1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2012 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND 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 COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40
41 /* This module contains the external function pcre_study(), along with local
42 supporting functions. */
43
44 /* %ExternalCopyright% */
45
46 #ifdef HAVE_CONFIG_H
47 #include "config.h"
48 #endif
49
50 #include "pcre_internal.h"
51
52 #define SET_BIT(c) start_bits[c/8] |= (1 << (c&7))
53
54 /* Returns from set_start_bits() */
55
56 enum { SSB_FAIL, SSB_DONE, SSB_CONTINUE, SSB_UNKNOWN };
57
58
59
60 /*************************************************
61 * Find the minimum subject length for a group *
62 *************************************************/
63
64 /* Scan a parenthesized group and compute the minimum length of subject that
65 is needed to match it. This is a lower bound; it does not mean there is a
66 string of that length that matches. In UTF8 mode, the result is in characters
67 rather than bytes.
68
69 Arguments:
70 re compiled pattern block
71 code pointer to start of group (the bracket)
72 startcode pointer to start of the whole pattern's code
73 options the compiling options
74 recurses chain of recurse_check to catch mutual recursion
75 countptr pointer to call count (to catch over complexity)
76
77 Returns: the minimum length
78 -1 if \C in UTF-8 mode or (*ACCEPT) was encountered
79 -2 internal error (missing capturing bracket)
80 -3 internal error (opcode not listed)
81 */
82
83 static int
find_minlength(const REAL_PCRE * re,const pcre_uchar * code,const pcre_uchar * startcode,int options,recurse_check * recurses,int * countptr)84 find_minlength(const REAL_PCRE *re, const pcre_uchar *code,
85 const pcre_uchar *startcode, int options, recurse_check *recurses,
86 int *countptr)
87 {
88 int length = -1;
89 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
90 BOOL utf = (options & PCRE_UTF8) != 0;
91 BOOL had_recurse = FALSE;
92 recurse_check this_recurse;
93 register int branchlength = 0;
94 register pcre_uchar *cc = (pcre_uchar *)code + 1 + LINK_SIZE;
95
96 if ((*countptr)++ > 1000) return -1; /* too complex */
97
98 if (*code == OP_CBRA || *code == OP_SCBRA ||
99 *code == OP_CBRAPOS || *code == OP_SCBRAPOS) cc += IMM2_SIZE;
100
101 /* Scan along the opcodes for this branch. If we get to the end of the
102 branch, check the length against that of the other branches. */
103
104 for (;;)
105 {
106 int d, min;
107 pcre_uchar *cs, *ce;
108 register pcre_uchar op = *cc;
109
110 switch (op)
111 {
112 case OP_COND:
113 case OP_SCOND:
114
115 /* If there is only one branch in a condition, the implied branch has zero
116 length, so we don't add anything. This covers the DEFINE "condition"
117 automatically. */
118
119 cs = cc + GET(cc, 1);
120 if (*cs != OP_ALT)
121 {
122 cc = cs + 1 + LINK_SIZE;
123 break;
124 }
125
126 /* Otherwise we can fall through and treat it the same as any other
127 subpattern. */
128
129 case OP_CBRA:
130 case OP_SCBRA:
131 case OP_BRA:
132 case OP_SBRA:
133 case OP_CBRAPOS:
134 case OP_SCBRAPOS:
135 case OP_BRAPOS:
136 case OP_SBRAPOS:
137 case OP_ONCE:
138 case OP_ONCE_NC:
139 d = find_minlength(re, cc, startcode, options, recurses, countptr);
140 if (d < 0) return d;
141 branchlength += d;
142 do cc += GET(cc, 1); while (*cc == OP_ALT);
143 cc += 1 + LINK_SIZE;
144 break;
145
146 /* ACCEPT makes things far too complicated; we have to give up. */
147
148 case OP_ACCEPT:
149 case OP_ASSERT_ACCEPT:
150 return -1;
151
152 /* Reached end of a branch; if it's a ket it is the end of a nested
153 call. If it's ALT it is an alternation in a nested call. If it is END it's
154 the end of the outer call. All can be handled by the same code. If an
155 ACCEPT was previously encountered, use the length that was in force at that
156 time, and pass back the shortest ACCEPT length. */
157
158 case OP_ALT:
159 case OP_KET:
160 case OP_KETRMAX:
161 case OP_KETRMIN:
162 case OP_KETRPOS:
163 case OP_END:
164 if (length < 0 || (!had_recurse && branchlength < length))
165 length = branchlength;
166 if (op != OP_ALT) return length;
167 cc += 1 + LINK_SIZE;
168 branchlength = 0;
169 had_recurse = FALSE;
170 break;
171
172 /* Skip over assertive subpatterns */
173
174 case OP_ASSERT:
175 case OP_ASSERT_NOT:
176 case OP_ASSERTBACK:
177 case OP_ASSERTBACK_NOT:
178 do cc += GET(cc, 1); while (*cc == OP_ALT);
179 /* Fall through */
180
181 /* Skip over things that don't match chars */
182
183 case OP_REVERSE:
184 case OP_CREF:
185 case OP_DNCREF:
186 case OP_RREF:
187 case OP_DNRREF:
188 case OP_DEF:
189 case OP_CALLOUT:
190 case OP_SOD:
191 case OP_SOM:
192 case OP_EOD:
193 case OP_EODN:
194 case OP_CIRC:
195 case OP_CIRCM:
196 case OP_DOLL:
197 case OP_DOLLM:
198 case OP_NOT_WORD_BOUNDARY:
199 case OP_WORD_BOUNDARY:
200 cc += PRIV(OP_lengths)[*cc];
201 break;
202
203 /* Skip over a subpattern that has a {0} or {0,x} quantifier */
204
205 case OP_BRAZERO:
206 case OP_BRAMINZERO:
207 case OP_BRAPOSZERO:
208 case OP_SKIPZERO:
209 cc += PRIV(OP_lengths)[*cc];
210 do cc += GET(cc, 1); while (*cc == OP_ALT);
211 cc += 1 + LINK_SIZE;
212 break;
213
214 /* Handle literal characters and + repetitions */
215
216 case OP_CHAR:
217 case OP_CHARI:
218 case OP_NOT:
219 case OP_NOTI:
220 case OP_PLUS:
221 case OP_PLUSI:
222 case OP_MINPLUS:
223 case OP_MINPLUSI:
224 case OP_POSPLUS:
225 case OP_POSPLUSI:
226 case OP_NOTPLUS:
227 case OP_NOTPLUSI:
228 case OP_NOTMINPLUS:
229 case OP_NOTMINPLUSI:
230 case OP_NOTPOSPLUS:
231 case OP_NOTPOSPLUSI:
232 branchlength++;
233 cc += 2;
234 #ifdef SUPPORT_UTF
235 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
236 #endif
237 break;
238
239 case OP_TYPEPLUS:
240 case OP_TYPEMINPLUS:
241 case OP_TYPEPOSPLUS:
242 branchlength++;
243 cc += (cc[1] == OP_PROP || cc[1] == OP_NOTPROP)? 4 : 2;
244 break;
245
246 /* Handle exact repetitions. The count is already in characters, but we
247 need to skip over a multibyte character in UTF8 mode. */
248
249 case OP_EXACT:
250 case OP_EXACTI:
251 case OP_NOTEXACT:
252 case OP_NOTEXACTI:
253 branchlength += GET2(cc,1);
254 cc += 2 + IMM2_SIZE;
255 #ifdef SUPPORT_UTF
256 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
257 #endif
258 break;
259
260 case OP_TYPEEXACT:
261 branchlength += GET2(cc,1);
262 cc += 2 + IMM2_SIZE + ((cc[1 + IMM2_SIZE] == OP_PROP
263 || cc[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
264 break;
265
266 /* Handle single-char non-literal matchers */
267
268 case OP_PROP:
269 case OP_NOTPROP:
270 cc += 2;
271 /* Fall through */
272
273 case OP_NOT_DIGIT:
274 case OP_DIGIT:
275 case OP_NOT_WHITESPACE:
276 case OP_WHITESPACE:
277 case OP_NOT_WORDCHAR:
278 case OP_WORDCHAR:
279 case OP_ANY:
280 case OP_ALLANY:
281 case OP_EXTUNI:
282 case OP_HSPACE:
283 case OP_NOT_HSPACE:
284 case OP_VSPACE:
285 case OP_NOT_VSPACE:
286 branchlength++;
287 cc++;
288 break;
289
290 /* "Any newline" might match two characters, but it also might match just
291 one. */
292
293 case OP_ANYNL:
294 branchlength += 1;
295 cc++;
296 break;
297
298 /* The single-byte matcher means we can't proceed in UTF-8 mode. (In
299 non-UTF-8 mode \C will actually be turned into OP_ALLANY, so won't ever
300 appear, but leave the code, just in case.) */
301
302 case OP_ANYBYTE:
303 #ifdef SUPPORT_UTF
304 if (utf) return -1;
305 #endif
306 branchlength++;
307 cc++;
308 break;
309
310 /* For repeated character types, we have to test for \p and \P, which have
311 an extra two bytes of parameters. */
312
313 case OP_TYPESTAR:
314 case OP_TYPEMINSTAR:
315 case OP_TYPEQUERY:
316 case OP_TYPEMINQUERY:
317 case OP_TYPEPOSSTAR:
318 case OP_TYPEPOSQUERY:
319 if (cc[1] == OP_PROP || cc[1] == OP_NOTPROP) cc += 2;
320 cc += PRIV(OP_lengths)[op];
321 break;
322
323 case OP_TYPEUPTO:
324 case OP_TYPEMINUPTO:
325 case OP_TYPEPOSUPTO:
326 if (cc[1 + IMM2_SIZE] == OP_PROP
327 || cc[1 + IMM2_SIZE] == OP_NOTPROP) cc += 2;
328 cc += PRIV(OP_lengths)[op];
329 break;
330
331 /* Check a class for variable quantification */
332
333 case OP_CLASS:
334 case OP_NCLASS:
335 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
336 case OP_XCLASS:
337 /* The original code caused an unsigned overflow in 64 bit systems,
338 so now we use a conditional statement. */
339 if (op == OP_XCLASS)
340 cc += GET(cc, 1);
341 else
342 cc += PRIV(OP_lengths)[OP_CLASS];
343 #else
344 cc += PRIV(OP_lengths)[OP_CLASS];
345 #endif
346
347 switch (*cc)
348 {
349 case OP_CRPLUS:
350 case OP_CRMINPLUS:
351 case OP_CRPOSPLUS:
352 branchlength++;
353 /* Fall through */
354
355 case OP_CRSTAR:
356 case OP_CRMINSTAR:
357 case OP_CRQUERY:
358 case OP_CRMINQUERY:
359 case OP_CRPOSSTAR:
360 case OP_CRPOSQUERY:
361 cc++;
362 break;
363
364 case OP_CRRANGE:
365 case OP_CRMINRANGE:
366 case OP_CRPOSRANGE:
367 branchlength += GET2(cc,1);
368 cc += 1 + 2 * IMM2_SIZE;
369 break;
370
371 default:
372 branchlength++;
373 break;
374 }
375 break;
376
377 /* Backreferences and subroutine calls are treated in the same way: we find
378 the minimum length for the subpattern. A recursion, however, causes an
379 a flag to be set that causes the length of this branch to be ignored. The
380 logic is that a recursion can only make sense if there is another
381 alternation that stops the recursing. That will provide the minimum length
382 (when no recursion happens). A backreference within the group that it is
383 referencing behaves in the same way.
384
385 If PCRE_JAVASCRIPT_COMPAT is set, a backreference to an unset bracket
386 matches an empty string (by default it causes a matching failure), so in
387 that case we must set the minimum length to zero. */
388
389 case OP_DNREF: /* Duplicate named pattern back reference */
390 case OP_DNREFI:
391 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
392 {
393 int count = GET2(cc, 1+IMM2_SIZE);
394 pcre_uchar *slot = (pcre_uchar *)re +
395 re->name_table_offset + GET2(cc, 1) * re->name_entry_size;
396 d = INT_MAX;
397 while (count-- > 0)
398 {
399 ce = cs = (pcre_uchar *)PRIV(find_bracket)(startcode, utf, GET2(slot, 0));
400 if (cs == NULL) return -2;
401 do ce += GET(ce, 1); while (*ce == OP_ALT);
402 if (cc > cs && cc < ce) /* Simple recursion */
403 {
404 d = 0;
405 had_recurse = TRUE;
406 break;
407 }
408 else
409 {
410 recurse_check *r = recurses;
411 for (r = recurses; r != NULL; r = r->prev) if (r->group == cs) break;
412 if (r != NULL) /* Mutual recursion */
413 {
414 d = 0;
415 had_recurse = TRUE;
416 break;
417 }
418 else
419 {
420 int dd;
421 this_recurse.prev = recurses;
422 this_recurse.group = cs;
423 dd = find_minlength(re, cs, startcode, options, &this_recurse,
424 countptr);
425 if (dd < d) d = dd;
426 }
427 }
428 slot += re->name_entry_size;
429 }
430 }
431 else d = 0;
432 cc += 1 + 2*IMM2_SIZE;
433 goto REPEAT_BACK_REFERENCE;
434
435 case OP_REF: /* Single back reference */
436 case OP_REFI:
437 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0)
438 {
439 ce = cs = (pcre_uchar *)PRIV(find_bracket)(startcode, utf, GET2(cc, 1));
440 if (cs == NULL) return -2;
441 do ce += GET(ce, 1); while (*ce == OP_ALT);
442 if (cc > cs && cc < ce) /* Simple recursion */
443 {
444 d = 0;
445 had_recurse = TRUE;
446 }
447 else
448 {
449 recurse_check *r = recurses;
450 for (r = recurses; r != NULL; r = r->prev) if (r->group == cs) break;
451 if (r != NULL) /* Mutual recursion */
452 {
453 d = 0;
454 had_recurse = TRUE;
455 }
456 else
457 {
458 this_recurse.prev = recurses;
459 this_recurse.group = cs;
460 d = find_minlength(re, cs, startcode, options, &this_recurse,
461 countptr);
462 }
463 }
464 }
465 else d = 0;
466 cc += 1 + IMM2_SIZE;
467
468 /* Handle repeated back references */
469
470 REPEAT_BACK_REFERENCE:
471 switch (*cc)
472 {
473 case OP_CRSTAR:
474 case OP_CRMINSTAR:
475 case OP_CRQUERY:
476 case OP_CRMINQUERY:
477 case OP_CRPOSSTAR:
478 case OP_CRPOSQUERY:
479 min = 0;
480 cc++;
481 break;
482
483 case OP_CRPLUS:
484 case OP_CRMINPLUS:
485 case OP_CRPOSPLUS:
486 min = 1;
487 cc++;
488 break;
489
490 case OP_CRRANGE:
491 case OP_CRMINRANGE:
492 case OP_CRPOSRANGE:
493 min = GET2(cc, 1);
494 cc += 1 + 2 * IMM2_SIZE;
495 break;
496
497 default:
498 min = 1;
499 break;
500 }
501
502 branchlength += min * d;
503 break;
504
505 /* We can easily detect direct recursion, but not mutual recursion. This is
506 caught by a recursion depth count. */
507
508 case OP_RECURSE:
509 cs = ce = (pcre_uchar *)startcode + GET(cc, 1);
510 do ce += GET(ce, 1); while (*ce == OP_ALT);
511 if (cc > cs && cc < ce) /* Simple recursion */
512 had_recurse = TRUE;
513 else
514 {
515 recurse_check *r = recurses;
516 for (r = recurses; r != NULL; r = r->prev) if (r->group == cs) break;
517 if (r != NULL) /* Mutual recursion */
518 had_recurse = TRUE;
519 else
520 {
521 this_recurse.prev = recurses;
522 this_recurse.group = cs;
523 branchlength += find_minlength(re, cs, startcode, options,
524 &this_recurse, countptr);
525 }
526 }
527 cc += 1 + LINK_SIZE;
528 break;
529
530 /* Anything else does not or need not match a character. We can get the
531 item's length from the table, but for those that can match zero occurrences
532 of a character, we must take special action for UTF-8 characters. As it
533 happens, the "NOT" versions of these opcodes are used at present only for
534 ASCII characters, so they could be omitted from this list. However, in
535 future that may change, so we include them here so as not to leave a
536 gotcha for a future maintainer. */
537
538 case OP_UPTO:
539 case OP_UPTOI:
540 case OP_NOTUPTO:
541 case OP_NOTUPTOI:
542 case OP_MINUPTO:
543 case OP_MINUPTOI:
544 case OP_NOTMINUPTO:
545 case OP_NOTMINUPTOI:
546 case OP_POSUPTO:
547 case OP_POSUPTOI:
548 case OP_NOTPOSUPTO:
549 case OP_NOTPOSUPTOI:
550
551 case OP_STAR:
552 case OP_STARI:
553 case OP_NOTSTAR:
554 case OP_NOTSTARI:
555 case OP_MINSTAR:
556 case OP_MINSTARI:
557 case OP_NOTMINSTAR:
558 case OP_NOTMINSTARI:
559 case OP_POSSTAR:
560 case OP_POSSTARI:
561 case OP_NOTPOSSTAR:
562 case OP_NOTPOSSTARI:
563
564 case OP_QUERY:
565 case OP_QUERYI:
566 case OP_NOTQUERY:
567 case OP_NOTQUERYI:
568 case OP_MINQUERY:
569 case OP_MINQUERYI:
570 case OP_NOTMINQUERY:
571 case OP_NOTMINQUERYI:
572 case OP_POSQUERY:
573 case OP_POSQUERYI:
574 case OP_NOTPOSQUERY:
575 case OP_NOTPOSQUERYI:
576
577 cc += PRIV(OP_lengths)[op];
578 #ifdef SUPPORT_UTF
579 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
580 #endif
581 break;
582
583 /* Skip these, but we need to add in the name length. */
584
585 case OP_MARK:
586 case OP_PRUNE_ARG:
587 case OP_SKIP_ARG:
588 case OP_THEN_ARG:
589 cc += PRIV(OP_lengths)[op] + cc[1];
590 break;
591
592 /* The remaining opcodes are just skipped over. */
593
594 case OP_CLOSE:
595 case OP_COMMIT:
596 case OP_FAIL:
597 case OP_PRUNE:
598 case OP_SET_SOM:
599 case OP_SKIP:
600 case OP_THEN:
601 cc += PRIV(OP_lengths)[op];
602 break;
603
604 /* This should not occur: we list all opcodes explicitly so that when
605 new ones get added they are properly considered. */
606
607 default:
608 return -3;
609 }
610 }
611 /* Control never gets here */
612 }
613
614
615
616 /*************************************************
617 * Set a bit and maybe its alternate case *
618 *************************************************/
619
620 /* Given a character, set its first byte's bit in the table, and also the
621 corresponding bit for the other version of a letter if we are caseless. In
622 UTF-8 mode, for characters greater than 127, we can only do the caseless thing
623 when Unicode property support is available.
624
625 Arguments:
626 start_bits points to the bit map
627 p points to the character
628 caseless the caseless flag
629 cd the block with char table pointers
630 utf TRUE for UTF-8 / UTF-16 / UTF-32 mode
631
632 Returns: pointer after the character
633 */
634
635 static const pcre_uchar *
set_table_bit(pcre_uint8 * start_bits,const pcre_uchar * p,BOOL caseless,compile_data * cd,BOOL utf)636 set_table_bit(pcre_uint8 *start_bits, const pcre_uchar *p, BOOL caseless,
637 compile_data *cd, BOOL utf)
638 {
639 pcre_uint32 c = *p;
640
641 #ifdef COMPILE_PCRE8
642 SET_BIT(c);
643
644 #ifdef SUPPORT_UTF
645 if (utf && c > 127)
646 {
647 GETCHARINC(c, p);
648 #ifdef SUPPORT_UCP
649 if (caseless)
650 {
651 pcre_uchar buff[6];
652 c = UCD_OTHERCASE(c);
653 (void)PRIV(ord2utf)(c, buff);
654 SET_BIT(buff[0]);
655 }
656 #endif /* Not SUPPORT_UCP */
657 return p;
658 }
659 #else /* Not SUPPORT_UTF */
660 (void)(utf); /* Stops warning for unused parameter */
661 #endif /* SUPPORT_UTF */
662
663 /* Not UTF-8 mode, or character is less than 127. */
664
665 if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
666 return p + 1;
667 #endif /* COMPILE_PCRE8 */
668
669 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
670 if (c > 0xff)
671 {
672 c = 0xff;
673 caseless = FALSE;
674 }
675 SET_BIT(c);
676
677 #ifdef SUPPORT_UTF
678 if (utf && c > 127)
679 {
680 GETCHARINC(c, p);
681 #ifdef SUPPORT_UCP
682 if (caseless)
683 {
684 c = UCD_OTHERCASE(c);
685 if (c > 0xff)
686 c = 0xff;
687 SET_BIT(c);
688 }
689 #endif /* SUPPORT_UCP */
690 return p;
691 }
692 #else /* Not SUPPORT_UTF */
693 (void)(utf); /* Stops warning for unused parameter */
694 #endif /* SUPPORT_UTF */
695
696 if (caseless && (cd->ctypes[c] & ctype_letter) != 0) SET_BIT(cd->fcc[c]);
697 return p + 1;
698 #endif
699 }
700
701
702
703 /*************************************************
704 * Set bits for a positive character type *
705 *************************************************/
706
707 /* This function sets starting bits for a character type. In UTF-8 mode, we can
708 only do a direct setting for bytes less than 128, as otherwise there can be
709 confusion with bytes in the middle of UTF-8 characters. In a "traditional"
710 environment, the tables will only recognize ASCII characters anyway, but in at
711 least one Windows environment, some higher bytes bits were set in the tables.
712 So we deal with that case by considering the UTF-8 encoding.
713
714 Arguments:
715 start_bits the starting bitmap
716 cbit type the type of character wanted
717 table_limit 32 for non-UTF-8; 16 for UTF-8
718 cd the block with char table pointers
719
720 Returns: nothing
721 */
722
723 static void
set_type_bits(pcre_uint8 * start_bits,int cbit_type,unsigned int table_limit,compile_data * cd)724 set_type_bits(pcre_uint8 *start_bits, int cbit_type, unsigned int table_limit,
725 compile_data *cd)
726 {
727 register pcre_uint32 c;
728 for (c = 0; c < table_limit; c++) start_bits[c] |= cd->cbits[c+cbit_type];
729 #if defined SUPPORT_UTF && defined COMPILE_PCRE8
730 if (table_limit == 32) return;
731 for (c = 128; c < 256; c++)
732 {
733 if ((cd->cbits[c/8] & (1 << (c&7))) != 0)
734 {
735 pcre_uchar buff[6];
736 (void)PRIV(ord2utf)(c, buff);
737 SET_BIT(buff[0]);
738 }
739 }
740 #endif
741 }
742
743
744 /*************************************************
745 * Set bits for a negative character type *
746 *************************************************/
747
748 /* This function sets starting bits for a negative character type such as \D.
749 In UTF-8 mode, we can only do a direct setting for bytes less than 128, as
750 otherwise there can be confusion with bytes in the middle of UTF-8 characters.
751 Unlike in the positive case, where we can set appropriate starting bits for
752 specific high-valued UTF-8 characters, in this case we have to set the bits for
753 all high-valued characters. The lowest is 0xc2, but we overkill by starting at
754 0xc0 (192) for simplicity.
755
756 Arguments:
757 start_bits the starting bitmap
758 cbit type the type of character wanted
759 table_limit 32 for non-UTF-8; 16 for UTF-8
760 cd the block with char table pointers
761
762 Returns: nothing
763 */
764
765 static void
set_nottype_bits(pcre_uint8 * start_bits,int cbit_type,unsigned int table_limit,compile_data * cd)766 set_nottype_bits(pcre_uint8 *start_bits, int cbit_type, unsigned int table_limit,
767 compile_data *cd)
768 {
769 register pcre_uint32 c;
770 for (c = 0; c < table_limit; c++) start_bits[c] |= ~cd->cbits[c+cbit_type];
771 #if defined SUPPORT_UTF && defined COMPILE_PCRE8
772 if (table_limit != 32) for (c = 24; c < 32; c++) start_bits[c] = 0xff;
773 #endif
774 }
775
776
777
778 /*************************************************
779 * Create bitmap of starting bytes *
780 *************************************************/
781
782 /* This function scans a compiled unanchored expression recursively and
783 attempts to build a bitmap of the set of possible starting bytes. As time goes
784 by, we may be able to get more clever at doing this. The SSB_CONTINUE return is
785 useful for parenthesized groups in patterns such as (a*)b where the group
786 provides some optional starting bytes but scanning must continue at the outer
787 level to find at least one mandatory byte. At the outermost level, this
788 function fails unless the result is SSB_DONE.
789
790 Arguments:
791 code points to an expression
792 start_bits points to a 32-byte table, initialized to 0
793 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
794 cd the block with char table pointers
795
796 Returns: SSB_FAIL => Failed to find any starting bytes
797 SSB_DONE => Found mandatory starting bytes
798 SSB_CONTINUE => Found optional starting bytes
799 SSB_UNKNOWN => Hit an unrecognized opcode
800 */
801
802 static int
set_start_bits(const pcre_uchar * code,pcre_uint8 * start_bits,BOOL utf,compile_data * cd)803 set_start_bits(const pcre_uchar *code, pcre_uint8 *start_bits, BOOL utf,
804 compile_data *cd)
805 {
806 register pcre_uint32 c;
807 int yield = SSB_DONE;
808 #if defined SUPPORT_UTF && defined COMPILE_PCRE8
809 int table_limit = utf? 16:32;
810 #else
811 int table_limit = 32;
812 #endif
813
814 #if 0
815 /* ========================================================================= */
816 /* The following comment and code was inserted in January 1999. In May 2006,
817 when it was observed to cause compiler warnings about unused values, I took it
818 out again. If anybody is still using OS/2, they will have to put it back
819 manually. */
820
821 /* This next statement and the later reference to dummy are here in order to
822 trick the optimizer of the IBM C compiler for OS/2 into generating correct
823 code. Apparently IBM isn't going to fix the problem, and we would rather not
824 disable optimization (in this module it actually makes a big difference, and
825 the pcre module can use all the optimization it can get). */
826
827 volatile int dummy;
828 /* ========================================================================= */
829 #endif
830
831 do
832 {
833 BOOL try_next = TRUE;
834 const pcre_uchar *tcode = code + 1 + LINK_SIZE;
835
836 if (*code == OP_CBRA || *code == OP_SCBRA ||
837 *code == OP_CBRAPOS || *code == OP_SCBRAPOS) tcode += IMM2_SIZE;
838
839 while (try_next) /* Loop for items in this branch */
840 {
841 int rc;
842
843 switch(*tcode)
844 {
845 /* If we reach something we don't understand, it means a new opcode has
846 been created that hasn't been added to this code. Hopefully this problem
847 will be discovered during testing. */
848
849 default:
850 return SSB_UNKNOWN;
851
852 /* Fail for a valid opcode that implies no starting bits. */
853
854 case OP_ACCEPT:
855 case OP_ASSERT_ACCEPT:
856 case OP_ALLANY:
857 case OP_ANY:
858 case OP_ANYBYTE:
859 case OP_CIRC:
860 case OP_CIRCM:
861 case OP_CLOSE:
862 case OP_COMMIT:
863 case OP_COND:
864 case OP_CREF:
865 case OP_DEF:
866 case OP_DNCREF:
867 case OP_DNREF:
868 case OP_DNREFI:
869 case OP_DNRREF:
870 case OP_DOLL:
871 case OP_DOLLM:
872 case OP_END:
873 case OP_EOD:
874 case OP_EODN:
875 case OP_EXTUNI:
876 case OP_FAIL:
877 case OP_MARK:
878 case OP_NOT:
879 case OP_NOTEXACT:
880 case OP_NOTEXACTI:
881 case OP_NOTI:
882 case OP_NOTMINPLUS:
883 case OP_NOTMINPLUSI:
884 case OP_NOTMINQUERY:
885 case OP_NOTMINQUERYI:
886 case OP_NOTMINSTAR:
887 case OP_NOTMINSTARI:
888 case OP_NOTMINUPTO:
889 case OP_NOTMINUPTOI:
890 case OP_NOTPLUS:
891 case OP_NOTPLUSI:
892 case OP_NOTPOSPLUS:
893 case OP_NOTPOSPLUSI:
894 case OP_NOTPOSQUERY:
895 case OP_NOTPOSQUERYI:
896 case OP_NOTPOSSTAR:
897 case OP_NOTPOSSTARI:
898 case OP_NOTPOSUPTO:
899 case OP_NOTPOSUPTOI:
900 case OP_NOTPROP:
901 case OP_NOTQUERY:
902 case OP_NOTQUERYI:
903 case OP_NOTSTAR:
904 case OP_NOTSTARI:
905 case OP_NOTUPTO:
906 case OP_NOTUPTOI:
907 case OP_NOT_HSPACE:
908 case OP_NOT_VSPACE:
909 case OP_PRUNE:
910 case OP_PRUNE_ARG:
911 case OP_RECURSE:
912 case OP_REF:
913 case OP_REFI:
914 case OP_REVERSE:
915 case OP_RREF:
916 case OP_SCOND:
917 case OP_SET_SOM:
918 case OP_SKIP:
919 case OP_SKIP_ARG:
920 case OP_SOD:
921 case OP_SOM:
922 case OP_THEN:
923 case OP_THEN_ARG:
924 return SSB_FAIL;
925
926 /* A "real" property test implies no starting bits, but the fake property
927 PT_CLIST identifies a list of characters. These lists are short, as they
928 are used for characters with more than one "other case", so there is no
929 point in recognizing them for OP_NOTPROP. */
930
931 case OP_PROP:
932 if (tcode[1] != PT_CLIST) return SSB_FAIL;
933 {
934 const pcre_uint32 *p = PRIV(ucd_caseless_sets) + tcode[2];
935 while ((c = *p++) < NOTACHAR)
936 {
937 #if defined SUPPORT_UTF && defined COMPILE_PCRE8
938 if (utf)
939 {
940 pcre_uchar buff[6];
941 (void)PRIV(ord2utf)(c, buff);
942 c = buff[0];
943 }
944 #endif
945 if (c > 0xff) SET_BIT(0xff); else SET_BIT(c);
946 }
947 }
948 try_next = FALSE;
949 break;
950
951 /* We can ignore word boundary tests. */
952
953 case OP_WORD_BOUNDARY:
954 case OP_NOT_WORD_BOUNDARY:
955 tcode++;
956 break;
957
958 /* If we hit a bracket or a positive lookahead assertion, recurse to set
959 bits from within the subpattern. If it can't find anything, we have to
960 give up. If it finds some mandatory character(s), we are done for this
961 branch. Otherwise, carry on scanning after the subpattern. */
962
963 case OP_BRA:
964 case OP_SBRA:
965 case OP_CBRA:
966 case OP_SCBRA:
967 case OP_BRAPOS:
968 case OP_SBRAPOS:
969 case OP_CBRAPOS:
970 case OP_SCBRAPOS:
971 case OP_ONCE:
972 case OP_ONCE_NC:
973 case OP_ASSERT:
974 rc = set_start_bits(tcode, start_bits, utf, cd);
975 if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
976 if (rc == SSB_DONE) try_next = FALSE; else
977 {
978 do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
979 tcode += 1 + LINK_SIZE;
980 }
981 break;
982
983 /* If we hit ALT or KET, it means we haven't found anything mandatory in
984 this branch, though we might have found something optional. For ALT, we
985 continue with the next alternative, but we have to arrange that the final
986 result from subpattern is SSB_CONTINUE rather than SSB_DONE. For KET,
987 return SSB_CONTINUE: if this is the top level, that indicates failure,
988 but after a nested subpattern, it causes scanning to continue. */
989
990 case OP_ALT:
991 yield = SSB_CONTINUE;
992 try_next = FALSE;
993 break;
994
995 case OP_KET:
996 case OP_KETRMAX:
997 case OP_KETRMIN:
998 case OP_KETRPOS:
999 return SSB_CONTINUE;
1000
1001 /* Skip over callout */
1002
1003 case OP_CALLOUT:
1004 tcode += 2 + 2*LINK_SIZE;
1005 break;
1006
1007 /* Skip over lookbehind and negative lookahead assertions */
1008
1009 case OP_ASSERT_NOT:
1010 case OP_ASSERTBACK:
1011 case OP_ASSERTBACK_NOT:
1012 do tcode += GET(tcode, 1); while (*tcode == OP_ALT);
1013 tcode += 1 + LINK_SIZE;
1014 break;
1015
1016 /* BRAZERO does the bracket, but carries on. */
1017
1018 case OP_BRAZERO:
1019 case OP_BRAMINZERO:
1020 case OP_BRAPOSZERO:
1021 rc = set_start_bits(++tcode, start_bits, utf, cd);
1022 if (rc == SSB_FAIL || rc == SSB_UNKNOWN) return rc;
1023 /* =========================================================================
1024 See the comment at the head of this function concerning the next line,
1025 which was an old fudge for the benefit of OS/2.
1026 dummy = 1;
1027 ========================================================================= */
1028 do tcode += GET(tcode,1); while (*tcode == OP_ALT);
1029 tcode += 1 + LINK_SIZE;
1030 break;
1031
1032 /* SKIPZERO skips the bracket. */
1033
1034 case OP_SKIPZERO:
1035 tcode++;
1036 do tcode += GET(tcode,1); while (*tcode == OP_ALT);
1037 tcode += 1 + LINK_SIZE;
1038 break;
1039
1040 /* Single-char * or ? sets the bit and tries the next item */
1041
1042 case OP_STAR:
1043 case OP_MINSTAR:
1044 case OP_POSSTAR:
1045 case OP_QUERY:
1046 case OP_MINQUERY:
1047 case OP_POSQUERY:
1048 tcode = set_table_bit(start_bits, tcode + 1, FALSE, cd, utf);
1049 break;
1050
1051 case OP_STARI:
1052 case OP_MINSTARI:
1053 case OP_POSSTARI:
1054 case OP_QUERYI:
1055 case OP_MINQUERYI:
1056 case OP_POSQUERYI:
1057 tcode = set_table_bit(start_bits, tcode + 1, TRUE, cd, utf);
1058 break;
1059
1060 /* Single-char upto sets the bit and tries the next */
1061
1062 case OP_UPTO:
1063 case OP_MINUPTO:
1064 case OP_POSUPTO:
1065 tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, FALSE, cd, utf);
1066 break;
1067
1068 case OP_UPTOI:
1069 case OP_MINUPTOI:
1070 case OP_POSUPTOI:
1071 tcode = set_table_bit(start_bits, tcode + 1 + IMM2_SIZE, TRUE, cd, utf);
1072 break;
1073
1074 /* At least one single char sets the bit and stops */
1075
1076 case OP_EXACT:
1077 tcode += IMM2_SIZE;
1078 /* Fall through */
1079 case OP_CHAR:
1080 case OP_PLUS:
1081 case OP_MINPLUS:
1082 case OP_POSPLUS:
1083 (void)set_table_bit(start_bits, tcode + 1, FALSE, cd, utf);
1084 try_next = FALSE;
1085 break;
1086
1087 case OP_EXACTI:
1088 tcode += IMM2_SIZE;
1089 /* Fall through */
1090 case OP_CHARI:
1091 case OP_PLUSI:
1092 case OP_MINPLUSI:
1093 case OP_POSPLUSI:
1094 (void)set_table_bit(start_bits, tcode + 1, TRUE, cd, utf);
1095 try_next = FALSE;
1096 break;
1097
1098 /* Special spacing and line-terminating items. These recognize specific
1099 lists of characters. The difference between VSPACE and ANYNL is that the
1100 latter can match the two-character CRLF sequence, but that is not
1101 relevant for finding the first character, so their code here is
1102 identical. */
1103
1104 case OP_HSPACE:
1105 SET_BIT(CHAR_HT);
1106 SET_BIT(CHAR_SPACE);
1107 #ifdef SUPPORT_UTF
1108 if (utf)
1109 {
1110 #ifdef COMPILE_PCRE8
1111 SET_BIT(0xC2); /* For U+00A0 */
1112 SET_BIT(0xE1); /* For U+1680, U+180E */
1113 SET_BIT(0xE2); /* For U+2000 - U+200A, U+202F, U+205F */
1114 SET_BIT(0xE3); /* For U+3000 */
1115 #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1116 SET_BIT(0xA0);
1117 SET_BIT(0xFF); /* For characters > 255 */
1118 #endif /* COMPILE_PCRE[8|16|32] */
1119 }
1120 else
1121 #endif /* SUPPORT_UTF */
1122 {
1123 #ifndef EBCDIC
1124 SET_BIT(0xA0);
1125 #endif /* Not EBCDIC */
1126 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1127 SET_BIT(0xFF); /* For characters > 255 */
1128 #endif /* COMPILE_PCRE[16|32] */
1129 }
1130 try_next = FALSE;
1131 break;
1132
1133 case OP_ANYNL:
1134 case OP_VSPACE:
1135 SET_BIT(CHAR_LF);
1136 SET_BIT(CHAR_VT);
1137 SET_BIT(CHAR_FF);
1138 SET_BIT(CHAR_CR);
1139 #ifdef SUPPORT_UTF
1140 if (utf)
1141 {
1142 #ifdef COMPILE_PCRE8
1143 SET_BIT(0xC2); /* For U+0085 */
1144 SET_BIT(0xE2); /* For U+2028, U+2029 */
1145 #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1146 SET_BIT(CHAR_NEL);
1147 SET_BIT(0xFF); /* For characters > 255 */
1148 #endif /* COMPILE_PCRE[8|16|32] */
1149 }
1150 else
1151 #endif /* SUPPORT_UTF */
1152 {
1153 SET_BIT(CHAR_NEL);
1154 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1155 SET_BIT(0xFF); /* For characters > 255 */
1156 #endif
1157 }
1158 try_next = FALSE;
1159 break;
1160
1161 /* Single character types set the bits and stop. Note that if PCRE_UCP
1162 is set, we do not see these op codes because \d etc are converted to
1163 properties. Therefore, these apply in the case when only characters less
1164 than 256 are recognized to match the types. */
1165
1166 case OP_NOT_DIGIT:
1167 set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
1168 try_next = FALSE;
1169 break;
1170
1171 case OP_DIGIT:
1172 set_type_bits(start_bits, cbit_digit, table_limit, cd);
1173 try_next = FALSE;
1174 break;
1175
1176 /* The cbit_space table has vertical tab as whitespace; we no longer
1177 have to play fancy tricks because Perl added VT to its whitespace at
1178 release 5.18. PCRE added it at release 8.34. */
1179
1180 case OP_NOT_WHITESPACE:
1181 set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1182 try_next = FALSE;
1183 break;
1184
1185 case OP_WHITESPACE:
1186 set_type_bits(start_bits, cbit_space, table_limit, cd);
1187 try_next = FALSE;
1188 break;
1189
1190 case OP_NOT_WORDCHAR:
1191 set_nottype_bits(start_bits, cbit_word, table_limit, cd);
1192 try_next = FALSE;
1193 break;
1194
1195 case OP_WORDCHAR:
1196 set_type_bits(start_bits, cbit_word, table_limit, cd);
1197 try_next = FALSE;
1198 break;
1199
1200 /* One or more character type fudges the pointer and restarts, knowing
1201 it will hit a single character type and stop there. */
1202
1203 case OP_TYPEPLUS:
1204 case OP_TYPEMINPLUS:
1205 case OP_TYPEPOSPLUS:
1206 tcode++;
1207 break;
1208
1209 case OP_TYPEEXACT:
1210 tcode += 1 + IMM2_SIZE;
1211 break;
1212
1213 /* Zero or more repeats of character types set the bits and then
1214 try again. */
1215
1216 case OP_TYPEUPTO:
1217 case OP_TYPEMINUPTO:
1218 case OP_TYPEPOSUPTO:
1219 tcode += IMM2_SIZE; /* Fall through */
1220
1221 case OP_TYPESTAR:
1222 case OP_TYPEMINSTAR:
1223 case OP_TYPEPOSSTAR:
1224 case OP_TYPEQUERY:
1225 case OP_TYPEMINQUERY:
1226 case OP_TYPEPOSQUERY:
1227 switch(tcode[1])
1228 {
1229 default:
1230 case OP_ANY:
1231 case OP_ALLANY:
1232 return SSB_FAIL;
1233
1234 case OP_HSPACE:
1235 SET_BIT(CHAR_HT);
1236 SET_BIT(CHAR_SPACE);
1237 #ifdef SUPPORT_UTF
1238 if (utf)
1239 {
1240 #ifdef COMPILE_PCRE8
1241 SET_BIT(0xC2); /* For U+00A0 */
1242 SET_BIT(0xE1); /* For U+1680, U+180E */
1243 SET_BIT(0xE2); /* For U+2000 - U+200A, U+202F, U+205F */
1244 SET_BIT(0xE3); /* For U+3000 */
1245 #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1246 SET_BIT(0xA0);
1247 SET_BIT(0xFF); /* For characters > 255 */
1248 #endif /* COMPILE_PCRE[8|16|32] */
1249 }
1250 else
1251 #endif /* SUPPORT_UTF */
1252 #ifndef EBCDIC
1253 SET_BIT(0xA0);
1254 #endif /* Not EBCDIC */
1255 break;
1256
1257 case OP_ANYNL:
1258 case OP_VSPACE:
1259 SET_BIT(CHAR_LF);
1260 SET_BIT(CHAR_VT);
1261 SET_BIT(CHAR_FF);
1262 SET_BIT(CHAR_CR);
1263 #ifdef SUPPORT_UTF
1264 if (utf)
1265 {
1266 #ifdef COMPILE_PCRE8
1267 SET_BIT(0xC2); /* For U+0085 */
1268 SET_BIT(0xE2); /* For U+2028, U+2029 */
1269 #elif defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1270 SET_BIT(CHAR_NEL);
1271 SET_BIT(0xFF); /* For characters > 255 */
1272 #endif /* COMPILE_PCRE16 */
1273 }
1274 else
1275 #endif /* SUPPORT_UTF */
1276 SET_BIT(CHAR_NEL);
1277 break;
1278
1279 case OP_NOT_DIGIT:
1280 set_nottype_bits(start_bits, cbit_digit, table_limit, cd);
1281 break;
1282
1283 case OP_DIGIT:
1284 set_type_bits(start_bits, cbit_digit, table_limit, cd);
1285 break;
1286
1287 /* The cbit_space table has vertical tab as whitespace; we no longer
1288 have to play fancy tricks because Perl added VT to its whitespace at
1289 release 5.18. PCRE added it at release 8.34. */
1290
1291 case OP_NOT_WHITESPACE:
1292 set_nottype_bits(start_bits, cbit_space, table_limit, cd);
1293 break;
1294
1295 case OP_WHITESPACE:
1296 set_type_bits(start_bits, cbit_space, table_limit, cd);
1297 break;
1298
1299 case OP_NOT_WORDCHAR:
1300 set_nottype_bits(start_bits, cbit_word, table_limit, cd);
1301 break;
1302
1303 case OP_WORDCHAR:
1304 set_type_bits(start_bits, cbit_word, table_limit, cd);
1305 break;
1306 }
1307
1308 tcode += 2;
1309 break;
1310
1311 /* Character class where all the information is in a bit map: set the
1312 bits and either carry on or not, according to the repeat count. If it was
1313 a negative class, and we are operating with UTF-8 characters, any byte
1314 with a value >= 0xc4 is a potentially valid starter because it starts a
1315 character with a value > 255. */
1316
1317 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1318 case OP_XCLASS:
1319 if ((tcode[1 + LINK_SIZE] & XCL_HASPROP) != 0)
1320 return SSB_FAIL;
1321 /* All bits are set. */
1322 if ((tcode[1 + LINK_SIZE] & XCL_MAP) == 0 && (tcode[1 + LINK_SIZE] & XCL_NOT) != 0)
1323 return SSB_FAIL;
1324 #endif
1325 /* Fall through */
1326
1327 case OP_NCLASS:
1328 #if defined SUPPORT_UTF && defined COMPILE_PCRE8
1329 if (utf)
1330 {
1331 start_bits[24] |= 0xf0; /* Bits for 0xc4 - 0xc8 */
1332 memset(start_bits+25, 0xff, 7); /* Bits for 0xc9 - 0xff */
1333 }
1334 #endif
1335 #if defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1336 SET_BIT(0xFF); /* For characters > 255 */
1337 #endif
1338 /* Fall through */
1339
1340 case OP_CLASS:
1341 {
1342 pcre_uint8 *map;
1343 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1344 map = NULL;
1345 if (*tcode == OP_XCLASS)
1346 {
1347 if ((tcode[1 + LINK_SIZE] & XCL_MAP) != 0)
1348 map = (pcre_uint8 *)(tcode + 1 + LINK_SIZE + 1);
1349 tcode += GET(tcode, 1);
1350 }
1351 else
1352 #endif
1353 {
1354 tcode++;
1355 map = (pcre_uint8 *)tcode;
1356 tcode += 32 / sizeof(pcre_uchar);
1357 }
1358
1359 /* In UTF-8 mode, the bits in a bit map correspond to character
1360 values, not to byte values. However, the bit map we are constructing is
1361 for byte values. So we have to do a conversion for characters whose
1362 value is > 127. In fact, there are only two possible starting bytes for
1363 characters in the range 128 - 255. */
1364
1365 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
1366 if (map != NULL)
1367 #endif
1368 {
1369 #if defined SUPPORT_UTF && defined COMPILE_PCRE8
1370 if (utf)
1371 {
1372 for (c = 0; c < 16; c++) start_bits[c] |= map[c];
1373 for (c = 128; c < 256; c++)
1374 {
1375 if ((map[c/8] & (1 << (c&7))) != 0)
1376 {
1377 int d = (c >> 6) | 0xc0; /* Set bit for this starter */
1378 start_bits[d/8] |= (1 << (d&7)); /* and then skip on to the */
1379 c = (c & 0xc0) + 0x40 - 1; /* next relevant character. */
1380 }
1381 }
1382 }
1383 else
1384 #endif
1385 {
1386 /* In non-UTF-8 mode, the two bit maps are completely compatible. */
1387 for (c = 0; c < 32; c++) start_bits[c] |= map[c];
1388 }
1389 }
1390
1391 /* Advance past the bit map, and act on what follows. For a zero
1392 minimum repeat, continue; otherwise stop processing. */
1393
1394 switch (*tcode)
1395 {
1396 case OP_CRSTAR:
1397 case OP_CRMINSTAR:
1398 case OP_CRQUERY:
1399 case OP_CRMINQUERY:
1400 case OP_CRPOSSTAR:
1401 case OP_CRPOSQUERY:
1402 tcode++;
1403 break;
1404
1405 case OP_CRRANGE:
1406 case OP_CRMINRANGE:
1407 case OP_CRPOSRANGE:
1408 if (GET2(tcode, 1) == 0) tcode += 1 + 2 * IMM2_SIZE;
1409 else try_next = FALSE;
1410 break;
1411
1412 default:
1413 try_next = FALSE;
1414 break;
1415 }
1416 }
1417 break; /* End of bitmap class handling */
1418
1419 } /* End of switch */
1420 } /* End of try_next loop */
1421
1422 code += GET(code, 1); /* Advance to next branch */
1423 }
1424 while (*code == OP_ALT);
1425 return yield;
1426 }
1427
1428
1429
1430
1431
1432 /*************************************************
1433 * Study a compiled expression *
1434 *************************************************/
1435
1436 /* This function is handed a compiled expression that it must study to produce
1437 information that will speed up the matching. It returns a pcre[16]_extra block
1438 which then gets handed back to pcre_exec().
1439
1440 Arguments:
1441 re points to the compiled expression
1442 options contains option bits
1443 errorptr points to where to place error messages;
1444 set NULL unless error
1445
1446 Returns: pointer to a pcre[16]_extra block, with study_data filled in and
1447 the appropriate flags set;
1448 NULL on error or if no optimization possible
1449 */
1450
1451 #if defined COMPILE_PCRE8
1452 #if defined(ERLANG_INTEGRATION)
1453 PCRE_EXP_DEFN erts_pcre_extra * PCRE_CALL_CONVENTION
erts_pcre_study(const pcre * external_re,int options,const char ** errorptr)1454 erts_pcre_study(const pcre *external_re, int options, const char **errorptr)
1455 #else
1456 PCRE_EXP_DEFN pcre_extra * PCRE_CALL_CONVENTION
1457 pcre_study(const pcre *external_re, int options, const char **errorptr)
1458 #endif
1459 #elif defined COMPILE_PCRE16
1460 PCRE_EXP_DEFN pcre16_extra * PCRE_CALL_CONVENTION
1461 pcre16_study(const pcre16 *external_re, int options, const char **errorptr)
1462 #elif defined COMPILE_PCRE32
1463 PCRE_EXP_DEFN pcre32_extra * PCRE_CALL_CONVENTION
1464 pcre32_study(const pcre32 *external_re, int options, const char **errorptr)
1465 #endif
1466 {
1467 int min;
1468 int count = 0;
1469 BOOL bits_set = FALSE;
1470 pcre_uint8 start_bits[32];
1471 PUBL(extra) *extra = NULL;
1472 pcre_study_data *study;
1473 const pcre_uint8 *tables;
1474 pcre_uchar *code;
1475 compile_data compile_block;
1476 const REAL_PCRE *re = (const REAL_PCRE *)external_re;
1477
1478
1479 *errorptr = NULL;
1480
1481 if (re == NULL || re->magic_number != MAGIC_NUMBER)
1482 {
1483 *errorptr = "argument is not a compiled regular expression";
1484 return NULL;
1485 }
1486
1487 if ((re->flags & PCRE_MODE) == 0)
1488 {
1489 #if defined COMPILE_PCRE8
1490 *errorptr = "argument not compiled in 8 bit mode";
1491 #elif defined COMPILE_PCRE16
1492 *errorptr = "argument not compiled in 16 bit mode";
1493 #elif defined COMPILE_PCRE32
1494 *errorptr = "argument not compiled in 32 bit mode";
1495 #endif
1496 return NULL;
1497 }
1498
1499 if ((options & ~PUBLIC_STUDY_OPTIONS) != 0)
1500 {
1501 *errorptr = "unknown or incorrect option bit(s) set";
1502 return NULL;
1503 }
1504
1505 code = (pcre_uchar *)re + re->name_table_offset +
1506 (re->name_count * re->name_entry_size);
1507
1508 /* For an anchored pattern, or an unanchored pattern that has a first char, or
1509 a multiline pattern that matches only at "line starts", there is no point in
1510 seeking a list of starting bytes. */
1511
1512 if ((re->options & PCRE_ANCHORED) == 0 &&
1513 (re->flags & (PCRE_FIRSTSET|PCRE_STARTLINE)) == 0)
1514 {
1515 int rc;
1516
1517 /* Set the character tables in the block that is passed around */
1518
1519 tables = re->tables;
1520
1521 #if defined COMPILE_PCRE8
1522 #if defined(ERLANG_INTEGRATION)
1523 if (tables == NULL)
1524 (void)erts_pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1525 (void *)(&tables));
1526 #else
1527 if (tables == NULL)
1528 (void)pcre_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1529 (void *)(&tables));
1530 #endif
1531 #elif defined COMPILE_PCRE16
1532 if (tables == NULL)
1533 (void)pcre16_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1534 (void *)(&tables));
1535 #elif defined COMPILE_PCRE32
1536 if (tables == NULL)
1537 (void)pcre32_fullinfo(external_re, NULL, PCRE_INFO_DEFAULT_TABLES,
1538 (void *)(&tables));
1539 #endif
1540
1541 compile_block.lcc = tables + lcc_offset;
1542 compile_block.fcc = tables + fcc_offset;
1543 compile_block.cbits = tables + cbits_offset;
1544 compile_block.ctypes = tables + ctypes_offset;
1545
1546 /* See if we can find a fixed set of initial characters for the pattern. */
1547
1548 memset(start_bits, 0, 32 * sizeof(pcre_uint8));
1549 rc = set_start_bits(code, start_bits, (re->options & PCRE_UTF8) != 0,
1550 &compile_block);
1551 bits_set = rc == SSB_DONE;
1552 if (rc == SSB_UNKNOWN)
1553 {
1554 *errorptr = "internal error: opcode not recognized";
1555 return NULL;
1556 }
1557 }
1558
1559 /* Find the minimum length of subject string. */
1560
1561 switch(min = find_minlength(re, code, code, re->options, NULL, &count))
1562 {
1563 case -2: *errorptr = "internal error: missing capturing bracket"; return NULL;
1564 case -3: *errorptr = "internal error: opcode not recognized"; return NULL;
1565 default: break;
1566 }
1567
1568 /* If a set of starting bytes has been identified, or if the minimum length is
1569 greater than zero, or if JIT optimization has been requested, or if
1570 PCRE_STUDY_EXTRA_NEEDED is set, get a pcre[16]_extra block and a
1571 pcre_study_data block. The study data is put in the latter, which is pointed to
1572 by the former, which may also get additional data set later by the calling
1573 program. At the moment, the size of pcre_study_data is fixed. We nevertheless
1574 save it in a field for returning via the pcre_fullinfo() function so that if it
1575 becomes variable in the future, we don't have to change that code. */
1576
1577 if (bits_set || min > 0 || (options & (
1578 #ifdef SUPPORT_JIT
1579 PCRE_STUDY_JIT_COMPILE | PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE |
1580 PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE |
1581 #endif
1582 PCRE_STUDY_EXTRA_NEEDED)) != 0)
1583 {
1584 extra = (PUBL(extra) *)(PUBL(malloc))
1585 (sizeof(PUBL(extra)) + sizeof(pcre_study_data));
1586 if (extra == NULL)
1587 {
1588 *errorptr = "failed to get memory";
1589 return NULL;
1590 }
1591
1592 study = (pcre_study_data *)((char *)extra + sizeof(PUBL(extra)));
1593 extra->flags = PCRE_EXTRA_STUDY_DATA;
1594 extra->study_data = study;
1595
1596 study->size = sizeof(pcre_study_data);
1597 study->flags = 0;
1598
1599 /* Set the start bits always, to avoid unset memory errors if the
1600 study data is written to a file, but set the flag only if any of the bits
1601 are set, to save time looking when none are. */
1602
1603 if (bits_set)
1604 {
1605 study->flags |= PCRE_STUDY_MAPPED;
1606 memcpy(study->start_bits, start_bits, sizeof(start_bits));
1607 }
1608 else memset(study->start_bits, 0, 32 * sizeof(pcre_uint8));
1609
1610 #ifdef PCRE_DEBUG
1611 if (bits_set)
1612 {
1613 pcre_uint8 *ptr = start_bits;
1614 int i;
1615
1616 printf("Start bits:\n");
1617 for (i = 0; i < 32; i++)
1618 printf("%3d: %02x%s", i * 8, *ptr++, ((i + 1) & 0x7) != 0? " " : "\n");
1619 }
1620 #endif
1621
1622 /* Always set the minlength value in the block, because the JIT compiler
1623 makes use of it. However, don't set the bit unless the length is greater than
1624 zero - the interpretive pcre_exec() and pcre_dfa_exec() needn't waste time
1625 checking the zero case. */
1626
1627 if (min > 0)
1628 {
1629 study->flags |= PCRE_STUDY_MINLEN;
1630 study->minlength = min;
1631 }
1632 else study->minlength = 0;
1633
1634 /* If JIT support was compiled and requested, attempt the JIT compilation.
1635 If no starting bytes were found, and the minimum length is zero, and JIT
1636 compilation fails, abandon the extra block and return NULL, unless
1637 PCRE_STUDY_EXTRA_NEEDED is set. */
1638
1639 #ifdef SUPPORT_JIT
1640 extra->executable_jit = NULL;
1641 if ((options & PCRE_STUDY_JIT_COMPILE) != 0)
1642 PRIV(jit_compile)(re, extra, JIT_COMPILE);
1643 if ((options & PCRE_STUDY_JIT_PARTIAL_SOFT_COMPILE) != 0)
1644 PRIV(jit_compile)(re, extra, JIT_PARTIAL_SOFT_COMPILE);
1645 if ((options & PCRE_STUDY_JIT_PARTIAL_HARD_COMPILE) != 0)
1646 PRIV(jit_compile)(re, extra, JIT_PARTIAL_HARD_COMPILE);
1647
1648 if (study->flags == 0 && (extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) == 0 &&
1649 (options & PCRE_STUDY_EXTRA_NEEDED) == 0)
1650 {
1651 #if defined COMPILE_PCRE8
1652 #if defined(ERLANG_INTEGRATION)
1653 erts_pcre_free_study(extra);
1654 #else
1655 pcre_free_study(extra);
1656 #endif
1657 #elif defined COMPILE_PCRE16
1658 pcre16_free_study(extra);
1659 #elif defined COMPILE_PCRE32
1660 pcre32_free_study(extra);
1661 #endif
1662 extra = NULL;
1663 }
1664 #endif
1665 }
1666
1667 return extra;
1668 }
1669
1670
1671 /*************************************************
1672 * Free the study data *
1673 *************************************************/
1674
1675 /* This function frees the memory that was obtained by pcre_study().
1676
1677 Argument: a pointer to the pcre[16]_extra block
1678 Returns: nothing
1679 */
1680
1681 #if defined COMPILE_PCRE8
1682 #if defined(ERLANG_INTEGRATION)
1683 PCRE_EXP_DEFN void
erts_pcre_free_study(erts_pcre_extra * extra)1684 erts_pcre_free_study(erts_pcre_extra *extra)
1685 #else
1686 PCRE_EXP_DEFN void
1687 pcre_free_study(pcre_extra *extra)
1688 #endif
1689 #elif defined COMPILE_PCRE16
1690 PCRE_EXP_DEFN void
1691 pcre16_free_study(pcre16_extra *extra)
1692 #elif defined COMPILE_PCRE32
1693 PCRE_EXP_DEFN void
1694 pcre32_free_study(pcre32_extra *extra)
1695 #endif
1696 {
1697 if (extra == NULL)
1698 return;
1699 #ifdef SUPPORT_JIT
1700 if ((extra->flags & PCRE_EXTRA_EXECUTABLE_JIT) != 0 &&
1701 extra->executable_jit != NULL)
1702 PRIV(jit_free)(extra->executable_jit);
1703 #endif
1704 PUBL(free)(extra);
1705 }
1706
1707 /* End of pcre_study.c */
1708