1 | /************************************************* |
2 | * Perl-Compatible Regular Expressions * |
3 | *************************************************/ |
4 | |
5 | /* PCRE is a library of functions to support regular expressions whose syntax |
6 | and semantics are as close as possible to those of the Perl 5 language. |
7 | |
8 | Written by Philip Hazel |
9 | Original API code Copyright (c) 1997-2012 University of Cambridge |
10 | New API code Copyright (c) 2016-2022 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 | |
42 | /* This module contains the external function pcre2_dfa_match(), which is an |
43 | alternative matching function that uses a sort of DFA algorithm (not a true |
44 | FSM). This is NOT Perl-compatible, but it has advantages in certain |
45 | applications. */ |
46 | |
47 | |
48 | /* NOTE ABOUT PERFORMANCE: A user of this function sent some code that improved |
49 | the performance of his patterns greatly. I could not use it as it stood, as it |
50 | was not thread safe, and made assumptions about pattern sizes. Also, it caused |
51 | test 7 to loop, and test 9 to crash with a segfault. |
52 | |
53 | The issue is the check for duplicate states, which is done by a simple linear |
54 | search up the state list. (Grep for "duplicate" below to find the code.) For |
55 | many patterns, there will never be many states active at one time, so a simple |
56 | linear search is fine. In patterns that have many active states, it might be a |
57 | bottleneck. The suggested code used an indexing scheme to remember which states |
58 | had previously been used for each character, and avoided the linear search when |
59 | it knew there was no chance of a duplicate. This was implemented when adding |
60 | states to the state lists. |
61 | |
62 | I wrote some thread-safe, not-limited code to try something similar at the time |
63 | of checking for duplicates (instead of when adding states), using index vectors |
64 | on the stack. It did give a 13% improvement with one specially constructed |
65 | pattern for certain subject strings, but on other strings and on many of the |
66 | simpler patterns in the test suite it did worse. The major problem, I think, |
67 | was the extra time to initialize the index. This had to be done for each call |
68 | of internal_dfa_match(). (The supplied patch used a static vector, initialized |
69 | only once - I suspect this was the cause of the problems with the tests.) |
70 | |
71 | Overall, I concluded that the gains in some cases did not outweigh the losses |
72 | in others, so I abandoned this code. */ |
73 | |
74 | |
75 | #ifdef HAVE_CONFIG_H |
76 | #include "config.h" |
77 | #endif |
78 | |
79 | #define NLBLOCK mb /* Block containing newline information */ |
80 | #define PSSTART start_subject /* Field containing processed string start */ |
81 | #define PSEND end_subject /* Field containing processed string end */ |
82 | |
83 | #include "pcre2_internal.h" |
84 | |
85 | #define PUBLIC_DFA_MATCH_OPTIONS \ |
86 | (PCRE2_ANCHORED|PCRE2_ENDANCHORED|PCRE2_NOTBOL|PCRE2_NOTEOL|PCRE2_NOTEMPTY| \ |
87 | PCRE2_NOTEMPTY_ATSTART|PCRE2_NO_UTF_CHECK|PCRE2_PARTIAL_HARD| \ |
88 | PCRE2_PARTIAL_SOFT|PCRE2_DFA_SHORTEST|PCRE2_DFA_RESTART| \ |
89 | PCRE2_COPY_MATCHED_SUBJECT) |
90 | |
91 | |
92 | /************************************************* |
93 | * Code parameters and static tables * |
94 | *************************************************/ |
95 | |
96 | /* These are offsets that are used to turn the OP_TYPESTAR and friends opcodes |
97 | into others, under special conditions. A gap of 20 between the blocks should be |
98 | enough. The resulting opcodes don't have to be less than 256 because they are |
99 | never stored, so we push them well clear of the normal opcodes. */ |
100 | |
101 | #define 300 |
102 | #define 320 |
103 | #define 340 |
104 | #define 360 |
105 | #define 380 |
106 | |
107 | |
108 | /* This table identifies those opcodes that are followed immediately by a |
109 | character that is to be tested in some way. This makes it possible to |
110 | centralize the loading of these characters. In the case of Type * etc, the |
111 | "character" is the opcode for \D, \d, \S, \s, \W, or \w, which will always be a |
112 | small value. Non-zero values in the table are the offsets from the opcode where |
113 | the character is to be found. ***NOTE*** If the start of this table is |
114 | modified, the three tables that follow must also be modified. */ |
115 | |
116 | static const uint8_t coptable[] = { |
117 | 0, /* End */ |
118 | 0, 0, 0, 0, 0, /* \A, \G, \K, \B, \b */ |
119 | 0, 0, 0, 0, 0, 0, /* \D, \d, \S, \s, \W, \w */ |
120 | 0, 0, 0, /* Any, AllAny, Anybyte */ |
121 | 0, 0, /* \P, \p */ |
122 | 0, 0, 0, 0, 0, /* \R, \H, \h, \V, \v */ |
123 | 0, /* \X */ |
124 | 0, 0, 0, 0, 0, 0, /* \Z, \z, $, $M, ^, ^M */ |
125 | 1, /* Char */ |
126 | 1, /* Chari */ |
127 | 1, /* not */ |
128 | 1, /* noti */ |
129 | /* Positive single-char repeats */ |
130 | 1, 1, 1, 1, 1, 1, /* *, *?, +, +?, ?, ?? */ |
131 | 1+IMM2_SIZE, 1+IMM2_SIZE, /* upto, minupto */ |
132 | 1+IMM2_SIZE, /* exact */ |
133 | 1, 1, 1, 1+IMM2_SIZE, /* *+, ++, ?+, upto+ */ |
134 | 1, 1, 1, 1, 1, 1, /* *I, *?I, +I, +?I, ?I, ??I */ |
135 | 1+IMM2_SIZE, 1+IMM2_SIZE, /* upto I, minupto I */ |
136 | 1+IMM2_SIZE, /* exact I */ |
137 | 1, 1, 1, 1+IMM2_SIZE, /* *+I, ++I, ?+I, upto+I */ |
138 | /* Negative single-char repeats - only for chars < 256 */ |
139 | 1, 1, 1, 1, 1, 1, /* NOT *, *?, +, +?, ?, ?? */ |
140 | 1+IMM2_SIZE, 1+IMM2_SIZE, /* NOT upto, minupto */ |
141 | 1+IMM2_SIZE, /* NOT exact */ |
142 | 1, 1, 1, 1+IMM2_SIZE, /* NOT *+, ++, ?+, upto+ */ |
143 | 1, 1, 1, 1, 1, 1, /* NOT *I, *?I, +I, +?I, ?I, ??I */ |
144 | 1+IMM2_SIZE, 1+IMM2_SIZE, /* NOT upto I, minupto I */ |
145 | 1+IMM2_SIZE, /* NOT exact I */ |
146 | 1, 1, 1, 1+IMM2_SIZE, /* NOT *+I, ++I, ?+I, upto+I */ |
147 | /* Positive type repeats */ |
148 | 1, 1, 1, 1, 1, 1, /* Type *, *?, +, +?, ?, ?? */ |
149 | 1+IMM2_SIZE, 1+IMM2_SIZE, /* Type upto, minupto */ |
150 | 1+IMM2_SIZE, /* Type exact */ |
151 | 1, 1, 1, 1+IMM2_SIZE, /* Type *+, ++, ?+, upto+ */ |
152 | /* Character class & ref repeats */ |
153 | 0, 0, 0, 0, 0, 0, /* *, *?, +, +?, ?, ?? */ |
154 | 0, 0, /* CRRANGE, CRMINRANGE */ |
155 | 0, 0, 0, 0, /* Possessive *+, ++, ?+, CRPOSRANGE */ |
156 | 0, /* CLASS */ |
157 | 0, /* NCLASS */ |
158 | 0, /* XCLASS - variable length */ |
159 | 0, /* REF */ |
160 | 0, /* REFI */ |
161 | 0, /* DNREF */ |
162 | 0, /* DNREFI */ |
163 | 0, /* RECURSE */ |
164 | 0, /* CALLOUT */ |
165 | 0, /* CALLOUT_STR */ |
166 | 0, /* Alt */ |
167 | 0, /* Ket */ |
168 | 0, /* KetRmax */ |
169 | 0, /* KetRmin */ |
170 | 0, /* KetRpos */ |
171 | 0, /* Reverse */ |
172 | 0, /* Assert */ |
173 | 0, /* Assert not */ |
174 | 0, /* Assert behind */ |
175 | 0, /* Assert behind not */ |
176 | 0, /* NA assert */ |
177 | 0, /* NA assert behind */ |
178 | 0, /* ONCE */ |
179 | 0, /* SCRIPT_RUN */ |
180 | 0, 0, 0, 0, 0, /* BRA, BRAPOS, CBRA, CBRAPOS, COND */ |
181 | 0, 0, 0, 0, 0, /* SBRA, SBRAPOS, SCBRA, SCBRAPOS, SCOND */ |
182 | 0, 0, /* CREF, DNCREF */ |
183 | 0, 0, /* RREF, DNRREF */ |
184 | 0, 0, /* FALSE, TRUE */ |
185 | 0, 0, 0, /* BRAZERO, BRAMINZERO, BRAPOSZERO */ |
186 | 0, 0, 0, /* MARK, PRUNE, PRUNE_ARG */ |
187 | 0, 0, 0, 0, /* SKIP, SKIP_ARG, THEN, THEN_ARG */ |
188 | 0, 0, /* COMMIT, COMMIT_ARG */ |
189 | 0, 0, 0, /* FAIL, ACCEPT, ASSERT_ACCEPT */ |
190 | 0, 0, 0 /* CLOSE, SKIPZERO, DEFINE */ |
191 | }; |
192 | |
193 | /* This table identifies those opcodes that inspect a character. It is used to |
194 | remember the fact that a character could have been inspected when the end of |
195 | the subject is reached. ***NOTE*** If the start of this table is modified, the |
196 | two tables that follow must also be modified. */ |
197 | |
198 | static const uint8_t poptable[] = { |
199 | 0, /* End */ |
200 | 0, 0, 0, 1, 1, /* \A, \G, \K, \B, \b */ |
201 | 1, 1, 1, 1, 1, 1, /* \D, \d, \S, \s, \W, \w */ |
202 | 1, 1, 1, /* Any, AllAny, Anybyte */ |
203 | 1, 1, /* \P, \p */ |
204 | 1, 1, 1, 1, 1, /* \R, \H, \h, \V, \v */ |
205 | 1, /* \X */ |
206 | 0, 0, 0, 0, 0, 0, /* \Z, \z, $, $M, ^, ^M */ |
207 | 1, /* Char */ |
208 | 1, /* Chari */ |
209 | 1, /* not */ |
210 | 1, /* noti */ |
211 | /* Positive single-char repeats */ |
212 | 1, 1, 1, 1, 1, 1, /* *, *?, +, +?, ?, ?? */ |
213 | 1, 1, 1, /* upto, minupto, exact */ |
214 | 1, 1, 1, 1, /* *+, ++, ?+, upto+ */ |
215 | 1, 1, 1, 1, 1, 1, /* *I, *?I, +I, +?I, ?I, ??I */ |
216 | 1, 1, 1, /* upto I, minupto I, exact I */ |
217 | 1, 1, 1, 1, /* *+I, ++I, ?+I, upto+I */ |
218 | /* Negative single-char repeats - only for chars < 256 */ |
219 | 1, 1, 1, 1, 1, 1, /* NOT *, *?, +, +?, ?, ?? */ |
220 | 1, 1, 1, /* NOT upto, minupto, exact */ |
221 | 1, 1, 1, 1, /* NOT *+, ++, ?+, upto+ */ |
222 | 1, 1, 1, 1, 1, 1, /* NOT *I, *?I, +I, +?I, ?I, ??I */ |
223 | 1, 1, 1, /* NOT upto I, minupto I, exact I */ |
224 | 1, 1, 1, 1, /* NOT *+I, ++I, ?+I, upto+I */ |
225 | /* Positive type repeats */ |
226 | 1, 1, 1, 1, 1, 1, /* Type *, *?, +, +?, ?, ?? */ |
227 | 1, 1, 1, /* Type upto, minupto, exact */ |
228 | 1, 1, 1, 1, /* Type *+, ++, ?+, upto+ */ |
229 | /* Character class & ref repeats */ |
230 | 1, 1, 1, 1, 1, 1, /* *, *?, +, +?, ?, ?? */ |
231 | 1, 1, /* CRRANGE, CRMINRANGE */ |
232 | 1, 1, 1, 1, /* Possessive *+, ++, ?+, CRPOSRANGE */ |
233 | 1, /* CLASS */ |
234 | 1, /* NCLASS */ |
235 | 1, /* XCLASS - variable length */ |
236 | 0, /* REF */ |
237 | 0, /* REFI */ |
238 | 0, /* DNREF */ |
239 | 0, /* DNREFI */ |
240 | 0, /* RECURSE */ |
241 | 0, /* CALLOUT */ |
242 | 0, /* CALLOUT_STR */ |
243 | 0, /* Alt */ |
244 | 0, /* Ket */ |
245 | 0, /* KetRmax */ |
246 | 0, /* KetRmin */ |
247 | 0, /* KetRpos */ |
248 | 0, /* Reverse */ |
249 | 0, /* Assert */ |
250 | 0, /* Assert not */ |
251 | 0, /* Assert behind */ |
252 | 0, /* Assert behind not */ |
253 | 0, /* NA assert */ |
254 | 0, /* NA assert behind */ |
255 | 0, /* ONCE */ |
256 | 0, /* SCRIPT_RUN */ |
257 | 0, 0, 0, 0, 0, /* BRA, BRAPOS, CBRA, CBRAPOS, COND */ |
258 | 0, 0, 0, 0, 0, /* SBRA, SBRAPOS, SCBRA, SCBRAPOS, SCOND */ |
259 | 0, 0, /* CREF, DNCREF */ |
260 | 0, 0, /* RREF, DNRREF */ |
261 | 0, 0, /* FALSE, TRUE */ |
262 | 0, 0, 0, /* BRAZERO, BRAMINZERO, BRAPOSZERO */ |
263 | 0, 0, 0, /* MARK, PRUNE, PRUNE_ARG */ |
264 | 0, 0, 0, 0, /* SKIP, SKIP_ARG, THEN, THEN_ARG */ |
265 | 0, 0, /* COMMIT, COMMIT_ARG */ |
266 | 0, 0, 0, /* FAIL, ACCEPT, ASSERT_ACCEPT */ |
267 | 0, 0, 0 /* CLOSE, SKIPZERO, DEFINE */ |
268 | }; |
269 | |
270 | /* These 2 tables allow for compact code for testing for \D, \d, \S, \s, \W, |
271 | and \w */ |
272 | |
273 | static const uint8_t toptable1[] = { |
274 | 0, 0, 0, 0, 0, 0, |
275 | ctype_digit, ctype_digit, |
276 | ctype_space, ctype_space, |
277 | ctype_word, ctype_word, |
278 | 0, 0 /* OP_ANY, OP_ALLANY */ |
279 | }; |
280 | |
281 | static const uint8_t toptable2[] = { |
282 | 0, 0, 0, 0, 0, 0, |
283 | ctype_digit, 0, |
284 | ctype_space, 0, |
285 | ctype_word, 0, |
286 | 1, 1 /* OP_ANY, OP_ALLANY */ |
287 | }; |
288 | |
289 | |
290 | /* Structure for holding data about a particular state, which is in effect the |
291 | current data for an active path through the match tree. It must consist |
292 | entirely of ints because the working vector we are passed, and which we put |
293 | these structures in, is a vector of ints. */ |
294 | |
295 | typedef struct stateblock { |
296 | int offset; /* Offset to opcode (-ve has meaning) */ |
297 | int count; /* Count for repeats */ |
298 | int data; /* Some use extra data */ |
299 | } stateblock; |
300 | |
301 | #define INTS_PER_STATEBLOCK (int)(sizeof(stateblock)/sizeof(int)) |
302 | |
303 | |
304 | /* Before version 10.32 the recursive calls of internal_dfa_match() were passed |
305 | local working space and output vectors that were created on the stack. This has |
306 | caused issues for some patterns, especially in small-stack environments such as |
307 | Windows. A new scheme is now in use which sets up a vector on the stack, but if |
308 | this is too small, heap memory is used, up to the heap_limit. The main |
309 | parameters are all numbers of ints because the workspace is a vector of ints. |
310 | |
311 | The size of the starting stack vector, DFA_START_RWS_SIZE, is in bytes, and is |
312 | defined in pcre2_internal.h so as to be available to pcre2test when it is |
313 | finding the minimum heap requirement for a match. */ |
314 | |
315 | #define OVEC_UNIT (sizeof(PCRE2_SIZE)/sizeof(int)) |
316 | |
317 | #define RWS_BASE_SIZE (DFA_START_RWS_SIZE/sizeof(int)) /* Stack vector */ |
318 | #define RWS_RSIZE 1000 /* Work size for recursion */ |
319 | #define RWS_OVEC_RSIZE (1000*OVEC_UNIT) /* Ovector for recursion */ |
320 | #define RWS_OVEC_OSIZE (2*OVEC_UNIT) /* Ovector in other cases */ |
321 | |
322 | /* This structure is at the start of each workspace block. */ |
323 | |
324 | typedef struct RWS_anchor { |
325 | struct RWS_anchor *next; |
326 | uint32_t size; /* Number of ints */ |
327 | uint32_t free; /* Number of ints */ |
328 | } RWS_anchor; |
329 | |
330 | #define RWS_ANCHOR_SIZE (sizeof(RWS_anchor)/sizeof(int)) |
331 | |
332 | |
333 | |
334 | /************************************************* |
335 | * Process a callout * |
336 | *************************************************/ |
337 | |
338 | /* This function is called to perform a callout. |
339 | |
340 | Arguments: |
341 | code current code pointer |
342 | offsets points to current capture offsets |
343 | current_subject start of current subject match |
344 | ptr current position in subject |
345 | mb the match block |
346 | extracode extra code offset when called from condition |
347 | lengthptr where to return the callout length |
348 | |
349 | Returns: the return from the callout |
350 | */ |
351 | |
352 | static int |
353 | do_callout_dfa(PCRE2_SPTR code, PCRE2_SIZE *offsets, PCRE2_SPTR current_subject, |
354 | PCRE2_SPTR ptr, dfa_match_block *mb, PCRE2_SIZE , |
355 | PCRE2_SIZE *lengthptr) |
356 | { |
357 | pcre2_callout_block *cb = mb->cb; |
358 | |
359 | *lengthptr = (code[extracode] == OP_CALLOUT)? |
360 | (PCRE2_SIZE)PRIV(OP_lengths)[OP_CALLOUT] : |
361 | (PCRE2_SIZE)GET(code, 1 + 2*LINK_SIZE + extracode); |
362 | |
363 | if (mb->callout == NULL) return 0; /* No callout provided */ |
364 | |
365 | /* Fixed fields in the callout block are set once and for all at the start of |
366 | matching. */ |
367 | |
368 | cb->offset_vector = offsets; |
369 | cb->start_match = (PCRE2_SIZE)(current_subject - mb->start_subject); |
370 | cb->current_position = (PCRE2_SIZE)(ptr - mb->start_subject); |
371 | cb->pattern_position = GET(code, 1 + extracode); |
372 | cb->next_item_length = GET(code, 1 + LINK_SIZE + extracode); |
373 | |
374 | if (code[extracode] == OP_CALLOUT) |
375 | { |
376 | cb->callout_number = code[1 + 2*LINK_SIZE + extracode]; |
377 | cb->callout_string_offset = 0; |
378 | cb->callout_string = NULL; |
379 | cb->callout_string_length = 0; |
380 | } |
381 | else |
382 | { |
383 | cb->callout_number = 0; |
384 | cb->callout_string_offset = GET(code, 1 + 3*LINK_SIZE + extracode); |
385 | cb->callout_string = code + (1 + 4*LINK_SIZE + extracode) + 1; |
386 | cb->callout_string_length = *lengthptr - (1 + 4*LINK_SIZE) - 2; |
387 | } |
388 | |
389 | return (mb->callout)(cb, mb->callout_data); |
390 | } |
391 | |
392 | |
393 | |
394 | /************************************************* |
395 | * Expand local workspace memory * |
396 | *************************************************/ |
397 | |
398 | /* This function is called when internal_dfa_match() is about to be called |
399 | recursively and there is insufficient working space left in the current |
400 | workspace block. If there's an existing next block, use it; otherwise get a new |
401 | block unless the heap limit is reached. |
402 | |
403 | Arguments: |
404 | rwsptr pointer to block pointer (updated) |
405 | ovecsize space needed for an ovector |
406 | mb the match block |
407 | |
408 | Returns: 0 rwsptr has been updated |
409 | !0 an error code |
410 | */ |
411 | |
412 | static int |
413 | more_workspace(RWS_anchor **rwsptr, unsigned int ovecsize, dfa_match_block *mb) |
414 | { |
415 | RWS_anchor *rws = *rwsptr; |
416 | RWS_anchor *new; |
417 | |
418 | if (rws->next != NULL) |
419 | { |
420 | new = rws->next; |
421 | } |
422 | |
423 | /* Sizes in the RWS_anchor blocks are in units of sizeof(int), but |
424 | mb->heap_limit and mb->heap_used are in kibibytes. Play carefully, to avoid |
425 | overflow. */ |
426 | |
427 | else |
428 | { |
429 | uint32_t newsize = (rws->size >= UINT32_MAX/2)? UINT32_MAX/2 : rws->size * 2; |
430 | uint32_t newsizeK = newsize/(1024/sizeof(int)); |
431 | |
432 | if (newsizeK + mb->heap_used > mb->heap_limit) |
433 | newsizeK = (uint32_t)(mb->heap_limit - mb->heap_used); |
434 | newsize = newsizeK*(1024/sizeof(int)); |
435 | |
436 | if (newsize < RWS_RSIZE + ovecsize + RWS_ANCHOR_SIZE) |
437 | return PCRE2_ERROR_HEAPLIMIT; |
438 | new = mb->memctl.malloc(newsize*sizeof(int), mb->memctl.memory_data); |
439 | if (new == NULL) return PCRE2_ERROR_NOMEMORY; |
440 | mb->heap_used += newsizeK; |
441 | new->next = NULL; |
442 | new->size = newsize; |
443 | rws->next = new; |
444 | } |
445 | |
446 | new->free = new->size - RWS_ANCHOR_SIZE; |
447 | *rwsptr = new; |
448 | return 0; |
449 | } |
450 | |
451 | |
452 | |
453 | /************************************************* |
454 | * Match a Regular Expression - DFA engine * |
455 | *************************************************/ |
456 | |
457 | /* This internal function applies a compiled pattern to a subject string, |
458 | starting at a given point, using a DFA engine. This function is called from the |
459 | external one, possibly multiple times if the pattern is not anchored. The |
460 | function calls itself recursively for some kinds of subpattern. |
461 | |
462 | Arguments: |
463 | mb the match_data block with fixed information |
464 | this_start_code the opening bracket of this subexpression's code |
465 | current_subject where we currently are in the subject string |
466 | start_offset start offset in the subject string |
467 | offsets vector to contain the matching string offsets |
468 | offsetcount size of same |
469 | workspace vector of workspace |
470 | wscount size of same |
471 | rlevel function call recursion level |
472 | |
473 | Returns: > 0 => number of match offset pairs placed in offsets |
474 | = 0 => offsets overflowed; longest matches are present |
475 | -1 => failed to match |
476 | < -1 => some kind of unexpected problem |
477 | |
478 | The following macros are used for adding states to the two state vectors (one |
479 | for the current character, one for the following character). */ |
480 | |
481 | #define ADD_ACTIVE(x,y) \ |
482 | if (active_count++ < wscount) \ |
483 | { \ |
484 | next_active_state->offset = (x); \ |
485 | next_active_state->count = (y); \ |
486 | next_active_state++; \ |
487 | } \ |
488 | else return PCRE2_ERROR_DFA_WSSIZE |
489 | |
490 | #define ADD_ACTIVE_DATA(x,y,z) \ |
491 | if (active_count++ < wscount) \ |
492 | { \ |
493 | next_active_state->offset = (x); \ |
494 | next_active_state->count = (y); \ |
495 | next_active_state->data = (z); \ |
496 | next_active_state++; \ |
497 | } \ |
498 | else return PCRE2_ERROR_DFA_WSSIZE |
499 | |
500 | #define ADD_NEW(x,y) \ |
501 | if (new_count++ < wscount) \ |
502 | { \ |
503 | next_new_state->offset = (x); \ |
504 | next_new_state->count = (y); \ |
505 | next_new_state++; \ |
506 | } \ |
507 | else return PCRE2_ERROR_DFA_WSSIZE |
508 | |
509 | #define ADD_NEW_DATA(x,y,z) \ |
510 | if (new_count++ < wscount) \ |
511 | { \ |
512 | next_new_state->offset = (x); \ |
513 | next_new_state->count = (y); \ |
514 | next_new_state->data = (z); \ |
515 | next_new_state++; \ |
516 | } \ |
517 | else return PCRE2_ERROR_DFA_WSSIZE |
518 | |
519 | /* And now, here is the code */ |
520 | |
521 | static int |
522 | internal_dfa_match( |
523 | dfa_match_block *mb, |
524 | PCRE2_SPTR this_start_code, |
525 | PCRE2_SPTR current_subject, |
526 | PCRE2_SIZE start_offset, |
527 | PCRE2_SIZE *offsets, |
528 | uint32_t offsetcount, |
529 | int *workspace, |
530 | int wscount, |
531 | uint32_t rlevel, |
532 | int *RWS) |
533 | { |
534 | stateblock *active_states, *new_states, *temp_states; |
535 | stateblock *next_active_state, *next_new_state; |
536 | const uint8_t *ctypes, *lcc, *fcc; |
537 | PCRE2_SPTR ptr; |
538 | PCRE2_SPTR end_code; |
539 | dfa_recursion_info new_recursive; |
540 | int active_count, new_count, match_count; |
541 | |
542 | /* Some fields in the mb block are frequently referenced, so we load them into |
543 | independent variables in the hope that this will perform better. */ |
544 | |
545 | PCRE2_SPTR start_subject = mb->start_subject; |
546 | PCRE2_SPTR end_subject = mb->end_subject; |
547 | PCRE2_SPTR start_code = mb->start_code; |
548 | |
549 | #ifdef SUPPORT_UNICODE |
550 | BOOL utf = (mb->poptions & PCRE2_UTF) != 0; |
551 | BOOL utf_or_ucp = utf || (mb->poptions & PCRE2_UCP) != 0; |
552 | #else |
553 | BOOL utf = FALSE; |
554 | #endif |
555 | |
556 | BOOL reset_could_continue = FALSE; |
557 | |
558 | if (mb->match_call_count++ >= mb->match_limit) return PCRE2_ERROR_MATCHLIMIT; |
559 | if (rlevel++ > mb->match_limit_depth) return PCRE2_ERROR_DEPTHLIMIT; |
560 | offsetcount &= (uint32_t)(-2); /* Round down */ |
561 | |
562 | wscount -= 2; |
563 | wscount = (wscount - (wscount % (INTS_PER_STATEBLOCK * 2))) / |
564 | (2 * INTS_PER_STATEBLOCK); |
565 | |
566 | ctypes = mb->tables + ctypes_offset; |
567 | lcc = mb->tables + lcc_offset; |
568 | fcc = mb->tables + fcc_offset; |
569 | |
570 | match_count = PCRE2_ERROR_NOMATCH; /* A negative number */ |
571 | |
572 | active_states = (stateblock *)(workspace + 2); |
573 | next_new_state = new_states = active_states + wscount; |
574 | new_count = 0; |
575 | |
576 | /* The first thing in any (sub) pattern is a bracket of some sort. Push all |
577 | the alternative states onto the list, and find out where the end is. This |
578 | makes is possible to use this function recursively, when we want to stop at a |
579 | matching internal ket rather than at the end. |
580 | |
581 | If we are dealing with a backward assertion we have to find out the maximum |
582 | amount to move back, and set up each alternative appropriately. */ |
583 | |
584 | if (*this_start_code == OP_ASSERTBACK || *this_start_code == OP_ASSERTBACK_NOT) |
585 | { |
586 | size_t max_back = 0; |
587 | size_t gone_back; |
588 | |
589 | end_code = this_start_code; |
590 | do |
591 | { |
592 | size_t back = (size_t)GET(end_code, 2+LINK_SIZE); |
593 | if (back > max_back) max_back = back; |
594 | end_code += GET(end_code, 1); |
595 | } |
596 | while (*end_code == OP_ALT); |
597 | |
598 | /* If we can't go back the amount required for the longest lookbehind |
599 | pattern, go back as far as we can; some alternatives may still be viable. */ |
600 | |
601 | #ifdef SUPPORT_UNICODE |
602 | /* In character mode we have to step back character by character */ |
603 | |
604 | if (utf) |
605 | { |
606 | for (gone_back = 0; gone_back < max_back; gone_back++) |
607 | { |
608 | if (current_subject <= start_subject) break; |
609 | current_subject--; |
610 | ACROSSCHAR(current_subject > start_subject, current_subject, |
611 | current_subject--); |
612 | } |
613 | } |
614 | else |
615 | #endif |
616 | |
617 | /* In byte-mode we can do this quickly. */ |
618 | |
619 | { |
620 | size_t current_offset = (size_t)(current_subject - start_subject); |
621 | gone_back = (current_offset < max_back)? current_offset : max_back; |
622 | current_subject -= gone_back; |
623 | } |
624 | |
625 | /* Save the earliest consulted character */ |
626 | |
627 | if (current_subject < mb->start_used_ptr) |
628 | mb->start_used_ptr = current_subject; |
629 | |
630 | /* Now we can process the individual branches. There will be an OP_REVERSE at |
631 | the start of each branch, except when the length of the branch is zero. */ |
632 | |
633 | end_code = this_start_code; |
634 | do |
635 | { |
636 | uint32_t revlen = (end_code[1+LINK_SIZE] == OP_REVERSE)? 1 + LINK_SIZE : 0; |
637 | size_t back = (revlen == 0)? 0 : (size_t)GET(end_code, 2+LINK_SIZE); |
638 | if (back <= gone_back) |
639 | { |
640 | int bstate = (int)(end_code - start_code + 1 + LINK_SIZE + revlen); |
641 | ADD_NEW_DATA(-bstate, 0, (int)(gone_back - back)); |
642 | } |
643 | end_code += GET(end_code, 1); |
644 | } |
645 | while (*end_code == OP_ALT); |
646 | } |
647 | |
648 | /* This is the code for a "normal" subpattern (not a backward assertion). The |
649 | start of a whole pattern is always one of these. If we are at the top level, |
650 | we may be asked to restart matching from the same point that we reached for a |
651 | previous partial match. We still have to scan through the top-level branches to |
652 | find the end state. */ |
653 | |
654 | else |
655 | { |
656 | end_code = this_start_code; |
657 | |
658 | /* Restarting */ |
659 | |
660 | if (rlevel == 1 && (mb->moptions & PCRE2_DFA_RESTART) != 0) |
661 | { |
662 | do { end_code += GET(end_code, 1); } while (*end_code == OP_ALT); |
663 | new_count = workspace[1]; |
664 | if (!workspace[0]) |
665 | memcpy(new_states, active_states, (size_t)new_count * sizeof(stateblock)); |
666 | } |
667 | |
668 | /* Not restarting */ |
669 | |
670 | else |
671 | { |
672 | int length = 1 + LINK_SIZE + |
673 | ((*this_start_code == OP_CBRA || *this_start_code == OP_SCBRA || |
674 | *this_start_code == OP_CBRAPOS || *this_start_code == OP_SCBRAPOS) |
675 | ? IMM2_SIZE:0); |
676 | do |
677 | { |
678 | ADD_NEW((int)(end_code - start_code + length), 0); |
679 | end_code += GET(end_code, 1); |
680 | length = 1 + LINK_SIZE; |
681 | } |
682 | while (*end_code == OP_ALT); |
683 | } |
684 | } |
685 | |
686 | workspace[0] = 0; /* Bit indicating which vector is current */ |
687 | |
688 | /* Loop for scanning the subject */ |
689 | |
690 | ptr = current_subject; |
691 | for (;;) |
692 | { |
693 | int i, j; |
694 | int clen, dlen; |
695 | uint32_t c, d; |
696 | int forced_fail = 0; |
697 | BOOL partial_newline = FALSE; |
698 | BOOL could_continue = reset_could_continue; |
699 | reset_could_continue = FALSE; |
700 | |
701 | if (ptr > mb->last_used_ptr) mb->last_used_ptr = ptr; |
702 | |
703 | /* Make the new state list into the active state list and empty the |
704 | new state list. */ |
705 | |
706 | temp_states = active_states; |
707 | active_states = new_states; |
708 | new_states = temp_states; |
709 | active_count = new_count; |
710 | new_count = 0; |
711 | |
712 | workspace[0] ^= 1; /* Remember for the restarting feature */ |
713 | workspace[1] = active_count; |
714 | |
715 | /* Set the pointers for adding new states */ |
716 | |
717 | next_active_state = active_states + active_count; |
718 | next_new_state = new_states; |
719 | |
720 | /* Load the current character from the subject outside the loop, as many |
721 | different states may want to look at it, and we assume that at least one |
722 | will. */ |
723 | |
724 | if (ptr < end_subject) |
725 | { |
726 | clen = 1; /* Number of data items in the character */ |
727 | #ifdef SUPPORT_UNICODE |
728 | GETCHARLENTEST(c, ptr, clen); |
729 | #else |
730 | c = *ptr; |
731 | #endif /* SUPPORT_UNICODE */ |
732 | } |
733 | else |
734 | { |
735 | clen = 0; /* This indicates the end of the subject */ |
736 | c = NOTACHAR; /* This value should never actually be used */ |
737 | } |
738 | |
739 | /* Scan up the active states and act on each one. The result of an action |
740 | may be to add more states to the currently active list (e.g. on hitting a |
741 | parenthesis) or it may be to put states on the new list, for considering |
742 | when we move the character pointer on. */ |
743 | |
744 | for (i = 0; i < active_count; i++) |
745 | { |
746 | stateblock *current_state = active_states + i; |
747 | BOOL caseless = FALSE; |
748 | PCRE2_SPTR code; |
749 | uint32_t codevalue; |
750 | int state_offset = current_state->offset; |
751 | int rrc; |
752 | int count; |
753 | |
754 | /* A negative offset is a special case meaning "hold off going to this |
755 | (negated) state until the number of characters in the data field have |
756 | been skipped". If the could_continue flag was passed over from a previous |
757 | state, arrange for it to passed on. */ |
758 | |
759 | if (state_offset < 0) |
760 | { |
761 | if (current_state->data > 0) |
762 | { |
763 | ADD_NEW_DATA(state_offset, current_state->count, |
764 | current_state->data - 1); |
765 | if (could_continue) reset_could_continue = TRUE; |
766 | continue; |
767 | } |
768 | else |
769 | { |
770 | current_state->offset = state_offset = -state_offset; |
771 | } |
772 | } |
773 | |
774 | /* Check for a duplicate state with the same count, and skip if found. |
775 | See the note at the head of this module about the possibility of improving |
776 | performance here. */ |
777 | |
778 | for (j = 0; j < i; j++) |
779 | { |
780 | if (active_states[j].offset == state_offset && |
781 | active_states[j].count == current_state->count) |
782 | goto NEXT_ACTIVE_STATE; |
783 | } |
784 | |
785 | /* The state offset is the offset to the opcode */ |
786 | |
787 | code = start_code + state_offset; |
788 | codevalue = *code; |
789 | |
790 | /* If this opcode inspects a character, but we are at the end of the |
791 | subject, remember the fact for use when testing for a partial match. */ |
792 | |
793 | if (clen == 0 && poptable[codevalue] != 0) |
794 | could_continue = TRUE; |
795 | |
796 | /* If this opcode is followed by an inline character, load it. It is |
797 | tempting to test for the presence of a subject character here, but that |
798 | is wrong, because sometimes zero repetitions of the subject are |
799 | permitted. |
800 | |
801 | We also use this mechanism for opcodes such as OP_TYPEPLUS that take an |
802 | argument that is not a data character - but is always one byte long because |
803 | the values are small. We have to take special action to deal with \P, \p, |
804 | \H, \h, \V, \v and \X in this case. To keep the other cases fast, convert |
805 | these ones to new opcodes. */ |
806 | |
807 | if (coptable[codevalue] > 0) |
808 | { |
809 | dlen = 1; |
810 | #ifdef SUPPORT_UNICODE |
811 | if (utf) { GETCHARLEN(d, (code + coptable[codevalue]), dlen); } else |
812 | #endif /* SUPPORT_UNICODE */ |
813 | d = code[coptable[codevalue]]; |
814 | if (codevalue >= OP_TYPESTAR) |
815 | { |
816 | switch(d) |
817 | { |
818 | case OP_ANYBYTE: return PCRE2_ERROR_DFA_UITEM; |
819 | case OP_NOTPROP: |
820 | case OP_PROP: codevalue += OP_PROP_EXTRA; break; |
821 | case OP_ANYNL: codevalue += OP_ANYNL_EXTRA; break; |
822 | case OP_EXTUNI: codevalue += OP_EXTUNI_EXTRA; break; |
823 | case OP_NOT_HSPACE: |
824 | case OP_HSPACE: codevalue += OP_HSPACE_EXTRA; break; |
825 | case OP_NOT_VSPACE: |
826 | case OP_VSPACE: codevalue += OP_VSPACE_EXTRA; break; |
827 | default: break; |
828 | } |
829 | } |
830 | } |
831 | else |
832 | { |
833 | dlen = 0; /* Not strictly necessary, but compilers moan */ |
834 | d = NOTACHAR; /* if these variables are not set. */ |
835 | } |
836 | |
837 | |
838 | /* Now process the individual opcodes */ |
839 | |
840 | switch (codevalue) |
841 | { |
842 | /* ========================================================================== */ |
843 | /* These cases are never obeyed. This is a fudge that causes a compile- |
844 | time error if the vectors coptable or poptable, which are indexed by |
845 | opcode, are not the correct length. It seems to be the only way to do |
846 | such a check at compile time, as the sizeof() operator does not work |
847 | in the C preprocessor. */ |
848 | |
849 | case OP_TABLE_LENGTH: |
850 | case OP_TABLE_LENGTH + |
851 | ((sizeof(coptable) == OP_TABLE_LENGTH) && |
852 | (sizeof(poptable) == OP_TABLE_LENGTH)): |
853 | return 0; |
854 | |
855 | /* ========================================================================== */ |
856 | /* Reached a closing bracket. If not at the end of the pattern, carry |
857 | on with the next opcode. For repeating opcodes, also add the repeat |
858 | state. Note that KETRPOS will always be encountered at the end of the |
859 | subpattern, because the possessive subpattern repeats are always handled |
860 | using recursive calls. Thus, it never adds any new states. |
861 | |
862 | At the end of the (sub)pattern, unless we have an empty string and |
863 | PCRE2_NOTEMPTY is set, or PCRE2_NOTEMPTY_ATSTART is set and we are at the |
864 | start of the subject, save the match data, shifting up all previous |
865 | matches so we always have the longest first. */ |
866 | |
867 | case OP_KET: |
868 | case OP_KETRMIN: |
869 | case OP_KETRMAX: |
870 | case OP_KETRPOS: |
871 | if (code != end_code) |
872 | { |
873 | ADD_ACTIVE(state_offset + 1 + LINK_SIZE, 0); |
874 | if (codevalue != OP_KET) |
875 | { |
876 | ADD_ACTIVE(state_offset - (int)GET(code, 1), 0); |
877 | } |
878 | } |
879 | else |
880 | { |
881 | if (ptr > current_subject || |
882 | ((mb->moptions & PCRE2_NOTEMPTY) == 0 && |
883 | ((mb->moptions & PCRE2_NOTEMPTY_ATSTART) == 0 || |
884 | current_subject > start_subject + mb->start_offset))) |
885 | { |
886 | if (match_count < 0) match_count = (offsetcount >= 2)? 1 : 0; |
887 | else if (match_count > 0 && ++match_count * 2 > (int)offsetcount) |
888 | match_count = 0; |
889 | count = ((match_count == 0)? (int)offsetcount : match_count * 2) - 2; |
890 | if (count > 0) (void)memmove(offsets + 2, offsets, |
891 | (size_t)count * sizeof(PCRE2_SIZE)); |
892 | if (offsetcount >= 2) |
893 | { |
894 | offsets[0] = (PCRE2_SIZE)(current_subject - start_subject); |
895 | offsets[1] = (PCRE2_SIZE)(ptr - start_subject); |
896 | } |
897 | if ((mb->moptions & PCRE2_DFA_SHORTEST) != 0) return match_count; |
898 | } |
899 | } |
900 | break; |
901 | |
902 | /* ========================================================================== */ |
903 | /* These opcodes add to the current list of states without looking |
904 | at the current character. */ |
905 | |
906 | /*-----------------------------------------------------------------*/ |
907 | case OP_ALT: |
908 | do { code += GET(code, 1); } while (*code == OP_ALT); |
909 | ADD_ACTIVE((int)(code - start_code), 0); |
910 | break; |
911 | |
912 | /*-----------------------------------------------------------------*/ |
913 | case OP_BRA: |
914 | case OP_SBRA: |
915 | do |
916 | { |
917 | ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0); |
918 | code += GET(code, 1); |
919 | } |
920 | while (*code == OP_ALT); |
921 | break; |
922 | |
923 | /*-----------------------------------------------------------------*/ |
924 | case OP_CBRA: |
925 | case OP_SCBRA: |
926 | ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE + IMM2_SIZE), 0); |
927 | code += GET(code, 1); |
928 | while (*code == OP_ALT) |
929 | { |
930 | ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0); |
931 | code += GET(code, 1); |
932 | } |
933 | break; |
934 | |
935 | /*-----------------------------------------------------------------*/ |
936 | case OP_BRAZERO: |
937 | case OP_BRAMINZERO: |
938 | ADD_ACTIVE(state_offset + 1, 0); |
939 | code += 1 + GET(code, 2); |
940 | while (*code == OP_ALT) code += GET(code, 1); |
941 | ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0); |
942 | break; |
943 | |
944 | /*-----------------------------------------------------------------*/ |
945 | case OP_SKIPZERO: |
946 | code += 1 + GET(code, 2); |
947 | while (*code == OP_ALT) code += GET(code, 1); |
948 | ADD_ACTIVE((int)(code - start_code + 1 + LINK_SIZE), 0); |
949 | break; |
950 | |
951 | /*-----------------------------------------------------------------*/ |
952 | case OP_CIRC: |
953 | if (ptr == start_subject && (mb->moptions & PCRE2_NOTBOL) == 0) |
954 | { ADD_ACTIVE(state_offset + 1, 0); } |
955 | break; |
956 | |
957 | /*-----------------------------------------------------------------*/ |
958 | case OP_CIRCM: |
959 | if ((ptr == start_subject && (mb->moptions & PCRE2_NOTBOL) == 0) || |
960 | ((ptr != end_subject || (mb->poptions & PCRE2_ALT_CIRCUMFLEX) != 0 ) |
961 | && WAS_NEWLINE(ptr))) |
962 | { ADD_ACTIVE(state_offset + 1, 0); } |
963 | break; |
964 | |
965 | /*-----------------------------------------------------------------*/ |
966 | case OP_EOD: |
967 | if (ptr >= end_subject) |
968 | { |
969 | if ((mb->moptions & PCRE2_PARTIAL_HARD) != 0) |
970 | return PCRE2_ERROR_PARTIAL; |
971 | else { ADD_ACTIVE(state_offset + 1, 0); } |
972 | } |
973 | break; |
974 | |
975 | /*-----------------------------------------------------------------*/ |
976 | case OP_SOD: |
977 | if (ptr == start_subject) { ADD_ACTIVE(state_offset + 1, 0); } |
978 | break; |
979 | |
980 | /*-----------------------------------------------------------------*/ |
981 | case OP_SOM: |
982 | if (ptr == start_subject + start_offset) { ADD_ACTIVE(state_offset + 1, 0); } |
983 | break; |
984 | |
985 | |
986 | /* ========================================================================== */ |
987 | /* These opcodes inspect the next subject character, and sometimes |
988 | the previous one as well, but do not have an argument. The variable |
989 | clen contains the length of the current character and is zero if we are |
990 | at the end of the subject. */ |
991 | |
992 | /*-----------------------------------------------------------------*/ |
993 | case OP_ANY: |
994 | if (clen > 0 && !IS_NEWLINE(ptr)) |
995 | { |
996 | if (ptr + 1 >= mb->end_subject && |
997 | (mb->moptions & (PCRE2_PARTIAL_HARD)) != 0 && |
998 | NLBLOCK->nltype == NLTYPE_FIXED && |
999 | NLBLOCK->nllen == 2 && |
1000 | c == NLBLOCK->nl[0]) |
1001 | { |
1002 | could_continue = partial_newline = TRUE; |
1003 | } |
1004 | else |
1005 | { |
1006 | ADD_NEW(state_offset + 1, 0); |
1007 | } |
1008 | } |
1009 | break; |
1010 | |
1011 | /*-----------------------------------------------------------------*/ |
1012 | case OP_ALLANY: |
1013 | if (clen > 0) |
1014 | { ADD_NEW(state_offset + 1, 0); } |
1015 | break; |
1016 | |
1017 | /*-----------------------------------------------------------------*/ |
1018 | case OP_EODN: |
1019 | if (clen == 0 || (IS_NEWLINE(ptr) && ptr == end_subject - mb->nllen)) |
1020 | { |
1021 | if ((mb->moptions & PCRE2_PARTIAL_HARD) != 0) |
1022 | return PCRE2_ERROR_PARTIAL; |
1023 | ADD_ACTIVE(state_offset + 1, 0); |
1024 | } |
1025 | break; |
1026 | |
1027 | /*-----------------------------------------------------------------*/ |
1028 | case OP_DOLL: |
1029 | if ((mb->moptions & PCRE2_NOTEOL) == 0) |
1030 | { |
1031 | if (clen == 0 && (mb->moptions & PCRE2_PARTIAL_HARD) != 0) |
1032 | could_continue = TRUE; |
1033 | else if (clen == 0 || |
1034 | ((mb->poptions & PCRE2_DOLLAR_ENDONLY) == 0 && IS_NEWLINE(ptr) && |
1035 | (ptr == end_subject - mb->nllen) |
1036 | )) |
1037 | { ADD_ACTIVE(state_offset + 1, 0); } |
1038 | else if (ptr + 1 >= mb->end_subject && |
1039 | (mb->moptions & (PCRE2_PARTIAL_HARD|PCRE2_PARTIAL_SOFT)) != 0 && |
1040 | NLBLOCK->nltype == NLTYPE_FIXED && |
1041 | NLBLOCK->nllen == 2 && |
1042 | c == NLBLOCK->nl[0]) |
1043 | { |
1044 | if ((mb->moptions & PCRE2_PARTIAL_HARD) != 0) |
1045 | { |
1046 | reset_could_continue = TRUE; |
1047 | ADD_NEW_DATA(-(state_offset + 1), 0, 1); |
1048 | } |
1049 | else could_continue = partial_newline = TRUE; |
1050 | } |
1051 | } |
1052 | break; |
1053 | |
1054 | /*-----------------------------------------------------------------*/ |
1055 | case OP_DOLLM: |
1056 | if ((mb->moptions & PCRE2_NOTEOL) == 0) |
1057 | { |
1058 | if (clen == 0 && (mb->moptions & PCRE2_PARTIAL_HARD) != 0) |
1059 | could_continue = TRUE; |
1060 | else if (clen == 0 || |
1061 | ((mb->poptions & PCRE2_DOLLAR_ENDONLY) == 0 && IS_NEWLINE(ptr))) |
1062 | { ADD_ACTIVE(state_offset + 1, 0); } |
1063 | else if (ptr + 1 >= mb->end_subject && |
1064 | (mb->moptions & (PCRE2_PARTIAL_HARD|PCRE2_PARTIAL_SOFT)) != 0 && |
1065 | NLBLOCK->nltype == NLTYPE_FIXED && |
1066 | NLBLOCK->nllen == 2 && |
1067 | c == NLBLOCK->nl[0]) |
1068 | { |
1069 | if ((mb->moptions & PCRE2_PARTIAL_HARD) != 0) |
1070 | { |
1071 | reset_could_continue = TRUE; |
1072 | ADD_NEW_DATA(-(state_offset + 1), 0, 1); |
1073 | } |
1074 | else could_continue = partial_newline = TRUE; |
1075 | } |
1076 | } |
1077 | else if (IS_NEWLINE(ptr)) |
1078 | { ADD_ACTIVE(state_offset + 1, 0); } |
1079 | break; |
1080 | |
1081 | /*-----------------------------------------------------------------*/ |
1082 | |
1083 | case OP_DIGIT: |
1084 | case OP_WHITESPACE: |
1085 | case OP_WORDCHAR: |
1086 | if (clen > 0 && c < 256 && |
1087 | ((ctypes[c] & toptable1[codevalue]) ^ toptable2[codevalue]) != 0) |
1088 | { ADD_NEW(state_offset + 1, 0); } |
1089 | break; |
1090 | |
1091 | /*-----------------------------------------------------------------*/ |
1092 | case OP_NOT_DIGIT: |
1093 | case OP_NOT_WHITESPACE: |
1094 | case OP_NOT_WORDCHAR: |
1095 | if (clen > 0 && (c >= 256 || |
1096 | ((ctypes[c] & toptable1[codevalue]) ^ toptable2[codevalue]) != 0)) |
1097 | { ADD_NEW(state_offset + 1, 0); } |
1098 | break; |
1099 | |
1100 | /*-----------------------------------------------------------------*/ |
1101 | case OP_WORD_BOUNDARY: |
1102 | case OP_NOT_WORD_BOUNDARY: |
1103 | { |
1104 | int left_word, right_word; |
1105 | |
1106 | if (ptr > start_subject) |
1107 | { |
1108 | PCRE2_SPTR temp = ptr - 1; |
1109 | if (temp < mb->start_used_ptr) mb->start_used_ptr = temp; |
1110 | #if defined SUPPORT_UNICODE && PCRE2_CODE_UNIT_WIDTH != 32 |
1111 | if (utf) { BACKCHAR(temp); } |
1112 | #endif |
1113 | GETCHARTEST(d, temp); |
1114 | #ifdef SUPPORT_UNICODE |
1115 | if ((mb->poptions & PCRE2_UCP) != 0) |
1116 | { |
1117 | if (d == '_') left_word = TRUE; else |
1118 | { |
1119 | uint32_t cat = UCD_CATEGORY(d); |
1120 | left_word = (cat == ucp_L || cat == ucp_N); |
1121 | } |
1122 | } |
1123 | else |
1124 | #endif |
1125 | left_word = d < 256 && (ctypes[d] & ctype_word) != 0; |
1126 | } |
1127 | else left_word = FALSE; |
1128 | |
1129 | if (clen > 0) |
1130 | { |
1131 | if (ptr >= mb->last_used_ptr) |
1132 | { |
1133 | PCRE2_SPTR temp = ptr + 1; |
1134 | #if defined SUPPORT_UNICODE && PCRE2_CODE_UNIT_WIDTH != 32 |
1135 | if (utf) { FORWARDCHARTEST(temp, mb->end_subject); } |
1136 | #endif |
1137 | mb->last_used_ptr = temp; |
1138 | } |
1139 | #ifdef SUPPORT_UNICODE |
1140 | if ((mb->poptions & PCRE2_UCP) != 0) |
1141 | { |
1142 | if (c == '_') right_word = TRUE; else |
1143 | { |
1144 | uint32_t cat = UCD_CATEGORY(c); |
1145 | right_word = (cat == ucp_L || cat == ucp_N); |
1146 | } |
1147 | } |
1148 | else |
1149 | #endif |
1150 | right_word = c < 256 && (ctypes[c] & ctype_word) != 0; |
1151 | } |
1152 | else right_word = FALSE; |
1153 | |
1154 | if ((left_word == right_word) == (codevalue == OP_NOT_WORD_BOUNDARY)) |
1155 | { ADD_ACTIVE(state_offset + 1, 0); } |
1156 | } |
1157 | break; |
1158 | |
1159 | |
1160 | /*-----------------------------------------------------------------*/ |
1161 | /* Check the next character by Unicode property. We will get here only |
1162 | if the support is in the binary; otherwise a compile-time error occurs. |
1163 | */ |
1164 | |
1165 | #ifdef SUPPORT_UNICODE |
1166 | case OP_PROP: |
1167 | case OP_NOTPROP: |
1168 | if (clen > 0) |
1169 | { |
1170 | BOOL OK; |
1171 | const uint32_t *cp; |
1172 | const ucd_record * prop = GET_UCD(c); |
1173 | switch(code[1]) |
1174 | { |
1175 | case PT_ANY: |
1176 | OK = TRUE; |
1177 | break; |
1178 | |
1179 | case PT_LAMP: |
1180 | OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || |
1181 | prop->chartype == ucp_Lt; |
1182 | break; |
1183 | |
1184 | case PT_GC: |
1185 | OK = PRIV(ucp_gentype)[prop->chartype] == code[2]; |
1186 | break; |
1187 | |
1188 | case PT_PC: |
1189 | OK = prop->chartype == code[2]; |
1190 | break; |
1191 | |
1192 | case PT_SC: |
1193 | OK = prop->script == code[2]; |
1194 | break; |
1195 | |
1196 | case PT_SCX: |
1197 | OK = (prop->script == code[2] || |
1198 | MAPBIT(PRIV(ucd_script_sets) + UCD_SCRIPTX_PROP(prop), code[2]) != 0); |
1199 | break; |
1200 | |
1201 | /* These are specials for combination cases. */ |
1202 | |
1203 | case PT_ALNUM: |
1204 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1205 | PRIV(ucp_gentype)[prop->chartype] == ucp_N; |
1206 | break; |
1207 | |
1208 | /* Perl space used to exclude VT, but from Perl 5.18 it is included, |
1209 | which means that Perl space and POSIX space are now identical. PCRE |
1210 | was changed at release 8.34. */ |
1211 | |
1212 | case PT_SPACE: /* Perl space */ |
1213 | case PT_PXSPACE: /* POSIX space */ |
1214 | switch(c) |
1215 | { |
1216 | HSPACE_CASES: |
1217 | VSPACE_CASES: |
1218 | OK = TRUE; |
1219 | break; |
1220 | |
1221 | default: |
1222 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_Z; |
1223 | break; |
1224 | } |
1225 | break; |
1226 | |
1227 | case PT_WORD: |
1228 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1229 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
1230 | c == CHAR_UNDERSCORE; |
1231 | break; |
1232 | |
1233 | case PT_CLIST: |
1234 | cp = PRIV(ucd_caseless_sets) + code[2]; |
1235 | for (;;) |
1236 | { |
1237 | if (c < *cp) { OK = FALSE; break; } |
1238 | if (c == *cp++) { OK = TRUE; break; } |
1239 | } |
1240 | break; |
1241 | |
1242 | case PT_UCNC: |
1243 | OK = c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
1244 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
1245 | c >= 0xe000; |
1246 | break; |
1247 | |
1248 | case PT_BIDICL: |
1249 | OK = UCD_BIDICLASS(c) == code[2]; |
1250 | break; |
1251 | |
1252 | case PT_BOOL: |
1253 | OK = MAPBIT(PRIV(ucd_boolprop_sets) + |
1254 | UCD_BPROPS_PROP(prop), code[2]) != 0; |
1255 | break; |
1256 | |
1257 | /* Should never occur, but keep compilers from grumbling. */ |
1258 | |
1259 | default: |
1260 | OK = codevalue != OP_PROP; |
1261 | break; |
1262 | } |
1263 | |
1264 | if (OK == (codevalue == OP_PROP)) { ADD_NEW(state_offset + 3, 0); } |
1265 | } |
1266 | break; |
1267 | #endif |
1268 | |
1269 | |
1270 | |
1271 | /* ========================================================================== */ |
1272 | /* These opcodes likewise inspect the subject character, but have an |
1273 | argument that is not a data character. It is one of these opcodes: |
1274 | OP_ANY, OP_ALLANY, OP_DIGIT, OP_NOT_DIGIT, OP_WHITESPACE, OP_NOT_SPACE, |
1275 | OP_WORDCHAR, OP_NOT_WORDCHAR. The value is loaded into d. */ |
1276 | |
1277 | case OP_TYPEPLUS: |
1278 | case OP_TYPEMINPLUS: |
1279 | case OP_TYPEPOSPLUS: |
1280 | count = current_state->count; /* Already matched */ |
1281 | if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); } |
1282 | if (clen > 0) |
1283 | { |
1284 | if (d == OP_ANY && ptr + 1 >= mb->end_subject && |
1285 | (mb->moptions & (PCRE2_PARTIAL_HARD)) != 0 && |
1286 | NLBLOCK->nltype == NLTYPE_FIXED && |
1287 | NLBLOCK->nllen == 2 && |
1288 | c == NLBLOCK->nl[0]) |
1289 | { |
1290 | could_continue = partial_newline = TRUE; |
1291 | } |
1292 | else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) || |
1293 | (c < 256 && |
1294 | (d != OP_ANY || !IS_NEWLINE(ptr)) && |
1295 | ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0)) |
1296 | { |
1297 | if (count > 0 && codevalue == OP_TYPEPOSPLUS) |
1298 | { |
1299 | active_count--; /* Remove non-match possibility */ |
1300 | next_active_state--; |
1301 | } |
1302 | count++; |
1303 | ADD_NEW(state_offset, count); |
1304 | } |
1305 | } |
1306 | break; |
1307 | |
1308 | /*-----------------------------------------------------------------*/ |
1309 | case OP_TYPEQUERY: |
1310 | case OP_TYPEMINQUERY: |
1311 | case OP_TYPEPOSQUERY: |
1312 | ADD_ACTIVE(state_offset + 2, 0); |
1313 | if (clen > 0) |
1314 | { |
1315 | if (d == OP_ANY && ptr + 1 >= mb->end_subject && |
1316 | (mb->moptions & (PCRE2_PARTIAL_HARD)) != 0 && |
1317 | NLBLOCK->nltype == NLTYPE_FIXED && |
1318 | NLBLOCK->nllen == 2 && |
1319 | c == NLBLOCK->nl[0]) |
1320 | { |
1321 | could_continue = partial_newline = TRUE; |
1322 | } |
1323 | else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) || |
1324 | (c < 256 && |
1325 | (d != OP_ANY || !IS_NEWLINE(ptr)) && |
1326 | ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0)) |
1327 | { |
1328 | if (codevalue == OP_TYPEPOSQUERY) |
1329 | { |
1330 | active_count--; /* Remove non-match possibility */ |
1331 | next_active_state--; |
1332 | } |
1333 | ADD_NEW(state_offset + 2, 0); |
1334 | } |
1335 | } |
1336 | break; |
1337 | |
1338 | /*-----------------------------------------------------------------*/ |
1339 | case OP_TYPESTAR: |
1340 | case OP_TYPEMINSTAR: |
1341 | case OP_TYPEPOSSTAR: |
1342 | ADD_ACTIVE(state_offset + 2, 0); |
1343 | if (clen > 0) |
1344 | { |
1345 | if (d == OP_ANY && ptr + 1 >= mb->end_subject && |
1346 | (mb->moptions & (PCRE2_PARTIAL_HARD)) != 0 && |
1347 | NLBLOCK->nltype == NLTYPE_FIXED && |
1348 | NLBLOCK->nllen == 2 && |
1349 | c == NLBLOCK->nl[0]) |
1350 | { |
1351 | could_continue = partial_newline = TRUE; |
1352 | } |
1353 | else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) || |
1354 | (c < 256 && |
1355 | (d != OP_ANY || !IS_NEWLINE(ptr)) && |
1356 | ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0)) |
1357 | { |
1358 | if (codevalue == OP_TYPEPOSSTAR) |
1359 | { |
1360 | active_count--; /* Remove non-match possibility */ |
1361 | next_active_state--; |
1362 | } |
1363 | ADD_NEW(state_offset, 0); |
1364 | } |
1365 | } |
1366 | break; |
1367 | |
1368 | /*-----------------------------------------------------------------*/ |
1369 | case OP_TYPEEXACT: |
1370 | count = current_state->count; /* Number already matched */ |
1371 | if (clen > 0) |
1372 | { |
1373 | if (d == OP_ANY && ptr + 1 >= mb->end_subject && |
1374 | (mb->moptions & (PCRE2_PARTIAL_HARD)) != 0 && |
1375 | NLBLOCK->nltype == NLTYPE_FIXED && |
1376 | NLBLOCK->nllen == 2 && |
1377 | c == NLBLOCK->nl[0]) |
1378 | { |
1379 | could_continue = partial_newline = TRUE; |
1380 | } |
1381 | else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) || |
1382 | (c < 256 && |
1383 | (d != OP_ANY || !IS_NEWLINE(ptr)) && |
1384 | ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0)) |
1385 | { |
1386 | if (++count >= (int)GET2(code, 1)) |
1387 | { ADD_NEW(state_offset + 1 + IMM2_SIZE + 1, 0); } |
1388 | else |
1389 | { ADD_NEW(state_offset, count); } |
1390 | } |
1391 | } |
1392 | break; |
1393 | |
1394 | /*-----------------------------------------------------------------*/ |
1395 | case OP_TYPEUPTO: |
1396 | case OP_TYPEMINUPTO: |
1397 | case OP_TYPEPOSUPTO: |
1398 | ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); |
1399 | count = current_state->count; /* Number already matched */ |
1400 | if (clen > 0) |
1401 | { |
1402 | if (d == OP_ANY && ptr + 1 >= mb->end_subject && |
1403 | (mb->moptions & (PCRE2_PARTIAL_HARD)) != 0 && |
1404 | NLBLOCK->nltype == NLTYPE_FIXED && |
1405 | NLBLOCK->nllen == 2 && |
1406 | c == NLBLOCK->nl[0]) |
1407 | { |
1408 | could_continue = partial_newline = TRUE; |
1409 | } |
1410 | else if ((c >= 256 && d != OP_DIGIT && d != OP_WHITESPACE && d != OP_WORDCHAR) || |
1411 | (c < 256 && |
1412 | (d != OP_ANY || !IS_NEWLINE(ptr)) && |
1413 | ((ctypes[c] & toptable1[d]) ^ toptable2[d]) != 0)) |
1414 | { |
1415 | if (codevalue == OP_TYPEPOSUPTO) |
1416 | { |
1417 | active_count--; /* Remove non-match possibility */ |
1418 | next_active_state--; |
1419 | } |
1420 | if (++count >= (int)GET2(code, 1)) |
1421 | { ADD_NEW(state_offset + 2 + IMM2_SIZE, 0); } |
1422 | else |
1423 | { ADD_NEW(state_offset, count); } |
1424 | } |
1425 | } |
1426 | break; |
1427 | |
1428 | /* ========================================================================== */ |
1429 | /* These are virtual opcodes that are used when something like |
1430 | OP_TYPEPLUS has OP_PROP, OP_NOTPROP, OP_ANYNL, or OP_EXTUNI as its |
1431 | argument. It keeps the code above fast for the other cases. The argument |
1432 | is in the d variable. */ |
1433 | |
1434 | #ifdef SUPPORT_UNICODE |
1435 | case OP_PROP_EXTRA + OP_TYPEPLUS: |
1436 | case OP_PROP_EXTRA + OP_TYPEMINPLUS: |
1437 | case OP_PROP_EXTRA + OP_TYPEPOSPLUS: |
1438 | count = current_state->count; /* Already matched */ |
1439 | if (count > 0) { ADD_ACTIVE(state_offset + 4, 0); } |
1440 | if (clen > 0) |
1441 | { |
1442 | BOOL OK; |
1443 | const uint32_t *cp; |
1444 | const ucd_record * prop = GET_UCD(c); |
1445 | switch(code[2]) |
1446 | { |
1447 | case PT_ANY: |
1448 | OK = TRUE; |
1449 | break; |
1450 | |
1451 | case PT_LAMP: |
1452 | OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || |
1453 | prop->chartype == ucp_Lt; |
1454 | break; |
1455 | |
1456 | case PT_GC: |
1457 | OK = PRIV(ucp_gentype)[prop->chartype] == code[3]; |
1458 | break; |
1459 | |
1460 | case PT_PC: |
1461 | OK = prop->chartype == code[3]; |
1462 | break; |
1463 | |
1464 | case PT_SC: |
1465 | OK = prop->script == code[3]; |
1466 | break; |
1467 | |
1468 | case PT_SCX: |
1469 | OK = (prop->script == code[3] || |
1470 | MAPBIT(PRIV(ucd_script_sets) + UCD_SCRIPTX_PROP(prop), code[3]) != 0); |
1471 | break; |
1472 | |
1473 | /* These are specials for combination cases. */ |
1474 | |
1475 | case PT_ALNUM: |
1476 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1477 | PRIV(ucp_gentype)[prop->chartype] == ucp_N; |
1478 | break; |
1479 | |
1480 | /* Perl space used to exclude VT, but from Perl 5.18 it is included, |
1481 | which means that Perl space and POSIX space are now identical. PCRE |
1482 | was changed at release 8.34. */ |
1483 | |
1484 | case PT_SPACE: /* Perl space */ |
1485 | case PT_PXSPACE: /* POSIX space */ |
1486 | switch(c) |
1487 | { |
1488 | HSPACE_CASES: |
1489 | VSPACE_CASES: |
1490 | OK = TRUE; |
1491 | break; |
1492 | |
1493 | default: |
1494 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_Z; |
1495 | break; |
1496 | } |
1497 | break; |
1498 | |
1499 | case PT_WORD: |
1500 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1501 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
1502 | c == CHAR_UNDERSCORE; |
1503 | break; |
1504 | |
1505 | case PT_CLIST: |
1506 | cp = PRIV(ucd_caseless_sets) + code[3]; |
1507 | for (;;) |
1508 | { |
1509 | if (c < *cp) { OK = FALSE; break; } |
1510 | if (c == *cp++) { OK = TRUE; break; } |
1511 | } |
1512 | break; |
1513 | |
1514 | case PT_UCNC: |
1515 | OK = c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
1516 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
1517 | c >= 0xe000; |
1518 | break; |
1519 | |
1520 | case PT_BIDICL: |
1521 | OK = UCD_BIDICLASS(c) == code[3]; |
1522 | break; |
1523 | |
1524 | case PT_BOOL: |
1525 | OK = MAPBIT(PRIV(ucd_boolprop_sets) + |
1526 | UCD_BPROPS_PROP(prop), code[3]) != 0; |
1527 | break; |
1528 | |
1529 | /* Should never occur, but keep compilers from grumbling. */ |
1530 | |
1531 | default: |
1532 | OK = codevalue != OP_PROP; |
1533 | break; |
1534 | } |
1535 | |
1536 | if (OK == (d == OP_PROP)) |
1537 | { |
1538 | if (count > 0 && codevalue == OP_PROP_EXTRA + OP_TYPEPOSPLUS) |
1539 | { |
1540 | active_count--; /* Remove non-match possibility */ |
1541 | next_active_state--; |
1542 | } |
1543 | count++; |
1544 | ADD_NEW(state_offset, count); |
1545 | } |
1546 | } |
1547 | break; |
1548 | |
1549 | /*-----------------------------------------------------------------*/ |
1550 | case OP_EXTUNI_EXTRA + OP_TYPEPLUS: |
1551 | case OP_EXTUNI_EXTRA + OP_TYPEMINPLUS: |
1552 | case OP_EXTUNI_EXTRA + OP_TYPEPOSPLUS: |
1553 | count = current_state->count; /* Already matched */ |
1554 | if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); } |
1555 | if (clen > 0) |
1556 | { |
1557 | int ncount = 0; |
1558 | if (count > 0 && codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSPLUS) |
1559 | { |
1560 | active_count--; /* Remove non-match possibility */ |
1561 | next_active_state--; |
1562 | } |
1563 | (void)PRIV(extuni)(c, ptr + clen, mb->start_subject, end_subject, utf, |
1564 | &ncount); |
1565 | count++; |
1566 | ADD_NEW_DATA(-state_offset, count, ncount); |
1567 | } |
1568 | break; |
1569 | #endif |
1570 | |
1571 | /*-----------------------------------------------------------------*/ |
1572 | case OP_ANYNL_EXTRA + OP_TYPEPLUS: |
1573 | case OP_ANYNL_EXTRA + OP_TYPEMINPLUS: |
1574 | case OP_ANYNL_EXTRA + OP_TYPEPOSPLUS: |
1575 | count = current_state->count; /* Already matched */ |
1576 | if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); } |
1577 | if (clen > 0) |
1578 | { |
1579 | int ncount = 0; |
1580 | switch (c) |
1581 | { |
1582 | case CHAR_VT: |
1583 | case CHAR_FF: |
1584 | case CHAR_NEL: |
1585 | #ifndef EBCDIC |
1586 | case 0x2028: |
1587 | case 0x2029: |
1588 | #endif /* Not EBCDIC */ |
1589 | if (mb->bsr_convention == PCRE2_BSR_ANYCRLF) break; |
1590 | goto ANYNL01; |
1591 | |
1592 | case CHAR_CR: |
1593 | if (ptr + 1 < end_subject && UCHAR21TEST(ptr + 1) == CHAR_LF) ncount = 1; |
1594 | /* Fall through */ |
1595 | |
1596 | ANYNL01: |
1597 | case CHAR_LF: |
1598 | if (count > 0 && codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSPLUS) |
1599 | { |
1600 | active_count--; /* Remove non-match possibility */ |
1601 | next_active_state--; |
1602 | } |
1603 | count++; |
1604 | ADD_NEW_DATA(-state_offset, count, ncount); |
1605 | break; |
1606 | |
1607 | default: |
1608 | break; |
1609 | } |
1610 | } |
1611 | break; |
1612 | |
1613 | /*-----------------------------------------------------------------*/ |
1614 | case OP_VSPACE_EXTRA + OP_TYPEPLUS: |
1615 | case OP_VSPACE_EXTRA + OP_TYPEMINPLUS: |
1616 | case OP_VSPACE_EXTRA + OP_TYPEPOSPLUS: |
1617 | count = current_state->count; /* Already matched */ |
1618 | if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); } |
1619 | if (clen > 0) |
1620 | { |
1621 | BOOL OK; |
1622 | switch (c) |
1623 | { |
1624 | VSPACE_CASES: |
1625 | OK = TRUE; |
1626 | break; |
1627 | |
1628 | default: |
1629 | OK = FALSE; |
1630 | break; |
1631 | } |
1632 | |
1633 | if (OK == (d == OP_VSPACE)) |
1634 | { |
1635 | if (count > 0 && codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSPLUS) |
1636 | { |
1637 | active_count--; /* Remove non-match possibility */ |
1638 | next_active_state--; |
1639 | } |
1640 | count++; |
1641 | ADD_NEW_DATA(-state_offset, count, 0); |
1642 | } |
1643 | } |
1644 | break; |
1645 | |
1646 | /*-----------------------------------------------------------------*/ |
1647 | case OP_HSPACE_EXTRA + OP_TYPEPLUS: |
1648 | case OP_HSPACE_EXTRA + OP_TYPEMINPLUS: |
1649 | case OP_HSPACE_EXTRA + OP_TYPEPOSPLUS: |
1650 | count = current_state->count; /* Already matched */ |
1651 | if (count > 0) { ADD_ACTIVE(state_offset + 2, 0); } |
1652 | if (clen > 0) |
1653 | { |
1654 | BOOL OK; |
1655 | switch (c) |
1656 | { |
1657 | HSPACE_CASES: |
1658 | OK = TRUE; |
1659 | break; |
1660 | |
1661 | default: |
1662 | OK = FALSE; |
1663 | break; |
1664 | } |
1665 | |
1666 | if (OK == (d == OP_HSPACE)) |
1667 | { |
1668 | if (count > 0 && codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSPLUS) |
1669 | { |
1670 | active_count--; /* Remove non-match possibility */ |
1671 | next_active_state--; |
1672 | } |
1673 | count++; |
1674 | ADD_NEW_DATA(-state_offset, count, 0); |
1675 | } |
1676 | } |
1677 | break; |
1678 | |
1679 | /*-----------------------------------------------------------------*/ |
1680 | #ifdef SUPPORT_UNICODE |
1681 | case OP_PROP_EXTRA + OP_TYPEQUERY: |
1682 | case OP_PROP_EXTRA + OP_TYPEMINQUERY: |
1683 | case OP_PROP_EXTRA + OP_TYPEPOSQUERY: |
1684 | count = 4; |
1685 | goto QS1; |
1686 | |
1687 | case OP_PROP_EXTRA + OP_TYPESTAR: |
1688 | case OP_PROP_EXTRA + OP_TYPEMINSTAR: |
1689 | case OP_PROP_EXTRA + OP_TYPEPOSSTAR: |
1690 | count = 0; |
1691 | |
1692 | QS1: |
1693 | |
1694 | ADD_ACTIVE(state_offset + 4, 0); |
1695 | if (clen > 0) |
1696 | { |
1697 | BOOL OK; |
1698 | const uint32_t *cp; |
1699 | const ucd_record * prop = GET_UCD(c); |
1700 | switch(code[2]) |
1701 | { |
1702 | case PT_ANY: |
1703 | OK = TRUE; |
1704 | break; |
1705 | |
1706 | case PT_LAMP: |
1707 | OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || |
1708 | prop->chartype == ucp_Lt; |
1709 | break; |
1710 | |
1711 | case PT_GC: |
1712 | OK = PRIV(ucp_gentype)[prop->chartype] == code[3]; |
1713 | break; |
1714 | |
1715 | case PT_PC: |
1716 | OK = prop->chartype == code[3]; |
1717 | break; |
1718 | |
1719 | case PT_SC: |
1720 | OK = prop->script == code[3]; |
1721 | break; |
1722 | |
1723 | case PT_SCX: |
1724 | OK = (prop->script == code[3] || |
1725 | MAPBIT(PRIV(ucd_script_sets) + UCD_SCRIPTX_PROP(prop), code[3]) != 0); |
1726 | break; |
1727 | |
1728 | /* These are specials for combination cases. */ |
1729 | |
1730 | case PT_ALNUM: |
1731 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1732 | PRIV(ucp_gentype)[prop->chartype] == ucp_N; |
1733 | break; |
1734 | |
1735 | /* Perl space used to exclude VT, but from Perl 5.18 it is included, |
1736 | which means that Perl space and POSIX space are now identical. PCRE |
1737 | was changed at release 8.34. */ |
1738 | |
1739 | case PT_SPACE: /* Perl space */ |
1740 | case PT_PXSPACE: /* POSIX space */ |
1741 | switch(c) |
1742 | { |
1743 | HSPACE_CASES: |
1744 | VSPACE_CASES: |
1745 | OK = TRUE; |
1746 | break; |
1747 | |
1748 | default: |
1749 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_Z; |
1750 | break; |
1751 | } |
1752 | break; |
1753 | |
1754 | case PT_WORD: |
1755 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
1756 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
1757 | c == CHAR_UNDERSCORE; |
1758 | break; |
1759 | |
1760 | case PT_CLIST: |
1761 | cp = PRIV(ucd_caseless_sets) + code[3]; |
1762 | for (;;) |
1763 | { |
1764 | if (c < *cp) { OK = FALSE; break; } |
1765 | if (c == *cp++) { OK = TRUE; break; } |
1766 | } |
1767 | break; |
1768 | |
1769 | case PT_UCNC: |
1770 | OK = c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
1771 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
1772 | c >= 0xe000; |
1773 | break; |
1774 | |
1775 | case PT_BIDICL: |
1776 | OK = UCD_BIDICLASS(c) == code[3]; |
1777 | break; |
1778 | |
1779 | case PT_BOOL: |
1780 | OK = MAPBIT(PRIV(ucd_boolprop_sets) + |
1781 | UCD_BPROPS_PROP(prop), code[3]) != 0; |
1782 | break; |
1783 | |
1784 | /* Should never occur, but keep compilers from grumbling. */ |
1785 | |
1786 | default: |
1787 | OK = codevalue != OP_PROP; |
1788 | break; |
1789 | } |
1790 | |
1791 | if (OK == (d == OP_PROP)) |
1792 | { |
1793 | if (codevalue == OP_PROP_EXTRA + OP_TYPEPOSSTAR || |
1794 | codevalue == OP_PROP_EXTRA + OP_TYPEPOSQUERY) |
1795 | { |
1796 | active_count--; /* Remove non-match possibility */ |
1797 | next_active_state--; |
1798 | } |
1799 | ADD_NEW(state_offset + count, 0); |
1800 | } |
1801 | } |
1802 | break; |
1803 | |
1804 | /*-----------------------------------------------------------------*/ |
1805 | case OP_EXTUNI_EXTRA + OP_TYPEQUERY: |
1806 | case OP_EXTUNI_EXTRA + OP_TYPEMINQUERY: |
1807 | case OP_EXTUNI_EXTRA + OP_TYPEPOSQUERY: |
1808 | count = 2; |
1809 | goto QS2; |
1810 | |
1811 | case OP_EXTUNI_EXTRA + OP_TYPESTAR: |
1812 | case OP_EXTUNI_EXTRA + OP_TYPEMINSTAR: |
1813 | case OP_EXTUNI_EXTRA + OP_TYPEPOSSTAR: |
1814 | count = 0; |
1815 | |
1816 | QS2: |
1817 | |
1818 | ADD_ACTIVE(state_offset + 2, 0); |
1819 | if (clen > 0) |
1820 | { |
1821 | int ncount = 0; |
1822 | if (codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSSTAR || |
1823 | codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSQUERY) |
1824 | { |
1825 | active_count--; /* Remove non-match possibility */ |
1826 | next_active_state--; |
1827 | } |
1828 | (void)PRIV(extuni)(c, ptr + clen, mb->start_subject, end_subject, utf, |
1829 | &ncount); |
1830 | ADD_NEW_DATA(-(state_offset + count), 0, ncount); |
1831 | } |
1832 | break; |
1833 | #endif |
1834 | |
1835 | /*-----------------------------------------------------------------*/ |
1836 | case OP_ANYNL_EXTRA + OP_TYPEQUERY: |
1837 | case OP_ANYNL_EXTRA + OP_TYPEMINQUERY: |
1838 | case OP_ANYNL_EXTRA + OP_TYPEPOSQUERY: |
1839 | count = 2; |
1840 | goto QS3; |
1841 | |
1842 | case OP_ANYNL_EXTRA + OP_TYPESTAR: |
1843 | case OP_ANYNL_EXTRA + OP_TYPEMINSTAR: |
1844 | case OP_ANYNL_EXTRA + OP_TYPEPOSSTAR: |
1845 | count = 0; |
1846 | |
1847 | QS3: |
1848 | ADD_ACTIVE(state_offset + 2, 0); |
1849 | if (clen > 0) |
1850 | { |
1851 | int ncount = 0; |
1852 | switch (c) |
1853 | { |
1854 | case CHAR_VT: |
1855 | case CHAR_FF: |
1856 | case CHAR_NEL: |
1857 | #ifndef EBCDIC |
1858 | case 0x2028: |
1859 | case 0x2029: |
1860 | #endif /* Not EBCDIC */ |
1861 | if (mb->bsr_convention == PCRE2_BSR_ANYCRLF) break; |
1862 | goto ANYNL02; |
1863 | |
1864 | case CHAR_CR: |
1865 | if (ptr + 1 < end_subject && UCHAR21TEST(ptr + 1) == CHAR_LF) ncount = 1; |
1866 | /* Fall through */ |
1867 | |
1868 | ANYNL02: |
1869 | case CHAR_LF: |
1870 | if (codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSSTAR || |
1871 | codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSQUERY) |
1872 | { |
1873 | active_count--; /* Remove non-match possibility */ |
1874 | next_active_state--; |
1875 | } |
1876 | ADD_NEW_DATA(-(state_offset + (int)count), 0, ncount); |
1877 | break; |
1878 | |
1879 | default: |
1880 | break; |
1881 | } |
1882 | } |
1883 | break; |
1884 | |
1885 | /*-----------------------------------------------------------------*/ |
1886 | case OP_VSPACE_EXTRA + OP_TYPEQUERY: |
1887 | case OP_VSPACE_EXTRA + OP_TYPEMINQUERY: |
1888 | case OP_VSPACE_EXTRA + OP_TYPEPOSQUERY: |
1889 | count = 2; |
1890 | goto QS4; |
1891 | |
1892 | case OP_VSPACE_EXTRA + OP_TYPESTAR: |
1893 | case OP_VSPACE_EXTRA + OP_TYPEMINSTAR: |
1894 | case OP_VSPACE_EXTRA + OP_TYPEPOSSTAR: |
1895 | count = 0; |
1896 | |
1897 | QS4: |
1898 | ADD_ACTIVE(state_offset + 2, 0); |
1899 | if (clen > 0) |
1900 | { |
1901 | BOOL OK; |
1902 | switch (c) |
1903 | { |
1904 | VSPACE_CASES: |
1905 | OK = TRUE; |
1906 | break; |
1907 | |
1908 | default: |
1909 | OK = FALSE; |
1910 | break; |
1911 | } |
1912 | if (OK == (d == OP_VSPACE)) |
1913 | { |
1914 | if (codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSSTAR || |
1915 | codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSQUERY) |
1916 | { |
1917 | active_count--; /* Remove non-match possibility */ |
1918 | next_active_state--; |
1919 | } |
1920 | ADD_NEW_DATA(-(state_offset + (int)count), 0, 0); |
1921 | } |
1922 | } |
1923 | break; |
1924 | |
1925 | /*-----------------------------------------------------------------*/ |
1926 | case OP_HSPACE_EXTRA + OP_TYPEQUERY: |
1927 | case OP_HSPACE_EXTRA + OP_TYPEMINQUERY: |
1928 | case OP_HSPACE_EXTRA + OP_TYPEPOSQUERY: |
1929 | count = 2; |
1930 | goto QS5; |
1931 | |
1932 | case OP_HSPACE_EXTRA + OP_TYPESTAR: |
1933 | case OP_HSPACE_EXTRA + OP_TYPEMINSTAR: |
1934 | case OP_HSPACE_EXTRA + OP_TYPEPOSSTAR: |
1935 | count = 0; |
1936 | |
1937 | QS5: |
1938 | ADD_ACTIVE(state_offset + 2, 0); |
1939 | if (clen > 0) |
1940 | { |
1941 | BOOL OK; |
1942 | switch (c) |
1943 | { |
1944 | HSPACE_CASES: |
1945 | OK = TRUE; |
1946 | break; |
1947 | |
1948 | default: |
1949 | OK = FALSE; |
1950 | break; |
1951 | } |
1952 | |
1953 | if (OK == (d == OP_HSPACE)) |
1954 | { |
1955 | if (codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSSTAR || |
1956 | codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSQUERY) |
1957 | { |
1958 | active_count--; /* Remove non-match possibility */ |
1959 | next_active_state--; |
1960 | } |
1961 | ADD_NEW_DATA(-(state_offset + (int)count), 0, 0); |
1962 | } |
1963 | } |
1964 | break; |
1965 | |
1966 | /*-----------------------------------------------------------------*/ |
1967 | #ifdef SUPPORT_UNICODE |
1968 | case OP_PROP_EXTRA + OP_TYPEEXACT: |
1969 | case OP_PROP_EXTRA + OP_TYPEUPTO: |
1970 | case OP_PROP_EXTRA + OP_TYPEMINUPTO: |
1971 | case OP_PROP_EXTRA + OP_TYPEPOSUPTO: |
1972 | if (codevalue != OP_PROP_EXTRA + OP_TYPEEXACT) |
1973 | { ADD_ACTIVE(state_offset + 1 + IMM2_SIZE + 3, 0); } |
1974 | count = current_state->count; /* Number already matched */ |
1975 | if (clen > 0) |
1976 | { |
1977 | BOOL OK; |
1978 | const uint32_t *cp; |
1979 | const ucd_record * prop = GET_UCD(c); |
1980 | switch(code[1 + IMM2_SIZE + 1]) |
1981 | { |
1982 | case PT_ANY: |
1983 | OK = TRUE; |
1984 | break; |
1985 | |
1986 | case PT_LAMP: |
1987 | OK = prop->chartype == ucp_Lu || prop->chartype == ucp_Ll || |
1988 | prop->chartype == ucp_Lt; |
1989 | break; |
1990 | |
1991 | case PT_GC: |
1992 | OK = PRIV(ucp_gentype)[prop->chartype] == code[1 + IMM2_SIZE + 2]; |
1993 | break; |
1994 | |
1995 | case PT_PC: |
1996 | OK = prop->chartype == code[1 + IMM2_SIZE + 2]; |
1997 | break; |
1998 | |
1999 | case PT_SC: |
2000 | OK = prop->script == code[1 + IMM2_SIZE + 2]; |
2001 | break; |
2002 | |
2003 | case PT_SCX: |
2004 | OK = (prop->script == code[1 + IMM2_SIZE + 2] || |
2005 | MAPBIT(PRIV(ucd_script_sets) + UCD_SCRIPTX_PROP(prop), |
2006 | code[1 + IMM2_SIZE + 2]) != 0); |
2007 | break; |
2008 | |
2009 | /* These are specials for combination cases. */ |
2010 | |
2011 | case PT_ALNUM: |
2012 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
2013 | PRIV(ucp_gentype)[prop->chartype] == ucp_N; |
2014 | break; |
2015 | |
2016 | /* Perl space used to exclude VT, but from Perl 5.18 it is included, |
2017 | which means that Perl space and POSIX space are now identical. PCRE |
2018 | was changed at release 8.34. */ |
2019 | |
2020 | case PT_SPACE: /* Perl space */ |
2021 | case PT_PXSPACE: /* POSIX space */ |
2022 | switch(c) |
2023 | { |
2024 | HSPACE_CASES: |
2025 | VSPACE_CASES: |
2026 | OK = TRUE; |
2027 | break; |
2028 | |
2029 | default: |
2030 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_Z; |
2031 | break; |
2032 | } |
2033 | break; |
2034 | |
2035 | case PT_WORD: |
2036 | OK = PRIV(ucp_gentype)[prop->chartype] == ucp_L || |
2037 | PRIV(ucp_gentype)[prop->chartype] == ucp_N || |
2038 | c == CHAR_UNDERSCORE; |
2039 | break; |
2040 | |
2041 | case PT_CLIST: |
2042 | cp = PRIV(ucd_caseless_sets) + code[1 + IMM2_SIZE + 2]; |
2043 | for (;;) |
2044 | { |
2045 | if (c < *cp) { OK = FALSE; break; } |
2046 | if (c == *cp++) { OK = TRUE; break; } |
2047 | } |
2048 | break; |
2049 | |
2050 | case PT_UCNC: |
2051 | OK = c == CHAR_DOLLAR_SIGN || c == CHAR_COMMERCIAL_AT || |
2052 | c == CHAR_GRAVE_ACCENT || (c >= 0xa0 && c <= 0xd7ff) || |
2053 | c >= 0xe000; |
2054 | break; |
2055 | |
2056 | case PT_BIDICL: |
2057 | OK = UCD_BIDICLASS(c) == code[1 + IMM2_SIZE + 2]; |
2058 | break; |
2059 | |
2060 | case PT_BOOL: |
2061 | OK = MAPBIT(PRIV(ucd_boolprop_sets) + |
2062 | UCD_BPROPS_PROP(prop), code[1 + IMM2_SIZE + 2]) != 0; |
2063 | break; |
2064 | |
2065 | /* Should never occur, but keep compilers from grumbling. */ |
2066 | |
2067 | default: |
2068 | OK = codevalue != OP_PROP; |
2069 | break; |
2070 | } |
2071 | |
2072 | if (OK == (d == OP_PROP)) |
2073 | { |
2074 | if (codevalue == OP_PROP_EXTRA + OP_TYPEPOSUPTO) |
2075 | { |
2076 | active_count--; /* Remove non-match possibility */ |
2077 | next_active_state--; |
2078 | } |
2079 | if (++count >= (int)GET2(code, 1)) |
2080 | { ADD_NEW(state_offset + 1 + IMM2_SIZE + 3, 0); } |
2081 | else |
2082 | { ADD_NEW(state_offset, count); } |
2083 | } |
2084 | } |
2085 | break; |
2086 | |
2087 | /*-----------------------------------------------------------------*/ |
2088 | case OP_EXTUNI_EXTRA + OP_TYPEEXACT: |
2089 | case OP_EXTUNI_EXTRA + OP_TYPEUPTO: |
2090 | case OP_EXTUNI_EXTRA + OP_TYPEMINUPTO: |
2091 | case OP_EXTUNI_EXTRA + OP_TYPEPOSUPTO: |
2092 | if (codevalue != OP_EXTUNI_EXTRA + OP_TYPEEXACT) |
2093 | { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); } |
2094 | count = current_state->count; /* Number already matched */ |
2095 | if (clen > 0) |
2096 | { |
2097 | PCRE2_SPTR nptr; |
2098 | int ncount = 0; |
2099 | if (codevalue == OP_EXTUNI_EXTRA + OP_TYPEPOSUPTO) |
2100 | { |
2101 | active_count--; /* Remove non-match possibility */ |
2102 | next_active_state--; |
2103 | } |
2104 | nptr = PRIV(extuni)(c, ptr + clen, mb->start_subject, end_subject, utf, |
2105 | &ncount); |
2106 | if (nptr >= end_subject && (mb->moptions & PCRE2_PARTIAL_HARD) != 0) |
2107 | reset_could_continue = TRUE; |
2108 | if (++count >= (int)GET2(code, 1)) |
2109 | { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, ncount); } |
2110 | else |
2111 | { ADD_NEW_DATA(-state_offset, count, ncount); } |
2112 | } |
2113 | break; |
2114 | #endif |
2115 | |
2116 | /*-----------------------------------------------------------------*/ |
2117 | case OP_ANYNL_EXTRA + OP_TYPEEXACT: |
2118 | case OP_ANYNL_EXTRA + OP_TYPEUPTO: |
2119 | case OP_ANYNL_EXTRA + OP_TYPEMINUPTO: |
2120 | case OP_ANYNL_EXTRA + OP_TYPEPOSUPTO: |
2121 | if (codevalue != OP_ANYNL_EXTRA + OP_TYPEEXACT) |
2122 | { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); } |
2123 | count = current_state->count; /* Number already matched */ |
2124 | if (clen > 0) |
2125 | { |
2126 | int ncount = 0; |
2127 | switch (c) |
2128 | { |
2129 | case CHAR_VT: |
2130 | case CHAR_FF: |
2131 | case CHAR_NEL: |
2132 | #ifndef EBCDIC |
2133 | case 0x2028: |
2134 | case 0x2029: |
2135 | #endif /* Not EBCDIC */ |
2136 | if (mb->bsr_convention == PCRE2_BSR_ANYCRLF) break; |
2137 | goto ANYNL03; |
2138 | |
2139 | case CHAR_CR: |
2140 | if (ptr + 1 < end_subject && UCHAR21TEST(ptr + 1) == CHAR_LF) ncount = 1; |
2141 | /* Fall through */ |
2142 | |
2143 | ANYNL03: |
2144 | case CHAR_LF: |
2145 | if (codevalue == OP_ANYNL_EXTRA + OP_TYPEPOSUPTO) |
2146 | { |
2147 | active_count--; /* Remove non-match possibility */ |
2148 | next_active_state--; |
2149 | } |
2150 | if (++count >= (int)GET2(code, 1)) |
2151 | { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, ncount); } |
2152 | else |
2153 | { ADD_NEW_DATA(-state_offset, count, ncount); } |
2154 | break; |
2155 | |
2156 | default: |
2157 | break; |
2158 | } |
2159 | } |
2160 | break; |
2161 | |
2162 | /*-----------------------------------------------------------------*/ |
2163 | case OP_VSPACE_EXTRA + OP_TYPEEXACT: |
2164 | case OP_VSPACE_EXTRA + OP_TYPEUPTO: |
2165 | case OP_VSPACE_EXTRA + OP_TYPEMINUPTO: |
2166 | case OP_VSPACE_EXTRA + OP_TYPEPOSUPTO: |
2167 | if (codevalue != OP_VSPACE_EXTRA + OP_TYPEEXACT) |
2168 | { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); } |
2169 | count = current_state->count; /* Number already matched */ |
2170 | if (clen > 0) |
2171 | { |
2172 | BOOL OK; |
2173 | switch (c) |
2174 | { |
2175 | VSPACE_CASES: |
2176 | OK = TRUE; |
2177 | break; |
2178 | |
2179 | default: |
2180 | OK = FALSE; |
2181 | } |
2182 | |
2183 | if (OK == (d == OP_VSPACE)) |
2184 | { |
2185 | if (codevalue == OP_VSPACE_EXTRA + OP_TYPEPOSUPTO) |
2186 | { |
2187 | active_count--; /* Remove non-match possibility */ |
2188 | next_active_state--; |
2189 | } |
2190 | if (++count >= (int)GET2(code, 1)) |
2191 | { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, 0); } |
2192 | else |
2193 | { ADD_NEW_DATA(-state_offset, count, 0); } |
2194 | } |
2195 | } |
2196 | break; |
2197 | |
2198 | /*-----------------------------------------------------------------*/ |
2199 | case OP_HSPACE_EXTRA + OP_TYPEEXACT: |
2200 | case OP_HSPACE_EXTRA + OP_TYPEUPTO: |
2201 | case OP_HSPACE_EXTRA + OP_TYPEMINUPTO: |
2202 | case OP_HSPACE_EXTRA + OP_TYPEPOSUPTO: |
2203 | if (codevalue != OP_HSPACE_EXTRA + OP_TYPEEXACT) |
2204 | { ADD_ACTIVE(state_offset + 2 + IMM2_SIZE, 0); } |
2205 | count = current_state->count; /* Number already matched */ |
2206 | if (clen > 0) |
2207 | { |
2208 | BOOL OK; |
2209 | switch (c) |
2210 | { |
2211 | HSPACE_CASES: |
2212 | OK = TRUE; |
2213 | break; |
2214 | |
2215 | default: |
2216 | OK = FALSE; |
2217 | break; |
2218 | } |
2219 | |
2220 | if (OK == (d == OP_HSPACE)) |
2221 | { |
2222 | if (codevalue == OP_HSPACE_EXTRA + OP_TYPEPOSUPTO) |
2223 | { |
2224 | active_count--; /* Remove non-match possibility */ |
2225 | next_active_state--; |
2226 | } |
2227 | if (++count >= (int)GET2(code, 1)) |
2228 | { ADD_NEW_DATA(-(state_offset + 2 + IMM2_SIZE), 0, 0); } |
2229 | else |
2230 | { ADD_NEW_DATA(-state_offset, count, 0); } |
2231 | } |
2232 | } |
2233 | break; |
2234 | |
2235 | /* ========================================================================== */ |
2236 | /* These opcodes are followed by a character that is usually compared |
2237 | to the current subject character; it is loaded into d. We still get |
2238 | here even if there is no subject character, because in some cases zero |
2239 | repetitions are permitted. */ |
2240 | |
2241 | /*-----------------------------------------------------------------*/ |
2242 | case OP_CHAR: |
2243 | if (clen > 0 && c == d) { ADD_NEW(state_offset + dlen + 1, 0); } |
2244 | break; |
2245 | |
2246 | /*-----------------------------------------------------------------*/ |
2247 | case OP_CHARI: |
2248 | if (clen == 0) break; |
2249 | |
2250 | #ifdef SUPPORT_UNICODE |
2251 | if (utf_or_ucp) |
2252 | { |
2253 | if (c == d) { ADD_NEW(state_offset + dlen + 1, 0); } else |
2254 | { |
2255 | unsigned int othercase; |
2256 | if (c < 128) |
2257 | othercase = fcc[c]; |
2258 | else |
2259 | othercase = UCD_OTHERCASE(c); |
2260 | if (d == othercase) { ADD_NEW(state_offset + dlen + 1, 0); } |
2261 | } |
2262 | } |
2263 | else |
2264 | #endif /* SUPPORT_UNICODE */ |
2265 | /* Not UTF or UCP mode */ |
2266 | { |
2267 | if (TABLE_GET(c, lcc, c) == TABLE_GET(d, lcc, d)) |
2268 | { ADD_NEW(state_offset + 2, 0); } |
2269 | } |
2270 | break; |
2271 | |
2272 | |
2273 | #ifdef SUPPORT_UNICODE |
2274 | /*-----------------------------------------------------------------*/ |
2275 | /* This is a tricky one because it can match more than one character. |
2276 | Find out how many characters to skip, and then set up a negative state |
2277 | to wait for them to pass before continuing. */ |
2278 | |
2279 | case OP_EXTUNI: |
2280 | if (clen > 0) |
2281 | { |
2282 | int ncount = 0; |
2283 | PCRE2_SPTR nptr = PRIV(extuni)(c, ptr + clen, mb->start_subject, |
2284 | end_subject, utf, &ncount); |
2285 | if (nptr >= end_subject && (mb->moptions & PCRE2_PARTIAL_HARD) != 0) |
2286 | reset_could_continue = TRUE; |
2287 | ADD_NEW_DATA(-(state_offset + 1), 0, ncount); |
2288 | } |
2289 | break; |
2290 | #endif |
2291 | |
2292 | /*-----------------------------------------------------------------*/ |
2293 | /* This is a tricky like EXTUNI because it too can match more than one |
2294 | character (when CR is followed by LF). In this case, set up a negative |
2295 | state to wait for one character to pass before continuing. */ |
2296 | |
2297 | case OP_ANYNL: |
2298 | if (clen > 0) switch(c) |
2299 | { |
2300 | case CHAR_VT: |
2301 | case CHAR_FF: |
2302 | case CHAR_NEL: |
2303 | #ifndef EBCDIC |
2304 | case 0x2028: |
2305 | case 0x2029: |
2306 | #endif /* Not EBCDIC */ |
2307 | if (mb->bsr_convention == PCRE2_BSR_ANYCRLF) break; |
2308 | /* Fall through */ |
2309 | |
2310 | case CHAR_LF: |
2311 | ADD_NEW(state_offset + 1, 0); |
2312 | break; |
2313 | |
2314 | case CHAR_CR: |
2315 | if (ptr + 1 >= end_subject) |
2316 | { |
2317 | ADD_NEW(state_offset + 1, 0); |
2318 | if ((mb->moptions & PCRE2_PARTIAL_HARD) != 0) |
2319 | reset_could_continue = TRUE; |
2320 | } |
2321 | else if (UCHAR21TEST(ptr + 1) == CHAR_LF) |
2322 | { |
2323 | ADD_NEW_DATA(-(state_offset + 1), 0, 1); |
2324 | } |
2325 | else |
2326 | { |
2327 | ADD_NEW(state_offset + 1, 0); |
2328 | } |
2329 | break; |
2330 | } |
2331 | break; |
2332 | |
2333 | /*-----------------------------------------------------------------*/ |
2334 | case OP_NOT_VSPACE: |
2335 | if (clen > 0) switch(c) |
2336 | { |
2337 | VSPACE_CASES: |
2338 | break; |
2339 | |
2340 | default: |
2341 | ADD_NEW(state_offset + 1, 0); |
2342 | break; |
2343 | } |
2344 | break; |
2345 | |
2346 | /*-----------------------------------------------------------------*/ |
2347 | case OP_VSPACE: |
2348 | if (clen > 0) switch(c) |
2349 | { |
2350 | VSPACE_CASES: |
2351 | ADD_NEW(state_offset + 1, 0); |
2352 | break; |
2353 | |
2354 | default: |
2355 | break; |
2356 | } |
2357 | break; |
2358 | |
2359 | /*-----------------------------------------------------------------*/ |
2360 | case OP_NOT_HSPACE: |
2361 | if (clen > 0) switch(c) |
2362 | { |
2363 | HSPACE_CASES: |
2364 | break; |
2365 | |
2366 | default: |
2367 | ADD_NEW(state_offset + 1, 0); |
2368 | break; |
2369 | } |
2370 | break; |
2371 | |
2372 | /*-----------------------------------------------------------------*/ |
2373 | case OP_HSPACE: |
2374 | if (clen > 0) switch(c) |
2375 | { |
2376 | HSPACE_CASES: |
2377 | ADD_NEW(state_offset + 1, 0); |
2378 | break; |
2379 | |
2380 | default: |
2381 | break; |
2382 | } |
2383 | break; |
2384 | |
2385 | /*-----------------------------------------------------------------*/ |
2386 | /* Match a negated single character casefully. */ |
2387 | |
2388 | case OP_NOT: |
2389 | if (clen > 0 && c != d) { ADD_NEW(state_offset + dlen + 1, 0); } |
2390 | break; |
2391 | |
2392 | /*-----------------------------------------------------------------*/ |
2393 | /* Match a negated single character caselessly. */ |
2394 | |
2395 | case OP_NOTI: |
2396 | if (clen > 0) |
2397 | { |
2398 | uint32_t otherd; |
2399 | #ifdef SUPPORT_UNICODE |
2400 | if (utf_or_ucp && d >= 128) |
2401 | otherd = UCD_OTHERCASE(d); |
2402 | else |
2403 | #endif /* SUPPORT_UNICODE */ |
2404 | otherd = TABLE_GET(d, fcc, d); |
2405 | if (c != d && c != otherd) |
2406 | { ADD_NEW(state_offset + dlen + 1, 0); } |
2407 | } |
2408 | break; |
2409 | |
2410 | /*-----------------------------------------------------------------*/ |
2411 | case OP_PLUSI: |
2412 | case OP_MINPLUSI: |
2413 | case OP_POSPLUSI: |
2414 | case OP_NOTPLUSI: |
2415 | case OP_NOTMINPLUSI: |
2416 | case OP_NOTPOSPLUSI: |
2417 | caseless = TRUE; |
2418 | codevalue -= OP_STARI - OP_STAR; |
2419 | |
2420 | /* Fall through */ |
2421 | case OP_PLUS: |
2422 | case OP_MINPLUS: |
2423 | case OP_POSPLUS: |
2424 | case OP_NOTPLUS: |
2425 | case OP_NOTMINPLUS: |
2426 | case OP_NOTPOSPLUS: |
2427 | count = current_state->count; /* Already matched */ |
2428 | if (count > 0) { ADD_ACTIVE(state_offset + dlen + 1, 0); } |
2429 | if (clen > 0) |
2430 | { |
2431 | uint32_t otherd = NOTACHAR; |
2432 | if (caseless) |
2433 | { |
2434 | #ifdef SUPPORT_UNICODE |
2435 | if (utf_or_ucp && d >= 128) |
2436 | otherd = UCD_OTHERCASE(d); |
2437 | else |
2438 | #endif /* SUPPORT_UNICODE */ |
2439 | otherd = TABLE_GET(d, fcc, d); |
2440 | } |
2441 | if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR)) |
2442 | { |
2443 | if (count > 0 && |
2444 | (codevalue == OP_POSPLUS || codevalue == OP_NOTPOSPLUS)) |
2445 | { |
2446 | active_count--; /* Remove non-match possibility */ |
2447 | next_active_state--; |
2448 | } |
2449 | count++; |
2450 | ADD_NEW(state_offset, count); |
2451 | } |
2452 | } |
2453 | break; |
2454 | |
2455 | /*-----------------------------------------------------------------*/ |
2456 | case OP_QUERYI: |
2457 | case OP_MINQUERYI: |
2458 | case OP_POSQUERYI: |
2459 | case OP_NOTQUERYI: |
2460 | case OP_NOTMINQUERYI: |
2461 | case OP_NOTPOSQUERYI: |
2462 | caseless = TRUE; |
2463 | codevalue -= OP_STARI - OP_STAR; |
2464 | /* Fall through */ |
2465 | case OP_QUERY: |
2466 | case OP_MINQUERY: |
2467 | case OP_POSQUERY: |
2468 | case OP_NOTQUERY: |
2469 | case OP_NOTMINQUERY: |
2470 | case OP_NOTPOSQUERY: |
2471 | ADD_ACTIVE(state_offset + dlen + 1, 0); |
2472 | if (clen > 0) |
2473 | { |
2474 | uint32_t otherd = NOTACHAR; |
2475 | if (caseless) |
2476 | { |
2477 | #ifdef SUPPORT_UNICODE |
2478 | if (utf_or_ucp && d >= 128) |
2479 | otherd = UCD_OTHERCASE(d); |
2480 | else |
2481 | #endif /* SUPPORT_UNICODE */ |
2482 | otherd = TABLE_GET(d, fcc, d); |
2483 | } |
2484 | if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR)) |
2485 | { |
2486 | if (codevalue == OP_POSQUERY || codevalue == OP_NOTPOSQUERY) |
2487 | { |
2488 | active_count--; /* Remove non-match possibility */ |
2489 | next_active_state--; |
2490 | } |
2491 | ADD_NEW(state_offset + dlen + 1, 0); |
2492 | } |
2493 | } |
2494 | break; |
2495 | |
2496 | /*-----------------------------------------------------------------*/ |
2497 | case OP_STARI: |
2498 | case OP_MINSTARI: |
2499 | case OP_POSSTARI: |
2500 | case OP_NOTSTARI: |
2501 | case OP_NOTMINSTARI: |
2502 | case OP_NOTPOSSTARI: |
2503 | caseless = TRUE; |
2504 | codevalue -= OP_STARI - OP_STAR; |
2505 | /* Fall through */ |
2506 | case OP_STAR: |
2507 | case OP_MINSTAR: |
2508 | case OP_POSSTAR: |
2509 | case OP_NOTSTAR: |
2510 | case OP_NOTMINSTAR: |
2511 | case OP_NOTPOSSTAR: |
2512 | ADD_ACTIVE(state_offset + dlen + 1, 0); |
2513 | if (clen > 0) |
2514 | { |
2515 | uint32_t otherd = NOTACHAR; |
2516 | if (caseless) |
2517 | { |
2518 | #ifdef SUPPORT_UNICODE |
2519 | if (utf_or_ucp && d >= 128) |
2520 | otherd = UCD_OTHERCASE(d); |
2521 | else |
2522 | #endif /* SUPPORT_UNICODE */ |
2523 | otherd = TABLE_GET(d, fcc, d); |
2524 | } |
2525 | if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR)) |
2526 | { |
2527 | if (codevalue == OP_POSSTAR || codevalue == OP_NOTPOSSTAR) |
2528 | { |
2529 | active_count--; /* Remove non-match possibility */ |
2530 | next_active_state--; |
2531 | } |
2532 | ADD_NEW(state_offset, 0); |
2533 | } |
2534 | } |
2535 | break; |
2536 | |
2537 | /*-----------------------------------------------------------------*/ |
2538 | case OP_EXACTI: |
2539 | case OP_NOTEXACTI: |
2540 | caseless = TRUE; |
2541 | codevalue -= OP_STARI - OP_STAR; |
2542 | /* Fall through */ |
2543 | case OP_EXACT: |
2544 | case OP_NOTEXACT: |
2545 | count = current_state->count; /* Number already matched */ |
2546 | if (clen > 0) |
2547 | { |
2548 | uint32_t otherd = NOTACHAR; |
2549 | if (caseless) |
2550 | { |
2551 | #ifdef SUPPORT_UNICODE |
2552 | if (utf_or_ucp && d >= 128) |
2553 | otherd = UCD_OTHERCASE(d); |
2554 | else |
2555 | #endif /* SUPPORT_UNICODE */ |
2556 | otherd = TABLE_GET(d, fcc, d); |
2557 | } |
2558 | if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR)) |
2559 | { |
2560 | if (++count >= (int)GET2(code, 1)) |
2561 | { ADD_NEW(state_offset + dlen + 1 + IMM2_SIZE, 0); } |
2562 | else |
2563 | { ADD_NEW(state_offset, count); } |
2564 | } |
2565 | } |
2566 | break; |
2567 | |
2568 | /*-----------------------------------------------------------------*/ |
2569 | case OP_UPTOI: |
2570 | case OP_MINUPTOI: |
2571 | case OP_POSUPTOI: |
2572 | case OP_NOTUPTOI: |
2573 | case OP_NOTMINUPTOI: |
2574 | case OP_NOTPOSUPTOI: |
2575 | caseless = TRUE; |
2576 | codevalue -= OP_STARI - OP_STAR; |
2577 | /* Fall through */ |
2578 | case OP_UPTO: |
2579 | case OP_MINUPTO: |
2580 | case OP_POSUPTO: |
2581 | case OP_NOTUPTO: |
2582 | case OP_NOTMINUPTO: |
2583 | case OP_NOTPOSUPTO: |
2584 | ADD_ACTIVE(state_offset + dlen + 1 + IMM2_SIZE, 0); |
2585 | count = current_state->count; /* Number already matched */ |
2586 | if (clen > 0) |
2587 | { |
2588 | uint32_t otherd = NOTACHAR; |
2589 | if (caseless) |
2590 | { |
2591 | #ifdef SUPPORT_UNICODE |
2592 | if (utf_or_ucp && d >= 128) |
2593 | otherd = UCD_OTHERCASE(d); |
2594 | else |
2595 | #endif /* SUPPORT_UNICODE */ |
2596 | otherd = TABLE_GET(d, fcc, d); |
2597 | } |
2598 | if ((c == d || c == otherd) == (codevalue < OP_NOTSTAR)) |
2599 | { |
2600 | if (codevalue == OP_POSUPTO || codevalue == OP_NOTPOSUPTO) |
2601 | { |
2602 | active_count--; /* Remove non-match possibility */ |
2603 | next_active_state--; |
2604 | } |
2605 | if (++count >= (int)GET2(code, 1)) |
2606 | { ADD_NEW(state_offset + dlen + 1 + IMM2_SIZE, 0); } |
2607 | else |
2608 | { ADD_NEW(state_offset, count); } |
2609 | } |
2610 | } |
2611 | break; |
2612 | |
2613 | |
2614 | /* ========================================================================== */ |
2615 | /* These are the class-handling opcodes */ |
2616 | |
2617 | case OP_CLASS: |
2618 | case OP_NCLASS: |
2619 | case OP_XCLASS: |
2620 | { |
2621 | BOOL isinclass = FALSE; |
2622 | int next_state_offset; |
2623 | PCRE2_SPTR ecode; |
2624 | |
2625 | /* For a simple class, there is always just a 32-byte table, and we |
2626 | can set isinclass from it. */ |
2627 | |
2628 | if (codevalue != OP_XCLASS) |
2629 | { |
2630 | ecode = code + 1 + (32 / sizeof(PCRE2_UCHAR)); |
2631 | if (clen > 0) |
2632 | { |
2633 | isinclass = (c > 255)? (codevalue == OP_NCLASS) : |
2634 | ((((uint8_t *)(code + 1))[c/8] & (1u << (c&7))) != 0); |
2635 | } |
2636 | } |
2637 | |
2638 | /* An extended class may have a table or a list of single characters, |
2639 | ranges, or both, and it may be positive or negative. There's a |
2640 | function that sorts all this out. */ |
2641 | |
2642 | else |
2643 | { |
2644 | ecode = code + GET(code, 1); |
2645 | if (clen > 0) isinclass = PRIV(xclass)(c, code + 1 + LINK_SIZE, utf); |
2646 | } |
2647 | |
2648 | /* At this point, isinclass is set for all kinds of class, and ecode |
2649 | points to the byte after the end of the class. If there is a |
2650 | quantifier, this is where it will be. */ |
2651 | |
2652 | next_state_offset = (int)(ecode - start_code); |
2653 | |
2654 | switch (*ecode) |
2655 | { |
2656 | case OP_CRSTAR: |
2657 | case OP_CRMINSTAR: |
2658 | case OP_CRPOSSTAR: |
2659 | ADD_ACTIVE(next_state_offset + 1, 0); |
2660 | if (isinclass) |
2661 | { |
2662 | if (*ecode == OP_CRPOSSTAR) |
2663 | { |
2664 | active_count--; /* Remove non-match possibility */ |
2665 | next_active_state--; |
2666 | } |
2667 | ADD_NEW(state_offset, 0); |
2668 | } |
2669 | break; |
2670 | |
2671 | case OP_CRPLUS: |
2672 | case OP_CRMINPLUS: |
2673 | case OP_CRPOSPLUS: |
2674 | count = current_state->count; /* Already matched */ |
2675 | if (count > 0) { ADD_ACTIVE(next_state_offset + 1, 0); } |
2676 | if (isinclass) |
2677 | { |
2678 | if (count > 0 && *ecode == OP_CRPOSPLUS) |
2679 | { |
2680 | active_count--; /* Remove non-match possibility */ |
2681 | next_active_state--; |
2682 | } |
2683 | count++; |
2684 | ADD_NEW(state_offset, count); |
2685 | } |
2686 | break; |
2687 | |
2688 | case OP_CRQUERY: |
2689 | case OP_CRMINQUERY: |
2690 | case OP_CRPOSQUERY: |
2691 | ADD_ACTIVE(next_state_offset + 1, 0); |
2692 | if (isinclass) |
2693 | { |
2694 | if (*ecode == OP_CRPOSQUERY) |
2695 | { |
2696 | active_count--; /* Remove non-match possibility */ |
2697 | next_active_state--; |
2698 | } |
2699 | ADD_NEW(next_state_offset + 1, 0); |
2700 | } |
2701 | break; |
2702 | |
2703 | case OP_CRRANGE: |
2704 | case OP_CRMINRANGE: |
2705 | case OP_CRPOSRANGE: |
2706 | count = current_state->count; /* Already matched */ |
2707 | if (count >= (int)GET2(ecode, 1)) |
2708 | { ADD_ACTIVE(next_state_offset + 1 + 2 * IMM2_SIZE, 0); } |
2709 | if (isinclass) |
2710 | { |
2711 | int max = (int)GET2(ecode, 1 + IMM2_SIZE); |
2712 | |
2713 | if (*ecode == OP_CRPOSRANGE && count >= (int)GET2(ecode, 1)) |
2714 | { |
2715 | active_count--; /* Remove non-match possibility */ |
2716 | next_active_state--; |
2717 | } |
2718 | |
2719 | if (++count >= max && max != 0) /* Max 0 => no limit */ |
2720 | { ADD_NEW(next_state_offset + 1 + 2 * IMM2_SIZE, 0); } |
2721 | else |
2722 | { ADD_NEW(state_offset, count); } |
2723 | } |
2724 | break; |
2725 | |
2726 | default: |
2727 | if (isinclass) { ADD_NEW(next_state_offset, 0); } |
2728 | break; |
2729 | } |
2730 | } |
2731 | break; |
2732 | |
2733 | /* ========================================================================== */ |
2734 | /* These are the opcodes for fancy brackets of various kinds. We have |
2735 | to use recursion in order to handle them. The "always failing" assertion |
2736 | (?!) is optimised to OP_FAIL when compiling, so we have to support that, |
2737 | though the other "backtracking verbs" are not supported. */ |
2738 | |
2739 | case OP_FAIL: |
2740 | forced_fail++; /* Count FAILs for multiple states */ |
2741 | break; |
2742 | |
2743 | case OP_ASSERT: |
2744 | case OP_ASSERT_NOT: |
2745 | case OP_ASSERTBACK: |
2746 | case OP_ASSERTBACK_NOT: |
2747 | { |
2748 | int rc; |
2749 | int *local_workspace; |
2750 | PCRE2_SIZE *local_offsets; |
2751 | PCRE2_SPTR endasscode = code + GET(code, 1); |
2752 | RWS_anchor *rws = (RWS_anchor *)RWS; |
2753 | |
2754 | if (rws->free < RWS_RSIZE + RWS_OVEC_OSIZE) |
2755 | { |
2756 | rc = more_workspace(&rws, RWS_OVEC_OSIZE, mb); |
2757 | if (rc != 0) return rc; |
2758 | RWS = (int *)rws; |
2759 | } |
2760 | |
2761 | local_offsets = (PCRE2_SIZE *)(RWS + rws->size - rws->free); |
2762 | local_workspace = ((int *)local_offsets) + RWS_OVEC_OSIZE; |
2763 | rws->free -= RWS_RSIZE + RWS_OVEC_OSIZE; |
2764 | |
2765 | while (*endasscode == OP_ALT) endasscode += GET(endasscode, 1); |
2766 | |
2767 | rc = internal_dfa_match( |
2768 | mb, /* static match data */ |
2769 | code, /* this subexpression's code */ |
2770 | ptr, /* where we currently are */ |
2771 | (PCRE2_SIZE)(ptr - start_subject), /* start offset */ |
2772 | local_offsets, /* offset vector */ |
2773 | RWS_OVEC_OSIZE/OVEC_UNIT, /* size of same */ |
2774 | local_workspace, /* workspace vector */ |
2775 | RWS_RSIZE, /* size of same */ |
2776 | rlevel, /* function recursion level */ |
2777 | RWS); /* recursion workspace */ |
2778 | |
2779 | rws->free += RWS_RSIZE + RWS_OVEC_OSIZE; |
2780 | |
2781 | if (rc < 0 && rc != PCRE2_ERROR_NOMATCH) return rc; |
2782 | if ((rc >= 0) == (codevalue == OP_ASSERT || codevalue == OP_ASSERTBACK)) |
2783 | { ADD_ACTIVE((int)(endasscode + LINK_SIZE + 1 - start_code), 0); } |
2784 | } |
2785 | break; |
2786 | |
2787 | /*-----------------------------------------------------------------*/ |
2788 | case OP_COND: |
2789 | case OP_SCOND: |
2790 | { |
2791 | int codelink = (int)GET(code, 1); |
2792 | PCRE2_UCHAR condcode; |
2793 | |
2794 | /* Because of the way auto-callout works during compile, a callout item |
2795 | is inserted between OP_COND and an assertion condition. This does not |
2796 | happen for the other conditions. */ |
2797 | |
2798 | if (code[LINK_SIZE + 1] == OP_CALLOUT |
2799 | || code[LINK_SIZE + 1] == OP_CALLOUT_STR) |
2800 | { |
2801 | PCRE2_SIZE callout_length; |
2802 | rrc = do_callout_dfa(code, offsets, current_subject, ptr, mb, |
2803 | 1 + LINK_SIZE, &callout_length); |
2804 | if (rrc < 0) return rrc; /* Abandon */ |
2805 | if (rrc > 0) break; /* Fail this thread */ |
2806 | code += callout_length; /* Skip callout data */ |
2807 | } |
2808 | |
2809 | condcode = code[LINK_SIZE+1]; |
2810 | |
2811 | /* Back reference conditions and duplicate named recursion conditions |
2812 | are not supported */ |
2813 | |
2814 | if (condcode == OP_CREF || condcode == OP_DNCREF || |
2815 | condcode == OP_DNRREF) |
2816 | return PCRE2_ERROR_DFA_UCOND; |
2817 | |
2818 | /* The DEFINE condition is always false, and the assertion (?!) is |
2819 | converted to OP_FAIL. */ |
2820 | |
2821 | if (condcode == OP_FALSE || condcode == OP_FAIL) |
2822 | { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); } |
2823 | |
2824 | /* There is also an always-true condition */ |
2825 | |
2826 | else if (condcode == OP_TRUE) |
2827 | { ADD_ACTIVE(state_offset + LINK_SIZE + 2, 0); } |
2828 | |
2829 | /* The only supported version of OP_RREF is for the value RREF_ANY, |
2830 | which means "test if in any recursion". We can't test for specifically |
2831 | recursed groups. */ |
2832 | |
2833 | else if (condcode == OP_RREF) |
2834 | { |
2835 | unsigned int value = GET2(code, LINK_SIZE + 2); |
2836 | if (value != RREF_ANY) return PCRE2_ERROR_DFA_UCOND; |
2837 | if (mb->recursive != NULL) |
2838 | { ADD_ACTIVE(state_offset + LINK_SIZE + 2 + IMM2_SIZE, 0); } |
2839 | else { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); } |
2840 | } |
2841 | |
2842 | /* Otherwise, the condition is an assertion */ |
2843 | |
2844 | else |
2845 | { |
2846 | int rc; |
2847 | int *local_workspace; |
2848 | PCRE2_SIZE *local_offsets; |
2849 | PCRE2_SPTR asscode = code + LINK_SIZE + 1; |
2850 | PCRE2_SPTR endasscode = asscode + GET(asscode, 1); |
2851 | RWS_anchor *rws = (RWS_anchor *)RWS; |
2852 | |
2853 | if (rws->free < RWS_RSIZE + RWS_OVEC_OSIZE) |
2854 | { |
2855 | rc = more_workspace(&rws, RWS_OVEC_OSIZE, mb); |
2856 | if (rc != 0) return rc; |
2857 | RWS = (int *)rws; |
2858 | } |
2859 | |
2860 | local_offsets = (PCRE2_SIZE *)(RWS + rws->size - rws->free); |
2861 | local_workspace = ((int *)local_offsets) + RWS_OVEC_OSIZE; |
2862 | rws->free -= RWS_RSIZE + RWS_OVEC_OSIZE; |
2863 | |
2864 | while (*endasscode == OP_ALT) endasscode += GET(endasscode, 1); |
2865 | |
2866 | rc = internal_dfa_match( |
2867 | mb, /* fixed match data */ |
2868 | asscode, /* this subexpression's code */ |
2869 | ptr, /* where we currently are */ |
2870 | (PCRE2_SIZE)(ptr - start_subject), /* start offset */ |
2871 | local_offsets, /* offset vector */ |
2872 | RWS_OVEC_OSIZE/OVEC_UNIT, /* size of same */ |
2873 | local_workspace, /* workspace vector */ |
2874 | RWS_RSIZE, /* size of same */ |
2875 | rlevel, /* function recursion level */ |
2876 | RWS); /* recursion workspace */ |
2877 | |
2878 | rws->free += RWS_RSIZE + RWS_OVEC_OSIZE; |
2879 | |
2880 | if (rc < 0 && rc != PCRE2_ERROR_NOMATCH) return rc; |
2881 | if ((rc >= 0) == |
2882 | (condcode == OP_ASSERT || condcode == OP_ASSERTBACK)) |
2883 | { ADD_ACTIVE((int)(endasscode + LINK_SIZE + 1 - start_code), 0); } |
2884 | else |
2885 | { ADD_ACTIVE(state_offset + codelink + LINK_SIZE + 1, 0); } |
2886 | } |
2887 | } |
2888 | break; |
2889 | |
2890 | /*-----------------------------------------------------------------*/ |
2891 | case OP_RECURSE: |
2892 | { |
2893 | int rc; |
2894 | int *local_workspace; |
2895 | PCRE2_SIZE *local_offsets; |
2896 | RWS_anchor *rws = (RWS_anchor *)RWS; |
2897 | dfa_recursion_info *ri; |
2898 | PCRE2_SPTR callpat = start_code + GET(code, 1); |
2899 | uint32_t recno = (callpat == mb->start_code)? 0 : |
2900 | GET2(callpat, 1 + LINK_SIZE); |
2901 | |
2902 | if (rws->free < RWS_RSIZE + RWS_OVEC_RSIZE) |
2903 | { |
2904 | rc = more_workspace(&rws, RWS_OVEC_RSIZE, mb); |
2905 | if (rc != 0) return rc; |
2906 | RWS = (int *)rws; |
2907 | } |
2908 | |
2909 | local_offsets = (PCRE2_SIZE *)(RWS + rws->size - rws->free); |
2910 | local_workspace = ((int *)local_offsets) + RWS_OVEC_RSIZE; |
2911 | rws->free -= RWS_RSIZE + RWS_OVEC_RSIZE; |
2912 | |
2913 | /* Check for repeating a recursion without advancing the subject |
2914 | pointer. This should catch convoluted mutual recursions. (Some simple |
2915 | cases are caught at compile time.) */ |
2916 | |
2917 | for (ri = mb->recursive; ri != NULL; ri = ri->prevrec) |
2918 | if (recno == ri->group_num && ptr == ri->subject_position) |
2919 | return PCRE2_ERROR_RECURSELOOP; |
2920 | |
2921 | /* Remember this recursion and where we started it so as to |
2922 | catch infinite loops. */ |
2923 | |
2924 | new_recursive.group_num = recno; |
2925 | new_recursive.subject_position = ptr; |
2926 | new_recursive.prevrec = mb->recursive; |
2927 | mb->recursive = &new_recursive; |
2928 | |
2929 | rc = internal_dfa_match( |
2930 | mb, /* fixed match data */ |
2931 | callpat, /* this subexpression's code */ |
2932 | ptr, /* where we currently are */ |
2933 | (PCRE2_SIZE)(ptr - start_subject), /* start offset */ |
2934 | local_offsets, /* offset vector */ |
2935 | RWS_OVEC_RSIZE/OVEC_UNIT, /* size of same */ |
2936 | local_workspace, /* workspace vector */ |
2937 | RWS_RSIZE, /* size of same */ |
2938 | rlevel, /* function recursion level */ |
2939 | RWS); /* recursion workspace */ |
2940 | |
2941 | rws->free += RWS_RSIZE + RWS_OVEC_RSIZE; |
2942 | mb->recursive = new_recursive.prevrec; /* Done this recursion */ |
2943 | |
2944 | /* Ran out of internal offsets */ |
2945 | |
2946 | if (rc == 0) return PCRE2_ERROR_DFA_RECURSE; |
2947 | |
2948 | /* For each successful matched substring, set up the next state with a |
2949 | count of characters to skip before trying it. Note that the count is in |
2950 | characters, not bytes. */ |
2951 | |
2952 | if (rc > 0) |
2953 | { |
2954 | for (rc = rc*2 - 2; rc >= 0; rc -= 2) |
2955 | { |
2956 | PCRE2_SIZE charcount = local_offsets[rc+1] - local_offsets[rc]; |
2957 | #if defined SUPPORT_UNICODE && PCRE2_CODE_UNIT_WIDTH != 32 |
2958 | if (utf) |
2959 | { |
2960 | PCRE2_SPTR p = start_subject + local_offsets[rc]; |
2961 | PCRE2_SPTR pp = start_subject + local_offsets[rc+1]; |
2962 | while (p < pp) if (NOT_FIRSTCU(*p++)) charcount--; |
2963 | } |
2964 | #endif |
2965 | if (charcount > 0) |
2966 | { |
2967 | ADD_NEW_DATA(-(state_offset + LINK_SIZE + 1), 0, |
2968 | (int)(charcount - 1)); |
2969 | } |
2970 | else |
2971 | { |
2972 | ADD_ACTIVE(state_offset + LINK_SIZE + 1, 0); |
2973 | } |
2974 | } |
2975 | } |
2976 | else if (rc != PCRE2_ERROR_NOMATCH) return rc; |
2977 | } |
2978 | break; |
2979 | |
2980 | /*-----------------------------------------------------------------*/ |
2981 | case OP_BRAPOS: |
2982 | case OP_SBRAPOS: |
2983 | case OP_CBRAPOS: |
2984 | case OP_SCBRAPOS: |
2985 | case OP_BRAPOSZERO: |
2986 | { |
2987 | int rc; |
2988 | int *local_workspace; |
2989 | PCRE2_SIZE *local_offsets; |
2990 | PCRE2_SIZE charcount, matched_count; |
2991 | PCRE2_SPTR local_ptr = ptr; |
2992 | RWS_anchor *rws = (RWS_anchor *)RWS; |
2993 | BOOL allow_zero; |
2994 | |
2995 | if (rws->free < RWS_RSIZE + RWS_OVEC_OSIZE) |
2996 | { |
2997 | rc = more_workspace(&rws, RWS_OVEC_OSIZE, mb); |
2998 | if (rc != 0) return rc; |
2999 | RWS = (int *)rws; |
3000 | } |
3001 | |
3002 | local_offsets = (PCRE2_SIZE *)(RWS + rws->size - rws->free); |
3003 | local_workspace = ((int *)local_offsets) + RWS_OVEC_OSIZE; |
3004 | rws->free -= RWS_RSIZE + RWS_OVEC_OSIZE; |
3005 | |
3006 | if (codevalue == OP_BRAPOSZERO) |
3007 | { |
3008 | allow_zero = TRUE; |
3009 | codevalue = *(++code); /* Codevalue will be one of above BRAs */ |
3010 | } |
3011 | else allow_zero = FALSE; |
3012 | |
3013 | /* Loop to match the subpattern as many times as possible as if it were |
3014 | a complete pattern. */ |
3015 | |
3016 | for (matched_count = 0;; matched_count++) |
3017 | { |
3018 | rc = internal_dfa_match( |
3019 | mb, /* fixed match data */ |
3020 | code, /* this subexpression's code */ |
3021 | local_ptr, /* where we currently are */ |
3022 | (PCRE2_SIZE)(ptr - start_subject), /* start offset */ |
3023 | local_offsets, /* offset vector */ |
3024 | RWS_OVEC_OSIZE/OVEC_UNIT, /* size of same */ |
3025 | local_workspace, /* workspace vector */ |
3026 | RWS_RSIZE, /* size of same */ |
3027 | rlevel, /* function recursion level */ |
3028 | RWS); /* recursion workspace */ |
3029 | |
3030 | /* Failed to match */ |
3031 | |
3032 | if (rc < 0) |
3033 | { |
3034 | if (rc != PCRE2_ERROR_NOMATCH) return rc; |
3035 | break; |
3036 | } |
3037 | |
3038 | /* Matched: break the loop if zero characters matched. */ |
3039 | |
3040 | charcount = local_offsets[1] - local_offsets[0]; |
3041 | if (charcount == 0) break; |
3042 | local_ptr += charcount; /* Advance temporary position ptr */ |
3043 | } |
3044 | |
3045 | rws->free += RWS_RSIZE + RWS_OVEC_OSIZE; |
3046 | |
3047 | /* At this point we have matched the subpattern matched_count |
3048 | times, and local_ptr is pointing to the character after the end of the |
3049 | last match. */ |
3050 | |
3051 | if (matched_count > 0 || allow_zero) |
3052 | { |
3053 | PCRE2_SPTR end_subpattern = code; |
3054 | int next_state_offset; |
3055 | |
3056 | do { end_subpattern += GET(end_subpattern, 1); } |
3057 | while (*end_subpattern == OP_ALT); |
3058 | next_state_offset = |
3059 | (int)(end_subpattern - start_code + LINK_SIZE + 1); |
3060 | |
3061 | /* Optimization: if there are no more active states, and there |
3062 | are no new states yet set up, then skip over the subject string |
3063 | right here, to save looping. Otherwise, set up the new state to swing |
3064 | into action when the end of the matched substring is reached. */ |
3065 | |
3066 | if (i + 1 >= active_count && new_count == 0) |
3067 | { |
3068 | ptr = local_ptr; |
3069 | clen = 0; |
3070 | ADD_NEW(next_state_offset, 0); |
3071 | } |
3072 | else |
3073 | { |
3074 | PCRE2_SPTR p = ptr; |
3075 | PCRE2_SPTR pp = local_ptr; |
3076 | charcount = (PCRE2_SIZE)(pp - p); |
3077 | #if defined SUPPORT_UNICODE && PCRE2_CODE_UNIT_WIDTH != 32 |
3078 | if (utf) while (p < pp) if (NOT_FIRSTCU(*p++)) charcount--; |
3079 | #endif |
3080 | ADD_NEW_DATA(-next_state_offset, 0, (int)(charcount - 1)); |
3081 | } |
3082 | } |
3083 | } |
3084 | break; |
3085 | |
3086 | /*-----------------------------------------------------------------*/ |
3087 | case OP_ONCE: |
3088 | { |
3089 | int rc; |
3090 | int *local_workspace; |
3091 | PCRE2_SIZE *local_offsets; |
3092 | RWS_anchor *rws = (RWS_anchor *)RWS; |
3093 | |
3094 | if (rws->free < RWS_RSIZE + RWS_OVEC_OSIZE) |
3095 | { |
3096 | rc = more_workspace(&rws, RWS_OVEC_OSIZE, mb); |
3097 | if (rc != 0) return rc; |
3098 | RWS = (int *)rws; |
3099 | } |
3100 | |
3101 | local_offsets = (PCRE2_SIZE *)(RWS + rws->size - rws->free); |
3102 | local_workspace = ((int *)local_offsets) + RWS_OVEC_OSIZE; |
3103 | rws->free -= RWS_RSIZE + RWS_OVEC_OSIZE; |
3104 | |
3105 | rc = internal_dfa_match( |
3106 | mb, /* fixed match data */ |
3107 | code, /* this subexpression's code */ |
3108 | ptr, /* where we currently are */ |
3109 | (PCRE2_SIZE)(ptr - start_subject), /* start offset */ |
3110 | local_offsets, /* offset vector */ |
3111 | RWS_OVEC_OSIZE/OVEC_UNIT, /* size of same */ |
3112 | local_workspace, /* workspace vector */ |
3113 | RWS_RSIZE, /* size of same */ |
3114 | rlevel, /* function recursion level */ |
3115 | RWS); /* recursion workspace */ |
3116 | |
3117 | rws->free += RWS_RSIZE + RWS_OVEC_OSIZE; |
3118 | |
3119 | if (rc >= 0) |
3120 | { |
3121 | PCRE2_SPTR end_subpattern = code; |
3122 | PCRE2_SIZE charcount = local_offsets[1] - local_offsets[0]; |
3123 | int next_state_offset, repeat_state_offset; |
3124 | |
3125 | do { end_subpattern += GET(end_subpattern, 1); } |
3126 | while (*end_subpattern == OP_ALT); |
3127 | next_state_offset = |
3128 | (int)(end_subpattern - start_code + LINK_SIZE + 1); |
3129 | |
3130 | /* If the end of this subpattern is KETRMAX or KETRMIN, we must |
3131 | arrange for the repeat state also to be added to the relevant list. |
3132 | Calculate the offset, or set -1 for no repeat. */ |
3133 | |
3134 | repeat_state_offset = (*end_subpattern == OP_KETRMAX || |
3135 | *end_subpattern == OP_KETRMIN)? |
3136 | (int)(end_subpattern - start_code - GET(end_subpattern, 1)) : -1; |
3137 | |
3138 | /* If we have matched an empty string, add the next state at the |
3139 | current character pointer. This is important so that the duplicate |
3140 | checking kicks in, which is what breaks infinite loops that match an |
3141 | empty string. */ |
3142 | |
3143 | if (charcount == 0) |
3144 | { |
3145 | ADD_ACTIVE(next_state_offset, 0); |
3146 | } |
3147 | |
3148 | /* Optimization: if there are no more active states, and there |
3149 | are no new states yet set up, then skip over the subject string |
3150 | right here, to save looping. Otherwise, set up the new state to swing |
3151 | into action when the end of the matched substring is reached. */ |
3152 | |
3153 | else if (i + 1 >= active_count && new_count == 0) |
3154 | { |
3155 | ptr += charcount; |
3156 | clen = 0; |
3157 | ADD_NEW(next_state_offset, 0); |
3158 | |
3159 | /* If we are adding a repeat state at the new character position, |
3160 | we must fudge things so that it is the only current state. |
3161 | Otherwise, it might be a duplicate of one we processed before, and |
3162 | that would cause it to be skipped. */ |
3163 | |
3164 | if (repeat_state_offset >= 0) |
3165 | { |
3166 | next_active_state = active_states; |
3167 | active_count = 0; |
3168 | i = -1; |
3169 | ADD_ACTIVE(repeat_state_offset, 0); |
3170 | } |
3171 | } |
3172 | else |
3173 | { |
3174 | #if defined SUPPORT_UNICODE && PCRE2_CODE_UNIT_WIDTH != 32 |
3175 | if (utf) |
3176 | { |
3177 | PCRE2_SPTR p = start_subject + local_offsets[0]; |
3178 | PCRE2_SPTR pp = start_subject + local_offsets[1]; |
3179 | while (p < pp) if (NOT_FIRSTCU(*p++)) charcount--; |
3180 | } |
3181 | #endif |
3182 | ADD_NEW_DATA(-next_state_offset, 0, (int)(charcount - 1)); |
3183 | if (repeat_state_offset >= 0) |
3184 | { ADD_NEW_DATA(-repeat_state_offset, 0, (int)(charcount - 1)); } |
3185 | } |
3186 | } |
3187 | else if (rc != PCRE2_ERROR_NOMATCH) return rc; |
3188 | } |
3189 | break; |
3190 | |
3191 | |
3192 | /* ========================================================================== */ |
3193 | /* Handle callouts */ |
3194 | |
3195 | case OP_CALLOUT: |
3196 | case OP_CALLOUT_STR: |
3197 | { |
3198 | PCRE2_SIZE callout_length; |
3199 | rrc = do_callout_dfa(code, offsets, current_subject, ptr, mb, 0, |
3200 | &callout_length); |
3201 | if (rrc < 0) return rrc; /* Abandon */ |
3202 | if (rrc == 0) |
3203 | { ADD_ACTIVE(state_offset + (int)callout_length, 0); } |
3204 | } |
3205 | break; |
3206 | |
3207 | |
3208 | /* ========================================================================== */ |
3209 | default: /* Unsupported opcode */ |
3210 | return PCRE2_ERROR_DFA_UITEM; |
3211 | } |
3212 | |
3213 | NEXT_ACTIVE_STATE: continue; |
3214 | |
3215 | } /* End of loop scanning active states */ |
3216 | |
3217 | /* We have finished the processing at the current subject character. If no |
3218 | new states have been set for the next character, we have found all the |
3219 | matches that we are going to find. If partial matching has been requested, |
3220 | check for appropriate conditions. |
3221 | |
3222 | The "forced_ fail" variable counts the number of (*F) encountered for the |
3223 | character. If it is equal to the original active_count (saved in |
3224 | workspace[1]) it means that (*F) was found on every active state. In this |
3225 | case we don't want to give a partial match. |
3226 | |
3227 | The "could_continue" variable is true if a state could have continued but |
3228 | for the fact that the end of the subject was reached. */ |
3229 | |
3230 | if (new_count <= 0) |
3231 | { |
3232 | if (could_continue && /* Some could go on, and */ |
3233 | forced_fail != workspace[1] && /* Not all forced fail & */ |
3234 | ( /* either... */ |
3235 | (mb->moptions & PCRE2_PARTIAL_HARD) != 0 /* Hard partial */ |
3236 | || /* or... */ |
3237 | ((mb->moptions & PCRE2_PARTIAL_SOFT) != 0 && /* Soft partial and */ |
3238 | match_count < 0) /* no matches */ |
3239 | ) && /* And... */ |
3240 | ( |
3241 | partial_newline || /* Either partial NL */ |
3242 | ( /* or ... */ |
3243 | ptr >= end_subject && /* End of subject and */ |
3244 | ( /* either */ |
3245 | ptr > mb->start_used_ptr || /* Inspected non-empty string */ |
3246 | mb->allowemptypartial /* or pattern has lookbehind */ |
3247 | ) /* or could match empty */ |
3248 | ) |
3249 | )) |
3250 | match_count = PCRE2_ERROR_PARTIAL; |
3251 | break; /* Exit from loop along the subject string */ |
3252 | } |
3253 | |
3254 | /* One or more states are active for the next character. */ |
3255 | |
3256 | ptr += clen; /* Advance to next subject character */ |
3257 | } /* Loop to move along the subject string */ |
3258 | |
3259 | /* Control gets here from "break" a few lines above. If we have a match and |
3260 | PCRE2_ENDANCHORED is set, the match fails. */ |
3261 | |
3262 | if (match_count >= 0 && |
3263 | ((mb->moptions | mb->poptions) & PCRE2_ENDANCHORED) != 0 && |
3264 | ptr < end_subject) |
3265 | match_count = PCRE2_ERROR_NOMATCH; |
3266 | |
3267 | return match_count; |
3268 | } |
3269 | |
3270 | |
3271 | |
3272 | /************************************************* |
3273 | * Match a pattern using the DFA algorithm * |
3274 | *************************************************/ |
3275 | |
3276 | /* This function matches a compiled pattern to a subject string, using the |
3277 | alternate matching algorithm that finds all matches at once. |
3278 | |
3279 | Arguments: |
3280 | code points to the compiled pattern |
3281 | subject subject string |
3282 | length length of subject string |
3283 | startoffset where to start matching in the subject |
3284 | options option bits |
3285 | match_data points to a match data structure |
3286 | gcontext points to a match context |
3287 | workspace pointer to workspace |
3288 | wscount size of workspace |
3289 | |
3290 | Returns: > 0 => number of match offset pairs placed in offsets |
3291 | = 0 => offsets overflowed; longest matches are present |
3292 | -1 => failed to match |
3293 | < -1 => some kind of unexpected problem |
3294 | */ |
3295 | |
3296 | PCRE2_EXP_DEFN int PCRE2_CALL_CONVENTION |
3297 | pcre2_dfa_match(const pcre2_code *code, PCRE2_SPTR subject, PCRE2_SIZE length, |
3298 | PCRE2_SIZE start_offset, uint32_t options, pcre2_match_data *match_data, |
3299 | pcre2_match_context *mcontext, int *workspace, PCRE2_SIZE wscount) |
3300 | { |
3301 | int rc; |
3302 | int was_zero_terminated = 0; |
3303 | |
3304 | const pcre2_real_code *re = (const pcre2_real_code *)code; |
3305 | |
3306 | PCRE2_SPTR start_match; |
3307 | PCRE2_SPTR end_subject; |
3308 | PCRE2_SPTR bumpalong_limit; |
3309 | PCRE2_SPTR req_cu_ptr; |
3310 | |
3311 | BOOL utf, anchored, startline, firstline; |
3312 | BOOL has_first_cu = FALSE; |
3313 | BOOL has_req_cu = FALSE; |
3314 | |
3315 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
3316 | PCRE2_SPTR memchr_found_first_cu = NULL; |
3317 | PCRE2_SPTR memchr_found_first_cu2 = NULL; |
3318 | #endif |
3319 | |
3320 | PCRE2_UCHAR first_cu = 0; |
3321 | PCRE2_UCHAR first_cu2 = 0; |
3322 | PCRE2_UCHAR req_cu = 0; |
3323 | PCRE2_UCHAR req_cu2 = 0; |
3324 | |
3325 | const uint8_t *start_bits = NULL; |
3326 | |
3327 | /* We need to have mb pointing to a match block, because the IS_NEWLINE macro |
3328 | is used below, and it expects NLBLOCK to be defined as a pointer. */ |
3329 | |
3330 | pcre2_callout_block cb; |
3331 | dfa_match_block actual_match_block; |
3332 | dfa_match_block *mb = &actual_match_block; |
3333 | |
3334 | /* Set up a starting block of memory for use during recursive calls to |
3335 | internal_dfa_match(). By putting this on the stack, it minimizes resource use |
3336 | in the case when it is not needed. If this is too small, more memory is |
3337 | obtained from the heap. At the start of each block is an anchor structure.*/ |
3338 | |
3339 | int base_recursion_workspace[RWS_BASE_SIZE]; |
3340 | RWS_anchor *rws = (RWS_anchor *)base_recursion_workspace; |
3341 | rws->next = NULL; |
3342 | rws->size = RWS_BASE_SIZE; |
3343 | rws->free = RWS_BASE_SIZE - RWS_ANCHOR_SIZE; |
3344 | |
3345 | /* Recognize NULL, length 0 as an empty string. */ |
3346 | |
3347 | if (subject == NULL && length == 0) subject = (PCRE2_SPTR)"" ; |
3348 | |
3349 | /* Plausibility checks */ |
3350 | |
3351 | if ((options & ~PUBLIC_DFA_MATCH_OPTIONS) != 0) return PCRE2_ERROR_BADOPTION; |
3352 | if (re == NULL || subject == NULL || workspace == NULL || match_data == NULL) |
3353 | return PCRE2_ERROR_NULL; |
3354 | |
3355 | if (length == PCRE2_ZERO_TERMINATED) |
3356 | { |
3357 | length = PRIV(strlen)(subject); |
3358 | was_zero_terminated = 1; |
3359 | } |
3360 | |
3361 | if (wscount < 20) return PCRE2_ERROR_DFA_WSSIZE; |
3362 | if (start_offset > length) return PCRE2_ERROR_BADOFFSET; |
3363 | |
3364 | /* Partial matching and PCRE2_ENDANCHORED are currently not allowed at the same |
3365 | time. */ |
3366 | |
3367 | if ((options & (PCRE2_PARTIAL_HARD|PCRE2_PARTIAL_SOFT)) != 0 && |
3368 | ((re->overall_options | options) & PCRE2_ENDANCHORED) != 0) |
3369 | return PCRE2_ERROR_BADOPTION; |
3370 | |
3371 | /* Invalid UTF support is not available for DFA matching. */ |
3372 | |
3373 | if ((re->overall_options & PCRE2_MATCH_INVALID_UTF) != 0) |
3374 | return PCRE2_ERROR_DFA_UINVALID_UTF; |
3375 | |
3376 | /* Check that the first field in the block is the magic number. If it is not, |
3377 | return with PCRE2_ERROR_BADMAGIC. */ |
3378 | |
3379 | if (re->magic_number != MAGIC_NUMBER) return PCRE2_ERROR_BADMAGIC; |
3380 | |
3381 | /* Check the code unit width. */ |
3382 | |
3383 | if ((re->flags & PCRE2_MODE_MASK) != PCRE2_CODE_UNIT_WIDTH/8) |
3384 | return PCRE2_ERROR_BADMODE; |
3385 | |
3386 | /* PCRE2_NOTEMPTY and PCRE2_NOTEMPTY_ATSTART are match-time flags in the |
3387 | options variable for this function. Users of PCRE2 who are not calling the |
3388 | function directly would like to have a way of setting these flags, in the same |
3389 | way that they can set pcre2_compile() flags like PCRE2_NO_AUTOPOSSESS with |
3390 | constructions like (*NO_AUTOPOSSESS). To enable this, (*NOTEMPTY) and |
3391 | (*NOTEMPTY_ATSTART) set bits in the pattern's "flag" function which can now be |
3392 | transferred to the options for this function. The bits are guaranteed to be |
3393 | adjacent, but do not have the same values. This bit of Boolean trickery assumes |
3394 | that the match-time bits are not more significant than the flag bits. If by |
3395 | accident this is not the case, a compile-time division by zero error will |
3396 | occur. */ |
3397 | |
3398 | #define FF (PCRE2_NOTEMPTY_SET|PCRE2_NE_ATST_SET) |
3399 | #define OO (PCRE2_NOTEMPTY|PCRE2_NOTEMPTY_ATSTART) |
3400 | options |= (re->flags & FF) / ((FF & (~FF+1)) / (OO & (~OO+1))); |
3401 | #undef FF |
3402 | #undef OO |
3403 | |
3404 | /* If restarting after a partial match, do some sanity checks on the contents |
3405 | of the workspace. */ |
3406 | |
3407 | if ((options & PCRE2_DFA_RESTART) != 0) |
3408 | { |
3409 | if ((workspace[0] & (-2)) != 0 || workspace[1] < 1 || |
3410 | workspace[1] > (int)((wscount - 2)/INTS_PER_STATEBLOCK)) |
3411 | return PCRE2_ERROR_DFA_BADRESTART; |
3412 | } |
3413 | |
3414 | /* Set some local values */ |
3415 | |
3416 | utf = (re->overall_options & PCRE2_UTF) != 0; |
3417 | start_match = subject + start_offset; |
3418 | end_subject = subject + length; |
3419 | req_cu_ptr = start_match - 1; |
3420 | anchored = (options & (PCRE2_ANCHORED|PCRE2_DFA_RESTART)) != 0 || |
3421 | (re->overall_options & PCRE2_ANCHORED) != 0; |
3422 | |
3423 | /* The "must be at the start of a line" flags are used in a loop when finding |
3424 | where to start. */ |
3425 | |
3426 | startline = (re->flags & PCRE2_STARTLINE) != 0; |
3427 | firstline = (re->overall_options & PCRE2_FIRSTLINE) != 0; |
3428 | bumpalong_limit = end_subject; |
3429 | |
3430 | /* Initialize and set up the fixed fields in the callout block, with a pointer |
3431 | in the match block. */ |
3432 | |
3433 | mb->cb = &cb; |
3434 | cb.version = 2; |
3435 | cb.subject = subject; |
3436 | cb.subject_length = (PCRE2_SIZE)(end_subject - subject); |
3437 | cb.callout_flags = 0; |
3438 | cb.capture_top = 1; /* No capture support */ |
3439 | cb.capture_last = 0; |
3440 | cb.mark = NULL; /* No (*MARK) support */ |
3441 | |
3442 | /* Get data from the match context, if present, and fill in the remaining |
3443 | fields in the match block. It is an error to set an offset limit without |
3444 | setting the flag at compile time. */ |
3445 | |
3446 | if (mcontext == NULL) |
3447 | { |
3448 | mb->callout = NULL; |
3449 | mb->memctl = re->memctl; |
3450 | mb->match_limit = PRIV(default_match_context).match_limit; |
3451 | mb->match_limit_depth = PRIV(default_match_context).depth_limit; |
3452 | mb->heap_limit = PRIV(default_match_context).heap_limit; |
3453 | } |
3454 | else |
3455 | { |
3456 | if (mcontext->offset_limit != PCRE2_UNSET) |
3457 | { |
3458 | if ((re->overall_options & PCRE2_USE_OFFSET_LIMIT) == 0) |
3459 | return PCRE2_ERROR_BADOFFSETLIMIT; |
3460 | bumpalong_limit = subject + mcontext->offset_limit; |
3461 | } |
3462 | mb->callout = mcontext->callout; |
3463 | mb->callout_data = mcontext->callout_data; |
3464 | mb->memctl = mcontext->memctl; |
3465 | mb->match_limit = mcontext->match_limit; |
3466 | mb->match_limit_depth = mcontext->depth_limit; |
3467 | mb->heap_limit = mcontext->heap_limit; |
3468 | } |
3469 | |
3470 | if (mb->match_limit > re->limit_match) |
3471 | mb->match_limit = re->limit_match; |
3472 | |
3473 | if (mb->match_limit_depth > re->limit_depth) |
3474 | mb->match_limit_depth = re->limit_depth; |
3475 | |
3476 | if (mb->heap_limit > re->limit_heap) |
3477 | mb->heap_limit = re->limit_heap; |
3478 | |
3479 | mb->start_code = (PCRE2_UCHAR *)((uint8_t *)re + sizeof(pcre2_real_code)) + |
3480 | re->name_count * re->name_entry_size; |
3481 | mb->tables = re->tables; |
3482 | mb->start_subject = subject; |
3483 | mb->end_subject = end_subject; |
3484 | mb->start_offset = start_offset; |
3485 | mb->allowemptypartial = (re->max_lookbehind > 0) || |
3486 | (re->flags & PCRE2_MATCH_EMPTY) != 0; |
3487 | mb->moptions = options; |
3488 | mb->poptions = re->overall_options; |
3489 | mb->match_call_count = 0; |
3490 | mb->heap_used = 0; |
3491 | |
3492 | /* Process the \R and newline settings. */ |
3493 | |
3494 | mb->bsr_convention = re->bsr_convention; |
3495 | mb->nltype = NLTYPE_FIXED; |
3496 | switch(re->newline_convention) |
3497 | { |
3498 | case PCRE2_NEWLINE_CR: |
3499 | mb->nllen = 1; |
3500 | mb->nl[0] = CHAR_CR; |
3501 | break; |
3502 | |
3503 | case PCRE2_NEWLINE_LF: |
3504 | mb->nllen = 1; |
3505 | mb->nl[0] = CHAR_NL; |
3506 | break; |
3507 | |
3508 | case PCRE2_NEWLINE_NUL: |
3509 | mb->nllen = 1; |
3510 | mb->nl[0] = CHAR_NUL; |
3511 | break; |
3512 | |
3513 | case PCRE2_NEWLINE_CRLF: |
3514 | mb->nllen = 2; |
3515 | mb->nl[0] = CHAR_CR; |
3516 | mb->nl[1] = CHAR_NL; |
3517 | break; |
3518 | |
3519 | case PCRE2_NEWLINE_ANY: |
3520 | mb->nltype = NLTYPE_ANY; |
3521 | break; |
3522 | |
3523 | case PCRE2_NEWLINE_ANYCRLF: |
3524 | mb->nltype = NLTYPE_ANYCRLF; |
3525 | break; |
3526 | |
3527 | default: return PCRE2_ERROR_INTERNAL; |
3528 | } |
3529 | |
3530 | /* Check a UTF string for validity if required. For 8-bit and 16-bit strings, |
3531 | we must also check that a starting offset does not point into the middle of a |
3532 | multiunit character. We check only the portion of the subject that is going to |
3533 | be inspected during matching - from the offset minus the maximum back reference |
3534 | to the given length. This saves time when a small part of a large subject is |
3535 | being matched by the use of a starting offset. Note that the maximum lookbehind |
3536 | is a number of characters, not code units. */ |
3537 | |
3538 | #ifdef SUPPORT_UNICODE |
3539 | if (utf && (options & PCRE2_NO_UTF_CHECK) == 0) |
3540 | { |
3541 | PCRE2_SPTR check_subject = start_match; /* start_match includes offset */ |
3542 | |
3543 | if (start_offset > 0) |
3544 | { |
3545 | #if PCRE2_CODE_UNIT_WIDTH != 32 |
3546 | unsigned int i; |
3547 | if (start_match < end_subject && NOT_FIRSTCU(*start_match)) |
3548 | return PCRE2_ERROR_BADUTFOFFSET; |
3549 | for (i = re->max_lookbehind; i > 0 && check_subject > subject; i--) |
3550 | { |
3551 | check_subject--; |
3552 | while (check_subject > subject && |
3553 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
3554 | (*check_subject & 0xc0) == 0x80) |
3555 | #else /* 16-bit */ |
3556 | (*check_subject & 0xfc00) == 0xdc00) |
3557 | #endif /* PCRE2_CODE_UNIT_WIDTH == 8 */ |
3558 | check_subject--; |
3559 | } |
3560 | #else /* In the 32-bit library, one code unit equals one character. */ |
3561 | check_subject -= re->max_lookbehind; |
3562 | if (check_subject < subject) check_subject = subject; |
3563 | #endif /* PCRE2_CODE_UNIT_WIDTH != 32 */ |
3564 | } |
3565 | |
3566 | /* Validate the relevant portion of the subject. After an error, adjust the |
3567 | offset to be an absolute offset in the whole string. */ |
3568 | |
3569 | match_data->rc = PRIV(valid_utf)(check_subject, |
3570 | length - (PCRE2_SIZE)(check_subject - subject), &(match_data->startchar)); |
3571 | if (match_data->rc != 0) |
3572 | { |
3573 | match_data->startchar += (PCRE2_SIZE)(check_subject - subject); |
3574 | return match_data->rc; |
3575 | } |
3576 | } |
3577 | #endif /* SUPPORT_UNICODE */ |
3578 | |
3579 | /* Set up the first code unit to match, if available. If there's no first code |
3580 | unit there may be a bitmap of possible first characters. */ |
3581 | |
3582 | if ((re->flags & PCRE2_FIRSTSET) != 0) |
3583 | { |
3584 | has_first_cu = TRUE; |
3585 | first_cu = first_cu2 = (PCRE2_UCHAR)(re->first_codeunit); |
3586 | if ((re->flags & PCRE2_FIRSTCASELESS) != 0) |
3587 | { |
3588 | first_cu2 = TABLE_GET(first_cu, mb->tables + fcc_offset, first_cu); |
3589 | #ifdef SUPPORT_UNICODE |
3590 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
3591 | if (first_cu > 127 && !utf && (re->overall_options & PCRE2_UCP) != 0) |
3592 | first_cu2 = (PCRE2_UCHAR)UCD_OTHERCASE(first_cu); |
3593 | #else |
3594 | if (first_cu > 127 && (utf || (re->overall_options & PCRE2_UCP) != 0)) |
3595 | first_cu2 = (PCRE2_UCHAR)UCD_OTHERCASE(first_cu); |
3596 | #endif |
3597 | #endif /* SUPPORT_UNICODE */ |
3598 | } |
3599 | } |
3600 | else |
3601 | if (!startline && (re->flags & PCRE2_FIRSTMAPSET) != 0) |
3602 | start_bits = re->start_bitmap; |
3603 | |
3604 | /* There may be a "last known required code unit" set. */ |
3605 | |
3606 | if ((re->flags & PCRE2_LASTSET) != 0) |
3607 | { |
3608 | has_req_cu = TRUE; |
3609 | req_cu = req_cu2 = (PCRE2_UCHAR)(re->last_codeunit); |
3610 | if ((re->flags & PCRE2_LASTCASELESS) != 0) |
3611 | { |
3612 | req_cu2 = TABLE_GET(req_cu, mb->tables + fcc_offset, req_cu); |
3613 | #ifdef SUPPORT_UNICODE |
3614 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
3615 | if (req_cu > 127 && !utf && (re->overall_options & PCRE2_UCP) != 0) |
3616 | req_cu2 = (PCRE2_UCHAR)UCD_OTHERCASE(req_cu); |
3617 | #else |
3618 | if (req_cu > 127 && (utf || (re->overall_options & PCRE2_UCP) != 0)) |
3619 | req_cu2 = (PCRE2_UCHAR)UCD_OTHERCASE(req_cu); |
3620 | #endif |
3621 | #endif /* SUPPORT_UNICODE */ |
3622 | } |
3623 | } |
3624 | |
3625 | /* If the match data block was previously used with PCRE2_COPY_MATCHED_SUBJECT, |
3626 | free the memory that was obtained. */ |
3627 | |
3628 | if ((match_data->flags & PCRE2_MD_COPIED_SUBJECT) != 0) |
3629 | { |
3630 | match_data->memctl.free((void *)match_data->subject, |
3631 | match_data->memctl.memory_data); |
3632 | match_data->flags &= ~PCRE2_MD_COPIED_SUBJECT; |
3633 | } |
3634 | |
3635 | /* Fill in fields that are always returned in the match data. */ |
3636 | |
3637 | match_data->code = re; |
3638 | match_data->subject = NULL; /* Default for no match */ |
3639 | match_data->mark = NULL; |
3640 | match_data->matchedby = PCRE2_MATCHEDBY_DFA_INTERPRETER; |
3641 | |
3642 | /* Call the main matching function, looping for a non-anchored regex after a |
3643 | failed match. If not restarting, perform certain optimizations at the start of |
3644 | a match. */ |
3645 | |
3646 | for (;;) |
3647 | { |
3648 | /* ----------------- Start of match optimizations ---------------- */ |
3649 | |
3650 | /* There are some optimizations that avoid running the match if a known |
3651 | starting point is not found, or if a known later code unit is not present. |
3652 | However, there is an option (settable at compile time) that disables |
3653 | these, for testing and for ensuring that all callouts do actually occur. |
3654 | The optimizations must also be avoided when restarting a DFA match. */ |
3655 | |
3656 | if ((re->overall_options & PCRE2_NO_START_OPTIMIZE) == 0 && |
3657 | (options & PCRE2_DFA_RESTART) == 0) |
3658 | { |
3659 | /* If firstline is TRUE, the start of the match is constrained to the first |
3660 | line of a multiline string. That is, the match must be before or at the |
3661 | first newline following the start of matching. Temporarily adjust |
3662 | end_subject so that we stop the optimization scans for a first code unit |
3663 | immediately after the first character of a newline (the first code unit can |
3664 | legitimately be a newline). If the match fails at the newline, later code |
3665 | breaks this loop. */ |
3666 | |
3667 | if (firstline) |
3668 | { |
3669 | PCRE2_SPTR t = start_match; |
3670 | #ifdef SUPPORT_UNICODE |
3671 | if (utf) |
3672 | { |
3673 | while (t < end_subject && !IS_NEWLINE(t)) |
3674 | { |
3675 | t++; |
3676 | ACROSSCHAR(t < end_subject, t, t++); |
3677 | } |
3678 | } |
3679 | else |
3680 | #endif |
3681 | while (t < end_subject && !IS_NEWLINE(t)) t++; |
3682 | end_subject = t; |
3683 | } |
3684 | |
3685 | /* Anchored: check the first code unit if one is recorded. This may seem |
3686 | pointless but it can help in detecting a no match case without scanning for |
3687 | the required code unit. */ |
3688 | |
3689 | if (anchored) |
3690 | { |
3691 | if (has_first_cu || start_bits != NULL) |
3692 | { |
3693 | BOOL ok = start_match < end_subject; |
3694 | if (ok) |
3695 | { |
3696 | PCRE2_UCHAR c = UCHAR21TEST(start_match); |
3697 | ok = has_first_cu && (c == first_cu || c == first_cu2); |
3698 | if (!ok && start_bits != NULL) |
3699 | { |
3700 | #if PCRE2_CODE_UNIT_WIDTH != 8 |
3701 | if (c > 255) c = 255; |
3702 | #endif |
3703 | ok = (start_bits[c/8] & (1u << (c&7))) != 0; |
3704 | } |
3705 | } |
3706 | if (!ok) break; |
3707 | } |
3708 | } |
3709 | |
3710 | /* Not anchored. Advance to a unique first code unit if there is one. */ |
3711 | |
3712 | else |
3713 | { |
3714 | if (has_first_cu) |
3715 | { |
3716 | if (first_cu != first_cu2) /* Caseless */ |
3717 | { |
3718 | /* In 16-bit and 32_bit modes we have to do our own search, so can |
3719 | look for both cases at once. */ |
3720 | |
3721 | #if PCRE2_CODE_UNIT_WIDTH != 8 |
3722 | PCRE2_UCHAR smc; |
3723 | while (start_match < end_subject && |
3724 | (smc = UCHAR21TEST(start_match)) != first_cu && |
3725 | smc != first_cu2) |
3726 | start_match++; |
3727 | #else |
3728 | /* In 8-bit mode, the use of memchr() gives a big speed up, even |
3729 | though we have to call it twice in order to find the earliest |
3730 | occurrence of the code unit in either of its cases. Caching is used |
3731 | to remember the positions of previously found code units. This can |
3732 | make a huge difference when the strings are very long and only one |
3733 | case is actually present. */ |
3734 | |
3735 | PCRE2_SPTR pp1 = NULL; |
3736 | PCRE2_SPTR pp2 = NULL; |
3737 | PCRE2_SIZE searchlength = end_subject - start_match; |
3738 | |
3739 | /* If we haven't got a previously found position for first_cu, or if |
3740 | the current starting position is later, we need to do a search. If |
3741 | the code unit is not found, set it to the end. */ |
3742 | |
3743 | if (memchr_found_first_cu == NULL || |
3744 | start_match > memchr_found_first_cu) |
3745 | { |
3746 | pp1 = memchr(start_match, first_cu, searchlength); |
3747 | memchr_found_first_cu = (pp1 == NULL)? end_subject : pp1; |
3748 | } |
3749 | |
3750 | /* If the start is before a previously found position, use the |
3751 | previous position, or NULL if a previous search failed. */ |
3752 | |
3753 | else pp1 = (memchr_found_first_cu == end_subject)? NULL : |
3754 | memchr_found_first_cu; |
3755 | |
3756 | /* Do the same thing for the other case. */ |
3757 | |
3758 | if (memchr_found_first_cu2 == NULL || |
3759 | start_match > memchr_found_first_cu2) |
3760 | { |
3761 | pp2 = memchr(start_match, first_cu2, searchlength); |
3762 | memchr_found_first_cu2 = (pp2 == NULL)? end_subject : pp2; |
3763 | } |
3764 | |
3765 | else pp2 = (memchr_found_first_cu2 == end_subject)? NULL : |
3766 | memchr_found_first_cu2; |
3767 | |
3768 | /* Set the start to the end of the subject if neither case was found. |
3769 | Otherwise, use the earlier found point. */ |
3770 | |
3771 | if (pp1 == NULL) |
3772 | start_match = (pp2 == NULL)? end_subject : pp2; |
3773 | else |
3774 | start_match = (pp2 == NULL || pp1 < pp2)? pp1 : pp2; |
3775 | |
3776 | #endif /* 8-bit handling */ |
3777 | } |
3778 | |
3779 | /* The caseful case is much simpler. */ |
3780 | |
3781 | else |
3782 | { |
3783 | #if PCRE2_CODE_UNIT_WIDTH != 8 |
3784 | while (start_match < end_subject && UCHAR21TEST(start_match) != |
3785 | first_cu) |
3786 | start_match++; |
3787 | #else /* 8-bit code units */ |
3788 | start_match = memchr(start_match, first_cu, end_subject - start_match); |
3789 | if (start_match == NULL) start_match = end_subject; |
3790 | #endif |
3791 | } |
3792 | |
3793 | /* If we can't find the required code unit, having reached the true end |
3794 | of the subject, break the bumpalong loop, to force a match failure, |
3795 | except when doing partial matching, when we let the next cycle run at |
3796 | the end of the subject. To see why, consider the pattern /(?<=abc)def/, |
3797 | which partially matches "abc", even though the string does not contain |
3798 | the starting character "d". If we have not reached the true end of the |
3799 | subject (PCRE2_FIRSTLINE caused end_subject to be temporarily modified) |
3800 | we also let the cycle run, because the matching string is legitimately |
3801 | allowed to start with the first code unit of a newline. */ |
3802 | |
3803 | if ((mb->moptions & (PCRE2_PARTIAL_HARD|PCRE2_PARTIAL_SOFT)) == 0 && |
3804 | start_match >= mb->end_subject) |
3805 | break; |
3806 | } |
3807 | |
3808 | /* If there's no first code unit, advance to just after a linebreak for a |
3809 | multiline match if required. */ |
3810 | |
3811 | else if (startline) |
3812 | { |
3813 | if (start_match > mb->start_subject + start_offset) |
3814 | { |
3815 | #ifdef SUPPORT_UNICODE |
3816 | if (utf) |
3817 | { |
3818 | while (start_match < end_subject && !WAS_NEWLINE(start_match)) |
3819 | { |
3820 | start_match++; |
3821 | ACROSSCHAR(start_match < end_subject, start_match, start_match++); |
3822 | } |
3823 | } |
3824 | else |
3825 | #endif |
3826 | while (start_match < end_subject && !WAS_NEWLINE(start_match)) |
3827 | start_match++; |
3828 | |
3829 | /* If we have just passed a CR and the newline option is ANY or |
3830 | ANYCRLF, and we are now at a LF, advance the match position by one |
3831 | more code unit. */ |
3832 | |
3833 | if (start_match[-1] == CHAR_CR && |
3834 | (mb->nltype == NLTYPE_ANY || mb->nltype == NLTYPE_ANYCRLF) && |
3835 | start_match < end_subject && |
3836 | UCHAR21TEST(start_match) == CHAR_NL) |
3837 | start_match++; |
3838 | } |
3839 | } |
3840 | |
3841 | /* If there's no first code unit or a requirement for a multiline line |
3842 | start, advance to a non-unique first code unit if any have been |
3843 | identified. The bitmap contains only 256 bits. When code units are 16 or |
3844 | 32 bits wide, all code units greater than 254 set the 255 bit. */ |
3845 | |
3846 | else if (start_bits != NULL) |
3847 | { |
3848 | while (start_match < end_subject) |
3849 | { |
3850 | uint32_t c = UCHAR21TEST(start_match); |
3851 | #if PCRE2_CODE_UNIT_WIDTH != 8 |
3852 | if (c > 255) c = 255; |
3853 | #endif |
3854 | if ((start_bits[c/8] & (1u << (c&7))) != 0) break; |
3855 | start_match++; |
3856 | } |
3857 | |
3858 | /* See comment above in first_cu checking about the next line. */ |
3859 | |
3860 | if ((mb->moptions & (PCRE2_PARTIAL_HARD|PCRE2_PARTIAL_SOFT)) == 0 && |
3861 | start_match >= mb->end_subject) |
3862 | break; |
3863 | } |
3864 | } /* End of first code unit handling */ |
3865 | |
3866 | /* Restore fudged end_subject */ |
3867 | |
3868 | end_subject = mb->end_subject; |
3869 | |
3870 | /* The following two optimizations are disabled for partial matching. */ |
3871 | |
3872 | if ((mb->moptions & (PCRE2_PARTIAL_HARD|PCRE2_PARTIAL_SOFT)) == 0) |
3873 | { |
3874 | PCRE2_SPTR p; |
3875 | |
3876 | /* The minimum matching length is a lower bound; no actual string of that |
3877 | length may actually match the pattern. Although the value is, strictly, |
3878 | in characters, we treat it as code units to avoid spending too much time |
3879 | in this optimization. */ |
3880 | |
3881 | if (end_subject - start_match < re->minlength) goto NOMATCH_EXIT; |
3882 | |
3883 | /* If req_cu is set, we know that that code unit must appear in the |
3884 | subject for the match to succeed. If the first code unit is set, req_cu |
3885 | must be later in the subject; otherwise the test starts at the match |
3886 | point. This optimization can save a huge amount of backtracking in |
3887 | patterns with nested unlimited repeats that aren't going to match. |
3888 | Writing separate code for cased/caseless versions makes it go faster, as |
3889 | does using an autoincrement and backing off on a match. As in the case of |
3890 | the first code unit, using memchr() in the 8-bit library gives a big |
3891 | speed up. Unlike the first_cu check above, we do not need to call |
3892 | memchr() twice in the caseless case because we only need to check for the |
3893 | presence of the character in either case, not find the first occurrence. |
3894 | |
3895 | The search can be skipped if the code unit was found later than the |
3896 | current starting point in a previous iteration of the bumpalong loop. |
3897 | |
3898 | HOWEVER: when the subject string is very, very long, searching to its end |
3899 | can take a long time, and give bad performance on quite ordinary |
3900 | patterns. This showed up when somebody was matching something like |
3901 | /^\d+C/ on a 32-megabyte string... so we don't do this when the string is |
3902 | sufficiently long, but it's worth searching a lot more for unanchored |
3903 | patterns. */ |
3904 | |
3905 | p = start_match + (has_first_cu? 1:0); |
3906 | if (has_req_cu && p > req_cu_ptr) |
3907 | { |
3908 | PCRE2_SIZE check_length = end_subject - start_match; |
3909 | |
3910 | if (check_length < REQ_CU_MAX || |
3911 | (!anchored && check_length < REQ_CU_MAX * 1000)) |
3912 | { |
3913 | if (req_cu != req_cu2) /* Caseless */ |
3914 | { |
3915 | #if PCRE2_CODE_UNIT_WIDTH != 8 |
3916 | while (p < end_subject) |
3917 | { |
3918 | uint32_t pp = UCHAR21INCTEST(p); |
3919 | if (pp == req_cu || pp == req_cu2) { p--; break; } |
3920 | } |
3921 | #else /* 8-bit code units */ |
3922 | PCRE2_SPTR pp = p; |
3923 | p = memchr(pp, req_cu, end_subject - pp); |
3924 | if (p == NULL) |
3925 | { |
3926 | p = memchr(pp, req_cu2, end_subject - pp); |
3927 | if (p == NULL) p = end_subject; |
3928 | } |
3929 | #endif /* PCRE2_CODE_UNIT_WIDTH != 8 */ |
3930 | } |
3931 | |
3932 | /* The caseful case */ |
3933 | |
3934 | else |
3935 | { |
3936 | #if PCRE2_CODE_UNIT_WIDTH != 8 |
3937 | while (p < end_subject) |
3938 | { |
3939 | if (UCHAR21INCTEST(p) == req_cu) { p--; break; } |
3940 | } |
3941 | |
3942 | #else /* 8-bit code units */ |
3943 | p = memchr(p, req_cu, end_subject - p); |
3944 | if (p == NULL) p = end_subject; |
3945 | #endif |
3946 | } |
3947 | |
3948 | /* If we can't find the required code unit, break the matching loop, |
3949 | forcing a match failure. */ |
3950 | |
3951 | if (p >= end_subject) break; |
3952 | |
3953 | /* If we have found the required code unit, save the point where we |
3954 | found it, so that we don't search again next time round the loop if |
3955 | the start hasn't passed this code unit yet. */ |
3956 | |
3957 | req_cu_ptr = p; |
3958 | } |
3959 | } |
3960 | } |
3961 | } |
3962 | |
3963 | /* ------------ End of start of match optimizations ------------ */ |
3964 | |
3965 | /* Give no match if we have passed the bumpalong limit. */ |
3966 | |
3967 | if (start_match > bumpalong_limit) break; |
3968 | |
3969 | /* OK, now we can do the business */ |
3970 | |
3971 | mb->start_used_ptr = start_match; |
3972 | mb->last_used_ptr = start_match; |
3973 | mb->recursive = NULL; |
3974 | |
3975 | rc = internal_dfa_match( |
3976 | mb, /* fixed match data */ |
3977 | mb->start_code, /* this subexpression's code */ |
3978 | start_match, /* where we currently are */ |
3979 | start_offset, /* start offset in subject */ |
3980 | match_data->ovector, /* offset vector */ |
3981 | (uint32_t)match_data->oveccount * 2, /* actual size of same */ |
3982 | workspace, /* workspace vector */ |
3983 | (int)wscount, /* size of same */ |
3984 | 0, /* function recurse level */ |
3985 | base_recursion_workspace); /* initial workspace for recursion */ |
3986 | |
3987 | /* Anything other than "no match" means we are done, always; otherwise, carry |
3988 | on only if not anchored. */ |
3989 | |
3990 | if (rc != PCRE2_ERROR_NOMATCH || anchored) |
3991 | { |
3992 | if (rc == PCRE2_ERROR_PARTIAL && match_data->oveccount > 0) |
3993 | { |
3994 | match_data->ovector[0] = (PCRE2_SIZE)(start_match - subject); |
3995 | match_data->ovector[1] = (PCRE2_SIZE)(end_subject - subject); |
3996 | } |
3997 | match_data->leftchar = (PCRE2_SIZE)(mb->start_used_ptr - subject); |
3998 | match_data->rightchar = (PCRE2_SIZE)( mb->last_used_ptr - subject); |
3999 | match_data->startchar = (PCRE2_SIZE)(start_match - subject); |
4000 | match_data->rc = rc; |
4001 | |
4002 | if (rc >= 0 &&(options & PCRE2_COPY_MATCHED_SUBJECT) != 0) |
4003 | { |
4004 | length = CU2BYTES(length + was_zero_terminated); |
4005 | match_data->subject = match_data->memctl.malloc(length, |
4006 | match_data->memctl.memory_data); |
4007 | if (match_data->subject == NULL) return PCRE2_ERROR_NOMEMORY; |
4008 | memcpy((void *)match_data->subject, subject, length); |
4009 | match_data->flags |= PCRE2_MD_COPIED_SUBJECT; |
4010 | } |
4011 | else |
4012 | { |
4013 | if (rc >= 0 || rc == PCRE2_ERROR_PARTIAL) match_data->subject = subject; |
4014 | } |
4015 | goto EXIT; |
4016 | } |
4017 | |
4018 | /* Advance to the next subject character unless we are at the end of a line |
4019 | and firstline is set. */ |
4020 | |
4021 | if (firstline && IS_NEWLINE(start_match)) break; |
4022 | start_match++; |
4023 | #ifdef SUPPORT_UNICODE |
4024 | if (utf) |
4025 | { |
4026 | ACROSSCHAR(start_match < end_subject, start_match, start_match++); |
4027 | } |
4028 | #endif |
4029 | if (start_match > end_subject) break; |
4030 | |
4031 | /* If we have just passed a CR and we are now at a LF, and the pattern does |
4032 | not contain any explicit matches for \r or \n, and the newline option is CRLF |
4033 | or ANY or ANYCRLF, advance the match position by one more character. */ |
4034 | |
4035 | if (UCHAR21TEST(start_match - 1) == CHAR_CR && |
4036 | start_match < end_subject && |
4037 | UCHAR21TEST(start_match) == CHAR_NL && |
4038 | (re->flags & PCRE2_HASCRORLF) == 0 && |
4039 | (mb->nltype == NLTYPE_ANY || |
4040 | mb->nltype == NLTYPE_ANYCRLF || |
4041 | mb->nllen == 2)) |
4042 | start_match++; |
4043 | |
4044 | } /* "Bumpalong" loop */ |
4045 | |
4046 | NOMATCH_EXIT: |
4047 | rc = PCRE2_ERROR_NOMATCH; |
4048 | |
4049 | EXIT: |
4050 | while (rws->next != NULL) |
4051 | { |
4052 | RWS_anchor *next = rws->next; |
4053 | rws->next = next->next; |
4054 | mb->memctl.free(next, mb->memctl.memory_data); |
4055 | } |
4056 | |
4057 | return rc; |
4058 | } |
4059 | |
4060 | /* These #undefs are here to enable unity builds with CMake. */ |
4061 | |
4062 | #undef NLBLOCK /* Block containing newline information */ |
4063 | #undef PSSTART /* Field containing processed string start */ |
4064 | #undef PSEND /* Field containing processed string end */ |
4065 | |
4066 | /* End of pcre2_dfa_match.c */ |
4067 | |