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

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1414 - (show annotations)
Sun Dec 22 16:27:35 2013 UTC (6 years, 1 month ago) by zherczeg
File MIME type: text/plain
File size: 310100 byte(s)
A new flag is set, when property checks are present in an XCLASS.
<
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) 1997-2013 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 processed string start */
51 #define PSEND end_pattern /* Field containing processed string 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 ;
551
552 /* Table to identify digits and hex digits. This is used when compiling
553 patterns. Note that the tables in chartables are dependent on the locale, and
554 may mark arbitrary characters as digits - but the PCRE compiling code expects
555 to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
556 a private table here. It costs 256 bytes, but it is a lot faster than doing
557 character value tests (at least in some simple cases I timed), and in some
558 applications one wants PCRE to compile efficiently as well as match
559 efficiently.
560
561 For convenience, we use the same bit definitions as in chartables:
562
563 0x04 decimal digit
564 0x08 hexadecimal digit
565
566 Then we can use ctype_digit and ctype_xdigit in the code. */
567
568 /* Using a simple comparison for decimal numbers rather than a memory read
569 is much faster, and the resulting code is simpler (the compiler turns it
570 into a subtraction and unsigned comparison). */
571
572 #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
573
574 #ifndef EBCDIC
575
576 /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
577 UTF-8 mode. */
578
579 static const pcre_uint8 digitab[] =
580 {
581 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
582 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
583 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
584 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
585 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
586 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
587 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
588 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
589 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* @ - G */
590 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H - O */
591 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* P - W */
592 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* X - _ */
593 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
594 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
595 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
596 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
597 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
598 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
599 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
600 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
601 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
602 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
603 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
604 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
605 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
606 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
607 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
608 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
609 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
610 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
611 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
612 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
613
614 #else
615
616 /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
617
618 static const pcre_uint8 digitab[] =
619 {
620 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
621 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
622 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
623 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
624 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
625 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
626 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
627 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
628 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
629 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
630 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
631 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
632 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
633 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
634 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
635 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
636 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
637 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
638 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
639 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
640 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
641 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
642 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
643 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
644 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
645 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
646 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
647 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
648 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
649 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
650 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
651 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
652
653 static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
654 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
655 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
656 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
657 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
658 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
659 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
660 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
661 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
662 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
663 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
664 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
665 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
666 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
667 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
668 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
669 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
670 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
671 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
672 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
673 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
674 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
675 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
676 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
677 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
678 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
679 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
680 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
681 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
682 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
683 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
684 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
685 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
686 #endif
687
688
689 /* This table is used to check whether auto-possessification is possible
690 between adjacent character-type opcodes. The left-hand (repeated) opcode is
691 used to select the row, and the right-hand opcode is use to select the column.
692 A value of 1 means that auto-possessification is OK. For example, the second
693 value in the first row means that \D+\d can be turned into \D++\d.
694
695 The Unicode property types (\P and \p) have to be present to fill out the table
696 because of what their opcode values are, but the table values should always be
697 zero because property types are handled separately in the code. The last four
698 columns apply to items that cannot be repeated, so there is no need to have
699 rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
700 *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
701
702 #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
703 #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
704
705 static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
706 /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
707 { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
708 { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
709 { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
710 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
711 { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
712 { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
713 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
714 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
715 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
716 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
717 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
718 { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
719 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
720 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
721 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
722 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
723 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
724 };
725
726
727 /* This table is used to check whether auto-possessification is possible
728 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
729 left-hand (repeated) opcode is used to select the row, and the right-hand
730 opcode is used to select the column. The values are as follows:
731
732 0 Always return FALSE (never auto-possessify)
733 1 Character groups are distinct (possessify if both are OP_PROP)
734 2 Check character categories in the same group (general or particular)
735 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
736
737 4 Check left general category vs right particular category
738 5 Check right general category vs left particular category
739
740 6 Left alphanum vs right general category
741 7 Left space vs right general category
742 8 Left word vs right general category
743
744 9 Right alphanum vs left general category
745 10 Right space vs left general category
746 11 Right word vs left general category
747
748 12 Left alphanum vs right particular category
749 13 Left space vs right particular category
750 14 Left word vs right particular category
751
752 15 Right alphanum vs left particular category
753 16 Right space vs left particular category
754 17 Right word vs left particular category
755 */
756
757 static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
758 /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
759 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
760 { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
761 { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
762 { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
763 { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
764 { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
765 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
766 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
767 { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
768 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
769 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
770 };
771
772 /* This table is used to check whether auto-possessification is possible
773 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
774 specifies a general category and the other specifies a particular category. The
775 row is selected by the general category and the column by the particular
776 category. The value is 1 if the particular category is not part of the general
777 category. */
778
779 static const pcre_uint8 catposstab[7][30] = {
780 /* 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 */
781 { 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 */
782 { 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 */
783 { 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 */
784 { 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 */
785 { 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 */
786 { 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 */
787 { 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 */
788 };
789
790 /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
791 a general or particular category. The properties in each row are those
792 that apply to the character set in question. Duplication means that a little
793 unnecessary work is done when checking, but this keeps things much simpler
794 because they can all use the same code. For more details see the comment where
795 this table is used.
796
797 Note: SPACE and PXSPACE used to be different because Perl excluded VT from
798 "space", but from Perl 5.18 it's included, so both categories are treated the
799 same here. */
800
801 static const pcre_uint8 posspropstab[3][4] = {
802 { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
803 { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
804 { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
805 };
806
807 /* This table is used when converting repeating opcodes into possessified
808 versions as a result of an explicit possessive quantifier such as ++. A zero
809 value means there is no possessified version - in those cases the item in
810 question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
811 because all relevant opcodes are less than that. */
812
813 static const pcre_uint8 opcode_possessify[] = {
814 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */
815 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */
816
817 0, /* NOTI */
818 OP_POSSTAR, 0, /* STAR, MINSTAR */
819 OP_POSPLUS, 0, /* PLUS, MINPLUS */
820 OP_POSQUERY, 0, /* QUERY, MINQUERY */
821 OP_POSUPTO, 0, /* UPTO, MINUPTO */
822 0, /* EXACT */
823 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */
824
825 OP_POSSTARI, 0, /* STARI, MINSTARI */
826 OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */
827 OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */
828 OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */
829 0, /* EXACTI */
830 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */
831
832 OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */
833 OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */
834 OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */
835 OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */
836 0, /* NOTEXACT */
837 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
838
839 OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */
840 OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */
841 OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */
842 OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */
843 0, /* NOTEXACTI */
844 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
845
846 OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */
847 OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */
848 OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */
849 OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */
850 0, /* TYPEEXACT */
851 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
852
853 OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */
854 OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */
855 OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */
856 OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */
857 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */
858
859 0, 0, 0, /* CLASS, NCLASS, XCLASS */
860 0, 0, /* REF, REFI */
861 0, 0, /* DNREF, DNREFI */
862 0, 0 /* RECURSE, CALLOUT */
863 };
864
865
866
867 /*************************************************
868 * Find an error text *
869 *************************************************/
870
871 /* The error texts are now all in one long string, to save on relocations. As
872 some of the text is of unknown length, we can't use a table of offsets.
873 Instead, just count through the strings. This is not a performance issue
874 because it happens only when there has been a compilation error.
875
876 Argument: the error number
877 Returns: pointer to the error string
878 */
879
880 static const char *
881 find_error_text(int n)
882 {
883 const char *s = error_texts;
884 for (; n > 0; n--)
885 {
886 while (*s++ != CHAR_NULL) {};
887 if (*s == CHAR_NULL) return "Error text not found (please report)";
888 }
889 return s;
890 }
891
892
893
894 /*************************************************
895 * Expand the workspace *
896 *************************************************/
897
898 /* This function is called during the second compiling phase, if the number of
899 forward references fills the existing workspace, which is originally a block on
900 the stack. A larger block is obtained from malloc() unless the ultimate limit
901 has been reached or the increase will be rather small.
902
903 Argument: pointer to the compile data block
904 Returns: 0 if all went well, else an error number
905 */
906
907 static int
908 expand_workspace(compile_data *cd)
909 {
910 pcre_uchar *newspace;
911 int newsize = cd->workspace_size * 2;
912
913 if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
914 if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
915 newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
916 return ERR72;
917
918 newspace = (PUBL(malloc))(IN_UCHARS(newsize));
919 if (newspace == NULL) return ERR21;
920 memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
921 cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
922 if (cd->workspace_size > COMPILE_WORK_SIZE)
923 (PUBL(free))((void *)cd->start_workspace);
924 cd->start_workspace = newspace;
925 cd->workspace_size = newsize;
926 return 0;
927 }
928
929
930
931 /*************************************************
932 * Check for counted repeat *
933 *************************************************/
934
935 /* This function is called when a '{' is encountered in a place where it might
936 start a quantifier. It looks ahead to see if it really is a quantifier or not.
937 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
938 where the ddds are digits.
939
940 Arguments:
941 p pointer to the first char after '{'
942
943 Returns: TRUE or FALSE
944 */
945
946 static BOOL
947 is_counted_repeat(const pcre_uchar *p)
948 {
949 if (!IS_DIGIT(*p)) return FALSE;
950 p++;
951 while (IS_DIGIT(*p)) p++;
952 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
953
954 if (*p++ != CHAR_COMMA) return FALSE;
955 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
956
957 if (!IS_DIGIT(*p)) return FALSE;
958 p++;
959 while (IS_DIGIT(*p)) p++;
960
961 return (*p == CHAR_RIGHT_CURLY_BRACKET);
962 }
963
964
965
966 /*************************************************
967 * Handle escapes *
968 *************************************************/
969
970 /* This function is called when a \ has been encountered. It either returns a
971 positive value for a simple escape such as \n, or 0 for a data character which
972 will be placed in chptr. A backreference to group n is returned as negative n.
973 When UTF-8 is enabled, a positive value greater than 255 may be returned in
974 chptr. On entry, ptr is pointing at the \. On exit, it is on the final
975 character of the escape sequence.
976
977 Arguments:
978 ptrptr points to the pattern position pointer
979 chptr points to a returned data character
980 errorcodeptr points to the errorcode variable
981 bracount number of previous extracting brackets
982 options the options bits
983 isclass TRUE if inside a character class
984
985 Returns: zero => a data character
986 positive => a special escape sequence
987 negative => a back reference
988 on error, errorcodeptr is set
989 */
990
991 static int
992 check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
993 int bracount, int options, BOOL isclass)
994 {
995 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
996 BOOL utf = (options & PCRE_UTF8) != 0;
997 const pcre_uchar *ptr = *ptrptr + 1;
998 pcre_uint32 c;
999 int escape = 0;
1000 int i;
1001
1002 GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
1003 ptr--; /* Set pointer back to the last byte */
1004
1005 /* If backslash is at the end of the pattern, it's an error. */
1006
1007 if (c == CHAR_NULL) *errorcodeptr = ERR1;
1008
1009 /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1010 in a table. A non-zero result is something that can be returned immediately.
1011 Otherwise further processing may be required. */
1012
1013 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1014 /* Not alphanumeric */
1015 else if (c < CHAR_0 || c > CHAR_z) {}
1016 else if ((i = escapes[c - CHAR_0]) != 0)
1017 { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1018
1019 #else /* EBCDIC coding */
1020 /* Not alphanumeric */
1021 else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1022 else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1023 #endif
1024
1025 /* Escapes that need further processing, or are illegal. */
1026
1027 else
1028 {
1029 const pcre_uchar *oldptr;
1030 BOOL braced, negated, overflow;
1031 int s;
1032
1033 switch (c)
1034 {
1035 /* A number of Perl escapes are not handled by PCRE. We give an explicit
1036 error. */
1037
1038 case CHAR_l:
1039 case CHAR_L:
1040 *errorcodeptr = ERR37;
1041 break;
1042
1043 case CHAR_u:
1044 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1045 {
1046 /* In JavaScript, \u must be followed by four hexadecimal numbers.
1047 Otherwise it is a lowercase u letter. */
1048 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1049 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1050 && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1051 && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1052 {
1053 c = 0;
1054 for (i = 0; i < 4; ++i)
1055 {
1056 register pcre_uint32 cc = *(++ptr);
1057 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1058 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1059 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1060 #else /* EBCDIC coding */
1061 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1062 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1063 #endif
1064 }
1065
1066 #if defined COMPILE_PCRE8
1067 if (c > (utf ? 0x10ffffU : 0xffU))
1068 #elif defined COMPILE_PCRE16
1069 if (c > (utf ? 0x10ffffU : 0xffffU))
1070 #elif defined COMPILE_PCRE32
1071 if (utf && c > 0x10ffffU)
1072 #endif
1073 {
1074 *errorcodeptr = ERR76;
1075 }
1076 else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1077 }
1078 }
1079 else
1080 *errorcodeptr = ERR37;
1081 break;
1082
1083 case CHAR_U:
1084 /* In JavaScript, \U is an uppercase U letter. */
1085 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1086 break;
1087
1088 /* In a character class, \g is just a literal "g". Outside a character
1089 class, \g must be followed by one of a number of specific things:
1090
1091 (1) A number, either plain or braced. If positive, it is an absolute
1092 backreference. If negative, it is a relative backreference. This is a Perl
1093 5.10 feature.
1094
1095 (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1096 is part of Perl's movement towards a unified syntax for back references. As
1097 this is synonymous with \k{name}, we fudge it up by pretending it really
1098 was \k.
1099
1100 (3) For Oniguruma compatibility we also support \g followed by a name or a
1101 number either in angle brackets or in single quotes. However, these are
1102 (possibly recursive) subroutine calls, _not_ backreferences. Just return
1103 the ESC_g code (cf \k). */
1104
1105 case CHAR_g:
1106 if (isclass) break;
1107 if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1108 {
1109 escape = ESC_g;
1110 break;
1111 }
1112
1113 /* Handle the Perl-compatible cases */
1114
1115 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1116 {
1117 const pcre_uchar *p;
1118 for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1119 if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1120 if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1121 {
1122 escape = ESC_k;
1123 break;
1124 }
1125 braced = TRUE;
1126 ptr++;
1127 }
1128 else braced = FALSE;
1129
1130 if (ptr[1] == CHAR_MINUS)
1131 {
1132 negated = TRUE;
1133 ptr++;
1134 }
1135 else negated = FALSE;
1136
1137 /* The integer range is limited by the machine's int representation. */
1138 s = 0;
1139 overflow = FALSE;
1140 while (IS_DIGIT(ptr[1]))
1141 {
1142 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1143 {
1144 overflow = TRUE;
1145 break;
1146 }
1147 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1148 }
1149 if (overflow) /* Integer overflow */
1150 {
1151 while (IS_DIGIT(ptr[1]))
1152 ptr++;
1153 *errorcodeptr = ERR61;
1154 break;
1155 }
1156
1157 if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1158 {
1159 *errorcodeptr = ERR57;
1160 break;
1161 }
1162
1163 if (s == 0)
1164 {
1165 *errorcodeptr = ERR58;
1166 break;
1167 }
1168
1169 if (negated)
1170 {
1171 if (s > bracount)
1172 {
1173 *errorcodeptr = ERR15;
1174 break;
1175 }
1176 s = bracount - (s - 1);
1177 }
1178
1179 escape = -s;
1180 break;
1181
1182 /* The handling of escape sequences consisting of a string of digits
1183 starting with one that is not zero is not straightforward. Perl has changed
1184 over the years. Nowadays \g{} for backreferences and \o{} for octal are
1185 recommended to avoid the ambiguities in the old syntax.
1186
1187 Outside a character class, the digits are read as a decimal number. If the
1188 number is less than 8 (used to be 10), or if there are that many previous
1189 extracting left brackets, then it is a back reference. Otherwise, up to
1190 three octal digits are read to form an escaped byte. Thus \123 is likely to
1191 be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1192 the octal value is greater than 377, the least significant 8 bits are
1193 taken. \8 and \9 are treated as the literal characters 8 and 9.
1194
1195 Inside a character class, \ followed by a digit is always either a literal
1196 8 or 9 or an octal number. */
1197
1198 case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1199 case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1200
1201 if (!isclass)
1202 {
1203 oldptr = ptr;
1204 /* The integer range is limited by the machine's int representation. */
1205 s = (int)(c -CHAR_0);
1206 overflow = FALSE;
1207 while (IS_DIGIT(ptr[1]))
1208 {
1209 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1210 {
1211 overflow = TRUE;
1212 break;
1213 }
1214 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1215 }
1216 if (overflow) /* Integer overflow */
1217 {
1218 while (IS_DIGIT(ptr[1]))
1219 ptr++;
1220 *errorcodeptr = ERR61;
1221 break;
1222 }
1223 if (s < 8 || s <= bracount) /* Check for back reference */
1224 {
1225 escape = -s;
1226 break;
1227 }
1228 ptr = oldptr; /* Put the pointer back and fall through */
1229 }
1230
1231 /* Handle a digit following \ when the number is not a back reference. If
1232 the first digit is 8 or 9, Perl used to generate a binary zero byte and
1233 then treat the digit as a following literal. At least by Perl 5.18 this
1234 changed so as not to insert the binary zero. */
1235
1236 if ((c = *ptr) >= CHAR_8) break;
1237
1238 /* Fall through with a digit less than 8 */
1239
1240 /* \0 always starts an octal number, but we may drop through to here with a
1241 larger first octal digit. The original code used just to take the least
1242 significant 8 bits of octal numbers (I think this is what early Perls used
1243 to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1244 but no more than 3 octal digits. */
1245
1246 case CHAR_0:
1247 c -= CHAR_0;
1248 while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1249 c = c * 8 + *(++ptr) - CHAR_0;
1250 #ifdef COMPILE_PCRE8
1251 if (!utf && c > 0xff) *errorcodeptr = ERR51;
1252 #endif
1253 break;
1254
1255 /* \o is a relatively new Perl feature, supporting a more general way of
1256 specifying character codes in octal. The only supported form is \o{ddd}. */
1257
1258 case CHAR_o:
1259 if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1260 {
1261 ptr += 2;
1262 c = 0;
1263 overflow = FALSE;
1264 while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1265 {
1266 register pcre_uint32 cc = *ptr++;
1267 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1268 #ifdef COMPILE_PCRE32
1269 if (c >= 0x20000000l) { overflow = TRUE; break; }
1270 #endif
1271 c = (c << 3) + cc - CHAR_0 ;
1272 #if defined COMPILE_PCRE8
1273 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1274 #elif defined COMPILE_PCRE16
1275 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1276 #elif defined COMPILE_PCRE32
1277 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1278 #endif
1279 }
1280 if (overflow)
1281 {
1282 while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1283 *errorcodeptr = ERR34;
1284 }
1285 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1286 {
1287 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1288 }
1289 else *errorcodeptr = ERR80;
1290 }
1291 break;
1292
1293 /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1294 numbers. Otherwise it is a lowercase x letter. */
1295
1296 case CHAR_x:
1297 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1298 {
1299 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1300 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1301 {
1302 c = 0;
1303 for (i = 0; i < 2; ++i)
1304 {
1305 register pcre_uint32 cc = *(++ptr);
1306 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1307 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1308 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1309 #else /* EBCDIC coding */
1310 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1311 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1312 #endif
1313 }
1314 }
1315 } /* End JavaScript handling */
1316
1317 /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1318 greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1319 digits. If not, { used to be treated as a data character. However, Perl
1320 seems to read hex digits up to the first non-such, and ignore the rest, so
1321 that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1322 now gives an error. */
1323
1324 else
1325 {
1326 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1327 {
1328 ptr += 2;
1329 c = 0;
1330 overflow = FALSE;
1331 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1332 {
1333 register pcre_uint32 cc = *ptr++;
1334 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1335
1336 #ifdef COMPILE_PCRE32
1337 if (c >= 0x10000000l) { overflow = TRUE; break; }
1338 #endif
1339
1340 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1341 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1342 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1343 #else /* EBCDIC coding */
1344 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1345 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1346 #endif
1347
1348 #if defined COMPILE_PCRE8
1349 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1350 #elif defined COMPILE_PCRE16
1351 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1352 #elif defined COMPILE_PCRE32
1353 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1354 #endif
1355 }
1356
1357 if (overflow)
1358 {
1359 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1360 *errorcodeptr = ERR34;
1361 }
1362
1363 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1364 {
1365 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1366 }
1367
1368 /* If the sequence of hex digits does not end with '}', give an error.
1369 We used just to recognize this construct and fall through to the normal
1370 \x handling, but nowadays Perl gives an error, which seems much more
1371 sensible, so we do too. */
1372
1373 else *errorcodeptr = ERR79;
1374 } /* End of \x{} processing */
1375
1376 /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1377
1378 else
1379 {
1380 c = 0;
1381 while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1382 {
1383 pcre_uint32 cc; /* Some compilers don't like */
1384 cc = *(++ptr); /* ++ in initializers */
1385 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1386 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1387 c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1388 #else /* EBCDIC coding */
1389 if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1390 c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1391 #endif
1392 }
1393 } /* End of \xdd handling */
1394 } /* End of Perl-style \x handling */
1395 break;
1396
1397 /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1398 An error is given if the byte following \c is not an ASCII character. This
1399 coding is ASCII-specific, but then the whole concept of \cx is
1400 ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1401
1402 case CHAR_c:
1403 c = *(++ptr);
1404 if (c == CHAR_NULL)
1405 {
1406 *errorcodeptr = ERR2;
1407 break;
1408 }
1409 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1410 if (c > 127) /* Excludes all non-ASCII in either mode */
1411 {
1412 *errorcodeptr = ERR68;
1413 break;
1414 }
1415 if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1416 c ^= 0x40;
1417 #else /* EBCDIC coding */
1418 if (c >= CHAR_a && c <= CHAR_z) c += 64;
1419 c ^= 0xC0;
1420 #endif
1421 break;
1422
1423 /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1424 other alphanumeric following \ is an error if PCRE_EXTRA was set;
1425 otherwise, for Perl compatibility, it is a literal. This code looks a bit
1426 odd, but there used to be some cases other than the default, and there may
1427 be again in future, so I haven't "optimized" it. */
1428
1429 default:
1430 if ((options & PCRE_EXTRA) != 0) switch(c)
1431 {
1432 default:
1433 *errorcodeptr = ERR3;
1434 break;
1435 }
1436 break;
1437 }
1438 }
1439
1440 /* Perl supports \N{name} for character names, as well as plain \N for "not
1441 newline". PCRE does not support \N{name}. However, it does support
1442 quantification such as \N{2,3}. */
1443
1444 if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1445 !is_counted_repeat(ptr+2))
1446 *errorcodeptr = ERR37;
1447
1448 /* If PCRE_UCP is set, we change the values for \d etc. */
1449
1450 if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1451 escape += (ESC_DU - ESC_D);
1452
1453 /* Set the pointer to the final character before returning. */
1454
1455 *ptrptr = ptr;
1456 *chptr = c;
1457 return escape;
1458 }
1459
1460
1461
1462 #ifdef SUPPORT_UCP
1463 /*************************************************
1464 * Handle \P and \p *
1465 *************************************************/
1466
1467 /* This function is called after \P or \p has been encountered, provided that
1468 PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1469 pointing at the P or p. On exit, it is pointing at the final character of the
1470 escape sequence.
1471
1472 Argument:
1473 ptrptr points to the pattern position pointer
1474 negptr points to a boolean that is set TRUE for negation else FALSE
1475 ptypeptr points to an unsigned int that is set to the type value
1476 pdataptr points to an unsigned int that is set to the detailed property value
1477 errorcodeptr points to the error code variable
1478
1479 Returns: TRUE if the type value was found, or FALSE for an invalid type
1480 */
1481
1482 static BOOL
1483 get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1484 unsigned int *pdataptr, int *errorcodeptr)
1485 {
1486 pcre_uchar c;
1487 int i, bot, top;
1488 const pcre_uchar *ptr = *ptrptr;
1489 pcre_uchar name[32];
1490
1491 c = *(++ptr);
1492 if (c == CHAR_NULL) goto ERROR_RETURN;
1493
1494 *negptr = FALSE;
1495
1496 /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1497 negation. */
1498
1499 if (c == CHAR_LEFT_CURLY_BRACKET)
1500 {
1501 if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1502 {
1503 *negptr = TRUE;
1504 ptr++;
1505 }
1506 for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1507 {
1508 c = *(++ptr);
1509 if (c == CHAR_NULL) goto ERROR_RETURN;
1510 if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1511 name[i] = c;
1512 }
1513 if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1514 name[i] = 0;
1515 }
1516
1517 /* Otherwise there is just one following character */
1518
1519 else
1520 {
1521 name[0] = c;
1522 name[1] = 0;
1523 }
1524
1525 *ptrptr = ptr;
1526
1527 /* Search for a recognized property name using binary chop */
1528
1529 bot = 0;
1530 top = PRIV(utt_size);
1531
1532 while (bot < top)
1533 {
1534 int r;
1535 i = (bot + top) >> 1;
1536 r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1537 if (r == 0)
1538 {
1539 *ptypeptr = PRIV(utt)[i].type;
1540 *pdataptr = PRIV(utt)[i].value;
1541 return TRUE;
1542 }
1543 if (r > 0) bot = i + 1; else top = i;
1544 }
1545
1546 *errorcodeptr = ERR47;
1547 *ptrptr = ptr;
1548 return FALSE;
1549
1550 ERROR_RETURN:
1551 *errorcodeptr = ERR46;
1552 *ptrptr = ptr;
1553 return FALSE;
1554 }
1555 #endif
1556
1557
1558
1559 /*************************************************
1560 * Read repeat counts *
1561 *************************************************/
1562
1563 /* Read an item of the form {n,m} and return the values. This is called only
1564 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1565 so the syntax is guaranteed to be correct, but we need to check the values.
1566
1567 Arguments:
1568 p pointer to first char after '{'
1569 minp pointer to int for min
1570 maxp pointer to int for max
1571 returned as -1 if no max
1572 errorcodeptr points to error code variable
1573
1574 Returns: pointer to '}' on success;
1575 current ptr on error, with errorcodeptr set non-zero
1576 */
1577
1578 static const pcre_uchar *
1579 read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1580 {
1581 int min = 0;
1582 int max = -1;
1583
1584 /* Read the minimum value and do a paranoid check: a negative value indicates
1585 an integer overflow. */
1586
1587 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1588 if (min < 0 || min > 65535)
1589 {
1590 *errorcodeptr = ERR5;
1591 return p;
1592 }
1593
1594 /* Read the maximum value if there is one, and again do a paranoid on its size.
1595 Also, max must not be less than min. */
1596
1597 if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1598 {
1599 if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1600 {
1601 max = 0;
1602 while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1603 if (max < 0 || max > 65535)
1604 {
1605 *errorcodeptr = ERR5;
1606 return p;
1607 }
1608 if (max < min)
1609 {
1610 *errorcodeptr = ERR4;
1611 return p;
1612 }
1613 }
1614 }
1615
1616 /* Fill in the required variables, and pass back the pointer to the terminating
1617 '}'. */
1618
1619 *minp = min;
1620 *maxp = max;
1621 return p;
1622 }
1623
1624
1625
1626 /*************************************************
1627 * Find first significant op code *
1628 *************************************************/
1629
1630 /* This is called by several functions that scan a compiled expression looking
1631 for a fixed first character, or an anchoring op code etc. It skips over things
1632 that do not influence this. For some calls, it makes sense to skip negative
1633 forward and all backward assertions, and also the \b assertion; for others it
1634 does not.
1635
1636 Arguments:
1637 code pointer to the start of the group
1638 skipassert TRUE if certain assertions are to be skipped
1639
1640 Returns: pointer to the first significant opcode
1641 */
1642
1643 static const pcre_uchar*
1644 first_significant_code(const pcre_uchar *code, BOOL skipassert)
1645 {
1646 for (;;)
1647 {
1648 switch ((int)*code)
1649 {
1650 case OP_ASSERT_NOT:
1651 case OP_ASSERTBACK:
1652 case OP_ASSERTBACK_NOT:
1653 if (!skipassert) return code;
1654 do code += GET(code, 1); while (*code == OP_ALT);
1655 code += PRIV(OP_lengths)[*code];
1656 break;
1657
1658 case OP_WORD_BOUNDARY:
1659 case OP_NOT_WORD_BOUNDARY:
1660 if (!skipassert) return code;
1661 /* Fall through */
1662
1663 case OP_CALLOUT:
1664 case OP_CREF:
1665 case OP_DNCREF:
1666 case OP_RREF:
1667 case OP_DNRREF:
1668 case OP_DEF:
1669 code += PRIV(OP_lengths)[*code];
1670 break;
1671
1672 default:
1673 return code;
1674 }
1675 }
1676 /* Control never reaches here */
1677 }
1678
1679
1680
1681 /*************************************************
1682 * Find the fixed length of a branch *
1683 *************************************************/
1684
1685 /* Scan a branch and compute the fixed length of subject that will match it,
1686 if the length is fixed. This is needed for dealing with backward assertions.
1687 In UTF8 mode, the result is in characters rather than bytes. The branch is
1688 temporarily terminated with OP_END when this function is called.
1689
1690 This function is called when a backward assertion is encountered, so that if it
1691 fails, the error message can point to the correct place in the pattern.
1692 However, we cannot do this when the assertion contains subroutine calls,
1693 because they can be forward references. We solve this by remembering this case
1694 and doing the check at the end; a flag specifies which mode we are running in.
1695
1696 Arguments:
1697 code points to the start of the pattern (the bracket)
1698 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1699 atend TRUE if called when the pattern is complete
1700 cd the "compile data" structure
1701
1702 Returns: the fixed length,
1703 or -1 if there is no fixed length,
1704 or -2 if \C was encountered (in UTF-8 mode only)
1705 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1706 or -4 if an unknown opcode was encountered (internal error)
1707 */
1708
1709 static int
1710 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1711 {
1712 int length = -1;
1713
1714 register int branchlength = 0;
1715 register pcre_uchar *cc = code + 1 + LINK_SIZE;
1716
1717 /* Scan along the opcodes for this branch. If we get to the end of the
1718 branch, check the length against that of the other branches. */
1719
1720 for (;;)
1721 {
1722 int d;
1723 pcre_uchar *ce, *cs;
1724 register pcre_uchar op = *cc;
1725
1726 switch (op)
1727 {
1728 /* We only need to continue for OP_CBRA (normal capturing bracket) and
1729 OP_BRA (normal non-capturing bracket) because the other variants of these
1730 opcodes are all concerned with unlimited repeated groups, which of course
1731 are not of fixed length. */
1732
1733 case OP_CBRA:
1734 case OP_BRA:
1735 case OP_ONCE:
1736 case OP_ONCE_NC:
1737 case OP_COND:
1738 d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1739 if (d < 0) return d;
1740 branchlength += d;
1741 do cc += GET(cc, 1); while (*cc == OP_ALT);
1742 cc += 1 + LINK_SIZE;
1743 break;
1744
1745 /* Reached end of a branch; if it's a ket it is the end of a nested call.
1746 If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1747 an ALT. If it is END it's the end of the outer call. All can be handled by
1748 the same code. Note that we must not include the OP_KETRxxx opcodes here,
1749 because they all imply an unlimited repeat. */
1750
1751 case OP_ALT:
1752 case OP_KET:
1753 case OP_END:
1754 case OP_ACCEPT:
1755 case OP_ASSERT_ACCEPT:
1756 if (length < 0) length = branchlength;
1757 else if (length != branchlength) return -1;
1758 if (*cc != OP_ALT) return length;
1759 cc += 1 + LINK_SIZE;
1760 branchlength = 0;
1761 break;
1762
1763 /* A true recursion implies not fixed length, but a subroutine call may
1764 be OK. If the subroutine is a forward reference, we can't deal with
1765 it until the end of the pattern, so return -3. */
1766
1767 case OP_RECURSE:
1768 if (!atend) return -3;
1769 cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1770 do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1771 if (cc > cs && cc < ce) return -1; /* Recursion */
1772 d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1773 if (d < 0) return d;
1774 branchlength += d;
1775 cc += 1 + LINK_SIZE;
1776 break;
1777
1778 /* Skip over assertive subpatterns */
1779
1780 case OP_ASSERT:
1781 case OP_ASSERT_NOT:
1782 case OP_ASSERTBACK:
1783 case OP_ASSERTBACK_NOT:
1784 do cc += GET(cc, 1); while (*cc == OP_ALT);
1785 cc += PRIV(OP_lengths)[*cc];
1786 break;
1787
1788 /* Skip over things that don't match chars */
1789
1790 case OP_MARK:
1791 case OP_PRUNE_ARG:
1792 case OP_SKIP_ARG:
1793 case OP_THEN_ARG:
1794 cc += cc[1] + PRIV(OP_lengths)[*cc];
1795 break;
1796
1797 case OP_CALLOUT:
1798 case OP_CIRC:
1799 case OP_CIRCM:
1800 case OP_CLOSE:
1801 case OP_COMMIT:
1802 case OP_CREF:
1803 case OP_DEF:
1804 case OP_DNCREF:
1805 case OP_DNRREF:
1806 case OP_DOLL:
1807 case OP_DOLLM:
1808 case OP_EOD:
1809 case OP_EODN:
1810 case OP_FAIL:
1811 case OP_NOT_WORD_BOUNDARY:
1812 case OP_PRUNE:
1813 case OP_REVERSE:
1814 case OP_RREF:
1815 case OP_SET_SOM:
1816 case OP_SKIP:
1817 case OP_SOD:
1818 case OP_SOM:
1819 case OP_THEN:
1820 case OP_WORD_BOUNDARY:
1821 cc += PRIV(OP_lengths)[*cc];
1822 break;
1823
1824 /* Handle literal characters */
1825
1826 case OP_CHAR:
1827 case OP_CHARI:
1828 case OP_NOT:
1829 case OP_NOTI:
1830 branchlength++;
1831 cc += 2;
1832 #ifdef SUPPORT_UTF
1833 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1834 #endif
1835 break;
1836
1837 /* Handle exact repetitions. The count is already in characters, but we
1838 need to skip over a multibyte character in UTF8 mode. */
1839
1840 case OP_EXACT:
1841 case OP_EXACTI:
1842 case OP_NOTEXACT:
1843 case OP_NOTEXACTI:
1844 branchlength += (int)GET2(cc,1);
1845 cc += 2 + IMM2_SIZE;
1846 #ifdef SUPPORT_UTF
1847 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1848 #endif
1849 break;
1850
1851 case OP_TYPEEXACT:
1852 branchlength += GET2(cc,1);
1853 if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1854 cc += 2;
1855 cc += 1 + IMM2_SIZE + 1;
1856 break;
1857
1858 /* Handle single-char matchers */
1859
1860 case OP_PROP:
1861 case OP_NOTPROP:
1862 cc += 2;
1863 /* Fall through */
1864
1865 case OP_HSPACE:
1866 case OP_VSPACE:
1867 case OP_NOT_HSPACE:
1868 case OP_NOT_VSPACE:
1869 case OP_NOT_DIGIT:
1870 case OP_DIGIT:
1871 case OP_NOT_WHITESPACE:
1872 case OP_WHITESPACE:
1873 case OP_NOT_WORDCHAR:
1874 case OP_WORDCHAR:
1875 case OP_ANY:
1876 case OP_ALLANY:
1877 branchlength++;
1878 cc++;
1879 break;
1880
1881 /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1882 otherwise \C is coded as OP_ALLANY. */
1883
1884 case OP_ANYBYTE:
1885 return -2;
1886
1887 /* Check a class for variable quantification */
1888
1889 case OP_CLASS:
1890 case OP_NCLASS:
1891 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1892 case OP_XCLASS:
1893 /* The original code caused an unsigned overflow in 64 bit systems,
1894 so now we use a conditional statement. */
1895 if (op == OP_XCLASS)
1896 cc += GET(cc, 1);
1897 else
1898 cc += PRIV(OP_lengths)[OP_CLASS];
1899 #else
1900 cc += PRIV(OP_lengths)[OP_CLASS];
1901 #endif
1902
1903 switch (*cc)
1904 {
1905 case OP_CRSTAR:
1906 case OP_CRMINSTAR:
1907 case OP_CRPLUS:
1908 case OP_CRMINPLUS:
1909 case OP_CRQUERY:
1910 case OP_CRMINQUERY:
1911 case OP_CRPOSSTAR:
1912 case OP_CRPOSPLUS:
1913 case OP_CRPOSQUERY:
1914 return -1;
1915
1916 case OP_CRRANGE:
1917 case OP_CRMINRANGE:
1918 case OP_CRPOSRANGE:
1919 if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1920 branchlength += (int)GET2(cc,1);
1921 cc += 1 + 2 * IMM2_SIZE;
1922 break;
1923
1924 default:
1925 branchlength++;
1926 }
1927 break;
1928
1929 /* Anything else is variable length */
1930
1931 case OP_ANYNL:
1932 case OP_BRAMINZERO:
1933 case OP_BRAPOS:
1934 case OP_BRAPOSZERO:
1935 case OP_BRAZERO:
1936 case OP_CBRAPOS:
1937 case OP_EXTUNI:
1938 case OP_KETRMAX:
1939 case OP_KETRMIN:
1940 case OP_KETRPOS:
1941 case OP_MINPLUS:
1942 case OP_MINPLUSI:
1943 case OP_MINQUERY:
1944 case OP_MINQUERYI:
1945 case OP_MINSTAR:
1946 case OP_MINSTARI:
1947 case OP_MINUPTO:
1948 case OP_MINUPTOI:
1949 case OP_NOTMINPLUS:
1950 case OP_NOTMINPLUSI:
1951 case OP_NOTMINQUERY:
1952 case OP_NOTMINQUERYI:
1953 case OP_NOTMINSTAR:
1954 case OP_NOTMINSTARI:
1955 case OP_NOTMINUPTO:
1956 case OP_NOTMINUPTOI:
1957 case OP_NOTPLUS:
1958 case OP_NOTPLUSI:
1959 case OP_NOTPOSPLUS:
1960 case OP_NOTPOSPLUSI:
1961 case OP_NOTPOSQUERY:
1962 case OP_NOTPOSQUERYI:
1963 case OP_NOTPOSSTAR:
1964 case OP_NOTPOSSTARI:
1965 case OP_NOTPOSUPTO:
1966 case OP_NOTPOSUPTOI:
1967 case OP_NOTQUERY:
1968 case OP_NOTQUERYI:
1969 case OP_NOTSTAR:
1970 case OP_NOTSTARI:
1971 case OP_NOTUPTO:
1972 case OP_NOTUPTOI:
1973 case OP_PLUS:
1974 case OP_PLUSI:
1975 case OP_POSPLUS:
1976 case OP_POSPLUSI:
1977 case OP_POSQUERY:
1978 case OP_POSQUERYI:
1979 case OP_POSSTAR:
1980 case OP_POSSTARI:
1981 case OP_POSUPTO:
1982 case OP_POSUPTOI:
1983 case OP_QUERY:
1984 case OP_QUERYI:
1985 case OP_REF:
1986 case OP_REFI:
1987 case OP_DNREF:
1988 case OP_DNREFI:
1989 case OP_SBRA:
1990 case OP_SBRAPOS:
1991 case OP_SCBRA:
1992 case OP_SCBRAPOS:
1993 case OP_SCOND:
1994 case OP_SKIPZERO:
1995 case OP_STAR:
1996 case OP_STARI:
1997 case OP_TYPEMINPLUS:
1998 case OP_TYPEMINQUERY:
1999 case OP_TYPEMINSTAR:
2000 case OP_TYPEMINUPTO:
2001 case OP_TYPEPLUS:
2002 case OP_TYPEPOSPLUS:
2003 case OP_TYPEPOSQUERY:
2004 case OP_TYPEPOSSTAR:
2005 case OP_TYPEPOSUPTO:
2006 case OP_TYPEQUERY:
2007 case OP_TYPESTAR:
2008 case OP_TYPEUPTO:
2009 case OP_UPTO:
2010 case OP_UPTOI:
2011 return -1;
2012
2013 /* Catch unrecognized opcodes so that when new ones are added they
2014 are not forgotten, as has happened in the past. */
2015
2016 default:
2017 return -4;
2018 }
2019 }
2020 /* Control never gets here */
2021 }
2022
2023
2024
2025 /*************************************************
2026 * Scan compiled regex for specific bracket *
2027 *************************************************/
2028
2029 /* This little function scans through a compiled pattern until it finds a
2030 capturing bracket with the given number, or, if the number is negative, an
2031 instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2032 so that it can be called from pcre_study() when finding the minimum matching
2033 length.
2034
2035 Arguments:
2036 code points to start of expression
2037 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2038 number the required bracket number or negative to find a lookbehind
2039
2040 Returns: pointer to the opcode for the bracket, or NULL if not found
2041 */
2042
2043 const pcre_uchar *
2044 PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2045 {
2046 for (;;)
2047 {
2048 register pcre_uchar c = *code;
2049
2050 if (c == OP_END) return NULL;
2051
2052 /* XCLASS is used for classes that cannot be represented just by a bit
2053 map. This includes negated single high-valued characters. The length in
2054 the table is zero; the actual length is stored in the compiled code. */
2055
2056 if (c == OP_XCLASS) code += GET(code, 1);
2057
2058 /* Handle recursion */
2059
2060 else if (c == OP_REVERSE)
2061 {
2062 if (number < 0) return (pcre_uchar *)code;
2063 code += PRIV(OP_lengths)[c];
2064 }
2065
2066 /* Handle capturing bracket */
2067
2068 else if (c == OP_CBRA || c == OP_SCBRA ||
2069 c == OP_CBRAPOS || c == OP_SCBRAPOS)
2070 {
2071 int n = (int)GET2(code, 1+LINK_SIZE);
2072 if (n == number) return (pcre_uchar *)code;
2073 code += PRIV(OP_lengths)[c];
2074 }
2075
2076 /* Otherwise, we can get the item's length from the table, except that for
2077 repeated character types, we have to test for \p and \P, which have an extra
2078 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2079 must add in its length. */
2080
2081 else
2082 {
2083 switch(c)
2084 {
2085 case OP_TYPESTAR:
2086 case OP_TYPEMINSTAR:
2087 case OP_TYPEPLUS:
2088 case OP_TYPEMINPLUS:
2089 case OP_TYPEQUERY:
2090 case OP_TYPEMINQUERY:
2091 case OP_TYPEPOSSTAR:
2092 case OP_TYPEPOSPLUS:
2093 case OP_TYPEPOSQUERY:
2094 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2095 break;
2096
2097 case OP_TYPEUPTO:
2098 case OP_TYPEMINUPTO:
2099 case OP_TYPEEXACT:
2100 case OP_TYPEPOSUPTO:
2101 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2102 code += 2;
2103 break;
2104
2105 case OP_MARK:
2106 case OP_PRUNE_ARG:
2107 case OP_SKIP_ARG:
2108 case OP_THEN_ARG:
2109 code += code[1];
2110 break;
2111 }
2112
2113 /* Add in the fixed length from the table */
2114
2115 code += PRIV(OP_lengths)[c];
2116
2117 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2118 a multi-byte character. The length in the table is a minimum, so we have to
2119 arrange to skip the extra bytes. */
2120
2121 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2122 if (utf) switch(c)
2123 {
2124 case OP_CHAR:
2125 case OP_CHARI:
2126 case OP_EXACT:
2127 case OP_EXACTI:
2128 case OP_UPTO:
2129 case OP_UPTOI:
2130 case OP_MINUPTO:
2131 case OP_MINUPTOI:
2132 case OP_POSUPTO:
2133 case OP_POSUPTOI:
2134 case OP_STAR:
2135 case OP_STARI:
2136 case OP_MINSTAR:
2137 case OP_MINSTARI:
2138 case OP_POSSTAR:
2139 case OP_POSSTARI:
2140 case OP_PLUS:
2141 case OP_PLUSI:
2142 case OP_MINPLUS:
2143 case OP_MINPLUSI:
2144 case OP_POSPLUS:
2145 case OP_POSPLUSI:
2146 case OP_QUERY:
2147 case OP_QUERYI:
2148 case OP_MINQUERY:
2149 case OP_MINQUERYI:
2150 case OP_POSQUERY:
2151 case OP_POSQUERYI:
2152 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2153 break;
2154 }
2155 #else
2156 (void)(utf); /* Keep compiler happy by referencing function argument */
2157 #endif
2158 }
2159 }
2160 }
2161
2162
2163
2164 /*************************************************
2165 * Scan compiled regex for recursion reference *
2166 *************************************************/
2167
2168 /* This little function scans through a compiled pattern until it finds an
2169 instance of OP_RECURSE.
2170
2171 Arguments:
2172 code points to start of expression
2173 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2174
2175 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2176 */
2177
2178 static const pcre_uchar *
2179 find_recurse(const pcre_uchar *code, BOOL utf)
2180 {
2181 for (;;)
2182 {
2183 register pcre_uchar c = *code;
2184 if (c == OP_END) return NULL;
2185 if (c == OP_RECURSE) return code;
2186
2187 /* XCLASS is used for classes that cannot be represented just by a bit
2188 map. This includes negated single high-valued characters. The length in
2189 the table is zero; the actual length is stored in the compiled code. */
2190
2191 if (c == OP_XCLASS) code += GET(code, 1);
2192
2193 /* Otherwise, we can get the item's length from the table, except that for
2194 repeated character types, we have to test for \p and \P, which have an extra
2195 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2196 must add in its length. */
2197
2198 else
2199 {
2200 switch(c)
2201 {
2202 case OP_TYPESTAR:
2203 case OP_TYPEMINSTAR:
2204 case OP_TYPEPLUS:
2205 case OP_TYPEMINPLUS:
2206 case OP_TYPEQUERY:
2207 case OP_TYPEMINQUERY:
2208 case OP_TYPEPOSSTAR:
2209 case OP_TYPEPOSPLUS:
2210 case OP_TYPEPOSQUERY:
2211 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2212 break;
2213
2214 case OP_TYPEPOSUPTO:
2215 case OP_TYPEUPTO:
2216 case OP_TYPEMINUPTO:
2217 case OP_TYPEEXACT:
2218 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2219 code += 2;
2220 break;
2221
2222 case OP_MARK:
2223 case OP_PRUNE_ARG:
2224 case OP_SKIP_ARG:
2225 case OP_THEN_ARG:
2226 code += code[1];
2227 break;
2228 }
2229
2230 /* Add in the fixed length from the table */
2231
2232 code += PRIV(OP_lengths)[c];
2233
2234 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2235 by a multi-byte character. The length in the table is a minimum, so we have
2236 to arrange to skip the extra bytes. */
2237
2238 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2239 if (utf) switch(c)
2240 {
2241 case OP_CHAR:
2242 case OP_CHARI:
2243 case OP_NOT:
2244 case OP_NOTI:
2245 case OP_EXACT:
2246 case OP_EXACTI:
2247 case OP_NOTEXACT:
2248 case OP_NOTEXACTI:
2249 case OP_UPTO:
2250 case OP_UPTOI:
2251 case OP_NOTUPTO:
2252 case OP_NOTUPTOI:
2253 case OP_MINUPTO:
2254 case OP_MINUPTOI:
2255 case OP_NOTMINUPTO:
2256 case OP_NOTMINUPTOI:
2257 case OP_POSUPTO:
2258 case OP_POSUPTOI:
2259 case OP_NOTPOSUPTO:
2260 case OP_NOTPOSUPTOI:
2261 case OP_STAR:
2262 case OP_STARI:
2263 case OP_NOTSTAR:
2264 case OP_NOTSTARI:
2265 case OP_MINSTAR:
2266 case OP_MINSTARI:
2267 case OP_NOTMINSTAR:
2268 case OP_NOTMINSTARI:
2269 case OP_POSSTAR:
2270 case OP_POSSTARI:
2271 case OP_NOTPOSSTAR:
2272 case OP_NOTPOSSTARI:
2273 case OP_PLUS:
2274 case OP_PLUSI:
2275 case OP_NOTPLUS:
2276 case OP_NOTPLUSI:
2277 case OP_MINPLUS:
2278 case OP_MINPLUSI:
2279 case OP_NOTMINPLUS:
2280 case OP_NOTMINPLUSI:
2281 case OP_POSPLUS:
2282 case OP_POSPLUSI:
2283 case OP_NOTPOSPLUS:
2284 case OP_NOTPOSPLUSI:
2285 case OP_QUERY:
2286 case OP_QUERYI:
2287 case OP_NOTQUERY:
2288 case OP_NOTQUERYI:
2289 case OP_MINQUERY:
2290 case OP_MINQUERYI:
2291 case OP_NOTMINQUERY:
2292 case OP_NOTMINQUERYI:
2293 case OP_POSQUERY:
2294 case OP_POSQUERYI:
2295 case OP_NOTPOSQUERY:
2296 case OP_NOTPOSQUERYI:
2297 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2298 break;
2299 }
2300 #else
2301 (void)(utf); /* Keep compiler happy by referencing function argument */
2302 #endif
2303 }
2304 }
2305 }
2306
2307
2308
2309 /*************************************************
2310 * Scan compiled branch for non-emptiness *
2311 *************************************************/
2312
2313 /* This function scans through a branch of a compiled pattern to see whether it
2314 can match the empty string or not. It is called from could_be_empty()
2315 below and from compile_branch() when checking for an unlimited repeat of a
2316 group that can match nothing. Note that first_significant_code() skips over
2317 backward and negative forward assertions when its final argument is TRUE. If we
2318 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2319 bracket whose current branch will already have been scanned.
2320
2321 Arguments:
2322 code points to start of search
2323 endcode points to where to stop
2324 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2325 cd contains pointers to tables etc.
2326 recurses chain of recurse_check to catch mutual recursion
2327
2328 Returns: TRUE if what is matched could be empty
2329 */
2330
2331 typedef struct recurse_check {
2332 struct recurse_check *prev;
2333 const pcre_uchar *group;
2334 } recurse_check;
2335
2336 static BOOL
2337 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2338 BOOL utf, compile_data *cd, recurse_check *recurses)
2339 {
2340 register pcre_uchar c;
2341 recurse_check this_recurse;
2342
2343 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2344 code < endcode;
2345 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2346 {
2347 const pcre_uchar *ccode;
2348
2349 c = *code;
2350
2351 /* Skip over forward assertions; the other assertions are skipped by
2352 first_significant_code() with a TRUE final argument. */
2353
2354 if (c == OP_ASSERT)
2355 {
2356 do code += GET(code, 1); while (*code == OP_ALT);
2357 c = *code;
2358 continue;
2359 }
2360
2361 /* For a recursion/subroutine call, if its end has been reached, which
2362 implies a backward reference subroutine call, we can scan it. If it's a
2363 forward reference subroutine call, we can't. To detect forward reference
2364 we have to scan up the list that is kept in the workspace. This function is
2365 called only when doing the real compile, not during the pre-compile that
2366 measures the size of the compiled pattern. */
2367
2368 if (c == OP_RECURSE)
2369 {
2370 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2371 BOOL empty_branch;
2372
2373 /* Test for forward reference or uncompleted reference. This is disabled
2374 when called to scan a completed pattern by setting cd->start_workspace to
2375 NULL. */
2376
2377 if (cd->start_workspace != NULL)
2378 {
2379 const pcre_uchar *tcode;
2380 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2381 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2382 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2383 }
2384
2385 /* If we are scanning a completed pattern, there are no forward references
2386 and all groups are complete. We need to detect whether this is a recursive
2387 call, as otherwise there will be an infinite loop. If it is a recursion,
2388 just skip over it. Simple recursions are easily detected. For mutual
2389 recursions we keep a chain on the stack. */
2390
2391 else
2392 {
2393 recurse_check *r = recurses;
2394 const pcre_uchar *endgroup = scode;
2395
2396 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2397 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2398
2399 for (r = recurses; r != NULL; r = r->prev)
2400 if (r->group == scode) break;
2401 if (r != NULL) continue; /* Mutual recursion */
2402 }
2403
2404 /* Completed reference; scan the referenced group, remembering it on the
2405 stack chain to detect mutual recursions. */
2406
2407 empty_branch = FALSE;
2408 this_recurse.prev = recurses;
2409 this_recurse.group = scode;
2410
2411 do
2412 {
2413 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2414 {
2415 empty_branch = TRUE;
2416 break;
2417 }
2418 scode += GET(scode, 1);
2419 }
2420 while (*scode == OP_ALT);
2421
2422 if (!empty_branch) return FALSE; /* All branches are non-empty */
2423 continue;
2424 }
2425
2426 /* Groups with zero repeats can of course be empty; skip them. */
2427
2428 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2429 c == OP_BRAPOSZERO)
2430 {
2431 code += PRIV(OP_lengths)[c];
2432 do code += GET(code, 1); while (*code == OP_ALT);
2433 c = *code;
2434 continue;
2435 }
2436
2437 /* A nested group that is already marked as "could be empty" can just be
2438 skipped. */
2439
2440 if (c == OP_SBRA || c == OP_SBRAPOS ||
2441 c == OP_SCBRA || c == OP_SCBRAPOS)
2442 {
2443 do code += GET(code, 1); while (*code == OP_ALT);
2444 c = *code;
2445 continue;
2446 }
2447
2448 /* For other groups, scan the branches. */
2449
2450 if (c == OP_BRA || c == OP_BRAPOS ||
2451 c == OP_CBRA || c == OP_CBRAPOS ||
2452 c == OP_ONCE || c == OP_ONCE_NC ||
2453 c == OP_COND)
2454 {
2455 BOOL empty_branch;
2456 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2457
2458 /* If a conditional group has only one branch, there is a second, implied,
2459 empty branch, so just skip over the conditional, because it could be empty.
2460 Otherwise, scan the individual branches of the group. */
2461
2462 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2463 code += GET(code, 1);
2464 else
2465 {
2466 empty_branch = FALSE;
2467 do
2468 {
2469 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2470 empty_branch = TRUE;
2471 code += GET(code, 1);
2472 }
2473 while (*code == OP_ALT);
2474 if (!empty_branch) return FALSE; /* All branches are non-empty */
2475 }
2476
2477 c = *code;
2478 continue;
2479 }
2480
2481 /* Handle the other opcodes */
2482
2483 switch (c)
2484 {
2485 /* Check for quantifiers after a class. XCLASS is used for classes that
2486 cannot be represented just by a bit map. This includes negated single
2487 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2488 actual length is stored in the compiled code, so we must update "code"
2489 here. */
2490
2491 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2492 case OP_XCLASS:
2493 ccode = code += GET(code, 1);
2494 goto CHECK_CLASS_REPEAT;
2495 #endif
2496
2497 case OP_CLASS:
2498 case OP_NCLASS:
2499 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2500
2501 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2502 CHECK_CLASS_REPEAT:
2503 #endif
2504
2505 switch (*ccode)
2506 {
2507 case OP_CRSTAR: /* These could be empty; continue */
2508 case OP_CRMINSTAR:
2509 case OP_CRQUERY:
2510 case OP_CRMINQUERY:
2511 case OP_CRPOSSTAR:
2512 case OP_CRPOSQUERY:
2513 break;
2514
2515 default: /* Non-repeat => class must match */
2516 case OP_CRPLUS: /* These repeats aren't empty */
2517 case OP_CRMINPLUS:
2518 case OP_CRPOSPLUS:
2519 return FALSE;
2520
2521 case OP_CRRANGE:
2522 case OP_CRMINRANGE:
2523 case OP_CRPOSRANGE:
2524 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2525 break;
2526 }
2527 break;
2528
2529 /* Opcodes that must match a character */
2530
2531 case OP_ANY:
2532 case OP_ALLANY:
2533 case OP_ANYBYTE:
2534
2535 case OP_PROP:
2536 case OP_NOTPROP:
2537 case OP_ANYNL:
2538
2539 case OP_NOT_HSPACE:
2540 case OP_HSPACE:
2541 case OP_NOT_VSPACE:
2542 case OP_VSPACE:
2543 case OP_EXTUNI:
2544
2545 case OP_NOT_DIGIT:
2546 case OP_DIGIT:
2547 case OP_NOT_WHITESPACE:
2548 case OP_WHITESPACE:
2549 case OP_NOT_WORDCHAR:
2550 case OP_WORDCHAR:
2551
2552 case OP_CHAR:
2553 case OP_CHARI:
2554 case OP_NOT:
2555 case OP_NOTI:
2556
2557 case OP_PLUS:
2558 case OP_PLUSI:
2559 case OP_MINPLUS:
2560 case OP_MINPLUSI:
2561
2562 case OP_NOTPLUS:
2563 case OP_NOTPLUSI:
2564 case OP_NOTMINPLUS:
2565 case OP_NOTMINPLUSI:
2566
2567 case OP_POSPLUS:
2568 case OP_POSPLUSI:
2569 case OP_NOTPOSPLUS:
2570 case OP_NOTPOSPLUSI:
2571
2572 case OP_EXACT:
2573 case OP_EXACTI:
2574 case OP_NOTEXACT:
2575 case OP_NOTEXACTI:
2576
2577 case OP_TYPEPLUS:
2578 case OP_TYPEMINPLUS:
2579 case OP_TYPEPOSPLUS:
2580 case OP_TYPEEXACT:
2581
2582 return FALSE;
2583
2584 /* These are going to continue, as they may be empty, but we have to
2585 fudge the length for the \p and \P cases. */
2586
2587 case OP_TYPESTAR:
2588 case OP_TYPEMINSTAR:
2589 case OP_TYPEPOSSTAR:
2590 case OP_TYPEQUERY:
2591 case OP_TYPEMINQUERY:
2592 case OP_TYPEPOSQUERY:
2593 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2594 break;
2595
2596 /* Same for these */
2597
2598 case OP_TYPEUPTO:
2599 case OP_TYPEMINUPTO:
2600 case OP_TYPEPOSUPTO:
2601 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2602 code += 2;
2603 break;
2604
2605 /* End of branch */
2606
2607 case OP_KET:
2608 case OP_KETRMAX:
2609 case OP_KETRMIN:
2610 case OP_KETRPOS:
2611 case OP_ALT:
2612 return TRUE;
2613
2614 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2615 MINUPTO, and POSUPTO and their caseless and negative versions may be
2616 followed by a multibyte character. */
2617
2618 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2619 case OP_STAR:
2620 case OP_STARI:
2621 case OP_NOTSTAR:
2622 case OP_NOTSTARI:
2623
2624 case OP_MINSTAR:
2625 case OP_MINSTARI:
2626 case OP_NOTMINSTAR:
2627 case OP_NOTMINSTARI:
2628
2629 case OP_POSSTAR:
2630 case OP_POSSTARI:
2631 case OP_NOTPOSSTAR:
2632 case OP_NOTPOSSTARI:
2633
2634 case OP_QUERY:
2635 case OP_QUERYI:
2636 case OP_NOTQUERY:
2637 case OP_NOTQUERYI:
2638
2639 case OP_MINQUERY:
2640 case OP_MINQUERYI:
2641 case OP_NOTMINQUERY:
2642 case OP_NOTMINQUERYI:
2643
2644 case OP_POSQUERY:
2645 case OP_POSQUERYI:
2646 case OP_NOTPOSQUERY:
2647 case OP_NOTPOSQUERYI:
2648
2649 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2650 break;
2651
2652 case OP_UPTO:
2653 case OP_UPTOI:
2654 case OP_NOTUPTO:
2655 case OP_NOTUPTOI:
2656
2657 case OP_MINUPTO:
2658 case OP_MINUPTOI:
2659 case OP_NOTMINUPTO:
2660 case OP_NOTMINUPTOI:
2661
2662 case OP_POSUPTO:
2663 case OP_POSUPTOI:
2664 case OP_NOTPOSUPTO:
2665 case OP_NOTPOSUPTOI:
2666
2667 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2668 break;
2669 #endif
2670
2671 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2672 string. */
2673
2674 case OP_MARK:
2675 case OP_PRUNE_ARG:
2676 case OP_SKIP_ARG:
2677 case OP_THEN_ARG:
2678 code += code[1];
2679 break;
2680
2681 /* None of the remaining opcodes are required to match a character. */
2682
2683 default:
2684 break;
2685 }
2686 }
2687
2688 return TRUE;
2689 }
2690
2691
2692
2693 /*************************************************
2694 * Scan compiled regex for non-emptiness *
2695 *************************************************/
2696
2697 /* This function is called to check for left recursive calls. We want to check
2698 the current branch of the current pattern to see if it could match the empty
2699 string. If it could, we must look outwards for branches at other levels,
2700 stopping when we pass beyond the bracket which is the subject of the recursion.
2701 This function is called only during the real compile, not during the
2702 pre-compile.
2703
2704 Arguments:
2705 code points to start of the recursion
2706 endcode points to where to stop (current RECURSE item)
2707 bcptr points to the chain of current (unclosed) branch starts
2708 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2709 cd pointers to tables etc
2710
2711 Returns: TRUE if what is matched could be empty
2712 */
2713
2714 static BOOL
2715 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2716 branch_chain *bcptr, BOOL utf, compile_data *cd)
2717 {
2718 while (bcptr != NULL && bcptr->current_branch >= code)
2719 {
2720 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2721 return FALSE;
2722 bcptr = bcptr->outer;
2723 }
2724 return TRUE;
2725 }
2726
2727
2728
2729 /*************************************************
2730 * Base opcode of repeated opcodes *
2731 *************************************************/
2732
2733 /* Returns the base opcode for repeated single character type opcodes. If the
2734 opcode is not a repeated character type, it returns with the original value.
2735
2736 Arguments: c opcode
2737 Returns: base opcode for the type
2738 */
2739
2740 static pcre_uchar
2741 get_repeat_base(pcre_uchar c)
2742 {
2743 return (c > OP_TYPEPOSUPTO)? c :
2744 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2745 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2746 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2747 (c >= OP_STARI)? OP_STARI :
2748 OP_STAR;
2749 }
2750
2751
2752
2753 #ifdef SUPPORT_UCP
2754 /*************************************************
2755 * Check a character and a property *
2756 *************************************************/
2757
2758 /* This function is called by check_auto_possessive() when a property item
2759 is adjacent to a fixed character.
2760
2761 Arguments:
2762 c the character
2763 ptype the property type
2764 pdata the data for the type
2765 negated TRUE if it's a negated property (\P or \p{^)
2766
2767 Returns: TRUE if auto-possessifying is OK
2768 */
2769
2770 static BOOL
2771 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2772 BOOL negated)
2773 {
2774 const pcre_uint32 *p;
2775 const ucd_record *prop = GET_UCD(c);
2776
2777 switch(ptype)
2778 {
2779 case PT_LAMP:
2780 return (prop->chartype == ucp_Lu ||
2781 prop->chartype == ucp_Ll ||
2782 prop->chartype == ucp_Lt) == negated;
2783
2784 case PT_GC:
2785 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2786
2787 case PT_PC:
2788 return (pdata == prop->chartype) == negated;
2789
2790 case PT_SC:
2791 return (pdata == prop->script) == negated;
2792
2793 /* These are specials */
2794
2795 case PT_ALNUM:
2796 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2797 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2798
2799 /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2800 means that Perl space and POSIX space are now identical. PCRE was changed
2801 at release 8.34. */
2802
2803 case PT_SPACE: /* Perl space */
2804 case PT_PXSPACE: /* POSIX space */
2805 switch(c)
2806 {
2807 HSPACE_CASES:
2808 VSPACE_CASES:
2809 return negated;
2810
2811 default:
2812 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2813 }
2814 break; /* Control never reaches here */
2815
2816 case PT_WORD:
2817 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2818 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2819 c == CHAR_UNDERSCORE) == negated;
2820
2821 case PT_CLIST:
2822 p = PRIV(ucd_caseless_sets) + prop->caseset;
2823 for (;;)
2824 {
2825 if (c < *p) return !negated;
2826 if (c == *p++) return negated;
2827 }
2828 break; /* Control never reaches here */
2829 }
2830
2831 return FALSE;
2832 }
2833 #endif /* SUPPORT_UCP */
2834
2835
2836
2837 /*************************************************
2838 * Fill the character property list *
2839 *************************************************/
2840
2841 /* Checks whether the code points to an opcode that can take part in auto-
2842 possessification, and if so, fills a list with its properties.
2843
2844 Arguments:
2845 code points to start of expression
2846 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2847 fcc points to case-flipping table
2848 list points to output list
2849 list[0] will be filled with the opcode
2850 list[1] will be non-zero if this opcode
2851 can match an empty character string
2852 list[2..7] depends on the opcode
2853
2854 Returns: points to the start of the next opcode if *code is accepted
2855 NULL if *code is not accepted
2856 */
2857
2858 static const pcre_uchar *
2859 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2860 const pcre_uint8 *fcc, pcre_uint32 *list)
2861 {
2862 pcre_uchar c = *code;
2863 pcre_uchar base;
2864 const pcre_uchar *end;
2865 pcre_uint32 chr;
2866
2867 #ifdef SUPPORT_UCP
2868 pcre_uint32 *clist_dest;
2869 const pcre_uint32 *clist_src;
2870 #else
2871 utf = utf; /* Suppress "unused parameter" compiler warning */
2872 #endif
2873
2874 list[0] = c;
2875 list[1] = FALSE;
2876 code++;
2877
2878 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2879 {
2880 base = get_repeat_base(c);
2881 c -= (base - OP_STAR);
2882
2883 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2884 code += IMM2_SIZE;
2885
2886 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2887
2888 switch(base)
2889 {
2890 case OP_STAR:
2891 list[0] = OP_CHAR;
2892 break;
2893
2894 case OP_STARI:
2895 list[0] = OP_CHARI;
2896 break;
2897
2898 case OP_NOTSTAR:
2899 list[0] = OP_NOT;
2900 break;
2901
2902 case OP_NOTSTARI:
2903 list[0] = OP_NOTI;
2904 break;
2905
2906 case OP_TYPESTAR:
2907 list[0] = *code;
2908 code++;
2909 break;
2910 }
2911 c = list[0];
2912 }
2913
2914 switch(c)
2915 {
2916 case OP_NOT_DIGIT:
2917 case OP_DIGIT:
2918 case OP_NOT_WHITESPACE:
2919 case OP_WHITESPACE:
2920 case OP_NOT_WORDCHAR:
2921 case OP_WORDCHAR:
2922 case OP_ANY:
2923 case OP_ALLANY:
2924 case OP_ANYNL:
2925 case OP_NOT_HSPACE:
2926 case OP_HSPACE:
2927 case OP_NOT_VSPACE:
2928 case OP_VSPACE:
2929 case OP_EXTUNI:
2930 case OP_EODN:
2931 case OP_EOD:
2932 case OP_DOLL:
2933 case OP_DOLLM:
2934 return code;
2935
2936 case OP_CHAR:
2937 case OP_NOT:
2938 GETCHARINCTEST(chr, code);
2939 list[2] = chr;
2940 list[3] = NOTACHAR;
2941 return code;
2942
2943 case OP_CHARI:
2944 case OP_NOTI:
2945 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2946 GETCHARINCTEST(chr, code);
2947 list[2] = chr;
2948
2949 #ifdef SUPPORT_UCP
2950 if (chr < 128 || (chr < 256 && !utf))
2951 list[3] = fcc[chr];
2952 else
2953 list[3] = UCD_OTHERCASE(chr);
2954 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2955 list[3] = (chr < 256) ? fcc[chr] : chr;
2956 #else
2957 list[3] = fcc[chr];
2958 #endif
2959
2960 /* The othercase might be the same value. */
2961
2962 if (chr == list[3])
2963 list[3] = NOTACHAR;
2964 else
2965 list[4] = NOTACHAR;
2966 return code;
2967
2968 #ifdef SUPPORT_UCP
2969 case OP_PROP:
2970 case OP_NOTPROP:
2971 if (code[0] != PT_CLIST)
2972 {
2973 list[2] = code[0];
2974 list[3] = code[1];
2975 return code + 2;
2976 }
2977
2978 /* Convert only if we have enough space. */
2979
2980 clist_src = PRIV(ucd_caseless_sets) + code[1];
2981 clist_dest = list + 2;
2982 code += 2;
2983
2984 do {
2985 if (clist_dest >= list + 8)
2986 {
2987 /* Early return if there is not enough space. This should never
2988 happen, since all clists are shorter than 5 character now. */
2989 list[2] = code[0];
2990 list[3] = code[1];
2991 return code;
2992 }
2993 *clist_dest++ = *clist_src;
2994 }
2995 while(*clist_src++ != NOTACHAR);
2996
2997 /* All characters are stored. The terminating NOTACHAR
2998 is copied form the clist itself. */
2999
3000 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3001 return code;
3002 #endif
3003
3004 case OP_NCLASS:
3005 case OP_CLASS:
3006 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3007 case OP_XCLASS:
3008 if (c == OP_XCLASS)
3009 end = code + GET(code, 0) - 1;
3010 else
3011 #endif
3012 end = code + 32 / sizeof(pcre_uchar);
3013
3014 switch(*end)
3015 {
3016 case OP_CRSTAR:
3017 case OP_CRMINSTAR:
3018 case OP_CRQUERY:
3019 case OP_CRMINQUERY:
3020 case OP_CRPOSSTAR:
3021 case OP_CRPOSQUERY:
3022 list[1] = TRUE;
3023 end++;
3024 break;
3025
3026 case OP_CRPLUS:
3027 case OP_CRMINPLUS:
3028 case OP_CRPOSPLUS:
3029 end++;
3030 break;
3031
3032 case OP_CRRANGE:
3033 case OP_CRMINRANGE:
3034 case OP_CRPOSRANGE:
3035 list[1] = (GET2(end, 1) == 0);
3036 end += 1 + 2 * IMM2_SIZE;
3037 break;
3038 }
3039 list[2] = end - code;
3040 return end;
3041 }
3042 return NULL; /* Opcode not accepted */
3043 }
3044
3045
3046
3047 /*************************************************
3048 * Scan further character sets for match *
3049 *************************************************/
3050
3051 /* Checks whether the base and the current opcode have a common character, in
3052 which case the base cannot be possessified.
3053
3054 Arguments:
3055 code points to the byte code
3056 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3057 cd static compile data
3058 base_list the data list of the base opcode
3059
3060 Returns: TRUE if the auto-possessification is possible
3061 */
3062
3063 static BOOL
3064 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3065 const pcre_uint32 *base_list, const pcre_uchar *base_end)
3066 {
3067 pcre_uchar c;
3068 pcre_uint32 list[8];
3069 const pcre_uint32 *chr_ptr;
3070 const pcre_uint32 *ochr_ptr;
3071 const pcre_uint32 *list_ptr;
3072 const pcre_uchar *next_code;
3073 const pcre_uint8 *class_bitset;
3074 const pcre_uint32 *set1, *set2, *set_end;
3075 pcre_uint32 chr;
3076 BOOL accepted, invert_bits;
3077
3078 /* Note: the base_list[1] contains whether the current opcode has greedy
3079 (represented by a non-zero value) quantifier. This is a different from
3080 other character type lists, which stores here that the character iterator
3081 matches to an empty string (also represented by a non-zero value). */
3082
3083 for(;;)
3084 {
3085 /* All operations move the code pointer forward.
3086 Therefore infinite recursions are not possible. */
3087
3088 c = *code;
3089
3090 /* Skip over callouts */
3091
3092 if (c == OP_CALLOUT)
3093 {
3094 code += PRIV(OP_lengths)[c];
3095 continue;
3096 }
3097
3098 if (c == OP_ALT)
3099 {
3100 do code += GET(code, 1); while (*code == OP_ALT);
3101 c = *code;
3102 }
3103
3104 switch(c)
3105 {
3106 case OP_END:
3107 case OP_KETRPOS:
3108 /* TRUE only in greedy case. The non-greedy case could be replaced by
3109 an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3110 uses more memory, which we cannot get at this stage.) */
3111
3112 return base_list[1] != 0;
3113
3114 case OP_KET:
3115 /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3116 it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3117 cannot be converted to a possessive form. */
3118
3119 if (base_list[1] == 0) return FALSE;
3120
3121 switch(*(code - GET(code, 1)))
3122 {
3123 case OP_ASSERT:
3124 case OP_ASSERT_NOT:
3125 case OP_ASSERTBACK:
3126 case OP_ASSERTBACK_NOT:
3127 case OP_ONCE:
3128 case OP_ONCE_NC:
3129 /* Atomic sub-patterns and assertions can always auto-possessify their
3130 last iterator. */
3131 return TRUE;
3132 }
3133
3134 code += PRIV(OP_lengths)[c];
3135 continue;
3136
3137 case OP_ONCE:
3138 case OP_ONCE_NC:
3139 case OP_BRA:
3140 case OP_CBRA:
3141 next_code = code + GET(code, 1);
3142 code += PRIV(OP_lengths)[c];
3143
3144 while (*next_code == OP_ALT)
3145 {
3146 if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3147 code = next_code + 1 + LINK_SIZE;
3148 next_code += GET(next_code, 1);
3149 }
3150 continue;
3151
3152 case OP_BRAZERO:
3153 case OP_BRAMINZERO:
3154
3155 next_code = code + 1;
3156 if (*next_code != OP_BRA && *next_code != OP_CBRA
3157 && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3158
3159 do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3160
3161 /* The bracket content will be checked by the
3162 OP_BRA/OP_CBRA case above. */
3163 next_code += 1 + LINK_SIZE;
3164 if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3165 return FALSE;
3166
3167 code += PRIV(OP_lengths)[c];
3168 continue;
3169 }
3170
3171 /* Check for a supported opcode, and load its properties. */
3172
3173 code = get_chr_property_list(code, utf, cd->fcc, list);
3174 if (code == NULL) return FALSE; /* Unsupported */
3175
3176 /* If either opcode is a small character list, set pointers for comparing
3177 characters from that list with another list, or with a property. */
3178
3179 if (base_list[0] == OP_CHAR)
3180 {
3181 chr_ptr = base_list + 2;
3182 list_ptr = list;
3183 }
3184 else if (list[0] == OP_CHAR)
3185 {
3186 chr_ptr = list + 2;
3187 list_ptr = base_list;
3188 }
3189
3190 /* Character bitsets can also be compared to certain opcodes. */
3191
3192 else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3193 #ifdef COMPILE_PCRE8
3194 /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3195 || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3196 #endif
3197 )
3198 {
3199 #ifdef COMPILE_PCRE8
3200 if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3201 #else
3202 if (base_list[0] == OP_CLASS)
3203 #endif
3204 {
3205 set1 = (pcre_uint32 *)(base_end - base_list[2]);
3206 list_ptr = list;
3207 }
3208 else
3209 {
3210 set1 = (pcre_uint32 *)(code - list[2]);
3211 list_ptr = base_list;
3212 }
3213
3214 invert_bits = FALSE;
3215 switch(list_ptr[0])
3216 {
3217 case OP_CLASS:
3218 case OP_NCLASS:
3219 set2 = (pcre_uint32 *)
3220 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3221 break;
3222
3223 /* OP_XCLASS cannot be supported here, because its bitset
3224 is not necessarily complete. E.g: [a-\0x{200}] is stored
3225 as a character range, and the appropriate bits are not set. */
3226
3227 case OP_NOT_DIGIT:
3228 invert_bits = TRUE;
3229 /* Fall through */
3230 case OP_DIGIT:
3231 set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3232 break;
3233
3234 case OP_NOT_WHITESPACE:
3235 invert_bits = TRUE;
3236 /* Fall through */
3237 case OP_WHITESPACE:
3238 set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3239 break;
3240
3241 case OP_NOT_WORDCHAR:
3242 invert_bits = TRUE;
3243 /* Fall through */
3244 case OP_WORDCHAR:
3245 set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3246 break;
3247
3248 default:
3249 return FALSE;
3250 }
3251
3252 /* Compare 4 bytes to improve speed. */
3253 set_end = set1 + (32 / 4);
3254 if (invert_bits)
3255 {
3256 do
3257 {
3258 if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3259 }
3260 while (set1 < set_end);
3261 }
3262 else
3263 {
3264 do
3265 {
3266 if ((*set1++ & *set2++) != 0) return FALSE;
3267 }
3268 while (set1 < set_end);
3269 }
3270
3271 if (list[1] == 0) return TRUE;
3272 /* Might be an empty repeat. */
3273 continue;
3274 }
3275
3276 /* Some property combinations also acceptable. Unicode property opcodes are
3277 processed specially; the rest can be handled with a lookup table. */
3278
3279 else
3280 {
3281 pcre_uint32 leftop, rightop;
3282
3283 leftop = base_list[0];
3284 rightop = list[0];
3285
3286 #ifdef SUPPORT_UCP
3287 accepted = FALSE; /* Always set in non-unicode case. */
3288 if (leftop == OP_PROP || leftop == OP_NOTPROP)
3289 {
3290 if (rightop == OP_EOD)
3291 accepted = TRUE;
3292 else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3293 {
3294 int n;
3295 const pcre_uint8 *p;
3296 BOOL same = leftop == rightop;
3297 BOOL lisprop = leftop == OP_PROP;
3298 BOOL risprop = rightop == OP_PROP;
3299 BOOL bothprop = lisprop && risprop;
3300
3301 /* There's a table that specifies how each combination is to be
3302 processed:
3303 0 Always return FALSE (never auto-possessify)
3304 1 Character groups are distinct (possessify if both are OP_PROP)
3305 2 Check character categories in the same group (general or particular)
3306 3 Return TRUE if the two opcodes are not the same
3307 ... see comments below
3308 */
3309
3310 n = propposstab[base_list[2]][list[2]];
3311 switch(n)
3312 {
3313 case 0: break;
3314 case 1: accepted = bothprop; break;
3315 case 2: accepted = (base_list[3] == list[3]) != same; break;
3316 case 3: accepted = !same; break;
3317
3318 case 4: /* Left general category, right particular category */
3319 accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3320 break;
3321
3322 case 5: /* Right general category, left particular category */
3323 accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3324 break;
3325
3326 /* This code is logically tricky. Think hard before fiddling with it.
3327 The posspropstab table has four entries per row. Each row relates to
3328 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3329 Only WORD actually needs all four entries, but using repeats for the
3330 others means they can all use the same code below.
3331
3332 The first two entries in each row are Unicode general categories, and
3333 apply always, because all the characters they include are part of the
3334 PCRE character set. The third and fourth entries are a general and a
3335 particular category, respectively, that include one or more relevant
3336 characters. One or the other is used, depending on whether the check
3337 is for a general or a particular category. However, in both cases the
3338 category contains more characters than the specials that are defined
3339 for the property being tested against. Therefore, it cannot be used
3340 in a NOTPROP case.
3341
3342 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3343 Underscore is covered by ucp_P or ucp_Po. */
3344
3345 case 6: /* Left alphanum vs right general category */
3346 case 7: /* Left space vs right general category */
3347 case 8: /* Left word vs right general category */
3348 p = posspropstab[n-6];
3349 accepted = risprop && lisprop ==
3350 (list[3] != p[0] &&
3351 list[3] != p[1] &&
3352 (list[3] != p[2] || !lisprop));
3353 break;
3354
3355 case 9: /* Right alphanum vs left general category */
3356 case 10: /* Right space vs left general category */
3357 case 11: /* Right word vs left general category */
3358 p = posspropstab[n-9];
3359 accepted = lisprop && risprop ==
3360 (base_list[3] != p[0] &&
3361 base_list[3] != p[1] &&
3362 (base_list[3] != p[2] || !risprop));
3363 break;
3364
3365 case 12: /* Left alphanum vs right particular category */
3366 case 13: /* Left space vs right particular category */
3367 case 14: /* Left word vs right particular category */
3368 p = posspropstab[n-12];
3369 accepted = risprop && lisprop ==
3370 (catposstab[p[0]][list[3]] &&
3371 catposstab[p[1]][list[3]] &&
3372 (list[3] != p[3] || !lisprop));
3373 break;
3374
3375 case 15: /* Right alphanum vs left particular category */
3376 case 16: /* Right space vs left particular category */
3377 case 17: /* Right word vs left particular category */
3378 p = posspropstab[n-15];
3379 accepted = lisprop && risprop ==
3380 (catposstab[p[0]][base_list[3]] &&
3381 catposstab[p[1]][base_list[3]] &&
3382 (base_list[3] != p[3] || !risprop));
3383 break;
3384 }
3385 }
3386 }
3387
3388 else
3389 #endif /* SUPPORT_UCP */
3390
3391 accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3392 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3393 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3394
3395 if (!accepted)
3396 return FALSE;
3397
3398 if (list[1] == 0) return TRUE;
3399 /* Might be an empty repeat. */
3400 continue;
3401 }
3402
3403 /* Control reaches here only if one of the items is a small character list.
3404 All characters are checked against the other side. */
3405
3406 do
3407 {
3408 chr = *chr_ptr;
3409
3410 switch(list_ptr[0])
3411 {
3412 case OP_CHAR:
3413 ochr_ptr = list_ptr + 2;
3414 do
3415 {
3416 if (chr == *ochr_ptr) return FALSE;
3417 ochr_ptr++;
3418 }
3419 while(*ochr_ptr != NOTACHAR);
3420 break;
3421
3422 case OP_NOT:
3423 ochr_ptr = list_ptr + 2;
3424 do
3425 {
3426 if (chr == *ochr_ptr)
3427 break;
3428 ochr_ptr++;
3429 }
3430 while(*ochr_ptr != NOTACHAR);
3431 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3432 break;
3433
3434 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3435 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3436
3437 case OP_DIGIT:
3438 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3439 break;
3440
3441 case OP_NOT_DIGIT:
3442 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3443 break;
3444
3445 case OP_WHITESPACE:
3446 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3447 break;
3448
3449 case OP_NOT_WHITESPACE:
3450 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3451 break;
3452
3453 case OP_WORDCHAR:
3454 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3455 break;
3456
3457 case OP_NOT_WORDCHAR:
3458 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3459 break;
3460
3461 case OP_HSPACE:
3462 switch(chr)
3463 {
3464 HSPACE_CASES: return FALSE;
3465 default: break;
3466 }
3467 break;
3468
3469 case OP_NOT_HSPACE:
3470 switch(chr)
3471 {
3472 HSPACE_CASES: break;
3473 default: return FALSE;
3474 }
3475 break;
3476
3477 case OP_ANYNL:
3478 case OP_VSPACE:
3479 switch(chr)
3480 {
3481 VSPACE_CASES: return FALSE;
3482 default: break;
3483 }
3484 break;
3485
3486 case OP_NOT_VSPACE:
3487 switch(chr)
3488 {
3489 VSPACE_CASES: break;
3490 default: return FALSE;
3491 }
3492 break;
3493
3494 case OP_DOLL:
3495 case OP_EODN:
3496 switch (chr)
3497 {
3498 case CHAR_CR:
3499 case CHAR_LF:
3500 case CHAR_VT:
3501 case CHAR_FF:
3502 case CHAR_NEL:
3503 #ifndef EBCDIC
3504 case 0x2028:
3505 case 0x2029:
3506 #endif /* Not EBCDIC */
3507 return FALSE;
3508 }
3509 break;
3510
3511 case OP_EOD: /* Can always possessify before \z */
3512 break;
3513
3514 #ifdef SUPPORT_UCP
3515 case OP_PROP:
3516 case OP_NOTPROP:
3517 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3518 list_ptr[0] == OP_NOTPROP))
3519 return FALSE;
3520 break;
3521 #endif
3522
3523 case OP_NCLASS:
3524 if (chr > 255) return FALSE;
3525 /* Fall through */
3526
3527 case OP_CLASS:
3528 if (chr > 255) break;
3529 class_bitset = (pcre_uint8 *)
3530 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3531 if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3532 break;
3533
3534 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3535 case OP_XCLASS:
3536 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3537 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3538 break;
3539 #endif
3540
3541 default:
3542 return FALSE;
3543 }
3544
3545 chr_ptr++;
3546 }
3547 while(*chr_ptr != NOTACHAR);
3548
3549 /* At least one character must be matched from this opcode. */
3550
3551 if (list[1] == 0) return TRUE;
3552 }
3553
3554 return FALSE;
3555 }
3556
3557
3558
3559 /*************************************************
3560 * Scan compiled regex for auto-possession *
3561 *************************************************/
3562
3563 /* Replaces single character iterations with their possessive alternatives
3564 if appropriate. This function modifies the compiled opcode!
3565
3566 Arguments:
3567 code points to start of the byte code
3568 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3569 cd static compile data
3570
3571 Returns: nothing
3572 */
3573
3574 static void
3575 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3576 {
3577 register pcre_uchar c;
3578 const pcre_uchar *end;
3579 pcre_uchar *repeat_opcode;
3580 pcre_uint32 list[8];
3581
3582 for (;;)
3583 {
3584 c = *code;
3585
3586 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3587 {
3588 c -= get_repeat_base(c) - OP_STAR;
3589 end = (c <= OP_MINUPTO) ?
3590 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3591 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3592
3593 if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3594 {
3595 switch(c)
3596 {
3597 case OP_STAR:
3598 *code += OP_POSSTAR - OP_STAR;
3599 break;
3600
3601 case OP_MINSTAR:
3602 *code += OP_POSSTAR - OP_MINSTAR;
3603 break;
3604
3605 case OP_PLUS:
3606 *code += OP_POSPLUS - OP_PLUS;
3607 break;
3608
3609 case OP_MINPLUS:
3610 *code += OP_POSPLUS - OP_MINPLUS;
3611 break;
3612
3613 case OP_QUERY:
3614 *code += OP_POSQUERY - OP_QUERY;
3615 break;
3616
3617 case OP_MINQUERY:
3618 *code += OP_POSQUERY - OP_MINQUERY;
3619 break;
3620
3621 case OP_UPTO:
3622 *code += OP_POSUPTO - OP_UPTO;
3623 break;
3624
3625 case OP_MINUPTO:
3626 *code += OP_MINUPTO - OP_UPTO;
3627 break;
3628 }
3629 }
3630 c = *code;
3631 }
3632 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3633 {
3634 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3635 if (c == OP_XCLASS)
3636 repeat_opcode = code + GET(code, 1);
3637 else
3638 #endif
3639 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3640
3641 c = *repeat_opcode;
3642 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3643 {
3644 /* end must not be NULL. */
3645 end = get_chr_property_list(code, utf, cd->fcc, list);
3646
3647 list[1] = (c & 1) == 0;
3648
3649 if (compare_opcodes(end, utf, cd, list, end))
3650 {
3651 switch (c)
3652 {
3653 case OP_CRSTAR:
3654 case OP_CRMINSTAR:
3655 *repeat_opcode = OP_CRPOSSTAR;
3656 break;
3657
3658 case OP_CRPLUS:
3659 case OP_CRMINPLUS:
3660 *repeat_opcode = OP_CRPOSPLUS;
3661 break;
3662
3663 case OP_CRQUERY:
3664 case OP_CRMINQUERY:
3665 *repeat_opcode = OP_CRPOSQUERY;
3666 break;
3667
3668 case OP_CRRANGE:
3669 case OP_CRMINRANGE:
3670 *repeat_opcode = OP_CRPOSRANGE;
3671 break;
3672 }
3673 }
3674 }
3675 c = *code;
3676 }
3677
3678 switch(c)
3679 {
3680 case OP_END:
3681 return;
3682
3683 case OP_TYPESTAR:
3684 case OP_TYPEMINSTAR:
3685 case OP_TYPEPLUS:
3686 case OP_TYPEMINPLUS:
3687 case OP_TYPEQUERY:
3688 case OP_TYPEMINQUERY:
3689 case OP_TYPEPOSSTAR:
3690 case OP_TYPEPOSPLUS:
3691 case OP_TYPEPOSQUERY:
3692 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3693 break;
3694
3695 case OP_TYPEUPTO:
3696 case OP_TYPEMINUPTO:
3697 case OP_TYPEEXACT:
3698 case OP_TYPEPOSUPTO:
3699 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3700 code += 2;
3701 break;
3702
3703 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3704 case OP_XCLASS:
3705 code += GET(code, 1);
3706 break;
3707 #endif
3708
3709 case OP_MARK:
3710 case OP_PRUNE_ARG:
3711 case OP_SKIP_ARG:
3712 case OP_THEN_ARG:
3713 code += code[1];
3714 break;
3715 }
3716
3717 /* Add in the fixed length from the table */
3718
3719 code += PRIV(OP_lengths)[c];
3720
3721 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3722 a multi-byte character. The length in the table is a minimum, so we have to
3723 arrange to skip the extra bytes. */
3724
3725 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3726 if (utf) switch(c)
3727 {
3728 case OP_CHAR:
3729 case OP_CHARI:
3730 case OP_NOT:
3731 case OP_NOTI:
3732 case OP_STAR:
3733 case OP_MINSTAR:
3734 case OP_PLUS:
3735 case OP_MINPLUS:
3736 case OP_QUERY:
3737 case OP_MINQUERY:
3738 case OP_UPTO:
3739 case OP_MINUPTO:
3740 case OP_EXACT:
3741 case OP_POSSTAR:
3742 case OP_POSPLUS:
3743 case OP_POSQUERY:
3744 case OP_POSUPTO:
3745 case OP_STARI:
3746 case OP_MINSTARI:
3747 case OP_PLUSI:
3748 case OP_MINPLUSI:
3749 case OP_QUERYI:
3750 case OP_MINQUERYI:
3751 case OP_UPTOI:
3752 case OP_MINUPTOI:
3753 case OP_EXACTI:
3754 case OP_POSSTARI:
3755 case OP_POSPLUSI:
3756 case OP_POSQUERYI:
3757 case OP_POSUPTOI:
3758 case OP_NOTSTAR:
3759 case OP_NOTMINSTAR:
3760 case OP_NOTPLUS:
3761 case OP_NOTMINPLUS:
3762 case OP_NOTQUERY:
3763 case OP_NOTMINQUERY:
3764 case OP_NOTUPTO:
3765 case OP_NOTMINUPTO:
3766 case OP_NOTEXACT:
3767 case OP_NOTPOSSTAR:
3768 case OP_NOTPOSPLUS:
3769 case OP_NOTPOSQUERY:
3770 case OP_NOTPOSUPTO:
3771 case OP_NOTSTARI:
3772 case OP_NOTMINSTARI:
3773 case OP_NOTPLUSI:
3774 case OP_NOTMINPLUSI:
3775 case OP_NOTQUERYI:
3776 case OP_NOTMINQUERYI:
3777 case OP_NOTUPTOI:
3778 case OP_NOTMINUPTOI:
3779 case OP_NOTEXACTI:
3780 case OP_NOTPOSSTARI:
3781 case OP_NOTPOSPLUSI:
3782 case OP_NOTPOSQUERYI:
3783 case OP_NOTPOSUPTOI:
3784 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3785 break;
3786 }
3787 #else
3788 (void)(utf); /* Keep compiler happy by referencing function argument */
3789 #endif
3790 }
3791 }
3792
3793
3794
3795 /*************************************************
3796 * Check for POSIX class syntax *
3797 *************************************************/
3798
3799 /* This function is called when the sequence "[:" or "[." or "[=" is
3800 encountered in a character class. It checks whether this is followed by a
3801 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3802 reach an unescaped ']' without the special preceding character, return FALSE.
3803
3804 Originally, this function only recognized a sequence of letters between the
3805 terminators, but it seems that Perl recognizes any sequence of characters,
3806 though of course unknown POSIX names are subsequently rejected. Perl gives an
3807 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3808 didn't consider this to be a POSIX class. Likewise for [:1234:].
3809
3810 The problem in trying to be exactly like Perl is in the handling of escapes. We
3811 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3812 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3813 below handles the special case of \], but does not try to do any other escape
3814 processing. This makes it different from Perl for cases such as [:l\ower:]
3815 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3816 "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3817 I think.
3818
3819 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3820 It seems that the appearance of a nested POSIX class supersedes an apparent
3821 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3822 a digit.
3823
3824 In Perl, unescaped square brackets may also appear as part of class names. For
3825 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3826 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3827 seem right at all. PCRE does not allow closing square brackets in POSIX class
3828 names.
3829
3830 Arguments:
3831 ptr pointer to the initial [
3832 endptr where to return the end pointer
3833
3834 Returns: TRUE or FALSE
3835 */
3836
3837 static BOOL
3838 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3839 {
3840 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3841 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3842 for (++ptr; *ptr != CHAR_NULL; ptr++)
3843 {
3844 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3845 ptr++;
3846 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3847 else
3848 {
3849 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3850 {
3851 *endptr = ptr;
3852 return TRUE;
3853 }
3854 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3855 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3856 ptr[1] == CHAR_EQUALS_SIGN) &&
3857 check_posix_syntax(ptr, endptr))
3858 return FALSE;
3859 }
3860 }
3861 return FALSE;
3862 }
3863
3864
3865
3866
3867 /*************************************************
3868 * Check POSIX class name *
3869 *************************************************/
3870
3871 /* This function is called to check the name given in a POSIX-style class entry
3872 such as [:alnum:].
3873
3874 Arguments:
3875 ptr points to the first letter
3876 len the length of the name
3877
3878 Returns: a value representing the name, or -1 if unknown
3879 */
3880
3881 static int
3882 check_posix_name(const pcre_uchar *ptr, int len)
3883 {
3884 const char *pn = posix_names;
3885 register int yield = 0;
3886 while (posix_name_lengths[yield] != 0)
3887 {
3888 if (len == posix_name_lengths[yield] &&
3889 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3890 pn += posix_name_lengths[yield] + 1;
3891 yield++;
3892 }
3893 return -1;
3894 }
3895
3896
3897 /*************************************************
3898 * Adjust OP_RECURSE items in repeated group *
3899 *************************************************/
3900
3901 /* OP_RECURSE items contain an offset from the start of the regex to the group
3902 that is referenced. This means that groups can be replicated for fixed
3903 repetition simply by copying (because the recursion is allowed to refer to
3904 earlier groups that are outside the current group). However, when a group is
3905 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3906 inserted before it, after it has been compiled. This means that any OP_RECURSE
3907 items within it that refer to the group itself or any contained groups have to
3908 have their offsets adjusted. That one of the jobs of this function. Before it
3909 is called, the partially compiled regex must be temporarily terminated with
3910 OP_END.
3911
3912 This function has been extended with the possibility of forward references for
3913 recursions and subroutine calls. It must also check the list of such references
3914 for the group we are dealing with. If it finds that one of the recursions in
3915 the current group is on this list, it adjusts the offset in the list, not the
3916 value in the reference (which is a group number).
3917
3918 Arguments:
3919 group points to the start of the group
3920 adjust the amount by which the group is to be moved
3921 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3922 cd contains pointers to tables etc.
3923 save_hwm the hwm forward reference pointer at the start of the group
3924
3925 Returns: nothing
3926 */
3927
3928 static void
3929 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3930 pcre_uchar *save_hwm)
3931 {
3932 pcre_uchar *ptr = group;
3933
3934 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3935 {
3936 int offset;
3937 pcre_uchar *hc;
3938
3939 /* See if this recursion is on the forward reference list. If so, adjust the
3940 reference. */
3941
3942 for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3943 {
3944 offset = (int)GET(hc, 0);
3945 if (cd->start_code + offset == ptr + 1)
3946 {
3947 PUT(hc, 0, offset + adjust);
3948 break;
3949 }
3950 }
3951
3952 /* Otherwise, adjust the recursion offset if it's after the start of this
3953 group. */
3954
3955 if (hc >= cd->hwm)
3956 {
3957 offset = (int)GET(ptr, 1);
3958 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3959 }
3960
3961 ptr += 1 + LINK_SIZE;
3962 }
3963 }
3964
3965
3966
3967 /*************************************************
3968 * Insert an automatic callout point *
3969 *************************************************/
3970
3971 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3972 callout points before each pattern item.
3973
3974 Arguments:
3975 code current code pointer
3976 ptr current pattern pointer
3977 cd pointers to tables etc
3978
3979 Returns: new code pointer
3980 */
3981
3982 static pcre_uchar *
3983 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3984 {
3985 *code++ = OP_CALLOUT;
3986 *code++ = 255;
3987 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
3988 PUT(code, LINK_SIZE, 0); /* Default length */
3989 return code + 2 * LINK_SIZE;
3990 }
3991
3992
3993
3994 /*************************************************
3995 * Complete a callout item *
3996 *************************************************/
3997
3998 /* A callout item contains the length of the next item in the pattern, which
3999 we can't fill in till after we have reached the relevant point. This is used
4000 for both automatic and manual callouts.
4001
4002 Arguments:
4003 previous_callout points to previous callout item
4004 ptr current pattern pointer
4005 cd pointers to tables etc
4006
4007 Returns: nothing
4008 */
4009
4010 static void
4011 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4012 {
4013 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4014 PUT(previous_callout, 2 + LINK_SIZE, length);
4015 }
4016
4017
4018
4019 #ifdef SUPPORT_UCP
4020 /*************************************************
4021 * Get othercase range *
4022 *************************************************/
4023
4024 /* This function is passed the start and end of a class range, in UTF-8 mode
4025 with UCP support. It searches up the characters, looking for ranges of
4026 characters in the "other" case. Each call returns the next one, updating the
4027 start address. A character with multiple other cases is returned on its own
4028 with a special return value.
4029
4030 Arguments:
4031 cptr points to starting character value; updated
4032 d end value
4033 ocptr where to put start of othercase range
4034 odptr where to put end of othercase range
4035
4036 Yield: -1 when no more
4037 0 when a range is returned
4038 >0 the CASESET offset for char with multiple other cases
4039 in this case, ocptr contains the original
4040 */
4041
4042 static int
4043 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4044 pcre_uint32 *odptr)
4045 {
4046 pcre_uint32 c, othercase, next;
4047 unsigned int co;
4048
4049 /* Find the first character that has an other case. If it has multiple other
4050 cases, return its case offset value. */
4051
4052 for (c = *cptr; c <= d; c++)
4053 {
4054 if ((co = UCD_CASESET(c)) != 0)
4055 {
4056 *ocptr = c++; /* Character that has the set */
4057 *cptr = c; /* Rest of input range */
4058 return (int)co;
4059 }
4060 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4061 }
4062
4063 if (c > d) return -1; /* Reached end of range */
4064
4065 *ocptr = othercase;
4066 next = othercase + 1;
4067
4068 for (++c; c <= d; c++)
4069 {
4070 if (UCD_OTHERCASE(c) != next) break;
4071 next++;
4072 }
4073
4074 *odptr = next - 1; /* End of othercase range */
4075 *cptr = c; /* Rest of input range */
4076 return 0;
4077 }
4078 #endif /* SUPPORT_UCP */
4079
4080
4081
4082 /*************************************************
4083 * Add a character or range to a class *
4084 *************************************************/
4085
4086 /* This function packages up the logic of adding a character or range of
4087 characters to a class. The character values in the arguments will be within the
4088 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4089 mutually recursive with the function immediately below.
4090
4091 Arguments:
4092 classbits the bit map for characters < 256
4093 uchardptr points to the pointer for extra data
4094 options the options word
4095 cd contains pointers to tables etc.
4096 start start of range character
4097 end end of range character
4098
4099 Returns: the number of < 256 characters added
4100 the pointer to extra data is updated
4101 */
4102
4103 static int
4104 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4105 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4106 {
4107 pcre_uint32 c;
4108 pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4109 int n8 = 0;
4110
4111 /* If caseless matching is required, scan the range and process alternate
4112 cases. In Unicode, there are 8-bit characters that have alternate cases that
4113 are greater than 255 and vice-versa. Sometimes we can just extend the original
4114 range. */
4115
4116 if ((options & PCRE_CASELESS) != 0)
4117 {
4118 #ifdef SUPPORT_UCP
4119 if ((options & PCRE_UTF8) != 0)
4120 {
4121 int rc;
4122 pcre_uint32 oc, od;
4123
4124 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4125 c = start;
4126
4127 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4128 {
4129 /* Handle a single character that has more than one other case. */
4130
4131 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4132 PRIV(ucd_caseless_sets) + rc, oc);
4133
4134 /* Do nothing if the other case range is within the original range. */
4135
4136 else if (oc >= start && od <= end) continue;
4137
4138 /* Extend the original range if there is overlap, noting that if oc < c, we
4139 can't have od > end because a subrange is always shorter than the basic
4140 range. Otherwise, use a recursive call to add the additional range. */
4141
4142 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4143 else if (od > end && oc <= end + 1) end = od; /* Extend upwards */
4144 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4145 }
4146 }
4147 else
4148 #endif /* SUPPORT_UCP */
4149
4150 /* Not UTF-mode, or no UCP */
4151
4152 for (c = start; c <= classbits_end; c++)
4153 {
4154 SETBIT(classbits, cd->fcc[c]);
4155 n8++;
4156 }
4157 }
4158
4159 /* Now handle the original range. Adjust the final value according to the bit
4160 length - this means that the same lists of (e.g.) horizontal spaces can be used
4161 in all cases. */
4162
4163 #if defined COMPILE_PCRE8
4164 #ifdef SUPPORT_UTF
4165 if ((options & PCRE_UTF8) == 0)
4166 #endif
4167 if (end > 0xff) end = 0xff;
4168
4169 #elif defined COMPILE_PCRE16
4170 #ifdef SUPPORT_UTF
4171 if ((options & PCRE_UTF16) == 0)
4172 #endif
4173 if (end > 0xffff) end = 0xffff;
4174
4175 #endif /* COMPILE_PCRE[8|16] */
4176
4177 /* Use the bitmap for characters < 256. Otherwise use extra data.*/
4178
4179 for (c = start; c <= classbits_end; c++)
4180 {
4181 /* Regardless of start, c will always be <= 255. */
4182 SETBIT(classbits, c);
4183 n8++;
4184 }
4185
4186 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4187 if (start <= 0xff) start = 0xff + 1;
4188
4189 if (end >= start) {
4190 pcre_uchar *uchardata = *uchardptr;
4191
4192 #ifdef SUPPORT_UTF
4193 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4194 {
4195 if (start < end)
4196 {
4197 *uchardata++ = XCL_RANGE;
4198 uchardata += PRIV(ord2utf)(start, uchardata);
4199 uchardata += PRIV(ord2utf)(end, uchardata);
4200 }
4201 else if (start == end)
4202 {
4203 *uchardata++ = XCL_SINGLE;
4204 uchardata += PRIV(ord2utf)(start, uchardata);
4205 }
4206 }
4207 else
4208 #endif /* SUPPORT_UTF */
4209
4210 /* Without UTF support, character values are constrained by the bit length,
4211 and can only be > 256 for 16-bit and 32-bit libraries. */
4212
4213 #ifdef COMPILE_PCRE8
4214 {}
4215 #else
4216 if (start < end)
4217 {
4218 *uchardata++ = XCL_RANGE;
4219 *uchardata++ = start;
4220 *uchardata++ = end;
4221 }
4222 else if (start == end)
4223 {
4224 *uchardata++ = XCL_SINGLE;
4225 *uchardata++ = start;
4226 }
4227 #endif
4228
4229 *uchardptr = uchardata; /* Updata extra data pointer */
4230 }
4231 #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4232
4233 return n8; /* Number of 8-bit characters */
4234 }
4235
4236
4237
4238
4239 /*************************************************
4240 * Add a list of characters to a class *
4241 *************************************************/
4242
4243 /* This function is used for adding a list of case-equivalent characters to a
4244 class, and also for adding a list of horizontal or vertical whitespace. If the
4245 list is in order (which it should be), ranges of characters are detected and
4246 handled appropriately. This function is mutually recursive with the function
4247 above.
4248
4249 Arguments:
4250 classbits the bit map for characters < 256
4251 uchardptr points to the pointer for extra data
4252 options the options word
4253 cd contains pointers to tables etc.
4254 p points to row of 32-bit values, terminated by NOTACHAR
4255 except character to omit; this is used when adding lists of
4256 case-equivalent characters to avoid including the one we
4257 already know about
4258
4259 Returns: the number of < 256 characters added
4260 the pointer to extra data is updated
4261 */
4262
4263 static int
4264 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4265 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4266 {
4267 int n8 = 0;
4268 while (p[0] < NOTACHAR)
4269 {
4270 int n = 0;
4271 if (p[0] != except)
4272 {
4273 while(p[n+1] == p[0] + n + 1) n++;
4274 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4275 }
4276 p += n + 1;
4277 }
4278 return n8;
4279 }
4280
4281
4282
4283 /*************************************************
4284 * Add characters not in a list to a class *
4285 *************************************************/
4286
4287 /* This function is used for adding the complement of a list of horizontal or
4288 vertical whitespace to a class. The list must be in order.
4289
4290 Arguments:
4291 classbits the bit map for characters < 256
4292 uchardptr points to the pointer for extra data
4293 options the options word
4294 cd contains pointers to tables etc.
4295 p points to row of 32-bit values, terminated by NOTACHAR
4296
4297 Returns: the number of < 256 characters added
4298 the pointer to extra data is updated
4299 */
4300
4301 static int
4302 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4303 int options, compile_data *cd, const pcre_uint32 *p)
4304 {
4305 BOOL utf = (options & PCRE_UTF8) != 0;
4306 int n8 = 0;
4307 if (p[0] > 0)
4308 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4309 while (p[0] < NOTACHAR)
4310 {
4311 while (p[1] == p[0] + 1) p++;
4312 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4313 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4314 p++;
4315 }
4316 return n8;
4317 }
4318
4319
4320
4321 /*************************************************
4322 * Compile one branch *
4323 *************************************************/
4324
4325 /* Scan the pattern, compiling it into the a vector. If the options are
4326 changed during the branch, the pointer is used to change the external options
4327 bits. This function is used during the pre-compile phase when we are trying
4328 to find out the amount of memory needed, as well as during the real compile
4329 phase. The value of lengthptr distinguishes the two phases.
4330
4331 Arguments:
4332 optionsptr pointer to the option bits
4333 codeptr points to the pointer to the current code point
4334 ptrptr points to the current pattern pointer
4335 errorcodeptr points to error code variable
4336 firstcharptr place to put the first required character
4337 firstcharflagsptr place to put the first character flags, or a negative number
4338 reqcharptr place to put the last required character
4339 reqcharflagsptr place to put the last required character flags, or a negative number
4340 bcptr points to current branch chain
4341 cond_depth conditional nesting depth
4342 cd contains pointers to tables etc.
4343 lengthptr NULL during the real compile phase
4344 points to length accumulator during pre-compile phase
4345
4346 Returns: TRUE on success
4347 FALSE, with *errorcodeptr set non-zero on error
4348 */
4349
4350 static BOOL
4351 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4352 const pcre_uchar **ptrptr, int *errorcodeptr,
4353 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4354 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4355 branch_chain *bcptr, int cond_depth,
4356 compile_data *cd, int *lengthptr)
4357 {
4358 int repeat_type, op_type;
4359 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4360 int bravalue = 0;
4361 int greedy_default, greedy_non_default;
4362 pcre_uint32 firstchar, reqchar;
4363 pcre_int32 firstcharflags, reqcharflags;
4364 pcre_uint32 zeroreqchar, zerofirstchar;
4365 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4366 pcre_int32 req_caseopt, reqvary, tempreqvary;
4367 int options = *optionsptr; /* May change dynamically */
4368 int after_manual_callout = 0;
4369 int length_prevgroup = 0;
4370 register pcre_uint32 c;
4371 int escape;
4372 register pcre_uchar *code = *codeptr;
4373 pcre_uchar *last_code = code;
4374 pcre_uchar *orig_code = code;
4375 pcre_uchar *tempcode;
4376 BOOL inescq = FALSE;
4377 BOOL groupsetfirstchar = FALSE;
4378 const pcre_uchar *ptr = *ptrptr;
4379 const pcre_uchar *tempptr;
4380 const pcre_uchar *nestptr = NULL;
4381 pcre_uchar *previous = NULL;
4382 pcre_uchar *previous_callout = NULL;
4383 pcre_uchar *save_hwm = NULL;
4384 pcre_uint8 classbits[32];
4385
4386 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4387 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4388 dynamically as we process the pattern. */
4389
4390 #ifdef SUPPORT_UTF
4391 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4392 BOOL utf = (options & PCRE_UTF8) != 0;
4393 #ifndef COMPILE_PCRE32
4394 pcre_uchar utf_chars[6];
4395 #endif
4396 #else
4397 BOOL utf = FALSE;
4398 #endif
4399
4400 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4401 class_uchardata always so that it can be passed to add_to_class() always,
4402 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4403 alternative calls for the different cases. */
4404
4405 pcre_uchar *class_uchardata;
4406 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4407 BOOL xclass;
4408 pcre_uchar *class_uchardata_base;
4409 #endif
4410
4411 #ifdef PCRE_DEBUG
4412 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4413 #endif
4414
4415 /* Set up the default and non-default settings for greediness */
4416
4417 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4418 greedy_non_default = greedy_default ^ 1;
4419
4420 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4421 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4422 matches a non-fixed char first char; reqchar just remains unset if we never
4423 find one.
4424
4425 When we hit a repeat whose minimum is zero, we may have to adjust these values
4426 to take the zero repeat into account. This is implemented by setting them to
4427 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4428 item types that can be repeated set these backoff variables appropriately. */
4429
4430 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4431 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4432
4433 /* The variable req_caseopt contains either the REQ_CASELESS value
4434 or zero, according to the current setting of the caseless flag. The
4435 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4436 firstchar or reqchar variables to record the case status of the
4437 value. This is used only for ASCII characters. */
4438
4439 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4440
4441 /* Switch on next character until the end of the branch */
4442
4443 for (;; ptr++)
4444 {
4445 BOOL negate_class;
4446 BOOL should_flip_negation;
4447 BOOL possessive_quantifier;
4448 BOOL is_quantifier;
4449 BOOL is_recurse;
4450 BOOL reset_bracount;
4451 int class_has_8bitchar;
4452 int class_one_char;
4453 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4454 BOOL xclass_has_prop;
4455 #endif
4456 int newoptions;
4457 int recno;
4458 int refsign;
4459 int skipbytes;
4460 pcre_uint32 subreqchar, subfirstchar;
4461 pcre_int32 subreqcharflags, subfirstcharflags;
4462 int terminator;
4463 unsigned int mclength;
4464 unsigned int tempbracount;
4465 pcre_uint32 ec;
4466 pcre_uchar mcbuffer[8];
4467
4468 /* Get next character in the pattern */
4469
4470 c = *ptr;
4471
4472 /* If we are at the end of a nested substitution, revert to the outer level
4473 string. Nesting only happens one level deep. */
4474
4475 if (c == CHAR_NULL && nestptr != NULL)
4476 {
4477 ptr = nestptr;
4478 nestptr = NULL;
4479 c = *ptr;
4480 }
4481
4482 /* If we are in the pre-compile phase, accumulate the length used for the
4483 previous cycle of this loop. */
4484
4485 if (lengthptr != NULL)
4486 {
4487 #ifdef PCRE_DEBUG
4488 if (code > cd->hwm) cd->hwm = code; /* High water info */
4489 #endif
4490 if (code > cd->start_workspace + cd->workspace_size -
4491 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4492 {
4493 *errorcodeptr = ERR52;
4494 goto FAILED;
4495 }
4496
4497 /* There is at least one situation where code goes backwards: this is the
4498 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4499 the class is simply eliminated. However, it is created first, so we have to
4500 allow memory for it. Therefore, don't ever reduce the length at this point.
4501 */
4502
4503 if (code < last_code) code = last_code;
4504
4505 /* Paranoid check for integer overflow */
4506
4507 if (OFLOW_MAX - *lengthptr < code - last_code)
4508 {
4509 *errorcodeptr = ERR20;
4510 goto FAILED;
4511 }
4512
4513 *lengthptr += (int)(code - last_code);
4514 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4515 (int)(code - last_code), c, c));
4516
4517 /* If "previous" is set and it is not at the start of the work space, move
4518 it back to there, in order to avoid filling up the work space. Otherwise,
4519 if "previous" is NULL, reset the current code pointer to the start. */
4520
4521 if (previous != NULL)
4522 {
4523 if (previous > orig_code)
4524 {
4525 memmove(orig_code, previous, IN_UCHARS(code - previous));
4526 code -= previous - orig_code;
4527 previous = orig_code;
4528 }
4529 }
4530 else code = orig_code;
4531
4532 /* Remember where this code item starts so we can pick up the length
4533 next time round. */
4534
4535 last_code = code;
4536 }
4537
4538 /* In the real compile phase, just check the workspace used by the forward
4539 reference list. */
4540
4541 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4542 WORK_SIZE_SAFETY_MARGIN)
4543 {
4544 *errorcodeptr = ERR52;
4545 goto FAILED;
4546 }
4547
4548 /* If in \Q...\E, check for the end; if not, we have a literal */
4549
4550 if (inescq && c != CHAR_NULL)
4551 {
4552 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4553 {
4554 inescq = FALSE;
4555 ptr++;
4556 continue;
4557 }
4558 else
4559 {
4560 if (previous_callout != NULL)
4561 {
4562 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4563 complete_callout(previous_callout, ptr, cd);
4564 previous_callout = NULL;
4565 }
4566 if ((options & PCRE_AUTO_CALLOUT) != 0)
4567 {
4568 previous_callout = code;
4569 code = auto_callout(code, ptr, cd);
4570 }
4571 goto NORMAL_CHAR;
4572 }
4573 /* Control does not reach here. */
4574 }
4575
4576 /* In extended mode, skip white space and comments. We need a loop in order
4577 to check for more white space and more comments after a comment. */
4578
4579 if ((options & PCRE_EXTENDED) != 0)
4580 {
4581 for (;;)
4582 {
4583 while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4584 if (c != CHAR_NUMBER_SIGN) break;
4585 ptr++;
4586 while (*ptr != CHAR_NULL)
4587 {
4588 if (IS_NEWLINE(ptr)) /* For non-fixed-length newline cases, */
4589 { /* IS_NEWLINE sets cd->nllen. */
4590 ptr += cd->nllen;
4591 break;
4592 }
4593 ptr++;
4594 #ifdef SUPPORT_UTF
4595 if (utf) FORWARDCHAR(ptr);
4596 #endif
4597 }
4598 c = *ptr; /* Either NULL or the char after a newline */
4599 }
4600 }
4601
4602 /* See if the next thing is a quantifier. */
4603
4604 is_quantifier =
4605 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4606 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4607
4608 /* Fill in length of a previous callout, except when the next thing is a
4609 quantifier or when processing a property substitution string in UCP mode. */
4610
4611 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4612 after_manual_callout-- <= 0)
4613 {
4614 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4615 complete_callout(previous_callout, ptr, cd);
4616 previous_callout = NULL;
4617 }
4618
4619 /* Create auto callout, except for quantifiers, or while processing property
4620 strings that are substituted for \w etc in UCP mode. */
4621
4622 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4623 {
4624 previous_callout = code;
4625 code = auto_callout(code, ptr, cd);
4626 }
4627
4628 /* Process the next pattern item. */
4629
4630 switch(c)
4631 {
4632 /* ===================================================================*/
4633 case CHAR_NULL: /* The branch terminates at string end */
4634 case CHAR_VERTICAL_LINE: /* or | or ) */
4635 case CHAR_RIGHT_PARENTHESIS:
4636 *firstcharptr = firstchar;
4637 *firstcharflagsptr = firstcharflags;
4638 *reqcharptr = reqchar;
4639 *reqcharflagsptr = reqcharflags;
4640 *codeptr = code;
4641 *ptrptr = ptr;
4642 if (lengthptr != NULL)
4643 {
4644 if (OFLOW_MAX - *lengthptr < code - last_code)
4645 {
4646 *errorcodeptr = ERR20;
4647 goto FAILED;
4648 }
4649 *lengthptr += (int)(code - last_code); /* To include callout length */
4650 DPRINTF((">> end branch\n"));
4651 }
4652 return TRUE;
4653
4654
4655 /* ===================================================================*/
4656 /* Handle single-character metacharacters. In multiline mode, ^ disables
4657 the setting of any following char as a first character. */
4658
4659 case CHAR_CIRCUMFLEX_ACCENT:
4660 previous = NULL;
4661 if ((options & PCRE_MULTILINE) != 0)
4662 {
4663 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4664 *code++ = OP_CIRCM;
4665 }
4666 else *code++ = OP_CIRC;
4667 break;
4668
4669 case CHAR_DOLLAR_SIGN:
4670 previous = NULL;
4671 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4672 break;
4673
4674 /* There can never be a first char if '.' is first, whatever happens about
4675 repeats. The value of reqchar doesn't change either. */
4676
4677 case CHAR_DOT:
4678 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4679 zerofirstchar = firstchar;
4680 zerofirstcharflags = firstcharflags;
4681 zeroreqchar = reqchar;
4682 zeroreqcharflags = reqcharflags;
4683 previous = code;
4684 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4685 break;
4686
4687
4688 /* ===================================================================*/
4689 /* Character classes. If the included characters are all < 256, we build a
4690 32-byte bitmap of the permitted characters, except in the special case
4691 where there is only one such character. For negated classes, we build the
4692 map as usual, then invert it at the end. However, we use a different opcode
4693 so that data characters > 255 can be handled correctly.
4694
4695 If the class contains characters outside the 0-255 range, a different
4696 opcode is compiled. It may optionally have a bit map for characters < 256,
4697 but those above are are explicitly listed afterwards. A flag byte tells
4698 whether the bitmap is present, and whether this is a negated class or not.
4699
4700 In JavaScript compatibility mode, an isolated ']' causes an error. In
4701 default (Perl) mode, it is treated as a data character. */
4702
4703 case CHAR_RIGHT_SQUARE_BRACKET:
4704 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4705 {
4706 *errorcodeptr = ERR64;
4707 goto FAILED;
4708 }
4709 goto NORMAL_CHAR;
4710
4711 /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4712 used for "start of word" and "end of word". As these are otherwise illegal
4713 sequences, we don't break anything by recognizing them. They are replaced
4714 by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4715 erroneous and are handled by the normal code below. */
4716
4717 case CHAR_LEFT_SQUARE_BRACKET:
4718 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4719 {
4720 nestptr = ptr + 7;
4721 ptr = sub_start_of_word - 1;
4722 continue;
4723 }
4724
4725 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4726 {
4727 nestptr = ptr + 7;
4728 ptr = sub_end_of_word - 1;
4729 continue;
4730 }
4731
4732 /* Handle a real character class. */
4733
4734 previous = code;
4735
4736 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4737 they are encountered at the top level, so we'll do that too. */
4738
4739 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4740 ptr[1] == CHAR_EQUALS_SIGN) &&
4741 check_posix_syntax(ptr, &tempptr))
4742 {
4743 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4744 goto FAILED;
4745 }
4746
4747 /* If the first character is '^', set the negation flag and skip it. Also,
4748 if the first few characters (either before or after ^) are \Q\E or \E we
4749 skip them too. This makes for compatibility with Perl. */
4750
4751 negate_class = FALSE;
4752 for (;;)
4753 {
4754 c = *(++ptr);
4755 if (c == CHAR_BACKSLASH)
4756 {
4757 if (ptr[1] == CHAR_E)
4758 ptr++;
4759 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4760 ptr += 3;
4761 else
4762 break;
4763 }
4764 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4765 negate_class = TRUE;
4766 else break;
4767 }
4768
4769 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4770 an initial ']' is taken as a data character -- the code below handles
4771 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4772 [^] must match any character, so generate OP_ALLANY. */
4773
4774 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4775 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4776 {
4777 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4778 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4779 zerofirstchar = firstchar;
4780 zerofirstcharflags = firstcharflags;
4781 break;
4782 }
4783
4784 /* If a class contains a negative special such as \S, we need to flip the
4785 negation flag at the end, so that support for characters > 255 works
4786 correctly (they are all included in the class). */
4787
4788 should_flip_negation = FALSE;
4789
4790 /* Extended class (xclass) will be used when characters > 255
4791 might match. */
4792
4793 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4794 xclass = FALSE;
4795 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4796 class_uchardata_base = class_uchardata; /* Save the start */
4797 #endif
4798
4799 /* For optimization purposes, we track some properties of the class:
4800 class_has_8bitchar will be non-zero if the class contains at least one <
4801 256 character; class_one_char will be 1 if the class contains just one
4802 character; xclass_has_prop will be TRUE if unicode property checks
4803 are present in the class. */
4804
4805 class_has_8bitchar = 0;
4806 class_one_char = 0;
4807 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4808 xclass_has_prop = FALSE;
4809 #endif
4810
4811 /* Initialize the 32-char bit map to all zeros. We build the map in a
4812 temporary bit of memory, in case the class contains fewer than two
4813 8-bit characters because in that case the compiled code doesn't use the bit
4814 map. */
4815
4816 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4817
4818 /* Process characters until ] is reached. By writing this as a "do" it
4819 means that an initial ] is taken as a data character. At the start of the
4820 loop, c contains the first byte of the character. */
4821
4822 if (c != CHAR_NULL) do
4823 {
4824 const pcre_uchar *oldptr;
4825
4826 #ifdef SUPPORT_UTF
4827 if (utf && HAS_EXTRALEN(c))
4828 { /* Braces are required because the */
4829 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4830 }
4831 #endif
4832
4833 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4834 /* In the pre-compile phase, accumulate the length of any extra
4835 data and reset the pointer. This is so that very large classes that
4836 contain a zillion > 255 characters no longer overwrite the work space
4837 (which is on the stack). We have to remember that there was XCLASS data,
4838 however. */
4839
4840 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4841 {
4842 xclass = TRUE;
4843 *lengthptr += class_uchardata - class_uchardata_base;
4844 class_uchardata = class_uchardata_base;
4845 }
4846 #endif
4847
4848 /* Inside \Q...\E everything is literal except \E */
4849
4850 if (inescq)
4851 {
4852 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4853 {
4854 inescq = FALSE; /* Reset literal state */
4855 ptr++; /* Skip the 'E' */
4856 continue; /* Carry on with next */
4857 }
4858 goto CHECK_RANGE; /* Could be range if \E follows */
4859 }
4860
4861 /* Handle POSIX class names. Perl allows a negation extension of the
4862 form [:^name:]. A square bracket that doesn't match the syntax is
4863 treated as a literal. We also recognize the POSIX constructions
4864 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4865 5.6 and 5.8 do. */
4866
4867 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4868 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4869 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4870 {
4871 BOOL local_negate = FALSE;
4872 int posix_class, taboffset, tabopt;
4873 register const pcre_uint8 *cbits = cd->cbits;
4874 pcre_uint8 pbits[32];
4875
4876 if (ptr[1] != CHAR_COLON)
4877 {
4878 *errorcodeptr = ERR31;
4879 goto FAILED;
4880 }
4881
4882 ptr += 2;
4883 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4884 {
4885 local_negate = TRUE;
4886 should_flip_negation = TRUE; /* Note negative special */
4887 ptr++;
4888 }
4889
4890 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4891 if (posix_class < 0)
4892 {
4893 *errorcodeptr = ERR30;
4894 goto FAILED;
4895 }
4896
4897 /* If matching is caseless, upper and lower are converted to
4898 alpha. This relies on the fact that the class table starts with
4899 alpha, lower, upper as the first 3 entries. */
4900
4901 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4902 posix_class = 0;
4903
4904 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4905 different escape sequences that use Unicode properties \p or \P. Others
4906 that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4907 directly. */
4908
4909 #ifdef SUPPORT_UCP
4910 if ((options & PCRE_UCP) != 0)
4911 {
4912 unsigned int ptype = 0;
4913 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4914
4915 /* The posix_substitutes table specifies which POSIX classes can be
4916 converted to \p or \P items. */
4917
4918 if (posix_substitutes[pc] != NULL)
4919 {
4920 nestptr = tempptr + 1;
4921 ptr = posix_substitutes[pc] - 1;
4922 continue;
4923 }
4924
4925 /* There are three other classes that generate special property calls
4926 that are recognized only in an XCLASS. */
4927
4928 else switch(posix_class)
4929 {
4930 case PC_GRAPH:
4931 ptype = PT_PXGRAPH;
4932 /* Fall through */
4933 case PC_PRINT:
4934 if (ptype == 0) ptype = PT_PXPRINT;
4935 /* Fall through */
4936 case PC_PUNCT:
4937 if (ptype == 0) ptype = PT_PXPUNCT;
4938 *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4939 *class_uchardata++ = ptype;
4940 *class_uchardata++ = 0;
4941 xclass_has_prop = TRUE;
4942 ptr = tempptr + 1;
4943 continue;
4944
4945 /* For all other POSIX classes, no special action is taken in UCP
4946 mode. Fall through to the non_UCP case. */
4947
4948 default:
4949 break;
4950 }
4951 }
4952 #endif
4953 /* In the non-UCP case, or when UCP makes no difference, we build the
4954 bit map for the POSIX class in a chunk of local store because we may be
4955 adding and subtracting from it, and we don't want to subtract bits that
4956 may be in the main map already. At the end we or the result into the
4957 bit map that is being built. */
4958
4959 posix_class *= 3;
4960
4961 /* Copy in the first table (always present) */
4962
4963 memcpy(pbits, cbits + posix_class_maps[posix_class],
4964 32 * sizeof(pcre_uint8));
4965
4966 /* If there is a second table, add or remove it as required. */
4967
4968 taboffset = posix_class_maps[posix_class + 1];
4969 tabopt = posix_class_maps[posix_class + 2];
4970
4971 if (taboffset >= 0)
4972 {
4973 if (tabopt >= 0)
4974 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
4975 else
4976 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
4977 }
4978
4979 /* Now see if we need to remove any special characters. An option
4980 value of 1 removes vertical space and 2 removes underscore. */
4981
4982 if (tabopt < 0) tabopt = -tabopt;
4983 if (tabopt == 1) pbits[1] &= ~0x3c;
4984 else if (tabopt == 2) pbits[11] &= 0x7f;
4985
4986 /* Add the POSIX table or its complement into the main table that is
4987 being built and we are done. */
4988
4989 if (local_negate)
4990 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
4991 else
4992 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
4993
4994 ptr = tempptr + 1;
4995 /* Every class contains at least one < 256 character. */
4996 class_has_8bitchar = 1;
4997 /* Every class contains at least two characters. */
4998 class_one_char = 2;
4999 continue; /* End of POSIX syntax handling */
5000 }
5001
5002 /* Backslash may introduce a single character, or it may introduce one
5003 of the specials, which just set a flag. The sequence \b is a special
5004 case. Inside a class (and only there) it is treated as backspace. We
5005 assume that other escapes have more than one character in them, so
5006 speculatively set both class_has_8bitchar and class_one_char bigger
5007 than one. Unrecognized escapes fall through and are either treated
5008 as literal characters (by default), or are faulted if
5009 PCRE_EXTRA is set. */
5010
5011 if (c == CHAR_BACKSLASH)
5012 {
5013 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
5014 TRUE);
5015 if (*errorcodeptr != 0) goto FAILED;
5016 if (escape == 0) c = ec;
5017 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
5018 else if (escape == ESC_N) /* \N is not supported in a class */
5019 {
5020 *errorcodeptr = ERR71;
5021 goto FAILED;
5022 }
5023 else if (escape == ESC_Q) /* Handle start of quoted string */
5024 {
5025 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5026 {
5027 ptr += 2; /* avoid empty string */
5028 }
5029 else inescq = TRUE;
5030 continue;
5031 }
5032 else if (escape == ESC_E) continue; /* Ignore orphan \E */
5033
5034 else
5035 {
5036 register const pcre_uint8 *cbits = cd->cbits;
5037 /* Every class contains at least two < 256 characters. */
5038 class_has_8bitchar++;
5039 /* Every class contains at least two characters. */
5040 class_one_char += 2;
5041
5042 switch (escape)
5043 {
5044 #ifdef SUPPORT_UCP
5045 case ESC_du: /* These are the values given for \d etc */
5046 case ESC_DU: /* when PCRE_UCP is set. We replace the */
5047 case ESC_wu: /* escape sequence with an appropriate \p */
5048 case ESC_WU: /* or \P to test Unicode properties instead */
5049 case ESC_su: /* of the default ASCII testing. */
5050 case ESC_SU:
5051 nestptr = ptr;
5052 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
5053 class_has_8bitchar--; /* Undo! */
5054 continue;
5055 #endif
5056 case ESC_d:
5057 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
5058 continue;
5059
5060 case ESC_D:
5061 should_flip_negation = TRUE;
5062 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
5063 continue;
5064
5065 case ESC_w:
5066 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
5067 continue;
5068
5069 case ESC_W:
5070 should_flip_negation = TRUE;
5071 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5072 continue;
5073
5074 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5075 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5076 previously set by something earlier in the character class.
5077 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5078 we could just adjust the appropriate bit. From PCRE 8.34 we no
5079 longer treat \s and \S specially. */
5080
5081 case ESC_s:
5082 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
5083 continue;
5084
5085 case ESC_S:
5086 should_flip_negation = TRUE;
5087 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
5088 continue;
5089
5090 /* The rest apply in both UCP and non-UCP cases. */
5091
5092 case ESC_h:
5093 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5094 PRIV(hspace_list), NOTACHAR);
5095 continue;
5096
5097 case ESC_H:
5098 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5099 cd, PRIV(hspace_list));
5100 continue;
5101
5102 case ESC_v:
5103 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5104 PRIV(vspace_list), NOTACHAR);
5105 continue;
5106
5107 case ESC_V:
5108 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5109 cd, PRIV(vspace_list));
5110 continue;
5111
5112 #ifdef SUPPORT_UCP
5113 case ESC_p:
5114 case ESC_P:
5115 {
5116 BOOL negated;
5117 unsigned int ptype = 0, pdata = 0;
5118 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5119 goto FAILED;
5120 *class_uchardata++ = ((escape == ESC_p) != negated)?
5121 XCL_PROP : XCL_NOTPROP;
5122 *class_uchardata++ = ptype;
5123 *class_uchardata++ = pdata;
5124 xclass_has_prop = TRUE;
5125 class_has_8bitchar--; /* Undo! */
5126 continue;
5127 }
5128 #endif
5129 /* Unrecognized escapes are faulted if PCRE is running in its
5130 strict mode. By default, for compatibility with Perl, they are
5131 treated as literals. */
5132
5133 default:
5134 if ((options & PCRE_EXTRA) != 0)
5135 {
5136 *errorcodeptr = ERR7;
5137 goto FAILED;
5138 }
5139 class_has_8bitchar--; /* Undo the speculative increase. */
5140 class_one_char -= 2; /* Undo the speculative increase. */
5141 c = *ptr; /* Get the final character and fall through */
5142 break;
5143 }
5144 }
5145
5146 /* Fall through if the escape just defined a single character (c >= 0).
5147 This may be greater than 256. */
5148
5149 escape = 0;
5150
5151 } /* End of backslash handling */
5152
5153 /* A character may be followed by '-' to form a range. However, Perl does
5154 not permit ']' to be the end of the range. A '-' character at the end is
5155 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5156 code for handling \Q and \E is messy. */
5157
5158 CHECK_RANGE:
5159 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5160 {
5161 inescq = FALSE;
5162 ptr += 2;
5163 }
5164 oldptr = ptr;
5165
5166 /* Remember if \r or \n were explicitly used */
5167
5168 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5169
5170 /* Check for range */
5171
5172 if (!inescq && ptr[1] == CHAR_MINUS)
5173 {
5174 pcre_uint32 d;
5175 ptr += 2;
5176 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5177
5178 /* If we hit \Q (not followed by \E) at this point, go into escaped
5179 mode. */
5180
5181 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5182 {
5183 ptr += 2;
5184 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5185 { ptr += 2; continue; }
5186 inescq = TRUE;
5187 break;
5188 }
5189
5190 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5191 back the pointer and jump to handle the character that preceded it. */
5192
5193 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5194 {
5195 ptr = oldptr;
5196 goto CLASS_SINGLE_CHARACTER;
5197 }
5198
5199 /* Otherwise, we have a potential range; pick up the next character */
5200
5201 #ifdef SUPPORT_UTF
5202 if (utf)
5203 { /* Braces are required because the */
5204 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5205 }
5206 else
5207 #endif
5208 d = *ptr; /* Not UTF-8 mode */
5209
5210 /* The second part of a range can be a single-character escape
5211 sequence, but not any of the other escapes. Perl treats a hyphen as a
5212 literal in such circumstances. However, in Perl's warning mode, a
5213 warning is given, so PCRE now faults it as it is almost certainly a
5214 mistake on the user's part. */
5215
5216 if (!inescq)
5217 {
5218 if (d == CHAR_BACKSLASH)
5219 {
5220 int descape;
5221 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5222 if (*errorcodeptr != 0) goto FAILED;
5223
5224 /* 0 means a character was put into d; \b is backspace; any other
5225 special causes an error. */
5226
5227 if (descape != 0)
5228 {
5229 if (descape == ESC_b) d = CHAR_BS; else
5230 {
5231 *errorcodeptr = ERR83;
5232 goto FAILED;
5233 }
5234 }
5235 }
5236
5237 /* A hyphen followed by a POSIX class is treated in the same way. */
5238
5239 else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5240 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5241 ptr[1] == CHAR_EQUALS_SIGN) &&
5242 check_posix_syntax(ptr, &tempptr))
5243 {
5244 *errorcodeptr = ERR83;
5245 goto FAILED;
5246 }
5247 }
5248
5249 /* Check that the two values are in the correct order. Optimize
5250 one-character ranges. */
5251
5252 if (d < c)
5253 {
5254 *errorcodeptr = ERR8;
5255 goto FAILED;
5256 }
5257 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5258
5259 /* We have found a character range, so single character optimizations
5260 cannot be done anymore. Any value greater than 1 indicates that there
5261 is more than one character. */
5262
5263 class_one_char = 2;
5264
5265 /* Remember an explicit \r or \n, and add the range to the class. */
5266
5267 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5268
5269 class_has_8bitchar +=
5270 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5271
5272 continue; /* Go get the next char in the class */
5273 }
5274
5275 /* Handle a single character - we can get here for a normal non-escape
5276 char, or after \ that introduces a single character or for an apparent
5277 range that isn't. Only the value 1 matters for class_one_char, so don't
5278 increase it if it is already 2 or more ... just in case there's a class
5279 with a zillion characters in it. */
5280
5281 CLASS_SINGLE_CHARACTER:
5282 if (class_one_char < 2) class_one_char++;
5283
5284 /* If class_one_char is 1, we have the first single character in the
5285 class, and there have been no prior ranges, or XCLASS items generated by
5286 escapes. If this is the final character in the class, we can optimize by
5287 turning the item into a 1-character OP_CHAR[I] if it's positive, or
5288 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5289 to be set. Otherwise, there can be no first char if this item is first,
5290 whatever repeat count may follow. In the case of reqchar, save the
5291 previous value for reinstating. */
5292
5293 if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5294 {
5295 ptr++;
5296 zeroreqchar = reqchar;
5297 zeroreqcharflags = reqcharflags;
5298
5299 if (negate_class)
5300 {
5301 #ifdef SUPPORT_UCP
5302 int d;
5303 #endif
5304 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5305 zerofirstchar = firstchar;
5306 zerofirstcharflags = firstcharflags;
5307
5308 /* For caseless UTF-8 mode when UCP support is available, check
5309 whether this character has more than one other case. If so, generate
5310 a special OP_NOTPROP item instead of OP_NOTI. */
5311
5312 #ifdef SUPPORT_UCP
5313 if (utf && (options & PCRE_CASELESS) != 0 &&
5314 (d = UCD_CASESET(c)) != 0)
5315 {
5316 *code++ = OP_NOTPROP;
5317 *code++ = PT_CLIST;
5318 *code++ = d;
5319 }
5320 else
5321 #endif
5322 /* Char has only one other case, or UCP not available */
5323
5324 {
5325 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5326 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5327 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5328 code += PRIV(ord2utf)(c, code);
5329 else
5330 #endif
5331 *code++ = c;
5332 }
5333
5334 /* We are finished with this character class */
5335
5336 goto END_CLASS;
5337 }
5338
5339 /* For a single, positive character, get the value into mcbuffer, and
5340 then we can handle this with the normal one-character code. */
5341
5342 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5343 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5344 mclength = PRIV(ord2utf)(c, mcbuffer);
5345 else
5346 #endif
5347 {
5348 mcbuffer[0] = c;
5349 mclength = 1;
5350 }
5351 goto ONE_CHAR;
5352 } /* End of 1-char optimization */
5353
5354 /* There is more than one character in the class, or an XCLASS item
5355 has been generated. Add this character to the class. */
5356
5357 class_has_8bitchar +=
5358 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5359 }
5360
5361 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5362 If we are at the end of an internal nested string, revert to the outer
5363 string. */
5364
5365 while (((c = *(++ptr)) != CHAR_NULL ||
5366 (nestptr != NULL &&
5367 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5368 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5369
5370 /* Check for missing terminating ']' */
5371
5372 if (c == CHAR_NULL)
5373 {
5374 *errorcodeptr = ERR6;
5375 goto FAILED;
5376 }
5377
5378 /* We will need an XCLASS if data has been placed in class_uchardata. In
5379 the second phase this is a sufficient test. However, in the pre-compile
5380 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5381 only if the very last character in the class needs XCLASS will it contain
5382 anything at this point. For this reason, xclass gets set TRUE above when
5383 uchar_classdata is emptied, and that's why this code is the way it is here
5384 instead of just doing a test on class_uchardata below. */
5385
5386 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5387 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5388 #endif
5389
5390 /* If this is the first thing in the branch, there can be no first char
5391 setting, whatever the repeat count. Any reqchar setting must remain
5392 unchanged after any kind of repeat. */
5393
5394 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5395 zerofirstchar = firstchar;
5396 zerofirstcharflags = firstcharflags;
5397 zeroreqchar = reqchar;
5398 zeroreqcharflags = reqcharflags;
5399
5400 /* If there are characters with values > 255, we have to compile an
5401 extended class, with its own opcode, unless there was a negated special
5402 such as \S in the class, and PCRE_UCP is not set, because in that case all
5403 characters > 255 are in the class, so any that were explicitly given as
5404 well can be ignored. If (when there are explicit characters > 255 that must
5405 be listed) there are no characters < 256, we can omit the bitmap in the
5406 actual compiled code. */
5407
5408 #ifdef SUPPORT_UTF
5409 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5410 #elif !defined COMPILE_PCRE8
5411 if (xclass && !should_flip_negation)
5412 #endif
5413 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5414 {
5415 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5416 *code++ = OP_XCLASS;
5417 code += LINK_SIZE;
5418 *code = negate_class? XCL_NOT:0;
5419 if (xclass_has_prop) *code |= XCL_HASPROP;
5420
5421 /* If the map is required, move up the extra data to make room for it;
5422 otherwise just move the code pointer to the end of the extra data. */
5423
5424 if (class_has_8bitchar > 0)
5425 {
5426 *code++ |= XCL_MAP;
5427 memmove(code + (32 / sizeof(pcre_uchar)), code,
5428 IN_UCHARS(class_uchardata - code));
5429 if (negate_class && !xclass_has_prop)
5430 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5431 memcpy(code, classbits, 32);
5432 code = class_uchardata + (32 / sizeof(pcre_uchar));
5433 }
5434 else code = class_uchardata;
5435
5436 /* Now fill in the complete length of the item */
5437
5438 PUT(previous, 1, (int)(code - previous));
5439 break; /* End of class handling */
5440 }
5441 #endif
5442
5443 /* If there are no characters > 255, or they are all to be included or
5444 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5445 whole class was negated and whether there were negative specials such as \S
5446 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5447 negating it if necessary. */
5448
5449 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5450 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5451 {
5452 if (negate_class)
5453 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5454 memcpy(code, classbits, 32);
5455 }
5456 code += 32 / sizeof(pcre_uchar);
5457
5458 END_CLASS:
5459 break;
5460
5461
5462 /* ===================================================================*/
5463 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5464 has been tested above. */
5465
5466 case CHAR_LEFT_CURLY_BRACKET:
5467 if (!is_quantifier) goto NORMAL_CHAR;
5468 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5469 if (*errorcodeptr != 0) goto FAILED;
5470 goto REPEAT;
5471
5472 case CHAR_ASTERISK:
5473 repeat_min = 0;
5474 repeat_max = -1;
5475 goto REPEAT;
5476
5477 case CHAR_PLUS:
5478 repeat_min = 1;
5479 repeat_max = -1;
5480 goto REPEAT;
5481
5482 case CHAR_QUESTION_MARK:
5483 repeat_min = 0;
5484 repeat_max = 1;
5485
5486 REPEAT:
5487 if (previous == NULL)
5488 {
5489 *errorcodeptr = ERR9;
5490 goto FAILED;
5491 }
5492
5493 if (repeat_min == 0)
5494 {
5495 firstchar = zerofirstchar; /* Adjust for zero repeat */
5496 firstcharflags = zerofirstcharflags;
5497 reqchar = zeroreqchar; /* Ditto */
5498 reqcharflags = zeroreqcharflags;
5499 }
5500
5501 /* Remember whether this is a variable length repeat */
5502
5503 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5504
5505 op_type = 0; /* Default single-char op codes */
5506 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5507
5508 /* Save start of previous item, in case we have to move it up in order to
5509 insert something before it. */
5510
5511 tempcode = previous;
5512
5513 /* Before checking for a possessive quantifier, we must skip over
5514 whitespace and comments in extended mode because Perl allows white space at
5515 this point. */
5516
5517 if ((options & PCRE_EXTENDED) != 0)
5518 {
5519 const pcre_uchar *p = ptr + 1;
5520 for (;;)
5521 {
5522 while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5523 if (*p != CHAR_NUMBER_SIGN) break;
5524 p++;
5525 while (*p != CHAR_NULL)
5526 {
5527 if (IS_NEWLINE(p)) /* For non-fixed-length newline cases, */
5528 { /* IS_NEWLINE sets cd->nllen. */
5529 p += cd->nllen;
5530 break;
5531 }
5532 p++;
5533 #ifdef SUPPORT_UTF
5534 if (utf) FORWARDCHAR(p);
5535 #endif
5536 } /* Loop for comment characters */
5537 } /* Loop for multiple comments */
5538 ptr = p - 1; /* Character before the next significant one. */
5539 }
5540
5541 /* If the next character is '+', we have a possessive quantifier. This
5542 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5543 If the next character is '?' this is a minimizing repeat, by default,
5544 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5545 repeat type to the non-default. */
5546
5547 if (ptr[1] == CHAR_PLUS)
5548 {
5549 repeat_type = 0; /* Force greedy */
5550 possessive_quantifier = TRUE;
5551 ptr++;
5552 }
5553 else if (ptr[1] == CHAR_QUESTION_MARK)
5554 {
5555 repeat_type = greedy_non_default;
5556 ptr++;
5557 }
5558 else repeat_type = greedy_default;
5559
5560 /* If previous was a recursion call, wrap it in atomic brackets so that
5561 previous becomes the atomic group. All recursions were so wrapped in the
5562 past, but it no longer happens for non-repeated recursions. In fact, the
5563 repeated ones could be re-implemented independently so as not to need this,
5564 but for the moment we rely on the code for repeating groups. */
5565
5566 if (*previous == OP_RECURSE)
5567 {
5568 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5569 *previous = OP_ONCE;
5570 PUT(previous, 1, 2 + 2*LINK_SIZE);
5571 previous[2 + 2*LINK_SIZE] = OP_KET;
5572 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5573 code += 2 + 2 * LINK_SIZE;
5574 length_prevgroup = 3 + 3*LINK_SIZE;
5575
5576 /* When actually compiling, we need to check whether this was a forward
5577 reference, and if so, adjust the offset. */
5578
5579 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5580 {
5581 int offset = GET(cd->hwm, -LINK_SIZE);
5582 if (offset == previous + 1 - cd->start_code)
5583 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5584 }
5585 }
5586
5587 /* Now handle repetition for the different types of item. */
5588
5589 /* If previous was a character or negated character match, abolish the item
5590 and generate a repeat item instead. If a char item has a minimum of more
5591 than one, ensure that it is set in reqchar - it might not be if a sequence
5592 such as x{3} is the first thing in a branch because the x will have gone
5593 into firstchar instead. */
5594
5595 if (*previous == OP_CHAR || *previous == OP_CHARI
5596 || *previous == OP_NOT || *previous == OP_NOTI)
5597 {
5598 switch (*previous)
5599 {
5600 default: /* Make compiler happy. */
5601 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5602 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5603 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5604 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5605 }
5606
5607 /* Deal with UTF characters that take up more than one character. It's
5608 easier to write this out separately than try to macrify it. Use c to
5609 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5610 it's a length rather than a small character. */
5611
5612 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5613 if (utf && NOT_FIRSTCHAR(code[-1]))
5614 {
5615 pcre_uchar *lastchar = code - 1;
5616 BACKCHAR(lastchar);
5617 c = (int)(code - lastchar); /* Length of UTF-8 character */
5618 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5619 c |= UTF_LENGTH; /* Flag c as a length */
5620 }
5621 else
5622 #endif /* SUPPORT_UTF */
5623
5624 /* Handle the case of a single charater - either with no UTF support, or
5625 with UTF disabled, or for a single character UTF character. */
5626 {
5627 c = code[-1];
5628 if (*previous <= OP_CHARI && repeat_min > 1)
5629 {
5630 reqchar = c;
5631 reqcharflags = req_caseopt | cd->req_varyopt;
5632 }
5633 }
5634
5635 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5636 }
5637
5638 /* If previous was a character type match (\d or similar), abolish it and
5639 create a suitable repeat item. The code is shared with single-character
5640 repeats by setting op_type to add a suitable offset into repeat_type. Note
5641 the the Unicode property types will be present only when SUPPORT_UCP is
5642 defined, but we don't wrap the little bits of code here because it just
5643 makes it horribly messy. */
5644
5645 else if (*previous < OP_EODN)
5646 {
5647 pcre_uchar *oldcode;
5648 int prop_type, prop_value;
5649 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5650 c = *previous;
5651
5652 OUTPUT_SINGLE_REPEAT:
5653 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5654 {
5655 prop_type = previous[1];
5656 prop_value = previous[2];
5657 }
5658 else prop_type = prop_value = -1;
5659
5660 oldcode = code;
5661 code = previous; /* Usually overwrite previous item */
5662
5663 /* If the maximum is zero then the minimum must also be zero; Perl allows
5664 this case, so we do too - by simply omitting the item altogether. */
5665
5666 if (repeat_max == 0) goto END_REPEAT;
5667
5668 /* Combine the op_type with the repeat_type */
5669
5670 repeat_type += op_type;
5671
5672 /* A minimum of zero is handled either as the special case * or ?, or as
5673 an UPTO, with the maximum given. */
5674
5675 if (repeat_min == 0)
5676 {
5677 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5678 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5679 else
5680 {
5681 *code++ = OP_UPTO + repeat_type;
5682 PUT2INC(code, 0, repeat_max);
5683 }
5684 }
5685
5686 /* A repeat minimum of 1 is optimized into some special cases. If the
5687 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5688 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5689 one less than the maximum. */
5690
5691 else if (repeat_min == 1)
5692 {
5693 if (repeat_max == -1)
5694 *code++ = OP_PLUS + repeat_type;
5695 else
5696 {
5697 code = oldcode; /* leave previous item in place */
5698 if (repeat_max == 1) goto END_REPEAT;
5699 *code++ = OP_UPTO + repeat_type;
5700 PUT2INC(code, 0, repeat_max - 1);
5701 }
5702 }
5703
5704 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5705 handled as an EXACT followed by an UPTO. */
5706
5707 else
5708 {
5709 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5710 PUT2INC(code, 0, repeat_min);
5711
5712 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5713 we have to insert the character for the previous code. For a repeated
5714 Unicode property match, there are two extra bytes that define the
5715 required property. In UTF-8 mode, long characters have their length in
5716 c, with the UTF_LENGTH bit as a flag. */
5717
5718 if (repeat_max < 0)
5719 {
5720 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5721 if (utf && (c & UTF_LENGTH) != 0)
5722 {
5723 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5724 code += c & 7;
5725 }
5726 else
5727 #endif
5728 {
5729 *code++ = c;
5730 if (prop_type >= 0)
5731 {
5732 *code++ = prop_type;
5733 *code++ = prop_value;
5734 }
5735 }
5736 *code++ = OP_STAR + repeat_type;
5737 }
5738
5739 /* Else insert an UPTO if the max is greater than the min, again
5740 preceded by the character, for the previously inserted code. If the
5741 UPTO is just for 1 instance, we can use QUERY instead. */
5742
5743 else if (repeat_max != repeat_min)
5744 {
5745 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5746 if (utf && (c & UTF_LENGTH) != 0)
5747 {
5748 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5749 code += c & 7;
5750 }
5751 else
5752 #endif
5753 *code++ = c;
5754 if (prop_type >= 0)
5755 {
5756 *code++ = prop_type;
5757 *code++ = prop_value;
5758 }
5759 repeat_max -= repeat_min;
5760
5761 if (repeat_max == 1)
5762 {
5763 *code++ = OP_QUERY + repeat_type;
5764 }
5765 else
5766 {
5767 *code++ = OP_UPTO + repeat_type;
5768 PUT2INC(code, 0, repeat_max);
5769 }
5770 }
5771 }
5772
5773 /* The character or character type itself comes last in all cases. */
5774
5775 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5776 if (utf && (c & UTF_LENGTH) != 0)
5777 {
5778 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5779 code += c & 7;
5780 }
5781 else
5782 #endif
5783 *code++ = c;
5784
5785 /* For a repeated Unicode property match, there are two extra bytes that
5786 define the required property. */
5787
5788 #ifdef SUPPORT_UCP
5789 if (prop_type >= 0)
5790 {
5791 *code++ = prop_type;
5792 *code++ = prop_value;
5793 }
5794 #endif
5795 }
5796
5797 /* If previous was a character class or a back reference, we put the repeat
5798 stuff after it, but just skip the item if the repeat was {0,0}. */
5799
5800 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5801 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5802 *previous == OP_XCLASS ||
5803 #endif
5804 *previous == OP_REF || *previous == OP_REFI ||
5805 *previous == OP_DNREF || *previous == OP_DNREFI)
5806 {
5807 if (repeat_max == 0)
5808 {
5809 code = previous;
5810 goto END_REPEAT;
5811 }
5812
5813 if (repeat_min == 0 && repeat_max == -1)
5814 *code++ = OP_CRSTAR + repeat_type;
5815 else if (repeat_min == 1 && repeat_max == -1)
5816 *code++ = OP_CRPLUS + repeat_type;
5817 else if (repeat_min == 0 && repeat_max == 1)
5818 *code++ = OP_CRQUERY + repeat_type;
5819 else
5820 {
5821 *code++ = OP_CRRANGE + repeat_type;
5822 PUT2INC(code, 0, repeat_min);
5823 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5824 PUT2INC(code, 0, repeat_max);
5825 }
5826 }
5827
5828 /* If previous was a bracket group, we may have to replicate it in certain
5829 cases. Note that at this point we can encounter only the "basic" bracket
5830 opcodes such as BRA and CBRA, as this is the place where they get converted
5831 into the more special varieties such as BRAPOS and SBRA. A test for >=
5832 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5833 ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5834 Originally, PCRE did not allow repetition of assertions, but now it does,
5835 for Perl compatibility. */
5836
5837 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5838 {
5839 register int i;
5840 int len = (int)(code - previous);
5841 pcre_uchar *bralink = NULL;
5842 pcre_uchar *brazeroptr = NULL;
5843
5844 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5845 we just ignore the repeat. */
5846
5847 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5848 goto END_REPEAT;
5849
5850 /* There is no sense in actually repeating assertions. The only potential
5851 use of repetition is in cases when the assertion is optional. Therefore,
5852 if the minimum is greater than zero, just ignore the repeat. If the
5853 maximum is not zero or one, set it to 1. */
5854
5855 if (*previous < OP_ONCE) /* Assertion */
5856 {
5857 if (repeat_min > 0) goto END_REPEAT;
5858 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5859 }
5860
5861 /* The case of a zero minimum is special because of the need to stick
5862 OP_BRAZERO in front of it, and because the group appears once in the
5863 data, whereas in other cases it appears the minimum number of times. For
5864 this reason, it is simplest to treat this case separately, as otherwise
5865 the code gets far too messy. There are several special subcases when the
5866 minimum is zero. */
5867
5868 if (repeat_min == 0)
5869 {
5870 /* If the maximum is also zero, we used to just omit the group from the
5871 output altogether, like this:
5872
5873 ** if (repeat_max == 0)
5874 ** {
5875 ** code = previous;
5876 ** goto END_REPEAT;
5877 ** }
5878
5879 However, that fails when a group or a subgroup within it is referenced
5880 as a subroutine from elsewhere in the pattern, so now we stick in
5881 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5882 don't have a list of which groups are referenced, we cannot do this
5883 selectively.
5884
5885 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5886 and do no more at this point. However, we do need to adjust any
5887 OP_RECURSE calls inside the group that refer to the group itself or any
5888 internal or forward referenced group, because the offset is from the
5889 start of the whole regex. Temporarily terminate the pattern while doing
5890 this. */
5891
5892 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5893 {
5894 *code = OP_END;
5895 adjust_recurse(previous, 1, utf, cd, save_hwm);
5896 memmove(previous + 1, previous, IN_UCHARS(len));
5897 code++;
5898 if (repeat_max == 0)
5899 {
5900 *previous++ = OP_SKIPZERO;
5901 goto END_REPEAT;
5902 }
5903 brazeroptr = previous; /* Save for possessive optimizing */
5904 *previous++ = OP_BRAZERO + repeat_type;
5905 }
5906
5907 /* If the maximum is greater than 1 and limited, we have to replicate
5908 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
5909 The first one has to be handled carefully because it's the original
5910 copy, which has to be moved up. The remainder can be handled by code
5911 that is common with the non-zero minimum case below. We have to
5912 adjust the value or repeat_max, since one less copy is required. Once
5913 again, we may have to adjust any OP_RECURSE calls inside the group. */
5914
5915 else
5916 {
5917 int offset;
5918 *code = OP_END;
5919 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5920 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5921 code += 2 + LINK_SIZE;
5922 *previous++ = OP_BRAZERO + repeat_type;
5923 *previous++ = OP_BRA;
5924
5925 /* We chain together the bracket offset fields that have to be
5926 filled in later when the ends of the brackets are reached. */
5927
5928 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
5929 bralink = previous;
5930 PUTINC(previous, 0, offset);
5931 }
5932
5933 repeat_max--;
5934 }
5935
5936 /* If the minimum is greater than zero, replicate the group as many
5937 times as necessary, and adjust the maximum to the number of subsequent
5938 copies that we need. If we set a first char from the group, and didn't
5939 set a required char, copy the latter from the former. If there are any
5940 forward reference subroutine calls in the group, there will be entries on
5941 the workspace list; replicate these with an appropriate increment. */
5942
5943 else
5944 {
5945 if (repeat_min > 1)
5946 {
5947 /* In the pre-compile phase, we don't actually do the replication. We
5948 just adjust the length as if we had. Do some paranoid checks for
5949 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
5950 integer type when available, otherwise double. */
5951
5952 if (lengthptr != NULL)
5953 {
5954 int delta = (repeat_min - 1)*length_prevgroup;
5955 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
5956 (INT64_OR_DOUBLE)length_prevgroup >
5957 (INT64_OR_DOUBLE)INT_MAX ||
5958 OFLOW_MAX - *lengthptr < delta)
5959 {
5960 *errorcodeptr = ERR20;
5961 goto FAILED;
5962 }
5963 *lengthptr += delta;
5964 }
5965
5966 /* This is compiling for real. If there is a set first byte for
5967 the group, and we have not yet set a "required byte", set it. Make
5968 sure there is enough workspace for copying forward references before
5969 doing the copy. */
5970
5971 else
5972 {
5973 if (groupsetfirstchar && reqcharflags < 0)
5974 {
5975 reqchar = firstchar;
5976 reqcharflags = firstcharflags;
5977 }
5978
5979 for (i = 1; i < repeat_min; i++)
5980 {
5981 pcre_uchar *hc;
5982 pcre_uchar *this_hwm = cd->hwm;
5983 memcpy(code, previous, IN_UCHARS(len));
5984
5985 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5986 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5987 {
5988 int save_offset = save_hwm - cd->start_workspace;
5989 int this_offset = this_hwm - cd->start_workspace;
5990 *errorcodeptr = expand_workspace(cd);
5991 if (*errorcodeptr != 0) goto FAILED;
5992 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5993 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5994 }
5995
5996 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5997 {
5998 PUT(cd->hwm, 0, GET(hc, 0) + len);
5999 cd->hwm += LINK_SIZE;
6000 }
6001 save_hwm = this_hwm;
6002 code += len;
6003 }
6004 }
6005 }
6006
6007 if (repeat_max > 0) repeat_max -= repeat_min;
6008 }
6009
6010 /* This code is common to both the zero and non-zero minimum cases. If
6011 the maximum is limited, it replicates the group in a nested fashion,
6012 remembering the bracket starts on a stack. In the case of a zero minimum,
6013 the first one was set up above. In all cases the repeat_max now specifies
6014 the number of additional copies needed. Again, we must remember to
6015 replicate entries on the forward reference list. */
6016
6017 if (repeat_max >= 0)
6018 {
6019 /* In the pre-compile phase, we don't actually do the replication. We
6020 just adjust the length as if we had. For each repetition we must add 1
6021 to the length for BRAZERO and for all but the last repetition we must
6022 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
6023 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
6024 a 64-bit integer type when available, otherwise double. */
6025
6026 if (lengthptr != NULL && repeat_max > 0)
6027 {
6028 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
6029 2 - 2*LINK_SIZE; /* Last one doesn't nest */
6030 if ((INT64_OR_DOUBLE)repeat_max *
6031 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
6032 > (INT64_OR_DOUBLE)INT_MAX ||
6033 OFLOW_MAX - *lengthptr < delta)
6034 {
6035 *errorcodeptr = ERR20;
6036 goto FAILED;
6037 }
6038 *lengthptr += delta;
6039 }
6040
6041 /* This is compiling for real */
6042
6043 else for (i = repeat_max - 1; i >= 0; i--)
6044 {
6045 pcre_uchar *hc;
6046 pcre_uchar *this_hwm = cd->hwm;
6047
6048 *code++ = OP_BRAZERO + repeat_type;
6049
6050 /* All but the final copy start a new nesting, maintaining the
6051 chain of brackets outstanding. */
6052
6053 if (i != 0)
6054 {
6055 int offset;
6056 *code++ = OP_BRA;
6057 offset = (bralink == NULL)? 0 : (int)(code - bralink);
6058 bralink = code;
6059 PUTINC(code, 0, offset);
6060 }
6061
6062 memcpy(code, previous, IN_UCHARS(len));
6063
6064 /* Ensure there is enough workspace for forward references before
6065 copying them. */
6066
6067 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6068 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
6069 {
6070 int save_offset = save_hwm - cd->start_workspace;
6071 int this_offset = this_hwm - cd->start_workspace;
6072 *errorcodeptr = expand_workspace(cd);
6073 if (*errorcodeptr != 0) goto FAILED;
6074 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
6075 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
6076 }
6077
6078 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
6079 {
6080 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
6081 cd->hwm += LINK_SIZE;
6082 }
6083 save_hwm = this_hwm;
6084 code += len;
6085 }
6086
6087 /* Now chain through the pending brackets, and fill in their length
6088 fields (which are holding the chain links pro tem). */
6089
6090 while (bralink != NULL)
6091 {
6092 int oldlinkoffset;
6093 int offset = (int)(code - bralink + 1);
6094 pcre_uchar *bra = code - offset;
6095 oldlinkoffset = GET(bra, 1);
6096 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
6097 *code++ = OP_KET;
6098 PUTINC(code, 0, offset);
6099 PUT(bra, 1, offset);
6100 }
6101 }
6102
6103 /* If the maximum is unlimited, set a repeater in the final copy. For
6104 ONCE brackets, that's all we need to do. However, possessively repeated
6105 ONCE brackets can be converted into non-capturing brackets, as the
6106 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
6107 deal with possessive ONCEs specially.
6108
6109 Otherwise, when we are doing the actual compile phase, check to see
6110 whether this group is one that could match an empty string. If so,
6111 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
6112 that runtime checking can be done. [This check is also applied to ONCE
6113 groups at runtime, but in a different way.]
6114
6115 Then, if the quantifier was possessive and the bracket is not a
6116 conditional, we convert the BRA code to the POS form, and the KET code to
6117 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
6118 subpattern at both the start and at the end.) The use of special opcodes
6119 makes it possible to reduce greatly the stack usage in pcre_exec(). If
6120 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
6121
6122 Then, if the minimum number of matches is 1 or 0, cancel the possessive
6123 flag so that the default action below, of wrapping everything inside
6124 atomic brackets, does not happen. When the minimum is greater than 1,
6125 there will be earlier copies of the group, and so we still have to wrap
6126 the whole thing. */
6127
6128 else
6129 {
6130 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
6131 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
6132
6133 /* Convert possessive ONCE brackets to non-capturing */
6134
6135 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
6136 possessive_quantifier) *bracode = OP_BRA;
6137
6138 /* For non-possessive ONCE brackets, all we need to do is to
6139 set the KET. */
6140
6141 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
6142 *ketcode = OP_KETRMAX + repeat_type;
6143
6144 /* Handle non-ONCE brackets and possessive ONCEs (which have been
6145 converted to non-capturing above). */
6146
6147 else
6148 {
6149 /* In the compile phase, check for empty string matching. */
6150
6151 if (lengthptr == NULL)
6152 {
6153 pcre_uchar *scode = bracode;
6154 do
6155 {
6156 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6157 {
6158 *bracode += OP_SBRA - OP_BRA;
6159 break;
6160 }
6161 scode += GET(scode, 1);
6162 }
6163 while (*scode == OP_ALT);
6164 }
6165
6166 /* Handle possessive quantifiers. */
6167
6168 if (possessive_quantifier)
6169 {
6170 /* For COND brackets, we wrap the whole thing in a possessively
6171 repeated non-capturing bracket, because we have not invented POS
6172 versions of the COND opcodes. Because we are moving code along, we
6173 must ensure that any pending recursive references are updated. */
6174
6175 if (*bracode == OP_COND || *bracode == OP_SCOND)
6176 {
6177 int nlen = (int)(code - bracode);
6178 *code = OP_END;
6179 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
6180 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6181 code += 1 + LINK_SIZE;
6182 nlen += 1 + LINK_SIZE;
6183 *bracode = OP_BRAPOS;
6184 *code++ = OP_KETRPOS;
6185 PUTINC(code, 0, nlen);
6186 PUT(bracode, 1, nlen);
6187 }
6188
6189 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6190
6191 else
6192 {
6193 *bracode += 1; /* Switch to xxxPOS opcodes */
6194 *ketcode = OP_KETRPOS;
6195 }
6196
6197 /* If the minimum is zero, mark it as possessive, then unset the
6198 possessive flag when the minimum is 0 or 1. */
6199
6200 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6201 if (repeat_min < 2) possessive_quantifier = FALSE;
6202 }
6203
6204 /* Non-possessive quantifier */
6205
6206 else *ketcode = OP_KETRMAX + repeat_type;
6207 }
6208 }
6209 }
6210
6211 /* If previous is OP_FAIL, it was generated by an empty class [] in
6212 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6213 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6214 error above. We can just ignore the repeat in JS case. */
6215
6216 else if (*previous == OP_FAIL) goto END_REPEAT;
6217
6218 /* Else there's some kind of shambles */
6219
6220 else
6221 {
6222 *errorcodeptr = ERR11;
6223 goto FAILED;
6224 }
6225
6226 /* If the character following a repeat is '+', possessive_quantifier is
6227 TRUE. For some opcodes, there are special alternative opcodes for this
6228 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6229 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6230 Sun's Java package, but the special opcodes can optimize it.
6231
6232 Some (but not all) possessively repeated subpatterns have already been
6233 completely handled in the code just above. For them, possessive_quantifier
6234 is always FALSE at this stage. Note that the repeated item starts at
6235 tempcode, not at previous, which might be the first part of a string whose
6236 (former) last char we repeated. */
6237
6238 if (possessive_quantifier)
6239 {
6240 int len;
6241
6242 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6243 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6244 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6245 remains is greater than zero, there's a further opcode that can be
6246 handled. If not, do nothing, leaving the EXACT alone. */
6247
6248 switch(*tempcode)
6249 {
6250 case OP_TYPEEXACT:
6251 tempcode += PRIV(OP_lengths)[*tempcode] +
6252 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6253 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6254 break;
6255
6256 /* CHAR opcodes are used for exacts whose count is 1. */
6257
6258 case OP_CHAR:
6259 case OP_CHARI:
6260 case OP_NOT:
6261 case OP_NOTI:
6262 case OP_EXACT:
6263 case OP_EXACTI:
6264 case OP_NOTEXACT:
6265 case OP_NOTEXACTI:
6266 tempcode += PRIV(OP_lengths)[*tempcode];
6267 #ifdef SUPPORT_UTF
6268 if (utf && HAS_EXTRALEN(tempcode[-1]))
6269 tempcode += GET_EXTRALEN(tempcode[-1]);
6270 #endif
6271 break;
6272
6273 /* For the class opcodes, the repeat operator appears at the end;
6274 adjust tempcode to point to it. */
6275
6276 case OP_CLASS:
6277 case OP_NCLASS:
6278 tempcode += 1 + 32/sizeof(pcre_uchar);
6279 break;
6280
6281 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6282 case OP_XCLASS:
6283 tempcode += GET(tempcode, 1);
6284 break;
6285 #endif
6286 }
6287
6288 /* If tempcode is equal to code (which points to the end of the repeated
6289 item), it means we have skipped an EXACT item but there is no following
6290 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6291 all other cases, tempcode will be pointing to the repeat opcode, and will
6292 be less than code, so the value of len will be greater than 0. */
6293
6294 len = (int)(code - tempcode);
6295 if (len > 0)
6296 {
6297 unsigned int repcode = *tempcode;
6298
6299 /* There is a table for possessifying opcodes, all of which are less
6300 than OP_CALLOUT. A zero entry means there is no possessified version.
6301 */
6302
6303 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6304 *tempcode = opcode_possessify[repcode];
6305
6306 /* For opcode without a special possessified version, wrap the item in
6307 ONCE brackets. Because we are moving code along, we must ensure that any
6308 pending recursive references are updated. */
6309
6310 else
6311 {
6312 *code = OP_END;
6313 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6314 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6315 code += 1 + LINK_SIZE;
6316 len += 1 + LINK_SIZE;
6317 tempcode[0] = OP_ONCE;
6318 *code++ = OP_KET;
6319 PUTINC(code, 0, len);
6320 PUT(tempcode, 1, len);
6321 }
6322 }
6323
6324 #ifdef NEVER
6325 if (len > 0) switch (*tempcode)
6326 {
6327 case OP_STAR: *tempcode = OP_POSSTAR; break;
6328 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6329 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6330 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6331
6332 case OP_STARI: *tempcode = OP_POSSTARI; break;
6333 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6334 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6335 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6336
6337 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6338 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6339 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6340 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6341
6342 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6343 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6344 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6345 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6346
6347 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6348 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6349 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6350 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6351
6352 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6353 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6354 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6355 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6356
6357 /* Because we are moving code along, we must ensure that any
6358 pending recursive references are updated. */
6359
6360 default:
6361 *code = OP_END;
6362 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6363 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6364 code += 1 + LINK_SIZE;
6365 len += 1 + LINK_SIZE;
6366 tempcode[0] = OP_ONCE;
6367 *code++ = OP_KET;
6368 PUTINC(code, 0, len);
6369 PUT(tempcode, 1, len);
6370 break;
6371 }
6372 #endif
6373 }
6374
6375 /* In all case we no longer have a previous item. We also set the
6376 "follows varying string" flag for subsequently encountered reqchars if
6377 it isn't already set and we have just passed a varying length item. */
6378
6379 END_REPEAT:
6380 previous = NULL;
6381 cd->req_varyopt |= reqvary;
6382 break;
6383
6384
6385 /* ===================================================================*/
6386 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6387 lookbehind or option setting or condition or all the other extended
6388 parenthesis forms. */
6389
6390 case CHAR_LEFT_PARENTHESIS:
6391 newoptions = options;
6392 skipbytes = 0;
6393 bravalue = OP_CBRA;
6394 save_hwm = cd->hwm;
6395 reset_bracount = FALSE;
6396
6397 /* First deal with various "verbs" that can be introduced by '*'. */
6398
6399 ptr++;
6400 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6401 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6402 {
6403 int i, namelen;
6404 int arglen = 0;
6405 const char *vn = verbnames;
6406 const pcre_uchar *name = ptr + 1;
6407 const pcre_uchar *arg = NULL;
6408 previous = NULL;
6409 ptr++;
6410 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6411 namelen = (int)(ptr - name);
6412
6413 /* It appears that Perl allows any characters whatsoever, other than
6414 a closing parenthesis, to appear in arguments, so we no longer insist on
6415 letters, digits, and underscores. */
6416
6417 if (*ptr == CHAR_COLON)
6418 {
6419 arg = ++ptr;
6420 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6421 arglen = (int)(ptr - arg);
6422 if ((unsigned int)arglen > MAX_MARK)
6423 {
6424 *errorcodeptr = ERR75;
6425 goto FAILED;
6426 }
6427 }
6428
6429 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6430 {
6431 *errorcodeptr = ERR60;
6432 goto FAILED;
6433 }
6434
6435 /* Scan the table of verb names */
6436
6437 for (i = 0; i < verbcount; i++)
6438 {
6439 if (namelen == verbs[i].len &&
6440 STRNCMP_UC_C8(name, vn, namelen) == 0)
6441 {
6442 int setverb;
6443
6444 /* Check for open captures before ACCEPT and convert it to
6445 ASSERT_ACCEPT if in an assertion. */
6446
6447 if (verbs[i].op == OP_ACCEPT)
6448 {
6449 open_capitem *oc;
6450 if (arglen != 0)
6451 {
6452 *errorcodeptr = ERR59;
6453 goto FAILED;
6454 }
6455 cd->had_accept = TRUE;
6456 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6457 {
6458 *code++ = OP_CLOSE;
6459 PUT2INC(code, 0, oc->number);
6460 }
6461 setverb = *code++ =
6462 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6463
6464 /* Do not set firstchar after *ACCEPT */
6465 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6466 }
6467
6468 /* Handle other cases with/without an argument */
6469
6470 else if (arglen == 0)
6471 {
6472 if (verbs[i].op < 0) /* Argument is mandatory */
6473 {
6474 *errorcodeptr = ERR66;
6475 goto FAILED;
6476 }
6477 setverb = *code++ = verbs[i].op;
6478 }
6479
6480 else
6481 {
6482 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6483 {
6484 *errorcodeptr = ERR59;
6485 goto FAILED;
6486 }
6487 setverb = *code++ = verbs[i].op_arg;
6488 *code++ = arglen;
6489 memcpy(code, arg, IN_UCHARS(arglen));
6490 code += arglen;
6491 *code++ = 0;
6492 }
6493
6494 switch (setverb)
6495 {
6496 case OP_THEN:
6497 case OP_THEN_ARG:
6498 cd->external_flags |= PCRE_HASTHEN;
6499 break;
6500
6501 case OP_PRUNE:
6502 case OP_PRUNE_ARG:
6503 case OP_SKIP:
6504 case OP_SKIP_ARG:
6505 cd->had_pruneorskip = TRUE;
6506 break;
6507 }
6508
6509 break; /* Found verb, exit loop */
6510 }
6511
6512 vn += verbs[i].len + 1;
6513 }
6514
6515 if (i < verbcount) continue; /* Successfully handled a verb */
6516 *errorcodeptr = ERR60; /* Verb not recognized */
6517 goto FAILED;
6518 }
6519
6520 /* Deal with the extended parentheses; all are introduced by '?', and the
6521 appearance of any of them means that this is not a capturing group. */
6522
6523 else if (*ptr == CHAR_QUESTION_MARK)
6524 {
6525 int i, set, unset, namelen;
6526 int *optset;
6527 const pcre_uchar *name;
6528 pcre_uchar *slot;
6529
6530 switch (*(++ptr))
6531 {
6532 case CHAR_NUMBER_SIGN: /* Comment; skip to ket */
6533 ptr++;
6534 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6535 if (*ptr == CHAR_NULL)
6536 {
6537 *errorcodeptr = ERR18;
6538 goto FAILED;
6539 }
6540 continue;
6541
6542
6543 /* ------------------------------------------------------------ */
6544 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6545 reset_bracount = TRUE;
6546 /* Fall through */
6547
6548 /* ------------------------------------------------------------ */
6549 case CHAR_COLON: /* Non-capturing bracket */
6550 bravalue = OP_BRA;
6551 ptr++;
6552 break;
6553
6554
6555 /* ------------------------------------------------------------ */
6556 case CHAR_LEFT_PARENTHESIS:
6557 bravalue = OP_COND; /* Conditional group */
6558 tempptr = ptr;
6559
6560 /* A condition can be an assertion, a number (referring to a numbered
6561 group's having been set), a name (referring to a named group), or 'R',
6562 referring to recursion. R<digits> and R&name are also permitted for
6563 recursion tests.
6564
6565 There are ways of testing a named group: (?(name)) is used by Python;
6566 Perl 5.10 onwards uses (?(<name>) or (?('name')).
6567
6568 There is one unfortunate ambiguity, caused by history. 'R' can be the
6569 recursive thing or the name 'R' (and similarly for 'R' followed by
6570 digits). We look for a name first; if not found, we try the other case.
6571
6572 For compatibility with auto-callouts, we allow a callout to be
6573 specified before a condition that is an assertion. First, check for the
6574 syntax of a callout; if found, adjust the temporary pointer that is
6575 used to check for an assertion condition. That's all that is needed! */
6576
6577 if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6578 {
6579 for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6580 if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6581 tempptr += i + 1;
6582 }
6583
6584 /* For conditions that are assertions, check the syntax, and then exit
6585 the switch. This will take control down to where bracketed groups,
6586 including assertions, are processed. */
6587
6588 if (tempptr[1] == CHAR_QUESTION_MARK &&
6589 (tempptr[2] == CHAR_EQUALS_SIGN ||
6590 tempptr[2] == CHAR_EXCLAMATION_MARK ||
6591 tempptr[2] == CHAR_LESS_THAN_SIGN))
6592 break;
6593
6594 /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6595 need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6596
6597 code[1+LINK_SIZE] = OP_CREF;
6598 skipbytes = 1+IMM2_SIZE;
6599 refsign = -1;
6600
6601 /* Check for a test for recursion in a named group. */
6602
6603 ptr++;
6604 if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6605 {
6606 terminator = -1;
6607 ptr += 2;
6608 code[1+LINK_SIZE] = OP_RREF; /* Change the type of test */
6609 }
6610
6611 /* Check for a test for a named group's having been set, using the Perl
6612 syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6613 syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6614
6615 else if (*ptr == CHAR_LESS_THAN_SIGN)
6616 {
6617 terminator = CHAR_GREATER_THAN_SIGN;
6618 ptr++;
6619 }
6620 else if (*ptr == CHAR_APOSTROPHE)
6621 {
6622 terminator = CHAR_APOSTROPHE;
6623 ptr++;
6624 }
6625 else
6626 {
6627 terminator = CHAR_NULL;
6628 if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6629 else if (IS_DIGIT(*ptr)) refsign = 0;
6630 }
6631
6632 /* Handle a number */
6633
6634 if (refsign >= 0)
6635 {
6636 recno = 0;
6637 while (IS_DIGIT(*ptr))
6638 {
6639 recno = recno * 10 + (int)(*ptr - CHAR_0);
6640 ptr++;
6641 }
6642 }
6643
6644 /* Otherwise we expect to read a name; anything else is an error. When
6645 a name is one of a number of duplicates, a different opcode is used and
6646 it needs more memory. Unfortunately we cannot tell whether a name is a
6647 duplicate in the first pass, so we have to allow for more memory. */
6648
6649 else
6650 {
6651 if (IS_DIGIT(*ptr))
6652 {
6653 *errorcodeptr = ERR84;
6654 goto FAILED;
6655 }
6656 if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6657 {
6658 *errorcodeptr = ERR28; /* Assertion expected */
6659 goto FAILED;
6660