g->sets));
1001 }
1002}
1003
1004/*
1005 - nonnewline - emit REG_NEWLINE version of OANY
1006 == static void nonnewline(struct parse *p);
1007 *
1008 * Boy, is this implementation ever a kludge...
1009 */
1010static void
1011nonnewline(struct parse *p)
1012{
1013 char *oldnext = p->next;
1014 char *oldend = p->end;
1015 char bracket[4];
1016
1017 p->next = bracket;
1018 p->end = bracket+3;
1019 bracket[0] = '^';
1020 bracket[1] = '\n';
1021 bracket[2] = ']';
1022 bracket[3] = '\0';
1023 p_bracket(p);
1024 assert(p->next == bracket+3);
1025 p->next = oldnext;
1026 p->end = oldend;
1027}
1028
1029/*
1030 - repeat - generate code for a bounded repetition, recursively if needed
1031 == static void repeat(struct parse *p, sopno start, int from, int to);
1032 */
1033static void
1034repeat(struct parse *p,
1035 sopno start, /* operand from here to end of strip */
1036 int from, /* repeated from this number */
1037 int to) /* to this number of times (maybe INFINITY) */
1038{
1039 sopno finish = HERE();
1040# define N 2
1041# define INF 3
1042# define REP(f, t) ((f)*8 + (t))
1043# define MAP(n) (((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1044 sopno copy;
1045
1046 if (p->error != 0) /* head off possible runaway recursion */
1047 return;
1048
1049 assert(from <= to);
1050
1051 switch (REP(MAP(from), MAP(to))) {
1052 case REP(0, 0): /* must be user doing this */
1053 DROP(finish-start); /* drop the operand */
1054 break;
1055 case REP(0, 1): /* as x{1,1}? */
1056 case REP(0, N): /* as x{1,n}? */
1057 case REP(0, INF): /* as x{1,}? */
1058 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1059 INSERT(OCH_, start); /* offset is wrong... */
1060 repeat(p, start+1, 1, to);
1061 ASTERN(OOR1, start);
1062 AHEAD(start); /* ... fix it */
1063 EMIT(OOR2, 0);
1064 AHEAD(THERE());
1065 ASTERN(O_CH, THERETHERE());
1066 break;
1067 case REP(1, 1): /* trivial case */
1068 /* done */
1069 break;
1070 case REP(1, N): /* as x?x{1,n-1} */
1071 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1072 INSERT(OCH_, start);
1073 ASTERN(OOR1, start);
1074 AHEAD(start);
1075 EMIT(OOR2, 0); /* offset very wrong... */
1076 AHEAD(THERE()); /* ...so fix it */
1077 ASTERN(O_CH, THERETHERE());
1078 copy = dupl(p, start+1, finish+1);
1079 assert(copy == finish+4);
1080 repeat(p, copy, 1, to-1);
1081 break;
1082 case REP(1, INF): /* as x+ */
1083 INSERT(OPLUS_, start);
1084 ASTERN(O_PLUS, start);
1085 break;
1086 case REP(N, N): /* as xx{m-1,n-1} */
1087 copy = dupl(p, start, finish);
1088 repeat(p, copy, from-1, to-1);
1089 break;
1090 case REP(N, INF): /* as xx{n-1,INF} */
1091 copy = dupl(p, start, finish);
1092 repeat(p, copy, from-1, to);
1093 break;
1094 default: /* "can't happen" */
1095 SETERROR(REG_ASSERT); /* just in case */
1096 break;
1097 }
1098}
1099
1100/*
1101 - wgetnext - helper function for WGETNEXT() macro. Gets the next wide
1102 - character from the parse struct, signals a REG_ILLSEQ error if the
1103 - character can't be converted. Returns the number of bytes consumed.
1104 */
1105static wint_t
1106wgetnext(struct parse *p)
1107{
1108 mbstate_t mbs;
1109 wchar_t wc;
1110 size_t n;
1111
1112 memset(&mbs, 0, sizeof(mbs));
1113 n = mbrtowc(&wc, p->next, p->end - p->next, &mbs);
1114 if (n == (size_t)-1 || n == (size_t)-2) {
1115 SETERROR(REG_ILLSEQ);
1116 return (0);
1117 }
1118 if (n == 0)
1119 n = 1;
1120 p->next += n;
1121 return (wc);
1122}
1123
1124/*
1125 - seterr - set an error condition
1126 == static int seterr(struct parse *p, int e);
1127 */
1128static int /* useless but makes type checking happy */
1129seterr(struct parse *p, int e)
1130{
1131 if (p->error == 0) /* keep earliest error condition */
1132 p->error = e;
1133 p->next = nuls; /* try to bring things to a halt */
1134 p->end = nuls;
1135 return(0); /* make the return value well-defined */
1136}
1137
1138/*
1139 - allocset - allocate a set of characters for []
1140 == static cset *allocset(struct parse *p);
1141 */
1142static cset *
1143allocset(struct parse *p)
1144{
1145 cset *cs, *ncs;
1146
1147 ncs = realloc(p->g->sets, (p->g->ncsets + 1) * sizeof(*ncs));
1148 if (ncs == NULL) {
1149 SETERROR(REG_ESPACE);
1150 return (NULL);
1151 }
1152 p->g->sets = ncs;
1153 cs = &p->g->sets[p->g->ncsets++];
1154 memset(cs, 0, sizeof(*cs));
1155
1156 return(cs);
1157}
1158
1159/*
1160 - freeset - free a now-unused set
1161 == static void freeset(struct parse *p, cset *cs);
1162 */
1163static void
1164freeset(struct parse *p, cset *cs)
1165{
1166 cset *top = &p->g->sets[p->g->ncsets];
1167
1168 free(cs->wides);
1169 free(cs->ranges);
1170 free(cs->types);
1171 memset(cs, 0, sizeof(*cs));
1172 if (cs == top-1) /* recover only the easy case */
1173 p->g->ncsets--;
1174}
1175
1176/*
1177 - singleton - Determine whether a set contains only one character,
1178 - returning it if so, otherwise returning OUT.
1179 */
1180static wint_t
1181singleton(cset *cs)
1182{
1183 wint_t i, s, n;
1184
1185 for (i = n = 0; i < NC; i++)
1186 if (CHIN(cs, i)) {
1187 n++;
1188 s = i;
1189 }
1190 if (n == 1)
1191 return (s);
1192 if (cs->nwides == 1 && cs->nranges == 0 && cs->ntypes == 0 &&
1193 cs->icase == 0)
1194 return (cs->wides[0]);
1195 /* Don't bother handling the other cases. */
1196 return (OUT);
1197}
1198
1199/*
1200 - CHadd - add character to character set.
1201 */
1202static void
1203CHadd(struct parse *p, cset *cs, wint_t ch)
1204{
1205 wint_t nch, *newwides;
1206 assert(ch >= 0);
1207 if (ch < NC)
1208 cs->bmp[ch >> 3] |= 1 << (ch & 7);
1209 else {
1210 newwides = realloc(cs->wides, (cs->nwides + 1) *
1211 sizeof(*cs->wides));
1212 if (newwides == NULL) {
1213 SETERROR(REG_ESPACE);
1214 return;
1215 }
1216 cs->wides = newwides;
1217 cs->wides[cs->nwides++] = ch;
1218 }
1219 if (cs->icase) {
1220 if ((nch = towlower(ch)) < NC)
1221 cs->bmp[nch >> 3] |= 1 << (nch & 7);
1222 if ((nch = towupper(ch)) < NC)
1223 cs->bmp[nch >> 3] |= 1 << (nch & 7);
1224 }
1225}
1226
1227/*
1228 - CHaddrange - add all characters in the range [min,max] to a character set.
1229 */
1230static void
1231CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max)
1232{
1233 crange *newranges;
1234
1235 for (; min < NC && min <= max; min++)
1236 CHadd(p, cs, min);
1237 if (min >= max)
1238 return;
1239 newranges = realloc(cs->ranges, (cs->nranges + 1) *
1240 sizeof(*cs->ranges));
1241 if (newranges == NULL) {
1242 SETERROR(REG_ESPACE);
1243 return;
1244 }
1245 cs->ranges = newranges;
1246 cs->ranges[cs->nranges].min = min;
1247 cs->ranges[cs->nranges].max = max;
1248 cs->nranges++;
1249}
1250
1251/*
1252 - CHaddtype - add all characters of a certain type to a character set.
1253 */
1254static void
1255CHaddtype(struct parse *p, cset *cs, wctype_t wct)
1256{
1257 wint_t i;
1258 wctype_t *newtypes;
1259
1260 for (i = 0; i < NC; i++)
1261 if (iswctype(i, wct))
1262 CHadd(p, cs, i);
1263 newtypes = realloc(cs->types, (cs->ntypes + 1) *
1264 sizeof(*cs->types));
1265 if (newtypes == NULL) {
1266 SETERROR(REG_ESPACE);
1267 return;
1268 }
1269 cs->types = newtypes;
1270 cs->types[cs->ntypes++] = wct;
1271}
1272
1273/*
1274 - dupl - emit a duplicate of a bunch of sops
1275 == static sopno dupl(struct parse *p, sopno start, sopno finish);
1276 */
1277static sopno /* start of duplicate */
1278dupl(struct parse *p,
1279 sopno start, /* from here */
1280 sopno finish) /* to this less one */
1281{
1282 sopno ret = HERE();
1283 sopno len = finish - start;
1284
1285 assert(finish >= start);
1286 if (len == 0)
1287 return(ret);
1288 if (!enlarge(p, p->ssize + len)) /* this many unexpected additions */
1289 return(ret);
1290 (void) memcpy((char *)(p->strip + p->slen),
1291 (char *)(p->strip + start), (size_t)len*sizeof(sop));
1292 p->slen += len;
1293 return(ret);
1294}
1295
1296/*
1297 - doemit - emit a strip operator
1298 == static void doemit(struct parse *p, sop op, size_t opnd);
1299 *
1300 * It might seem better to implement this as a macro with a function as
1301 * hard-case backup, but it's just too big and messy unless there are
1302 * some changes to the data structures. Maybe later.
1303 */
1304static void
1305doemit(struct parse *p, sop op, size_t opnd)
1306{
1307 /* avoid making error situations worse */
1308 if (p->error != 0)
1309 return;
1310
1311 /* deal with oversize operands ("can't happen", more or less) */
1312 assert(opnd < 1<<OPSHIFT);
1313
1314 /* deal with undersized strip */
1315 if (p->slen >= p->ssize)
1316 if (!enlarge(p, (p->ssize+1) / 2 * 3)) /* +50% */
1317 return;
1318
1319 /* finally, it's all reduced to the easy case */
1320 p->strip[p->slen++] = SOP(op, opnd);
1321}
1322
1323/*
1324 - doinsert - insert a sop into the strip
1325 == static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
1326 */
1327static void
1328doinsert(struct parse *p, sop op, size_t opnd, sopno pos)
1329{
1330 sopno sn;
1331 sop s;
1332 int i;
1333
1334 /* avoid making error situations worse */
1335 if (p->error != 0)
1336 return;
1337
1338 sn = HERE();
1339 EMIT(op, opnd); /* do checks, ensure space */
1340 assert(HERE() == sn+1);
1341 s = p->strip[sn];
1342
1343 /* adjust paren pointers */
1344 assert(pos > 0);
1345 for (i = 1; i < NPAREN; i++) {
1346 if (p->pbegin[i] >= pos) {
1347 p->pbegin[i]++;
1348 }
1349 if (p->pend[i] >= pos) {
1350 p->pend[i]++;
1351 }
1352 }
1353
1354 memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1355 (HERE()-pos-1)*sizeof(sop));
1356 p->strip[pos] = s;
1357}
1358
1359/*
1360 - dofwd - complete a forward reference
1361 == static void dofwd(struct parse *p, sopno pos, sop value);
1362 */
1363static void
1364dofwd(struct parse *p, sopno pos, sop value)
1365{
1366 /* avoid making error situations worse */
1367 if (p->error != 0)
1368 return;
1369
1370 assert(value < 1<<OPSHIFT);
1371 p->strip[pos] = OP(p->strip[pos]) | value;
1372}
1373
1374/*
1375 - enlarge - enlarge the strip
1376 == static int enlarge(struct parse *p, sopno size);
1377 */
1378static int
1379enlarge(struct parse *p, sopno size)
1380{
1381 sop *sp;
1382
1383 if (p->ssize >= size)
1384 return 1;
1385
1386 sp = (sop *)realloc(p->strip, size*sizeof(sop));
1387 if (sp == NULL) {
1388 SETERROR(REG_ESPACE);
1389 return 0;
1390 }
1391 p->strip = sp;
1392 p->ssize = size;
1393 return 1;
1394}
1395
1396/*
1397 - stripsnug - compact the strip
1398 == static void stripsnug(struct parse *p, struct re_guts *g);
1399 */
1400static void
1401stripsnug(struct parse *p, struct re_guts *g)
1402{
1403 g->nstates = p->slen;
1404 g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1405 if (g->strip == NULL) {
1406 SETERROR(REG_ESPACE);
1407 g->strip = p->strip;
1408 }
1409}
1410
1411/*
1412 - findmust - fill in must and mlen with longest mandatory literal string
1413 == static void findmust(struct parse *p, struct re_guts *g);
1414 *
1415 * This algorithm could do fancy things like analyzing the operands of |
1416 * for common subsequences. Someday. This code is simple and finds most
1417 * of the interesting cases.
1418 *
1419 * Note that must and mlen got initialized during setup.
1420 */
1421static void
1422findmust(struct parse *p, struct re_guts *g)
1423{
1424 sop *scan;
1425 sop *start = NULL;
1426 sop *newstart = NULL;
1427 sopno newlen;
1428 sop s;
1429 char *cp;
1430 int offset;
1431 char buf[MB_LEN_MAX];
1432 size_t clen;
1433 mbstate_t mbs;
1434
1435 /* avoid making error situations worse */
1436 if (p->error != 0)
1437 return;
1438
1439 /*
1440 * It's not generally safe to do a ``char'' substring search on
1441 * multibyte character strings, but it's safe for at least
1442 * UTF-8 (see RFC 3629).
1443 */
1444 if (MB_CUR_MAX > 1 &&
1445 strcmp(_CurrentRuneLocale->__encoding, "UTF-8") != 0)
1446 return;
1447
1448 /* find the longest OCHAR sequence in strip */
1449 newlen = 0;
1450 offset = 0;
1451 g->moffset = 0;
1452 scan = g->strip + 1;
1453 do {
1454 s = *scan++;
1455 switch (OP(s)) {
1456 case OCHAR: /* sequence member */
1457 if (newlen == 0) { /* new sequence */
1458 memset(&mbs, 0, sizeof(mbs));
1459 newstart = scan - 1;
1460 }
1461 clen = wcrtomb(buf, OPND(s), &mbs);
1462 if (clen == (size_t)-1)
1463 goto toohard;
1464 newlen += clen;
1465 break;
1466 case OPLUS_: /* things that don't break one */
1467 case OLPAREN:
1468 case ORPAREN:
1469 break;
1470 case OQUEST_: /* things that must be skipped */
1471 case OCH_:
1472 offset = altoffset(scan, offset);
1473 scan--;
1474 do {
1475 scan += OPND(s);
1476 s = *scan;
1477 /* assert() interferes w debug printouts */
1478 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1479 OP(s) != OOR2) {
1480 g->iflags |= BAD;
1481 return;
1482 }
1483 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1484 /* FALLTHROUGH */
1485 case OBOW: /* things that break a sequence */
1486 case OEOW:
1487 case OBOL:
1488 case OEOL:
1489 case O_QUEST:
1490 case O_CH:
1491 case OEND:
1492 if (newlen > g->mlen) { /* ends one */
1493 start = newstart;
1494 g->mlen = newlen;
1495 if (offset > -1) {
1496 g->moffset += offset;
1497 offset = newlen;
1498 } else
1499 g->moffset = offset;
1500 } else {
1501 if (offset > -1)
1502 offset += newlen;
1503 }
1504 newlen = 0;
1505 break;
1506 case OANY:
1507 if (newlen > g->mlen) { /* ends one */
1508 start = newstart;
1509 g->mlen = newlen;
1510 if (offset > -1) {
1511 g->moffset += offset;
1512 offset = newlen;
1513 } else
1514 g->moffset = offset;
1515 } else {
1516 if (offset > -1)
1517 offset += newlen;
1518 }
1519 if (offset > -1)
1520 offset++;
1521 newlen = 0;
1522 break;
1523 case OANYOF: /* may or may not invalidate offset */
1524 /* First, everything as OANY */
1525 if (newlen > g->mlen) { /* ends one */
1526 start = newstart;
1527 g->mlen = newlen;
1528 if (offset > -1) {
1529 g->moffset += offset;
1530 offset = newlen;
1531 } else
1532 g->moffset = offset;
1533 } else {
1534 if (offset > -1)
1535 offset += newlen;
1536 }
1537 if (offset > -1)
1538 offset++;
1539 newlen = 0;
1540 break;
1541 toohard:
1542 default:
1543 /* Anything here makes it impossible or too hard
1544 * to calculate the offset -- so we give up;
1545 * save the last known good offset, in case the
1546 * must sequence doesn't occur later.
1547 */
1548 if (newlen > g->mlen) { /* ends one */
1549 start = newstart;
1550 g->mlen = newlen;
1551 if (offset > -1)
1552 g->moffset += offset;
1553 else
1554 g->moffset = offset;
1555 }
1556 offset = -1;
1557 newlen = 0;
1558 break;
1559 }
1560 } while (OP(s) != OEND);
1561
1562 if (g->mlen == 0) { /* there isn't one */
1563 g->moffset = -1;
1564 return;
1565 }
1566
1567 /* turn it into a character string */
1568 g->must = malloc((size_t)g->mlen + 1);
1569 if (g->must == NULL) { /* argh; just forget it */
1570 g->mlen = 0;
1571 g->moffset = -1;
1572 return;
1573 }
1574 cp = g->must;
1575 scan = start;
1576 memset(&mbs, 0, sizeof(mbs));
1577 while (cp < g->must + g->mlen) {
1578 while (OP(s = *scan++) != OCHAR)
1579 continue;
1580 clen = wcrtomb(cp, OPND(s), &mbs);
1581 assert(clen != (size_t)-1);
1582 cp += clen;
1583 }
1584 assert(cp == g->must + g->mlen);
1585 *cp++ = '\0'; /* just on general principles */
1586}
1587
1588/*
1589 - altoffset - choose biggest offset among multiple choices
1590 == static int altoffset(sop *scan, int offset);
1591 *
1592 * Compute, recursively if necessary, the largest offset among multiple
1593 * re paths.
1594 */
1595static int
1596altoffset(sop *scan, int offset)
1597{
1598 int largest;
1599 int try;
1600 sop s;
1601
1602 /* If we gave up already on offsets, return */
1603 if (offset == -1)
1604 return -1;
1605
1606 largest = 0;
1607 try = 0;
1608 s = *scan++;
1609 while (OP(s) != O_QUEST && OP(s) != O_CH) {
1610 switch (OP(s)) {
1611 case OOR1:
1612 if (try > largest)
1613 largest = try;
1614 try = 0;
1615 break;
1616 case OQUEST_:
1617 case OCH_:
1618 try = altoffset(scan, try);
1619 if (try == -1)
1620 return -1;
1621 scan--;
1622 do {
1623 scan += OPND(s);
1624 s = *scan;
1625 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1626 OP(s) != OOR2)
1627 return -1;
1628 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1629 /* We must skip to the next position, or we'll
1630 * leave altoffset() too early.
1631 */
1632 scan++;
1633 break;
1634 case OANYOF:
1635 case OCHAR:
1636 case OANY:
1637 try++;
1638 case OBOW:
1639 case OEOW:
1640 case OLPAREN:
1641 case ORPAREN:
1642 case OOR2:
1643 break;
1644 default:
1645 try = -1;
1646 break;
1647 }
1648 if (try == -1)
1649 return -1;
1650 s = *scan++;
1651 }
1652
1653 if (try > largest)
1654 largest = try;
1655
1656 return largest+offset;
1657}
1658
1659/*
1660 - computejumps - compute char jumps for BM scan
1661 == static void computejumps(struct parse *p, struct re_guts *g);
1662 *
1663 * This algorithm assumes g->must exists and is has size greater than
1664 * zero. It's based on the algorithm found on Computer Algorithms by
1665 * Sara Baase.
1666 *
1667 * A char jump is the number of characters one needs to jump based on
1668 * the value of the character from the text that was mismatched.
1669 */
1670static void
1671computejumps(struct parse *p, struct re_guts *g)
1672{
1673 int ch;
1674 int mindex;
1675
1676 /* Avoid making errors worse */
1677 if (p->error != 0)
1678 return;
1679
1680 g->charjump = (int*) malloc((NC + 1) * sizeof(int));
1681 if (g->charjump == NULL) /* Not a fatal error */
1682 return;
1683 /* Adjust for signed chars, if necessary */
1684 g->charjump = &g->charjump[-(CHAR_MIN)];
1685
1686 /* If the character does not exist in the pattern, the jump
1687 * is equal to the number of characters in the pattern.
1688 */
1689 for (ch = CHAR_MIN; ch < (CHAR_MAX + 1); ch++)
1690 g->charjump[ch] = g->mlen;
1691
1692 /* If the character does exist, compute the jump that would
1693 * take us to the last character in the pattern equal to it
1694 * (notice that we match right to left, so that last character
1695 * is the first one that would be matched).
1696 */
1697 for (mindex = 0; mindex < g->mlen; mindex++)
1698 g->charjump[(int)g->must[mindex]] = g->mlen - mindex - 1;
1699}
1700
1701/*
1702 - computematchjumps - compute match jumps for BM scan
1703 == static void computematchjumps(struct parse *p, struct re_guts *g);
1704 *
1705 * This algorithm assumes g->must exists and is has size greater than
1706 * zero. It's based on the algorithm found on Computer Algorithms by
1707 * Sara Baase.
1708 *
1709 * A match jump is the number of characters one needs to advance based
1710 * on the already-matched suffix.
1711 * Notice that all values here are minus (g->mlen-1), because of the way
1712 * the search algorithm works.
1713 */
1714static void
1715computematchjumps(struct parse *p, struct re_guts *g)
1716{
1717 int mindex; /* General "must" iterator */
1718 int suffix; /* Keeps track of matching suffix */
1719 int ssuffix; /* Keeps track of suffixes' suffix */
1720 int* pmatches; /* pmatches[k] points to the next i
1721 * such that i+1...mlen is a substring
1722 * of k+1...k+mlen-i-1
1723 */
1724
1725 /* Avoid making errors worse */
1726 if (p->error != 0)
1727 return;
1728
1729 pmatches = (int*) malloc(g->mlen * sizeof(int));
1730 if (pmatches == NULL) {
1731 g->matchjump = NULL;
1732 return;
1733 }
1734
1735 g->matchjump = (int*) malloc(g->mlen * sizeof(int));
1736 if (g->matchjump == NULL) { /* Not a fatal error */
1737 free(pmatches);
1738 return;
1739 }
1740
1741 /* Set maximum possible jump for each character in the pattern */
1742 for (mindex = 0; mindex < g->mlen; mindex++)
1743 g->matchjump[mindex] = 2*g->mlen - mindex - 1;
1744
1745 /* Compute pmatches[] */
1746 for (mindex = g->mlen - 1, suffix = g->mlen; mindex >= 0;
1747 mindex--, suffix--) {
1748 pmatches[mindex] = suffix;
1749
1750 /* If a mismatch is found, interrupting the substring,
1751 * compute the matchjump for that position. If no
1752 * mismatch is found, then a text substring mismatched
1753 * against the suffix will also mismatch against the
1754 * substring.
1755 */
1756 while (suffix < g->mlen
1757 && g->must[mindex] != g->must[suffix]) {
1758 g->matchjump[suffix] = MIN(g->matchjump[suffix],
1759 g->mlen - mindex - 1);
1760 suffix = pmatches[suffix];
1761 }
1762 }
1763
1764 /* Compute the matchjump up to the last substring found to jump
1765 * to the beginning of the largest must pattern prefix matching
1766 * it's own suffix.
1767 */
1768 for (mindex = 0; mindex <= suffix; mindex++)
1769 g->matchjump[mindex] = MIN(g->matchjump[mindex],
1770 g->mlen + suffix - mindex);
1771
1772 ssuffix = pmatches[suffix];
1773 while (suffix < g->mlen) {
1774 while (suffix <= ssuffix && suffix < g->mlen) {
1775 g->matchjump[suffix] = MIN(g->matchjump[suffix],
1776 g->mlen + ssuffix - suffix);
1777 suffix++;
1778 }
1779 if (suffix < g->mlen)
1780 ssuffix = pmatches[ssuffix];
1781 }
1782
1783 free(pmatches);
1784}
1785
1786/*
1787 - pluscount - count + nesting
1788 == static sopno pluscount(struct parse *p, struct re_guts *g);
1789 */
1790static sopno /* nesting depth */
1791pluscount(struct parse *p, struct re_guts *g)
1792{
1793 sop *scan;
1794 sop s;
1795 sopno plusnest = 0;
1796 sopno maxnest = 0;
1797
1798 if (p->error != 0)
1799 return(0); /* there may not be an OEND */
1800
1801 scan = g->strip + 1;
1802 do {
1803 s = *scan++;
1804 switch (OP(s)) {
1805 case OPLUS_:
1806 plusnest++;
1807 break;
1808 case O_PLUS:
1809 if (plusnest > maxnest)
1810 maxnest = plusnest;
1811 plusnest--;
1812 break;
1813 }
1814 } while (OP(s) != OEND);
1815 if (plusnest != 0)
1816 g->iflags |= BAD;
1817 return(maxnest);
1818}
| 828 )
829 CHadd(p, cs, i);
830 }
831 }
832 }
833 break;
834 }
835}
836
837/*
838 - p_b_cclass - parse a character-class name and deal with it
839 == static void p_b_cclass(struct parse *p, cset *cs);
840 */
841static void
842p_b_cclass(struct parse *p, cset *cs)
843{
844 char *sp = p->next;
845 size_t len;
846 wctype_t wct;
847 char clname[16];
848
849 while (MORE() && isalpha((uch)PEEK()))
850 NEXT();
851 len = p->next - sp;
852 if (len >= sizeof(clname) - 1) {
853 SETERROR(REG_ECTYPE);
854 return;
855 }
856 memcpy(clname, sp, len);
857 clname[len] = '\0';
858 if ((wct = wctype(clname)) == 0) {
859 SETERROR(REG_ECTYPE);
860 return;
861 }
862 CHaddtype(p, cs, wct);
863}
864
865/*
866 - p_b_eclass - parse an equivalence-class name and deal with it
867 == static void p_b_eclass(struct parse *p, cset *cs);
868 *
869 * This implementation is incomplete. xxx
870 */
871static void
872p_b_eclass(struct parse *p, cset *cs)
873{
874 wint_t c;
875
876 c = p_b_coll_elem(p, '=');
877 CHadd(p, cs, c);
878}
879
880/*
881 - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
882 == static wint_t p_b_symbol(struct parse *p);
883 */
884static wint_t /* value of symbol */
885p_b_symbol(struct parse *p)
886{
887 wint_t value;
888
889 (void)REQUIRE(MORE(), REG_EBRACK);
890 if (!EATTWO('[', '.'))
891 return(WGETNEXT());
892
893 /* collating symbol */
894 value = p_b_coll_elem(p, '.');
895 (void)REQUIRE(EATTWO('.', ']'), REG_ECOLLATE);
896 return(value);
897}
898
899/*
900 - p_b_coll_elem - parse a collating-element name and look it up
901 == static wint_t p_b_coll_elem(struct parse *p, wint_t endc);
902 */
903static wint_t /* value of collating element */
904p_b_coll_elem(struct parse *p,
905 wint_t endc) /* name ended by endc,']' */
906{
907 char *sp = p->next;
908 struct cname *cp;
909 int len;
910 mbstate_t mbs;
911 wchar_t wc;
912 size_t clen;
913
914 while (MORE() && !SEETWO(endc, ']'))
915 NEXT();
916 if (!MORE()) {
917 SETERROR(REG_EBRACK);
918 return(0);
919 }
920 len = p->next - sp;
921 for (cp = cnames; cp->name != NULL; cp++)
922 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
923 return(cp->code); /* known name */
924 memset(&mbs, 0, sizeof(mbs));
925 if ((clen = mbrtowc(&wc, sp, len, &mbs)) == len)
926 return (wc); /* single character */
927 else if (clen == (size_t)-1 || clen == (size_t)-2)
928 SETERROR(REG_ILLSEQ);
929 else
930 SETERROR(REG_ECOLLATE); /* neither */
931 return(0);
932}
933
934/*
935 - othercase - return the case counterpart of an alphabetic
936 == static wint_t othercase(wint_t ch);
937 */
938static wint_t /* if no counterpart, return ch */
939othercase(wint_t ch)
940{
941 assert(iswalpha(ch));
942 if (iswupper(ch))
943 return(towlower(ch));
944 else if (iswlower(ch))
945 return(towupper(ch));
946 else /* peculiar, but could happen */
947 return(ch);
948}
949
950/*
951 - bothcases - emit a dualcase version of a two-case character
952 == static void bothcases(struct parse *p, wint_t ch);
953 *
954 * Boy, is this implementation ever a kludge...
955 */
956static void
957bothcases(struct parse *p, wint_t ch)
958{
959 char *oldnext = p->next;
960 char *oldend = p->end;
961 char bracket[3 + MB_LEN_MAX];
962 size_t n;
963 mbstate_t mbs;
964
965 assert(othercase(ch) != ch); /* p_bracket() would recurse */
966 p->next = bracket;
967 memset(&mbs, 0, sizeof(mbs));
968 n = wcrtomb(bracket, ch, &mbs);
969 assert(n != (size_t)-1);
970 bracket[n] = ']';
971 bracket[n + 1] = '\0';
972 p->end = bracket+n+1;
973 p_bracket(p);
974 assert(p->next == p->end);
975 p->next = oldnext;
976 p->end = oldend;
977}
978
979/*
980 - ordinary - emit an ordinary character
981 == static void ordinary(struct parse *p, wint_t ch);
982 */
983static void
984ordinary(struct parse *p, wint_t ch)
985{
986 cset *cs;
987
988 if ((p->g->cflags®_ICASE) && iswalpha(ch) && othercase(ch) != ch)
989 bothcases(p, ch);
990 else if ((ch & OPDMASK) == ch)
991 EMIT(OCHAR, ch);
992 else {
993 /*
994 * Kludge: character is too big to fit into an OCHAR operand.
995 * Emit a singleton set.
996 */
997 if ((cs = allocset(p)) == NULL)
998 return;
999 CHadd(p, cs, ch);
1000 EMIT(OANYOF, (int)(cs - p->g->sets));
1001 }
1002}
1003
1004/*
1005 - nonnewline - emit REG_NEWLINE version of OANY
1006 == static void nonnewline(struct parse *p);
1007 *
1008 * Boy, is this implementation ever a kludge...
1009 */
1010static void
1011nonnewline(struct parse *p)
1012{
1013 char *oldnext = p->next;
1014 char *oldend = p->end;
1015 char bracket[4];
1016
1017 p->next = bracket;
1018 p->end = bracket+3;
1019 bracket[0] = '^';
1020 bracket[1] = '\n';
1021 bracket[2] = ']';
1022 bracket[3] = '\0';
1023 p_bracket(p);
1024 assert(p->next == bracket+3);
1025 p->next = oldnext;
1026 p->end = oldend;
1027}
1028
1029/*
1030 - repeat - generate code for a bounded repetition, recursively if needed
1031 == static void repeat(struct parse *p, sopno start, int from, int to);
1032 */
1033static void
1034repeat(struct parse *p,
1035 sopno start, /* operand from here to end of strip */
1036 int from, /* repeated from this number */
1037 int to) /* to this number of times (maybe INFINITY) */
1038{
1039 sopno finish = HERE();
1040# define N 2
1041# define INF 3
1042# define REP(f, t) ((f)*8 + (t))
1043# define MAP(n) (((n) <= 1) ? (n) : ((n) == INFINITY) ? INF : N)
1044 sopno copy;
1045
1046 if (p->error != 0) /* head off possible runaway recursion */
1047 return;
1048
1049 assert(from <= to);
1050
1051 switch (REP(MAP(from), MAP(to))) {
1052 case REP(0, 0): /* must be user doing this */
1053 DROP(finish-start); /* drop the operand */
1054 break;
1055 case REP(0, 1): /* as x{1,1}? */
1056 case REP(0, N): /* as x{1,n}? */
1057 case REP(0, INF): /* as x{1,}? */
1058 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1059 INSERT(OCH_, start); /* offset is wrong... */
1060 repeat(p, start+1, 1, to);
1061 ASTERN(OOR1, start);
1062 AHEAD(start); /* ... fix it */
1063 EMIT(OOR2, 0);
1064 AHEAD(THERE());
1065 ASTERN(O_CH, THERETHERE());
1066 break;
1067 case REP(1, 1): /* trivial case */
1068 /* done */
1069 break;
1070 case REP(1, N): /* as x?x{1,n-1} */
1071 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1072 INSERT(OCH_, start);
1073 ASTERN(OOR1, start);
1074 AHEAD(start);
1075 EMIT(OOR2, 0); /* offset very wrong... */
1076 AHEAD(THERE()); /* ...so fix it */
1077 ASTERN(O_CH, THERETHERE());
1078 copy = dupl(p, start+1, finish+1);
1079 assert(copy == finish+4);
1080 repeat(p, copy, 1, to-1);
1081 break;
1082 case REP(1, INF): /* as x+ */
1083 INSERT(OPLUS_, start);
1084 ASTERN(O_PLUS, start);
1085 break;
1086 case REP(N, N): /* as xx{m-1,n-1} */
1087 copy = dupl(p, start, finish);
1088 repeat(p, copy, from-1, to-1);
1089 break;
1090 case REP(N, INF): /* as xx{n-1,INF} */
1091 copy = dupl(p, start, finish);
1092 repeat(p, copy, from-1, to);
1093 break;
1094 default: /* "can't happen" */
1095 SETERROR(REG_ASSERT); /* just in case */
1096 break;
1097 }
1098}
1099
1100/*
1101 - wgetnext - helper function for WGETNEXT() macro. Gets the next wide
1102 - character from the parse struct, signals a REG_ILLSEQ error if the
1103 - character can't be converted. Returns the number of bytes consumed.
1104 */
1105static wint_t
1106wgetnext(struct parse *p)
1107{
1108 mbstate_t mbs;
1109 wchar_t wc;
1110 size_t n;
1111
1112 memset(&mbs, 0, sizeof(mbs));
1113 n = mbrtowc(&wc, p->next, p->end - p->next, &mbs);
1114 if (n == (size_t)-1 || n == (size_t)-2) {
1115 SETERROR(REG_ILLSEQ);
1116 return (0);
1117 }
1118 if (n == 0)
1119 n = 1;
1120 p->next += n;
1121 return (wc);
1122}
1123
1124/*
1125 - seterr - set an error condition
1126 == static int seterr(struct parse *p, int e);
1127 */
1128static int /* useless but makes type checking happy */
1129seterr(struct parse *p, int e)
1130{
1131 if (p->error == 0) /* keep earliest error condition */
1132 p->error = e;
1133 p->next = nuls; /* try to bring things to a halt */
1134 p->end = nuls;
1135 return(0); /* make the return value well-defined */
1136}
1137
1138/*
1139 - allocset - allocate a set of characters for []
1140 == static cset *allocset(struct parse *p);
1141 */
1142static cset *
1143allocset(struct parse *p)
1144{
1145 cset *cs, *ncs;
1146
1147 ncs = realloc(p->g->sets, (p->g->ncsets + 1) * sizeof(*ncs));
1148 if (ncs == NULL) {
1149 SETERROR(REG_ESPACE);
1150 return (NULL);
1151 }
1152 p->g->sets = ncs;
1153 cs = &p->g->sets[p->g->ncsets++];
1154 memset(cs, 0, sizeof(*cs));
1155
1156 return(cs);
1157}
1158
1159/*
1160 - freeset - free a now-unused set
1161 == static void freeset(struct parse *p, cset *cs);
1162 */
1163static void
1164freeset(struct parse *p, cset *cs)
1165{
1166 cset *top = &p->g->sets[p->g->ncsets];
1167
1168 free(cs->wides);
1169 free(cs->ranges);
1170 free(cs->types);
1171 memset(cs, 0, sizeof(*cs));
1172 if (cs == top-1) /* recover only the easy case */
1173 p->g->ncsets--;
1174}
1175
1176/*
1177 - singleton - Determine whether a set contains only one character,
1178 - returning it if so, otherwise returning OUT.
1179 */
1180static wint_t
1181singleton(cset *cs)
1182{
1183 wint_t i, s, n;
1184
1185 for (i = n = 0; i < NC; i++)
1186 if (CHIN(cs, i)) {
1187 n++;
1188 s = i;
1189 }
1190 if (n == 1)
1191 return (s);
1192 if (cs->nwides == 1 && cs->nranges == 0 && cs->ntypes == 0 &&
1193 cs->icase == 0)
1194 return (cs->wides[0]);
1195 /* Don't bother handling the other cases. */
1196 return (OUT);
1197}
1198
1199/*
1200 - CHadd - add character to character set.
1201 */
1202static void
1203CHadd(struct parse *p, cset *cs, wint_t ch)
1204{
1205 wint_t nch, *newwides;
1206 assert(ch >= 0);
1207 if (ch < NC)
1208 cs->bmp[ch >> 3] |= 1 << (ch & 7);
1209 else {
1210 newwides = realloc(cs->wides, (cs->nwides + 1) *
1211 sizeof(*cs->wides));
1212 if (newwides == NULL) {
1213 SETERROR(REG_ESPACE);
1214 return;
1215 }
1216 cs->wides = newwides;
1217 cs->wides[cs->nwides++] = ch;
1218 }
1219 if (cs->icase) {
1220 if ((nch = towlower(ch)) < NC)
1221 cs->bmp[nch >> 3] |= 1 << (nch & 7);
1222 if ((nch = towupper(ch)) < NC)
1223 cs->bmp[nch >> 3] |= 1 << (nch & 7);
1224 }
1225}
1226
1227/*
1228 - CHaddrange - add all characters in the range [min,max] to a character set.
1229 */
1230static void
1231CHaddrange(struct parse *p, cset *cs, wint_t min, wint_t max)
1232{
1233 crange *newranges;
1234
1235 for (; min < NC && min <= max; min++)
1236 CHadd(p, cs, min);
1237 if (min >= max)
1238 return;
1239 newranges = realloc(cs->ranges, (cs->nranges + 1) *
1240 sizeof(*cs->ranges));
1241 if (newranges == NULL) {
1242 SETERROR(REG_ESPACE);
1243 return;
1244 }
1245 cs->ranges = newranges;
1246 cs->ranges[cs->nranges].min = min;
1247 cs->ranges[cs->nranges].max = max;
1248 cs->nranges++;
1249}
1250
1251/*
1252 - CHaddtype - add all characters of a certain type to a character set.
1253 */
1254static void
1255CHaddtype(struct parse *p, cset *cs, wctype_t wct)
1256{
1257 wint_t i;
1258 wctype_t *newtypes;
1259
1260 for (i = 0; i < NC; i++)
1261 if (iswctype(i, wct))
1262 CHadd(p, cs, i);
1263 newtypes = realloc(cs->types, (cs->ntypes + 1) *
1264 sizeof(*cs->types));
1265 if (newtypes == NULL) {
1266 SETERROR(REG_ESPACE);
1267 return;
1268 }
1269 cs->types = newtypes;
1270 cs->types[cs->ntypes++] = wct;
1271}
1272
1273/*
1274 - dupl - emit a duplicate of a bunch of sops
1275 == static sopno dupl(struct parse *p, sopno start, sopno finish);
1276 */
1277static sopno /* start of duplicate */
1278dupl(struct parse *p,
1279 sopno start, /* from here */
1280 sopno finish) /* to this less one */
1281{
1282 sopno ret = HERE();
1283 sopno len = finish - start;
1284
1285 assert(finish >= start);
1286 if (len == 0)
1287 return(ret);
1288 if (!enlarge(p, p->ssize + len)) /* this many unexpected additions */
1289 return(ret);
1290 (void) memcpy((char *)(p->strip + p->slen),
1291 (char *)(p->strip + start), (size_t)len*sizeof(sop));
1292 p->slen += len;
1293 return(ret);
1294}
1295
1296/*
1297 - doemit - emit a strip operator
1298 == static void doemit(struct parse *p, sop op, size_t opnd);
1299 *
1300 * It might seem better to implement this as a macro with a function as
1301 * hard-case backup, but it's just too big and messy unless there are
1302 * some changes to the data structures. Maybe later.
1303 */
1304static void
1305doemit(struct parse *p, sop op, size_t opnd)
1306{
1307 /* avoid making error situations worse */
1308 if (p->error != 0)
1309 return;
1310
1311 /* deal with oversize operands ("can't happen", more or less) */
1312 assert(opnd < 1<<OPSHIFT);
1313
1314 /* deal with undersized strip */
1315 if (p->slen >= p->ssize)
1316 if (!enlarge(p, (p->ssize+1) / 2 * 3)) /* +50% */
1317 return;
1318
1319 /* finally, it's all reduced to the easy case */
1320 p->strip[p->slen++] = SOP(op, opnd);
1321}
1322
1323/*
1324 - doinsert - insert a sop into the strip
1325 == static void doinsert(struct parse *p, sop op, size_t opnd, sopno pos);
1326 */
1327static void
1328doinsert(struct parse *p, sop op, size_t opnd, sopno pos)
1329{
1330 sopno sn;
1331 sop s;
1332 int i;
1333
1334 /* avoid making error situations worse */
1335 if (p->error != 0)
1336 return;
1337
1338 sn = HERE();
1339 EMIT(op, opnd); /* do checks, ensure space */
1340 assert(HERE() == sn+1);
1341 s = p->strip[sn];
1342
1343 /* adjust paren pointers */
1344 assert(pos > 0);
1345 for (i = 1; i < NPAREN; i++) {
1346 if (p->pbegin[i] >= pos) {
1347 p->pbegin[i]++;
1348 }
1349 if (p->pend[i] >= pos) {
1350 p->pend[i]++;
1351 }
1352 }
1353
1354 memmove((char *)&p->strip[pos+1], (char *)&p->strip[pos],
1355 (HERE()-pos-1)*sizeof(sop));
1356 p->strip[pos] = s;
1357}
1358
1359/*
1360 - dofwd - complete a forward reference
1361 == static void dofwd(struct parse *p, sopno pos, sop value);
1362 */
1363static void
1364dofwd(struct parse *p, sopno pos, sop value)
1365{
1366 /* avoid making error situations worse */
1367 if (p->error != 0)
1368 return;
1369
1370 assert(value < 1<<OPSHIFT);
1371 p->strip[pos] = OP(p->strip[pos]) | value;
1372}
1373
1374/*
1375 - enlarge - enlarge the strip
1376 == static int enlarge(struct parse *p, sopno size);
1377 */
1378static int
1379enlarge(struct parse *p, sopno size)
1380{
1381 sop *sp;
1382
1383 if (p->ssize >= size)
1384 return 1;
1385
1386 sp = (sop *)realloc(p->strip, size*sizeof(sop));
1387 if (sp == NULL) {
1388 SETERROR(REG_ESPACE);
1389 return 0;
1390 }
1391 p->strip = sp;
1392 p->ssize = size;
1393 return 1;
1394}
1395
1396/*
1397 - stripsnug - compact the strip
1398 == static void stripsnug(struct parse *p, struct re_guts *g);
1399 */
1400static void
1401stripsnug(struct parse *p, struct re_guts *g)
1402{
1403 g->nstates = p->slen;
1404 g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1405 if (g->strip == NULL) {
1406 SETERROR(REG_ESPACE);
1407 g->strip = p->strip;
1408 }
1409}
1410
1411/*
1412 - findmust - fill in must and mlen with longest mandatory literal string
1413 == static void findmust(struct parse *p, struct re_guts *g);
1414 *
1415 * This algorithm could do fancy things like analyzing the operands of |
1416 * for common subsequences. Someday. This code is simple and finds most
1417 * of the interesting cases.
1418 *
1419 * Note that must and mlen got initialized during setup.
1420 */
1421static void
1422findmust(struct parse *p, struct re_guts *g)
1423{
1424 sop *scan;
1425 sop *start = NULL;
1426 sop *newstart = NULL;
1427 sopno newlen;
1428 sop s;
1429 char *cp;
1430 int offset;
1431 char buf[MB_LEN_MAX];
1432 size_t clen;
1433 mbstate_t mbs;
1434
1435 /* avoid making error situations worse */
1436 if (p->error != 0)
1437 return;
1438
1439 /*
1440 * It's not generally safe to do a ``char'' substring search on
1441 * multibyte character strings, but it's safe for at least
1442 * UTF-8 (see RFC 3629).
1443 */
1444 if (MB_CUR_MAX > 1 &&
1445 strcmp(_CurrentRuneLocale->__encoding, "UTF-8") != 0)
1446 return;
1447
1448 /* find the longest OCHAR sequence in strip */
1449 newlen = 0;
1450 offset = 0;
1451 g->moffset = 0;
1452 scan = g->strip + 1;
1453 do {
1454 s = *scan++;
1455 switch (OP(s)) {
1456 case OCHAR: /* sequence member */
1457 if (newlen == 0) { /* new sequence */
1458 memset(&mbs, 0, sizeof(mbs));
1459 newstart = scan - 1;
1460 }
1461 clen = wcrtomb(buf, OPND(s), &mbs);
1462 if (clen == (size_t)-1)
1463 goto toohard;
1464 newlen += clen;
1465 break;
1466 case OPLUS_: /* things that don't break one */
1467 case OLPAREN:
1468 case ORPAREN:
1469 break;
1470 case OQUEST_: /* things that must be skipped */
1471 case OCH_:
1472 offset = altoffset(scan, offset);
1473 scan--;
1474 do {
1475 scan += OPND(s);
1476 s = *scan;
1477 /* assert() interferes w debug printouts */
1478 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1479 OP(s) != OOR2) {
1480 g->iflags |= BAD;
1481 return;
1482 }
1483 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1484 /* FALLTHROUGH */
1485 case OBOW: /* things that break a sequence */
1486 case OEOW:
1487 case OBOL:
1488 case OEOL:
1489 case O_QUEST:
1490 case O_CH:
1491 case OEND:
1492 if (newlen > g->mlen) { /* ends one */
1493 start = newstart;
1494 g->mlen = newlen;
1495 if (offset > -1) {
1496 g->moffset += offset;
1497 offset = newlen;
1498 } else
1499 g->moffset = offset;
1500 } else {
1501 if (offset > -1)
1502 offset += newlen;
1503 }
1504 newlen = 0;
1505 break;
1506 case OANY:
1507 if (newlen > g->mlen) { /* ends one */
1508 start = newstart;
1509 g->mlen = newlen;
1510 if (offset > -1) {
1511 g->moffset += offset;
1512 offset = newlen;
1513 } else
1514 g->moffset = offset;
1515 } else {
1516 if (offset > -1)
1517 offset += newlen;
1518 }
1519 if (offset > -1)
1520 offset++;
1521 newlen = 0;
1522 break;
1523 case OANYOF: /* may or may not invalidate offset */
1524 /* First, everything as OANY */
1525 if (newlen > g->mlen) { /* ends one */
1526 start = newstart;
1527 g->mlen = newlen;
1528 if (offset > -1) {
1529 g->moffset += offset;
1530 offset = newlen;
1531 } else
1532 g->moffset = offset;
1533 } else {
1534 if (offset > -1)
1535 offset += newlen;
1536 }
1537 if (offset > -1)
1538 offset++;
1539 newlen = 0;
1540 break;
1541 toohard:
1542 default:
1543 /* Anything here makes it impossible or too hard
1544 * to calculate the offset -- so we give up;
1545 * save the last known good offset, in case the
1546 * must sequence doesn't occur later.
1547 */
1548 if (newlen > g->mlen) { /* ends one */
1549 start = newstart;
1550 g->mlen = newlen;
1551 if (offset > -1)
1552 g->moffset += offset;
1553 else
1554 g->moffset = offset;
1555 }
1556 offset = -1;
1557 newlen = 0;
1558 break;
1559 }
1560 } while (OP(s) != OEND);
1561
1562 if (g->mlen == 0) { /* there isn't one */
1563 g->moffset = -1;
1564 return;
1565 }
1566
1567 /* turn it into a character string */
1568 g->must = malloc((size_t)g->mlen + 1);
1569 if (g->must == NULL) { /* argh; just forget it */
1570 g->mlen = 0;
1571 g->moffset = -1;
1572 return;
1573 }
1574 cp = g->must;
1575 scan = start;
1576 memset(&mbs, 0, sizeof(mbs));
1577 while (cp < g->must + g->mlen) {
1578 while (OP(s = *scan++) != OCHAR)
1579 continue;
1580 clen = wcrtomb(cp, OPND(s), &mbs);
1581 assert(clen != (size_t)-1);
1582 cp += clen;
1583 }
1584 assert(cp == g->must + g->mlen);
1585 *cp++ = '\0'; /* just on general principles */
1586}
1587
1588/*
1589 - altoffset - choose biggest offset among multiple choices
1590 == static int altoffset(sop *scan, int offset);
1591 *
1592 * Compute, recursively if necessary, the largest offset among multiple
1593 * re paths.
1594 */
1595static int
1596altoffset(sop *scan, int offset)
1597{
1598 int largest;
1599 int try;
1600 sop s;
1601
1602 /* If we gave up already on offsets, return */
1603 if (offset == -1)
1604 return -1;
1605
1606 largest = 0;
1607 try = 0;
1608 s = *scan++;
1609 while (OP(s) != O_QUEST && OP(s) != O_CH) {
1610 switch (OP(s)) {
1611 case OOR1:
1612 if (try > largest)
1613 largest = try;
1614 try = 0;
1615 break;
1616 case OQUEST_:
1617 case OCH_:
1618 try = altoffset(scan, try);
1619 if (try == -1)
1620 return -1;
1621 scan--;
1622 do {
1623 scan += OPND(s);
1624 s = *scan;
1625 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1626 OP(s) != OOR2)
1627 return -1;
1628 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1629 /* We must skip to the next position, or we'll
1630 * leave altoffset() too early.
1631 */
1632 scan++;
1633 break;
1634 case OANYOF:
1635 case OCHAR:
1636 case OANY:
1637 try++;
1638 case OBOW:
1639 case OEOW:
1640 case OLPAREN:
1641 case ORPAREN:
1642 case OOR2:
1643 break;
1644 default:
1645 try = -1;
1646 break;
1647 }
1648 if (try == -1)
1649 return -1;
1650 s = *scan++;
1651 }
1652
1653 if (try > largest)
1654 largest = try;
1655
1656 return largest+offset;
1657}
1658
1659/*
1660 - computejumps - compute char jumps for BM scan
1661 == static void computejumps(struct parse *p, struct re_guts *g);
1662 *
1663 * This algorithm assumes g->must exists and is has size greater than
1664 * zero. It's based on the algorithm found on Computer Algorithms by
1665 * Sara Baase.
1666 *
1667 * A char jump is the number of characters one needs to jump based on
1668 * the value of the character from the text that was mismatched.
1669 */
1670static void
1671computejumps(struct parse *p, struct re_guts *g)
1672{
1673 int ch;
1674 int mindex;
1675
1676 /* Avoid making errors worse */
1677 if (p->error != 0)
1678 return;
1679
1680 g->charjump = (int*) malloc((NC + 1) * sizeof(int));
1681 if (g->charjump == NULL) /* Not a fatal error */
1682 return;
1683 /* Adjust for signed chars, if necessary */
1684 g->charjump = &g->charjump[-(CHAR_MIN)];
1685
1686 /* If the character does not exist in the pattern, the jump
1687 * is equal to the number of characters in the pattern.
1688 */
1689 for (ch = CHAR_MIN; ch < (CHAR_MAX + 1); ch++)
1690 g->charjump[ch] = g->mlen;
1691
1692 /* If the character does exist, compute the jump that would
1693 * take us to the last character in the pattern equal to it
1694 * (notice that we match right to left, so that last character
1695 * is the first one that would be matched).
1696 */
1697 for (mindex = 0; mindex < g->mlen; mindex++)
1698 g->charjump[(int)g->must[mindex]] = g->mlen - mindex - 1;
1699}
1700
1701/*
1702 - computematchjumps - compute match jumps for BM scan
1703 == static void computematchjumps(struct parse *p, struct re_guts *g);
1704 *
1705 * This algorithm assumes g->must exists and is has size greater than
1706 * zero. It's based on the algorithm found on Computer Algorithms by
1707 * Sara Baase.
1708 *
1709 * A match jump is the number of characters one needs to advance based
1710 * on the already-matched suffix.
1711 * Notice that all values here are minus (g->mlen-1), because of the way
1712 * the search algorithm works.
1713 */
1714static void
1715computematchjumps(struct parse *p, struct re_guts *g)
1716{
1717 int mindex; /* General "must" iterator */
1718 int suffix; /* Keeps track of matching suffix */
1719 int ssuffix; /* Keeps track of suffixes' suffix */
1720 int* pmatches; /* pmatches[k] points to the next i
1721 * such that i+1...mlen is a substring
1722 * of k+1...k+mlen-i-1
1723 */
1724
1725 /* Avoid making errors worse */
1726 if (p->error != 0)
1727 return;
1728
1729 pmatches = (int*) malloc(g->mlen * sizeof(int));
1730 if (pmatches == NULL) {
1731 g->matchjump = NULL;
1732 return;
1733 }
1734
1735 g->matchjump = (int*) malloc(g->mlen * sizeof(int));
1736 if (g->matchjump == NULL) { /* Not a fatal error */
1737 free(pmatches);
1738 return;
1739 }
1740
1741 /* Set maximum possible jump for each character in the pattern */
1742 for (mindex = 0; mindex < g->mlen; mindex++)
1743 g->matchjump[mindex] = 2*g->mlen - mindex - 1;
1744
1745 /* Compute pmatches[] */
1746 for (mindex = g->mlen - 1, suffix = g->mlen; mindex >= 0;
1747 mindex--, suffix--) {
1748 pmatches[mindex] = suffix;
1749
1750 /* If a mismatch is found, interrupting the substring,
1751 * compute the matchjump for that position. If no
1752 * mismatch is found, then a text substring mismatched
1753 * against the suffix will also mismatch against the
1754 * substring.
1755 */
1756 while (suffix < g->mlen
1757 && g->must[mindex] != g->must[suffix]) {
1758 g->matchjump[suffix] = MIN(g->matchjump[suffix],
1759 g->mlen - mindex - 1);
1760 suffix = pmatches[suffix];
1761 }
1762 }
1763
1764 /* Compute the matchjump up to the last substring found to jump
1765 * to the beginning of the largest must pattern prefix matching
1766 * it's own suffix.
1767 */
1768 for (mindex = 0; mindex <= suffix; mindex++)
1769 g->matchjump[mindex] = MIN(g->matchjump[mindex],
1770 g->mlen + suffix - mindex);
1771
1772 ssuffix = pmatches[suffix];
1773 while (suffix < g->mlen) {
1774 while (suffix <= ssuffix && suffix < g->mlen) {
1775 g->matchjump[suffix] = MIN(g->matchjump[suffix],
1776 g->mlen + ssuffix - suffix);
1777 suffix++;
1778 }
1779 if (suffix < g->mlen)
1780 ssuffix = pmatches[ssuffix];
1781 }
1782
1783 free(pmatches);
1784}
1785
1786/*
1787 - pluscount - count + nesting
1788 == static sopno pluscount(struct parse *p, struct re_guts *g);
1789 */
1790static sopno /* nesting depth */
1791pluscount(struct parse *p, struct re_guts *g)
1792{
1793 sop *scan;
1794 sop s;
1795 sopno plusnest = 0;
1796 sopno maxnest = 0;
1797
1798 if (p->error != 0)
1799 return(0); /* there may not be an OEND */
1800
1801 scan = g->strip + 1;
1802 do {
1803 s = *scan++;
1804 switch (OP(s)) {
1805 case OPLUS_:
1806 plusnest++;
1807 break;
1808 case O_PLUS:
1809 if (plusnest > maxnest)
1810 maxnest = plusnest;
1811 plusnest--;
1812 break;
1813 }
1814 } while (OP(s) != OEND);
1815 if (plusnest != 0)
1816 g->iflags |= BAD;
1817 return(maxnest);
1818}
|