/[pcre]/code/trunk/pcre_compile.c
ViewVC logotype

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1542 - (show annotations)
Wed Apr 1 15:43:53 2015 UTC (4 years, 10 months ago) by ph10
File MIME type: text/plain
File size: 315050 byte(s)
Fix stack overflow instead of diagnostic for mutual recursion inside a 
lookbehind assertion.
1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2014 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40
41 /* This module contains the external function pcre_compile(), along with
42 supporting internal functions that are not used by other modules. */
43
44
45 #ifdef HAVE_CONFIG_H
46 #include "config.h"
47 #endif
48
49 #define NLBLOCK cd /* Block containing newline information */
50 #define PSSTART start_pattern /* Field containing pattern start */
51 #define PSEND end_pattern /* Field containing pattern end */
52
53 #include "pcre_internal.h"
54
55
56 /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57 is also used by pcretest. PCRE_DEBUG is not defined when building a production
58 library. We do not need to select pcre16_printint.c specially, because the
59 COMPILE_PCREx macro will already be appropriately set. */
60
61 #ifdef PCRE_DEBUG
62 /* pcre_printint.c should not include any headers */
63 #define PCRE_INCLUDED
64 #include "pcre_printint.c"
65 #undef PCRE_INCLUDED
66 #endif
67
68
69 /* Macro for setting individual bits in class bitmaps. */
70
71 #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72
73 /* Maximum length value to check against when making sure that the integer that
74 holds the compiled pattern length does not overflow. We make it a bit less than
75 INT_MAX to allow for adding in group terminating bytes, so that we don't have
76 to check them every time. */
77
78 #define OFLOW_MAX (INT_MAX - 20)
79
80 /* Definitions to allow mutual recursion */
81
82 static int
83 add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84 const pcre_uint32 *, unsigned int);
85
86 static BOOL
87 compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88 pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89 compile_data *, int *);
90
91
92
93 /*************************************************
94 * Code parameters and static tables *
95 *************************************************/
96
97 /* This value specifies the size of stack workspace that is used during the
98 first pre-compile phase that determines how much memory is required. The regex
99 is partly compiled into this space, but the compiled parts are discarded as
100 soon as they can be, so that hopefully there will never be an overrun. The code
101 does, however, check for an overrun. The largest amount I've seen used is 218,
102 so this number is very generous.
103
104 The same workspace is used during the second, actual compile phase for
105 remembering forward references to groups so that they can be filled in at the
106 end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107 is 4 there is plenty of room for most patterns. However, the memory can get
108 filled up by repetitions of forward references, for example patterns like
109 /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110 that the workspace is expanded using malloc() in this situation. The value
111 below is therefore a minimum, and we put a maximum on it for safety. The
112 minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113 kicks in at the same number of forward references in all cases. */
114
115 #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116 #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117
118 /* This value determines the size of the initial vector that is used for
119 remembering named groups during the pre-compile. It is allocated on the stack,
120 but if it is too small, it is expanded using malloc(), in a similar way to the
121 workspace. The value is the number of slots in the list. */
122
123 #define NAMED_GROUP_LIST_SIZE 20
124
125 /* The overrun tests check for a slightly smaller size so that they detect the
126 overrun before it actually does run off the end of the data block. */
127
128 #define WORK_SIZE_SAFETY_MARGIN (100)
129
130 /* Private flags added to firstchar and reqchar. */
131
132 #define REQ_CASELESS (1 << 0) /* Indicates caselessness */
133 #define REQ_VARY (1 << 1) /* Reqchar followed non-literal item */
134 /* Negative values for the firstchar and reqchar flags */
135 #define REQ_UNSET (-2)
136 #define REQ_NONE (-1)
137
138 /* Repeated character flags. */
139
140 #define UTF_LENGTH 0x10000000l /* The char contains its length. */
141
142 /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143 are simple data values; negative values are for special things like \d and so
144 on. Zero means further processing is needed (for things like \x), or the escape
145 is invalid. */
146
147 #ifndef EBCDIC
148
149 /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150 in UTF-8 mode. */
151
152 static const short int escapes[] = {
153 0, 0,
154 0, 0,
155 0, 0,
156 0, 0,
157 0, 0,
158 CHAR_COLON, CHAR_SEMICOLON,
159 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN,
160 CHAR_GREATER_THAN_SIGN, CHAR_QUESTION_MARK,
161 CHAR_COMMERCIAL_AT, -ESC_A,
162 -ESC_B, -ESC_C,
163 -ESC_D, -ESC_E,
164 0, -ESC_G,
165 -ESC_H, 0,
166 0, -ESC_K,
167 0, 0,
168 -ESC_N, 0,
169 -ESC_P, -ESC_Q,
170 -ESC_R, -ESC_S,
171 0, 0,
172 -ESC_V, -ESC_W,
173 -ESC_X, 0,
174 -ESC_Z, CHAR_LEFT_SQUARE_BRACKET,
175 CHAR_BACKSLASH, CHAR_RIGHT_SQUARE_BRACKET,
176 CHAR_CIRCUMFLEX_ACCENT, CHAR_UNDERSCORE,
177 CHAR_GRAVE_ACCENT, 7,
178 -ESC_b, 0,
179 -ESC_d, ESC_e,
180 ESC_f, 0,
181 -ESC_h, 0,
182 0, -ESC_k,
183 0, 0,
184 ESC_n, 0,
185 -ESC_p, 0,
186 ESC_r, -ESC_s,
187 ESC_tee, 0,
188 -ESC_v, -ESC_w,
189 0, 0,
190 -ESC_z
191 };
192
193 #else
194
195 /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
196
197 static const short int escapes[] = {
198 /* 48 */ 0, 0, 0, '.', '<', '(', '+', '|',
199 /* 50 */ '&', 0, 0, 0, 0, 0, 0, 0,
200 /* 58 */ 0, 0, '!', '$', '*', ')', ';', '~',
201 /* 60 */ '-', '/', 0, 0, 0, 0, 0, 0,
202 /* 68 */ 0, 0, '|', ',', '%', '_', '>', '?',
203 /* 70 */ 0, 0, 0, 0, 0, 0, 0, 0,
204 /* 78 */ 0, '`', ':', '#', '@', '\'', '=', '"',
205 /* 80 */ 0, 7, -ESC_b, 0, -ESC_d, ESC_e, ESC_f, 0,
206 /* 88 */-ESC_h, 0, 0, '{', 0, 0, 0, 0,
207 /* 90 */ 0, 0, -ESC_k, 'l', 0, ESC_n, 0, -ESC_p,
208 /* 98 */ 0, ESC_r, 0, '}', 0, 0, 0, 0,
209 /* A0 */ 0, '~', -ESC_s, ESC_tee, 0,-ESC_v, -ESC_w, 0,
210 /* A8 */ 0,-ESC_z, 0, 0, 0, '[', 0, 0,
211 /* B0 */ 0, 0, 0, 0, 0, 0, 0, 0,
212 /* B8 */ 0, 0, 0, 0, 0, ']', '=', '-',
213 /* C0 */ '{',-ESC_A, -ESC_B, -ESC_C, -ESC_D,-ESC_E, 0, -ESC_G,
214 /* C8 */-ESC_H, 0, 0, 0, 0, 0, 0, 0,
215 /* D0 */ '}', 0, -ESC_K, 0, 0,-ESC_N, 0, -ESC_P,
216 /* D8 */-ESC_Q,-ESC_R, 0, 0, 0, 0, 0, 0,
217 /* E0 */ '\\', 0, -ESC_S, 0, 0,-ESC_V, -ESC_W, -ESC_X,
218 /* E8 */ 0,-ESC_Z, 0, 0, 0, 0, 0, 0,
219 /* F0 */ 0, 0, 0, 0, 0, 0, 0, 0,
220 /* F8 */ 0, 0, 0, 0, 0, 0, 0, 0
221 };
222 #endif
223
224
225 /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226 searched linearly. Put all the names into a single string, in order to reduce
227 the number of relocations when a shared library is dynamically linked. The
228 string is built from string macros so that it works in UTF-8 mode on EBCDIC
229 platforms. */
230
231 typedef struct verbitem {
232 int len; /* Length of verb name */
233 int op; /* Op when no arg, or -1 if arg mandatory */
234 int op_arg; /* Op when arg present, or -1 if not allowed */
235 } verbitem;
236
237 static const char verbnames[] =
238 "\0" /* Empty name is a shorthand for MARK */
239 STRING_MARK0
240 STRING_ACCEPT0
241 STRING_COMMIT0
242 STRING_F0
243 STRING_FAIL0
244 STRING_PRUNE0
245 STRING_SKIP0
246 STRING_THEN;
247
248 static const verbitem verbs[] = {
249 { 0, -1, OP_MARK },
250 { 4, -1, OP_MARK },
251 { 6, OP_ACCEPT, -1 },
252 { 6, OP_COMMIT, -1 },
253 { 1, OP_FAIL, -1 },
254 { 4, OP_FAIL, -1 },
255 { 5, OP_PRUNE, OP_PRUNE_ARG },
256 { 4, OP_SKIP, OP_SKIP_ARG },
257 { 4, OP_THEN, OP_THEN_ARG }
258 };
259
260 static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261
262
263 /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264 another regex library. */
265
266 static const pcre_uchar sub_start_of_word[] = {
267 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268 CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269
270 static const pcre_uchar sub_end_of_word[] = {
271 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273 CHAR_RIGHT_PARENTHESIS, '\0' };
274
275
276 /* Tables of names of POSIX character classes and their lengths. The names are
277 now all in a single string, to reduce the number of relocations when a shared
278 library is dynamically loaded. The list of lengths is terminated by a zero
279 length entry. The first three must be alpha, lower, upper, as this is assumed
280 for handling case independence. The indices for graph, print, and punct are
281 needed, so identify them. */
282
283 static const char posix_names[] =
284 STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
285 STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
286 STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287 STRING_word0 STRING_xdigit;
288
289 static const pcre_uint8 posix_name_lengths[] = {
290 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291
292 #define PC_GRAPH 8
293 #define PC_PRINT 9
294 #define PC_PUNCT 10
295
296
297 /* Table of class bit maps for each POSIX class. Each class is formed from a
298 base map, with an optional addition or removal of another map. Then, for some
299 classes, there is some additional tweaking: for [:blank:] the vertical space
300 characters are removed, and for [:alpha:] and [:alnum:] the underscore
301 character is removed. The triples in the table consist of the base map offset,
302 second map offset or -1 if no second map, and a non-negative value for map
303 addition or a negative value for map subtraction (if there are two maps). The
304 absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
305 remove vertical space characters, 2 => remove underscore. */
306
307 static const int posix_class_maps[] = {
308 cbit_word, cbit_digit, -2, /* alpha */
309 cbit_lower, -1, 0, /* lower */
310 cbit_upper, -1, 0, /* upper */
311 cbit_word, -1, 2, /* alnum - word without underscore */
312 cbit_print, cbit_cntrl, 0, /* ascii */
313 cbit_space, -1, 1, /* blank - a GNU extension */
314 cbit_cntrl, -1, 0, /* cntrl */
315 cbit_digit, -1, 0, /* digit */
316 cbit_graph, -1, 0, /* graph */
317 cbit_print, -1, 0, /* print */
318 cbit_punct, -1, 0, /* punct */
319 cbit_space, -1, 0, /* space */
320 cbit_word, -1, 0, /* word - a Perl extension */
321 cbit_xdigit,-1, 0 /* xdigit */
322 };
323
324 /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325 Unicode property escapes. */
326
327 #ifdef SUPPORT_UCP
328 static const pcre_uchar string_PNd[] = {
329 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331 static const pcre_uchar string_pNd[] = {
332 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334 static const pcre_uchar string_PXsp[] = {
335 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337 static const pcre_uchar string_pXsp[] = {
338 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340 static const pcre_uchar string_PXwd[] = {
341 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343 static const pcre_uchar string_pXwd[] = {
344 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346
347 static const pcre_uchar *substitutes[] = {
348 string_PNd, /* \D */
349 string_pNd, /* \d */
350 string_PXsp, /* \S */ /* Xsp is Perl space, but from 8.34, Perl */
351 string_pXsp, /* \s */ /* space and POSIX space are the same. */
352 string_PXwd, /* \W */
353 string_pXwd /* \w */
354 };
355
356 /* The POSIX class substitutes must be in the order of the POSIX class names,
357 defined above, and there are both positive and negative cases. NULL means no
358 general substitute of a Unicode property escape (\p or \P). However, for some
359 POSIX classes (e.g. graph, print, punct) a special property code is compiled
360 directly. */
361
362 static const pcre_uchar string_pL[] = {
363 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365 static const pcre_uchar string_pLl[] = {
366 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368 static const pcre_uchar string_pLu[] = {
369 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371 static const pcre_uchar string_pXan[] = {
372 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374 static const pcre_uchar string_h[] = {
375 CHAR_BACKSLASH, CHAR_h, '\0' };
376 static const pcre_uchar string_pXps[] = {
377 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379 static const pcre_uchar string_PL[] = {
380 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382 static const pcre_uchar string_PLl[] = {
383 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385 static const pcre_uchar string_PLu[] = {
386 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388 static const pcre_uchar string_PXan[] = {
389 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391 static const pcre_uchar string_H[] = {
392 CHAR_BACKSLASH, CHAR_H, '\0' };
393 static const pcre_uchar string_PXps[] = {
394 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396
397 static const pcre_uchar *posix_substitutes[] = {
398 string_pL, /* alpha */
399 string_pLl, /* lower */
400 string_pLu, /* upper */
401 string_pXan, /* alnum */
402 NULL, /* ascii */
403 string_h, /* blank */
404 NULL, /* cntrl */
405 string_pNd, /* digit */
406 NULL, /* graph */
407 NULL, /* print */
408 NULL, /* punct */
409 string_pXps, /* space */ /* Xps is POSIX space, but from 8.34 */
410 string_pXwd, /* word */ /* Perl and POSIX space are the same */
411 NULL, /* xdigit */
412 /* Negated cases */
413 string_PL, /* ^alpha */
414 string_PLl, /* ^lower */
415 string_PLu, /* ^upper */
416 string_PXan, /* ^alnum */
417 NULL, /* ^ascii */
418 string_H, /* ^blank */
419 NULL, /* ^cntrl */
420 string_PNd, /* ^digit */
421 NULL, /* ^graph */
422 NULL, /* ^print */
423 NULL, /* ^punct */
424 string_PXps, /* ^space */ /* Xps is POSIX space, but from 8.34 */
425 string_PXwd, /* ^word */ /* Perl and POSIX space are the same */
426 NULL /* ^xdigit */
427 };
428 #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429 #endif
430
431 #define STRING(a) # a
432 #define XSTRING(s) STRING(s)
433
434 /* The texts of compile-time error messages. These are "char *" because they
435 are passed to the outside world. Do not ever re-use any error number, because
436 they are documented. Always add a new error instead. Messages marked DEAD below
437 are no longer used. This used to be a table of strings, but in order to reduce
438 the number of relocations needed when a shared library is loaded dynamically,
439 it is now one long string. We cannot use a table of offsets, because the
440 lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
441 simply count through to the one we want - this isn't a performance issue
442 because these strings are used only when there is a compilation error.
443
444 Each substring ends with \0 to insert a null character. This includes the final
445 substring, so that the whole string ends with \0\0, which can be detected when
446 counting through. */
447
448 static const char error_texts[] =
449 "no error\0"
450 "\\ at end of pattern\0"
451 "\\c at end of pattern\0"
452 "unrecognized character follows \\\0"
453 "numbers out of order in {} quantifier\0"
454 /* 5 */
455 "number too big in {} quantifier\0"
456 "missing terminating ] for character class\0"
457 "invalid escape sequence in character class\0"
458 "range out of order in character class\0"
459 "nothing to repeat\0"
460 /* 10 */
461 "operand of unlimited repeat could match the empty string\0" /** DEAD **/
462 "internal error: unexpected repeat\0"
463 "unrecognized character after (? or (?-\0"
464 "POSIX named classes are supported only within a class\0"
465 "missing )\0"
466 /* 15 */
467 "reference to non-existent subpattern\0"
468 "erroffset passed as NULL\0"
469 "unknown option bit(s) set\0"
470 "missing ) after comment\0"
471 "parentheses nested too deeply\0" /** DEAD **/
472 /* 20 */
473 "regular expression is too large\0"
474 "failed to get memory\0"
475 "unmatched parentheses\0"
476 "internal error: code overflow\0"
477 "unrecognized character after (?<\0"
478 /* 25 */
479 "lookbehind assertion is not fixed length\0"
480 "malformed number or name after (?(\0"
481 "conditional group contains more than two branches\0"
482 "assertion expected after (?(\0"
483 "(?R or (?[+-]digits must be followed by )\0"
484 /* 30 */
485 "unknown POSIX class name\0"
486 "POSIX collating elements are not supported\0"
487 "this version of PCRE is compiled without UTF support\0"
488 "spare error\0" /** DEAD **/
489 "character value in \\x{} or \\o{} is too large\0"
490 /* 35 */
491 "invalid condition (?(0)\0"
492 "\\C not allowed in lookbehind assertion\0"
493 "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
494 "number after (?C is > 255\0"
495 "closing ) for (?C expected\0"
496 /* 40 */
497 "recursive call could loop indefinitely\0"
498 "unrecognized character after (?P\0"
499 "syntax error in subpattern name (missing terminator)\0"
500 "two named subpatterns have the same name\0"
501 "invalid UTF-8 string\0"
502 /* 45 */
503 "support for \\P, \\p, and \\X has not been compiled\0"
504 "malformed \\P or \\p sequence\0"
505 "unknown property name after \\P or \\p\0"
506 "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
507 "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508 /* 50 */
509 "repeated subpattern is too long\0" /** DEAD **/
510 "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
511 "internal error: overran compiling workspace\0"
512 "internal error: previously-checked referenced subpattern not found\0"
513 "DEFINE group contains more than one branch\0"
514 /* 55 */
515 "repeating a DEFINE group is not allowed\0" /** DEAD **/
516 "inconsistent NEWLINE options\0"
517 "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
518 "a numbered reference must not be zero\0"
519 "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520 /* 60 */
521 "(*VERB) not recognized or malformed\0"
522 "number is too big\0"
523 "subpattern name expected\0"
524 "digit expected after (?+\0"
525 "] is an invalid data character in JavaScript compatibility mode\0"
526 /* 65 */
527 "different names for subpatterns of the same number are not allowed\0"
528 "(*MARK) must have an argument\0"
529 "this version of PCRE is not compiled with Unicode property support\0"
530 "\\c must be followed by an ASCII character\0"
531 "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532 /* 70 */
533 "internal error: unknown opcode in find_fixedlength()\0"
534 "\\N is not supported in a class\0"
535 "too many forward references\0"
536 "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537 "invalid UTF-16 string\0"
538 /* 75 */
539 "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540 "character value in \\u.... sequence is too large\0"
541 "invalid UTF-32 string\0"
542 "setting UTF is disabled by the application\0"
543 "non-hex character in \\x{} (closing brace missing?)\0"
544 /* 80 */
545 "non-octal character in \\o{} (closing brace missing?)\0"
546 "missing opening brace after \\o\0"
547 "parentheses are too deeply nested\0"
548 "invalid range in character class\0"
549 "group name must start with a non-digit\0"
550 /* 85 */
551 "parentheses are too deeply nested (stack check)\0"
552 "digits missing in \\x{} or \\o{}\0"
553 ;
554
555 /* Table to identify digits and hex digits. This is used when compiling
556 patterns. Note that the tables in chartables are dependent on the locale, and
557 may mark arbitrary characters as digits - but the PCRE compiling code expects
558 to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
559 a private table here. It costs 256 bytes, but it is a lot faster than doing
560 character value tests (at least in some simple cases I timed), and in some
561 applications one wants PCRE to compile efficiently as well as match
562 efficiently.
563
564 For convenience, we use the same bit definitions as in chartables:
565
566 0x04 decimal digit
567 0x08 hexadecimal digit
568
569 Then we can use ctype_digit and ctype_xdigit in the code. */
570
571 /* Using a simple comparison for decimal numbers rather than a memory read
572 is much faster, and the resulting code is simpler (the compiler turns it
573 into a subtraction and unsigned comparison). */
574
575 #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
576
577 #ifndef EBCDIC
578
579 /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
580 UTF-8 mode. */
581
582 static const pcre_uint8 digitab[] =
583 {
584 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
585 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
586 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
587 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
588 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
589 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
590 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
591 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
592 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* @ - G */
593 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H - O */
594 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* P - W */
595 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* X - _ */
596 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
597 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
598 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
599 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
600 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
601 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
602 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
603 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
604 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
605 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
606 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
607 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
608 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
609 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
610 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
611 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
612 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
613 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
614 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
615 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
616
617 #else
618
619 /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
620
621 static const pcre_uint8 digitab[] =
622 {
623 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
624 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
625 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
626 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
627 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
628 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
629 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
630 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
631 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
632 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
633 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
634 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
635 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
636 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
637 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
638 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
639 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
640 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
641 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
642 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
643 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
644 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
645 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
646 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
647 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
648 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
649 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
650 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
651 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
652 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
653 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
654 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
655
656 static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
657 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
658 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
659 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
660 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
661 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
662 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
663 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
664 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
665 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
666 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
667 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
668 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
669 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
670 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
671 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
672 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
673 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
674 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
675 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
676 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
677 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
678 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
679 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
680 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
681 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
682 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
683 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
684 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
685 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
686 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
687 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
688 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
689 #endif
690
691
692 /* This table is used to check whether auto-possessification is possible
693 between adjacent character-type opcodes. The left-hand (repeated) opcode is
694 used to select the row, and the right-hand opcode is use to select the column.
695 A value of 1 means that auto-possessification is OK. For example, the second
696 value in the first row means that \D+\d can be turned into \D++\d.
697
698 The Unicode property types (\P and \p) have to be present to fill out the table
699 because of what their opcode values are, but the table values should always be
700 zero because property types are handled separately in the code. The last four
701 columns apply to items that cannot be repeated, so there is no need to have
702 rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
703 *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
704
705 #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
706 #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
707
708 static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
709 /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
710 { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
711 { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
712 { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
713 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
714 { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
715 { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
716 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
717 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
718 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
719 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
720 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
721 { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
722 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
723 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
724 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
725 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
726 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
727 };
728
729
730 /* This table is used to check whether auto-possessification is possible
731 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
732 left-hand (repeated) opcode is used to select the row, and the right-hand
733 opcode is used to select the column. The values are as follows:
734
735 0 Always return FALSE (never auto-possessify)
736 1 Character groups are distinct (possessify if both are OP_PROP)
737 2 Check character categories in the same group (general or particular)
738 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
739
740 4 Check left general category vs right particular category
741 5 Check right general category vs left particular category
742
743 6 Left alphanum vs right general category
744 7 Left space vs right general category
745 8 Left word vs right general category
746
747 9 Right alphanum vs left general category
748 10 Right space vs left general category
749 11 Right word vs left general category
750
751 12 Left alphanum vs right particular category
752 13 Left space vs right particular category
753 14 Left word vs right particular category
754
755 15 Right alphanum vs left particular category
756 16 Right space vs left particular category
757 17 Right word vs left particular category
758 */
759
760 static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
761 /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
762 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
763 { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
764 { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
765 { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
766 { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
767 { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
768 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
769 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
770 { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
771 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
772 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
773 };
774
775 /* This table is used to check whether auto-possessification is possible
776 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
777 specifies a general category and the other specifies a particular category. The
778 row is selected by the general category and the column by the particular
779 category. The value is 1 if the particular category is not part of the general
780 category. */
781
782 static const pcre_uint8 catposstab[7][30] = {
783 /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
784 { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* C */
785 { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* L */
786 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* M */
787 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* N */
788 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 }, /* P */
789 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 }, /* S */
790 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 } /* Z */
791 };
792
793 /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
794 a general or particular category. The properties in each row are those
795 that apply to the character set in question. Duplication means that a little
796 unnecessary work is done when checking, but this keeps things much simpler
797 because they can all use the same code. For more details see the comment where
798 this table is used.
799
800 Note: SPACE and PXSPACE used to be different because Perl excluded VT from
801 "space", but from Perl 5.18 it's included, so both categories are treated the
802 same here. */
803
804 static const pcre_uint8 posspropstab[3][4] = {
805 { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
806 { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
807 { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
808 };
809
810 /* This table is used when converting repeating opcodes into possessified
811 versions as a result of an explicit possessive quantifier such as ++. A zero
812 value means there is no possessified version - in those cases the item in
813 question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
814 because all relevant opcodes are less than that. */
815
816 static const pcre_uint8 opcode_possessify[] = {
817 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */
818 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */
819
820 0, /* NOTI */
821 OP_POSSTAR, 0, /* STAR, MINSTAR */
822 OP_POSPLUS, 0, /* PLUS, MINPLUS */
823 OP_POSQUERY, 0, /* QUERY, MINQUERY */
824 OP_POSUPTO, 0, /* UPTO, MINUPTO */
825 0, /* EXACT */
826 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */
827
828 OP_POSSTARI, 0, /* STARI, MINSTARI */
829 OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */
830 OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */
831 OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */
832 0, /* EXACTI */
833 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */
834
835 OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */
836 OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */
837 OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */
838 OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */
839 0, /* NOTEXACT */
840 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
841
842 OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */
843 OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */
844 OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */
845 OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */
846 0, /* NOTEXACTI */
847 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
848
849 OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */
850 OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */
851 OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */
852 OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */
853 0, /* TYPEEXACT */
854 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
855
856 OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */
857 OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */
858 OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */
859 OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */
860 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */
861
862 0, 0, 0, /* CLASS, NCLASS, XCLASS */
863 0, 0, /* REF, REFI */
864 0, 0, /* DNREF, DNREFI */
865 0, 0 /* RECURSE, CALLOUT */
866 };
867
868
869 /* Structure for mutual recursion detection. */
870
871 typedef struct recurse_check {
872 struct recurse_check *prev;
873 const pcre_uchar *group;
874 } recurse_check;
875
876
877
878 /*************************************************
879 * Find an error text *
880 *************************************************/
881
882 /* The error texts are now all in one long string, to save on relocations. As
883 some of the text is of unknown length, we can't use a table of offsets.
884 Instead, just count through the strings. This is not a performance issue
885 because it happens only when there has been a compilation error.
886
887 Argument: the error number
888 Returns: pointer to the error string
889 */
890
891 static const char *
892 find_error_text(int n)
893 {
894 const char *s = error_texts;
895 for (; n > 0; n--)
896 {
897 while (*s++ != CHAR_NULL) {};
898 if (*s == CHAR_NULL) return "Error text not found (please report)";
899 }
900 return s;
901 }
902
903
904
905 /*************************************************
906 * Expand the workspace *
907 *************************************************/
908
909 /* This function is called during the second compiling phase, if the number of
910 forward references fills the existing workspace, which is originally a block on
911 the stack. A larger block is obtained from malloc() unless the ultimate limit
912 has been reached or the increase will be rather small.
913
914 Argument: pointer to the compile data block
915 Returns: 0 if all went well, else an error number
916 */
917
918 static int
919 expand_workspace(compile_data *cd)
920 {
921 pcre_uchar *newspace;
922 int newsize = cd->workspace_size * 2;
923
924 if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
925 if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
926 newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
927 return ERR72;
928
929 newspace = (PUBL(malloc))(IN_UCHARS(newsize));
930 if (newspace == NULL) return ERR21;
931 memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
932 cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
933 if (cd->workspace_size > COMPILE_WORK_SIZE)
934 (PUBL(free))((void *)cd->start_workspace);
935 cd->start_workspace = newspace;
936 cd->workspace_size = newsize;
937 return 0;
938 }
939
940
941
942 /*************************************************
943 * Check for counted repeat *
944 *************************************************/
945
946 /* This function is called when a '{' is encountered in a place where it might
947 start a quantifier. It looks ahead to see if it really is a quantifier or not.
948 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
949 where the ddds are digits.
950
951 Arguments:
952 p pointer to the first char after '{'
953
954 Returns: TRUE or FALSE
955 */
956
957 static BOOL
958 is_counted_repeat(const pcre_uchar *p)
959 {
960 if (!IS_DIGIT(*p)) return FALSE;
961 p++;
962 while (IS_DIGIT(*p)) p++;
963 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
964
965 if (*p++ != CHAR_COMMA) return FALSE;
966 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
967
968 if (!IS_DIGIT(*p)) return FALSE;
969 p++;
970 while (IS_DIGIT(*p)) p++;
971
972 return (*p == CHAR_RIGHT_CURLY_BRACKET);
973 }
974
975
976
977 /*************************************************
978 * Handle escapes *
979 *************************************************/
980
981 /* This function is called when a \ has been encountered. It either returns a
982 positive value for a simple escape such as \n, or 0 for a data character which
983 will be placed in chptr. A backreference to group n is returned as negative n.
984 When UTF-8 is enabled, a positive value greater than 255 may be returned in
985 chptr. On entry, ptr is pointing at the \. On exit, it is on the final
986 character of the escape sequence.
987
988 Arguments:
989 ptrptr points to the pattern position pointer
990 chptr points to a returned data character
991 errorcodeptr points to the errorcode variable
992 bracount number of previous extracting brackets
993 options the options bits
994 isclass TRUE if inside a character class
995
996 Returns: zero => a data character
997 positive => a special escape sequence
998 negative => a back reference
999 on error, errorcodeptr is set
1000 */
1001
1002 static int
1003 check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
1004 int bracount, int options, BOOL isclass)
1005 {
1006 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
1007 BOOL utf = (options & PCRE_UTF8) != 0;
1008 const pcre_uchar *ptr = *ptrptr + 1;
1009 pcre_uint32 c;
1010 int escape = 0;
1011 int i;
1012
1013 GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
1014 ptr--; /* Set pointer back to the last byte */
1015
1016 /* If backslash is at the end of the pattern, it's an error. */
1017
1018 if (c == CHAR_NULL) *errorcodeptr = ERR1;
1019
1020 /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1021 in a table. A non-zero result is something that can be returned immediately.
1022 Otherwise further processing may be required. */
1023
1024 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1025 /* Not alphanumeric */
1026 else if (c < CHAR_0 || c > CHAR_z) {}
1027 else if ((i = escapes[c - CHAR_0]) != 0)
1028 { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1029
1030 #else /* EBCDIC coding */
1031 /* Not alphanumeric */
1032 else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1033 else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1034 #endif
1035
1036 /* Escapes that need further processing, or are illegal. */
1037
1038 else
1039 {
1040 const pcre_uchar *oldptr;
1041 BOOL braced, negated, overflow;
1042 int s;
1043
1044 switch (c)
1045 {
1046 /* A number of Perl escapes are not handled by PCRE. We give an explicit
1047 error. */
1048
1049 case CHAR_l:
1050 case CHAR_L:
1051 *errorcodeptr = ERR37;
1052 break;
1053
1054 case CHAR_u:
1055 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1056 {
1057 /* In JavaScript, \u must be followed by four hexadecimal numbers.
1058 Otherwise it is a lowercase u letter. */
1059 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1060 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1061 && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1062 && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1063 {
1064 c = 0;
1065 for (i = 0; i < 4; ++i)
1066 {
1067 register pcre_uint32 cc = *(++ptr);
1068 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1069 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1070 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1071 #else /* EBCDIC coding */
1072 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1073 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1074 #endif
1075 }
1076
1077 #if defined COMPILE_PCRE8
1078 if (c > (utf ? 0x10ffffU : 0xffU))
1079 #elif defined COMPILE_PCRE16
1080 if (c > (utf ? 0x10ffffU : 0xffffU))
1081 #elif defined COMPILE_PCRE32
1082 if (utf && c > 0x10ffffU)
1083 #endif
1084 {
1085 *errorcodeptr = ERR76;
1086 }
1087 else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1088 }
1089 }
1090 else
1091 *errorcodeptr = ERR37;
1092 break;
1093
1094 case CHAR_U:
1095 /* In JavaScript, \U is an uppercase U letter. */
1096 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1097 break;
1098
1099 /* In a character class, \g is just a literal "g". Outside a character
1100 class, \g must be followed by one of a number of specific things:
1101
1102 (1) A number, either plain or braced. If positive, it is an absolute
1103 backreference. If negative, it is a relative backreference. This is a Perl
1104 5.10 feature.
1105
1106 (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1107 is part of Perl's movement towards a unified syntax for back references. As
1108 this is synonymous with \k{name}, we fudge it up by pretending it really
1109 was \k.
1110
1111 (3) For Oniguruma compatibility we also support \g followed by a name or a
1112 number either in angle brackets or in single quotes. However, these are
1113 (possibly recursive) subroutine calls, _not_ backreferences. Just return
1114 the ESC_g code (cf \k). */
1115
1116 case CHAR_g:
1117 if (isclass) break;
1118 if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1119 {
1120 escape = ESC_g;
1121 break;
1122 }
1123
1124 /* Handle the Perl-compatible cases */
1125
1126 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1127 {
1128 const pcre_uchar *p;
1129 for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1130 if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1131 if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1132 {
1133 escape = ESC_k;
1134 break;
1135 }
1136 braced = TRUE;
1137 ptr++;
1138 }
1139 else braced = FALSE;
1140
1141 if (ptr[1] == CHAR_MINUS)
1142 {
1143 negated = TRUE;
1144 ptr++;
1145 }
1146 else negated = FALSE;
1147
1148 /* The integer range is limited by the machine's int representation. */
1149 s = 0;
1150 overflow = FALSE;
1151 while (IS_DIGIT(ptr[1]))
1152 {
1153 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1154 {
1155 overflow = TRUE;
1156 break;
1157 }
1158 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1159 }
1160 if (overflow) /* Integer overflow */
1161 {
1162 while (IS_DIGIT(ptr[1]))
1163 ptr++;
1164 *errorcodeptr = ERR61;
1165 break;
1166 }
1167
1168 if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1169 {
1170 *errorcodeptr = ERR57;
1171 break;
1172 }
1173
1174 if (s == 0)
1175 {
1176 *errorcodeptr = ERR58;
1177 break;
1178 }
1179
1180 if (negated)
1181 {
1182 if (s > bracount)
1183 {
1184 *errorcodeptr = ERR15;
1185 break;
1186 }
1187 s = bracount - (s - 1);
1188 }
1189
1190 escape = -s;
1191 break;
1192
1193 /* The handling of escape sequences consisting of a string of digits
1194 starting with one that is not zero is not straightforward. Perl has changed
1195 over the years. Nowadays \g{} for backreferences and \o{} for octal are
1196 recommended to avoid the ambiguities in the old syntax.
1197
1198 Outside a character class, the digits are read as a decimal number. If the
1199 number is less than 8 (used to be 10), or if there are that many previous
1200 extracting left brackets, then it is a back reference. Otherwise, up to
1201 three octal digits are read to form an escaped byte. Thus \123 is likely to
1202 be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1203 the octal value is greater than 377, the least significant 8 bits are
1204 taken. \8 and \9 are treated as the literal characters 8 and 9.
1205
1206 Inside a character class, \ followed by a digit is always either a literal
1207 8 or 9 or an octal number. */
1208
1209 case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1210 case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1211
1212 if (!isclass)
1213 {
1214 oldptr = ptr;
1215 /* The integer range is limited by the machine's int representation. */
1216 s = (int)(c -CHAR_0);
1217 overflow = FALSE;
1218 while (IS_DIGIT(ptr[1]))
1219 {
1220 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1221 {
1222 overflow = TRUE;
1223 break;
1224 }
1225 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1226 }
1227 if (overflow) /* Integer overflow */
1228 {
1229 while (IS_DIGIT(ptr[1]))
1230 ptr++;
1231 *errorcodeptr = ERR61;
1232 break;
1233 }
1234 if (s < 8 || s <= bracount) /* Check for back reference */
1235 {
1236 escape = -s;
1237 break;
1238 }
1239 ptr = oldptr; /* Put the pointer back and fall through */
1240 }
1241
1242 /* Handle a digit following \ when the number is not a back reference. If
1243 the first digit is 8 or 9, Perl used to generate a binary zero byte and
1244 then treat the digit as a following literal. At least by Perl 5.18 this
1245 changed so as not to insert the binary zero. */
1246
1247 if ((c = *ptr) >= CHAR_8) break;
1248
1249 /* Fall through with a digit less than 8 */
1250
1251 /* \0 always starts an octal number, but we may drop through to here with a
1252 larger first octal digit. The original code used just to take the least
1253 significant 8 bits of octal numbers (I think this is what early Perls used
1254 to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1255 but no more than 3 octal digits. */
1256
1257 case CHAR_0:
1258 c -= CHAR_0;
1259 while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1260 c = c * 8 + *(++ptr) - CHAR_0;
1261 #ifdef COMPILE_PCRE8
1262 if (!utf && c > 0xff) *errorcodeptr = ERR51;
1263 #endif
1264 break;
1265
1266 /* \o is a relatively new Perl feature, supporting a more general way of
1267 specifying character codes in octal. The only supported form is \o{ddd}. */
1268
1269 case CHAR_o:
1270 if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1271 if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1272 {
1273 ptr += 2;
1274 c = 0;
1275 overflow = FALSE;
1276 while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1277 {
1278 register pcre_uint32 cc = *ptr++;
1279 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1280 #ifdef COMPILE_PCRE32
1281 if (c >= 0x20000000l) { overflow = TRUE; break; }
1282 #endif
1283 c = (c << 3) + cc - CHAR_0 ;
1284 #if defined COMPILE_PCRE8
1285 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1286 #elif defined COMPILE_PCRE16
1287 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1288 #elif defined COMPILE_PCRE32
1289 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1290 #endif
1291 }
1292 if (overflow)
1293 {
1294 while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1295 *errorcodeptr = ERR34;
1296 }
1297 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1298 {
1299 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1300 }
1301 else *errorcodeptr = ERR80;
1302 }
1303 break;
1304
1305 /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1306 numbers. Otherwise it is a lowercase x letter. */
1307
1308 case CHAR_x:
1309 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1310 {
1311 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1312 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1313 {
1314 c = 0;
1315 for (i = 0; i < 2; ++i)
1316 {
1317 register pcre_uint32 cc = *(++ptr);
1318 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1319 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1320 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1321 #else /* EBCDIC coding */
1322 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1323 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1324 #endif
1325 }
1326 }
1327 } /* End JavaScript handling */
1328
1329 /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1330 greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1331 digits. If not, { used to be treated as a data character. However, Perl
1332 seems to read hex digits up to the first non-such, and ignore the rest, so
1333 that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1334 now gives an error. */
1335
1336 else
1337 {
1338 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1339 {
1340 ptr += 2;
1341 if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1342 {
1343 *errorcodeptr = ERR86;
1344 break;
1345 }
1346 c = 0;
1347 overflow = FALSE;
1348 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1349 {
1350 register pcre_uint32 cc = *ptr++;
1351 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1352
1353 #ifdef COMPILE_PCRE32
1354 if (c >= 0x10000000l) { overflow = TRUE; break; }
1355 #endif
1356
1357 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1358 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1359 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1360 #else /* EBCDIC coding */
1361 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1362 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1363 #endif
1364
1365 #if defined COMPILE_PCRE8
1366 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1367 #elif defined COMPILE_PCRE16
1368 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1369 #elif defined COMPILE_PCRE32
1370 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1371 #endif
1372 }
1373
1374 if (overflow)
1375 {
1376 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1377 *errorcodeptr = ERR34;
1378 }
1379
1380 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1381 {
1382 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1383 }
1384
1385 /* If the sequence of hex digits does not end with '}', give an error.
1386 We used just to recognize this construct and fall through to the normal
1387 \x handling, but nowadays Perl gives an error, which seems much more
1388 sensible, so we do too. */
1389
1390 else *errorcodeptr = ERR79;
1391 } /* End of \x{} processing */
1392
1393 /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1394
1395 else
1396 {
1397 c = 0;
1398 while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1399 {
1400 pcre_uint32 cc; /* Some compilers don't like */
1401 cc = *(++ptr); /* ++ in initializers */
1402 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1403 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1404 c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1405 #else /* EBCDIC coding */
1406 if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1407 c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1408 #endif
1409 }
1410 } /* End of \xdd handling */
1411 } /* End of Perl-style \x handling */
1412 break;
1413
1414 /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1415 An error is given if the byte following \c is not an ASCII character. This
1416 coding is ASCII-specific, but then the whole concept of \cx is
1417 ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1418
1419 case CHAR_c:
1420 c = *(++ptr);
1421 if (c == CHAR_NULL)
1422 {
1423 *errorcodeptr = ERR2;
1424 break;
1425 }
1426 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1427 if (c > 127) /* Excludes all non-ASCII in either mode */
1428 {
1429 *errorcodeptr = ERR68;
1430 break;
1431 }
1432 if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1433 c ^= 0x40;
1434 #else /* EBCDIC coding */
1435 if (c >= CHAR_a && c <= CHAR_z) c += 64;
1436 c ^= 0xC0;
1437 #endif
1438 break;
1439
1440 /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1441 other alphanumeric following \ is an error if PCRE_EXTRA was set;
1442 otherwise, for Perl compatibility, it is a literal. This code looks a bit
1443 odd, but there used to be some cases other than the default, and there may
1444 be again in future, so I haven't "optimized" it. */
1445
1446 default:
1447 if ((options & PCRE_EXTRA) != 0) switch(c)
1448 {
1449 default:
1450 *errorcodeptr = ERR3;
1451 break;
1452 }
1453 break;
1454 }
1455 }
1456
1457 /* Perl supports \N{name} for character names, as well as plain \N for "not
1458 newline". PCRE does not support \N{name}. However, it does support
1459 quantification such as \N{2,3}. */
1460
1461 if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1462 !is_counted_repeat(ptr+2))
1463 *errorcodeptr = ERR37;
1464
1465 /* If PCRE_UCP is set, we change the values for \d etc. */
1466
1467 if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1468 escape += (ESC_DU - ESC_D);
1469
1470 /* Set the pointer to the final character before returning. */
1471
1472 *ptrptr = ptr;
1473 *chptr = c;
1474 return escape;
1475 }
1476
1477
1478
1479 #ifdef SUPPORT_UCP
1480 /*************************************************
1481 * Handle \P and \p *
1482 *************************************************/
1483
1484 /* This function is called after \P or \p has been encountered, provided that
1485 PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1486 pointing at the P or p. On exit, it is pointing at the final character of the
1487 escape sequence.
1488
1489 Argument:
1490 ptrptr points to the pattern position pointer
1491 negptr points to a boolean that is set TRUE for negation else FALSE
1492 ptypeptr points to an unsigned int that is set to the type value
1493 pdataptr points to an unsigned int that is set to the detailed property value
1494 errorcodeptr points to the error code variable
1495
1496 Returns: TRUE if the type value was found, or FALSE for an invalid type
1497 */
1498
1499 static BOOL
1500 get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1501 unsigned int *pdataptr, int *errorcodeptr)
1502 {
1503 pcre_uchar c;
1504 int i, bot, top;
1505 const pcre_uchar *ptr = *ptrptr;
1506 pcre_uchar name[32];
1507
1508 c = *(++ptr);
1509 if (c == CHAR_NULL) goto ERROR_RETURN;
1510
1511 *negptr = FALSE;
1512
1513 /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1514 negation. */
1515
1516 if (c == CHAR_LEFT_CURLY_BRACKET)
1517 {
1518 if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1519 {
1520 *negptr = TRUE;
1521 ptr++;
1522 }
1523 for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1524 {
1525 c = *(++ptr);
1526 if (c == CHAR_NULL) goto ERROR_RETURN;
1527 if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1528 name[i] = c;
1529 }
1530 if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1531 name[i] = 0;
1532 }
1533
1534 /* Otherwise there is just one following character */
1535
1536 else
1537 {
1538 name[0] = c;
1539 name[1] = 0;
1540 }
1541
1542 *ptrptr = ptr;
1543
1544 /* Search for a recognized property name using binary chop */
1545
1546 bot = 0;
1547 top = PRIV(utt_size);
1548
1549 while (bot < top)
1550 {
1551 int r;
1552 i = (bot + top) >> 1;
1553 r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1554 if (r == 0)
1555 {
1556 *ptypeptr = PRIV(utt)[i].type;
1557 *pdataptr = PRIV(utt)[i].value;
1558 return TRUE;
1559 }
1560 if (r > 0) bot = i + 1; else top = i;
1561 }
1562
1563 *errorcodeptr = ERR47;
1564 *ptrptr = ptr;
1565 return FALSE;
1566
1567 ERROR_RETURN:
1568 *errorcodeptr = ERR46;
1569 *ptrptr = ptr;
1570 return FALSE;
1571 }
1572 #endif
1573
1574
1575
1576 /*************************************************
1577 * Read repeat counts *
1578 *************************************************/
1579
1580 /* Read an item of the form {n,m} and return the values. This is called only
1581 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1582 so the syntax is guaranteed to be correct, but we need to check the values.
1583
1584 Arguments:
1585 p pointer to first char after '{'
1586 minp pointer to int for min
1587 maxp pointer to int for max
1588 returned as -1 if no max
1589 errorcodeptr points to error code variable
1590
1591 Returns: pointer to '}' on success;
1592 current ptr on error, with errorcodeptr set non-zero
1593 */
1594
1595 static const pcre_uchar *
1596 read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1597 {
1598 int min = 0;
1599 int max = -1;
1600
1601 while (IS_DIGIT(*p))
1602 {
1603 min = min * 10 + (int)(*p++ - CHAR_0);
1604 if (min > 65535)
1605 {
1606 *errorcodeptr = ERR5;
1607 return p;
1608 }
1609 }
1610
1611 if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1612 {
1613 if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1614 {
1615 max = 0;
1616 while(IS_DIGIT(*p))
1617 {
1618 max = max * 10 + (int)(*p++ - CHAR_0);
1619 if (max > 65535)
1620 {
1621 *errorcodeptr = ERR5;
1622 return p;
1623 }
1624 }
1625 if (max < min)
1626 {
1627 *errorcodeptr = ERR4;
1628 return p;
1629 }
1630 }
1631 }
1632
1633 *minp = min;
1634 *maxp = max;
1635 return p;
1636 }
1637
1638
1639
1640 /*************************************************
1641 * Find first significant op code *
1642 *************************************************/
1643
1644 /* This is called by several functions that scan a compiled expression looking
1645 for a fixed first character, or an anchoring op code etc. It skips over things
1646 that do not influence this. For some calls, it makes sense to skip negative
1647 forward and all backward assertions, and also the \b assertion; for others it
1648 does not.
1649
1650 Arguments:
1651 code pointer to the start of the group
1652 skipassert TRUE if certain assertions are to be skipped
1653
1654 Returns: pointer to the first significant opcode
1655 */
1656
1657 static const pcre_uchar*
1658 first_significant_code(const pcre_uchar *code, BOOL skipassert)
1659 {
1660 for (;;)
1661 {
1662 switch ((int)*code)
1663 {
1664 case OP_ASSERT_NOT:
1665 case OP_ASSERTBACK:
1666 case OP_ASSERTBACK_NOT:
1667 if (!skipassert) return code;
1668 do code += GET(code, 1); while (*code == OP_ALT);
1669 code += PRIV(OP_lengths)[*code];
1670 break;
1671
1672 case OP_WORD_BOUNDARY:
1673 case OP_NOT_WORD_BOUNDARY:
1674 if (!skipassert) return code;
1675 /* Fall through */
1676
1677 case OP_CALLOUT:
1678 case OP_CREF:
1679 case OP_DNCREF:
1680 case OP_RREF:
1681 case OP_DNRREF:
1682 case OP_DEF:
1683 code += PRIV(OP_lengths)[*code];
1684 break;
1685
1686 default:
1687 return code;
1688 }
1689 }
1690 /* Control never reaches here */
1691 }
1692
1693
1694
1695 /*************************************************
1696 * Find the fixed length of a branch *
1697 *************************************************/
1698
1699 /* Scan a branch and compute the fixed length of subject that will match it,
1700 if the length is fixed. This is needed for dealing with backward assertions.
1701 In UTF8 mode, the result is in characters rather than bytes. The branch is
1702 temporarily terminated with OP_END when this function is called.
1703
1704 This function is called when a backward assertion is encountered, so that if it
1705 fails, the error message can point to the correct place in the pattern.
1706 However, we cannot do this when the assertion contains subroutine calls,
1707 because they can be forward references. We solve this by remembering this case
1708 and doing the check at the end; a flag specifies which mode we are running in.
1709
1710 Arguments:
1711 code points to the start of the pattern (the bracket)
1712 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1713 atend TRUE if called when the pattern is complete
1714 cd the "compile data" structure
1715 recurses chain of recurse_check to catch mutual recursion
1716
1717 Returns: the fixed length,
1718 or -1 if there is no fixed length,
1719 or -2 if \C was encountered (in UTF-8 mode only)
1720 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1721 or -4 if an unknown opcode was encountered (internal error)
1722 */
1723
1724 static int
1725 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd,
1726 recurse_check *recurses)
1727 {
1728 int length = -1;
1729 recurse_check this_recurse;
1730 register int branchlength = 0;
1731 register pcre_uchar *cc = code + 1 + LINK_SIZE;
1732
1733 /* Scan along the opcodes for this branch. If we get to the end of the
1734 branch, check the length against that of the other branches. */
1735
1736 for (;;)
1737 {
1738 int d;
1739 pcre_uchar *ce, *cs;
1740 register pcre_uchar op = *cc;
1741
1742 switch (op)
1743 {
1744 /* We only need to continue for OP_CBRA (normal capturing bracket) and
1745 OP_BRA (normal non-capturing bracket) because the other variants of these
1746 opcodes are all concerned with unlimited repeated groups, which of course
1747 are not of fixed length. */
1748
1749 case OP_CBRA:
1750 case OP_BRA:
1751 case OP_ONCE:
1752 case OP_ONCE_NC:
1753 case OP_COND:
1754 d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd,
1755 recurses);
1756 if (d < 0) return d;
1757 branchlength += d;
1758 do cc += GET(cc, 1); while (*cc == OP_ALT);
1759 cc += 1 + LINK_SIZE;
1760 break;
1761
1762 /* Reached end of a branch; if it's a ket it is the end of a nested call.
1763 If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1764 an ALT. If it is END it's the end of the outer call. All can be handled by
1765 the same code. Note that we must not include the OP_KETRxxx opcodes here,
1766 because they all imply an unlimited repeat. */
1767
1768 case OP_ALT:
1769 case OP_KET:
1770 case OP_END:
1771 case OP_ACCEPT:
1772 case OP_ASSERT_ACCEPT:
1773 if (length < 0) length = branchlength;
1774 else if (length != branchlength) return -1;
1775 if (*cc != OP_ALT) return length;
1776 cc += 1 + LINK_SIZE;
1777 branchlength = 0;
1778 break;
1779
1780 /* A true recursion implies not fixed length, but a subroutine call may
1781 be OK. If the subroutine is a forward reference, we can't deal with
1782 it until the end of the pattern, so return -3. */
1783
1784 case OP_RECURSE:
1785 if (!atend) return -3;
1786 cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1787 do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1788 if (cc > cs && cc < ce) return -1; /* Recursion */
1789 else /* Check for mutual recursion */
1790 {
1791 recurse_check *r = recurses;
1792 for (r = recurses; r != NULL; r = r->prev) if (r->group == cs) break;
1793 if (r != NULL) return -1; /* Mutual recursion */
1794 }
1795 this_recurse.prev = recurses;
1796 this_recurse.group = cs;
1797 d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd, &this_recurse);
1798 if (d < 0) return d;
1799 branchlength += d;
1800 cc += 1 + LINK_SIZE;
1801 break;
1802
1803 /* Skip over assertive subpatterns */
1804
1805 case OP_ASSERT:
1806 case OP_ASSERT_NOT:
1807 case OP_ASSERTBACK:
1808 case OP_ASSERTBACK_NOT:
1809 do cc += GET(cc, 1); while (*cc == OP_ALT);
1810 cc += PRIV(OP_lengths)[*cc];
1811 break;
1812
1813 /* Skip over things that don't match chars */
1814
1815 case OP_MARK:
1816 case OP_PRUNE_ARG:
1817 case OP_SKIP_ARG:
1818 case OP_THEN_ARG:
1819 cc += cc[1] + PRIV(OP_lengths)[*cc];
1820 break;
1821
1822 case OP_CALLOUT:
1823 case OP_CIRC:
1824 case OP_CIRCM:
1825 case OP_CLOSE:
1826 case OP_COMMIT:
1827 case OP_CREF:
1828 case OP_DEF:
1829 case OP_DNCREF:
1830 case OP_DNRREF:
1831 case OP_DOLL:
1832 case OP_DOLLM:
1833 case OP_EOD:
1834 case OP_EODN:
1835 case OP_FAIL:
1836 case OP_NOT_WORD_BOUNDARY:
1837 case OP_PRUNE:
1838 case OP_REVERSE:
1839 case OP_RREF:
1840 case OP_SET_SOM:
1841 case OP_SKIP:
1842 case OP_SOD:
1843 case OP_SOM:
1844 case OP_THEN:
1845 case OP_WORD_BOUNDARY:
1846 cc += PRIV(OP_lengths)[*cc];
1847 break;
1848
1849 /* Handle literal characters */
1850
1851 case OP_CHAR:
1852 case OP_CHARI:
1853 case OP_NOT:
1854 case OP_NOTI:
1855 branchlength++;
1856 cc += 2;
1857 #ifdef SUPPORT_UTF
1858 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1859 #endif
1860 break;
1861
1862 /* Handle exact repetitions. The count is already in characters, but we
1863 need to skip over a multibyte character in UTF8 mode. */
1864
1865 case OP_EXACT:
1866 case OP_EXACTI:
1867 case OP_NOTEXACT:
1868 case OP_NOTEXACTI:
1869 branchlength += (int)GET2(cc,1);
1870 cc += 2 + IMM2_SIZE;
1871 #ifdef SUPPORT_UTF
1872 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1873 #endif
1874 break;
1875
1876 case OP_TYPEEXACT:
1877 branchlength += GET2(cc,1);
1878 if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1879 cc += 2;
1880 cc += 1 + IMM2_SIZE + 1;
1881 break;
1882
1883 /* Handle single-char matchers */
1884
1885 case OP_PROP:
1886 case OP_NOTPROP:
1887 cc += 2;
1888 /* Fall through */
1889
1890 case OP_HSPACE:
1891 case OP_VSPACE:
1892 case OP_NOT_HSPACE:
1893 case OP_NOT_VSPACE:
1894 case OP_NOT_DIGIT:
1895 case OP_DIGIT:
1896 case OP_NOT_WHITESPACE:
1897 case OP_WHITESPACE:
1898 case OP_NOT_WORDCHAR:
1899 case OP_WORDCHAR:
1900 case OP_ANY:
1901 case OP_ALLANY:
1902 branchlength++;
1903 cc++;
1904 break;
1905
1906 /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1907 otherwise \C is coded as OP_ALLANY. */
1908
1909 case OP_ANYBYTE:
1910 return -2;
1911
1912 /* Check a class for variable quantification */
1913
1914 case OP_CLASS:
1915 case OP_NCLASS:
1916 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1917 case OP_XCLASS:
1918 /* The original code caused an unsigned overflow in 64 bit systems,
1919 so now we use a conditional statement. */
1920 if (op == OP_XCLASS)
1921 cc += GET(cc, 1);
1922 else
1923 cc += PRIV(OP_lengths)[OP_CLASS];
1924 #else
1925 cc += PRIV(OP_lengths)[OP_CLASS];
1926 #endif
1927
1928 switch (*cc)
1929 {
1930 case OP_CRSTAR:
1931 case OP_CRMINSTAR:
1932 case OP_CRPLUS:
1933 case OP_CRMINPLUS:
1934 case OP_CRQUERY:
1935 case OP_CRMINQUERY:
1936 case OP_CRPOSSTAR:
1937 case OP_CRPOSPLUS:
1938 case OP_CRPOSQUERY:
1939 return -1;
1940
1941 case OP_CRRANGE:
1942 case OP_CRMINRANGE:
1943 case OP_CRPOSRANGE:
1944 if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1945 branchlength += (int)GET2(cc,1);
1946 cc += 1 + 2 * IMM2_SIZE;
1947 break;
1948
1949 default:
1950 branchlength++;
1951 }
1952 break;
1953
1954 /* Anything else is variable length */
1955
1956 case OP_ANYNL:
1957 case OP_BRAMINZERO:
1958 case OP_BRAPOS:
1959 case OP_BRAPOSZERO:
1960 case OP_BRAZERO:
1961 case OP_CBRAPOS:
1962 case OP_EXTUNI:
1963 case OP_KETRMAX:
1964 case OP_KETRMIN:
1965 case OP_KETRPOS:
1966 case OP_MINPLUS:
1967 case OP_MINPLUSI:
1968 case OP_MINQUERY:
1969 case OP_MINQUERYI:
1970 case OP_MINSTAR:
1971 case OP_MINSTARI:
1972 case OP_MINUPTO:
1973 case OP_MINUPTOI:
1974 case OP_NOTMINPLUS:
1975 case OP_NOTMINPLUSI:
1976 case OP_NOTMINQUERY:
1977 case OP_NOTMINQUERYI:
1978 case OP_NOTMINSTAR:
1979 case OP_NOTMINSTARI:
1980 case OP_NOTMINUPTO:
1981 case OP_NOTMINUPTOI:
1982 case OP_NOTPLUS:
1983 case OP_NOTPLUSI:
1984 case OP_NOTPOSPLUS:
1985 case OP_NOTPOSPLUSI:
1986 case OP_NOTPOSQUERY:
1987 case OP_NOTPOSQUERYI:
1988 case OP_NOTPOSSTAR:
1989 case OP_NOTPOSSTARI:
1990 case OP_NOTPOSUPTO:
1991 case OP_NOTPOSUPTOI:
1992 case OP_NOTQUERY:
1993 case OP_NOTQUERYI:
1994 case OP_NOTSTAR:
1995 case OP_NOTSTARI:
1996 case OP_NOTUPTO:
1997 case OP_NOTUPTOI:
1998 case OP_PLUS:
1999 case OP_PLUSI:
2000 case OP_POSPLUS:
2001 case OP_POSPLUSI:
2002 case OP_POSQUERY:
2003 case OP_POSQUERYI:
2004 case OP_POSSTAR:
2005 case OP_POSSTARI:
2006 case OP_POSUPTO:
2007 case OP_POSUPTOI:
2008 case OP_QUERY:
2009 case OP_QUERYI:
2010 case OP_REF:
2011 case OP_REFI:
2012 case OP_DNREF:
2013 case OP_DNREFI:
2014 case OP_SBRA:
2015 case OP_SBRAPOS:
2016 case OP_SCBRA:
2017 case OP_SCBRAPOS:
2018 case OP_SCOND:
2019 case OP_SKIPZERO:
2020 case OP_STAR:
2021 case OP_STARI:
2022 case OP_TYPEMINPLUS:
2023 case OP_TYPEMINQUERY:
2024 case OP_TYPEMINSTAR:
2025 case OP_TYPEMINUPTO:
2026 case OP_TYPEPLUS:
2027 case OP_TYPEPOSPLUS:
2028 case OP_TYPEPOSQUERY:
2029 case OP_TYPEPOSSTAR:
2030 case OP_TYPEPOSUPTO:
2031 case OP_TYPEQUERY:
2032 case OP_TYPESTAR:
2033 case OP_TYPEUPTO:
2034 case OP_UPTO:
2035 case OP_UPTOI:
2036 return -1;
2037
2038 /* Catch unrecognized opcodes so that when new ones are added they
2039 are not forgotten, as has happened in the past. */
2040
2041 default:
2042 return -4;
2043 }
2044 }
2045 /* Control never gets here */
2046 }
2047
2048
2049
2050 /*************************************************
2051 * Scan compiled regex for specific bracket *
2052 *************************************************/
2053
2054 /* This little function scans through a compiled pattern until it finds a
2055 capturing bracket with the given number, or, if the number is negative, an
2056 instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2057 so that it can be called from pcre_study() when finding the minimum matching
2058 length.
2059
2060 Arguments:
2061 code points to start of expression
2062 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2063 number the required bracket number or negative to find a lookbehind
2064
2065 Returns: pointer to the opcode for the bracket, or NULL if not found
2066 */
2067
2068 const pcre_uchar *
2069 PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2070 {
2071 for (;;)
2072 {
2073 register pcre_uchar c = *code;
2074
2075 if (c == OP_END) return NULL;
2076
2077 /* XCLASS is used for classes that cannot be represented just by a bit
2078 map. This includes negated single high-valued characters. The length in
2079 the table is zero; the actual length is stored in the compiled code. */
2080
2081 if (c == OP_XCLASS) code += GET(code, 1);
2082
2083 /* Handle recursion */
2084
2085 else if (c == OP_REVERSE)
2086 {
2087 if (number < 0) return (pcre_uchar *)code;
2088 code += PRIV(OP_lengths)[c];
2089 }
2090
2091 /* Handle capturing bracket */
2092
2093 else if (c == OP_CBRA || c == OP_SCBRA ||
2094 c == OP_CBRAPOS || c == OP_SCBRAPOS)
2095 {
2096 int n = (int)GET2(code, 1+LINK_SIZE);
2097 if (n == number) return (pcre_uchar *)code;
2098 code += PRIV(OP_lengths)[c];
2099 }
2100
2101 /* Otherwise, we can get the item's length from the table, except that for
2102 repeated character types, we have to test for \p and \P, which have an extra
2103 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2104 must add in its length. */
2105
2106 else
2107 {
2108 switch(c)
2109 {
2110 case OP_TYPESTAR:
2111 case OP_TYPEMINSTAR:
2112 case OP_TYPEPLUS:
2113 case OP_TYPEMINPLUS:
2114 case OP_TYPEQUERY:
2115 case OP_TYPEMINQUERY:
2116 case OP_TYPEPOSSTAR:
2117 case OP_TYPEPOSPLUS:
2118 case OP_TYPEPOSQUERY:
2119 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2120 break;
2121
2122 case OP_TYPEUPTO:
2123 case OP_TYPEMINUPTO:
2124 case OP_TYPEEXACT:
2125 case OP_TYPEPOSUPTO:
2126 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2127 code += 2;
2128 break;
2129
2130 case OP_MARK:
2131 case OP_PRUNE_ARG:
2132 case OP_SKIP_ARG:
2133 case OP_THEN_ARG:
2134 code += code[1];
2135 break;
2136 }
2137
2138 /* Add in the fixed length from the table */
2139
2140 code += PRIV(OP_lengths)[c];
2141
2142 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2143 a multi-byte character. The length in the table is a minimum, so we have to
2144 arrange to skip the extra bytes. */
2145
2146 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2147 if (utf) switch(c)
2148 {
2149 case OP_CHAR:
2150 case OP_CHARI:
2151 case OP_NOT:
2152 case OP_NOTI:
2153 case OP_EXACT:
2154 case OP_EXACTI:
2155 case OP_NOTEXACT:
2156 case OP_NOTEXACTI:
2157 case OP_UPTO:
2158 case OP_UPTOI:
2159 case OP_NOTUPTO:
2160 case OP_NOTUPTOI:
2161 case OP_MINUPTO:
2162 case OP_MINUPTOI:
2163 case OP_NOTMINUPTO:
2164 case OP_NOTMINUPTOI:
2165 case OP_POSUPTO:
2166 case OP_POSUPTOI:
2167 case OP_NOTPOSUPTO:
2168 case OP_NOTPOSUPTOI:
2169 case OP_STAR:
2170 case OP_STARI:
2171 case OP_NOTSTAR:
2172 case OP_NOTSTARI:
2173 case OP_MINSTAR:
2174 case OP_MINSTARI:
2175 case OP_NOTMINSTAR:
2176 case OP_NOTMINSTARI:
2177 case OP_POSSTAR:
2178 case OP_POSSTARI:
2179 case OP_NOTPOSSTAR:
2180 case OP_NOTPOSSTARI:
2181 case OP_PLUS:
2182 case OP_PLUSI:
2183 case OP_NOTPLUS:
2184 case OP_NOTPLUSI:
2185 case OP_MINPLUS:
2186 case OP_MINPLUSI:
2187 case OP_NOTMINPLUS:
2188 case OP_NOTMINPLUSI:
2189 case OP_POSPLUS:
2190 case OP_POSPLUSI:
2191 case OP_NOTPOSPLUS:
2192 case OP_NOTPOSPLUSI:
2193 case OP_QUERY:
2194 case OP_QUERYI:
2195 case OP_NOTQUERY:
2196 case OP_NOTQUERYI:
2197 case OP_MINQUERY:
2198 case OP_MINQUERYI:
2199 case OP_NOTMINQUERY:
2200 case OP_NOTMINQUERYI:
2201 case OP_POSQUERY:
2202 case OP_POSQUERYI:
2203 case OP_NOTPOSQUERY:
2204 case OP_NOTPOSQUERYI:
2205 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2206 break;
2207 }
2208 #else
2209 (void)(utf); /* Keep compiler happy by referencing function argument */
2210 #endif
2211 }
2212 }
2213 }
2214
2215
2216
2217 /*************************************************
2218 * Scan compiled regex for recursion reference *
2219 *************************************************/
2220
2221 /* This little function scans through a compiled pattern until it finds an
2222 instance of OP_RECURSE.
2223
2224 Arguments:
2225 code points to start of expression
2226 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2227
2228 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2229 */
2230
2231 static const pcre_uchar *
2232 find_recurse(const pcre_uchar *code, BOOL utf)
2233 {
2234 for (;;)
2235 {
2236 register pcre_uchar c = *code;
2237 if (c == OP_END) return NULL;
2238 if (c == OP_RECURSE) return code;
2239
2240 /* XCLASS is used for classes that cannot be represented just by a bit
2241 map. This includes negated single high-valued characters. The length in
2242 the table is zero; the actual length is stored in the compiled code. */
2243
2244 if (c == OP_XCLASS) code += GET(code, 1);
2245
2246 /* Otherwise, we can get the item's length from the table, except that for
2247 repeated character types, we have to test for \p and \P, which have an extra
2248 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2249 must add in its length. */
2250
2251 else
2252 {
2253 switch(c)
2254 {
2255 case OP_TYPESTAR:
2256 case OP_TYPEMINSTAR:
2257 case OP_TYPEPLUS:
2258 case OP_TYPEMINPLUS:
2259 case OP_TYPEQUERY:
2260 case OP_TYPEMINQUERY:
2261 case OP_TYPEPOSSTAR:
2262 case OP_TYPEPOSPLUS:
2263 case OP_TYPEPOSQUERY:
2264 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2265 break;
2266
2267 case OP_TYPEPOSUPTO:
2268 case OP_TYPEUPTO:
2269 case OP_TYPEMINUPTO:
2270 case OP_TYPEEXACT:
2271 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2272 code += 2;
2273 break;
2274
2275 case OP_MARK:
2276 case OP_PRUNE_ARG:
2277 case OP_SKIP_ARG:
2278 case OP_THEN_ARG:
2279 code += code[1];
2280 break;
2281 }
2282
2283 /* Add in the fixed length from the table */
2284
2285 code += PRIV(OP_lengths)[c];
2286
2287 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2288 by a multi-byte character. The length in the table is a minimum, so we have
2289 to arrange to skip the extra bytes. */
2290
2291 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2292 if (utf) switch(c)
2293 {
2294 case OP_CHAR:
2295 case OP_CHARI:
2296 case OP_NOT:
2297 case OP_NOTI:
2298 case OP_EXACT:
2299 case OP_EXACTI:
2300 case OP_NOTEXACT:
2301 case OP_NOTEXACTI:
2302 case OP_UPTO:
2303 case OP_UPTOI:
2304 case OP_NOTUPTO:
2305 case OP_NOTUPTOI:
2306 case OP_MINUPTO:
2307 case OP_MINUPTOI:
2308 case OP_NOTMINUPTO:
2309 case OP_NOTMINUPTOI:
2310 case OP_POSUPTO:
2311 case OP_POSUPTOI:
2312 case OP_NOTPOSUPTO:
2313 case OP_NOTPOSUPTOI:
2314 case OP_STAR:
2315 case OP_STARI:
2316 case OP_NOTSTAR:
2317 case OP_NOTSTARI:
2318 case OP_MINSTAR:
2319 case OP_MINSTARI:
2320 case OP_NOTMINSTAR:
2321 case OP_NOTMINSTARI:
2322 case OP_POSSTAR:
2323 case OP_POSSTARI:
2324 case OP_NOTPOSSTAR:
2325 case OP_NOTPOSSTARI:
2326 case OP_PLUS:
2327 case OP_PLUSI:
2328 case OP_NOTPLUS:
2329 case OP_NOTPLUSI:
2330 case OP_MINPLUS:
2331 case OP_MINPLUSI:
2332 case OP_NOTMINPLUS:
2333 case OP_NOTMINPLUSI:
2334 case OP_POSPLUS:
2335 case OP_POSPLUSI:
2336 case OP_NOTPOSPLUS:
2337 case OP_NOTPOSPLUSI:
2338 case OP_QUERY:
2339 case OP_QUERYI:
2340 case OP_NOTQUERY:
2341 case OP_NOTQUERYI:
2342 case OP_MINQUERY:
2343 case OP_MINQUERYI:
2344 case OP_NOTMINQUERY:
2345 case OP_NOTMINQUERYI:
2346 case OP_POSQUERY:
2347 case OP_POSQUERYI:
2348 case OP_NOTPOSQUERY:
2349 case OP_NOTPOSQUERYI:
2350 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2351 break;
2352 }
2353 #else
2354 (void)(utf); /* Keep compiler happy by referencing function argument */
2355 #endif
2356 }
2357 }
2358 }
2359
2360
2361
2362 /*************************************************
2363 * Scan compiled branch for non-emptiness *
2364 *************************************************/
2365
2366 /* This function scans through a branch of a compiled pattern to see whether it
2367 can match the empty string or not. It is called from could_be_empty()
2368 below and from compile_branch() when checking for an unlimited repeat of a
2369 group that can match nothing. Note that first_significant_code() skips over
2370 backward and negative forward assertions when its final argument is TRUE. If we
2371 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2372 bracket whose current branch will already have been scanned.
2373
2374 Arguments:
2375 code points to start of search
2376 endcode points to where to stop
2377 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2378 cd contains pointers to tables etc.
2379 recurses chain of recurse_check to catch mutual recursion
2380
2381 Returns: TRUE if what is matched could be empty
2382 */
2383
2384 static BOOL
2385 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2386 BOOL utf, compile_data *cd, recurse_check *recurses)
2387 {
2388 register pcre_uchar c;
2389 recurse_check this_recurse;
2390
2391 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2392 code < endcode;
2393 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2394 {
2395 const pcre_uchar *ccode;
2396
2397 c = *code;
2398
2399 /* Skip over forward assertions; the other assertions are skipped by
2400 first_significant_code() with a TRUE final argument. */
2401
2402 if (c == OP_ASSERT)
2403 {
2404 do code += GET(code, 1); while (*code == OP_ALT);
2405 c = *code;
2406 continue;
2407 }
2408
2409 /* For a recursion/subroutine call, if its end has been reached, which
2410 implies a backward reference subroutine call, we can scan it. If it's a
2411 forward reference subroutine call, we can't. To detect forward reference
2412 we have to scan up the list that is kept in the workspace. This function is
2413 called only when doing the real compile, not during the pre-compile that
2414 measures the size of the compiled pattern. */
2415
2416 if (c == OP_RECURSE)
2417 {
2418 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2419 const pcre_uchar *endgroup = scode;
2420 BOOL empty_branch;
2421
2422 /* Test for forward reference or uncompleted reference. This is disabled
2423 when called to scan a completed pattern by setting cd->start_workspace to
2424 NULL. */
2425
2426 if (cd->start_workspace != NULL)
2427 {
2428 const pcre_uchar *tcode;
2429 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2430 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2431 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2432 }
2433
2434 /* If the reference is to a completed group, we need to detect whether this
2435 is a recursive call, as otherwise there will be an infinite loop. If it is
2436 a recursion, just skip over it. Simple recursions are easily detected. For
2437 mutual recursions we keep a chain on the stack. */
2438
2439 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2440 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2441 else
2442 {
2443 recurse_check *r = recurses;
2444 for (r = recurses; r != NULL; r = r->prev)
2445 if (r->group == scode) break;
2446 if (r != NULL) continue; /* Mutual recursion */
2447 }
2448
2449 /* Completed reference; scan the referenced group, remembering it on the
2450 stack chain to detect mutual recursions. */
2451
2452 empty_branch = FALSE;
2453 this_recurse.prev = recurses;
2454 this_recurse.group = scode;
2455
2456 do
2457 {
2458 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2459 {
2460 empty_branch = TRUE;
2461 break;
2462 }
2463 scode += GET(scode, 1);
2464 }
2465 while (*scode == OP_ALT);
2466
2467 if (!empty_branch) return FALSE; /* All branches are non-empty */
2468 continue;
2469 }
2470
2471 /* Groups with zero repeats can of course be empty; skip them. */
2472
2473 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2474 c == OP_BRAPOSZERO)
2475 {
2476 code += PRIV(OP_lengths)[c];
2477 do code += GET(code, 1); while (*code == OP_ALT);
2478 c = *code;
2479 continue;
2480 }
2481
2482 /* A nested group that is already marked as "could be empty" can just be
2483 skipped. */
2484
2485 if (c == OP_SBRA || c == OP_SBRAPOS ||
2486 c == OP_SCBRA || c == OP_SCBRAPOS)
2487 {
2488 do code += GET(code, 1); while (*code == OP_ALT);
2489 c = *code;
2490 continue;
2491 }
2492
2493 /* For other groups, scan the branches. */
2494
2495 if (c == OP_BRA || c == OP_BRAPOS ||
2496 c == OP_CBRA || c == OP_CBRAPOS ||
2497 c == OP_ONCE || c == OP_ONCE_NC ||
2498 c == OP_COND)
2499 {
2500 BOOL empty_branch;
2501 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2502
2503 /* If a conditional group has only one branch, there is a second, implied,
2504 empty branch, so just skip over the conditional, because it could be empty.
2505 Otherwise, scan the individual branches of the group. */
2506
2507 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2508 code += GET(code, 1);
2509 else
2510 {
2511 empty_branch = FALSE;
2512 do
2513 {
2514 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd,
2515 recurses)) empty_branch = TRUE;
2516 code += GET(code, 1);
2517 }
2518 while (*code == OP_ALT);
2519 if (!empty_branch) return FALSE; /* All branches are non-empty */
2520 }
2521
2522 c = *code;
2523 continue;
2524 }
2525
2526 /* Handle the other opcodes */
2527
2528 switch (c)
2529 {
2530 /* Check for quantifiers after a class. XCLASS is used for classes that
2531 cannot be represented just by a bit map. This includes negated single
2532 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2533 actual length is stored in the compiled code, so we must update "code"
2534 here. */
2535
2536 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2537 case OP_XCLASS:
2538 ccode = code += GET(code, 1);
2539 goto CHECK_CLASS_REPEAT;
2540 #endif
2541
2542 case OP_CLASS:
2543 case OP_NCLASS:
2544 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2545
2546 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2547 CHECK_CLASS_REPEAT:
2548 #endif
2549
2550 switch (*ccode)
2551 {
2552 case OP_CRSTAR: /* These could be empty; continue */
2553 case OP_CRMINSTAR:
2554 case OP_CRQUERY:
2555 case OP_CRMINQUERY:
2556 case OP_CRPOSSTAR:
2557 case OP_CRPOSQUERY:
2558 break;
2559
2560 default: /* Non-repeat => class must match */
2561 case OP_CRPLUS: /* These repeats aren't empty */
2562 case OP_CRMINPLUS:
2563 case OP_CRPOSPLUS:
2564 return FALSE;
2565
2566 case OP_CRRANGE:
2567 case OP_CRMINRANGE:
2568 case OP_CRPOSRANGE:
2569 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2570 break;
2571 }
2572 break;
2573
2574 /* Opcodes that must match a character */
2575
2576 case OP_ANY:
2577 case OP_ALLANY:
2578 case OP_ANYBYTE:
2579
2580 case OP_PROP:
2581 case OP_NOTPROP:
2582 case OP_ANYNL:
2583
2584 case OP_NOT_HSPACE:
2585 case OP_HSPACE:
2586 case OP_NOT_VSPACE:
2587 case OP_VSPACE:
2588 case OP_EXTUNI:
2589
2590 case OP_NOT_DIGIT:
2591 case OP_DIGIT:
2592 case OP_NOT_WHITESPACE:
2593 case OP_WHITESPACE:
2594 case OP_NOT_WORDCHAR:
2595 case OP_WORDCHAR:
2596
2597 case OP_CHAR:
2598 case OP_CHARI:
2599 case OP_NOT:
2600 case OP_NOTI:
2601
2602 case OP_PLUS:
2603 case OP_PLUSI:
2604 case OP_MINPLUS:
2605 case OP_MINPLUSI:
2606
2607 case OP_NOTPLUS:
2608 case OP_NOTPLUSI:
2609 case OP_NOTMINPLUS:
2610 case OP_NOTMINPLUSI:
2611
2612 case OP_POSPLUS:
2613 case OP_POSPLUSI:
2614 case OP_NOTPOSPLUS:
2615 case OP_NOTPOSPLUSI:
2616
2617 case OP_EXACT:
2618 case OP_EXACTI:
2619 case OP_NOTEXACT:
2620 case OP_NOTEXACTI:
2621
2622 case OP_TYPEPLUS:
2623 case OP_TYPEMINPLUS:
2624 case OP_TYPEPOSPLUS:
2625 case OP_TYPEEXACT:
2626
2627 return FALSE;
2628
2629 /* These are going to continue, as they may be empty, but we have to
2630 fudge the length for the \p and \P cases. */
2631
2632 case OP_TYPESTAR:
2633 case OP_TYPEMINSTAR:
2634 case OP_TYPEPOSSTAR:
2635 case OP_TYPEQUERY:
2636 case OP_TYPEMINQUERY:
2637 case OP_TYPEPOSQUERY:
2638 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2639 break;
2640
2641 /* Same for these */
2642
2643 case OP_TYPEUPTO:
2644 case OP_TYPEMINUPTO:
2645 case OP_TYPEPOSUPTO:
2646 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2647 code += 2;
2648 break;
2649
2650 /* End of branch */
2651
2652 case OP_KET:
2653 case OP_KETRMAX:
2654 case OP_KETRMIN:
2655 case OP_KETRPOS:
2656 case OP_ALT:
2657 return TRUE;
2658
2659 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2660 MINUPTO, and POSUPTO and their caseless and negative versions may be
2661 followed by a multibyte character. */
2662
2663 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2664 case OP_STAR:
2665 case OP_STARI:
2666 case OP_NOTSTAR:
2667 case OP_NOTSTARI:
2668
2669 case OP_MINSTAR:
2670 case OP_MINSTARI:
2671 case OP_NOTMINSTAR:
2672 case OP_NOTMINSTARI:
2673
2674 case OP_POSSTAR:
2675 case OP_POSSTARI:
2676 case OP_NOTPOSSTAR:
2677 case OP_NOTPOSSTARI:
2678
2679 case OP_QUERY:
2680 case OP_QUERYI:
2681 case OP_NOTQUERY:
2682 case OP_NOTQUERYI:
2683
2684 case OP_MINQUERY:
2685 case OP_MINQUERYI:
2686 case OP_NOTMINQUERY:
2687 case OP_NOTMINQUERYI:
2688
2689 case OP_POSQUERY:
2690 case OP_POSQUERYI:
2691 case OP_NOTPOSQUERY:
2692 case OP_NOTPOSQUERYI:
2693
2694 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2695 break;
2696
2697 case OP_UPTO:
2698 case OP_UPTOI:
2699 case OP_NOTUPTO:
2700 case OP_NOTUPTOI:
2701
2702 case OP_MINUPTO:
2703 case OP_MINUPTOI:
2704 case OP_NOTMINUPTO:
2705 case OP_NOTMINUPTOI:
2706
2707 case OP_POSUPTO:
2708 case OP_POSUPTOI:
2709 case OP_NOTPOSUPTO:
2710 case OP_NOTPOSUPTOI:
2711
2712 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2713 break;
2714 #endif
2715
2716 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2717 string. */
2718
2719 case OP_MARK:
2720 case OP_PRUNE_ARG:
2721 case OP_SKIP_ARG:
2722 case OP_THEN_ARG:
2723 code += code[1];
2724 break;
2725
2726 /* None of the remaining opcodes are required to match a character. */
2727
2728 default:
2729 break;
2730 }
2731 }
2732
2733 return TRUE;
2734 }
2735
2736
2737
2738 /*************************************************
2739 * Scan compiled regex for non-emptiness *
2740 *************************************************/
2741
2742 /* This function is called to check for left recursive calls. We want to check
2743 the current branch of the current pattern to see if it could match the empty
2744 string. If it could, we must look outwards for branches at other levels,
2745 stopping when we pass beyond the bracket which is the subject of the recursion.
2746 This function is called only during the real compile, not during the
2747 pre-compile.
2748
2749 Arguments:
2750 code points to start of the recursion
2751 endcode points to where to stop (current RECURSE item)
2752 bcptr points to the chain of current (unclosed) branch starts
2753 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2754 cd pointers to tables etc
2755
2756 Returns: TRUE if what is matched could be empty
2757 */
2758
2759 static BOOL
2760 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2761 branch_chain *bcptr, BOOL utf, compile_data *cd)
2762 {
2763 while (bcptr != NULL && bcptr->current_branch >= code)
2764 {
2765 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2766 return FALSE;
2767 bcptr = bcptr->outer;
2768 }
2769 return TRUE;
2770 }
2771
2772
2773
2774 /*************************************************
2775 * Base opcode of repeated opcodes *
2776 *************************************************/
2777
2778 /* Returns the base opcode for repeated single character type opcodes. If the
2779 opcode is not a repeated character type, it returns with the original value.
2780
2781 Arguments: c opcode
2782 Returns: base opcode for the type
2783 */
2784
2785 static pcre_uchar
2786 get_repeat_base(pcre_uchar c)
2787 {
2788 return (c > OP_TYPEPOSUPTO)? c :
2789 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2790 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2791 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2792 (c >= OP_STARI)? OP_STARI :
2793 OP_STAR;
2794 }
2795
2796
2797
2798 #ifdef SUPPORT_UCP
2799 /*************************************************
2800 * Check a character and a property *
2801 *************************************************/
2802
2803 /* This function is called by check_auto_possessive() when a property item
2804 is adjacent to a fixed character.
2805
2806 Arguments:
2807 c the character
2808 ptype the property type
2809 pdata the data for the type
2810 negated TRUE if it's a negated property (\P or \p{^)
2811
2812 Returns: TRUE if auto-possessifying is OK
2813 */
2814
2815 static BOOL
2816 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2817 BOOL negated)
2818 {
2819 const pcre_uint32 *p;
2820 const ucd_record *prop = GET_UCD(c);
2821
2822 switch(ptype)
2823 {
2824 case PT_LAMP:
2825 return (prop->chartype == ucp_Lu ||
2826 prop->chartype == ucp_Ll ||
2827 prop->chartype == ucp_Lt) == negated;
2828
2829 case PT_GC:
2830 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2831
2832 case PT_PC:
2833 return (pdata == prop->chartype) == negated;
2834
2835 case PT_SC:
2836 return (pdata == prop->script) == negated;
2837
2838 /* These are specials */
2839
2840 case PT_ALNUM:
2841 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2842 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2843
2844 /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2845 means that Perl space and POSIX space are now identical. PCRE was changed
2846 at release 8.34. */
2847
2848 case PT_SPACE: /* Perl space */
2849 case PT_PXSPACE: /* POSIX space */
2850 switch(c)
2851 {
2852 HSPACE_CASES:
2853 VSPACE_CASES:
2854 return negated;
2855
2856 default:
2857 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2858 }
2859 break; /* Control never reaches here */
2860
2861 case PT_WORD:
2862 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2863 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2864 c == CHAR_UNDERSCORE) == negated;
2865
2866 case PT_CLIST:
2867 p = PRIV(ucd_caseless_sets) + prop->caseset;
2868 for (;;)
2869 {
2870 if (c < *p) return !negated;
2871 if (c == *p++) return negated;
2872 }
2873 break; /* Control never reaches here */
2874 }
2875
2876 return FALSE;
2877 }
2878 #endif /* SUPPORT_UCP */
2879
2880
2881
2882 /*************************************************
2883 * Fill the character property list *
2884 *************************************************/
2885
2886 /* Checks whether the code points to an opcode that can take part in auto-
2887 possessification, and if so, fills a list with its properties.
2888
2889 Arguments:
2890 code points to start of expression
2891 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2892 fcc points to case-flipping table
2893 list points to output list
2894 list[0] will be filled with the opcode
2895 list[1] will be non-zero if this opcode
2896 can match an empty character string
2897 list[2..7] depends on the opcode
2898
2899 Returns: points to the start of the next opcode if *code is accepted
2900 NULL if *code is not accepted
2901 */
2902
2903 static const pcre_uchar *
2904 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2905 const pcre_uint8 *fcc, pcre_uint32 *list)
2906 {
2907 pcre_uchar c = *code;
2908 pcre_uchar base;
2909 const pcre_uchar *end;
2910 pcre_uint32 chr;
2911
2912 #ifdef SUPPORT_UCP
2913 pcre_uint32 *clist_dest;
2914 const pcre_uint32 *clist_src;
2915 #else
2916 utf = utf; /* Suppress "unused parameter" compiler warning */
2917 #endif
2918
2919 list[0] = c;
2920 list[1] = FALSE;
2921 code++;
2922
2923 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2924 {
2925 base = get_repeat_base(c);
2926 c -= (base - OP_STAR);
2927
2928 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2929 code += IMM2_SIZE;
2930
2931 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2932
2933 switch(base)
2934 {
2935 case OP_STAR:
2936 list[0] = OP_CHAR;
2937 break;
2938
2939 case OP_STARI:
2940 list[0] = OP_CHARI;
2941 break;
2942
2943 case OP_NOTSTAR:
2944 list[0] = OP_NOT;
2945 break;
2946
2947 case OP_NOTSTARI:
2948 list[0] = OP_NOTI;
2949 break;
2950
2951 case OP_TYPESTAR:
2952 list[0] = *code;
2953 code++;
2954 break;
2955 }
2956 c = list[0];
2957 }
2958
2959 switch(c)
2960 {
2961 case OP_NOT_DIGIT:
2962 case OP_DIGIT:
2963 case OP_NOT_WHITESPACE:
2964 case OP_WHITESPACE:
2965 case OP_NOT_WORDCHAR:
2966 case OP_WORDCHAR:
2967 case OP_ANY:
2968 case OP_ALLANY:
2969 case OP_ANYNL:
2970 case OP_NOT_HSPACE:
2971 case OP_HSPACE:
2972 case OP_NOT_VSPACE:
2973 case OP_VSPACE:
2974 case OP_EXTUNI:
2975 case OP_EODN:
2976 case OP_EOD:
2977 case OP_DOLL:
2978 case OP_DOLLM:
2979 return code;
2980
2981 case OP_CHAR:
2982 case OP_NOT:
2983 GETCHARINCTEST(chr, code);
2984 list[2] = chr;
2985 list[3] = NOTACHAR;
2986 return code;
2987
2988 case OP_CHARI:
2989 case OP_NOTI:
2990 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2991 GETCHARINCTEST(chr, code);
2992 list[2] = chr;
2993
2994 #ifdef SUPPORT_UCP
2995 if (chr < 128 || (chr < 256 && !utf))
2996 list[3] = fcc[chr];
2997 else
2998 list[3] = UCD_OTHERCASE(chr);
2999 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
3000 list[3] = (chr < 256) ? fcc[chr] : chr;
3001 #else
3002 list[3] = fcc[chr];
3003 #endif
3004
3005 /* The othercase might be the same value. */
3006
3007 if (chr == list[3])
3008 list[3] = NOTACHAR;
3009 else
3010 list[4] = NOTACHAR;
3011 return code;
3012
3013 #ifdef SUPPORT_UCP
3014 case OP_PROP:
3015 case OP_NOTPROP:
3016 if (code[0] != PT_CLIST)
3017 {
3018 list[2] = code[0];
3019 list[3] = code[1];
3020 return code + 2;
3021 }
3022
3023 /* Convert only if we have enough space. */
3024
3025 clist_src = PRIV(ucd_caseless_sets) + code[1];
3026 clist_dest = list + 2;
3027 code += 2;
3028
3029 do {
3030 if (clist_dest >= list + 8)
3031 {
3032 /* Early return if there is not enough space. This should never
3033 happen, since all clists are shorter than 5 character now. */
3034 list[2] = code[0];
3035 list[3] = code[1];
3036 return code;
3037 }
3038 *clist_dest++ = *clist_src;
3039 }
3040 while(*clist_src++ != NOTACHAR);
3041
3042 /* All characters are stored. The terminating NOTACHAR
3043 is copied form the clist itself. */
3044
3045 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3046 return code;
3047 #endif
3048
3049 case OP_NCLASS:
3050 case OP_CLASS:
3051 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3052 case OP_XCLASS:
3053 if (c == OP_XCLASS)
3054 end = code + GET(code, 0) - 1;
3055 else
3056 #endif
3057 end = code + 32 / sizeof(pcre_uchar);
3058
3059 switch(*end)
3060 {
3061 case OP_CRSTAR:
3062 case OP_CRMINSTAR:
3063 case OP_CRQUERY:
3064 case OP_CRMINQUERY:
3065 case OP_CRPOSSTAR:
3066 case OP_CRPOSQUERY:
3067 list[1] = TRUE;
3068 end++;
3069 break;
3070
3071 case OP_CRPLUS:
3072 case OP_CRMINPLUS:
3073 case OP_CRPOSPLUS:
3074 end++;
3075 break;
3076
3077 case OP_CRRANGE:
3078 case OP_CRMINRANGE:
3079 case OP_CRPOSRANGE:
3080 list[1] = (GET2(end, 1) == 0);
3081 end += 1 + 2 * IMM2_SIZE;
3082 break;
3083 }
3084 list[2] = (pcre_uint32)(end - code);
3085 return end;
3086 }
3087 return NULL; /* Opcode not accepted */
3088 }
3089
3090
3091
3092 /*************************************************
3093 * Scan further character sets for match *
3094 *************************************************/
3095
3096 /* Checks whether the base and the current opcode have a common character, in
3097 which case the base cannot be possessified.
3098
3099 Arguments:
3100 code points to the byte code
3101 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3102 cd static compile data
3103 base_list the data list of the base opcode
3104
3105 Returns: TRUE if the auto-possessification is possible
3106 */
3107
3108 static BOOL
3109 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3110 const pcre_uint32 *base_list, const pcre_uchar *base_end, int *rec_limit)
3111 {
3112 pcre_uchar c;
3113 pcre_uint32 list[8];
3114 const pcre_uint32 *chr_ptr;
3115 const pcre_uint32 *ochr_ptr;
3116 const pcre_uint32 *list_ptr;
3117 const pcre_uchar *next_code;
3118 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3119 const pcre_uchar *xclass_flags;
3120 #endif
3121 const pcre_uint8 *class_bitset;
3122 const pcre_uint8 *set1, *set2, *set_end;
3123 pcre_uint32 chr;
3124 BOOL accepted, invert_bits;
3125 BOOL entered_a_group = FALSE;
3126
3127 if (*rec_limit == 0) return FALSE;
3128 --(*rec_limit);
3129
3130 /* Note: the base_list[1] contains whether the current opcode has greedy
3131 (represented by a non-zero value) quantifier. This is a different from
3132 other character type lists, which stores here that the character iterator
3133 matches to an empty string (also represented by a non-zero value). */
3134
3135 for(;;)
3136 {
3137 /* All operations move the code pointer forward.
3138 Therefore infinite recursions are not possible. */
3139
3140 c = *code;
3141
3142 /* Skip over callouts */
3143
3144 if (c == OP_CALLOUT)
3145 {
3146 code += PRIV(OP_lengths)[c];
3147 continue;
3148 }
3149
3150 if (c == OP_ALT)
3151 {
3152 do code += GET(code, 1); while (*code == OP_ALT);
3153 c = *code;
3154 }
3155
3156 switch(c)
3157 {
3158 case OP_END:
3159 case OP_KETRPOS:
3160 /* TRUE only in greedy case. The non-greedy case could be replaced by
3161 an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3162 uses more memory, which we cannot get at this stage.) */
3163
3164 return base_list[1] != 0;
3165
3166 case OP_KET:
3167 /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3168 it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3169 cannot be converted to a possessive form. */
3170
3171 if (base_list[1] == 0) return FALSE;
3172
3173 switch(*(code - GET(code, 1)))
3174 {
3175 case OP_ASSERT:
3176 case OP_ASSERT_NOT:
3177 case OP_ASSERTBACK:
3178 case OP_ASSERTBACK_NOT:
3179 case OP_ONCE:
3180 case OP_ONCE_NC:
3181 /* Atomic sub-patterns and assertions can always auto-possessify their
3182 last iterator. However, if the group was entered as a result of checking
3183 a previous iterator, this is not possible. */
3184
3185 return !entered_a_group;
3186 }
3187
3188 code += PRIV(OP_lengths)[c];
3189 continue;
3190
3191 case OP_ONCE:
3192 case OP_ONCE_NC:
3193 case OP_BRA:
3194 case OP_CBRA:
3195 next_code = code + GET(code, 1);
3196 code += PRIV(OP_lengths)[c];
3197
3198 while (*next_code == OP_ALT)
3199 {
3200 if (!compare_opcodes(code, utf, cd, base_list, base_end, rec_limit))
3201 return FALSE;
3202 code = next_code + 1 + LINK_SIZE;
3203 next_code += GET(next_code, 1);
3204 }
3205
3206 entered_a_group = TRUE;
3207 continue;
3208
3209 case OP_BRAZERO:
3210 case OP_BRAMINZERO:
3211
3212 next_code = code + 1;
3213 if (*next_code != OP_BRA && *next_code != OP_CBRA
3214 && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3215
3216 do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3217
3218 /* The bracket content will be checked by the
3219 OP_BRA/OP_CBRA case above. */
3220 next_code += 1 + LINK_SIZE;
3221 if (!compare_opcodes(next_code, utf, cd, base_list, base_end, rec_limit))
3222 return FALSE;
3223
3224 code += PRIV(OP_lengths)[c];
3225 continue;
3226
3227 default:
3228 break;
3229 }
3230
3231 /* Check for a supported opcode, and load its properties. */
3232
3233 code = get_chr_property_list(code, utf, cd->fcc, list);
3234 if (code == NULL) return FALSE; /* Unsupported */
3235
3236 /* If either opcode is a small character list, set pointers for comparing
3237 characters from that list with another list, or with a property. */
3238
3239 if (base_list[0] == OP_CHAR)
3240 {
3241 chr_ptr = base_list + 2;
3242 list_ptr = list;
3243 }
3244 else if (list[0] == OP_CHAR)
3245 {
3246 chr_ptr = list + 2;
3247 list_ptr = base_list;
3248 }
3249
3250 /* Character bitsets can also be compared to certain opcodes. */
3251
3252 else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3253 #ifdef COMPILE_PCRE8
3254 /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3255 || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3256 #endif
3257 )
3258 {
3259 #ifdef COMPILE_PCRE8
3260 if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3261 #else
3262 if (base_list[0] == OP_CLASS)
3263 #endif
3264 {
3265 set1 = (pcre_uint8 *)(base_end - base_list[2]);
3266 list_ptr = list;
3267 }
3268 else
3269 {
3270 set1 = (pcre_uint8 *)(code - list[2]);
3271 list_ptr = base_list;
3272 }
3273
3274 invert_bits = FALSE;
3275 switch(list_ptr[0])
3276 {
3277 case OP_CLASS:
3278 case OP_NCLASS:
3279 set2 = (pcre_uint8 *)
3280 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3281 break;
3282
3283 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3284 case OP_XCLASS:
3285 xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3286 if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3287 if ((*xclass_flags & XCL_MAP) == 0)
3288 {
3289 /* No bits are set for characters < 256. */
3290 if (list[1] == 0) return TRUE;
3291 /* Might be an empty repeat. */
3292 continue;
3293 }
3294 set2 = (pcre_uint8 *)(xclass_flags + 1);
3295 break;
3296 #endif
3297
3298 case OP_NOT_DIGIT:
3299 invert_bits = TRUE;
3300 /* Fall through */
3301 case OP_DIGIT:
3302 set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3303 break;
3304
3305 case OP_NOT_WHITESPACE:
3306 invert_bits = TRUE;
3307 /* Fall through */
3308 case OP_WHITESPACE:
3309 set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3310 break;
3311
3312 case OP_NOT_WORDCHAR:
3313 invert_bits = TRUE;
3314 /* Fall through */
3315 case OP_WORDCHAR:
3316 set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3317 break;
3318
3319 default:
3320 return FALSE;
3321 }
3322
3323 /* Because the sets are unaligned, we need
3324 to perform byte comparison here. */
3325 set_end = set1 + 32;
3326 if (invert_bits)
3327 {
3328 do
3329 {
3330 if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3331 }
3332 while (set1 < set_end);
3333 }
3334 else
3335 {
3336 do
3337 {
3338 if ((*set1++ & *set2++) != 0) return FALSE;
3339 }
3340 while (set1 < set_end);
3341 }
3342
3343 if (list[1] == 0) return TRUE;
3344 /* Might be an empty repeat. */
3345 continue;
3346 }
3347
3348 /* Some property combinations also acceptable. Unicode property opcodes are
3349 processed specially; the rest can be handled with a lookup table. */
3350
3351 else
3352 {
3353 pcre_uint32 leftop, rightop;
3354
3355 leftop = base_list[0];
3356 rightop = list[0];
3357
3358 #ifdef SUPPORT_UCP
3359 accepted = FALSE; /* Always set in non-unicode case. */
3360 if (leftop == OP_PROP || leftop == OP_NOTPROP)
3361 {
3362 if (rightop == OP_EOD)
3363 accepted = TRUE;
3364 else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3365 {
3366 int n;
3367 const pcre_uint8 *p;
3368 BOOL same = leftop == rightop;
3369 BOOL lisprop = leftop == OP_PROP;
3370 BOOL risprop = rightop == OP_PROP;
3371 BOOL bothprop = lisprop && risprop;
3372
3373 /* There's a table that specifies how each combination is to be
3374 processed:
3375 0 Always return FALSE (never auto-possessify)
3376 1 Character groups are distinct (possessify if both are OP_PROP)
3377 2 Check character categories in the same group (general or particular)
3378 3 Return TRUE if the two opcodes are not the same
3379 ... see comments below
3380 */
3381
3382 n = propposstab[base_list[2]][list[2]];
3383 switch(n)
3384 {
3385 case 0: break;
3386 case 1: accepted = bothprop; break;
3387 case 2: accepted = (base_list[3] == list[3]) != same; break;
3388 case 3: accepted = !same; break;
3389
3390 case 4: /* Left general category, right particular category */
3391 accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3392 break;
3393
3394 case 5: /* Right general category, left particular category */
3395 accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3396 break;
3397
3398 /* This code is logically tricky. Think hard before fiddling with it.
3399 The posspropstab table has four entries per row. Each row relates to
3400 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3401 Only WORD actually needs all four entries, but using repeats for the
3402 others means they can all use the same code below.
3403
3404 The first two entries in each row are Unicode general categories, and
3405 apply always, because all the characters they include are part of the
3406 PCRE character set. The third and fourth entries are a general and a
3407 particular category, respectively, that include one or more relevant
3408 characters. One or the other is used, depending on whether the check
3409 is for a general or a particular category. However, in both cases the
3410 category contains more characters than the specials that are defined
3411 for the property being tested against. Therefore, it cannot be used
3412 in a NOTPROP case.
3413
3414 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3415 Underscore is covered by ucp_P or ucp_Po. */
3416
3417 case 6: /* Left alphanum vs right general category */
3418 case 7: /* Left space vs right general category */
3419 case 8: /* Left word vs right general category */
3420 p = posspropstab[n-6];
3421 accepted = risprop && lisprop ==
3422 (list[3] != p[0] &&
3423 list[3] != p[1] &&
3424 (list[3] != p[2] || !lisprop));
3425 break;
3426
3427 case 9: /* Right alphanum vs left general category */
3428 case 10: /* Right space vs left general category */
3429 case 11: /* Right word vs left general category */
3430 p = posspropstab[n-9];
3431 accepted = lisprop && risprop ==
3432 (base_list[3] != p[0] &&
3433 base_list[3] != p[1] &&
3434 (base_list[3] != p[2] || !risprop));
3435 break;
3436
3437 case 12: /* Left alphanum vs right particular category */
3438 case 13: /* Left space vs right particular category */
3439 case 14: /* Left word vs right particular category */
3440 p = posspropstab[n-12];
3441 accepted = risprop && lisprop ==
3442 (catposstab[p[0]][list[3]] &&
3443 catposstab[p[1]][list[3]] &&
3444 (list[3] != p[3] || !lisprop));
3445 break;
3446
3447 case 15: /* Right alphanum vs left particular category */
3448 case 16: /* Right space vs left particular category */
3449 case 17: /* Right word vs left particular category */
3450 p = posspropstab[n-15];
3451 accepted = lisprop && risprop ==
3452 (catposstab[p[0]][base_list[3]] &&
3453 catposstab[p[1]][base_list[3]] &&
3454 (base_list[3] != p[3] || !risprop));
3455 break;
3456 }
3457 }
3458 }
3459
3460 else
3461 #endif /* SUPPORT_UCP */
3462
3463 accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3464 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3465 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3466
3467 if (!accepted) return FALSE;
3468
3469 if (list[1] == 0) return TRUE;
3470 /* Might be an empty repeat. */
3471 continue;
3472 }
3473
3474 /* Control reaches here only if one of the items is a small character list.
3475 All characters are checked against the other side. */
3476
3477 do
3478 {
3479 chr = *chr_ptr;
3480
3481 switch(list_ptr[0])
3482 {
3483 case OP_CHAR:
3484 ochr_ptr = list_ptr + 2;
3485 do
3486 {
3487 if (chr == *ochr_ptr) return FALSE;
3488 ochr_ptr++;
3489 }
3490 while(*ochr_ptr != NOTACHAR);
3491 break;
3492
3493 case OP_NOT:
3494 ochr_ptr = list_ptr + 2;
3495 do
3496 {
3497 if (chr == *ochr_ptr)
3498 break;
3499 ochr_ptr++;
3500 }
3501 while(*ochr_ptr != NOTACHAR);
3502 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3503 break;
3504
3505 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3506 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3507
3508 case OP_DIGIT:
3509 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3510 break;
3511
3512 case OP_NOT_DIGIT:
3513 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3514 break;
3515
3516 case OP_WHITESPACE:
3517 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3518 break;
3519
3520 case OP_NOT_WHITESPACE:
3521 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3522 break;
3523
3524 case OP_WORDCHAR:
3525 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3526 break;
3527
3528 case OP_NOT_WORDCHAR:
3529 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3530 break;
3531
3532 case OP_HSPACE:
3533 switch(chr)
3534 {
3535 HSPACE_CASES: return FALSE;
3536 default: break;
3537 }
3538 break;
3539
3540 case OP_NOT_HSPACE:
3541 switch(chr)
3542 {
3543 HSPACE_CASES: break;
3544 default: return FALSE;
3545 }
3546 break;
3547
3548 case OP_ANYNL:
3549 case OP_VSPACE:
3550 switch(chr)
3551 {
3552 VSPACE_CASES: return FALSE;
3553 default: break;
3554 }
3555 break;
3556
3557 case OP_NOT_VSPACE:
3558 switch(chr)
3559 {
3560 VSPACE_CASES: break;
3561 default: return FALSE;
3562 }
3563 break;
3564
3565 case OP_DOLL:
3566 case OP_EODN:
3567 switch (chr)
3568 {
3569 case CHAR_CR:
3570 case CHAR_LF:
3571 case CHAR_VT:
3572 case CHAR_FF:
3573 case CHAR_NEL:
3574 #ifndef EBCDIC
3575 case 0x2028:
3576 case 0x2029:
3577 #endif /* Not EBCDIC */
3578 return FALSE;
3579 }
3580 break;
3581
3582 case OP_EOD: /* Can always possessify before \z */
3583 break;
3584
3585 #ifdef SUPPORT_UCP
3586 case OP_PROP:
3587 case OP_NOTPROP:
3588 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3589 list_ptr[0] == OP_NOTPROP))
3590 return FALSE;
3591 break;
3592 #endif
3593
3594 case OP_NCLASS:
3595 if (chr > 255) return FALSE;
3596 /* Fall through */
3597
3598 case OP_CLASS:
3599 if (chr > 255) break;
3600 class_bitset = (pcre_uint8 *)
3601 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3602 if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3603 break;
3604
3605 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3606 case OP_XCLASS:
3607 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3608 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3609 break;
3610 #endif
3611
3612 default:
3613 return FALSE;
3614 }
3615
3616 chr_ptr++;
3617 }
3618 while(*chr_ptr != NOTACHAR);
3619
3620 /* At least one character must be matched from this opcode. */
3621
3622 if (list[1] == 0) return TRUE;
3623 }
3624
3625 /* Control never reaches here. There used to be a fail-save return FALSE; here,
3626 but some compilers complain about an unreachable statement. */
3627
3628 }
3629
3630
3631
3632 /*************************************************
3633 * Scan compiled regex for auto-possession *
3634 *************************************************/
3635
3636 /* Replaces single character iterations with their possessive alternatives
3637 if appropriate. This function modifies the compiled opcode!
3638
3639 Arguments:
3640 code points to start of the byte code
3641 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3642 cd static compile data
3643
3644 Returns: nothing
3645 */
3646
3647 static void
3648 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3649 {
3650 register pcre_uchar c;
3651 const pcre_uchar *end;
3652 pcre_uchar *repeat_opcode;
3653 pcre_uint32 list[8];
3654 int rec_limit;
3655
3656 for (;;)
3657 {
3658 c = *code;
3659
3660 /* When a pattern with bad UTF-8 encoding is compiled with NO_UTF_CHECK,
3661 it may compile without complaining, but may get into a loop here if the code
3662 pointer points to a bad value. This is, of course a documentated possibility,
3663 when NO_UTF_CHECK is set, so it isn't a bug, but we can detect this case and
3664 just give up on this optimization. */
3665
3666 if (c >= OP_TABLE_LENGTH) return;
3667
3668 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3669 {
3670 c -= get_repeat_base(c) - OP_STAR;
3671 end = (c <= OP_MINUPTO) ?
3672 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3673 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3674
3675 rec_limit = 1000;
3676 if (end != NULL && compare_opcodes(end, utf, cd, list, end, &rec_limit))
3677 {
3678 switch(c)
3679 {
3680 case OP_STAR:
3681 *code += OP_POSSTAR - OP_STAR;
3682 break;
3683
3684 case OP_MINSTAR:
3685 *code += OP_POSSTAR - OP_MINSTAR;
3686 break;
3687
3688 case OP_PLUS:
3689 *code += OP_POSPLUS - OP_PLUS;
3690 break;
3691
3692 case OP_MINPLUS:
3693 *code += OP_POSPLUS - OP_MINPLUS;
3694 break;
3695
3696 case OP_QUERY:
3697 *code += OP_POSQUERY - OP_QUERY;
3698 break;
3699
3700 case OP_MINQUERY:
3701 *code += OP_POSQUERY - OP_MINQUERY;
3702 break;
3703
3704 case OP_UPTO:
3705 *code += OP_POSUPTO - OP_UPTO;
3706 break;
3707
3708 case OP_MINUPTO:
3709 *code += OP_POSUPTO - OP_MINUPTO;
3710 break;
3711 }
3712 }
3713 c = *code;
3714 }
3715 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3716 {
3717 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3718 if (c == OP_XCLASS)
3719 repeat_opcode = code + GET(code, 1);
3720 else
3721 #endif
3722 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3723
3724 c = *repeat_opcode;
3725 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3726 {
3727 /* end must not be NULL. */
3728 end = get_chr_property_list(code, utf, cd->fcc, list);
3729
3730 list[1] = (c & 1) == 0;
3731
3732 rec_limit = 1000;
3733 if (compare_opcodes(end, utf, cd, list, end, &rec_limit))
3734 {
3735 switch (c)
3736 {
3737 case OP_CRSTAR:
3738 case OP_CRMINSTAR:
3739 *repeat_opcode = OP_CRPOSSTAR;
3740 break;
3741
3742 case OP_CRPLUS:
3743 case OP_CRMINPLUS:
3744 *repeat_opcode = OP_CRPOSPLUS;
3745 break;
3746
3747 case OP_CRQUERY:
3748 case OP_CRMINQUERY:
3749 *repeat_opcode = OP_CRPOSQUERY;
3750 break;
3751
3752 case OP_CRRANGE:
3753 case OP_CRMINRANGE:
3754 *repeat_opcode = OP_CRPOSRANGE;
3755 break;
3756 }
3757 }
3758 }
3759 c = *code;
3760 }
3761
3762 switch(c)
3763 {
3764 case OP_END:
3765 return;
3766
3767 case OP_TYPESTAR:
3768 case OP_TYPEMINSTAR:
3769 case OP_TYPEPLUS:
3770 case OP_TYPEMINPLUS:
3771 case OP_TYPEQUERY:
3772 case OP_TYPEMINQUERY:
3773 case OP_TYPEPOSSTAR:
3774 case OP_TYPEPOSPLUS:
3775 case OP_TYPEPOSQUERY:
3776 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3777 break;
3778
3779 case OP_TYPEUPTO:
3780 case OP_TYPEMINUPTO:
3781 case OP_TYPEEXACT:
3782 case OP_TYPEPOSUPTO:
3783 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3784 code += 2;
3785 break;
3786
3787 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3788 case OP_XCLASS:
3789 code += GET(code, 1);
3790 break;
3791 #endif
3792
3793 case OP_MARK:
3794 case OP_PRUNE_ARG:
3795 case OP_SKIP_ARG:
3796 case OP_THEN_ARG:
3797 code += code[1];
3798 break;
3799 }
3800
3801 /* Add in the fixed length from the table */
3802
3803 code += PRIV(OP_lengths)[c];
3804
3805 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3806 a multi-byte character. The length in the table is a minimum, so we have to
3807 arrange to skip the extra bytes. */
3808
3809 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3810 if (utf) switch(c)
3811 {
3812 case OP_CHAR:
3813 case OP_CHARI:
3814 case OP_NOT:
3815 case OP_NOTI:
3816 case OP_STAR:
3817 case OP_MINSTAR:
3818 case OP_PLUS:
3819 case OP_MINPLUS:
3820 case OP_QUERY:
3821 case OP_MINQUERY:
3822 case OP_UPTO:
3823 case OP_MINUPTO:
3824 case OP_EXACT:
3825 case OP_POSSTAR:
3826 case OP_POSPLUS:
3827 case OP_POSQUERY:
3828 case OP_POSUPTO:
3829 case OP_STARI:
3830 case OP_MINSTARI:
3831 case OP_PLUSI:
3832 case OP_MINPLUSI:
3833 case OP_QUERYI:
3834 case OP_MINQUERYI:
3835 case OP_UPTOI:
3836 case OP_MINUPTOI:
3837 case OP_EXACTI:
3838 case OP_POSSTARI:
3839 case OP_POSPLUSI:
3840 case OP_POSQUERYI:
3841 case OP_POSUPTOI:
3842 case OP_NOTSTAR:
3843 case OP_NOTMINSTAR:
3844 case OP_NOTPLUS:
3845 case OP_NOTMINPLUS:
3846 case OP_NOTQUERY:
3847 case OP_NOTMINQUERY:
3848 case OP_NOTUPTO:
3849 case OP_NOTMINUPTO:
3850 case OP_NOTEXACT:
3851 case OP_NOTPOSSTAR:
3852 case OP_NOTPOSPLUS:
3853 case OP_NOTPOSQUERY:
3854 case OP_NOTPOSUPTO:
3855 case OP_NOTSTARI:
3856 case OP_NOTMINSTARI:
3857 case OP_NOTPLUSI:
3858 case OP_NOTMINPLUSI:
3859 case OP_NOTQUERYI:
3860 case OP_NOTMINQUERYI:
3861 case OP_NOTUPTOI:
3862 case OP_NOTMINUPTOI:
3863 case OP_NOTEXACTI:
3864 case OP_NOTPOSSTARI:
3865 case OP_NOTPOSPLUSI:
3866 case OP_NOTPOSQUERYI:
3867 case OP_NOTPOSUPTOI:
3868 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3869 break;
3870 }
3871 #else
3872 (void)(utf); /* Keep compiler happy by referencing function argument */
3873 #endif
3874 }
3875 }
3876
3877
3878
3879 /*************************************************
3880 * Check for POSIX class syntax *
3881 *************************************************/
3882
3883 /* This function is called when the sequence "[:" or "[." or "[=" is
3884 encountered in a character class. It checks whether this is followed by a
3885 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3886 reach an unescaped ']' without the special preceding character, return FALSE.
3887
3888 Originally, this function only recognized a sequence of letters between the
3889 terminators, but it seems that Perl recognizes any sequence of characters,
3890 though of course unknown POSIX names are subsequently rejected. Perl gives an
3891 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3892 didn't consider this to be a POSIX class. Likewise for [:1234:].
3893
3894 The problem in trying to be exactly like Perl is in the handling of escapes. We
3895 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3896 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3897 below handles the special case of \], but does not try to do any other escape
3898 processing. This makes it different from Perl for cases such as [:l\ower:]
3899 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3900 "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3901 I think.
3902
3903 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3904 It seems that the appearance of a nested POSIX class supersedes an apparent
3905 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3906 a digit.
3907
3908 In Perl, unescaped square brackets may also appear as part of class names. For
3909 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3910 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3911 seem right at all. PCRE does not allow closing square brackets in POSIX class
3912 names.
3913
3914 Arguments:
3915 ptr pointer to the initial [
3916 endptr where to return the end pointer
3917
3918 Returns: TRUE or FALSE
3919 */
3920
3921 static BOOL
3922 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3923 {
3924 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3925 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3926 for (++ptr; *ptr != CHAR_NULL; ptr++)
3927 {
3928 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3929 ptr++;
3930 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3931 else
3932 {
3933 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3934 {
3935 *endptr = ptr;
3936 return TRUE;
3937 }
3938 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3939 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3940 ptr[1] == CHAR_EQUALS_SIGN) &&
3941 check_posix_syntax(ptr, endptr))
3942 return FALSE;
3943 }
3944 }
3945 return FALSE;
3946 }
3947
3948
3949
3950
3951 /*************************************************
3952 * Check POSIX class name *
3953 *************************************************/
3954
3955 /* This function is called to check the name given in a POSIX-style class entry
3956 such as [:alnum:].
3957
3958 Arguments:
3959 ptr points to the first letter
3960 len the length of the name
3961
3962 Returns: a value representing the name, or -1 if unknown
3963 */
3964
3965 static int
3966 check_posix_name(const pcre_uchar *ptr, int len)
3967 {
3968 const char *pn = posix_names;
3969 register int yield = 0;
3970 while (posix_name_lengths[yield] != 0)
3971 {
3972 if (len == posix_name_lengths[yield] &&
3973 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3974 pn += posix_name_lengths[yield] + 1;
3975 yield++;
3976 }
3977 return -1;
3978 }
3979
3980
3981 /*************************************************
3982 * Adjust OP_RECURSE items in repeated group *
3983 *************************************************/
3984
3985 /* OP_RECURSE items contain an offset from the start of the regex to the group
3986 that is referenced. This means that groups can be replicated for fixed
3987 repetition simply by copying (because the recursion is allowed to refer to
3988 earlier groups that are outside the current group). However, when a group is
3989 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3990 inserted before it, after it has been compiled. This means that any OP_RECURSE
3991 items within it that refer to the group itself or any contained groups have to
3992 have their offsets adjusted. That one of the jobs of this function. Before it
3993 is called, the partially compiled regex must be temporarily terminated with
3994 OP_END.
3995
3996 This function has been extended with the possibility of forward references for
3997 recursions and subroutine calls. It must also check the list of such references
3998 for the group we are dealing with. If it finds that one of the recursions in
3999 the current group is on this list, it adjusts the offset in the list, not the
4000 value in the reference (which is a group number).
4001
4002 Arguments:
4003 group points to the start of the group
4004 adjust the amount by which the group is to be moved
4005 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
4006 cd contains pointers to tables etc.
4007 save_hwm_offset the hwm forward reference offset at the start of the group
4008
4009 Returns: nothing
4010 */
4011
4012 static void
4013 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
4014 size_t save_hwm_offset)
4015 {
4016 pcre_uchar *ptr = group;
4017
4018 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
4019 {
4020 int offset;
4021 pcre_uchar *hc;
4022
4023 /* See if this recursion is on the forward reference list. If so, adjust the
4024 reference. */
4025
4026 for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4027 hc += LINK_SIZE)
4028 {
4029 offset = (int)GET(hc, 0);
4030 if (cd->start_code + offset == ptr + 1)
4031 {
4032 PUT(hc, 0, offset + adjust);
4033 break;
4034 }
4035 }
4036
4037 /* Otherwise, adjust the recursion offset if it's after the start of this
4038 group. */
4039
4040 if (hc >= cd->hwm)
4041 {
4042 offset = (int)GET(ptr, 1);
4043 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
4044 }
4045
4046 ptr += 1 + LINK_SIZE;
4047 }
4048 }
4049
4050
4051
4052 /*************************************************
4053 * Insert an automatic callout point *
4054 *************************************************/
4055
4056 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
4057 callout points before each pattern item.
4058
4059 Arguments:
4060 code current code pointer
4061 ptr current pattern pointer
4062 cd pointers to tables etc
4063
4064 Returns: new code pointer
4065 */
4066
4067 static pcre_uchar *
4068 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4069 {
4070 *code++ = OP_CALLOUT;
4071 *code++ = 255;
4072 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
4073 PUT(code, LINK_SIZE, 0); /* Default length */
4074 return code + 2 * LINK_SIZE;
4075 }
4076
4077
4078
4079 /*************************************************
4080 * Complete a callout item *
4081 *************************************************/
4082
4083 /* A callout item contains the length of the next item in the pattern, which
4084 we can't fill in till after we have reached the relevant point. This is used
4085 for both automatic and manual callouts.
4086
4087 Arguments:
4088 previous_callout points to previous callout item
4089 ptr current pattern pointer
4090 cd pointers to tables etc
4091
4092 Returns: nothing
4093 */
4094
4095 static void
4096 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4097 {
4098 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4099 PUT(previous_callout, 2 + LINK_SIZE, length);
4100 }
4101
4102
4103
4104 #ifdef SUPPORT_UCP
4105 /*************************************************
4106 * Get othercase range *
4107 *************************************************/
4108
4109 /* This function is passed the start and end of a class range, in UTF-8 mode
4110 with UCP support. It searches up the characters, looking for ranges of
4111 characters in the "other" case. Each call returns the next one, updating the
4112 start address. A character with multiple other cases is returned on its own
4113 with a special return value.
4114
4115 Arguments:
4116 cptr points to starting character value; updated
4117 d end value
4118 ocptr where to put start of othercase range
4119 odptr where to put end of othercase range
4120
4121 Yield: -1 when no more
4122 0 when a range is returned
4123 >0 the CASESET offset for char with multiple other cases
4124 in this case, ocptr contains the original
4125 */
4126
4127 static int
4128 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4129 pcre_uint32 *odptr)
4130 {
4131 pcre_uint32 c, othercase, next;
4132 unsigned int co;
4133
4134 /* Find the first character that has an other case. If it has multiple other
4135 cases, return its case offset value. */
4136
4137 for (c = *cptr; c <= d; c++)
4138 {
4139 if ((co = UCD_CASESET(c)) != 0)
4140 {
4141 *ocptr = c++; /* Character that has the set */
4142 *cptr = c; /* Rest of input range */
4143 return (int)co;
4144 }
4145 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4146 }
4147
4148 if (c > d) return -1; /* Reached end of range */
4149
4150 /* Found a character that has a single other case. Search for the end of the
4151 range, which is either the end of the input range, or a character that has zero
4152 or more than one other cases. */
4153
4154 *ocptr = othercase;
4155 next = othercase + 1;
4156
4157 for (++c; c <= d; c++)
4158 {
4159 if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4160 next++;
4161 }
4162
4163 *odptr = next - 1; /* End of othercase range */
4164 *cptr = c; /* Rest of input range */
4165 return 0;
4166 }
4167 #endif /* SUPPORT_UCP */
4168
4169
4170
4171 /*************************************************
4172 * Add a character or range to a class *
4173 *************************************************/
4174
4175 /* This function packages up the logic of adding a character or range of
4176 characters to a class. The character values in the arguments will be within the
4177 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4178 mutually recursive with the function immediately below.
4179
4180 Arguments:
4181 classbits the bit map for characters < 256
4182 uchardptr points to the pointer for extra data
4183 options the options word
4184 cd contains pointers to tables etc.
4185 start start of range character
4186 end end of range character
4187
4188 Returns: the number of < 256 characters added
4189 the pointer to extra data is updated
4190 */
4191
4192 static int
4193 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4194 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4195 {
4196 pcre_uint32 c;
4197 pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4198 int n8 = 0;
4199
4200 /* If caseless matching is required, scan the range and process alternate
4201 cases. In Unicode, there are 8-bit characters that have alternate cases that
4202 are greater than 255 and vice-versa. Sometimes we can just extend the original
4203 range. */
4204
4205 if ((options & PCRE_CASELESS) != 0)
4206 {
4207 #ifdef SUPPORT_UCP
4208 if ((options & PCRE_UTF8) != 0)
4209 {
4210 int rc;
4211 pcre_uint32 oc, od;
4212
4213 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4214 c = start;
4215
4216 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4217 {
4218 /* Handle a single character that has more than one other case. */
4219
4220 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4221 PRIV(ucd_caseless_sets) + rc, oc);
4222
4223 /* Do nothing if the other case range is within the original range. */
4224
4225 else if (oc >= start && od <= end) continue;
4226
4227 /* Extend the original range if there is overlap, noting that if oc < c, we
4228 can't have od > end because a subrange is always shorter than the basic
4229 range. Otherwise, use a recursive call to add the additional range. */
4230
4231 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4232 else if (od > end && oc <= end + 1)
4233 {
4234 end = od; /* Extend upwards */
4235 if (end > classbits_end) classbits_end = (end <= 0xff ? end : 0xff);
4236 }
4237 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4238 }
4239 }
4240 else
4241 #endif /* SUPPORT_UCP */
4242
4243 /* Not UTF-mode, or no UCP */
4244
4245 for (c = start; c <= classbits_end; c++)
4246 {
4247 SETBIT(classbits, cd->fcc[c]);
4248 n8++;
4249 }
4250 }
4251
4252 /* Now handle the original range. Adjust the final value according to the bit
4253 length - this means that the same lists of (e.g.) horizontal spaces can be used
4254 in all cases. */
4255
4256 #if defined COMPILE_PCRE8
4257 #ifdef SUPPORT_UTF
4258 if ((options & PCRE_UTF8) == 0)
4259 #endif
4260 if (end > 0xff) end = 0xff;
4261
4262 #elif defined COMPILE_PCRE16
4263 #ifdef SUPPORT_UTF
4264 if ((options & PCRE_UTF16) == 0)
4265 #endif
4266 if (end > 0xffff) end = 0xffff;
4267
4268 #endif /* COMPILE_PCRE[8|16] */
4269
4270 /* Use the bitmap for characters < 256. Otherwise use extra data.*/
4271
4272 for (c = start; c <= classbits_end; c++)
4273 {
4274 /* Regardless of start, c will always be <= 255. */
4275 SETBIT(classbits, c);
4276 n8++;
4277 }
4278
4279 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4280 if (start <= 0xff) start = 0xff + 1;
4281
4282 if (end >= start)
4283 {
4284 pcre_uchar *uchardata = *uchardptr;
4285 #ifdef SUPPORT_UTF
4286 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4287 {
4288 if (start < end)
4289 {
4290 *uchardata++ = XCL_RANGE;
4291 uchardata += PRIV(ord2utf)(start, uchardata);
4292 uchardata += PRIV(ord2utf)(end, uchardata);
4293 }
4294 else if (start == end)
4295 {
4296 *uchardata++ = XCL_SINGLE;
4297 uchardata += PRIV(ord2utf)(start, uchardata);
4298 }
4299 }
4300 else
4301 #endif /* SUPPORT_UTF */
4302
4303 /* Without UTF support, character values are constrained by the bit length,
4304 and can only be > 256 for 16-bit and 32-bit libraries. */
4305
4306 #ifdef COMPILE_PCRE8
4307 {}
4308 #else
4309 if (start < end)
4310 {
4311 *uchardata++ = XCL_RANGE;
4312 *uchardata++ = start;
4313 *uchardata++ = end;
4314 }
4315 else if (start == end)
4316 {
4317 *uchardata++ = XCL_SINGLE;
4318 *uchardata++ = start;
4319 }
4320 #endif
4321
4322 *uchardptr = uchardata; /* Updata extra data pointer */
4323 }
4324 #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4325
4326 return n8; /* Number of 8-bit characters */
4327 }
4328
4329
4330
4331
4332 /*************************************************
4333 * Add a list of characters to a class *
4334 *************************************************/
4335
4336 /* This function is used for adding a list of case-equivalent characters to a
4337 class, and also for adding a list of horizontal or vertical whitespace. If the
4338 list is in order (which it should be), ranges of characters are detected and
4339 handled appropriately. This function is mutually recursive with the function
4340 above.
4341
4342 Arguments:
4343 classbits the bit map for characters < 256
4344 uchardptr points to the pointer for extra data
4345 options the options word
4346 cd contains pointers to tables etc.
4347 p points to row of 32-bit values, terminated by NOTACHAR
4348 except character to omit; this is used when adding lists of
4349 case-equivalent characters to avoid including the one we
4350 already know about
4351
4352 Returns: the number of < 256 characters added
4353 the pointer to extra data is updated
4354 */
4355
4356 static int
4357 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4358 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4359 {
4360 int n8 = 0;
4361 while (p[0] < NOTACHAR)
4362 {
4363 int n = 0;
4364 if (p[0] != except)
4365 {
4366 while(p[n+1] == p[0] + n + 1) n++;
4367 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4368 }
4369 p += n + 1;
4370 }
4371 return n8;
4372 }
4373
4374
4375
4376 /*************************************************
4377 * Add characters not in a list to a class *
4378 *************************************************/
4379
4380 /* This function is used for adding the complement of a list of horizontal or
4381 vertical whitespace to a class. The list must be in order.
4382
4383 Arguments:
4384 classbits the bit map for characters < 256
4385 uchardptr points to the pointer for extra data
4386 options the options word
4387 cd contains pointers to tables etc.
4388 p points to row of 32-bit values, terminated by NOTACHAR
4389
4390 Returns: the number of < 256 characters added
4391 the pointer to extra data is updated
4392 */
4393
4394 static int
4395 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4396 int options, compile_data *cd, const pcre_uint32 *p)
4397 {
4398 BOOL utf = (options & PCRE_UTF8) != 0;
4399 int n8 = 0;
4400 if (p[0] > 0)
4401 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4402 while (p[0] < NOTACHAR)
4403 {
4404 while (p[1] == p[0] + 1) p++;
4405 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4406 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4407 p++;
4408 }
4409 return n8;
4410 }
4411
4412
4413
4414 /*************************************************
4415 * Compile one branch *
4416 *************************************************/
4417
4418 /* Scan the pattern, compiling it into the a vector. If the options are
4419 changed during the branch, the pointer is used to change the external options
4420 bits. This function is used during the pre-compile phase when we are trying
4421 to find out the amount of memory needed, as well as during the real compile
4422 phase. The value of lengthptr distinguishes the two phases.
4423
4424 Arguments:
4425 optionsptr pointer to the option bits
4426 codeptr points to the pointer to the current code point
4427 ptrptr points to the current pattern pointer
4428 errorcodeptr points to error code variable
4429 firstcharptr place to put the first required character
4430 firstcharflagsptr place to put the first character flags, or a negative number
4431 reqcharptr place to put the last required character
4432 reqcharflagsptr place to put the last required character flags, or a negative number
4433 bcptr points to current branch chain
4434 cond_depth conditional nesting depth
4435 cd contains pointers to tables etc.
4436 lengthptr NULL during the real compile phase
4437 points to length accumulator during pre-compile phase
4438
4439 Returns: TRUE on success
4440 FALSE, with *errorcodeptr set non-zero on error
4441 */
4442
4443 static BOOL
4444 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4445 const pcre_uchar **ptrptr, int *errorcodeptr,
4446 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4447 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4448 branch_chain *bcptr, int cond_depth,
4449 compile_data *cd, int *lengthptr)
4450 {
4451 int repeat_type, op_type;
4452 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4453 int bravalue = 0;
4454 int greedy_default, greedy_non_default;
4455 pcre_uint32 firstchar, reqchar;
4456 pcre_int32 firstcharflags, reqcharflags;
4457 pcre_uint32 zeroreqchar, zerofirstchar;
4458 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4459 pcre_int32 req_caseopt, reqvary, tempreqvary;
4460 int options = *optionsptr; /* May change dynamically */
4461 int after_manual_callout = 0;
4462 int length_prevgroup = 0;
4463 register pcre_uint32 c;
4464 int escape;
4465 register pcre_uchar *code = *codeptr;
4466 pcre_uchar *last_code = code;
4467 pcre_uchar *orig_code = code;
4468 pcre_uchar *tempcode;
4469 BOOL inescq = FALSE;
4470 BOOL groupsetfirstchar = FALSE;
4471 const pcre_uchar *ptr = *ptrptr;
4472 const pcre_uchar *tempptr;
4473 const pcre_uchar *nestptr = NULL;
4474 pcre_uchar *previous = NULL;
4475 pcre_uchar *previous_callout = NULL;
4476 size_t save_hwm_offset = 0;
4477 pcre_uint8 classbits[32];
4478
4479 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4480 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4481 dynamically as we process the pattern. */
4482
4483 #ifdef SUPPORT_UTF
4484 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4485 BOOL utf = (options & PCRE_UTF8) != 0;
4486 #ifndef COMPILE_PCRE32
4487 pcre_uchar utf_chars[6];
4488 #endif
4489 #else
4490 BOOL utf = FALSE;
4491 #endif
4492
4493 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4494 class_uchardata always so that it can be passed to add_to_class() always,
4495 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4496 alternative calls for the different cases. */
4497
4498 pcre_uchar *class_uchardata;
4499 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4500 BOOL xclass;
4501 pcre_uchar *class_uchardata_base;
4502 #endif
4503
4504 #ifdef PCRE_DEBUG
4505 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4506 #endif
4507
4508 /* Set up the default and non-default settings for greediness */
4509
4510 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4511 greedy_non_default = greedy_default ^ 1;
4512
4513 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4514 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4515 matches a non-fixed char first char; reqchar just remains unset if we never
4516 find one.
4517
4518 When we hit a repeat whose minimum is zero, we may have to adjust these values
4519 to take the zero repeat into account. This is implemented by setting them to
4520 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4521 item types that can be repeated set these backoff variables appropriately. */
4522
4523 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4524 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4525
4526 /* The variable req_caseopt contains either the REQ_CASELESS value
4527 or zero, according to the current setting of the caseless flag. The
4528 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4529 firstchar or reqchar variables to record the case status of the
4530 value. This is used only for ASCII characters. */
4531
4532 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4533
4534 /* Switch on next character until the end of the branch */
4535
4536 for (;; ptr++)
4537 {
4538 BOOL negate_class;
4539 BOOL should_flip_negation;
4540 BOOL possessive_quantifier;
4541 BOOL is_quantifier;
4542 BOOL is_recurse;
4543 BOOL reset_bracount;
4544 int class_has_8bitchar;
4545 int class_one_char;
4546 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4547 BOOL xclass_has_prop;
4548 #endif
4549 int newoptions;
4550 int recno;
4551 int refsign;
4552 int skipbytes;
4553 pcre_uint32 subreqchar, subfirstchar;
4554 pcre_int32 subreqcharflags, subfirstcharflags;
4555 int terminator;
4556 unsigned int mclength;
4557 unsigned int tempbracount;
4558 pcre_uint32 ec;
4559 pcre_uchar mcbuffer[8];
4560
4561 /* Get next character in the pattern */
4562
4563 c = *ptr;
4564
4565 /* If we are at the end of a nested substitution, revert to the outer level
4566 string. Nesting only happens one level deep. */
4567
4568 if (c == CHAR_NULL && nestptr != NULL)
4569 {
4570 ptr = nestptr;
4571 nestptr = NULL;
4572 c = *ptr;
4573 }
4574
4575 /* If we are in the pre-compile phase, accumulate the length used for the
4576 previous cycle of this loop. */
4577
4578 if (lengthptr != NULL)
4579 {
4580 #ifdef PCRE_DEBUG
4581 if (code > cd->hwm) cd->hwm = code; /* High water info */
4582 #endif
4583 if (code > cd->start_workspace + cd->workspace_size -
4584 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4585 {
4586 *errorcodeptr = ERR52;
4587 goto FAILED;
4588 }
4589
4590 /* There is at least one situation where code goes backwards: this is the
4591 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4592 the class is simply eliminated. However, it is created first, so we have to
4593 allow memory for it. Therefore, don't ever reduce the length at this point.
4594 */
4595
4596 if (code < last_code) code = last_code;
4597
4598 /* Paranoid check for integer overflow */
4599
4600 if (OFLOW_MAX - *lengthptr < code - last_code)
4601 {
4602 *errorcodeptr = ERR20;
4603 goto FAILED;
4604 }
4605
4606 *lengthptr += (int)(code - last_code);
4607 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4608 (int)(code - last_code), c, c));
4609
4610 /* If "previous" is set and it is not at the start of the work space, move
4611 it back to there, in order to avoid filling up the work space. Otherwise,
4612 if "previous" is NULL, reset the current code pointer to the start. */
4613
4614 if (previous != NULL)
4615 {
4616 if (previous > orig_code)
4617 {
4618 memmove(orig_code, previous, IN_UCHARS(code - previous));
4619 code -= previous - orig_code;
4620 previous = orig_code;
4621 }
4622 }
4623 else code = orig_code;
4624
4625 /* Remember where this code item starts so we can pick up the length
4626 next time round. */
4627
4628 last_code = code;
4629 }
4630
4631 /* In the real compile phase, just check the workspace used by the forward
4632 reference list. */
4633
4634 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4635 WORK_SIZE_SAFETY_MARGIN)
4636 {
4637 *errorcodeptr = ERR52;
4638 goto FAILED;
4639 }
4640
4641 /* If in \Q...\E, check for the end; if not, we have a literal */
4642
4643 if (inescq && c != CHAR_NULL)
4644 {
4645 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4646 {
4647 inescq = FALSE;
4648 ptr++;
4649 continue;
4650 }
4651 else
4652 {
4653 if (previous_callout != NULL)
4654 {
4655 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4656 complete_callout(previous_callout, ptr, cd);
4657 previous_callout = NULL;
4658 }
4659 if ((options & PCRE_AUTO_CALLOUT) != 0)
4660 {
4661 previous_callout = code;
4662 code = auto_callout(code, ptr, cd);
4663 }
4664 goto NORMAL_CHAR;
4665 }
4666 /* Control does not reach here. */
4667 }
4668
4669 /* In extended mode, skip white space and comments. We need a loop in order
4670 to check for more white space and more comments after a comment. */
4671
4672 if ((options & PCRE_EXTENDED) != 0)
4673 {
4674 for (;;)
4675 {
4676 while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4677 if (c != CHAR_NUMBER_SIGN) break;
4678 ptr++;
4679 while (*ptr != CHAR_NULL)
4680 {
4681 if (IS_NEWLINE(ptr)) /* For non-fixed-length newline cases, */
4682 { /* IS_NEWLINE sets cd->nllen. */
4683 ptr += cd->nllen;
4684 break;
4685 }
4686 ptr++;
4687 #ifdef SUPPORT_UTF
4688 if (utf) FORWARDCHAR(ptr);
4689 #endif
4690 }
4691 c = *ptr; /* Either NULL or the char after a newline */
4692 }
4693 }
4694
4695 /* See if the next thing is a quantifier. */
4696
4697 is_quantifier =
4698 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4699 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4700
4701 /* Fill in length of a previous callout, except when the next thing is a
4702 quantifier or when processing a property substitution string in UCP mode. */
4703
4704 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4705 after_manual_callout-- <= 0)
4706 {
4707 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4708 complete_callout(previous_callout, ptr, cd);
4709 previous_callout = NULL;
4710 }
4711
4712 /* Create auto callout, except for quantifiers, or while processing property
4713 strings that are substituted for \w etc in UCP mode. */
4714
4715 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4716 {
4717 previous_callout = code;
4718 code = auto_callout(code, ptr, cd);
4719 }
4720
4721 /* Process the next pattern item. */
4722
4723 switch(c)
4724 {
4725 /* ===================================================================*/
4726 case CHAR_NULL: /* The branch terminates at string end */
4727 case CHAR_VERTICAL_LINE: /* or | or ) */
4728 case CHAR_RIGHT_PARENTHESIS:
4729 *firstcharptr = firstchar;
4730 *firstcharflagsptr = firstcharflags;
4731 *reqcharptr = reqchar;
4732 *reqcharflagsptr = reqcharflags;
4733 *codeptr = code;
4734 *ptrptr = ptr;
4735 if (lengthptr != NULL)
4736 {
4737 if (OFLOW_MAX - *lengthptr < code - last_code)
4738 {
4739 *errorcodeptr = ERR20;
4740 goto FAILED;
4741 }
4742 *lengthptr += (int)(code - last_code); /* To include callout length */
4743 DPRINTF((">> end branch\n"));
4744 }
4745 return TRUE;
4746
4747
4748 /* ===================================================================*/
4749 /* Handle single-character metacharacters. In multiline mode, ^ disables
4750 the setting of any following char as a first character. */
4751
4752 case CHAR_CIRCUMFLEX_ACCENT:
4753 previous = NULL;
4754 if ((options & PCRE_MULTILINE) != 0)
4755 {
4756 if (firstcharflags == REQ_UNSET)
4757 zerofirstcharflags = firstcharflags = REQ_NONE;
4758 *code++ = OP_CIRCM;
4759 }
4760 else *code++ = OP_CIRC;
4761 break;
4762
4763 case CHAR_DOLLAR_SIGN:
4764 previous = NULL;
4765 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4766 break;
4767
4768 /* There can never be a first char if '.' is first, whatever happens about
4769 repeats. The value of reqchar doesn't change either. */
4770
4771 case CHAR_DOT:
4772 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4773 zerofirstchar = firstchar;
4774 zerofirstcharflags = firstcharflags;
4775 zeroreqchar = reqchar;
4776 zeroreqcharflags = reqcharflags;
4777 previous = code;
4778 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4779 break;
4780
4781
4782 /* ===================================================================*/
4783 /* Character classes. If the included characters are all < 256, we build a
4784 32-byte bitmap of the permitted characters, except in the special case
4785 where there is only one such character. For negated classes, we build the
4786 map as usual, then invert it at the end. However, we use a different opcode
4787 so that data characters > 255 can be handled correctly.
4788
4789 If the class contains characters outside the 0-255 range, a different
4790 opcode is compiled. It may optionally have a bit map for characters < 256,
4791 but those above are are explicitly listed afterwards. A flag byte tells
4792 whether the bitmap is present, and whether this is a negated class or not.
4793
4794 In JavaScript compatibility mode, an isolated ']' causes an error. In
4795 default (Perl) mode, it is treated as a data character. */
4796
4797 case CHAR_RIGHT_SQUARE_BRACKET:
4798 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4799 {
4800 *errorcodeptr = ERR64;
4801 goto FAILED;
4802 }
4803 goto NORMAL_CHAR;
4804
4805 /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4806 used for "start of word" and "end of word". As these are otherwise illegal
4807 sequences, we don't break anything by recognizing them. They are replaced
4808 by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4809 erroneous and are handled by the normal code below. */
4810
4811 case CHAR_LEFT_SQUARE_BRACKET:
4812 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4813 {
4814 nestptr = ptr + 7;
4815 ptr = sub_start_of_word - 1;
4816 continue;
4817 }
4818
4819 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4820 {
4821 nestptr = ptr + 7;
4822 ptr = sub_end_of_word - 1;
4823 continue;
4824 }
4825
4826 /* Handle a real character class. */
4827
4828 previous = code;
4829
4830 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4831 they are encountered at the top level, so we'll do that too. */
4832
4833 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4834 ptr[1] == CHAR_EQUALS_SIGN) &&
4835 check_posix_syntax(ptr, &tempptr))
4836 {
4837 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4838 goto FAILED;
4839 }
4840
4841 /* If the first character is '^', set the negation flag and skip it. Also,
4842 if the first few characters (either before or after ^) are \Q\E or \E we
4843 skip them too. This makes for compatibility with Perl. */
4844
4845 negate_class = FALSE;
4846 for (;;)
4847 {
4848 c = *(++ptr);
4849 if (c == CHAR_BACKSLASH)
4850 {
4851 if (ptr[1] == CHAR_E)
4852 ptr++;
4853 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4854 ptr += 3;
4855 else
4856 break;
4857 }
4858 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4859 negate_class = TRUE;
4860 else break;
4861 }
4862
4863 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4864 an initial ']' is taken as a data character -- the code below handles
4865 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4866 [^] must match any character, so generate OP_ALLANY. */
4867
4868 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4869 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4870 {
4871 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4872 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4873 zerofirstchar = firstchar;
4874 zerofirstcharflags = firstcharflags;
4875 break;
4876 }
4877
4878 /* If a class contains a negative special such as \S, we need to flip the
4879 negation flag at the end, so that support for characters > 255 works
4880 correctly (they are all included in the class). */
4881
4882 should_flip_negation = FALSE;
4883
4884 /* Extended class (xclass) will be used when characters > 255
4885 might match. */
4886
4887 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4888 xclass = FALSE;
4889 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4890 class_uchardata_base = class_uchardata; /* Save the start */
4891 #endif
4892
4893 /* For optimization purposes, we track some properties of the class:
4894 class_has_8bitchar will be non-zero if the class contains at least one <
4895 256 character; class_one_char will be 1 if the class contains just one
4896 character; xclass_has_prop will be TRUE if unicode property checks
4897 are present in the class. */
4898
4899 class_has_8bitchar = 0;
4900 class_one_char = 0;
4901 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4902 xclass_has_prop = FALSE;
4903 #endif
4904
4905 /* Initialize the 32-char bit map to all zeros. We build the map in a
4906 temporary bit of memory, in case the class contains fewer than two
4907 8-bit characters because in that case the compiled code doesn't use the bit
4908 map. */
4909
4910 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4911
4912 /* Process characters until ] is reached. By writing this as a "do" it
4913 means that an initial ] is taken as a data character. At the start of the
4914 loop, c contains the first byte of the character. */
4915
4916 if (c != CHAR_NULL) do
4917 {
4918 const pcre_uchar *oldptr;
4919
4920 #ifdef SUPPORT_UTF
4921 if (utf && HAS_EXTRALEN(c))
4922 { /* Braces are required because the */
4923 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4924 }
4925 #endif
4926
4927 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4928 /* In the pre-compile phase, accumulate the length of any extra
4929 data and reset the pointer. This is so that very large classes that
4930 contain a zillion > 255 characters no longer overwrite the work space
4931 (which is on the stack). We have to remember that there was XCLASS data,
4932 however. */
4933
4934 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4935 {
4936 xclass = TRUE;
4937 *lengthptr += (int)(class_uchardata - class_uchardata_base);
4938 class_uchardata = class_uchardata_base;
4939 }
4940 #endif
4941
4942 /* Inside \Q...\E everything is literal except \E */
4943
4944 if (inescq)
4945 {
4946 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4947 {
4948 inescq = FALSE; /* Reset literal state */
4949 ptr++; /* Skip the 'E' */
4950 continue; /* Carry on with next */
4951 }
4952 goto CHECK_RANGE; /* Could be range if \E follows */
4953 }
4954
4955 /* Handle POSIX class names. Perl allows a negation extension of the
4956 form [:^name:]. A square bracket that doesn't match the syntax is
4957 treated as a literal. We also recognize the POSIX constructions
4958 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4959 5.6 and 5.8 do. */
4960
4961 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4962 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4963 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4964 {
4965 BOOL local_negate = FALSE;
4966 int posix_class, taboffset, tabopt;
4967 register const pcre_uint8 *cbits = cd->cbits;
4968 pcre_uint8 pbits[32];
4969
4970 if (ptr[1] != CHAR_COLON)
4971 {
4972 *errorcodeptr = ERR31;
4973 goto FAILED;
4974 }
4975
4976 ptr += 2;
4977 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4978 {
4979 local_negate = TRUE;
4980 should_flip_negation = TRUE; /* Note negative special */
4981 ptr++;
4982 }
4983
4984 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4985 if (posix_class < 0)
4986 {
4987 *errorcodeptr = ERR30;
4988 goto FAILED;
4989 }
4990
4991 /* If matching is caseless, upper and lower are converted to
4992 alpha. This relies on the fact that the class table starts with
4993 alpha, lower, upper as the first 3 entries. */
4994
4995 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4996 posix_class = 0;
4997
4998 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4999 different escape sequences that use Unicode properties \p or \P. Others
5000 that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
5001 directly. */
5002
5003 #ifdef SUPPORT_UCP
5004 if ((options & PCRE_UCP) != 0)
5005 {
5006 unsigned int ptype = 0;
5007 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
5008
5009 /* The posix_substitutes table specifies which POSIX classes can be
5010 converted to \p or \P items. */
5011
5012 if (posix_substitutes[pc] != NULL)
5013 {
5014 nestptr = tempptr + 1;
5015 ptr = posix_substitutes[pc] - 1;
5016 continue;
5017 }
5018
5019 /* There are three other classes that generate special property calls
5020 that are recognized only in an XCLASS. */
5021
5022 else switch(posix_class)
5023 {
5024 case PC_GRAPH:
5025 ptype = PT_PXGRAPH;
5026 /* Fall through */
5027 case PC_PRINT:
5028 if (ptype == 0) ptype = PT_PXPRINT;
5029 /* Fall through */
5030 case PC_PUNCT:
5031 if (ptype == 0) ptype = PT_PXPUNCT;
5032 *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
5033 *class_uchardata++ = ptype;
5034 *class_uchardata++ = 0;
5035 xclass_has_prop = TRUE;
5036 ptr = tempptr + 1;
5037 continue;
5038
5039 /* For all other POSIX classes, no special action is taken in UCP
5040 mode. Fall through to the non_UCP case. */
5041
5042 default:
5043 break;
5044 }
5045 }
5046 #endif
5047 /* In the non-UCP case, or when UCP makes no difference, we build the
5048 bit map for the POSIX class in a chunk of local store because we may be
5049 adding and subtracting from it, and we don't want to subtract bits that
5050 may be in the main map already. At the end we or the result into the
5051 bit map that is being built. */
5052
5053 posix_class *= 3;
5054
5055 /* Copy in the first table (always present) */
5056
5057 memcpy(pbits, cbits + posix_class_maps[posix_class],
5058 32 * sizeof(pcre_uint8));
5059
5060 /* If there is a second table, add or remove it as required. */
5061
5062 taboffset = posix_class_maps[posix_class + 1];
5063 tabopt = posix_class_maps[posix_class + 2];
5064
5065 if (taboffset >= 0)
5066 {
5067 if (tabopt >= 0)
5068 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
5069 else
5070 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
5071 }
5072
5073 /* Now see if we need to remove any special characters. An option
5074 value of 1 removes vertical space and 2 removes underscore. */
5075
5076 if (tabopt < 0) tabopt = -tabopt;
5077 if (tabopt == 1) pbits[1] &= ~0x3c;
5078 else if (tabopt == 2) pbits[11] &= 0x7f;
5079
5080 /* Add the POSIX table or its complement into the main table that is
5081 being built and we are done. */
5082
5083 if (local_negate)
5084 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
5085 else
5086 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
5087
5088 ptr = tempptr + 1;
5089 /* Every class contains at least one < 256 character. */
5090 class_has_8bitchar = 1;
5091 /* Every class contains at least two characters. */
5092 class_one_char = 2;
5093 continue; /* End of POSIX syntax handling */
5094 }
5095
5096 /* Backslash may introduce a single character, or it may introduce one
5097 of the specials, which just set a flag. The sequence \b is a special
5098 case. Inside a class (and only there) it is treated as backspace. We
5099 assume that other escapes have more than one character in them, so
5100 speculatively set both class_has_8bitchar and class_one_char bigger
5101 than one. Unrecognized escapes fall through and are either treated
5102 as literal characters (by default), or are faulted if
5103 PCRE_EXTRA is set. */
5104
5105 if (c == CHAR_BACKSLASH)
5106 {
5107 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
5108 TRUE);
5109 if (*errorcodeptr != 0) goto FAILED;
5110 if (escape == 0) c = ec;
5111 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
5112 else if (escape == ESC_N) /* \N is not supported in a class */
5113 {
5114 *errorcodeptr = ERR71;
5115 goto FAILED;
5116 }
5117 else if (escape == ESC_Q) /* Handle start of quoted string */
5118 {
5119 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5120 {
5121 ptr += 2; /* avoid empty string */
5122 }
5123 else inescq = TRUE;
5124 continue;
5125 }
5126 else if (escape == ESC_E) continue; /* Ignore orphan \E */
5127
5128 else
5129 {
5130 register const pcre_uint8 *cbits = cd->cbits;
5131 /* Every class contains at least two < 256 characters. */
5132 class_has_8bitchar++;
5133 /* Every class contains at least two characters. */
5134 class_one_char += 2;
5135
5136 switch (escape)
5137 {
5138 #ifdef SUPPORT_UCP
5139 case ESC_du: /* These are the values given for \d etc */
5140 case ESC_DU: /* when PCRE_UCP is set. We replace the */
5141 case ESC_wu: /* escape sequence with an appropriate \p */
5142 case ESC_WU: /* or \P to test Unicode properties instead */
5143 case ESC_su: /* of the default ASCII testing. */
5144 case ESC_SU:
5145 nestptr = ptr;
5146 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
5147 class_has_8bitchar--; /* Undo! */
5148 continue;
5149 #endif
5150 case ESC_d:
5151 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
5152 continue;
5153
5154 case ESC_D:
5155 should_flip_negation = TRUE;
5156 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
5157 continue;
5158
5159 case ESC_w:
5160 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
5161 continue;
5162
5163 case ESC_W:
5164 should_flip_negation = TRUE;
5165 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5166 continue;
5167
5168 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5169 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5170 previously set by something earlier in the character class.
5171 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5172 we could just adjust the appropriate bit. From PCRE 8.34 we no
5173 longer treat \s and \S specially. */
5174
5175 case ESC_s:
5176 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
5177 continue;
5178
5179 case ESC_S:
5180 should_flip_negation = TRUE;
5181 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
5182 continue;
5183
5184 /* The rest apply in both UCP and non-UCP cases. */
5185
5186 case ESC_h:
5187 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5188 PRIV(hspace_list), NOTACHAR);
5189 continue;
5190
5191 case ESC_H:
5192 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5193 cd, PRIV(hspace_list));
5194 continue;
5195
5196 case ESC_v:
5197 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5198 PRIV(vspace_list), NOTACHAR);
5199 continue;
5200
5201 case ESC_V:
5202 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5203 cd, PRIV(vspace_list));
5204 continue;
5205
5206 #ifdef SUPPORT_UCP
5207 case ESC_p:
5208 case ESC_P:
5209 {
5210 BOOL negated;
5211 unsigned int ptype = 0, pdata = 0;
5212 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5213 goto FAILED;
5214 *class_uchardata++ = ((escape == ESC_p) != negated)?
5215 XCL_PROP : XCL_NOTPROP;
5216 *class_uchardata++ = ptype;
5217 *class_uchardata++ = pdata;
5218 xclass_has_prop = TRUE;
5219 class_has_8bitchar--; /* Undo! */
5220 continue;
5221 }
5222 #endif
5223 /* Unrecognized escapes are faulted if PCRE is running in its
5224 strict mode. By default, for compatibility with Perl, they are
5225 treated as literals. */
5226
5227 default:
5228 if ((options & PCRE_EXTRA) != 0)
5229 {
5230 *errorcodeptr = ERR7;
5231 goto FAILED;
5232 }
5233 class_has_8bitchar--; /* Undo the speculative increase. */
5234 class_one_char -= 2; /* Undo the speculative increase. */
5235 c = *ptr; /* Get the final character and fall through */
5236 break;
5237 }
5238 }
5239
5240 /* Fall through if the escape just defined a single character (c >= 0).
5241 This may be greater than 256. */
5242
5243 escape = 0;
5244
5245 } /* End of backslash handling */
5246
5247 /* A character may be followed by '-' to form a range. However, Perl does
5248 not permit ']' to be the end of the range. A '-' character at the end is
5249 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5250 code for handling \Q and \E is messy. */
5251
5252 CHECK_RANGE:
5253 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5254 {
5255 inescq = FALSE;
5256 ptr += 2;
5257 }
5258 oldptr = ptr;
5259
5260 /* Remember if \r or \n were explicitly used */
5261
5262 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5263
5264 /* Check for range */
5265
5266 if (!inescq && ptr[1] == CHAR_MINUS)
5267 {
5268 pcre_uint32 d;
5269 ptr += 2;
5270 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5271
5272 /* If we hit \Q (not followed by \E) at this point, go into escaped
5273 mode. */
5274
5275 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5276 {
5277 ptr += 2;
5278 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5279 { ptr += 2; continue; }
5280 inescq = TRUE;
5281 break;
5282 }
5283
5284 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5285 back the pointer and jump to handle the character that preceded it. */
5286
5287 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5288 {
5289 ptr = oldptr;
5290 goto CLASS_SINGLE_CHARACTER;
5291 }
5292
5293 /* Otherwise, we have a potential range; pick up the next character */
5294
5295 #ifdef SUPPORT_UTF
5296 if (utf)
5297 { /* Braces are required because the */
5298 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5299 }
5300 else
5301 #endif
5302 d = *ptr; /* Not UTF-8 mode */
5303
5304 /* The second part of a range can be a single-character escape
5305 sequence, but not any of the other escapes. Perl treats a hyphen as a
5306 literal in such circumstances. However, in Perl's warning mode, a
5307 warning is given, so PCRE now faults it as it is almost certainly a
5308 mistake on the user's part. */
5309
5310 if (!inescq)
5311 {
5312 if (d == CHAR_BACKSLASH)
5313 {
5314 int descape;
5315 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5316 if (*errorcodeptr != 0) goto FAILED;
5317
5318 /* 0 means a character was put into d; \b is backspace; any other
5319 special causes an error. */
5320
5321 if (descape != 0)
5322 {
5323 if (descape == ESC_b) d = CHAR_BS; else
5324 {
5325 *errorcodeptr = ERR83;
5326 goto FAILED;
5327 }
5328 }
5329 }
5330
5331 /* A hyphen followed by a POSIX class is treated in the same way. */
5332
5333 else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5334 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5335 ptr[1] == CHAR_EQUALS_SIGN) &&
5336 check_posix_syntax(ptr, &tempptr))
5337 {
5338 *errorcodeptr = ERR83;
5339 goto FAILED;
5340 }
5341 }
5342
5343 /* Check that the two values are in the correct order. Optimize
5344 one-character ranges. */
5345
5346 if (d < c)
5347 {
5348 *errorcodeptr = ERR8;
5349 goto FAILED;
5350 }
5351 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5352
5353 /* We have found a character range, so single character optimizations
5354 cannot be done anymore. Any value greater than 1 indicates that there
5355 is more than one character. */
5356
5357 class_one_char = 2;
5358
5359 /* Remember an explicit \r or \n, and add the range to the class. */
5360
5361 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5362
5363 class_has_8bitchar +=
5364 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5365
5366 continue; /* Go get the next char in the class */
5367 }
5368
5369 /* Handle a single character - we can get here for a normal non-escape
5370 char, or after \ that introduces a single character or for an apparent
5371 range that isn't. Only the value 1 matters for class_one_char, so don't
5372 increase it if it is already 2 or more ... just in case there's a class
5373 with a zillion characters in it. */
5374
5375 CLASS_SINGLE_CHARACTER:
5376 if (class_one_char < 2) class_one_char++;
5377
5378 /* If class_one_char is 1, we have the first single character in the
5379 class, and there have been no prior ranges, or XCLASS items generated by
5380 escapes. If this is the final character in the class, we can optimize by
5381 turning the item into a 1-character OP_CHAR[I] if it's positive, or
5382 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5383 to be set. Otherwise, there can be no first char if this item is first,
5384 whatever repeat count may follow. In the case of reqchar, save the
5385 previous value for reinstating. */
5386
5387 if (!inescq && class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5388 {
5389 ptr++;
5390 zeroreqchar = reqchar;
5391 zeroreqcharflags = reqcharflags;
5392
5393 if (negate_class)
5394 {
5395 #ifdef SUPPORT_UCP
5396 int d;
5397 #endif
5398 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5399 zerofirstchar = firstchar;
5400 zerofirstcharflags = firstcharflags;
5401
5402 /* For caseless UTF-8 mode when UCP support is available, check
5403 whether this character has more than one other case. If so, generate
5404 a special OP_NOTPROP item instead of OP_NOTI. */
5405
5406 #ifdef SUPPORT_UCP
5407 if (utf && (options & PCRE_CASELESS) != 0 &&
5408 (d = UCD_CASESET(c)) != 0)
5409 {
5410 *code++ = OP_NOTPROP;
5411 *code++ = PT_CLIST;
5412 *code++ = d;
5413 }
5414 else
5415 #endif
5416 /* Char has only one other case, or UCP not available */
5417
5418 {
5419 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5420 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5421 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5422 code += PRIV(ord2utf)(c, code);
5423 else
5424 #endif
5425 *code++ = c;
5426 }
5427
5428 /* We are finished with this character class */
5429
5430 goto END_CLASS;
5431 }
5432
5433 /* For a single, positive character, get the value into mcbuffer, and
5434 then we can handle this with the normal one-character code. */
5435
5436 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5437 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5438 mclength = PRIV(ord2utf)(c, mcbuffer);
5439 else
5440 #endif
5441 {
5442 mcbuffer[0] = c;
5443 mclength = 1;
5444 }
5445 goto ONE_CHAR;
5446 } /* End of 1-char optimization */
5447
5448 /* There is more than one character in the class, or an XCLASS item
5449 has been generated. Add this character to the class. */
5450
5451 class_has_8bitchar +=
5452 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5453 }
5454
5455 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5456 If we are at the end of an internal nested string, revert to the outer
5457 string. */
5458
5459 while (((c = *(++ptr)) != CHAR_NULL ||
5460 (nestptr != NULL &&
5461 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5462 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5463
5464 /* Check for missing terminating ']' */
5465
5466 if (c == CHAR_NULL)
5467 {
5468 *errorcodeptr = ERR6;
5469 goto FAILED;
5470 }
5471
5472 /* We will need an XCLASS if data has been placed in class_uchardata. In
5473 the second phase this is a sufficient test. However, in the pre-compile
5474 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5475 only if the very last character in the class needs XCLASS will it contain
5476 anything at this point. For this reason, xclass gets set TRUE above when
5477 uchar_classdata is emptied, and that's why this code is the way it is here
5478 instead of just doing a test on class_uchardata below. */
5479
5480 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5481 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5482 #endif
5483
5484 /* If this is the first thing in the branch, there can be no first char
5485 setting, whatever the repeat count. Any reqchar setting must remain
5486 unchanged after any kind of repeat. */
5487
5488 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5489 zerofirstchar = firstchar;
5490 zerofirstcharflags = firstcharflags;
5491 zeroreqchar = reqchar;
5492 zeroreqcharflags = reqcharflags;
5493
5494 /* If there are characters with values > 255, we have to compile an
5495 extended class, with its own opcode, unless there was a negated special
5496 such as \S in the class, and PCRE_UCP is not set, because in that case all
5497 characters > 255 are in the class, so any that were explicitly given as
5498 well can be ignored. If (when there are explicit characters > 255 that must
5499 be listed) there are no characters < 256, we can omit the bitmap in the
5500 actual compiled code. */
5501
5502 #ifdef SUPPORT_UTF
5503 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5504 #elif !defined COMPILE_PCRE8
5505 if (xclass && !should_flip_negation)
5506 #endif
5507 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5508 {
5509 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5510 *code++ = OP_XCLASS;
5511 code += LINK_SIZE;
5512 *code = negate_class? XCL_NOT:0;
5513 if (xclass_has_prop) *code |= XCL_HASPROP;
5514
5515 /* If the map is required, move up the extra data to make room for it;
5516 otherwise just move the code pointer to the end of the extra data. */
5517
5518 if (class_has_8bitchar > 0)
5519 {
5520 *code++ |= XCL_MAP;
5521 memmove(code + (32 / sizeof(pcre_uchar)), code,
5522 IN_UCHARS(class_uchardata - code));
5523 if (negate_class && !xclass_has_prop)
5524 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5525 memcpy(code, classbits, 32);
5526 code = class_uchardata + (32 / sizeof(pcre_uchar));
5527 }
5528 else code = class_uchardata;
5529
5530 /* Now fill in the complete length of the item */
5531
5532 PUT(previous, 1, (int)(code - previous));
5533 break; /* End of class handling */
5534 }
5535 #endif
5536
5537 /* Even though any XCLASS list is now discarded, we must allow for
5538 its memory. */
5539
5540 if (lengthptr != NULL)
5541 *lengthptr += (int)(class_uchardata - class_uchardata_base);
5542
5543 /* If there are no characters > 255, or they are all to be included or
5544 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5545 whole class was negated and whether there were negative specials such as \S
5546 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5547 negating it if necessary. */
5548
5549 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5550 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5551 {
5552 if (negate_class)
5553 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5554 memcpy(code, classbits, 32);
5555 }
5556 code += 32 / sizeof(pcre_uchar);
5557
5558 END_CLASS:
5559 break;
5560
5561
5562 /* ===================================================================*/
5563 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5564 has been tested above. */
5565
5566 case CHAR_LEFT_CURLY_BRACKET:
5567 if (!is_quantifier) goto NORMAL_CHAR;
5568 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5569 if (*errorcodeptr != 0) goto FAILED;
5570 goto REPEAT;
5571
5572 case CHAR_ASTERISK:
5573 repeat_min = 0;
5574 repeat_max = -1;
5575 goto REPEAT;
5576
5577 case CHAR_PLUS:
5578 repeat_min = 1;
5579 repeat_max = -1;
5580 goto REPEAT;
5581
5582 case CHAR_QUESTION_MARK:
5583 repeat_min = 0;
5584 repeat_max = 1;
5585
5586 REPEAT:
5587 if (previous == NULL)
5588 {
5589 *errorcodeptr = ERR9;
5590 goto FAILED;
5591 }
5592
5593 if (repeat_min == 0)
5594 {
5595 firstchar = zerofirstchar; /* Adjust for zero repeat */
5596 firstcharflags = zerofirstcharflags;
5597 reqchar = zeroreqchar; /* Ditto */
5598 reqcharflags = zeroreqcharflags;
5599 }
5600
5601 /* Remember whether this is a variable length repeat */
5602
5603 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5604
5605 op_type = 0; /* Default single-char op codes */
5606 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5607
5608 /* Save start of previous item, in case we have to move it up in order to
5609 insert something before it. */
5610
5611 tempcode = previous;
5612
5613 /* Before checking for a possessive quantifier, we must skip over
5614 whitespace and comments in extended mode because Perl allows white space at
5615 this point. */
5616
5617 if ((options & PCRE_EXTENDED) != 0)
5618 {
5619 const pcre_uchar *p = ptr + 1;
5620 for (;;)
5621 {
5622 while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5623 if (*p != CHAR_NUMBER_SIGN) break;
5624 p++;
5625 while (*p != CHAR_NULL)
5626 {
5627 if (IS_NEWLINE(p)) /* For non-fixed-length newline cases, */
5628 { /* IS_NEWLINE sets cd->nllen. */
5629 p += cd->nllen;
5630 break;
5631 }
5632 p++;
5633 #ifdef SUPPORT_UTF
5634 if (utf) FORWARDCHAR(p);
5635 #endif
5636 } /* Loop for comment characters */
5637 } /* Loop for multiple comments */
5638 ptr = p - 1; /* Character before the next significant one. */
5639 }
5640
5641 /* If the next character is '+', we have a possessive quantifier. This
5642 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5643 If the next character is '?' this is a minimizing repeat, by default,
5644 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5645 repeat type to the non-default. */
5646
5647 if (ptr[1] == CHAR_PLUS)
5648 {
5649 repeat_type = 0; /* Force greedy */
5650 possessive_quantifier = TRUE;
5651 ptr++;
5652 }
5653 else if (ptr[1] == CHAR_QUESTION_MARK)
5654 {
5655 repeat_type = greedy_non_default;
5656 ptr++;
5657 }
5658 else repeat_type = greedy_default;
5659
5660 /* If previous was a recursion call, wrap it in atomic brackets so that
5661 previous becomes the atomic group. All recursions were so wrapped in the
5662 past, but it no longer happens for non-repeated recursions. In fact, the
5663 repeated ones could be re-implemented independently so as not to need this,
5664 but for the moment we rely on the code for repeating groups. */
5665
5666 if (*previous == OP_RECURSE)
5667 {
5668 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5669 *previous = OP_ONCE;
5670 PUT(previous, 1, 2 + 2*LINK_SIZE);
5671 previous[2 + 2*LINK_SIZE] = OP_KET;
5672 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5673 code += 2 + 2 * LINK_SIZE;
5674 length_prevgroup = 3 + 3*LINK_SIZE;
5675
5676 /* When actually compiling, we need to check whether this was a forward
5677 reference, and if so, adjust the offset. */
5678
5679 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5680 {
5681 int offset = GET(cd->hwm, -LINK_SIZE);
5682 if (offset == previous + 1 - cd->start_code)
5683 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5684 }
5685 }
5686
5687 /* Now handle repetition for the different types of item. */
5688
5689 /* If previous was a character or negated character match, abolish the item
5690 and generate a repeat item instead. If a char item has a minimum of more
5691 than one, ensure that it is set in reqchar - it might not be if a sequence
5692 such as x{3} is the first thing in a branch because the x will have gone
5693 into firstchar instead. */
5694
5695 if (*previous == OP_CHAR || *previous == OP_CHARI
5696 || *previous == OP_NOT || *previous == OP_NOTI)
5697 {
5698 switch (*previous)
5699 {
5700 default: /* Make compiler happy. */
5701 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5702 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5703 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5704 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5705 }
5706
5707 /* Deal with UTF characters that take up more than one character. It's
5708 easier to write this out separately than try to macrify it. Use c to
5709 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5710 it's a length rather than a small character. */
5711
5712 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5713 if (utf && NOT_FIRSTCHAR(code[-1]))
5714 {
5715 pcre_uchar *lastchar = code - 1;
5716 BACKCHAR(lastchar);
5717 c = (int)(code - lastchar); /* Length of UTF-8 character */
5718 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5719 c |= UTF_LENGTH; /* Flag c as a length */
5720 }
5721 else
5722 #endif /* SUPPORT_UTF */
5723
5724 /* Handle the case of a single charater - either with no UTF support, or
5725 with UTF disabled, or for a single character UTF character. */
5726 {
5727 c = code[-1];
5728 if (*previous <= OP_CHARI && repeat_min > 1)
5729 {
5730 reqchar = c;
5731 reqcharflags = req_caseopt | cd->req_varyopt;
5732 }
5733 }
5734
5735 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5736 }
5737
5738 /* If previous was a character type match (\d or similar), abolish it and
5739 create a suitable repeat item. The code is shared with single-character
5740 repeats by setting op_type to add a suitable offset into repeat_type. Note
5741 the the Unicode property types will be present only when SUPPORT_UCP is
5742 defined, but we don't wrap the little bits of code here because it just
5743 makes it horribly messy. */
5744
5745 else if (*previous < OP_EODN)
5746 {
5747 pcre_uchar *oldcode;
5748 int prop_type, prop_value;
5749 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5750 c = *previous;
5751
5752 OUTPUT_SINGLE_REPEAT:
5753 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5754 {
5755 prop_type = previous[1];
5756 prop_value = previous[2];
5757 }
5758 else prop_type = prop_value = -1;
5759
5760 oldcode = code;
5761 code = previous; /* Usually overwrite previous item */
5762
5763 /* If the maximum is zero then the minimum must also be zero; Perl allows
5764 this case, so we do too - by simply omitting the item altogether. */
5765
5766 if (repeat_max == 0) goto END_REPEAT;
5767
5768 /* Combine the op_type with the repeat_type */
5769
5770 repeat_type += op_type;
5771
5772 /* A minimum of zero is handled either as the special case * or ?, or as
5773 an UPTO, with the maximum given. */
5774
5775 if (repeat_min == 0)
5776 {
5777 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5778 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5779 else
5780 {
5781 *code++ = OP_UPTO + repeat_type;
5782 PUT2INC(code, 0, repeat_max);
5783 }
5784 }
5785
5786 /* A repeat minimum of 1 is optimized into some special cases. If the
5787 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5788 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5789 one less than the maximum. */
5790
5791 else if (repeat_min == 1)
5792 {
5793 if (repeat_max == -1)
5794 *code++ = OP_PLUS + repeat_type;
5795 else
5796 {
5797 code = oldcode; /* leave previous item in place */
5798 if (repeat_max == 1) goto END_REPEAT;
5799 *code++ = OP_UPTO + repeat_type;
5800 PUT2INC(code, 0, repeat_max - 1);
5801 }
5802 }
5803
5804 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5805 handled as an EXACT followed by an UPTO. */
5806
5807 else
5808 {
5809 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5810 PUT2INC(code, 0, repeat_min);
5811
5812 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5813 we have to insert the character for the previous code. For a repeated
5814 Unicode property match, there are two extra bytes that define the
5815 required property. In UTF-8 mode, long characters have their length in
5816 c, with the UTF_LENGTH bit as a flag. */
5817
5818 if (repeat_max < 0)
5819 {
5820 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5821 if (utf && (c & UTF_LENGTH) != 0)
5822 {
5823 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5824 code += c & 7;
5825 }
5826 else
5827 #endif
5828 {
5829 *code++ = c;
5830 if (prop_type >= 0)
5831 {
5832 *code++ = prop_type;
5833 *code++ = prop_value;
5834 }
5835 }
5836 *code++ = OP_STAR + repeat_type;
5837 }
5838
5839 /* Else insert an UPTO if the max is greater than the min, again
5840 preceded by the character, for the previously inserted code. If the
5841 UPTO is just for 1 instance, we can use QUERY instead. */
5842
5843 else if (repeat_max != repeat_min)
5844 {
5845 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5846 if (utf && (c & UTF_LENGTH) != 0)
5847 {
5848 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5849 code += c & 7;
5850 }
5851 else
5852 #endif
5853 *code++ = c;
5854 if (prop_type >= 0)
5855 {
5856 *code++ = prop_type;
5857 *code++ = prop_value;
5858 }
5859 repeat_max -= repeat_min;
5860
5861 if (repeat_max == 1)
5862 {
5863 *code++ = OP_QUERY + repeat_type;
5864 }
5865 else
5866 {
5867 *code++ = OP_UPTO + repeat_type;
5868 PUT2INC(code, 0, repeat_max);
5869 }
5870 }
5871 }
5872
5873 /* The character or character type itself comes last in all cases. */
5874
5875 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5876 if (utf && (c & UTF_LENGTH) != 0)
5877 {
5878 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5879 code += c & 7;
5880 }
5881 else
5882 #endif
5883 *code++ = c;
5884
5885 /* For a repeated Unicode property match, there are two extra bytes that
5886 define the required property. */
5887
5888 #ifdef SUPPORT_UCP
5889 if (prop_type >= 0)
5890 {
5891 *code++ = prop_type;
5892 *code++ = prop_value;
5893 }
5894 #endif
5895 }
5896
5897 /* If previous was a character class or a back reference, we put the repeat
5898 stuff after it, but just skip the item if the repeat was {0,0}. */
5899
5900 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5901 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5902 *previous == OP_XCLASS ||
5903 #endif
5904 *previous == OP_REF || *previous == OP_REFI ||
5905 *previous == OP_DNREF || *previous == OP_DNREFI)
5906 {
5907 if (repeat_max == 0)
5908 {
5909 code = previous;
5910 goto END_REPEAT;
5911 }
5912
5913 if (repeat_min == 0 && repeat_max == -1)
5914 *code++ = OP_CRSTAR + repeat_type;
5915 else if (repeat_min == 1 && repeat_max == -1)
5916 *code++ = OP_CRPLUS + repeat_type;
5917 else if (repeat_min == 0 && repeat_max == 1)
5918 *code++ = OP_CRQUERY + repeat_type;
5919 else
5920 {
5921 *code++ = OP_CRRANGE + repeat_type;
5922 PUT2INC(code, 0, repeat_min);
5923 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5924 PUT2INC(code, 0, repeat_max);
5925 }
5926 }
5927
5928 /* If previous was a bracket group, we may have to replicate it in certain
5929 cases. Note that at this point we can encounter only the "basic" bracket
5930 opcodes such as BRA and CBRA, as this is the place where they get converted
5931 into the more special varieties such as BRAPOS and SBRA. A test for >=
5932 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5933 ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5934 Originally, PCRE did not allow repetition of assertions, but now it does,
5935 for Perl compatibility. */
5936
5937 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5938 {
5939 register int i;
5940 int len = (int)(code - previous);
5941 size_t base_hwm_offset = save_hwm_offset;
5942 pcre_uchar *bralink = NULL;
5943 pcre_uchar *brazeroptr = NULL;
5944
5945 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5946 we just ignore the repeat. */
5947
5948 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5949 goto END_REPEAT;
5950
5951 /* There is no sense in actually repeating assertions. The only potential
5952 use of repetition is in cases when the assertion is optional. Therefore,
5953 if the minimum is greater than zero, just ignore the repeat. If the
5954 maximum is not zero or one, set it to 1. */
5955
5956 if (*previous < OP_ONCE) /* Assertion */
5957 {
5958 if (repeat_min > 0) goto END_REPEAT;
5959 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5960 }
5961
5962 /* The case of a zero minimum is special because of the need to stick
5963 OP_BRAZERO in front of it, and because the group appears once in the
5964 data, whereas in other cases it appears the minimum number of times. For
5965 this reason, it is simplest to treat this case separately, as otherwise
5966 the code gets far too messy. There are several special subcases when the
5967 minimum is zero. */
5968
5969 if (repeat_min == 0)
5970 {
5971 /* If the maximum is also zero, we used to just omit the group from the
5972 output altogether, like this:
5973
5974 ** if (repeat_max == 0)
5975 ** {
5976 ** code = previous;
5977 ** goto END_REPEAT;
5978 ** }
5979
5980 However, that fails when a group or a subgroup within it is referenced
5981 as a subroutine from elsewhere in the pattern, so now we stick in
5982 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5983 don't have a list of which groups are referenced, we cannot do this
5984 selectively.
5985
5986 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5987 and do no more at this point. However, we do need to adjust any
5988 OP_RECURSE calls inside the group that refer to the group itself or any
5989 internal or forward referenced group, because the offset is from the
5990 start of the whole regex. Temporarily terminate the pattern while doing
5991 this. */
5992
5993 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5994 {
5995 *code = OP_END;
5996 adjust_recurse(previous, 1, utf, cd, save_hwm_offset);
5997 memmove(previous + 1, previous, IN_UCHARS(len));
5998 code++;
5999 if (repeat_max == 0)
6000 {
6001 *previous++ = OP_SKIPZERO;
6002 goto END_REPEAT;
6003 }
6004 brazeroptr = previous; /* Save for possessive optimizing */
6005 *previous++ = OP_BRAZERO + repeat_type;
6006 }
6007
6008 /* If the maximum is greater than 1 and limited, we have to replicate
6009 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
6010 The first one has to be handled carefully because it's the original
6011 copy, which has to be moved up. The remainder can be handled by code
6012 that is common with the non-zero minimum case below. We have to
6013 adjust the value or repeat_max, since one less copy is required. Once
6014 again, we may have to adjust any OP_RECURSE calls inside the group. */
6015
6016 else
6017 {
6018 int offset;
6019 *code = OP_END;
6020 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm_offset);
6021 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
6022 code += 2 + LINK_SIZE;
6023 *previous++ = OP_BRAZERO + repeat_type;
6024 *previous++ = OP_BRA;
6025
6026 /* We chain together the bracket offset fields that have to be
6027 filled in later when the ends of the brackets are reached. */
6028
6029 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
6030 bralink = previous;
6031 PUTINC(previous, 0, offset);
6032 }
6033
6034 repeat_max--;
6035 }
6036
6037 /* If the minimum is greater than zero, replicate the group as many
6038 times as necessary, and adjust the maximum to the number of subsequent
6039 copies that we need. If we set a first char from the group, and didn't
6040 set a required char, copy the latter from the former. If there are any
6041 forward reference subroutine calls in the group, there will be entries on
6042 the workspace list; replicate these with an appropriate increment. */
6043
6044 else
6045 {
6046 if (repeat_min > 1)
6047 {
6048 /* In the pre-compile phase, we don't actually do the replication. We
6049 just adjust the length as if we had. Do some paranoid checks for
6050 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
6051 integer type when available, otherwise double. */
6052
6053 if (lengthptr != NULL)
6054 {
6055 int delta = (repeat_min - 1)*length_prevgroup;
6056 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
6057 (INT64_OR_DOUBLE)length_prevgroup >
6058 (INT64_OR_DOUBLE)INT_MAX ||
6059 OFLOW_MAX - *lengthptr < delta)
6060 {
6061 *errorcodeptr = ERR20;
6062 goto FAILED;
6063 }
6064 *lengthptr += delta;
6065 }
6066
6067 /* This is compiling for real. If there is a set first byte for
6068 the group, and we have not yet set a "required byte", set it. Make
6069 sure there is enough workspace for copying forward references before
6070 doing the copy. */
6071
6072 else
6073 {
6074 if (groupsetfirstchar && reqcharflags < 0)
6075 {
6076 reqchar = firstchar;
6077 reqcharflags = firstcharflags;
6078 }
6079
6080 for (i = 1; i < repeat_min; i++)
6081 {
6082 pcre_uchar *hc;
6083 size_t this_hwm_offset = cd->hwm - cd->start_workspace;
6084 memcpy(code, previous, IN_UCHARS(len));
6085
6086 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6087 WORK_SIZE_SAFETY_MARGIN -
6088 (this_hwm_offset - base_hwm_offset))
6089 {
6090 *errorcodeptr = expand_workspace(cd);
6091 if (*errorcodeptr != 0) goto FAILED;
6092 }
6093
6094 for (hc = (pcre_uchar *)cd->start_workspace + base_hwm_offset;
6095 hc < (pcre_uchar *)cd->start_workspace + this_hwm_offset;
6096 hc += LINK_SIZE)
6097 {
6098 PUT(cd->hwm, 0, GET(hc, 0) + len);
6099 cd->hwm += LINK_SIZE;
6100 }
6101 base_hwm_offset = this_hwm_offset;
6102 code += len;
6103 }
6104 }
6105 }
6106
6107 if (repeat_max > 0) repeat_max -= repeat_min;
6108 }
6109
6110 /* This code is common to both the zero and non-zero minimum cases. If
6111 the maximum is limited, it replicates the group in a nested fashion,
6112 remembering the bracket starts on a stack. In the case of a zero minimum,
6113 the first one was set up above. In all cases the repeat_max now specifies
6114 the number of additional copies needed. Again, we must remember to
6115 replicate entries on the forward reference list. */
6116
6117 if (repeat_max >= 0)
6118 {
6119 /* In the pre-compile phase, we don't actually do the replication. We
6120 just adjust the length as if we had. For each repetition we must add 1
6121 to the length for BRAZERO and for all but the last repetition we must
6122 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
6123 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
6124 a 64-bit integer type when available, otherwise double. */
6125
6126 if (lengthptr != NULL && repeat_max > 0)
6127 {
6128 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
6129 2 - 2*LINK_SIZE; /* Last one doesn't nest */
6130 if ((INT64_OR_DOUBLE)repeat_max *
6131 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
6132 > (INT64_OR_DOUBLE)INT_MAX ||
6133 OFLOW_MAX - *lengthptr < delta)
6134 {
6135 *errorcodeptr = ERR20;
6136 goto FAILED;
6137 }
6138 *lengthptr += delta;
6139 }
6140
6141 /* This is compiling for real */
6142
6143 else for (i = repeat_max - 1; i >= 0; i--)
6144 {
6145 pcre_uchar *hc;
6146 size_t this_hwm_offset = cd->hwm - cd->start_workspace;
6147
6148 *code++ = OP_BRAZERO + repeat_type;
6149
6150 /* All but the final copy start a new nesting, maintaining the
6151 chain of brackets outstanding. */
6152
6153 if (i != 0)
6154 {
6155 int offset;
6156 *code++ = OP_BRA;
6157 offset = (bralink == NULL)? 0 : (int)(code - bralink);
6158 bralink = code;
6159 PUTINC(code, 0, offset);
6160 }
6161
6162 memcpy(code, previous, IN_UCHARS(len));
6163
6164 /* Ensure there is enough workspace for forward references before
6165 copying them. */
6166
6167 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6168 WORK_SIZE_SAFETY_MARGIN -
6169 (this_hwm_offset - base_hwm_offset))
6170 {
6171 *errorcodeptr = expand_workspace(cd);
6172 if (*errorcodeptr != 0) goto FAILED;
6173 }
6174
6175 for (hc = (pcre_uchar *)cd->start_workspace + base_hwm_offset;
6176 hc < (pcre_uchar *)cd->start_workspace + this_hwm_offset;
6177 hc += LINK_SIZE)
6178 {
6179 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
6180 cd->hwm += LINK_SIZE;
6181 }
6182 base_hwm_offset = this_hwm_offset;
6183 code += len;
6184 }
6185
6186 /* Now chain through the pending brackets, and fill in their length
6187 fields (which are holding the chain links pro tem). */
6188
6189 while (bralink != NULL)
6190 {
6191 int oldlinkoffset;
6192 int offset = (int)(code - bralink + 1);
6193 pcre_uchar *bra = code - offset;
6194 oldlinkoffset = GET(bra, 1);
6195 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
6196 *code++ = OP_KET;
6197 PUTINC(code, 0, offset);
6198 PUT(bra, 1, offset);
6199 }
6200 }
6201
6202 /* If the maximum is unlimited, set a repeater in the final copy. For
6203 ONCE brackets, that's all we need to do. However, possessively repeated
6204 ONCE brackets can be converted into non-capturing brackets, as the
6205 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
6206 deal with possessive ONCEs specially.
6207
6208 Otherwise, when we are doing the actual compile phase, check to see
6209 whether this group is one that could match an empty string. If so,
6210 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
6211 that runtime checking can be done. [This check is also applied to ONCE
6212 groups at runtime, but in a different way.]
6213
6214 Then, if the quantifier was possessive and the bracket is not a
6215 conditional, we convert the BRA code to the POS form, and the KET code to
6216 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
6217 subpattern at both the start and at the end.) The use of special opcodes
6218 makes it possible to reduce greatly the stack usage in pcre_exec(). If
6219 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
6220
6221 Then, if the minimum number of matches is 1 or 0, cancel the possessive
6222 flag so that the default action below, of wrapping everything inside
6223 atomic brackets, does not happen. When the minimum is greater than 1,
6224 there will be earlier copies of the group, and so we still have to wrap
6225 the whole thing. */
6226
6227 else
6228 {
6229 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
6230 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
6231
6232 /* Convert possessive ONCE brackets to non-capturing */
6233
6234 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
6235 possessive_quantifier) *bracode = OP_BRA;
6236
6237 /* For non-possessive ONCE brackets, all we need to do is to
6238 set the KET. */
6239
6240 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
6241 *ketcode = OP_KETRMAX + repeat_type;
6242
6243 /* Handle non-ONCE brackets and possessive ONCEs (which have been
6244 converted to non-capturing above). */
6245
6246 else
6247 {
6248 /* In the compile phase, check for empty string matching. */
6249
6250 if (lengthptr == NULL)
6251 {
6252 pcre_uchar *scode = bracode;
6253 do
6254 {
6255 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6256 {
6257 *bracode += OP_SBRA - OP_BRA;
6258 break;
6259 }
6260 scode += GET(scode, 1);
6261 }
6262 while (*scode == OP_ALT);
6263 }
6264
6265 /* Handle possessive quantifiers. */
6266
6267 if (possessive_quantifier)
6268 {
6269 /* For COND brackets, we wrap the whole thing in a possessively
6270 repeated non-capturing bracket, because we have not invented POS
6271 versions of the COND opcodes. Because we are moving code along, we
6272 must ensure that any pending recursive references are updated. */
6273
6274 if (*bracode == OP_COND || *bracode == OP_SCOND)
6275 {
6276 int nlen = (int)(code - bracode);
6277 *code = OP_END;
6278 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm_offset);
6279 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6280 code += 1 + LINK_SIZE;
6281 nlen += 1 + LINK_SIZE;
6282 *bracode = OP_BRAPOS;
6283 *code++ = OP_KETRPOS;
6284 PUTINC(code, 0, nlen);
6285 PUT(bracode, 1, nlen);
6286 }
6287
6288 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6289
6290 else
6291 {
6292 *bracode += 1; /* Switch to xxxPOS opcodes */
6293 *ketcode = OP_KETRPOS;
6294 }
6295
6296 /* If the minimum is zero, mark it as possessive, then unset the
6297 possessive flag when the minimum is 0 or 1. */
6298
6299 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6300 if (repeat_min < 2) possessive_quantifier = FALSE;
6301 }
6302
6303 /* Non-possessive quantifier */
6304
6305 else *ketcode = OP_KETRMAX + repeat_type;
6306 }
6307 }
6308 }
6309
6310 /* If previous is OP_FAIL, it was generated by an empty class [] in
6311 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6312 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6313 error above. We can just ignore the repeat in JS case. */
6314
6315 else if (*previous == OP_FAIL) goto END_REPEAT;
6316
6317 /* Else there's some kind of shambles */
6318
6319 else
6320 {
6321 *errorcodeptr = ERR11;
6322 goto FAILED;
6323 }
6324
6325 /* If the character following a repeat is '+', possessive_quantifier is
6326 TRUE. For some opcodes, there are special alternative opcodes for this
6327 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6328 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6329 Sun's Java package, but the special opcodes can optimize it.
6330
6331 Some (but not all) possessively repeated subpatterns have already been
6332 completely handled in the code just above. For them, possessive_quantifier
6333 is always FALSE at this stage. Note that the repeated item starts at
6334 tempcode, not at previous, which might be the first part of a string whose
6335 (former) last char we repeated. */
6336
6337 if (possessive_quantifier)
6338 {
6339 int len;
6340
6341 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6342 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6343 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6344 remains is greater than zero, there's a further opcode that can be
6345 handled. If not, do nothing, leaving the EXACT alone. */
6346
6347 switch(*tempcode)
6348 {
6349 case OP_TYPEEXACT:
6350 tempcode += PRIV(OP_lengths)[*tempcode] +
6351 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6352 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6353 break;
6354
6355 /* CHAR opcodes are used for exacts whose count is 1. */
6356
6357 case OP_CHAR:
6358 case OP_CHARI:
6359 case OP_NOT:
6360 case OP_NOTI:
6361 case OP_EXACT:
6362 case OP_EXACTI:
6363 case OP_NOTEXACT:
6364 case OP_NOTEXACTI:
6365 tempcode += PRIV(OP_lengths)[*tempcode];
6366 #ifdef SUPPORT_UTF
6367 if (utf && HAS_EXTRALEN(tempcode[-1]))
6368 tempcode += GET_EXTRALEN(tempcode[-1]);
6369 #endif
6370 break;
6371
6372 /* For the class opcodes, the repeat operator appears at the end;
6373 adjust tempcode to point to it. */
6374
6375 case OP_CLASS:
6376 case OP_NCLASS:
6377 tempcode += 1 + 32/sizeof(pcre_uchar);
6378 break;
6379
6380 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6381 case OP_XCLASS:
6382 tempcode += GET(tempcode, 1);
6383 break;
6384 #endif
6385 }
6386
6387 /* If tempcode is equal to code (which points to the end of the repeated
6388 item), it means we have skipped an EXACT item but there is no following
6389 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6390 all other cases, tempcode will be pointing to the repeat opcode, and will
6391 be less than code, so the value of len will be greater than 0. */
6392
6393 len = (int)(code - tempcode);
6394 if (len > 0)
6395 {
6396 unsigned int repcode = *tempcode;
6397
6398 /* There is a table for possessifying opcodes, all of which are less
6399 than OP_CALLOUT. A zero entry means there is no possessified version.
6400 */
6401
6402 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6403 *tempcode = opcode_possessify[repcode];
6404
6405 /* For opcode without a special possessified version, wrap the item in
6406 ONCE brackets. Because we are moving code along, we must ensure that any
6407 pending recursive references are updated. */
6408
6409 else
6410 {
6411 *code = OP_END;
6412 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm_offset);
6413 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6414 code += 1 + LINK_SIZE;
6415 len += 1 + LINK_SIZE;
6416 tempcode[0] = OP_ONCE;
6417 *code++ = OP_KET;
6418 PUTINC(code, 0, len);
6419 PUT(tempcode, 1, len);
6420 }
6421 }
6422
6423 #ifdef NEVER
6424 if (len > 0) switch (*tempcode)
6425 {
6426 case OP_STAR: *tempcode = OP_POSSTAR; break;
6427 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6428 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6429 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6430
6431 case OP_STARI: *tempcode = OP_POSSTARI; break;
6432 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6433 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6434 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6435
6436 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6437 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6438 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6439 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6440
6441 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6442 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6443 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6444 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6445
6446 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6447 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6448 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6449 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6450
6451 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6452 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6453 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6454 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6455
6456 /* Because we are moving code along, we must ensure that any
6457 pending recursive references are updated. */
6458
6459 default:
6460 *code = OP_END;
6461 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm_offset);
6462 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6463 code += 1 + LINK_SIZE;
6464 len += 1 + LINK_SIZE;
6465 tempcode[0] = OP_ONCE;
6466 *code++ = OP_KET;
6467 PUTINC(code, 0, len);
6468 PUT(tempcode, 1, len);
6469 break;
6470 }
6471 #endif
6472 }
6473
6474 /* In all case we no longer have a previous item. We also set the
6475 "follows varying string" flag for subsequently encountered reqchars if
6476 it isn't already set and we have just passed a varying length item. */
6477
6478 END_REPEAT:
6479 previous = NULL;
6480 cd->req_varyopt |= reqvary;
6481 break;
6482
6483
6484 /* ===================================================================*/
6485 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6486 lookbehind or option setting or condition or all the other extended
6487 parenthesis forms. */
6488
6489 case CHAR_LEFT_PARENTHESIS:
6490 ptr++;
6491
6492 /* First deal with comments. Putting this code right at the start ensures
6493 that comments have no bad side effects. */
6494
6495 if (ptr[0] == CHAR_QUESTION_MARK && ptr[1] == CHAR_NUMBER_SIGN)
6496 {
6497 ptr += 2;
6498 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6499 if (*ptr == CHAR_NULL)
6500 {
6501 *errorcodeptr = ERR18;
6502 goto FAILED;
6503 }
6504 continue;
6505 }
6506
6507 /* Now deal with various "verbs" that can be introduced by '*'. */
6508
6509 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6510 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6511 {
6512 int i, namelen;
6513 int arglen = 0;
6514 const char *vn = verbnames;
6515 const pcre_uchar *name = ptr + 1;
6516 const pcre_uchar *arg = NULL;
6517 previous = NULL;
6518 ptr++;
6519 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6520 namelen = (int)(ptr - name);
6521
6522 /* It appears that Perl allows any characters whatsoever, other than
6523 a closing parenthesis, to appear in arguments, so we no longer insist on
6524 letters, digits, and underscores. */
6525
6526 if (*ptr == CHAR_COLON)
6527 {
6528 arg = ++ptr;
6529 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6530 arglen = (int)(ptr - arg);
6531 if ((unsigned int)arglen > MAX_MARK)
6532 {
6533 *errorcodeptr = ERR75;
6534 goto FAILED;
6535 }
6536 }
6537
6538 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6539 {
6540 *errorcodeptr = ERR60;
6541 goto FAILED;
6542 }
6543
6544 /* Scan the table of verb names */
6545
6546 for (i = 0; i < verbcount; i++)
6547 {
6548 if (namelen == verbs[i].len &&
6549 STRNCMP_UC_C8(name, vn, namelen) == 0)
6550 {
6551 int setverb;
6552
6553 /* Check for open captures before ACCEPT and convert it to
6554 ASSERT_ACCEPT if in an assertion. */
6555
6556 if (verbs[i].op == OP_ACCEPT)
6557 {
6558 open_capitem *oc;
6559 if (arglen != 0)
6560 {
6561 *errorcodeptr = ERR59;
6562 goto FAILED;
6563 }
6564 cd->had_accept = TRUE;
6565 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6566 {
6567 *code++ = OP_CLOSE;
6568 PUT2INC(code, 0, oc->number);
6569 }
6570 setverb = *code++ =
6571 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6572
6573 /* Do not set firstchar after *ACCEPT */
6574 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6575 }
6576
6577 /* Handle other cases with/without an argument */
6578
6579 else if (arglen == 0)
6580 {
6581 if (verbs[i].op < 0) /* Argument is mandatory */
6582 {
6583 *errorcodeptr = ERR66;
6584 goto FAILED;
6585 }
6586 setverb = *code++ = verbs[i].op;
6587 }
6588
6589 else
6590 {
6591 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6592 {
6593 *errorcodeptr = ERR59;
6594 goto FAILED;
6595 }
6596 setverb = *code++ = verbs[i].op_arg;
6597 *code++ = arglen;
6598 memcpy(code, arg, IN_UCHARS(arglen));
6599 code += arglen;
6600 *code++ = 0;
6601 }
6602
6603 switch (setverb)
6604 {
6605 case OP_THEN:
6606 case OP_THEN_ARG:
6607 cd->external_flags |= PCRE_HASTHEN;
6608 break;
6609
6610 case OP_PRUNE:
6611 case OP_PRUNE_ARG:
6612 case OP_SKIP:
6613 case OP_SKIP_ARG:
6614 cd->had_pruneorskip = TRUE;
6615 break;
6616 }
6617
6618 break; /* Found verb, exit loop */
6619 }
6620
6621 vn += verbs[i].len + 1;
6622 }
6623
6624 if (i < verbcount) continue; /* Successfully handled a verb */
6625 *errorcodeptr = ERR60; /* Verb not recognized */
6626 goto FAILED;
6627 }
6628
6629 /* Initialize for "real" parentheses */
6630
6631 newoptions = options;
6632 skipbytes = 0;
6633 bravalue = OP_CBRA;
6634 save_hwm_offset = cd->hwm - cd->start_workspace;
6635 reset_bracount = FALSE;
6636
6637 /* Deal with the extended parentheses; all are introduced by '?', and the
6638 appearance of any of them means that this is not a capturing group. */
6639
6640 if (*ptr == CHAR_QUESTION_MARK)
6641 {
6642 int i, set, unset, namelen;
6643 int *optset;
6644 const pcre_uchar *name;
6645 pcre_uchar *slot;
6646
6647 switch (*(++ptr))
6648 {
6649 /* ------------------------------------------------------------ */
6650 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6651 reset_bracount = TRUE;
6652 /* Fall through */
6653
6654 /* ------------------------------------------------------------ */
6655 case CHAR_COLON: /* Non-capturing bracket */
6656 bravalue = OP_BRA;
6657 ptr++;
6658 break;
6659
6660
6661 /* ------------------------------------------------------------ */
6662 case CHAR_LEFT_PARENTHESIS:
6663 bravalue = OP_COND; /* Conditional group */
6664 tempptr =