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) 1997-2017 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 | #pragma warning( disable : 4244) // conversion from 'int' to 'unsigned short', possible loss of data |
74 | #pragma warning( disable : 4146) // unary minus operator applied to unsigned type, result still unsigned |
75 | |
76 | #include "pcre_config.h" |
77 | |
78 | #define NLBLOCK md /* Block containing newline information */ |
79 | #define PSSTART start_subject /* Field containing processed string start */ |
80 | #define PSEND end_subject /* Field containing processed string end */ |
81 | |
82 | #include "pcre_internal.h" |
83 | |
84 | |
85 | /* For use to indent debugging output */ |
86 | |
87 | #define SP " " |
88 | |
89 | |
90 | /************************************************* |
91 | * Code parameters and static tables * |
92 | *************************************************/ |
93 | |
94 | /* These are offsets that are used to turn the OP_TYPESTAR and friends opcodes |
95 | into others, under special conditions. A gap of 20 between the blocks should be |
96 | enough. The resulting opcodes don't have to be less than 256 because they are |
97 | never stored, so we push them well clear of the normal opcodes. */ |
98 | |
99 | #define 300 |
100 | #define 320 |
101 | #define 340 |
102 | #define 360 |
103 | #define 380 |
104 | |
105 | |
106 | /* This table identifies those opcodes that are followed immediately by a |
107 | character that is to be tested in some way. This makes it possible to |
108 | centralize the loading of these characters. In the case of Type * etc, the |
109 | "character" is the opcode for \D, \d, \S, \s, \W, or \w, which will always be a |
110 | small value. Non-zero values in the table are the offsets from the opcode where |
111 | the character is to be found. ***NOTE*** If the start of this table is |
112 | modified, the three tables that follow must also be modified. */ |
113 | |
114 | static const pcre_uint8 coptable[] = { |
115 | 0, /* End */ |
116 | 0, 0, 0, 0, 0, /* \A, \G, \K, \B, \b */ |
117 | 0, 0, 0, 0, 0, 0, /* \D, \d, \S, \s, \W, \w */ |
118 | 0, 0, 0, /* Any, AllAny, Anybyte */ |
119 | 0, 0, /* \P, \p */ |
120 | 0, 0, 0, 0, 0, /* \R, \H, \h, \V, \v */ |
121 | 0, /* \X */ |
122 | 0, 0, 0, 0, 0, 0, /* \Z, \z, $, $M, ^, ^M */ |
123 | 1, /* Char */ |
124 | 1, /* Chari */ |
125 | 1, /* not */ |
126 | 1, /* noti */ |
127 | /* Positive single-char repeats */ |
128 | 1, 1, 1, 1, 1, 1, /* *, *?, +, +?, ?, ?? */ |
129 | 1+IMM2_SIZE, 1+IMM2_SIZE, /* upto, minupto */ |
130 | 1+IMM2_SIZE, /* exact */ |
131 | 1, 1, 1, 1+IMM2_SIZE, /* *+, ++, ?+, upto+ */ |
132 | 1, 1, 1, 1, 1, 1, /* *I, *?I, +I, +?I, ?I, ??I */ |
133 | 1+IMM2_SIZE, 1+IMM2_SIZE, /* upto I, minupto I */ |
134 | 1+IMM2_SIZE, /* exact I */ |
135 | 1, 1, 1, 1+IMM2_SIZE, /* *+I, ++I, ?+I, upto+I */ |
136 | /* Negative single-char repeats - only for chars < 256 */ |
137 | 1, 1, 1, 1, 1, 1, /* NOT *, *?, +, +?, ?, ?? */ |
138 | 1+IMM2_SIZE, 1+IMM2_SIZE, /* NOT upto, minupto */ |
139 | 1+IMM2_SIZE, /* NOT exact */ |
140 | 1, 1, 1, 1+IMM2_SIZE, /* NOT *+, ++, ?+, upto+ */ |
141 | 1, 1, 1, 1, 1, 1, /* NOT *I, *?I, +I, +?I, ?I, ??I */ |
142 | 1+IMM2_SIZE, 1+IMM2_SIZE, /* NOT upto I, minupto I */ |
143 | 1+IMM2_SIZE, /* NOT exact I */ |
144 | 1, 1, 1, 1+IMM2_SIZE, /* NOT *+I, ++I, ?+I, upto+I */ |
145 | /* Positive type repeats */ |
146 | 1, 1, 1, 1, 1, 1, /* Type *, *?, +, +?, ?, ?? */ |
147 | 1+IMM2_SIZE, 1+IMM2_SIZE, /* Type upto, minupto */ |
148 | 1+IMM2_SIZE, /* Type exact */ |
149 | 1, 1, 1, 1+IMM2_SIZE, /* Type *+, ++, ?+, upto+ */ |
150 | /* Character class & ref repeats */ |
151 | 0, 0, 0, 0, 0, 0, /* *, *?, +, +?, ?, ?? */ |
152 | 0, 0, /* CRRANGE, CRMINRANGE */ |
153 | 0, 0, 0, 0, /* Possessive *+, ++, ?+, CRPOSRANGE */ |
154 | 0, /* CLASS */ |
155 | 0, /* NCLASS */ |
156 | 0, /* XCLASS - variable length */ |
157 | 0, /* REF */ |
158 | 0, /* REFI */ |
159 | 0, /* DNREF */ |
160 | 0, /* DNREFI */ |
161 | 0, /* RECURSE */ |
162 | 0, /* CALLOUT */ |
163 | 0, /* Alt */ |
164 | 0, /* Ket */ |
165 | 0, /* KetRmax */ |
166 | 0, /* KetRmin */ |
167 | 0, /* KetRpos */ |
168 | 0, /* Reverse */ |
169 | 0, /* Assert */ |
170 | 0, /* Assert not */ |
171 | 0, /* Assert behind */ |
172 | 0, /* Assert behind not */ |
173 | 0, 0, /* ONCE, ONCE_NC */ |
174 | 0, 0, 0, 0, 0, /* BRA, BRAPOS, CBRA, CBRAPOS, COND */ |
175 | 0, 0, 0, 0, 0, /* SBRA, SBRAPOS, SCBRA, SCBRAPOS, SCOND */ |
176 | 0, 0, /* CREF, DNCREF */ |
177 | 0, 0, /* RREF, DNRREF */ |
178 | 0, /* DEF */ |
179 | 0, 0, 0, /* BRAZERO, BRAMINZERO, BRAPOSZERO */ |
180 | 0, 0, 0, /* MARK, PRUNE, PRUNE_ARG */ |
181 | 0, 0, 0, 0, /* SKIP, SKIP_ARG, THEN, THEN_ARG */ |
182 | 0, 0, 0, 0, /* COMMIT, FAIL, ACCEPT, ASSERT_ACCEPT */ |
183 | 0, 0 /* CLOSE, SKIPZERO */ |
184 | }; |
185 | |
186 | /* This table identifies those opcodes that inspect a character. It is used to |
187 | remember the fact that a character could have been inspected when the end of |
188 | the subject is reached. ***NOTE*** If the start of this table is modified, the |
189 | two tables that follow must also be modified. */ |
190 | |
191 | static const pcre_uint8 poptable[] = { |
192 | 0, /* End */ |
193 | 0, 0, 0, 1, 1, /* \A, \G, \K, \B, \b */ |
194 | 1, 1, 1, 1, 1, 1, /* \D, \d, \S, \s, \W, \w */ |
195 | 1, 1, 1, /* Any, AllAny, Anybyte */ |
196 | 1, 1, /* \P, \p */ |
197 | 1, 1, 1, 1, 1, /* \R, \H, \h, \V, \v */ |
198 | 1, /* \X */ |
199 | 0, 0, 0, 0, 0, 0, /* \Z, \z, $, $M, ^, ^M */ |
200 | 1, /* Char */ |
201 | 1, /* Chari */ |
202 | 1, /* not */ |
203 | 1, /* noti */ |
204 | /* Positive single-char repeats */ |
205 | 1, 1, 1, 1, 1, 1, /* *, *?, +, +?, ?, ?? */ |
206 | 1, 1, 1, /* upto, minupto, exact */ |
207 | 1, 1, 1, 1, /* *+, ++, ?+, upto+ */ |
208 | 1, 1, 1, 1, 1, 1, /* *I, *?I, +I, +?I, ?I, ??I */ |
209 | 1, 1, 1, /* upto I, minupto I, exact I */ |
210 | 1, 1, 1, 1, /* *+I, ++I, ?+I, upto+I */ |
211 | /* Negative single-char repeats - only for chars < 256 */ |
212 | 1, 1, 1, 1, 1, 1, /* NOT *, *?, +, +?, ?, ?? */ |
213 | 1, 1, 1, /* NOT upto, minupto, exact */ |
214 | 1, 1, 1, 1, /* NOT *+, ++, ?+, upto+ */ |
215 | 1, 1, 1, 1, 1, 1, /* NOT *I, *?I, +I, +?I, ?I, ??I */ |
216 | 1, 1, 1, /* NOT upto I, minupto I, exact I */ |
217 | 1, 1, 1, 1, /* NOT *+I, ++I, ?+I, upto+I */ |
218 | /* Positive type repeats */ |
219 | 1, 1, 1, 1, 1, 1, /* Type *, *?, +, +?, ?, ?? */ |
220 | 1, 1, 1, /* Type upto, minupto, exact */ |
221 | 1, 1, 1, 1, /* Type *+, ++, ?+, upto+ */ |
222 | /* Character class & ref repeats */ |
223 | 1, 1, 1, 1, 1, 1, /* *, *?, +, +?, ?, ?? */ |
224 | 1, 1, /* CRRANGE, CRMINRANGE */ |
225 | 1, 1, 1, 1, /* Possessive *+, ++, ?+, CRPOSRANGE */ |
226 | 1, /* CLASS */ |
227 | 1, /* NCLASS */ |
228 | 1, /* XCLASS - variable length */ |
229 | 0, /* REF */ |
230 | 0, /* REFI */ |
231 | 0, /* DNREF */ |
232 | 0, /* DNREFI */ |
233 | 0, /* RECURSE */ |
234 | 0, /* CALLOUT */ |
235 | 0, /* Alt */ |
236 | 0, /* Ket */ |
237 | 0, /* KetRmax */ |
238 | 0, /* KetRmin */ |
239 | 0, /* KetRpos */ |
240 | 0, /* Reverse */ |
241 | 0, /* Assert */ |
242 | 0, /* Assert not */ |
243 | 0, /* Assert behind */ |
244 | 0, /* Assert behind not */ |
245 | 0, 0, /* ONCE, ONCE_NC */ |
246 | 0, 0, 0, 0, 0, /* BRA, BRAPOS, CBRA, CBRAPOS, COND */ |
247 | 0, 0, 0, 0, 0, /* SBRA, SBRAPOS, SCBRA, SCBRAPOS, SCOND */ |
248 | 0, 0, /* CREF, DNCREF */ |
249 | 0, 0, /* RREF, DNRREF */ |
250 | 0, /* DEF */ |
251 | 0, 0, 0, /* BRAZERO, BRAMINZERO, BRAPOSZERO */ |
252 | 0, 0, 0, /* MARK, PRUNE, PRUNE_ARG */ |
253 | 0, 0, 0, 0, /* SKIP, SKIP_ARG, THEN, THEN_ARG */ |
254 | 0, 0, 0, 0, /* COMMIT, FAIL, ACCEPT, ASSERT_ACCEPT */ |
255 | 0, 0 /* CLOSE, SKIPZERO */ |
256 | }; |
257 | |
258 | /* These 2 tables allow for compact code for testing for \D, \d, \S, \s, \W, |
259 | and \w */ |
260 | |
261 | static const pcre_uint8 toptable1[] = { |
262 | 0, 0, 0, 0, 0, 0, |
263 | ctype_digit, ctype_digit, |
264 | ctype_space, ctype_space, |
265 | ctype_word, ctype_word, |
266 | 0, 0 /* OP_ANY, OP_ALLANY */ |
267 | }; |
268 | |
269 | static const pcre_uint8 toptable2[] = { |
270 | 0, 0, 0, 0, 0, 0, |
271 | ctype_digit, 0, |
272 | ctype_space, 0, |
273 | ctype_word, 0, |
274 | 1, 1 /* OP_ANY, OP_ALLANY */ |
275 | }; |
276 | |
277 | |
278 | /* Structure for holding data about a particular state, which is in effect the |
279 | current data for an active path through the match tree. It must consist |
280 | entirely of ints because the working vector we are passed, and which we put |
281 | these structures in, is a vector of ints. */ |
282 | |
283 | typedef struct stateblock { |
284 | int offset; /* Offset to opcode */ |
285 | int count; /* Count for repeats */ |
286 | int data; /* Some use extra data */ |
287 | } stateblock; |
288 | |
289 | #define INTS_PER_STATEBLOCK (int)(sizeof(stateblock)/sizeof(int)) |
290 | |
291 | |
292 | #ifdef PCRE_DEBUG |
293 | /************************************************* |
294 | * Print character string * |
295 | *************************************************/ |
296 | |
297 | /* Character string printing function for debugging. |
298 | |
299 | Arguments: |
300 | p points to string |
301 | length number of bytes |
302 | f where to print |
303 | |
304 | Returns: nothing |
305 | */ |
306 | |
307 | static void |
308 | pchars(const pcre_uchar *p, int length, FILE *f) |
309 | { |
310 | pcre_uint32 c; |
311 | while (length-- > 0) |
312 | { |
313 | if (isprint(c = *(p++))) |
314 | fprintf(f, "%c" , c); |
315 | else |
316 | fprintf(f, "\\x{%02x}" , c); |
317 | } |
318 | } |
319 | #endif |
320 | |
321 | |
322 | |
323 | /************************************************* |
324 | * Execute a Regular Expression - DFA engine * |
325 | *************************************************/ |
326 | |
327 | /* This internal function applies a compiled pattern to a subject string, |
328 | starting at a given point, using a DFA engine. This function is called from the |
329 | external one, possibly multiple times if the pattern is not anchored. The |
330 | function calls itself recursively for some kinds of subpattern. |
331 | |
332 | Arguments: |
333 | md the match_data block with fixed information |
334 | this_start_code the opening bracket of this subexpression's code |
335 | current_subject where we currently are in the subject string |
336 | start_offset start offset in the subject string |
337 | offsets vector to contain the matching string offsets |
338 | offsetcount size of same |
339 | workspace vector of workspace |
340 | wscount size of same |
341 | rlevel function call recursion level |
342 | |
343 | Returns: > 0 => number of match offset pairs placed in offsets |
344 | = 0 => offsets overflowed; longest matches are present |
345 | -1 => failed to match |
346 | < -1 => some kind of unexpected problem |
347 | |
348 | The following macros are used for adding states to the two state vectors (one |
349 | for the current character, one for the following character). */ |
350 | |
351 | #define ADD_ACTIVE(x,y) \ |
352 | if (active_count++ < wscount) \ |
353 | { \ |
354 | next_active_state->offset = (x); \ |
355 | next_active_state->count = (y); \ |
356 | next_active_state++; \ |
357 | DPRINTF(("%.*sADD_ACTIVE(%d,%d)\n", rlevel*2-2, SP, (x), (y))); \ |
358 | } \ |
359 | else return PCRE_ERROR_DFA_WSSIZE |
360 | |
361 | #define ADD_ACTIVE_DATA(x,y,z) \ |
362 | if (active_count++ < wscount) \ |
363 | { \ |
364 | next_active_state->offset = (x); \ |
365 | next_active_state->count = (y); \ |
366 | next_active_state->data = (z); \ |
367 | next_active_state++; \ |
368 | DPRINTF(("%.*sADD_ACTIVE_DATA(%d,%d,%d)\n", rlevel*2-2, SP, (x), (y), (z))); \ |
369 | } \ |
370 | else return PCRE_ERROR_DFA_WSSIZE |
371 | |
372 | #define ADD_NEW(x,y) \ |
373 | if (new_count++ < wscount) \ |
374 | { \ |
375 | next_new_state->offset = (x); \ |
376 | next_new_state->count = (y); \ |
377 | next_new_state++; \ |
378 | DPRINTF(("%.*sADD_NEW(%d,%d)\n", rlevel*2-2, SP, (x), (y))); \ |
379 | } \ |
380 | else return PCRE_ERROR_DFA_WSSIZE |
381 | |
382 | #define ADD_NEW_DATA(x,y,z) \ |
383 | if (new_count++ < wscount) \ |
384 | { \ |
385 | next_new_state->offset = (x); \ |
386 | next_new_state->count = (y); \ |
387 | next_new_state->data = (z); \ |
388 | next_new_state++; \ |
389 | DPRINTF(("%.*sADD_NEW_DATA(%d,%d,%d) line %d\n", rlevel*2-2, SP, \ |
390 | (x), (y), (z), __LINE__)); \ |
391 | } \ |
392 | else return PCRE_ERROR_DFA_WSSIZE |
393 | |
394 | /* And now, here is the code */ |
395 | |
396 | static int |
397 | internal_dfa_exec( |
398 | dfa_match_data *md, |
399 | const pcre_uchar *this_start_code, |
400 | const pcre_uchar *current_subject, |
401 | int start_offset, |
402 | int *offsets, |
403 | int offsetcount, |
404 | int *workspace, |
405 | int wscount, |
406 | int rlevel) |
407 | { |
408 | stateblock *active_states, *new_states, *temp_states; |
409 | stateblock *next_active_state, *next_new_state; |
410 | |
411 | const pcre_uint8 *ctypes, *lcc, *fcc; |
412 | const pcre_uchar *ptr; |
413 | const pcre_uchar *end_code, *first_op; |
414 | |
415 | dfa_recursion_info new_recursive; |
416 | |
417 | int active_count, new_count, match_count; |
418 | |
419 | /* Some fields in the md block are frequently referenced, so we load them into |
420 | independent variables in the hope that this will perform better. */ |
421 | |
422 | const pcre_uchar *start_subject = md->start_subject; |
423 | const pcre_uchar *end_subject = md->end_subject; |
424 | const pcre_uchar *start_code = md->start_code; |
425 | |
426 | #ifdef SUPPORT_UTF |
427 | BOOL utf = (md->poptions & PCRE_UTF8) != 0; |
428 | #else |
429 | BOOL utf = FALSE; |
430 | #endif |
431 | |
432 | BOOL reset_could_continue = FALSE; |
433 | |
434 | rlevel++; |
435 | offsetcount &= (-2); |
436 | |
437 | wscount -= 2; |
438 | wscount = (wscount - (wscount % (INTS_PER_STATEBLOCK * 2))) / |
439 | (2 * INTS_PER_STATEBLOCK); |
440 | |
441 | DPRINTF(("\n%.*s---------------------\n" |
442 | "%.*sCall to internal_dfa_exec f=%d\n" , |
443 | rlevel*2-2, SP, rlevel*2-2, SP, rlevel)); |
444 | |
445 | ctypes = md->tables + ctypes_offset; |
446 | lcc = md->tables + lcc_offset; |
447 | fcc = md->tables + fcc_offset; |
448 | |
449 | match_count = PCRE_ERROR_NOMATCH; /* A negative number */ |
450 | |
451 | active_states = (stateblock *)(workspace + 2); |
452 | next_new_state = new_states = active_states + wscount; |
453 | new_count = 0; |
454 | |
455 | first_op = this_start_code + 1 + LINK_SIZE + |
456 | ((*this_start_code == OP_CBRA || *this_start_code == OP_SCBRA || |
457 | *this_start_code == OP_CBRAPOS || *this_start_code == OP_SCBRAPOS) |
458 | ? IMM2_SIZE:0); |
459 | |
460 | /* The first thing in any (sub) pattern is a bracket of some sort. Push all |
461 | the alternative states onto the list, and find out where the end is. This |
462 | makes is possible to use this function recursively, when we want to stop at a |
463 | matching internal ket rather than at the end. |
464 | |
465 | If the first opcode in the first alternative is OP_REVERSE, we are dealing with |
466 | a backward assertion. In that case, we have to find out the maximum amount to |
467 | move back, and set up each alternative appropriately. */ |
468 | |
469 | if (*first_op == OP_REVERSE) |
470 | { |
471 | int max_back = 0; |
472 | int gone_back; |
473 | |
474 | end_code = this_start_code; |
475 | do |
476 | { |
477 | int back = GET(end_code, 2+LINK_SIZE); |
478 | if (back > max_back) max_back = back; |
479 | end_code += GET(end_code, 1); |
480 | } |
481 | while (*end_code == OP_ALT); |
482 | |
483 | /* If we can't go back the amount required for the longest lookbehind |
484 | pattern, go back as far as we can; some alternatives may still be viable. */ |
485 | |
486 | #ifdef SUPPORT_UTF |
487 | /* In character mode we have to step back character by character */ |
488 | |
489 | if (utf) |
490 | { |
491 | for (gone_back = 0; gone_back < max_back; gone_back++) |
492 | { |
493 | if (current_subject <= start_subject) break; |
494 | current_subject--; |
495 | ACROSSCHAR(current_subject > start_subject, *current_subject, current_subject--); |
496 | } |
497 | } |
498 | else |
499 | #endif |
500 | |
501 | /* In byte-mode we can do this quickly. */ |
502 | |
503 | { |
504 | gone_back = (current_subject - max_back < start_subject)? |
505 | (int)(current_subject - start_subject) : max_back; |
506 | current_subject -= gone_back; |
507 | } |
508 | |
509 | /* Save the earliest consulted character */ |
510 | |
511 | if (current_subject < md->start_used_ptr) |
512 | md->start_used_ptr = current_subject; |
513 | |
514 | /* Now we can process the individual branches. */ |
515 | |
516 | end_code = this_start_code; |
517 | do |
518 | { |
519 | int back = GET(end_code, 2+LINK_SIZE); |
520 | if (back <= gone_back) |
521 | { |
522 | int bstate = (int)(end_code - start_code + 2 + 2*LINK_SIZE); |
523 | ADD_NEW_DATA(-bstate, 0, gone_back - back); |
524 | } |
525 | end_code += GET(end_code, 1); |
526 | } |
527 | while (*end_code == OP_ALT); |
528 | } |
529 | |
530 | /* This is the code for a "normal" subpattern (not a backward assertion). The |
531 | start of a whole pattern is always one of these. If we are at the top level, |
532 | we may be asked to restart matching from the same point that we reached for a |
533 | previous partial match. We still have to scan through the top-level branches to |
534 | find the end state. */ |
535 | |
536 | else |
537 | { |
538 | end_code = this_start_code; |
539 | |
540 | /* Restarting */ |
541 | |
542 | if (rlevel == 1 && (md->moptions & PCRE_DFA_RESTART) != 0) |
543 | { |
544 | do { end_code += GET(end_code, 1); } while (*end_code == OP_ALT); |
545 | new_count = workspace[1]; |
546 | if (!workspace[0]) |
547 | memcpy(new_states, active_states, new_count * sizeof(stateblock)); |
548 | } |
549 | |
550 | /* Not restarting */ |
551 | |
552 | else |
553 | { |
554 | int length = 1 + LINK_SIZE + |
555 | ((*this_start_code == OP_CBRA || *this_start_code == OP_SCBRA || |
556 | *this_start_code == OP_CBRAPOS || *this_start_code == OP_SCBRAPOS) |
557 | ? IMM2_SIZE:0); |
558 | do |
559 | { |
560 | ADD_NEW((int)(end_code - start_code + length), 0); |
561 | end_code += GET(end_code, 1); |
562 | length = 1 + LINK_SIZE; |
563 | } |
564 | while (*end_code == OP_ALT); |
565 | } |
566 | } |
567 | |
568 | workspace[0] = 0; /* Bit indicating which vector is current */ |
569 | |
570 | DPRINTF(("%.*sEnd state = %d\n" , rlevel*2-2, SP, (int)(end_code - start_code))); |
571 | |
572 | /* Loop for scanning the subject */ |
573 | |
574 | ptr = current_subject; |
575 | for (;;) |
576 | { |
577 | int i, j; |
578 | int clen, dlen; |
579 | pcre_uint32 c, d; |
580 | int forced_fail = 0; |
581 | BOOL partial_newline = FALSE; |
582 | BOOL could_continue = reset_could_continue; |
583 | reset_could_continue = FALSE; |
584 | |
585 | /* Make the new state list into the active state list and empty the |
586 | new state list. */ |
587 | |
588 | temp_states = active_states; |
589 | active_states = new_states; |
590 | new_states = temp_states; |
591 | active_count = new_count; |
592 | new_count = 0; |
593 | |
594 | workspace[0] ^= 1; /* Remember for the restarting feature */ |
595 | workspace[1] = active_count; |
596 | |
597 | #ifdef PCRE_DEBUG |
598 | printf("%.*sNext character: rest of subject = \"" , rlevel*2-2, SP); |
599 | pchars(ptr, STRLEN_UC(ptr), stdout); |
600 | printf("\"\n" ); |
601 | |
602 | printf("%.*sActive states: " , rlevel*2-2, SP); |
603 | for (i = 0; i < active_count; i++) |
604 | printf("%d/%d " , active_states[i].offset, active_states[i].count); |
605 | printf("\n" ); |
606 | #endif |
607 | |
608 | /* Set the pointers for adding new states */ |
609 | |
610 | next_active_state = active_states + active_count; |
611 | next_new_state = new_states; |
612 | |
613 | /* Load the current character from the subject outside the loop, as many |
614 | different states may want to look at it, and we assume that at least one |
615 | will. */ |
616 | |
617 | if (ptr < end_subject) |
618 | { |
619 | clen = 1; /* Number of data items in the character */ |
620 | #ifdef SUPPORT_UTF |
621 | GETCHARLENTEST(c, ptr, clen); |
622 | #else |
623 | c = *ptr; |
624 | #endif /* SUPPORT_UTF */ |
625 | } |
626 | else |
627 | { |
628 | clen = 0; /* This indicates the end of the subject */ |
629 | c = NOTACHAR; /* This value should never actually be used */ |
630 | } |
631 | |
632 | /* Scan up the active states and act on each one. The result of an action |
633 | may be to add more states to the currently active list (e.g. on hitting a |
634 | parenthesis) or it may be to put states on the new list, for considering |
635 | when we move the character pointer on. */ |
636 | |
637 | for (i = 0; i < active_count; i++) |
638 | { |
639 | stateblock *current_state = active_states + i; |
640 | BOOL caseless = FALSE; |
641 | const pcre_uchar *code; |
642 | int state_offset = current_state->offset; |
643 | int codevalue, rrc; |
644 | int count; |
645 | |
646 | #ifdef PCRE_DEBUG |
647 | printf ("%.*sProcessing state %d c=" , rlevel*2-2, SP, state_offset); |
648 | if (clen == 0) printf("EOL\n" ); |
649 | else if (c > 32 && c < 127) printf("'%c'\n" , c); |
650 | else printf("0x%02x\n" , c); |
651 | #endif |
652 | |
653 | /* A negative offset is a special case meaning "hold off going to this |
654 | (negated) state until the number of characters in the data field have |
655 | been skipped". If the could_continue flag was passed over from a previous |
656 | state, arrange for it to passed on. */ |
657 | |
658 | if (state_offset < 0) |
659 | { |
660 | if (current_state->data > 0) |
661 | { |
662 | DPRINTF(("%.*sSkipping this character\n" , rlevel*2-2, SP)); |
663 | ADD_NEW_DATA(state_offset, current_state->count, |
664 | current_state->data - 1); |
665 | if (could_continue) reset_could_continue = TRUE; |
666 | continue; |
667 | } |
668 | else |
669 | { |
670 | current_state->offset = state_offset = -state_offset; |
671 | } |
672 | } |
673 | |
674 | /* Check for a duplicate state with the same count, and skip if found. |
675 | See the note at the head of this module about the possibility of improving |
676 | performance here. */ |
677 | |
678 | for (j = 0; j < i; j++) |
679 | { |
680 | if (active_states[j].offset == state_offset && |
681 | active_states[j].count == current_state->count) |
682 | { |
683 | DPRINTF(("%.*sDuplicate state: skipped\n" , rlevel*2-2, SP)); |
684 | goto NEXT_ACTIVE_STATE; |
685 | } |
686 | } |
687 | |
688 | /* The state offset is the offset to the opcode */ |
689 | |
690 | code = start_code + state_offset; |
691 | codevalue = *code; |
692 | |
693 | /* If this opcode inspects a character, but we are at the end of the |
694 | subject, remember the fact for use when testing for a partial match. */ |
695 | |
696 | if (clen == 0 && poptable[codevalue] != 0) |
697 | could_continue = TRUE; |
698 | |
699 | /* If this opcode is followed by an inline character, load it. It is |
700 | tempting to test for the presence of a subject character here, but that |
701 | is wrong, because sometimes zero repetitions of the subject are |
702 | permitted. |
703 | |
704 | We also use this mechanism for opcodes such as OP_TYPEPLUS that take an |
705 | argument that is not a data character - but is always one byte long because |
706 | the values are small. We have to take special action to deal with \P, \p, |
707 | \H, \h, \V, \v and \X in this case. To keep the other cases fast, convert |
708 | these ones to new opcodes. */ |
709 | |
710 | if (coptable[codevalue] > 0) |
711 | { |
712 | dlen = 1; |
713 | #ifdef SUPPORT_UTF |
714 | if (utf) { GETCHARLEN(d, (code + coptable[codevalue]), dlen); } else |
715 | #endif /* SUPPORT_UTF */ |
716 | d = code[coptable[codevalue]]; |
717 | if (codevalue >= OP_TYPESTAR) |
718 | { |
719 | switch(d) |
720 | { |
721 | case OP_ANYBYTE: return PCRE_ERROR_DFA_UITEM; |
722 | case OP_NOTPROP: |
723 | case OP_PROP: codevalue += OP_PROP_EXTRA; break; |
724 | case OP_ANYNL: codevalue += OP_ANYNL_EXTRA; break; |
725 | case OP_EXTUNI: codevalue += OP_EXTUNI_EXTRA; break; |
726 | case OP_NOT_HSPACE: |
727 | case OP_HSPACE: codevalue += OP_HSPACE_EXTRA; break; |
728 | case OP_NOT_VSPACE: |
729 | case OP_VSPACE: codevalue += OP_VSPACE_EXTRA; break; |
730 | default: break; |
731 | } |
732 | } |
733 | } |
734 | else |
735 | { |
736 | dlen = 0; /* Not strictly necessary, but compilers moan */ |
737 | d = NOTACHAR; /* if these variables are not set. */ |
738 | } |
739 | |
740 | |
741 | /* Now process the individual opcodes */ |
742 | |
743 | switch (codevalue) |
744 | { |
745 | /* ========================================================================== */ |
746 | /* These cases are never obeyed. This is a fudge that causes a compile- |
747 | time error if the vectors coptable or poptable, which are indexed by |
748 | opcode, are not the correct length. It seems to be the only way to do |
749 | such a check at compile time, as the sizeof() operator does not work |
750 | in the C preprocessor. */ |
751 | |
752 | case OP_TABLE_LENGTH: |
753 | case OP_TABLE_LENGTH + |
754 | ((sizeof(coptable) == OP_TABLE_LENGTH) && |
755 | (sizeof(poptable) == OP_TABLE_LENGTH)): |
756 | break; |
757 | |
758 | /* ========================================================================== */ |
759 | /* Reached a closing bracket. If not at the end of the pattern, carry |
760 | on with the next opcode. For repeating opcodes, also add the repeat |
761 | state. Note that KETRPOS will always be encountered at the end of the |
762 | subpattern, because the possessive subpattern repeats are always handled |
763 | using recursive calls. Thus, it never adds any new states. |
764 | |
765 | At the end of the (sub)pattern, unless we have an empty string and |
766 | PCRE_NOTEMPTY is set, or PCRE_NOTEMPTY_ATSTART is set and we are at the |
767 | start of the subject, save the match data, shifting up all previous |
768 | matches so we always have the longest first. */ |
769 | |
770 | case OP_KET: |
771 | case OP_KETRMIN: |
772 | case OP_KETRMAX: |
773 | case OP_KETRPOS: |
774 | if (code != end_code) |
775 | { |
776 | ADD_ACTIVE(state_offset + 1 + LINK_SIZE, 0); |
777 | if (codevalue != OP_KET) |
778 | { |
779 | ADD_ACTIVE(state_offset - GET(code, 1), 0); |
780 | } |
781 | } |
782 | else |
783 | { |
784 | if (ptr > current_subject || |
785 | ((md->moptions & PCRE_NOTEMPTY) == 0 && |
786 | ((md->moptions & PCRE_NOTEMPTY_ATSTART) == 0 || |
787 | current_subject > start_subject + md->start_offset))) |
788 | { |
789 | if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0; |
790 | else if (match_count > 0 && ++match_count * 2 > offsetcount) |
791 | match_count = 0; |
792 | count = ((match_count == 0)? offsetcount : match_count * 2) - 2; |
793 | if (count > 0) memmove(offsets + 2, offsets, count * sizeof(int)); |
794 | if (offsetcount >= 2) |
795 | { |
796 | offsets[0] = (int)(current_subject - start_subject); |
797 | offsets[1] = (int)(ptr - start_subject); |
798 | DPRINTF(("%.*sSet matched string = \"%.*s\"\n" , rlevel*2-2, SP, |
799 | offsets[1] - offsets[0], (char *)current_subject)); |
800 | } |
801 | if ((md->moptions & PCRE_DFA_SHORTEST) != 0) |
802 | { |
803 | DPRINTF(("%.*sEnd of internal_dfa_exec %d: returning %d\n" |
804 | "%.*s---------------------\n\n" , rlevel*2-2, SP, rlevel, |
805 | match_count, rlevel*2-2, SP)); |
806 | return match_count; |
807 | } |
808 | } |
809 | } |
810 | break; |
811 | |
812 | /* ========================================================================== */ |
813 | /* These opcodes add to the current list of states without looking |
814 | at the current character. */ |
815 | |
816 | /*-----------------------------------------------------------------*/ |
817 | case OP_ALT: |
818 | do { code += GET(code, 1); } while (*code == OP_ALT); |
819 | ADD_ACTIVE((int)(code - start_code), 0); |
820 | break; |
821 | |
822 | /*-----------------------------------------------------------------*/ |
823 | case OP_BRA: |
824 | case OP_SBRA: |
825 | do |
826 | { |
827 | ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0); |
828 | code += GET(code, 1); |
829 | } |
830 | while (*code == OP_ALT); |
831 | break; |
832 | |
833 | /*-----------------------------------------------------------------*/ |
834 | case OP_CBRA: |
835 | case OP_SCBRA: |
836 | ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE + IMM2_SIZE), 0); |
837 | code += GET(code, 1); |
838 | while (*code == OP_ALT) |
839 | { |
840 | ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0); |
841 | code += GET(code, 1); |
842 | } |
843 | break; |
844 | |
845 | /*-----------------------------------------------------------------*/ |
846 | case OP_BRAZERO: |
847 | case OP_BRAMINZERO: |
848 | ADD_ACTIVE(state_offset + 1, 0); |
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_SKIPZERO: |
856 | code += 1 + GET(code, 2); |
857 | while (*code == OP_ALT) code += GET(code, 1); |
858 | ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0); |
859 | break; |
860 | |
861 | /*-----------------------------------------------------------------*/ |
862 | case OP_CIRC: |
863 | if (ptr == start_subject && (md->moptions & PCRE_NOTBOL) == 0) |
864 | { ADD_ACTIVE(state_offset + 1, 0); } |
865 | break; |
866 | |
867 | /*-----------------------------------------------------------------*/ |
868 | case OP_CIRCM: |
869 | if ((ptr == start_subject && (md->moptions & PCRE_NOTBOL) == 0) || |
870 | (ptr != end_subject && WAS_NEWLINE(ptr))) |
871 | { ADD_ACTIVE(state_offset + 1, 0); } |
872 | break; |
873 | |
874 | /*-----------------------------------------------------------------*/ |
875 | case OP_EOD: |
876 | if (ptr >= end_subject) |
877 | { |
878 | if ((md->moptions & PCRE_PARTIAL_HARD) != 0) |
879 | could_continue = TRUE; |
880 | else { ADD_ACTIVE(state_offset + 1, 0); } |
881 | } |
882 | break; |
883 | |
884 | /*-----------------------------------------------------------------*/ |
885 | case OP_SOD: |
886 | if (ptr == start_subject) { ADD_ACTIVE(state_offset + 1, 0); } |
887 | break; |
888 | |
889 | /*-----------------------------------------------------------------*/ |
890 | case OP_SOM: |
891 | if (ptr == start_subject + start_offset) { ADD_ACTIVE(state_offset + 1, 0); } |
892 | break; |
893 | |
894 | |
895 | /* ========================================================================== */ |
896 | /* These opcodes inspect the next subject character, and sometimes |
897 | the previous one as well, but do not have an argument. The variable |
898 | clen contains the length of the current character and is zero if we are |
899 | at the end of the subject. */ |
900 | |
901 | /*-----------------------------------------------------------------*/ |
902 | case OP_ANY: |
903 | if (clen > 0 && !IS_NEWLINE(ptr)) |
904 | { |
905 | if (ptr + 1 >= md->end_subject && |
906 | (md->moptions & (PCRE_PARTIAL_HARD)) != 0 && |
907 | NLBLOCK->nltype == NLTYPE_FIXED && |
908 | NLBLOCK->nllen == 2 && |
909 | c == NLBLOCK->nl[0]) |
910 | { |
911 | could_continue = partial_newline = TRUE; |
912 | } |
913 | else |
914 | { |
915 | ADD_NEW(state_offset + 1, 0); |
916 | } |
917 | } |
918 | break; |
919 | |
920 | /*-----------------------------------------------------------------*/ |
921 | case OP_ALLANY: |
922 | if (clen > 0) |
923 | { ADD_NEW(state_offset + 1, 0); } |
924 | break; |
925 | |
926 | /*-----------------------------------------------------------------*/ |
927 | case OP_EODN: |
928 | if (clen == 0 && (md->moptions & PCRE_PARTIAL_HARD) != 0) |
929 | could_continue = TRUE; |
930 | else if (clen == 0 || (IS_NEWLINE(ptr) && ptr == end_subject - md->nllen)) |
931 | { ADD_ACTIVE(state_offset + 1, 0); } |
932 | break; |
933 | |
934 | /*-----------------------------------------------------------------*/ |
935 | case OP_DOLL: |
936 | if ((md->moptions & PCRE_NOTEOL) == 0) |
937 | { |
938 | if (clen == 0 && (md->moptions & PCRE_PARTIAL_HARD) != 0) |
939 | could_continue = TRUE; |
940 | else if (clen == 0 || |
941 | ((md->poptions & PCRE_DOLLAR_ENDONLY) == 0 && IS_NEWLINE(ptr) && |
942 | (ptr == end_subject - md->nllen) |
943 | )) |
944 | { ADD_ACTIVE(state_offset + 1, 0); } |
945 | else if (ptr + 1 >= md->end_subject && |
946 | (md->moptions & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) != 0 && |
947 | NLBLOCK->nltype == NLTYPE_FIXED && |
948 | NLBLOCK->nllen == 2 && |
949 | c == NLBLOCK->nl[0]) |
950 | { |
951 | if ((md->moptions & PCRE_PARTIAL_HARD) != 0) |
952 | { |
953 | reset_could_continue = TRUE; |
954 | ADD_NEW_DATA(-(state_offset + 1), 0, 1); |
955 | } |
956 | else could_continue = partial_newline = TRUE; |
957 | } |
958 | } |
959 | break; |
960 | |
961 | /*-----------------------------------------------------------------*/ |
962 | case OP_DOLLM: |
963 | if ((md->moptions & PCRE_NOTEOL) == 0) |
964 | { |
965 | if (clen == 0 && (md->moptions & PCRE_PARTIAL_HARD) != 0) |
966 | could_continue = TRUE; |
967 | else if (clen == 0 || |
968 | ((md->poptions & PCRE_DOLLAR_ENDONLY) == 0 && IS_NEWLINE(ptr))) |
969 | { ADD_ACTIVE(state_offset + 1, 0); } |
970 | else if (ptr + 1 >= md->end_subject && |
971 | (md->moptions & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) != 0 && |
972 | NLBLOCK->nltype == NLTYPE_FIXED && |
973 | NLBLOCK->nllen == 2 && |
974 | c == NLBLOCK->nl[0]) |
975 | { |
976 | if ((md->moptions & PCRE_PARTIAL_HARD) != 0) |
977 | { |
978 | reset_could_continue = TRUE; |
979 | ADD_NEW_DATA(-(state_offset + 1), 0, 1); |
980 | } |
981 | else could_continue = partial_newline = TRUE; |
982 | } |
983 | } |
984 | else if (IS_NEWLINE(ptr)) |
985 | { ADD_ACTIVE(state_offset + 1, 0); } |
986 | break; |
987 | |
988 | /*-----------------------------------------------------------------*/ |
989 | |
990 | case OP_DIGIT: |
991 | case OP_WHITESPACE: |
992 | case OP_WORDCHAR: |
993 | if (clen > 0 && c < 256 && |
994 | ((ctypes[c] & toptable1[codevalue]) ^ toptable2[codevalue]) != 0) |
995 | { ADD_NEW(state_offset + 1, 0); } |
996 | break; |
997 | |
998 | /*-----------------------------------------------------------------*/ |
999 | case OP_NOT_DIGIT: |
1000 | case OP_NOT_WHITESPACE: |
1001 | case OP_NOT_WORDCHAR: |
1002 | if (clen > 0 && (c >= 256 || |
1003 | ((ctypes[c] & toptable1[codevalue]) ^ toptable2[codevalue]) != 0)) |
1004 | { ADD_NEW(state_offset + 1, 0); } |
1005 | break; |
1006 | |
1007 | /*-----------------------------------------------------------------*/ |
1008 | case OP_WORD_BOUNDARY: |
1009 | case OP_NOT_WORD_BOUNDARY: |
1010 | { |
1011 | int left_word, right_word; |
1012 | |
1013 | if (ptr > start_subject) |
1014 | { |
1015 | const pcre_uchar *temp = ptr - 1; |
1016 | if (temp < md->start_used_ptr) md->start_used_ptr = temp; |
1017 | #if defined SUPPORT_UTF && !defined COMPILE_PCRE32 |
1018 | if (utf) { BACKCHAR(temp); } |
1019 | #endif |
1020 | GETCHARTEST(d, temp); |
1021 | #ifdef SUPPORT_UCP |
1022 | if ((md->poptions & PCRE_UCP) != 0) |
1023 | { |
1024 | if (d == '_') left_word = TRUE; else |
1025 | { |
1026 | int cat = UCD_CATEGORY(d); |
1027 | left_word = (cat == ucp_L || cat == ucp_N); |
1028 | } |
1029 | } |
1030 | else |
1031 | #endif |
1032 | left_word = d < 256 && (ctypes[d] & ctype_word) != 0; |
1033 | } |
1034 | else left_word = FALSE; |
1035 | |
1036 | if (clen > 0) |
1037 | { |
1038 | #ifdef SUPPORT_UCP |
1039 | if ((md->poptions & PCRE_UCP) != 0) |
1040 | { |
1041 | if (c == '_') right_word = TRUE; else |
1042 | { |
1043 | int cat = UCD_CATEGORY(c); |
1044 | right_word = (cat == ucp_L || cat == ucp_N); |
1045 | } |
1046 | } |
1047 | else |
1048 | #endif |
1049 | right_word = c < 256 && (ctypes[c] & ctype_word) != 0; |
1050 | } |
1051 | else right_word = FALSE; |
1052 | |
1053 | if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY)) |
1054 | { ADD_ACTIVE(state_offset + 1, 0); } |
1055 | } |
1056 | break; |
1057 | |
1058 | |
1059 | /*-----------------------------------------------------------------*/ |
1060 | /* Check the next character by Unicode property. We will get here only |
1061 | if the support is in the binary; otherwise a compile-time error occurs. |
1062 | */ |
1063 | |
1064 | #ifdef SUPPORT_UCP |
1065 | case OP_PROP: |
1066 | case OP_NOTPROP: |
1067 | if (clen > 0) |
1068 | { |
1069 | BOOL OK; |
1070 | const pcre_uint32 *cp; |
1071 | const ucd_record * prop = GET_UCD(c); |
1072 | switch(code[1]) |
1073 | { |
1074 | case PT_ANY: |
1075 | OK = TRUE; |
1076 | break; |
1077 | |
1078 | case PT_LAMP: |
1079 | OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || |
1080 | prop->chartype == ucp_Lt; |
1081 | break; |
1082 | |
1083 | case PT_GC: |
1084 | OK = PRIV(ucp_gentype)[prop->chartype] == code[2]; |
1085 | break; |
1086 | |
1087 | case PT_PC: |
1088 | OK = prop->chartype == code[2]; |
1089 | break; |
1090 | |
1091 | case PT_SC: |
1092 | OK = prop->script == code[2]; |
1093 | break; |
1094 | |
1095 | /* These are specials for combination cases. */ |
1096 | |
1097 | case PT_ALNUM: |
1098 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1099 | PRIV(ucp_gentype)[prop->chartype] == ucp_N; |
1100 | break; |
1101 | |
1102 | /* Perl space used to exclude VT, but from Perl 5.18 it is included, |
1103 | which means that Perl space and POSIX space are now identical. PCRE |
1104 | was changed at release 8.34. */ |
1105 | |
1106 | case PT_SPACE: /* Perl space */ |
1107 | case PT_PXSPACE: /* POSIX space */ |
1108 | switch(c) |
1109 | { |
1110 | HSPACE_CASES: |
1111 | VSPACE_CASES: |
1112 | OK = TRUE; |
1113 | break; |
1114 | |
1115 | default: |
1116 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_Z; |
1117 | break; |
1118 | } |
1119 | break; |
1120 | |
1121 | case PT_WORD: |
1122 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1123 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
1124 | c == CHAR_UNDERSCORE; |
1125 | break; |
1126 | |
1127 | case PT_CLIST: |
1128 | cp = PRIV(ucd_caseless_sets) + code[2]; |
1129 | for (;;) |
1130 | { |
1131 | if (c < *cp) { OK = FALSE; break; } |
1132 | if (c == *cp++) { OK = TRUE; break; } |
1133 | } |
1134 | break; |
1135 | |
1136 | case PT_UCNC: |
1137 | OK = c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
1138 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
1139 | c >= 0xe000; |
1140 | break; |
1141 | |
1142 | /* Should never occur, but keep compilers from grumbling. */ |
1143 | |
1144 | default: |
1145 | OK = codevalue != OP_PROP; |
1146 | break; |
1147 | } |
1148 | |
1149 | if (OK == (codevalue == OP_PROP)) { ADD_NEW(state_offset + 3, 0); } |
1150 | } |
1151 | break; |
1152 | #endif |
1153 | |
1154 | |
1155 | |
1156 | /* ========================================================================== */ |
1157 | /* These opcodes likewise inspect the subject character, but have an |
1158 | argument that is not a data character. It is one of these opcodes: |
1159 | OP_ANY, OP_ALLANY, OP_DIGIT, OP_NOT_DIGIT, OP_WHITESPACE, OP_NOT_SPACE, |
1160 | OP_WORDCHAR, OP_NOT_WORDCHAR. The value is loaded into d. */ |
1161 | |
1162 | case OP_TYPEPLUS: |
1163 | case OP_TYPEMINPLUS: |
1164 | case OP_TYPEPOSPLUS: |
1165 | count = current_state->count; /* Already matched */ |
1166 | if (count > 0) { 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 (count > 0 && codevalue == OP_TYPEPOSPLUS) |
1183 | { |
1184 | active_count--; /* Remove non-match possibility */ |
1185 | next_active_state--; |
1186 | } |
1187 | count++; |
1188 | ADD_NEW(state_offset, count); |
1189 | } |
1190 | } |
1191 | break; |
1192 | |
1193 | /*-----------------------------------------------------------------*/ |
1194 | case OP_TYPEQUERY: |
1195 | case OP_TYPEMINQUERY: |
1196 | case OP_TYPEPOSQUERY: |
1197 | ADD_ACTIVE(state_offset + 2, 0); |
1198 | if (clen > 0) |
1199 | { |
1200 | if (d == OP_ANY && ptr + 1 >= md->end_subject && |
1201 | (md->moptions & (PCRE_PARTIAL_HARD)) != 0 && |
1202 | NLBLOCK->nltype == NLTYPE_FIXED && |
1203 | NLBLOCK->nllen == 2 && |
1204 | c == NLBLOCK->nl[0]) |
1205 | { |
1206 | could_continue = partial_newline = TRUE; |
1207 | } |
1208 | else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) || |
1209 | (c < 256 && |
1210 | (d != OP_ANY || !IS_NEWLINE(ptr)) && |
1211 | ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0)) |
1212 | { |
1213 | if (codevalue == OP_TYPEPOSQUERY) |
1214 | { |
1215 | active_count--; /* Remove non-match possibility */ |
1216 | next_active_state--; |
1217 | } |
1218 | ADD_NEW(state_offset + 2, 0); |
1219 | } |
1220 | } |
1221 | break; |
1222 | |
1223 | /*-----------------------------------------------------------------*/ |
1224 | case OP_TYPESTAR: |
1225 | case OP_TYPEMINSTAR: |
1226 | case OP_TYPEPOSSTAR: |
1227 | ADD_ACTIVE(state_offset + 2, 0); |
1228 | if (clen > 0) |
1229 | { |
1230 | if (d == OP_ANY && ptr + 1 >= md->end_subject && |
1231 | (md->moptions & (PCRE_PARTIAL_HARD)) != 0 && |
1232 | NLBLOCK->nltype == NLTYPE_FIXED && |
1233 | NLBLOCK->nllen == 2 && |
1234 | c == NLBLOCK->nl[0]) |
1235 | { |
1236 | could_continue = partial_newline = TRUE; |
1237 | } |
1238 | else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) || |
1239 | (c < 256 && |
1240 | (d != OP_ANY || !IS_NEWLINE(ptr)) && |
1241 | ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0)) |
1242 | { |
1243 | if (codevalue == OP_TYPEPOSSTAR) |
1244 | { |
1245 | active_count--; /* Remove non-match possibility */ |
1246 | next_active_state--; |
1247 | } |
1248 | ADD_NEW(state_offset, 0); |
1249 | } |
1250 | } |
1251 | break; |
1252 | |
1253 | /*-----------------------------------------------------------------*/ |
1254 | case OP_TYPEEXACT: |
1255 | count = current_state->count; /* Number already matched */ |
1256 | if (clen > 0) |
1257 | { |
1258 | if (d == OP_ANY && ptr + 1 >= md->end_subject && |
1259 | (md->moptions & (PCRE_PARTIAL_HARD)) != 0 && |
1260 | NLBLOCK->nltype == NLTYPE_FIXED && |
1261 | NLBLOCK->nllen == 2 && |
1262 | c == NLBLOCK->nl[0]) |
1263 | { |
1264 | could_continue = partial_newline = TRUE; |
1265 | } |
1266 | else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) || |
1267 | (c < 256 && |
1268 | (d != OP_ANY || !IS_NEWLINE(ptr)) && |
1269 | ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0)) |
1270 | { |
1271 | if (++count >= (int)GET2(code, 1)) |
1272 | { ADD_NEW(state_offset + 1 + IMM2_SIZE + 1, 0); } |
1273 | else |
1274 | { ADD_NEW(state_offset, count); } |
1275 | } |
1276 | } |
1277 | break; |
1278 | |
1279 | /*-----------------------------------------------------------------*/ |
1280 | case OP_TYPEUPTO: |
1281 | case OP_TYPEMINUPTO: |
1282 | case OP_TYPEPOSUPTO: |
1283 | ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); |
1284 | count = current_state->count; /* Number already matched */ |
1285 | if (clen > 0) |
1286 | { |
1287 | if (d == OP_ANY && ptr + 1 >= md->end_subject && |
1288 | (md->moptions & (PCRE_PARTIAL_HARD)) != 0 && |
1289 | NLBLOCK->nltype == NLTYPE_FIXED && |
1290 | NLBLOCK->nllen == 2 && |
1291 | c == NLBLOCK->nl[0]) |
1292 | { |
1293 | could_continue = partial_newline = TRUE; |
1294 | } |
1295 | else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) || |
1296 | (c < 256 && |
1297 | (d != OP_ANY || !IS_NEWLINE(ptr)) && |
1298 | ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0)) |
1299 | { |
1300 | if (codevalue == OP_TYPEPOSUPTO) |
1301 | { |
1302 | active_count--; /* Remove non-match possibility */ |
1303 | next_active_state--; |
1304 | } |
1305 | if (++count >= (int)GET2(code, 1)) |
1306 | { ADD_NEW(state_offset + 2 + IMM2_SIZE, 0); } |
1307 | else |
1308 | { ADD_NEW(state_offset, count); } |
1309 | } |
1310 | } |
1311 | break; |
1312 | |
1313 | /* ========================================================================== */ |
1314 | /* These are virtual opcodes that are used when something like |
1315 | OP_TYPEPLUS has OP_PROP, OP_NOTPROP, OP_ANYNL, or OP_EXTUNI as its |
1316 | argument. It keeps the code above fast for the other cases. The argument |
1317 | is in the d variable. */ |
1318 | |
1319 | #ifdef SUPPORT_UCP |
1320 | case OP_PROP_EXTRA + OP_TYPEPLUS: |
1321 | case OP_PROP_EXTRA + OP_TYPEMINPLUS: |
1322 | case OP_PROP_EXTRA + OP_TYPEPOSPLUS: |
1323 | count = current_state->count; /* Already matched */ |
1324 | if (count > 0) { ADD_ACTIVE(state_offset + 4, 0); } |
1325 | if (clen > 0) |
1326 | { |
1327 | BOOL OK; |
1328 | const pcre_uint32 *cp; |
1329 | const ucd_record * prop = GET_UCD(c); |
1330 | switch(code[2]) |
1331 | { |
1332 | case PT_ANY: |
1333 | OK = TRUE; |
1334 | break; |
1335 | |
1336 | case PT_LAMP: |
1337 | OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || |
1338 | prop->chartype == ucp_Lt; |
1339 | break; |
1340 | |
1341 | case PT_GC: |
1342 | OK = PRIV(ucp_gentype)[prop->chartype] == code[3]; |
1343 | break; |
1344 | |
1345 | case PT_PC: |
1346 | OK = prop->chartype == code[3]; |
1347 | break; |
1348 | |
1349 | case PT_SC: |
1350 | OK = prop->script == code[3]; |
1351 | break; |
1352 | |
1353 | /* These are specials for combination cases. */ |
1354 | |
1355 | case PT_ALNUM: |
1356 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1357 | PRIV(ucp_gentype)[prop->chartype] == ucp_N; |
1358 | break; |
1359 | |
1360 | /* Perl space used to exclude VT, but from Perl 5.18 it is included, |
1361 | which means that Perl space and POSIX space are now identical. PCRE |
1362 | was changed at release 8.34. */ |
1363 | |
1364 | case PT_SPACE: /* Perl space */ |
1365 | case PT_PXSPACE: /* POSIX space */ |
1366 | switch(c) |
1367 | { |
1368 | HSPACE_CASES: |
1369 | VSPACE_CASES: |
1370 | OK = TRUE; |
1371 | break; |
1372 | |
1373 | default: |
1374 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_Z; |
1375 | break; |
1376 | } |
1377 | break; |
1378 | |
1379 | case PT_WORD: |
1380 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1381 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
1382 | c == CHAR_UNDERSCORE; |
1383 | break; |
1384 | |
1385 | case PT_CLIST: |
1386 | cp = PRIV(ucd_caseless_sets) + code[3]; |
1387 | for (;;) |
1388 | { |
1389 | if (c < *cp) { OK = FALSE; break; } |
1390 | if (c == *cp++) { OK = TRUE; break; } |
1391 | } |
1392 | break; |
1393 | |
1394 | case PT_UCNC: |
1395 | OK = c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
1396 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
1397 | c >= 0xe000; |
1398 | break; |
1399 | |
1400 | /* Should never occur, but keep compilers from grumbling. */ |
1401 | |
1402 | default: |
1403 | OK = codevalue != OP_PROP; |
1404 | break; |
1405 | } |
1406 | |
1407 | if (OK == (d == OP_PROP)) |
1408 | { |
1409 | if (count > 0 && codevalue == OP_PROP_EXTRA + OP_TYPEPOSPLUS) |
1410 | { |
1411 | active_count--; /* Remove non-match possibility */ |
1412 | next_active_state--; |
1413 | } |
1414 | count++; |
1415 | ADD_NEW(state_offset, count); |
1416 | } |
1417 | } |
1418 | break; |
1419 | |
1420 | /*-----------------------------------------------------------------*/ |
1421 | case OP_EXTUNI_EXTRA + OP_TYPEPLUS: |
1422 | case OP_EXTUNI_EXTRA + OP_TYPEMINPLUS: |
1423 | case OP_EXTUNI_EXTRA + OP_TYPEPOSPLUS: |
1424 | count = current_state->count; /* Already matched */ |
1425 | if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); } |
1426 | if (clen > 0) |
1427 | { |
1428 | int lgb, rgb; |
1429 | const pcre_uchar *nptr = ptr + clen; |
1430 | int ncount = 0; |
1431 | if (count > 0 && codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSPLUS) |
1432 | { |
1433 | active_count--; /* Remove non-match possibility */ |
1434 | next_active_state--; |
1435 | } |
1436 | lgb = UCD_GRAPHBREAK(c); |
1437 | while (nptr < end_subject) |
1438 | { |
1439 | dlen = 1; |
1440 | if (!utf) d = *nptr; else { GETCHARLEN(d, nptr, dlen); } |
1441 | rgb = UCD_GRAPHBREAK(d); |
1442 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
1443 | ncount++; |
1444 | lgb = rgb; |
1445 | nptr += dlen; |
1446 | } |
1447 | count++; |
1448 | ADD_NEW_DATA(-state_offset, count, ncount); |
1449 | } |
1450 | break; |
1451 | #endif |
1452 | |
1453 | /*-----------------------------------------------------------------*/ |
1454 | case OP_ANYNL_EXTRA + OP_TYPEPLUS: |
1455 | case OP_ANYNL_EXTRA + OP_TYPEMINPLUS: |
1456 | case OP_ANYNL_EXTRA + OP_TYPEPOSPLUS: |
1457 | count = current_state->count; /* Already matched */ |
1458 | if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); } |
1459 | if (clen > 0) |
1460 | { |
1461 | int ncount = 0; |
1462 | switch (c) |
1463 | { |
1464 | case CHAR_VT: |
1465 | case CHAR_FF: |
1466 | case CHAR_NEL: |
1467 | #ifndef EBCDIC |
1468 | case 0x2028: |
1469 | case 0x2029: |
1470 | #endif /* Not EBCDIC */ |
1471 | if ((md->moptions & PCRE_BSR_ANYCRLF) != 0) break; |
1472 | goto ANYNL01; |
1473 | |
1474 | case CHAR_CR: |
1475 | if (ptr + 1 < end_subject && UCHAR21TEST(ptr + 1) == CHAR_LF) ncount = 1; |
1476 | /* Fall through */ |
1477 | |
1478 | ANYNL01: |
1479 | case CHAR_LF: |
1480 | if (count > 0 && codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSPLUS) |
1481 | { |
1482 | active_count--; /* Remove non-match possibility */ |
1483 | next_active_state--; |
1484 | } |
1485 | count++; |
1486 | ADD_NEW_DATA(-state_offset, count, ncount); |
1487 | break; |
1488 | |
1489 | default: |
1490 | break; |
1491 | } |
1492 | } |
1493 | break; |
1494 | |
1495 | /*-----------------------------------------------------------------*/ |
1496 | case OP_VSPACE_EXTRA + OP_TYPEPLUS: |
1497 | case OP_VSPACE_EXTRA + OP_TYPEMINPLUS: |
1498 | case OP_VSPACE_EXTRA + OP_TYPEPOSPLUS: |
1499 | count = current_state->count; /* Already matched */ |
1500 | if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); } |
1501 | if (clen > 0) |
1502 | { |
1503 | BOOL OK; |
1504 | switch (c) |
1505 | { |
1506 | VSPACE_CASES: |
1507 | OK = TRUE; |
1508 | break; |
1509 | |
1510 | default: |
1511 | OK = FALSE; |
1512 | break; |
1513 | } |
1514 | |
1515 | if (OK == (d == OP_VSPACE)) |
1516 | { |
1517 | if (count > 0 && codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSPLUS) |
1518 | { |
1519 | active_count--; /* Remove non-match possibility */ |
1520 | next_active_state--; |
1521 | } |
1522 | count++; |
1523 | ADD_NEW_DATA(-state_offset, count, 0); |
1524 | } |
1525 | } |
1526 | break; |
1527 | |
1528 | /*-----------------------------------------------------------------*/ |
1529 | case OP_HSPACE_EXTRA + OP_TYPEPLUS: |
1530 | case OP_HSPACE_EXTRA + OP_TYPEMINPLUS: |
1531 | case OP_HSPACE_EXTRA + OP_TYPEPOSPLUS: |
1532 | count = current_state->count; /* Already matched */ |
1533 | if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); } |
1534 | if (clen > 0) |
1535 | { |
1536 | BOOL OK; |
1537 | switch (c) |
1538 | { |
1539 | HSPACE_CASES: |
1540 | OK = TRUE; |
1541 | break; |
1542 | |
1543 | default: |
1544 | OK = FALSE; |
1545 | break; |
1546 | } |
1547 | |
1548 | if (OK == (d == OP_HSPACE)) |
1549 | { |
1550 | if (count > 0 && codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSPLUS) |
1551 | { |
1552 | active_count--; /* Remove non-match possibility */ |
1553 | next_active_state--; |
1554 | } |
1555 | count++; |
1556 | ADD_NEW_DATA(-state_offset, count, 0); |
1557 | } |
1558 | } |
1559 | break; |
1560 | |
1561 | /*-----------------------------------------------------------------*/ |
1562 | #ifdef SUPPORT_UCP |
1563 | case OP_PROP_EXTRA + OP_TYPEQUERY: |
1564 | case OP_PROP_EXTRA + OP_TYPEMINQUERY: |
1565 | case OP_PROP_EXTRA + OP_TYPEPOSQUERY: |
1566 | count = 4; |
1567 | goto QS1; |
1568 | |
1569 | case OP_PROP_EXTRA + OP_TYPESTAR: |
1570 | case OP_PROP_EXTRA + OP_TYPEMINSTAR: |
1571 | case OP_PROP_EXTRA + OP_TYPEPOSSTAR: |
1572 | count = 0; |
1573 | |
1574 | QS1: |
1575 | |
1576 | ADD_ACTIVE(state_offset + 4, 0); |
1577 | if (clen > 0) |
1578 | { |
1579 | BOOL OK; |
1580 | const pcre_uint32 *cp; |
1581 | const ucd_record * prop = GET_UCD(c); |
1582 | switch(code[2]) |
1583 | { |
1584 | case PT_ANY: |
1585 | OK = TRUE; |
1586 | break; |
1587 | |
1588 | case PT_LAMP: |
1589 | OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || |
1590 | prop->chartype == ucp_Lt; |
1591 | break; |
1592 | |
1593 | case PT_GC: |
1594 | OK = PRIV(ucp_gentype)[prop->chartype] == code[3]; |
1595 | break; |
1596 | |
1597 | case PT_PC: |
1598 | OK = prop->chartype == code[3]; |
1599 | break; |
1600 | |
1601 | case PT_SC: |
1602 | OK = prop->script == code[3]; |
1603 | break; |
1604 | |
1605 | /* These are specials for combination cases. */ |
1606 | |
1607 | case PT_ALNUM: |
1608 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1609 | PRIV(ucp_gentype)[prop->chartype] == ucp_N; |
1610 | break; |
1611 | |
1612 | /* Perl space used to exclude VT, but from Perl 5.18 it is included, |
1613 | which means that Perl space and POSIX space are now identical. PCRE |
1614 | was changed at release 8.34. */ |
1615 | |
1616 | case PT_SPACE: /* Perl space */ |
1617 | case PT_PXSPACE: /* POSIX space */ |
1618 | switch(c) |
1619 | { |
1620 | HSPACE_CASES: |
1621 | VSPACE_CASES: |
1622 | OK = TRUE; |
1623 | break; |
1624 | |
1625 | default: |
1626 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_Z; |
1627 | break; |
1628 | } |
1629 | break; |
1630 | |
1631 | case PT_WORD: |
1632 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1633 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
1634 | c == CHAR_UNDERSCORE; |
1635 | break; |
1636 | |
1637 | case PT_CLIST: |
1638 | cp = PRIV(ucd_caseless_sets) + code[3]; |
1639 | for (;;) |
1640 | { |
1641 | if (c < *cp) { OK = FALSE; break; } |
1642 | if (c == *cp++) { OK = TRUE; break; } |
1643 | } |
1644 | break; |
1645 | |
1646 | case PT_UCNC: |
1647 | OK = c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
1648 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
1649 | c >= 0xe000; |
1650 | break; |
1651 | |
1652 | /* Should never occur, but keep compilers from grumbling. */ |
1653 | |
1654 | default: |
1655 | OK = codevalue != OP_PROP; |
1656 | break; |
1657 | } |
1658 | |
1659 | if (OK == (d == OP_PROP)) |
1660 | { |
1661 | if (codevalue == OP_PROP_EXTRA + OP_TYPEPOSSTAR || |
1662 | codevalue == OP_PROP_EXTRA + OP_TYPEPOSQUERY) |
1663 | { |
1664 | active_count--; /* Remove non-match possibility */ |
1665 | next_active_state--; |
1666 | } |
1667 | ADD_NEW(state_offset + count, 0); |
1668 | } |
1669 | } |
1670 | break; |
1671 | |
1672 | /*-----------------------------------------------------------------*/ |
1673 | case OP_EXTUNI_EXTRA + OP_TYPEQUERY: |
1674 | case OP_EXTUNI_EXTRA + OP_TYPEMINQUERY: |
1675 | case OP_EXTUNI_EXTRA + OP_TYPEPOSQUERY: |
1676 | count = 2; |
1677 | goto QS2; |
1678 | |
1679 | case OP_EXTUNI_EXTRA + OP_TYPESTAR: |
1680 | case OP_EXTUNI_EXTRA + OP_TYPEMINSTAR: |
1681 | case OP_EXTUNI_EXTRA + OP_TYPEPOSSTAR: |
1682 | count = 0; |
1683 | |
1684 | QS2: |
1685 | |
1686 | ADD_ACTIVE(state_offset + 2, 0); |
1687 | if (clen > 0) |
1688 | { |
1689 | int lgb, rgb; |
1690 | const pcre_uchar *nptr = ptr + clen; |
1691 | int ncount = 0; |
1692 | if (codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSSTAR || |
1693 | codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSQUERY) |
1694 | { |
1695 | active_count--; /* Remove non-match possibility */ |
1696 | next_active_state--; |
1697 | } |
1698 | lgb = UCD_GRAPHBREAK(c); |
1699 | while (nptr < end_subject) |
1700 | { |
1701 | dlen = 1; |
1702 | if (!utf) d = *nptr; else { GETCHARLEN(d, nptr, dlen); } |
1703 | rgb = UCD_GRAPHBREAK(d); |
1704 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
1705 | ncount++; |
1706 | lgb = rgb; |
1707 | nptr += dlen; |
1708 | } |
1709 | ADD_NEW_DATA(-(state_offset + count), 0, ncount); |
1710 | } |
1711 | break; |
1712 | #endif |
1713 | |
1714 | /*-----------------------------------------------------------------*/ |
1715 | case OP_ANYNL_EXTRA + OP_TYPEQUERY: |
1716 | case OP_ANYNL_EXTRA + OP_TYPEMINQUERY: |
1717 | case OP_ANYNL_EXTRA + OP_TYPEPOSQUERY: |
1718 | count = 2; |
1719 | goto QS3; |
1720 | |
1721 | case OP_ANYNL_EXTRA + OP_TYPESTAR: |
1722 | case OP_ANYNL_EXTRA + OP_TYPEMINSTAR: |
1723 | case OP_ANYNL_EXTRA + OP_TYPEPOSSTAR: |
1724 | count = 0; |
1725 | |
1726 | QS3: |
1727 | ADD_ACTIVE(state_offset + 2, 0); |
1728 | if (clen > 0) |
1729 | { |
1730 | int ncount = 0; |
1731 | switch (c) |
1732 | { |
1733 | case CHAR_VT: |
1734 | case CHAR_FF: |
1735 | case CHAR_NEL: |
1736 | #ifndef EBCDIC |
1737 | case 0x2028: |
1738 | case 0x2029: |
1739 | #endif /* Not EBCDIC */ |
1740 | if ((md->moptions & PCRE_BSR_ANYCRLF) != 0) break; |
1741 | goto ANYNL02; |
1742 | |
1743 | case CHAR_CR: |
1744 | if (ptr + 1 < end_subject && UCHAR21TEST(ptr + 1) == CHAR_LF) ncount = 1; |
1745 | /* Fall through */ |
1746 | |
1747 | ANYNL02: |
1748 | case CHAR_LF: |
1749 | if (codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSSTAR || |
1750 | codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSQUERY) |
1751 | { |
1752 | active_count--; /* Remove non-match possibility */ |
1753 | next_active_state--; |
1754 | } |
1755 | ADD_NEW_DATA(-(state_offset + (int)count), 0, ncount); |
1756 | break; |
1757 | |
1758 | default: |
1759 | break; |
1760 | } |
1761 | } |
1762 | break; |
1763 | |
1764 | /*-----------------------------------------------------------------*/ |
1765 | case OP_VSPACE_EXTRA + OP_TYPEQUERY: |
1766 | case OP_VSPACE_EXTRA + OP_TYPEMINQUERY: |
1767 | case OP_VSPACE_EXTRA + OP_TYPEPOSQUERY: |
1768 | count = 2; |
1769 | goto QS4; |
1770 | |
1771 | case OP_VSPACE_EXTRA + OP_TYPESTAR: |
1772 | case OP_VSPACE_EXTRA + OP_TYPEMINSTAR: |
1773 | case OP_VSPACE_EXTRA + OP_TYPEPOSSTAR: |
1774 | count = 0; |
1775 | |
1776 | QS4: |
1777 | ADD_ACTIVE(state_offset + 2, 0); |
1778 | if (clen > 0) |
1779 | { |
1780 | BOOL OK; |
1781 | switch (c) |
1782 | { |
1783 | VSPACE_CASES: |
1784 | OK = TRUE; |
1785 | break; |
1786 | |
1787 | default: |
1788 | OK = FALSE; |
1789 | break; |
1790 | } |
1791 | if (OK == (d == OP_VSPACE)) |
1792 | { |
1793 | if (codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSSTAR || |
1794 | codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSQUERY) |
1795 | { |
1796 | active_count--; /* Remove non-match possibility */ |
1797 | next_active_state--; |
1798 | } |
1799 | ADD_NEW_DATA(-(state_offset + (int)count), 0, 0); |
1800 | } |
1801 | } |
1802 | break; |
1803 | |
1804 | /*-----------------------------------------------------------------*/ |
1805 | case OP_HSPACE_EXTRA + OP_TYPEQUERY: |
1806 | case OP_HSPACE_EXTRA + OP_TYPEMINQUERY: |
1807 | case OP_HSPACE_EXTRA + OP_TYPEPOSQUERY: |
1808 | count = 2; |
1809 | goto QS5; |
1810 | |
1811 | case OP_HSPACE_EXTRA + OP_TYPESTAR: |
1812 | case OP_HSPACE_EXTRA + OP_TYPEMINSTAR: |
1813 | case OP_HSPACE_EXTRA + OP_TYPEPOSSTAR: |
1814 | count = 0; |
1815 | |
1816 | QS5: |
1817 | ADD_ACTIVE(state_offset + 2, 0); |
1818 | if (clen > 0) |
1819 | { |
1820 | BOOL OK; |
1821 | switch (c) |
1822 | { |
1823 | HSPACE_CASES: |
1824 | OK = TRUE; |
1825 | break; |
1826 | |
1827 | default: |
1828 | OK = FALSE; |
1829 | break; |
1830 | } |
1831 | |
1832 | if (OK == (d == OP_HSPACE)) |
1833 | { |
1834 | if (codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSSTAR || |
1835 | codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSQUERY) |
1836 | { |
1837 | active_count--; /* Remove non-match possibility */ |
1838 | next_active_state--; |
1839 | } |
1840 | ADD_NEW_DATA(-(state_offset + (int)count), 0, 0); |
1841 | } |
1842 | } |
1843 | break; |
1844 | |
1845 | /*-----------------------------------------------------------------*/ |
1846 | #ifdef SUPPORT_UCP |
1847 | case OP_PROP_EXTRA + OP_TYPEEXACT: |
1848 | case OP_PROP_EXTRA + OP_TYPEUPTO: |
1849 | case OP_PROP_EXTRA + OP_TYPEMINUPTO: |
1850 | case OP_PROP_EXTRA + OP_TYPEPOSUPTO: |
1851 | if (codevalue != OP_PROP_EXTRA + OP_TYPEEXACT) |
1852 | { ADD_ACTIVE(state_offset + 1 + IMM2_SIZE + 3, 0); } |
1853 | count = current_state->count; /* Number already matched */ |
1854 | if (clen > 0) |
1855 | { |
1856 | BOOL OK; |
1857 | const pcre_uint32 *cp; |
1858 | const ucd_record * prop = GET_UCD(c); |
1859 | switch(code[1 + IMM2_SIZE + 1]) |
1860 | { |
1861 | case PT_ANY: |
1862 | OK = TRUE; |
1863 | break; |
1864 | |
1865 | case PT_LAMP: |
1866 | OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || |
1867 | prop->chartype == ucp_Lt; |
1868 | break; |
1869 | |
1870 | case PT_GC: |
1871 | OK = PRIV(ucp_gentype)[prop->chartype] == code[1 + IMM2_SIZE + 2]; |
1872 | break; |
1873 | |
1874 | case PT_PC: |
1875 | OK = prop->chartype == code[1 + IMM2_SIZE + 2]; |
1876 | break; |
1877 | |
1878 | case PT_SC: |
1879 | OK = prop->script == code[1 + IMM2_SIZE + 2]; |
1880 | break; |
1881 | |
1882 | /* These are specials for combination cases. */ |
1883 | |
1884 | case PT_ALNUM: |
1885 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1886 | PRIV(ucp_gentype)[prop->chartype] == ucp_N; |
1887 | break; |
1888 | |
1889 | /* Perl space used to exclude VT, but from Perl 5.18 it is included, |
1890 | which means that Perl space and POSIX space are now identical. PCRE |
1891 | was changed at release 8.34. */ |
1892 | |
1893 | case PT_SPACE: /* Perl space */ |
1894 | case PT_PXSPACE: /* POSIX space */ |
1895 | switch(c) |
1896 | { |
1897 | HSPACE_CASES: |
1898 | VSPACE_CASES: |
1899 | OK = TRUE; |
1900 | break; |
1901 | |
1902 | default: |
1903 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_Z; |
1904 | break; |
1905 | } |
1906 | break; |
1907 | |
1908 | case PT_WORD: |
1909 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1910 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
1911 | c == CHAR_UNDERSCORE; |
1912 | break; |
1913 | |
1914 | case PT_CLIST: |
1915 | cp = PRIV(ucd_caseless_sets) + code[1 + IMM2_SIZE + 2]; |
1916 | for (;;) |
1917 | { |
1918 | if (c < *cp) { OK = FALSE; break; } |
1919 | if (c == *cp++) { OK = TRUE; break; } |
1920 | } |
1921 | break; |
1922 | |
1923 | case PT_UCNC: |
1924 | OK = c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
1925 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
1926 | c >= 0xe000; |
1927 | break; |
1928 | |
1929 | /* Should never occur, but keep compilers from grumbling. */ |
1930 | |
1931 | default: |
1932 | OK = codevalue != OP_PROP; |
1933 | break; |
1934 | } |
1935 | |
1936 | if (OK == (d == OP_PROP)) |
1937 | { |
1938 | if (codevalue == OP_PROP_EXTRA + OP_TYPEPOSUPTO) |
1939 | { |
1940 | active_count--; /* Remove non-match possibility */ |
1941 | next_active_state--; |
1942 | } |
1943 | if (++count >= (int)GET2(code, 1)) |
1944 | { ADD_NEW(state_offset + 1 + IMM2_SIZE + 3, 0); } |
1945 | else |
1946 | { ADD_NEW(state_offset, count); } |
1947 | } |
1948 | } |
1949 | break; |
1950 | |
1951 | /*-----------------------------------------------------------------*/ |
1952 | case OP_EXTUNI_EXTRA + OP_TYPEEXACT: |
1953 | case OP_EXTUNI_EXTRA + OP_TYPEUPTO: |
1954 | case OP_EXTUNI_EXTRA + OP_TYPEMINUPTO: |
1955 | case OP_EXTUNI_EXTRA + OP_TYPEPOSUPTO: |
1956 | if (codevalue != OP_EXTUNI_EXTRA + OP_TYPEEXACT) |
1957 | { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); } |
1958 | count = current_state->count; /* Number already matched */ |
1959 | if (clen > 0) |
1960 | { |
1961 | int lgb, rgb; |
1962 | const pcre_uchar *nptr = ptr + clen; |
1963 | int ncount = 0; |
1964 | if (codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSUPTO) |
1965 | { |
1966 | active_count--; /* Remove non-match possibility */ |
1967 | next_active_state--; |
1968 | } |
1969 | lgb = UCD_GRAPHBREAK(c); |
1970 | while (nptr < end_subject) |
1971 | { |
1972 | dlen = 1; |
1973 | if (!utf) d = *nptr; else { GETCHARLEN(d, nptr, dlen); } |
1974 | rgb = UCD_GRAPHBREAK(d); |
1975 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
1976 | ncount++; |
1977 | lgb = rgb; |
1978 | nptr += dlen; |
1979 | } |
1980 | if (nptr >= end_subject && (md->moptions & PCRE_PARTIAL_HARD) != 0) |
1981 | reset_could_continue = TRUE; |
1982 | if (++count >= (int)GET2(code, 1)) |
1983 | { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, ncount); } |
1984 | else |
1985 | { ADD_NEW_DATA(-state_offset, count, ncount); } |
1986 | } |
1987 | break; |
1988 | #endif |
1989 | |
1990 | /*-----------------------------------------------------------------*/ |
1991 | case OP_ANYNL_EXTRA + OP_TYPEEXACT: |
1992 | case OP_ANYNL_EXTRA + OP_TYPEUPTO: |
1993 | case OP_ANYNL_EXTRA + OP_TYPEMINUPTO: |
1994 | case OP_ANYNL_EXTRA + OP_TYPEPOSUPTO: |
1995 | if (codevalue != OP_ANYNL_EXTRA + OP_TYPEEXACT) |
1996 | { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); } |
1997 | count = current_state->count; /* Number already matched */ |
1998 | if (clen > 0) |
1999 | { |
2000 | int ncount = 0; |
2001 | switch (c) |
2002 | { |
2003 | case CHAR_VT: |
2004 | case CHAR_FF: |
2005 | case CHAR_NEL: |
2006 | #ifndef EBCDIC |
2007 | case 0x2028: |
2008 | case 0x2029: |
2009 | #endif /* Not EBCDIC */ |
2010 | if ((md->moptions & PCRE_BSR_ANYCRLF) != 0) break; |
2011 | goto ANYNL03; |
2012 | |
2013 | case CHAR_CR: |
2014 | if (ptr + 1 < end_subject && UCHAR21TEST(ptr + 1) == CHAR_LF) ncount = 1; |
2015 | /* Fall through */ |
2016 | |
2017 | ANYNL03: |
2018 | case CHAR_LF: |
2019 | if (codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSUPTO) |
2020 | { |
2021 | active_count--; /* Remove non-match possibility */ |
2022 | next_active_state--; |
2023 | } |
2024 | if (++count >= (int)GET2(code, 1)) |
2025 | { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, ncount); } |
2026 | else |
2027 | { ADD_NEW_DATA(-state_offset, count, ncount); } |
2028 | break; |
2029 | |
2030 | default: |
2031 | break; |
2032 | } |
2033 | } |
2034 | break; |
2035 | |
2036 | /*-----------------------------------------------------------------*/ |
2037 | case OP_VSPACE_EXTRA + OP_TYPEEXACT: |
2038 | case OP_VSPACE_EXTRA + OP_TYPEUPTO: |
2039 | case OP_VSPACE_EXTRA + OP_TYPEMINUPTO: |
2040 | case OP_VSPACE_EXTRA + OP_TYPEPOSUPTO: |
2041 | if (codevalue != OP_VSPACE_EXTRA + OP_TYPEEXACT) |
2042 | { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); } |
2043 | count = current_state->count; /* Number already matched */ |
2044 | if (clen > 0) |
2045 | { |
2046 | BOOL OK; |
2047 | switch (c) |
2048 | { |
2049 | VSPACE_CASES: |
2050 | OK = TRUE; |
2051 | break; |
2052 | |
2053 | default: |
2054 | OK = FALSE; |
2055 | } |
2056 | |
2057 | if (OK == (d == OP_VSPACE)) |
2058 | { |
2059 | if (codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSUPTO) |
2060 | { |
2061 | active_count--; /* Remove non-match possibility */ |
2062 | next_active_state--; |
2063 | } |
2064 | if (++count >= (int)GET2(code, 1)) |
2065 | { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, 0); } |
2066 | else |
2067 | { ADD_NEW_DATA(-state_offset, count, 0); } |
2068 | } |
2069 | } |
2070 | break; |
2071 | |
2072 | /*-----------------------------------------------------------------*/ |
2073 | case OP_HSPACE_EXTRA + OP_TYPEEXACT: |
2074 | case OP_HSPACE_EXTRA + OP_TYPEUPTO: |
2075 | case OP_HSPACE_EXTRA + OP_TYPEMINUPTO: |
2076 | case OP_HSPACE_EXTRA + OP_TYPEPOSUPTO: |
2077 | if (codevalue != OP_HSPACE_EXTRA + OP_TYPEEXACT) |
2078 | { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); } |
2079 | count = current_state->count; /* Number already matched */ |
2080 | if (clen > 0) |
2081 | { |
2082 | BOOL OK; |
2083 | switch (c) |
2084 | { |
2085 | HSPACE_CASES: |
2086 | OK = TRUE; |
2087 | break; |
2088 | |
2089 | default: |
2090 | OK = FALSE; |
2091 | break; |
2092 | } |
2093 | |
2094 | if (OK == (d == OP_HSPACE)) |
2095 | { |
2096 | if (codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSUPTO) |
2097 | { |
2098 | active_count--; /* Remove non-match possibility */ |
2099 | next_active_state--; |
2100 | } |
2101 | if (++count >= (int)GET2(code, 1)) |
2102 | { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, 0); } |
2103 | else |
2104 | { ADD_NEW_DATA(-state_offset, count, 0); } |
2105 | } |
2106 | } |
2107 | break; |
2108 | |
2109 | /* ========================================================================== */ |
2110 | /* These opcodes are followed by a character that is usually compared |
2111 | to the current subject character; it is loaded into d. We still get |
2112 | here even if there is no subject character, because in some cases zero |
2113 | repetitions are permitted. */ |
2114 | |
2115 | /*-----------------------------------------------------------------*/ |
2116 | case OP_CHAR: |
2117 | if (clen > 0 && c == d) { ADD_NEW(state_offset + dlen + 1, 0); } |
2118 | break; |
2119 | |
2120 | /*-----------------------------------------------------------------*/ |
2121 | case OP_CHARI: |
2122 | if (clen == 0) break; |
2123 | |
2124 | #ifdef SUPPORT_UTF |
2125 | if (utf) |
2126 | { |
2127 | if (c == d) { ADD_NEW(state_offset + dlen + 1, 0); } else |
2128 | { |
2129 | unsigned int othercase; |
2130 | if (c < 128) |
2131 | othercase = fcc[c]; |
2132 | else |
2133 | /* If we have Unicode property support, we can use it to test the |
2134 | other case of the character. */ |
2135 | #ifdef SUPPORT_UCP |
2136 | othercase = UCD_OTHERCASE(c); |
2137 | #else |
2138 | othercase = NOTACHAR; |
2139 | #endif |
2140 | |
2141 | if (d == othercase) { ADD_NEW(state_offset + dlen + 1, 0); } |
2142 | } |
2143 | } |
2144 | else |
2145 | #endif /* SUPPORT_UTF */ |
2146 | /* Not UTF mode */ |
2147 | { |
2148 | if (TABLE_GET(c, lcc, c) == TABLE_GET(d, lcc, d)) |
2149 | { ADD_NEW(state_offset + 2, 0); } |
2150 | } |
2151 | break; |
2152 | |
2153 | |
2154 | #ifdef SUPPORT_UCP |
2155 | /*-----------------------------------------------------------------*/ |
2156 | /* This is a tricky one because it can match more than one character. |
2157 | Find out how many characters to skip, and then set up a negative state |
2158 | to wait for them to pass before continuing. */ |
2159 | |
2160 | case OP_EXTUNI: |
2161 | if (clen > 0) |
2162 | { |
2163 | int lgb, rgb; |
2164 | const pcre_uchar *nptr = ptr + clen; |
2165 | int ncount = 0; |
2166 | lgb = UCD_GRAPHBREAK(c); |
2167 | while (nptr < end_subject) |
2168 | { |
2169 | dlen = 1; |
2170 | if (!utf) d = *nptr; else { GETCHARLEN(d, nptr, dlen); } |
2171 | rgb = UCD_GRAPHBREAK(d); |
2172 | if ((PRIV(ucp_gbtable)[lgb] & (1 << rgb)) == 0) break; |
2173 | ncount++; |
2174 | lgb = rgb; |
2175 | nptr += dlen; |
2176 | } |
2177 | if (nptr >= end_subject && (md->moptions & PCRE_PARTIAL_HARD) != 0) |
2178 | reset_could_continue = TRUE; |
2179 | ADD_NEW_DATA(-(state_offset + 1), 0, ncount); |
2180 | } |
2181 | break; |
2182 | #endif |
2183 | |
2184 | /*-----------------------------------------------------------------*/ |
2185 | /* This is a tricky like EXTUNI because it too can match more than one |
2186 | character (when CR is followed by LF). In this case, set up a negative |
2187 | state to wait for one character to pass before continuing. */ |
2188 | |
2189 | case OP_ANYNL: |
2190 | if (clen > 0) switch(c) |
2191 | { |
2192 | case CHAR_VT: |
2193 | case CHAR_FF: |
2194 | case CHAR_NEL: |
2195 | #ifndef EBCDIC |
2196 | case 0x2028: |
2197 | case 0x2029: |
2198 | #endif /* Not EBCDIC */ |
2199 | if ((md->moptions & PCRE_BSR_ANYCRLF) != 0) break; |
2200 | |
2201 | case CHAR_LF: |
2202 | ADD_NEW(state_offset + 1, 0); |
2203 | break; |
2204 | |
2205 | case CHAR_CR: |
2206 | if (ptr + 1 >= end_subject) |
2207 | { |
2208 | ADD_NEW(state_offset + 1, 0); |
2209 | if ((md->moptions & PCRE_PARTIAL_HARD) != 0) |
2210 | reset_could_continue = TRUE; |
2211 | } |
2212 | else if (UCHAR21TEST(ptr + 1) == CHAR_LF) |
2213 | { |
2214 | ADD_NEW_DATA(-(state_offset + 1), 0, 1); |
2215 | } |
2216 | else |
2217 | { |
2218 | ADD_NEW(state_offset + 1, 0); |
2219 | } |
2220 | break; |
2221 | } |
2222 | break; |
2223 | |
2224 | /*-----------------------------------------------------------------*/ |
2225 | case OP_NOT_VSPACE: |
2226 | if (clen > 0) switch(c) |
2227 | { |
2228 | VSPACE_CASES: |
2229 | break; |
2230 | |
2231 | default: |
2232 | ADD_NEW(state_offset + 1, 0); |
2233 | break; |
2234 | } |
2235 | break; |
2236 | |
2237 | /*-----------------------------------------------------------------*/ |
2238 | case OP_VSPACE: |
2239 | if (clen > 0) switch(c) |
2240 | { |
2241 | VSPACE_CASES: |
2242 | ADD_NEW(state_offset + 1, 0); |
2243 | break; |
2244 | |
2245 | default: |
2246 | break; |
2247 | } |
2248 | break; |
2249 | |
2250 | /*-----------------------------------------------------------------*/ |
2251 | case OP_NOT_HSPACE: |
2252 | if (clen > 0) switch(c) |
2253 | { |
2254 | HSPACE_CASES: |
2255 | break; |
2256 | |
2257 | default: |
2258 | ADD_NEW(state_offset + 1, 0); |
2259 | break; |
2260 | } |
2261 | break; |
2262 | |
2263 | /*-----------------------------------------------------------------*/ |
2264 | case OP_HSPACE: |
2265 | if (clen > 0) switch(c) |
2266 | { |
2267 | HSPACE_CASES: |
2268 | ADD_NEW(state_offset + 1, 0); |
2269 | break; |
2270 | |
2271 | default: |
2272 | break; |
2273 | } |
2274 | break; |
2275 | |
2276 | /*-----------------------------------------------------------------*/ |
2277 | /* Match a negated single character casefully. */ |
2278 | |
2279 | case OP_NOT: |
2280 | if (clen > 0 && c != d) { ADD_NEW(state_offset + dlen + 1, 0); } |
2281 | break; |
2282 | |
2283 | /*-----------------------------------------------------------------*/ |
2284 | /* Match a negated single character caselessly. */ |
2285 | |
2286 | case OP_NOTI: |
2287 | if (clen > 0) |
2288 | { |
2289 | pcre_uint32 otherd; |
2290 | #ifdef SUPPORT_UTF |
2291 | if (utf && d >= 128) |
2292 | { |
2293 | #ifdef SUPPORT_UCP |
2294 | otherd = UCD_OTHERCASE(d); |
2295 | #else |
2296 | otherd = d; |
2297 | #endif /* SUPPORT_UCP */ |
2298 | } |
2299 | else |
2300 | #endif /* SUPPORT_UTF */ |
2301 | otherd = TABLE_GET(d, fcc, d); |
2302 | if (c != d && c != otherd) |
2303 | { ADD_NEW(state_offset + dlen + 1, 0); } |
2304 | } |
2305 | break; |
2306 | |
2307 | /*-----------------------------------------------------------------*/ |
2308 | case OP_PLUSI: |
2309 | case OP_MINPLUSI: |
2310 | case OP_POSPLUSI: |
2311 | case OP_NOTPLUSI: |
2312 | case OP_NOTMINPLUSI: |
2313 | case OP_NOTPOSPLUSI: |
2314 | caseless = TRUE; |
2315 | codevalue -= OP_STARI - OP_STAR; |
2316 | |
2317 | /* Fall through */ |
2318 | case OP_PLUS: |
2319 | case OP_MINPLUS: |
2320 | case OP_POSPLUS: |
2321 | case OP_NOTPLUS: |
2322 | case OP_NOTMINPLUS: |
2323 | case OP_NOTPOSPLUS: |
2324 | count = current_state->count; /* Already matched */ |
2325 | if (count > 0) { ADD_ACTIVE(state_offset + dlen + 1, 0); } |
2326 | if (clen > 0) |
2327 | { |
2328 | pcre_uint32 otherd = NOTACHAR; |
2329 | if (caseless) |
2330 | { |
2331 | #ifdef SUPPORT_UTF |
2332 | if (utf && d >= 128) |
2333 | { |
2334 | #ifdef SUPPORT_UCP |
2335 | otherd = UCD_OTHERCASE(d); |
2336 | #endif /* SUPPORT_UCP */ |
2337 | } |
2338 | else |
2339 | #endif /* SUPPORT_UTF */ |
2340 | otherd = TABLE_GET(d, fcc, d); |
2341 | } |
2342 | if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR)) |
2343 | { |
2344 | if (count > 0 && |
2345 | (codevalue == OP_POSPLUS || codevalue == OP_NOTPOSPLUS)) |
2346 | { |
2347 | active_count--; /* Remove non-match possibility */ |
2348 | next_active_state--; |
2349 | } |
2350 | count++; |
2351 | ADD_NEW(state_offset, count); |
2352 | } |
2353 | } |
2354 | break; |
2355 | |
2356 | /*-----------------------------------------------------------------*/ |
2357 | case OP_QUERYI: |
2358 | case OP_MINQUERYI: |
2359 | case OP_POSQUERYI: |
2360 | case OP_NOTQUERYI: |
2361 | case OP_NOTMINQUERYI: |
2362 | case OP_NOTPOSQUERYI: |
2363 | caseless = TRUE; |
2364 | codevalue -= OP_STARI - OP_STAR; |
2365 | /* Fall through */ |
2366 | case OP_QUERY: |
2367 | case OP_MINQUERY: |
2368 | case OP_POSQUERY: |
2369 | case OP_NOTQUERY: |
2370 | case OP_NOTMINQUERY: |
2371 | case OP_NOTPOSQUERY: |
2372 | ADD_ACTIVE(state_offset + dlen + 1, 0); |
2373 | if (clen > 0) |
2374 | { |
2375 | pcre_uint32 otherd = NOTACHAR; |
2376 | if (caseless) |
2377 | { |
2378 | #ifdef SUPPORT_UTF |
2379 | if (utf && d >= 128) |
2380 | { |
2381 | #ifdef SUPPORT_UCP |
2382 | otherd = UCD_OTHERCASE(d); |
2383 | #endif /* SUPPORT_UCP */ |
2384 | } |
2385 | else |
2386 | #endif /* SUPPORT_UTF */ |
2387 | otherd = TABLE_GET(d, fcc, d); |
2388 | } |
2389 | if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR)) |
2390 | { |
2391 | if (codevalue == OP_POSQUERY || codevalue == OP_NOTPOSQUERY) |
2392 | { |
2393 | active_count--; /* Remove non-match possibility */ |
2394 | next_active_state--; |
2395 | } |
2396 | ADD_NEW(state_offset + dlen + 1, 0); |
2397 | } |
2398 | } |
2399 | break; |
2400 | |
2401 | /*-----------------------------------------------------------------*/ |
2402 | case OP_STARI: |
2403 | case OP_MINSTARI: |
2404 | case OP_POSSTARI: |
2405 | case OP_NOTSTARI: |
2406 | case OP_NOTMINSTARI: |
2407 | case OP_NOTPOSSTARI: |
2408 | caseless = TRUE; |
2409 | codevalue -= OP_STARI - OP_STAR; |
2410 | /* Fall through */ |
2411 | case OP_STAR: |
2412 | case OP_MINSTAR: |
2413 | case OP_POSSTAR: |
2414 | case OP_NOTSTAR: |
2415 | case OP_NOTMINSTAR: |
2416 | case OP_NOTPOSSTAR: |
2417 | ADD_ACTIVE(state_offset + dlen + 1, 0); |
2418 | if (clen > 0) |
2419 | { |
2420 | pcre_uint32 otherd = NOTACHAR; |
2421 | if (caseless) |
2422 | { |
2423 | #ifdef SUPPORT_UTF |
2424 | if (utf && d >= 128) |
2425 | { |
2426 | #ifdef SUPPORT_UCP |
2427 | otherd = UCD_OTHERCASE(d); |
2428 | #endif /* SUPPORT_UCP */ |
2429 | } |
2430 | else |
2431 | #endif /* SUPPORT_UTF */ |
2432 | otherd = TABLE_GET(d, fcc, d); |
2433 | } |
2434 | if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR)) |
2435 | { |
2436 | if (codevalue == OP_POSSTAR || codevalue == OP_NOTPOSSTAR) |
2437 | { |
2438 | active_count--; /* Remove non-match possibility */ |
2439 | next_active_state--; |
2440 | } |
2441 | ADD_NEW(state_offset, 0); |
2442 | } |
2443 | } |
2444 | break; |
2445 | |
2446 | /*-----------------------------------------------------------------*/ |
2447 | case OP_EXACTI: |
2448 | case OP_NOTEXACTI: |
2449 | caseless = TRUE; |
2450 | codevalue -= OP_STARI - OP_STAR; |
2451 | /* Fall through */ |
2452 | case OP_EXACT: |
2453 | case OP_NOTEXACT: |
2454 | count = current_state->count; /* Number already matched */ |
2455 | if (clen > 0) |
2456 | { |
2457 | pcre_uint32 otherd = NOTACHAR; |
2458 | if (caseless) |
2459 | { |
2460 | #ifdef SUPPORT_UTF |
2461 | if (utf && d >= 128) |
2462 | { |
2463 | #ifdef SUPPORT_UCP |
2464 | otherd = UCD_OTHERCASE(d); |
2465 | #endif /* SUPPORT_UCP */ |
2466 | } |
2467 | else |
2468 | #endif /* SUPPORT_UTF */ |
2469 | otherd = TABLE_GET(d, fcc, d); |
2470 | } |
2471 | if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR)) |
2472 | { |
2473 | if (++count >= (int)GET2(code, 1)) |
2474 | { ADD_NEW(state_offset + dlen + 1 + IMM2_SIZE, 0); } |
2475 | else |
2476 | { ADD_NEW(state_offset, count); } |
2477 | } |
2478 | } |
2479 | break; |
2480 | |
2481 | /*-----------------------------------------------------------------*/ |
2482 | case OP_UPTOI: |
2483 | case OP_MINUPTOI: |
2484 | case OP_POSUPTOI: |
2485 | case OP_NOTUPTOI: |
2486 | case OP_NOTMINUPTOI: |
2487 | case OP_NOTPOSUPTOI: |
2488 | caseless = TRUE; |
2489 | codevalue -= OP_STARI - OP_STAR; |
2490 | /* Fall through */ |
2491 | case OP_UPTO: |
2492 | case OP_MINUPTO: |
2493 | case OP_POSUPTO: |
2494 | case OP_NOTUPTO: |
2495 | case OP_NOTMINUPTO: |
2496 | case OP_NOTPOSUPTO: |
2497 | ADD_ACTIVE(state_offset + dlen + 1 + IMM2_SIZE, 0); |
2498 | count = current_state->count; /* Number already matched */ |
2499 | if (clen > 0) |
2500 | { |
2501 | pcre_uint32 otherd = NOTACHAR; |
2502 | if (caseless) |
2503 | { |
2504 | #ifdef SUPPORT_UTF |
2505 | if (utf && d >= 128) |
2506 | { |
2507 | #ifdef SUPPORT_UCP |
2508 | otherd = UCD_OTHERCASE(d); |
2509 | #endif /* SUPPORT_UCP */ |
2510 | } |
2511 | else |
2512 | #endif /* SUPPORT_UTF */ |
2513 | otherd = TABLE_GET(d, fcc, d); |
2514 | } |
2515 | if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR)) |
2516 | { |
2517 | if (codevalue == OP_POSUPTO || codevalue == OP_NOTPOSUPTO) |
2518 | { |
2519 | active_count--; /* Remove non-match possibility */ |
2520 | next_active_state--; |
2521 | } |
2522 | if (++count >= (int)GET2(code, 1)) |
2523 | { ADD_NEW(state_offset + dlen + 1 + IMM2_SIZE, 0); } |
2524 | else |
2525 | { ADD_NEW(state_offset, count); } |
2526 | } |
2527 | } |
2528 | break; |
2529 | |
2530 | |
2531 | /* ========================================================================== */ |
2532 | /* These are the class-handling opcodes */ |
2533 | |
2534 | case OP_CLASS: |
2535 | case OP_NCLASS: |
2536 | case OP_XCLASS: |
2537 | { |
2538 | BOOL isinclass = FALSE; |
2539 | int next_state_offset; |
2540 | const pcre_uchar *ecode; |
2541 | |
2542 | /* For a simple class, there is always just a 32-byte table, and we |
2543 | can set isinclass from it. */ |
2544 | |
2545 | if (codevalue != OP_XCLASS) |
2546 | { |
2547 | ecode = code + 1 + (32 / sizeof(pcre_uchar)); |
2548 | if (clen > 0) |
2549 | { |
2550 | isinclass = (c > 255)? (codevalue == OP_NCLASS) : |
2551 | ((((pcre_uint8 *)(code + 1))[c/8] & (1 << (c&7))) != 0); |
2552 | } |
2553 | } |
2554 | |
2555 | /* An extended class may have a table or a list of single characters, |
2556 | ranges, or both, and it may be positive or negative. There's a |
2557 | function that sorts all this out. */ |
2558 | |
2559 | else |
2560 | { |
2561 | ecode = code + GET(code, 1); |
2562 | if (clen > 0) isinclass = PRIV(xclass)(c, code + 1 + LINK_SIZE, utf); |
2563 | } |
2564 | |
2565 | /* At this point, isinclass is set for all kinds of class, and ecode |
2566 | points to the byte after the end of the class. If there is a |
2567 | quantifier, this is where it will be. */ |
2568 | |
2569 | next_state_offset = (int)(ecode - start_code); |
2570 | |
2571 | switch (*ecode) |
2572 | { |
2573 | case OP_CRSTAR: |
2574 | case OP_CRMINSTAR: |
2575 | case OP_CRPOSSTAR: |
2576 | ADD_ACTIVE(next_state_offset + 1, 0); |
2577 | if (isinclass) |
2578 | { |
2579 | if (*ecode == OP_CRPOSSTAR) |
2580 | { |
2581 | active_count--; /* Remove non-match possibility */ |
2582 | next_active_state--; |
2583 | } |
2584 | ADD_NEW(state_offset, 0); |
2585 | } |
2586 | break; |
2587 | |
2588 | case OP_CRPLUS: |
2589 | case OP_CRMINPLUS: |
2590 | case OP_CRPOSPLUS: |
2591 | count = current_state->count; /* Already matched */ |
2592 | if (count > 0) { ADD_ACTIVE(next_state_offset + 1, 0); } |
2593 | if (isinclass) |
2594 | { |
2595 | if (count > 0 && *ecode == OP_CRPOSPLUS) |
2596 | { |
2597 | active_count--; /* Remove non-match possibility */ |
2598 | next_active_state--; |
2599 | } |
2600 | count++; |
2601 | ADD_NEW(state_offset, count); |
2602 | } |
2603 | break; |
2604 | |
2605 | case OP_CRQUERY: |
2606 | case OP_CRMINQUERY: |
2607 | case OP_CRPOSQUERY: |
2608 | ADD_ACTIVE(next_state_offset + 1, 0); |
2609 | if (isinclass) |
2610 | { |
2611 | if (*ecode == OP_CRPOSQUERY) |
2612 | { |
2613 | active_count--; /* Remove non-match possibility */ |
2614 | next_active_state--; |
2615 | } |
2616 | ADD_NEW(next_state_offset + 1, 0); |
2617 | } |
2618 | break; |
2619 | |
2620 | case OP_CRRANGE: |
2621 | case OP_CRMINRANGE: |
2622 | case OP_CRPOSRANGE: |
2623 | count = current_state->count; /* Already matched */ |
2624 | if (count >= (int)GET2(ecode, 1)) |
2625 | { ADD_ACTIVE(next_state_offset + 1 + 2 * IMM2_SIZE, 0); } |
2626 | if (isinclass) |
2627 | { |
2628 | int max = (int)GET2(ecode, 1 + IMM2_SIZE); |
2629 | if (*ecode == OP_CRPOSRANGE && count >= (int)GET2(ecode, 1)) |
2630 | { |
2631 | active_count--; /* Remove non-match possibility */ |
2632 | next_active_state--; |
2633 | } |
2634 | if (++count >= max && max != 0) /* Max 0 => no limit */ |
2635 | { ADD_NEW(next_state_offset + 1 + 2 * IMM2_SIZE, 0); } |
2636 | else |
2637 | { ADD_NEW(state_offset, count); } |
2638 | } |
2639 | break; |
2640 | |
2641 | default: |
2642 | if (isinclass) { ADD_NEW(next_state_offset, 0); } |
2643 | break; |
2644 | } |
2645 | } |
2646 | break; |
2647 | |
2648 | /* ========================================================================== */ |
2649 | /* These are the opcodes for fancy brackets of various kinds. We have |
2650 | to use recursion in order to handle them. The "always failing" assertion |
2651 | (?!) is optimised to OP_FAIL when compiling, so we have to support that, |
2652 | though the other "backtracking verbs" are not supported. */ |
2653 | |
2654 | case OP_FAIL: |
2655 | forced_fail++; /* Count FAILs for multiple states */ |
2656 | break; |
2657 | |
2658 | case OP_ASSERT: |
2659 | case OP_ASSERT_NOT: |
2660 | case OP_ASSERTBACK: |
2661 | case OP_ASSERTBACK_NOT: |
2662 | { |
2663 | int rc; |
2664 | int local_offsets[2]; |
2665 | int local_workspace[1000]; |
2666 | const pcre_uchar *endasscode = code + GET(code, 1); |
2667 | |
2668 | while (*endasscode == OP_ALT) endasscode += GET(endasscode, 1); |
2669 | |
2670 | rc = internal_dfa_exec( |
2671 | md, /* static match data */ |
2672 | code, /* this subexpression's code */ |
2673 | ptr, /* where we currently are */ |
2674 | (int)(ptr - start_subject), /* start offset */ |
2675 | local_offsets, /* offset vector */ |
2676 | sizeof(local_offsets)/sizeof(int), /* size of same */ |
2677 | local_workspace, /* workspace vector */ |
2678 | sizeof(local_workspace)/sizeof(int), /* size of same */ |
2679 | rlevel); /* function recursion level */ |
2680 | |
2681 | if (rc == PCRE_ERROR_DFA_UITEM) return rc; |
2682 | if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK)) |
2683 | { ADD_ACTIVE((int)(endasscode + LINK_SIZE + 1 - start_code), 0); } |
2684 | } |
2685 | break; |
2686 | |
2687 | /*-----------------------------------------------------------------*/ |
2688 | case OP_COND: |
2689 | case OP_SCOND: |
2690 | { |
2691 | int local_offsets[1000]; |
2692 | int local_workspace[1000]; |
2693 | int codelink = GET(code, 1); |
2694 | int condcode; |
2695 | |
2696 | /* Because of the way auto-callout works during compile, a callout item |
2697 | is inserted between OP_COND and an assertion condition. This does not |
2698 | happen for the other conditions. */ |
2699 | |
2700 | if (code[LINK_SIZE+1] == OP_CALLOUT) |
2701 | { |
2702 | rrc = 0; |
2703 | if (PUBL(callout) != NULL) |
2704 | { |
2705 | PUBL(callout_block) cb; |
2706 | cb.version = 1; /* Version 1 of the callout block */ |
2707 | cb.callout_number = code[LINK_SIZE+2]; |
2708 | cb.offset_vector = offsets; |
2709 | #if defined COMPILE_PCRE8 |
2710 | cb.subject = (PCRE_SPTR)start_subject; |
2711 | #elif defined COMPILE_PCRE16 |
2712 | cb.subject = (PCRE_SPTR16)start_subject; |
2713 | #elif defined COMPILE_PCRE32 |
2714 | cb.subject = (PCRE_SPTR32)start_subject; |
2715 | #endif |
2716 | cb.subject_length = (int)(end_subject - start_subject); |
2717 | cb.start_match = (int)(current_subject - start_subject); |
2718 | cb.current_position = (int)(ptr - start_subject); |
2719 | cb.pattern_position = GET(code, LINK_SIZE + 3); |
2720 | cb.next_item_length = GET(code, 3 + 2*LINK_SIZE); |
2721 | cb.capture_top = 1; |
2722 | cb.capture_last = -1; |
2723 | cb.callout_data = md->callout_data; |
2724 | cb.mark = NULL; /* No (*MARK) support */ |
2725 | if ((rrc = (*PUBL(callout))(&cb)) < 0) return rrc; /* Abandon */ |
2726 | } |
2727 | if (rrc > 0) break; /* Fail this thread */ |
2728 | code += PRIV(OP_lengths)[OP_CALLOUT]; /* Skip callout data */ |
2729 | } |
2730 | |
2731 | condcode = code[LINK_SIZE+1]; |
2732 | |
2733 | /* Back reference conditions and duplicate named recursion conditions |
2734 | are not supported */ |
2735 | |
2736 | if (condcode == OP_CREF || condcode == OP_DNCREF || |
2737 | condcode == OP_DNRREF) |
2738 | return PCRE_ERROR_DFA_UCOND; |
2739 | |
2740 | /* The DEFINE condition is always false, and the assertion (?!) is |
2741 | converted to OP_FAIL. */ |
2742 | |
2743 | if (condcode == OP_DEF || condcode == OP_FAIL) |
2744 | { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); } |
2745 | |
2746 | /* The only supported version of OP_RREF is for the value RREF_ANY, |
2747 | which means "test if in any recursion". We can't test for specifically |
2748 | recursed groups. */ |
2749 | |
2750 | else if (condcode == OP_RREF) |
2751 | { |
2752 | int value = GET2(code, LINK_SIZE + 2); |
2753 | if (value != RREF_ANY) return PCRE_ERROR_DFA_UCOND; |
2754 | if (md->recursive != NULL) |
2755 | { ADD_ACTIVE(state_offset + LINK_SIZE + 2 + IMM2_SIZE, 0); } |
2756 | else { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); } |
2757 | } |
2758 | |
2759 | /* Otherwise, the condition is an assertion */ |
2760 | |
2761 | else |
2762 | { |
2763 | int rc; |
2764 | const pcre_uchar *asscode = code + LINK_SIZE + 1; |
2765 | const pcre_uchar *endasscode = asscode + GET(asscode, 1); |
2766 | |
2767 | while (*endasscode == OP_ALT) endasscode += GET(endasscode, 1); |
2768 | |
2769 | rc = internal_dfa_exec( |
2770 | md, /* fixed match data */ |
2771 | asscode, /* this subexpression's code */ |
2772 | ptr, /* where we currently are */ |
2773 | (int)(ptr - start_subject), /* start offset */ |
2774 | local_offsets, /* offset vector */ |
2775 | sizeof(local_offsets)/sizeof(int), /* size of same */ |
2776 | local_workspace, /* workspace vector */ |
2777 | sizeof(local_workspace)/sizeof(int), /* size of same */ |
2778 | rlevel); /* function recursion level */ |
2779 | |
2780 | if (rc == PCRE_ERROR_DFA_UITEM) return rc; |
2781 | if ((rc >= 0) == |
2782 | (condcode == OP_ASSERT || condcode == OP_ASSERTBACK)) |
2783 | { ADD_ACTIVE((int)(endasscode + LINK_SIZE + 1 - start_code), 0); } |
2784 | else |
2785 | { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); } |
2786 | } |
2787 | } |
2788 | break; |
2789 | |
2790 | /*-----------------------------------------------------------------*/ |
2791 | case OP_RECURSE: |
2792 | { |
2793 | dfa_recursion_info *ri; |
2794 | int local_offsets[1000]; |
2795 | int local_workspace[1000]; |
2796 | const pcre_uchar *callpat = start_code + GET(code, 1); |
2797 | int recno = (callpat == md->start_code)? 0 : |
2798 | GET2(callpat, 1 + LINK_SIZE); |
2799 | int rc; |
2800 | |
2801 | DPRINTF(("%.*sStarting regex recursion\n" , rlevel*2-2, SP)); |
2802 | |
2803 | /* Check for repeating a recursion without advancing the subject |
2804 | pointer. This should catch convoluted mutual recursions. (Some simple |
2805 | cases are caught at compile time.) */ |
2806 | |
2807 | for (ri = md->recursive; ri != NULL; ri = ri->prevrec) |
2808 | if (recno == ri->group_num && ptr == ri->subject_position) |
2809 | return PCRE_ERROR_RECURSELOOP; |
2810 | |
2811 | /* Remember this recursion and where we started it so as to |
2812 | catch infinite loops. */ |
2813 | |
2814 | new_recursive.group_num = recno; |
2815 | new_recursive.subject_position = ptr; |
2816 | new_recursive.prevrec = md->recursive; |
2817 | md->recursive = &new_recursive; |
2818 | |
2819 | rc = internal_dfa_exec( |
2820 | md, /* fixed match data */ |
2821 | callpat, /* this subexpression's code */ |
2822 | ptr, /* where we currently are */ |
2823 | (int)(ptr - start_subject), /* start offset */ |
2824 | local_offsets, /* offset vector */ |
2825 | sizeof(local_offsets)/sizeof(int), /* size of same */ |
2826 | local_workspace, /* workspace vector */ |
2827 | sizeof(local_workspace)/sizeof(int), /* size of same */ |
2828 | rlevel); /* function recursion level */ |
2829 | |
2830 | md->recursive = new_recursive.prevrec; /* Done this recursion */ |
2831 | |
2832 | DPRINTF(("%.*sReturn from regex recursion: rc=%d\n" , rlevel*2-2, SP, |
2833 | rc)); |
2834 | |
2835 | /* Ran out of internal offsets */ |
2836 | |
2837 | if (rc == 0) return PCRE_ERROR_DFA_RECURSE; |
2838 | |
2839 | /* For each successful matched substring, set up the next state with a |
2840 | count of characters to skip before trying it. Note that the count is in |
2841 | characters, not bytes. */ |
2842 | |
2843 | if (rc > 0) |
2844 | { |
2845 | for (rc = rc*2 - 2; rc >= 0; rc -= 2) |
2846 | { |
2847 | int charcount = local_offsets[rc+1] - local_offsets[rc]; |
2848 | #if defined SUPPORT_UTF && !defined COMPILE_PCRE32 |
2849 | if (utf) |
2850 | { |
2851 | const pcre_uchar *p = start_subject + local_offsets[rc]; |
2852 | const pcre_uchar *pp = start_subject + local_offsets[rc+1]; |
2853 | while (p < pp) if (NOT_FIRSTCHAR(*p++)) charcount--; |
2854 | } |
2855 | #endif |
2856 | if (charcount > 0) |
2857 | { |
2858 | ADD_NEW_DATA(-(state_offset + LINK_SIZE + 1), 0, (charcount - 1)); |
2859 | } |
2860 | else |
2861 | { |
2862 | ADD_ACTIVE(state_offset + LINK_SIZE + 1, 0); |
2863 | } |
2864 | } |
2865 | } |
2866 | else if (rc != PCRE_ERROR_NOMATCH) return rc; |
2867 | } |
2868 | break; |
2869 | |
2870 | /*-----------------------------------------------------------------*/ |
2871 | case OP_BRAPOS: |
2872 | case OP_SBRAPOS: |
2873 | case OP_CBRAPOS: |
2874 | case OP_SCBRAPOS: |
2875 | case OP_BRAPOSZERO: |
2876 | { |
2877 | int charcount, matched_count; |
2878 | const pcre_uchar *local_ptr = ptr; |
2879 | BOOL allow_zero; |
2880 | |
2881 | if (codevalue == OP_BRAPOSZERO) |
2882 | { |
2883 | allow_zero = TRUE; |
2884 | codevalue = *(++code); /* Codevalue will be one of above BRAs */ |
2885 | } |
2886 | else allow_zero = FALSE; |
2887 | |
2888 | /* Loop to match the subpattern as many times as possible as if it were |
2889 | a complete pattern. */ |
2890 | |
2891 | for (matched_count = 0;; matched_count++) |
2892 | { |
2893 | int local_offsets[2]; |
2894 | int local_workspace[1000]; |
2895 | |
2896 | int rc = internal_dfa_exec( |
2897 | md, /* fixed match data */ |
2898 | code, /* this subexpression's code */ |
2899 | local_ptr, /* where we currently are */ |
2900 | (int)(ptr - start_subject), /* start offset */ |
2901 | local_offsets, /* offset vector */ |
2902 | sizeof(local_offsets)/sizeof(int), /* size of same */ |
2903 | local_workspace, /* workspace vector */ |
2904 | sizeof(local_workspace)/sizeof(int), /* size of same */ |
2905 | rlevel); /* function recursion level */ |
2906 | |
2907 | /* Failed to match */ |
2908 | |
2909 | if (rc < 0) |
2910 | { |
2911 | if (rc != PCRE_ERROR_NOMATCH) return rc; |
2912 | break; |
2913 | } |
2914 | |
2915 | /* Matched: break the loop if zero characters matched. */ |
2916 | |
2917 | charcount = local_offsets[1] - local_offsets[0]; |
2918 | if (charcount == 0) break; |
2919 | local_ptr += charcount; /* Advance temporary position ptr */ |
2920 | } |
2921 | |
2922 | /* At this point we have matched the subpattern matched_count |
2923 | times, and local_ptr is pointing to the character after the end of the |
2924 | last match. */ |
2925 | |
2926 | if (matched_count > 0 || allow_zero) |
2927 | { |
2928 | const pcre_uchar *end_subpattern = code; |
2929 | int next_state_offset; |
2930 | |
2931 | do { end_subpattern += GET(end_subpattern, 1); } |
2932 | while (*end_subpattern == OP_ALT); |
2933 | next_state_offset = |
2934 | (int)(end_subpattern - start_code + LINK_SIZE + 1); |
2935 | |
2936 | /* Optimization: if there are no more active states, and there |
2937 | are no new states yet set up, then skip over the subject string |
2938 | right here, to save looping. Otherwise, set up the new state to swing |
2939 | into action when the end of the matched substring is reached. */ |
2940 | |
2941 | if (i + 1 >= active_count && new_count == 0) |
2942 | { |
2943 | ptr = local_ptr; |
2944 | clen = 0; |
2945 | ADD_NEW(next_state_offset, 0); |
2946 | } |
2947 | else |
2948 | { |
2949 | const pcre_uchar *p = ptr; |
2950 | const pcre_uchar *pp = local_ptr; |
2951 | charcount = (int)(pp - p); |
2952 | #if defined SUPPORT_UTF && !defined COMPILE_PCRE32 |
2953 | if (utf) while (p < pp) if (NOT_FIRSTCHAR(*p++)) charcount--; |
2954 | #endif |
2955 | ADD_NEW_DATA(-next_state_offset, 0, (charcount - 1)); |
2956 | } |
2957 | } |
2958 | } |
2959 | break; |
2960 | |
2961 | /*-----------------------------------------------------------------*/ |
2962 | case OP_ONCE: |
2963 | case OP_ONCE_NC: |
2964 | { |
2965 | int local_offsets[2]; |
2966 | int local_workspace[1000]; |
2967 | |
2968 | int rc = internal_dfa_exec( |
2969 | md, /* fixed match data */ |
2970 | code, /* this subexpression's code */ |
2971 | ptr, /* where we currently are */ |
2972 | (int)(ptr - start_subject), /* start offset */ |
2973 | local_offsets, /* offset vector */ |
2974 | sizeof(local_offsets)/sizeof(int), /* size of same */ |
2975 | local_workspace, /* workspace vector */ |
2976 | sizeof(local_workspace)/sizeof(int), /* size of same */ |
2977 | rlevel); /* function recursion level */ |
2978 | |
2979 | if (rc >= 0) |
2980 | { |
2981 | const pcre_uchar *end_subpattern = code; |
2982 | int charcount = local_offsets[1] - local_offsets[0]; |
2983 | int next_state_offset, repeat_state_offset; |
2984 | |
2985 | do { end_subpattern += GET(end_subpattern, 1); } |
2986 | while (*end_subpattern == OP_ALT); |
2987 | next_state_offset = |
2988 | (int)(end_subpattern - start_code + LINK_SIZE + 1); |
2989 | |
2990 | /* If the end of this subpattern is KETRMAX or KETRMIN, we must |
2991 | arrange for the repeat state also to be added to the relevant list. |
2992 | Calculate the offset, or set -1 for no repeat. */ |
2993 | |
2994 | repeat_state_offset = (*end_subpattern == OP_KETRMAX || |
2995 | *end_subpattern == OP_KETRMIN)? |
2996 | (int)(end_subpattern - start_code - GET(end_subpattern, 1)) : -1; |
2997 | |
2998 | /* If we have matched an empty string, add the next state at the |
2999 | current character pointer. This is important so that the duplicate |
3000 | checking kicks in, which is what breaks infinite loops that match an |
3001 | empty string. */ |
3002 | |
3003 | if (charcount == 0) |
3004 | { |
3005 | ADD_ACTIVE(next_state_offset, 0); |
3006 | } |
3007 | |
3008 | /* Optimization: if there are no more active states, and there |
3009 | are no new states yet set up, then skip over the subject string |
3010 | right here, to save looping. Otherwise, set up the new state to swing |
3011 | into action when the end of the matched substring is reached. */ |
3012 | |
3013 | else if (i + 1 >= active_count && new_count == 0) |
3014 | { |
3015 | ptr += charcount; |
3016 | clen = 0; |
3017 | ADD_NEW(next_state_offset, 0); |
3018 | |
3019 | /* If we are adding a repeat state at the new character position, |
3020 | we must fudge things so that it is the only current state. |
3021 | Otherwise, it might be a duplicate of one we processed before, and |
3022 | that would cause it to be skipped. */ |
3023 | |
3024 | if (repeat_state_offset >= 0) |
3025 | { |
3026 | next_active_state = active_states; |
3027 | active_count = 0; |
3028 | i = -1; |
3029 | ADD_ACTIVE(repeat_state_offset, 0); |
3030 | } |
3031 | } |
3032 | else |
3033 | { |
3034 | #if defined SUPPORT_UTF && !defined COMPILE_PCRE32 |
3035 | if (utf) |
3036 | { |
3037 | const pcre_uchar *p = start_subject + local_offsets[0]; |
3038 | const pcre_uchar *pp = start_subject + local_offsets[1]; |
3039 | while (p < pp) if (NOT_FIRSTCHAR(*p++)) charcount--; |
3040 | } |
3041 | #endif |
3042 | ADD_NEW_DATA(-next_state_offset, 0, (charcount - 1)); |
3043 | if (repeat_state_offset >= 0) |
3044 | { ADD_NEW_DATA(-repeat_state_offset, 0, (charcount - 1)); } |
3045 | } |
3046 | } |
3047 | else if (rc != PCRE_ERROR_NOMATCH) return rc; |
3048 | } |
3049 | break; |
3050 | |
3051 | |
3052 | /* ========================================================================== */ |
3053 | /* Handle callouts */ |
3054 | |
3055 | case OP_CALLOUT: |
3056 | rrc = 0; |
3057 | if (PUBL(callout) != NULL) |
3058 | { |
3059 | PUBL(callout_block) cb; |
3060 | cb.version = 1; /* Version 1 of the callout block */ |
3061 | cb.callout_number = code[1]; |
3062 | cb.offset_vector = offsets; |
3063 | #if defined COMPILE_PCRE8 |
3064 | cb.subject = (PCRE_SPTR)start_subject; |
3065 | #elif defined COMPILE_PCRE16 |
3066 | cb.subject = (PCRE_SPTR16)start_subject; |
3067 | #elif defined COMPILE_PCRE32 |
3068 | cb.subject = (PCRE_SPTR32)start_subject; |
3069 | #endif |
3070 | cb.subject_length = (int)(end_subject - start_subject); |
3071 | cb.start_match = (int)(current_subject - start_subject); |
3072 | cb.current_position = (int)(ptr - start_subject); |
3073 | cb.pattern_position = GET(code, 2); |
3074 | cb.next_item_length = GET(code, 2 + LINK_SIZE); |
3075 | cb.capture_top = 1; |
3076 | cb.capture_last = -1; |
3077 | cb.callout_data = md->callout_data; |
3078 | cb.mark = NULL; /* No (*MARK) support */ |
3079 | if ((rrc = (*PUBL(callout))(&cb)) < 0) return rrc; /* Abandon */ |
3080 | } |
3081 | if (rrc == 0) |
3082 | { ADD_ACTIVE(state_offset + PRIV(OP_lengths)[OP_CALLOUT], 0); } |
3083 | break; |
3084 | |
3085 | |
3086 | /* ========================================================================== */ |
3087 | default: /* Unsupported opcode */ |
3088 | return PCRE_ERROR_DFA_UITEM; |
3089 | } |
3090 | |
3091 | NEXT_ACTIVE_STATE: continue; |
3092 | |
3093 | } /* End of loop scanning active states */ |
3094 | |
3095 | /* We have finished the processing at the current subject character. If no |
3096 | new states have been set for the next character, we have found all the |
3097 | matches that we are going to find. If we are at the top level and partial |
3098 | matching has been requested, check for appropriate conditions. |
3099 | |
3100 | The "forced_ fail" variable counts the number of (*F) encountered for the |
3101 | character. If it is equal to the original active_count (saved in |
3102 | workspace[1]) it means that (*F) was found on every active state. In this |
3103 | case we don't want to give a partial match. |
3104 | |
3105 | The "could_continue" variable is true if a state could have continued but |
3106 | for the fact that the end of the subject was reached. */ |
3107 | |
3108 | if (new_count <= 0) |
3109 | { |
3110 | if (rlevel == 1 && /* Top level, and */ |
3111 | could_continue && /* Some could go on, and */ |
3112 | forced_fail != workspace[1] && /* Not all forced fail & */ |
3113 | ( /* either... */ |
3114 | (md->moptions & PCRE_PARTIAL_HARD) != 0 /* Hard partial */ |
3115 | || /* or... */ |
3116 | ((md->moptions & PCRE_PARTIAL_SOFT) != 0 && /* Soft partial and */ |
3117 | match_count < 0) /* no matches */ |
3118 | ) && /* And... */ |
3119 | ( |
3120 | partial_newline || /* Either partial NL */ |
3121 | ( /* or ... */ |
3122 | ptr >= end_subject && /* End of subject and */ |
3123 | ptr > md->start_used_ptr) /* Inspected non-empty string */ |
3124 | ) |
3125 | ) |
3126 | match_count = PCRE_ERROR_PARTIAL; |
3127 | DPRINTF(("%.*sEnd of internal_dfa_exec %d: returning %d\n" |
3128 | "%.*s---------------------\n\n" , rlevel*2-2, SP, rlevel, match_count, |
3129 | rlevel*2-2, SP)); |
3130 | break; /* In effect, "return", but see the comment below */ |
3131 | } |
3132 | |
3133 | /* One or more states are active for the next character. */ |
3134 | |
3135 | ptr += clen; /* Advance to next subject character */ |
3136 | } /* Loop to move along the subject string */ |
3137 | |
3138 | /* Control gets here from "break" a few lines above. We do it this way because |
3139 | if we use "return" above, we have compiler trouble. Some compilers warn if |
3140 | there's nothing here because they think the function doesn't return a value. On |
3141 | the other hand, if we put a dummy statement here, some more clever compilers |
3142 | complain that it can't be reached. Sigh. */ |
3143 | |
3144 | return match_count; |
3145 | } |
3146 | |
3147 | |
3148 | |
3149 | |
3150 | /************************************************* |
3151 | * Execute a Regular Expression - DFA engine * |
3152 | *************************************************/ |
3153 | |
3154 | /* This external function applies a compiled re to a subject string using a DFA |
3155 | engine. This function calls the internal function multiple times if the pattern |
3156 | is not anchored. |
3157 | |
3158 | Arguments: |
3159 | argument_re points to the compiled expression |
3160 | extra_data points to extra data or is NULL |
3161 | subject points to the subject string |
3162 | length length of subject string (may contain binary zeros) |
3163 | start_offset where to start in the subject string |
3164 | options option bits |
3165 | offsets vector of match offsets |
3166 | offsetcount size of same |
3167 | workspace workspace vector |
3168 | wscount size of same |
3169 | |
3170 | Returns: > 0 => number of match offset pairs placed in offsets |
3171 | = 0 => offsets overflowed; longest matches are present |
3172 | -1 => failed to match |
3173 | < -1 => some kind of unexpected problem |
3174 | */ |
3175 | |
3176 | #if defined COMPILE_PCRE8 |
3177 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
3178 | pcre_dfa_exec(const pcre *argument_re, const pcre_extra *, |
3179 | const char *subject, int length, int start_offset, int options, int *offsets, |
3180 | int offsetcount, int *workspace, int wscount) |
3181 | #elif defined COMPILE_PCRE16 |
3182 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
3183 | pcre16_dfa_exec(const pcre16 *argument_re, const pcre16_extra *extra_data, |
3184 | PCRE_SPTR16 subject, int length, int start_offset, int options, int *offsets, |
3185 | int offsetcount, int *workspace, int wscount) |
3186 | #elif defined COMPILE_PCRE32 |
3187 | PCRE_EXP_DEFN int PCRE_CALL_CONVENTION |
3188 | pcre32_dfa_exec(const pcre32 *argument_re, const pcre32_extra *extra_data, |
3189 | PCRE_SPTR32 subject, int length, int start_offset, int options, int *offsets, |
3190 | int offsetcount, int *workspace, int wscount) |
3191 | #endif |
3192 | { |
3193 | REAL_PCRE *re = (REAL_PCRE *)argument_re; |
3194 | dfa_match_data match_block; |
3195 | dfa_match_data *md = &match_block; |
3196 | BOOL utf, anchored, startline, firstline; |
3197 | const pcre_uchar *current_subject, *end_subject; |
3198 | const pcre_study_data *study = NULL; |
3199 | |
3200 | const pcre_uchar *req_char_ptr; |
3201 | const pcre_uint8 *start_bits = NULL; |
3202 | BOOL has_first_char = FALSE; |
3203 | BOOL has_req_char = FALSE; |
3204 | pcre_uchar first_char = 0; |
3205 | pcre_uchar first_char2 = 0; |
3206 | pcre_uchar req_char = 0; |
3207 | pcre_uchar req_char2 = 0; |
3208 | int newline; |
3209 | |
3210 | /* Plausibility checks */ |
3211 | |
3212 | if ((options & ~PUBLIC_DFA_EXEC_OPTIONS) != 0) return PCRE_ERROR_BADOPTION; |
3213 | if (re == NULL || subject == NULL || workspace == NULL || |
3214 | (offsets == NULL && offsetcount > 0)) return PCRE_ERROR_NULL; |
3215 | if (offsetcount < 0) return PCRE_ERROR_BADCOUNT; |
3216 | if (wscount < 20) return PCRE_ERROR_DFA_WSSIZE; |
3217 | if (length < 0) return PCRE_ERROR_BADLENGTH; |
3218 | if (start_offset < 0 || start_offset > length) return PCRE_ERROR_BADOFFSET; |
3219 | |
3220 | /* Check that the first field in the block is the magic number. If it is not, |
3221 | return with PCRE_ERROR_BADMAGIC. However, if the magic number is equal to |
3222 | REVERSED_MAGIC_NUMBER we return with PCRE_ERROR_BADENDIANNESS, which |
3223 | means that the pattern is likely compiled with different endianness. */ |
3224 | |
3225 | if (re->magic_number != MAGIC_NUMBER) |
3226 | return re->magic_number == REVERSED_MAGIC_NUMBER? |
3227 | PCRE_ERROR_BADENDIANNESS:PCRE_ERROR_BADMAGIC; |
3228 | if ((re->flags & PCRE_MODE) == 0) return PCRE_ERROR_BADMODE; |
3229 | |
3230 | /* If restarting after a partial match, do some sanity checks on the contents |
3231 | of the workspace. */ |
3232 | |
3233 | if ((options & PCRE_DFA_RESTART) != 0) |
3234 | { |
3235 | if ((workspace[0] & (-2)) != 0 || workspace[1] < 1 || |
3236 | workspace[1] > (wscount - 2)/INTS_PER_STATEBLOCK) |
3237 | return PCRE_ERROR_DFA_BADRESTART; |
3238 | } |
3239 | |
3240 | /* Set up study, callout, and table data */ |
3241 | |
3242 | md->tables = re->tables; |
3243 | md->callout_data = NULL; |
3244 | |
3245 | if (extra_data != NULL) |
3246 | { |
3247 | unsigned long int flags = extra_data->flags; |
3248 | if ((flags & PCRE_EXTRA_STUDY_DATA) != 0) |
3249 | study = (const pcre_study_data *)extra_data->study_data; |
3250 | if ((flags & PCRE_EXTRA_MATCH_LIMIT) != 0) return PCRE_ERROR_DFA_UMLIMIT; |
3251 | if ((flags & PCRE_EXTRA_MATCH_LIMIT_RECURSION) != 0) |
3252 | return PCRE_ERROR_DFA_UMLIMIT; |
3253 | if ((flags & PCRE_EXTRA_CALLOUT_DATA) != 0) |
3254 | md->callout_data = extra_data->callout_data; |
3255 | if ((flags & PCRE_EXTRA_TABLES) != 0) |
3256 | md->tables = extra_data->tables; |
3257 | } |
3258 | |
3259 | /* Set some local values */ |
3260 | |
3261 | current_subject = (const pcre_uchar *)subject + start_offset; |
3262 | end_subject = (const pcre_uchar *)subject + length; |
3263 | req_char_ptr = current_subject - 1; |
3264 | |
3265 | #ifdef SUPPORT_UTF |
3266 | /* PCRE_UTF(16|32) have the same value as PCRE_UTF8. */ |
3267 | utf = (re->options & PCRE_UTF8) != 0; |
3268 | #else |
3269 | utf = FALSE; |
3270 | #endif |
3271 | |
3272 | anchored = (options & (PCRE_ANCHORED|PCRE_DFA_RESTART)) != 0 || |
3273 | (re->options & PCRE_ANCHORED) != 0; |
3274 | |
3275 | /* The remaining fixed data for passing around. */ |
3276 | |
3277 | md->start_code = (const pcre_uchar *)argument_re + |
3278 | re->name_table_offset + re->name_count * re->name_entry_size; |
3279 | md->start_subject = (const pcre_uchar *)subject; |
3280 | md->end_subject = end_subject; |
3281 | md->start_offset = start_offset; |
3282 | md->moptions = options; |
3283 | md->poptions = re->options; |
3284 | |
3285 | /* If the BSR option is not set at match time, copy what was set |
3286 | at compile time. */ |
3287 | |
3288 | if ((md->moptions & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) == 0) |
3289 | { |
3290 | if ((re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) != 0) |
3291 | md->moptions |= re->options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE); |
3292 | #ifdef BSR_ANYCRLF |
3293 | else md->moptions |= PCRE_BSR_ANYCRLF; |
3294 | #endif |
3295 | } |
3296 | |
3297 | /* Handle different types of newline. The three bits give eight cases. If |
3298 | nothing is set at run time, whatever was used at compile time applies. */ |
3299 | |
3300 | switch ((((options & PCRE_NEWLINE_BITS) == 0)? re->options : (pcre_uint32)options) & |
3301 | PCRE_NEWLINE_BITS) |
3302 | { |
3303 | case 0: newline = NEWLINE; break; /* Compile-time default */ |
3304 | case PCRE_NEWLINE_CR: newline = CHAR_CR; break; |
3305 | case PCRE_NEWLINE_LF: newline = CHAR_NL; break; |
3306 | case PCRE_NEWLINE_CR+ |
3307 | PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break; |
3308 | case PCRE_NEWLINE_ANY: newline = -1; break; |
3309 | case PCRE_NEWLINE_ANYCRLF: newline = -2; break; |
3310 | default: return PCRE_ERROR_BADNEWLINE; |
3311 | } |
3312 | |
3313 | if (newline == -2) |
3314 | { |
3315 | md->nltype = NLTYPE_ANYCRLF; |
3316 | } |
3317 | else if (newline < 0) |
3318 | { |
3319 | md->nltype = NLTYPE_ANY; |
3320 | } |
3321 | else |
3322 | { |
3323 | md->nltype = NLTYPE_FIXED; |
3324 | if (newline > 255) |
3325 | { |
3326 | md->nllen = 2; |
3327 | md->nl[0] = (newline >> 8) & 255; |
3328 | md->nl[1] = newline & 255; |
3329 | } |
3330 | else |
3331 | { |
3332 | md->nllen = 1; |
3333 | md->nl[0] = newline; |
3334 | } |
3335 | } |
3336 | |
3337 | /* Check a UTF-8 string if required. Unfortunately there's no way of passing |
3338 | back the character offset. */ |
3339 | |
3340 | #ifdef SUPPORT_UTF |
3341 | if (utf && (options & PCRE_NO_UTF8_CHECK) == 0) |
3342 | { |
3343 | int erroroffset; |
3344 | int errorcode = PRIV(valid_utf)((pcre_uchar *)subject, length, &erroroffset); |
3345 | if (errorcode != 0) |
3346 | { |
3347 | if (offsetcount >= 2) |
3348 | { |
3349 | offsets[0] = erroroffset; |
3350 | offsets[1] = errorcode; |
3351 | } |
3352 | #if defined COMPILE_PCRE8 |
3353 | return (errorcode <= PCRE_UTF8_ERR5 && (options & PCRE_PARTIAL_HARD) != 0) ? |
3354 | PCRE_ERROR_SHORTUTF8 : PCRE_ERROR_BADUTF8; |
3355 | #elif defined COMPILE_PCRE16 |
3356 | return (errorcode <= PCRE_UTF16_ERR1 && (options & PCRE_PARTIAL_HARD) != 0) ? |
3357 | PCRE_ERROR_SHORTUTF16 : PCRE_ERROR_BADUTF16; |
3358 | #elif defined COMPILE_PCRE32 |
3359 | return PCRE_ERROR_BADUTF32; |
3360 | #endif |
3361 | } |
3362 | #if defined COMPILE_PCRE8 || defined COMPILE_PCRE16 |
3363 | if (start_offset > 0 && start_offset < length && |
3364 | NOT_FIRSTCHAR(((PCRE_PUCHAR)subject)[start_offset])) |
3365 | return PCRE_ERROR_BADUTF8_OFFSET; |
3366 | #endif |
3367 | } |
3368 | #endif |
3369 | |
3370 | /* If the exec call supplied NULL for tables, use the inbuilt ones. This |
3371 | is a feature that makes it possible to save compiled regex and re-use them |
3372 | in other programs later. */ |
3373 | |
3374 | if (md->tables == NULL) md->tables = PRIV(default_tables); |
3375 | |
3376 | /* The "must be at the start of a line" flags are used in a loop when finding |
3377 | where to start. */ |
3378 | |
3379 | startline = (re->flags & PCRE_STARTLINE) != 0; |
3380 | firstline = (re->options & PCRE_FIRSTLINE) != 0; |
3381 | |
3382 | /* Set up the first character to match, if available. The first_byte value is |
3383 | never set for an anchored regular expression, but the anchoring may be forced |
3384 | at run time, so we have to test for anchoring. The first char may be unset for |
3385 | an unanchored pattern, of course. If there's no first char and the pattern was |
3386 | studied, there may be a bitmap of possible first characters. */ |
3387 | |
3388 | if (!anchored) |
3389 | { |
3390 | if ((re->flags & PCRE_FIRSTSET) != 0) |
3391 | { |
3392 | has_first_char = TRUE; |
3393 | first_char = first_char2 = (pcre_uchar)(re->first_char); |
3394 | if ((re->flags & PCRE_FCH_CASELESS) != 0) |
3395 | { |
3396 | first_char2 = TABLE_GET(first_char, md->tables + fcc_offset, first_char); |
3397 | #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8) |
3398 | if (utf && first_char > 127) |
3399 | first_char2 = UCD_OTHERCASE(first_char); |
3400 | #endif |
3401 | } |
3402 | } |
3403 | else |
3404 | { |
3405 | if (!startline && study != NULL && |
3406 | (study->flags & PCRE_STUDY_MAPPED) != 0) |
3407 | start_bits = study->start_bits; |
3408 | } |
3409 | } |
3410 | |
3411 | /* For anchored or unanchored matches, there may be a "last known required |
3412 | character" set. */ |
3413 | |
3414 | if ((re->flags & PCRE_REQCHSET) != 0) |
3415 | { |
3416 | has_req_char = TRUE; |
3417 | req_char = req_char2 = (pcre_uchar)(re->req_char); |
3418 | if ((re->flags & PCRE_RCH_CASELESS) != 0) |
3419 | { |
3420 | req_char2 = TABLE_GET(req_char, md->tables + fcc_offset, req_char); |
3421 | #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8) |
3422 | if (utf && req_char > 127) |
3423 | req_char2 = UCD_OTHERCASE(req_char); |
3424 | #endif |
3425 | } |
3426 | } |
3427 | |
3428 | /* Call the main matching function, looping for a non-anchored regex after a |
3429 | failed match. If not restarting, perform certain optimizations at the start of |
3430 | a match. */ |
3431 | |
3432 | for (;;) |
3433 | { |
3434 | int rc; |
3435 | |
3436 | if ((options & PCRE_DFA_RESTART) == 0) |
3437 | { |
3438 | const pcre_uchar *save_end_subject = end_subject; |
3439 | |
3440 | /* If firstline is TRUE, the start of the match is constrained to the first |
3441 | line of a multiline string. Implement this by temporarily adjusting |
3442 | end_subject so that we stop scanning at a newline. If the match fails at |
3443 | the newline, later code breaks this loop. */ |
3444 | |
3445 | if (firstline) |
3446 | { |
3447 | PCRE_PUCHAR t = current_subject; |
3448 | #ifdef SUPPORT_UTF |
3449 | if (utf) |
3450 | { |
3451 | while (t < md->end_subject && !IS_NEWLINE(t)) |
3452 | { |
3453 | t++; |
3454 | ACROSSCHAR(t < end_subject, *t, t++); |
3455 | } |
3456 | } |
3457 | else |
3458 | #endif |
3459 | while (t < md->end_subject && !IS_NEWLINE(t)) t++; |
3460 | end_subject = t; |
3461 | } |
3462 | |
3463 | /* There are some optimizations that avoid running the match if a known |
3464 | starting point is not found. However, there is an option that disables |
3465 | these, for testing and for ensuring that all callouts do actually occur. |
3466 | The option can be set in the regex by (*NO_START_OPT) or passed in |
3467 | match-time options. */ |
3468 | |
3469 | if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0) |
3470 | { |
3471 | /* Advance to a known first pcre_uchar (i.e. data item) */ |
3472 | |
3473 | if (has_first_char) |
3474 | { |
3475 | if (first_char != first_char2) |
3476 | { |
3477 | pcre_uchar csc; |
3478 | while (current_subject < end_subject && |
3479 | (csc = UCHAR21TEST(current_subject)) != first_char && csc != first_char2) |
3480 | current_subject++; |
3481 | } |
3482 | else |
3483 | while (current_subject < end_subject && |
3484 | UCHAR21TEST(current_subject) != first_char) |
3485 | current_subject++; |
3486 | } |
3487 | |
3488 | /* Or to just after a linebreak for a multiline match if possible */ |
3489 | |
3490 | else if (startline) |
3491 | { |
3492 | if (current_subject > md->start_subject + start_offset) |
3493 | { |
3494 | #ifdef SUPPORT_UTF |
3495 | if (utf) |
3496 | { |
3497 | while (current_subject < end_subject && |
3498 | !WAS_NEWLINE(current_subject)) |
3499 | { |
3500 | current_subject++; |
3501 | ACROSSCHAR(current_subject < end_subject, *current_subject, |
3502 | current_subject++); |
3503 | } |
3504 | } |
3505 | else |
3506 | #endif |
3507 | while (current_subject < end_subject && !WAS_NEWLINE(current_subject)) |
3508 | current_subject++; |
3509 | |
3510 | /* If we have just passed a CR and the newline option is ANY or |
3511 | ANYCRLF, and we are now at a LF, advance the match position by one |
3512 | more character. */ |
3513 | |
3514 | if (UCHAR21TEST(current_subject - 1) == CHAR_CR && |
3515 | (md->nltype == NLTYPE_ANY || md->nltype == NLTYPE_ANYCRLF) && |
3516 | current_subject < end_subject && |
3517 | UCHAR21TEST(current_subject) == CHAR_NL) |
3518 | current_subject++; |
3519 | } |
3520 | } |
3521 | |
3522 | /* Advance to a non-unique first pcre_uchar after study */ |
3523 | |
3524 | else if (start_bits != NULL) |
3525 | { |
3526 | while (current_subject < end_subject) |
3527 | { |
3528 | register pcre_uint32 c = UCHAR21TEST(current_subject); |
3529 | #ifndef COMPILE_PCRE8 |
3530 | if (c > 255) c = 255; |
3531 | #endif |
3532 | if ((start_bits[c/8] & (1 << (c&7))) != 0) break; |
3533 | current_subject++; |
3534 | } |
3535 | } |
3536 | } |
3537 | |
3538 | /* Restore fudged end_subject */ |
3539 | |
3540 | end_subject = save_end_subject; |
3541 | |
3542 | /* The following two optimizations are disabled for partial matching or if |
3543 | disabling is explicitly requested (and of course, by the test above, this |
3544 | code is not obeyed when restarting after a partial match). */ |
3545 | |
3546 | if (((options | re->options) & PCRE_NO_START_OPTIMIZE) == 0 && |
3547 | (options & (PCRE_PARTIAL_HARD|PCRE_PARTIAL_SOFT)) == 0) |
3548 | { |
3549 | /* If the pattern was studied, a minimum subject length may be set. This |
3550 | is a lower bound; no actual string of that length may actually match the |
3551 | pattern. Although the value is, strictly, in characters, we treat it as |
3552 | in pcre_uchar units to avoid spending too much time in this optimization. |
3553 | */ |
3554 | |
3555 | if (study != NULL && (study->flags & PCRE_STUDY_MINLEN) != 0 && |
3556 | (pcre_uint32)(end_subject - current_subject) < study->minlength) |
3557 | return PCRE_ERROR_NOMATCH; |
3558 | |
3559 | /* If req_char is set, we know that that pcre_uchar must appear in the |
3560 | subject for the match to succeed. If the first pcre_uchar is set, |
3561 | req_char must be later in the subject; otherwise the test starts at the |
3562 | match point. This optimization can save a huge amount of work in patterns |
3563 | with nested unlimited repeats that aren't going to match. Writing |
3564 | separate code for cased/caseless versions makes it go faster, as does |
3565 | using an autoincrement and backing off on a match. |
3566 | |
3567 | HOWEVER: when the subject string is very, very long, searching to its end |
3568 | can take a long time, and give bad performance on quite ordinary |
3569 | patterns. This showed up when somebody was matching /^C/ on a 32-megabyte |
3570 | string... so we don't do this when the string is sufficiently long. */ |
3571 | |
3572 | if (has_req_char && end_subject - current_subject < REQ_BYTE_MAX) |
3573 | { |
3574 | register PCRE_PUCHAR p = current_subject + (has_first_char? 1:0); |
3575 | |
3576 | /* We don't need to repeat the search if we haven't yet reached the |
3577 | place we found it at last time. */ |
3578 | |
3579 | if (p > req_char_ptr) |
3580 | { |
3581 | if (req_char != req_char2) |
3582 | { |
3583 | while (p < end_subject) |
3584 | { |
3585 | register pcre_uint32 pp = UCHAR21INCTEST(p); |
3586 | if (pp == req_char || pp == req_char2) { p--; break; } |
3587 | } |
3588 | } |
3589 | else |
3590 | { |
3591 | while (p < end_subject) |
3592 | { |
3593 | if (UCHAR21INCTEST(p) == req_char) { p--; break; } |
3594 | } |
3595 | } |
3596 | |
3597 | /* If we can't find the required pcre_uchar, break the matching loop, |
3598 | which will cause a return or PCRE_ERROR_NOMATCH. */ |
3599 | |
3600 | if (p >= end_subject) break; |
3601 | |
3602 | /* If we have found the required pcre_uchar, save the point where we |
3603 | found it, so that we don't search again next time round the loop if |
3604 | the start hasn't passed this point yet. */ |
3605 | |
3606 | req_char_ptr = p; |
3607 | } |
3608 | } |
3609 | } |
3610 | } /* End of optimizations that are done when not restarting */ |
3611 | |
3612 | /* OK, now we can do the business */ |
3613 | |
3614 | md->start_used_ptr = current_subject; |
3615 | md->recursive = NULL; |
3616 | |
3617 | rc = internal_dfa_exec( |
3618 | md, /* fixed match data */ |
3619 | md->start_code, /* this subexpression's code */ |
3620 | current_subject, /* where we currently are */ |
3621 | start_offset, /* start offset in subject */ |
3622 | offsets, /* offset vector */ |
3623 | offsetcount, /* size of same */ |
3624 | workspace, /* workspace vector */ |
3625 | wscount, /* size of same */ |
3626 | 0); /* function recurse level */ |
3627 | |
3628 | /* Anything other than "no match" means we are done, always; otherwise, carry |
3629 | on only if not anchored. */ |
3630 | |
3631 | if (rc != PCRE_ERROR_NOMATCH || anchored) |
3632 | { |
3633 | if (rc == PCRE_ERROR_PARTIAL && offsetcount >= 2) |
3634 | { |
3635 | offsets[0] = (int)(md->start_used_ptr - (PCRE_PUCHAR)subject); |
3636 | offsets[1] = (int)(end_subject - (PCRE_PUCHAR)subject); |
3637 | if (offsetcount > 2) |
3638 | offsets[2] = (int)(current_subject - (PCRE_PUCHAR)subject); |
3639 | } |
3640 | return rc; |
3641 | } |
3642 | |
3643 | /* Advance to the next subject character unless we are at the end of a line |
3644 | and firstline is set. */ |
3645 | |
3646 | if (firstline && IS_NEWLINE(current_subject)) break; |
3647 | current_subject++; |
3648 | #ifdef SUPPORT_UTF |
3649 | if (utf) |
3650 | { |
3651 | ACROSSCHAR(current_subject < end_subject, *current_subject, |
3652 | current_subject++); |
3653 | } |
3654 | #endif |
3655 | if (current_subject > end_subject) break; |
3656 | |
3657 | /* If we have just passed a CR and we are now at a LF, and the pattern does |
3658 | not contain any explicit matches for \r or \n, and the newline option is CRLF |
3659 | or ANY or ANYCRLF, advance the match position by one more character. */ |
3660 | |
3661 | if (UCHAR21TEST(current_subject - 1) == CHAR_CR && |
3662 | current_subject < end_subject && |
3663 | UCHAR21TEST(current_subject) == CHAR_NL && |
3664 | (re->flags & PCRE_HASCRORLF) == 0 && |
3665 | (md->nltype == NLTYPE_ANY || |
3666 | md->nltype == NLTYPE_ANYCRLF || |
3667 | md->nllen == 2)) |
3668 | current_subject++; |
3669 | |
3670 | } /* "Bumpalong" loop */ |
3671 | |
3672 | return PCRE_ERROR_NOMATCH; |
3673 | } |
3674 | |
3675 | /* End of pcre_dfa_exec.c */ |
3676 | |