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 | #ifdef HAVE_CONFIG_H |
43 | #include "config.h" |
44 | #endif |
45 | |
46 | #define NLBLOCK cb /* Block containing newline information */ |
47 | #define PSSTART start_pattern /* Field containing processed string start */ |
48 | #define PSEND end_pattern /* Field containing processed string end */ |
49 | |
50 | #include "pcre2_internal.h" |
51 | |
52 | /* In rare error cases debugging might require calling pcre2_printint(). */ |
53 | |
54 | #if 0 |
55 | #ifdef EBCDIC |
56 | #define PRINTABLE(c) ((c) >= 64 && (c) < 255) |
57 | #else |
58 | #define PRINTABLE(c) ((c) >= 32 && (c) < 127) |
59 | #endif |
60 | #include "pcre2_printint.c" |
61 | #define DEBUG_CALL_PRINTINT |
62 | #endif |
63 | |
64 | /* Other debugging code can be enabled by these defines. */ |
65 | |
66 | /* #define DEBUG_SHOW_CAPTURES */ |
67 | /* #define DEBUG_SHOW_PARSED */ |
68 | |
69 | /* There are a few things that vary with different code unit sizes. Handle them |
70 | by defining macros in order to minimize #if usage. */ |
71 | |
72 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
73 | #define STRING_UTFn_RIGHTPAR STRING_UTF8_RIGHTPAR, 5 |
74 | #define XDIGIT(c) xdigitab[c] |
75 | |
76 | #else /* Either 16-bit or 32-bit */ |
77 | #define XDIGIT(c) (MAX_255(c)? xdigitab[c] : 0xff) |
78 | |
79 | #if PCRE2_CODE_UNIT_WIDTH == 16 |
80 | #define STRING_UTFn_RIGHTPAR STRING_UTF16_RIGHTPAR, 6 |
81 | |
82 | #else /* 32-bit */ |
83 | #define STRING_UTFn_RIGHTPAR STRING_UTF32_RIGHTPAR, 6 |
84 | #endif |
85 | #endif |
86 | |
87 | /* Macros to store and retrieve a PCRE2_SIZE value in the parsed pattern, which |
88 | consists of uint32_t elements. Assume that if uint32_t can't hold it, two of |
89 | them will be able to (i.e. assume a 64-bit world). */ |
90 | |
91 | #if PCRE2_SIZE_MAX <= UINT32_MAX |
92 | #define PUTOFFSET(s,p) *p++ = s |
93 | #define GETOFFSET(s,p) s = *p++ |
94 | #define GETPLUSOFFSET(s,p) s = *(++p) |
95 | #define READPLUSOFFSET(s,p) s = p[1] |
96 | #define SKIPOFFSET(p) p++ |
97 | #define SIZEOFFSET 1 |
98 | #else |
99 | #define PUTOFFSET(s,p) \ |
100 | { *p++ = (uint32_t)(s >> 32); *p++ = (uint32_t)(s & 0xffffffff); } |
101 | #define GETOFFSET(s,p) \ |
102 | { s = ((PCRE2_SIZE)p[0] << 32) | (PCRE2_SIZE)p[1]; p += 2; } |
103 | #define GETPLUSOFFSET(s,p) \ |
104 | { s = ((PCRE2_SIZE)p[1] << 32) | (PCRE2_SIZE)p[2]; p += 2; } |
105 | #define READPLUSOFFSET(s,p) \ |
106 | { s = ((PCRE2_SIZE)p[1] << 32) | (PCRE2_SIZE)p[2]; } |
107 | #define SKIPOFFSET(p) p += 2 |
108 | #define SIZEOFFSET 2 |
109 | #endif |
110 | |
111 | /* Macros for manipulating elements of the parsed pattern vector. */ |
112 | |
113 | #define META_CODE(x) (x & 0xffff0000u) |
114 | #define META_DATA(x) (x & 0x0000ffffu) |
115 | #define META_DIFF(x,y) ((x-y)>>16) |
116 | |
117 | /* Function definitions to allow mutual recursion */ |
118 | |
119 | #ifdef SUPPORT_UNICODE |
120 | static unsigned int |
121 | add_list_to_class_internal(uint8_t *, PCRE2_UCHAR **, uint32_t, |
122 | compile_block *, const uint32_t *, unsigned int); |
123 | #endif |
124 | |
125 | static int |
126 | compile_regex(uint32_t, PCRE2_UCHAR **, uint32_t **, int *, uint32_t, |
127 | uint32_t *, uint32_t *, uint32_t *, uint32_t *, branch_chain *, |
128 | compile_block *, PCRE2_SIZE *); |
129 | |
130 | static int |
131 | get_branchlength(uint32_t **, int *, int *, parsed_recurse_check *, |
132 | compile_block *); |
133 | |
134 | static BOOL |
135 | set_lookbehind_lengths(uint32_t **, int *, int *, parsed_recurse_check *, |
136 | compile_block *); |
137 | |
138 | static int |
139 | check_lookbehinds(uint32_t *, uint32_t **, parsed_recurse_check *, |
140 | compile_block *, int *); |
141 | |
142 | |
143 | /************************************************* |
144 | * Code parameters and static tables * |
145 | *************************************************/ |
146 | |
147 | #define MAX_GROUP_NUMBER 65535u |
148 | #define MAX_REPEAT_COUNT 65535u |
149 | #define REPEAT_UNLIMITED (MAX_REPEAT_COUNT+1) |
150 | |
151 | /* COMPILE_WORK_SIZE specifies the size of stack workspace, which is used in |
152 | different ways in the different pattern scans. The parsing and group- |
153 | identifying pre-scan uses it to handle nesting, and needs it to be 16-bit |
154 | aligned for this. Having defined the size in code units, we set up |
155 | C16_WORK_SIZE as the number of elements in the 16-bit vector. |
156 | |
157 | During the first compiling phase, when determining how much memory is required, |
158 | the regex is partly compiled into this space, but the compiled parts are |
159 | discarded as soon as they can be, so that hopefully there will never be an |
160 | overrun. The code does, however, check for an overrun, which can occur for |
161 | pathological patterns. The size of the workspace depends on LINK_SIZE because |
162 | the length of compiled items varies with this. |
163 | |
164 | In the real compile phase, this workspace is not currently used. */ |
165 | |
166 | #define COMPILE_WORK_SIZE (3000*LINK_SIZE) /* Size in code units */ |
167 | |
168 | #define C16_WORK_SIZE \ |
169 | ((COMPILE_WORK_SIZE * sizeof(PCRE2_UCHAR))/sizeof(uint16_t)) |
170 | |
171 | /* A uint32_t vector is used for caching information about the size of |
172 | capturing groups, to improve performance. A default is created on the stack of |
173 | this size. */ |
174 | |
175 | #define GROUPINFO_DEFAULT_SIZE 256 |
176 | |
177 | /* The overrun tests check for a slightly smaller size so that they detect the |
178 | overrun before it actually does run off the end of the data block. */ |
179 | |
180 | #define WORK_SIZE_SAFETY_MARGIN (100) |
181 | |
182 | /* This value determines the size of the initial vector that is used for |
183 | remembering named groups during the pre-compile. It is allocated on the stack, |
184 | but if it is too small, it is expanded, in a similar way to the workspace. The |
185 | value is the number of slots in the list. */ |
186 | |
187 | #define NAMED_GROUP_LIST_SIZE 20 |
188 | |
189 | /* The pre-compiling pass over the pattern creates a parsed pattern in a vector |
190 | of uint32_t. For short patterns this lives on the stack, with this size. Heap |
191 | memory is used for longer patterns. */ |
192 | |
193 | #define PARSED_PATTERN_DEFAULT_SIZE 1024 |
194 | |
195 | /* Maximum length value to check against when making sure that the variable |
196 | that holds the compiled pattern length does not overflow. We make it a bit less |
197 | than INT_MAX to allow for adding in group terminating code units, so that we |
198 | don't have to check them every time. */ |
199 | |
200 | #define OFLOW_MAX (INT_MAX - 20) |
201 | |
202 | /* Code values for parsed patterns, which are stored in a vector of 32-bit |
203 | unsigned ints. Values less than META_END are literal data values. The coding |
204 | for identifying the item is in the top 16-bits, leaving 16 bits for the |
205 | additional data that some of them need. The META_CODE, META_DATA, and META_DIFF |
206 | macros are used to manipulate parsed pattern elements. |
207 | |
208 | NOTE: When these definitions are changed, the table of extra lengths for each |
209 | code (meta_extra_lengths, just below) must be updated to remain in step. */ |
210 | |
211 | #define META_END 0x80000000u /* End of pattern */ |
212 | |
213 | #define META_ALT 0x80010000u /* alternation */ |
214 | #define META_ATOMIC 0x80020000u /* atomic group */ |
215 | #define META_BACKREF 0x80030000u /* Back ref */ |
216 | #define META_BACKREF_BYNAME 0x80040000u /* \k'name' */ |
217 | #define META_BIGVALUE 0x80050000u /* Next is a literal > META_END */ |
218 | #define META_CALLOUT_NUMBER 0x80060000u /* (?C with numerical argument */ |
219 | #define META_CALLOUT_STRING 0x80070000u /* (?C with string argument */ |
220 | #define META_CAPTURE 0x80080000u /* Capturing parenthesis */ |
221 | #define META_CIRCUMFLEX 0x80090000u /* ^ metacharacter */ |
222 | #define META_CLASS 0x800a0000u /* start non-empty class */ |
223 | #define META_CLASS_EMPTY 0x800b0000u /* empty class */ |
224 | #define META_CLASS_EMPTY_NOT 0x800c0000u /* negative empty class */ |
225 | #define META_CLASS_END 0x800d0000u /* end of non-empty class */ |
226 | #define META_CLASS_NOT 0x800e0000u /* start non-empty negative class */ |
227 | #define META_COND_ASSERT 0x800f0000u /* (?(?assertion)... */ |
228 | #define META_COND_DEFINE 0x80100000u /* (?(DEFINE)... */ |
229 | #define META_COND_NAME 0x80110000u /* (?(<name>)... */ |
230 | #define META_COND_NUMBER 0x80120000u /* (?(digits)... */ |
231 | #define META_COND_RNAME 0x80130000u /* (?(R&name)... */ |
232 | #define META_COND_RNUMBER 0x80140000u /* (?(Rdigits)... */ |
233 | #define META_COND_VERSION 0x80150000u /* (?(VERSION<op>x.y)... */ |
234 | #define META_DOLLAR 0x80160000u /* $ metacharacter */ |
235 | #define META_DOT 0x80170000u /* . metacharacter */ |
236 | #define META_ESCAPE 0x80180000u /* \d and friends */ |
237 | #define META_KET 0x80190000u /* closing parenthesis */ |
238 | #define META_NOCAPTURE 0x801a0000u /* no capture parens */ |
239 | #define META_OPTIONS 0x801b0000u /* (?i) and friends */ |
240 | #define META_POSIX 0x801c0000u /* POSIX class item */ |
241 | #define META_POSIX_NEG 0x801d0000u /* negative POSIX class item */ |
242 | #define META_RANGE_ESCAPED 0x801e0000u /* range with at least one escape */ |
243 | #define META_RANGE_LITERAL 0x801f0000u /* range defined literally */ |
244 | #define META_RECURSE 0x80200000u /* Recursion */ |
245 | #define META_RECURSE_BYNAME 0x80210000u /* (?&name) */ |
246 | #define META_SCRIPT_RUN 0x80220000u /* (*script_run:...) */ |
247 | |
248 | /* These must be kept together to make it easy to check that an assertion |
249 | is present where expected in a conditional group. */ |
250 | |
251 | #define META_LOOKAHEAD 0x80230000u /* (?= */ |
252 | #define META_LOOKAHEADNOT 0x80240000u /* (?! */ |
253 | #define META_LOOKBEHIND 0x80250000u /* (?<= */ |
254 | #define META_LOOKBEHINDNOT 0x80260000u /* (?<! */ |
255 | |
256 | /* These cannot be conditions */ |
257 | |
258 | #define META_LOOKAHEAD_NA 0x80270000u /* (*napla: */ |
259 | #define META_LOOKBEHIND_NA 0x80280000u /* (*naplb: */ |
260 | |
261 | /* These must be kept in this order, with consecutive values, and the _ARG |
262 | versions of COMMIT, PRUNE, SKIP, and THEN immediately after their non-argument |
263 | versions. */ |
264 | |
265 | #define META_MARK 0x80290000u /* (*MARK) */ |
266 | #define META_ACCEPT 0x802a0000u /* (*ACCEPT) */ |
267 | #define META_FAIL 0x802b0000u /* (*FAIL) */ |
268 | #define META_COMMIT 0x802c0000u /* These */ |
269 | #define META_COMMIT_ARG 0x802d0000u /* pairs */ |
270 | #define META_PRUNE 0x802e0000u /* must */ |
271 | #define META_PRUNE_ARG 0x802f0000u /* be */ |
272 | #define META_SKIP 0x80300000u /* kept */ |
273 | #define META_SKIP_ARG 0x80310000u /* in */ |
274 | #define META_THEN 0x80320000u /* this */ |
275 | #define META_THEN_ARG 0x80330000u /* order */ |
276 | |
277 | /* These must be kept in groups of adjacent 3 values, and all together. */ |
278 | |
279 | #define META_ASTERISK 0x80340000u /* * */ |
280 | #define META_ASTERISK_PLUS 0x80350000u /* *+ */ |
281 | #define META_ASTERISK_QUERY 0x80360000u /* *? */ |
282 | #define META_PLUS 0x80370000u /* + */ |
283 | #define META_PLUS_PLUS 0x80380000u /* ++ */ |
284 | #define META_PLUS_QUERY 0x80390000u /* +? */ |
285 | #define META_QUERY 0x803a0000u /* ? */ |
286 | #define META_QUERY_PLUS 0x803b0000u /* ?+ */ |
287 | #define META_QUERY_QUERY 0x803c0000u /* ?? */ |
288 | #define META_MINMAX 0x803d0000u /* {n,m} repeat */ |
289 | #define META_MINMAX_PLUS 0x803e0000u /* {n,m}+ repeat */ |
290 | #define META_MINMAX_QUERY 0x803f0000u /* {n,m}? repeat */ |
291 | |
292 | #define META_FIRST_QUANTIFIER META_ASTERISK |
293 | #define META_LAST_QUANTIFIER META_MINMAX_QUERY |
294 | |
295 | /* This is a special "meta code" that is used only to distinguish (*asr: from |
296 | (*sr: in the table of aphabetic assertions. It is never stored in the parsed |
297 | pattern because (*asr: is turned into (*sr:(*atomic: at that stage. There is |
298 | therefore no need for it to have a length entry, so use a high value. */ |
299 | |
300 | #define META_ATOMIC_SCRIPT_RUN 0x8fff0000u |
301 | |
302 | /* Table of extra lengths for each of the meta codes. Must be kept in step with |
303 | the definitions above. For some items these values are a basic length to which |
304 | a variable amount has to be added. */ |
305 | |
306 | static unsigned char [] = { |
307 | 0, /* META_END */ |
308 | 0, /* META_ALT */ |
309 | 0, /* META_ATOMIC */ |
310 | 0, /* META_BACKREF - more if group is >= 10 */ |
311 | 1+SIZEOFFSET, /* META_BACKREF_BYNAME */ |
312 | 1, /* META_BIGVALUE */ |
313 | 3, /* META_CALLOUT_NUMBER */ |
314 | 3+SIZEOFFSET, /* META_CALLOUT_STRING */ |
315 | 0, /* META_CAPTURE */ |
316 | 0, /* META_CIRCUMFLEX */ |
317 | 0, /* META_CLASS */ |
318 | 0, /* META_CLASS_EMPTY */ |
319 | 0, /* META_CLASS_EMPTY_NOT */ |
320 | 0, /* META_CLASS_END */ |
321 | 0, /* META_CLASS_NOT */ |
322 | 0, /* META_COND_ASSERT */ |
323 | SIZEOFFSET, /* META_COND_DEFINE */ |
324 | 1+SIZEOFFSET, /* META_COND_NAME */ |
325 | 1+SIZEOFFSET, /* META_COND_NUMBER */ |
326 | 1+SIZEOFFSET, /* META_COND_RNAME */ |
327 | 1+SIZEOFFSET, /* META_COND_RNUMBER */ |
328 | 3, /* META_COND_VERSION */ |
329 | 0, /* META_DOLLAR */ |
330 | 0, /* META_DOT */ |
331 | 0, /* META_ESCAPE - more for ESC_P, ESC_p, ESC_g, ESC_k */ |
332 | 0, /* META_KET */ |
333 | 0, /* META_NOCAPTURE */ |
334 | 1, /* META_OPTIONS */ |
335 | 1, /* META_POSIX */ |
336 | 1, /* META_POSIX_NEG */ |
337 | 0, /* META_RANGE_ESCAPED */ |
338 | 0, /* META_RANGE_LITERAL */ |
339 | SIZEOFFSET, /* META_RECURSE */ |
340 | 1+SIZEOFFSET, /* META_RECURSE_BYNAME */ |
341 | 0, /* META_SCRIPT_RUN */ |
342 | 0, /* META_LOOKAHEAD */ |
343 | 0, /* META_LOOKAHEADNOT */ |
344 | SIZEOFFSET, /* META_LOOKBEHIND */ |
345 | SIZEOFFSET, /* META_LOOKBEHINDNOT */ |
346 | 0, /* META_LOOKAHEAD_NA */ |
347 | SIZEOFFSET, /* META_LOOKBEHIND_NA */ |
348 | 1, /* META_MARK - plus the string length */ |
349 | 0, /* META_ACCEPT */ |
350 | 0, /* META_FAIL */ |
351 | 0, /* META_COMMIT */ |
352 | 1, /* META_COMMIT_ARG - plus the string length */ |
353 | 0, /* META_PRUNE */ |
354 | 1, /* META_PRUNE_ARG - plus the string length */ |
355 | 0, /* META_SKIP */ |
356 | 1, /* META_SKIP_ARG - plus the string length */ |
357 | 0, /* META_THEN */ |
358 | 1, /* META_THEN_ARG - plus the string length */ |
359 | 0, /* META_ASTERISK */ |
360 | 0, /* META_ASTERISK_PLUS */ |
361 | 0, /* META_ASTERISK_QUERY */ |
362 | 0, /* META_PLUS */ |
363 | 0, /* META_PLUS_PLUS */ |
364 | 0, /* META_PLUS_QUERY */ |
365 | 0, /* META_QUERY */ |
366 | 0, /* META_QUERY_PLUS */ |
367 | 0, /* META_QUERY_QUERY */ |
368 | 2, /* META_MINMAX */ |
369 | 2, /* META_MINMAX_PLUS */ |
370 | 2 /* META_MINMAX_QUERY */ |
371 | }; |
372 | |
373 | /* Types for skipping parts of a parsed pattern. */ |
374 | |
375 | enum { PSKIP_ALT, PSKIP_CLASS, PSKIP_KET }; |
376 | |
377 | /* Macro for setting individual bits in class bitmaps. It took some |
378 | experimenting to figure out how to stop gcc 5.3.0 from warning with |
379 | -Wconversion. This version gets a warning: |
380 | |
381 | #define SETBIT(a,b) a[(b)/8] |= (uint8_t)(1u << ((b)&7)) |
382 | |
383 | Let's hope the apparently less efficient version isn't actually so bad if the |
384 | compiler is clever with identical subexpressions. */ |
385 | |
386 | #define SETBIT(a,b) a[(b)/8] = (uint8_t)(a[(b)/8] | (1u << ((b)&7))) |
387 | |
388 | /* Values and flags for the unsigned xxcuflags variables that accompany xxcu |
389 | variables, which are concerned with first and required code units. A value |
390 | greater than or equal to REQ_NONE means "no code unit set"; otherwise the |
391 | matching xxcu variable is set, and the low valued bits are relevant. */ |
392 | |
393 | #define REQ_UNSET 0xffffffffu /* Not yet found anything */ |
394 | #define REQ_NONE 0xfffffffeu /* Found not fixed character */ |
395 | #define REQ_CASELESS 0x00000001u /* Code unit in xxcu is caseless */ |
396 | #define REQ_VARY 0x00000002u /* Code unit is followed by non-literal */ |
397 | |
398 | /* These flags are used in the groupinfo vector. */ |
399 | |
400 | #define GI_SET_FIXED_LENGTH 0x80000000u |
401 | #define GI_NOT_FIXED_LENGTH 0x40000000u |
402 | #define GI_FIXED_LENGTH_MASK 0x0000ffffu |
403 | |
404 | /* This simple test for a decimal digit works for both ASCII/Unicode and EBCDIC |
405 | and is fast (a good compiler can turn it into a subtraction and unsigned |
406 | comparison). */ |
407 | |
408 | #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9) |
409 | |
410 | /* Table to identify hex digits. The tables in chartables are dependent on the |
411 | locale, and may mark arbitrary characters as digits. We want to recognize only |
412 | 0-9, a-z, and A-Z as hex digits, which is why we have a private table here. It |
413 | costs 256 bytes, but it is a lot faster than doing character value tests (at |
414 | least in some simple cases I timed), and in some applications one wants PCRE2 |
415 | to compile efficiently as well as match efficiently. The value in the table is |
416 | the binary hex digit value, or 0xff for non-hex digits. */ |
417 | |
418 | /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in |
419 | UTF-8 mode. */ |
420 | |
421 | #ifndef EBCDIC |
422 | static const uint8_t xdigitab[] = |
423 | { |
424 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 0- 7 */ |
425 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 8- 15 */ |
426 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 16- 23 */ |
427 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 24- 31 */ |
428 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* - ' */ |
429 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* ( - / */ |
430 | 0x00,0x01,0x02,0x03,0x04,0x05,0x06,0x07, /* 0 - 7 */ |
431 | 0x08,0x09,0xff,0xff,0xff,0xff,0xff,0xff, /* 8 - ? */ |
432 | 0xff,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0xff, /* @ - G */ |
433 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* H - O */ |
434 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* P - W */ |
435 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* X - _ */ |
436 | 0xff,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0xff, /* ` - g */ |
437 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* h - o */ |
438 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* p - w */ |
439 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* x -127 */ |
440 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 128-135 */ |
441 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 136-143 */ |
442 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 144-151 */ |
443 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 152-159 */ |
444 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 160-167 */ |
445 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 168-175 */ |
446 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 176-183 */ |
447 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 184-191 */ |
448 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 192-199 */ |
449 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 2ff-207 */ |
450 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 208-215 */ |
451 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 216-223 */ |
452 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 224-231 */ |
453 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 232-239 */ |
454 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 240-247 */ |
455 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff};/* 248-255 */ |
456 | |
457 | #else |
458 | |
459 | /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */ |
460 | |
461 | static const uint8_t xdigitab[] = |
462 | { |
463 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 0- 7 0 */ |
464 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 8- 15 */ |
465 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 16- 23 10 */ |
466 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 24- 31 */ |
467 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 32- 39 20 */ |
468 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 40- 47 */ |
469 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 48- 55 30 */ |
470 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 56- 63 */ |
471 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* - 71 40 */ |
472 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 72- | */ |
473 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* & - 87 50 */ |
474 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 88- 95 */ |
475 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* - -103 60 */ |
476 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 104- ? */ |
477 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 112-119 70 */ |
478 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 120- " */ |
479 | 0xff,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0xff, /* 128- g 80 */ |
480 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* h -143 */ |
481 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 144- p 90 */ |
482 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* q -159 */ |
483 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 160- x A0 */ |
484 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* y -175 */ |
485 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* ^ -183 B0 */ |
486 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* 184-191 */ |
487 | 0xff,0x0a,0x0b,0x0c,0x0d,0x0e,0x0f,0xff, /* { - G C0 */ |
488 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* H -207 */ |
489 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* } - P D0 */ |
490 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* Q -223 */ |
491 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* \ - X E0 */ |
492 | 0xff,0xff,0xff,0xff,0xff,0xff,0xff,0xff, /* Y -239 */ |
493 | 0x00,0x01,0x02,0x03,0x04,0x05,0x06,0x07, /* 0 - 7 F0 */ |
494 | 0x08,0x09,0xff,0xff,0xff,0xff,0xff,0xff};/* 8 -255 */ |
495 | #endif /* EBCDIC */ |
496 | |
497 | |
498 | /* Table for handling alphanumeric escaped characters. Positive returns are |
499 | simple data values; negative values are for special things like \d and so on. |
500 | Zero means further processing is needed (for things like \x), or the escape is |
501 | invalid. */ |
502 | |
503 | /* This is the "normal" table for ASCII systems or for EBCDIC systems running |
504 | in UTF-8 mode. It runs from '0' to 'z'. */ |
505 | |
506 | #ifndef EBCDIC |
507 | #define ESCAPES_FIRST CHAR_0 |
508 | #define ESCAPES_LAST CHAR_z |
509 | #define UPPER_CASE(c) (c-32) |
510 | |
511 | static const short int escapes[] = { |
512 | 0, 0, |
513 | 0, 0, |
514 | 0, 0, |
515 | 0, 0, |
516 | 0, 0, |
517 | CHAR_COLON, CHAR_SEMICOLON, |
518 | CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, |
519 | CHAR_GREATER_THAN_SIGN, CHAR_QUESTION_MARK, |
520 | CHAR_COMMERCIAL_AT, -ESC_A, |
521 | -ESC_B, -ESC_C, |
522 | -ESC_D, -ESC_E, |
523 | 0, -ESC_G, |
524 | -ESC_H, 0, |
525 | 0, -ESC_K, |
526 | 0, 0, |
527 | -ESC_N, 0, |
528 | -ESC_P, -ESC_Q, |
529 | -ESC_R, -ESC_S, |
530 | 0, 0, |
531 | -ESC_V, -ESC_W, |
532 | -ESC_X, 0, |
533 | -ESC_Z, CHAR_LEFT_SQUARE_BRACKET, |
534 | CHAR_BACKSLASH, CHAR_RIGHT_SQUARE_BRACKET, |
535 | CHAR_CIRCUMFLEX_ACCENT, CHAR_UNDERSCORE, |
536 | CHAR_GRAVE_ACCENT, CHAR_BEL, |
537 | -ESC_b, 0, |
538 | -ESC_d, CHAR_ESC, |
539 | CHAR_FF, 0, |
540 | -ESC_h, 0, |
541 | 0, -ESC_k, |
542 | 0, 0, |
543 | CHAR_LF, 0, |
544 | -ESC_p, 0, |
545 | CHAR_CR, -ESC_s, |
546 | CHAR_HT, 0, |
547 | -ESC_v, -ESC_w, |
548 | 0, 0, |
549 | -ESC_z |
550 | }; |
551 | |
552 | #else |
553 | |
554 | /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. |
555 | It runs from 'a' to '9'. For some minimal testing of EBCDIC features, the code |
556 | is sometimes compiled on an ASCII system. In this case, we must not use CHAR_a |
557 | because it is defined as 'a', which of course picks up the ASCII value. */ |
558 | |
559 | #if 'a' == 0x81 /* Check for a real EBCDIC environment */ |
560 | #define ESCAPES_FIRST CHAR_a |
561 | #define ESCAPES_LAST CHAR_9 |
562 | #define UPPER_CASE(c) (c+64) |
563 | #else /* Testing in an ASCII environment */ |
564 | #define ESCAPES_FIRST ((unsigned char)'\x81') /* EBCDIC 'a' */ |
565 | #define ESCAPES_LAST ((unsigned char)'\xf9') /* EBCDIC '9' */ |
566 | #define UPPER_CASE(c) (c-32) |
567 | #endif |
568 | |
569 | static const short int escapes[] = { |
570 | /* 80 */ CHAR_BEL, -ESC_b, 0, -ESC_d, CHAR_ESC, CHAR_FF, 0, |
571 | /* 88 */ -ESC_h, 0, 0, '{', 0, 0, 0, 0, |
572 | /* 90 */ 0, 0, -ESC_k, 0, 0, CHAR_LF, 0, -ESC_p, |
573 | /* 98 */ 0, CHAR_CR, 0, '}', 0, 0, 0, 0, |
574 | /* A0 */ 0, '~', -ESC_s, CHAR_HT, 0, -ESC_v, -ESC_w, 0, |
575 | /* A8 */ 0, -ESC_z, 0, 0, 0, '[', 0, 0, |
576 | /* B0 */ 0, 0, 0, 0, 0, 0, 0, 0, |
577 | /* B8 */ 0, 0, 0, 0, 0, ']', '=', '-', |
578 | /* C0 */ '{', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E, 0, -ESC_G, |
579 | /* C8 */ -ESC_H, 0, 0, 0, 0, 0, 0, 0, |
580 | /* D0 */ '}', 0, -ESC_K, 0, 0, -ESC_N, 0, -ESC_P, |
581 | /* D8 */ -ESC_Q, -ESC_R, 0, 0, 0, 0, 0, 0, |
582 | /* E0 */ '\\', 0, -ESC_S, 0, 0, -ESC_V, -ESC_W, -ESC_X, |
583 | /* E8 */ 0, -ESC_Z, 0, 0, 0, 0, 0, 0, |
584 | /* F0 */ 0, 0, 0, 0, 0, 0, 0, 0, |
585 | /* F8 */ 0, 0 |
586 | }; |
587 | |
588 | /* We also need a table of characters that may follow \c in an EBCDIC |
589 | environment for characters 0-31. */ |
590 | |
591 | static unsigned char ebcdic_escape_c[] = "@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\\]^_" ; |
592 | |
593 | #endif /* EBCDIC */ |
594 | |
595 | |
596 | /* Table of special "verbs" like (*PRUNE). This is a short table, so it is |
597 | searched linearly. Put all the names into a single string, in order to reduce |
598 | the number of relocations when a shared library is dynamically linked. The |
599 | string is built from string macros so that it works in UTF-8 mode on EBCDIC |
600 | platforms. */ |
601 | |
602 | typedef struct verbitem { |
603 | unsigned int len; /* Length of verb name */ |
604 | uint32_t meta; /* Base META_ code */ |
605 | int has_arg; /* Argument requirement */ |
606 | } verbitem; |
607 | |
608 | static const char verbnames[] = |
609 | "\0" /* Empty name is a shorthand for MARK */ |
610 | STRING_MARK0 |
611 | STRING_ACCEPT0 |
612 | STRING_F0 |
613 | STRING_FAIL0 |
614 | STRING_COMMIT0 |
615 | STRING_PRUNE0 |
616 | STRING_SKIP0 |
617 | STRING_THEN; |
618 | |
619 | static const verbitem verbs[] = { |
620 | { 0, META_MARK, +1 }, /* > 0 => must have an argument */ |
621 | { 4, META_MARK, +1 }, |
622 | { 6, META_ACCEPT, -1 }, /* < 0 => Optional argument, convert to pre-MARK */ |
623 | { 1, META_FAIL, -1 }, |
624 | { 4, META_FAIL, -1 }, |
625 | { 6, META_COMMIT, 0 }, |
626 | { 5, META_PRUNE, 0 }, /* Optional argument; bump META code if found */ |
627 | { 4, META_SKIP, 0 }, |
628 | { 4, META_THEN, 0 } |
629 | }; |
630 | |
631 | static const int verbcount = sizeof(verbs)/sizeof(verbitem); |
632 | |
633 | /* Verb opcodes, indexed by their META code offset from META_MARK. */ |
634 | |
635 | static const uint32_t verbops[] = { |
636 | OP_MARK, OP_ACCEPT, OP_FAIL, OP_COMMIT, OP_COMMIT_ARG, OP_PRUNE, |
637 | OP_PRUNE_ARG, OP_SKIP, OP_SKIP_ARG, OP_THEN, OP_THEN_ARG }; |
638 | |
639 | /* Table of "alpha assertions" like (*pla:...), similar to the (*VERB) table. */ |
640 | |
641 | typedef struct alasitem { |
642 | unsigned int len; /* Length of name */ |
643 | uint32_t meta; /* Base META_ code */ |
644 | } alasitem; |
645 | |
646 | static const char alasnames[] = |
647 | STRING_pla0 |
648 | STRING_plb0 |
649 | STRING_napla0 |
650 | STRING_naplb0 |
651 | STRING_nla0 |
652 | STRING_nlb0 |
653 | STRING_positive_lookahead0 |
654 | STRING_positive_lookbehind0 |
655 | STRING_non_atomic_positive_lookahead0 |
656 | STRING_non_atomic_positive_lookbehind0 |
657 | STRING_negative_lookahead0 |
658 | STRING_negative_lookbehind0 |
659 | STRING_atomic0 |
660 | STRING_sr0 |
661 | STRING_asr0 |
662 | STRING_script_run0 |
663 | STRING_atomic_script_run; |
664 | |
665 | static const alasitem alasmeta[] = { |
666 | { 3, META_LOOKAHEAD }, |
667 | { 3, META_LOOKBEHIND }, |
668 | { 5, META_LOOKAHEAD_NA }, |
669 | { 5, META_LOOKBEHIND_NA }, |
670 | { 3, META_LOOKAHEADNOT }, |
671 | { 3, META_LOOKBEHINDNOT }, |
672 | { 18, META_LOOKAHEAD }, |
673 | { 19, META_LOOKBEHIND }, |
674 | { 29, META_LOOKAHEAD_NA }, |
675 | { 30, META_LOOKBEHIND_NA }, |
676 | { 18, META_LOOKAHEADNOT }, |
677 | { 19, META_LOOKBEHINDNOT }, |
678 | { 6, META_ATOMIC }, |
679 | { 2, META_SCRIPT_RUN }, /* sr = script run */ |
680 | { 3, META_ATOMIC_SCRIPT_RUN }, /* asr = atomic script run */ |
681 | { 10, META_SCRIPT_RUN }, /* script run */ |
682 | { 17, META_ATOMIC_SCRIPT_RUN } /* atomic script run */ |
683 | }; |
684 | |
685 | static const int alascount = sizeof(alasmeta)/sizeof(alasitem); |
686 | |
687 | /* Offsets from OP_STAR for case-independent and negative repeat opcodes. */ |
688 | |
689 | static uint32_t chartypeoffset[] = { |
690 | OP_STAR - OP_STAR, OP_STARI - OP_STAR, |
691 | OP_NOTSTAR - OP_STAR, OP_NOTSTARI - OP_STAR }; |
692 | |
693 | /* Tables of names of POSIX character classes and their lengths. The names are |
694 | now all in a single string, to reduce the number of relocations when a shared |
695 | library is dynamically loaded. The list of lengths is terminated by a zero |
696 | length entry. The first three must be alpha, lower, upper, as this is assumed |
697 | for handling case independence. The indices for graph, print, and punct are |
698 | needed, so identify them. */ |
699 | |
700 | static const char posix_names[] = |
701 | STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0 |
702 | STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0 |
703 | STRING_graph0 STRING_print0 STRING_punct0 STRING_space0 |
704 | STRING_word0 STRING_xdigit; |
705 | |
706 | static const uint8_t posix_name_lengths[] = { |
707 | 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 }; |
708 | |
709 | #define PC_GRAPH 8 |
710 | #define PC_PRINT 9 |
711 | #define PC_PUNCT 10 |
712 | |
713 | /* Table of class bit maps for each POSIX class. Each class is formed from a |
714 | base map, with an optional addition or removal of another map. Then, for some |
715 | classes, there is some additional tweaking: for [:blank:] the vertical space |
716 | characters are removed, and for [:alpha:] and [:alnum:] the underscore |
717 | character is removed. The triples in the table consist of the base map offset, |
718 | second map offset or -1 if no second map, and a non-negative value for map |
719 | addition or a negative value for map subtraction (if there are two maps). The |
720 | absolute value of the third field has these meanings: 0 => no tweaking, 1 => |
721 | remove vertical space characters, 2 => remove underscore. */ |
722 | |
723 | static const int posix_class_maps[] = { |
724 | cbit_word, cbit_digit, -2, /* alpha */ |
725 | cbit_lower, -1, 0, /* lower */ |
726 | cbit_upper, -1, 0, /* upper */ |
727 | cbit_word, -1, 2, /* alnum - word without underscore */ |
728 | cbit_print, cbit_cntrl, 0, /* ascii */ |
729 | cbit_space, -1, 1, /* blank - a GNU extension */ |
730 | cbit_cntrl, -1, 0, /* cntrl */ |
731 | cbit_digit, -1, 0, /* digit */ |
732 | cbit_graph, -1, 0, /* graph */ |
733 | cbit_print, -1, 0, /* print */ |
734 | cbit_punct, -1, 0, /* punct */ |
735 | cbit_space, -1, 0, /* space */ |
736 | cbit_word, -1, 0, /* word - a Perl extension */ |
737 | cbit_xdigit,-1, 0 /* xdigit */ |
738 | }; |
739 | |
740 | #ifdef SUPPORT_UNICODE |
741 | |
742 | /* The POSIX class Unicode property substitutes that are used in UCP mode must |
743 | be in the order of the POSIX class names, defined above. */ |
744 | |
745 | static int posix_substitutes[] = { |
746 | PT_GC, ucp_L, /* alpha */ |
747 | PT_PC, ucp_Ll, /* lower */ |
748 | PT_PC, ucp_Lu, /* upper */ |
749 | PT_ALNUM, 0, /* alnum */ |
750 | -1, 0, /* ascii, treat as non-UCP */ |
751 | -1, 1, /* blank, treat as \h */ |
752 | PT_PC, ucp_Cc, /* cntrl */ |
753 | PT_PC, ucp_Nd, /* digit */ |
754 | PT_PXGRAPH, 0, /* graph */ |
755 | PT_PXPRINT, 0, /* print */ |
756 | PT_PXPUNCT, 0, /* punct */ |
757 | PT_PXSPACE, 0, /* space */ /* Xps is POSIX space, but from 8.34 */ |
758 | PT_WORD, 0, /* word */ /* Perl and POSIX space are the same */ |
759 | -1, 0 /* xdigit, treat as non-UCP */ |
760 | }; |
761 | #define POSIX_SUBSIZE (sizeof(posix_substitutes) / (2*sizeof(uint32_t))) |
762 | #endif /* SUPPORT_UNICODE */ |
763 | |
764 | /* Masks for checking option settings. When PCRE2_LITERAL is set, only a subset |
765 | are allowed. */ |
766 | |
767 | #define PUBLIC_LITERAL_COMPILE_OPTIONS \ |
768 | (PCRE2_ANCHORED|PCRE2_AUTO_CALLOUT|PCRE2_CASELESS|PCRE2_ENDANCHORED| \ |
769 | PCRE2_FIRSTLINE|PCRE2_LITERAL|PCRE2_MATCH_INVALID_UTF| \ |
770 | PCRE2_NO_START_OPTIMIZE|PCRE2_NO_UTF_CHECK|PCRE2_USE_OFFSET_LIMIT|PCRE2_UTF) |
771 | |
772 | #define PUBLIC_COMPILE_OPTIONS \ |
773 | (PUBLIC_LITERAL_COMPILE_OPTIONS| \ |
774 | PCRE2_ALLOW_EMPTY_CLASS|PCRE2_ALT_BSUX|PCRE2_ALT_CIRCUMFLEX| \ |
775 | PCRE2_ALT_VERBNAMES|PCRE2_DOLLAR_ENDONLY|PCRE2_DOTALL|PCRE2_DUPNAMES| \ |
776 | PCRE2_EXTENDED|PCRE2_EXTENDED_MORE|PCRE2_MATCH_UNSET_BACKREF| \ |
777 | PCRE2_MULTILINE|PCRE2_NEVER_BACKSLASH_C|PCRE2_NEVER_UCP| \ |
778 | PCRE2_NEVER_UTF|PCRE2_NO_AUTO_CAPTURE|PCRE2_NO_AUTO_POSSESS| \ |
779 | PCRE2_NO_DOTSTAR_ANCHOR|PCRE2_UCP|PCRE2_UNGREEDY) |
780 | |
781 | #define \ |
782 | (PCRE2_EXTRA_MATCH_LINE|PCRE2_EXTRA_MATCH_WORD) |
783 | |
784 | #define \ |
785 | (PUBLIC_LITERAL_COMPILE_EXTRA_OPTIONS| \ |
786 | PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES|PCRE2_EXTRA_BAD_ESCAPE_IS_LITERAL| \ |
787 | PCRE2_EXTRA_ESCAPED_CR_IS_LF|PCRE2_EXTRA_ALT_BSUX| \ |
788 | PCRE2_EXTRA_ALLOW_LOOKAROUND_BSK) |
789 | |
790 | /* Compile time error code numbers. They are given names so that they can more |
791 | easily be tracked. When a new number is added, the tables called eint1 and |
792 | eint2 in pcre2posix.c may need to be updated, and a new error text must be |
793 | added to compile_error_texts in pcre2_error.c. Also, the error codes in |
794 | pcre2.h.in must be updated - their values are exactly 100 greater than these |
795 | values. */ |
796 | |
797 | enum { ERR0 = COMPILE_ERROR_BASE, |
798 | ERR1, ERR2, ERR3, ERR4, ERR5, ERR6, ERR7, ERR8, ERR9, ERR10, |
799 | ERR11, ERR12, ERR13, ERR14, ERR15, ERR16, ERR17, ERR18, ERR19, ERR20, |
800 | ERR21, ERR22, ERR23, ERR24, ERR25, ERR26, ERR27, ERR28, ERR29, ERR30, |
801 | ERR31, ERR32, ERR33, ERR34, ERR35, ERR36, ERR37, ERR38, ERR39, ERR40, |
802 | ERR41, ERR42, ERR43, ERR44, ERR45, ERR46, ERR47, ERR48, ERR49, ERR50, |
803 | ERR51, ERR52, ERR53, ERR54, ERR55, ERR56, ERR57, ERR58, ERR59, ERR60, |
804 | ERR61, ERR62, ERR63, ERR64, ERR65, ERR66, ERR67, ERR68, ERR69, ERR70, |
805 | ERR71, ERR72, ERR73, ERR74, ERR75, ERR76, ERR77, ERR78, ERR79, ERR80, |
806 | ERR81, ERR82, ERR83, ERR84, ERR85, ERR86, ERR87, ERR88, ERR89, ERR90, |
807 | ERR91, ERR92, ERR93, ERR94, ERR95, ERR96, ERR97, ERR98, ERR99 }; |
808 | |
809 | /* This is a table of start-of-pattern options such as (*UTF) and settings such |
810 | as (*LIMIT_MATCH=nnnn) and (*CRLF). For completeness and backward |
811 | compatibility, (*UTFn) is supported in the relevant libraries, but (*UTF) is |
812 | generic and always supported. */ |
813 | |
814 | enum { PSO_OPT, /* Value is an option bit */ |
815 | PSO_FLG, /* Value is a flag bit */ |
816 | PSO_NL, /* Value is a newline type */ |
817 | PSO_BSR, /* Value is a \R type */ |
818 | PSO_LIMH, /* Read integer value for heap limit */ |
819 | PSO_LIMM, /* Read integer value for match limit */ |
820 | PSO_LIMD }; /* Read integer value for depth limit */ |
821 | |
822 | typedef struct pso { |
823 | const uint8_t *name; |
824 | uint16_t length; |
825 | uint16_t type; |
826 | uint32_t value; |
827 | } pso; |
828 | |
829 | /* NB: STRING_UTFn_RIGHTPAR contains the length as well */ |
830 | |
831 | static pso pso_list[] = { |
832 | { (uint8_t *)STRING_UTFn_RIGHTPAR, PSO_OPT, PCRE2_UTF }, |
833 | { (uint8_t *)STRING_UTF_RIGHTPAR, 4, PSO_OPT, PCRE2_UTF }, |
834 | { (uint8_t *)STRING_UCP_RIGHTPAR, 4, PSO_OPT, PCRE2_UCP }, |
835 | { (uint8_t *)STRING_NOTEMPTY_RIGHTPAR, 9, PSO_FLG, PCRE2_NOTEMPTY_SET }, |
836 | { (uint8_t *)STRING_NOTEMPTY_ATSTART_RIGHTPAR, 17, PSO_FLG, PCRE2_NE_ATST_SET }, |
837 | { (uint8_t *)STRING_NO_AUTO_POSSESS_RIGHTPAR, 16, PSO_OPT, PCRE2_NO_AUTO_POSSESS }, |
838 | { (uint8_t *)STRING_NO_DOTSTAR_ANCHOR_RIGHTPAR, 18, PSO_OPT, PCRE2_NO_DOTSTAR_ANCHOR }, |
839 | { (uint8_t *)STRING_NO_JIT_RIGHTPAR, 7, PSO_FLG, PCRE2_NOJIT }, |
840 | { (uint8_t *)STRING_NO_START_OPT_RIGHTPAR, 13, PSO_OPT, PCRE2_NO_START_OPTIMIZE }, |
841 | { (uint8_t *)STRING_LIMIT_HEAP_EQ, 11, PSO_LIMH, 0 }, |
842 | { (uint8_t *)STRING_LIMIT_MATCH_EQ, 12, PSO_LIMM, 0 }, |
843 | { (uint8_t *)STRING_LIMIT_DEPTH_EQ, 12, PSO_LIMD, 0 }, |
844 | { (uint8_t *)STRING_LIMIT_RECURSION_EQ, 16, PSO_LIMD, 0 }, |
845 | { (uint8_t *)STRING_CR_RIGHTPAR, 3, PSO_NL, PCRE2_NEWLINE_CR }, |
846 | { (uint8_t *)STRING_LF_RIGHTPAR, 3, PSO_NL, PCRE2_NEWLINE_LF }, |
847 | { (uint8_t *)STRING_CRLF_RIGHTPAR, 5, PSO_NL, PCRE2_NEWLINE_CRLF }, |
848 | { (uint8_t *)STRING_ANY_RIGHTPAR, 4, PSO_NL, PCRE2_NEWLINE_ANY }, |
849 | { (uint8_t *)STRING_NUL_RIGHTPAR, 4, PSO_NL, PCRE2_NEWLINE_NUL }, |
850 | { (uint8_t *)STRING_ANYCRLF_RIGHTPAR, 8, PSO_NL, PCRE2_NEWLINE_ANYCRLF }, |
851 | { (uint8_t *)STRING_BSR_ANYCRLF_RIGHTPAR, 12, PSO_BSR, PCRE2_BSR_ANYCRLF }, |
852 | { (uint8_t *)STRING_BSR_UNICODE_RIGHTPAR, 12, PSO_BSR, PCRE2_BSR_UNICODE } |
853 | }; |
854 | |
855 | /* This table is used when converting repeating opcodes into possessified |
856 | versions as a result of an explicit possessive quantifier such as ++. A zero |
857 | value means there is no possessified version - in those cases the item in |
858 | question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT |
859 | because all relevant opcodes are less than that. */ |
860 | |
861 | static const uint8_t opcode_possessify[] = { |
862 | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */ |
863 | 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */ |
864 | |
865 | 0, /* NOTI */ |
866 | OP_POSSTAR, 0, /* STAR, MINSTAR */ |
867 | OP_POSPLUS, 0, /* PLUS, MINPLUS */ |
868 | OP_POSQUERY, 0, /* QUERY, MINQUERY */ |
869 | OP_POSUPTO, 0, /* UPTO, MINUPTO */ |
870 | 0, /* EXACT */ |
871 | 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */ |
872 | |
873 | OP_POSSTARI, 0, /* STARI, MINSTARI */ |
874 | OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */ |
875 | OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */ |
876 | OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */ |
877 | 0, /* EXACTI */ |
878 | 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */ |
879 | |
880 | OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */ |
881 | OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */ |
882 | OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */ |
883 | OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */ |
884 | 0, /* NOTEXACT */ |
885 | 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */ |
886 | |
887 | OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */ |
888 | OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */ |
889 | OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */ |
890 | OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */ |
891 | 0, /* NOTEXACTI */ |
892 | 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */ |
893 | |
894 | OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */ |
895 | OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */ |
896 | OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */ |
897 | OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */ |
898 | 0, /* TYPEEXACT */ |
899 | 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */ |
900 | |
901 | OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */ |
902 | OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */ |
903 | OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */ |
904 | OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */ |
905 | 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */ |
906 | |
907 | 0, 0, 0, /* CLASS, NCLASS, XCLASS */ |
908 | 0, 0, /* REF, REFI */ |
909 | 0, 0, /* DNREF, DNREFI */ |
910 | 0, 0 /* RECURSE, CALLOUT */ |
911 | }; |
912 | |
913 | |
914 | #ifdef DEBUG_SHOW_PARSED |
915 | /************************************************* |
916 | * Show the parsed pattern for debugging * |
917 | *************************************************/ |
918 | |
919 | /* For debugging the pre-scan, this code, which outputs the parsed data vector, |
920 | can be enabled. */ |
921 | |
922 | static void show_parsed(compile_block *cb) |
923 | { |
924 | uint32_t *pptr = cb->parsed_pattern; |
925 | |
926 | for (;;) |
927 | { |
928 | int max, min; |
929 | PCRE2_SIZE offset; |
930 | uint32_t i; |
931 | uint32_t length; |
932 | uint32_t meta_arg = META_DATA(*pptr); |
933 | |
934 | fprintf(stderr, "+++ %02d %.8x " , (int)(pptr - cb->parsed_pattern), *pptr); |
935 | |
936 | if (*pptr < META_END) |
937 | { |
938 | if (*pptr > 32 && *pptr < 128) fprintf(stderr, "%c" , *pptr); |
939 | pptr++; |
940 | } |
941 | |
942 | else switch (META_CODE(*pptr++)) |
943 | { |
944 | default: |
945 | fprintf(stderr, "**** OOPS - unknown META value - giving up ****\n" ); |
946 | return; |
947 | |
948 | case META_END: |
949 | fprintf(stderr, "META_END\n" ); |
950 | return; |
951 | |
952 | case META_CAPTURE: |
953 | fprintf(stderr, "META_CAPTURE %d" , meta_arg); |
954 | break; |
955 | |
956 | case META_RECURSE: |
957 | GETOFFSET(offset, pptr); |
958 | fprintf(stderr, "META_RECURSE %d %zd" , meta_arg, offset); |
959 | break; |
960 | |
961 | case META_BACKREF: |
962 | if (meta_arg < 10) |
963 | offset = cb->small_ref_offset[meta_arg]; |
964 | else |
965 | GETOFFSET(offset, pptr); |
966 | fprintf(stderr, "META_BACKREF %d %zd" , meta_arg, offset); |
967 | break; |
968 | |
969 | case META_ESCAPE: |
970 | if (meta_arg == ESC_P || meta_arg == ESC_p) |
971 | { |
972 | uint32_t ptype = *pptr >> 16; |
973 | uint32_t pvalue = *pptr++ & 0xffff; |
974 | fprintf(stderr, "META \\%c %d %d" , (meta_arg == ESC_P)? 'P':'p', |
975 | ptype, pvalue); |
976 | } |
977 | else |
978 | { |
979 | uint32_t cc; |
980 | /* There's just one escape we might have here that isn't negated in the |
981 | escapes table. */ |
982 | if (meta_arg == ESC_g) cc = CHAR_g; |
983 | else for (cc = ESCAPES_FIRST; cc <= ESCAPES_LAST; cc++) |
984 | { |
985 | if (meta_arg == (uint32_t)(-escapes[cc - ESCAPES_FIRST])) break; |
986 | } |
987 | if (cc > ESCAPES_LAST) cc = CHAR_QUESTION_MARK; |
988 | fprintf(stderr, "META \\%c" , cc); |
989 | } |
990 | break; |
991 | |
992 | case META_MINMAX: |
993 | min = *pptr++; |
994 | max = *pptr++; |
995 | if (max != REPEAT_UNLIMITED) |
996 | fprintf(stderr, "META {%d,%d}" , min, max); |
997 | else |
998 | fprintf(stderr, "META {%d,}" , min); |
999 | break; |
1000 | |
1001 | case META_MINMAX_QUERY: |
1002 | min = *pptr++; |
1003 | max = *pptr++; |
1004 | if (max != REPEAT_UNLIMITED) |
1005 | fprintf(stderr, "META {%d,%d}?" , min, max); |
1006 | else |
1007 | fprintf(stderr, "META {%d,}?" , min); |
1008 | break; |
1009 | |
1010 | case META_MINMAX_PLUS: |
1011 | min = *pptr++; |
1012 | max = *pptr++; |
1013 | if (max != REPEAT_UNLIMITED) |
1014 | fprintf(stderr, "META {%d,%d}+" , min, max); |
1015 | else |
1016 | fprintf(stderr, "META {%d,}+" , min); |
1017 | break; |
1018 | |
1019 | case META_BIGVALUE: fprintf(stderr, "META_BIGVALUE %.8x" , *pptr++); break; |
1020 | case META_CIRCUMFLEX: fprintf(stderr, "META_CIRCUMFLEX" ); break; |
1021 | case META_COND_ASSERT: fprintf(stderr, "META_COND_ASSERT" ); break; |
1022 | case META_DOLLAR: fprintf(stderr, "META_DOLLAR" ); break; |
1023 | case META_DOT: fprintf(stderr, "META_DOT" ); break; |
1024 | case META_ASTERISK: fprintf(stderr, "META *" ); break; |
1025 | case META_ASTERISK_QUERY: fprintf(stderr, "META *?" ); break; |
1026 | case META_ASTERISK_PLUS: fprintf(stderr, "META *+" ); break; |
1027 | case META_PLUS: fprintf(stderr, "META +" ); break; |
1028 | case META_PLUS_QUERY: fprintf(stderr, "META +?" ); break; |
1029 | case META_PLUS_PLUS: fprintf(stderr, "META ++" ); break; |
1030 | case META_QUERY: fprintf(stderr, "META ?" ); break; |
1031 | case META_QUERY_QUERY: fprintf(stderr, "META ??" ); break; |
1032 | case META_QUERY_PLUS: fprintf(stderr, "META ?+" ); break; |
1033 | |
1034 | case META_ATOMIC: fprintf(stderr, "META (?>" ); break; |
1035 | case META_NOCAPTURE: fprintf(stderr, "META (?:" ); break; |
1036 | case META_LOOKAHEAD: fprintf(stderr, "META (?=" ); break; |
1037 | case META_LOOKAHEADNOT: fprintf(stderr, "META (?!" ); break; |
1038 | case META_LOOKAHEAD_NA: fprintf(stderr, "META (*napla:" ); break; |
1039 | case META_SCRIPT_RUN: fprintf(stderr, "META (*sr:" ); break; |
1040 | case META_KET: fprintf(stderr, "META )" ); break; |
1041 | case META_ALT: fprintf(stderr, "META | %d" , meta_arg); break; |
1042 | |
1043 | case META_CLASS: fprintf(stderr, "META [" ); break; |
1044 | case META_CLASS_NOT: fprintf(stderr, "META [^" ); break; |
1045 | case META_CLASS_END: fprintf(stderr, "META ]" ); break; |
1046 | case META_CLASS_EMPTY: fprintf(stderr, "META []" ); break; |
1047 | case META_CLASS_EMPTY_NOT: fprintf(stderr, "META [^]" ); break; |
1048 | |
1049 | case META_RANGE_LITERAL: fprintf(stderr, "META - (literal)" ); break; |
1050 | case META_RANGE_ESCAPED: fprintf(stderr, "META - (escaped)" ); break; |
1051 | |
1052 | case META_POSIX: fprintf(stderr, "META_POSIX %d" , *pptr++); break; |
1053 | case META_POSIX_NEG: fprintf(stderr, "META_POSIX_NEG %d" , *pptr++); break; |
1054 | |
1055 | case META_ACCEPT: fprintf(stderr, "META (*ACCEPT)" ); break; |
1056 | case META_FAIL: fprintf(stderr, "META (*FAIL)" ); break; |
1057 | case META_COMMIT: fprintf(stderr, "META (*COMMIT)" ); break; |
1058 | case META_PRUNE: fprintf(stderr, "META (*PRUNE)" ); break; |
1059 | case META_SKIP: fprintf(stderr, "META (*SKIP)" ); break; |
1060 | case META_THEN: fprintf(stderr, "META (*THEN)" ); break; |
1061 | |
1062 | case META_OPTIONS: fprintf(stderr, "META_OPTIONS 0x%02x" , *pptr++); break; |
1063 | |
1064 | case META_LOOKBEHIND: |
1065 | fprintf(stderr, "META (?<= %d offset=" , meta_arg); |
1066 | GETOFFSET(offset, pptr); |
1067 | fprintf(stderr, "%zd" , offset); |
1068 | break; |
1069 | |
1070 | case META_LOOKBEHIND_NA: |
1071 | fprintf(stderr, "META (*naplb: %d offset=" , meta_arg); |
1072 | GETOFFSET(offset, pptr); |
1073 | fprintf(stderr, "%zd" , offset); |
1074 | break; |
1075 | |
1076 | case META_LOOKBEHINDNOT: |
1077 | fprintf(stderr, "META (?<! %d offset=" , meta_arg); |
1078 | GETOFFSET(offset, pptr); |
1079 | fprintf(stderr, "%zd" , offset); |
1080 | break; |
1081 | |
1082 | case META_CALLOUT_NUMBER: |
1083 | fprintf(stderr, "META (?C%d) next=%d/%d" , pptr[2], pptr[0], |
1084 | pptr[1]); |
1085 | pptr += 3; |
1086 | break; |
1087 | |
1088 | case META_CALLOUT_STRING: |
1089 | { |
1090 | uint32_t patoffset = *pptr++; /* Offset of next pattern item */ |
1091 | uint32_t patlength = *pptr++; /* Length of next pattern item */ |
1092 | fprintf(stderr, "META (?Cstring) length=%d offset=" , *pptr++); |
1093 | GETOFFSET(offset, pptr); |
1094 | fprintf(stderr, "%zd next=%d/%d" , offset, patoffset, patlength); |
1095 | } |
1096 | break; |
1097 | |
1098 | case META_RECURSE_BYNAME: |
1099 | fprintf(stderr, "META (?(&name) length=%d offset=" , *pptr++); |
1100 | GETOFFSET(offset, pptr); |
1101 | fprintf(stderr, "%zd" , offset); |
1102 | break; |
1103 | |
1104 | case META_BACKREF_BYNAME: |
1105 | fprintf(stderr, "META_BACKREF_BYNAME length=%d offset=" , *pptr++); |
1106 | GETOFFSET(offset, pptr); |
1107 | fprintf(stderr, "%zd" , offset); |
1108 | break; |
1109 | |
1110 | case META_COND_NUMBER: |
1111 | fprintf(stderr, "META_COND_NUMBER %d offset=" , pptr[SIZEOFFSET]); |
1112 | GETOFFSET(offset, pptr); |
1113 | fprintf(stderr, "%zd" , offset); |
1114 | pptr++; |
1115 | break; |
1116 | |
1117 | case META_COND_DEFINE: |
1118 | fprintf(stderr, "META (?(DEFINE) offset=" ); |
1119 | GETOFFSET(offset, pptr); |
1120 | fprintf(stderr, "%zd" , offset); |
1121 | break; |
1122 | |
1123 | case META_COND_VERSION: |
1124 | fprintf(stderr, "META (?(VERSION%s" , (*pptr++ == 0)? "=" : ">=" ); |
1125 | fprintf(stderr, "%d." , *pptr++); |
1126 | fprintf(stderr, "%d)" , *pptr++); |
1127 | break; |
1128 | |
1129 | case META_COND_NAME: |
1130 | fprintf(stderr, "META (?(<name>) length=%d offset=" , *pptr++); |
1131 | GETOFFSET(offset, pptr); |
1132 | fprintf(stderr, "%zd" , offset); |
1133 | break; |
1134 | |
1135 | case META_COND_RNAME: |
1136 | fprintf(stderr, "META (?(R&name) length=%d offset=" , *pptr++); |
1137 | GETOFFSET(offset, pptr); |
1138 | fprintf(stderr, "%zd" , offset); |
1139 | break; |
1140 | |
1141 | /* This is kept as a name, because it might be. */ |
1142 | |
1143 | case META_COND_RNUMBER: |
1144 | fprintf(stderr, "META (?(Rnumber) length=%d offset=" , *pptr++); |
1145 | GETOFFSET(offset, pptr); |
1146 | fprintf(stderr, "%zd" , offset); |
1147 | break; |
1148 | |
1149 | case META_MARK: |
1150 | fprintf(stderr, "META (*MARK:" ); |
1151 | goto SHOWARG; |
1152 | |
1153 | case META_COMMIT_ARG: |
1154 | fprintf(stderr, "META (*COMMIT:" ); |
1155 | goto SHOWARG; |
1156 | |
1157 | case META_PRUNE_ARG: |
1158 | fprintf(stderr, "META (*PRUNE:" ); |
1159 | goto SHOWARG; |
1160 | |
1161 | case META_SKIP_ARG: |
1162 | fprintf(stderr, "META (*SKIP:" ); |
1163 | goto SHOWARG; |
1164 | |
1165 | case META_THEN_ARG: |
1166 | fprintf(stderr, "META (*THEN:" ); |
1167 | SHOWARG: |
1168 | length = *pptr++; |
1169 | for (i = 0; i < length; i++) |
1170 | { |
1171 | uint32_t cc = *pptr++; |
1172 | if (cc > 32 && cc < 128) fprintf(stderr, "%c" , cc); |
1173 | else fprintf(stderr, "\\x{%x}" , cc); |
1174 | } |
1175 | fprintf(stderr, ") length=%u" , length); |
1176 | break; |
1177 | } |
1178 | fprintf(stderr, "\n" ); |
1179 | } |
1180 | return; |
1181 | } |
1182 | #endif /* DEBUG_SHOW_PARSED */ |
1183 | |
1184 | |
1185 | |
1186 | /************************************************* |
1187 | * Copy compiled code * |
1188 | *************************************************/ |
1189 | |
1190 | /* Compiled JIT code cannot be copied, so the new compiled block has no |
1191 | associated JIT data. */ |
1192 | |
1193 | PCRE2_EXP_DEFN pcre2_code * PCRE2_CALL_CONVENTION |
1194 | pcre2_code_copy(const pcre2_code *code) |
1195 | { |
1196 | PCRE2_SIZE* ref_count; |
1197 | pcre2_code *newcode; |
1198 | |
1199 | if (code == NULL) return NULL; |
1200 | newcode = code->memctl.malloc(code->blocksize, code->memctl.memory_data); |
1201 | if (newcode == NULL) return NULL; |
1202 | memcpy(newcode, code, code->blocksize); |
1203 | newcode->executable_jit = NULL; |
1204 | |
1205 | /* If the code is one that has been deserialized, increment the reference count |
1206 | in the decoded tables. */ |
1207 | |
1208 | if ((code->flags & PCRE2_DEREF_TABLES) != 0) |
1209 | { |
1210 | ref_count = (PCRE2_SIZE *)(code->tables + TABLES_LENGTH); |
1211 | (*ref_count)++; |
1212 | } |
1213 | |
1214 | return newcode; |
1215 | } |
1216 | |
1217 | |
1218 | |
1219 | /************************************************* |
1220 | * Copy compiled code and character tables * |
1221 | *************************************************/ |
1222 | |
1223 | /* Compiled JIT code cannot be copied, so the new compiled block has no |
1224 | associated JIT data. This version of code_copy also makes a separate copy of |
1225 | the character tables. */ |
1226 | |
1227 | PCRE2_EXP_DEFN pcre2_code * PCRE2_CALL_CONVENTION |
1228 | pcre2_code_copy_with_tables(const pcre2_code *code) |
1229 | { |
1230 | PCRE2_SIZE* ref_count; |
1231 | pcre2_code *newcode; |
1232 | uint8_t *newtables; |
1233 | |
1234 | if (code == NULL) return NULL; |
1235 | newcode = code->memctl.malloc(code->blocksize, code->memctl.memory_data); |
1236 | if (newcode == NULL) return NULL; |
1237 | memcpy(newcode, code, code->blocksize); |
1238 | newcode->executable_jit = NULL; |
1239 | |
1240 | newtables = code->memctl.malloc(TABLES_LENGTH + sizeof(PCRE2_SIZE), |
1241 | code->memctl.memory_data); |
1242 | if (newtables == NULL) |
1243 | { |
1244 | code->memctl.free((void *)newcode, code->memctl.memory_data); |
1245 | return NULL; |
1246 | } |
1247 | memcpy(newtables, code->tables, TABLES_LENGTH); |
1248 | ref_count = (PCRE2_SIZE *)(newtables + TABLES_LENGTH); |
1249 | *ref_count = 1; |
1250 | |
1251 | newcode->tables = newtables; |
1252 | newcode->flags |= PCRE2_DEREF_TABLES; |
1253 | return newcode; |
1254 | } |
1255 | |
1256 | |
1257 | |
1258 | /************************************************* |
1259 | * Free compiled code * |
1260 | *************************************************/ |
1261 | |
1262 | PCRE2_EXP_DEFN void PCRE2_CALL_CONVENTION |
1263 | pcre2_code_free(pcre2_code *code) |
1264 | { |
1265 | PCRE2_SIZE* ref_count; |
1266 | |
1267 | if (code != NULL) |
1268 | { |
1269 | #ifdef SUPPORT_JIT |
1270 | if (code->executable_jit != NULL) |
1271 | PRIV(jit_free)(code->executable_jit, &code->memctl); |
1272 | #endif |
1273 | |
1274 | if ((code->flags & PCRE2_DEREF_TABLES) != 0) |
1275 | { |
1276 | /* Decoded tables belong to the codes after deserialization, and they must |
1277 | be freed when there are no more references to them. The *ref_count should |
1278 | always be > 0. */ |
1279 | |
1280 | ref_count = (PCRE2_SIZE *)(code->tables + TABLES_LENGTH); |
1281 | if (*ref_count > 0) |
1282 | { |
1283 | (*ref_count)--; |
1284 | if (*ref_count == 0) |
1285 | code->memctl.free((void *)code->tables, code->memctl.memory_data); |
1286 | } |
1287 | } |
1288 | |
1289 | code->memctl.free(code, code->memctl.memory_data); |
1290 | } |
1291 | } |
1292 | |
1293 | |
1294 | |
1295 | /************************************************* |
1296 | * Read a number, possibly signed * |
1297 | *************************************************/ |
1298 | |
1299 | /* This function is used to read numbers in the pattern. The initial pointer |
1300 | must be the sign or first digit of the number. When relative values (introduced |
1301 | by + or -) are allowed, they are relative group numbers, and the result must be |
1302 | greater than zero. |
1303 | |
1304 | Arguments: |
1305 | ptrptr points to the character pointer variable |
1306 | ptrend points to the end of the input string |
1307 | allow_sign if < 0, sign not allowed; if >= 0, sign is relative to this |
1308 | max_value the largest number allowed |
1309 | max_error the error to give for an over-large number |
1310 | intptr where to put the result |
1311 | errcodeptr where to put an error code |
1312 | |
1313 | Returns: TRUE - a number was read |
1314 | FALSE - errorcode == 0 => no number was found |
1315 | errorcode != 0 => an error occurred |
1316 | */ |
1317 | |
1318 | static BOOL |
1319 | read_number(PCRE2_SPTR *ptrptr, PCRE2_SPTR ptrend, int32_t allow_sign, |
1320 | uint32_t max_value, uint32_t max_error, int *intptr, int *errorcodeptr) |
1321 | { |
1322 | int sign = 0; |
1323 | uint32_t n = 0; |
1324 | PCRE2_SPTR ptr = *ptrptr; |
1325 | BOOL yield = FALSE; |
1326 | |
1327 | *errorcodeptr = 0; |
1328 | |
1329 | if (allow_sign >= 0 && ptr < ptrend) |
1330 | { |
1331 | if (*ptr == CHAR_PLUS) |
1332 | { |
1333 | sign = +1; |
1334 | max_value -= allow_sign; |
1335 | ptr++; |
1336 | } |
1337 | else if (*ptr == CHAR_MINUS) |
1338 | { |
1339 | sign = -1; |
1340 | ptr++; |
1341 | } |
1342 | } |
1343 | |
1344 | if (ptr >= ptrend || !IS_DIGIT(*ptr)) return FALSE; |
1345 | while (ptr < ptrend && IS_DIGIT(*ptr)) |
1346 | { |
1347 | n = n * 10 + *ptr++ - CHAR_0; |
1348 | if (n > max_value) |
1349 | { |
1350 | *errorcodeptr = max_error; |
1351 | goto EXIT; |
1352 | } |
1353 | } |
1354 | |
1355 | if (allow_sign >= 0 && sign != 0) |
1356 | { |
1357 | if (n == 0) |
1358 | { |
1359 | *errorcodeptr = ERR26; /* +0 and -0 are not allowed */ |
1360 | goto EXIT; |
1361 | } |
1362 | |
1363 | if (sign > 0) n += allow_sign; |
1364 | else if ((int)n > allow_sign) |
1365 | { |
1366 | *errorcodeptr = ERR15; /* Non-existent subpattern */ |
1367 | goto EXIT; |
1368 | } |
1369 | else n = allow_sign + 1 - n; |
1370 | } |
1371 | |
1372 | yield = TRUE; |
1373 | |
1374 | EXIT: |
1375 | *intptr = n; |
1376 | *ptrptr = ptr; |
1377 | return yield; |
1378 | } |
1379 | |
1380 | |
1381 | |
1382 | /************************************************* |
1383 | * Read repeat counts * |
1384 | *************************************************/ |
1385 | |
1386 | /* Read an item of the form {n,m} and return the values if non-NULL pointers |
1387 | are supplied. Repeat counts must be less than 65536 (MAX_REPEAT_COUNT); a |
1388 | larger value is used for "unlimited". We have to use signed arguments for |
1389 | read_number() because it is capable of returning a signed value. |
1390 | |
1391 | Arguments: |
1392 | ptrptr points to pointer to character after'{' |
1393 | ptrend pointer to end of input |
1394 | minp if not NULL, pointer to int for min |
1395 | maxp if not NULL, pointer to int for max (-1 if no max) |
1396 | returned as -1 if no max |
1397 | errorcodeptr points to error code variable |
1398 | |
1399 | Returns: FALSE if not a repeat quantifier, errorcode set zero |
1400 | FALSE on error, with errorcode set non-zero |
1401 | TRUE on success, with pointer updated to point after '}' |
1402 | */ |
1403 | |
1404 | static BOOL |
1405 | read_repeat_counts(PCRE2_SPTR *ptrptr, PCRE2_SPTR ptrend, uint32_t *minp, |
1406 | uint32_t *maxp, int *errorcodeptr) |
1407 | { |
1408 | PCRE2_SPTR p; |
1409 | BOOL yield = FALSE; |
1410 | BOOL had_comma = FALSE; |
1411 | int32_t min = 0; |
1412 | int32_t max = REPEAT_UNLIMITED; /* This value is larger than MAX_REPEAT_COUNT */ |
1413 | |
1414 | /* Check the syntax */ |
1415 | |
1416 | *errorcodeptr = 0; |
1417 | for (p = *ptrptr;; p++) |
1418 | { |
1419 | uint32_t c; |
1420 | if (p >= ptrend) return FALSE; |
1421 | c = *p; |
1422 | if (IS_DIGIT(c)) continue; |
1423 | if (c == CHAR_RIGHT_CURLY_BRACKET) break; |
1424 | if (c == CHAR_COMMA) |
1425 | { |
1426 | if (had_comma) return FALSE; |
1427 | had_comma = TRUE; |
1428 | } |
1429 | else return FALSE; |
1430 | } |
1431 | |
1432 | /* The only error from read_number() is for a number that is too big. */ |
1433 | |
1434 | p = *ptrptr; |
1435 | if (!read_number(&p, ptrend, -1, MAX_REPEAT_COUNT, ERR5, &min, errorcodeptr)) |
1436 | goto EXIT; |
1437 | |
1438 | if (*p == CHAR_RIGHT_CURLY_BRACKET) |
1439 | { |
1440 | p++; |
1441 | max = min; |
1442 | } |
1443 | else |
1444 | { |
1445 | if (*(++p) != CHAR_RIGHT_CURLY_BRACKET) |
1446 | { |
1447 | if (!read_number(&p, ptrend, -1, MAX_REPEAT_COUNT, ERR5, &max, |
1448 | errorcodeptr)) |
1449 | goto EXIT; |
1450 | if (max < min) |
1451 | { |
1452 | *errorcodeptr = ERR4; |
1453 | goto EXIT; |
1454 | } |
1455 | } |
1456 | p++; |
1457 | } |
1458 | |
1459 | yield = TRUE; |
1460 | if (minp != NULL) *minp = (uint32_t)min; |
1461 | if (maxp != NULL) *maxp = (uint32_t)max; |
1462 | |
1463 | /* Update the pattern pointer */ |
1464 | |
1465 | EXIT: |
1466 | *ptrptr = p; |
1467 | return yield; |
1468 | } |
1469 | |
1470 | |
1471 | |
1472 | /************************************************* |
1473 | * Handle escapes * |
1474 | *************************************************/ |
1475 | |
1476 | /* This function is called when a \ has been encountered. It either returns a |
1477 | positive value for a simple escape such as \d, or 0 for a data character, which |
1478 | is placed in chptr. A backreference to group n is returned as negative n. On |
1479 | entry, ptr is pointing at the character after \. On exit, it points after the |
1480 | final code unit of the escape sequence. |
1481 | |
1482 | This function is also called from pcre2_substitute() to handle escape sequences |
1483 | in replacement strings. In this case, the cb argument is NULL, and in the case |
1484 | of escapes that have further processing, only sequences that define a data |
1485 | character are recognised. The isclass argument is not relevant; the options |
1486 | argument is the final value of the compiled pattern's options. |
1487 | |
1488 | Arguments: |
1489 | ptrptr points to the input position pointer |
1490 | ptrend points to the end of the input |
1491 | chptr points to a returned data character |
1492 | errorcodeptr points to the errorcode variable (containing zero) |
1493 | options the current options bits |
1494 | isclass TRUE if inside a character class |
1495 | cb compile data block or NULL when called from pcre2_substitute() |
1496 | |
1497 | Returns: zero => a data character |
1498 | positive => a special escape sequence |
1499 | negative => a numerical back reference |
1500 | on error, errorcodeptr is set non-zero |
1501 | */ |
1502 | |
1503 | int |
1504 | PRIV(check_escape)(PCRE2_SPTR *ptrptr, PCRE2_SPTR ptrend, uint32_t *chptr, |
1505 | int *errorcodeptr, uint32_t options, uint32_t , BOOL isclass, |
1506 | compile_block *cb) |
1507 | { |
1508 | BOOL utf = (options & PCRE2_UTF) != 0; |
1509 | PCRE2_SPTR ptr = *ptrptr; |
1510 | uint32_t c, cc; |
1511 | int escape = 0; |
1512 | int i; |
1513 | |
1514 | /* If backslash is at the end of the string, it's an error. */ |
1515 | |
1516 | if (ptr >= ptrend) |
1517 | { |
1518 | *errorcodeptr = ERR1; |
1519 | return 0; |
1520 | } |
1521 | |
1522 | GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */ |
1523 | *errorcodeptr = 0; /* Be optimistic */ |
1524 | |
1525 | /* Non-alphanumerics are literals, so we just leave the value in c. An initial |
1526 | value test saves a memory lookup for code points outside the alphanumeric |
1527 | range. */ |
1528 | |
1529 | if (c < ESCAPES_FIRST || c > ESCAPES_LAST) {} /* Definitely literal */ |
1530 | |
1531 | /* Otherwise, do a table lookup. Non-zero values need little processing here. A |
1532 | positive value is a literal value for something like \n. A negative value is |
1533 | the negation of one of the ESC_ macros that is passed back for handling by the |
1534 | calling function. Some extra checking is needed for \N because only \N{U+dddd} |
1535 | is supported. If the value is zero, further processing is handled below. */ |
1536 | |
1537 | else if ((i = escapes[c - ESCAPES_FIRST]) != 0) |
1538 | { |
1539 | if (i > 0) |
1540 | { |
1541 | c = (uint32_t)i; |
1542 | if (c == CHAR_CR && (extra_options & PCRE2_EXTRA_ESCAPED_CR_IS_LF) != 0) |
1543 | c = CHAR_LF; |
1544 | } |
1545 | else /* Negative table entry */ |
1546 | { |
1547 | escape = -i; /* Else return a special escape */ |
1548 | if (cb != NULL && (escape == ESC_P || escape == ESC_p || escape == ESC_X)) |
1549 | cb->external_flags |= PCRE2_HASBKPORX; /* Note \P, \p, or \X */ |
1550 | |
1551 | /* Perl supports \N{name} for character names and \N{U+dddd} for numerical |
1552 | Unicode code points, as well as plain \N for "not newline". PCRE does not |
1553 | support \N{name}. However, it does support quantification such as \N{2,3}, |
1554 | so if \N{ is not followed by U+dddd we check for a quantifier. */ |
1555 | |
1556 | if (escape == ESC_N && ptr < ptrend && *ptr == CHAR_LEFT_CURLY_BRACKET) |
1557 | { |
1558 | PCRE2_SPTR p = ptr + 1; |
1559 | |
1560 | /* \N{U+ can be handled by the \x{ code. However, this construction is |
1561 | not valid in EBCDIC environments because it specifies a Unicode |
1562 | character, not a codepoint in the local code. For example \N{U+0041} |
1563 | must be "A" in all environments. Also, in Perl, \N{U+ forces Unicode |
1564 | casing semantics for the entire pattern, so allow it only in UTF (i.e. |
1565 | Unicode) mode. */ |
1566 | |
1567 | if (ptrend - p > 1 && *p == CHAR_U && p[1] == CHAR_PLUS) |
1568 | { |
1569 | #ifdef EBCDIC |
1570 | *errorcodeptr = ERR93; |
1571 | #else |
1572 | if (utf) |
1573 | { |
1574 | ptr = p + 1; |
1575 | escape = 0; /* Not a fancy escape after all */ |
1576 | goto COME_FROM_NU; |
1577 | } |
1578 | else *errorcodeptr = ERR93; |
1579 | #endif |
1580 | } |
1581 | |
1582 | /* Give an error if what follows is not a quantifier, but don't override |
1583 | an error set by the quantifier reader (e.g. number overflow). */ |
1584 | |
1585 | else |
1586 | { |
1587 | if (!read_repeat_counts(&p, ptrend, NULL, NULL, errorcodeptr) && |
1588 | *errorcodeptr == 0) |
1589 | *errorcodeptr = ERR37; |
1590 | } |
1591 | } |
1592 | } |
1593 | } |
1594 | |
1595 | /* Escapes that need further processing, including those that are unknown, have |
1596 | a zero entry in the lookup table. When called from pcre2_substitute(), only \c, |
1597 | \o, and \x are recognized (\u and \U can never appear as they are used for case |
1598 | forcing). */ |
1599 | |
1600 | else |
1601 | { |
1602 | int s; |
1603 | PCRE2_SPTR oldptr; |
1604 | BOOL overflow; |
1605 | BOOL alt_bsux = |
1606 | ((options & PCRE2_ALT_BSUX) | (extra_options & PCRE2_EXTRA_ALT_BSUX)) != 0; |
1607 | |
1608 | /* Filter calls from pcre2_substitute(). */ |
1609 | |
1610 | if (cb == NULL) |
1611 | { |
1612 | if (c != CHAR_c && c != CHAR_o && c != CHAR_x) |
1613 | { |
1614 | *errorcodeptr = ERR3; |
1615 | return 0; |
1616 | } |
1617 | alt_bsux = FALSE; /* Do not modify \x handling */ |
1618 | } |
1619 | |
1620 | switch (c) |
1621 | { |
1622 | /* A number of Perl escapes are not handled by PCRE. We give an explicit |
1623 | error. */ |
1624 | |
1625 | case CHAR_F: |
1626 | case CHAR_l: |
1627 | case CHAR_L: |
1628 | *errorcodeptr = ERR37; |
1629 | break; |
1630 | |
1631 | /* \u is unrecognized when neither PCRE2_ALT_BSUX nor PCRE2_EXTRA_ALT_BSUX |
1632 | is set. Otherwise, \u must be followed by exactly four hex digits or, if |
1633 | PCRE2_EXTRA_ALT_BSUX is set, by any number of hex digits in braces. |
1634 | Otherwise it is a lowercase u letter. This gives some compatibility with |
1635 | ECMAScript (aka JavaScript). */ |
1636 | |
1637 | case CHAR_u: |
1638 | if (!alt_bsux) *errorcodeptr = ERR37; else |
1639 | { |
1640 | uint32_t xc; |
1641 | |
1642 | if (ptr >= ptrend) break; |
1643 | if (*ptr == CHAR_LEFT_CURLY_BRACKET && |
1644 | (extra_options & PCRE2_EXTRA_ALT_BSUX) != 0) |
1645 | { |
1646 | PCRE2_SPTR hptr = ptr + 1; |
1647 | cc = 0; |
1648 | |
1649 | while (hptr < ptrend && (xc = XDIGIT(*hptr)) != 0xff) |
1650 | { |
1651 | if ((cc & 0xf0000000) != 0) /* Test for 32-bit overflow */ |
1652 | { |
1653 | *errorcodeptr = ERR77; |
1654 | ptr = hptr; /* Show where */ |
1655 | break; /* *hptr != } will cause another break below */ |
1656 | } |
1657 | cc = (cc << 4) | xc; |
1658 | hptr++; |
1659 | } |
1660 | |
1661 | if (hptr == ptr + 1 || /* No hex digits */ |
1662 | hptr >= ptrend || /* Hit end of input */ |
1663 | *hptr != CHAR_RIGHT_CURLY_BRACKET) /* No } terminator */ |
1664 | break; /* Hex escape not recognized */ |
1665 | |
1666 | c = cc; /* Accept the code point */ |
1667 | ptr = hptr + 1; |
1668 | } |
1669 | |
1670 | else /* Must be exactly 4 hex digits */ |
1671 | { |
1672 | if (ptrend - ptr < 4) break; /* Less than 4 chars */ |
1673 | if ((cc = XDIGIT(ptr[0])) == 0xff) break; /* Not a hex digit */ |
1674 | if ((xc = XDIGIT(ptr[1])) == 0xff) break; /* Not a hex digit */ |
1675 | cc = (cc << 4) | xc; |
1676 | if ((xc = XDIGIT(ptr[2])) == 0xff) break; /* Not a hex digit */ |
1677 | cc = (cc << 4) | xc; |
1678 | if ((xc = XDIGIT(ptr[3])) == 0xff) break; /* Not a hex digit */ |
1679 | c = (cc << 4) | xc; |
1680 | ptr += 4; |
1681 | } |
1682 | |
1683 | if (utf) |
1684 | { |
1685 | if (c > 0x10ffffU) *errorcodeptr = ERR77; |
1686 | else |
1687 | if (c >= 0xd800 && c <= 0xdfff && |
1688 | (extra_options & PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES) == 0) |
1689 | *errorcodeptr = ERR73; |
1690 | } |
1691 | else if (c > MAX_NON_UTF_CHAR) *errorcodeptr = ERR77; |
1692 | } |
1693 | break; |
1694 | |
1695 | /* \U is unrecognized unless PCRE2_ALT_BSUX or PCRE2_EXTRA_ALT_BSUX is set, |
1696 | in which case it is an upper case letter. */ |
1697 | |
1698 | case CHAR_U: |
1699 | if (!alt_bsux) *errorcodeptr = ERR37; |
1700 | break; |
1701 | |
1702 | /* In a character class, \g is just a literal "g". Outside a character |
1703 | class, \g must be followed by one of a number of specific things: |
1704 | |
1705 | (1) A number, either plain or braced. If positive, it is an absolute |
1706 | backreference. If negative, it is a relative backreference. This is a Perl |
1707 | 5.10 feature. |
1708 | |
1709 | (2) Perl 5.10 also supports \g{name} as a reference to a named group. This |
1710 | is part of Perl's movement towards a unified syntax for back references. As |
1711 | this is synonymous with \k{name}, we fudge it up by pretending it really |
1712 | was \k{name}. |
1713 | |
1714 | (3) For Oniguruma compatibility we also support \g followed by a name or a |
1715 | number either in angle brackets or in single quotes. However, these are |
1716 | (possibly recursive) subroutine calls, _not_ backreferences. We return |
1717 | the ESC_g code. |
1718 | |
1719 | Summary: Return a negative number for a numerical back reference, ESC_k for |
1720 | a named back reference, and ESC_g for a named or numbered subroutine call. |
1721 | */ |
1722 | |
1723 | case CHAR_g: |
1724 | if (isclass) break; |
1725 | |
1726 | if (ptr >= ptrend) |
1727 | { |
1728 | *errorcodeptr = ERR57; |
1729 | break; |
1730 | } |
1731 | |
1732 | if (*ptr == CHAR_LESS_THAN_SIGN || *ptr == CHAR_APOSTROPHE) |
1733 | { |
1734 | escape = ESC_g; |
1735 | break; |
1736 | } |
1737 | |
1738 | /* If there is a brace delimiter, try to read a numerical reference. If |
1739 | there isn't one, assume we have a name and treat it as \k. */ |
1740 | |
1741 | if (*ptr == CHAR_LEFT_CURLY_BRACKET) |
1742 | { |
1743 | PCRE2_SPTR p = ptr + 1; |
1744 | if (!read_number(&p, ptrend, cb->bracount, MAX_GROUP_NUMBER, ERR61, &s, |
1745 | errorcodeptr)) |
1746 | { |
1747 | if (*errorcodeptr == 0) escape = ESC_k; /* No number found */ |
1748 | break; |
1749 | } |
1750 | if (p >= ptrend || *p != CHAR_RIGHT_CURLY_BRACKET) |
1751 | { |
1752 | *errorcodeptr = ERR57; |
1753 | break; |
1754 | } |
1755 | ptr = p + 1; |
1756 | } |
1757 | |
1758 | /* Read an undelimited number */ |
1759 | |
1760 | else |
1761 | { |
1762 | if (!read_number(&ptr, ptrend, cb->bracount, MAX_GROUP_NUMBER, ERR61, &s, |
1763 | errorcodeptr)) |
1764 | { |
1765 | if (*errorcodeptr == 0) *errorcodeptr = ERR57; /* No number found */ |
1766 | break; |
1767 | } |
1768 | } |
1769 | |
1770 | if (s <= 0) |
1771 | { |
1772 | *errorcodeptr = ERR15; |
1773 | break; |
1774 | } |
1775 | |
1776 | escape = -s; |
1777 | break; |
1778 | |
1779 | /* The handling of escape sequences consisting of a string of digits |
1780 | starting with one that is not zero is not straightforward. Perl has changed |
1781 | over the years. Nowadays \g{} for backreferences and \o{} for octal are |
1782 | recommended to avoid the ambiguities in the old syntax. |
1783 | |
1784 | Outside a character class, the digits are read as a decimal number. If the |
1785 | number is less than 10, or if there are that many previous extracting left |
1786 | brackets, it is a back reference. Otherwise, up to three octal digits are |
1787 | read to form an escaped character code. Thus \123 is likely to be octal 123 |
1788 | (cf \0123, which is octal 012 followed by the literal 3). |
1789 | |
1790 | Inside a character class, \ followed by a digit is always either a literal |
1791 | 8 or 9 or an octal number. */ |
1792 | |
1793 | case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5: |
1794 | case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9: |
1795 | |
1796 | if (!isclass) |
1797 | { |
1798 | oldptr = ptr; |
1799 | ptr--; /* Back to the digit */ |
1800 | |
1801 | /* As we know we are at a digit, the only possible error from |
1802 | read_number() is a number that is too large to be a group number. In this |
1803 | case we fall through handle this as not a group reference. If we have |
1804 | read a small enough number, check for a back reference. |
1805 | |
1806 | \1 to \9 are always back references. \8x and \9x are too; \1x to \7x |
1807 | are octal escapes if there are not that many previous captures. */ |
1808 | |
1809 | if (read_number(&ptr, ptrend, -1, INT_MAX/10 - 1, 0, &s, errorcodeptr) && |
1810 | (s < 10 || oldptr[-1] >= CHAR_8 || s <= (int)cb->bracount)) |
1811 | { |
1812 | if (s > (int)MAX_GROUP_NUMBER) *errorcodeptr = ERR61; |
1813 | else escape = -s; /* Indicates a back reference */ |
1814 | break; |
1815 | } |
1816 | |
1817 | ptr = oldptr; /* Put the pointer back and fall through */ |
1818 | } |
1819 | |
1820 | /* Handle a digit following \ when the number is not a back reference, or |
1821 | we are within a character class. If the first digit is 8 or 9, Perl used to |
1822 | generate a binary zero and then treat the digit as a following literal. At |
1823 | least by Perl 5.18 this changed so as not to insert the binary zero. */ |
1824 | |
1825 | if (c >= CHAR_8) break; |
1826 | |
1827 | /* Fall through */ |
1828 | |
1829 | /* \0 always starts an octal number, but we may drop through to here with a |
1830 | larger first octal digit. The original code used just to take the least |
1831 | significant 8 bits of octal numbers (I think this is what early Perls used |
1832 | to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode, |
1833 | but no more than 3 octal digits. */ |
1834 | |
1835 | case CHAR_0: |
1836 | c -= CHAR_0; |
1837 | while(i++ < 2 && ptr < ptrend && *ptr >= CHAR_0 && *ptr <= CHAR_7) |
1838 | c = c * 8 + *ptr++ - CHAR_0; |
1839 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
1840 | if (!utf && c > 0xff) *errorcodeptr = ERR51; |
1841 | #endif |
1842 | break; |
1843 | |
1844 | /* \o is a relatively new Perl feature, supporting a more general way of |
1845 | specifying character codes in octal. The only supported form is \o{ddd}. */ |
1846 | |
1847 | case CHAR_o: |
1848 | if (ptr >= ptrend || *ptr++ != CHAR_LEFT_CURLY_BRACKET) |
1849 | { |
1850 | ptr--; |
1851 | *errorcodeptr = ERR55; |
1852 | } |
1853 | else if (ptr >= ptrend || *ptr == CHAR_RIGHT_CURLY_BRACKET) |
1854 | *errorcodeptr = ERR78; |
1855 | else |
1856 | { |
1857 | c = 0; |
1858 | overflow = FALSE; |
1859 | while (ptr < ptrend && *ptr >= CHAR_0 && *ptr <= CHAR_7) |
1860 | { |
1861 | cc = *ptr++; |
1862 | if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */ |
1863 | #if PCRE2_CODE_UNIT_WIDTH == 32 |
1864 | if (c >= 0x20000000l) { overflow = TRUE; break; } |
1865 | #endif |
1866 | c = (c << 3) + (cc - CHAR_0); |
1867 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
1868 | if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; } |
1869 | #elif PCRE2_CODE_UNIT_WIDTH == 16 |
1870 | if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; } |
1871 | #elif PCRE2_CODE_UNIT_WIDTH == 32 |
1872 | if (utf && c > 0x10ffffU) { overflow = TRUE; break; } |
1873 | #endif |
1874 | } |
1875 | if (overflow) |
1876 | { |
1877 | while (ptr < ptrend && *ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++; |
1878 | *errorcodeptr = ERR34; |
1879 | } |
1880 | else if (ptr < ptrend && *ptr++ == CHAR_RIGHT_CURLY_BRACKET) |
1881 | { |
1882 | if (utf && c >= 0xd800 && c <= 0xdfff && |
1883 | (extra_options & PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES) == 0) |
1884 | { |
1885 | ptr--; |
1886 | *errorcodeptr = ERR73; |
1887 | } |
1888 | } |
1889 | else |
1890 | { |
1891 | ptr--; |
1892 | *errorcodeptr = ERR64; |
1893 | } |
1894 | } |
1895 | break; |
1896 | |
1897 | /* When PCRE2_ALT_BSUX or PCRE2_EXTRA_ALT_BSUX is set, \x must be followed |
1898 | by two hexadecimal digits. Otherwise it is a lowercase x letter. */ |
1899 | |
1900 | case CHAR_x: |
1901 | if (alt_bsux) |
1902 | { |
1903 | uint32_t xc; |
1904 | if (ptrend - ptr < 2) break; /* Less than 2 characters */ |
1905 | if ((cc = XDIGIT(ptr[0])) == 0xff) break; /* Not a hex digit */ |
1906 | if ((xc = XDIGIT(ptr[1])) == 0xff) break; /* Not a hex digit */ |
1907 | c = (cc << 4) | xc; |
1908 | ptr += 2; |
1909 | } |
1910 | |
1911 | /* Handle \x in Perl's style. \x{ddd} is a character code which can be |
1912 | greater than 0xff in UTF-8 or non-8bit mode, but only if the ddd are hex |
1913 | digits. If not, { used to be treated as a data character. However, Perl |
1914 | seems to read hex digits up to the first non-such, and ignore the rest, so |
1915 | that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE |
1916 | now gives an error. */ |
1917 | |
1918 | else |
1919 | { |
1920 | if (ptr < ptrend && *ptr == CHAR_LEFT_CURLY_BRACKET) |
1921 | { |
1922 | #ifndef EBCDIC |
1923 | COME_FROM_NU: |
1924 | #endif |
1925 | if (++ptr >= ptrend || *ptr == CHAR_RIGHT_CURLY_BRACKET) |
1926 | { |
1927 | *errorcodeptr = ERR78; |
1928 | break; |
1929 | } |
1930 | c = 0; |
1931 | overflow = FALSE; |
1932 | |
1933 | while (ptr < ptrend && (cc = XDIGIT(*ptr)) != 0xff) |
1934 | { |
1935 | ptr++; |
1936 | if (c == 0 && cc == 0) continue; /* Leading zeroes */ |
1937 | #if PCRE2_CODE_UNIT_WIDTH == 32 |
1938 | if (c >= 0x10000000l) { overflow = TRUE; break; } |
1939 | #endif |
1940 | c = (c << 4) | cc; |
1941 | if ((utf && c > 0x10ffffU) || (!utf && c > MAX_NON_UTF_CHAR)) |
1942 | { |
1943 | overflow = TRUE; |
1944 | break; |
1945 | } |
1946 | } |
1947 | |
1948 | if (overflow) |
1949 | { |
1950 | while (ptr < ptrend && XDIGIT(*ptr) != 0xff) ptr++; |
1951 | *errorcodeptr = ERR34; |
1952 | } |
1953 | else if (ptr < ptrend && *ptr++ == CHAR_RIGHT_CURLY_BRACKET) |
1954 | { |
1955 | if (utf && c >= 0xd800 && c <= 0xdfff && |
1956 | (extra_options & PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES) == 0) |
1957 | { |
1958 | ptr--; |
1959 | *errorcodeptr = ERR73; |
1960 | } |
1961 | } |
1962 | |
1963 | /* If the sequence of hex digits does not end with '}', give an error. |
1964 | We used just to recognize this construct and fall through to the normal |
1965 | \x handling, but nowadays Perl gives an error, which seems much more |
1966 | sensible, so we do too. */ |
1967 | |
1968 | else |
1969 | { |
1970 | ptr--; |
1971 | *errorcodeptr = ERR67; |
1972 | } |
1973 | } /* End of \x{} processing */ |
1974 | |
1975 | /* Read a up to two hex digits after \x */ |
1976 | |
1977 | else |
1978 | { |
1979 | c = 0; |
1980 | if (ptr >= ptrend || (cc = XDIGIT(*ptr)) == 0xff) break; /* Not a hex digit */ |
1981 | ptr++; |
1982 | c = cc; |
1983 | if (ptr >= ptrend || (cc = XDIGIT(*ptr)) == 0xff) break; /* Not a hex digit */ |
1984 | ptr++; |
1985 | c = (c << 4) | cc; |
1986 | } /* End of \xdd handling */ |
1987 | } /* End of Perl-style \x handling */ |
1988 | break; |
1989 | |
1990 | /* The handling of \c is different in ASCII and EBCDIC environments. In an |
1991 | ASCII (or Unicode) environment, an error is given if the character |
1992 | following \c is not a printable ASCII character. Otherwise, the following |
1993 | character is upper-cased if it is a letter, and after that the 0x40 bit is |
1994 | flipped. The result is the value of the escape. |
1995 | |
1996 | In an EBCDIC environment the handling of \c is compatible with the |
1997 | specification in the perlebcdic document. The following character must be |
1998 | a letter or one of small number of special characters. These provide a |
1999 | means of defining the character values 0-31. |
2000 | |
2001 | For testing the EBCDIC handling of \c in an ASCII environment, recognize |
2002 | the EBCDIC value of 'c' explicitly. */ |
2003 | |
2004 | #if defined EBCDIC && 'a' != 0x81 |
2005 | case 0x83: |
2006 | #else |
2007 | case CHAR_c: |
2008 | #endif |
2009 | if (ptr >= ptrend) |
2010 | { |
2011 | *errorcodeptr = ERR2; |
2012 | break; |
2013 | } |
2014 | c = *ptr; |
2015 | if (c >= CHAR_a && c <= CHAR_z) c = UPPER_CASE(c); |
2016 | |
2017 | /* Handle \c in an ASCII/Unicode environment. */ |
2018 | |
2019 | #ifndef EBCDIC /* ASCII/UTF-8 coding */ |
2020 | if (c < 32 || c > 126) /* Excludes all non-printable ASCII */ |
2021 | { |
2022 | *errorcodeptr = ERR68; |
2023 | break; |
2024 | } |
2025 | c ^= 0x40; |
2026 | |
2027 | /* Handle \c in an EBCDIC environment. The special case \c? is converted to |
2028 | 255 (0xff) or 95 (0x5f) if other characters suggest we are using the |
2029 | POSIX-BC encoding. (This is the way Perl indicates that it handles \c?.) |
2030 | The other valid sequences correspond to a list of specific characters. */ |
2031 | |
2032 | #else |
2033 | if (c == CHAR_QUESTION_MARK) |
2034 | c = ('\\' == 188 && '`' == 74)? 0x5f : 0xff; |
2035 | else |
2036 | { |
2037 | for (i = 0; i < 32; i++) |
2038 | { |
2039 | if (c == ebcdic_escape_c[i]) break; |
2040 | } |
2041 | if (i < 32) c = i; else *errorcodeptr = ERR68; |
2042 | } |
2043 | #endif /* EBCDIC */ |
2044 | |
2045 | ptr++; |
2046 | break; |
2047 | |
2048 | /* Any other alphanumeric following \ is an error. Perl gives an error only |
2049 | if in warning mode, but PCRE doesn't have a warning mode. */ |
2050 | |
2051 | default: |
2052 | *errorcodeptr = ERR3; |
2053 | *ptrptr = ptr - 1; /* Point to the character at fault */ |
2054 | return 0; |
2055 | } |
2056 | } |
2057 | |
2058 | /* Set the pointer to the next character before returning. */ |
2059 | |
2060 | *ptrptr = ptr; |
2061 | *chptr = c; |
2062 | return escape; |
2063 | } |
2064 | |
2065 | |
2066 | |
2067 | #ifdef SUPPORT_UNICODE |
2068 | /************************************************* |
2069 | * Handle \P and \p * |
2070 | *************************************************/ |
2071 | |
2072 | /* This function is called after \P or \p has been encountered, provided that |
2073 | PCRE2 is compiled with support for UTF and Unicode properties. On entry, the |
2074 | contents of ptrptr are pointing after the P or p. On exit, it is left pointing |
2075 | after the final code unit of the escape sequence. |
2076 | |
2077 | Arguments: |
2078 | ptrptr the pattern position pointer |
2079 | negptr a boolean that is set TRUE for negation else FALSE |
2080 | ptypeptr an unsigned int that is set to the type value |
2081 | pdataptr an unsigned int that is set to the detailed property value |
2082 | errorcodeptr the error code variable |
2083 | cb the compile data |
2084 | |
2085 | Returns: TRUE if the type value was found, or FALSE for an invalid type |
2086 | */ |
2087 | |
2088 | static BOOL |
2089 | get_ucp(PCRE2_SPTR *ptrptr, BOOL *negptr, uint16_t *ptypeptr, |
2090 | uint16_t *pdataptr, int *errorcodeptr, compile_block *cb) |
2091 | { |
2092 | PCRE2_UCHAR c; |
2093 | PCRE2_SIZE i, bot, top; |
2094 | PCRE2_SPTR ptr = *ptrptr; |
2095 | PCRE2_UCHAR name[50]; |
2096 | PCRE2_UCHAR *vptr = NULL; |
2097 | uint16_t ptscript = PT_NOTSCRIPT; |
2098 | |
2099 | if (ptr >= cb->end_pattern) goto ERROR_RETURN; |
2100 | c = *ptr++; |
2101 | *negptr = FALSE; |
2102 | |
2103 | /* \P or \p can be followed by a name in {}, optionally preceded by ^ for |
2104 | negation. */ |
2105 | |
2106 | if (c == CHAR_LEFT_CURLY_BRACKET) |
2107 | { |
2108 | if (ptr >= cb->end_pattern) goto ERROR_RETURN; |
2109 | |
2110 | if (*ptr == CHAR_CIRCUMFLEX_ACCENT) |
2111 | { |
2112 | *negptr = TRUE; |
2113 | ptr++; |
2114 | } |
2115 | |
2116 | for (i = 0; i < (int)(sizeof(name) / sizeof(PCRE2_UCHAR)) - 1; i++) |
2117 | { |
2118 | if (ptr >= cb->end_pattern) goto ERROR_RETURN; |
2119 | c = *ptr++; |
2120 | while (c == '_' || c == '-' || isspace(c)) |
2121 | { |
2122 | if (ptr >= cb->end_pattern) goto ERROR_RETURN; |
2123 | c = *ptr++; |
2124 | } |
2125 | if (c == CHAR_NUL) goto ERROR_RETURN; |
2126 | if (c == CHAR_RIGHT_CURLY_BRACKET) break; |
2127 | name[i] = tolower(c); |
2128 | if ((c == ':' || c == '=') && vptr == NULL) vptr = name + i; |
2129 | } |
2130 | |
2131 | if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN; |
2132 | name[i] = 0; |
2133 | } |
2134 | |
2135 | /* If { doesn't follow \p or \P there is just one following character, which |
2136 | must be an ASCII letter. */ |
2137 | |
2138 | else if (MAX_255(c) && (cb->ctypes[c] & ctype_letter) != 0) |
2139 | { |
2140 | name[0] = tolower(c); |
2141 | name[1] = 0; |
2142 | } |
2143 | else goto ERROR_RETURN; |
2144 | |
2145 | *ptrptr = ptr; |
2146 | |
2147 | /* If the property contains ':' or '=' we have class name and value separately |
2148 | specified. The following are supported: |
2149 | |
2150 | . Bidi_Class (synonym bc), for which the property names are "bidi<name>". |
2151 | . Script (synonym sc) for which the property name is the script name |
2152 | . Script_Extensions (synonym scx), ditto |
2153 | |
2154 | As this is a small number, we currently just check the names directly. If this |
2155 | grows, a sorted table and a switch will be neater. |
2156 | |
2157 | For both the script properties, set a PT_xxx value so that (1) they can be |
2158 | distinguished and (2) invalid script names that happen to be the name of |
2159 | another property can be diagnosed. */ |
2160 | |
2161 | if (vptr != NULL) |
2162 | { |
2163 | int offset = 0; |
2164 | PCRE2_UCHAR sname[8]; |
2165 | |
2166 | *vptr = 0; /* Terminate property name */ |
2167 | if (PRIV(strcmp_c8)(name, STRING_bidiclass) == 0 || |
2168 | PRIV(strcmp_c8)(name, STRING_bc) == 0) |
2169 | { |
2170 | offset = 4; |
2171 | sname[0] = CHAR_b; |
2172 | sname[1] = CHAR_i; /* There is no strcpy_c8 function */ |
2173 | sname[2] = CHAR_d; |
2174 | sname[3] = CHAR_i; |
2175 | } |
2176 | |
2177 | else if (PRIV(strcmp_c8)(name, STRING_script) == 0 || |
2178 | PRIV(strcmp_c8)(name, STRING_sc) == 0) |
2179 | ptscript = PT_SC; |
2180 | |
2181 | else if (PRIV(strcmp_c8)(name, STRING_scriptextensions) == 0 || |
2182 | PRIV(strcmp_c8)(name, STRING_scx) == 0) |
2183 | ptscript = PT_SCX; |
2184 | |
2185 | else |
2186 | { |
2187 | *errorcodeptr = ERR47; |
2188 | return FALSE; |
2189 | } |
2190 | |
2191 | /* Adjust the string in name[] as needed */ |
2192 | |
2193 | memmove(name + offset, vptr + 1, (name + i - vptr)*sizeof(PCRE2_UCHAR)); |
2194 | if (offset != 0) memmove(name, sname, offset*sizeof(PCRE2_UCHAR)); |
2195 | } |
2196 | |
2197 | /* Search for a recognized property using binary chop. */ |
2198 | |
2199 | bot = 0; |
2200 | top = PRIV(utt_size); |
2201 | |
2202 | while (bot < top) |
2203 | { |
2204 | int r; |
2205 | i = (bot + top) >> 1; |
2206 | r = PRIV(strcmp_c8)(name, PRIV(utt_names) + PRIV(utt)[i].name_offset); |
2207 | |
2208 | /* When a matching property is found, some extra checking is needed when the |
2209 | \p{xx:yy} syntax is used and xx is either sc or scx. */ |
2210 | |
2211 | if (r == 0) |
2212 | { |
2213 | *pdataptr = PRIV(utt)[i].value; |
2214 | if (vptr == NULL || ptscript == PT_NOTSCRIPT) |
2215 | { |
2216 | *ptypeptr = PRIV(utt)[i].type; |
2217 | return TRUE; |
2218 | } |
2219 | |
2220 | switch (PRIV(utt)[i].type) |
2221 | { |
2222 | case PT_SC: |
2223 | *ptypeptr = PT_SC; |
2224 | return TRUE; |
2225 | |
2226 | case PT_SCX: |
2227 | *ptypeptr = ptscript; |
2228 | return TRUE; |
2229 | } |
2230 | |
2231 | break; /* Non-script found */ |
2232 | } |
2233 | |
2234 | if (r > 0) bot = i + 1; else top = i; |
2235 | } |
2236 | |
2237 | *errorcodeptr = ERR47; /* Unrecognized property */ |
2238 | return FALSE; |
2239 | |
2240 | ERROR_RETURN: /* Malformed \P or \p */ |
2241 | *errorcodeptr = ERR46; |
2242 | *ptrptr = ptr; |
2243 | return FALSE; |
2244 | } |
2245 | #endif |
2246 | |
2247 | |
2248 | |
2249 | /************************************************* |
2250 | * Check for POSIX class syntax * |
2251 | *************************************************/ |
2252 | |
2253 | /* This function is called when the sequence "[:" or "[." or "[=" is |
2254 | encountered in a character class. It checks whether this is followed by a |
2255 | sequence of characters terminated by a matching ":]" or ".]" or "=]". If we |
2256 | reach an unescaped ']' without the special preceding character, return FALSE. |
2257 | |
2258 | Originally, this function only recognized a sequence of letters between the |
2259 | terminators, but it seems that Perl recognizes any sequence of characters, |
2260 | though of course unknown POSIX names are subsequently rejected. Perl gives an |
2261 | "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE |
2262 | didn't consider this to be a POSIX class. Likewise for [:1234:]. |
2263 | |
2264 | The problem in trying to be exactly like Perl is in the handling of escapes. We |
2265 | have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX |
2266 | class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code |
2267 | below handles the special cases \\ and \], but does not try to do any other |
2268 | escape processing. This makes it different from Perl for cases such as |
2269 | [:l\ower:] where Perl recognizes it as the POSIX class "lower" but PCRE does |
2270 | not recognize "l\ower". This is a lesser evil than not diagnosing bad classes |
2271 | when Perl does, I think. |
2272 | |
2273 | A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not. |
2274 | It seems that the appearance of a nested POSIX class supersedes an apparent |
2275 | external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or |
2276 | a digit. This is handled by returning FALSE if the start of a new group with |
2277 | the same terminator is encountered, since the next closing sequence must close |
2278 | the nested group, not the outer one. |
2279 | |
2280 | In Perl, unescaped square brackets may also appear as part of class names. For |
2281 | example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for |
2282 | [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not |
2283 | seem right at all. PCRE does not allow closing square brackets in POSIX class |
2284 | names. |
2285 | |
2286 | Arguments: |
2287 | ptr pointer to the character after the initial [ (colon, dot, equals) |
2288 | ptrend pointer to the end of the pattern |
2289 | endptr where to return a pointer to the terminating ':', '.', or '=' |
2290 | |
2291 | Returns: TRUE or FALSE |
2292 | */ |
2293 | |
2294 | static BOOL |
2295 | check_posix_syntax(PCRE2_SPTR ptr, PCRE2_SPTR ptrend, PCRE2_SPTR *endptr) |
2296 | { |
2297 | PCRE2_UCHAR terminator; /* Don't combine these lines; the Solaris cc */ |
2298 | terminator = *ptr++; /* compiler warns about "non-constant" initializer. */ |
2299 | |
2300 | for (; ptrend - ptr >= 2; ptr++) |
2301 | { |
2302 | if (*ptr == CHAR_BACKSLASH && |
2303 | (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET || ptr[1] == CHAR_BACKSLASH)) |
2304 | ptr++; |
2305 | |
2306 | else if ((*ptr == CHAR_LEFT_SQUARE_BRACKET && ptr[1] == terminator) || |
2307 | *ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE; |
2308 | |
2309 | else if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) |
2310 | { |
2311 | *endptr = ptr; |
2312 | return TRUE; |
2313 | } |
2314 | } |
2315 | |
2316 | return FALSE; |
2317 | } |
2318 | |
2319 | |
2320 | |
2321 | /************************************************* |
2322 | * Check POSIX class name * |
2323 | *************************************************/ |
2324 | |
2325 | /* This function is called to check the name given in a POSIX-style class entry |
2326 | such as [:alnum:]. |
2327 | |
2328 | Arguments: |
2329 | ptr points to the first letter |
2330 | len the length of the name |
2331 | |
2332 | Returns: a value representing the name, or -1 if unknown |
2333 | */ |
2334 | |
2335 | static int |
2336 | check_posix_name(PCRE2_SPTR ptr, int len) |
2337 | { |
2338 | const char *pn = posix_names; |
2339 | int yield = 0; |
2340 | while (posix_name_lengths[yield] != 0) |
2341 | { |
2342 | if (len == posix_name_lengths[yield] && |
2343 | PRIV(strncmp_c8)(ptr, pn, (unsigned int)len) == 0) return yield; |
2344 | pn += posix_name_lengths[yield] + 1; |
2345 | yield++; |
2346 | } |
2347 | return -1; |
2348 | } |
2349 | |
2350 | |
2351 | |
2352 | /************************************************* |
2353 | * Read a subpattern or VERB name * |
2354 | *************************************************/ |
2355 | |
2356 | /* This function is called from parse_regex() below whenever it needs to read |
2357 | the name of a subpattern or a (*VERB) or an (*alpha_assertion). The initial |
2358 | pointer must be to the character before the name. If that character is '*' we |
2359 | are reading a verb or alpha assertion name. The pointer is updated to point |
2360 | after the name, for a VERB or alpha assertion name, or after tha name's |
2361 | terminator for a subpattern name. Returning both the offset and the name |
2362 | pointer is redundant information, but some callers use one and some the other, |
2363 | so it is simplest just to return both. |
2364 | |
2365 | Arguments: |
2366 | ptrptr points to the character pointer variable |
2367 | ptrend points to the end of the input string |
2368 | utf true if the input is UTF-encoded |
2369 | terminator the terminator of a subpattern name must be this |
2370 | offsetptr where to put the offset from the start of the pattern |
2371 | nameptr where to put a pointer to the name in the input |
2372 | namelenptr where to put the length of the name |
2373 | errcodeptr where to put an error code |
2374 | cb pointer to the compile data block |
2375 | |
2376 | Returns: TRUE if a name was read |
2377 | FALSE otherwise, with error code set |
2378 | */ |
2379 | |
2380 | static BOOL |
2381 | read_name(PCRE2_SPTR *ptrptr, PCRE2_SPTR ptrend, BOOL utf, uint32_t terminator, |
2382 | PCRE2_SIZE *offsetptr, PCRE2_SPTR *nameptr, uint32_t *namelenptr, |
2383 | int *errorcodeptr, compile_block *cb) |
2384 | { |
2385 | PCRE2_SPTR ptr = *ptrptr; |
2386 | BOOL is_group = (*ptr != CHAR_ASTERISK); |
2387 | |
2388 | if (++ptr >= ptrend) /* No characters in name */ |
2389 | { |
2390 | *errorcodeptr = is_group? ERR62: /* Subpattern name expected */ |
2391 | ERR60; /* Verb not recognized or malformed */ |
2392 | goto FAILED; |
2393 | } |
2394 | |
2395 | *nameptr = ptr; |
2396 | *offsetptr = (PCRE2_SIZE)(ptr - cb->start_pattern); |
2397 | |
2398 | /* In UTF mode, a group name may contain letters and decimal digits as defined |
2399 | by Unicode properties, and underscores, but must not start with a digit. */ |
2400 | |
2401 | #ifdef SUPPORT_UNICODE |
2402 | if (utf && is_group) |
2403 | { |
2404 | uint32_t c, type; |
2405 | |
2406 | GETCHAR(c, ptr); |
2407 | type = UCD_CHARTYPE(c); |
2408 | |
2409 | if (type == ucp_Nd) |
2410 | { |
2411 | *errorcodeptr = ERR44; |
2412 | goto FAILED; |
2413 | } |
2414 | |
2415 | for(;;) |
2416 | { |
2417 | if (type != ucp_Nd && PRIV(ucp_gentype)[type] != ucp_L && |
2418 | c != CHAR_UNDERSCORE) break; |
2419 | ptr++; |
2420 | FORWARDCHARTEST(ptr, ptrend); |
2421 | if (ptr >= ptrend) break; |
2422 | GETCHAR(c, ptr); |
2423 | type = UCD_CHARTYPE(c); |
2424 | } |
2425 | } |
2426 | else |
2427 | #else |
2428 | (void)utf; /* Avoid compiler warning */ |
2429 | #endif /* SUPPORT_UNICODE */ |
2430 | |
2431 | /* Handle non-group names and group names in non-UTF modes. A group name must |
2432 | not start with a digit. If either of the others start with a digit it just |
2433 | won't be recognized. */ |
2434 | |
2435 | { |
2436 | if (is_group && IS_DIGIT(*ptr)) |
2437 | { |
2438 | *errorcodeptr = ERR44; |
2439 | goto FAILED; |
2440 | } |
2441 | |
2442 | while (ptr < ptrend && MAX_255(*ptr) && (cb->ctypes[*ptr] & ctype_word) != 0) |
2443 | { |
2444 | ptr++; |
2445 | } |
2446 | } |
2447 | |
2448 | /* Check name length */ |
2449 | |
2450 | if (ptr > *nameptr + MAX_NAME_SIZE) |
2451 | { |
2452 | *errorcodeptr = ERR48; |
2453 | goto FAILED; |
2454 | } |
2455 | *namelenptr = (uint32_t)(ptr - *nameptr); |
2456 | |
2457 | /* Subpattern names must not be empty, and their terminator is checked here. |
2458 | (What follows a verb or alpha assertion name is checked separately.) */ |
2459 | |
2460 | if (is_group) |
2461 | { |
2462 | if (ptr == *nameptr) |
2463 | { |
2464 | *errorcodeptr = ERR62; /* Subpattern name expected */ |
2465 | goto FAILED; |
2466 | } |
2467 | if (ptr >= ptrend || *ptr != (PCRE2_UCHAR)terminator) |
2468 | { |
2469 | *errorcodeptr = ERR42; |
2470 | goto FAILED; |
2471 | } |
2472 | ptr++; |
2473 | } |
2474 | |
2475 | *ptrptr = ptr; |
2476 | return TRUE; |
2477 | |
2478 | FAILED: |
2479 | *ptrptr = ptr; |
2480 | return FALSE; |
2481 | } |
2482 | |
2483 | |
2484 | |
2485 | /************************************************* |
2486 | * Manage callouts at start of cycle * |
2487 | *************************************************/ |
2488 | |
2489 | /* At the start of a new item in parse_regex() we are able to record the |
2490 | details of the previous item in a prior callout, and also to set up an |
2491 | automatic callout if enabled. Avoid having two adjacent automatic callouts, |
2492 | which would otherwise happen for items such as \Q that contribute nothing to |
2493 | the parsed pattern. |
2494 | |
2495 | Arguments: |
2496 | ptr current pattern pointer |
2497 | pcalloutptr points to a pointer to previous callout, or NULL |
2498 | auto_callout TRUE if auto_callouts are enabled |
2499 | parsed_pattern the parsed pattern pointer |
2500 | cb compile block |
2501 | |
2502 | Returns: possibly updated parsed_pattern pointer. |
2503 | */ |
2504 | |
2505 | static uint32_t * |
2506 | manage_callouts(PCRE2_SPTR ptr, uint32_t **pcalloutptr, BOOL auto_callout, |
2507 | uint32_t *parsed_pattern, compile_block *cb) |
2508 | { |
2509 | uint32_t *previous_callout = *pcalloutptr; |
2510 | |
2511 | if (previous_callout != NULL) previous_callout[2] = (uint32_t)(ptr - |
2512 | cb->start_pattern - (PCRE2_SIZE)previous_callout[1]); |
2513 | |
2514 | if (!auto_callout) previous_callout = NULL; else |
2515 | { |
2516 | if (previous_callout == NULL || |
2517 | previous_callout != parsed_pattern - 4 || |
2518 | previous_callout[3] != 255) |
2519 | { |
2520 | previous_callout = parsed_pattern; /* Set up new automatic callout */ |
2521 | parsed_pattern += 4; |
2522 | previous_callout[0] = META_CALLOUT_NUMBER; |
2523 | previous_callout[2] = 0; |
2524 | previous_callout[3] = 255; |
2525 | } |
2526 | previous_callout[1] = (uint32_t)(ptr - cb->start_pattern); |
2527 | } |
2528 | |
2529 | *pcalloutptr = previous_callout; |
2530 | return parsed_pattern; |
2531 | } |
2532 | |
2533 | |
2534 | |
2535 | /************************************************* |
2536 | * Parse regex and identify named groups * |
2537 | *************************************************/ |
2538 | |
2539 | /* This function is called first of all. It scans the pattern and does two |
2540 | things: (1) It identifies capturing groups and makes a table of named capturing |
2541 | groups so that information about them is fully available to both the compiling |
2542 | scans. (2) It writes a parsed version of the pattern with comments omitted and |
2543 | escapes processed into the parsed_pattern vector. |
2544 | |
2545 | Arguments: |
2546 | ptr points to the start of the pattern |
2547 | options compiling dynamic options (may change during the scan) |
2548 | has_lookbehind points to a boolean, set TRUE if a lookbehind is found |
2549 | cb pointer to the compile data block |
2550 | |
2551 | Returns: zero on success or a non-zero error code, with the |
2552 | error offset placed in the cb field |
2553 | */ |
2554 | |
2555 | /* A structure and some flags for dealing with nested groups. */ |
2556 | |
2557 | typedef struct nest_save { |
2558 | uint16_t nest_depth; |
2559 | uint16_t reset_group; |
2560 | uint16_t max_group; |
2561 | uint16_t flags; |
2562 | uint32_t options; |
2563 | } nest_save; |
2564 | |
2565 | #define NSF_RESET 0x0001u |
2566 | #define NSF_CONDASSERT 0x0002u |
2567 | #define NSF_ATOMICSR 0x0004u |
2568 | |
2569 | /* Options that are changeable within the pattern must be tracked during |
2570 | parsing. Some (e.g. PCRE2_EXTENDED) are implemented entirely during parsing, |
2571 | but all must be tracked so that META_OPTIONS items set the correct values for |
2572 | the main compiling phase. */ |
2573 | |
2574 | #define PARSE_TRACKED_OPTIONS (PCRE2_CASELESS|PCRE2_DOTALL|PCRE2_DUPNAMES| \ |
2575 | PCRE2_EXTENDED|PCRE2_EXTENDED_MORE|PCRE2_MULTILINE|PCRE2_NO_AUTO_CAPTURE| \ |
2576 | PCRE2_UNGREEDY) |
2577 | |
2578 | /* States used for analyzing ranges in character classes. The two OK values |
2579 | must be last. */ |
2580 | |
2581 | enum { RANGE_NO, RANGE_STARTED, RANGE_OK_ESCAPED, RANGE_OK_LITERAL }; |
2582 | |
2583 | /* Only in 32-bit mode can there be literals > META_END. A macro encapsulates |
2584 | the storing of literal values in the main parsed pattern, where they can always |
2585 | be quantified. */ |
2586 | |
2587 | #if PCRE2_CODE_UNIT_WIDTH == 32 |
2588 | #define PARSED_LITERAL(c, p) \ |
2589 | { \ |
2590 | if (c >= META_END) *p++ = META_BIGVALUE; \ |
2591 | *p++ = c; \ |
2592 | okquantifier = TRUE; \ |
2593 | } |
2594 | #else |
2595 | #define PARSED_LITERAL(c, p) *p++ = c; okquantifier = TRUE; |
2596 | #endif |
2597 | |
2598 | /* Here's the actual function. */ |
2599 | |
2600 | static int parse_regex(PCRE2_SPTR ptr, uint32_t options, BOOL *has_lookbehind, |
2601 | compile_block *cb) |
2602 | { |
2603 | uint32_t c; |
2604 | uint32_t delimiter; |
2605 | uint32_t namelen; |
2606 | uint32_t class_range_state; |
2607 | uint32_t *verblengthptr = NULL; /* Value avoids compiler warning */ |
2608 | uint32_t *verbstartptr = NULL; |
2609 | uint32_t *previous_callout = NULL; |
2610 | uint32_t *parsed_pattern = cb->parsed_pattern; |
2611 | uint32_t *parsed_pattern_end = cb->parsed_pattern_end; |
2612 | uint32_t meta_quantifier = 0; |
2613 | uint32_t add_after_mark = 0; |
2614 | uint32_t = cb->cx->extra_options; |
2615 | uint16_t nest_depth = 0; |
2616 | int after_manual_callout = 0; |
2617 | int expect_cond_assert = 0; |
2618 | int errorcode = 0; |
2619 | int escape; |
2620 | int i; |
2621 | BOOL inescq = FALSE; |
2622 | BOOL inverbname = FALSE; |
2623 | BOOL utf = (options & PCRE2_UTF) != 0; |
2624 | BOOL auto_callout = (options & PCRE2_AUTO_CALLOUT) != 0; |
2625 | BOOL isdupname; |
2626 | BOOL negate_class; |
2627 | BOOL okquantifier = FALSE; |
2628 | PCRE2_SPTR thisptr; |
2629 | PCRE2_SPTR name; |
2630 | PCRE2_SPTR ptrend = cb->end_pattern; |
2631 | PCRE2_SPTR verbnamestart = NULL; /* Value avoids compiler warning */ |
2632 | named_group *ng; |
2633 | nest_save *top_nest, *end_nests; |
2634 | |
2635 | /* Insert leading items for word and line matching (features provided for the |
2636 | benefit of pcre2grep). */ |
2637 | |
2638 | if ((extra_options & PCRE2_EXTRA_MATCH_LINE) != 0) |
2639 | { |
2640 | *parsed_pattern++ = META_CIRCUMFLEX; |
2641 | *parsed_pattern++ = META_NOCAPTURE; |
2642 | } |
2643 | else if ((extra_options & PCRE2_EXTRA_MATCH_WORD) != 0) |
2644 | { |
2645 | *parsed_pattern++ = META_ESCAPE + ESC_b; |
2646 | *parsed_pattern++ = META_NOCAPTURE; |
2647 | } |
2648 | |
2649 | /* If the pattern is actually a literal string, process it separately to avoid |
2650 | cluttering up the main loop. */ |
2651 | |
2652 | if ((options & PCRE2_LITERAL) != 0) |
2653 | { |
2654 | while (ptr < ptrend) |
2655 | { |
2656 | if (parsed_pattern >= parsed_pattern_end) |
2657 | { |
2658 | errorcode = ERR63; /* Internal error (parsed pattern overflow) */ |
2659 | goto FAILED; |
2660 | } |
2661 | thisptr = ptr; |
2662 | GETCHARINCTEST(c, ptr); |
2663 | if (auto_callout) |
2664 | parsed_pattern = manage_callouts(thisptr, &previous_callout, |
2665 | auto_callout, parsed_pattern, cb); |
2666 | PARSED_LITERAL(c, parsed_pattern); |
2667 | } |
2668 | goto PARSED_END; |
2669 | } |
2670 | |
2671 | /* Process a real regex which may contain meta-characters. */ |
2672 | |
2673 | top_nest = NULL; |
2674 | end_nests = (nest_save *)(cb->start_workspace + cb->workspace_size); |
2675 | |
2676 | /* The size of the nest_save structure might not be a factor of the size of the |
2677 | workspace. Therefore we must round down end_nests so as to correctly avoid |
2678 | creating a nest_save that spans the end of the workspace. */ |
2679 | |
2680 | end_nests = (nest_save *)((char *)end_nests - |
2681 | ((cb->workspace_size * sizeof(PCRE2_UCHAR)) % sizeof(nest_save))); |
2682 | |
2683 | /* PCRE2_EXTENDED_MORE implies PCRE2_EXTENDED */ |
2684 | |
2685 | if ((options & PCRE2_EXTENDED_MORE) != 0) options |= PCRE2_EXTENDED; |
2686 | |
2687 | /* Now scan the pattern */ |
2688 | |
2689 | while (ptr < ptrend) |
2690 | { |
2691 | int prev_expect_cond_assert; |
2692 | uint32_t min_repeat = 0, max_repeat = 0; |
2693 | uint32_t set, unset, *optset; |
2694 | uint32_t terminator; |
2695 | uint32_t prev_meta_quantifier; |
2696 | BOOL prev_okquantifier; |
2697 | PCRE2_SPTR tempptr; |
2698 | PCRE2_SIZE offset; |
2699 | |
2700 | if (parsed_pattern >= parsed_pattern_end) |
2701 | { |
2702 | errorcode = ERR63; /* Internal error (parsed pattern overflow) */ |
2703 | goto FAILED; |
2704 | } |
2705 | |
2706 | if (nest_depth > cb->cx->parens_nest_limit) |
2707 | { |
2708 | errorcode = ERR19; |
2709 | goto FAILED; /* Parentheses too deeply nested */ |
2710 | } |
2711 | |
2712 | /* Get next input character, save its position for callout handling. */ |
2713 | |
2714 | thisptr = ptr; |
2715 | GETCHARINCTEST(c, ptr); |
2716 | |
2717 | /* Copy quoted literals until \E, allowing for the possibility of automatic |
2718 | callouts, except when processing a (*VERB) "name". */ |
2719 | |
2720 | if (inescq) |
2721 | { |
2722 | if (c == CHAR_BACKSLASH && ptr < ptrend && *ptr == CHAR_E) |
2723 | { |
2724 | inescq = FALSE; |
2725 | ptr++; /* Skip E */ |
2726 | } |
2727 | else |
2728 | { |
2729 | if (expect_cond_assert > 0) /* A literal is not allowed if we are */ |
2730 | { /* expecting a conditional assertion, */ |
2731 | ptr--; /* but an empty \Q\E sequence is OK. */ |
2732 | errorcode = ERR28; |
2733 | goto FAILED; |
2734 | } |
2735 | if (inverbname) |
2736 | { /* Don't use PARSED_LITERAL() because it */ |
2737 | #if PCRE2_CODE_UNIT_WIDTH == 32 /* sets okquantifier. */ |
2738 | if (c >= META_END) *parsed_pattern++ = META_BIGVALUE; |
2739 | #endif |
2740 | *parsed_pattern++ = c; |
2741 | } |
2742 | else |
2743 | { |
2744 | if (after_manual_callout-- <= 0) |
2745 | parsed_pattern = manage_callouts(thisptr, &previous_callout, |
2746 | auto_callout, parsed_pattern, cb); |
2747 | PARSED_LITERAL(c, parsed_pattern); |
2748 | } |
2749 | meta_quantifier = 0; |
2750 | } |
2751 | continue; /* Next character */ |
2752 | } |
2753 | |
2754 | /* If we are processing the "name" part of a (*VERB:NAME) item, all |
2755 | characters up to the closing parenthesis are literals except when |
2756 | PCRE2_ALT_VERBNAMES is set. That causes backslash interpretation, but only \Q |
2757 | and \E and escaped characters are allowed (no character types such as \d). If |
2758 | PCRE2_EXTENDED is also set, we must ignore white space and # comments. Do |
2759 | this by not entering the special (*VERB:NAME) processing - they are then |
2760 | picked up below. Note that c is a character, not a code unit, so we must not |
2761 | use MAX_255 to test its size because MAX_255 tests code units and is assumed |
2762 | TRUE in 8-bit mode. */ |
2763 | |
2764 | if (inverbname && |
2765 | ( |
2766 | /* EITHER: not both options set */ |
2767 | ((options & (PCRE2_EXTENDED | PCRE2_ALT_VERBNAMES)) != |
2768 | (PCRE2_EXTENDED | PCRE2_ALT_VERBNAMES)) || |
2769 | #ifdef SUPPORT_UNICODE |
2770 | /* OR: character > 255 AND not Unicode Pattern White Space */ |
2771 | (c > 255 && (c|1) != 0x200f && (c|1) != 0x2029) || |
2772 | #endif |
2773 | /* OR: not a # comment or isspace() white space */ |
2774 | (c < 256 && c != CHAR_NUMBER_SIGN && (cb->ctypes[c] & ctype_space) == 0 |
2775 | #ifdef SUPPORT_UNICODE |
2776 | /* and not CHAR_NEL when Unicode is supported */ |
2777 | && c != CHAR_NEL |
2778 | #endif |
2779 | ))) |
2780 | { |
2781 | PCRE2_SIZE verbnamelength; |
2782 | |
2783 | switch(c) |
2784 | { |
2785 | default: /* Don't use PARSED_LITERAL() because it */ |
2786 | #if PCRE2_CODE_UNIT_WIDTH == 32 /* sets okquantifier. */ |
2787 | if (c >= META_END) *parsed_pattern++ = META_BIGVALUE; |
2788 | #endif |
2789 | *parsed_pattern++ = c; |
2790 | break; |
2791 | |
2792 | case CHAR_RIGHT_PARENTHESIS: |
2793 | inverbname = FALSE; |
2794 | /* This is the length in characters */ |
2795 | verbnamelength = (PCRE2_SIZE)(parsed_pattern - verblengthptr - 1); |
2796 | /* But the limit on the length is in code units */ |
2797 | if (ptr - verbnamestart - 1 > (int)MAX_MARK) |
2798 | { |
2799 | ptr--; |
2800 | errorcode = ERR76; |
2801 | goto FAILED; |
2802 | } |
2803 | *verblengthptr = (uint32_t)verbnamelength; |
2804 | |
2805 | /* If this name was on a verb such as (*ACCEPT) which does not continue, |
2806 | a (*MARK) was generated for the name. We now add the original verb as the |
2807 | next item. */ |
2808 | |
2809 | if (add_after_mark != 0) |
2810 | { |
2811 | *parsed_pattern++ = add_after_mark; |
2812 | add_after_mark = 0; |
2813 | } |
2814 | break; |
2815 | |
2816 | case CHAR_BACKSLASH: |
2817 | if ((options & PCRE2_ALT_VERBNAMES) != 0) |
2818 | { |
2819 | escape = PRIV(check_escape)(&ptr, ptrend, &c, &errorcode, options, |
2820 | cb->cx->extra_options, FALSE, cb); |
2821 | if (errorcode != 0) goto FAILED; |
2822 | } |
2823 | else escape = 0; /* Treat all as literal */ |
2824 | |
2825 | switch(escape) |
2826 | { |
2827 | case 0: /* Don't use PARSED_LITERAL() because it */ |
2828 | #if PCRE2_CODE_UNIT_WIDTH == 32 /* sets okquantifier. */ |
2829 | if (c >= META_END) *parsed_pattern++ = META_BIGVALUE; |
2830 | #endif |
2831 | *parsed_pattern++ = c; |
2832 | break; |
2833 | |
2834 | case ESC_Q: |
2835 | inescq = TRUE; |
2836 | break; |
2837 | |
2838 | case ESC_E: /* Ignore */ |
2839 | break; |
2840 | |
2841 | default: |
2842 | errorcode = ERR40; /* Invalid in verb name */ |
2843 | goto FAILED; |
2844 | } |
2845 | } |
2846 | continue; /* Next character in pattern */ |
2847 | } |
2848 | |
2849 | /* Not a verb name character. At this point we must process everything that |
2850 | must not change the quantification state. This is mainly comments, but we |
2851 | handle \Q and \E here as well, so that an item such as A\Q\E+ is treated as |
2852 | A+, as in Perl. An isolated \E is ignored. */ |
2853 | |
2854 | if (c == CHAR_BACKSLASH && ptr < ptrend) |
2855 | { |
2856 | if (*ptr == CHAR_Q || *ptr == CHAR_E) |
2857 | { |
2858 | inescq = *ptr == CHAR_Q; |
2859 | ptr++; |
2860 | continue; |
2861 | } |
2862 | } |
2863 | |
2864 | /* Skip over whitespace and # comments in extended mode. Note that c is a |
2865 | character, not a code unit, so we must not use MAX_255 to test its size |
2866 | because MAX_255 tests code units and is assumed TRUE in 8-bit mode. The |
2867 | whitespace characters are those designated as "Pattern White Space" by |
2868 | Unicode, which are the isspace() characters plus CHAR_NEL (newline), which is |
2869 | U+0085 in Unicode, plus U+200E, U+200F, U+2028, and U+2029. These are a |
2870 | subset of space characters that match \h and \v. */ |
2871 | |
2872 | if ((options & PCRE2_EXTENDED) != 0) |
2873 | { |
2874 | if (c < 256 && (cb->ctypes[c] & ctype_space) != 0) continue; |
2875 | #ifdef SUPPORT_UNICODE |
2876 | if (c == CHAR_NEL || (c|1) == 0x200f || (c|1) == 0x2029) continue; |
2877 | #endif |
2878 | if (c == CHAR_NUMBER_SIGN) |
2879 | { |
2880 | while (ptr < ptrend) |
2881 | { |
2882 | if (IS_NEWLINE(ptr)) /* For non-fixed-length newline cases, */ |
2883 | { /* IS_NEWLINE sets cb->nllen. */ |
2884 | ptr += cb->nllen; |
2885 | break; |
2886 | } |
2887 | ptr++; |
2888 | #ifdef SUPPORT_UNICODE |
2889 | if (utf) FORWARDCHARTEST(ptr, ptrend); |
2890 | #endif |
2891 | } |
2892 | continue; /* Next character in pattern */ |
2893 | } |
2894 | } |
2895 | |
2896 | /* Skip over bracketed comments */ |
2897 | |
2898 | if (c == CHAR_LEFT_PARENTHESIS && ptrend - ptr >= 2 && |
2899 | ptr[0] == CHAR_QUESTION_MARK && ptr[1] == CHAR_NUMBER_SIGN) |
2900 | { |
2901 | while (++ptr < ptrend && *ptr != CHAR_RIGHT_PARENTHESIS); |
2902 | if (ptr >= ptrend) |
2903 | { |
2904 | errorcode = ERR18; /* A special error for missing ) in a comment */ |
2905 | goto FAILED; /* to make it easier to debug. */ |
2906 | } |
2907 | ptr++; |
2908 | continue; /* Next character in pattern */ |
2909 | } |
2910 | |
2911 | /* If the next item is not a quantifier, fill in length of any previous |
2912 | callout and create an auto callout if required. */ |
2913 | |
2914 | if (c != CHAR_ASTERISK && c != CHAR_PLUS && c != CHAR_QUESTION_MARK && |
2915 | (c != CHAR_LEFT_CURLY_BRACKET || |
2916 | (tempptr = ptr, |
2917 | !read_repeat_counts(&tempptr, ptrend, NULL, NULL, &errorcode)))) |
2918 | { |
2919 | if (after_manual_callout-- <= 0) |
2920 | parsed_pattern = manage_callouts(thisptr, &previous_callout, auto_callout, |
2921 | parsed_pattern, cb); |
2922 | } |
2923 | |
2924 | /* If expect_cond_assert is 2, we have just passed (?( and are expecting an |
2925 | assertion, possibly preceded by a callout. If the value is 1, we have just |
2926 | had the callout and expect an assertion. There must be at least 3 more |
2927 | characters in all cases. When expect_cond_assert is 2, we know that the |
2928 | current character is an opening parenthesis, as otherwise we wouldn't be |
2929 | here. However, when it is 1, we need to check, and it's easiest just to check |
2930 | always. Note that expect_cond_assert may be negative, since all callouts just |
2931 | decrement it. */ |
2932 | |
2933 | if (expect_cond_assert > 0) |
2934 | { |
2935 | BOOL ok = c == CHAR_LEFT_PARENTHESIS && ptrend - ptr >= 3 && |
2936 | (ptr[0] == CHAR_QUESTION_MARK || ptr[0] == CHAR_ASTERISK); |
2937 | if (ok) |
2938 | { |
2939 | if (ptr[0] == CHAR_ASTERISK) /* New alpha assertion format, possibly */ |
2940 | { |
2941 | ok = MAX_255(ptr[1]) && (cb->ctypes[ptr[1]] & ctype_lcletter) != 0; |
2942 | } |
2943 | else switch(ptr[1]) /* Traditional symbolic format */ |
2944 | { |
2945 | case CHAR_C: |
2946 | ok = expect_cond_assert == 2; |
2947 | break; |
2948 | |
2949 | case CHAR_EQUALS_SIGN: |
2950 | case CHAR_EXCLAMATION_MARK: |
2951 | break; |
2952 | |
2953 | case CHAR_LESS_THAN_SIGN: |
2954 | ok = ptr[2] == CHAR_EQUALS_SIGN || ptr[2] == CHAR_EXCLAMATION_MARK; |
2955 | break; |
2956 | |
2957 | default: |
2958 | ok = FALSE; |
2959 | } |
2960 | } |
2961 | |
2962 | if (!ok) |
2963 | { |
2964 | ptr--; /* Adjust error offset */ |
2965 | errorcode = ERR28; |
2966 | goto FAILED; |
2967 | } |
2968 | } |
2969 | |
2970 | /* Remember whether we are expecting a conditional assertion, and set the |
2971 | default for this item. */ |
2972 | |
2973 | prev_expect_cond_assert = expect_cond_assert; |
2974 | expect_cond_assert = 0; |
2975 | |
2976 | /* Remember quantification status for the previous significant item, then set |
2977 | default for this item. */ |
2978 | |
2979 | prev_okquantifier = okquantifier; |
2980 | prev_meta_quantifier = meta_quantifier; |
2981 | okquantifier = FALSE; |
2982 | meta_quantifier = 0; |
2983 | |
2984 | /* If the previous significant item was a quantifier, adjust the parsed code |
2985 | if there is a following modifier. The base meta value is always followed by |
2986 | the PLUS and QUERY values, in that order. We do this here rather than after |
2987 | reading a quantifier so that intervening comments and /x whitespace can be |
2988 | ignored without having to replicate code. */ |
2989 | |
2990 | if (prev_meta_quantifier != 0 && (c == CHAR_QUESTION_MARK || c == CHAR_PLUS)) |
2991 | { |
2992 | parsed_pattern[(prev_meta_quantifier == META_MINMAX)? -3 : -1] = |
2993 | prev_meta_quantifier + ((c == CHAR_QUESTION_MARK)? |
2994 | 0x00020000u : 0x00010000u); |
2995 | continue; /* Next character in pattern */ |
2996 | } |
2997 | |
2998 | |
2999 | /* Process the next item in the main part of a pattern. */ |
3000 | |
3001 | switch(c) |
3002 | { |
3003 | default: /* Non-special character */ |
3004 | PARSED_LITERAL(c, parsed_pattern); |
3005 | break; |
3006 | |
3007 | |
3008 | /* ---- Escape sequence ---- */ |
3009 | |
3010 | case CHAR_BACKSLASH: |
3011 | tempptr = ptr; |
3012 | escape = PRIV(check_escape)(&ptr, ptrend, &c, &errorcode, options, |
3013 | cb->cx->extra_options, FALSE, cb); |
3014 | if (errorcode != 0) |
3015 | { |
3016 | ESCAPE_FAILED: |
3017 | if ((extra_options & PCRE2_EXTRA_BAD_ESCAPE_IS_LITERAL) == 0) |
3018 | goto FAILED; |
3019 | ptr = tempptr; |
3020 | if (ptr >= ptrend) c = CHAR_BACKSLASH; else |
3021 | { |
3022 | GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */ |
3023 | } |
3024 | escape = 0; /* Treat as literal character */ |
3025 | } |
3026 | |
3027 | /* The escape was a data escape or literal character. */ |
3028 | |
3029 | if (escape == 0) |
3030 | { |
3031 | PARSED_LITERAL(c, parsed_pattern); |
3032 | } |
3033 | |
3034 | /* The escape was a back (or forward) reference. We keep the offset in |
3035 | order to give a more useful diagnostic for a bad forward reference. For |
3036 | references to groups numbered less than 10 we can't use more than two items |
3037 | in parsed_pattern because they may be just two characters in the input (and |
3038 | in a 64-bit world an offset may need two elements). So for them, the offset |
3039 | of the first occurrent is held in a special vector. */ |
3040 | |
3041 | else if (escape < 0) |
3042 | { |
3043 | offset = (PCRE2_SIZE)(ptr - cb->start_pattern - 1); |
3044 | escape = -escape; |
3045 | *parsed_pattern++ = META_BACKREF | (uint32_t)escape; |
3046 | if (escape < 10) |
3047 | { |
3048 | if (cb->small_ref_offset[escape] == PCRE2_UNSET) |
3049 | cb->small_ref_offset[escape] = offset; |
3050 | } |
3051 | else |
3052 | { |
3053 | PUTOFFSET(offset, parsed_pattern); |
3054 | } |
3055 | okquantifier = TRUE; |
3056 | } |
3057 | |
3058 | /* The escape was a character class such as \d etc. or other special |
3059 | escape indicator such as \A or \X. Most of them generate just a single |
3060 | parsed item, but \P and \p are followed by a 16-bit type and a 16-bit |
3061 | value. They are supported only when Unicode is available. The type and |
3062 | value are packed into a single 32-bit value so that the whole sequences |
3063 | uses only two elements in the parsed_vector. This is because the same |
3064 | coding is used if \d (for example) is turned into \p{Nd} when PCRE2_UCP is |
3065 | set. |
3066 | |
3067 | There are also some cases where the escape sequence is followed by a name: |
3068 | \k{name}, \k<name>, and \k'name' are backreferences by name, and \g<name> |
3069 | and \g'name' are subroutine calls by name; \g{name} is a synonym for |
3070 | \k{name}. Note that \g<number> and \g'number' are handled by check_escape() |
3071 | and returned as a negative value (handled above). A name is coded as an |
3072 | offset into the pattern and a length. */ |
3073 | |
3074 | else switch (escape) |
3075 | { |
3076 | case ESC_C: |
3077 | #ifdef NEVER_BACKSLASH_C |
3078 | errorcode = ERR85; |
3079 | goto ESCAPE_FAILED; |
3080 | #else |
3081 | if ((options & PCRE2_NEVER_BACKSLASH_C) != 0) |
3082 | { |
3083 | errorcode = ERR83; |
3084 | goto ESCAPE_FAILED; |
3085 | } |
3086 | #endif |
3087 | okquantifier = TRUE; |
3088 | *parsed_pattern++ = META_ESCAPE + escape; |
3089 | break; |
3090 | |
3091 | case ESC_X: |
3092 | #ifndef SUPPORT_UNICODE |
3093 | errorcode = ERR45; /* Supported only with Unicode support */ |
3094 | goto ESCAPE_FAILED; |
3095 | #endif |
3096 | case ESC_H: |
3097 | case ESC_h: |
3098 | case ESC_N: |
3099 | case ESC_R: |
3100 | case ESC_V: |
3101 | case ESC_v: |
3102 | okquantifier = TRUE; |
3103 | *parsed_pattern++ = META_ESCAPE + escape; |
3104 | break; |
3105 | |
3106 | default: /* \A, \B, \b, \G, \K, \Z, \z cannot be quantified. */ |
3107 | *parsed_pattern++ = META_ESCAPE + escape; |
3108 | break; |
3109 | |
3110 | /* Escapes that change in UCP mode. Note that PCRE2_UCP will never be set |
3111 | without Unicode support because it is checked when pcre2_compile() is |
3112 | called. */ |
3113 | |
3114 | case ESC_d: |
3115 | case ESC_D: |
3116 | case ESC_s: |
3117 | case ESC_S: |
3118 | case ESC_w: |
3119 | case ESC_W: |
3120 | okquantifier = TRUE; |
3121 | if ((options & PCRE2_UCP) == 0) |
3122 | { |
3123 | *parsed_pattern++ = META_ESCAPE + escape; |
3124 | } |
3125 | else |
3126 | { |
3127 | *parsed_pattern++ = META_ESCAPE + |
3128 | ((escape == ESC_d || escape == ESC_s || escape == ESC_w)? |
3129 | ESC_p : ESC_P); |
3130 | switch(escape) |
3131 | { |
3132 | case ESC_d: |
3133 | case ESC_D: |
3134 | *parsed_pattern++ = (PT_PC << 16) | ucp_Nd; |
3135 | break; |
3136 | |
3137 | case ESC_s: |
3138 | case ESC_S: |
3139 | *parsed_pattern++ = PT_SPACE << 16; |
3140 | break; |
3141 | |
3142 | case ESC_w: |
3143 | case ESC_W: |
3144 | *parsed_pattern++ = PT_WORD << 16; |
3145 | break; |
3146 | } |
3147 | } |
3148 | break; |
3149 | |
3150 | /* Unicode property matching */ |
3151 | |
3152 | case ESC_P: |
3153 | case ESC_p: |
3154 | #ifdef SUPPORT_UNICODE |
3155 | { |
3156 | BOOL negated; |
3157 | uint16_t ptype = 0, pdata = 0; |
3158 | if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcode, cb)) |
3159 | goto ESCAPE_FAILED; |
3160 | if (negated) escape = (escape == ESC_P)? ESC_p : ESC_P; |
3161 | *parsed_pattern++ = META_ESCAPE + escape; |
3162 | *parsed_pattern++ = (ptype << 16) | pdata; |
3163 | okquantifier = TRUE; |
3164 | } |
3165 | #else |
3166 | errorcode = ERR45; |
3167 | goto ESCAPE_FAILED; |
3168 | #endif |
3169 | break; /* End \P and \p */ |
3170 | |
3171 | /* When \g is used with quotes or angle brackets as delimiters, it is a |
3172 | numerical or named subroutine call, and control comes here. When used |
3173 | with brace delimiters it is a numberical back reference and does not come |
3174 | here because check_escape() returns it directly as a reference. \k is |
3175 | always a named back reference. */ |
3176 | |
3177 | case ESC_g: |
3178 | case ESC_k: |
3179 | if (ptr >= ptrend || (*ptr != CHAR_LEFT_CURLY_BRACKET && |
3180 | *ptr != CHAR_LESS_THAN_SIGN && *ptr != CHAR_APOSTROPHE)) |
3181 | { |
3182 | errorcode = (escape == ESC_g)? ERR57 : ERR69; |
3183 | goto ESCAPE_FAILED; |
3184 | } |
3185 | terminator = (*ptr == CHAR_LESS_THAN_SIGN)? |
3186 | CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)? |
3187 | CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET; |
3188 | |
3189 | /* For a non-braced \g, check for a numerical recursion. */ |
3190 | |
3191 | if (escape == ESC_g && terminator != CHAR_RIGHT_CURLY_BRACKET) |
3192 | { |
3193 | PCRE2_SPTR p = ptr + 1; |
3194 | |
3195 | if (read_number(&p, ptrend, cb->bracount, MAX_GROUP_NUMBER, ERR61, &i, |
3196 | &errorcode)) |
3197 | { |
3198 | if (p >= ptrend || *p != terminator) |
3199 | { |
3200 | errorcode = ERR57; |
3201 | goto ESCAPE_FAILED; |
3202 | } |
3203 | ptr = p; |
3204 | goto SET_RECURSION; |
3205 | } |
3206 | if (errorcode != 0) goto ESCAPE_FAILED; |
3207 | } |
3208 | |
3209 | /* Not a numerical recursion */ |
3210 | |
3211 | if (!read_name(&ptr, ptrend, utf, terminator, &offset, &name, &namelen, |
3212 | &errorcode, cb)) goto ESCAPE_FAILED; |
3213 | |
3214 | /* \k and \g when used with braces are back references, whereas \g used |
3215 | with quotes or angle brackets is a recursion */ |
3216 | |
3217 | *parsed_pattern++ = |
3218 | (escape == ESC_k || terminator == CHAR_RIGHT_CURLY_BRACKET)? |
3219 | META_BACKREF_BYNAME : META_RECURSE_BYNAME; |
3220 | *parsed_pattern++ = namelen; |
3221 | |
3222 | PUTOFFSET(offset, parsed_pattern); |
3223 | okquantifier = TRUE; |
3224 | break; /* End special escape processing */ |
3225 | } |
3226 | break; /* End escape sequence processing */ |
3227 | |
3228 | |
3229 | /* ---- Single-character special items ---- */ |
3230 | |
3231 | case CHAR_CIRCUMFLEX_ACCENT: |
3232 | *parsed_pattern++ = META_CIRCUMFLEX; |
3233 | break; |
3234 | |
3235 | case CHAR_DOLLAR_SIGN: |
3236 | *parsed_pattern++ = META_DOLLAR; |
3237 | break; |
3238 | |
3239 | case CHAR_DOT: |
3240 | *parsed_pattern++ = META_DOT; |
3241 | okquantifier = TRUE; |
3242 | break; |
3243 | |
3244 | |
3245 | /* ---- Single-character quantifiers ---- */ |
3246 | |
3247 | case CHAR_ASTERISK: |
3248 | meta_quantifier = META_ASTERISK; |
3249 | goto CHECK_QUANTIFIER; |
3250 | |
3251 | case CHAR_PLUS: |
3252 | meta_quantifier = META_PLUS; |
3253 | goto CHECK_QUANTIFIER; |
3254 | |
3255 | case CHAR_QUESTION_MARK: |
3256 | meta_quantifier = META_QUERY; |
3257 | goto CHECK_QUANTIFIER; |
3258 | |
3259 | |
3260 | /* ---- Potential {n,m} quantifier ---- */ |
3261 | |
3262 | case CHAR_LEFT_CURLY_BRACKET: |
3263 | if (!read_repeat_counts(&ptr, ptrend, &min_repeat, &max_repeat, |
3264 | &errorcode)) |
3265 | { |
3266 | if (errorcode != 0) goto FAILED; /* Error in quantifier. */ |
3267 | PARSED_LITERAL(c, parsed_pattern); /* Not a quantifier */ |
3268 | break; /* No more quantifier processing */ |
3269 | } |
3270 | meta_quantifier = META_MINMAX; |
3271 | /* Fall through */ |
3272 | |
3273 | |
3274 | /* ---- Quantifier post-processing ---- */ |
3275 | |
3276 | /* Check that a quantifier is allowed after the previous item. */ |
3277 | |
3278 | CHECK_QUANTIFIER: |
3279 | if (!prev_okquantifier) |
3280 | { |
3281 | errorcode = ERR9; |
3282 | goto FAILED_BACK; |
3283 | } |
3284 | |
3285 | /* Most (*VERB)s are not allowed to be quantified, but an ungreedy |
3286 | quantifier can be useful for (*ACCEPT) - meaning "succeed on backtrack", a |
3287 | sort of negated (*COMMIT). We therefore allow (*ACCEPT) to be quantified by |
3288 | wrapping it in non-capturing brackets, but we have to allow for a preceding |
3289 | (*MARK) for when (*ACCEPT) has an argument. */ |
3290 | |
3291 | if (parsed_pattern[-1] == META_ACCEPT) |
3292 | { |
3293 | uint32_t *p; |
3294 | for (p = parsed_pattern - 1; p >= verbstartptr; p--) p[1] = p[0]; |
3295 | *verbstartptr = META_NOCAPTURE; |
3296 | parsed_pattern[1] = META_KET; |
3297 | parsed_pattern += 2; |
3298 | } |
3299 | |
3300 | /* Now we can put the quantifier into the parsed pattern vector. At this |
3301 | stage, we have only the basic quantifier. The check for a following + or ? |
3302 | modifier happens at the top of the loop, after any intervening comments |
3303 | have been removed. */ |
3304 | |
3305 | *parsed_pattern++ = meta_quantifier; |
3306 | if (c == CHAR_LEFT_CURLY_BRACKET) |
3307 | { |
3308 | *parsed_pattern++ = min_repeat; |
3309 | *parsed_pattern++ = max_repeat; |
3310 | } |
3311 | break; |
3312 | |
3313 | |
3314 | /* ---- Character class ---- */ |
3315 | |
3316 | case CHAR_LEFT_SQUARE_BRACKET: |
3317 | okquantifier = TRUE; |
3318 | |
3319 | /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is |
3320 | used for "start of word" and "end of word". As these are otherwise illegal |
3321 | sequences, we don't break anything by recognizing them. They are replaced |
3322 | by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are |
3323 | erroneous and are handled by the normal code below. */ |
3324 | |
3325 | if (ptrend - ptr >= 6 && |
3326 | (PRIV(strncmp_c8)(ptr, STRING_WEIRD_STARTWORD, 6) == 0 || |
3327 | PRIV(strncmp_c8)(ptr, STRING_WEIRD_ENDWORD, 6) == 0)) |
3328 | { |
3329 | *parsed_pattern++ = META_ESCAPE + ESC_b; |
3330 | |
3331 | if (ptr[2] == CHAR_LESS_THAN_SIGN) |
3332 | { |
3333 | *parsed_pattern++ = META_LOOKAHEAD; |
3334 | } |
3335 | else |
3336 | { |
3337 | *parsed_pattern++ = META_LOOKBEHIND; |
3338 | *has_lookbehind = TRUE; |
3339 | |
3340 | /* The offset is used only for the "non-fixed length" error; this won't |
3341 | occur here, so just store zero. */ |
3342 | |
3343 | PUTOFFSET((PCRE2_SIZE)0, parsed_pattern); |
3344 | } |
3345 | |
3346 | if ((options & PCRE2_UCP) == 0) |
3347 | *parsed_pattern++ = META_ESCAPE + ESC_w; |
3348 | else |
3349 | { |
3350 | *parsed_pattern++ = META_ESCAPE + ESC_p; |
3351 | *parsed_pattern++ = PT_WORD << 16; |
3352 | } |
3353 | *parsed_pattern++ = META_KET; |
3354 | ptr += 6; |
3355 | break; |
3356 | } |
3357 | |
3358 | /* PCRE supports POSIX class stuff inside a class. Perl gives an error if |
3359 | they are encountered at the top level, so we'll do that too. */ |
3360 | |
3361 | if (ptr < ptrend && (*ptr == CHAR_COLON || *ptr == CHAR_DOT || |
3362 | *ptr == CHAR_EQUALS_SIGN) && |
3363 | check_posix_syntax(ptr, ptrend, &tempptr)) |
3364 | { |
3365 | errorcode = (*ptr-- == CHAR_COLON)? ERR12 : ERR13; |
3366 | goto FAILED; |
3367 | } |
3368 | |
3369 | /* Process a regular character class. If the first character is '^', set |
3370 | the negation flag. If the first few characters (either before or after ^) |
3371 | are \Q\E or \E or space or tab in extended-more mode, we skip them too. |
3372 | This makes for compatibility with Perl. */ |
3373 | |
3374 | negate_class = FALSE; |
3375 | while (ptr < ptrend) |
3376 | { |
3377 | GETCHARINCTEST(c, ptr); |
3378 | if (c == CHAR_BACKSLASH) |
3379 | { |
3380 | if (ptr < ptrend && *ptr == CHAR_E) ptr++; |
3381 | else if (ptrend - ptr >= 3 && |
3382 | PRIV(strncmp_c8)(ptr, STR_Q STR_BACKSLASH STR_E, 3) == 0) |
3383 | ptr += 3; |
3384 | else |
3385 | break; |
3386 | } |
3387 | else if ((options & PCRE2_EXTENDED_MORE) != 0 && |
3388 | (c == CHAR_SPACE || c == CHAR_HT)) /* Note: just these two */ |
3389 | continue; |
3390 | else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT) |
3391 | negate_class = TRUE; |
3392 | else break; |
3393 | } |
3394 | |
3395 | /* Now the real contents of the class; c has the first "real" character. |
3396 | Empty classes are permitted only if the option is set. */ |
3397 | |
3398 | if (c == CHAR_RIGHT_SQUARE_BRACKET && |
3399 | (cb->external_options & PCRE2_ALLOW_EMPTY_CLASS) != 0) |
3400 | { |
3401 | *parsed_pattern++ = negate_class? META_CLASS_EMPTY_NOT : META_CLASS_EMPTY; |
3402 | break; /* End of class processing */ |
3403 | } |
3404 | |
3405 | /* Process a non-empty class. */ |
3406 | |
3407 | *parsed_pattern++ = negate_class? META_CLASS_NOT : META_CLASS; |
3408 | class_range_state = RANGE_NO; |
3409 | |
3410 | /* In an EBCDIC environment, Perl treats alphabetic ranges specially |
3411 | because there are holes in the encoding, and simply using the range A-Z |
3412 | (for example) would include the characters in the holes. This applies only |
3413 | to ranges where both values are literal; [\xC1-\xE9] is different to [A-Z] |
3414 | in this respect. In order to accommodate this, we keep track of whether |
3415 | character values are literal or not, and a state variable for handling |
3416 | ranges. */ |
3417 | |
3418 | /* Loop for the contents of the class */ |
3419 | |
3420 | for (;;) |
3421 | { |
3422 | BOOL char_is_literal = TRUE; |
3423 | |
3424 | /* Inside \Q...\E everything is literal except \E */ |
3425 | |
3426 | if (inescq) |
3427 | { |
3428 | if (c == CHAR_BACKSLASH && ptr < ptrend && *ptr == CHAR_E) |
3429 | { |
3430 | inescq = FALSE; /* Reset literal state */ |
3431 | ptr++; /* Skip the 'E' */ |
3432 | goto CLASS_CONTINUE; |
3433 | } |
3434 | goto CLASS_LITERAL; |
3435 | } |
3436 | |
3437 | /* Skip over space and tab (only) in extended-more mode. */ |
3438 | |
3439 | if ((options & PCRE2_EXTENDED_MORE) != 0 && |
3440 | (c == CHAR_SPACE || c == CHAR_HT)) |
3441 | goto CLASS_CONTINUE; |
3442 | |
3443 | /* Handle POSIX class names. Perl allows a negation extension of the |
3444 | form [:^name:]. A square bracket that doesn't match the syntax is |
3445 | treated as a literal. We also recognize the POSIX constructions |
3446 | [.ch.] and [=ch=] ("collating elements") and fault them, as Perl |
3447 | 5.6 and 5.8 do. */ |
3448 | |
3449 | if (c == CHAR_LEFT_SQUARE_BRACKET && |
3450 | ptrend - ptr >= 3 && |
3451 | (*ptr == CHAR_COLON || *ptr == CHAR_DOT || |
3452 | *ptr == CHAR_EQUALS_SIGN) && |
3453 | check_posix_syntax(ptr, ptrend, &tempptr)) |
3454 | { |
3455 | BOOL posix_negate = FALSE; |
3456 | int posix_class; |
3457 | |
3458 | /* Perl treats a hyphen before a POSIX class as a literal, not the |
3459 | start of a range. However, it gives a warning in its warning mode. PCRE |
3460 | does not have a warning mode, so we give an error, because this is |
3461 | likely an error on the user's part. */ |
3462 | |
3463 | if (class_range_state == RANGE_STARTED) |
3464 | { |
3465 | errorcode = ERR50; |
3466 | goto FAILED; |
3467 | } |
3468 | |
3469 | if (*ptr != CHAR_COLON) |
3470 | { |
3471 | errorcode = ERR13; |
3472 | goto FAILED_BACK; |
3473 | } |
3474 | |
3475 | if (*(++ptr) == CHAR_CIRCUMFLEX_ACCENT) |
3476 | { |
3477 | posix_negate = TRUE; |
3478 | ptr++; |
3479 | } |
3480 | |
3481 | posix_class = check_posix_name(ptr, (int)(tempptr - ptr)); |
3482 | if (posix_class < 0) |
3483 | { |
3484 | errorcode = ERR30; |
3485 | goto FAILED; |
3486 | } |
3487 | ptr = tempptr + 2; |
3488 | |
3489 | /* Perl treats a hyphen after a POSIX class as a literal, not the |
3490 | start of a range. However, it gives a warning in its warning mode |
3491 | unless the hyphen is the last character in the class. PCRE does not |
3492 | have a warning mode, so we give an error, because this is likely an |
3493 | error on the user's part. */ |
3494 | |
3495 | if (ptr < ptrend - 1 && *ptr == CHAR_MINUS && |
3496 | ptr[1] != CHAR_RIGHT_SQUARE_BRACKET) |
3497 | { |
3498 | errorcode = ERR50; |
3499 | goto FAILED; |
3500 | } |
3501 | |
3502 | /* Set "a hyphen is not the start of a range" for the -] case, and also |
3503 | in case the POSIX class is followed by \E or \Q\E (possibly repeated - |
3504 | fuzzers do that kind of thing) and *then* a hyphen. This causes that |
3505 | hyphen to be treated as a literal. I don't think it's worth setting up |
3506 | special apparatus to do otherwise. */ |
3507 | |
3508 | class_range_state = RANGE_NO; |
3509 | |
3510 | /* When PCRE2_UCP is set, some of the POSIX classes are converted to |
3511 | use Unicode properties \p or \P or, in one case, \h or \H. The |
3512 | substitutes table has two values per class, containing the type and |
3513 | value of a \p or \P item. The special cases are specified with a |
3514 | negative type: a non-zero value causes \h or \H to be used, and a zero |
3515 | value falls through to behave like a non-UCP POSIX class. */ |
3516 | |
3517 | #ifdef SUPPORT_UNICODE |
3518 | if ((options & PCRE2_UCP) != 0) |
3519 | { |
3520 | int ptype = posix_substitutes[2*posix_class]; |
3521 | int pvalue = posix_substitutes[2*posix_class + 1]; |
3522 | if (ptype >= 0) |
3523 | { |
3524 | *parsed_pattern++ = META_ESCAPE + (posix_negate? ESC_P : ESC_p); |
3525 | *parsed_pattern++ = (ptype << 16) | pvalue; |
3526 | goto CLASS_CONTINUE; |
3527 | } |
3528 | |
3529 | if (pvalue != 0) |
3530 | { |
3531 | *parsed_pattern++ = META_ESCAPE + (posix_negate? ESC_H : ESC_h); |
3532 | goto CLASS_CONTINUE; |
3533 | } |
3534 | |
3535 | /* Fall through */ |
3536 | } |
3537 | #endif /* SUPPORT_UNICODE */ |
3538 | |
3539 | /* Non-UCP POSIX class */ |
3540 | |
3541 | *parsed_pattern++ = posix_negate? META_POSIX_NEG : META_POSIX; |
3542 | *parsed_pattern++ = posix_class; |
3543 | } |
3544 | |
3545 | /* Handle potential start of range */ |
3546 | |
3547 | else if (c == CHAR_MINUS && class_range_state >= RANGE_OK_ESCAPED) |
3548 | { |
3549 | *parsed_pattern++ = (class_range_state == RANGE_OK_LITERAL)? |
3550 | META_RANGE_LITERAL : META_RANGE_ESCAPED; |
3551 | class_range_state = RANGE_STARTED; |
3552 | } |
3553 | |
3554 | /* Handle a literal character */ |
3555 | |
3556 | else if (c != CHAR_BACKSLASH) |
3557 | { |
3558 | CLASS_LITERAL: |
3559 | if (class_range_state == RANGE_STARTED) |
3560 | { |
3561 | if (c == parsed_pattern[-2]) /* Optimize one-char range */ |
3562 | parsed_pattern--; |
3563 | else if (parsed_pattern[-2] > c) /* Check range is in order */ |
3564 | { |
3565 | errorcode = ERR8; |
3566 | goto FAILED_BACK; |
3567 | } |
3568 | else |
3569 | { |
3570 | if (!char_is_literal && parsed_pattern[-1] == META_RANGE_LITERAL) |
3571 | parsed_pattern[-1] = META_RANGE_ESCAPED; |
3572 | PARSED_LITERAL(c, parsed_pattern); |
3573 | } |
3574 | class_range_state = RANGE_NO; |
3575 | } |
3576 | else /* Potential start of range */ |
3577 | { |
3578 | class_range_state = char_is_literal? |
3579 | RANGE_OK_LITERAL : RANGE_OK_ESCAPED; |
3580 | PARSED_LITERAL(c, parsed_pattern); |
3581 | } |
3582 | } |
3583 | |
3584 | /* Handle escapes in a class */ |
3585 | |
3586 | else |
3587 | { |
3588 | tempptr = ptr; |
3589 | escape = PRIV(check_escape)(&ptr, ptrend, &c, &errorcode, options, |
3590 | cb->cx->extra_options, TRUE, cb); |
3591 | |
3592 | if (errorcode != 0) |
3593 | { |
3594 | if ((extra_options & PCRE2_EXTRA_BAD_ESCAPE_IS_LITERAL) == 0) |
3595 | goto FAILED; |
3596 | ptr = tempptr; |
3597 | if (ptr >= ptrend) c = CHAR_BACKSLASH; else |
3598 | { |
3599 | GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */ |
3600 | } |
3601 | escape = 0; /* Treat as literal character */ |
3602 | } |
3603 | |
3604 | switch(escape) |
3605 | { |
3606 | case 0: /* Escaped character code point is in c */ |
3607 | char_is_literal = FALSE; |
3608 | goto CLASS_LITERAL; |
3609 | |
3610 | case ESC_b: |
3611 | c = CHAR_BS; /* \b is backspace in a class */ |
3612 | char_is_literal = FALSE; |
3613 | goto CLASS_LITERAL; |
3614 | |
3615 | case ESC_Q: |
3616 | inescq = TRUE; /* Enter literal mode */ |
3617 | goto CLASS_CONTINUE; |
3618 | |
3619 | case ESC_E: /* Ignore orphan \E */ |
3620 | goto CLASS_CONTINUE; |
3621 | |
3622 | case ESC_B: /* Always an error in a class */ |
3623 | case ESC_R: |
3624 | case ESC_X: |
3625 | errorcode = ERR7; |
3626 | ptr--; |
3627 | goto FAILED; |
3628 | } |
3629 | |
3630 | /* The second part of a range can be a single-character escape |
3631 | sequence (detected above), but not any of the other escapes. Perl |
3632 | treats a hyphen as a literal in such circumstances. However, in Perl's |
3633 | warning mode, a warning is given, so PCRE now faults it, as it is |
3634 | almost certainly a mistake on the user's part. */ |
3635 | |
3636 | if (class_range_state == RANGE_STARTED) |
3637 | { |
3638 | errorcode = ERR50; |
3639 | goto FAILED; /* Not CLASS_ESCAPE_FAILED; always an error */ |
3640 | } |
3641 | |
3642 | /* Of the remaining escapes, only those that define characters are |
3643 | allowed in a class. None may start a range. */ |
3644 | |
3645 | class_range_state = RANGE_NO; |
3646 | switch(escape) |
3647 | { |
3648 | case ESC_N: |
3649 | errorcode = ERR71; |
3650 | goto FAILED; |
3651 | |
3652 | case ESC_H: |
3653 | case ESC_h: |
3654 | case ESC_V: |
3655 | case ESC_v: |
3656 | *parsed_pattern++ = META_ESCAPE + escape; |
3657 | break; |
3658 | |
3659 | /* These escapes are converted to Unicode property tests when |
3660 | PCRE2_UCP is set. */ |
3661 | |
3662 | case ESC_d: |
3663 | case ESC_D: |
3664 | case ESC_s: |
3665 | case ESC_S: |
3666 | case ESC_w: |
3667 | case ESC_W: |
3668 | if ((options & PCRE2_UCP) == 0) |
3669 | { |
3670 | *parsed_pattern++ = META_ESCAPE + escape; |
3671 | } |
3672 | else |
3673 | { |
3674 | *parsed_pattern++ = META_ESCAPE + |
3675 | ((escape == ESC_d || escape == ESC_s || escape == ESC_w)? |
3676 | ESC_p : ESC_P); |
3677 | switch(escape) |
3678 | { |
3679 | case ESC_d: |
3680 | case ESC_D: |
3681 | *parsed_pattern++ = (PT_PC << 16) | ucp_Nd; |
3682 | break; |
3683 | |
3684 | case ESC_s: |
3685 | case ESC_S: |
3686 | *parsed_pattern++ = PT_SPACE << 16; |
3687 | break; |
3688 | |
3689 | case ESC_w: |
3690 | case ESC_W: |
3691 | *parsed_pattern++ = PT_WORD << 16; |
3692 | break; |
3693 | } |
3694 | } |
3695 | break; |
3696 | |
3697 | /* Explicit Unicode property matching */ |
3698 | |
3699 | case ESC_P: |
3700 | case ESC_p: |
3701 | #ifdef SUPPORT_UNICODE |
3702 | { |
3703 | BOOL negated; |
3704 | uint16_t ptype = 0, pdata = 0; |
3705 | if (!get_ucp(&ptr, &negated, &ptype, &pdata, &errorcode, cb)) |
3706 | goto FAILED; |
3707 | if (negated) escape = (escape == ESC_P)? ESC_p : ESC_P; |
3708 | *parsed_pattern++ = META_ESCAPE + escape; |
3709 | *parsed_pattern++ = (ptype << 16) | pdata; |
3710 | } |
3711 | #else |
3712 | errorcode = ERR45; |
3713 | goto FAILED; |
3714 | #endif |
3715 | break; /* End \P and \p */ |
3716 | |
3717 | default: /* All others are not allowed in a class */ |
3718 | errorcode = ERR7; |
3719 | ptr--; |
3720 | goto FAILED; |
3721 | } |
3722 | |
3723 | /* Perl gives a warning unless a following hyphen is the last character |
3724 | in the class. PCRE throws an error. */ |
3725 | |
3726 | if (ptr < ptrend - 1 && *ptr == CHAR_MINUS && |
3727 | ptr[1] != CHAR_RIGHT_SQUARE_BRACKET) |
3728 | { |
3729 | errorcode = ERR50; |
3730 | goto FAILED; |
3731 | } |
3732 | } |
3733 | |
3734 | /* Proceed to next thing in the class. */ |
3735 | |
3736 | CLASS_CONTINUE: |
3737 | if (ptr >= ptrend) |
3738 | { |
3739 | errorcode = ERR6; /* Missing terminating ']' */ |
3740 | goto FAILED; |
3741 | } |
3742 | GETCHARINCTEST(c, ptr); |
3743 | if (c == CHAR_RIGHT_SQUARE_BRACKET && !inescq) break; |
3744 | } /* End of class-processing loop */ |
3745 | |
3746 | /* -] at the end of a class is a literal '-' */ |
3747 | |
3748 | if (class_range_state == RANGE_STARTED) |
3749 | { |
3750 | parsed_pattern[-1] = CHAR_MINUS; |
3751 | class_range_state = RANGE_NO; |
3752 | } |
3753 | |
3754 | *parsed_pattern++ = META_CLASS_END; |
3755 | break; /* End of character class */ |
3756 | |
3757 | |
3758 | /* ---- Opening parenthesis ---- */ |
3759 | |
3760 | case CHAR_LEFT_PARENTHESIS: |
3761 | if (ptr >= ptrend) goto UNCLOSED_PARENTHESIS; |
3762 | |
3763 | /* If ( is not followed by ? it is either a capture or a special verb or an |
3764 | alpha assertion or a positive non-atomic lookahead. */ |
3765 | |
3766 | if (*ptr != CHAR_QUESTION_MARK) |
3767 | { |
3768 | const char *vn; |
3769 | |
3770 | /* Handle capturing brackets (or non-capturing if auto-capture is turned |
3771 | off). */ |
3772 | |
3773 | if (*ptr != CHAR_ASTERISK) |
3774 | { |
3775 | nest_depth++; |
3776 | if ((options & PCRE2_NO_AUTO_CAPTURE) == 0) |
3777 | { |
3778 | if (cb->bracount >= MAX_GROUP_NUMBER) |
3779 | { |
3780 | errorcode = ERR97; |
3781 | goto FAILED; |
3782 | } |
3783 | cb->bracount++; |
3784 | *parsed_pattern++ = META_CAPTURE | cb->bracount; |
3785 | } |
3786 | else *parsed_pattern++ = META_NOCAPTURE; |
3787 | } |
3788 | |
3789 | /* Do nothing for (* followed by end of pattern or ) so it gives a "bad |
3790 | quantifier" error rather than "(*MARK) must have an argument". */ |
3791 | |
3792 | else if (ptrend - ptr <= 1 || (c = ptr[1]) == CHAR_RIGHT_PARENTHESIS) |
3793 | break; |
3794 | |
3795 | /* Handle "alpha assertions" such as (*pla:...). Most of these are |
3796 | synonyms for the historical symbolic assertions, but the script run and |
3797 | non-atomic lookaround ones are new. They are distinguished by starting |
3798 | with a lower case letter. Checking both ends of the alphabet makes this |
3799 | work in all character codes. */ |
3800 | |
3801 | else if (CHMAX_255(c) && (cb->ctypes[c] & ctype_lcletter) != 0) |
3802 | { |
3803 | uint32_t meta; |
3804 | |
3805 | vn = alasnames; |
3806 | if (!read_name(&ptr, ptrend, utf, 0, &offset, &name, &namelen, |
3807 | &errorcode, cb)) goto FAILED; |
3808 | if (ptr >= ptrend || *ptr != CHAR_COLON) |
3809 | { |
3810 | errorcode = ERR95; /* Malformed */ |
3811 | goto FAILED; |
3812 | } |
3813 | |
3814 | /* Scan the table of alpha assertion names */ |
3815 | |
3816 | for (i = 0; i < alascount; i++) |
3817 | { |
3818 | if (namelen == alasmeta[i].len && |
3819 | PRIV(strncmp_c8)(name, vn, namelen) == 0) |
3820 | break; |
3821 | vn += alasmeta[i].len + 1; |
3822 | } |
3823 | |
3824 | if (i >= alascount) |
3825 | { |
3826 | errorcode = ERR95; /* Alpha assertion not recognized */ |
3827 | goto FAILED; |
3828 | } |
3829 | |
3830 | /* Check for expecting an assertion condition. If so, only atomic |
3831 | lookaround assertions are valid. */ |
3832 | |
3833 | meta = alasmeta[i].meta; |
3834 | if (prev_expect_cond_assert > 0 && |
3835 | (meta < META_LOOKAHEAD || meta > META_LOOKBEHINDNOT)) |
3836 | { |
3837 | errorcode = (meta == META_LOOKAHEAD_NA || meta == META_LOOKBEHIND_NA)? |
3838 | ERR98 : ERR28; /* (Atomic) assertion expected */ |
3839 | goto FAILED; |
3840 | } |
3841 | |
3842 | /* The lookaround alphabetic synonyms can mostly be handled by jumping |
3843 | to the code that handles the traditional symbolic forms. */ |
3844 | |
3845 | switch(meta) |
3846 | { |
3847 | default: |
3848 | errorcode = ERR89; /* Unknown code; should never occur because */ |
3849 | goto FAILED; /* the meta values come from a table above. */ |
3850 | |
3851 | case META_ATOMIC: |
3852 | goto ATOMIC_GROUP; |
3853 | |
3854 | case META_LOOKAHEAD: |
3855 | goto POSITIVE_LOOK_AHEAD; |
3856 | |
3857 | case META_LOOKAHEAD_NA: |
3858 | goto POSITIVE_NONATOMIC_LOOK_AHEAD; |
3859 | |
3860 | case META_LOOKAHEADNOT: |
3861 | goto NEGATIVE_LOOK_AHEAD; |
3862 | |
3863 | case META_LOOKBEHIND: |
3864 | case META_LOOKBEHINDNOT: |
3865 | case META_LOOKBEHIND_NA: |
3866 | *parsed_pattern++ = meta; |
3867 | ptr--; |
3868 | goto POST_LOOKBEHIND; |
3869 | |
3870 | /* The script run facilities are handled here. Unicode support is |
3871 | required (give an error if not, as this is a security issue). Always |
3872 | record a META_SCRIPT_RUN item. Then, for the atomic version, insert |
3873 | META_ATOMIC and remember that we need two META_KETs at the end. */ |
3874 | |
3875 | case META_SCRIPT_RUN: |
3876 | case META_ATOMIC_SCRIPT_RUN: |
3877 | #ifdef SUPPORT_UNICODE |
3878 | *parsed_pattern++ = META_SCRIPT_RUN; |
3879 | nest_depth++; |
3880 | ptr++; |
3881 | if (meta == META_ATOMIC_SCRIPT_RUN) |
3882 | { |
3883 | *parsed_pattern++ = META_ATOMIC; |
3884 | if (top_nest == NULL) top_nest = (nest_save *)(cb->start_workspace); |
3885 | else if (++top_nest >= end_nests) |
3886 | { |
3887 | errorcode = ERR84; |
3888 | goto FAILED; |
3889 | } |
3890 | top_nest->nest_depth = nest_depth; |
3891 | top_nest->flags = NSF_ATOMICSR; |
3892 | top_nest->options = options & PARSE_TRACKED_OPTIONS; |
3893 | } |
3894 | break; |
3895 | #else /* SUPPORT_UNICODE */ |
3896 | errorcode = ERR96; |
3897 | goto FAILED; |
3898 | #endif |
3899 | } |
3900 | } |
3901 | |
3902 | |
3903 | /* ---- Handle (*VERB) and (*VERB:NAME) ---- */ |
3904 | |
3905 | else |
3906 | { |
3907 | vn = verbnames; |
3908 | if (!read_name(&ptr, ptrend, utf, 0, &offset, &name, &namelen, |
3909 | &errorcode, cb)) goto FAILED; |
3910 | if (ptr >= ptrend || (*ptr != CHAR_COLON && |
3911 | *ptr != CHAR_RIGHT_PARENTHESIS)) |
3912 | { |
3913 | errorcode = ERR60; /* Malformed */ |
3914 | goto FAILED; |
3915 | } |
3916 | |
3917 | /* Scan the table of verb names */ |
3918 | |
3919 | for (i = 0; i < verbcount; i++) |
3920 | { |
3921 | if (namelen == verbs[i].len && |
3922 | PRIV(strncmp_c8)(name, vn, namelen) == 0) |
3923 | break; |
3924 | vn += verbs[i].len + 1; |
3925 | } |
3926 | |
3927 | if (i >= verbcount) |
3928 | { |
3929 | errorcode = ERR60; /* Verb not recognized */ |
3930 | goto FAILED; |
3931 | } |
3932 | |
3933 | /* An empty argument is treated as no argument. */ |
3934 | |
3935 | if (*ptr == CHAR_COLON && ptr + 1 < ptrend && |
3936 | ptr[1] == CHAR_RIGHT_PARENTHESIS) |
3937 | ptr++; /* Advance to the closing parens */ |
3938 | |
3939 | /* Check for mandatory non-empty argument; this is (*MARK) */ |
3940 | |
3941 | if (verbs[i].has_arg > 0 && *ptr != CHAR_COLON) |
3942 | { |
3943 | errorcode = ERR66; |
3944 | goto FAILED; |
3945 | } |
3946 | |
3947 | /* Remember where this verb, possibly with a preceding (*MARK), starts, |
3948 | for handling quantified (*ACCEPT). */ |
3949 | |
3950 | verbstartptr = parsed_pattern; |
3951 | okquantifier = (verbs[i].meta == META_ACCEPT); |
3952 | |
3953 | /* It appears that Perl allows any characters whatsoever, other than a |
3954 | closing parenthesis, to appear in arguments ("names"), so we no longer |
3955 | insist on letters, digits, and underscores. Perl does not, however, do |
3956 | any interpretation within arguments, and has no means of including a |
3957 | closing parenthesis. PCRE supports escape processing but only when it |
3958 | is requested by an option. We set inverbname TRUE here, and let the |
3959 | main loop take care of this so that escape and \x processing is done by |
3960 | the main code above. */ |
3961 | |
3962 | if (*ptr++ == CHAR_COLON) /* Skip past : or ) */ |
3963 | { |
3964 | /* Some optional arguments can be treated as a preceding (*MARK) */ |
3965 | |
3966 | if (verbs[i].has_arg < 0) |
3967 | { |
3968 | add_after_mark = verbs[i].meta; |
3969 | *parsed_pattern++ = META_MARK; |
3970 | } |
3971 | |
3972 | /* The remaining verbs with arguments (except *MARK) need a different |
3973 | opcode. */ |
3974 | |
3975 | else |
3976 | { |
3977 | *parsed_pattern++ = verbs[i].meta + |
3978 | ((verbs[i].meta != META_MARK)? 0x00010000u:0); |
3979 | } |
3980 | |
3981 | /* Set up for reading the name in the main loop. */ |
3982 | |
3983 | verblengthptr = parsed_pattern++; |
3984 | verbnamestart = ptr; |
3985 | inverbname = TRUE; |
3986 | } |
3987 | else /* No verb "name" argument */ |
3988 | { |
3989 | *parsed_pattern++ = verbs[i].meta; |
3990 | } |
3991 | } /* End of (*VERB) handling */ |
3992 | break; /* Done with this parenthesis */ |
3993 | } /* End of groups that don't start with (? */ |
3994 | |
3995 | |
3996 | /* ---- Items starting (? ---- */ |
3997 | |
3998 | /* The type of item is determined by what follows (?. Handle (?| and option |
3999 | changes under "default" because both need a new block on the nest stack. |
4000 | Comments starting with (?# are handled above. Note that there is some |
4001 | ambiguity about the sequence (?- because if a digit follows it's a relative |
4002 | recursion or subroutine call whereas otherwise it's an option unsetting. */ |
4003 | |
4004 | if (++ptr >= ptrend) goto UNCLOSED_PARENTHESIS; |
4005 | |
4006 | switch(*ptr) |
4007 | { |
4008 | default: |
4009 | if (*ptr == CHAR_MINUS && ptrend - ptr > 1 && IS_DIGIT(ptr[1])) |
4010 | goto RECURSION_BYNUMBER; /* The + case is handled by CHAR_PLUS */ |
4011 | |
4012 | /* We now have either (?| or a (possibly empty) option setting, |
4013 | optionally followed by a non-capturing group. */ |
4014 | |
4015 | nest_depth++; |
4016 | if (top_nest == NULL) top_nest = (nest_save *)(cb->start_workspace); |
4017 | else if (++top_nest >= end_nests) |
4018 | { |
4019 | errorcode = ERR84; |
4020 | goto FAILED; |
4021 | } |
4022 | top_nest->nest_depth = nest_depth; |
4023 | top_nest->flags = 0; |
4024 | top_nest->options = options & PARSE_TRACKED_OPTIONS; |
4025 | |
4026 | /* Start of non-capturing group that resets the capture count for each |
4027 | branch. */ |
4028 | |
4029 | if (*ptr == CHAR_VERTICAL_LINE) |
4030 | { |
4031 | top_nest->reset_group = (uint16_t)cb->bracount; |
4032 | top_nest->max_group = (uint16_t)cb->bracount; |
4033 | top_nest->flags |= NSF_RESET; |
4034 | cb->external_flags |= PCRE2_DUPCAPUSED; |
4035 | *parsed_pattern++ = META_NOCAPTURE; |
4036 | ptr++; |
4037 | } |
4038 | |
4039 | /* Scan for options imnsxJU to be set or unset. */ |
4040 | |
4041 | else |
4042 | { |
4043 | BOOL hyphenok = TRUE; |
4044 | uint32_t oldoptions = options; |
4045 | |
4046 | top_nest->reset_group = 0; |
4047 | top_nest->max_group = 0; |
4048 | set = unset = 0; |
4049 | optset = &set; |
4050 | |
4051 | /* ^ at the start unsets imnsx and disables the subsequent use of - */ |
4052 | |
4053 | if (ptr < ptrend && *ptr == CHAR_CIRCUMFLEX_ACCENT) |
4054 | { |
4055 | options &= ~(PCRE2_CASELESS|PCRE2_MULTILINE|PCRE2_NO_AUTO_CAPTURE| |
4056 | PCRE2_DOTALL|PCRE2_EXTENDED|PCRE2_EXTENDED_MORE); |
4057 | hyphenok = FALSE; |
4058 | ptr++; |
4059 | } |
4060 | |
4061 | while (ptr < ptrend && *ptr != CHAR_RIGHT_PARENTHESIS && |
4062 | *ptr != CHAR_COLON) |
4063 | { |
4064 | switch (*ptr++) |
4065 | { |
4066 | case CHAR_MINUS: |
4067 | if (!hyphenok) |
4068 | { |
4069 | errorcode = ERR94; |
4070 | ptr--; /* Correct the offset */ |
4071 | goto FAILED; |
4072 | } |
4073 | optset = &unset; |
4074 | hyphenok = FALSE; |
4075 | break; |
4076 | |
4077 | case CHAR_J: /* Record that it changed in the external options */ |
4078 | *optset |= PCRE2_DUPNAMES; |
4079 | cb->external_flags |= PCRE2_JCHANGED; |
4080 | break; |
4081 | |
4082 | case CHAR_i: *optset |= PCRE2_CASELESS; break; |
4083 | case CHAR_m: *optset |= PCRE2_MULTILINE; break; |
4084 | case CHAR_n: *optset |= PCRE2_NO_AUTO_CAPTURE; break; |
4085 | case CHAR_s: *optset |= PCRE2_DOTALL; break; |
4086 | case CHAR_U: *optset |= PCRE2_UNGREEDY; break; |
4087 | |
4088 | /* If x appears twice it sets the extended extended option. */ |
4089 | |
4090 | case CHAR_x: |
4091 | *optset |= PCRE2_EXTENDED; |
4092 | if (ptr < ptrend && *ptr == CHAR_x) |
4093 | { |
4094 | *optset |= PCRE2_EXTENDED_MORE; |
4095 | ptr++; |
4096 | } |
4097 | break; |
4098 | |
4099 | default: |
4100 | errorcode = ERR11; |
4101 | ptr--; /* Correct the offset */ |
4102 | goto FAILED; |
4103 | } |
4104 | } |
4105 | |
4106 | /* If we are setting extended without extended-more, ensure that any |
4107 | existing extended-more gets unset. Also, unsetting extended must also |
4108 | unset extended-more. */ |
4109 | |
4110 | if ((set & (PCRE2_EXTENDED|PCRE2_EXTENDED_MORE)) == PCRE2_EXTENDED || |
4111 | (unset & PCRE2_EXTENDED) != 0) |
4112 | unset |= PCRE2_EXTENDED_MORE; |
4113 | |
4114 | options = (options | set) & (~unset); |
4115 | |
4116 | /* If the options ended with ')' this is not the start of a nested |
4117 | group with option changes, so the options change at this level. |
4118 | In this case, if the previous level set up a nest block, discard the |
4119 | one we have just created. Otherwise adjust it for the previous level. |
4120 | If the options ended with ':' we are starting a non-capturing group, |
4121 | possibly with an options setting. */ |
4122 | |
4123 | if (ptr >= ptrend) goto UNCLOSED_PARENTHESIS; |
4124 | if (*ptr++ == CHAR_RIGHT_PARENTHESIS) |
4125 | { |
4126 | nest_depth--; /* This is not a nested group after all. */ |
4127 | if (top_nest > (nest_save *)(cb->start_workspace) && |
4128 | (top_nest-1)->nest_depth == nest_depth) top_nest--; |
4129 | else top_nest->nest_depth = nest_depth; |
4130 | } |
4131 | else *parsed_pattern++ = META_NOCAPTURE; |
4132 | |
4133 | /* If nothing changed, no need to record. */ |
4134 | |
4135 | if (options != oldoptions) |
4136 | { |
4137 | *parsed_pattern++ = META_OPTIONS; |
4138 | *parsed_pattern++ = options; |
4139 | } |
4140 | } /* End options processing */ |
4141 | break; /* End default case after (? */ |
4142 | |
4143 | |
4144 | /* ---- Python syntax support ---- */ |
4145 | |
4146 | case CHAR_P: |
4147 | if (++ptr >= ptrend) goto UNCLOSED_PARENTHESIS; |
4148 | |
4149 | /* (?P<name> is the same as (?<name>, which defines a named group. */ |
4150 | |
4151 | if (*ptr == CHAR_LESS_THAN_SIGN) |
4152 | { |
4153 | terminator = CHAR_GREATER_THAN_SIGN; |
4154 | goto DEFINE_NAME; |
4155 | } |
4156 | |
4157 | /* (?P>name) is the same as (?&name), which is a recursion or subroutine |
4158 | call. */ |
4159 | |
4160 | if (*ptr == CHAR_GREATER_THAN_SIGN) goto RECURSE_BY_NAME; |
4161 | |
4162 | /* (?P=name) is the same as \k<name>, a back reference by name. Anything |
4163 | else after (?P is an error. */ |
4164 | |
4165 | if (*ptr != CHAR_EQUALS_SIGN) |
4166 | { |
4167 | errorcode = ERR41; |
4168 | goto FAILED; |
4169 | } |
4170 | if (!read_name(&ptr, ptrend, utf, CHAR_RIGHT_PARENTHESIS, &offset, &name, |
4171 | &namelen, &errorcode, cb)) goto FAILED; |
4172 | *parsed_pattern++ = META_BACKREF_BYNAME; |
4173 | *parsed_pattern++ = namelen; |
4174 | PUTOFFSET(offset, parsed_pattern); |
4175 | okquantifier = TRUE; |
4176 | break; /* End of (?P processing */ |
4177 | |
4178 | |
4179 | /* ---- Recursion/subroutine calls by number ---- */ |
4180 | |
4181 | case CHAR_R: |
4182 | i = 0; /* (?R) == (?R0) */ |
4183 | ptr++; |
4184 | if (ptr >= ptrend || *ptr != CHAR_RIGHT_PARENTHESIS) |
4185 | { |
4186 | errorcode = ERR58; |
4187 | goto FAILED; |
4188 | } |
4189 | goto SET_RECURSION; |
4190 | |
4191 | /* An item starting (?- followed by a digit comes here via the "default" |
4192 | case because (?- followed by a non-digit is an options setting. */ |
4193 | |
4194 | case CHAR_PLUS: |
4195 | if (ptrend - ptr < 2 || !IS_DIGIT(ptr[1])) |
4196 | { |
4197 | errorcode = ERR29; /* Missing number */ |
4198 | goto FAILED; |
4199 | } |
4200 | /* Fall through */ |
4201 | |
4202 | case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: |
4203 | case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9: |
4204 | RECURSION_BYNUMBER: |
4205 | if (!read_number(&ptr, ptrend, |
4206 | (IS_DIGIT(*ptr))? -1:(int)(cb->bracount), /* + and - are relative */ |
4207 | MAX_GROUP_NUMBER, ERR61, |
4208 | &i, &errorcode)) goto FAILED; |
4209 | if (i < 0) /* NB (?0) is permitted */ |
4210 | { |
4211 | errorcode = ERR15; /* Unknown group */ |
4212 | goto FAILED_BACK; |
4213 | } |
4214 | if (ptr >= ptrend || *ptr != CHAR_RIGHT_PARENTHESIS) |
4215 | goto UNCLOSED_PARENTHESIS; |
4216 | |
4217 | SET_RECURSION: |
4218 | *parsed_pattern++ = META_RECURSE | (uint32_t)i; |
4219 | offset = (PCRE2_SIZE)(ptr - cb->start_pattern); |
4220 | ptr++; |
4221 | PUTOFFSET(offset, parsed_pattern); |
4222 | okquantifier = TRUE; |
4223 | break; /* End of recursive call by number handling */ |
4224 | |
4225 | |
4226 | /* ---- Recursion/subroutine calls by name ---- */ |
4227 | |
4228 | case CHAR_AMPERSAND: |
4229 | RECURSE_BY_NAME: |
4230 | if (!read_name(&ptr, ptrend, utf, CHAR_RIGHT_PARENTHESIS, &offset, &name, |
4231 | &namelen, &errorcode, cb)) goto FAILED; |
4232 | *parsed_pattern++ = META_RECURSE_BYNAME; |
4233 | *parsed_pattern++ = namelen; |
4234 | PUTOFFSET(offset, parsed_pattern); |
4235 | okquantifier = TRUE; |
4236 | break; |
4237 | |
4238 | /* ---- Callout with numerical or string argument ---- */ |
4239 | |
4240 | case CHAR_C: |
4241 | if (++ptr >= ptrend) goto UNCLOSED_PARENTHESIS; |
4242 | |
4243 | /* If the previous item was a condition starting (?(? an assertion, |
4244 | optionally preceded by a callout, is expected. This is checked later on, |
4245 | during actual compilation. However we need to identify this kind of |
4246 | assertion in this pass because it must not be qualified. The value of |
4247 | expect_cond_assert is set to 2 after (?(? is processed. We decrement it |
4248 | for a callout - still leaving a positive value that identifies the |
4249 | assertion. Multiple callouts or any other items will make it zero or |
4250 | less, which doesn't matter because they will cause an error later. */ |
4251 | |
4252 | expect_cond_assert = prev_expect_cond_assert - 1; |
4253 | |
4254 | /* If previous_callout is not NULL, it means this follows a previous |
4255 | callout. If it was a manual callout, do nothing; this means its "length |
4256 | of next pattern item" field will remain zero. If it was an automatic |
4257 | callout, abolish it. */ |
4258 | |
4259 | if (previous_callout != NULL && (options & PCRE2_AUTO_CALLOUT) != 0 && |
4260 | previous_callout == parsed_pattern - 4 && |
4261 | parsed_pattern[-1] == 255) |
4262 | parsed_pattern = previous_callout; |
4263 | |
4264 | /* Save for updating next pattern item length, and skip one item before |
4265 | completing. */ |
4266 | |
4267 | previous_callout = parsed_pattern; |
4268 | after_manual_callout = 1; |
4269 | |
4270 | /* Handle a string argument; specific delimiter is required. */ |
4271 | |
4272 | if (*ptr != CHAR_RIGHT_PARENTHESIS && !IS_DIGIT(*ptr)) |
4273 | { |
4274 | PCRE2_SIZE calloutlength; |
4275 | PCRE2_SPTR startptr = ptr; |
4276 | |
4277 | delimiter = 0; |
4278 | for (i = 0; PRIV(callout_start_delims)[i] != 0; i++) |
4279 | { |
4280 | if (*ptr == PRIV(callout_start_delims)[i]) |
4281 | { |
4282 | delimiter = PRIV(callout_end_delims)[i]; |
4283 | break; |
4284 | } |
4285 | } |
4286 | if (delimiter == 0) |
4287 | { |
4288 | errorcode = ERR82; |
4289 | goto FAILED; |
4290 | } |
4291 | |
4292 | *parsed_pattern = META_CALLOUT_STRING; |
4293 | parsed_pattern += 3; /* Skip pattern info */ |
4294 | |
4295 | for (;;) |
4296 | { |
4297 | if (++ptr >= ptrend) |
4298 | { |
4299 | errorcode = ERR81; |
4300 | ptr = startptr; /* To give a more useful message */ |
4301 | goto FAILED; |
4302 | } |
4303 | if (*ptr == delimiter && (++ptr >= ptrend || *ptr != delimiter)) |
4304 | break; |
4305 | } |
4306 | |
4307 | calloutlength = (PCRE2_SIZE)(ptr - startptr); |
4308 | if (calloutlength > UINT32_MAX) |
4309 | { |
4310 | errorcode = ERR72; |
4311 | goto FAILED; |
4312 | } |
4313 | *parsed_pattern++ = (uint32_t)calloutlength; |
4314 | offset = (PCRE2_SIZE)(startptr - cb->start_pattern); |
4315 | PUTOFFSET(offset, parsed_pattern); |
4316 | } |
4317 | |
4318 | /* Handle a callout with an optional numerical argument, which must be |
4319 | less than or equal to 255. A missing argument gives 0. */ |
4320 | |
4321 | else |
4322 | { |
4323 | int n = 0; |
4324 | *parsed_pattern = META_CALLOUT_NUMBER; /* Numerical callout */ |
4325 | parsed_pattern += 3; /* Skip pattern info */ |
4326 | while (ptr < ptrend && IS_DIGIT(*ptr)) |
4327 | { |
4328 | n = n * 10 + *ptr++ - CHAR_0; |
4329 | if (n > 255) |
4330 | { |
4331 | errorcode = ERR38; |
4332 | goto FAILED; |
4333 | } |
4334 | } |
4335 | *parsed_pattern++ = n; |
4336 | } |
4337 | |
4338 | /* Both formats must have a closing parenthesis */ |
4339 | |
4340 | if (ptr >= ptrend || *ptr != CHAR_RIGHT_PARENTHESIS) |
4341 | { |
4342 | errorcode = ERR39; |
4343 | goto FAILED; |
4344 | } |
4345 | ptr++; |
4346 | |
4347 | /* Remember the offset to the next item in the pattern, and set a default |
4348 | length. This should get updated after the next item is read. */ |
4349 | |
4350 | previous_callout[1] = (uint32_t)(ptr - cb->start_pattern); |
4351 | previous_callout[2] = 0; |
4352 | break; /* End callout */ |
4353 | |
4354 | |
4355 | /* ---- Conditional group ---- */ |
4356 | |
4357 | /* A condition can be an assertion, a number (referring to a numbered |
4358 | group's having been set), a name (referring to a named group), or 'R', |
4359 | referring to overall recursion. R<digits> and R&name are also permitted |
4360 | for recursion state tests. Numbers may be preceded by + or - to specify a |
4361 | relative group number. |
4362 | |
4363 | There are several syntaxes for testing a named group: (?(name)) is used |
4364 | by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')). |
4365 | |
4366 | There are two unfortunate ambiguities. 'R' can be the recursive thing or |
4367 | the name 'R' (and similarly for 'R' followed by digits). 'DEFINE' can be |
4368 | the Perl DEFINE feature or the Python named test. We look for a name |
4369 | first; if not found, we try the other case. |
4370 | |
4371 | For compatibility with auto-callouts, we allow a callout to be specified |
4372 | before a condition that is an assertion. */ |
4373 | |
4374 | case CHAR_LEFT_PARENTHESIS: |
4375 | if (++ptr >= ptrend) goto UNCLOSED_PARENTHESIS; |
4376 | nest_depth++; |
4377 | |
4378 | /* If the next character is ? or * there must be an assertion next |
4379 | (optionally preceded by a callout). We do not check this here, but |
4380 | instead we set expect_cond_assert to 2. If this is still greater than |
4381 | zero (callouts decrement it) when the next assertion is read, it will be |
4382 | marked as a condition that must not be repeated. A value greater than |
4383 | zero also causes checking that an assertion (possibly with callout) |
4384 | follows. */ |
4385 | |
4386 | if (*ptr == CHAR_QUESTION_MARK || *ptr == CHAR_ASTERISK) |
4387 | { |
4388 | *parsed_pattern++ = META_COND_ASSERT; |
4389 | ptr--; /* Pull pointer back to the opening parenthesis. */ |
4390 | expect_cond_assert = 2; |
4391 | break; /* End of conditional */ |
4392 | } |
4393 | |
4394 | /* Handle (?([+-]number)... */ |
4395 | |
4396 | if (read_number(&ptr, ptrend, cb->bracount, MAX_GROUP_NUMBER, ERR61, &i, |
4397 | &errorcode)) |
4398 | { |
4399 | if (i <= 0) |
4400 | { |
4401 | errorcode = ERR15; |
4402 | goto FAILED; |
4403 | } |
4404 | *parsed_pattern++ = META_COND_NUMBER; |
4405 | offset = (PCRE2_SIZE)(ptr - cb->start_pattern - 2); |
4406 | PUTOFFSET(offset, parsed_pattern); |
4407 | *parsed_pattern++ = i; |
4408 | } |
4409 | else if (errorcode != 0) goto FAILED; /* Number too big */ |
4410 | |
4411 | /* No number found. Handle the special case (?(VERSION[>]=n.m)... */ |
4412 | |
4413 | else if (ptrend - ptr >= 10 && |
4414 | PRIV(strncmp_c8)(ptr, STRING_VERSION, 7) == 0 && |
4415 | ptr[7] != CHAR_RIGHT_PARENTHESIS) |
4416 | { |
4417 | uint32_t ge = 0; |
4418 | int major = 0; |
4419 | int minor = 0; |
4420 | |
4421 | ptr += 7; |
4422 | if (*ptr == CHAR_GREATER_THAN_SIGN) |
4423 | { |
4424 | ge = 1; |
4425 | ptr++; |
4426 | } |
4427 | |
4428 | /* NOTE: cannot write IS_DIGIT(*(++ptr)) here because IS_DIGIT |
4429 | references its argument twice. */ |
4430 | |
4431 | if (*ptr != CHAR_EQUALS_SIGN || (ptr++, !IS_DIGIT(*ptr))) |
4432 | goto BAD_VERSION_CONDITION; |
4433 | |
4434 | if (!read_number(&ptr, ptrend, -1, 1000, ERR79, &major, &errorcode)) |
4435 | goto FAILED; |
4436 | |
4437 | if (ptr >= ptrend) goto BAD_VERSION_CONDITION; |
4438 | if (*ptr == CHAR_DOT) |
4439 | { |
4440 | if (++ptr >= ptrend || !IS_DIGIT(*ptr)) goto BAD_VERSION_CONDITION; |
4441 | minor = (*ptr++ - CHAR_0) * 10; |
4442 | if (ptr >= ptrend) goto BAD_VERSION_CONDITION; |
4443 | if (IS_DIGIT(*ptr)) minor += *ptr++ - CHAR_0; |
4444 | if (ptr >= ptrend || *ptr != CHAR_RIGHT_PARENTHESIS) |
4445 | goto BAD_VERSION_CONDITION; |
4446 | } |
4447 | |
4448 | *parsed_pattern++ = META_COND_VERSION; |
4449 | *parsed_pattern++ = ge; |
4450 | *parsed_pattern++ = major; |
4451 | *parsed_pattern++ = minor; |
4452 | } |
4453 | |
4454 | /* All the remaining cases now require us to read a name. We cannot at |
4455 | this stage distinguish ambiguous cases such as (?(R12) which might be a |
4456 | recursion test by number or a name, because the named groups have not yet |
4457 | all been identified. Those cases are treated as names, but given a |
4458 | different META code. */ |
4459 | |
4460 | else |
4461 | { |
4462 | BOOL was_r_ampersand = FALSE; |
4463 | |
4464 | if (*ptr == CHAR_R && ptrend - ptr > 1 && ptr[1] == CHAR_AMPERSAND) |
4465 | { |
4466 | terminator = CHAR_RIGHT_PARENTHESIS; |
4467 | was_r_ampersand = TRUE; |
4468 | ptr++; |
4469 | } |
4470 | else if (*ptr == CHAR_LESS_THAN_SIGN) |
4471 | terminator = CHAR_GREATER_THAN_SIGN; |
4472 | else if (*ptr == CHAR_APOSTROPHE) |
4473 | terminator = CHAR_APOSTROPHE; |
4474 | else |
4475 | { |
4476 | terminator = CHAR_RIGHT_PARENTHESIS; |
4477 | ptr--; /* Point to char before name */ |
4478 | } |
4479 | if (!read_name(&ptr, ptrend, utf, terminator, &offset, &name, &namelen, |
4480 | &errorcode, cb)) goto FAILED; |
4481 | |
4482 | /* Handle (?(R&name) */ |
4483 | |
4484 | if (was_r_ampersand) |
4485 | { |
4486 | *parsed_pattern = META_COND_RNAME; |
4487 | ptr--; /* Back to closing parens */ |
4488 | } |
4489 | |
4490 | /* Handle (?(name). If the name is "DEFINE" we identify it with a |
4491 | special code. Likewise if the name consists of R followed only by |
4492 | digits. Otherwise, handle it like a quoted name. */ |
4493 | |
4494 | else if (terminator == CHAR_RIGHT_PARENTHESIS) |
4495 | { |
4496 | if (namelen == 6 && PRIV(strncmp_c8)(name, STRING_DEFINE, 6) == 0) |
4497 | *parsed_pattern = META_COND_DEFINE; |
4498 | else |
4499 | { |
4500 | for (i = 1; i < (int)namelen; i++) |
4501 | if (!IS_DIGIT(name[i])) break; |
4502 | *parsed_pattern = (*name == CHAR_R && i >= (int)namelen)? |
4503 | META_COND_RNUMBER : META_COND_NAME; |
4504 | } |
4505 | ptr--; /* Back to closing parens */ |
4506 | } |
4507 | |
4508 | /* Handle (?('name') or (?(<name>) */ |
4509 | |
4510 | else *parsed_pattern = META_COND_NAME; |
4511 | |
4512 | /* All these cases except DEFINE end with the name length and offset; |
4513 | DEFINE just has an offset (for the "too many branches" error). */ |
4514 | |
4515 | if (*parsed_pattern++ != META_COND_DEFINE) *parsed_pattern++ = namelen; |
4516 | PUTOFFSET(offset, parsed_pattern); |
4517 | } /* End cases that read a name */ |
4518 | |
4519 | /* Check the closing parenthesis of the condition */ |
4520 | |
4521 | if (ptr >= ptrend || *ptr != CHAR_RIGHT_PARENTHESIS) |
4522 | { |
4523 | errorcode = ERR24; |
4524 | goto FAILED; |
4525 | } |
4526 | ptr++; |
4527 | break; /* End of condition processing */ |
4528 | |
4529 | |
4530 | /* ---- Atomic group ---- */ |
4531 | |
4532 | case CHAR_GREATER_THAN_SIGN: |
4533 | ATOMIC_GROUP: /* Come from (*atomic: */ |
4534 | *parsed_pattern++ = META_ATOMIC; |
4535 | nest_depth++; |
4536 | ptr++; |
4537 | break; |
4538 | |
4539 | |
4540 | /* ---- Lookahead assertions ---- */ |
4541 | |
4542 | case CHAR_EQUALS_SIGN: |
4543 | POSITIVE_LOOK_AHEAD: /* Come from (*pla: */ |
4544 | *parsed_pattern++ = META_LOOKAHEAD; |
4545 | ptr++; |
4546 | goto POST_ASSERTION; |
4547 | |
4548 | case CHAR_ASTERISK: |
4549 | POSITIVE_NONATOMIC_LOOK_AHEAD: /* Come from (?* */ |
4550 | *parsed_pattern++ = META_LOOKAHEAD_NA; |
4551 | ptr++; |
4552 | goto POST_ASSERTION; |
4553 | |
4554 | case CHAR_EXCLAMATION_MARK: |
4555 | NEGATIVE_LOOK_AHEAD: /* Come from (*nla: */ |
4556 | *parsed_pattern++ = META_LOOKAHEADNOT; |
4557 | ptr++; |
4558 | goto POST_ASSERTION; |
4559 | |
4560 | |
4561 | /* ---- Lookbehind assertions ---- */ |
4562 | |
4563 | /* (?< followed by = or ! or * is a lookbehind assertion. Otherwise (?< |
4564 | is the start of the name of a capturing group. */ |
4565 | |
4566 | case CHAR_LESS_THAN_SIGN: |
4567 | if (ptrend - ptr <= 1 || |
4568 | (ptr[1] != CHAR_EQUALS_SIGN && |
4569 | ptr[1] != CHAR_EXCLAMATION_MARK && |
4570 | ptr[1] != CHAR_ASTERISK)) |
4571 | { |
4572 | terminator = CHAR_GREATER_THAN_SIGN; |
4573 | goto DEFINE_NAME; |
4574 | } |
4575 | *parsed_pattern++ = (ptr[1] == CHAR_EQUALS_SIGN)? |
4576 | META_LOOKBEHIND : (ptr[1] == CHAR_EXCLAMATION_MARK)? |
4577 | META_LOOKBEHINDNOT : META_LOOKBEHIND_NA; |
4578 | |
4579 | POST_LOOKBEHIND: /* Come from (*plb: (*naplb: and (*nlb: */ |
4580 | *has_lookbehind = TRUE; |
4581 | offset = (PCRE2_SIZE)(ptr - cb->start_pattern - 2); |
4582 | PUTOFFSET(offset, parsed_pattern); |
4583 | ptr += 2; |
4584 | /* Fall through */ |
4585 | |
4586 | /* If the previous item was a condition starting (?(? an assertion, |
4587 | optionally preceded by a callout, is expected. This is checked later on, |
4588 | during actual compilation. However we need to identify this kind of |
4589 | assertion in this pass because it must not be qualified. The value of |
4590 | expect_cond_assert is set to 2 after (?(? is processed. We decrement it |
4591 | for a callout - still leaving a positive value that identifies the |
4592 | assertion. Multiple callouts or any other items will make it zero or |
4593 | less, which doesn't matter because they will cause an error later. */ |
4594 | |
4595 | POST_ASSERTION: |
4596 | nest_depth++; |
4597 | if (prev_expect_cond_assert > 0) |
4598 | { |
4599 | if (top_nest == NULL) top_nest = (nest_save *)(cb->start_workspace); |
4600 | else if (++top_nest >= end_nests) |
4601 | { |
4602 | errorcode = ERR84; |
4603 | goto FAILED; |
4604 | } |
4605 | top_nest->nest_depth = nest_depth; |
4606 | top_nest->flags = NSF_CONDASSERT; |
4607 | top_nest->options = options & PARSE_TRACKED_OPTIONS; |
4608 | } |
4609 | break; |
4610 | |
4611 | |
4612 | /* ---- Define a named group ---- */ |
4613 | |
4614 | /* A named group may be defined as (?'name') or (?<name>). In the latter |
4615 | case we jump to DEFINE_NAME from the disambiguation of (?< above with the |
4616 | terminator set to '>'. */ |
4617 | |
4618 | case CHAR_APOSTROPHE: |
4619 | terminator = CHAR_APOSTROPHE; /* Terminator */ |
4620 | |
4621 | DEFINE_NAME: |
4622 | if (!read_name(&ptr, ptrend, utf, terminator, &offset, &name, &namelen, |
4623 | &errorcode, cb)) goto FAILED; |
4624 | |
4625 | /* We have a name for this capturing group. It is also assigned a number, |
4626 | which is its primary means of identification. */ |
4627 | |
4628 | if (cb->bracount >= MAX_GROUP_NUMBER) |
4629 | { |
4630 | errorcode = ERR97; |
4631 | goto FAILED; |
4632 | } |
4633 | cb->bracount++; |
4634 | *parsed_pattern++ = META_CAPTURE | cb->bracount; |
4635 | nest_depth++; |
4636 | |
4637 | /* Check not too many names */ |
4638 | |
4639 | if (cb->names_found >= MAX_NAME_COUNT) |
4640 | { |
4641 | errorcode = ERR49; |
4642 | goto FAILED; |
4643 | } |
4644 | |
4645 | /* Adjust the entry size to accommodate the longest name found. */ |
4646 | |
4647 | if (namelen + IMM2_SIZE + 1 > cb->name_entry_size) |
4648 | cb->name_entry_size = (uint16_t)(namelen + IMM2_SIZE + 1); |
4649 | |
4650 | /* Scan the list to check for duplicates. For duplicate names, if the |
4651 | number is the same, break the loop, which causes the name to be |
4652 | discarded; otherwise, if DUPNAMES is not set, give an error. |
4653 | If it is set, allow the name with a different number, but continue |
4654 | scanning in case this is a duplicate with the same number. For |
4655 | non-duplicate names, give an error if the number is duplicated. */ |
4656 | |
4657 | isdupname = FALSE; |
4658 | ng = cb->named_groups; |
4659 | for (i = 0; i < cb->names_found; i++, ng++) |
4660 | { |
4661 | if (namelen == ng->length && |
4662 | PRIV(strncmp)(name, ng->name, (PCRE2_SIZE)namelen) == 0) |
4663 | { |
4664 | if (ng->number == cb->bracount) break; |
4665 | if ((options & PCRE2_DUPNAMES) == 0) |
4666 | { |
4667 | errorcode = ERR43; |
4668 | goto FAILED; |
4669 | } |
4670 | isdupname = ng->isdup = TRUE; /* Mark as a duplicate */ |
4671 | cb->dupnames = TRUE; /* Duplicate names exist */ |
4672 | } |
4673 | else if (ng->number == cb->bracount) |
4674 | { |
4675 | errorcode = ERR65; |
4676 | goto FAILED; |
4677 | } |
4678 | } |
4679 | |
4680 | if (i < cb->names_found) break; /* Ignore duplicate with same number */ |
4681 | |
4682 | /* Increase the list size if necessary */ |
4683 | |
4684 | if (cb->names_found >= cb->named_group_list_size) |
4685 | { |
4686 | uint32_t newsize = cb->named_group_list_size * 2; |
4687 | named_group *newspace = |
4688 | cb->cx->memctl.malloc(newsize * sizeof(named_group), |
4689 | cb->cx->memctl.memory_data); |
4690 | if (newspace == NULL) |
4691 | { |
4692 | errorcode = ERR21; |
4693 | goto FAILED; |
4694 | } |
4695 | |
4696 | memcpy(newspace, cb->named_groups, |
4697 | cb->named_group_list_size * sizeof(named_group)); |
4698 | if (cb->named_group_list_size > NAMED_GROUP_LIST_SIZE) |
4699 | cb->cx->memctl.free((void *)cb->named_groups, |
4700 | cb->cx->memctl.memory_data); |
4701 | cb->named_groups = newspace; |
4702 | cb->named_group_list_size = newsize; |
4703 | } |
4704 | |
4705 | /* Add this name to the list */ |
4706 | |
4707 | cb->named_groups[cb->names_found].name = name; |
4708 | cb->named_groups[cb->names_found].length = (uint16_t)namelen; |
4709 | cb->named_groups[cb->names_found].number = cb->bracount; |
4710 | cb->named_groups[cb->names_found].isdup = (uint16_t)isdupname; |
4711 | cb->names_found++; |
4712 | break; |
4713 | } /* End of (? switch */ |
4714 | break; /* End of ( handling */ |
4715 | |
4716 | |
4717 | /* ---- Branch terminators ---- */ |
4718 | |
4719 | /* Alternation: reset the capture count if we are in a (?| group. */ |
4720 | |
4721 | case CHAR_VERTICAL_LINE: |
4722 | if (top_nest != NULL && top_nest->nest_depth == nest_depth && |
4723 | (top_nest->flags & NSF_RESET) != 0) |
4724 | { |
4725 | if (cb->bracount > top_nest->max_group) |
4726 | top_nest->max_group = (uint16_t)cb->bracount; |
4727 | cb->bracount = top_nest->reset_group; |
4728 | } |
4729 | *parsed_pattern++ = META_ALT; |
4730 | break; |
4731 | |
4732 | /* End of group; reset the capture count to the maximum if we are in a (?| |
4733 | group and/or reset the options that are tracked during parsing. Disallow |
4734 | quantifier for a condition that is an assertion. */ |
4735 | |
4736 | case CHAR_RIGHT_PARENTHESIS: |
4737 | okquantifier = TRUE; |
4738 | if (top_nest != NULL && top_nest->nest_depth == nest_depth) |
4739 | { |
4740 | options = (options & ~PARSE_TRACKED_OPTIONS) | top_nest->options; |
4741 | if ((top_nest->flags & NSF_RESET) != 0 && |
4742 | top_nest->max_group > cb->bracount) |
4743 | cb->bracount = top_nest->max_group; |
4744 | if ((top_nest->flags & NSF_CONDASSERT) != 0) |
4745 | okquantifier = FALSE; |
4746 | |
4747 | if ((top_nest->flags & NSF_ATOMICSR) != 0) |
4748 | { |
4749 | *parsed_pattern++ = META_KET; |
4750 | } |
4751 | |
4752 | if (top_nest == (nest_save *)(cb->start_workspace)) top_nest = NULL; |
4753 | else top_nest--; |
4754 | } |
4755 | if (nest_depth == 0) /* Unmatched closing parenthesis */ |
4756 | { |
4757 | errorcode = ERR22; |
4758 | goto FAILED_BACK; |
4759 | } |
4760 | nest_depth--; |
4761 | *parsed_pattern++ = META_KET; |
4762 | break; |
4763 | } /* End of switch on pattern character */ |
4764 | } /* End of main character scan loop */ |
4765 | |
4766 | /* End of pattern reached. Check for missing ) at the end of a verb name. */ |
4767 | |
4768 | if (inverbname && ptr >= ptrend) |
4769 | { |
4770 | errorcode = ERR60; |
4771 | goto FAILED; |
4772 | } |
4773 | |
4774 | /* Manage callout for the final item */ |
4775 | |
4776 | PARSED_END: |
4777 | parsed_pattern = manage_callouts(ptr, &previous_callout, auto_callout, |
4778 | parsed_pattern, cb); |
4779 | |
4780 | /* Insert trailing items for word and line matching (features provided for the |
4781 | benefit of pcre2grep). */ |
4782 | |
4783 | if ((extra_options & PCRE2_EXTRA_MATCH_LINE) != 0) |
4784 | { |
4785 | *parsed_pattern++ = META_KET; |
4786 | *parsed_pattern++ = META_DOLLAR; |
4787 | } |
4788 | else if ((extra_options & PCRE2_EXTRA_MATCH_WORD) != 0) |
4789 | { |
4790 | *parsed_pattern++ = META_KET; |
4791 | *parsed_pattern++ = META_ESCAPE + ESC_b; |
4792 | } |
4793 | |
4794 | /* Terminate the parsed pattern, then return success if all groups are closed. |
4795 | Otherwise we have unclosed parentheses. */ |
4796 | |
4797 | if (parsed_pattern >= parsed_pattern_end) |
4798 | { |
4799 | errorcode = ERR63; /* Internal error (parsed pattern overflow) */ |
4800 | goto FAILED; |
4801 | } |
4802 | |
4803 | *parsed_pattern = META_END; |
4804 | if (nest_depth == 0) return 0; |
4805 | |
4806 | UNCLOSED_PARENTHESIS: |
4807 | errorcode = ERR14; |
4808 | |
4809 | /* Come here for all failures. */ |
4810 | |
4811 | FAILED: |
4812 | cb->erroroffset = (PCRE2_SIZE)(ptr - cb->start_pattern); |
4813 | return errorcode; |
4814 | |
4815 | /* Some errors need to indicate the previous character. */ |
4816 | |
4817 | FAILED_BACK: |
4818 | ptr--; |
4819 | goto FAILED; |
4820 | |
4821 | /* This failure happens several times. */ |
4822 | |
4823 | BAD_VERSION_CONDITION: |
4824 | errorcode = ERR79; |
4825 | goto FAILED; |
4826 | } |
4827 | |
4828 | |
4829 | |
4830 | /************************************************* |
4831 | * Find first significant opcode * |
4832 | *************************************************/ |
4833 | |
4834 | /* This is called by several functions that scan a compiled expression looking |
4835 | for a fixed first character, or an anchoring opcode etc. It skips over things |
4836 | that do not influence this. For some calls, it makes sense to skip negative |
4837 | forward and all backward assertions, and also the \b assertion; for others it |
4838 | does not. |
4839 | |
4840 | Arguments: |
4841 | code pointer to the start of the group |
4842 | skipassert TRUE if certain assertions are to be skipped |
4843 | |
4844 | Returns: pointer to the first significant opcode |
4845 | */ |
4846 | |
4847 | static const PCRE2_UCHAR* |
4848 | first_significant_code(PCRE2_SPTR code, BOOL skipassert) |
4849 | { |
4850 | for (;;) |
4851 | { |
4852 | switch ((int)*code) |
4853 | { |
4854 | case OP_ASSERT_NOT: |
4855 | case OP_ASSERTBACK: |
4856 | case OP_ASSERTBACK_NOT: |
4857 | case OP_ASSERTBACK_NA: |
4858 | if (!skipassert) return code; |
4859 | do code += GET(code, 1); while (*code == OP_ALT); |
4860 | code += PRIV(OP_lengths)[*code]; |
4861 | break; |
4862 | |
4863 | case OP_WORD_BOUNDARY: |
4864 | case OP_NOT_WORD_BOUNDARY: |
4865 | if (!skipassert) return code; |
4866 | /* Fall through */ |
4867 | |
4868 | case OP_CALLOUT: |
4869 | case OP_CREF: |
4870 | case OP_DNCREF: |
4871 | case OP_RREF: |
4872 | case OP_DNRREF: |
4873 | case OP_FALSE: |
4874 | case OP_TRUE: |
4875 | code += PRIV(OP_lengths)[*code]; |
4876 | break; |
4877 | |
4878 | case OP_CALLOUT_STR: |
4879 | code += GET(code, 1 + 2*LINK_SIZE); |
4880 | break; |
4881 | |
4882 | case OP_SKIPZERO: |
4883 | code += 2 + GET(code, 2) + LINK_SIZE; |
4884 | break; |
4885 | |
4886 | case OP_COND: |
4887 | case OP_SCOND: |
4888 | if (code[1+LINK_SIZE] != OP_FALSE || /* Not DEFINE */ |
4889 | code[GET(code, 1)] != OP_KET) /* More than one branch */ |
4890 | return code; |
4891 | code += GET(code, 1) + 1 + LINK_SIZE; |
4892 | break; |
4893 | |
4894 | case OP_MARK: |
4895 | case OP_COMMIT_ARG: |
4896 | case OP_PRUNE_ARG: |
4897 | case OP_SKIP_ARG: |
4898 | case OP_THEN_ARG: |
4899 | code += code[1] + PRIV(OP_lengths)[*code]; |
4900 | break; |
4901 | |
4902 | default: |
4903 | return code; |
4904 | } |
4905 | } |
4906 | /* Control never reaches here */ |
4907 | } |
4908 | |
4909 | |
4910 | |
4911 | #ifdef SUPPORT_UNICODE |
4912 | /************************************************* |
4913 | * Get othercase range * |
4914 | *************************************************/ |
4915 | |
4916 | /* This function is passed the start and end of a class range in UCP mode. It |
4917 | searches up the characters, looking for ranges of characters in the "other" |
4918 | case. Each call returns the next one, updating the start address. A character |
4919 | with multiple other cases is returned on its own with a special return value. |
4920 | |
4921 | Arguments: |
4922 | cptr points to starting character value; updated |
4923 | d end value |
4924 | ocptr where to put start of othercase range |
4925 | odptr where to put end of othercase range |
4926 | |
4927 | Yield: -1 when no more |
4928 | 0 when a range is returned |
4929 | >0 the CASESET offset for char with multiple other cases |
4930 | in this case, ocptr contains the original |
4931 | */ |
4932 | |
4933 | static int |
4934 | get_othercase_range(uint32_t *cptr, uint32_t d, uint32_t *ocptr, |
4935 | uint32_t *odptr) |
4936 | { |
4937 | uint32_t c, othercase, next; |
4938 | unsigned int co; |
4939 | |
4940 | /* Find the first character that has an other case. If it has multiple other |
4941 | cases, return its case offset value. */ |
4942 | |
4943 | for (c = *cptr; c <= d; c++) |
4944 | { |
4945 | if ((co = UCD_CASESET(c)) != 0) |
4946 | { |
4947 | *ocptr = c++; /* Character that has the set */ |
4948 | *cptr = c; /* Rest of input range */ |
4949 | return (int)co; |
4950 | } |
4951 | if ((othercase = UCD_OTHERCASE(c)) != c) break; |
4952 | } |
4953 | |
4954 | if (c > d) return -1; /* Reached end of range */ |
4955 | |
4956 | /* Found a character that has a single other case. Search for the end of the |
4957 | range, which is either the end of the input range, or a character that has zero |
4958 | or more than one other cases. */ |
4959 | |
4960 | *ocptr = othercase; |
4961 | next = othercase + 1; |
4962 | |
4963 | for (++c; c <= d; c++) |
4964 | { |
4965 | if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break; |
4966 | next++; |
4967 | } |
4968 | |
4969 | *odptr = next - 1; /* End of othercase range */ |
4970 | *cptr = c; /* Rest of input range */ |
4971 | return 0; |
4972 | } |
4973 | #endif /* SUPPORT_UNICODE */ |
4974 | |
4975 | |
4976 | |
4977 | /************************************************* |
4978 | * Add a character or range to a class (internal) * |
4979 | *************************************************/ |
4980 | |
4981 | /* This function packages up the logic of adding a character or range of |
4982 | characters to a class. The character values in the arguments will be within the |
4983 | valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is |
4984 | called only from within the "add to class" group of functions, some of which |
4985 | are recursive and mutually recursive. The external entry point is |
4986 | add_to_class(). |
4987 | |
4988 | Arguments: |
4989 | classbits the bit map for characters < 256 |
4990 | uchardptr points to the pointer for extra data |
4991 | options the options word |
4992 | cb compile data |
4993 | start start of range character |
4994 | end end of range character |
4995 | |
4996 | Returns: the number of < 256 characters added |
4997 | the pointer to extra data is updated |
4998 | */ |
4999 | |
5000 | static unsigned int |
5001 | add_to_class_internal(uint8_t *classbits, PCRE2_UCHAR **uchardptr, |
5002 | uint32_t options, compile_block *cb, uint32_t start, uint32_t end) |
5003 | { |
5004 | uint32_t c; |
5005 | uint32_t classbits_end = (end <= 0xff ? end : 0xff); |
5006 | unsigned int n8 = 0; |
5007 | |
5008 | /* If caseless matching is required, scan the range and process alternate |
5009 | cases. In Unicode, there are 8-bit characters that have alternate cases that |
5010 | are greater than 255 and vice-versa. Sometimes we can just extend the original |
5011 | range. */ |
5012 | |
5013 | if ((options & PCRE2_CASELESS) != 0) |
5014 | { |
5015 | #ifdef SUPPORT_UNICODE |
5016 | if ((options & (PCRE2_UTF|PCRE2_UCP)) != 0) |
5017 | { |
5018 | int rc; |
5019 | uint32_t oc, od; |
5020 | |
5021 | options &= ~PCRE2_CASELESS; /* Remove for recursive calls */ |
5022 | c = start; |
5023 | |
5024 | while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0) |
5025 | { |
5026 | /* Handle a single character that has more than one other case. */ |
5027 | |
5028 | if (rc > 0) n8 += add_list_to_class_internal(classbits, uchardptr, options, cb, |
5029 | PRIV(ucd_caseless_sets) + rc, oc); |
5030 | |
5031 | /* Do nothing if the other case range is within the original range. */ |
5032 | |
5033 | else if (oc >= cb->class_range_start && od <= cb->class_range_end) continue; |
5034 | |
5035 | /* Extend the original range if there is overlap, noting that if oc < c, we |
5036 | can't have od > end because a subrange is always shorter than the basic |
5037 | range. Otherwise, use a recursive call to add the additional range. */ |
5038 | |
5039 | else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */ |
5040 | else if (od > end && oc <= end + 1) |
5041 | { |
5042 | end = od; /* Extend upwards */ |
5043 | if (end > classbits_end) classbits_end = (end <= 0xff ? end : 0xff); |
5044 | } |
5045 | else n8 += add_to_class_internal(classbits, uchardptr, options, cb, oc, od); |
5046 | } |
5047 | } |
5048 | else |
5049 | #endif /* SUPPORT_UNICODE */ |
5050 | |
5051 | /* Not UTF mode */ |
5052 | |
5053 | for (c = start; c <= classbits_end; c++) |
5054 | { |
5055 | SETBIT(classbits, cb->fcc[c]); |
5056 | n8++; |
5057 | } |
5058 | } |
5059 | |
5060 | /* Now handle the originally supplied range. Adjust the final value according |
5061 | to the bit length - this means that the same lists of (e.g.) horizontal spaces |
5062 | can be used in all cases. */ |
5063 | |
5064 | if ((options & PCRE2_UTF) == 0 && end > MAX_NON_UTF_CHAR) |
5065 | end = MAX_NON_UTF_CHAR; |
5066 | |
5067 | if (start > cb->class_range_start && end < cb->class_range_end) return n8; |
5068 | |
5069 | /* Use the bitmap for characters < 256. Otherwise use extra data.*/ |
5070 | |
5071 | for (c = start; c <= classbits_end; c++) |
5072 | { |
5073 | /* Regardless of start, c will always be <= 255. */ |
5074 | SETBIT(classbits, c); |
5075 | n8++; |
5076 | } |
5077 | |
5078 | #ifdef SUPPORT_WIDE_CHARS |
5079 | if (start <= 0xff) start = 0xff + 1; |
5080 | |
5081 | if (end >= start) |
5082 | { |
5083 | PCRE2_UCHAR *uchardata = *uchardptr; |
5084 | |
5085 | #ifdef SUPPORT_UNICODE |
5086 | if ((options & PCRE2_UTF) != 0) |
5087 | { |
5088 | if (start < end) |
5089 | { |
5090 | *uchardata++ = XCL_RANGE; |
5091 | uchardata += PRIV(ord2utf)(start, uchardata); |
5092 | uchardata += PRIV(ord2utf)(end, uchardata); |
5093 | } |
5094 | else if (start == end) |
5095 | { |
5096 | *uchardata++ = XCL_SINGLE; |
5097 | uchardata += PRIV(ord2utf)(start, uchardata); |
5098 | } |
5099 | } |
5100 | else |
5101 | #endif /* SUPPORT_UNICODE */ |
5102 | |
5103 | /* Without UTF support, character values are constrained by the bit length, |
5104 | and can only be > 256 for 16-bit and 32-bit libraries. */ |
5105 | |
5106 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
5107 | {} |
5108 | #else |
5109 | if (start < end) |
5110 | { |
5111 | *uchardata++ = XCL_RANGE; |
5112 | *uchardata++ = start; |
5113 | *uchardata++ = end; |
5114 | } |
5115 | else if (start == end) |
5116 | { |
5117 | *uchardata++ = XCL_SINGLE; |
5118 | *uchardata++ = start; |
5119 | } |
5120 | #endif /* PCRE2_CODE_UNIT_WIDTH == 8 */ |
5121 | *uchardptr = uchardata; /* Updata extra data pointer */ |
5122 | } |
5123 | #else /* SUPPORT_WIDE_CHARS */ |
5124 | (void)uchardptr; /* Avoid compiler warning */ |
5125 | #endif /* SUPPORT_WIDE_CHARS */ |
5126 | |
5127 | return n8; /* Number of 8-bit characters */ |
5128 | } |
5129 | |
5130 | |
5131 | |
5132 | #ifdef SUPPORT_UNICODE |
5133 | /************************************************* |
5134 | * Add a list of characters to a class (internal) * |
5135 | *************************************************/ |
5136 | |
5137 | /* This function is used for adding a list of case-equivalent characters to a |
5138 | class when in UTF mode. This function is called only from within |
5139 | add_to_class_internal(), with which it is mutually recursive. |
5140 | |
5141 | Arguments: |
5142 | classbits the bit map for characters < 256 |
5143 | uchardptr points to the pointer for extra data |
5144 | options the options word |
5145 | cb contains pointers to tables etc. |
5146 | p points to row of 32-bit values, terminated by NOTACHAR |
5147 | except character to omit; this is used when adding lists of |
5148 | case-equivalent characters to avoid including the one we |
5149 | already know about |
5150 | |
5151 | Returns: the number of < 256 characters added |
5152 | the pointer to extra data is updated |
5153 | */ |
5154 | |
5155 | static unsigned int |
5156 | add_list_to_class_internal(uint8_t *classbits, PCRE2_UCHAR **uchardptr, |
5157 | uint32_t options, compile_block *cb, const uint32_t *p, unsigned int except) |
5158 | { |
5159 | unsigned int n8 = 0; |
5160 | while (p[0] < NOTACHAR) |
5161 | { |
5162 | unsigned int n = 0; |
5163 | if (p[0] != except) |
5164 | { |
5165 | while(p[n+1] == p[0] + n + 1) n++; |
5166 | n8 += add_to_class_internal(classbits, uchardptr, options, cb, p[0], p[n]); |
5167 | } |
5168 | p += n + 1; |
5169 | } |
5170 | return n8; |
5171 | } |
5172 | #endif |
5173 | |
5174 | |
5175 | |
5176 | /************************************************* |
5177 | * External entry point for add range to class * |
5178 | *************************************************/ |
5179 | |
5180 | /* This function sets the overall range so that the internal functions can try |
5181 | to avoid duplication when handling case-independence. |
5182 | |
5183 | Arguments: |
5184 | classbits the bit map for characters < 256 |
5185 | uchardptr points to the pointer for extra data |
5186 | options the options word |
5187 | cb compile data |
5188 | start start of range character |
5189 | end end of range character |
5190 | |
5191 | Returns: the number of < 256 characters added |
5192 | the pointer to extra data is updated |
5193 | */ |
5194 | |
5195 | static unsigned int |
5196 | add_to_class(uint8_t *classbits, PCRE2_UCHAR **uchardptr, uint32_t options, |
5197 | compile_block *cb, uint32_t start, uint32_t end) |
5198 | { |
5199 | cb->class_range_start = start; |
5200 | cb->class_range_end = end; |
5201 | return add_to_class_internal(classbits, uchardptr, options, cb, start, end); |
5202 | } |
5203 | |
5204 | |
5205 | /************************************************* |
5206 | * External entry point for add list to class * |
5207 | *************************************************/ |
5208 | |
5209 | /* This function is used for adding a list of horizontal or vertical whitespace |
5210 | characters to a class. The list must be in order so that ranges of characters |
5211 | can be detected and handled appropriately. This function sets the overall range |
5212 | so that the internal functions can try to avoid duplication when handling |
5213 | case-independence. |
5214 | |
5215 | Arguments: |
5216 | classbits the bit map for characters < 256 |
5217 | uchardptr points to the pointer for extra data |
5218 | options the options word |
5219 | cb contains pointers to tables etc. |
5220 | p points to row of 32-bit values, terminated by NOTACHAR |
5221 | except character to omit; this is used when adding lists of |
5222 | case-equivalent characters to avoid including the one we |
5223 | already know about |
5224 | |
5225 | Returns: the number of < 256 characters added |
5226 | the pointer to extra data is updated |
5227 | */ |
5228 | |
5229 | static unsigned int |
5230 | add_list_to_class(uint8_t *classbits, PCRE2_UCHAR **uchardptr, uint32_t options, |
5231 | compile_block *cb, const uint32_t *p, unsigned int except) |
5232 | { |
5233 | unsigned int n8 = 0; |
5234 | while (p[0] < NOTACHAR) |
5235 | { |
5236 | unsigned int n = 0; |
5237 | if (p[0] != except) |
5238 | { |
5239 | while(p[n+1] == p[0] + n + 1) n++; |
5240 | cb->class_range_start = p[0]; |
5241 | cb->class_range_end = p[n]; |
5242 | n8 += add_to_class_internal(classbits, uchardptr, options, cb, p[0], p[n]); |
5243 | } |
5244 | p += n + 1; |
5245 | } |
5246 | return n8; |
5247 | } |
5248 | |
5249 | |
5250 | |
5251 | /************************************************* |
5252 | * Add characters not in a list to a class * |
5253 | *************************************************/ |
5254 | |
5255 | /* This function is used for adding the complement of a list of horizontal or |
5256 | vertical whitespace to a class. The list must be in order. |
5257 | |
5258 | Arguments: |
5259 | classbits the bit map for characters < 256 |
5260 | uchardptr points to the pointer for extra data |
5261 | options the options word |
5262 | cb contains pointers to tables etc. |
5263 | p points to row of 32-bit values, terminated by NOTACHAR |
5264 | |
5265 | Returns: the number of < 256 characters added |
5266 | the pointer to extra data is updated |
5267 | */ |
5268 | |
5269 | static unsigned int |
5270 | add_not_list_to_class(uint8_t *classbits, PCRE2_UCHAR **uchardptr, |
5271 | uint32_t options, compile_block *cb, const uint32_t *p) |
5272 | { |
5273 | BOOL utf = (options & PCRE2_UTF) != 0; |
5274 | unsigned int n8 = 0; |
5275 | if (p[0] > 0) |
5276 | n8 += add_to_class(classbits, uchardptr, options, cb, 0, p[0] - 1); |
5277 | while (p[0] < NOTACHAR) |
5278 | { |
5279 | while (p[1] == p[0] + 1) p++; |
5280 | n8 += add_to_class(classbits, uchardptr, options, cb, p[0] + 1, |
5281 | (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1); |
5282 | p++; |
5283 | } |
5284 | return n8; |
5285 | } |
5286 | |
5287 | |
5288 | |
5289 | /************************************************* |
5290 | * Find details of duplicate group names * |
5291 | *************************************************/ |
5292 | |
5293 | /* This is called from compile_branch() when it needs to know the index and |
5294 | count of duplicates in the names table when processing named backreferences, |
5295 | either directly, or as conditions. |
5296 | |
5297 | Arguments: |
5298 | name points to the name |
5299 | length the length of the name |
5300 | indexptr where to put the index |
5301 | countptr where to put the count of duplicates |
5302 | errorcodeptr where to put an error code |
5303 | cb the compile block |
5304 | |
5305 | Returns: TRUE if OK, FALSE if not, error code set |
5306 | */ |
5307 | |
5308 | static BOOL |
5309 | find_dupname_details(PCRE2_SPTR name, uint32_t length, int *indexptr, |
5310 | int *countptr, int *errorcodeptr, compile_block *cb) |
5311 | { |
5312 | uint32_t i, groupnumber; |
5313 | int count; |
5314 | PCRE2_UCHAR *slot = cb->name_table; |
5315 | |
5316 | /* Find the first entry in the table */ |
5317 | |
5318 | for (i = 0; i < cb->names_found; i++) |
5319 | { |
5320 | if (PRIV(strncmp)(name, slot+IMM2_SIZE, length) == 0 && |
5321 | slot[IMM2_SIZE+length] == 0) break; |
5322 | slot += cb->name_entry_size; |
5323 | } |
5324 | |
5325 | /* This should not occur, because this function is called only when we know we |
5326 | have duplicate names. Give an internal error. */ |
5327 | |
5328 | if (i >= cb->names_found) |
5329 | { |
5330 | *errorcodeptr = ERR53; |
5331 | cb->erroroffset = name - cb->start_pattern; |
5332 | return FALSE; |
5333 | } |
5334 | |
5335 | /* Record the index and then see how many duplicates there are, updating the |
5336 | backref map and maximum back reference as we do. */ |
5337 | |
5338 | *indexptr = i; |
5339 | count = 0; |
5340 | |
5341 | for (;;) |
5342 | { |
5343 | count++; |
5344 | groupnumber = GET2(slot,0); |
5345 | cb->backref_map |= (groupnumber < 32)? (1u << groupnumber) : 1; |
5346 | if (groupnumber > cb->top_backref) cb->top_backref = groupnumber; |
5347 | if (++i >= cb->names_found) break; |
5348 | slot += cb->name_entry_size; |
5349 | if (PRIV(strncmp)(name, slot+IMM2_SIZE, length) != 0 || |
5350 | (slot+IMM2_SIZE)[length] != 0) break; |
5351 | } |
5352 | |
5353 | *countptr = count; |
5354 | return TRUE; |
5355 | } |
5356 | |
5357 | |
5358 | |
5359 | /************************************************* |
5360 | * Compile one branch * |
5361 | *************************************************/ |
5362 | |
5363 | /* Scan the parsed pattern, compiling it into the a vector of PCRE2_UCHAR. If |
5364 | the options are changed during the branch, the pointer is used to change the |
5365 | external options bits. This function is used during the pre-compile phase when |
5366 | we are trying to find out the amount of memory needed, as well as during the |
5367 | real compile phase. The value of lengthptr distinguishes the two phases. |
5368 | |
5369 | Arguments: |
5370 | optionsptr pointer to the option bits |
5371 | codeptr points to the pointer to the current code point |
5372 | pptrptr points to the current parsed pattern pointer |
5373 | errorcodeptr points to error code variable |
5374 | firstcuptr place to put the first required code unit |
5375 | firstcuflagsptr place to put the first code unit flags |
5376 | reqcuptr place to put the last required code unit |
5377 | reqcuflagsptr place to put the last required code unit flags |
5378 | bcptr points to current branch chain |
5379 | cb contains pointers to tables etc. |
5380 | lengthptr NULL during the real compile phase |
5381 | points to length accumulator during pre-compile phase |
5382 | |
5383 | Returns: 0 There's been an error, *errorcodeptr is non-zero |
5384 | +1 Success, this branch must match at least one character |
5385 | -1 Success, this branch may match an empty string |
5386 | */ |
5387 | |
5388 | static int |
5389 | compile_branch(uint32_t *optionsptr, PCRE2_UCHAR **codeptr, uint32_t **pptrptr, |
5390 | int *errorcodeptr, uint32_t *firstcuptr, uint32_t *firstcuflagsptr, |
5391 | uint32_t *reqcuptr, uint32_t *reqcuflagsptr, branch_chain *bcptr, |
5392 | compile_block *cb, PCRE2_SIZE *lengthptr) |
5393 | { |
5394 | int bravalue = 0; |
5395 | int okreturn = -1; |
5396 | int group_return = 0; |
5397 | uint32_t repeat_min = 0, repeat_max = 0; /* To please picky compilers */ |
5398 | uint32_t greedy_default, greedy_non_default; |
5399 | uint32_t repeat_type, op_type; |
5400 | uint32_t options = *optionsptr; /* May change dynamically */ |
5401 | uint32_t firstcu, reqcu; |
5402 | uint32_t zeroreqcu, zerofirstcu; |
5403 | uint32_t escape; |
5404 | uint32_t *pptr = *pptrptr; |
5405 | uint32_t meta, meta_arg; |
5406 | uint32_t firstcuflags, reqcuflags; |
5407 | uint32_t zeroreqcuflags, zerofirstcuflags; |
5408 | uint32_t req_caseopt, reqvary, tempreqvary; |
5409 | PCRE2_SIZE offset = 0; |
5410 | PCRE2_SIZE length_prevgroup = 0; |
5411 | PCRE2_UCHAR *code = *codeptr; |
5412 | PCRE2_UCHAR *last_code = code; |
5413 | PCRE2_UCHAR *orig_code = code; |
5414 | PCRE2_UCHAR *tempcode; |
5415 | PCRE2_UCHAR *previous = NULL; |
5416 | PCRE2_UCHAR op_previous; |
5417 | BOOL groupsetfirstcu = FALSE; |
5418 | BOOL had_accept = FALSE; |
5419 | BOOL matched_char = FALSE; |
5420 | BOOL previous_matched_char = FALSE; |
5421 | BOOL reset_caseful = FALSE; |
5422 | const uint8_t *cbits = cb->cbits; |
5423 | uint8_t classbits[32]; |
5424 | |
5425 | /* We can fish out the UTF setting once and for all into a BOOL, but we must |
5426 | not do this for other options (e.g. PCRE2_EXTENDED) because they may change |
5427 | dynamically as we process the pattern. */ |
5428 | |
5429 | #ifdef SUPPORT_UNICODE |
5430 | BOOL utf = (options & PCRE2_UTF) != 0; |
5431 | BOOL ucp = (options & PCRE2_UCP) != 0; |
5432 | #else /* No Unicode support */ |
5433 | BOOL utf = FALSE; |
5434 | #endif |
5435 | |
5436 | /* Helper variables for OP_XCLASS opcode (for characters > 255). We define |
5437 | class_uchardata always so that it can be passed to add_to_class() always, |
5438 | though it will not be used in non-UTF 8-bit cases. This avoids having to supply |
5439 | alternative calls for the different cases. */ |
5440 | |
5441 | PCRE2_UCHAR *class_uchardata; |
5442 | #ifdef SUPPORT_WIDE_CHARS |
5443 | BOOL xclass; |
5444 | PCRE2_UCHAR *class_uchardata_base; |
5445 | #endif |
5446 | |
5447 | /* Set up the default and non-default settings for greediness */ |
5448 | |
5449 | greedy_default = ((options & PCRE2_UNGREEDY) != 0); |
5450 | greedy_non_default = greedy_default ^ 1; |
5451 | |
5452 | /* Initialize no first unit, no required unit. REQ_UNSET means "no char |
5453 | matching encountered yet". It gets changed to REQ_NONE if we hit something that |
5454 | matches a non-fixed first unit; reqcu just remains unset if we never find one. |
5455 | |
5456 | When we hit a repeat whose minimum is zero, we may have to adjust these values |
5457 | to take the zero repeat into account. This is implemented by setting them to |
5458 | zerofirstcu and zeroreqcu when such a repeat is encountered. The individual |
5459 | item types that can be repeated set these backoff variables appropriately. */ |
5460 | |
5461 | firstcu = reqcu = zerofirstcu = zeroreqcu = 0; |
5462 | firstcuflags = reqcuflags = zerofirstcuflags = zeroreqcuflags = REQ_UNSET; |
5463 | |
5464 | /* The variable req_caseopt contains either the REQ_CASELESS bit or zero, |
5465 | according to the current setting of the caseless flag. The REQ_CASELESS value |
5466 | leaves the lower 28 bit empty. It is added into the firstcu or reqcu variables |
5467 | to record the case status of the value. This is used only for ASCII characters. |
5468 | */ |
5469 | |
5470 | req_caseopt = ((options & PCRE2_CASELESS) != 0)? REQ_CASELESS : 0; |
5471 | |
5472 | /* Switch on next META item until the end of the branch */ |
5473 | |
5474 | for (;; pptr++) |
5475 | { |
5476 | #ifdef SUPPORT_WIDE_CHARS |
5477 | BOOL xclass_has_prop; |
5478 | #endif |
5479 | BOOL negate_class; |
5480 | BOOL should_flip_negation; |
5481 | BOOL match_all_or_no_wide_chars; |
5482 | BOOL possessive_quantifier; |
5483 | BOOL note_group_empty; |
5484 | int class_has_8bitchar; |
5485 | uint32_t mclength; |
5486 | uint32_t skipunits; |
5487 | uint32_t subreqcu, subfirstcu; |
5488 | uint32_t groupnumber; |
5489 | uint32_t verbarglen, verbculen; |
5490 | uint32_t subreqcuflags, subfirstcuflags; |
5491 | open_capitem *oc; |
5492 | PCRE2_UCHAR mcbuffer[8]; |
5493 | |
5494 | /* Get next META item in the pattern and its potential argument. */ |
5495 | |
5496 | meta = META_CODE(*pptr); |
5497 | meta_arg = META_DATA(*pptr); |
5498 | |
5499 | /* If we are in the pre-compile phase, accumulate the length used for the |
5500 | previous cycle of this loop, unless the next item is a quantifier. */ |
5501 | |
5502 | if (lengthptr != NULL) |
5503 | { |
5504 | if (code > cb->start_workspace + cb->workspace_size - |
5505 | WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */ |
5506 | { |
5507 | *errorcodeptr = (code >= cb->start_workspace + cb->workspace_size)? |
5508 | ERR52 : ERR86; |
5509 | return 0; |
5510 | } |
5511 | |
5512 | /* There is at least one situation where code goes backwards: this is the |
5513 | case of a zero quantifier after a class (e.g. [ab]{0}). When the quantifier |
5514 | is processed, the whole class is eliminated. However, it is created first, |
5515 | so we have to allow memory for it. Therefore, don't ever reduce the length |
5516 | at this point. */ |
5517 | |
5518 | if (code < last_code) code = last_code; |
5519 | |
5520 | /* If the next thing is not a quantifier, we add the length of the previous |
5521 | item into the total, and reset the code pointer to the start of the |
5522 | workspace. Otherwise leave the previous item available to be quantified. */ |
5523 | |
5524 | if (meta < META_ASTERISK || meta > META_MINMAX_QUERY) |
5525 | { |
5526 | if (OFLOW_MAX - *lengthptr < (PCRE2_SIZE)(code - orig_code)) |
5527 | { |
5528 | *errorcodeptr = ERR20; /* Integer overflow */ |
5529 | return 0; |
5530 | } |
5531 | *lengthptr += (PCRE2_SIZE)(code - orig_code); |
5532 | if (*lengthptr > MAX_PATTERN_SIZE) |
5533 | { |
5534 | *errorcodeptr = ERR20; /* Pattern is too large */ |
5535 | return 0; |
5536 | } |
5537 | code = orig_code; |
5538 | } |
5539 | |
5540 | /* Remember where this code item starts so we can catch the "backwards" |
5541 | case above next time round. */ |
5542 | |
5543 | last_code = code; |
5544 | } |
5545 | |
5546 | /* Process the next parsed pattern item. If it is not a quantifier, remember |
5547 | where it starts so that it can be quantified when a quantifier follows. |
5548 | Checking for the legality of quantifiers happens in parse_regex(), except for |
5549 | a quantifier after an assertion that is a condition. */ |
5550 | |
5551 | if (meta < META_ASTERISK || meta > META_MINMAX_QUERY) |
5552 | { |
5553 | previous = code; |
5554 | if (matched_char && !had_accept) okreturn = 1; |
5555 | } |
5556 | |
5557 | previous_matched_char = matched_char; |
5558 | matched_char = FALSE; |
5559 | note_group_empty = FALSE; |
5560 | skipunits = 0; /* Default value for most subgroups */ |
5561 | |
5562 | switch(meta) |
5563 | { |
5564 | /* ===================================================================*/ |
5565 | /* The branch terminates at pattern end or | or ) */ |
5566 | |
5567 | case META_END: |
5568 | case META_ALT: |
5569 | case META_KET: |
5570 | *firstcuptr = firstcu; |
5571 | *firstcuflagsptr = firstcuflags; |
5572 | *reqcuptr = reqcu; |
5573 | *reqcuflagsptr = reqcuflags; |
5574 | *codeptr = code; |
5575 | *pptrptr = pptr; |
5576 | return okreturn; |
5577 | |
5578 | |
5579 | /* ===================================================================*/ |
5580 | /* Handle single-character metacharacters. In multiline mode, ^ disables |
5581 | the setting of any following char as a first character. */ |
5582 | |
5583 | case META_CIRCUMFLEX: |
5584 | if ((options & PCRE2_MULTILINE) != 0) |
5585 | { |
5586 | if (firstcuflags == REQ_UNSET) |
5587 | zerofirstcuflags = firstcuflags = REQ_NONE; |
5588 | *code++ = OP_CIRCM; |
5589 | } |
5590 | else *code++ = OP_CIRC; |
5591 | break; |
5592 | |
5593 | case META_DOLLAR: |
5594 | *code++ = ((options & PCRE2_MULTILINE) != 0)? OP_DOLLM : OP_DOLL; |
5595 | break; |
5596 | |
5597 | /* There can never be a first char if '.' is first, whatever happens about |
5598 | repeats. The value of reqcu doesn't change either. */ |
5599 | |
5600 | case META_DOT: |
5601 | matched_char = TRUE; |
5602 | if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE; |
5603 | zerofirstcu = firstcu; |
5604 | zerofirstcuflags = firstcuflags; |
5605 | zeroreqcu = reqcu; |
5606 | zeroreqcuflags = reqcuflags; |
5607 | *code++ = ((options & PCRE2_DOTALL) != 0)? OP_ALLANY: OP_ANY; |
5608 | break; |
5609 | |
5610 | |
5611 | /* ===================================================================*/ |
5612 | /* Empty character classes are allowed if PCRE2_ALLOW_EMPTY_CLASS is set. |
5613 | Otherwise, an initial ']' is taken as a data character. When empty classes |
5614 | are allowed, [] must always fail, so generate OP_FAIL, whereas [^] must |
5615 | match any character, so generate OP_ALLANY. */ |
5616 | |
5617 | case META_CLASS_EMPTY: |
5618 | case META_CLASS_EMPTY_NOT: |
5619 | matched_char = TRUE; |
5620 | *code++ = (meta == META_CLASS_EMPTY_NOT)? OP_ALLANY : OP_FAIL; |
5621 | if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE; |
5622 | zerofirstcu = firstcu; |
5623 | zerofirstcuflags = firstcuflags; |
5624 | break; |
5625 | |
5626 | |
5627 | /* ===================================================================*/ |
5628 | /* Non-empty character class. If the included characters are all < 256, we |
5629 | build a 32-byte bitmap of the permitted characters, except in the special |
5630 | case where there is only one such character. For negated classes, we build |
5631 | the map as usual, then invert it at the end. However, we use a different |
5632 | opcode so that data characters > 255 can be handled correctly. |
5633 | |
5634 | If the class contains characters outside the 0-255 range, a different |
5635 | opcode is compiled. It may optionally have a bit map for characters < 256, |
5636 | but those above are are explicitly listed afterwards. A flag code unit |
5637 | tells whether the bitmap is present, and whether this is a negated class or |
5638 | not. */ |
5639 | |
5640 | case META_CLASS_NOT: |
5641 | case META_CLASS: |
5642 | matched_char = TRUE; |
5643 | negate_class = meta == META_CLASS_NOT; |
5644 | |
5645 | /* We can optimize the case of a single character in a class by generating |
5646 | OP_CHAR or OP_CHARI if it's positive, or OP_NOT or OP_NOTI if it's |
5647 | negative. In the negative case there can be no first char if this item is |
5648 | first, whatever repeat count may follow. In the case of reqcu, save the |
5649 | previous value for reinstating. */ |
5650 | |
5651 | /* NOTE: at present this optimization is not effective if the only |
5652 | character in a class in 32-bit, non-UCP mode has its top bit set. */ |
5653 | |
5654 | if (pptr[1] < META_END && pptr[2] == META_CLASS_END) |
5655 | { |
5656 | #ifdef SUPPORT_UNICODE |
5657 | uint32_t d; |
5658 | #endif |
5659 | uint32_t c = pptr[1]; |
5660 | |
5661 | pptr += 2; /* Move on to class end */ |
5662 | if (meta == META_CLASS) /* A positive one-char class can be */ |
5663 | { /* handled as a normal literal character. */ |
5664 | meta = c; /* Set up the character */ |
5665 | goto NORMAL_CHAR_SET; |
5666 | } |
5667 | |
5668 | /* Handle a negative one-character class */ |
5669 | |
5670 | zeroreqcu = reqcu; |
5671 | zeroreqcuflags = reqcuflags; |
5672 | if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE; |
5673 | zerofirstcu = firstcu; |
5674 | zerofirstcuflags = firstcuflags; |
5675 | |
5676 | /* For caseless UTF or UCP mode, check whether this character has more |
5677 | than one other case. If so, generate a special OP_NOTPROP item instead of |
5678 | OP_NOTI. */ |
5679 | |
5680 | #ifdef SUPPORT_UNICODE |
5681 | if ((utf||ucp) && (options & PCRE2_CASELESS) != 0 && |
5682 | (d = UCD_CASESET(c)) != 0) |
5683 | { |
5684 | *code++ = OP_NOTPROP; |
5685 | *code++ = PT_CLIST; |
5686 | *code++ = d; |
5687 | break; /* We are finished with this class */ |
5688 | } |
5689 | #endif |
5690 | /* Char has only one other case, or UCP not available */ |
5691 | |
5692 | *code++ = ((options & PCRE2_CASELESS) != 0)? OP_NOTI: OP_NOT; |
5693 | code += PUTCHAR(c, code); |
5694 | break; /* We are finished with this class */ |
5695 | } /* End of 1-char optimization */ |
5696 | |
5697 | /* Handle character classes that contain more than just one literal |
5698 | character. If there are exactly two characters in a positive class, see if |
5699 | they are case partners. This can be optimized to generate a caseless single |
5700 | character match (which also sets first/required code units if relevant). */ |
5701 | |
5702 | if (meta == META_CLASS && pptr[1] < META_END && pptr[2] < META_END && |
5703 | pptr[3] == META_CLASS_END) |
5704 | { |
5705 | uint32_t c = pptr[1]; |
5706 | |
5707 | #ifdef SUPPORT_UNICODE |
5708 | if (UCD_CASESET(c) == 0) |
5709 | #endif |
5710 | { |
5711 | uint32_t d; |
5712 | |
5713 | #ifdef SUPPORT_UNICODE |
5714 | if ((utf || ucp) && c > 127) d = UCD_OTHERCASE(c); else |
5715 | #endif |
5716 | { |
5717 | #if PCRE2_CODE_UNIT_WIDTH != 8 |
5718 | if (c > 255) d = c; else |
5719 | #endif |
5720 | d = TABLE_GET(c, cb->fcc, c); |
5721 | } |
5722 | |
5723 | if (c != d && pptr[2] == d) |
5724 | { |
5725 | pptr += 3; /* Move on to class end */ |
5726 | meta = c; |
5727 | if ((options & PCRE2_CASELESS) == 0) |
5728 | { |
5729 | reset_caseful = TRUE; |
5730 | options |= PCRE2_CASELESS; |
5731 | req_caseopt = REQ_CASELESS; |
5732 | } |
5733 | goto CLASS_CASELESS_CHAR; |
5734 | } |
5735 | } |
5736 | } |
5737 | |
5738 | /* If a non-extended class contains a negative special such as \S, we need |
5739 | to flip the negation flag at the end, so that support for characters > 255 |
5740 | works correctly (they are all included in the class). An extended class may |
5741 | need to insert specific matching or non-matching code for wide characters. |
5742 | */ |
5743 | |
5744 | should_flip_negation = match_all_or_no_wide_chars = FALSE; |
5745 | |
5746 | /* Extended class (xclass) will be used when characters > 255 |
5747 | might match. */ |
5748 | |
5749 | #ifdef SUPPORT_WIDE_CHARS |
5750 | xclass = FALSE; |
5751 | class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */ |
5752 | class_uchardata_base = class_uchardata; /* Save the start */ |
5753 | #endif |
5754 | |
5755 | /* For optimization purposes, we track some properties of the class: |
5756 | class_has_8bitchar will be non-zero if the class contains at least one |
5757 | character with a code point less than 256; xclass_has_prop will be TRUE if |
5758 | Unicode property checks are present in the class. */ |
5759 | |
5760 | class_has_8bitchar = 0; |
5761 | #ifdef SUPPORT_WIDE_CHARS |
5762 | xclass_has_prop = FALSE; |
5763 | #endif |
5764 | |
5765 | /* Initialize the 256-bit (32-byte) bit map to all zeros. We build the map |
5766 | in a temporary bit of memory, in case the class contains fewer than two |
5767 | 8-bit characters because in that case the compiled code doesn't use the bit |
5768 | map. */ |
5769 | |
5770 | memset(classbits, 0, 32 * sizeof(uint8_t)); |
5771 | |
5772 | /* Process items until META_CLASS_END is reached. */ |
5773 | |
5774 | while ((meta = *(++pptr)) != META_CLASS_END) |
5775 | { |
5776 | /* Handle POSIX classes such as [:alpha:] etc. */ |
5777 | |
5778 | if (meta == META_POSIX || meta == META_POSIX_NEG) |
5779 | { |
5780 | BOOL local_negate = (meta == META_POSIX_NEG); |
5781 | int posix_class = *(++pptr); |
5782 | int taboffset, tabopt; |
5783 | uint8_t pbits[32]; |
5784 | |
5785 | should_flip_negation = local_negate; /* Note negative special */ |
5786 | |
5787 | /* If matching is caseless, upper and lower are converted to alpha. |
5788 | This relies on the fact that the class table starts with alpha, |
5789 | lower, upper as the first 3 entries. */ |
5790 | |
5791 | if ((options & PCRE2_CASELESS) != 0 && posix_class <= 2) |
5792 | posix_class = 0; |
5793 | |
5794 | /* When PCRE2_UCP is set, some of the POSIX classes are converted to |
5795 | different escape sequences that use Unicode properties \p or \P. |
5796 | Others that are not available via \p or \P have to generate |
5797 | XCL_PROP/XCL_NOTPROP directly, which is done here. */ |
5798 | |
5799 | #ifdef SUPPORT_UNICODE |
5800 | if ((options & PCRE2_UCP) != 0) switch(posix_class) |
5801 | { |
5802 | case PC_GRAPH: |
5803 | case PC_PRINT: |
5804 | case PC_PUNCT: |
5805 | *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP; |
5806 | *class_uchardata++ = (PCRE2_UCHAR) |
5807 | ((posix_class == PC_GRAPH)? PT_PXGRAPH : |
5808 | (posix_class == PC_PRINT)? PT_PXPRINT : PT_PXPUNCT); |
5809 | *class_uchardata++ = 0; |
5810 | xclass_has_prop = TRUE; |
5811 | goto CONTINUE_CLASS; |
5812 | |
5813 | /* For the other POSIX classes (ascii, xdigit) we are going to |
5814 | fall through to the non-UCP case and build a bit map for |
5815 | characters with code points less than 256. However, if we are in |
5816 | a negated POSIX class, characters with code points greater than |
5817 | 255 must either all match or all not match, depending on whether |
5818 | the whole class is not or is negated. For example, for |
5819 | [[:^ascii:]... they must all match, whereas for [^[:^xdigit:]... |
5820 | they must not. |
5821 | |
5822 | In the special case where there are no xclass items, this is |
5823 | automatically handled by the use of OP_CLASS or OP_NCLASS, but an |
5824 | explicit range is needed for OP_XCLASS. Setting a flag here |
5825 | causes the range to be generated later when it is known that |
5826 | OP_XCLASS is required. In the 8-bit library this is relevant only in |
5827 | utf mode, since no wide characters can exist otherwise. */ |
5828 | |
5829 | default: |
5830 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
5831 | if (utf) |
5832 | #endif |
5833 | match_all_or_no_wide_chars |= local_negate; |
5834 | break; |
5835 | } |
5836 | #endif /* SUPPORT_UNICODE */ |
5837 | |
5838 | /* In the non-UCP case, or when UCP makes no difference, we build the |
5839 | bit map for the POSIX class in a chunk of local store because we may |
5840 | be adding and subtracting from it, and we don't want to subtract bits |
5841 | that may be in the main map already. At the end we or the result into |
5842 | the bit map that is being built. */ |
5843 | |
5844 | posix_class *= 3; |
5845 | |
5846 | /* Copy in the first table (always present) */ |
5847 | |
5848 | memcpy(pbits, cbits + posix_class_maps[posix_class], |
5849 | 32 * sizeof(uint8_t)); |
5850 | |
5851 | /* If there is a second table, add or remove it as required. */ |
5852 | |
5853 | taboffset = posix_class_maps[posix_class + 1]; |
5854 | tabopt = posix_class_maps[posix_class + 2]; |
5855 | |
5856 | if (taboffset >= 0) |
5857 | { |
5858 | if (tabopt >= 0) |
5859 | for (int i = 0; i < 32; i++) pbits[i] |= cbits[(int)i + taboffset]; |
5860 | else |
5861 | for (int i = 0; i < 32; i++) pbits[i] &= ~cbits[(int)i + taboffset]; |
5862 | } |
5863 | |
5864 | /* Now see if we need to remove any special characters. An option |
5865 | value of 1 removes vertical space and 2 removes underscore. */ |
5866 | |
5867 | if (tabopt < 0) tabopt = -tabopt; |
5868 | if (tabopt == 1) pbits[1] &= ~0x3c; |
5869 | else if (tabopt == 2) pbits[11] &= 0x7f; |
5870 | |
5871 | /* Add the POSIX table or its complement into the main table that is |
5872 | being built and we are done. */ |
5873 | |
5874 | if (local_negate) |
5875 | for (int i = 0; i < 32; i++) classbits[i] |= (uint8_t)(~pbits[i]); |
5876 | else |
5877 | for (int i = 0; i < 32; i++) classbits[i] |= pbits[i]; |
5878 | |
5879 | /* Every class contains at least one < 256 character. */ |
5880 | |
5881 | class_has_8bitchar = 1; |
5882 | goto CONTINUE_CLASS; /* End of POSIX handling */ |
5883 | } |
5884 | |
5885 | /* Other than POSIX classes, the only items we should encounter are |
5886 | \d-type escapes and literal characters (possibly as ranges). */ |
5887 | |
5888 | if (meta == META_BIGVALUE) |
5889 | { |
5890 | meta = *(++pptr); |
5891 | goto CLASS_LITERAL; |
5892 | } |
5893 | |
5894 | /* Any other non-literal must be an escape */ |
5895 | |
5896 | if (meta >= META_END) |
5897 | { |
5898 | if (META_CODE(meta) != META_ESCAPE) |
5899 | { |
5900 | #ifdef DEBUG_SHOW_PARSED |
5901 | fprintf(stderr, "** Unrecognized parsed pattern item 0x%.8x " |
5902 | "in character class\n" , meta); |
5903 | #endif |
5904 | *errorcodeptr = ERR89; /* Internal error - unrecognized. */ |
5905 | return 0; |
5906 | } |
5907 | escape = META_DATA(meta); |
5908 | |
5909 | /* Every class contains at least one < 256 character. */ |
5910 | |
5911 | class_has_8bitchar++; |
5912 | |
5913 | switch(escape) |
5914 | { |
5915 | case ESC_d: |
5916 | for (int i = 0; i < 32; i++) classbits[i] |= cbits[i+cbit_digit]; |
5917 | break; |
5918 | |
5919 | case ESC_D: |
5920 | should_flip_negation = TRUE; |
5921 | for (int i = 0; i < 32; i++) |
5922 | classbits[i] |= (uint8_t)(~cbits[i+cbit_digit]); |
5923 | break; |
5924 | |
5925 | case ESC_w: |
5926 | for (int i = 0; i < 32; i++) classbits[i] |= cbits[i+cbit_word]; |
5927 | break; |
5928 | |
5929 | case ESC_W: |
5930 | should_flip_negation = TRUE; |
5931 | for (int i = 0; i < 32; i++) |
5932 | classbits[i] |= (uint8_t)(~cbits[i+cbit_word]); |
5933 | break; |
5934 | |
5935 | /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl |
5936 | 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was |
5937 | previously set by something earlier in the character class. |
5938 | Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so |
5939 | we could just adjust the appropriate bit. From PCRE 8.34 we no |
5940 | longer treat \s and \S specially. */ |
5941 | |
5942 | case ESC_s: |
5943 | for (int i = 0; i < 32; i++) classbits[i] |= cbits[i+cbit_space]; |
5944 | break; |
5945 | |
5946 | case ESC_S: |
5947 | should_flip_negation = TRUE; |
5948 | for (int i = 0; i < 32; i++) |
5949 | classbits[i] |= (uint8_t)(~cbits[i+cbit_space]); |
5950 | break; |
5951 | |
5952 | /* When adding the horizontal or vertical space lists to a class, or |
5953 | their complements, disable PCRE2_CASELESS, because it justs wastes |
5954 | time, and in the "not-x" UTF cases can create unwanted duplicates in |
5955 | the XCLASS list (provoked by characters that have more than one other |
5956 | case and by both cases being in the same "not-x" sublist). */ |
5957 | |
5958 | case ESC_h: |
5959 | (void)add_list_to_class(classbits, &class_uchardata, |
5960 | options & ~PCRE2_CASELESS, cb, PRIV(hspace_list), NOTACHAR); |
5961 | break; |
5962 | |
5963 | case ESC_H: |
5964 | (void)add_not_list_to_class(classbits, &class_uchardata, |
5965 | options & ~PCRE2_CASELESS, cb, PRIV(hspace_list)); |
5966 | break; |
5967 | |
5968 | case ESC_v: |
5969 | (void)add_list_to_class(classbits, &class_uchardata, |
5970 | options & ~PCRE2_CASELESS, cb, PRIV(vspace_list), NOTACHAR); |
5971 | break; |
5972 | |
5973 | case ESC_V: |
5974 | (void)add_not_list_to_class(classbits, &class_uchardata, |
5975 | options & ~PCRE2_CASELESS, cb, PRIV(vspace_list)); |
5976 | break; |
5977 | |
5978 | /* If Unicode is not supported, \P and \p are not allowed and are |
5979 | faulted at parse time, so will never appear here. */ |
5980 | |
5981 | #ifdef SUPPORT_UNICODE |
5982 | case ESC_p: |
5983 | case ESC_P: |
5984 | { |
5985 | uint32_t ptype = *(++pptr) >> 16; |
5986 | uint32_t pdata = *pptr & 0xffff; |
5987 | *class_uchardata++ = (escape == ESC_p)? XCL_PROP : XCL_NOTPROP; |
5988 | *class_uchardata++ = ptype; |
5989 | *class_uchardata++ = pdata; |
5990 | xclass_has_prop = TRUE; |
5991 | class_has_8bitchar--; /* Undo! */ |
5992 | } |
5993 | break; |
5994 | #endif |
5995 | } |
5996 | |
5997 | goto CONTINUE_CLASS; |
5998 | } /* End handling \d-type escapes */ |
5999 | |
6000 | /* A literal character may be followed by a range meta. At parse time |
6001 | there are checks for out-of-order characters, for ranges where the two |
6002 | characters are equal, and for hyphens that cannot indicate a range. At |
6003 | this point, therefore, no checking is needed. */ |
6004 | |
6005 | else |
6006 | { |
6007 | uint32_t c, d; |
6008 | |
6009 | CLASS_LITERAL: |
6010 | c = d = meta; |
6011 | |
6012 | /* Remember if \r or \n were explicitly used */ |
6013 | |
6014 | if (c == CHAR_CR || c == CHAR_NL) cb->external_flags |= PCRE2_HASCRORLF; |
6015 | |
6016 | /* Process a character range */ |
6017 | |
6018 | if (pptr[1] == META_RANGE_LITERAL || pptr[1] == META_RANGE_ESCAPED) |
6019 | { |
6020 | #ifdef EBCDIC |
6021 | BOOL range_is_literal = (pptr[1] == META_RANGE_LITERAL); |
6022 | #endif |
6023 | pptr += 2; |
6024 | d = *pptr; |
6025 | if (d == META_BIGVALUE) d = *(++pptr); |
6026 | |
6027 | /* Remember an explicit \r or \n, and add the range to the class. */ |
6028 | |
6029 | if (d == CHAR_CR || d == CHAR_NL) cb->external_flags |= PCRE2_HASCRORLF; |
6030 | |
6031 | /* In an EBCDIC environment, Perl treats alphabetic ranges specially |
6032 | because there are holes in the encoding, and simply using the range |
6033 | A-Z (for example) would include the characters in the holes. This |
6034 | applies only to literal ranges; [\xC1-\xE9] is different to [A-Z]. */ |
6035 | |
6036 | #ifdef EBCDIC |
6037 | if (range_is_literal && |
6038 | (cb->ctypes[c] & ctype_letter) != 0 && |
6039 | (cb->ctypes[d] & ctype_letter) != 0 && |
6040 | (c <= CHAR_z) == (d <= CHAR_z)) |
6041 | { |
6042 | uint32_t uc = (d <= CHAR_z)? 0 : 64; |
6043 | uint32_t C = c - uc; |
6044 | uint32_t D = d - uc; |
6045 | |
6046 | if (C <= CHAR_i) |
6047 | { |
6048 | class_has_8bitchar += |
6049 | add_to_class(classbits, &class_uchardata, options, cb, C + uc, |
6050 | ((D < CHAR_i)? D : CHAR_i) + uc); |
6051 | C = CHAR_j; |
6052 | } |
6053 | |
6054 | if (C <= D && C <= CHAR_r) |
6055 | { |
6056 | class_has_8bitchar += |
6057 | add_to_class(classbits, &class_uchardata, options, cb, C + uc, |
6058 | ((D < CHAR_r)? D : CHAR_r) + uc); |
6059 | C = CHAR_s; |
6060 | } |
6061 | |
6062 | if (C <= D) |
6063 | { |
6064 | class_has_8bitchar += |
6065 | add_to_class(classbits, &class_uchardata, options, cb, C + uc, |
6066 | D + uc); |
6067 | } |
6068 | } |
6069 | else |
6070 | #endif |
6071 | /* Not an EBCDIC special range */ |
6072 | |
6073 | class_has_8bitchar += |
6074 | add_to_class(classbits, &class_uchardata, options, cb, c, d); |
6075 | goto CONTINUE_CLASS; /* Go get the next char in the class */ |
6076 | } /* End of range handling */ |
6077 | |
6078 | |
6079 | /* Handle a single character. */ |
6080 | |
6081 | class_has_8bitchar += |
6082 | add_to_class(classbits, &class_uchardata, options, cb, meta, meta); |
6083 | } |
6084 | |
6085 | /* Continue to the next item in the class. */ |
6086 | |
6087 | CONTINUE_CLASS: |
6088 | |
6089 | #ifdef SUPPORT_WIDE_CHARS |
6090 | /* If any wide characters or Unicode properties have been encountered, |
6091 | set xclass = TRUE. Then, in the pre-compile phase, accumulate the length |
6092 | of the extra data and reset the pointer. This is so that very large |
6093 | classes that contain a zillion wide characters or Unicode property tests |
6094 | do not overwrite the workspace (which is on the stack). */ |
6095 | |
6096 | if (class_uchardata > class_uchardata_base) |
6097 | { |
6098 | xclass = TRUE; |
6099 | if (lengthptr != NULL) |
6100 | { |
6101 | *lengthptr += class_uchardata - class_uchardata_base; |
6102 | class_uchardata = class_uchardata_base; |
6103 | } |
6104 | } |
6105 | #endif |
6106 | |
6107 | continue; /* Needed to avoid error when not supporting wide chars */ |
6108 | } /* End of main class-processing loop */ |
6109 | |
6110 | /* If this class is the first thing in the branch, there can be no first |
6111 | char setting, whatever the repeat count. Any reqcu setting must remain |
6112 | unchanged after any kind of repeat. */ |
6113 | |
6114 | if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE; |
6115 | zerofirstcu = firstcu; |
6116 | zerofirstcuflags = firstcuflags; |
6117 | zeroreqcu = reqcu; |
6118 | zeroreqcuflags = reqcuflags; |
6119 | |
6120 | /* If there are characters with values > 255, or Unicode property settings |
6121 | (\p or \P), we have to compile an extended class, with its own opcode, |
6122 | unless there were no property settings and there was a negated special such |
6123 | as \S in the class, and PCRE2_UCP is not set, because in that case all |
6124 | characters > 255 are in or not in the class, so any that were explicitly |
6125 | given as well can be ignored. |
6126 | |
6127 | In the UCP case, if certain negated POSIX classes ([:^ascii:] or |
6128 | [^:xdigit:]) were present in a class, we either have to match or not match |
6129 | all wide characters (depending on whether the whole class is or is not |
6130 | negated). This requirement is indicated by match_all_or_no_wide_chars being |
6131 | true. We do this by including an explicit range, which works in both cases. |
6132 | This applies only in UTF and 16-bit and 32-bit non-UTF modes, since there |
6133 | cannot be any wide characters in 8-bit non-UTF mode. |
6134 | |
6135 | When there *are* properties in a positive UTF-8 or any 16-bit or 32_bit |
6136 | class where \S etc is present without PCRE2_UCP, causing an extended class |
6137 | to be compiled, we make sure that all characters > 255 are included by |
6138 | forcing match_all_or_no_wide_chars to be true. |
6139 | |
6140 | If, when generating an xclass, there are no characters < 256, we can omit |
6141 | the bitmap in the actual compiled code. */ |
6142 | |
6143 | #ifdef SUPPORT_WIDE_CHARS /* Defined for 16/32 bits, or 8-bit with Unicode */ |
6144 | if (xclass && ( |
6145 | #ifdef SUPPORT_UNICODE |
6146 | (options & PCRE2_UCP) != 0 || |
6147 | #endif |
6148 | xclass_has_prop || !should_flip_negation)) |
6149 | { |
6150 | if (match_all_or_no_wide_chars || ( |
6151 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
6152 | utf && |
6153 | #endif |
6154 | should_flip_negation && !negate_class && (options & PCRE2_UCP) == 0)) |
6155 | { |
6156 | *class_uchardata++ = XCL_RANGE; |
6157 | if (utf) /* Will always be utf in the 8-bit library */ |
6158 | { |
6159 | class_uchardata += PRIV(ord2utf)(0x100, class_uchardata); |
6160 | class_uchardata += PRIV(ord2utf)(MAX_UTF_CODE_POINT, class_uchardata); |
6161 | } |
6162 | else /* Can only happen for the 16-bit & 32-bit libraries */ |
6163 | { |
6164 | #if PCRE2_CODE_UNIT_WIDTH == 16 |
6165 | *class_uchardata++ = 0x100; |
6166 | *class_uchardata++ = 0xffffu; |
6167 | #elif PCRE2_CODE_UNIT_WIDTH == 32 |
6168 | *class_uchardata++ = 0x100; |
6169 | *class_uchardata++ = 0xffffffffu; |
6170 | #endif |
6171 | } |
6172 | } |
6173 | *class_uchardata++ = XCL_END; /* Marks the end of extra data */ |
6174 | *code++ = OP_XCLASS; |
6175 | code += LINK_SIZE; |
6176 | *code = negate_class? XCL_NOT:0; |
6177 | if (xclass_has_prop) *code |= XCL_HASPROP; |
6178 | |
6179 | /* If the map is required, move up the extra data to make room for it; |
6180 | otherwise just move the code pointer to the end of the extra data. */ |
6181 | |
6182 | if (class_has_8bitchar > 0) |
6183 | { |
6184 | *code++ |= XCL_MAP; |
6185 | (void)memmove(code + (32 / sizeof(PCRE2_UCHAR)), code, |
6186 | CU2BYTES(class_uchardata - code)); |
6187 | if (negate_class && !xclass_has_prop) |
6188 | { |
6189 | /* Using 255 ^ instead of ~ avoids clang sanitize warning. */ |
6190 | for (int i = 0; i < 32; i++) classbits[i] = 255 ^ classbits[i]; |
6191 | } |
6192 | memcpy(code, classbits, 32); |
6193 | code = class_uchardata + (32 / sizeof(PCRE2_UCHAR)); |
6194 | } |
6195 | else code = class_uchardata; |
6196 | |
6197 | /* Now fill in the complete length of the item */ |
6198 | |
6199 | PUT(previous, 1, (int)(code - previous)); |
6200 | break; /* End of class handling */ |
6201 | } |
6202 | #endif /* SUPPORT_WIDE_CHARS */ |
6203 | |
6204 | /* If there are no characters > 255, or they are all to be included or |
6205 | excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the |
6206 | whole class was negated and whether there were negative specials such as \S |
6207 | (non-UCP) in the class. Then copy the 32-byte map into the code vector, |
6208 | negating it if necessary. */ |
6209 | |
6210 | *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS; |
6211 | if (lengthptr == NULL) /* Save time in the pre-compile phase */ |
6212 | { |
6213 | if (negate_class) |
6214 | { |
6215 | /* Using 255 ^ instead of ~ avoids clang sanitize warning. */ |
6216 | for (int i = 0; i < 32; i++) classbits[i] = 255 ^ classbits[i]; |
6217 | } |
6218 | memcpy(code, classbits, 32); |
6219 | } |
6220 | code += 32 / sizeof(PCRE2_UCHAR); |
6221 | break; /* End of class processing */ |
6222 | |
6223 | |
6224 | /* ===================================================================*/ |
6225 | /* Deal with (*VERB)s. */ |
6226 | |
6227 | /* Check for open captures before ACCEPT and close those that are within |
6228 | the same assertion level, also converting ACCEPT to ASSERT_ACCEPT in an |
6229 | assertion. In the first pass, just accumulate the length required; |
6230 | otherwise hitting (*ACCEPT) inside many nested parentheses can cause |
6231 | workspace overflow. Do not set firstcu after *ACCEPT. */ |
6232 | |
6233 | case META_ACCEPT: |
6234 | cb->had_accept = had_accept = TRUE; |
6235 | for (oc = cb->open_caps; |
6236 | oc != NULL && oc->assert_depth >= cb->assert_depth; |
6237 | oc = oc->next) |
6238 | { |
6239 | if (lengthptr != NULL) |
6240 | { |
6241 | *lengthptr += CU2BYTES(1) + IMM2_SIZE; |
6242 | } |
6243 | else |
6244 | { |
6245 | *code++ = OP_CLOSE; |
6246 | PUT2INC(code, 0, oc->number); |
6247 | } |
6248 | } |
6249 | *code++ = (cb->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT; |
6250 | if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE; |
6251 | break; |
6252 | |
6253 | case META_PRUNE: |
6254 | case META_SKIP: |
6255 | cb->had_pruneorskip = TRUE; |
6256 | /* Fall through */ |
6257 | case META_COMMIT: |
6258 | case META_FAIL: |
6259 | *code++ = verbops[(meta - META_MARK) >> 16]; |
6260 | break; |
6261 | |
6262 | case META_THEN: |
6263 | cb->external_flags |= PCRE2_HASTHEN; |
6264 | *code++ = OP_THEN; |
6265 | break; |
6266 | |
6267 | /* Handle verbs with arguments. Arguments can be very long, especially in |
6268 | 16- and 32-bit modes, and can overflow the workspace in the first pass. |
6269 | However, the argument length is constrained to be small enough to fit in |
6270 | one code unit. This check happens in parse_regex(). In the first pass, |
6271 | instead of putting the argument into memory, we just update the length |
6272 | counter and set up an empty argument. */ |
6273 | |
6274 | case META_THEN_ARG: |
6275 | cb->external_flags |= PCRE2_HASTHEN; |
6276 | goto VERB_ARG; |
6277 | |
6278 | case META_PRUNE_ARG: |
6279 | case META_SKIP_ARG: |
6280 | cb->had_pruneorskip = TRUE; |
6281 | /* Fall through */ |
6282 | case META_MARK: |
6283 | case META_COMMIT_ARG: |
6284 | VERB_ARG: |
6285 | *code++ = verbops[(meta - META_MARK) >> 16]; |
6286 | /* The length is in characters. */ |
6287 | verbarglen = *(++pptr); |
6288 | verbculen = 0; |
6289 | tempcode = code++; |
6290 | for (int i = 0; i < (int)verbarglen; i++) |
6291 | { |
6292 | meta = *(++pptr); |
6293 | #ifdef SUPPORT_UNICODE |
6294 | if (utf) mclength = PRIV(ord2utf)(meta, mcbuffer); else |
6295 | #endif |
6296 | { |
6297 | mclength = 1; |
6298 | mcbuffer[0] = meta; |
6299 | } |
6300 | if (lengthptr != NULL) *lengthptr += mclength; else |
6301 | { |
6302 | memcpy(code, mcbuffer, CU2BYTES(mclength)); |
6303 | code += mclength; |
6304 | verbculen += mclength; |
6305 | } |
6306 | } |
6307 | |
6308 | *tempcode = verbculen; /* Fill in the code unit length */ |
6309 | *code++ = 0; /* Terminating zero */ |
6310 | break; |
6311 | |
6312 | |
6313 | /* ===================================================================*/ |
6314 | /* Handle options change. The new setting must be passed back for use in |
6315 | subsequent branches. Reset the greedy defaults and the case value for |
6316 | firstcu and reqcu. */ |
6317 | |
6318 | case META_OPTIONS: |
6319 | *optionsptr = options = *(++pptr); |
6320 | greedy_default = ((options & PCRE2_UNGREEDY) != 0); |
6321 | greedy_non_default = greedy_default ^ 1; |
6322 | req_caseopt = ((options & PCRE2_CASELESS) != 0)? REQ_CASELESS : 0; |
6323 | break; |
6324 | |
6325 | |
6326 | /* ===================================================================*/ |
6327 | /* Handle conditional subpatterns. The case of (?(Rdigits) is ambiguous |
6328 | because it could be a numerical check on recursion, or a name check on a |
6329 | group's being set. The pre-pass sets up META_COND_RNUMBER as a name so that |
6330 | we can handle it either way. We first try for a name; if not found, process |
6331 | the number. */ |
6332 | |
6333 | case META_COND_RNUMBER: /* (?(Rdigits) */ |
6334 | case META_COND_NAME: /* (?(name) or (?'name') or ?(<name>) */ |
6335 | case META_COND_RNAME: /* (?(R&name) - test for recursion */ |
6336 | bravalue = OP_COND; |
6337 | { |
6338 | int count, index; |
6339 | unsigned int i; |
6340 | PCRE2_SPTR name; |
6341 | named_group *ng = cb->named_groups; |
6342 | uint32_t length = *(++pptr); |
6343 | |
6344 | GETPLUSOFFSET(offset, pptr); |
6345 | name = cb->start_pattern + offset; |
6346 | |
6347 | /* In the first pass, the names generated in the pre-pass are available, |
6348 | but the main name table has not yet been created. Scan the list of names |
6349 | generated in the pre-pass in order to get a number and whether or not |
6350 | this name is duplicated. If it is not duplicated, we can handle it as a |
6351 | numerical group. */ |
6352 | |
6353 | for (i = 0; i < cb->names_found; i++, ng++) |
6354 | { |
6355 | if (length == ng->length && |
6356 | PRIV(strncmp)(name, ng->name, length) == 0) |
6357 | { |
6358 | if (!ng->isdup) |
6359 | { |
6360 | code[1+LINK_SIZE] = (meta == META_COND_RNAME)? OP_RREF : OP_CREF; |
6361 | PUT2(code, 2+LINK_SIZE, ng->number); |
6362 | if (ng->number > cb->top_backref) cb->top_backref = ng->number; |
6363 | skipunits = 1+IMM2_SIZE; |
6364 | goto GROUP_PROCESS_NOTE_EMPTY; |
6365 | } |
6366 | break; /* Found a duplicated name */ |
6367 | } |
6368 | } |
6369 | |
6370 | /* If the name was not found we have a bad reference, unless we are |
6371 | dealing with R<digits>, which is treated as a recursion test by number. |
6372 | */ |
6373 | |
6374 | if (i >= cb->names_found) |
6375 | { |
6376 | groupnumber = 0; |
6377 | if (meta == META_COND_RNUMBER) |
6378 | { |
6379 | for (i = 1; i < length; i++) |
6380 | { |
6381 | groupnumber = groupnumber * 10 + name[i] - CHAR_0; |
6382 | if (groupnumber > MAX_GROUP_NUMBER) |
6383 | { |
6384 | *errorcodeptr = ERR61; |
6385 | cb->erroroffset = offset + i; |
6386 | return 0; |
6387 | } |
6388 | } |
6389 | } |
6390 | |
6391 | if (meta != META_COND_RNUMBER || groupnumber > cb->bracount) |
6392 | { |
6393 | *errorcodeptr = ERR15; |
6394 | cb->erroroffset = offset; |
6395 | return 0; |
6396 | } |
6397 | |
6398 | /* (?Rdigits) treated as a recursion reference by number. A value of |
6399 | zero (which is the result of both (?R) and (?R0)) means "any", and is |
6400 | translated into RREF_ANY (which is 0xffff). */ |
6401 | |
6402 | if (groupnumber == 0) groupnumber = RREF_ANY; |
6403 | code[1+LINK_SIZE] = OP_RREF; |
6404 | PUT2(code, 2+LINK_SIZE, groupnumber); |
6405 | skipunits = 1+IMM2_SIZE; |
6406 | goto GROUP_PROCESS_NOTE_EMPTY; |
6407 | } |
6408 | |
6409 | /* A duplicated name was found. Note that if an R<digits> name is found |
6410 | (META_COND_RNUMBER), it is a reference test, not a recursion test. */ |
6411 | |
6412 | code[1+LINK_SIZE] = (meta == META_COND_RNAME)? OP_RREF : OP_CREF; |
6413 | |
6414 | /* We have a duplicated name. In the compile pass we have to search the |
6415 | main table in order to get the index and count values. */ |
6416 | |
6417 | count = 0; /* Values for first pass (avoids compiler warning) */ |
6418 | index = 0; |
6419 | if (lengthptr == NULL && !find_dupname_details(name, length, &index, |
6420 | &count, errorcodeptr, cb)) return 0; |
6421 | |
6422 | /* Add one to the opcode to change CREF/RREF into DNCREF/DNRREF and |
6423 | insert appropriate data values. */ |
6424 | |
6425 | code[1+LINK_SIZE]++; |
6426 | skipunits = 1+2*IMM2_SIZE; |
6427 | PUT2(code, 2+LINK_SIZE, index); |
6428 | PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count); |
6429 | } |
6430 | goto GROUP_PROCESS_NOTE_EMPTY; |
6431 | |
6432 | /* The DEFINE condition is always false. Its internal groups may never |
6433 | be called, so matched_char must remain false, hence the jump to |
6434 | GROUP_PROCESS rather than GROUP_PROCESS_NOTE_EMPTY. */ |
6435 | |
6436 | case META_COND_DEFINE: |
6437 | bravalue = OP_COND; |
6438 | GETPLUSOFFSET(offset, pptr); |
6439 | code[1+LINK_SIZE] = OP_DEFINE; |
6440 | skipunits = 1; |
6441 | goto GROUP_PROCESS; |
6442 | |
6443 | /* Conditional test of a group's being set. */ |
6444 | |
6445 | case META_COND_NUMBER: |
6446 | bravalue = OP_COND; |
6447 | GETPLUSOFFSET(offset, pptr); |
6448 | groupnumber = *(++pptr); |
6449 | if (groupnumber > cb->bracount) |
6450 | { |
6451 | *errorcodeptr = ERR15; |
6452 | cb->erroroffset = offset; |
6453 | return 0; |
6454 | } |
6455 | if (groupnumber > cb->top_backref) cb->top_backref = groupnumber; |
6456 | offset -= 2; /* Point at initial ( for too many branches error */ |
6457 | code[1+LINK_SIZE] = OP_CREF; |
6458 | skipunits = 1+IMM2_SIZE; |
6459 | PUT2(code, 2+LINK_SIZE, groupnumber); |
6460 | goto GROUP_PROCESS_NOTE_EMPTY; |
6461 | |
6462 | /* Test for the PCRE2 version. */ |
6463 | |
6464 | case META_COND_VERSION: |
6465 | bravalue = OP_COND; |
6466 | if (pptr[1] > 0) |
6467 | code[1+LINK_SIZE] = ((PCRE2_MAJOR > pptr[2]) || |
6468 | (PCRE2_MAJOR == pptr[2] && PCRE2_MINOR >= pptr[3]))? |
6469 | OP_TRUE : OP_FALSE; |
6470 | else |
6471 | code[1+LINK_SIZE] = (PCRE2_MAJOR == pptr[2] && PCRE2_MINOR == pptr[3])? |
6472 | OP_TRUE : OP_FALSE; |
6473 | skipunits = 1; |
6474 | pptr += 3; |
6475 | goto GROUP_PROCESS_NOTE_EMPTY; |
6476 | |
6477 | /* The condition is an assertion, possibly preceded by a callout. */ |
6478 | |
6479 | case META_COND_ASSERT: |
6480 | bravalue = OP_COND; |
6481 | goto GROUP_PROCESS_NOTE_EMPTY; |
6482 | |
6483 | |
6484 | /* ===================================================================*/ |
6485 | /* Handle all kinds of nested bracketed groups. The non-capturing, |
6486 | non-conditional cases are here; others come to GROUP_PROCESS via goto. */ |
6487 | |
6488 | case META_LOOKAHEAD: |
6489 | bravalue = OP_ASSERT; |
6490 | cb->assert_depth += 1; |
6491 | goto GROUP_PROCESS; |
6492 | |
6493 | case META_LOOKAHEAD_NA: |
6494 | bravalue = OP_ASSERT_NA; |
6495 | cb->assert_depth += 1; |
6496 | goto GROUP_PROCESS; |
6497 | |
6498 | /* Optimize (?!) to (*FAIL) unless it is quantified - which is a weird |
6499 | thing to do, but Perl allows all assertions to be quantified, and when |
6500 | they contain capturing parentheses there may be a potential use for |
6501 | this feature. Not that that applies to a quantified (?!) but we allow |
6502 | it for uniformity. */ |
6503 | |
6504 | case META_LOOKAHEADNOT: |
6505 | if (pptr[1] == META_KET && |
6506 | (pptr[2] < META_ASTERISK || pptr[2] > META_MINMAX_QUERY)) |
6507 | { |
6508 | *code++ = OP_FAIL; |
6509 | pptr++; |
6510 | } |
6511 | else |
6512 | { |
6513 | bravalue = OP_ASSERT_NOT; |
6514 | cb->assert_depth += 1; |
6515 | goto GROUP_PROCESS; |
6516 | } |
6517 | break; |
6518 | |
6519 | case META_LOOKBEHIND: |
6520 | bravalue = OP_ASSERTBACK; |
6521 | cb->assert_depth += 1; |
6522 | goto GROUP_PROCESS; |
6523 | |
6524 | case META_LOOKBEHINDNOT: |
6525 | bravalue = OP_ASSERTBACK_NOT; |
6526 | cb->assert_depth += 1; |
6527 | goto GROUP_PROCESS; |
6528 | |
6529 | case META_LOOKBEHIND_NA: |
6530 | bravalue = OP_ASSERTBACK_NA; |
6531 | cb->assert_depth += 1; |
6532 | goto GROUP_PROCESS; |
6533 | |
6534 | case META_ATOMIC: |
6535 | bravalue = OP_ONCE; |
6536 | goto GROUP_PROCESS_NOTE_EMPTY; |
6537 | |
6538 | case META_SCRIPT_RUN: |
6539 | bravalue = OP_SCRIPT_RUN; |
6540 | goto GROUP_PROCESS_NOTE_EMPTY; |
6541 | |
6542 | case META_NOCAPTURE: |
6543 | bravalue = OP_BRA; |
6544 | /* Fall through */ |
6545 | |
6546 | /* Process nested bracketed regex. The nesting depth is maintained for the |
6547 | benefit of the stackguard function. The test for too deep nesting is now |
6548 | done in parse_regex(). Assertion and DEFINE groups come to GROUP_PROCESS; |
6549 | others come to GROUP_PROCESS_NOTE_EMPTY, to indicate that we need to take |
6550 | note of whether or not they may match an empty string. */ |
6551 | |
6552 | GROUP_PROCESS_NOTE_EMPTY: |
6553 | note_group_empty = TRUE; |
6554 | |
6555 | GROUP_PROCESS: |
6556 | cb->parens_depth += 1; |
6557 | *code = bravalue; |
6558 | pptr++; |
6559 | tempcode = code; |
6560 | tempreqvary = cb->req_varyopt; /* Save value before group */ |
6561 | length_prevgroup = 0; /* Initialize for pre-compile phase */ |
6562 | |
6563 | if ((group_return = |
6564 | compile_regex( |
6565 | options, /* The option state */ |
6566 | &tempcode, /* Where to put code (updated) */ |
6567 | &pptr, /* Input pointer (updated) */ |
6568 | errorcodeptr, /* Where to put an error message */ |
6569 | skipunits, /* Skip over bracket number */ |
6570 | &subfirstcu, /* For possible first char */ |
6571 | &subfirstcuflags, |
6572 | &subreqcu, /* For possible last char */ |
6573 | &subreqcuflags, |
6574 | bcptr, /* Current branch chain */ |
6575 | cb, /* Compile data block */ |
6576 | (lengthptr == NULL)? NULL : /* Actual compile phase */ |
6577 | &length_prevgroup /* Pre-compile phase */ |
6578 | )) == 0) |
6579 | return 0; /* Error */ |
6580 | |
6581 | cb->parens_depth -= 1; |
6582 | |
6583 | /* If that was a non-conditional significant group (not an assertion, not a |
6584 | DEFINE) that matches at least one character, then the current item matches |
6585 | a character. Conditionals are handled below. */ |
6586 | |
6587 | if (note_group_empty && bravalue != OP_COND && group_return > 0) |
6588 | matched_char = TRUE; |
6589 | |
6590 | /* If we've just compiled an assertion, pop the assert depth. */ |
6591 | |
6592 | if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NA) |
6593 | cb->assert_depth -= 1; |
6594 | |
6595 | /* At the end of compiling, code is still pointing to the start of the |
6596 | group, while tempcode has been updated to point past the end of the group. |
6597 | The parsed pattern pointer (pptr) is on the closing META_KET. |
6598 | |
6599 | If this is a conditional bracket, check that there are no more than |
6600 | two branches in the group, or just one if it's a DEFINE group. We do this |
6601 | in the real compile phase, not in the pre-pass, where the whole group may |
6602 | not be available. */ |
6603 | |
6604 | if (bravalue == OP_COND && lengthptr == NULL) |
6605 | { |
6606 | PCRE2_UCHAR *tc = code; |
6607 | int condcount = 0; |
6608 | |
6609 | do { |
6610 | condcount++; |
6611 | tc += GET(tc,1); |
6612 | } |
6613 | while (*tc != OP_KET); |
6614 | |
6615 | /* A DEFINE group is never obeyed inline (the "condition" is always |
6616 | false). It must have only one branch. Having checked this, change the |
6617 | opcode to OP_FALSE. */ |
6618 | |
6619 | if (code[LINK_SIZE+1] == OP_DEFINE) |
6620 | { |
6621 | if (condcount > 1) |
6622 | { |
6623 | cb->erroroffset = offset; |
6624 | *errorcodeptr = ERR54; |
6625 | return 0; |
6626 | } |
6627 | code[LINK_SIZE+1] = OP_FALSE; |
6628 | bravalue = OP_DEFINE; /* A flag to suppress char handling below */ |
6629 | } |
6630 | |
6631 | /* A "normal" conditional group. If there is just one branch, we must not |
6632 | make use of its firstcu or reqcu, because this is equivalent to an |
6633 | empty second branch. Also, it may match an empty string. If there are two |
6634 | branches, this item must match a character if the group must. */ |
6635 | |
6636 | else |
6637 | { |
6638 | if (condcount > 2) |
6639 | { |
6640 | cb->erroroffset = offset; |
6641 | *errorcodeptr = ERR27; |
6642 | return 0; |
6643 | } |
6644 | if (condcount == 1) subfirstcuflags = subreqcuflags = REQ_NONE; |
6645 | else if (group_return > 0) matched_char = TRUE; |
6646 | } |
6647 | } |
6648 | |
6649 | /* In the pre-compile phase, update the length by the length of the group, |
6650 | less the brackets at either end. Then reduce the compiled code to just a |
6651 | set of non-capturing brackets so that it doesn't use much memory if it is |
6652 | duplicated by a quantifier.*/ |
6653 | |
6654 | if (lengthptr != NULL) |
6655 | { |
6656 | if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE) |
6657 | { |
6658 | *errorcodeptr = ERR20; |
6659 | return 0; |
6660 | } |
6661 | *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE; |
6662 | code++; /* This already contains bravalue */ |
6663 | PUTINC(code, 0, 1 + LINK_SIZE); |
6664 | *code++ = OP_KET; |
6665 | PUTINC(code, 0, 1 + LINK_SIZE); |
6666 | break; /* No need to waste time with special character handling */ |
6667 | } |
6668 | |
6669 | /* Otherwise update the main code pointer to the end of the group. */ |
6670 | |
6671 | code = tempcode; |
6672 | |
6673 | /* For a DEFINE group, required and first character settings are not |
6674 | relevant. */ |
6675 | |
6676 | if (bravalue == OP_DEFINE) break; |
6677 | |
6678 | /* Handle updating of the required and first code units for other types of |
6679 | group. Update for normal brackets of all kinds, and conditions with two |
6680 | branches (see code above). If the bracket is followed by a quantifier with |
6681 | zero repeat, we have to back off. Hence the definition of zeroreqcu and |
6682 | zerofirstcu outside the main loop so that they can be accessed for the back |
6683 | off. */ |
6684 | |
6685 | zeroreqcu = reqcu; |
6686 | zeroreqcuflags = reqcuflags; |
6687 | zerofirstcu = firstcu; |
6688 | zerofirstcuflags = firstcuflags; |
6689 | groupsetfirstcu = FALSE; |
6690 | |
6691 | if (bravalue >= OP_ONCE) /* Not an assertion */ |
6692 | { |
6693 | /* If we have not yet set a firstcu in this branch, take it from the |
6694 | subpattern, remembering that it was set here so that a repeat of more |
6695 | than one can replicate it as reqcu if necessary. If the subpattern has |
6696 | no firstcu, set "none" for the whole branch. In both cases, a zero |
6697 | repeat forces firstcu to "none". */ |
6698 | |
6699 | if (firstcuflags == REQ_UNSET && subfirstcuflags != REQ_UNSET) |
6700 | { |
6701 | if (subfirstcuflags < REQ_NONE) |
6702 | { |
6703 | firstcu = subfirstcu; |
6704 | firstcuflags = subfirstcuflags; |
6705 | groupsetfirstcu = TRUE; |
6706 | } |
6707 | else firstcuflags = REQ_NONE; |
6708 | zerofirstcuflags = REQ_NONE; |
6709 | } |
6710 | |
6711 | /* If firstcu was previously set, convert the subpattern's firstcu |
6712 | into reqcu if there wasn't one, using the vary flag that was in |
6713 | existence beforehand. */ |
6714 | |
6715 | else if (subfirstcuflags < REQ_NONE && subreqcuflags >= REQ_NONE) |
6716 | { |
6717 | subreqcu = subfirstcu; |
6718 | subreqcuflags = subfirstcuflags | tempreqvary; |
6719 | } |
6720 | |
6721 | /* If the subpattern set a required code unit (or set a first code unit |
6722 | that isn't really the first code unit - see above), set it. */ |
6723 | |
6724 | if (subreqcuflags < REQ_NONE) |
6725 | { |
6726 | reqcu = subreqcu; |
6727 | reqcuflags = subreqcuflags; |
6728 | } |
6729 | } |
6730 | |
6731 | /* For a forward assertion, we take the reqcu, if set, provided that the |
6732 | group has also set a firstcu. This can be helpful if the pattern that |
6733 | follows the assertion doesn't set a different char. For example, it's |
6734 | useful for /(?=abcde).+/. We can't set firstcu for an assertion, however |
6735 | because it leads to incorrect effect for patterns such as /(?=a)a.+/ when |
6736 | the "real" "a" would then become a reqcu instead of a firstcu. This is |
6737 | overcome by a scan at the end if there's no firstcu, looking for an |
6738 | asserted first char. A similar effect for patterns like /(?=.*X)X$/ means |
6739 | we must only take the reqcu when the group also set a firstcu. Otherwise, |
6740 | in that example, 'X' ends up set for both. */ |
6741 | |
6742 | else if ((bravalue == OP_ASSERT || bravalue == OP_ASSERT_NA) && |
6743 | subreqcuflags < REQ_NONE && subfirstcuflags < REQ_NONE) |
6744 | { |
6745 | reqcu = subreqcu; |
6746 | reqcuflags = subreqcuflags; |
6747 | } |
6748 | |
6749 | break; /* End of nested group handling */ |
6750 | |
6751 | |
6752 | /* ===================================================================*/ |
6753 | /* Handle named backreferences and recursions. */ |
6754 | |
6755 | case META_BACKREF_BYNAME: |
6756 | case META_RECURSE_BYNAME: |
6757 | { |
6758 | int count, index; |
6759 | PCRE2_SPTR name; |
6760 | BOOL is_dupname = FALSE; |
6761 | named_group *ng = cb->named_groups; |
6762 | uint32_t length = *(++pptr); |
6763 | |
6764 | GETPLUSOFFSET(offset, pptr); |
6765 | name = cb->start_pattern + offset; |
6766 | |
6767 | /* In the first pass, the names generated in the pre-pass are available, |
6768 | but the main name table has not yet been created. Scan the list of names |
6769 | generated in the pre-pass in order to get a number and whether or not |
6770 | this name is duplicated. */ |
6771 | |
6772 | groupnumber = 0; |
6773 | for (unsigned int i = 0; i < cb->names_found; i++, ng++) |
6774 | { |
6775 | if (length == ng->length && |
6776 | PRIV(strncmp)(name, ng->name, length) == 0) |
6777 | { |
6778 | is_dupname = ng->isdup; |
6779 | groupnumber = ng->number; |
6780 | |
6781 | /* For a recursion, that's all that is needed. We can now go to |
6782 | the code that handles numerical recursion, applying it to the first |
6783 | group with the given name. */ |
6784 | |
6785 | if (meta == META_RECURSE_BYNAME) |
6786 | { |
6787 | meta_arg = groupnumber; |
6788 | goto HANDLE_NUMERICAL_RECURSION; |
6789 | } |
6790 | |
6791 | /* For a back reference, update the back reference map and the |
6792 | maximum back reference. */ |
6793 | |
6794 | cb->backref_map |= (groupnumber < 32)? (1u << groupnumber) : 1; |
6795 | if (groupnumber > cb->top_backref) |
6796 | cb->top_backref = groupnumber; |
6797 | } |
6798 | } |
6799 | |
6800 | /* If the name was not found we have a bad reference. */ |
6801 | |
6802 | if (groupnumber == 0) |
6803 | { |
6804 | *errorcodeptr = ERR15; |
6805 | cb->erroroffset = offset; |
6806 | return 0; |
6807 | } |
6808 | |
6809 | /* If a back reference name is not duplicated, we can handle it as |
6810 | a numerical reference. */ |
6811 | |
6812 | if (!is_dupname) |
6813 | { |
6814 | meta_arg = groupnumber; |
6815 | goto HANDLE_SINGLE_REFERENCE; |
6816 | } |
6817 | |
6818 | /* If a back reference name is duplicated, we generate a different |
6819 | opcode to a numerical back reference. In the second pass we must |
6820 | search for the index and count in the final name table. */ |
6821 | |
6822 | count = 0; /* Values for first pass (avoids compiler warning) */ |
6823 | index = 0; |
6824 | if (lengthptr == NULL && !find_dupname_details(name, length, &index, |
6825 | &count, errorcodeptr, cb)) return 0; |
6826 | |
6827 | if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE; |
6828 | *code++ = ((options & PCRE2_CASELESS) != 0)? OP_DNREFI : OP_DNREF; |
6829 | PUT2INC(code, 0, index); |
6830 | PUT2INC(code, 0, count); |
6831 | } |
6832 | break; |
6833 | |
6834 | |
6835 | /* ===================================================================*/ |
6836 | /* Handle a numerical callout. */ |
6837 | |
6838 | case META_CALLOUT_NUMBER: |
6839 | code[0] = OP_CALLOUT; |
6840 | PUT(code, 1, pptr[1]); /* Offset to next pattern item */ |
6841 | PUT(code, 1 + LINK_SIZE, pptr[2]); /* Length of next pattern item */ |
6842 | code[1 + 2*LINK_SIZE] = pptr[3]; |
6843 | pptr += 3; |
6844 | code += PRIV(OP_lengths)[OP_CALLOUT]; |
6845 | break; |
6846 | |
6847 | |
6848 | /* ===================================================================*/ |
6849 | /* Handle a callout with a string argument. In the pre-pass we just compute |
6850 | the length without generating anything. The length in pptr[3] includes both |
6851 | delimiters; in the actual compile only the first one is copied, but a |
6852 | terminating zero is added. Any doubled delimiters within the string make |
6853 | this an overestimate, but it is not worth bothering about. */ |
6854 | |
6855 | case META_CALLOUT_STRING: |
6856 | if (lengthptr != NULL) |
6857 | { |
6858 | *lengthptr += pptr[3] + (1 + 4*LINK_SIZE); |
6859 | pptr += 3; |
6860 | SKIPOFFSET(pptr); |
6861 | } |
6862 | |
6863 | /* In the real compile we can copy the string. The starting delimiter is |
6864 | included so that the client can discover it if they want. We also pass the |
6865 | start offset to help a script language give better error messages. */ |
6866 | |
6867 | else |
6868 | { |
6869 | PCRE2_SPTR pp; |
6870 | uint32_t delimiter; |
6871 | uint32_t length = pptr[3]; |
6872 | PCRE2_UCHAR *callout_string = code + (1 + 4*LINK_SIZE); |
6873 | |
6874 | code[0] = OP_CALLOUT_STR; |
6875 | PUT(code, 1, pptr[1]); /* Offset to next pattern item */ |
6876 | PUT(code, 1 + LINK_SIZE, pptr[2]); /* Length of next pattern item */ |
6877 | |
6878 | pptr += 3; |
6879 | GETPLUSOFFSET(offset, pptr); /* Offset to string in pattern */ |
6880 | pp = cb->start_pattern + offset; |
6881 | delimiter = *callout_string++ = *pp++; |
6882 | if (delimiter == CHAR_LEFT_CURLY_BRACKET) |
6883 | delimiter = CHAR_RIGHT_CURLY_BRACKET; |
6884 | PUT(code, 1 + 3*LINK_SIZE, (int)(offset + 1)); /* One after delimiter */ |
6885 | |
6886 | /* The syntax of the pattern was checked in the parsing scan. The length |
6887 | includes both delimiters, but we have passed the opening one just above, |
6888 | so we reduce length before testing it. The test is for > 1 because we do |
6889 | not want to copy the final delimiter. This also ensures that pp[1] is |
6890 | accessible. */ |
6891 | |
6892 | while (--length > 1) |
6893 | { |
6894 | if (*pp == delimiter && pp[1] == delimiter) |
6895 | { |
6896 | *callout_string++ = delimiter; |
6897 | pp += 2; |
6898 | length--; |
6899 | } |
6900 | else *callout_string++ = *pp++; |
6901 | } |
6902 | *callout_string++ = CHAR_NUL; |
6903 | |
6904 | /* Set the length of the entire item, the advance to its end. */ |
6905 | |
6906 | PUT(code, 1 + 2*LINK_SIZE, (int)(callout_string - code)); |
6907 | code = callout_string; |
6908 | } |
6909 | break; |
6910 | |
6911 | |
6912 | /* ===================================================================*/ |
6913 | /* Handle repetition. The different types are all sorted out in the parsing |
6914 | pass. */ |
6915 | |
6916 | case META_MINMAX_PLUS: |
6917 | case META_MINMAX_QUERY: |
6918 | case META_MINMAX: |
6919 | repeat_min = *(++pptr); |
6920 | repeat_max = *(++pptr); |
6921 | goto REPEAT; |
6922 | |
6923 | case META_ASTERISK: |
6924 | case META_ASTERISK_PLUS: |
6925 | case META_ASTERISK_QUERY: |
6926 | repeat_min = 0; |
6927 | repeat_max = REPEAT_UNLIMITED; |
6928 | goto REPEAT; |
6929 | |
6930 | case META_PLUS: |
6931 | case META_PLUS_PLUS: |
6932 | case META_PLUS_QUERY: |
6933 | repeat_min = 1; |
6934 | repeat_max = REPEAT_UNLIMITED; |
6935 | goto REPEAT; |
6936 | |
6937 | case META_QUERY: |
6938 | case META_QUERY_PLUS: |
6939 | case META_QUERY_QUERY: |
6940 | repeat_min = 0; |
6941 | repeat_max = 1; |
6942 | |
6943 | REPEAT: |
6944 | if (previous_matched_char && repeat_min > 0) matched_char = TRUE; |
6945 | |
6946 | /* Remember whether this is a variable length repeat, and default to |
6947 | single-char opcodes. */ |
6948 | |
6949 | reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY; |
6950 | op_type = 0; |
6951 | |
6952 | /* Adjust first and required code units for a zero repeat. */ |
6953 | |
6954 | if (repeat_min == 0) |
6955 | { |
6956 | firstcu = zerofirstcu; |
6957 | firstcuflags = zerofirstcuflags; |
6958 | reqcu = zeroreqcu; |
6959 | reqcuflags = zeroreqcuflags; |
6960 | } |
6961 | |
6962 | /* Note the greediness and possessiveness. */ |
6963 | |
6964 | switch (meta) |
6965 | { |
6966 | case META_MINMAX_PLUS: |
6967 | case META_ASTERISK_PLUS: |
6968 | case META_PLUS_PLUS: |
6969 | case META_QUERY_PLUS: |
6970 | repeat_type = 0; /* Force greedy */ |
6971 | possessive_quantifier = TRUE; |
6972 | break; |
6973 | |
6974 | case META_MINMAX_QUERY: |
6975 | case META_ASTERISK_QUERY: |
6976 | case META_PLUS_QUERY: |
6977 | case META_QUERY_QUERY: |
6978 | repeat_type = greedy_non_default; |
6979 | possessive_quantifier = FALSE; |
6980 | break; |
6981 | |
6982 | default: |
6983 | repeat_type = greedy_default; |
6984 | possessive_quantifier = FALSE; |
6985 | break; |
6986 | } |
6987 | |
6988 | /* Save start of previous item, in case we have to move it up in order to |
6989 | insert something before it, and remember what it was. */ |
6990 | |
6991 | tempcode = previous; |
6992 | op_previous = *previous; |
6993 | |
6994 | /* Now handle repetition for the different types of item. If the repeat |
6995 | minimum and the repeat maximum are both 1, we can ignore the quantifier for |
6996 | non-parenthesized items, as they have only one alternative. For anything in |
6997 | parentheses, we must not ignore if {1} is possessive. */ |
6998 | |
6999 | switch (op_previous) |
7000 | { |
7001 | /* If previous was a character or negated character match, abolish the |
7002 | item and generate a repeat item instead. If a char item has a minimum of |
7003 | more than one, ensure that it is set in reqcu - it might not be if a |
7004 | sequence such as x{3} is the first thing in a branch because the x will |
7005 | have gone into firstcu instead. */ |
7006 | |
7007 | case OP_CHAR: |
7008 | case OP_CHARI: |
7009 | case OP_NOT: |
7010 | case OP_NOTI: |
7011 | if (repeat_max == 1 && repeat_min == 1) goto END_REPEAT; |
7012 | op_type = chartypeoffset[op_previous - OP_CHAR]; |
7013 | |
7014 | /* Deal with UTF characters that take up more than one code unit. */ |
7015 | |
7016 | #ifdef MAYBE_UTF_MULTI |
7017 | if (utf && NOT_FIRSTCU(code[-1])) |
7018 | { |
7019 | PCRE2_UCHAR *lastchar = code - 1; |
7020 | BACKCHAR(lastchar); |
7021 | mclength = (uint32_t)(code - lastchar); /* Length of UTF character */ |
7022 | memcpy(mcbuffer, lastchar, CU2BYTES(mclength)); /* Save the char */ |
7023 | } |
7024 | else |
7025 | #endif /* MAYBE_UTF_MULTI */ |
7026 | |
7027 | /* Handle the case of a single code unit - either with no UTF support, or |
7028 | with UTF disabled, or for a single-code-unit UTF character. In the latter |
7029 | case, for a repeated positive match, get the caseless flag for the |
7030 | required code unit from the previous character, because a class like [Aa] |
7031 | sets a caseless A but by now the req_caseopt flag has been reset. */ |
7032 | |
7033 | { |
7034 | mcbuffer[0] = code[-1]; |
7035 | mclength = 1; |
7036 | if (op_previous <= OP_CHARI && repeat_min > 1) |
7037 | { |
7038 | reqcu = mcbuffer[0]; |
7039 | reqcuflags = cb->req_varyopt; |
7040 | if (op_previous == OP_CHARI) reqcuflags |= REQ_CASELESS; |
7041 | } |
7042 | } |
7043 | goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */ |
7044 | |
7045 | /* If previous was a character class or a back reference, we put the |
7046 | repeat stuff after it, but just skip the item if the repeat was {0,0}. */ |
7047 | |
7048 | #ifdef SUPPORT_WIDE_CHARS |
7049 | case OP_XCLASS: |
7050 | #endif |
7051 | case OP_CLASS: |
7052 | case OP_NCLASS: |
7053 | case OP_REF: |
7054 | case OP_REFI: |
7055 | case OP_DNREF: |
7056 | case OP_DNREFI: |
7057 | |
7058 | if (repeat_max == 0) |
7059 | { |
7060 | code = previous; |
7061 | goto END_REPEAT; |
7062 | } |
7063 | if (repeat_max == 1 && repeat_min == 1) goto END_REPEAT; |
7064 | |
7065 | if (repeat_min == 0 && repeat_max == REPEAT_UNLIMITED) |
7066 | *code++ = OP_CRSTAR + repeat_type; |
7067 | else if (repeat_min == 1 && repeat_max == REPEAT_UNLIMITED) |
7068 | *code++ = OP_CRPLUS + repeat_type; |
7069 | else if (repeat_min == 0 && repeat_max == 1) |
7070 | *code++ = OP_CRQUERY + repeat_type; |
7071 | else |
7072 | { |
7073 | *code++ = OP_CRRANGE + repeat_type; |
7074 | PUT2INC(code, 0, repeat_min); |
7075 | if (repeat_max == REPEAT_UNLIMITED) repeat_max = 0; /* 2-byte encoding for max */ |
7076 | PUT2INC(code, 0, repeat_max); |
7077 | } |
7078 | break; |
7079 | |
7080 | /* If previous is OP_FAIL, it was generated by an empty class [] |
7081 | (PCRE2_ALLOW_EMPTY_CLASS is set). The other ways in which OP_FAIL can be |
7082 | generated, that is by (*FAIL) or (?!), disallow a quantifier at parse |
7083 | time. We can just ignore this repeat. */ |
7084 | |
7085 | case OP_FAIL: |
7086 | goto END_REPEAT; |
7087 | |
7088 | /* Prior to 10.30, repeated recursions were wrapped in OP_ONCE brackets |
7089 | because pcre2_match() could not handle backtracking into recursively |
7090 | called groups. Now that this backtracking is available, we no longer need |
7091 | to do this. However, we still need to replicate recursions as we do for |
7092 | groups so as to have independent backtracking points. We can replicate |
7093 | for the minimum number of repeats directly. For optional repeats we now |
7094 | wrap the recursion in OP_BRA brackets and make use of the bracket |
7095 | repetition. */ |
7096 | |
7097 | case OP_RECURSE: |
7098 | if (repeat_max == 1 && repeat_min == 1 && !possessive_quantifier) |
7099 | goto END_REPEAT; |
7100 | |
7101 | /* Generate unwrapped repeats for a non-zero minimum, except when the |
7102 | minimum is 1 and the maximum unlimited, because that can be handled with |
7103 | OP_BRA terminated by OP_KETRMAX/MIN. When the maximum is equal to the |
7104 | minimum, we just need to generate the appropriate additional copies. |
7105 | Otherwise we need to generate one more, to simulate the situation when |
7106 | the minimum is zero. */ |
7107 | |
7108 | if (repeat_min > 0 && (repeat_min != 1 || repeat_max != REPEAT_UNLIMITED)) |
7109 | { |
7110 | int replicate = repeat_min; |
7111 | if (repeat_min == repeat_max) replicate--; |
7112 | |
7113 | /* In the pre-compile phase, we don't actually do the replication. We |
7114 | just adjust the length as if we had. Do some paranoid checks for |
7115 | potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit |
7116 | integer type when available, otherwise double. */ |
7117 | |
7118 | if (lengthptr != NULL) |
7119 | { |
7120 | PCRE2_SIZE delta = replicate*(1 + LINK_SIZE); |
7121 | if ((INT64_OR_DOUBLE)replicate* |
7122 | (INT64_OR_DOUBLE)(1 + LINK_SIZE) > |
7123 | (INT64_OR_DOUBLE)INT_MAX || |
7124 | OFLOW_MAX - *lengthptr < delta) |
7125 | { |
7126 | *errorcodeptr = ERR20; |
7127 | return 0; |
7128 | } |
7129 | *lengthptr += delta; |
7130 | } |
7131 | |
7132 | else for (int i = 0; i < replicate; i++) |
7133 | { |
7134 | memcpy(code, previous, CU2BYTES(1 + LINK_SIZE)); |
7135 | previous = code; |
7136 | code += 1 + LINK_SIZE; |
7137 | } |
7138 | |
7139 | /* If the number of repeats is fixed, we are done. Otherwise, adjust |
7140 | the counts and fall through. */ |
7141 | |
7142 | if (repeat_min == repeat_max) break; |
7143 | if (repeat_max != REPEAT_UNLIMITED) repeat_max -= repeat_min; |
7144 | repeat_min = 0; |
7145 | } |
7146 | |
7147 | /* Wrap the recursion call in OP_BRA brackets. */ |
7148 | |
7149 | (void)memmove(previous + 1 + LINK_SIZE, previous, CU2BYTES(1 + LINK_SIZE)); |
7150 | op_previous = *previous = OP_BRA; |
7151 | PUT(previous, 1, 2 + 2*LINK_SIZE); |
7152 | previous[2 + 2*LINK_SIZE] = OP_KET; |
7153 | PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE); |
7154 | code += 2 + 2 * LINK_SIZE; |
7155 | length_prevgroup = 3 + 3*LINK_SIZE; |
7156 | group_return = -1; /* Set "may match empty string" */ |
7157 | |
7158 | /* Now treat as a repeated OP_BRA. */ |
7159 | /* Fall through */ |
7160 | |
7161 | /* If previous was a bracket group, we may have to replicate it in |
7162 | certain cases. Note that at this point we can encounter only the "basic" |
7163 | bracket opcodes such as BRA and CBRA, as this is the place where they get |
7164 | converted into the more special varieties such as BRAPOS and SBRA. |
7165 | Originally, PCRE did not allow repetition of assertions, but now it does, |
7166 | for Perl compatibility. */ |
7167 | |
7168 | case OP_ASSERT: |
7169 | case OP_ASSERT_NOT: |
7170 | case OP_ASSERT_NA: |
7171 | case OP_ASSERTBACK: |
7172 | case OP_ASSERTBACK_NOT: |
7173 | case OP_ASSERTBACK_NA: |
7174 | case OP_ONCE: |
7175 | case OP_SCRIPT_RUN: |
7176 | case OP_BRA: |
7177 | case OP_CBRA: |
7178 | case OP_COND: |
7179 | { |
7180 | int len = (int)(code - previous); |
7181 | PCRE2_UCHAR *bralink = NULL; |
7182 | PCRE2_UCHAR *brazeroptr = NULL; |
7183 | |
7184 | if (repeat_max == 1 && repeat_min == 1 && !possessive_quantifier) |
7185 | goto END_REPEAT; |
7186 | |
7187 | /* Repeating a DEFINE group (or any group where the condition is always |
7188 | FALSE and there is only one branch) is pointless, but Perl allows the |
7189 | syntax, so we just ignore the repeat. */ |
7190 | |
7191 | if (op_previous == OP_COND && previous[LINK_SIZE+1] == OP_FALSE && |
7192 | previous[GET(previous, 1)] != OP_ALT) |
7193 | goto END_REPEAT; |
7194 | |
7195 | /* Perl allows all assertions to be quantified, and when they contain |
7196 | capturing parentheses and/or are optional there are potential uses for |
7197 | this feature. PCRE2 used to force the maximum quantifier to 1 on the |
7198 | invalid grounds that further repetition was never useful. This was |
7199 | always a bit pointless, since an assertion could be wrapped with a |
7200 | repeated group to achieve the effect. General repetition is now |
7201 | permitted, but if the maximum is unlimited it is set to one more than |
7202 | the minimum. */ |
7203 | |
7204 | if (op_previous < OP_ONCE) /* Assertion */ |
7205 | { |
7206 | if (repeat_max == REPEAT_UNLIMITED) repeat_max = repeat_min + 1; |
7207 | } |
7208 | |
7209 | /* The case of a zero minimum is special because of the need to stick |
7210 | OP_BRAZERO in front of it, and because the group appears once in the |
7211 | data, whereas in other cases it appears the minimum number of times. For |
7212 | this reason, it is simplest to treat this case separately, as otherwise |
7213 | the code gets far too messy. There are several special subcases when the |
7214 | minimum is zero. */ |
7215 | |
7216 | if (repeat_min == 0) |
7217 | { |
7218 | /* If the maximum is also zero, we used to just omit the group from |
7219 | the output altogether, like this: |
7220 | |
7221 | ** if (repeat_max == 0) |
7222 | ** { |
7223 | ** code = previous; |
7224 | ** goto END_REPEAT; |
7225 | ** } |
7226 | |
7227 | However, that fails when a group or a subgroup within it is |
7228 | referenced as a subroutine from elsewhere in the pattern, so now we |
7229 | stick in OP_SKIPZERO in front of it so that it is skipped on |
7230 | execution. As we don't have a list of which groups are referenced, we |
7231 | cannot do this selectively. |
7232 | |
7233 | If the maximum is 1 or unlimited, we just have to stick in the |
7234 | BRAZERO and do no more at this point. */ |
7235 | |
7236 | if (repeat_max <= 1 || repeat_max == REPEAT_UNLIMITED) |
7237 | { |
7238 | (void)memmove(previous + 1, previous, CU2BYTES(len)); |
7239 | code++; |
7240 | if (repeat_max == 0) |
7241 | { |
7242 | *previous++ = OP_SKIPZERO; |
7243 | goto END_REPEAT; |
7244 | } |
7245 | brazeroptr = previous; /* Save for possessive optimizing */ |
7246 | *previous++ = OP_BRAZERO + repeat_type; |
7247 | } |
7248 | |
7249 | /* If the maximum is greater than 1 and limited, we have to replicate |
7250 | in a nested fashion, sticking OP_BRAZERO before each set of brackets. |
7251 | The first one has to be handled carefully because it's the original |
7252 | copy, which has to be moved up. The remainder can be handled by code |
7253 | that is common with the non-zero minimum case below. We have to |
7254 | adjust the value or repeat_max, since one less copy is required. */ |
7255 | |
7256 | else |
7257 | { |
7258 | int linkoffset; |
7259 | (void)memmove(previous + 2 + LINK_SIZE, previous, CU2BYTES(len)); |
7260 | code += 2 + LINK_SIZE; |
7261 | *previous++ = OP_BRAZERO + repeat_type; |
7262 | *previous++ = OP_BRA; |
7263 | |
7264 | /* We chain together the bracket link offset fields that have to be |
7265 | filled in later when the ends of the brackets are reached. */ |
7266 | |
7267 | linkoffset = (bralink == NULL)? 0 : (int)(previous - bralink); |
7268 | bralink = previous; |
7269 | PUTINC(previous, 0, linkoffset); |
7270 | } |
7271 | |
7272 | if (repeat_max != REPEAT_UNLIMITED) repeat_max--; |
7273 | } |
7274 | |
7275 | /* If the minimum is greater than zero, replicate the group as many |
7276 | times as necessary, and adjust the maximum to the number of subsequent |
7277 | copies that we need. */ |
7278 | |
7279 | else |
7280 | { |
7281 | if (repeat_min > 1) |
7282 | { |
7283 | /* In the pre-compile phase, we don't actually do the replication. |
7284 | We just adjust the length as if we had. Do some paranoid checks for |
7285 | potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit |
7286 | integer type when available, otherwise double. */ |
7287 | |
7288 | if (lengthptr != NULL) |
7289 | { |
7290 | PCRE2_SIZE delta = (repeat_min - 1)*length_prevgroup; |
7291 | if ((INT64_OR_DOUBLE)(repeat_min - 1)* |
7292 | (INT64_OR_DOUBLE)length_prevgroup > |
7293 | (INT64_OR_DOUBLE)INT_MAX || |
7294 | OFLOW_MAX - *lengthptr < delta) |
7295 | { |
7296 | *errorcodeptr = ERR20; |
7297 | return 0; |
7298 | } |
7299 | *lengthptr += delta; |
7300 | } |
7301 | |
7302 | /* This is compiling for real. If there is a set first code unit |
7303 | for the group, and we have not yet set a "required code unit", set |
7304 | it. */ |
7305 | |
7306 | else |
7307 | { |
7308 | if (groupsetfirstcu && reqcuflags >= REQ_NONE) |
7309 | { |
7310 | reqcu = firstcu; |
7311 | reqcuflags = firstcuflags; |
7312 | } |
7313 | for (uint32_t i = 1; i < repeat_min; i++) |
7314 | { |
7315 | memcpy(code, previous, CU2BYTES(len)); |
7316 | code += len; |
7317 | } |
7318 | } |
7319 | } |
7320 | |
7321 | if (repeat_max != REPEAT_UNLIMITED) repeat_max -= repeat_min; |
7322 | } |
7323 | |
7324 | /* This code is common to both the zero and non-zero minimum cases. If |
7325 | the maximum is limited, it replicates the group in a nested fashion, |
7326 | remembering the bracket starts on a stack. In the case of a zero |
7327 | minimum, the first one was set up above. In all cases the repeat_max |
7328 | now specifies the number of additional copies needed. Again, we must |
7329 | remember to replicate entries on the forward reference list. */ |
7330 | |
7331 | if (repeat_max != REPEAT_UNLIMITED) |
7332 | { |
7333 | /* In the pre-compile phase, we don't actually do the replication. We |
7334 | just adjust the length as if we had. For each repetition we must add |
7335 | 1 to the length for BRAZERO and for all but the last repetition we |
7336 | must add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some |
7337 | paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type |
7338 | is a 64-bit integer type when available, otherwise double. */ |
7339 | |
7340 | if (lengthptr != NULL && repeat_max > 0) |
7341 | { |
7342 | PCRE2_SIZE delta = repeat_max*(length_prevgroup + 1 + 2 + 2*LINK_SIZE) - |
7343 | 2 - 2*LINK_SIZE; /* Last one doesn't nest */ |
7344 | if ((INT64_OR_DOUBLE)repeat_max * |
7345 | (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE) |
7346 | > (INT64_OR_DOUBLE)INT_MAX || |
7347 | OFLOW_MAX - *lengthptr < delta) |
7348 | { |
7349 | *errorcodeptr = ERR20; |
7350 | return 0; |
7351 | } |
7352 | *lengthptr += delta; |
7353 | } |
7354 | |
7355 | /* This is compiling for real */ |
7356 | |
7357 | else for (uint32_t i = repeat_max; i >= 1; i--) |
7358 | { |
7359 | *code++ = OP_BRAZERO + repeat_type; |
7360 | |
7361 | /* All but the final copy start a new nesting, maintaining the |
7362 | chain of brackets outstanding. */ |
7363 | |
7364 | if (i != 1) |
7365 | { |
7366 | int linkoffset; |
7367 | *code++ = OP_BRA; |
7368 | linkoffset = (bralink == NULL)? 0 : (int)(code - bralink); |
7369 | bralink = code; |
7370 | PUTINC(code, 0, linkoffset); |
7371 | } |
7372 | |
7373 | memcpy(code, previous, CU2BYTES(len)); |
7374 | code += len; |
7375 | } |
7376 | |
7377 | /* Now chain through the pending brackets, and fill in their length |
7378 | fields (which are holding the chain links pro tem). */ |
7379 | |
7380 | while (bralink != NULL) |
7381 | { |
7382 | int oldlinkoffset; |
7383 | int linkoffset = (int)(code - bralink + 1); |
7384 | PCRE2_UCHAR *bra = code - linkoffset; |
7385 | oldlinkoffset = GET(bra, 1); |
7386 | bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset; |
7387 | *code++ = OP_KET; |
7388 | PUTINC(code, 0, linkoffset); |
7389 | PUT(bra, 1, linkoffset); |
7390 | } |
7391 | } |
7392 | |
7393 | /* If the maximum is unlimited, set a repeater in the final copy. For |
7394 | SCRIPT_RUN and ONCE brackets, that's all we need to do. However, |
7395 | possessively repeated ONCE brackets can be converted into non-capturing |
7396 | brackets, as the behaviour of (?:xx)++ is the same as (?>xx)++ and this |
7397 | saves having to deal with possessive ONCEs specially. |
7398 | |
7399 | Otherwise, when we are doing the actual compile phase, check to see |
7400 | whether this group is one that could match an empty string. If so, |
7401 | convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so |
7402 | that runtime checking can be done. [This check is also applied to ONCE |
7403 | and SCRIPT_RUN groups at runtime, but in a different way.] |
7404 | |
7405 | Then, if the quantifier was possessive and the bracket is not a |
7406 | conditional, we convert the BRA code to the POS form, and the KET code |
7407 | to KETRPOS. (It turns out to be convenient at runtime to detect this |
7408 | kind of subpattern at both the start and at the end.) The use of |
7409 | special opcodes makes it possible to reduce greatly the stack usage in |
7410 | pcre2_match(). If the group is preceded by OP_BRAZERO, convert this to |
7411 | OP_BRAPOSZERO. |
7412 | |
7413 | Then, if the minimum number of matches is 1 or 0, cancel the possessive |
7414 | flag so that the default action below, of wrapping everything inside |
7415 | atomic brackets, does not happen. When the minimum is greater than 1, |
7416 | there will be earlier copies of the group, and so we still have to wrap |
7417 | the whole thing. */ |
7418 | |
7419 | else |
7420 | { |
7421 | PCRE2_UCHAR *ketcode = code - 1 - LINK_SIZE; |
7422 | PCRE2_UCHAR *bracode = ketcode - GET(ketcode, 1); |
7423 | |
7424 | /* Convert possessive ONCE brackets to non-capturing */ |
7425 | |
7426 | if (*bracode == OP_ONCE && possessive_quantifier) *bracode = OP_BRA; |
7427 | |
7428 | /* For non-possessive ONCE and for SCRIPT_RUN brackets, all we need |
7429 | to do is to set the KET. */ |
7430 | |
7431 | if (*bracode == OP_ONCE || *bracode == OP_SCRIPT_RUN) |
7432 | *ketcode = OP_KETRMAX + repeat_type; |
7433 | |
7434 | /* Handle non-SCRIPT_RUN and non-ONCE brackets and possessive ONCEs |
7435 | (which have been converted to non-capturing above). */ |
7436 | |
7437 | else |
7438 | { |
7439 | /* In the compile phase, adjust the opcode if the group can match |
7440 | an empty string. For a conditional group with only one branch, the |
7441 | value of group_return will not show "could be empty", so we must |
7442 | check that separately. */ |
7443 | |
7444 | if (lengthptr == NULL) |
7445 | { |
7446 | if (group_return < 0) *bracode += OP_SBRA - OP_BRA; |
7447 | if (*bracode == OP_COND && bracode[GET(bracode,1)] != OP_ALT) |
7448 | *bracode = OP_SCOND; |
7449 | } |
7450 | |
7451 | /* Handle possessive quantifiers. */ |
7452 | |
7453 | if (possessive_quantifier) |
7454 | { |
7455 | /* For COND brackets, we wrap the whole thing in a possessively |
7456 | repeated non-capturing bracket, because we have not invented POS |
7457 | versions of the COND opcodes. */ |
7458 | |
7459 | if (*bracode == OP_COND || *bracode == OP_SCOND) |
7460 | { |
7461 | int nlen = (int)(code - bracode); |
7462 | (void)memmove(bracode + 1 + LINK_SIZE, bracode, CU2BYTES(nlen)); |
7463 | code += 1 + LINK_SIZE; |
7464 | nlen += 1 + LINK_SIZE; |
7465 | *bracode = (*bracode == OP_COND)? OP_BRAPOS : OP_SBRAPOS; |
7466 | *code++ = OP_KETRPOS; |
7467 | PUTINC(code, 0, nlen); |
7468 | PUT(bracode, 1, nlen); |
7469 | } |
7470 | |
7471 | /* For non-COND brackets, we modify the BRA code and use KETRPOS. */ |
7472 | |
7473 | else |
7474 | { |
7475 | *bracode += 1; /* Switch to xxxPOS opcodes */ |
7476 | *ketcode = OP_KETRPOS; |
7477 | } |
7478 | |
7479 | /* If the minimum is zero, mark it as possessive, then unset the |
7480 | possessive flag when the minimum is 0 or 1. */ |
7481 | |
7482 | if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO; |
7483 | if (repeat_min < 2) possessive_quantifier = FALSE; |
7484 | } |
7485 | |
7486 | /* Non-possessive quantifier */ |
7487 | |
7488 | else *ketcode = OP_KETRMAX + repeat_type; |
7489 | } |
7490 | } |
7491 | } |
7492 | break; |
7493 | |
7494 | /* If previous was a character type match (\d or similar), abolish it and |
7495 | create a suitable repeat item. The code is shared with single-character |
7496 | repeats by setting op_type to add a suitable offset into repeat_type. |
7497 | Note the the Unicode property types will be present only when |
7498 | SUPPORT_UNICODE is defined, but we don't wrap the little bits of code |
7499 | here because it just makes it horribly messy. */ |
7500 | |
7501 | default: |
7502 | if (op_previous >= OP_EODN) /* Not a character type - internal error */ |
7503 | { |
7504 | *errorcodeptr = ERR10; |
7505 | return 0; |
7506 | } |
7507 | else |
7508 | { |
7509 | int prop_type, prop_value; |
7510 | PCRE2_UCHAR *oldcode; |
7511 | |
7512 | if (repeat_max == 1 && repeat_min == 1) goto END_REPEAT; |
7513 | |
7514 | op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */ |
7515 | mclength = 0; /* Not a character */ |
7516 | |
7517 | if (op_previous == OP_PROP || op_previous == OP_NOTPROP) |
7518 | { |
7519 | prop_type = previous[1]; |
7520 | prop_value = previous[2]; |
7521 | } |
7522 | else |
7523 | { |
7524 | /* Come here from just above with a character in mcbuffer/mclength. */ |
7525 | OUTPUT_SINGLE_REPEAT: |
7526 | prop_type = prop_value = -1; |
7527 | } |
7528 | |
7529 | /* At this point, if prop_type == prop_value == -1 we either have a |
7530 | character in mcbuffer when mclength is greater than zero, or we have |
7531 | mclength zero, in which case there is a non-property character type in |
7532 | op_previous. If prop_type/value are not negative, we have a property |
7533 | character type in op_previous. */ |
7534 | |
7535 | oldcode = code; /* Save where we were */ |
7536 | code = previous; /* Usually overwrite previous item */ |
7537 | |
7538 | /* If the maximum is zero then the minimum must also be zero; Perl allows |
7539 | this case, so we do too - by simply omitting the item altogether. */ |
7540 | |
7541 | if (repeat_max == 0) goto END_REPEAT; |
7542 | |
7543 | /* Combine the op_type with the repeat_type */ |
7544 | |
7545 | repeat_type += op_type; |
7546 | |
7547 | /* A minimum of zero is handled either as the special case * or ?, or as |
7548 | an UPTO, with the maximum given. */ |
7549 | |
7550 | if (repeat_min == 0) |
7551 | { |
7552 | if (repeat_max == REPEAT_UNLIMITED) *code++ = OP_STAR + repeat_type; |
7553 | else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type; |
7554 | else |
7555 | { |
7556 | *code++ = OP_UPTO + repeat_type; |
7557 | PUT2INC(code, 0, repeat_max); |
7558 | } |
7559 | } |
7560 | |
7561 | /* A repeat minimum of 1 is optimized into some special cases. If the |
7562 | maximum is unlimited, we use OP_PLUS. Otherwise, the original item is |
7563 | left in place and, if the maximum is greater than 1, we use OP_UPTO with |
7564 | one less than the maximum. */ |
7565 | |
7566 | else if (repeat_min == 1) |
7567 | { |
7568 | if (repeat_max == REPEAT_UNLIMITED) |
7569 | *code++ = OP_PLUS + repeat_type; |
7570 | else |
7571 | { |
7572 | code = oldcode; /* Leave previous item in place */ |
7573 | if (repeat_max == 1) goto END_REPEAT; |
7574 | *code++ = OP_UPTO + repeat_type; |
7575 | PUT2INC(code, 0, repeat_max - 1); |
7576 | } |
7577 | } |
7578 | |
7579 | /* The case {n,n} is just an EXACT, while the general case {n,m} is |
7580 | handled as an EXACT followed by an UPTO or STAR or QUERY. */ |
7581 | |
7582 | else |
7583 | { |
7584 | *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */ |
7585 | PUT2INC(code, 0, repeat_min); |
7586 | |
7587 | /* Unless repeat_max equals repeat_min, fill in the data for EXACT, |
7588 | and then generate the second opcode. For a repeated Unicode property |
7589 | match, there are two extra values that define the required property, |
7590 | and mclength is set zero to indicate this. */ |
7591 | |
7592 | if (repeat_max != repeat_min) |
7593 | { |
7594 | if (mclength > 0) |
7595 | { |
7596 | memcpy(code, mcbuffer, CU2BYTES(mclength)); |
7597 | code += mclength; |
7598 | } |
7599 | else |
7600 | { |
7601 | *code++ = op_previous; |
7602 | if (prop_type >= 0) |
7603 | { |
7604 | *code++ = prop_type; |
7605 | *code++ = prop_value; |
7606 | } |
7607 | } |
7608 | |
7609 | /* Now set up the following opcode */ |
7610 | |
7611 | if (repeat_max == REPEAT_UNLIMITED) |
7612 | *code++ = OP_STAR + repeat_type; |
7613 | else |
7614 | { |
7615 | repeat_max -= repeat_min; |
7616 | if (repeat_max == 1) |
7617 | { |
7618 | *code++ = OP_QUERY + repeat_type; |
7619 | } |
7620 | else |
7621 | { |
7622 | *code++ = OP_UPTO + repeat_type; |
7623 | PUT2INC(code, 0, repeat_max); |
7624 | } |
7625 | } |
7626 | } |
7627 | } |
7628 | |
7629 | /* Fill in the character or character type for the final opcode. */ |
7630 | |
7631 | if (mclength > 0) |
7632 | { |
7633 | memcpy(code, mcbuffer, CU2BYTES(mclength)); |
7634 | code += mclength; |
7635 | } |
7636 | else |
7637 | { |
7638 | *code++ = op_previous; |
7639 | if (prop_type >= 0) |
7640 | { |
7641 | *code++ = prop_type; |
7642 | *code++ = prop_value; |
7643 | } |
7644 | } |
7645 | } |
7646 | break; |
7647 | } /* End of switch on different op_previous values */ |
7648 | |
7649 | |
7650 | /* If the character following a repeat is '+', possessive_quantifier is |
7651 | TRUE. For some opcodes, there are special alternative opcodes for this |
7652 | case. For anything else, we wrap the entire repeated item inside OP_ONCE |
7653 | brackets. Logically, the '+' notation is just syntactic sugar, taken from |
7654 | Sun's Java package, but the special opcodes can optimize it. |
7655 | |
7656 | Some (but not all) possessively repeated subpatterns have already been |
7657 | completely handled in the code just above. For them, possessive_quantifier |
7658 | is always FALSE at this stage. Note that the repeated item starts at |
7659 | tempcode, not at previous, which might be the first part of a string whose |
7660 | (former) last char we repeated. */ |
7661 | |
7662 | if (possessive_quantifier) |
7663 | { |
7664 | int len; |
7665 | |
7666 | /* Possessifying an EXACT quantifier has no effect, so we can ignore it. |
7667 | However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6}, |
7668 | {5,}, or {5,10}). We skip over an EXACT item; if the length of what |
7669 | remains is greater than zero, there's a further opcode that can be |
7670 | handled. If not, do nothing, leaving the EXACT alone. */ |
7671 | |
7672 | switch(*tempcode) |
7673 | { |
7674 | case OP_TYPEEXACT: |
7675 | tempcode += PRIV(OP_lengths)[*tempcode] + |
7676 | ((tempcode[1 + IMM2_SIZE] == OP_PROP |
7677 | || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0); |
7678 | break; |
7679 | |
7680 | /* CHAR opcodes are used for exacts whose count is 1. */ |
7681 | |
7682 | case OP_CHAR: |
7683 | case OP_CHARI: |
7684 | case OP_NOT: |
7685 | case OP_NOTI: |
7686 | case OP_EXACT: |
7687 | case OP_EXACTI: |
7688 | case OP_NOTEXACT: |
7689 | case OP_NOTEXACTI: |
7690 | tempcode += PRIV(OP_lengths)[*tempcode]; |
7691 | #ifdef SUPPORT_UNICODE |
7692 | if (utf && HAS_EXTRALEN(tempcode[-1])) |
7693 | tempcode += GET_EXTRALEN(tempcode[-1]); |
7694 | #endif |
7695 | break; |
7696 | |
7697 | /* For the class opcodes, the repeat operator appears at the end; |
7698 | adjust tempcode to point to it. */ |
7699 | |
7700 | case OP_CLASS: |
7701 | case OP_NCLASS: |
7702 | tempcode += 1 + 32/sizeof(PCRE2_UCHAR); |
7703 | break; |
7704 | |
7705 | #ifdef SUPPORT_WIDE_CHARS |
7706 | case OP_XCLASS: |
7707 | tempcode += GET(tempcode, 1); |
7708 | break; |
7709 | #endif |
7710 | } |
7711 | |
7712 | /* If tempcode is equal to code (which points to the end of the repeated |
7713 | item), it means we have skipped an EXACT item but there is no following |
7714 | QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In |
7715 | all other cases, tempcode will be pointing to the repeat opcode, and will |
7716 | be less than code, so the value of len will be greater than 0. */ |
7717 | |
7718 | len = (int)(code - tempcode); |
7719 | if (len > 0) |
7720 | { |
7721 | unsigned int repcode = *tempcode; |
7722 | |
7723 | /* There is a table for possessifying opcodes, all of which are less |
7724 | than OP_CALLOUT. A zero entry means there is no possessified version. |
7725 | */ |
7726 | |
7727 | if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0) |
7728 | *tempcode = opcode_possessify[repcode]; |
7729 | |
7730 | /* For opcode without a special possessified version, wrap the item in |
7731 | ONCE brackets. */ |
7732 | |
7733 | else |
7734 | { |
7735 | (void)memmove(tempcode + 1 + LINK_SIZE, tempcode, CU2BYTES(len)); |
7736 | code += 1 + LINK_SIZE; |
7737 | len += 1 + LINK_SIZE; |
7738 | tempcode[0] = OP_ONCE; |
7739 | *code++ = OP_KET; |
7740 | PUTINC(code, 0, len); |
7741 | PUT(tempcode, 1, len); |
7742 | } |
7743 | } |
7744 | } |
7745 | |
7746 | /* We set the "follows varying string" flag for subsequently encountered |
7747 | reqcus if it isn't already set and we have just passed a varying length |
7748 | item. */ |
7749 | |
7750 | END_REPEAT: |
7751 | cb->req_varyopt |= reqvary; |
7752 | break; |
7753 | |
7754 | |
7755 | /* ===================================================================*/ |
7756 | /* Handle a 32-bit data character with a value greater than META_END. */ |
7757 | |
7758 | case META_BIGVALUE: |
7759 | pptr++; |
7760 | goto NORMAL_CHAR; |
7761 | |
7762 | |
7763 | /* ===============================================================*/ |
7764 | /* Handle a back reference by number, which is the meta argument. The |
7765 | pattern offsets for back references to group numbers less than 10 are held |
7766 | in a special vector, to avoid using more than two parsed pattern elements |
7767 | in 64-bit environments. We only need the offset to the first occurrence, |
7768 | because if that doesn't fail, subsequent ones will also be OK. */ |
7769 | |
7770 | case META_BACKREF: |
7771 | if (meta_arg < 10) offset = cb->small_ref_offset[meta_arg]; |
7772 | else GETPLUSOFFSET(offset, pptr); |
7773 | |
7774 | if (meta_arg > cb->bracount) |
7775 | { |
7776 | cb->erroroffset = offset; |
7777 | *errorcodeptr = ERR15; /* Non-existent subpattern */ |
7778 | return 0; |
7779 | } |
7780 | |
7781 | /* Come here from named backref handling when the reference is to a |
7782 | single group (that is, not to a duplicated name). The back reference |
7783 | data will have already been updated. We must disable firstcu if not |
7784 | set, to cope with cases like (?=(\w+))\1: which would otherwise set ':' |
7785 | later. */ |
7786 | |
7787 | HANDLE_SINGLE_REFERENCE: |
7788 | if (firstcuflags == REQ_UNSET) zerofirstcuflags = firstcuflags = REQ_NONE; |
7789 | *code++ = ((options & PCRE2_CASELESS) != 0)? OP_REFI : OP_REF; |
7790 | PUT2INC(code, 0, meta_arg); |
7791 | |
7792 | /* Update the map of back references, and keep the highest one. We |
7793 | could do this in parse_regex() for numerical back references, but not |
7794 | for named back references, because we don't know the numbers to which |
7795 | named back references refer. So we do it all in this function. */ |
7796 | |
7797 | cb->backref_map |= (meta_arg < 32)? (1u << meta_arg) : 1; |
7798 | if (meta_arg > cb->top_backref) cb->top_backref = meta_arg; |
7799 | break; |
7800 | |
7801 | |
7802 | /* ===============================================================*/ |
7803 | /* Handle recursion by inserting the number of the called group (which is |
7804 | the meta argument) after OP_RECURSE. At the end of compiling the pattern is |
7805 | scanned and these numbers are replaced by offsets within the pattern. It is |
7806 | done like this to avoid problems with forward references and adjusting |
7807 | offsets when groups are duplicated and moved (as discovered in previous |
7808 | implementations). Note that a recursion does not have a set first |
7809 | character. */ |
7810 | |
7811 | case META_RECURSE: |
7812 | GETPLUSOFFSET(offset, pptr); |
7813 | if (meta_arg > cb->bracount) |
7814 | { |
7815 | cb->erroroffset = offset; |
7816 | *errorcodeptr = ERR15; /* Non-existent subpattern */ |
7817 | return 0; |
7818 | } |
7819 | HANDLE_NUMERICAL_RECURSION: |
7820 | *code = OP_RECURSE; |
7821 | PUT(code, 1, meta_arg); |
7822 | code += 1 + LINK_SIZE; |
7823 | groupsetfirstcu = FALSE; |
7824 | cb->had_recurse = TRUE; |
7825 | if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE; |
7826 | zerofirstcu = firstcu; |
7827 | zerofirstcuflags = firstcuflags; |
7828 | break; |
7829 | |
7830 | |
7831 | /* ===============================================================*/ |
7832 | /* Handle capturing parentheses; the number is the meta argument. */ |
7833 | |
7834 | case META_CAPTURE: |
7835 | bravalue = OP_CBRA; |
7836 | skipunits = IMM2_SIZE; |
7837 | PUT2(code, 1+LINK_SIZE, meta_arg); |
7838 | cb->lastcapture = meta_arg; |
7839 | goto GROUP_PROCESS_NOTE_EMPTY; |
7840 | |
7841 | |
7842 | /* ===============================================================*/ |
7843 | /* Handle escape sequence items. For ones like \d, the ESC_values are |
7844 | arranged to be the same as the corresponding OP_values in the default case |
7845 | when PCRE2_UCP is not set (which is the only case in which they will appear |
7846 | here). |
7847 | |
7848 | Note: \Q and \E are never seen here, as they were dealt with in |
7849 | parse_pattern(). Neither are numerical back references or recursions, which |
7850 | were turned into META_BACKREF or META_RECURSE items, respectively. \k and |
7851 | \g, when followed by names, are turned into META_BACKREF_BYNAME or |
7852 | META_RECURSE_BYNAME. */ |
7853 | |
7854 | case META_ESCAPE: |
7855 | |
7856 | /* We can test for escape sequences that consume a character because their |
7857 | values lie between ESC_b and ESC_Z; this may have to change if any new ones |
7858 | are ever created. For these sequences, we disable the setting of a first |
7859 | character if it hasn't already been set. */ |
7860 | |
7861 | if (meta_arg > ESC_b && meta_arg < ESC_Z) |
7862 | { |
7863 | matched_char = TRUE; |
7864 | if (firstcuflags == REQ_UNSET) firstcuflags = REQ_NONE; |
7865 | } |
7866 | |
7867 | /* Set values to reset to if this is followed by a zero repeat. */ |
7868 | |
7869 | zerofirstcu = firstcu; |
7870 | zerofirstcuflags = firstcuflags; |
7871 | zeroreqcu = reqcu; |
7872 | zeroreqcuflags = reqcuflags; |
7873 | |
7874 | /* If Unicode is not supported, \P and \p are not allowed and are |
7875 | faulted at parse time, so will never appear here. */ |
7876 | |
7877 | #ifdef SUPPORT_UNICODE |
7878 | if (meta_arg == ESC_P || meta_arg == ESC_p) |
7879 | { |
7880 | uint32_t ptype = *(++pptr) >> 16; |
7881 | uint32_t pdata = *pptr & 0xffff; |
7882 | |
7883 | /* The special case of \p{Any} is compiled to OP_ALLANY so as to benefit |
7884 | from the auto-anchoring code. */ |
7885 | |
7886 | if (meta_arg == ESC_p && ptype == PT_ANY) |
7887 | { |
7888 | *code++ = OP_ALLANY; |
7889 | } |
7890 | else |
7891 | { |
7892 | *code++ = (meta_arg == ESC_p)? OP_PROP : OP_NOTPROP; |
7893 | *code++ = ptype; |
7894 | *code++ = pdata; |
7895 | } |
7896 | break; /* End META_ESCAPE */ |
7897 | } |
7898 | #endif |
7899 | |
7900 | /* \K is forbidden in lookarounds since 10.38 because that's what Perl has |
7901 | done. However, there's an option, in case anyone was relying on it. */ |
7902 | |
7903 | if (cb->assert_depth > 0 && meta_arg == ESC_K && |
7904 | (cb->cx->extra_options & PCRE2_EXTRA_ALLOW_LOOKAROUND_BSK) == 0) |
7905 | { |
7906 | *errorcodeptr = ERR99; |
7907 | return 0; |
7908 | } |
7909 | |
7910 | /* For the rest (including \X when Unicode is supported - if not it's |
7911 | faulted at parse time), the OP value is the escape value when PCRE2_UCP is |
7912 | not set; if it is set, these escapes do not show up here because they are |
7913 | converted into Unicode property tests in parse_regex(). Note that \b and \B |
7914 | do a one-character lookbehind, and \A also behaves as if it does. */ |
7915 | |
7916 | if (meta_arg == ESC_C) cb->external_flags |= PCRE2_HASBKC; /* Record */ |
7917 | if ((meta_arg == ESC_b || meta_arg == ESC_B || meta_arg == ESC_A) && |
7918 | cb->max_lookbehind == 0) |
7919 | cb->max_lookbehind = 1; |
7920 | |
7921 | /* In non-UTF mode, and for both 32-bit modes, we turn \C into OP_ALLANY |
7922 | instead of OP_ANYBYTE so that it works in DFA mode and in lookbehinds. */ |
7923 | |
7924 | #if PCRE2_CODE_UNIT_WIDTH == 32 |
7925 | *code++ = (meta_arg == ESC_C)? OP_ALLANY : meta_arg; |
7926 | #else |
7927 | *code++ = (!utf && meta_arg == ESC_C)? OP_ALLANY : meta_arg; |
7928 | #endif |
7929 | break; /* End META_ESCAPE */ |
7930 | |
7931 | |
7932 | /* ===================================================================*/ |
7933 | /* Handle an unrecognized meta value. A parsed pattern value less than |
7934 | META_END is a literal. Otherwise we have a problem. */ |
7935 | |
7936 | default: |
7937 | if (meta >= META_END) |
7938 | { |
7939 | #ifdef DEBUG_SHOW_PARSED |
7940 | fprintf(stderr, "** Unrecognized parsed pattern item 0x%.8x\n" , *pptr); |
7941 | #endif |
7942 | *errorcodeptr = ERR89; /* Internal error - unrecognized. */ |
7943 | return 0; |
7944 | } |
7945 | |
7946 | /* Handle a literal character. We come here by goto in the case of a |
7947 | 32-bit, non-UTF character whose value is greater than META_END. */ |
7948 | |
7949 | NORMAL_CHAR: |
7950 | meta = *pptr; /* Get the full 32 bits */ |
7951 | NORMAL_CHAR_SET: /* Character is already in meta */ |
7952 | matched_char = TRUE; |
7953 | |
7954 | /* For caseless UTF or UCP mode, check whether this character has more than |
7955 | one other case. If so, generate a special OP_PROP item instead of OP_CHARI. |
7956 | */ |
7957 | |
7958 | #ifdef SUPPORT_UNICODE |
7959 | if ((utf||ucp) && (options & PCRE2_CASELESS) != 0) |
7960 | { |
7961 | uint32_t caseset = UCD_CASESET(meta); |
7962 | if (caseset != 0) |
7963 | { |
7964 | *code++ = OP_PROP; |
7965 | *code++ = PT_CLIST; |
7966 | *code++ = caseset; |
7967 | if (firstcuflags == REQ_UNSET) |
7968 | firstcuflags = zerofirstcuflags = REQ_NONE; |
7969 | break; /* End handling this meta item */ |
7970 | } |
7971 | } |
7972 | #endif |
7973 | |
7974 | /* Caseful matches, or caseless and not one of the multicase characters. We |
7975 | come here by goto in the case of a positive class that contains only |
7976 | case-partners of a character with just two cases; matched_char has already |
7977 | been set TRUE and options fudged if necessary. */ |
7978 | |
7979 | CLASS_CASELESS_CHAR: |
7980 | |
7981 | /* Get the character's code units into mcbuffer, with the length in |
7982 | mclength. When not in UTF mode, the length is always 1. */ |
7983 | |
7984 | #ifdef SUPPORT_UNICODE |
7985 | if (utf) mclength = PRIV(ord2utf)(meta, mcbuffer); else |
7986 | #endif |
7987 | { |
7988 | mclength = 1; |
7989 | mcbuffer[0] = meta; |
7990 | } |
7991 | |
7992 | /* Generate the appropriate code */ |
7993 | |
7994 | *code++ = ((options & PCRE2_CASELESS) != 0)? OP_CHARI : OP_CHAR; |
7995 | memcpy(code, mcbuffer, CU2BYTES(mclength)); |
7996 | code += mclength; |
7997 | |
7998 | /* Remember if \r or \n were seen */ |
7999 | |
8000 | if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL) |
8001 | cb->external_flags |= PCRE2_HASCRORLF; |
8002 | |
8003 | /* Set the first and required code units appropriately. If no previous |
8004 | first code unit, set it from this character, but revert to none on a zero |
8005 | repeat. Otherwise, leave the firstcu value alone, and don't change it on |
8006 | a zero repeat. */ |
8007 | |
8008 | if (firstcuflags == REQ_UNSET) |
8009 | { |
8010 | zerofirstcuflags = REQ_NONE; |
8011 | zeroreqcu = reqcu; |
8012 | zeroreqcuflags = reqcuflags; |
8013 | |
8014 | /* If the character is more than one code unit long, we can set a single |
8015 | firstcu only if it is not to be matched caselessly. Multiple possible |
8016 | starting code units may be picked up later in the studying code. */ |
8017 | |
8018 | if (mclength == 1 || req_caseopt == 0) |
8019 | { |
8020 | firstcu = mcbuffer[0]; |
8021 | firstcuflags = req_caseopt; |
8022 | if (mclength != 1) |
8023 | { |
8024 | reqcu = code[-1]; |
8025 | reqcuflags = cb->req_varyopt; |
8026 | } |
8027 | } |
8028 | else firstcuflags = reqcuflags = REQ_NONE; |
8029 | } |
8030 | |
8031 | /* firstcu was previously set; we can set reqcu only if the length is |
8032 | 1 or the matching is caseful. */ |
8033 | |
8034 | else |
8035 | { |
8036 | zerofirstcu = firstcu; |
8037 | zerofirstcuflags = firstcuflags; |
8038 | zeroreqcu = reqcu; |
8039 | zeroreqcuflags = reqcuflags; |
8040 | if (mclength == 1 || req_caseopt == 0) |
8041 | { |
8042 | reqcu = code[-1]; |
8043 | reqcuflags = req_caseopt | cb->req_varyopt; |
8044 | } |
8045 | } |
8046 | |
8047 | /* If caselessness was temporarily instated, reset it. */ |
8048 | |
8049 | if (reset_caseful) |
8050 | { |
8051 | options &= ~PCRE2_CASELESS; |
8052 | req_caseopt = 0; |
8053 | reset_caseful = FALSE; |
8054 | } |
8055 | |
8056 | break; /* End literal character handling */ |
8057 | } /* End of big switch */ |
8058 | } /* End of big loop */ |
8059 | |
8060 | /* Control never reaches here. */ |
8061 | } |
8062 | |
8063 | |
8064 | |
8065 | /************************************************* |
8066 | * Compile regex: a sequence of alternatives * |
8067 | *************************************************/ |
8068 | |
8069 | /* On entry, pptr is pointing past the bracket meta, but on return it points to |
8070 | the closing bracket or META_END. The code variable is pointing at the code unit |
8071 | into which the BRA operator has been stored. This function is used during the |
8072 | pre-compile phase when we are trying to find out the amount of memory needed, |
8073 | as well as during the real compile phase. The value of lengthptr distinguishes |
8074 | the two phases. |
8075 | |
8076 | Arguments: |
8077 | options option bits, including any changes for this subpattern |
8078 | codeptr -> the address of the current code pointer |
8079 | pptrptr -> the address of the current parsed pattern pointer |
8080 | errorcodeptr -> pointer to error code variable |
8081 | skipunits skip this many code units at start (for brackets and OP_COND) |
8082 | firstcuptr place to put the first required code unit |
8083 | firstcuflagsptr place to put the first code unit flags |
8084 | reqcuptr place to put the last required code unit |
8085 | reqcuflagsptr place to put the last required code unit flags |
8086 | bcptr pointer to the chain of currently open branches |
8087 | cb points to the data block with tables pointers etc. |
8088 | lengthptr NULL during the real compile phase |
8089 | points to length accumulator during pre-compile phase |
8090 | |
8091 | Returns: 0 There has been an error |
8092 | +1 Success, this group must match at least one character |
8093 | -1 Success, this group may match an empty string |
8094 | */ |
8095 | |
8096 | static int |
8097 | compile_regex(uint32_t options, PCRE2_UCHAR **codeptr, uint32_t **pptrptr, |
8098 | int *errorcodeptr, uint32_t skipunits, uint32_t *firstcuptr, |
8099 | uint32_t *firstcuflagsptr, uint32_t *reqcuptr, uint32_t *reqcuflagsptr, |
8100 | branch_chain *bcptr, compile_block *cb, PCRE2_SIZE *lengthptr) |
8101 | { |
8102 | PCRE2_UCHAR *code = *codeptr; |
8103 | PCRE2_UCHAR *last_branch = code; |
8104 | PCRE2_UCHAR *start_bracket = code; |
8105 | BOOL lookbehind; |
8106 | open_capitem capitem; |
8107 | int capnumber = 0; |
8108 | int okreturn = 1; |
8109 | uint32_t *pptr = *pptrptr; |
8110 | uint32_t firstcu, reqcu; |
8111 | uint32_t lookbehindlength; |
8112 | uint32_t firstcuflags, reqcuflags; |
8113 | uint32_t branchfirstcu, branchreqcu; |
8114 | uint32_t branchfirstcuflags, branchreqcuflags; |
8115 | PCRE2_SIZE length; |
8116 | branch_chain bc; |
8117 | |
8118 | /* If set, call the external function that checks for stack availability. */ |
8119 | |
8120 | if (cb->cx->stack_guard != NULL && |
8121 | cb->cx->stack_guard(cb->parens_depth, cb->cx->stack_guard_data)) |
8122 | { |
8123 | *errorcodeptr= ERR33; |
8124 | return 0; |
8125 | } |
8126 | |
8127 | /* Miscellaneous initialization */ |
8128 | |
8129 | bc.outer = bcptr; |
8130 | bc.current_branch = code; |
8131 | |
8132 | firstcu = reqcu = 0; |
8133 | firstcuflags = reqcuflags = REQ_UNSET; |
8134 | |
8135 | /* Accumulate the length for use in the pre-compile phase. Start with the |
8136 | length of the BRA and KET and any extra code units that are required at the |
8137 | beginning. We accumulate in a local variable to save frequent testing of |
8138 | lengthptr for NULL. We cannot do this by looking at the value of 'code' at the |
8139 | start and end of each alternative, because compiled items are discarded during |
8140 | the pre-compile phase so that the workspace is not exceeded. */ |
8141 | |
8142 | length = 2 + 2*LINK_SIZE + skipunits; |
8143 | |
8144 | /* Remember if this is a lookbehind assertion, and if it is, save its length |
8145 | and skip over the pattern offset. */ |
8146 | |
8147 | lookbehind = *code == OP_ASSERTBACK || |
8148 | *code == OP_ASSERTBACK_NOT || |
8149 | *code == OP_ASSERTBACK_NA; |
8150 | |
8151 | if (lookbehind) |
8152 | { |
8153 | lookbehindlength = META_DATA(pptr[-1]); |
8154 | pptr += SIZEOFFSET; |
8155 | } |
8156 | else lookbehindlength = 0; |
8157 | |
8158 | /* If this is a capturing subpattern, add to the chain of open capturing items |
8159 | so that we can detect them if (*ACCEPT) is encountered. Note that only OP_CBRA |
8160 | need be tested here; changing this opcode to one of its variants, e.g. |
8161 | OP_SCBRAPOS, happens later, after the group has been compiled. */ |
8162 | |
8163 | if (*code == OP_CBRA) |
8164 | { |
8165 | capnumber = GET2(code, 1 + LINK_SIZE); |
8166 | capitem.number = capnumber; |
8167 | capitem.next = cb->open_caps; |
8168 | capitem.assert_depth = cb->assert_depth; |
8169 | cb->open_caps = &capitem; |
8170 | } |
8171 | |
8172 | /* Offset is set zero to mark that this bracket is still open */ |
8173 | |
8174 | PUT(code, 1, 0); |
8175 | code += 1 + LINK_SIZE + skipunits; |
8176 | |
8177 | /* Loop for each alternative branch */ |
8178 | |
8179 | for (;;) |
8180 | { |
8181 | int branch_return; |
8182 | |
8183 | /* Insert OP_REVERSE if this is as lookbehind assertion. */ |
8184 | |
8185 | if (lookbehind && lookbehindlength > 0) |
8186 | { |
8187 | *code++ = OP_REVERSE; |
8188 | PUTINC(code, 0, lookbehindlength); |
8189 | length += 1 + LINK_SIZE; |
8190 | } |
8191 | |
8192 | /* Now compile the branch; in the pre-compile phase its length gets added |
8193 | into the length. */ |
8194 | |
8195 | if ((branch_return = |
8196 | compile_branch(&options, &code, &pptr, errorcodeptr, &branchfirstcu, |
8197 | &branchfirstcuflags, &branchreqcu, &branchreqcuflags, &bc, |
8198 | cb, (lengthptr == NULL)? NULL : &length)) == 0) |
8199 | return 0; |
8200 | |
8201 | /* If a branch can match an empty string, so can the whole group. */ |
8202 | |
8203 | if (branch_return < 0) okreturn = -1; |
8204 | |
8205 | /* In the real compile phase, there is some post-processing to be done. */ |
8206 | |
8207 | if (lengthptr == NULL) |
8208 | { |
8209 | /* If this is the first branch, the firstcu and reqcu values for the |
8210 | branch become the values for the regex. */ |
8211 | |
8212 | if (*last_branch != OP_ALT) |
8213 | { |
8214 | firstcu = branchfirstcu; |
8215 | firstcuflags = branchfirstcuflags; |
8216 | reqcu = branchreqcu; |
8217 | reqcuflags = branchreqcuflags; |
8218 | } |
8219 | |
8220 | /* If this is not the first branch, the first char and reqcu have to |
8221 | match the values from all the previous branches, except that if the |
8222 | previous value for reqcu didn't have REQ_VARY set, it can still match, |
8223 | and we set REQ_VARY for the group from this branch's value. */ |
8224 | |
8225 | else |
8226 | { |
8227 | /* If we previously had a firstcu, but it doesn't match the new branch, |
8228 | we have to abandon the firstcu for the regex, but if there was |
8229 | previously no reqcu, it takes on the value of the old firstcu. */ |
8230 | |
8231 | if (firstcuflags != branchfirstcuflags || firstcu != branchfirstcu) |
8232 | { |
8233 | if (firstcuflags < REQ_NONE) |
8234 | { |
8235 | if (reqcuflags >= REQ_NONE) |
8236 | { |
8237 | reqcu = firstcu; |
8238 | reqcuflags = firstcuflags; |
8239 | } |
8240 | } |
8241 | firstcuflags = REQ_NONE; |
8242 | } |
8243 | |
8244 | /* If we (now or from before) have no firstcu, a firstcu from the |
8245 | branch becomes a reqcu if there isn't a branch reqcu. */ |
8246 | |
8247 | if (firstcuflags >= REQ_NONE && branchfirstcuflags < REQ_NONE && |
8248 | branchreqcuflags >= REQ_NONE) |
8249 | { |
8250 | branchreqcu = branchfirstcu; |
8251 | branchreqcuflags = branchfirstcuflags; |
8252 | } |
8253 | |
8254 | /* Now ensure that the reqcus match */ |
8255 | |
8256 | if (((reqcuflags & ~REQ_VARY) != (branchreqcuflags & ~REQ_VARY)) || |
8257 | reqcu != branchreqcu) |
8258 | reqcuflags = REQ_NONE; |
8259 | else |
8260 | { |
8261 | reqcu = branchreqcu; |
8262 | reqcuflags |= branchreqcuflags; /* To "or" REQ_VARY if present */ |
8263 | } |
8264 | } |
8265 | } |
8266 | |
8267 | /* Handle reaching the end of the expression, either ')' or end of pattern. |
8268 | In the real compile phase, go back through the alternative branches and |
8269 | reverse the chain of offsets, with the field in the BRA item now becoming an |
8270 | offset to the first alternative. If there are no alternatives, it points to |
8271 | the end of the group. The length in the terminating ket is always the length |
8272 | of the whole bracketed item. Return leaving the pointer at the terminating |
8273 | char. */ |
8274 | |
8275 | if (META_CODE(*pptr) != META_ALT) |
8276 | { |
8277 | if (lengthptr == NULL) |
8278 | { |
8279 | PCRE2_SIZE branch_length = code - last_branch; |
8280 | do |
8281 | { |
8282 | PCRE2_SIZE prev_length = GET(last_branch, 1); |
8283 | PUT(last_branch, 1, branch_length); |
8284 | branch_length = prev_length; |
8285 | last_branch -= branch_length; |
8286 | } |
8287 | while (branch_length > 0); |
8288 | } |
8289 | |
8290 | /* Fill in the ket */ |
8291 | |
8292 | *code = OP_KET; |
8293 | PUT(code, 1, (int)(code - start_bracket)); |
8294 | code += 1 + LINK_SIZE; |
8295 | |
8296 | /* If it was a capturing subpattern, remove the block from the chain. */ |
8297 | |
8298 | if (capnumber > 0) cb->open_caps = cb->open_caps->next; |
8299 | |
8300 | /* Set values to pass back */ |
8301 | |
8302 | *codeptr = code; |
8303 | *pptrptr = pptr; |
8304 | *firstcuptr = firstcu; |
8305 | *firstcuflagsptr = firstcuflags; |
8306 | *reqcuptr = reqcu; |
8307 | *reqcuflagsptr = reqcuflags; |
8308 | if (lengthptr != NULL) |
8309 | { |
8310 | if (OFLOW_MAX - *lengthptr < length) |
8311 | { |
8312 | *errorcodeptr = ERR20; |
8313 | return 0; |
8314 | } |
8315 | *lengthptr += length; |
8316 | } |
8317 | return okreturn; |
8318 | } |
8319 | |
8320 | /* Another branch follows. In the pre-compile phase, we can move the code |
8321 | pointer back to where it was for the start of the first branch. (That is, |
8322 | pretend that each branch is the only one.) |
8323 | |
8324 | In the real compile phase, insert an ALT node. Its length field points back |
8325 | to the previous branch while the bracket remains open. At the end the chain |
8326 | is reversed. It's done like this so that the start of the bracket has a |
8327 | zero offset until it is closed, making it possible to detect recursion. */ |
8328 | |
8329 | if (lengthptr != NULL) |
8330 | { |
8331 | code = *codeptr + 1 + LINK_SIZE + skipunits; |
8332 | length += 1 + LINK_SIZE; |
8333 | } |
8334 | else |
8335 | { |
8336 | *code = OP_ALT; |
8337 | PUT(code, 1, (int)(code - last_branch)); |
8338 | bc.current_branch = last_branch = code; |
8339 | code += 1 + LINK_SIZE; |
8340 | } |
8341 | |
8342 | /* Set the lookbehind length (if not in a lookbehind the value will be zero) |
8343 | and then advance past the vertical bar. */ |
8344 | |
8345 | lookbehindlength = META_DATA(*pptr); |
8346 | pptr++; |
8347 | } |
8348 | /* Control never reaches here */ |
8349 | } |
8350 | |
8351 | |
8352 | |
8353 | /************************************************* |
8354 | * Check for anchored pattern * |
8355 | *************************************************/ |
8356 | |
8357 | /* Try to find out if this is an anchored regular expression. Consider each |
8358 | alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket |
8359 | all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then |
8360 | it's anchored. However, if this is a multiline pattern, then only OP_SOD will |
8361 | be found, because ^ generates OP_CIRCM in that mode. |
8362 | |
8363 | We can also consider a regex to be anchored if OP_SOM starts all its branches. |
8364 | This is the code for \G, which means "match at start of match position, taking |
8365 | into account the match offset". |
8366 | |
8367 | A branch is also implicitly anchored if it starts with .* and DOTALL is set, |
8368 | because that will try the rest of the pattern at all possible matching points, |
8369 | so there is no point trying again.... er .... |
8370 | |
8371 | .... except when the .* appears inside capturing parentheses, and there is a |
8372 | subsequent back reference to those parentheses. We haven't enough information |
8373 | to catch that case precisely. |
8374 | |
8375 | At first, the best we could do was to detect when .* was in capturing brackets |
8376 | and the highest back reference was greater than or equal to that level. |
8377 | However, by keeping a bitmap of the first 31 back references, we can catch some |
8378 | of the more common cases more precisely. |
8379 | |
8380 | ... A second exception is when the .* appears inside an atomic group, because |
8381 | this prevents the number of characters it matches from being adjusted. |
8382 | |
8383 | Arguments: |
8384 | code points to start of the compiled pattern |
8385 | bracket_map a bitmap of which brackets we are inside while testing; this |
8386 | handles up to substring 31; after that we just have to take |
8387 | the less precise approach |
8388 | cb points to the compile data block |
8389 | atomcount atomic group level |
8390 | inassert TRUE if in an assertion |
8391 | |
8392 | Returns: TRUE or FALSE |
8393 | */ |
8394 | |
8395 | static BOOL |
8396 | is_anchored(PCRE2_SPTR code, uint32_t bracket_map, compile_block *cb, |
8397 | int atomcount, BOOL inassert) |
8398 | { |
8399 | do { |
8400 | PCRE2_SPTR scode = first_significant_code( |
8401 | code + PRIV(OP_lengths)[*code], FALSE); |
8402 | int op = *scode; |
8403 | |
8404 | /* Non-capturing brackets */ |
8405 | |
8406 | if (op == OP_BRA || op == OP_BRAPOS || |
8407 | op == OP_SBRA || op == OP_SBRAPOS) |
8408 | { |
8409 | if (!is_anchored(scode, bracket_map, cb, atomcount, inassert)) |
8410 | return FALSE; |
8411 | } |
8412 | |
8413 | /* Capturing brackets */ |
8414 | |
8415 | else if (op == OP_CBRA || op == OP_CBRAPOS || |
8416 | op == OP_SCBRA || op == OP_SCBRAPOS) |
8417 | { |
8418 | int n = GET2(scode, 1+LINK_SIZE); |
8419 | uint32_t new_map = bracket_map | ((n < 32)? (1u << n) : 1); |
8420 | if (!is_anchored(scode, new_map, cb, atomcount, inassert)) return FALSE; |
8421 | } |
8422 | |
8423 | /* Positive forward assertion */ |
8424 | |
8425 | else if (op == OP_ASSERT || op == OP_ASSERT_NA) |
8426 | { |
8427 | if (!is_anchored(scode, bracket_map, cb, atomcount, TRUE)) return FALSE; |
8428 | } |
8429 | |
8430 | /* Condition. If there is no second branch, it can't be anchored. */ |
8431 | |
8432 | else if (op == OP_COND || op == OP_SCOND) |
8433 | { |
8434 | if (scode[GET(scode,1)] != OP_ALT) return FALSE; |
8435 | if (!is_anchored(scode, bracket_map, cb, atomcount, inassert)) |
8436 | return FALSE; |
8437 | } |
8438 | |
8439 | /* Atomic groups */ |
8440 | |
8441 | else if (op == OP_ONCE) |
8442 | { |
8443 | if (!is_anchored(scode, bracket_map, cb, atomcount + 1, inassert)) |
8444 | return FALSE; |
8445 | } |
8446 | |
8447 | /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and |
8448 | it isn't in brackets that are or may be referenced or inside an atomic |
8449 | group or an assertion. Also the pattern must not contain *PRUNE or *SKIP, |
8450 | because these break the feature. Consider, for example, /(?s).*?(*PRUNE)b/ |
8451 | with the subject "aab", which matches "b", i.e. not at the start of a line. |
8452 | There is also an option that disables auto-anchoring. */ |
8453 | |
8454 | else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR || |
8455 | op == OP_TYPEPOSSTAR)) |
8456 | { |
8457 | if (scode[1] != OP_ALLANY || (bracket_map & cb->backref_map) != 0 || |
8458 | atomcount > 0 || cb->had_pruneorskip || inassert || |
8459 | (cb->external_options & PCRE2_NO_DOTSTAR_ANCHOR) != 0) |
8460 | return FALSE; |
8461 | } |
8462 | |
8463 | /* Check for explicit anchoring */ |
8464 | |
8465 | else if (op != OP_SOD && op != OP_SOM && op != OP_CIRC) return FALSE; |
8466 | |
8467 | code += GET(code, 1); |
8468 | } |
8469 | while (*code == OP_ALT); /* Loop for each alternative */ |
8470 | return TRUE; |
8471 | } |
8472 | |
8473 | |
8474 | |
8475 | /************************************************* |
8476 | * Check for starting with ^ or .* * |
8477 | *************************************************/ |
8478 | |
8479 | /* This is called to find out if every branch starts with ^ or .* so that |
8480 | "first char" processing can be done to speed things up in multiline |
8481 | matching and for non-DOTALL patterns that start with .* (which must start at |
8482 | the beginning or after \n). As in the case of is_anchored() (see above), we |
8483 | have to take account of back references to capturing brackets that contain .* |
8484 | because in that case we can't make the assumption. Also, the appearance of .* |
8485 | inside atomic brackets or in an assertion, or in a pattern that contains *PRUNE |
8486 | or *SKIP does not count, because once again the assumption no longer holds. |
8487 | |
8488 | Arguments: |
8489 | code points to start of the compiled pattern or a group |
8490 | bracket_map a bitmap of which brackets we are inside while testing; this |
8491 | handles up to substring 31; after that we just have to take |
8492 | the less precise approach |
8493 | cb points to the compile data |
8494 | atomcount atomic group level |
8495 | inassert TRUE if in an assertion |
8496 | |
8497 | Returns: TRUE or FALSE |
8498 | */ |
8499 | |
8500 | static BOOL |
8501 | is_startline(PCRE2_SPTR code, unsigned int bracket_map, compile_block *cb, |
8502 | int atomcount, BOOL inassert) |
8503 | { |
8504 | do { |
8505 | PCRE2_SPTR scode = first_significant_code( |
8506 | code + PRIV(OP_lengths)[*code], FALSE); |
8507 | int op = *scode; |
8508 | |
8509 | /* If we are at the start of a conditional assertion group, *both* the |
8510 | conditional assertion *and* what follows the condition must satisfy the test |
8511 | for start of line. Other kinds of condition fail. Note that there may be an |
8512 | auto-callout at the start of a condition. */ |
8513 | |
8514 | if (op == OP_COND) |
8515 | { |
8516 | scode += 1 + LINK_SIZE; |
8517 | |
8518 | if (*scode == OP_CALLOUT) scode += PRIV(OP_lengths)[OP_CALLOUT]; |
8519 | else if (*scode == OP_CALLOUT_STR) scode += GET(scode, 1 + 2*LINK_SIZE); |
8520 | |
8521 | switch (*scode) |
8522 | { |
8523 | case OP_CREF: |
8524 | case OP_DNCREF: |
8525 | case OP_RREF: |
8526 | case OP_DNRREF: |
8527 | case OP_FAIL: |
8528 | case OP_FALSE: |
8529 | case OP_TRUE: |
8530 | return FALSE; |
8531 | |
8532 | default: /* Assertion */ |
8533 | if (!is_startline(scode, bracket_map, cb, atomcount, TRUE)) return FALSE; |
8534 | do scode += GET(scode, 1); while (*scode == OP_ALT); |
8535 | scode += 1 + LINK_SIZE; |
8536 | break; |
8537 | } |
8538 | scode = first_significant_code(scode, FALSE); |
8539 | op = *scode; |
8540 | } |
8541 | |
8542 | /* Non-capturing brackets */ |
8543 | |
8544 | if (op == OP_BRA || op == OP_BRAPOS || |
8545 | op == OP_SBRA || op == OP_SBRAPOS) |
8546 | { |
8547 | if (!is_startline(scode, bracket_map, cb, atomcount, inassert)) |
8548 | return FALSE; |
8549 | } |
8550 | |
8551 | /* Capturing brackets */ |
8552 | |
8553 | else if (op == OP_CBRA || op == OP_CBRAPOS || |
8554 | op == OP_SCBRA || op == OP_SCBRAPOS) |
8555 | { |
8556 | int n = GET2(scode, 1+LINK_SIZE); |
8557 | unsigned int new_map = bracket_map | ((n < 32)? (1u << n) : 1); |
8558 | if (!is_startline(scode, new_map, cb, atomcount, inassert)) return FALSE; |
8559 | } |
8560 | |
8561 | /* Positive forward assertions */ |
8562 | |
8563 | else if (op == OP_ASSERT || op == OP_ASSERT_NA) |
8564 | { |
8565 | if (!is_startline(scode, bracket_map, cb, atomcount, TRUE)) |
8566 | return FALSE; |
8567 | } |
8568 | |
8569 | /* Atomic brackets */ |
8570 | |
8571 | else if (op == OP_ONCE) |
8572 | { |
8573 | if (!is_startline(scode, bracket_map, cb, atomcount + 1, inassert)) |
8574 | return FALSE; |
8575 | } |
8576 | |
8577 | /* .* means "start at start or after \n" if it isn't in atomic brackets or |
8578 | brackets that may be referenced or an assertion, and as long as the pattern |
8579 | does not contain *PRUNE or *SKIP, because these break the feature. Consider, |
8580 | for example, /.*?a(*PRUNE)b/ with the subject "aab", which matches "ab", |
8581 | i.e. not at the start of a line. There is also an option that disables this |
8582 | optimization. */ |
8583 | |
8584 | else if (op == OP_TYPESTAR || op == OP_TYPEMINSTAR || op == OP_TYPEPOSSTAR) |
8585 | { |
8586 | if (scode[1] != OP_ANY || (bracket_map & cb->backref_map) != 0 || |
8587 | atomcount > 0 || cb->had_pruneorskip || inassert || |
8588 | (cb->external_options & PCRE2_NO_DOTSTAR_ANCHOR) != 0) |
8589 | return FALSE; |
8590 | } |
8591 | |
8592 | /* Check for explicit circumflex; anything else gives a FALSE result. Note |
8593 | in particular that this includes atomic brackets OP_ONCE because the number |
8594 | of characters matched by .* cannot be adjusted inside them. */ |
8595 | |
8596 | else if (op != OP_CIRC && op != OP_CIRCM) return FALSE; |
8597 | |
8598 | /* Move on to the next alternative */ |
8599 | |
8600 | code += GET(code, 1); |
8601 | } |
8602 | while (*code == OP_ALT); /* Loop for each alternative */ |
8603 | return TRUE; |
8604 | } |
8605 | |
8606 | |
8607 | |
8608 | /************************************************* |
8609 | * Scan compiled regex for recursion reference * |
8610 | *************************************************/ |
8611 | |
8612 | /* This function scans through a compiled pattern until it finds an instance of |
8613 | OP_RECURSE. |
8614 | |
8615 | Arguments: |
8616 | code points to start of expression |
8617 | utf TRUE in UTF mode |
8618 | |
8619 | Returns: pointer to the opcode for OP_RECURSE, or NULL if not found |
8620 | */ |
8621 | |
8622 | static PCRE2_SPTR |
8623 | find_recurse(PCRE2_SPTR code, BOOL utf) |
8624 | { |
8625 | for (;;) |
8626 | { |
8627 | PCRE2_UCHAR c = *code; |
8628 | if (c == OP_END) return NULL; |
8629 | if (c == OP_RECURSE) return code; |
8630 | |
8631 | /* XCLASS is used for classes that cannot be represented just by a bit map. |
8632 | This includes negated single high-valued characters. CALLOUT_STR is used for |
8633 | callouts with string arguments. In both cases the length in the table is |
8634 | zero; the actual length is stored in the compiled code. */ |
8635 | |
8636 | if (c == OP_XCLASS) code += GET(code, 1); |
8637 | else if (c == OP_CALLOUT_STR) code += GET(code, 1 + 2*LINK_SIZE); |
8638 | |
8639 | /* Otherwise, we can get the item's length from the table, except that for |
8640 | repeated character types, we have to test for \p and \P, which have an extra |
8641 | two code units of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, |
8642 | we must add in its length. */ |
8643 | |
8644 | else |
8645 | { |
8646 | switch(c) |
8647 | { |
8648 | case OP_TYPESTAR: |
8649 | case OP_TYPEMINSTAR: |
8650 | case OP_TYPEPLUS: |
8651 | case OP_TYPEMINPLUS: |
8652 | case OP_TYPEQUERY: |
8653 | case OP_TYPEMINQUERY: |
8654 | case OP_TYPEPOSSTAR: |
8655 | case OP_TYPEPOSPLUS: |
8656 | case OP_TYPEPOSQUERY: |
8657 | if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2; |
8658 | break; |
8659 | |
8660 | case OP_TYPEPOSUPTO: |
8661 | case OP_TYPEUPTO: |
8662 | case OP_TYPEMINUPTO: |
8663 | case OP_TYPEEXACT: |
8664 | if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP) |
8665 | code += 2; |
8666 | break; |
8667 | |
8668 | case OP_MARK: |
8669 | case OP_COMMIT_ARG: |
8670 | case OP_PRUNE_ARG: |
8671 | case OP_SKIP_ARG: |
8672 | case OP_THEN_ARG: |
8673 | code += code[1]; |
8674 | break; |
8675 | } |
8676 | |
8677 | /* Add in the fixed length from the table */ |
8678 | |
8679 | code += PRIV(OP_lengths)[c]; |
8680 | |
8681 | /* In UTF-8 and UTF-16 modes, opcodes that are followed by a character may |
8682 | be followed by a multi-unit character. The length in the table is a |
8683 | minimum, so we have to arrange to skip the extra units. */ |
8684 | |
8685 | #ifdef MAYBE_UTF_MULTI |
8686 | if (utf) switch(c) |
8687 | { |
8688 | case OP_CHAR: |
8689 | case OP_CHARI: |
8690 | case OP_NOT: |
8691 | case OP_NOTI: |
8692 | case OP_EXACT: |
8693 | case OP_EXACTI: |
8694 | case OP_NOTEXACT: |
8695 | case OP_NOTEXACTI: |
8696 | case OP_UPTO: |
8697 | case OP_UPTOI: |
8698 | case OP_NOTUPTO: |
8699 | case OP_NOTUPTOI: |
8700 | case OP_MINUPTO: |
8701 | case OP_MINUPTOI: |
8702 | case OP_NOTMINUPTO: |
8703 | case OP_NOTMINUPTOI: |
8704 | case OP_POSUPTO: |
8705 | case OP_POSUPTOI: |
8706 | case OP_NOTPOSUPTO: |
8707 | case OP_NOTPOSUPTOI: |
8708 | case OP_STAR: |
8709 | case OP_STARI: |
8710 | case OP_NOTSTAR: |
8711 | case OP_NOTSTARI: |
8712 | case OP_MINSTAR: |
8713 | case OP_MINSTARI: |
8714 | case OP_NOTMINSTAR: |
8715 | case OP_NOTMINSTARI: |
8716 | case OP_POSSTAR: |
8717 | case OP_POSSTARI: |
8718 | case OP_NOTPOSSTAR: |
8719 | case OP_NOTPOSSTARI: |
8720 | case OP_PLUS: |
8721 | case OP_PLUSI: |
8722 | case OP_NOTPLUS: |
8723 | case OP_NOTPLUSI: |
8724 | case OP_MINPLUS: |
8725 | case OP_MINPLUSI: |
8726 | case OP_NOTMINPLUS: |
8727 | case OP_NOTMINPLUSI: |
8728 | case OP_POSPLUS: |
8729 | case OP_POSPLUSI: |
8730 | case OP_NOTPOSPLUS: |
8731 | case OP_NOTPOSPLUSI: |
8732 | case OP_QUERY: |
8733 | case OP_QUERYI: |
8734 | case OP_NOTQUERY: |
8735 | case OP_NOTQUERYI: |
8736 | case OP_MINQUERY: |
8737 | case OP_MINQUERYI: |
8738 | case OP_NOTMINQUERY: |
8739 | case OP_NOTMINQUERYI: |
8740 | case OP_POSQUERY: |
8741 | case OP_POSQUERYI: |
8742 | case OP_NOTPOSQUERY: |
8743 | case OP_NOTPOSQUERYI: |
8744 | if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]); |
8745 | break; |
8746 | } |
8747 | #else |
8748 | (void)(utf); /* Keep compiler happy by referencing function argument */ |
8749 | #endif /* MAYBE_UTF_MULTI */ |
8750 | } |
8751 | } |
8752 | } |
8753 | |
8754 | |
8755 | |
8756 | /************************************************* |
8757 | * Check for asserted fixed first code unit * |
8758 | *************************************************/ |
8759 | |
8760 | /* During compilation, the "first code unit" settings from forward assertions |
8761 | are discarded, because they can cause conflicts with actual literals that |
8762 | follow. However, if we end up without a first code unit setting for an |
8763 | unanchored pattern, it is worth scanning the regex to see if there is an |
8764 | initial asserted first code unit. If all branches start with the same asserted |
8765 | code unit, or with a non-conditional bracket all of whose alternatives start |
8766 | with the same asserted code unit (recurse ad lib), then we return that code |
8767 | unit, with the flags set to zero or REQ_CASELESS; otherwise return zero with |
8768 | REQ_NONE in the flags. |
8769 | |
8770 | Arguments: |
8771 | code points to start of compiled pattern |
8772 | flags points to the first code unit flags |
8773 | inassert non-zero if in an assertion |
8774 | |
8775 | Returns: the fixed first code unit, or 0 with REQ_NONE in flags |
8776 | */ |
8777 | |
8778 | static uint32_t |
8779 | find_firstassertedcu(PCRE2_SPTR code, uint32_t *flags, uint32_t inassert) |
8780 | { |
8781 | uint32_t c = 0; |
8782 | uint32_t cflags = REQ_NONE; |
8783 | |
8784 | *flags = REQ_NONE; |
8785 | do { |
8786 | uint32_t d; |
8787 | uint32_t dflags; |
8788 | int xl = (*code == OP_CBRA || *code == OP_SCBRA || |
8789 | *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? IMM2_SIZE:0; |
8790 | PCRE2_SPTR scode = first_significant_code(code + 1+LINK_SIZE + xl, TRUE); |
8791 | PCRE2_UCHAR op = *scode; |
8792 | |
8793 | switch(op) |
8794 | { |
8795 | default: |
8796 | return 0; |
8797 | |
8798 | case OP_BRA: |
8799 | case OP_BRAPOS: |
8800 | case OP_CBRA: |
8801 | case OP_SCBRA: |
8802 | case OP_CBRAPOS: |
8803 | case OP_SCBRAPOS: |
8804 | case OP_ASSERT: |
8805 | case OP_ASSERT_NA: |
8806 | case OP_ONCE: |
8807 | case OP_SCRIPT_RUN: |
8808 | d = find_firstassertedcu(scode, &dflags, inassert + |
8809 | ((op == OP_ASSERT || op == OP_ASSERT_NA)?1:0)); |
8810 | if (dflags >= REQ_NONE) return 0; |
8811 | if (cflags >= REQ_NONE) { c = d; cflags = dflags; } |
8812 | else if (c != d || cflags != dflags) return 0; |
8813 | break; |
8814 | |
8815 | case OP_EXACT: |
8816 | scode += IMM2_SIZE; |
8817 | /* Fall through */ |
8818 | |
8819 | case OP_CHAR: |
8820 | case OP_PLUS: |
8821 | case OP_MINPLUS: |
8822 | case OP_POSPLUS: |
8823 | if (inassert == 0) return 0; |
8824 | if (cflags >= REQ_NONE) { c = scode[1]; cflags = 0; } |
8825 | else if (c != scode[1]) return 0; |
8826 | break; |
8827 | |
8828 | case OP_EXACTI: |
8829 | scode += IMM2_SIZE; |
8830 | /* Fall through */ |
8831 | |
8832 | case OP_CHARI: |
8833 | case OP_PLUSI: |
8834 | case OP_MINPLUSI: |
8835 | case OP_POSPLUSI: |
8836 | if (inassert == 0) return 0; |
8837 | |
8838 | /* If the character is more than one code unit long, we cannot set its |
8839 | first code unit when matching caselessly. Later scanning may pick up |
8840 | multiple code units. */ |
8841 | |
8842 | #ifdef SUPPORT_UNICODE |
8843 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
8844 | if (scode[1] >= 0x80) return 0; |
8845 | #elif PCRE2_CODE_UNIT_WIDTH == 16 |
8846 | if (scode[1] >= 0xd800 && scode[1] <= 0xdfff) return 0; |
8847 | #endif |
8848 | #endif |
8849 | |
8850 | if (cflags >= REQ_NONE) { c = scode[1]; cflags = REQ_CASELESS; } |
8851 | else if (c != scode[1]) return 0; |
8852 | break; |
8853 | } |
8854 | |
8855 | code += GET(code, 1); |
8856 | } |
8857 | while (*code == OP_ALT); |
8858 | |
8859 | *flags = cflags; |
8860 | return c; |
8861 | } |
8862 | |
8863 | |
8864 | |
8865 | /************************************************* |
8866 | * Add an entry to the name/number table * |
8867 | *************************************************/ |
8868 | |
8869 | /* This function is called between compiling passes to add an entry to the |
8870 | name/number table, maintaining alphabetical order. Checking for permitted |
8871 | and forbidden duplicates has already been done. |
8872 | |
8873 | Arguments: |
8874 | cb the compile data block |
8875 | name the name to add |
8876 | length the length of the name |
8877 | groupno the group number |
8878 | tablecount the count of names in the table so far |
8879 | |
8880 | Returns: nothing |
8881 | */ |
8882 | |
8883 | static void |
8884 | add_name_to_table(compile_block *cb, PCRE2_SPTR name, int length, |
8885 | unsigned int groupno, uint32_t tablecount) |
8886 | { |
8887 | uint32_t i; |
8888 | PCRE2_UCHAR *slot = cb->name_table; |
8889 | |
8890 | for (i = 0; i < tablecount; i++) |
8891 | { |
8892 | int crc = memcmp(name, slot+IMM2_SIZE, CU2BYTES(length)); |
8893 | if (crc == 0 && slot[IMM2_SIZE+length] != 0) |
8894 | crc = -1; /* Current name is a substring */ |
8895 | |
8896 | /* Make space in the table and break the loop for an earlier name. For a |
8897 | duplicate or later name, carry on. We do this for duplicates so that in the |
8898 | simple case (when ?(| is not used) they are in order of their numbers. In all |
8899 | cases they are in the order in which they appear in the pattern. */ |
8900 | |
8901 | if (crc < 0) |
8902 | { |
8903 | (void)memmove(slot + cb->name_entry_size, slot, |
8904 | CU2BYTES((tablecount - i) * cb->name_entry_size)); |
8905 | break; |
8906 | } |
8907 | |
8908 | /* Continue the loop for a later or duplicate name */ |
8909 | |
8910 | slot += cb->name_entry_size; |
8911 | } |
8912 | |
8913 | PUT2(slot, 0, groupno); |
8914 | memcpy(slot + IMM2_SIZE, name, CU2BYTES(length)); |
8915 | |
8916 | /* Add a terminating zero and fill the rest of the slot with zeroes so that |
8917 | the memory is all initialized. Otherwise valgrind moans about uninitialized |
8918 | memory when saving serialized compiled patterns. */ |
8919 | |
8920 | memset(slot + IMM2_SIZE + length, 0, |
8921 | CU2BYTES(cb->name_entry_size - length - IMM2_SIZE)); |
8922 | } |
8923 | |
8924 | |
8925 | |
8926 | /************************************************* |
8927 | * Skip in parsed pattern * |
8928 | *************************************************/ |
8929 | |
8930 | /* This function is called to skip parts of the parsed pattern when finding the |
8931 | length of a lookbehind branch. It is called after (*ACCEPT) and (*FAIL) to find |
8932 | the end of the branch, it is called to skip over an internal lookaround or |
8933 | (DEFINE) group, and it is also called to skip to the end of a class, during |
8934 | which it will never encounter nested groups (but there's no need to have |
8935 | special code for that). |
8936 | |
8937 | When called to find the end of a branch or group, pptr must point to the first |
8938 | meta code inside the branch, not the branch-starting code. In other cases it |
8939 | can point to the item that causes the function to be called. |
8940 | |
8941 | Arguments: |
8942 | pptr current pointer to skip from |
8943 | skiptype PSKIP_CLASS when skipping to end of class |
8944 | PSKIP_ALT when META_ALT ends the skip |
8945 | PSKIP_KET when only META_KET ends the skip |
8946 | |
8947 | Returns: new value of pptr |
8948 | NULL if META_END is reached - should never occur |
8949 | or for an unknown meta value - likewise |
8950 | */ |
8951 | |
8952 | static uint32_t * |
8953 | parsed_skip(uint32_t *pptr, uint32_t skiptype) |
8954 | { |
8955 | uint32_t nestlevel = 0; |
8956 | |
8957 | for (;; pptr++) |
8958 | { |
8959 | uint32_t meta = META_CODE(*pptr); |
8960 | |
8961 | switch(meta) |
8962 | { |
8963 | default: /* Just skip over most items */ |
8964 | if (meta < META_END) continue; /* Literal */ |
8965 | break; |
8966 | |
8967 | /* This should never occur. */ |
8968 | |
8969 | case META_END: |
8970 | return NULL; |
8971 | |
8972 | /* The data for these items is variable in length. */ |
8973 | |
8974 | case META_BACKREF: /* Offset is present only if group >= 10 */ |
8975 | if (META_DATA(*pptr) >= 10) pptr += SIZEOFFSET; |
8976 | break; |
8977 | |
8978 | case META_ESCAPE: /* A few escapes are followed by data items. */ |
8979 | switch (META_DATA(*pptr)) |
8980 | { |
8981 | case ESC_P: |
8982 | case ESC_p: |
8983 | pptr += 1; |
8984 | break; |
8985 | |
8986 | case ESC_g: |
8987 | case ESC_k: |
8988 | pptr += 1 + SIZEOFFSET; |
8989 | break; |
8990 | } |
8991 | break; |
8992 | |
8993 | case META_MARK: /* Add the length of the name. */ |
8994 | case META_COMMIT_ARG: |
8995 | case META_PRUNE_ARG: |
8996 | case META_SKIP_ARG: |
8997 | case META_THEN_ARG: |
8998 | pptr += pptr[1]; |
8999 | break; |
9000 | |
9001 | /* These are the "active" items in this loop. */ |
9002 | |
9003 | case META_CLASS_END: |
9004 | if (skiptype == PSKIP_CLASS) return pptr; |
9005 | break; |
9006 | |
9007 | case META_ATOMIC: |
9008 | case META_CAPTURE: |
9009 | case META_COND_ASSERT: |
9010 | case META_COND_DEFINE: |
9011 | case META_COND_NAME: |
9012 | case META_COND_NUMBER: |
9013 | case META_COND_RNAME: |
9014 | case META_COND_RNUMBER: |
9015 | case META_COND_VERSION: |
9016 | case META_LOOKAHEAD: |
9017 | case META_LOOKAHEADNOT: |
9018 | case META_LOOKAHEAD_NA: |
9019 | case META_LOOKBEHIND: |
9020 | case META_LOOKBEHINDNOT: |
9021 | case META_LOOKBEHIND_NA: |
9022 | case META_NOCAPTURE: |
9023 | case META_SCRIPT_RUN: |
9024 | nestlevel++; |
9025 | break; |
9026 | |
9027 | case META_ALT: |
9028 | if (nestlevel == 0 && skiptype == PSKIP_ALT) return pptr; |
9029 | break; |
9030 | |
9031 | case META_KET: |
9032 | if (nestlevel == 0) return pptr; |
9033 | nestlevel--; |
9034 | break; |
9035 | } |
9036 | |
9037 | /* The extra data item length for each meta is in a table. */ |
9038 | |
9039 | meta = (meta >> 16) & 0x7fff; |
9040 | if (meta >= sizeof(meta_extra_lengths)) return NULL; |
9041 | pptr += meta_extra_lengths[meta]; |
9042 | } |
9043 | /* Control never reaches here */ |
9044 | return pptr; |
9045 | } |
9046 | |
9047 | |
9048 | |
9049 | /************************************************* |
9050 | * Find length of a parsed group * |
9051 | *************************************************/ |
9052 | |
9053 | /* This is called for nested groups within a branch of a lookbehind whose |
9054 | length is being computed. If all the branches in the nested group have the same |
9055 | length, that is OK. On entry, the pointer must be at the first element after |
9056 | the group initializing code. On exit it points to OP_KET. Caching is used to |
9057 | improve processing speed when the same capturing group occurs many times. |
9058 | |
9059 | Arguments: |
9060 | pptrptr pointer to pointer in the parsed pattern |
9061 | isinline FALSE if a reference or recursion; TRUE for inline group |
9062 | errcodeptr pointer to the errorcode |
9063 | lcptr pointer to the loop counter |
9064 | group number of captured group or -1 for a non-capturing group |
9065 | recurses chain of recurse_check to catch mutual recursion |
9066 | cb pointer to the compile data |
9067 | |
9068 | Returns: the group length or a negative number |
9069 | */ |
9070 | |
9071 | static int |
9072 | get_grouplength(uint32_t **pptrptr, BOOL isinline, int *errcodeptr, int *lcptr, |
9073 | int group, parsed_recurse_check *recurses, compile_block *cb) |
9074 | { |
9075 | int branchlength; |
9076 | int grouplength = -1; |
9077 | |
9078 | /* The cache can be used only if there is no possibility of there being two |
9079 | groups with the same number. We do not need to set the end pointer for a group |
9080 | that is being processed as a back reference or recursion, but we must do so for |
9081 | an inline group. */ |
9082 | |
9083 | if (group > 0 && (cb->external_flags & PCRE2_DUPCAPUSED) == 0) |
9084 | { |
9085 | uint32_t groupinfo = cb->groupinfo[group]; |
9086 | if ((groupinfo & GI_NOT_FIXED_LENGTH) != 0) return -1; |
9087 | if ((groupinfo & GI_SET_FIXED_LENGTH) != 0) |
9088 | { |
9089 | if (isinline) *pptrptr = parsed_skip(*pptrptr, PSKIP_KET); |
9090 | return groupinfo & GI_FIXED_LENGTH_MASK; |
9091 | } |
9092 | } |
9093 | |
9094 | /* Scan the group. In this case we find the end pointer of necessity. */ |
9095 | |
9096 | for(;;) |
9097 | { |
9098 | branchlength = get_branchlength(pptrptr, errcodeptr, lcptr, recurses, cb); |
9099 | if (branchlength < 0) goto ISNOTFIXED; |
9100 | if (grouplength == -1) grouplength = branchlength; |
9101 | else if (grouplength != branchlength) goto ISNOTFIXED; |
9102 | if (**pptrptr == META_KET) break; |
9103 | *pptrptr += 1; /* Skip META_ALT */ |
9104 | } |
9105 | |
9106 | if (group > 0) |
9107 | cb->groupinfo[group] |= (uint32_t)(GI_SET_FIXED_LENGTH | grouplength); |
9108 | return grouplength; |
9109 | |
9110 | ISNOTFIXED: |
9111 | if (group > 0) cb->groupinfo[group] |= GI_NOT_FIXED_LENGTH; |
9112 | return -1; |
9113 | } |
9114 | |
9115 | |
9116 | |
9117 | /************************************************* |
9118 | * Find length of a parsed branch * |
9119 | *************************************************/ |
9120 | |
9121 | /* Return a fixed length for a branch in a lookbehind, giving an error if the |
9122 | length is not fixed. On entry, *pptrptr points to the first element inside the |
9123 | branch. On exit it is set to point to the ALT or KET. |
9124 | |
9125 | Arguments: |
9126 | pptrptr pointer to pointer in the parsed pattern |
9127 | errcodeptr pointer to error code |
9128 | lcptr pointer to loop counter |
9129 | recurses chain of recurse_check to catch mutual recursion |
9130 | cb pointer to compile block |
9131 | |
9132 | Returns: the length, or a negative value on error |
9133 | */ |
9134 | |
9135 | static int |
9136 | get_branchlength(uint32_t **pptrptr, int *errcodeptr, int *lcptr, |
9137 | parsed_recurse_check *recurses, compile_block *cb) |
9138 | { |
9139 | int branchlength = 0; |
9140 | int grouplength; |
9141 | uint32_t lastitemlength = 0; |
9142 | uint32_t *pptr = *pptrptr; |
9143 | PCRE2_SIZE offset; |
9144 | parsed_recurse_check this_recurse; |
9145 | |
9146 | /* A large and/or complex regex can take too long to process. This can happen |
9147 | more often when (?| groups are present in the pattern because their length |
9148 | cannot be cached. */ |
9149 | |
9150 | if ((*lcptr)++ > 2000) |
9151 | { |
9152 | *errcodeptr = ERR35; /* Lookbehind is too complicated */ |
9153 | return -1; |
9154 | } |
9155 | |
9156 | /* Scan the branch, accumulating the length. */ |
9157 | |
9158 | for (;; pptr++) |
9159 | { |
9160 | parsed_recurse_check *r; |
9161 | uint32_t *gptr, *gptrend; |
9162 | uint32_t escape; |
9163 | uint32_t group = 0; |
9164 | uint32_t itemlength = 0; |
9165 | |
9166 | if (*pptr < META_END) |
9167 | { |
9168 | itemlength = 1; |
9169 | } |
9170 | |
9171 | else switch (META_CODE(*pptr)) |
9172 | { |
9173 | case META_KET: |
9174 | case META_ALT: |
9175 | goto EXIT; |
9176 | |
9177 | /* (*ACCEPT) and (*FAIL) terminate the branch, but we must skip to the |
9178 | actual termination. */ |
9179 | |
9180 | case META_ACCEPT: |
9181 | case META_FAIL: |
9182 | pptr = parsed_skip(pptr, PSKIP_ALT); |
9183 | if (pptr == NULL) goto PARSED_SKIP_FAILED; |
9184 | goto EXIT; |
9185 | |
9186 | case META_MARK: |
9187 | case META_COMMIT_ARG: |
9188 | case META_PRUNE_ARG: |
9189 | case META_SKIP_ARG: |
9190 | case META_THEN_ARG: |
9191 | pptr += pptr[1] + 1; |
9192 | break; |
9193 | |
9194 | case META_CIRCUMFLEX: |
9195 | case META_COMMIT: |
9196 | case META_DOLLAR: |
9197 | case META_PRUNE: |
9198 | case META_SKIP: |
9199 | case META_THEN: |
9200 | break; |
9201 | |
9202 | case META_OPTIONS: |
9203 | pptr += 1; |
9204 | break; |
9205 | |
9206 | case META_BIGVALUE: |
9207 | itemlength = 1; |
9208 | pptr += 1; |
9209 | break; |
9210 | |
9211 | case META_CLASS: |
9212 | case META_CLASS_NOT: |
9213 | itemlength = 1; |
9214 | pptr = parsed_skip(pptr, PSKIP_CLASS); |
9215 | if (pptr == NULL) goto PARSED_SKIP_FAILED; |
9216 | break; |
9217 | |
9218 | case META_CLASS_EMPTY_NOT: |
9219 | case META_DOT: |
9220 | itemlength = 1; |
9221 | break; |
9222 | |
9223 | case META_CALLOUT_NUMBER: |
9224 | pptr += 3; |
9225 | break; |
9226 | |
9227 | case META_CALLOUT_STRING: |
9228 | pptr += 3 + SIZEOFFSET; |
9229 | break; |
9230 | |
9231 | /* Only some escapes consume a character. Of those, \R and \X are never |
9232 | allowed because they might match more than character. \C is allowed only in |
9233 | 32-bit and non-UTF 8/16-bit modes. */ |
9234 | |
9235 | case META_ESCAPE: |
9236 | escape = META_DATA(*pptr); |
9237 | if (escape == ESC_R || escape == ESC_X) return -1; |
9238 | if (escape > ESC_b && escape < ESC_Z) |
9239 | { |
9240 | #if PCRE2_CODE_UNIT_WIDTH != 32 |
9241 | if ((cb->external_options & PCRE2_UTF) != 0 && escape == ESC_C) |
9242 | { |
9243 | *errcodeptr = ERR36; |
9244 | return -1; |
9245 | } |
9246 | #endif |
9247 | itemlength = 1; |
9248 | if (escape == ESC_p || escape == ESC_P) pptr++; /* Skip prop data */ |
9249 | } |
9250 | break; |
9251 | |
9252 | /* Lookaheads do not contribute to the length of this branch, but they may |
9253 | contain lookbehinds within them whose lengths need to be set. */ |
9254 | |
9255 | case META_LOOKAHEAD: |
9256 | case META_LOOKAHEADNOT: |
9257 | case META_LOOKAHEAD_NA: |
9258 | *errcodeptr = check_lookbehinds(pptr + 1, &pptr, recurses, cb, lcptr); |
9259 | if (*errcodeptr != 0) return -1; |
9260 | |
9261 | /* Ignore any qualifiers that follow a lookahead assertion. */ |
9262 | |
9263 | switch (pptr[1]) |
9264 | { |
9265 | case META_ASTERISK: |
9266 | case META_ASTERISK_PLUS: |
9267 | case META_ASTERISK_QUERY: |
9268 | case META_PLUS: |
9269 | case META_PLUS_PLUS: |
9270 | case META_PLUS_QUERY: |
9271 | case META_QUERY: |
9272 | case META_QUERY_PLUS: |
9273 | case META_QUERY_QUERY: |
9274 | pptr++; |
9275 | break; |
9276 | |
9277 | case META_MINMAX: |
9278 | case META_MINMAX_PLUS: |
9279 | case META_MINMAX_QUERY: |
9280 | pptr += 3; |
9281 | break; |
9282 | |
9283 | default: |
9284 | break; |
9285 | } |
9286 | break; |
9287 | |
9288 | /* A nested lookbehind does not contribute any length to this lookbehind, |
9289 | but must itself be checked and have its lengths set. */ |
9290 | |
9291 | case META_LOOKBEHIND: |
9292 | case META_LOOKBEHINDNOT: |
9293 | case META_LOOKBEHIND_NA: |
9294 | if (!set_lookbehind_lengths(&pptr, errcodeptr, lcptr, recurses, cb)) |
9295 | return -1; |
9296 | break; |
9297 | |
9298 | /* Back references and recursions are handled by very similar code. At this |
9299 | stage, the names generated in the parsing pass are available, but the main |
9300 | name table has not yet been created. So for the named varieties, scan the |
9301 | list of names in order to get the number of the first one in the pattern, |
9302 | and whether or not this name is duplicated. */ |
9303 | |
9304 | case META_BACKREF_BYNAME: |
9305 | if ((cb->external_options & PCRE2_MATCH_UNSET_BACKREF) != 0) |
9306 | goto ISNOTFIXED; |
9307 | /* Fall through */ |
9308 | |
9309 | case META_RECURSE_BYNAME: |
9310 | { |
9311 | int i; |
9312 | PCRE2_SPTR name; |
9313 | BOOL is_dupname = FALSE; |
9314 | named_group *ng = cb->named_groups; |
9315 | uint32_t meta_code = META_CODE(*pptr); |
9316 | uint32_t length = *(++pptr); |
9317 | |
9318 | GETPLUSOFFSET(offset, pptr); |
9319 | name = cb->start_pattern + offset; |
9320 | for (i = 0; i < cb->names_found; i++, ng++) |
9321 | { |
9322 | if (length == ng->length && PRIV(strncmp)(name, ng->name, length) == 0) |
9323 | { |
9324 | group = ng->number; |
9325 | is_dupname = ng->isdup; |
9326 | break; |
9327 | } |
9328 | } |
9329 | |
9330 | if (group == 0) |
9331 | { |
9332 | *errcodeptr = ERR15; /* Non-existent subpattern */ |
9333 | cb->erroroffset = offset; |
9334 | return -1; |
9335 | } |
9336 | |
9337 | /* A numerical back reference can be fixed length if duplicate capturing |
9338 | groups are not being used. A non-duplicate named back reference can also |
9339 | be handled. */ |
9340 | |
9341 | if (meta_code == META_RECURSE_BYNAME || |
9342 | (!is_dupname && (cb->external_flags & PCRE2_DUPCAPUSED) == 0)) |
9343 | goto RECURSE_OR_BACKREF_LENGTH; /* Handle as a numbered version. */ |
9344 | } |
9345 | goto ISNOTFIXED; /* Duplicate name or number */ |
9346 | |
9347 | /* The offset values for back references < 10 are in a separate vector |
9348 | because otherwise they would use more than two parsed pattern elements on |
9349 | 64-bit systems. */ |
9350 | |
9351 | case META_BACKREF: |
9352 | if ((cb->external_options & PCRE2_MATCH_UNSET_BACKREF) != 0 || |
9353 | (cb->external_flags & PCRE2_DUPCAPUSED) != 0) |
9354 | goto ISNOTFIXED; |
9355 | group = META_DATA(*pptr); |
9356 | if (group < 10) |
9357 | { |
9358 | offset = cb->small_ref_offset[group]; |
9359 | goto RECURSE_OR_BACKREF_LENGTH; |
9360 | } |
9361 | |
9362 | /* Fall through */ |
9363 | /* For groups >= 10 - picking up group twice does no harm. */ |
9364 | |
9365 | /* A true recursion implies not fixed length, but a subroutine call may |
9366 | be OK. Back reference "recursions" are also failed. */ |
9367 | |
9368 | case META_RECURSE: |
9369 | group = META_DATA(*pptr); |
9370 | GETPLUSOFFSET(offset, pptr); |
9371 | |
9372 | RECURSE_OR_BACKREF_LENGTH: |
9373 | if (group > cb->bracount) |
9374 | { |
9375 | cb->erroroffset = offset; |
9376 | *errcodeptr = ERR15; /* Non-existent subpattern */ |
9377 | return -1; |
9378 | } |
9379 | if (group == 0) goto ISNOTFIXED; /* Local recursion */ |
9380 | for (gptr = cb->parsed_pattern; *gptr != META_END; gptr++) |
9381 | { |
9382 | if (META_CODE(*gptr) == META_BIGVALUE) gptr++; |
9383 | else if (*gptr == (META_CAPTURE | group)) break; |
9384 | } |
9385 | |
9386 | /* We must start the search for the end of the group at the first meta code |
9387 | inside the group. Otherwise it will be treated as an enclosed group. */ |
9388 | |
9389 | gptrend = parsed_skip(gptr + 1, PSKIP_KET); |
9390 | if (gptrend == NULL) goto PARSED_SKIP_FAILED; |
9391 | if (pptr > gptr && pptr < gptrend) goto ISNOTFIXED; /* Local recursion */ |
9392 | for (r = recurses; r != NULL; r = r->prev) if (r->groupptr == gptr) break; |
9393 | if (r != NULL) goto ISNOTFIXED; /* Mutual recursion */ |
9394 | this_recurse.prev = recurses; |
9395 | this_recurse.groupptr = gptr; |
9396 | |
9397 | /* We do not need to know the position of the end of the group, that is, |
9398 | gptr is not used after the call to get_grouplength(). Setting the second |
9399 | argument FALSE stops it scanning for the end when the length can be found |
9400 | in the cache. */ |
9401 | |
9402 | gptr++; |
9403 | grouplength = get_grouplength(&gptr, FALSE, errcodeptr, lcptr, group, |
9404 | &this_recurse, cb); |
9405 | if (grouplength < 0) |
9406 | { |
9407 | if (*errcodeptr == 0) goto ISNOTFIXED; |
9408 | return -1; /* Error already set */ |
9409 | } |
9410 | itemlength = grouplength; |
9411 | break; |
9412 | |
9413 | /* A (DEFINE) group is never obeyed inline and so it does not contribute to |
9414 | the length of this branch. Skip from the following item to the next |
9415 | unpaired ket. */ |
9416 | |
9417 | case META_COND_DEFINE: |
9418 | pptr = parsed_skip(pptr + 1, PSKIP_KET); |
9419 | break; |
9420 | |
9421 | /* Check other nested groups - advance past the initial data for each type |
9422 | and then seek a fixed length with get_grouplength(). */ |
9423 | |
9424 | case META_COND_NAME: |
9425 | case META_COND_NUMBER: |
9426 | case META_COND_RNAME: |
9427 | case META_COND_RNUMBER: |
9428 | pptr += 2 + SIZEOFFSET; |
9429 | goto CHECK_GROUP; |
9430 | |
9431 | case META_COND_ASSERT: |
9432 | pptr += 1; |
9433 | goto CHECK_GROUP; |
9434 | |
9435 | case META_COND_VERSION: |
9436 | pptr += 4; |
9437 | goto CHECK_GROUP; |
9438 | |
9439 | case META_CAPTURE: |
9440 | group = META_DATA(*pptr); |
9441 | /* Fall through */ |
9442 | |
9443 | case META_ATOMIC: |
9444 | case META_NOCAPTURE: |
9445 | case META_SCRIPT_RUN: |
9446 | pptr++; |
9447 | CHECK_GROUP: |
9448 | grouplength = get_grouplength(&pptr, TRUE, errcodeptr, lcptr, group, |
9449 | recurses, cb); |
9450 | if (grouplength < 0) return -1; |
9451 | itemlength = grouplength; |
9452 | break; |
9453 | |
9454 | /* Exact repetition is OK; variable repetition is not. A repetition of zero |
9455 | must subtract the length that has already been added. */ |
9456 | |
9457 | case META_MINMAX: |
9458 | case META_MINMAX_PLUS: |
9459 | case META_MINMAX_QUERY: |
9460 | if (pptr[1] == pptr[2]) |
9461 | { |
9462 | switch(pptr[1]) |
9463 | { |
9464 | case 0: |
9465 | branchlength -= lastitemlength; |
9466 | break; |
9467 | |
9468 | case 1: |
9469 | itemlength = 0; |
9470 | break; |
9471 | |
9472 | default: /* Check for integer overflow */ |
9473 | if (lastitemlength != 0 && /* Should not occur, but just in case */ |
9474 | INT_MAX/lastitemlength < pptr[1] - 1) |
9475 | { |
9476 | *errcodeptr = ERR87; /* Integer overflow; lookbehind too big */ |
9477 | return -1; |
9478 | } |
9479 | itemlength = (pptr[1] - 1) * lastitemlength; |
9480 | break; |
9481 | } |
9482 | pptr += 2; |
9483 | break; |
9484 | } |
9485 | /* Fall through */ |
9486 | |
9487 | /* Any other item means this branch does not have a fixed length. */ |
9488 | |
9489 | default: |
9490 | ISNOTFIXED: |
9491 | *errcodeptr = ERR25; /* Not fixed length */ |
9492 | return -1; |
9493 | } |
9494 | |
9495 | /* Add the item length to the branchlength, checking for integer overflow and |
9496 | for the branch length exceeding the limit. */ |
9497 | |
9498 | if (INT_MAX - branchlength < (int)itemlength || |
9499 | (branchlength += itemlength) > LOOKBEHIND_MAX) |
9500 | { |
9501 | *errcodeptr = ERR87; |
9502 | return -1; |
9503 | } |
9504 | |
9505 | /* Save this item length for use if the next item is a quantifier. */ |
9506 | |
9507 | lastitemlength = itemlength; |
9508 | } |
9509 | |
9510 | EXIT: |
9511 | *pptrptr = pptr; |
9512 | return branchlength; |
9513 | |
9514 | PARSED_SKIP_FAILED: |
9515 | *errcodeptr = ERR90; |
9516 | return -1; |
9517 | } |
9518 | |
9519 | |
9520 | |
9521 | /************************************************* |
9522 | * Set lengths in a lookbehind * |
9523 | *************************************************/ |
9524 | |
9525 | /* This function is called for each lookbehind, to set the lengths in its |
9526 | branches. An error occurs if any branch does not have a fixed length that is |
9527 | less than the maximum (65535). On exit, the pointer must be left on the final |
9528 | ket. |
9529 | |
9530 | The function also maintains the max_lookbehind value. Any lookbehind branch |
9531 | that contains a nested lookbehind may actually look further back than the |
9532 | length of the branch. The additional amount is passed back from |
9533 | get_branchlength() as an "extra" value. |
9534 | |
9535 | Arguments: |
9536 | pptrptr pointer to pointer in the parsed pattern |
9537 | errcodeptr pointer to error code |
9538 | lcptr pointer to loop counter |
9539 | recurses chain of recurse_check to catch mutual recursion |
9540 | cb pointer to compile block |
9541 | |
9542 | Returns: TRUE if all is well |
9543 | FALSE otherwise, with error code and offset set |
9544 | */ |
9545 | |
9546 | static BOOL |
9547 | set_lookbehind_lengths(uint32_t **pptrptr, int *errcodeptr, int *lcptr, |
9548 | parsed_recurse_check *recurses, compile_block *cb) |
9549 | { |
9550 | PCRE2_SIZE offset; |
9551 | int branchlength; |
9552 | uint32_t *bptr = *pptrptr; |
9553 | |
9554 | READPLUSOFFSET(offset, bptr); /* Offset for error messages */ |
9555 | *pptrptr += SIZEOFFSET; |
9556 | |
9557 | do |
9558 | { |
9559 | *pptrptr += 1; |
9560 | branchlength = get_branchlength(pptrptr, errcodeptr, lcptr, recurses, cb); |
9561 | if (branchlength < 0) |
9562 | { |
9563 | /* The errorcode and offset may already be set from a nested lookbehind. */ |
9564 | if (*errcodeptr == 0) *errcodeptr = ERR25; |
9565 | if (cb->erroroffset == PCRE2_UNSET) cb->erroroffset = offset; |
9566 | return FALSE; |
9567 | } |
9568 | if (branchlength > cb->max_lookbehind) cb->max_lookbehind = branchlength; |
9569 | *bptr |= branchlength; /* branchlength never more than 65535 */ |
9570 | bptr = *pptrptr; |
9571 | } |
9572 | while (*bptr == META_ALT); |
9573 | |
9574 | return TRUE; |
9575 | } |
9576 | |
9577 | |
9578 | |
9579 | /************************************************* |
9580 | * Check parsed pattern lookbehinds * |
9581 | *************************************************/ |
9582 | |
9583 | /* This function is called at the end of parsing a pattern if any lookbehinds |
9584 | were encountered. It scans the parsed pattern for them, calling |
9585 | set_lookbehind_lengths() for each one. At the start, the errorcode is zero and |
9586 | the error offset is marked unset. The enables the functions above not to |
9587 | override settings from deeper nestings. |
9588 | |
9589 | This function is called recursively from get_branchlength() for lookaheads in |
9590 | order to process any lookbehinds that they may contain. It stops when it hits a |
9591 | non-nested closing parenthesis in this case, returning a pointer to it. |
9592 | |
9593 | Arguments |
9594 | pptr points to where to start (start of pattern or start of lookahead) |
9595 | retptr if not NULL, return the ket pointer here |
9596 | recurses chain of recurse_check to catch mutual recursion |
9597 | cb points to the compile block |
9598 | lcptr points to loop counter |
9599 | |
9600 | Returns: 0 on success, or an errorcode (cb->erroroffset will be set) |
9601 | */ |
9602 | |
9603 | static int |
9604 | check_lookbehinds(uint32_t *pptr, uint32_t **retptr, |
9605 | parsed_recurse_check *recurses, compile_block *cb, int *lcptr) |
9606 | { |
9607 | int errorcode = 0; |
9608 | int nestlevel = 0; |
9609 | |
9610 | cb->erroroffset = PCRE2_UNSET; |
9611 | |
9612 | for (; *pptr != META_END; pptr++) |
9613 | { |
9614 | if (*pptr < META_END) continue; /* Literal */ |
9615 | |
9616 | switch (META_CODE(*pptr)) |
9617 | { |
9618 | default: |
9619 | return ERR70; /* Unrecognized meta code */ |
9620 | |
9621 | case META_ESCAPE: |
9622 | if (*pptr - META_ESCAPE == ESC_P || *pptr - META_ESCAPE == ESC_p) |
9623 | pptr += 1; |
9624 | break; |
9625 | |
9626 | case META_KET: |
9627 | if (--nestlevel < 0) |
9628 | { |
9629 | if (retptr != NULL) *retptr = pptr; |
9630 | return 0; |
9631 | } |
9632 | break; |
9633 | |
9634 | case META_ATOMIC: |
9635 | case META_CAPTURE: |
9636 | case META_COND_ASSERT: |
9637 | case META_LOOKAHEAD: |
9638 | case META_LOOKAHEADNOT: |
9639 | case META_LOOKAHEAD_NA: |
9640 | case META_NOCAPTURE: |
9641 | case META_SCRIPT_RUN: |
9642 | nestlevel++; |
9643 | break; |
9644 | |
9645 | case META_ACCEPT: |
9646 | case META_ALT: |
9647 | case META_ASTERISK: |
9648 | case META_ASTERISK_PLUS: |
9649 | case META_ASTERISK_QUERY: |
9650 | case META_BACKREF: |
9651 | case META_CIRCUMFLEX: |
9652 | case META_CLASS: |
9653 | case META_CLASS_EMPTY: |
9654 | case META_CLASS_EMPTY_NOT: |
9655 | case META_CLASS_END: |
9656 | case META_CLASS_NOT: |
9657 | case META_COMMIT: |
9658 | case META_DOLLAR: |
9659 | case META_DOT: |
9660 | case META_FAIL: |
9661 | case META_PLUS: |
9662 | case META_PLUS_PLUS: |
9663 | case META_PLUS_QUERY: |
9664 | case META_PRUNE: |
9665 | case META_QUERY: |
9666 | case META_QUERY_PLUS: |
9667 | case META_QUERY_QUERY: |
9668 | case META_RANGE_ESCAPED: |
9669 | case META_RANGE_LITERAL: |
9670 | case META_SKIP: |
9671 | case META_THEN: |
9672 | break; |
9673 | |
9674 | case META_RECURSE: |
9675 | pptr += SIZEOFFSET; |
9676 | break; |
9677 | |
9678 | case META_BACKREF_BYNAME: |
9679 | case META_RECURSE_BYNAME: |
9680 | pptr += 1 + SIZEOFFSET; |
9681 | break; |
9682 | |
9683 | case META_COND_DEFINE: |
9684 | pptr += SIZEOFFSET; |
9685 | nestlevel++; |
9686 | break; |
9687 | |
9688 | case META_COND_NAME: |
9689 | case META_COND_NUMBER: |
9690 | case META_COND_RNAME: |
9691 | case META_COND_RNUMBER: |
9692 | pptr += 1 + SIZEOFFSET; |
9693 | nestlevel++; |
9694 | break; |
9695 | |
9696 | case META_COND_VERSION: |
9697 | pptr += 3; |
9698 | nestlevel++; |
9699 | break; |
9700 | |
9701 | case META_CALLOUT_STRING: |
9702 | pptr += 3 + SIZEOFFSET; |
9703 | break; |
9704 | |
9705 | case META_BIGVALUE: |
9706 | case META_OPTIONS: |
9707 | case META_POSIX: |
9708 | case META_POSIX_NEG: |
9709 | pptr += 1; |
9710 | break; |
9711 | |
9712 | case META_MINMAX: |
9713 | case META_MINMAX_QUERY: |
9714 | case META_MINMAX_PLUS: |
9715 | pptr += 2; |
9716 | break; |
9717 | |
9718 | case META_CALLOUT_NUMBER: |
9719 | pptr += 3; |
9720 | break; |
9721 | |
9722 | case META_MARK: |
9723 | case META_COMMIT_ARG: |
9724 | case META_PRUNE_ARG: |
9725 | case META_SKIP_ARG: |
9726 | case META_THEN_ARG: |
9727 | pptr += 1 + pptr[1]; |
9728 | break; |
9729 | |
9730 | case META_LOOKBEHIND: |
9731 | case META_LOOKBEHINDNOT: |
9732 | case META_LOOKBEHIND_NA: |
9733 | if (!set_lookbehind_lengths(&pptr, &errorcode, lcptr, recurses, cb)) |
9734 | return errorcode; |
9735 | break; |
9736 | } |
9737 | } |
9738 | |
9739 | return 0; |
9740 | } |
9741 | |
9742 | |
9743 | |
9744 | /************************************************* |
9745 | * External function to compile a pattern * |
9746 | *************************************************/ |
9747 | |
9748 | /* This function reads a regular expression in the form of a string and returns |
9749 | a pointer to a block of store holding a compiled version of the expression. |
9750 | |
9751 | Arguments: |
9752 | pattern the regular expression |
9753 | patlen the length of the pattern, or PCRE2_ZERO_TERMINATED |
9754 | options option bits |
9755 | errorptr pointer to errorcode |
9756 | erroroffset pointer to error offset |
9757 | ccontext points to a compile context or is NULL |
9758 | |
9759 | Returns: pointer to compiled data block, or NULL on error, |
9760 | with errorcode and erroroffset set |
9761 | */ |
9762 | |
9763 | PCRE2_EXP_DEFN pcre2_code * PCRE2_CALL_CONVENTION |
9764 | pcre2_compile(PCRE2_SPTR pattern, PCRE2_SIZE patlen, uint32_t options, |
9765 | int *errorptr, PCRE2_SIZE *erroroffset, pcre2_compile_context *ccontext) |
9766 | { |
9767 | BOOL utf; /* Set TRUE for UTF mode */ |
9768 | BOOL ucp; /* Set TRUE for UCP mode */ |
9769 | BOOL has_lookbehind = FALSE; /* Set TRUE if a lookbehind is found */ |
9770 | BOOL zero_terminated; /* Set TRUE for zero-terminated pattern */ |
9771 | pcre2_real_code *re = NULL; /* What we will return */ |
9772 | compile_block cb; /* "Static" compile-time data */ |
9773 | const uint8_t *tables; /* Char tables base pointer */ |
9774 | |
9775 | PCRE2_UCHAR *code; /* Current pointer in compiled code */ |
9776 | PCRE2_SPTR codestart; /* Start of compiled code */ |
9777 | PCRE2_SPTR ptr; /* Current pointer in pattern */ |
9778 | uint32_t *pptr; /* Current pointer in parsed pattern */ |
9779 | |
9780 | PCRE2_SIZE length = 1; /* Allow for final END opcode */ |
9781 | PCRE2_SIZE usedlength; /* Actual length used */ |
9782 | PCRE2_SIZE re_blocksize; /* Size of memory block */ |
9783 | PCRE2_SIZE big32count = 0; /* 32-bit literals >= 0x80000000 */ |
9784 | PCRE2_SIZE parsed_size_needed; /* Needed for parsed pattern */ |
9785 | |
9786 | uint32_t firstcuflags, reqcuflags; /* Type of first/req code unit */ |
9787 | uint32_t firstcu, reqcu; /* Value of first/req code unit */ |
9788 | uint32_t setflags = 0; /* NL and BSR set flags */ |
9789 | |
9790 | uint32_t skipatstart; /* When checking (*UTF) etc */ |
9791 | uint32_t limit_heap = UINT32_MAX; |
9792 | uint32_t limit_match = UINT32_MAX; /* Unset match limits */ |
9793 | uint32_t limit_depth = UINT32_MAX; |
9794 | |
9795 | int newline = 0; /* Unset; can be set by the pattern */ |
9796 | int bsr = 0; /* Unset; can be set by the pattern */ |
9797 | int errorcode = 0; /* Initialize to avoid compiler warn */ |
9798 | int regexrc; /* Return from compile */ |
9799 | |
9800 | uint32_t i; /* Local loop counter */ |
9801 | |
9802 | /* Comments at the head of this file explain about these variables. */ |
9803 | |
9804 | uint32_t stack_groupinfo[GROUPINFO_DEFAULT_SIZE]; |
9805 | uint32_t stack_parsed_pattern[PARSED_PATTERN_DEFAULT_SIZE]; |
9806 | named_group named_groups[NAMED_GROUP_LIST_SIZE]; |
9807 | |
9808 | /* The workspace is used in different ways in the different compiling phases. |
9809 | It needs to be 16-bit aligned for the preliminary parsing scan. */ |
9810 | |
9811 | uint32_t c16workspace[C16_WORK_SIZE]; |
9812 | PCRE2_UCHAR *cworkspace = (PCRE2_UCHAR *)c16workspace; |
9813 | |
9814 | |
9815 | /* -------------- Check arguments and set up the pattern ----------------- */ |
9816 | |
9817 | /* There must be error code and offset pointers. */ |
9818 | |
9819 | if (errorptr == NULL || erroroffset == NULL) return NULL; |
9820 | *errorptr = ERR0; |
9821 | *erroroffset = 0; |
9822 | |
9823 | /* There must be a pattern! */ |
9824 | |
9825 | if (pattern == NULL) |
9826 | { |
9827 | *errorptr = ERR16; |
9828 | return NULL; |
9829 | } |
9830 | |
9831 | /* A NULL compile context means "use a default context" */ |
9832 | |
9833 | if (ccontext == NULL) |
9834 | ccontext = (pcre2_compile_context *)(&PRIV(default_compile_context)); |
9835 | |
9836 | /* PCRE2_MATCH_INVALID_UTF implies UTF */ |
9837 | |
9838 | if ((options & PCRE2_MATCH_INVALID_UTF) != 0) options |= PCRE2_UTF; |
9839 | |
9840 | /* Check that all undefined public option bits are zero. */ |
9841 | |
9842 | if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0 || |
9843 | (ccontext->extra_options & ~PUBLIC_COMPILE_EXTRA_OPTIONS) != 0) |
9844 | { |
9845 | *errorptr = ERR17; |
9846 | return NULL; |
9847 | } |
9848 | |
9849 | if ((options & PCRE2_LITERAL) != 0 && |
9850 | ((options & ~PUBLIC_LITERAL_COMPILE_OPTIONS) != 0 || |
9851 | (ccontext->extra_options & ~PUBLIC_LITERAL_COMPILE_EXTRA_OPTIONS) != 0)) |
9852 | { |
9853 | *errorptr = ERR92; |
9854 | return NULL; |
9855 | } |
9856 | |
9857 | /* A zero-terminated pattern is indicated by the special length value |
9858 | PCRE2_ZERO_TERMINATED. Check for an overlong pattern. */ |
9859 | |
9860 | if ((zero_terminated = (patlen == PCRE2_ZERO_TERMINATED))) |
9861 | patlen = PRIV(strlen)(pattern); |
9862 | |
9863 | if (patlen > ccontext->max_pattern_length) |
9864 | { |
9865 | *errorptr = ERR88; |
9866 | return NULL; |
9867 | } |
9868 | |
9869 | /* From here on, all returns from this function should end up going via the |
9870 | EXIT label. */ |
9871 | |
9872 | |
9873 | /* ------------ Initialize the "static" compile data -------------- */ |
9874 | |
9875 | tables = (ccontext->tables != NULL)? ccontext->tables : PRIV(default_tables); |
9876 | |
9877 | cb.lcc = tables + lcc_offset; /* Individual */ |
9878 | cb.fcc = tables + fcc_offset; /* character */ |
9879 | cb.cbits = tables + cbits_offset; /* tables */ |
9880 | cb.ctypes = tables + ctypes_offset; |
9881 | |
9882 | cb.assert_depth = 0; |
9883 | cb.bracount = 0; |
9884 | cb.cx = ccontext; |
9885 | cb.dupnames = FALSE; |
9886 | cb.end_pattern = pattern + patlen; |
9887 | cb.erroroffset = 0; |
9888 | cb.external_flags = 0; |
9889 | cb.external_options = options; |
9890 | cb.groupinfo = stack_groupinfo; |
9891 | cb.had_recurse = FALSE; |
9892 | cb.lastcapture = 0; |
9893 | cb.max_lookbehind = 0; |
9894 | cb.name_entry_size = 0; |
9895 | cb.name_table = NULL; |
9896 | cb.named_groups = named_groups; |
9897 | cb.named_group_list_size = NAMED_GROUP_LIST_SIZE; |
9898 | cb.names_found = 0; |
9899 | cb.open_caps = NULL; |
9900 | cb.parens_depth = 0; |
9901 | cb.parsed_pattern = stack_parsed_pattern; |
9902 | cb.req_varyopt = 0; |
9903 | cb.start_code = cworkspace; |
9904 | cb.start_pattern = pattern; |
9905 | cb.start_workspace = cworkspace; |
9906 | cb.workspace_size = COMPILE_WORK_SIZE; |
9907 | |
9908 | /* Maximum back reference and backref bitmap. The bitmap records up to 31 back |
9909 | references to help in deciding whether (.*) can be treated as anchored or not. |
9910 | */ |
9911 | |
9912 | cb.top_backref = 0; |
9913 | cb.backref_map = 0; |
9914 | |
9915 | /* Escape sequences \1 to \9 are always back references, but as they are only |
9916 | two characters long, only two elements can be used in the parsed_pattern |
9917 | vector. The first contains the reference, and we'd like to use the second to |
9918 | record the offset in the pattern, so that forward references to non-existent |
9919 | groups can be diagnosed later with an offset. However, on 64-bit systems, |
9920 | PCRE2_SIZE won't fit. Instead, we have a vector of offsets for the first |
9921 | occurrence of \1 to \9, indexed by the second parsed_pattern value. All other |
9922 | references have enough space for the offset to be put into the parsed pattern. |
9923 | */ |
9924 | |
9925 | for (i = 0; i < 10; i++) cb.small_ref_offset[i] = PCRE2_UNSET; |
9926 | |
9927 | |
9928 | /* --------------- Start looking at the pattern --------------- */ |
9929 | |
9930 | /* Unless PCRE2_LITERAL is set, check for global one-time option settings at |
9931 | the start of the pattern, and remember the offset to the actual regex. With |
9932 | valgrind support, make the terminator of a zero-terminated pattern |
9933 | inaccessible. This catches bugs that would otherwise only show up for |
9934 | non-zero-terminated patterns. */ |
9935 | |
9936 | #ifdef SUPPORT_VALGRIND |
9937 | if (zero_terminated) VALGRIND_MAKE_MEM_NOACCESS(pattern + patlen, CU2BYTES(1)); |
9938 | #endif |
9939 | |
9940 | ptr = pattern; |
9941 | skipatstart = 0; |
9942 | |
9943 | if ((options & PCRE2_LITERAL) == 0) |
9944 | { |
9945 | while (patlen - skipatstart >= 2 && |
9946 | ptr[skipatstart] == CHAR_LEFT_PARENTHESIS && |
9947 | ptr[skipatstart+1] == CHAR_ASTERISK) |
9948 | { |
9949 | for (i = 0; i < sizeof(pso_list)/sizeof(pso); i++) |
9950 | { |
9951 | uint32_t c, pp; |
9952 | pso *p = pso_list + i; |
9953 | |
9954 | if (patlen - skipatstart - 2 >= p->length && |
9955 | PRIV(strncmp_c8)(ptr + skipatstart + 2, (char *)(p->name), |
9956 | p->length) == 0) |
9957 | { |
9958 | skipatstart += p->length + 2; |
9959 | switch(p->type) |
9960 | { |
9961 | case PSO_OPT: |
9962 | cb.external_options |= p->value; |
9963 | break; |
9964 | |
9965 | case PSO_FLG: |
9966 | setflags |= p->value; |
9967 | break; |
9968 | |
9969 | case PSO_NL: |
9970 | newline = p->value; |
9971 | setflags |= PCRE2_NL_SET; |
9972 | break; |
9973 | |
9974 | case PSO_BSR: |
9975 | bsr = p->value; |
9976 | setflags |= PCRE2_BSR_SET; |
9977 | break; |
9978 | |
9979 | case PSO_LIMM: |
9980 | case PSO_LIMD: |
9981 | case PSO_LIMH: |
9982 | c = 0; |
9983 | pp = skipatstart; |
9984 | if (!IS_DIGIT(ptr[pp])) |
9985 | { |
9986 | errorcode = ERR60; |
9987 | ptr += pp; |
9988 | goto HAD_EARLY_ERROR; |
9989 | } |
9990 | while (IS_DIGIT(ptr[pp])) |
9991 | { |
9992 | if (c > UINT32_MAX / 10 - 1) break; /* Integer overflow */ |
9993 | c = c*10 + (ptr[pp++] - CHAR_0); |
9994 | } |
9995 | if (ptr[pp++] != CHAR_RIGHT_PARENTHESIS) |
9996 | { |
9997 | errorcode = ERR60; |
9998 | ptr += pp; |
9999 | goto HAD_EARLY_ERROR; |
10000 | } |
10001 | if (p->type == PSO_LIMH) limit_heap = c; |
10002 | else if (p->type == PSO_LIMM) limit_match = c; |
10003 | else limit_depth = c; |
10004 | skipatstart += pp - skipatstart; |
10005 | break; |
10006 | } |
10007 | break; /* Out of the table scan loop */ |
10008 | } |
10009 | } |
10010 | if (i >= sizeof(pso_list)/sizeof(pso)) break; /* Out of pso loop */ |
10011 | } |
10012 | } |
10013 | |
10014 | /* End of pattern-start options; advance to start of real regex. */ |
10015 | |
10016 | ptr += skipatstart; |
10017 | |
10018 | /* Can't support UTF or UCP if PCRE2 was built without Unicode support. */ |
10019 | |
10020 | #ifndef SUPPORT_UNICODE |
10021 | if ((cb.external_options & (PCRE2_UTF|PCRE2_UCP)) != 0) |
10022 | { |
10023 | errorcode = ERR32; |
10024 | goto HAD_EARLY_ERROR; |
10025 | } |
10026 | #endif |
10027 | |
10028 | /* Check UTF. We have the original options in 'options', with that value as |
10029 | modified by (*UTF) etc in cb->external_options. The extra option |
10030 | PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES is not permitted in UTF-16 mode because the |
10031 | surrogate code points cannot be represented in UTF-16. */ |
10032 | |
10033 | utf = (cb.external_options & PCRE2_UTF) != 0; |
10034 | if (utf) |
10035 | { |
10036 | if ((options & PCRE2_NEVER_UTF) != 0) |
10037 | { |
10038 | errorcode = ERR74; |
10039 | goto HAD_EARLY_ERROR; |
10040 | } |
10041 | if ((options & PCRE2_NO_UTF_CHECK) == 0 && |
10042 | (errorcode = PRIV(valid_utf)(pattern, patlen, erroroffset)) != 0) |
10043 | goto HAD_ERROR; /* Offset was set by valid_utf() */ |
10044 | |
10045 | #if PCRE2_CODE_UNIT_WIDTH == 16 |
10046 | if ((ccontext->extra_options & PCRE2_EXTRA_ALLOW_SURROGATE_ESCAPES) != 0) |
10047 | { |
10048 | errorcode = ERR91; |
10049 | goto HAD_EARLY_ERROR; |
10050 | } |
10051 | #endif |
10052 | } |
10053 | |
10054 | /* Check UCP lockout. */ |
10055 | |
10056 | ucp = (cb.external_options & PCRE2_UCP) != 0; |
10057 | if (ucp && (cb.external_options & PCRE2_NEVER_UCP) != 0) |
10058 | { |
10059 | errorcode = ERR75; |
10060 | goto HAD_EARLY_ERROR; |
10061 | } |
10062 | |
10063 | /* Process the BSR setting. */ |
10064 | |
10065 | if (bsr == 0) bsr = ccontext->bsr_convention; |
10066 | |
10067 | /* Process the newline setting. */ |
10068 | |
10069 | if (newline == 0) newline = ccontext->newline_convention; |
10070 | cb.nltype = NLTYPE_FIXED; |
10071 | switch(newline) |
10072 | { |
10073 | case PCRE2_NEWLINE_CR: |
10074 | cb.nllen = 1; |
10075 | cb.nl[0] = CHAR_CR; |
10076 | break; |
10077 | |
10078 | case PCRE2_NEWLINE_LF: |
10079 | cb.nllen = 1; |
10080 | cb.nl[0] = CHAR_NL; |
10081 | break; |
10082 | |
10083 | case PCRE2_NEWLINE_NUL: |
10084 | cb.nllen = 1; |
10085 | cb.nl[0] = CHAR_NUL; |
10086 | break; |
10087 | |
10088 | case PCRE2_NEWLINE_CRLF: |
10089 | cb.nllen = 2; |
10090 | cb.nl[0] = CHAR_CR; |
10091 | cb.nl[1] = CHAR_NL; |
10092 | break; |
10093 | |
10094 | case PCRE2_NEWLINE_ANY: |
10095 | cb.nltype = NLTYPE_ANY; |
10096 | break; |
10097 | |
10098 | case PCRE2_NEWLINE_ANYCRLF: |
10099 | cb.nltype = NLTYPE_ANYCRLF; |
10100 | break; |
10101 | |
10102 | default: |
10103 | errorcode = ERR56; |
10104 | goto HAD_EARLY_ERROR; |
10105 | } |
10106 | |
10107 | /* Pre-scan the pattern to do two things: (1) Discover the named groups and |
10108 | their numerical equivalents, so that this information is always available for |
10109 | the remaining processing. (2) At the same time, parse the pattern and put a |
10110 | processed version into the parsed_pattern vector. This has escapes interpreted |
10111 | and comments removed (amongst other things). |
10112 | |
10113 | In all but one case, when PCRE2_AUTO_CALLOUT is not set, the number of unsigned |
10114 | 32-bit ints in the parsed pattern is bounded by the length of the pattern plus |
10115 | one (for the terminator) plus four if PCRE2_EXTRA_WORD or PCRE2_EXTRA_LINE is |
10116 | set. The exceptional case is when running in 32-bit, non-UTF mode, when literal |
10117 | characters greater than META_END (0x80000000) have to be coded as two units. In |
10118 | this case, therefore, we scan the pattern to check for such values. */ |
10119 | |
10120 | #if PCRE2_CODE_UNIT_WIDTH == 32 |
10121 | if (!utf) |
10122 | { |
10123 | PCRE2_SPTR p; |
10124 | for (p = ptr; p < cb.end_pattern; p++) if (*p >= META_END) big32count++; |
10125 | } |
10126 | #endif |
10127 | |
10128 | /* Ensure that the parsed pattern buffer is big enough. When PCRE2_AUTO_CALLOUT |
10129 | is set we have to assume a numerical callout (4 elements) for each character |
10130 | plus one at the end. This is overkill, but memory is plentiful these days. For |
10131 | many smaller patterns the vector on the stack (which was set up above) can be |
10132 | used. */ |
10133 | |
10134 | parsed_size_needed = patlen - skipatstart + big32count; |
10135 | |
10136 | if ((ccontext->extra_options & |
10137 | (PCRE2_EXTRA_MATCH_WORD|PCRE2_EXTRA_MATCH_LINE)) != 0) |
10138 | parsed_size_needed += 4; |
10139 | |
10140 | if ((options & PCRE2_AUTO_CALLOUT) != 0) |
10141 | parsed_size_needed = (parsed_size_needed + 1) * 5; |
10142 | |
10143 | if (parsed_size_needed >= PARSED_PATTERN_DEFAULT_SIZE) |
10144 | { |
10145 | uint32_t *heap_parsed_pattern = ccontext->memctl.malloc( |
10146 | (parsed_size_needed + 1) * sizeof(uint32_t), ccontext->memctl.memory_data); |
10147 | if (heap_parsed_pattern == NULL) |
10148 | { |
10149 | *errorptr = ERR21; |
10150 | goto EXIT; |
10151 | } |
10152 | cb.parsed_pattern = heap_parsed_pattern; |
10153 | } |
10154 | cb.parsed_pattern_end = cb.parsed_pattern + parsed_size_needed + 1; |
10155 | |
10156 | /* Do the parsing scan. */ |
10157 | |
10158 | errorcode = parse_regex(ptr, cb.external_options, &has_lookbehind, &cb); |
10159 | if (errorcode != 0) goto HAD_CB_ERROR; |
10160 | |
10161 | /* Workspace is needed to remember information about numbered groups: whether a |
10162 | group can match an empty string and what its fixed length is. This is done to |
10163 | avoid the possibility of recursive references causing very long compile times |
10164 | when checking these features. Unnumbered groups do not have this exposure since |
10165 | they cannot be referenced. We use an indexed vector for this purpose. If there |
10166 | are sufficiently few groups, the default vector on the stack, as set up above, |
10167 | can be used. Otherwise we have to get/free a special vector. The vector must be |
10168 | initialized to zero. */ |
10169 | |
10170 | if (cb.bracount >= GROUPINFO_DEFAULT_SIZE) |
10171 | { |
10172 | cb.groupinfo = ccontext->memctl.malloc( |
10173 | (cb.bracount + 1)*sizeof(uint32_t), ccontext->memctl.memory_data); |
10174 | if (cb.groupinfo == NULL) |
10175 | { |
10176 | errorcode = ERR21; |
10177 | cb.erroroffset = 0; |
10178 | goto HAD_CB_ERROR; |
10179 | } |
10180 | } |
10181 | memset(cb.groupinfo, 0, (cb.bracount + 1) * sizeof(uint32_t)); |
10182 | |
10183 | /* If there were any lookbehinds, scan the parsed pattern to figure out their |
10184 | lengths. */ |
10185 | |
10186 | if (has_lookbehind) |
10187 | { |
10188 | int loopcount = 0; |
10189 | errorcode = check_lookbehinds(cb.parsed_pattern, NULL, NULL, &cb, &loopcount); |
10190 | if (errorcode != 0) goto HAD_CB_ERROR; |
10191 | } |
10192 | |
10193 | /* For debugging, there is a function that shows the parsed data vector. */ |
10194 | |
10195 | #ifdef DEBUG_SHOW_PARSED |
10196 | fprintf(stderr, "+++ Pre-scan complete:\n" ); |
10197 | show_parsed(&cb); |
10198 | #endif |
10199 | |
10200 | /* For debugging capturing information this code can be enabled. */ |
10201 | |
10202 | #ifdef DEBUG_SHOW_CAPTURES |
10203 | { |
10204 | named_group *ng = cb.named_groups; |
10205 | fprintf(stderr, "+++Captures: %d\n" , cb.bracount); |
10206 | for (i = 0; i < cb.names_found; i++, ng++) |
10207 | { |
10208 | fprintf(stderr, "+++%3d %.*s\n" , ng->number, ng->length, ng->name); |
10209 | } |
10210 | } |
10211 | #endif |
10212 | |
10213 | /* Pretend to compile the pattern while actually just accumulating the amount |
10214 | of memory required in the 'length' variable. This behaviour is triggered by |
10215 | passing a non-NULL final argument to compile_regex(). We pass a block of |
10216 | workspace (cworkspace) for it to compile parts of the pattern into; the |
10217 | compiled code is discarded when it is no longer needed, so hopefully this |
10218 | workspace will never overflow, though there is a test for its doing so. |
10219 | |
10220 | On error, errorcode will be set non-zero, so we don't need to look at the |
10221 | result of the function. The initial options have been put into the cb block, |
10222 | but we still have to pass a separate options variable (the first argument) |
10223 | because the options may change as the pattern is processed. */ |
10224 | |
10225 | cb.erroroffset = patlen; /* For any subsequent errors that do not set it */ |
10226 | pptr = cb.parsed_pattern; |
10227 | code = cworkspace; |
10228 | *code = OP_BRA; |
10229 | |
10230 | (void)compile_regex(cb.external_options, &code, &pptr, &errorcode, 0, &firstcu, |
10231 | &firstcuflags, &reqcu, &reqcuflags, NULL, &cb, &length); |
10232 | |
10233 | if (errorcode != 0) goto HAD_CB_ERROR; /* Offset is in cb.erroroffset */ |
10234 | |
10235 | /* This should be caught in compile_regex(), but just in case... */ |
10236 | |
10237 | if (length > MAX_PATTERN_SIZE) |
10238 | { |
10239 | errorcode = ERR20; |
10240 | goto HAD_CB_ERROR; |
10241 | } |
10242 | |
10243 | /* Compute the size of, and then get and initialize, the data block for storing |
10244 | the compiled pattern and names table. Integer overflow should no longer be |
10245 | possible because nowadays we limit the maximum value of cb.names_found and |
10246 | cb.name_entry_size. */ |
10247 | |
10248 | re_blocksize = sizeof(pcre2_real_code) + |
10249 | CU2BYTES(length + |
10250 | (PCRE2_SIZE)cb.names_found * (PCRE2_SIZE)cb.name_entry_size); |
10251 | re = (pcre2_real_code *) |
10252 | ccontext->memctl.malloc(re_blocksize, ccontext->memctl.memory_data); |
10253 | if (re == NULL) |
10254 | { |
10255 | errorcode = ERR21; |
10256 | goto HAD_CB_ERROR; |
10257 | } |
10258 | |
10259 | /* The compiler may put padding at the end of the pcre2_real_code structure in |
10260 | order to round it up to a multiple of 4 or 8 bytes. This means that when a |
10261 | compiled pattern is copied (for example, when serialized) undefined bytes are |
10262 | read, and this annoys debuggers such as valgrind. To avoid this, we explicitly |
10263 | write to the last 8 bytes of the structure before setting the fields. */ |
10264 | |
10265 | memset((char *)re + sizeof(pcre2_real_code) - 8, 0, 8); |
10266 | re->memctl = ccontext->memctl; |
10267 | re->tables = tables; |
10268 | re->executable_jit = NULL; |
10269 | memset(re->start_bitmap, 0, 32 * sizeof(uint8_t)); |
10270 | re->blocksize = re_blocksize; |
10271 | re->magic_number = MAGIC_NUMBER; |
10272 | re->compile_options = options; |
10273 | re->overall_options = cb.external_options; |
10274 | re->extra_options = ccontext->extra_options; |
10275 | re->flags = PCRE2_CODE_UNIT_WIDTH/8 | cb.external_flags | setflags; |
10276 | re->limit_heap = limit_heap; |
10277 | re->limit_match = limit_match; |
10278 | re->limit_depth = limit_depth; |
10279 | re->first_codeunit = 0; |
10280 | re->last_codeunit = 0; |
10281 | re->bsr_convention = bsr; |
10282 | re->newline_convention = newline; |
10283 | re->max_lookbehind = 0; |
10284 | re->minlength = 0; |
10285 | re->top_bracket = 0; |
10286 | re->top_backref = 0; |
10287 | re->name_entry_size = cb.name_entry_size; |
10288 | re->name_count = cb.names_found; |
10289 | |
10290 | /* The basic block is immediately followed by the name table, and the compiled |
10291 | code follows after that. */ |
10292 | |
10293 | codestart = (PCRE2_SPTR)((uint8_t *)re + sizeof(pcre2_real_code)) + |
10294 | re->name_entry_size * re->name_count; |
10295 | |
10296 | /* Update the compile data block for the actual compile. The starting points of |
10297 | the name/number translation table and of the code are passed around in the |
10298 | compile data block. The start/end pattern and initial options are already set |
10299 | from the pre-compile phase, as is the name_entry_size field. */ |
10300 | |
10301 | cb.parens_depth = 0; |
10302 | cb.assert_depth = 0; |
10303 | cb.lastcapture = 0; |
10304 | cb.name_table = (PCRE2_UCHAR *)((uint8_t *)re + sizeof(pcre2_real_code)); |
10305 | cb.start_code = codestart; |
10306 | cb.req_varyopt = 0; |
10307 | cb.had_accept = FALSE; |
10308 | cb.had_pruneorskip = FALSE; |
10309 | cb.open_caps = NULL; |
10310 | |
10311 | /* If any named groups were found, create the name/number table from the list |
10312 | created in the pre-pass. */ |
10313 | |
10314 | if (cb.names_found > 0) |
10315 | { |
10316 | named_group *ng = cb.named_groups; |
10317 | for (i = 0; i < cb.names_found; i++, ng++) |
10318 | add_name_to_table(&cb, ng->name, ng->length, ng->number, i); |
10319 | } |
10320 | |
10321 | /* Set up a starting, non-extracting bracket, then compile the expression. On |
10322 | error, errorcode will be set non-zero, so we don't need to look at the result |
10323 | of the function here. */ |
10324 | |
10325 | pptr = cb.parsed_pattern; |
10326 | code = (PCRE2_UCHAR *)codestart; |
10327 | *code = OP_BRA; |
10328 | regexrc = compile_regex(re->overall_options, &code, &pptr, &errorcode, 0, |
10329 | &firstcu, &firstcuflags, &reqcu, &reqcuflags, NULL, &cb, NULL); |
10330 | if (regexrc < 0) re->flags |= PCRE2_MATCH_EMPTY; |
10331 | re->top_bracket = cb.bracount; |
10332 | re->top_backref = cb.top_backref; |
10333 | re->max_lookbehind = cb.max_lookbehind; |
10334 | |
10335 | if (cb.had_accept) |
10336 | { |
10337 | reqcu = 0; /* Must disable after (*ACCEPT) */ |
10338 | reqcuflags = REQ_NONE; |
10339 | re->flags |= PCRE2_HASACCEPT; /* Disables minimum length */ |
10340 | } |
10341 | |
10342 | /* Fill in the final opcode and check for disastrous overflow. If no overflow, |
10343 | but the estimated length exceeds the really used length, adjust the value of |
10344 | re->blocksize, and if valgrind support is configured, mark the extra allocated |
10345 | memory as unaddressable, so that any out-of-bound reads can be detected. */ |
10346 | |
10347 | *code++ = OP_END; |
10348 | usedlength = code - codestart; |
10349 | if (usedlength > length) errorcode = ERR23; else |
10350 | { |
10351 | re->blocksize -= CU2BYTES(length - usedlength); |
10352 | #ifdef SUPPORT_VALGRIND |
10353 | VALGRIND_MAKE_MEM_NOACCESS(code, CU2BYTES(length - usedlength)); |
10354 | #endif |
10355 | } |
10356 | |
10357 | /* Scan the pattern for recursion/subroutine calls and convert the group |
10358 | numbers into offsets. Maintain a small cache so that repeated groups containing |
10359 | recursions are efficiently handled. */ |
10360 | |
10361 | #define RSCAN_CACHE_SIZE 8 |
10362 | |
10363 | if (errorcode == 0 && cb.had_recurse) |
10364 | { |
10365 | PCRE2_UCHAR *rcode; |
10366 | PCRE2_SPTR rgroup; |
10367 | unsigned int ccount = 0; |
10368 | int start = RSCAN_CACHE_SIZE; |
10369 | recurse_cache rc[RSCAN_CACHE_SIZE]; |
10370 | |
10371 | for (rcode = (PCRE2_UCHAR *)find_recurse(codestart, utf); |
10372 | rcode != NULL; |
10373 | rcode = (PCRE2_UCHAR *)find_recurse(rcode + 1 + LINK_SIZE, utf)) |
10374 | { |
10375 | int p, groupnumber; |
10376 | |
10377 | groupnumber = (int)GET(rcode, 1); |
10378 | if (groupnumber == 0) rgroup = codestart; else |
10379 | { |
10380 | PCRE2_SPTR search_from = codestart; |
10381 | rgroup = NULL; |
10382 | for (i = 0, p = start; i < ccount; i++, p = (p + 1) & 7) |
10383 | { |
10384 | if (groupnumber == rc[p].groupnumber) |
10385 | { |
10386 | rgroup = rc[p].group; |
10387 | break; |
10388 | } |
10389 | |
10390 | /* Group n+1 must always start to the right of group n, so we can save |
10391 | search time below when the new group number is greater than any of the |
10392 | previously found groups. */ |
10393 | |
10394 | if (groupnumber > rc[p].groupnumber) search_from = rc[p].group; |
10395 | } |
10396 | |
10397 | if (rgroup == NULL) |
10398 | { |
10399 | rgroup = PRIV(find_bracket)(search_from, utf, groupnumber); |
10400 | if (rgroup == NULL) |
10401 | { |
10402 | errorcode = ERR53; |
10403 | break; |
10404 | } |
10405 | if (--start < 0) start = RSCAN_CACHE_SIZE - 1; |
10406 | rc[start].groupnumber = groupnumber; |
10407 | rc[start].group = rgroup; |
10408 | if (ccount < RSCAN_CACHE_SIZE) ccount++; |
10409 | } |
10410 | } |
10411 | |
10412 | PUT(rcode, 1, rgroup - codestart); |
10413 | } |
10414 | } |
10415 | |
10416 | /* In rare debugging situations we sometimes need to look at the compiled code |
10417 | at this stage. */ |
10418 | |
10419 | #ifdef DEBUG_CALL_PRINTINT |
10420 | pcre2_printint(re, stderr, TRUE); |
10421 | fprintf(stderr, "Length=%lu Used=%lu\n" , length, usedlength); |
10422 | #endif |
10423 | |
10424 | /* Unless disabled, check whether any single character iterators can be |
10425 | auto-possessified. The function overwrites the appropriate opcode values, so |
10426 | the type of the pointer must be cast. NOTE: the intermediate variable "temp" is |
10427 | used in this code because at least one compiler gives a warning about loss of |
10428 | "const" attribute if the cast (PCRE2_UCHAR *)codestart is used directly in the |
10429 | function call. */ |
10430 | |
10431 | if (errorcode == 0 && (re->overall_options & PCRE2_NO_AUTO_POSSESS) == 0) |
10432 | { |
10433 | PCRE2_UCHAR *temp = (PCRE2_UCHAR *)codestart; |
10434 | if (PRIV(auto_possessify)(temp, &cb) != 0) errorcode = ERR80; |
10435 | } |
10436 | |
10437 | /* Failed to compile, or error while post-processing. */ |
10438 | |
10439 | if (errorcode != 0) goto HAD_CB_ERROR; |
10440 | |
10441 | /* Successful compile. If the anchored option was not passed, set it if |
10442 | we can determine that the pattern is anchored by virtue of ^ characters or \A |
10443 | or anything else, such as starting with non-atomic .* when DOTALL is set and |
10444 | there are no occurrences of *PRUNE or *SKIP (though there is an option to |
10445 | disable this case). */ |
10446 | |
10447 | if ((re->overall_options & PCRE2_ANCHORED) == 0 && |
10448 | is_anchored(codestart, 0, &cb, 0, FALSE)) |
10449 | re->overall_options |= PCRE2_ANCHORED; |
10450 | |
10451 | /* Set up the first code unit or startline flag, the required code unit, and |
10452 | then study the pattern. This code need not be obeyed if PCRE2_NO_START_OPTIMIZE |
10453 | is set, as the data it would create will not be used. Note that a first code |
10454 | unit (but not the startline flag) is useful for anchored patterns because it |
10455 | can still give a quick "no match" and also avoid searching for a last code |
10456 | unit. */ |
10457 | |
10458 | if ((re->overall_options & PCRE2_NO_START_OPTIMIZE) == 0) |
10459 | { |
10460 | int minminlength = 0; /* For minimal minlength from first/required CU */ |
10461 | |
10462 | /* If we do not have a first code unit, see if there is one that is asserted |
10463 | (these are not saved during the compile because they can cause conflicts with |
10464 | actual literals that follow). */ |
10465 | |
10466 | if (firstcuflags >= REQ_NONE) |
10467 | firstcu = find_firstassertedcu(codestart, &firstcuflags, 0); |
10468 | |
10469 | /* Save the data for a first code unit. The existence of one means the |
10470 | minimum length must be at least 1. */ |
10471 | |
10472 | if (firstcuflags < REQ_NONE) |
10473 | { |
10474 | re->first_codeunit = firstcu; |
10475 | re->flags |= PCRE2_FIRSTSET; |
10476 | minminlength++; |
10477 | |
10478 | /* Handle caseless first code units. */ |
10479 | |
10480 | if ((firstcuflags & REQ_CASELESS) != 0) |
10481 | { |
10482 | if (firstcu < 128 || (!utf && !ucp && firstcu < 255)) |
10483 | { |
10484 | if (cb.fcc[firstcu] != firstcu) re->flags |= PCRE2_FIRSTCASELESS; |
10485 | } |
10486 | |
10487 | /* The first code unit is > 128 in UTF or UCP mode, or > 255 otherwise. |
10488 | In 8-bit UTF mode, codepoints in the range 128-255 are introductory code |
10489 | points and cannot have another case, but if UCP is set they may do. */ |
10490 | |
10491 | #ifdef SUPPORT_UNICODE |
10492 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
10493 | else if (ucp && !utf && UCD_OTHERCASE(firstcu) != firstcu) |
10494 | re->flags |= PCRE2_FIRSTCASELESS; |
10495 | #else |
10496 | else if ((utf || ucp) && firstcu <= MAX_UTF_CODE_POINT && |
10497 | UCD_OTHERCASE(firstcu) != firstcu) |
10498 | re->flags |= PCRE2_FIRSTCASELESS; |
10499 | #endif |
10500 | #endif /* SUPPORT_UNICODE */ |
10501 | } |
10502 | } |
10503 | |
10504 | /* When there is no first code unit, for non-anchored patterns, see if we can |
10505 | set the PCRE2_STARTLINE flag. This is helpful for multiline matches when all |
10506 | branches start with ^ and also when all branches start with non-atomic .* for |
10507 | non-DOTALL matches when *PRUNE and SKIP are not present. (There is an option |
10508 | that disables this case.) */ |
10509 | |
10510 | else if ((re->overall_options & PCRE2_ANCHORED) == 0 && |
10511 | is_startline(codestart, 0, &cb, 0, FALSE)) |
10512 | re->flags |= PCRE2_STARTLINE; |
10513 | |
10514 | /* Handle the "required code unit", if one is set. In the UTF case we can |
10515 | increment the minimum minimum length only if we are sure this really is a |
10516 | different character and not a non-starting code unit of the first character, |
10517 | because the minimum length count is in characters, not code units. */ |
10518 | |
10519 | if (reqcuflags < REQ_NONE) |
10520 | { |
10521 | #if PCRE2_CODE_UNIT_WIDTH == 16 |
10522 | if ((re->overall_options & PCRE2_UTF) == 0 || /* Not UTF */ |
10523 | firstcuflags >= REQ_NONE || /* First not set */ |
10524 | (firstcu & 0xf800) != 0xd800 || /* First not surrogate */ |
10525 | (reqcu & 0xfc00) != 0xdc00) /* Req not low surrogate */ |
10526 | #elif PCRE2_CODE_UNIT_WIDTH == 8 |
10527 | if ((re->overall_options & PCRE2_UTF) == 0 || /* Not UTF */ |
10528 | firstcuflags >= REQ_NONE || /* First not set */ |
10529 | (firstcu & 0x80) == 0 || /* First is ASCII */ |
10530 | (reqcu & 0x80) == 0) /* Req is ASCII */ |
10531 | #endif |
10532 | { |
10533 | minminlength++; |
10534 | } |
10535 | |
10536 | /* In the case of an anchored pattern, set up the value only if it follows |
10537 | a variable length item in the pattern. */ |
10538 | |
10539 | if ((re->overall_options & PCRE2_ANCHORED) == 0 || |
10540 | (reqcuflags & REQ_VARY) != 0) |
10541 | { |
10542 | re->last_codeunit = reqcu; |
10543 | re->flags |= PCRE2_LASTSET; |
10544 | |
10545 | /* Handle caseless required code units as for first code units (above). */ |
10546 | |
10547 | if ((reqcuflags & REQ_CASELESS) != 0) |
10548 | { |
10549 | if (reqcu < 128 || (!utf && !ucp && reqcu < 255)) |
10550 | { |
10551 | if (cb.fcc[reqcu] != reqcu) re->flags |= PCRE2_LASTCASELESS; |
10552 | } |
10553 | #ifdef SUPPORT_UNICODE |
10554 | #if PCRE2_CODE_UNIT_WIDTH == 8 |
10555 | else if (ucp && !utf && UCD_OTHERCASE(reqcu) != reqcu) |
10556 | re->flags |= PCRE2_LASTCASELESS; |
10557 | #else |
10558 | else if ((utf || ucp) && reqcu <= MAX_UTF_CODE_POINT && |
10559 | UCD_OTHERCASE(reqcu) != reqcu) |
10560 | re->flags |= PCRE2_LASTCASELESS; |
10561 | #endif |
10562 | #endif /* SUPPORT_UNICODE */ |
10563 | } |
10564 | } |
10565 | } |
10566 | |
10567 | /* Study the compiled pattern to set up information such as a bitmap of |
10568 | starting code units and a minimum matching length. */ |
10569 | |
10570 | if (PRIV(study)(re) != 0) |
10571 | { |
10572 | errorcode = ERR31; |
10573 | goto HAD_CB_ERROR; |
10574 | } |
10575 | |
10576 | /* If study() set a bitmap of starting code units, it implies a minimum |
10577 | length of at least one. */ |
10578 | |
10579 | if ((re->flags & PCRE2_FIRSTMAPSET) != 0 && minminlength == 0) |
10580 | minminlength = 1; |
10581 | |
10582 | /* If the minimum length set (or not set) by study() is less than the minimum |
10583 | implied by required code units, override it. */ |
10584 | |
10585 | if (re->minlength < minminlength) re->minlength = minminlength; |
10586 | } /* End of start-of-match optimizations. */ |
10587 | |
10588 | /* Control ends up here in all cases. When running under valgrind, make a |
10589 | pattern's terminating zero defined again. If memory was obtained for the parsed |
10590 | version of the pattern, free it before returning. Also free the list of named |
10591 | groups if a larger one had to be obtained, and likewise the group information |
10592 | vector. */ |
10593 | |
10594 | EXIT: |
10595 | #ifdef SUPPORT_VALGRIND |
10596 | if (zero_terminated) VALGRIND_MAKE_MEM_DEFINED(pattern + patlen, CU2BYTES(1)); |
10597 | #endif |
10598 | if (cb.parsed_pattern != stack_parsed_pattern) |
10599 | ccontext->memctl.free(cb.parsed_pattern, ccontext->memctl.memory_data); |
10600 | if (cb.named_group_list_size > NAMED_GROUP_LIST_SIZE) |
10601 | ccontext->memctl.free((void *)cb.named_groups, ccontext->memctl.memory_data); |
10602 | if (cb.groupinfo != stack_groupinfo) |
10603 | ccontext->memctl.free((void *)cb.groupinfo, ccontext->memctl.memory_data); |
10604 | return re; /* Will be NULL after an error */ |
10605 | |
10606 | /* Errors discovered in parse_regex() set the offset value in the compile |
10607 | block. Errors discovered before it is called must compute it from the ptr |
10608 | value. After parse_regex() is called, the offset in the compile block is set to |
10609 | the end of the pattern, but certain errors in compile_regex() may reset it if |
10610 | an offset is available in the parsed pattern. */ |
10611 | |
10612 | HAD_CB_ERROR: |
10613 | ptr = pattern + cb.erroroffset; |
10614 | |
10615 | HAD_EARLY_ERROR: |
10616 | *erroroffset = ptr - pattern; |
10617 | |
10618 | HAD_ERROR: |
10619 | *errorptr = errorcode; |
10620 | pcre2_code_free(re); |
10621 | re = NULL; |
10622 | goto EXIT; |
10623 | } |
10624 | |
10625 | /* These #undefs are here to enable unity builds with CMake. */ |
10626 | |
10627 | #undef NLBLOCK /* Block containing newline information */ |
10628 | #undef PSSTART /* Field containing processed string start */ |
10629 | #undef PSEND /* Field containing processed string end */ |
10630 | |
10631 | /* End of pcre2_compile.c */ |
10632 | |