1 /* Copyright 1992, 1993, 1994 Henry Spencer. All rights reserved.
2 See file COPYRIGHT for details.
3
4 This file was modified by Oracle on 2015-05-18 for 32-bit compatibility.
5
6 Modifications Copyright (c) 2015, 2021, Oracle and/or its affiliates. All rights
7 reserved. */
8
9 #include <my_global.h>
10 #include <m_string.h>
11 #include <m_ctype.h>
12 #ifdef _WIN32
13 #include <limits.h>
14 #endif
15
16 #include "my_regex.h"
17 #include "utils.h"
18 #include "regex2.h"
19
20 #include "cclass.h"
21 #include "cname.h"
22
23 /*
24 * parse structure, passed up and down to avoid global variables and
25 * other clumsinesses
26 */
27 struct parse {
28 char *next; /* next character in RE */
29 char *end; /* end of string (-> NUL normally) */
30 int error; /* has an error been seen? */
31 sop *strip; /* malloced strip */
32 sopno ssize; /* malloced strip size (allocated) */
33 sopno slen; /* malloced strip length (used) */
34 int ncsalloc; /* number of csets allocated */
35 struct re_guts *g;
36 # define NPAREN 10 /* we need to remember () 1-9 for back refs */
37 sopno pbegin[NPAREN]; /* -> ( ([0] unused) */
38 sopno pend[NPAREN]; /* -> ) ([0] unused) */
39 const CHARSET_INFO *charset; /* for ctype things */
40 };
41
42 /* Check if there is enough stack space for recursion. */
43 my_regex_stack_check_t my_regex_enough_mem_in_stack= NULL;
44
45 #include "regcomp.ih"
46
47 static char nuls[10]; /* place to point scanner in event of error */
48
49 struct cclass cclasses[CCLASS_LAST+1]= {
50 { "alnum", "","", _MY_U | _MY_L | _MY_NMR},
51 { "alpha", "","", _MY_U | _MY_L },
52 { "blank", "","", _MY_B },
53 { "cntrl", "","", _MY_CTR },
54 { "digit", "","", _MY_NMR },
55 { "graph", "","", _MY_PNT | _MY_U | _MY_L | _MY_NMR},
56 { "lower", "","", _MY_L },
57 { "print", "","", _MY_PNT | _MY_U | _MY_L | _MY_NMR | _MY_B },
58 { "punct", "","", _MY_PNT },
59 { "space", "","", _MY_SPC },
60 { "upper", "","", _MY_U },
61 { "xdigit", "","", _MY_X },
62 { NULL,NULL,NULL, 0 }
63 };
64
65 /*
66 * macros for use with parse structure
67 * BEWARE: these know that the parse structure is named `p' !!!
68 */
69 #define PEEK() (*p->next)
70 #define PEEK2() (*(p->next+1))
71 #define MORE() (p->next < p->end)
72 #define MORE2() (p->next+1 < p->end)
73 #define SEE(c) (MORE() && PEEK() == (c))
74 #define SEETWO(a, b) (MORE() && MORE2() && PEEK() == (a) && PEEK2() == (b))
75 #define EAT(c) ((SEE(c)) ? (NEXT(), 1) : 0)
76 #define EATTWO(a, b) ((SEETWO(a, b)) ? (NEXT2(), 1) : 0)
77 #define NEXT() (p->next++)
78 #define NEXT2() (p->next += 2)
79 #define NEXTn(n) (p->next += (n))
80 #define GETNEXT() (*p->next++)
81 #define SETERROR(e) seterr(p, (e))
82 #define REQUIRE(co, e) ((co) || SETERROR(e))
83 #define MUSTSEE(c, e) (REQUIRE(MORE() && PEEK() == (c), e))
84 #define MUSTEAT(c, e) (REQUIRE(MORE() && GETNEXT() == (c), e))
85 #define MUSTNOTSEE(c, e) (REQUIRE(!MORE() || PEEK() != (c), e))
86 #define EMIT(op, sopnd) doemit(p, (sop)(op), (size_t)(sopnd))
87 #define INSERT(op, pos) doinsert(p, (sop)(op), HERE()-(pos)+1, pos)
88 #define AHEAD(pos) dofwd(p, pos, HERE()-(pos))
89 #define ASTERN(sop, pos) EMIT(sop, HERE()-pos)
90 #define HERE() (p->slen)
91 #define THERE() (p->slen - 1)
92 #define THERETHERE() (p->slen - 2)
93 #define DROP(n) (p->slen -= (n))
94
95 #ifndef NDEBUG
96 static int never = 0; /* for use in asserts; shuts lint up */
97 #else
98 #define never 0 /* some <assert.h>s have bugs too */
99 #endif
100
101 /*
102 - regcomp - interface for parser and compilation
103 = extern int regcomp(regex_t *, const char *, int);
104 = #define MY_REG_BASIC 0000
105 = #define MY_REG_EXTENDED 0001
106 = #define MY_REG_ICASE 0002
107 = #define MY_REG_NOSUB 0004
108 = #define MY_REG_NEWLINE 0010
109 = #define MY_REG_NOSPEC 0020
110 = #define MY_REG_PEND 0040
111 = #define MY_REG_DUMP 0200
112 */
113 int /* 0 success, otherwise MY_REG_something */
my_regcomp(preg,pattern,cflags,charset)114 my_regcomp(preg, pattern, cflags, charset)
115 my_regex_t *preg;
116 const char *pattern;
117 int cflags;
118 const CHARSET_INFO *charset;
119 {
120 struct parse pa;
121 struct re_guts *g;
122 struct parse *p = &pa;
123 int i;
124 size_t len;
125 #ifdef REDEBUG
126 # define GOODFLAGS(f) (f)
127 #else
128 # define GOODFLAGS(f) ((f)&~MY_REG_DUMP)
129 #endif
130
131 my_regex_init(charset, NULL); /* Init cclass if neaded */
132 preg->charset=charset;
133 cflags = GOODFLAGS(cflags);
134 if ((cflags&MY_REG_EXTENDED) && (cflags&MY_REG_NOSPEC))
135 return(MY_REG_INVARG);
136
137 if (cflags&MY_REG_PEND) {
138 if (preg->re_endp < pattern)
139 return(MY_REG_INVARG);
140 len = preg->re_endp - pattern;
141 } else
142 len = strlen((char *)pattern);
143
144 /*
145 Find the maximum len we can safely process
146 without a rollover and a mis-malloc.
147 p->ssize is a sopno is a long (32+ bit signed);
148 size_t is 16+ bit unsigned.
149 */
150 {
151 size_t new_ssize = len / (size_t)2 * (size_t)3 + (size_t)1; /* ugh */
152 if ((new_ssize < len) || /* size_t rolled over */
153 ((SIZE_T_MAX / sizeof(sop)) < new_ssize) || /* malloc arg */
154 (new_ssize > LONG_MAX)) /* won't fit in ssize */
155 return(MY_REG_ESPACE); /* MY_REG_ESPACE or MY_REG_INVARG */
156 p->ssize = new_ssize;
157 }
158
159 /* do the mallocs early so failure handling is easy */
160 g = (struct re_guts *)malloc(sizeof(struct re_guts) +
161 (NC-1)*sizeof(cat_t));
162 if (g == NULL)
163 return(MY_REG_ESPACE);
164 p->strip = (sop *)malloc(p->ssize * sizeof(sop));
165 p->slen = 0;
166 if (p->strip == NULL) {
167 free((char *)g);
168 return(MY_REG_ESPACE);
169 }
170
171 /* set things up */
172 p->g = g;
173 p->next = (char *)pattern; /* convenience; we do not modify it */
174 p->end = p->next + len;
175 p->error = 0;
176 p->ncsalloc = 0;
177 p->charset = preg->charset;
178 for (i = 0; i < NPAREN; i++) {
179 p->pbegin[i] = 0;
180 p->pend[i] = 0;
181 }
182 g->csetsize = NC;
183 g->sets = NULL;
184 g->setbits = NULL;
185 g->ncsets = 0;
186 g->cflags = cflags;
187 g->iflags = 0;
188 g->nbol = 0;
189 g->neol = 0;
190 g->must = NULL;
191 g->mlen = 0;
192 g->nsub = 0;
193 g->ncategories = 1; /* category 0 is "everything else" */
194 g->categories = &g->catspace[-(CHAR_MIN)];
195 (void) memset((char *)g->catspace, 0, NC*sizeof(cat_t));
196 g->backrefs = 0;
197
198 /* do it */
199 EMIT(OEND, 0);
200 g->firststate = THERE();
201 if (cflags&MY_REG_EXTENDED)
202 p_ere(p, OUT);
203 else if (cflags&MY_REG_NOSPEC)
204 p_str(p);
205 else
206 p_bre(p, OUT, OUT);
207 EMIT(OEND, 0);
208 g->laststate = THERE();
209
210 /* tidy up loose ends and fill things in */
211 categorize(p, g);
212 stripsnug(p, g);
213 findmust(p, g);
214 g->nplus = pluscount(p, g);
215 g->magic = MAGIC2;
216 preg->re_nsub = g->nsub;
217 preg->re_g = g;
218 preg->re_magic = MAGIC1;
219 #ifndef REDEBUG
220 /* not debugging, so can't rely on the assert() in regexec() */
221 if (g->iflags&BAD)
222 SETERROR(MY_REG_ASSERT);
223 #endif
224
225 /* win or lose, we're done */
226 if (p->error != 0) /* lose */
227 my_regfree(preg);
228 return(p->error);
229 }
230
231 /*
232 - p_ere - ERE parser top level, concatenation and alternation
233 == static void p_ere(register struct parse *p, int stop);
234 */
235 static void
p_ere(p,stop)236 p_ere(p, stop)
237 struct parse *p;
238 int stop; /* character this ERE should end at */
239 {
240 char c;
241 sopno prevback= 0;
242 sopno prevfwd= 0;
243 sopno conc;
244 int first = 1; /* is this the first alternative? */
245
246 for (;;) {
247 /* do a bunch of concatenated expressions */
248 conc = HERE();
249 while (MORE() && (c = PEEK()) != '|' && c != stop)
250 {
251 if (my_regex_enough_mem_in_stack &&
252 my_regex_enough_mem_in_stack(0))
253 {
254 SETERROR(MY_REG_ESPACE);
255 return;
256 }
257 p_ere_exp(p);
258 }
259 if(REQUIRE(HERE() != conc, MY_REG_EMPTY)) {}/* require nonempty */
260
261 if (!EAT('|'))
262 break; /* NOTE BREAK OUT */
263
264 if (first) {
265 INSERT(OCH_, conc); /* offset is wrong */
266 prevfwd = conc;
267 prevback = conc;
268 first = 0;
269 }
270 ASTERN(OOR1, prevback);
271 prevback = THERE();
272 AHEAD(prevfwd); /* fix previous offset */
273 prevfwd = HERE();
274 EMIT(OOR2, 0); /* offset is very wrong */
275 }
276
277 if (!first) { /* tail-end fixups */
278 AHEAD(prevfwd);
279 ASTERN(O_CH, prevback);
280 }
281
282 assert(!MORE() || SEE(stop));
283 }
284
285 /*
286 - p_ere_exp - parse one subERE, an atom possibly followed by a repetition op
287 == static void p_ere_exp(register struct parse *p);
288 */
289 static void
p_ere_exp(p)290 p_ere_exp(p)
291 struct parse *p;
292 {
293 char c;
294 sopno pos;
295 int count;
296 int count2;
297 sopno subno;
298 int wascaret = 0;
299
300 assert(MORE()); /* caller should have ensured this */
301 c = GETNEXT();
302
303 pos = HERE();
304 switch (c) {
305 case '(':
306 if(REQUIRE(MORE(), MY_REG_EPAREN)) {}
307 p->g->nsub++;
308 subno = (sopno) p->g->nsub;
309 if (subno < NPAREN)
310 p->pbegin[subno] = HERE();
311 EMIT(OLPAREN, subno);
312 if (!SEE(')'))
313 p_ere(p, ')');
314 if (subno < NPAREN) {
315 p->pend[subno] = HERE();
316 assert(p->pend[subno] != 0);
317 }
318 EMIT(ORPAREN, subno);
319 if(MUSTEAT(')', MY_REG_EPAREN)) {}
320 break;
321 case '^':
322 EMIT(OBOL, 0);
323 p->g->iflags |= USEBOL;
324 p->g->nbol++;
325 wascaret = 1;
326 break;
327 case '$':
328 EMIT(OEOL, 0);
329 p->g->iflags |= USEEOL;
330 p->g->neol++;
331 break;
332 case '|':
333 SETERROR(MY_REG_EMPTY);
334 break;
335 case '*':
336 case '+':
337 case '?':
338 SETERROR(MY_REG_BADRPT);
339 break;
340 case '.':
341 if (p->g->cflags&MY_REG_NEWLINE)
342 nonnewline(p);
343 else
344 EMIT(OANY, 0);
345 break;
346 case '[':
347 p_bracket(p);
348 break;
349 case '\\':
350 if(REQUIRE(MORE(), MY_REG_EESCAPE)) {}
351 c = GETNEXT();
352 ordinary(p, c);
353 break;
354 case '{': /* okay as ordinary except if digit follows */
355 if(REQUIRE(!MORE() || !my_isdigit(p->charset,PEEK()), MY_REG_BADRPT)) {}
356 /* FALLTHROUGH */
357 default:
358 ordinary(p, c);
359 break;
360 }
361
362 if (!MORE())
363 return;
364 c = PEEK();
365 /* we call { a repetition if followed by a digit */
366 if (!( c == '*' || c == '+' || c == '?' ||
367 (c == '{' && MORE2() &&
368 my_isdigit(p->charset,PEEK2())) ))
369 return; /* no repetition, we're done */
370 NEXT();
371
372 if(REQUIRE(!wascaret, MY_REG_BADRPT)) {}
373 switch (c) {
374 case '*': /* implemented as +? */
375 /* this case does not require the (y|) trick, noKLUDGE */
376 INSERT(OPLUS_, pos);
377 ASTERN(O_PLUS, pos);
378 INSERT(OQUEST_, pos);
379 ASTERN(O_QUEST, pos);
380 break;
381 case '+':
382 INSERT(OPLUS_, pos);
383 ASTERN(O_PLUS, pos);
384 break;
385 case '?':
386 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
387 INSERT(OCH_, pos); /* offset slightly wrong */
388 ASTERN(OOR1, pos); /* this one's right */
389 AHEAD(pos); /* fix the OCH_ */
390 EMIT(OOR2, 0); /* offset very wrong... */
391 AHEAD(THERE()); /* ...so fix it */
392 ASTERN(O_CH, THERETHERE());
393 break;
394 case '{':
395 count = p_count(p);
396 if (EAT(',')) {
397 if (my_isdigit(p->charset,PEEK())) {
398 count2 = p_count(p);
399 if(REQUIRE(count <= count2, MY_REG_BADBR)) {}
400 } else /* single number with comma */
401 count2 = RE_INFINITY;
402 } else /* just a single number */
403 count2 = count;
404 repeat(p, pos, count, count2);
405 if (!EAT('}')) { /* error heuristics */
406 while (MORE() && PEEK() != '}')
407 NEXT();
408 if(REQUIRE(MORE(), MY_REG_EBRACE)) {}
409 SETERROR(MY_REG_BADBR);
410 }
411 break;
412 }
413
414 if (!MORE())
415 return;
416 c = PEEK();
417 if (!( c == '*' || c == '+' || c == '?' ||
418 (c == '{' && MORE2() &&
419 my_isdigit(p->charset,PEEK2())) ) )
420 return;
421 SETERROR(MY_REG_BADRPT);
422 }
423
424 /*
425 - p_str - string (no metacharacters) "parser"
426 == static void p_str(register struct parse *p);
427 */
428 static void
p_str(p)429 p_str(p)
430 struct parse *p;
431 {
432 if(REQUIRE(MORE(), MY_REG_EMPTY)) {}
433 while (MORE())
434 ordinary(p, GETNEXT());
435 }
436
437 /*
438 - p_bre - BRE parser top level, anchoring and concatenation
439 == static void p_bre(register struct parse *p, register int end1, \
440 == register int end2);
441 * Giving end1 as OUT essentially eliminates the end1/end2 check.
442 *
443 * This implementation is a bit of a kludge, in that a trailing $ is first
444 * taken as an ordinary character and then revised to be an anchor. The
445 * only undesirable side effect is that '$' gets included as a character
446 * category in such cases. This is fairly harmless; not worth fixing.
447 * The amount of lookahead needed to avoid this kludge is excessive.
448 */
449 static void
p_bre(p,end1,end2)450 p_bre(p, end1, end2)
451 struct parse *p;
452 int end1; /* first terminating character */
453 int end2; /* second terminating character */
454 {
455 sopno start = HERE();
456 int first = 1; /* first subexpression? */
457 int wasdollar = 0;
458
459 if (EAT('^')) {
460 EMIT(OBOL, 0);
461 p->g->iflags |= USEBOL;
462 p->g->nbol++;
463 }
464 while (MORE() && !SEETWO(end1, end2)) {
465 wasdollar = p_simp_re(p, first);
466 first = 0;
467 }
468 if (wasdollar) { /* oops, that was a trailing anchor */
469 DROP(1);
470 EMIT(OEOL, 0);
471 p->g->iflags |= USEEOL;
472 p->g->neol++;
473 }
474
475 if(REQUIRE(HERE() != start, MY_REG_EMPTY)) {} /* require nonempty */
476 }
477
478 /*
479 - p_simp_re - parse a simple RE, an atom possibly followed by a repetition
480 == static int p_simp_re(register struct parse *p, int starordinary);
481 */
482 static int /* was the simple RE an unbackslashed $? */
p_simp_re(p,starordinary)483 p_simp_re(p, starordinary)
484 struct parse *p;
485 int starordinary; /* is a leading * an ordinary character? */
486 {
487 int c;
488 int count;
489 int count2;
490 sopno pos;
491 int i;
492 sopno subno;
493 # define BACKSL (1<<CHAR_BIT)
494
495 pos = HERE(); /* repetion op, if any, covers from here */
496
497 assert(MORE()); /* caller should have ensured this */
498 c = GETNEXT();
499 if (c == '\\') {
500 if(REQUIRE(MORE(), MY_REG_EESCAPE)) {}
501 c = BACKSL | (unsigned char)GETNEXT();
502 }
503 switch (c) {
504 case '.':
505 if (p->g->cflags&MY_REG_NEWLINE)
506 nonnewline(p);
507 else
508 EMIT(OANY, 0);
509 break;
510 case '[':
511 p_bracket(p);
512 break;
513 case BACKSL|'{':
514 SETERROR(MY_REG_BADRPT);
515 break;
516 case BACKSL|'(':
517 p->g->nsub++;
518 subno = (sopno) p->g->nsub;
519 if (subno < NPAREN)
520 p->pbegin[subno] = HERE();
521 EMIT(OLPAREN, subno);
522 /* the MORE here is an error heuristic */
523 if (MORE() && !SEETWO('\\', ')'))
524 p_bre(p, '\\', ')');
525 if (subno < NPAREN) {
526 p->pend[subno] = HERE();
527 assert(p->pend[subno] != 0);
528 }
529 EMIT(ORPAREN, subno);
530 if(REQUIRE(EATTWO('\\', ')'), MY_REG_EPAREN)) {}
531 break;
532 case BACKSL|')': /* should not get here -- must be user */
533 case BACKSL|'}':
534 SETERROR(MY_REG_EPAREN);
535 break;
536 case BACKSL|'1':
537 case BACKSL|'2':
538 case BACKSL|'3':
539 case BACKSL|'4':
540 case BACKSL|'5':
541 case BACKSL|'6':
542 case BACKSL|'7':
543 case BACKSL|'8':
544 case BACKSL|'9':
545 i = (c&~BACKSL) - '0';
546 assert(i < NPAREN);
547 if (p->pend[i] != 0) {
548 assert((uint) i <= p->g->nsub);
549 EMIT(OBACK_, i);
550 assert(p->pbegin[i] != 0);
551 assert(OP(p->strip[p->pbegin[i]]) == OLPAREN);
552 assert(OP(p->strip[p->pend[i]]) == ORPAREN);
553 (void) dupl(p, p->pbegin[i]+1, p->pend[i]);
554 if (p->error != 0)
555 break; /* purecov: inspected */
556 EMIT(O_BACK, i);
557 } else
558 SETERROR(MY_REG_ESUBREG);
559 p->g->backrefs = 1;
560 break;
561 case '*':
562 if(REQUIRE(starordinary, MY_REG_BADRPT)) {}
563 /* FALLTHROUGH */
564 default:
565 ordinary(p, c &~ BACKSL);
566 break;
567 }
568
569 if (EAT('*')) { /* implemented as +? */
570 /* this case does not require the (y|) trick, noKLUDGE */
571 INSERT(OPLUS_, pos);
572 ASTERN(O_PLUS, pos);
573 INSERT(OQUEST_, pos);
574 ASTERN(O_QUEST, pos);
575 } else if (EATTWO('\\', '{')) {
576 count = p_count(p);
577 if (EAT(',')) {
578 if (MORE() && my_isdigit(p->charset,PEEK())) {
579 count2 = p_count(p);
580 if(REQUIRE(count <= count2, MY_REG_BADBR)) {}
581 } else /* single number with comma */
582 count2 = RE_INFINITY;
583 } else /* just a single number */
584 count2 = count;
585 repeat(p, pos, count, count2);
586 if (!EATTWO('\\', '}')) { /* error heuristics */
587 while (MORE() && !SEETWO('\\', '}'))
588 NEXT();
589 if(REQUIRE(MORE(), MY_REG_EBRACE)) {}
590 SETERROR(MY_REG_BADBR);
591 }
592 } else if (c == (unsigned char)'$') /* $ (but not \$) ends it */
593 return(1);
594
595 return(0);
596 }
597
598 /*
599 - p_count - parse a repetition count
600 == static int p_count(register struct parse *p);
601 */
602 static int /* the value */
p_count(p)603 p_count(p)
604 struct parse *p;
605 {
606 int count = 0;
607 int ndigits = 0;
608
609 while (MORE() && my_isdigit(p->charset,PEEK()) && count <= DUPMAX) {
610 count = count*10 + (GETNEXT() - '0');
611 ndigits++;
612 }
613
614 if(REQUIRE(ndigits > 0 && count <= DUPMAX, MY_REG_BADBR)) {}
615 return(count);
616 }
617
618 /*
619 - p_bracket - parse a bracketed character list
620 == static void p_bracket(register struct parse *p);
621 *
622 * Note a significant property of this code: if the allocset() did SETERROR,
623 * no set operations are done.
624 */
625 static void
p_bracket(p)626 p_bracket(p)
627 struct parse *p;
628 {
629 cset *cs = allocset(p);
630 int invert = 0;
631
632 /* Dept of Truly Sickening Special-Case Kludges */
633 if (p->next + 5 < p->end && strncmp(p->next, "[:<:]]", 6) == 0) {
634 EMIT(OBOW, 0);
635 NEXTn(6);
636 return;
637 }
638 if (p->next + 5 < p->end && strncmp(p->next, "[:>:]]", 6) == 0) {
639 EMIT(OEOW, 0);
640 NEXTn(6);
641 return;
642 }
643
644 if (EAT('^'))
645 invert++; /* make note to invert set at end */
646 if (EAT(']'))
647 CHadd(cs, ']');
648 else if (EAT('-'))
649 CHadd(cs, '-');
650 while (MORE() && PEEK() != ']' && !SEETWO('-', ']'))
651 p_b_term(p, cs);
652 if (EAT('-'))
653 CHadd(cs, '-');
654 if(MUSTEAT(']', MY_REG_EBRACK)) {}
655
656 if (p->error != 0) /* don't mess things up further */
657 return;
658
659 if (p->g->cflags&MY_REG_ICASE) {
660 int i;
661 int ci;
662
663 for (i = p->g->csetsize - 1; i >= 0; i--)
664 if (CHIN(cs, i) && my_isalpha(p->charset,i)) {
665 ci = othercase(p->charset,i);
666 if (ci != i)
667 CHadd(cs, ci);
668 }
669 if (cs->multis != NULL)
670 mccase(p, cs);
671 }
672 if (invert) {
673 int i;
674
675 for (i = p->g->csetsize - 1; i >= 0; i--)
676 if (CHIN(cs, i))
677 CHsub(cs, i);
678 else
679 CHadd(cs, i);
680 if (p->g->cflags&MY_REG_NEWLINE)
681 CHsub(cs, '\n');
682 if (cs->multis != NULL)
683 mcinvert(p, cs);
684 }
685
686 assert(cs->multis == NULL); /* xxx */
687
688 if (nch(p, cs) == 1) { /* optimize singleton sets */
689 ordinary(p, firstch(p, cs));
690 freeset(p, cs);
691 } else
692 EMIT(OANYOF, freezeset(p, cs));
693 }
694
695 /*
696 - p_b_term - parse one term of a bracketed character list
697 == static void p_b_term(register struct parse *p, register cset *cs);
698 */
699 static void
p_b_term(p,cs)700 p_b_term(p, cs)
701 struct parse *p;
702 cset *cs;
703 {
704 char c;
705 char start, finish;
706 int i;
707
708 /* classify what we've got */
709 switch ((MORE()) ? PEEK() : '\0') {
710 case '[':
711 c = (MORE2()) ? PEEK2() : '\0';
712 break;
713 case '-':
714 SETERROR(MY_REG_ERANGE);
715 return; /* NOTE RETURN */
716 default:
717 c = '\0';
718 break;
719 }
720
721 switch (c) {
722 case ':': /* character class */
723 NEXT2();
724 if(REQUIRE(MORE(), MY_REG_EBRACK)) {}
725 c = PEEK();
726 if(REQUIRE(c != '-' && c != ']', MY_REG_ECTYPE)) {}
727 p_b_cclass(p, cs);
728 if(REQUIRE(MORE(), MY_REG_EBRACK)) {}
729 if(REQUIRE(EATTWO(':', ']'), MY_REG_ECTYPE)) {}
730 break;
731 case '=': /* equivalence class */
732 NEXT2();
733 if(REQUIRE(MORE(), MY_REG_EBRACK)) {}
734 c = PEEK();
735 if(REQUIRE(c != '-' && c != ']', MY_REG_ECOLLATE)) {}
736 p_b_eclass(p, cs);
737 if(REQUIRE(MORE(), MY_REG_EBRACK)) {}
738 if(REQUIRE(EATTWO('=', ']'), MY_REG_ECOLLATE)) {}
739 break;
740 default: /* symbol, ordinary character, or range */
741 /* xxx revision needed for multichar stuff */
742 start = p_b_symbol(p);
743 if (SEE('-') && MORE2() && PEEK2() != ']') {
744 /* range */
745 NEXT();
746 if (EAT('-'))
747 finish = '-';
748 else
749 finish = p_b_symbol(p);
750 } else
751 finish = start;
752 /* xxx what about signed chars here... */
753 if(REQUIRE(start <= finish, MY_REG_ERANGE)) {}
754 for (i = start; i <= finish; i++)
755 CHadd(cs, i);
756 break;
757 }
758 }
759
760 /*
761 - p_b_cclass - parse a character-class name and deal with it
762 == static void p_b_cclass(register struct parse *p, register cset *cs);
763 */
764 static void
p_b_cclass(p,cs)765 p_b_cclass(p, cs)
766 struct parse *p;
767 cset *cs;
768 {
769 char *sp = p->next;
770 struct cclass *cp;
771 size_t len;
772
773 while (MORE() && my_isalpha(p->charset,PEEK()))
774 NEXT();
775 len = p->next - sp;
776 for (cp = cclasses; cp->name != NULL; cp++)
777 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
778 break;
779 if (cp->name == NULL) {
780 /* oops, didn't find it */
781 SETERROR(MY_REG_ECTYPE);
782 return;
783 }
784
785 #ifndef USE_ORIG_REGEX_CODE
786 {
787 size_t i;
788 for (i=1 ; i<256 ; i++)
789 if (p->charset->ctype[i+1] & cp->mask)
790 CHadd(cs, i);
791 }
792 #else
793 {
794 char *u = (char*) cp->chars;
795 char c;
796
797 while ((c = *u++) != '\0')
798 CHadd(cs, c);
799
800 for (u = (char*) cp->multis; *u != '\0'; u += strlen(u) + 1)
801 MCadd(p, cs, u);
802 }
803 #endif
804
805 }
806
807 /*
808 - p_b_eclass - parse an equivalence-class name and deal with it
809 == static void p_b_eclass(register struct parse *p, register cset *cs);
810 *
811 * This implementation is incomplete. xxx
812 */
813 static void
p_b_eclass(p,cs)814 p_b_eclass(p, cs)
815 struct parse *p;
816 cset *cs;
817 {
818 char c;
819
820 c = p_b_coll_elem(p, '=');
821 CHadd(cs, c);
822 }
823
824 /*
825 - p_b_symbol - parse a character or [..]ed multicharacter collating symbol
826 == static char p_b_symbol(register struct parse *p);
827 */
828 static char /* value of symbol */
p_b_symbol(p)829 p_b_symbol(p)
830 struct parse *p;
831 {
832 char value;
833
834 if(REQUIRE(MORE(), MY_REG_EBRACK)) {}
835 if (!EATTWO('[', '.'))
836 return(GETNEXT());
837
838 /* collating symbol */
839 value = p_b_coll_elem(p, '.');
840 if(REQUIRE(EATTWO('.', ']'), MY_REG_ECOLLATE)) {}
841 return(value);
842 }
843
844 /*
845 - p_b_coll_elem - parse a collating-element name and look it up
846 == static char p_b_coll_elem(register struct parse *p, int endc);
847 */
848 static char /* value of collating element */
p_b_coll_elem(p,endc)849 p_b_coll_elem(p, endc)
850 struct parse *p;
851 int endc; /* name ended by endc,']' */
852 {
853 char *sp = p->next;
854 struct cname *cp;
855 #ifdef _WIN64
856 __int64 len;
857 #else
858 int len;
859 #endif
860 while (MORE() && !SEETWO(endc, ']'))
861 NEXT();
862 if (!MORE()) {
863 SETERROR(MY_REG_EBRACK);
864 return(0);
865 }
866 len = p->next - sp;
867 for (cp = cnames; cp->name != NULL; cp++)
868 if (strncmp(cp->name, sp, len) == 0 && cp->name[len] == '\0')
869 return(cp->code); /* known name */
870 if (len == 1)
871 return(*sp); /* single character */
872 SETERROR(MY_REG_ECOLLATE); /* neither */
873 return(0);
874 }
875
876 /*
877 - othercase - return the case counterpart of an alphabetic
878 == static char othercase(int ch);
879 */
880 static char /* if no counterpart, return ch */
othercase(charset,ch)881 othercase(charset,ch)
882 const CHARSET_INFO *charset;
883 int ch;
884 {
885 /*
886 In MySQL some multi-byte character sets
887 have 'ctype' array but don't have 'to_lower'
888 and 'to_upper' arrays. In this case we handle
889 only basic latin letters a..z and A..Z.
890
891 If 'to_lower' and 'to_upper' arrays are empty in a character set,
892 then my_isalpha(cs, ch) should never return TRUE for characters
893 other than basic latin letters. Otherwise it should be
894 considered as a mistake in character set definition.
895 */
896 assert(my_isalpha(charset,ch));
897 if (my_isupper(charset,ch))
898 {
899 return(charset->to_lower ? my_tolower(charset,ch) :
900 ch - 'A' + 'a');
901 }
902 else if (my_islower(charset,ch))
903 {
904 return(charset->to_upper ? my_toupper(charset,ch) :
905 ch - 'a' + 'A');
906 }
907 else /* peculiar, but could happen */
908 return(ch);
909 }
910
911 /*
912 - bothcases - emit a dualcase version of a two-case character
913 == static void bothcases(register struct parse *p, int ch);
914 *
915 * Boy, is this implementation ever a kludge...
916 */
917 static void
bothcases(p,ch)918 bothcases(p, ch)
919 struct parse *p;
920 int ch;
921 {
922 char *oldnext = p->next;
923 char *oldend = p->end;
924 char bracket[3];
925
926 assert(othercase(p->charset, ch) != ch); /* p_bracket() would recurse */
927 p->next = bracket;
928 p->end = bracket+2;
929 bracket[0] = ch;
930 bracket[1] = ']';
931 bracket[2] = '\0';
932 p_bracket(p);
933 assert(p->next == bracket+2);
934 p->next = oldnext;
935 p->end = oldend;
936 }
937
938 /*
939 - ordinary - emit an ordinary character
940 == static void ordinary(register struct parse *p, register int ch);
941 */
942 static void
ordinary(p,ch)943 ordinary(p, ch)
944 struct parse *p;
945 int ch;
946 {
947 cat_t *cap = p->g->categories;
948
949 if ((p->g->cflags&MY_REG_ICASE) && my_isalpha(p->charset,ch) &&
950 othercase(p->charset,ch) != ch)
951 bothcases(p, ch);
952 else {
953 EMIT(OCHAR, (unsigned char)ch);
954 if (cap[ch] == 0)
955 cap[ch] = p->g->ncategories++;
956 }
957 }
958
959 /*
960 - nonnewline - emit MY_REG_NEWLINE version of OANY
961 == static void nonnewline(register struct parse *p);
962 *
963 * Boy, is this implementation ever a kludge...
964 */
965 static void
nonnewline(p)966 nonnewline(p)
967 struct parse *p;
968 {
969 char *oldnext = p->next;
970 char *oldend = p->end;
971 char bracket[4];
972
973 p->next = bracket;
974 p->end = bracket+3;
975 bracket[0] = '^';
976 bracket[1] = '\n';
977 bracket[2] = ']';
978 bracket[3] = '\0';
979 p_bracket(p);
980 assert(p->next == bracket+3);
981 p->next = oldnext;
982 p->end = oldend;
983 }
984
985 /*
986 - repeat - generate code for a bounded repetition, recursively if needed
987 == static void repeat(register struct parse *p, sopno start, int from, int to);
988 */
989 static void
repeat(p,start,from,to)990 repeat(p, start, from, to)
991 struct parse *p;
992 sopno start; /* operand from here to end of strip */
993 int from; /* repeated from this number */
994 int to; /* to this number of times (maybe RE_INFINITY) */
995 {
996 sopno finish = HERE();
997 # define N 2
998 # define INF 3
999 # define REP(f, t) ((f)*8 + (t))
1000 # define MAP(n) (((n) <= 1) ? (n) : ((n) == RE_INFINITY) ? INF : N)
1001 sopno copy;
1002
1003 if (p->error != 0) /* head off possible runaway recursion */
1004 return;
1005
1006 assert(from <= to);
1007
1008 switch (REP(MAP(from), MAP(to))) {
1009 case REP(0, 0): /* must be user doing this */
1010 DROP(finish-start); /* drop the operand */
1011 break;
1012 case REP(0, 1): /* as x{1,1}? */
1013 case REP(0, N): /* as x{1,n}? */
1014 case REP(0, INF): /* as x{1,}? */
1015 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1016 INSERT(OCH_, start); /* offset is wrong... */
1017 repeat(p, start+1, 1, to);
1018 ASTERN(OOR1, start);
1019 AHEAD(start); /* ... fix it */
1020 EMIT(OOR2, 0);
1021 AHEAD(THERE());
1022 ASTERN(O_CH, THERETHERE());
1023 break;
1024 case REP(1, 1): /* trivial case */
1025 /* done */
1026 break;
1027 case REP(1, N): /* as x?x{1,n-1} */
1028 /* KLUDGE: emit y? as (y|) until subtle bug gets fixed */
1029 INSERT(OCH_, start);
1030 ASTERN(OOR1, start);
1031 AHEAD(start);
1032 EMIT(OOR2, 0); /* offset very wrong... */
1033 AHEAD(THERE()); /* ...so fix it */
1034 ASTERN(O_CH, THERETHERE());
1035 copy = dupl(p, start+1, finish+1);
1036 if (p->error != 0)
1037 return; /* purecov: inspected */
1038 assert(copy == finish+4);
1039 repeat(p, copy, 1, to-1);
1040 break;
1041 case REP(1, INF): /* as x+ */
1042 INSERT(OPLUS_, start);
1043 ASTERN(O_PLUS, start);
1044 break;
1045 case REP(N, N): /* as xx{m-1,n-1} */
1046 copy = dupl(p, start, finish);
1047 if (p->error != 0)
1048 return;
1049 repeat(p, copy, from-1, to-1);
1050 break;
1051 case REP(N, INF): /* as xx{n-1,INF} */
1052 copy = dupl(p, start, finish);
1053 if (p->error != 0)
1054 return; /* purecov: inspected */
1055 repeat(p, copy, from-1, to);
1056 break;
1057 default: /* "can't happen" */
1058 SETERROR(MY_REG_ASSERT); /* just in case */
1059 break;
1060 }
1061 }
1062
1063 /*
1064 - seterr - set an error condition
1065 == static int seterr(register struct parse *p, int e);
1066 */
1067 static int /* useless but makes type checking happy */
seterr(p,e)1068 seterr(p, e)
1069 struct parse *p;
1070 int e;
1071 {
1072 if (p->error == 0) /* keep earliest error condition */
1073 p->error = e;
1074 p->next = nuls; /* try to bring things to a halt */
1075 p->end = nuls;
1076 return(0); /* make the return value well-defined */
1077 }
1078
1079 /*
1080 - allocset - allocate a set of characters for []
1081 == static cset *allocset(register struct parse *p);
1082 */
1083 static cset *
allocset(p)1084 allocset(p)
1085 struct parse *p;
1086 {
1087 int no = p->g->ncsets++;
1088 size_t nc;
1089 size_t nbytes;
1090 cset *cs;
1091 size_t css = (size_t)p->g->csetsize;
1092 int i;
1093
1094 if (no >= p->ncsalloc) { /* need another column of space */
1095 p->ncsalloc += CHAR_BIT;
1096 nc = p->ncsalloc;
1097 assert(nc % CHAR_BIT == 0);
1098 nbytes = nc / CHAR_BIT * css;
1099 if (p->g->sets == NULL)
1100 p->g->sets = (cset *)malloc(nc * sizeof(cset));
1101 else
1102 p->g->sets = (cset *)realloc((char *)p->g->sets,
1103 nc * sizeof(cset));
1104 if (p->g->setbits == NULL)
1105 p->g->setbits = (uch *)malloc(nbytes);
1106 else {
1107 p->g->setbits = (uch *)realloc((char *)p->g->setbits,
1108 nbytes);
1109 /* xxx this isn't right if setbits is now NULL */
1110 for (i = 0; i < no; i++)
1111 p->g->sets[i].ptr = p->g->setbits + css*(i/CHAR_BIT);
1112 }
1113 if (p->g->sets != NULL && p->g->setbits != NULL)
1114 (void) memset((char *)p->g->setbits + (nbytes - css),
1115 0, css);
1116 else {
1117 no = 0;
1118 SETERROR(MY_REG_ESPACE);
1119 /* caller's responsibility not to do set ops */
1120 }
1121 }
1122
1123 assert(p->g->sets != NULL); /* xxx */
1124 cs = &p->g->sets[no];
1125 cs->ptr = p->g->setbits + css*((no)/CHAR_BIT);
1126 cs->mask = 1 << ((no) % CHAR_BIT);
1127 cs->hash = 0;
1128 cs->smultis = 0;
1129 cs->multis = NULL;
1130
1131 return(cs);
1132 }
1133
1134 /*
1135 - freeset - free a now-unused set
1136 == static void freeset(register struct parse *p, register cset *cs);
1137 */
1138 static void
freeset(p,cs)1139 freeset(p, cs)
1140 struct parse *p;
1141 cset *cs;
1142 {
1143 size_t i;
1144 cset *top = &p->g->sets[p->g->ncsets];
1145 size_t css = (size_t)p->g->csetsize;
1146
1147 for (i = 0; i < css; i++)
1148 CHsub(cs, i);
1149 if (cs == top-1) /* recover only the easy case */
1150 p->g->ncsets--;
1151 }
1152
1153 /*
1154 - freezeset - final processing on a set of characters
1155 == static int freezeset(register struct parse *p, register cset *cs);
1156 *
1157 * The main task here is merging identical sets. This is usually a waste
1158 * of time (although the hash code minimizes the overhead), but can win
1159 * big if MY_REG_ICASE is being used. MY_REG_ICASE, by the way, is why the hash
1160 * is done using addition rather than xor -- all ASCII [aA] sets xor to
1161 * the same value!
1162 */
1163 static int /* set number */
freezeset(p,cs)1164 freezeset(p, cs)
1165 struct parse *p;
1166 cset *cs;
1167 {
1168 uch h = cs->hash;
1169 size_t i;
1170 cset *top = &p->g->sets[p->g->ncsets];
1171 cset *cs2;
1172 size_t css = (size_t)p->g->csetsize;
1173
1174 /* look for an earlier one which is the same */
1175 for (cs2 = &p->g->sets[0]; cs2 < top; cs2++)
1176 if (cs2->hash == h && cs2 != cs) {
1177 /* maybe */
1178 for (i = 0; i < css; i++)
1179 if (!!CHIN(cs2, i) != !!CHIN(cs, i))
1180 break; /* no */
1181 if (i == css)
1182 break; /* yes */
1183 }
1184
1185 if (cs2 < top) { /* found one */
1186 freeset(p, cs);
1187 cs = cs2;
1188 }
1189
1190 return((int)(cs - p->g->sets));
1191 }
1192
1193 /*
1194 - firstch - return first character in a set (which must have at least one)
1195 == static int firstch(register struct parse *p, register cset *cs);
1196 */
1197 static int /* character; there is no "none" value */
firstch(p,cs)1198 firstch(p, cs)
1199 struct parse *p;
1200 cset *cs;
1201 {
1202 size_t i;
1203 size_t css = (size_t)p->g->csetsize;
1204
1205 for (i = 0; i < css; i++)
1206 if (CHIN(cs, i))
1207 return((char)i);
1208 assert(never);
1209 return(0); /* arbitrary */
1210 }
1211
1212 /*
1213 - nch - number of characters in a set
1214 == static int nch(register struct parse *p, register cset *cs);
1215 */
1216 static int
nch(p,cs)1217 nch(p, cs)
1218 struct parse *p;
1219 cset *cs;
1220 {
1221 size_t i;
1222 size_t css = (size_t)p->g->csetsize;
1223 int n = 0;
1224
1225 for (i = 0; i < css; i++)
1226 if (CHIN(cs, i))
1227 n++;
1228 return(n);
1229 }
1230
1231 #ifdef USE_ORIG_REGEX_CODE
1232 /*
1233 - mcadd - add a collating element to a cset
1234 == static void mcadd(register struct parse *p, register cset *cs, \
1235 == register char *cp);
1236 */
1237 static void
mcadd(p,cs,cp)1238 mcadd(p, cs, cp)
1239 struct parse *p;
1240 cset *cs;
1241 char *cp;
1242 {
1243 size_t oldend = cs->smultis;
1244
1245 cs->smultis += strlen(cp) + 1;
1246 if (cs->multis == NULL)
1247 cs->multis = malloc(cs->smultis);
1248 else
1249 cs->multis = realloc(cs->multis, cs->smultis);
1250 if (cs->multis == NULL) {
1251 SETERROR(MY_REG_ESPACE);
1252 return;
1253 }
1254
1255 (void) strcpy(cs->multis + oldend - 1, cp);
1256 cs->multis[cs->smultis - 1] = '\0';
1257 }
1258 #endif
1259
1260 /*
1261 - mcinvert - invert the list of collating elements in a cset
1262 == static void mcinvert(register struct parse *p, register cset *cs);
1263 *
1264 * This would have to know the set of possibilities. Implementation
1265 * is deferred.
1266 */
1267 static void
1268 mcinvert(p, cs)
1269 struct parse *p MY_ATTRIBUTE((unused));
1270 cset *cs MY_ATTRIBUTE((unused));
1271 {
1272 assert(cs->multis == NULL); /* xxx */
1273 }
1274
1275 /*
1276 - mccase - add case counterparts of the list of collating elements in a cset
1277 == static void mccase(register struct parse *p, register cset *cs);
1278 *
1279 * This would have to know the set of possibilities. Implementation
1280 * is deferred.
1281 */
1282 static void
1283 mccase(p, cs)
1284 struct parse *p MY_ATTRIBUTE((unused));
1285 cset *cs MY_ATTRIBUTE((unused));
1286 {
1287 assert(cs->multis == NULL); /* xxx */
1288 }
1289
1290 /*
1291 - isinsets - is this character in any sets?
1292 == static int isinsets(register struct re_guts *g, int c);
1293 */
1294 static int /* predicate */
isinsets(g,c)1295 isinsets(g, c)
1296 struct re_guts *g;
1297 int c;
1298 {
1299 uch *col;
1300 int i;
1301 int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1302 unsigned uc = (unsigned char)c;
1303
1304 for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1305 if (col[uc] != 0)
1306 return(1);
1307 return(0);
1308 }
1309
1310 /*
1311 - samesets - are these two characters in exactly the same sets?
1312 == static int samesets(register struct re_guts *g, int c1, int c2);
1313 */
1314 static int /* predicate */
samesets(g,c1,c2)1315 samesets(g, c1, c2)
1316 struct re_guts *g;
1317 int c1;
1318 int c2;
1319 {
1320 uch *col;
1321 int i;
1322 int ncols = (g->ncsets+(CHAR_BIT-1)) / CHAR_BIT;
1323 unsigned uc1 = (unsigned char)c1;
1324 unsigned uc2 = (unsigned char)c2;
1325
1326 for (i = 0, col = g->setbits; i < ncols; i++, col += g->csetsize)
1327 if (col[uc1] != col[uc2])
1328 return(0);
1329 return(1);
1330 }
1331
1332 /*
1333 - categorize - sort out character categories
1334 == static void categorize(struct parse *p, register struct re_guts *g);
1335 */
1336 static void
categorize(p,g)1337 categorize(p, g)
1338 struct parse *p;
1339 struct re_guts *g;
1340 {
1341 cat_t *cats = g->categories;
1342 int c;
1343 int c2;
1344 cat_t cat;
1345
1346 /* avoid making error situations worse */
1347 if (p->error != 0)
1348 return;
1349
1350 for (c = CHAR_MIN; c <= CHAR_MAX; c++)
1351 if (cats[c] == 0 && isinsets(g, c)) {
1352 cat = g->ncategories++;
1353 cats[c] = cat;
1354 for (c2 = c+1; c2 <= CHAR_MAX; c2++)
1355 if (cats[c2] == 0 && samesets(g, c, c2))
1356 cats[c2] = cat;
1357 }
1358 }
1359
1360 /*
1361 - dupl - emit a duplicate of a bunch of sops
1362 == static sopno dupl(register struct parse *p, sopno start, sopno finish);
1363 */
1364 static sopno /* start of duplicate */
dupl(p,start,finish)1365 dupl(p, start, finish)
1366 struct parse *p;
1367 sopno start; /* from here */
1368 sopno finish; /* to this less one */
1369 {
1370 sopno ret = HERE();
1371 sopno len = finish - start;
1372
1373 assert(finish >= start);
1374 if (len == 0)
1375 return(ret);
1376 enlarge(p, p->ssize + len); /* this many unexpected additions */
1377 if (p->error != 0)
1378 return(p->error);
1379
1380 assert(p->ssize >= p->slen + len);
1381 (void) memcpy((char *)(p->strip + p->slen),
1382 (char *)(p->strip + start), (size_t)len*sizeof(sop));
1383 p->slen += len;
1384 return(ret);
1385 }
1386
1387 /*
1388 - doemit - emit a strip operator
1389 == static void doemit(register struct parse *p, sop op, size_t opnd);
1390 *
1391 * It might seem better to implement this as a macro with a function as
1392 * hard-case backup, but it's just too big and messy unless there are
1393 * some changes to the data structures. Maybe later.
1394 */
1395 static void
doemit(p,op,opnd)1396 doemit(p, op, opnd)
1397 struct parse *p;
1398 sop op;
1399 size_t opnd;
1400 {
1401 /* avoid making error situations worse */
1402 if (p->error != 0)
1403 return;
1404
1405 /* deal with oversize operands ("can't happen", more or less) */
1406 assert(opnd < 1<<OPSHIFT);
1407
1408 /* deal with undersized strip */
1409 if (p->slen >= p->ssize)
1410 enlarge(p, (p->ssize+1) / 2 * 3); /* +50% */
1411 assert(p->slen < p->ssize);
1412
1413 /* finally, it's all reduced to the easy case */
1414 p->strip[p->slen++] = SOP(op, opnd);
1415 }
1416
1417 /*
1418 - doinsert - insert a sop into the strip
1419 == static void doinsert(register struct parse *p, sop op, size_t opnd, sopno pos);
1420 */
1421 static void
doinsert(p,op,opnd,pos)1422 doinsert(p, op, opnd, pos)
1423 struct parse *p;
1424 sop op;
1425 size_t opnd;
1426 sopno pos;
1427 {
1428 sopno sn;
1429 sop s;
1430 int i;
1431
1432 /* avoid making error situations worse */
1433 if (p->error != 0)
1434 return;
1435
1436 sn = HERE();
1437 EMIT(op, opnd); /* do checks, ensure space */
1438 assert(HERE() == sn+1);
1439 s = p->strip[sn];
1440
1441 /* adjust paren pointers */
1442 assert(pos > 0);
1443 for (i = 1; i < NPAREN; i++) {
1444 if (p->pbegin[i] >= pos) {
1445 p->pbegin[i]++;
1446 }
1447 if (p->pend[i] >= pos) {
1448 p->pend[i]++;
1449 }
1450 }
1451 {
1452 size_t length=(HERE()-pos-1)*sizeof(sop);
1453 memmove((uchar *) &p->strip[pos+1],
1454 (uchar *) &p->strip[pos],
1455 length);
1456 }
1457 p->strip[pos] = s;
1458 }
1459
1460 /*
1461 - dofwd - complete a forward reference
1462 == static void dofwd(register struct parse *p, sopno pos, sop value);
1463 */
1464 static void
dofwd(p,pos,value)1465 dofwd(p, pos, value)
1466 struct parse *p;
1467 sopno pos;
1468 sop value;
1469 {
1470 /* avoid making error situations worse */
1471 if (p->error != 0)
1472 return;
1473
1474 assert(value < 1<<OPSHIFT);
1475 p->strip[pos] = OP(p->strip[pos]) | value;
1476 }
1477
1478 /*
1479 - enlarge - enlarge the strip
1480 == static void enlarge(register struct parse *p, sopno size);
1481 */
1482 static void
enlarge(p,size)1483 enlarge(p, size)
1484 struct parse *p;
1485 sopno size;
1486 {
1487 sop *sp;
1488
1489 if (p->ssize >= size)
1490 return;
1491
1492 DBUG_EXECUTE_IF("bug24449090_simulate_oom",
1493 {
1494 free(p->strip);
1495 p->strip= NULL;
1496 p->ssize= 0;
1497 SETERROR(MY_REG_ESPACE);
1498 return;
1499 });
1500
1501 sp = (sop *)realloc(p->strip, size*sizeof(sop));
1502 if (sp == NULL) {
1503 SETERROR(MY_REG_ESPACE);
1504 return;
1505 }
1506 p->strip = sp;
1507 p->ssize = size;
1508 }
1509
1510 /*
1511 - stripsnug - compact the strip
1512 == static void stripsnug(register struct parse *p, register struct re_guts *g);
1513 */
1514 static void
stripsnug(p,g)1515 stripsnug(p, g)
1516 struct parse *p;
1517 struct re_guts *g;
1518 {
1519 g->nstates = p->slen;
1520 g->strip = (sop *)realloc((char *)p->strip, p->slen * sizeof(sop));
1521 if (g->strip == NULL) {
1522 SETERROR(MY_REG_ESPACE);
1523 g->strip = p->strip;
1524 }
1525 }
1526
1527 /*
1528 - findmust - fill in must and mlen with longest mandatory literal string
1529 == static void findmust(register struct parse *p, register struct re_guts *g);
1530 *
1531 * This algorithm could do fancy things like analyzing the operands of |
1532 * for common subsequences. Someday. This code is simple and finds most
1533 * of the interesting cases.
1534 *
1535 * Note that must and mlen got initialized during setup.
1536 */
1537 static void
findmust(p,g)1538 findmust(p, g)
1539 struct parse *p;
1540 struct re_guts *g;
1541 {
1542 sop *scan;
1543 sop *start= NULL;
1544 sop *newstart= NULL;
1545 sopno newlen;
1546 sop s;
1547 char *cp;
1548 sopno i;
1549
1550 /* avoid making error situations worse */
1551 if (p->error != 0)
1552 return;
1553
1554 /* find the longest OCHAR sequence in strip */
1555 newlen = 0;
1556 scan = g->strip + 1;
1557 do {
1558 s = *scan++;
1559 switch (OP(s)) {
1560 case OCHAR: /* sequence member */
1561 if (newlen == 0) /* new sequence */
1562 newstart = scan - 1;
1563 newlen++;
1564 break;
1565 case OPLUS_: /* things that don't break one */
1566 case OLPAREN:
1567 case ORPAREN:
1568 break;
1569 case OQUEST_: /* things that must be skipped */
1570 case OCH_:
1571 scan--;
1572 do {
1573 scan += OPND(s);
1574 s = *scan;
1575 /* assert() interferes w debug printouts */
1576 if (OP(s) != O_QUEST && OP(s) != O_CH &&
1577 OP(s) != OOR2) {
1578 g->iflags |= BAD;
1579 return;
1580 }
1581 } while (OP(s) != O_QUEST && OP(s) != O_CH);
1582 /* fallthrough */
1583 default: /* things that break a sequence */
1584 if (newlen > g->mlen) { /* ends one */
1585 start = newstart;
1586 g->mlen = newlen;
1587 }
1588 newlen = 0;
1589 break;
1590 }
1591 } while (OP(s) != OEND);
1592
1593 if (g->mlen == 0) /* there isn't one */
1594 return;
1595
1596 /* turn it into a character string */
1597 g->must = malloc((size_t)g->mlen + 1);
1598 if (g->must == NULL) { /* argh; just forget it */
1599 g->mlen = 0;
1600 return;
1601 }
1602 cp = g->must;
1603 scan = start;
1604 for (i = g->mlen; i > 0; i--) {
1605 while (OP(s = *scan++) != OCHAR)
1606 continue;
1607 assert(cp < g->must + g->mlen);
1608 *cp++ = (char)OPND(s);
1609 }
1610 assert(cp == g->must + g->mlen);
1611 *cp++ = '\0'; /* just on general principles */
1612 }
1613
1614 /*
1615 - pluscount - count + nesting
1616 == static sopno pluscount(register struct parse *p, register struct re_guts *g);
1617 */
1618 static sopno /* nesting depth */
pluscount(p,g)1619 pluscount(p, g)
1620 struct parse *p;
1621 struct re_guts *g;
1622 {
1623 sop *scan;
1624 sop s;
1625 sopno plusnest = 0;
1626 sopno maxnest = 0;
1627
1628 if (p->error != 0)
1629 return(0); /* there may not be an OEND */
1630
1631 scan = g->strip + 1;
1632 do {
1633 s = *scan++;
1634 switch (OP(s)) {
1635 case OPLUS_:
1636 plusnest++;
1637 break;
1638 case O_PLUS:
1639 if (plusnest > maxnest)
1640 maxnest = plusnest;
1641 plusnest--;
1642 break;
1643 }
1644 } while (OP(s) != OEND);
1645 if (plusnest != 0)
1646 g->iflags |= BAD;
1647 return(maxnest);
1648 }
1649