?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Add some more cases to the app-id unit tests
[glib.git] / glib / pcre / pcre_dfa_exec.c
blobf8c887fb3578291f30fac14f9c7efee0d57e4c0f
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 (but see
7 below for why this module is different).
8
9 Written by Philip Hazel
10 Copyright (c) 1997-2012 University of Cambridge
11
12 -----------------------------------------------------------------------------
13 Redistribution and use in source and binary forms, with or without
14 modification, are permitted provided that the following conditions are met:
15
16 * Redistributions of source code must retain the above copyright notice,
17 this list of conditions and the following disclaimer.
18
19 * Redistributions in binary form must reproduce the above copyright
20 notice, this list of conditions and the following disclaimer in the
21 documentation and/or other materials provided with the distribution.
22
23 * Neither the name of the University of Cambridge nor the names of its
24 contributors may be used to endorse or promote products derived from
25 this software without specific prior written permission.
26
27 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
28 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
31 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
32 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
33 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
34 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
35 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
36 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
37 POSSIBILITY OF SUCH DAMAGE.
38 -----------------------------------------------------------------------------
39 */
40
41 /* This module contains the external function pcre_dfa_exec(), which is an
42 alternative matching function that uses a sort of DFA algorithm (not a true
43 FSM). This is NOT Perl-compatible, but it has advantages in certain
44 applications. */
45
46
47 /* NOTE ABOUT PERFORMANCE: A user of this function sent some code that improved
48 the performance of his patterns greatly. I could not use it as it stood, as it
49 was not thread safe, and made assumptions about pattern sizes. Also, it caused
50 test 7 to loop, and test 9 to crash with a segfault.
51
52 The issue is the check for duplicate states, which is done by a simple linear
53 search up the state list. (Grep for "duplicate" below to find the code.) For
54 many patterns, there will never be many states active at one time, so a simple
55 linear search is fine. In patterns that have many active states, it might be a
56 bottleneck. The suggested code used an indexing scheme to remember which states
57 had previously been used for each character, and avoided the linear search when
58 it knew there was no chance of a duplicate. This was implemented when adding
59 states to the state lists.
60
61 I wrote some thread-safe, not-limited code to try something similar at the time
62 of checking for duplicates (instead of when adding states), using index vectors
63 on the stack. It did give a 13% improvement with one specially constructed
64 pattern for certain subject strings, but on other strings and on many of the
65 simpler patterns in the test suite it did worse. The major problem, I think,
66 was the extra time to initialize the index. This had to be done for each call
67 of internal_dfa_exec(). (The supplied patch used a static vector, initialized
68 only once - I suspect this was the cause of the problems with the tests.)
69
70 Overall, I concluded that the gains in some cases did not outweigh the losses
71 in others, so I abandoned this code. */
72
73
74
75 #ifdef HAVE_CONFIG_H
76 #include "config.h"
77 #endif
78
79 #define NLBLOCK md /* Block containing newline information */
80 #define PSSTART start_subject /* Field containing processed string start */
81 #define PSEND end_subject /* Field containing processed string end */
82
83 #include "pcre_internal.h"
84
85
86 /* For use to indent debugging output */
87
88 #define SP " "
89
90
91 /*************************************************
92 * Code parameters and static tables *
93 *************************************************/
94
95 /* These are offsets that are used to turn the OP_TYPESTAR and friends opcodes
96 into others, under special conditions. A gap of 20 between the blocks should be
97 enough. The resulting opcodes don't have to be less than 256 because they are
98 never stored, so we push them well clear of the normal opcodes. */
99
100 #define OP_PROP_EXTRA 300
101 #define OP_EXTUNI_EXTRA 320
102 #define OP_ANYNL_EXTRA 340
103 #define OP_HSPACE_EXTRA 360
104 #define OP_VSPACE_EXTRA 380
105
106
107 /* This table identifies those opcodes that are followed immediately by a
108 character that is to be tested in some way. This makes it possible to
109 centralize the loading of these characters. In the case of Type * etc, the
110 "character" is the opcode for \D, \d, \S, \s, \W, or \w, which will always be a
111 small value. Non-zero values in the table are the offsets from the opcode where
112 the character is to be found. ***NOTE*** If the start of this table is
113 modified, the three tables that follow must also be modified. */
114
115 static const pcre_uint8 coptable[] = {
116 0, /* End */
117 0, 0, 0, 0, 0, /* \A, \G, \K, \B, \b */
118 0, 0, 0, 0, 0, 0, /* \D, \d, \S, \s, \W, \w */
119 0, 0, 0, /* Any, AllAny, Anybyte */
120 0, 0, /* \P, \p */
121 0, 0, 0, 0, 0, /* \R, \H, \h, \V, \v */
122 0, /* \X */
123 0, 0, 0, 0, 0, 0, /* \Z, \z, ^, ^M, $, $M */
124 1, /* Char */
125 1, /* Chari */
126 1, /* not */
127 1, /* noti */
128 /* Positive single-char repeats */
129 1, 1, 1, 1, 1, 1, /* *, *?, +, +?, ?, ?? */
130 1+IMM2_SIZE, 1+IMM2_SIZE, /* upto, minupto */
131 1+IMM2_SIZE, /* exact */
132 1, 1, 1, 1+IMM2_SIZE, /* *+, ++, ?+, upto+ */
133 1, 1, 1, 1, 1, 1, /* *I, *?I, +I, +?I, ?I, ??I */
134 1+IMM2_SIZE, 1+IMM2_SIZE, /* upto I, minupto I */
135 1+IMM2_SIZE, /* exact I */
136 1, 1, 1, 1+IMM2_SIZE, /* *+I, ++I, ?+I, upto+I */
137 /* Negative single-char repeats - only for chars < 256 */
138 1, 1, 1, 1, 1, 1, /* NOT *, *?, +, +?, ?, ?? */
139 1+IMM2_SIZE, 1+IMM2_SIZE, /* NOT upto, minupto */
140 1+IMM2_SIZE, /* NOT exact */
141 1, 1, 1, 1+IMM2_SIZE, /* NOT *+, ++, ?+, upto+ */
142 1, 1, 1, 1, 1, 1, /* NOT *I, *?I, +I, +?I, ?I, ??I */
143 1+IMM2_SIZE, 1+IMM2_SIZE, /* NOT upto I, minupto I */
144 1+IMM2_SIZE, /* NOT exact I */
145 1, 1, 1, 1+IMM2_SIZE, /* NOT *+I, ++I, ?+I, upto+I */
146 /* Positive type repeats */
147 1, 1, 1, 1, 1, 1, /* Type *, *?, +, +?, ?, ?? */
148 1+IMM2_SIZE, 1+IMM2_SIZE, /* Type upto, minupto */
149 1+IMM2_SIZE, /* Type exact */
150 1, 1, 1, 1+IMM2_SIZE, /* Type *+, ++, ?+, upto+ */
151 /* Character class & ref repeats */
152 0, 0, 0, 0, 0, 0, /* *, *?, +, +?, ?, ?? */
153 0, 0, /* CRRANGE, CRMINRANGE */
154 0, /* CLASS */
155 0, /* NCLASS */
156 0, /* XCLASS - variable length */
157 0, /* REF */
158 0, /* REFI */
159 0, /* RECURSE */
160 0, /* CALLOUT */
161 0, /* Alt */
162 0, /* Ket */
163 0, /* KetRmax */
164 0, /* KetRmin */
165 0, /* KetRpos */
166 0, /* Reverse */
167 0, /* Assert */
168 0, /* Assert not */
169 0, /* Assert behind */
170 0, /* Assert behind not */
171 0, 0, /* ONCE, ONCE_NC */
172 0, 0, 0, 0, 0, /* BRA, BRAPOS, CBRA, CBRAPOS, COND */
173 0, 0, 0, 0, 0, /* SBRA, SBRAPOS, SCBRA, SCBRAPOS, SCOND */
174 0, 0, /* CREF, NCREF */
175 0, 0, /* RREF, NRREF */
176 0, /* DEF */
177 0, 0, 0, /* BRAZERO, BRAMINZERO, BRAPOSZERO */
178 0, 0, 0, /* MARK, PRUNE, PRUNE_ARG */
179 0, 0, 0, 0, /* SKIP, SKIP_ARG, THEN, THEN_ARG */
180 0, 0, 0, 0, /* COMMIT, FAIL, ACCEPT, ASSERT_ACCEPT */
181 0, 0 /* CLOSE, SKIPZERO */
182 };
183
184 /* This table identifies those opcodes that inspect a character. It is used to
185 remember the fact that a character could have been inspected when the end of
186 the subject is reached. ***NOTE*** If the start of this table is modified, the
187 two tables that follow must also be modified. */
188
189 static const pcre_uint8 poptable[] = {
190 0, /* End */
191 0, 0, 0, 1, 1, /* \A, \G, \K, \B, \b */
192 1, 1, 1, 1, 1, 1, /* \D, \d, \S, \s, \W, \w */
193 1, 1, 1, /* Any, AllAny, Anybyte */
194 1, 1, /* \P, \p */
195 1, 1, 1, 1, 1, /* \R, \H, \h, \V, \v */
196 1, /* \X */
197 0, 0, 0, 0, 0, 0, /* \Z, \z, ^, ^M, $, $M */
198 1, /* Char */
199 1, /* Chari */
200 1, /* not */
201 1, /* noti */
202 /* Positive single-char repeats */
203 1, 1, 1, 1, 1, 1, /* *, *?, +, +?, ?, ?? */
204 1, 1, 1, /* upto, minupto, exact */
205 1, 1, 1, 1, /* *+, ++, ?+, upto+ */
206 1, 1, 1, 1, 1, 1, /* *I, *?I, +I, +?I, ?I, ??I */
207 1, 1, 1, /* upto I, minupto I, exact I */
208 1, 1, 1, 1, /* *+I, ++I, ?+I, upto+I */
209 /* Negative single-char repeats - only for chars < 256 */
210 1, 1, 1, 1, 1, 1, /* NOT *, *?, +, +?, ?, ?? */
211 1, 1, 1, /* NOT upto, minupto, exact */
212 1, 1, 1, 1, /* NOT *+, ++, ?+, upto+ */
213 1, 1, 1, 1, 1, 1, /* NOT *I, *?I, +I, +?I, ?I, ??I */
214 1, 1, 1, /* NOT upto I, minupto I, exact I */
215 1, 1, 1, 1, /* NOT *+I, ++I, ?+I, upto+I */
216 /* Positive type repeats */
217 1, 1, 1, 1, 1, 1, /* Type *, *?, +, +?, ?, ?? */
218 1, 1, 1, /* Type upto, minupto, exact */
219 1, 1, 1, 1, /* Type *+, ++, ?+, upto+ */
220 /* Character class & ref repeats */
221 1, 1, 1, 1, 1, 1, /* *, *?, +, +?, ?, ?? */
222 1, 1, /* CRRANGE, CRMINRANGE */
223 1, /* CLASS */
224 1, /* NCLASS */
225 1, /* XCLASS - variable length */
226 0, /* REF */
227 0, /* REFI */
228 0, /* RECURSE */
229 0, /* CALLOUT */
230 0, /* Alt */
231 0, /* Ket */
232 0, /* KetRmax */
233 0, /* KetRmin */
234 0, /* KetRpos */
235 0, /* Reverse */
236 0, /* Assert */
237 0, /* Assert not */
238 0, /* Assert behind */
239 0, /* Assert behind not */
240 0, 0, /* ONCE, ONCE_NC */
241 0, 0, 0, 0, 0, /* BRA, BRAPOS, CBRA, CBRAPOS, COND */
242 0, 0, 0, 0, 0, /* SBRA, SBRAPOS, SCBRA, SCBRAPOS, SCOND */
243 0, 0, /* CREF, NCREF */
244 0, 0, /* RREF, NRREF */
245 0, /* DEF */
246 0, 0, 0, /* BRAZERO, BRAMINZERO, BRAPOSZERO */
247 0, 0, 0, /* MARK, PRUNE, PRUNE_ARG */
248 0, 0, 0, 0, /* SKIP, SKIP_ARG, THEN, THEN_ARG */
249 0, 0, 0, 0, /* COMMIT, FAIL, ACCEPT, ASSERT_ACCEPT */
250 0, 0 /* CLOSE, SKIPZERO */
251 };
252
253 /* These 2 tables allow for compact code for testing for \D, \d, \S, \s, \W,
254 and \w */
255
256 static const pcre_uint8 toptable1[] = {
257 0, 0, 0, 0, 0, 0,
258 ctype_digit, ctype_digit,
259 ctype_space, ctype_space,
260 ctype_word, ctype_word,
261 0, 0 /* OP_ANY, OP_ALLANY */
262 };
263
264 static const pcre_uint8 toptable2[] = {
265 0, 0, 0, 0, 0, 0,
266 ctype_digit, 0,
267 ctype_space, 0,
268 ctype_word, 0,
269 1, 1 /* OP_ANY, OP_ALLANY */
270 };
271
272
273 /* Structure for holding data about a particular state, which is in effect the
274 current data for an active path through the match tree. It must consist
275 entirely of ints because the working vector we are passed, and which we put
276 these structures in, is a vector of ints. */
277
278 typedef struct stateblock {
279 int offset; /* Offset to opcode */
280 int count; /* Count for repeats */
281 int data; /* Some use extra data */
282 } stateblock;
283
284 #define INTS_PER_STATEBLOCK (int)(sizeof(stateblock)/sizeof(int))
285
286
287 #ifdef PCRE_DEBUG
288 /*************************************************
289 * Print character string *
290 *************************************************/
291
292 /* Character string printing function for debugging.
293
294 Arguments:
295 p points to string
296 length number of bytes
297 f where to print
298
299 Returns: nothing
300 */
301
302 static void
303 pchars(const pcre_uchar *p, int length, FILE *f)
304 {
305 int c;
306 while (length-- > 0)
307 {
308 if (isprint(c = *(p++)))
309 fprintf(f, "%c", c);
310 else
311 fprintf(f, "\\x%02x", c);
312 }
313 }
314 #endif
315
316
317
318 /*************************************************
319 * Execute a Regular Expression - DFA engine *
320 *************************************************/
321
322 /* This internal function applies a compiled pattern to a subject string,
323 starting at a given point, using a DFA engine. This function is called from the
324 external one, possibly multiple times if the pattern is not anchored. The
325 function calls itself recursively for some kinds of subpattern.
326
327 Arguments:
328 md the match_data block with fixed information
329 this_start_code the opening bracket of this subexpression's code
330 current_subject where we currently are in the subject string
331 start_offset start offset in the subject string
332 offsets vector to contain the matching string offsets
333 offsetcount size of same
334 workspace vector of workspace
335 wscount size of same
336 rlevel function call recursion level
337
338 Returns: > 0 => number of match offset pairs placed in offsets
339 = 0 => offsets overflowed; longest matches are present
340 -1 => failed to match
341 < -1 => some kind of unexpected problem
342
343 The following macros are used for adding states to the two state vectors (one
344 for the current character, one for the following character). */
345
346 #define ADD_ACTIVE(x,y) \
347 if (active_count++ < wscount) \
348 { \
349 next_active_state->offset = (x); \
350 next_active_state->count = (y); \
351 next_active_state++; \
352 DPRINTF(("%.*sADD_ACTIVE(%d,%d)\n", rlevel*2-2, SP, (x), (y))); \
353 } \
354 else return PCRE_ERROR_DFA_WSSIZE
355
356 #define ADD_ACTIVE_DATA(x,y,z) \
357 if (active_count++ < wscount) \
358 { \
359 next_active_state->offset = (x); \
360 next_active_state->count = (y); \
361 next_active_state->data = (z); \
362 next_active_state++; \
363 DPRINTF(("%.*sADD_ACTIVE_DATA(%d,%d,%d)\n", rlevel*2-2, SP, (x), (y), (z))); \
364 } \
365 else return PCRE_ERROR_DFA_WSSIZE
366
367 #define ADD_NEW(x,y) \
368 if (new_count++ < wscount) \
369 { \
370 next_new_state->offset = (x); \
371 next_new_state->count = (y); \
372 next_new_state++; \
373 DPRINTF(("%.*sADD_NEW(%d,%d)\n", rlevel*2-2, SP, (x), (y))); \
374 } \
375 else return PCRE_ERROR_DFA_WSSIZE
376
377 #define ADD_NEW_DATA(x,y,z) \
378 if (new_count++ < wscount) \
379 { \
380 next_new_state->offset = (x); \
381 next_new_state->count = (y); \
382 next_new_state->data = (z); \
383 next_new_state++; \
384 DPRINTF(("%.*sADD_NEW_DATA(%d,%d,%d) line %d\n", rlevel*2-2, SP, \
385 (x), (y), (z), __LINE__)); \
386 } \
387 else return PCRE_ERROR_DFA_WSSIZE
388
389 /* And now, here is the code */
390
391 static int
392 internal_dfa_exec(
393 dfa_match_data *md,
394 const pcre_uchar *this_start_code,
395 const pcre_uchar *current_subject,
396 int start_offset,
397 int *offsets,
398 int offsetcount,
399 int *workspace,
400 int wscount,
401 int rlevel)
402 {
403 stateblock *active_states, *new_states, *temp_states;
404 stateblock *next_active_state, *next_new_state;
405
406 const pcre_uint8 *ctypes, *lcc, *fcc;
407 const pcre_uchar *ptr;
408 const pcre_uchar *end_code, *first_op;
409
410 dfa_recursion_info new_recursive;
411
412 int active_count, new_count, match_count;
413
414 /* Some fields in the md block are frequently referenced, so we load them into
415 independent variables in the hope that this will perform better. */
416
417 const pcre_uchar *start_subject = md->start_subject;
418 const pcre_uchar *end_subject = md->end_subject;
419 const pcre_uchar *start_code = md->start_code;
420
421 #ifdef SUPPORT_UTF
422 BOOL utf = (md->poptions & PCRE_UTF8) != 0;
423 #else
424 BOOL utf = FALSE;
425 #endif
426
427 BOOL reset_could_continue = FALSE;
428
429 rlevel++;
430 offsetcount &= (-2);
431
432 wscount -= 2;
433 wscount = (wscount - (wscount % (INTS_PER_STATEBLOCK * 2))) /
434 (2 * INTS_PER_STATEBLOCK);
435
436 DPRINTF(("\n%.*s---------------------\n"
437 "%.*sCall to internal_dfa_exec f=%d\n",
438 rlevel*2-2, SP, rlevel*2-2, SP, rlevel));
439
440 ctypes = md->tables + ctypes_offset;
441 lcc = md->tables + lcc_offset;
442 fcc = md->tables + fcc_offset;
443
444 match_count = PCRE_ERROR_NOMATCH; /* A negative number */
445
446 active_states = (stateblock *)(workspace + 2);
447 next_new_state = new_states = active_states + wscount;
448 new_count = 0;
449
450 first_op = this_start_code + 1 + LINK_SIZE +
451 ((*this_start_code == OP_CBRA || *this_start_code == OP_SCBRA ||
452 *this_start_code == OP_CBRAPOS || *this_start_code == OP_SCBRAPOS)
453 ? IMM2_SIZE:0);
454
455 /* The first thing in any (sub) pattern is a bracket of some sort. Push all
456 the alternative states onto the list, and find out where the end is. This
457 makes is possible to use this function recursively, when we want to stop at a
458 matching internal ket rather than at the end.
459
460 If the first opcode in the first alternative is OP_REVERSE, we are dealing with
461 a backward assertion. In that case, we have to find out the maximum amount to
462 move back, and set up each alternative appropriately. */
463
464 if (*first_op == OP_REVERSE)
465 {
466 int max_back = 0;
467 int gone_back;
468
469 end_code = this_start_code;
470 do
471 {
472 int back = GET(end_code, 2+LINK_SIZE);
473 if (back > max_back) max_back = back;
474 end_code += GET(end_code, 1);
475 }
476 while (*end_code == OP_ALT);
477
478 /* If we can't go back the amount required for the longest lookbehind
479 pattern, go back as far as we can; some alternatives may still be viable. */
480
481 #ifdef SUPPORT_UTF
482 /* In character mode we have to step back character by character */
483
484 if (utf)
485 {
486 for (gone_back = 0; gone_back < max_back; gone_back++)
487 {
488 if (current_subject <= start_subject) break;
489 current_subject--;
490 ACROSSCHAR(current_subject > start_subject, *current_subject, current_subject--);
491 }
492 }
493 else
494 #endif
495
496 /* In byte-mode we can do this quickly. */
497
498 {
499 gone_back = (current_subject - max_back < start_subject)?
500 (int)(current_subject - start_subject) : max_back;
501 current_subject -= gone_back;
502 }
503
504 /* Save the earliest consulted character */
505
506 if (current_subject < md->start_used_ptr)
507 md->start_used_ptr = current_subject;
508
509 /* Now we can process the individual branches. */
510
511 end_code = this_start_code;
512 do
513 {
514 int back = GET(end_code, 2+LINK_SIZE);
515 if (back <= gone_back)
516 {
517 int bstate = (int)(end_code - start_code + 2 + 2*LINK_SIZE);
518 ADD_NEW_DATA(-bstate, 0, gone_back - back);
519 }
520 end_code += GET(end_code, 1);
521 }
522 while (*end_code == OP_ALT);
523 }
524
525 /* This is the code for a "normal" subpattern (not a backward assertion). The
526 start of a whole pattern is always one of these. If we are at the top level,
527 we may be asked to restart matching from the same point that we reached for a
528 previous partial match. We still have to scan through the top-level branches to
529 find the end state. */
530
531 else
532 {
533 end_code = this_start_code;
534
535 /* Restarting */
536
537 if (rlevel == 1 && (md->moptions & PCRE_DFA_RESTART) != 0)
538 {
539 do { end_code += GET(end_code, 1); } while (*end_code == OP_ALT);
540 new_count = workspace[1];
541 if (!workspace[0])
542 memcpy(new_states, active_states, new_count * sizeof(stateblock));
543 }
544
545 /* Not restarting */
546
547 else
548 {
549 int length = 1 + LINK_SIZE +
550 ((*this_start_code == OP_CBRA || *this_start_code == OP_SCBRA ||
551 *this_start_code == OP_CBRAPOS || *this_start_code == OP_SCBRAPOS)
552 ? IMM2_SIZE:0);
553 do
554 {
555 ADD_NEW((int)(end_code - start_code + length), 0);
556 end_code += GET(end_code, 1);
557 length = 1 + LINK_SIZE;
558 }
559 while (*end_code == OP_ALT);
560 }
561 }
562
563 workspace[0] = 0; /* Bit indicating which vector is current */
564
565 DPRINTF(("%.*sEnd state = %d\n", rlevel*2-2, SP, (int)(end_code - start_code)));
566
567 /* Loop for scanning the subject */
568
569 ptr = current_subject;
570 for (;;)
571 {
572 int i, j;
573 int clen, dlen;
574 unsigned int c, d;
575 int forced_fail = 0;
576 BOOL partial_newline = FALSE;
577 BOOL could_continue = reset_could_continue;
578 reset_could_continue = FALSE;
579
580 /* Make the new state list into the active state list and empty the
581 new state list. */
582
583 temp_states = active_states;
584 active_states = new_states;
585 new_states = temp_states;
586 active_count = new_count;
587 new_count = 0;
588
589 workspace[0] ^= 1; /* Remember for the restarting feature */
590 workspace[1] = active_count;
591
592 #ifdef PCRE_DEBUG
593 printf("%.*sNext character: rest of subject = \"", rlevel*2-2, SP);
594 pchars(ptr, STRLEN_UC(ptr), stdout);
595 printf("\"\n");
596
597 printf("%.*sActive states: ", rlevel*2-2, SP);
598 for (i = 0; i < active_count; i++)
599 printf("%d/%d ", active_states[i].offset, active_states[i].count);
600 printf("\n");
601 #endif
602
603 /* Set the pointers for adding new states */
604
605 next_active_state = active_states + active_count;
606 next_new_state = new_states;
607
608 /* Load the current character from the subject outside the loop, as many
609 different states may want to look at it, and we assume that at least one
610 will. */
611
612 if (ptr < end_subject)
613 {
614 clen = 1; /* Number of data items in the character */
615 #ifdef SUPPORT_UTF
616 if (utf) { GETCHARLEN(c, ptr, clen); } else
617 #endif /* SUPPORT_UTF */
618 c = *ptr;
619 }
620 else
621 {
622 clen = 0; /* This indicates the end of the subject */
623 c = NOTACHAR; /* This value should never actually be used */
624 }
625
626 /* Scan up the active states and act on each one. The result of an action
627 may be to add more states to the currently active list (e.g. on hitting a
628 parenthesis) or it may be to put states on the new list, for considering
629 when we move the character pointer on. */
630
631 for (i = 0; i < active_count; i++)
632 {
633 stateblock *current_state = active_states + i;
634 BOOL caseless = FALSE;
635 const pcre_uchar *code;
636 int state_offset = current_state->offset;
637 int count, codevalue, rrc;
638
639 #ifdef PCRE_DEBUG
640 printf ("%.*sProcessing state %d c=", rlevel*2-2, SP, state_offset);
641 if (clen == 0) printf("EOL\n");
642 else if (c > 32 && c < 127) printf("'%c'\n", c);
643 else printf("0x%02x\n", c);
644 #endif
645
646 /* A negative offset is a special case meaning "hold off going to this
647 (negated) state until the number of characters in the data field have
648 been skipped". If the could_continue flag was passed over from a previous
649 state, arrange for it to passed on. */
650
651 if (state_offset < 0)
652 {
653 if (current_state->data > 0)
654 {
655 DPRINTF(("%.*sSkipping this character\n", rlevel*2-2, SP));
656 ADD_NEW_DATA(state_offset, current_state->count,
657 current_state->data - 1);
658 if (could_continue) reset_could_continue = TRUE;
659 continue;
660 }
661 else
662 {
663 current_state->offset = state_offset = -state_offset;
664 }
665 }
666
667 /* Check for a duplicate state with the same count, and skip if found.
668 See the note at the head of this module about the possibility of improving
669 performance here. */
670
671 for (j = 0; j < i; j++)
672 {
673 if (active_states[j].offset == state_offset &&
674 active_states[j].count == current_state->count)
675 {
676 DPRINTF(("%.*sDuplicate state: skipped\n", rlevel*2-2, SP));
677 goto NEXT_ACTIVE_STATE;
678 }
679 }
680
681 /* The state offset is the offset to the opcode */
682
683 code = start_code + state_offset;
684 codevalue = *code;
685
686 /* If this opcode inspects a character, but we are at the end of the
687 subject, remember the fact for use when testing for a partial match. */
688
689 if (clen == 0 && poptable[codevalue] != 0)
690 could_continue = TRUE;
691
692 /* If this opcode is followed by an inline character, load it. It is
693 tempting to test for the presence of a subject character here, but that
694 is wrong, because sometimes zero repetitions of the subject are
695 permitted.
696
697 We also use this mechanism for opcodes such as OP_TYPEPLUS that take an
698 argument that is not a data character - but is always one byte long because
699 the values are small. We have to take special action to deal with \P, \p,
700 \H, \h, \V, \v and \X in this case. To keep the other cases fast, convert
701 these ones to new opcodes. */
702
703 if (coptable[codevalue] > 0)
704 {
705 dlen = 1;
706 #ifdef SUPPORT_UTF
707 if (utf) { GETCHARLEN(d, (code + coptable[codevalue]), dlen); } else
708 #endif /* SUPPORT_UTF */
709 d = code[coptable[codevalue]];
710 if (codevalue >= OP_TYPESTAR)
711 {
712 switch(d)
713 {
714 case OP_ANYBYTE: return PCRE_ERROR_DFA_UITEM;
715 case OP_NOTPROP:
716 case OP_PROP: codevalue += OP_PROP_EXTRA; break;
717 case OP_ANYNL: codevalue += OP_ANYNL_EXTRA; break;
718 case OP_EXTUNI: codevalue += OP_EXTUNI_EXTRA; break;
719 case OP_NOT_HSPACE:
720 case OP_HSPACE: codevalue += OP_HSPACE_EXTRA; break;
721 case OP_NOT_VSPACE:
722 case OP_VSPACE: codevalue += OP_VSPACE_EXTRA; break;
723 default: break;
724 }
725 }
726 }
727 else
728 {
729 dlen = 0; /* Not strictly necessary, but compilers moan */
730 d = NOTACHAR; /* if these variables are not set. */
731 }
732
733
734 /* Now process the individual opcodes */
735
736 switch (codevalue)
737 {
738 /* ========================================================================== */
739 /* These cases are never obeyed. This is a fudge that causes a compile-
740 time error if the vectors coptable or poptable, which are indexed by
741 opcode, are not the correct length. It seems to be the only way to do
742 such a check at compile time, as the sizeof() operator does not work
743 in the C preprocessor. */
744
745 case OP_TABLE_LENGTH:
746 case OP_TABLE_LENGTH +
747 ((sizeof(coptable) == OP_TABLE_LENGTH) &&
748 (sizeof(poptable) == OP_TABLE_LENGTH)):
749 break;
750
751 /* ========================================================================== */
752 /* Reached a closing bracket. If not at the end of the pattern, carry
753 on with the next opcode. For repeating opcodes, also add the repeat
754 state. Note that KETRPOS will always be encountered at the end of the
755 subpattern, because the possessive subpattern repeats are always handled
756 using recursive calls. Thus, it never adds any new states.
757
758 At the end of the (sub)pattern, unless we have an empty string and
759 PCRE_NOTEMPTY is set, or PCRE_NOTEMPTY_ATSTART is set and we are at the
760 start of the subject, save the match data, shifting up all previous
761 matches so we always have the longest first. */
762
763 case OP_KET:
764 case OP_KETRMIN:
765 case OP_KETRMAX:
766 case OP_KETRPOS:
767 if (code != end_code)
768 {
769 ADD_ACTIVE(state_offset + 1 + LINK_SIZE, 0);
770 if (codevalue != OP_KET)
771 {
772 ADD_ACTIVE(state_offset - GET(code, 1), 0);
773 }
774 }
775 else
776 {
777 if (ptr > current_subject ||
778 ((md->moptions & PCRE_NOTEMPTY) == 0 &&
779 ((md->moptions & PCRE_NOTEMPTY_ATSTART) == 0 ||
780 current_subject > start_subject + md->start_offset)))
781 {
782 if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0;
783 else if (match_count > 0 && ++match_count * 2 > offsetcount)
784 match_count = 0;
785 count = ((match_count == 0)? offsetcount : match_count * 2) - 2;
786 if (count > 0) memmove(offsets + 2, offsets, count * sizeof(int));
787 if (offsetcount >= 2)
788 {
789 offsets[0] = (int)(current_subject - start_subject);
790 offsets[1] = (int)(ptr - start_subject);
791 DPRINTF(("%.*sSet matched string = \"%.*s\"\n", rlevel*2-2, SP,
792 offsets[1] - offsets[0], (char *)current_subject));
793 }
794 if ((md->moptions & PCRE_DFA_SHORTEST) != 0)
795 {
796 DPRINTF(("%.*sEnd of internal_dfa_exec %d: returning %d\n"
797 "%.*s---------------------\n\n", rlevel*2-2, SP, rlevel,
798 match_count, rlevel*2-2, SP));
799 return match_count;
800 }
801 }
802 }
803 break;
804
805 /* ========================================================================== */
806 /* These opcodes add to the current list of states without looking
807 at the current character. */
808
809 /*-----------------------------------------------------------------*/
810 case OP_ALT:
811 do { code += GET(code, 1); } while (*code == OP_ALT);
812 ADD_ACTIVE((int)(code - start_code), 0);
813 break;
814
815 /*-----------------------------------------------------------------*/
816 case OP_BRA:
817 case OP_SBRA:
818 do
819 {
820 ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0);
821 code += GET(code, 1);
822 }
823 while (*code == OP_ALT);
824 break;
825
826 /*-----------------------------------------------------------------*/
827 case OP_CBRA:
828 case OP_SCBRA:
829 ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE + IMM2_SIZE), 0);
830 code += GET(code, 1);
831 while (*code == OP_ALT)
832 {
833 ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0);
834 code += GET(code, 1);
835 }
836 break;
837
838 /*-----------------------------------------------------------------*/
839 case OP_BRAZERO:
840 case OP_BRAMINZERO:
841 ADD_ACTIVE(state_offset + 1, 0);
842 code += 1 + GET(code, 2);
843 while (*code == OP_ALT) code += GET(code, 1);
844 ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0);
845 break;
846
847 /*-----------------------------------------------------------------*/
848 case OP_SKIPZERO:
849 code += 1 + GET(code, 2);
850 while (*code == OP_ALT) code += GET(code, 1);
851 ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0);
852 break;
853
854 /*-----------------------------------------------------------------*/
855 case OP_CIRC:
856 if (ptr == start_subject && (md->moptions & PCRE_NOTBOL) == 0)
857 { ADD_ACTIVE(state_offset + 1, 0); }
858 break;
859
860 /*-----------------------------------------------------------------*/
861 case OP_CIRCM:
862 if ((ptr == start_subject && (md->moptions & PCRE_NOTBOL) == 0) ||
863 (ptr != end_subject && WAS_NEWLINE(ptr)))
864 { ADD_ACTIVE(state_offset + 1, 0); }
865 break;
866
867 /*-----------------------------------------------------------------*/
868 case OP_EOD:
869 if (ptr >= end_subject)
870 {
871 if ((md->moptions & PCRE_PARTIAL_HARD) != 0)
872 could_continue = TRUE;
873 else { ADD_ACTIVE(state_offset + 1, 0); }
874 }
875 break;
876
877 /*-----------------------------------------------------------------*/
878 case OP_SOD:
879 if (ptr == start_subject) { ADD_ACTIVE(state_offset + 1, 0); }
880 break;
881
882 /*-----------------------------------------------------------------*/
883 case OP_SOM:
884 if (ptr == start_subject + start_offset) { ADD_ACTIVE(state_offset + 1, 0); }
885 break;
886
887
888 /* ========================================================================== */
889 /* These opcodes inspect the next subject character, and sometimes
890 the previous one as well, but do not have an argument. The variable
891 clen contains the length of the current character and is zero if we are
892 at the end of the subject. */
893
894 /*-----------------------------------------------------------------*/
895 case OP_ANY:
896 if (clen > 0 && !IS_NEWLINE(ptr))
897 {
898 if (ptr + 1 >= md->end_subject &&
899 (md->moptions & (PCRE_PARTIAL_HARD)) != 0 &&
900 NLBLOCK->nltype == NLTYPE_FIXED &&
901 NLBLOCK->nllen == 2 &&
902 c == NLBLOCK->nl[0])
903 {
904 could_continue = partial_newline = TRUE;
905 }
906 else
907 {
908 ADD_NEW(state_offset + 1, 0);
909 }
910 }
911 break;
912
913 /*-----------------------------------------------------------------*/
914 case OP_ALLANY:
915 if (clen > 0)
916 { ADD_NEW(state_offset + 1, 0); }
917 break;
918
919 /*-----------------------------------------------------------------*/
920 case OP_EODN:
921 if (clen == 0 && (md->moptions & PCRE_PARTIAL_HARD) != 0)
922 could_continue = TRUE;
923 else if (clen == 0 || (IS_NEWLINE(ptr) && ptr == end_subject - md->nllen))
924 { ADD_ACTIVE(state_offset + 1, 0); }
925 break;
926
927 /*-----------------------------------------------------------------*/
928 case OP_DOLL:
929 if ((md->moptions & PCRE_NOTEOL) == 0)
930 {
931 if (clen == 0 && (md->moptions & PCRE_PARTIAL_HARD) != 0)
932 could_continue = TRUE;
933 else if (clen == 0 ||
934 ((md->poptions & PCRE_DOLLAR_ENDONLY) == 0 && IS_NEWLINE(ptr) &&
935 (ptr == end_subject - md->nllen)
936 ))
937 { ADD_ACTIVE(state_offset + 1, 0); }
938 else if (ptr + 1 >= md->end_subject &&
939 (md->moptions & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) != 0 &&
940 NLBLOCK->nltype == NLTYPE_FIXED &&
941 NLBLOCK->nllen == 2 &&
942 c == NLBLOCK->nl[0])
943 {
944 if ((md->moptions & PCRE_PARTIAL_HARD) != 0)
945 {
946 reset_could_continue = TRUE;
947 ADD_NEW_DATA(-(state_offset + 1), 0, 1);
948 }
949 else could_continue = partial_newline = TRUE;
950 }
951 }
952 break;
953
954 /*-----------------------------------------------------------------*/
955 case OP_DOLLM:
956 if ((md->moptions & PCRE_NOTEOL) == 0)
957 {
958 if (clen == 0 && (md->moptions & PCRE_PARTIAL_HARD) != 0)
959 could_continue = TRUE;
960 else if (clen == 0 ||
961 ((md->poptions & PCRE_DOLLAR_ENDONLY) == 0 && IS_NEWLINE(ptr)))
962 { ADD_ACTIVE(state_offset + 1, 0); }
963 else if (ptr + 1 >= md->end_subject &&
964 (md->moptions & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) != 0 &&
965 NLBLOCK->nltype == NLTYPE_FIXED &&
966 NLBLOCK->nllen == 2 &&
967 c == NLBLOCK->nl[0])
968 {
969 if ((md->moptions & PCRE_PARTIAL_HARD) != 0)
970 {
971 reset_could_continue = TRUE;
972 ADD_NEW_DATA(-(state_offset + 1), 0, 1);
973 }
974 else could_continue = partial_newline = TRUE;
975 }
976 }
977 else if (IS_NEWLINE(ptr))
978 { ADD_ACTIVE(state_offset + 1, 0); }
979 break;
980
981 /*-----------------------------------------------------------------*/
982
983 case OP_DIGIT:
984 case OP_WHITESPACE:
985 case OP_WORDCHAR:
986 if (clen > 0 && c < 256 &&
987 ((ctypes[c] & toptable1[codevalue]) ^ toptable2[codevalue]) != 0)
988 { ADD_NEW(state_offset + 1, 0); }
989 break;
990
991 /*-----------------------------------------------------------------*/
992 case OP_NOT_DIGIT:
993 case OP_NOT_WHITESPACE:
994 case OP_NOT_WORDCHAR:
995 if (clen > 0 && (c >= 256 ||
996 ((ctypes[c] & toptable1[codevalue]) ^ toptable2[codevalue]) != 0))
997 { ADD_NEW(state_offset + 1, 0); }
998 break;
999
1000 /*-----------------------------------------------------------------*/
1001 case OP_WORD_BOUNDARY:
1002 case OP_NOT_WORD_BOUNDARY:
1003 {
1004 int left_word, right_word;
1005
1006 if (ptr > start_subject)
1007 {
1008 const pcre_uchar *temp = ptr - 1;
1009 if (temp < md->start_used_ptr) md->start_used_ptr = temp;
1010 #ifdef SUPPORT_UTF
1011 if (utf) { BACKCHAR(temp); }
1012 #endif
1013 GETCHARTEST(d, temp);
1014 #ifdef SUPPORT_UCP
1015 if ((md->poptions & PCRE_UCP) != 0)
1016 {
1017 if (d == '_') left_word = TRUE; else
1018 {
1019 int cat = UCD_CATEGORY(d);
1020 left_word = (cat == ucp_L || cat == ucp_N);
1021 }
1022 }
1023 else
1024 #endif
1025 left_word = d < 256 && (ctypes[d] & ctype_word) != 0;
1026 }
1027 else left_word = FALSE;
1028
1029 if (clen > 0)
1030 {
1031 #ifdef SUPPORT_UCP
1032 if ((md->poptions & PCRE_UCP) != 0)
1033 {
1034 if (c == '_') right_word = TRUE; else
1035 {
1036 int cat = UCD_CATEGORY(c);
1037 right_word = (cat == ucp_L || cat == ucp_N);
1038 }
1039 }
1040 else
1041 #endif
1042 right_word = c < 256 && (ctypes[c] & ctype_word) != 0;
1043 }
1044 else right_word = FALSE;
1045
1046 if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY))
1047 { ADD_ACTIVE(state_offset + 1, 0); }
1048 }
1049 break;
1050
1051
1052 /*-----------------------------------------------------------------*/
1053 /* Check the next character by Unicode property. We will get here only
1054 if the support is in the binary; otherwise a compile-time error occurs.
1055 */
1056
1057 #ifdef SUPPORT_UCP
1058 case OP_PROP:
1059 case OP_NOTPROP:
1060 if (clen > 0)
1061 {
1062 BOOL OK;
1063 const pcre_uint8 chartype = UCD_CHARTYPE(c);
1064 switch(code[1])
1065 {
1066 case PT_ANY:
1067 OK = TRUE;
1068 break;
1069
1070 case PT_LAMP:
1071 OK = chartype == ucp_Lu || chartype == ucp_Ll ||
1072 chartype == ucp_Lt;
1073 break;
1074
1075 case PT_GC:
1076 OK = PRIV(ucp_gentype)[chartype] == code[2];
1077 break;
1078
1079 case PT_PC:
1080 OK = chartype == code[2];
1081 break;
1082
1083 case PT_SC:
1084 OK = UCD_SCRIPT(c) == code[2];
1085 break;
1086
1087 /* These are specials for combination cases. */
1088
1089 case PT_ALNUM:
1090 OK = PRIV(ucp_gentype)[chartype] == ucp_L ||
1091 PRIV(ucp_gentype)[chartype] == ucp_N;
1092 break;
1093
1094 case PT_SPACE: /* Perl space */
1095 OK = PRIV(ucp_gentype)[chartype] == ucp_Z ||
1096 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1097 break;
1098
1099 case PT_PXSPACE: /* POSIX space */
1100 OK = PRIV(ucp_gentype)[chartype] == ucp_Z ||
1101 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1102 c == CHAR_FF || c == CHAR_CR;
1103 break;
1104
1105 case PT_WORD:
1106 OK = PRIV(ucp_gentype)[chartype] == ucp_L ||
1107 PRIV(ucp_gentype)[chartype] == ucp_N ||
1108 c == CHAR_UNDERSCORE;
1109 break;
1110
1111 /* Should never occur, but keep compilers from grumbling. */
1112
1113 default:
1114 OK = codevalue != OP_PROP;
1115 break;
1116 }
1117
1118 if (OK == (codevalue == OP_PROP)) { ADD_NEW(state_offset + 3, 0); }
1119 }
1120 break;
1121 #endif
1122
1123
1124
1125 /* ========================================================================== */
1126 /* These opcodes likewise inspect the subject character, but have an
1127 argument that is not a data character. It is one of these opcodes:
1128 OP_ANY, OP_ALLANY, OP_DIGIT, OP_NOT_DIGIT, OP_WHITESPACE, OP_NOT_SPACE,
1129 OP_WORDCHAR, OP_NOT_WORDCHAR. The value is loaded into d. */
1130
1131 case OP_TYPEPLUS:
1132 case OP_TYPEMINPLUS:
1133 case OP_TYPEPOSPLUS:
1134 count = current_state->count; /* Already matched */
1135 if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); }
1136 if (clen > 0)
1137 {
1138 if (d == OP_ANY && ptr + 1 >= md->end_subject &&
1139 (md->moptions & (PCRE_PARTIAL_HARD)) != 0 &&
1140 NLBLOCK->nltype == NLTYPE_FIXED &&
1141 NLBLOCK->nllen == 2 &&
1142 c == NLBLOCK->nl[0])
1143 {
1144 could_continue = partial_newline = TRUE;
1145 }
1146 else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) ||
1147 (c < 256 &&
1148 (d != OP_ANY || !IS_NEWLINE(ptr)) &&
1149 ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0))
1150 {
1151 if (count > 0 && codevalue == OP_TYPEPOSPLUS)
1152 {
1153 active_count--; /* Remove non-match possibility */
1154 next_active_state--;
1155 }
1156 count++;
1157 ADD_NEW(state_offset, count);
1158 }
1159 }
1160 break;
1161
1162 /*-----------------------------------------------------------------*/
1163 case OP_TYPEQUERY:
1164 case OP_TYPEMINQUERY:
1165 case OP_TYPEPOSQUERY:
1166 ADD_ACTIVE(state_offset + 2, 0);
1167 if (clen > 0)
1168 {
1169 if (d == OP_ANY && ptr + 1 >= md->end_subject &&
1170 (md->moptions & (PCRE_PARTIAL_HARD)) != 0 &&
1171 NLBLOCK->nltype == NLTYPE_FIXED &&
1172 NLBLOCK->nllen == 2 &&
1173 c == NLBLOCK->nl[0])
1174 {
1175 could_continue = partial_newline = TRUE;
1176 }
1177 else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) ||
1178 (c < 256 &&
1179 (d != OP_ANY || !IS_NEWLINE(ptr)) &&
1180 ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0))
1181 {
1182 if (codevalue == OP_TYPEPOSQUERY)
1183 {
1184 active_count--; /* Remove non-match possibility */
1185 next_active_state--;
1186 }
1187 ADD_NEW(state_offset + 2, 0);
1188 }
1189 }
1190 break;
1191
1192 /*-----------------------------------------------------------------*/
1193 case OP_TYPESTAR:
1194 case OP_TYPEMINSTAR:
1195 case OP_TYPEPOSSTAR:
1196 ADD_ACTIVE(state_offset + 2, 0);
1197 if (clen > 0)
1198 {
1199 if (d == OP_ANY && ptr + 1 >= md->end_subject &&
1200 (md->moptions & (PCRE_PARTIAL_HARD)) != 0 &&
1201 NLBLOCK->nltype == NLTYPE_FIXED &&
1202 NLBLOCK->nllen == 2 &&
1203 c == NLBLOCK->nl[0])
1204 {
1205 could_continue = partial_newline = TRUE;
1206 }
1207 else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) ||
1208 (c < 256 &&
1209 (d != OP_ANY || !IS_NEWLINE(ptr)) &&
1210 ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0))
1211 {
1212 if (codevalue == OP_TYPEPOSSTAR)
1213 {
1214 active_count--; /* Remove non-match possibility */
1215 next_active_state--;
1216 }
1217 ADD_NEW(state_offset, 0);
1218 }
1219 }
1220 break;
1221
1222 /*-----------------------------------------------------------------*/
1223 case OP_TYPEEXACT:
1224 count = current_state->count; /* Number already matched */
1225 if (clen > 0)
1226 {
1227 if (d == OP_ANY && ptr + 1 >= md->end_subject &&
1228 (md->moptions & (PCRE_PARTIAL_HARD)) != 0 &&
1229 NLBLOCK->nltype == NLTYPE_FIXED &&
1230 NLBLOCK->nllen == 2 &&
1231 c == NLBLOCK->nl[0])
1232 {
1233 could_continue = partial_newline = TRUE;
1234 }
1235 else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) ||
1236 (c < 256 &&
1237 (d != OP_ANY || !IS_NEWLINE(ptr)) &&
1238 ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0))
1239 {
1240 if (++count >= GET2(code, 1))
1241 { ADD_NEW(state_offset + 1 + IMM2_SIZE + 1, 0); }
1242 else
1243 { ADD_NEW(state_offset, count); }
1244 }
1245 }
1246 break;
1247
1248 /*-----------------------------------------------------------------*/
1249 case OP_TYPEUPTO:
1250 case OP_TYPEMINUPTO:
1251 case OP_TYPEPOSUPTO:
1252 ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0);
1253 count = current_state->count; /* Number already matched */
1254 if (clen > 0)
1255 {
1256 if (d == OP_ANY && ptr + 1 >= md->end_subject &&
1257 (md->moptions & (PCRE_PARTIAL_HARD)) != 0 &&
1258 NLBLOCK->nltype == NLTYPE_FIXED &&
1259 NLBLOCK->nllen == 2 &&
1260 c == NLBLOCK->nl[0])
1261 {
1262 could_continue = partial_newline = TRUE;
1263 }
1264 else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) ||
1265 (c < 256 &&
1266 (d != OP_ANY || !IS_NEWLINE(ptr)) &&
1267 ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0))
1268 {
1269 if (codevalue == OP_TYPEPOSUPTO)
1270 {
1271 active_count--; /* Remove non-match possibility */
1272 next_active_state--;
1273 }
1274 if (++count >= GET2(code, 1))
1275 { ADD_NEW(state_offset + 2 + IMM2_SIZE, 0); }
1276 else
1277 { ADD_NEW(state_offset, count); }
1278 }
1279 }
1280 break;
1281
1282 /* ========================================================================== */
1283 /* These are virtual opcodes that are used when something like
1284 OP_TYPEPLUS has OP_PROP, OP_NOTPROP, OP_ANYNL, or OP_EXTUNI as its
1285 argument. It keeps the code above fast for the other cases. The argument
1286 is in the d variable. */
1287
1288 #ifdef SUPPORT_UCP
1289 case OP_PROP_EXTRA + OP_TYPEPLUS:
1290 case OP_PROP_EXTRA + OP_TYPEMINPLUS:
1291 case OP_PROP_EXTRA + OP_TYPEPOSPLUS:
1292 count = current_state->count; /* Already matched */
1293 if (count > 0) { ADD_ACTIVE(state_offset + 4, 0); }
1294 if (clen > 0)
1295 {
1296 BOOL OK;
1297 const pcre_uint8 chartype = UCD_CHARTYPE(c);
1298 switch(code[2])
1299 {
1300 case PT_ANY:
1301 OK = TRUE;
1302 break;
1303
1304 case PT_LAMP:
1305 OK = chartype == ucp_Lu || chartype == ucp_Ll ||
1306 chartype == ucp_Lt;
1307 break;
1308
1309 case PT_GC:
1310 OK = PRIV(ucp_gentype)[chartype] == code[3];
1311 break;
1312
1313 case PT_PC:
1314 OK = chartype == code[3];
1315 break;
1316
1317 case PT_SC:
1318 OK = UCD_SCRIPT(c) == code[3];
1319 break;
1320
1321 /* These are specials for combination cases. */
1322
1323 case PT_ALNUM:
1324 OK = PRIV(ucp_gentype)[chartype] == ucp_L ||
1325 PRIV(ucp_gentype)[chartype] == ucp_N;
1326 break;
1327
1328 case PT_SPACE: /* Perl space */
1329 OK = PRIV(ucp_gentype)[chartype] == ucp_Z ||
1330 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1331 break;
1332
1333 case PT_PXSPACE: /* POSIX space */
1334 OK = PRIV(ucp_gentype)[chartype] == ucp_Z ||
1335 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1336 c == CHAR_FF || c == CHAR_CR;
1337 break;
1338
1339 case PT_WORD:
1340 OK = PRIV(ucp_gentype)[chartype] == ucp_L ||
1341 PRIV(ucp_gentype)[chartype] == ucp_N ||
1342 c == CHAR_UNDERSCORE;
1343 break;
1344
1345 /* Should never occur, but keep compilers from grumbling. */
1346
1347 default:
1348 OK = codevalue != OP_PROP;
1349 break;
1350 }
1351
1352 if (OK == (d == OP_PROP))
1353 {
1354 if (count > 0 && codevalue == OP_PROP_EXTRA + OP_TYPEPOSPLUS)
1355 {
1356 active_count--; /* Remove non-match possibility */
1357 next_active_state--;
1358 }
1359 count++;
1360 ADD_NEW(state_offset, count);
1361 }
1362 }
1363 break;
1364
1365 /*-----------------------------------------------------------------*/
1366 case OP_EXTUNI_EXTRA + OP_TYPEPLUS:
1367 case OP_EXTUNI_EXTRA + OP_TYPEMINPLUS:
1368 case OP_EXTUNI_EXTRA + OP_TYPEPOSPLUS:
1369 count = current_state->count; /* Already matched */
1370 if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); }
1371 if (clen > 0 && UCD_CATEGORY(c) != ucp_M)
1372 {
1373 const pcre_uchar *nptr = ptr + clen;
1374 int ncount = 0;
1375 if (count > 0 && codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSPLUS)
1376 {
1377 active_count--; /* Remove non-match possibility */
1378 next_active_state--;
1379 }
1380 while (nptr < end_subject)
1381 {
1382 int nd;
1383 int ndlen = 1;
1384 GETCHARLEN(nd, nptr, ndlen);
1385 if (UCD_CATEGORY(nd) != ucp_M) break;
1386 ncount++;
1387 nptr += ndlen;
1388 }
1389 count++;
1390 ADD_NEW_DATA(-state_offset, count, ncount);
1391 }
1392 break;
1393 #endif
1394
1395 /*-----------------------------------------------------------------*/
1396 case OP_ANYNL_EXTRA + OP_TYPEPLUS:
1397 case OP_ANYNL_EXTRA + OP_TYPEMINPLUS:
1398 case OP_ANYNL_EXTRA + OP_TYPEPOSPLUS:
1399 count = current_state->count; /* Already matched */
1400 if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); }
1401 if (clen > 0)
1402 {
1403 int ncount = 0;
1404 switch (c)
1405 {
1406 case 0x000b:
1407 case 0x000c:
1408 case 0x0085:
1409 case 0x2028:
1410 case 0x2029:
1411 if ((md->moptions & PCRE_BSR_ANYCRLF) != 0) break;
1412 goto ANYNL01;
1413
1414 case 0x000d:
1415 if (ptr + 1 < end_subject && ptr[1] == 0x0a) ncount = 1;
1416 /* Fall through */
1417
1418 ANYNL01:
1419 case 0x000a:
1420 if (count > 0 && codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSPLUS)
1421 {
1422 active_count--; /* Remove non-match possibility */
1423 next_active_state--;
1424 }
1425 count++;
1426 ADD_NEW_DATA(-state_offset, count, ncount);
1427 break;
1428
1429 default:
1430 break;
1431 }
1432 }
1433 break;
1434
1435 /*-----------------------------------------------------------------*/
1436 case OP_VSPACE_EXTRA + OP_TYPEPLUS:
1437 case OP_VSPACE_EXTRA + OP_TYPEMINPLUS:
1438 case OP_VSPACE_EXTRA + OP_TYPEPOSPLUS:
1439 count = current_state->count; /* Already matched */
1440 if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); }
1441 if (clen > 0)
1442 {
1443 BOOL OK;
1444 switch (c)
1445 {
1446 case 0x000a:
1447 case 0x000b:
1448 case 0x000c:
1449 case 0x000d:
1450 case 0x0085:
1451 case 0x2028:
1452 case 0x2029:
1453 OK = TRUE;
1454 break;
1455
1456 default:
1457 OK = FALSE;
1458 break;
1459 }
1460
1461 if (OK == (d == OP_VSPACE))
1462 {
1463 if (count > 0 && codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSPLUS)
1464 {
1465 active_count--; /* Remove non-match possibility */
1466 next_active_state--;
1467 }
1468 count++;
1469 ADD_NEW_DATA(-state_offset, count, 0);
1470 }
1471 }
1472 break;
1473
1474 /*-----------------------------------------------------------------*/
1475 case OP_HSPACE_EXTRA + OP_TYPEPLUS:
1476 case OP_HSPACE_EXTRA + OP_TYPEMINPLUS:
1477 case OP_HSPACE_EXTRA + OP_TYPEPOSPLUS:
1478 count = current_state->count; /* Already matched */
1479 if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); }
1480 if (clen > 0)
1481 {
1482 BOOL OK;
1483 switch (c)
1484 {
1485 case 0x09: /* HT */
1486 case 0x20: /* SPACE */
1487 case 0xa0: /* NBSP */
1488 case 0x1680: /* OGHAM SPACE MARK */
1489 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
1490 case 0x2000: /* EN QUAD */
1491 case 0x2001: /* EM QUAD */
1492 case 0x2002: /* EN SPACE */
1493 case 0x2003: /* EM SPACE */
1494 case 0x2004: /* THREE-PER-EM SPACE */
1495 case 0x2005: /* FOUR-PER-EM SPACE */
1496 case 0x2006: /* SIX-PER-EM SPACE */
1497 case 0x2007: /* FIGURE SPACE */
1498 case 0x2008: /* PUNCTUATION SPACE */
1499 case 0x2009: /* THIN SPACE */
1500 case 0x200A: /* HAIR SPACE */
1501 case 0x202f: /* NARROW NO-BREAK SPACE */
1502 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
1503 case 0x3000: /* IDEOGRAPHIC SPACE */
1504 OK = TRUE;
1505 break;
1506
1507 default:
1508 OK = FALSE;
1509 break;
1510 }
1511
1512 if (OK == (d == OP_HSPACE))
1513 {
1514 if (count > 0 && codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSPLUS)
1515 {
1516 active_count--; /* Remove non-match possibility */
1517 next_active_state--;
1518 }
1519 count++;
1520 ADD_NEW_DATA(-state_offset, count, 0);
1521 }
1522 }
1523 break;
1524
1525 /*-----------------------------------------------------------------*/
1526 #ifdef SUPPORT_UCP
1527 case OP_PROP_EXTRA + OP_TYPEQUERY:
1528 case OP_PROP_EXTRA + OP_TYPEMINQUERY:
1529 case OP_PROP_EXTRA + OP_TYPEPOSQUERY:
1530 count = 4;
1531 goto QS1;
1532
1533 case OP_PROP_EXTRA + OP_TYPESTAR:
1534 case OP_PROP_EXTRA + OP_TYPEMINSTAR:
1535 case OP_PROP_EXTRA + OP_TYPEPOSSTAR:
1536 count = 0;
1537
1538 QS1:
1539
1540 ADD_ACTIVE(state_offset + 4, 0);
1541 if (clen > 0)
1542 {
1543 BOOL OK;
1544 const pcre_uint8 chartype = UCD_CHARTYPE(c);
1545 switch(code[2])
1546 {
1547 case PT_ANY:
1548 OK = TRUE;
1549 break;
1550
1551 case PT_LAMP:
1552 OK = chartype == ucp_Lu || chartype == ucp_Ll ||
1553 chartype == ucp_Lt;
1554 break;
1555
1556 case PT_GC:
1557 OK = PRIV(ucp_gentype)[chartype] == code[3];
1558 break;
1559
1560 case PT_PC:
1561 OK = chartype == code[3];
1562 break;
1563
1564 case PT_SC:
1565 OK = UCD_SCRIPT(c) == code[3];
1566 break;
1567
1568 /* These are specials for combination cases. */
1569
1570 case PT_ALNUM:
1571 OK = PRIV(ucp_gentype)[chartype] == ucp_L ||
1572 PRIV(ucp_gentype)[chartype] == ucp_N;
1573 break;
1574
1575 case PT_SPACE: /* Perl space */
1576 OK = PRIV(ucp_gentype)[chartype] == ucp_Z ||
1577 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1578 break;
1579
1580 case PT_PXSPACE: /* POSIX space */
1581 OK = PRIV(ucp_gentype)[chartype] == ucp_Z ||
1582 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1583 c == CHAR_FF || c == CHAR_CR;
1584 break;
1585
1586 case PT_WORD:
1587 OK = PRIV(ucp_gentype)[chartype] == ucp_L ||
1588 PRIV(ucp_gentype)[chartype] == ucp_N ||
1589 c == CHAR_UNDERSCORE;
1590 break;
1591
1592 /* Should never occur, but keep compilers from grumbling. */
1593
1594 default:
1595 OK = codevalue != OP_PROP;
1596 break;
1597 }
1598
1599 if (OK == (d == OP_PROP))
1600 {
1601 if (codevalue == OP_PROP_EXTRA + OP_TYPEPOSSTAR ||
1602 codevalue == OP_PROP_EXTRA + OP_TYPEPOSQUERY)
1603 {
1604 active_count--; /* Remove non-match possibility */
1605 next_active_state--;
1606 }
1607 ADD_NEW(state_offset + count, 0);
1608 }
1609 }
1610 break;
1611
1612 /*-----------------------------------------------------------------*/
1613 case OP_EXTUNI_EXTRA + OP_TYPEQUERY:
1614 case OP_EXTUNI_EXTRA + OP_TYPEMINQUERY:
1615 case OP_EXTUNI_EXTRA + OP_TYPEPOSQUERY:
1616 count = 2;
1617 goto QS2;
1618
1619 case OP_EXTUNI_EXTRA + OP_TYPESTAR:
1620 case OP_EXTUNI_EXTRA + OP_TYPEMINSTAR:
1621 case OP_EXTUNI_EXTRA + OP_TYPEPOSSTAR:
1622 count = 0;
1623
1624 QS2:
1625
1626 ADD_ACTIVE(state_offset + 2, 0);
1627 if (clen > 0 && UCD_CATEGORY(c) != ucp_M)
1628 {
1629 const pcre_uchar *nptr = ptr + clen;
1630 int ncount = 0;
1631 if (codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSSTAR ||
1632 codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSQUERY)
1633 {
1634 active_count--; /* Remove non-match possibility */
1635 next_active_state--;
1636 }
1637 while (nptr < end_subject)
1638 {
1639 int nd;
1640 int ndlen = 1;
1641 GETCHARLEN(nd, nptr, ndlen);
1642 if (UCD_CATEGORY(nd) != ucp_M) break;
1643 ncount++;
1644 nptr += ndlen;
1645 }
1646 ADD_NEW_DATA(-(state_offset + count), 0, ncount);
1647 }
1648 break;
1649 #endif
1650
1651 /*-----------------------------------------------------------------*/
1652 case OP_ANYNL_EXTRA + OP_TYPEQUERY:
1653 case OP_ANYNL_EXTRA + OP_TYPEMINQUERY:
1654 case OP_ANYNL_EXTRA + OP_TYPEPOSQUERY:
1655 count = 2;
1656 goto QS3;
1657
1658 case OP_ANYNL_EXTRA + OP_TYPESTAR:
1659 case OP_ANYNL_EXTRA + OP_TYPEMINSTAR:
1660 case OP_ANYNL_EXTRA + OP_TYPEPOSSTAR:
1661 count = 0;
1662
1663 QS3:
1664 ADD_ACTIVE(state_offset + 2, 0);
1665 if (clen > 0)
1666 {
1667 int ncount = 0;
1668 switch (c)
1669 {
1670 case 0x000b:
1671 case 0x000c:
1672 case 0x0085:
1673 case 0x2028:
1674 case 0x2029:
1675 if ((md->moptions & PCRE_BSR_ANYCRLF) != 0) break;
1676 goto ANYNL02;
1677
1678 case 0x000d:
1679 if (ptr + 1 < end_subject && ptr[1] == 0x0a) ncount = 1;
1680 /* Fall through */
1681
1682 ANYNL02:
1683 case 0x000a:
1684 if (codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSSTAR ||
1685 codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSQUERY)
1686 {
1687 active_count--; /* Remove non-match possibility */
1688 next_active_state--;
1689 }
1690 ADD_NEW_DATA(-(state_offset + count), 0, ncount);
1691 break;
1692
1693 default:
1694 break;
1695 }
1696 }
1697 break;
1698
1699 /*-----------------------------------------------------------------*/
1700 case OP_VSPACE_EXTRA + OP_TYPEQUERY:
1701 case OP_VSPACE_EXTRA + OP_TYPEMINQUERY:
1702 case OP_VSPACE_EXTRA + OP_TYPEPOSQUERY:
1703 count = 2;
1704 goto QS4;
1705
1706 case OP_VSPACE_EXTRA + OP_TYPESTAR:
1707 case OP_VSPACE_EXTRA + OP_TYPEMINSTAR:
1708 case OP_VSPACE_EXTRA + OP_TYPEPOSSTAR:
1709 count = 0;
1710
1711 QS4:
1712 ADD_ACTIVE(state_offset + 2, 0);
1713 if (clen > 0)
1714 {
1715 BOOL OK;
1716 switch (c)
1717 {
1718 case 0x000a:
1719 case 0x000b:
1720 case 0x000c:
1721 case 0x000d:
1722 case 0x0085:
1723 case 0x2028:
1724 case 0x2029:
1725 OK = TRUE;
1726 break;
1727
1728 default:
1729 OK = FALSE;
1730 break;
1731 }
1732 if (OK == (d == OP_VSPACE))
1733 {
1734 if (codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSSTAR ||
1735 codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSQUERY)
1736 {
1737 active_count--; /* Remove non-match possibility */
1738 next_active_state--;
1739 }
1740 ADD_NEW_DATA(-(state_offset + count), 0, 0);
1741 }
1742 }
1743 break;
1744
1745 /*-----------------------------------------------------------------*/
1746 case OP_HSPACE_EXTRA + OP_TYPEQUERY:
1747 case OP_HSPACE_EXTRA + OP_TYPEMINQUERY:
1748 case OP_HSPACE_EXTRA + OP_TYPEPOSQUERY:
1749 count = 2;
1750 goto QS5;
1751
1752 case OP_HSPACE_EXTRA + OP_TYPESTAR:
1753 case OP_HSPACE_EXTRA + OP_TYPEMINSTAR:
1754 case OP_HSPACE_EXTRA + OP_TYPEPOSSTAR:
1755 count = 0;
1756
1757 QS5:
1758 ADD_ACTIVE(state_offset + 2, 0);
1759 if (clen > 0)
1760 {
1761 BOOL OK;
1762 switch (c)
1763 {
1764 case 0x09: /* HT */
1765 case 0x20: /* SPACE */
1766 case 0xa0: /* NBSP */
1767 case 0x1680: /* OGHAM SPACE MARK */
1768 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
1769 case 0x2000: /* EN QUAD */
1770 case 0x2001: /* EM QUAD */
1771 case 0x2002: /* EN SPACE */
1772 case 0x2003: /* EM SPACE */
1773 case 0x2004: /* THREE-PER-EM SPACE */
1774 case 0x2005: /* FOUR-PER-EM SPACE */
1775 case 0x2006: /* SIX-PER-EM SPACE */
1776 case 0x2007: /* FIGURE SPACE */
1777 case 0x2008: /* PUNCTUATION SPACE */
1778 case 0x2009: /* THIN SPACE */
1779 case 0x200A: /* HAIR SPACE */
1780 case 0x202f: /* NARROW NO-BREAK SPACE */
1781 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
1782 case 0x3000: /* IDEOGRAPHIC SPACE */
1783 OK = TRUE;
1784 break;
1785
1786 default:
1787 OK = FALSE;
1788 break;
1789 }
1790
1791 if (OK == (d == OP_HSPACE))
1792 {
1793 if (codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSSTAR ||
1794 codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSQUERY)
1795 {
1796 active_count--; /* Remove non-match possibility */
1797 next_active_state--;
1798 }
1799 ADD_NEW_DATA(-(state_offset + count), 0, 0);
1800 }
1801 }
1802 break;
1803
1804 /*-----------------------------------------------------------------*/
1805 #ifdef SUPPORT_UCP
1806 case OP_PROP_EXTRA + OP_TYPEEXACT:
1807 case OP_PROP_EXTRA + OP_TYPEUPTO:
1808 case OP_PROP_EXTRA + OP_TYPEMINUPTO:
1809 case OP_PROP_EXTRA + OP_TYPEPOSUPTO:
1810 if (codevalue != OP_PROP_EXTRA + OP_TYPEEXACT)
1811 { ADD_ACTIVE(state_offset + 1 + IMM2_SIZE + 3, 0); }
1812 count = current_state->count; /* Number already matched */
1813 if (clen > 0)
1814 {
1815 BOOL OK;
1816 const pcre_uint8 chartype = UCD_CHARTYPE(c);
1817 switch(code[1 + IMM2_SIZE + 1])
1818 {
1819 case PT_ANY:
1820 OK = TRUE;
1821 break;
1822
1823 case PT_LAMP:
1824 OK = chartype == ucp_Lu || chartype == ucp_Ll ||
1825 chartype == ucp_Lt;
1826 break;
1827
1828 case PT_GC:
1829 OK = PRIV(ucp_gentype)[chartype] == code[1 + IMM2_SIZE + 2];
1830 break;
1831
1832 case PT_PC:
1833 OK = chartype == code[1 + IMM2_SIZE + 2];
1834 break;
1835
1836 case PT_SC:
1837 OK = UCD_SCRIPT(c) == code[1 + IMM2_SIZE + 2];
1838 break;
1839
1840 /* These are specials for combination cases. */
1841
1842 case PT_ALNUM:
1843 OK = PRIV(ucp_gentype)[chartype] == ucp_L ||
1844 PRIV(ucp_gentype)[chartype] == ucp_N;
1845 break;
1846
1847 case PT_SPACE: /* Perl space */
1848 OK = PRIV(ucp_gentype)[chartype] == ucp_Z ||
1849 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR;
1850 break;
1851
1852 case PT_PXSPACE: /* POSIX space */
1853 OK = PRIV(ucp_gentype)[chartype] == ucp_Z ||
1854 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
1855 c == CHAR_FF || c == CHAR_CR;
1856 break;
1857
1858 case PT_WORD:
1859 OK = PRIV(ucp_gentype)[chartype] == ucp_L ||
1860 PRIV(ucp_gentype)[chartype] == ucp_N ||
1861 c == CHAR_UNDERSCORE;
1862 break;
1863
1864 /* Should never occur, but keep compilers from grumbling. */
1865
1866 default:
1867 OK = codevalue != OP_PROP;
1868 break;
1869 }
1870
1871 if (OK == (d == OP_PROP))
1872 {
1873 if (codevalue == OP_PROP_EXTRA + OP_TYPEPOSUPTO)
1874 {
1875 active_count--; /* Remove non-match possibility */
1876 next_active_state--;
1877 }
1878 if (++count >= GET2(code, 1))
1879 { ADD_NEW(state_offset + 1 + IMM2_SIZE + 3, 0); }
1880 else
1881 { ADD_NEW(state_offset, count); }
1882 }
1883 }
1884 break;
1885
1886 /*-----------------------------------------------------------------*/
1887 case OP_EXTUNI_EXTRA + OP_TYPEEXACT:
1888 case OP_EXTUNI_EXTRA + OP_TYPEUPTO:
1889 case OP_EXTUNI_EXTRA + OP_TYPEMINUPTO:
1890 case OP_EXTUNI_EXTRA + OP_TYPEPOSUPTO:
1891 if (codevalue != OP_EXTUNI_EXTRA + OP_TYPEEXACT)
1892 { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); }
1893 count = current_state->count; /* Number already matched */
1894 if (clen > 0 && UCD_CATEGORY(c) != ucp_M)
1895 {
1896 const pcre_uchar *nptr = ptr + clen;
1897 int ncount = 0;
1898 if (codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSUPTO)
1899 {
1900 active_count--; /* Remove non-match possibility */
1901 next_active_state--;
1902 }
1903 while (nptr < end_subject)
1904 {
1905 int nd;
1906 int ndlen = 1;
1907 GETCHARLEN(nd, nptr, ndlen);
1908 if (UCD_CATEGORY(nd) != ucp_M) break;
1909 ncount++;
1910 nptr += ndlen;
1911 }
1912 if (nptr >= end_subject && (md->moptions & PCRE_PARTIAL_HARD) != 0)
1913 reset_could_continue = TRUE;
1914 if (++count >= GET2(code, 1))
1915 { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, ncount); }
1916 else
1917 { ADD_NEW_DATA(-state_offset, count, ncount); }
1918 }
1919 break;
1920 #endif
1921
1922 /*-----------------------------------------------------------------*/
1923 case OP_ANYNL_EXTRA + OP_TYPEEXACT:
1924 case OP_ANYNL_EXTRA + OP_TYPEUPTO:
1925 case OP_ANYNL_EXTRA + OP_TYPEMINUPTO:
1926 case OP_ANYNL_EXTRA + OP_TYPEPOSUPTO:
1927 if (codevalue != OP_ANYNL_EXTRA + OP_TYPEEXACT)
1928 { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); }
1929 count = current_state->count; /* Number already matched */
1930 if (clen > 0)
1931 {
1932 int ncount = 0;
1933 switch (c)
1934 {
1935 case 0x000b:
1936 case 0x000c:
1937 case 0x0085:
1938 case 0x2028:
1939 case 0x2029:
1940 if ((md->moptions & PCRE_BSR_ANYCRLF) != 0) break;
1941 goto ANYNL03;
1942
1943 case 0x000d:
1944 if (ptr + 1 < end_subject && ptr[1] == 0x0a) ncount = 1;
1945 /* Fall through */
1946
1947 ANYNL03:
1948 case 0x000a:
1949 if (codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSUPTO)
1950 {
1951 active_count--; /* Remove non-match possibility */
1952 next_active_state--;
1953 }
1954 if (++count >= GET2(code, 1))
1955 { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, ncount); }
1956 else
1957 { ADD_NEW_DATA(-state_offset, count, ncount); }
1958 break;
1959
1960 default:
1961 break;
1962 }
1963 }
1964 break;
1965
1966 /*-----------------------------------------------------------------*/
1967 case OP_VSPACE_EXTRA + OP_TYPEEXACT:
1968 case OP_VSPACE_EXTRA + OP_TYPEUPTO:
1969 case OP_VSPACE_EXTRA + OP_TYPEMINUPTO:
1970 case OP_VSPACE_EXTRA + OP_TYPEPOSUPTO:
1971 if (codevalue != OP_VSPACE_EXTRA + OP_TYPEEXACT)
1972 { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); }
1973 count = current_state->count; /* Number already matched */
1974 if (clen > 0)
1975 {
1976 BOOL OK;
1977 switch (c)
1978 {
1979 case 0x000a:
1980 case 0x000b:
1981 case 0x000c:
1982 case 0x000d:
1983 case 0x0085:
1984 case 0x2028:
1985 case 0x2029:
1986 OK = TRUE;
1987 break;
1988
1989 default:
1990 OK = FALSE;
1991 }
1992
1993 if (OK == (d == OP_VSPACE))
1994 {
1995 if (codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSUPTO)
1996 {
1997 active_count--; /* Remove non-match possibility */
1998 next_active_state--;
1999 }
2000 if (++count >= GET2(code, 1))
2001 { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, 0); }
2002 else
2003 { ADD_NEW_DATA(-state_offset, count, 0); }
2004 }
2005 }
2006 break;
2007
2008 /*-----------------------------------------------------------------*/
2009 case OP_HSPACE_EXTRA + OP_TYPEEXACT:
2010 case OP_HSPACE_EXTRA + OP_TYPEUPTO:
2011 case OP_HSPACE_EXTRA + OP_TYPEMINUPTO:
2012 case OP_HSPACE_EXTRA + OP_TYPEPOSUPTO:
2013 if (codevalue != OP_HSPACE_EXTRA + OP_TYPEEXACT)
2014 { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); }
2015 count = current_state->count; /* Number already matched */
2016 if (clen > 0)
2017 {
2018 BOOL OK;
2019 switch (c)
2020 {
2021 case 0x09: /* HT */
2022 case 0x20: /* SPACE */
2023 case 0xa0: /* NBSP */
2024 case 0x1680: /* OGHAM SPACE MARK */
2025 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
2026 case 0x2000: /* EN QUAD */
2027 case 0x2001: /* EM QUAD */
2028 case 0x2002: /* EN SPACE */
2029 case 0x2003: /* EM SPACE */
2030 case 0x2004: /* THREE-PER-EM SPACE */
2031 case 0x2005: /* FOUR-PER-EM SPACE */
2032 case 0x2006: /* SIX-PER-EM SPACE */
2033 case 0x2007: /* FIGURE SPACE */
2034 case 0x2008: /* PUNCTUATION SPACE */
2035 case 0x2009: /* THIN SPACE */
2036 case 0x200A: /* HAIR SPACE */
2037 case 0x202f: /* NARROW NO-BREAK SPACE */
2038 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
2039 case 0x3000: /* IDEOGRAPHIC SPACE */
2040 OK = TRUE;
2041 break;
2042
2043 default:
2044 OK = FALSE;
2045 break;
2046 }
2047
2048 if (OK == (d == OP_HSPACE))
2049 {
2050 if (codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSUPTO)
2051 {
2052 active_count--; /* Remove non-match possibility */
2053 next_active_state--;
2054 }
2055 if (++count >= GET2(code, 1))
2056 { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, 0); }
2057 else
2058 { ADD_NEW_DATA(-state_offset, count, 0); }
2059 }
2060 }
2061 break;
2062
2063 /* ========================================================================== */
2064 /* These opcodes are followed by a character that is usually compared
2065 to the current subject character; it is loaded into d. We still get
2066 here even if there is no subject character, because in some cases zero
2067 repetitions are permitted. */
2068
2069 /*-----------------------------------------------------------------*/
2070 case OP_CHAR:
2071 if (clen > 0 && c == d) { ADD_NEW(state_offset + dlen + 1, 0); }
2072 break;
2073
2074 /*-----------------------------------------------------------------*/
2075 case OP_CHARI:
2076 if (clen == 0) break;
2077
2078 #ifdef SUPPORT_UTF
2079 if (utf)
2080 {
2081 if (c == d) { ADD_NEW(state_offset + dlen + 1, 0); } else
2082 {
2083 unsigned int othercase;
2084 if (c < 128)
2085 othercase = fcc[c];
2086 else
2087 /* If we have Unicode property support, we can use it to test the
2088 other case of the character. */
2089 #ifdef SUPPORT_UCP
2090 othercase = UCD_OTHERCASE(c);
2091 #else
2092 othercase = NOTACHAR;
2093 #endif
2094
2095 if (d == othercase) { ADD_NEW(state_offset + dlen + 1, 0); }
2096 }
2097 }
2098 else
2099 #endif /* SUPPORT_UTF */
2100 /* Not UTF mode */
2101 {
2102 if (TABLE_GET(c, lcc, c) == TABLE_GET(d, lcc, d))
2103 { ADD_NEW(state_offset + 2, 0); }
2104 }
2105 break;
2106
2107
2108 #ifdef SUPPORT_UCP
2109 /*-----------------------------------------------------------------*/
2110 /* This is a tricky one because it can match more than one character.
2111 Find out how many characters to skip, and then set up a negative state
2112 to wait for them to pass before continuing. */
2113
2114 case OP_EXTUNI:
2115 if (clen > 0 && UCD_CATEGORY(c) != ucp_M)
2116 {
2117 const pcre_uchar *nptr = ptr + clen;
2118 int ncount = 0;
2119 while (nptr < end_subject)
2120 {
2121 int nclen = 1;
2122 GETCHARLEN(c, nptr, nclen);
2123 if (UCD_CATEGORY(c) != ucp_M) break;
2124 ncount++;
2125 nptr += nclen;
2126 }
2127 if (nptr >= end_subject && (md->moptions & PCRE_PARTIAL_HARD) != 0)
2128 reset_could_continue = TRUE;
2129 ADD_NEW_DATA(-(state_offset + 1), 0, ncount);
2130 }
2131 break;
2132 #endif
2133
2134 /*-----------------------------------------------------------------*/
2135 /* This is a tricky like EXTUNI because it too can match more than one
2136 character (when CR is followed by LF). In this case, set up a negative
2137 state to wait for one character to pass before continuing. */
2138
2139 case OP_ANYNL:
2140 if (clen > 0) switch(c)
2141 {
2142 case 0x000b:
2143 case 0x000c:
2144 case 0x0085:
2145 case 0x2028:
2146 case 0x2029:
2147 if ((md->moptions & PCRE_BSR_ANYCRLF) != 0) break;
2148
2149 case 0x000a:
2150 ADD_NEW(state_offset + 1, 0);
2151 break;
2152
2153 case 0x000d:
2154 if (ptr + 1 >= end_subject)
2155 {
2156 ADD_NEW(state_offset + 1, 0);
2157 if ((md->moptions & PCRE_PARTIAL_HARD) != 0)
2158 reset_could_continue = TRUE;
2159 }
2160 else if (ptr[1] == 0x0a)
2161 {
2162 ADD_NEW_DATA(-(state_offset + 1), 0, 1);
2163 }
2164 else
2165 {
2166 ADD_NEW(state_offset + 1, 0);
2167 }
2168 break;
2169 }
2170 break;
2171
2172 /*-----------------------------------------------------------------*/
2173 case OP_NOT_VSPACE:
2174 if (clen > 0) switch(c)
2175 {
2176 case 0x000a:
2177 case 0x000b:
2178 case 0x000c:
2179 case 0x000d:
2180 case 0x0085:
2181 case 0x2028:
2182 case 0x2029:
2183 break;
2184
2185 default:
2186 ADD_NEW(state_offset + 1, 0);
2187 break;
2188 }
2189 break;
2190
2191 /*-----------------------------------------------------------------*/
2192 case OP_VSPACE:
2193 if (clen > 0) switch(c)
2194 {
2195 case 0x000a:
2196 case 0x000b:
2197 case 0x000c:
2198 case 0x000d:
2199 case 0x0085:
2200 case 0x2028:
2201 case 0x2029:
2202 ADD_NEW(state_offset + 1, 0);
2203 break;
2204
2205 default: break;
2206 }
2207 break;
2208
2209 /*-----------------------------------------------------------------*/
2210 case OP_NOT_HSPACE:
2211 if (clen > 0) switch(c)
2212 {
2213 case 0x09: /* HT */
2214 case 0x20: /* SPACE */
2215 case 0xa0: /* NBSP */
2216 case 0x1680: /* OGHAM SPACE MARK */
2217 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
2218 case 0x2000: /* EN QUAD */
2219 case 0x2001: /* EM QUAD */
2220 case 0x2002: /* EN SPACE */
2221 case 0x2003: /* EM SPACE */
2222 case 0x2004: /* THREE-PER-EM SPACE */
2223 case 0x2005: /* FOUR-PER-EM SPACE */
2224 case 0x2006: /* SIX-PER-EM SPACE */
2225 case 0x2007: /* FIGURE SPACE */
2226 case 0x2008: /* PUNCTUATION SPACE */
2227 case 0x2009: /* THIN SPACE */
2228 case 0x200A: /* HAIR SPACE */
2229 case 0x202f: /* NARROW NO-BREAK SPACE */
2230 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
2231 case 0x3000: /* IDEOGRAPHIC SPACE */
2232 break;
2233
2234 default:
2235 ADD_NEW(state_offset + 1, 0);
2236 break;
2237 }
2238 break;
2239
2240 /*-----------------------------------------------------------------*/
2241 case OP_HSPACE:
2242 if (clen > 0) switch(c)
2243 {
2244 case 0x09: /* HT */
2245 case 0x20: /* SPACE */
2246 case 0xa0: /* NBSP */
2247 case 0x1680: /* OGHAM SPACE MARK */
2248 case 0x180e: /* MONGOLIAN VOWEL SEPARATOR */
2249 case 0x2000: /* EN QUAD */
2250 case 0x2001: /* EM QUAD */
2251 case 0x2002: /* EN SPACE */
2252 case 0x2003: /* EM SPACE */
2253 case 0x2004: /* THREE-PER-EM SPACE */
2254 case 0x2005: /* FOUR-PER-EM SPACE */
2255 case 0x2006: /* SIX-PER-EM SPACE */
2256 case 0x2007: /* FIGURE SPACE */
2257 case 0x2008: /* PUNCTUATION SPACE */
2258 case 0x2009: /* THIN SPACE */
2259 case 0x200A: /* HAIR SPACE */
2260 case 0x202f: /* NARROW NO-BREAK SPACE */
2261 case 0x205f: /* MEDIUM MATHEMATICAL SPACE */
2262 case 0x3000: /* IDEOGRAPHIC SPACE */
2263 ADD_NEW(state_offset + 1, 0);
2264 break;
2265 }
2266 break;
2267
2268 /*-----------------------------------------------------------------*/
2269 /* Match a negated single character casefully. */
2270
2271 case OP_NOT:
2272 if (clen > 0 && c != d) { ADD_NEW(state_offset + dlen + 1, 0); }
2273 break;
2274
2275 /*-----------------------------------------------------------------*/
2276 /* Match a negated single character caselessly. */
2277
2278 case OP_NOTI:
2279 if (clen > 0)
2280 {
2281 unsigned int otherd;
2282 #ifdef SUPPORT_UTF
2283 if (utf && d >= 128)
2284 {
2285 #ifdef SUPPORT_UCP
2286 otherd = UCD_OTHERCASE(d);
2287 #endif /* SUPPORT_UCP */
2288 }
2289 else
2290 #endif /* SUPPORT_UTF */
2291 otherd = TABLE_GET(d, fcc, d);
2292 if (c != d && c != otherd)
2293 { ADD_NEW(state_offset + dlen + 1, 0); }
2294 }
2295 break;
2296
2297 /*-----------------------------------------------------------------*/
2298 case OP_PLUSI:
2299 case OP_MINPLUSI:
2300 case OP_POSPLUSI:
2301 case OP_NOTPLUSI:
2302 case OP_NOTMINPLUSI:
2303 case OP_NOTPOSPLUSI:
2304 caseless = TRUE;
2305 codevalue -= OP_STARI - OP_STAR;
2306
2307 /* Fall through */
2308 case OP_PLUS:
2309 case OP_MINPLUS:
2310 case OP_POSPLUS:
2311 case OP_NOTPLUS:
2312 case OP_NOTMINPLUS:
2313 case OP_NOTPOSPLUS:
2314 count = current_state->count; /* Already matched */
2315 if (count > 0) { ADD_ACTIVE(state_offset + dlen + 1, 0); }
2316 if (clen > 0)
2317 {
2318 unsigned int otherd = NOTACHAR;
2319 if (caseless)
2320 {
2321 #ifdef SUPPORT_UTF
2322 if (utf && d >= 128)
2323 {
2324 #ifdef SUPPORT_UCP
2325 otherd = UCD_OTHERCASE(d);
2326 #endif /* SUPPORT_UCP */
2327 }
2328 else
2329 #endif /* SUPPORT_UTF */
2330 otherd = TABLE_GET(d, fcc, d);
2331 }
2332 if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR))
2333 {
2334 if (count > 0 &&
2335 (codevalue == OP_POSPLUS || codevalue == OP_NOTPOSPLUS))
2336 {
2337 active_count--; /* Remove non-match possibility */
2338 next_active_state--;
2339 }
2340 count++;
2341 ADD_NEW(state_offset, count);
2342 }
2343 }
2344 break;
2345
2346 /*-----------------------------------------------------------------*/
2347 case OP_QUERYI:
2348 case OP_MINQUERYI:
2349 case OP_POSQUERYI:
2350 case OP_NOTQUERYI:
2351 case OP_NOTMINQUERYI:
2352 case OP_NOTPOSQUERYI:
2353 caseless = TRUE;
2354 codevalue -= OP_STARI - OP_STAR;
2355 /* Fall through */
2356 case OP_QUERY:
2357 case OP_MINQUERY:
2358 case OP_POSQUERY:
2359 case OP_NOTQUERY:
2360 case OP_NOTMINQUERY:
2361 case OP_NOTPOSQUERY:
2362 ADD_ACTIVE(state_offset + dlen + 1, 0);
2363 if (clen > 0)
2364 {
2365 unsigned int otherd = NOTACHAR;
2366 if (caseless)
2367 {
2368 #ifdef SUPPORT_UTF
2369 if (utf && d >= 128)
2370 {
2371 #ifdef SUPPORT_UCP
2372 otherd = UCD_OTHERCASE(d);
2373 #endif /* SUPPORT_UCP */
2374 }
2375 else
2376 #endif /* SUPPORT_UTF */
2377 otherd = TABLE_GET(d, fcc, d);
2378 }
2379 if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR))
2380 {
2381 if (codevalue == OP_POSQUERY || codevalue == OP_NOTPOSQUERY)
2382 {
2383 active_count--; /* Remove non-match possibility */
2384 next_active_state--;
2385 }
2386 ADD_NEW(state_offset + dlen + 1, 0);
2387 }
2388 }
2389 break;
2390
2391 /*-----------------------------------------------------------------*/
2392 case OP_STARI:
2393 case OP_MINSTARI:
2394 case OP_POSSTARI:
2395 case OP_NOTSTARI:
2396 case OP_NOTMINSTARI:
2397 case OP_NOTPOSSTARI:
2398 caseless = TRUE;
2399 codevalue -= OP_STARI - OP_STAR;
2400 /* Fall through */
2401 case OP_STAR:
2402 case OP_MINSTAR:
2403 case OP_POSSTAR:
2404 case OP_NOTSTAR:
2405 case OP_NOTMINSTAR:
2406 case OP_NOTPOSSTAR:
2407 ADD_ACTIVE(state_offset + dlen + 1, 0);
2408 if (clen > 0)
2409 {
2410 unsigned int otherd = NOTACHAR;
2411 if (caseless)
2412 {
2413 #ifdef SUPPORT_UTF
2414 if (utf && d >= 128)
2415 {
2416 #ifdef SUPPORT_UCP
2417 otherd = UCD_OTHERCASE(d);
2418 #endif /* SUPPORT_UCP */
2419 }
2420 else
2421 #endif /* SUPPORT_UTF */
2422 otherd = TABLE_GET(d, fcc, d);
2423 }
2424 if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR))
2425 {
2426 if (codevalue == OP_POSSTAR || codevalue == OP_NOTPOSSTAR)
2427 {
2428 active_count--; /* Remove non-match possibility */
2429 next_active_state--;
2430 }
2431 ADD_NEW(state_offset, 0);
2432 }
2433 }
2434 break;
2435
2436 /*-----------------------------------------------------------------*/
2437 case OP_EXACTI:
2438 case OP_NOTEXACTI:
2439 caseless = TRUE;
2440 codevalue -= OP_STARI - OP_STAR;
2441 /* Fall through */
2442 case OP_EXACT:
2443 case OP_NOTEXACT:
2444 count = current_state->count; /* Number already matched */
2445 if (clen > 0)
2446 {
2447 unsigned int otherd = NOTACHAR;
2448 if (caseless)
2449 {
2450 #ifdef SUPPORT_UTF
2451 if (utf && d >= 128)
2452 {
2453 #ifdef SUPPORT_UCP
2454 otherd = UCD_OTHERCASE(d);
2455 #endif /* SUPPORT_UCP */
2456 }
2457 else
2458 #endif /* SUPPORT_UTF */
2459 otherd = TABLE_GET(d, fcc, d);
2460 }
2461 if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR))
2462 {
2463 if (++count >= GET2(code, 1))
2464 { ADD_NEW(state_offset + dlen + 1 + IMM2_SIZE, 0); }
2465 else
2466 { ADD_NEW(state_offset, count); }
2467 }
2468 }
2469 break;
2470
2471 /*-----------------------------------------------------------------*/
2472 case OP_UPTOI:
2473 case OP_MINUPTOI:
2474 case OP_POSUPTOI:
2475 case OP_NOTUPTOI:
2476 case OP_NOTMINUPTOI:
2477 case OP_NOTPOSUPTOI:
2478 caseless = TRUE;
2479 codevalue -= OP_STARI - OP_STAR;
2480 /* Fall through */
2481 case OP_UPTO:
2482 case OP_MINUPTO:
2483 case OP_POSUPTO:
2484 case OP_NOTUPTO:
2485 case OP_NOTMINUPTO:
2486 case OP_NOTPOSUPTO:
2487 ADD_ACTIVE(state_offset + dlen + 1 + IMM2_SIZE, 0);
2488 count = current_state->count; /* Number already matched */
2489 if (clen > 0)
2490 {
2491 unsigned int otherd = NOTACHAR;
2492 if (caseless)
2493 {
2494 #ifdef SUPPORT_UTF
2495 if (utf && d >= 128)
2496 {
2497 #ifdef SUPPORT_UCP
2498 otherd = UCD_OTHERCASE(d);
2499 #endif /* SUPPORT_UCP */
2500 }
2501 else
2502 #endif /* SUPPORT_UTF */
2503 otherd = TABLE_GET(d, fcc, d);
2504 }
2505 if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR))
2506 {
2507 if (codevalue == OP_POSUPTO || codevalue == OP_NOTPOSUPTO)
2508 {
2509 active_count--; /* Remove non-match possibility */
2510 next_active_state--;
2511 }
2512 if (++count >= GET2(code, 1))
2513 { ADD_NEW(state_offset + dlen + 1 + IMM2_SIZE, 0); }
2514 else
2515 { ADD_NEW(state_offset, count); }
2516 }
2517 }
2518 break;
2519
2520
2521 /* ========================================================================== */
2522 /* These are the class-handling opcodes */
2523
2524 case OP_CLASS:
2525 case OP_NCLASS:
2526 case OP_XCLASS:
2527 {
2528 BOOL isinclass = FALSE;
2529 int next_state_offset;
2530 const pcre_uchar *ecode;
2531
2532 /* For a simple class, there is always just a 32-byte table, and we
2533 can set isinclass from it. */
2534
2535 if (codevalue != OP_XCLASS)
2536 {
2537 ecode = code + 1 + (32 / sizeof(pcre_uchar));
2538 if (clen > 0)
2539 {
2540 isinclass = (c > 255)? (codevalue == OP_NCLASS) :
2541 ((((pcre_uint8 *)(code + 1))[c/8] & (1 << (c&7))) != 0);
2542 }
2543 }
2544
2545 /* An extended class may have a table or a list of single characters,
2546 ranges, or both, and it may be positive or negative. There's a
2547 function that sorts all this out. */
2548
2549 else
2550 {
2551 ecode = code + GET(code, 1);
2552 if (clen > 0) isinclass = PRIV(xclass)(c, code + 1 + LINK_SIZE, utf);
2553 }
2554
2555 /* At this point, isinclass is set for all kinds of class, and ecode
2556 points to the byte after the end of the class. If there is a
2557 quantifier, this is where it will be. */
2558
2559 next_state_offset = (int)(ecode - start_code);
2560
2561 switch (*ecode)
2562 {
2563 case OP_CRSTAR:
2564 case OP_CRMINSTAR:
2565 ADD_ACTIVE(next_state_offset + 1, 0);
2566 if (isinclass) { ADD_NEW(state_offset, 0); }
2567 break;
2568
2569 case OP_CRPLUS:
2570 case OP_CRMINPLUS:
2571 count = current_state->count; /* Already matched */
2572 if (count > 0) { ADD_ACTIVE(next_state_offset + 1, 0); }
2573 if (isinclass) { count++; ADD_NEW(state_offset, count); }
2574 break;
2575
2576 case OP_CRQUERY:
2577 case OP_CRMINQUERY:
2578 ADD_ACTIVE(next_state_offset + 1, 0);
2579 if (isinclass) { ADD_NEW(next_state_offset + 1, 0); }
2580 break;
2581
2582 case OP_CRRANGE:
2583 case OP_CRMINRANGE:
2584 count = current_state->count; /* Already matched */
2585 if (count >= GET2(ecode, 1))
2586 { ADD_ACTIVE(next_state_offset + 1 + 2 * IMM2_SIZE, 0); }
2587 if (isinclass)
2588 {
2589 int max = GET2(ecode, 1 + IMM2_SIZE);
2590 if (++count >= max && max != 0) /* Max 0 => no limit */
2591 { ADD_NEW(next_state_offset + 1 + 2 * IMM2_SIZE, 0); }
2592 else
2593 { ADD_NEW(state_offset, count); }
2594 }
2595 break;
2596
2597 default:
2598 if (isinclass) { ADD_NEW(next_state_offset, 0); }
2599 break;
2600 }
2601 }
2602 break;
2603
2604 /* ========================================================================== */
2605 /* These are the opcodes for fancy brackets of various kinds. We have
2606 to use recursion in order to handle them. The "always failing" assertion
2607 (?!) is optimised to OP_FAIL when compiling, so we have to support that,
2608 though the other "backtracking verbs" are not supported. */
2609
2610 case OP_FAIL:
2611 forced_fail++; /* Count FAILs for multiple states */
2612 break;
2613
2614 case OP_ASSERT:
2615 case OP_ASSERT_NOT:
2616 case OP_ASSERTBACK:
2617 case OP_ASSERTBACK_NOT:
2618 {
2619 int rc;
2620 int local_offsets[2];
2621 int local_workspace[1000];
2622 const pcre_uchar *endasscode = code + GET(code, 1);
2623
2624 while (*endasscode == OP_ALT) endasscode += GET(endasscode, 1);
2625
2626 rc = internal_dfa_exec(
2627 md, /* static match data */
2628 code, /* this subexpression's code */
2629 ptr, /* where we currently are */
2630 (int)(ptr - start_subject), /* start offset */
2631 local_offsets, /* offset vector */
2632 sizeof(local_offsets)/sizeof(int), /* size of same */
2633 local_workspace, /* workspace vector */
2634 sizeof(local_workspace)/sizeof(int), /* size of same */
2635 rlevel); /* function recursion level */
2636
2637 if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2638 if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK))
2639 { ADD_ACTIVE((int)(endasscode + LINK_SIZE + 1 - start_code), 0); }
2640 }
2641 break;
2642
2643 /*-----------------------------------------------------------------*/
2644 case OP_COND:
2645 case OP_SCOND:
2646 {
2647 int local_offsets[1000];
2648 int local_workspace[1000];
2649 int codelink = GET(code, 1);
2650 int condcode;
2651
2652 /* Because of the way auto-callout works during compile, a callout item
2653 is inserted between OP_COND and an assertion condition. This does not
2654 happen for the other conditions. */
2655
2656 if (code[LINK_SIZE+1] == OP_CALLOUT)
2657 {
2658 rrc = 0;
2659 if (PUBL(callout) != NULL)
2660 {
2661 PUBL(callout_block) cb;
2662 cb.version = 1; /* Version 1 of the callout block */
2663 cb.callout_number = code[LINK_SIZE+2];
2664 cb.offset_vector = offsets;
2665 #ifdef COMPILE_PCRE8
2666 cb.subject = (PCRE_SPTR)start_subject;
2667 #else
2668 cb.subject = (PCRE_SPTR16)start_subject;
2669 #endif
2670 cb.subject_length = (int)(end_subject - start_subject);
2671 cb.start_match = (int)(current_subject - start_subject);
2672 cb.current_position = (int)(ptr - start_subject);
2673 cb.pattern_position = GET(code, LINK_SIZE + 3);
2674 cb.next_item_length = GET(code, 3 + 2*LINK_SIZE);
2675 cb.capture_top = 1;
2676 cb.capture_last = -1;
2677 cb.callout_data = md->callout_data;
2678 cb.mark = NULL; /* No (*MARK) support */
2679 if ((rrc = (*PUBL(callout))(&cb)) < 0) return rrc; /* Abandon */
2680 }
2681 if (rrc > 0) break; /* Fail this thread */
2682 code += PRIV(OP_lengths)[OP_CALLOUT]; /* Skip callout data */
2683 }
2684
2685 condcode = code[LINK_SIZE+1];
2686
2687 /* Back reference conditions are not supported */
2688
2689 if (condcode == OP_CREF || condcode == OP_NCREF)
2690 return PCRE_ERROR_DFA_UCOND;
2691
2692 /* The DEFINE condition is always false */
2693
2694 if (condcode == OP_DEF)
2695 { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); }
2696
2697 /* The only supported version of OP_RREF is for the value RREF_ANY,
2698 which means "test if in any recursion". We can't test for specifically
2699 recursed groups. */
2700
2701 else if (condcode == OP_RREF || condcode == OP_NRREF)
2702 {
2703 int value = GET2(code, LINK_SIZE + 2);
2704 if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND;
2705 if (md->recursive != NULL)
2706 { ADD_ACTIVE(state_offset + LINK_SIZE + 2 + IMM2_SIZE, 0); }
2707 else { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); }
2708 }
2709
2710 /* Otherwise, the condition is an assertion */
2711
2712 else
2713 {
2714 int rc;
2715 const pcre_uchar *asscode = code + LINK_SIZE + 1;
2716 const pcre_uchar *endasscode = asscode + GET(asscode, 1);
2717
2718 while (*endasscode == OP_ALT) endasscode += GET(endasscode, 1);
2719
2720 rc = internal_dfa_exec(
2721 md, /* fixed match data */
2722 asscode, /* this subexpression's code */
2723 ptr, /* where we currently are */
2724 (int)(ptr - start_subject), /* start offset */
2725 local_offsets, /* offset vector */
2726 sizeof(local_offsets)/sizeof(int), /* size of same */
2727 local_workspace, /* workspace vector */
2728 sizeof(local_workspace)/sizeof(int), /* size of same */
2729 rlevel); /* function recursion level */
2730
2731 if (rc == PCRE_ERROR_DFA_UITEM) return rc;
2732 if ((rc >= 0) ==
2733 (condcode == OP_ASSERT || condcode == OP_ASSERTBACK))
2734 { ADD_ACTIVE((int)(endasscode + LINK_SIZE + 1 - start_code), 0); }
2735 else
2736 { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); }
2737 }
2738 }
2739 break;
2740
2741 /*-----------------------------------------------------------------*/
2742 case OP_RECURSE:
2743 {
2744 dfa_recursion_info *ri;
2745 int local_offsets[1000];
2746 int local_workspace[1000];
2747 const pcre_uchar *callpat = start_code + GET(code, 1);
2748 int recno = (callpat == md->start_code)? 0 :
2749 GET2(callpat, 1 + LINK_SIZE);
2750 int rc;
2751
2752 DPRINTF(("%.*sStarting regex recursion\n", rlevel*2-2, SP));
2753
2754 /* Check for repeating a recursion without advancing the subject
2755 pointer. This should catch convoluted mutual recursions. (Some simple
2756 cases are caught at compile time.) */
2757
2758 for (ri = md->recursive; ri != NULL; ri = ri->prevrec)
2759 if (recno == ri->group_num && ptr == ri->subject_position)
2760 return PCRE_ERROR_RECURSELOOP;
2761
2762 /* Remember this recursion and where we started it so as to
2763 catch infinite loops. */
2764
2765 new_recursive.group_num = recno;
2766 new_recursive.subject_position = ptr;
2767 new_recursive.prevrec = md->recursive;
2768 md->recursive = &new_recursive;
2769
2770 rc = internal_dfa_exec(
2771 md, /* fixed match data */
2772 callpat, /* this subexpression's code */
2773 ptr, /* where we currently are */
2774 (int)(ptr - start_subject), /* start offset */
2775 local_offsets, /* offset vector */
2776 sizeof(local_offsets)/sizeof(int), /* size of same */
2777 local_workspace, /* workspace vector */
2778 sizeof(local_workspace)/sizeof(int), /* size of same */
2779 rlevel); /* function recursion level */
2780
2781 md->recursive = new_recursive.prevrec; /* Done this recursion */
2782
2783 DPRINTF(("%.*sReturn from regex recursion: rc=%d\n", rlevel*2-2, SP,
2784 rc));
2785
2786 /* Ran out of internal offsets */
2787
2788 if (rc == 0) return PCRE_ERROR_DFA_RECURSE;
2789
2790 /* For each successful matched substring, set up the next state with a
2791 count of characters to skip before trying it. Note that the count is in
2792 characters, not bytes. */
2793
2794 if (rc > 0)
2795 {
2796 for (rc = rc*2 - 2; rc >= 0; rc -= 2)
2797 {
2798 int charcount = local_offsets[rc+1] - local_offsets[rc];
2799 #ifdef SUPPORT_UTF
2800 if (utf)
2801 {
2802 const pcre_uchar *p = start_subject + local_offsets[rc];
2803 const pcre_uchar *pp = start_subject + local_offsets[rc+1];
2804 while (p < pp) if (NOT_FIRSTCHAR(*p++)) charcount--;
2805 }
2806 #endif
2807 if (charcount > 0)
2808 {
2809 ADD_NEW_DATA(-(state_offset + LINK_SIZE + 1), 0, (charcount - 1));
2810 }
2811 else
2812 {
2813 ADD_ACTIVE(state_offset + LINK_SIZE + 1, 0);
2814 }
2815 }
2816 }
2817 else if (rc != PCRE_ERROR_NOMATCH) return rc;
2818 }
2819 break;
2820
2821 /*-----------------------------------------------------------------*/
2822 case OP_BRAPOS:
2823 case OP_SBRAPOS:
2824 case OP_CBRAPOS:
2825 case OP_SCBRAPOS:
2826 case OP_BRAPOSZERO:
2827 {
2828 int charcount, matched_count;
2829 const pcre_uchar *local_ptr = ptr;
2830 BOOL allow_zero;
2831
2832 if (codevalue == OP_BRAPOSZERO)
2833 {
2834 allow_zero = TRUE;
2835 codevalue = *(++code); /* Codevalue will be one of above BRAs */
2836 }
2837 else allow_zero = FALSE;
2838
2839 /* Loop to match the subpattern as many times as possible as if it were
2840 a complete pattern. */
2841
2842 for (matched_count = 0;; matched_count++)
2843 {
2844 int local_offsets[2];
2845 int local_workspace[1000];
2846
2847 int rc = internal_dfa_exec(
2848 md, /* fixed match data */
2849 code, /* this subexpression's code */
2850 local_ptr, /* where we currently are */
2851 (int)(ptr - start_subject), /* start offset */
2852 local_offsets, /* offset vector */
2853 sizeof(local_offsets)/sizeof(int), /* size of same */
2854 local_workspace, /* workspace vector */
2855 sizeof(local_workspace)/sizeof(int), /* size of same */
2856 rlevel); /* function recursion level */
2857
2858 /* Failed to match */
2859
2860 if (rc < 0)
2861 {
2862 if (rc != PCRE_ERROR_NOMATCH) return rc;
2863 break;
2864 }
2865
2866 /* Matched: break the loop if zero characters matched. */
2867
2868 charcount = local_offsets[1] - local_offsets[0];
2869 if (charcount == 0) break;
2870 local_ptr += charcount; /* Advance temporary position ptr */
2871 }
2872
2873 /* At this point we have matched the subpattern matched_count
2874 times, and local_ptr is pointing to the character after the end of the
2875 last match. */
2876
2877 if (matched_count > 0 || allow_zero)
2878 {
2879 const pcre_uchar *end_subpattern = code;
2880 int next_state_offset;
2881
2882 do { end_subpattern += GET(end_subpattern, 1); }
2883 while (*end_subpattern == OP_ALT);
2884 next_state_offset =
2885 (int)(end_subpattern - start_code + LINK_SIZE + 1);
2886
2887 /* Optimization: if there are no more active states, and there
2888 are no new states yet set up, then skip over the subject string
2889 right here, to save looping. Otherwise, set up the new state to swing
2890 into action when the end of the matched substring is reached. */
2891
2892 if (i + 1 >= active_count && new_count == 0)
2893 {
2894 ptr = local_ptr;
2895 clen = 0;
2896 ADD_NEW(next_state_offset, 0);
2897 }
2898 else
2899 {
2900 const pcre_uchar *p = ptr;
2901 const pcre_uchar *pp = local_ptr;
2902 charcount = (int)(pp - p);
2903 #ifdef SUPPORT_UTF
2904 if (utf) while (p < pp) if (NOT_FIRSTCHAR(*p++)) charcount--;
2905 #endif
2906 ADD_NEW_DATA(-next_state_offset, 0, (charcount - 1));
2907 }
2908 }
2909 }
2910 break;
2911
2912 /*-----------------------------------------------------------------*/
2913 case OP_ONCE:
2914 case OP_ONCE_NC:
2915 {
2916 int local_offsets[2];
2917 int local_workspace[1000];
2918
2919 int rc = internal_dfa_exec(
2920 md, /* fixed match data */
2921 code, /* this subexpression's code */
2922 ptr, /* where we currently are */
2923 (int)(ptr - start_subject), /* start offset */
2924 local_offsets, /* offset vector */
2925 sizeof(local_offsets)/sizeof(int), /* size of same */
2926 local_workspace, /* workspace vector */
2927 sizeof(local_workspace)/sizeof(int), /* size of same */
2928 rlevel); /* function recursion level */
2929
2930 if (rc >= 0)
2931 {
2932 const pcre_uchar *end_subpattern = code;
2933 int charcount = local_offsets[1] - local_offsets[0];
2934 int next_state_offset, repeat_state_offset;
2935
2936 do { end_subpattern += GET(end_subpattern, 1); }
2937 while (*end_subpattern == OP_ALT);
2938 next_state_offset =
2939 (int)(end_subpattern - start_code + LINK_SIZE + 1);
2940
2941 /* If the end of this subpattern is KETRMAX or KETRMIN, we must
2942 arrange for the repeat state also to be added to the relevant list.
2943 Calculate the offset, or set -1 for no repeat. */
2944
2945 repeat_state_offset = (*end_subpattern == OP_KETRMAX ||
2946 *end_subpattern == OP_KETRMIN)?
2947 (int)(end_subpattern - start_code - GET(end_subpattern, 1)) : -1;
2948
2949 /* If we have matched an empty string, add the next state at the
2950 current character pointer. This is important so that the duplicate
2951 checking kicks in, which is what breaks infinite loops that match an
2952 empty string. */
2953
2954 if (charcount == 0)
2955 {
2956 ADD_ACTIVE(next_state_offset, 0);
2957 }
2958
2959 /* Optimization: if there are no more active states, and there
2960 are no new states yet set up, then skip over the subject string
2961 right here, to save looping. Otherwise, set up the new state to swing
2962 into action when the end of the matched substring is reached. */
2963
2964 else if (i + 1 >= active_count && new_count == 0)
2965 {
2966 ptr += charcount;
2967 clen = 0;
2968 ADD_NEW(next_state_offset, 0);
2969
2970 /* If we are adding a repeat state at the new character position,
2971 we must fudge things so that it is the only current state.
2972 Otherwise, it might be a duplicate of one we processed before, and
2973 that would cause it to be skipped. */
2974
2975 if (repeat_state_offset >= 0)
2976 {
2977 next_active_state = active_states;
2978 active_count = 0;
2979 i = -1;
2980 ADD_ACTIVE(repeat_state_offset, 0);
2981 }
2982 }
2983 else
2984 {
2985 #ifdef SUPPORT_UTF
2986 if (utf)
2987 {
2988 const pcre_uchar *p = start_subject + local_offsets[0];
2989 const pcre_uchar *pp = start_subject + local_offsets[1];
2990 while (p < pp) if (NOT_FIRSTCHAR(*p++)) charcount--;
2991 }
2992 #endif
2993 ADD_NEW_DATA(-next_state_offset, 0, (charcount - 1));
2994 if (repeat_state_offset >= 0)
2995 { ADD_NEW_DATA(-repeat_state_offset, 0, (charcount - 1)); }
2996 }
2997 }
2998 else if (rc != PCRE_ERROR_NOMATCH) return rc;
2999 }
3000 break;
3001
3002
3003 /* ========================================================================== */
3004 /* Handle callouts */
3005
3006 case OP_CALLOUT:
3007 rrc = 0;
3008 if (PUBL(callout) != NULL)
3009 {
3010 PUBL(callout_block) cb;
3011 cb.version = 1; /* Version 1 of the callout block */
3012 cb.callout_number = code[1];
3013 cb.offset_vector = offsets;
3014 #ifdef COMPILE_PCRE8
3015 cb.subject = (PCRE_SPTR)start_subject;
3016 #else
3017 cb.subject = (PCRE_SPTR16)start_subject;
3018 #endif
3019 cb.subject_length = (int)(end_subject - start_subject);
3020 cb.start_match = (int)(current_subject - start_subject);
3021 cb.current_position = (int)(ptr - start_subject);
3022 cb.pattern_position = GET(code, 2);
3023 cb.next_item_length = GET(code, 2 + LINK_SIZE);
3024 cb.capture_top = 1;
3025 cb.capture_last = -1;
3026 cb.callout_data = md->callout_data;
3027 cb.mark = NULL; /* No (*MARK) support */
3028 if ((rrc = (*PUBL(callout))(&cb)) < 0) return rrc; /* Abandon */
3029 }
3030 if (rrc == 0)
3031 { ADD_ACTIVE(state_offset + PRIV(OP_lengths)[OP_CALLOUT], 0); }
3032 break;
3033
3034
3035 /* ========================================================================== */
3036 default: /* Unsupported opcode */
3037 return PCRE_ERROR_DFA_UITEM;
3038 }
3039
3040 NEXT_ACTIVE_STATE: continue;
3041
3042 } /* End of loop scanning active states */
3043
3044 /* We have finished the processing at the current subject character. If no
3045 new states have been set for the next character, we have found all the
3046 matches that we are going to find. If we are at the top level and partial
3047 matching has been requested, check for appropriate conditions.
3048
3049 The "forced_ fail" variable counts the number of (*F) encountered for the
3050 character. If it is equal to the original active_count (saved in
3051 workspace[1]) it means that (*F) was found on every active state. In this
3052 case we don't want to give a partial match.
3053
3054 The "could_continue" variable is true if a state could have continued but
3055 for the fact that the end of the subject was reached. */
3056
3057 if (new_count <= 0)
3058 {
3059 if (rlevel == 1 && /* Top level, and */
3060 could_continue && /* Some could go on, and */
3061 forced_fail != workspace[1] && /* Not all forced fail & */
3062 ( /* either... */
3063 (md->moptions & PCRE_PARTIAL_HARD) != 0 /* Hard partial */
3064 || /* or... */
3065 ((md->moptions & PCRE_PARTIAL_SOFT) != 0 && /* Soft partial and */
3066 match_count < 0) /* no matches */
3067 ) && /* And... */
3068 (
3069 partial_newline || /* Either partial NL */
3070 ( /* or ... */
3071 ptr >= end_subject && /* End of subject and */
3072 ptr > md->start_used_ptr) /* Inspected non-empty string */
3073 )
3074 )
3075 {
3076 if (offsetcount >= 2)
3077 {
3078 offsets[0] = (int)(md->start_used_ptr - start_subject);
3079 offsets[1] = (int)(end_subject - start_subject);
3080 }
3081 match_count = PCRE_ERROR_PARTIAL;
3082 }
3083
3084 DPRINTF(("%.*sEnd of internal_dfa_exec %d: returning %d\n"
3085 "%.*s---------------------\n\n", rlevel*2-2, SP, rlevel, match_count,
3086 rlevel*2-2, SP));
3087 break; /* In effect, "return", but see the comment below */
3088 }
3089
3090 /* One or more states are active for the next character. */
3091
3092 ptr += clen; /* Advance to next subject character */
3093 } /* Loop to move along the subject string */
3094
3095 /* Control gets here from "break" a few lines above. We do it this way because
3096 if we use "return" above, we have compiler trouble. Some compilers warn if
3097 there's nothing here because they think the function doesn't return a value. On
3098 the other hand, if we put a dummy statement here, some more clever compilers
3099 complain that it can't be reached. Sigh. */
3100
3101 return match_count;
3102 }
3103
3104
3105
3106
3107 /*************************************************
3108 * Execute a Regular Expression - DFA engine *
3109 *************************************************/
3110
3111 /* This external function applies a compiled re to a subject string using a DFA
3112 engine. This function calls the internal function multiple times if the pattern
3113 is not anchored.
3114
3115 Arguments:
3116 argument_re points to the compiled expression
3117 extra_data points to extra data or is NULL
3118 subject points to the subject string
3119 length length of subject string (may contain binary zeros)
3120 start_offset where to start in the subject string
3121 options option bits
3122 offsets vector of match offsets
3123 offsetcount size of same
3124 workspace workspace vector
3125 wscount size of same
3126
3127 Returns: > 0 => number of match offset pairs placed in offsets
3128 = 0 => offsets overflowed; longest matches are present
3129 -1 => failed to match
3130 < -1 => some kind of unexpected problem
3131 */
3132
3133 #ifdef COMPILE_PCRE8
3134 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
3135 pcre_dfa_exec(const pcre *argument_re, const pcre_extra *extra_data,
3136 const char *subject, int length, int start_offset, int options, int *offsets,
3137 int offsetcount, int *workspace, int wscount)
3138 #else
3139 PCRE_EXP_DEFN int PCRE_CALL_CONVENTION
3140 pcre16_dfa_exec(const pcre16 *argument_re, const pcre16_extra *extra_data,
3141 PCRE_SPTR16 subject, int length, int start_offset, int options, int *offsets,
3142 int offsetcount, int *workspace, int wscount)
3143 #endif
3144 {
3145 REAL_PCRE *re = (REAL_PCRE *)argument_re;
3146 dfa_match_data match_block;
3147 dfa_match_data *md = &match_block;
3148 BOOL utf, anchored, startline, firstline;
3149 const pcre_uchar *current_subject, *end_subject;
3150 const pcre_study_data *study = NULL;
3151
3152 const pcre_uchar *req_char_ptr;
3153 const pcre_uint8 *start_bits = NULL;
3154 BOOL has_first_char = FALSE;
3155 BOOL has_req_char = FALSE;
3156 pcre_uchar first_char = 0;
3157 pcre_uchar first_char2 = 0;
3158 pcre_uchar req_char = 0;
3159 pcre_uchar req_char2 = 0;
3160 int newline;
3161
3162 /* Plausibility checks */
3163
3164 if ((options & ~PUBLIC_DFA_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION;
3165 if (re == NULL || subject == NULL || workspace == NULL ||
3166 (offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL;
3167 if (offsetcount < 0) return PCRE_ERROR_BADCOUNT;
3168 if (wscount < 20) return PCRE_ERROR_DFA_WSSIZE;
3169 if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET;
3170
3171 /* Check that the first field in the block is the magic number. If it is not,
3172 return with PCRE_ERROR_BADMAGIC. However, if the magic number is equal to
3173 REVERSED_MAGIC_NUMBER we return with PCRE_ERROR_BADENDIANNESS, which
3174 means that the pattern is likely compiled with different endianness. */
3175
3176 if (re->magic_number != MAGIC_NUMBER)
3177 return re->magic_number == REVERSED_MAGIC_NUMBER?
3178 PCRE_ERROR_BADENDIANNESS:PCRE_ERROR_BADMAGIC;
3179 if ((re->flags & PCRE_MODE) == 0) return PCRE_ERROR_BADMODE;
3180
3181 /* If restarting after a partial match, do some sanity checks on the contents
3182 of the workspace. */
3183
3184 if ((options & PCRE_DFA_RESTART) != 0)
3185 {
3186 if ((workspace[0] & (-2)) != 0 || workspace[1] < 1 ||
3187 workspace[1] > (wscount - 2)/INTS_PER_STATEBLOCK)
3188 return PCRE_ERROR_DFA_BADRESTART;
3189 }
3190
3191 /* Set up study, callout, and table data */
3192
3193 md->tables = re->tables;
3194 md->callout_data = NULL;
3195
3196 if (extra_data != NULL)
3197 {
3198 unsigned int flags = extra_data->flags;
3199 if ((flags & PCRE_EXTRA_STUDY_DATA) != 0)
3200 study = (const pcre_study_data *)extra_data->study_data;
3201 if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0) return PCRE_ERROR_DFA_UMLIMIT;
3202 if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0)
3203 return PCRE_ERROR_DFA_UMLIMIT;
3204 if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0)
3205 md->callout_data = extra_data->callout_data;
3206 if ((flags & PCRE_EXTRA_TABLES) != 0)
3207 md->tables = extra_data->tables;
3208 }
3209
3210 /* Set some local values */
3211
3212 current_subject = (const pcre_uchar *)subject + start_offset;
3213 end_subject = (const pcre_uchar *)subject + length;
3214 req_char_ptr = current_subject - 1;
3215
3216 #ifdef SUPPORT_UTF
3217 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
3218 utf = (re->options & PCRE_UTF8) != 0;
3219 #else
3220 utf = FALSE;
3221 #endif
3222
3223 anchored = (options & (PCRE_ANCHORED|PCRE_DFA_RESTART)) != 0 ||
3224 (re->options & PCRE_ANCHORED) != 0;
3225
3226 /* The remaining fixed data for passing around. */
3227
3228 md->start_code = (const pcre_uchar *)argument_re +
3229 re->name_table_offset + re->name_count * re->name_entry_size;
3230 md->start_subject = (const pcre_uchar *)subject;
3231 md->end_subject = end_subject;
3232 md->start_offset = start_offset;
3233 md->moptions = options;
3234 md->poptions = re->options;
3235
3236 /* If the BSR option is not set at match time, copy what was set
3237 at compile time. */
3238
3239 if ((md->moptions & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) == 0)
3240 {
3241 if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0)
3242 md->moptions |= re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE);
3243 #ifdef BSR_ANYCRLF
3244 else md->moptions |= PCRE_BSR_ANYCRLF;
3245 #endif
3246 }
3247
3248 /* Handle different types of newline. The three bits give eight cases. If
3249 nothing is set at run time, whatever was used at compile time applies. */
3250
3251 switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options : (pcre_uint32)options) &
3252 PCRE_NEWLINE_BITS)
3253 {
3254 case 0: newline = NEWLINE; break; /* Compile-time default */
3255 case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
3256 case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
3257 case PCRE_NEWLINE_CR+
3258 PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
3259 case PCRE_NEWLINE_ANY: newline = -1; break;
3260 case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
3261 default: return PCRE_ERROR_BADNEWLINE;
3262 }
3263
3264 if (newline == -2)
3265 {
3266 md->nltype = NLTYPE_ANYCRLF;
3267 }
3268 else if (newline < 0)
3269 {
3270 md->nltype = NLTYPE_ANY;
3271 }
3272 else
3273 {
3274 md->nltype = NLTYPE_FIXED;
3275 if (newline > 255)
3276 {
3277 md->nllen = 2;
3278 md->nl[0] = (newline >> 8) & 255;
3279 md->nl[1] = newline & 255;
3280 }
3281 else
3282 {
3283 md->nllen = 1;
3284 md->nl[0] = newline;
3285 }
3286 }
3287
3288 /* Check a UTF-8 string if required. Unfortunately there's no way of passing
3289 back the character offset. */
3290
3291 #ifdef SUPPORT_UTF
3292 if (utf && (options & PCRE_NO_UTF8_CHECK) == 0)
3293 {
3294 int erroroffset;
3295 int errorcode = PRIV(valid_utf)((pcre_uchar *)subject, length, &erroroffset);
3296 if (errorcode != 0)
3297 {
3298 if (offsetcount >= 2)
3299 {
3300 offsets[0] = erroroffset;
3301 offsets[1] = errorcode;
3302 }
3303 return (errorcode <= PCRE_UTF8_ERR5 && (options & PCRE_PARTIAL_HARD) != 0)?
3304 PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8;
3305 }
3306 if (start_offset > 0 && start_offset < length &&
3307 NOT_FIRSTCHAR(((PCRE_PUCHAR)subject)[start_offset]))
3308 return PCRE_ERROR_BADUTF8_OFFSET;
3309 }
3310 #endif
3311
3312 /* If the exec call supplied NULL for tables, use the inbuilt ones. This
3313 is a feature that makes it possible to save compiled regex and re-use them
3314 in other programs later. */
3315
3316 if (md->tables == NULL) md->tables = PRIV(default_tables);
3317
3318 /* The "must be at the start of a line" flags are used in a loop when finding
3319 where to start. */
3320
3321 startline = (re->flags & PCRE_STARTLINE) != 0;
3322 firstline = (re->options & PCRE_FIRSTLINE) != 0;
3323
3324 /* Set up the first character to match, if available. The first_byte value is
3325 never set for an anchored regular expression, but the anchoring may be forced
3326 at run time, so we have to test for anchoring. The first char may be unset for
3327 an unanchored pattern, of course. If there's no first char and the pattern was
3328 studied, there may be a bitmap of possible first characters. */
3329
3330 if (!anchored)
3331 {
3332 if ((re->flags & PCRE_FIRSTSET) != 0)
3333 {
3334 has_first_char = TRUE;
3335 first_char = first_char2 = (pcre_uchar)(re->first_char);
3336 if ((re->flags & PCRE_FCH_CASELESS) != 0)
3337 {
3338 first_char2 = TABLE_GET(first_char, md->tables + fcc_offset, first_char);
3339 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
3340 if (utf && first_char > 127)
3341 first_char2 = UCD_OTHERCASE(first_char);
3342 #endif
3343 }
3344 }
3345 else
3346 {
3347 if (!startline && study != NULL &&
3348 (study->flags & PCRE_STUDY_MAPPED) != 0)
3349 start_bits = study->start_bits;
3350 }
3351 }
3352
3353 /* For anchored or unanchored matches, there may be a "last known required
3354 character" set. */
3355
3356 if ((re->flags & PCRE_REQCHSET) != 0)
3357 {
3358 has_req_char = TRUE;
3359 req_char = req_char2 = (pcre_uchar)(re->req_char);
3360 if ((re->flags & PCRE_RCH_CASELESS) != 0)
3361 {
3362 req_char2 = TABLE_GET(req_char, md->tables + fcc_offset, req_char);
3363 #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
3364 if (utf && req_char > 127)
3365 req_char2 = UCD_OTHERCASE(req_char);
3366 #endif
3367 }
3368 }
3369
3370 /* Call the main matching function, looping for a non-anchored regex after a
3371 failed match. If not restarting, perform certain optimizations at the start of
3372 a match. */
3373
3374 for (;;)
3375 {
3376 int rc;
3377
3378 if ((options & PCRE_DFA_RESTART) == 0)
3379 {
3380 const pcre_uchar *save_end_subject = end_subject;
3381
3382 /* If firstline is TRUE, the start of the match is constrained to the first
3383 line of a multiline string. Implement this by temporarily adjusting
3384 end_subject so that we stop scanning at a newline. If the match fails at
3385 the newline, later code breaks this loop. */
3386
3387 if (firstline)
3388 {
3389 PCRE_PUCHAR t = current_subject;
3390 #ifdef SUPPORT_UTF
3391 if (utf)
3392 {
3393 while (t < md->end_subject && !IS_NEWLINE(t))
3394 {
3395 t++;
3396 ACROSSCHAR(t < end_subject, *t, t++);
3397 }
3398 }
3399 else
3400 #endif
3401 while (t < md->end_subject && !IS_NEWLINE(t)) t++;
3402 end_subject = t;
3403 }
3404
3405 /* There are some optimizations that avoid running the match if a known
3406 starting point is not found. However, there is an option that disables
3407 these, for testing and for ensuring that all callouts do actually occur.
3408 The option can be set in the regex by (*NO_START_OPT) or passed in
3409 match-time options. */
3410
3411 if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0)
3412 {
3413 /* Advance to a known first char. */
3414
3415 if (has_first_char)
3416 {
3417 if (first_char != first_char2)
3418 while (current_subject < end_subject &&
3419 *current_subject != first_char && *current_subject != first_char2)
3420 current_subject++;
3421 else
3422 while (current_subject < end_subject &&
3423 *current_subject != first_char)
3424 current_subject++;
3425 }
3426
3427 /* Or to just after a linebreak for a multiline match if possible */
3428
3429 else if (startline)
3430 {
3431 if (current_subject > md->start_subject + start_offset)
3432 {
3433 #ifdef SUPPORT_UTF
3434 if (utf)
3435 {
3436 while (current_subject < end_subject &&
3437 !WAS_NEWLINE(current_subject))
3438 {
3439 current_subject++;
3440 ACROSSCHAR(current_subject < end_subject, *current_subject,
3441 current_subject++);
3442 }
3443 }
3444 else
3445 #endif
3446 while (current_subject < end_subject && !WAS_NEWLINE(current_subject))
3447 current_subject++;
3448
3449 /* If we have just passed a CR and the newline option is ANY or
3450 ANYCRLF, and we are now at a LF, advance the match position by one
3451 more character. */
3452
3453 if (current_subject[-1] == CHAR_CR &&
3454 (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) &&
3455 current_subject < end_subject &&
3456 *current_subject == CHAR_NL)
3457 current_subject++;
3458 }
3459 }
3460
3461 /* Or to a non-unique first char after study */
3462
3463 else if (start_bits != NULL)
3464 {
3465 while (current_subject < end_subject)
3466 {
3467 unsigned int c = *current_subject;
3468 #ifndef COMPILE_PCRE8
3469 if (c > 255) c = 255;
3470 #endif
3471 if ((start_bits[c/8] & (1 << (c&7))) == 0)
3472 {
3473 current_subject++;
3474 #if defined SUPPORT_UTF && defined COMPILE_PCRE8
3475 /* In non 8-bit mode, the iteration will stop for
3476 characters > 255 at the beginning or not stop at all. */
3477 if (utf)
3478 ACROSSCHAR(current_subject < end_subject, *current_subject,
3479 current_subject++);
3480 #endif
3481 }
3482 else break;
3483 }
3484 }
3485 }
3486
3487 /* Restore fudged end_subject */
3488
3489 end_subject = save_end_subject;
3490
3491 /* The following two optimizations are disabled for partial matching or if
3492 disabling is explicitly requested (and of course, by the test above, this
3493 code is not obeyed when restarting after a partial match). */
3494
3495 if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0 &&
3496 (options & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) == 0)
3497 {
3498 /* If the pattern was studied, a minimum subject length may be set. This
3499 is a lower bound; no actual string of that length may actually match the
3500 pattern. Although the value is, strictly, in characters, we treat it as
3501 bytes to avoid spending too much time in this optimization. */
3502
3503 if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 &&
3504 (pcre_uint32)(end_subject - current_subject) < study->minlength)
3505 return PCRE_ERROR_NOMATCH;
3506
3507 /* If req_char is set, we know that that character must appear in the
3508 subject for the match to succeed. If the first character is set, req_char
3509 must be later in the subject; otherwise the test starts at the match
3510 point. This optimization can save a huge amount of work in patterns with
3511 nested unlimited repeats that aren't going to match. Writing separate
3512 code for cased/caseless versions makes it go faster, as does using an
3513 autoincrement and backing off on a match.
3514
3515 HOWEVER: when the subject string is very, very long, searching to its end
3516 can take a long time, and give bad performance on quite ordinary
3517 patterns. This showed up when somebody was matching /^C/ on a 32-megabyte
3518 string... so we don't do this when the string is sufficiently long. */
3519
3520 if (has_req_char && end_subject - current_subject < REQ_BYTE_MAX)
3521 {
3522 PCRE_PUCHAR p = current_subject + (has_first_char? 1:0);
3523
3524 /* We don't need to repeat the search if we haven't yet reached the
3525 place we found it at last time. */
3526
3527 if (p > req_char_ptr)
3528 {
3529 if (req_char != req_char2)
3530 {
3531 while (p < end_subject)
3532 {
3533 int pp = *p++;
3534 if (pp == req_char || pp == req_char2) { p--; break; }
3535 }
3536 }
3537 else
3538 {
3539 while (p < end_subject)
3540 {
3541 if (*p++ == req_char) { p--; break; }
3542 }
3543 }
3544
3545 /* If we can't find the required character, break the matching loop,
3546 which will cause a return or PCRE_ERROR_NOMATCH. */
3547
3548 if (p >= end_subject) break;
3549
3550 /* If we have found the required character, save the point where we
3551 found it, so that we don't search again next time round the loop if
3552 the start hasn't passed this character yet. */
3553
3554 req_char_ptr = p;
3555 }
3556 }
3557 }
3558 } /* End of optimizations that are done when not restarting */
3559
3560 /* OK, now we can do the business */
3561
3562 md->start_used_ptr = current_subject;
3563 md->recursive = NULL;
3564
3565 rc = internal_dfa_exec(
3566 md, /* fixed match data */
3567 md->start_code, /* this subexpression's code */
3568 current_subject, /* where we currently are */
3569 start_offset, /* start offset in subject */
3570 offsets, /* offset vector */
3571 offsetcount, /* size of same */
3572 workspace, /* workspace vector */
3573 wscount, /* size of same */
3574 0); /* function recurse level */
3575
3576 /* Anything other than "no match" means we are done, always; otherwise, carry
3577 on only if not anchored. */
3578
3579 if (rc != PCRE_ERROR_NOMATCH || anchored) return rc;
3580
3581 /* Advance to the next subject character unless we are at the end of a line
3582 and firstline is set. */
3583
3584 if (firstline && IS_NEWLINE(current_subject)) break;
3585 current_subject++;
3586 #ifdef SUPPORT_UTF
3587 if (utf)
3588 {
3589 ACROSSCHAR(current_subject < end_subject, *current_subject,
3590 current_subject++);
3591 }
3592 #endif
3593 if (current_subject > end_subject) break;
3594
3595 /* If we have just passed a CR and we are now at a LF, and the pattern does
3596 not contain any explicit matches for \r or \n, and the newline option is CRLF
3597 or ANY or ANYCRLF, advance the match position by one more character. */
3598
3599 if (current_subject[-1] == CHAR_CR &&
3600 current_subject < end_subject &&
3601 *current_subject == CHAR_NL &&
3602 (re->flags & PCRE_HASCRORLF) == 0 &&
3603 (md->nltype == NLTYPE_ANY ||
3604 md->nltype == NLTYPE_ANYCRLF ||
3605 md->nllen == 2))
3606 current_subject++;
3607
3608 } /* "Bumpalong" loop */
3609
3610 return PCRE_ERROR_NOMATCH;
3611 }
3612
3613 /* End of pcre_dfa_exec.c */
glib mirror