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

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1532 - (show annotations)
Fri Mar 6 11:59:39 2015 UTC (4 years, 8 months ago) by ph10
File MIME type: text/plain
File size: 313828 byte(s)
Fix bad compile of patterns like /[A-`]/i8 where the range contains 
characters with multiple other cases and the ranges adjoin.
1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2014 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40
41 /* This module contains the external function pcre_compile(), along with
42 supporting internal functions that are not used by other modules. */
43
44
45 #ifdef HAVE_CONFIG_H
46 #include "config.h"
47 #endif
48
49 #define NLBLOCK cd /* Block containing newline information */
50 #define PSSTART start_pattern /* Field containing pattern start */
51 #define PSEND end_pattern /* Field containing pattern end */
52
53 #include "pcre_internal.h"
54
55
56 /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57 is also used by pcretest. PCRE_DEBUG is not defined when building a production
58 library. We do not need to select pcre16_printint.c specially, because the
59 COMPILE_PCREx macro will already be appropriately set. */
60
61 #ifdef PCRE_DEBUG
62 /* pcre_printint.c should not include any headers */
63 #define PCRE_INCLUDED
64 #include "pcre_printint.c"
65 #undef PCRE_INCLUDED
66 #endif
67
68
69 /* Macro for setting individual bits in class bitmaps. */
70
71 #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72
73 /* Maximum length value to check against when making sure that the integer that
74 holds the compiled pattern length does not overflow. We make it a bit less than
75 INT_MAX to allow for adding in group terminating bytes, so that we don't have
76 to check them every time. */
77
78 #define OFLOW_MAX (INT_MAX - 20)
79
80 /* Definitions to allow mutual recursion */
81
82 static int
83 add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84 const pcre_uint32 *, unsigned int);
85
86 static BOOL
87 compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88 pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89 compile_data *, int *);
90
91
92
93 /*************************************************
94 * Code parameters and static tables *
95 *************************************************/
96
97 /* This value specifies the size of stack workspace that is used during the
98 first pre-compile phase that determines how much memory is required. The regex
99 is partly compiled into this space, but the compiled parts are discarded as
100 soon as they can be, so that hopefully there will never be an overrun. The code
101 does, however, check for an overrun. The largest amount I've seen used is 218,
102 so this number is very generous.
103
104 The same workspace is used during the second, actual compile phase for
105 remembering forward references to groups so that they can be filled in at the
106 end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107 is 4 there is plenty of room for most patterns. However, the memory can get
108 filled up by repetitions of forward references, for example patterns like
109 /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110 that the workspace is expanded using malloc() in this situation. The value
111 below is therefore a minimum, and we put a maximum on it for safety. The
112 minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113 kicks in at the same number of forward references in all cases. */
114
115 #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116 #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117
118 /* This value determines the size of the initial vector that is used for
119 remembering named groups during the pre-compile. It is allocated on the stack,
120 but if it is too small, it is expanded using malloc(), in a similar way to the
121 workspace. The value is the number of slots in the list. */
122
123 #define NAMED_GROUP_LIST_SIZE 20
124
125 /* The overrun tests check for a slightly smaller size so that they detect the
126 overrun before it actually does run off the end of the data block. */
127
128 #define WORK_SIZE_SAFETY_MARGIN (100)
129
130 /* Private flags added to firstchar and reqchar. */
131
132 #define REQ_CASELESS (1 << 0) /* Indicates caselessness */
133 #define REQ_VARY (1 << 1) /* Reqchar followed non-literal item */
134 /* Negative values for the firstchar and reqchar flags */
135 #define REQ_UNSET (-2)
136 #define REQ_NONE (-1)
137
138 /* Repeated character flags. */
139
140 #define UTF_LENGTH 0x10000000l /* The char contains its length. */
141
142 /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143 are simple data values; negative values are for special things like \d and so
144 on. Zero means further processing is needed (for things like \x), or the escape
145 is invalid. */
146
147 #ifndef EBCDIC
148
149 /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150 in UTF-8 mode. */
151
152 static const short int escapes[] = {
153 0, 0,
154 0, 0,
155 0, 0,
156 0, 0,
157 0, 0,
158 CHAR_COLON, CHAR_SEMICOLON,
159 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN,
160 CHAR_GREATER_THAN_SIGN, CHAR_QUESTION_MARK,
161 CHAR_COMMERCIAL_AT, -ESC_A,
162 -ESC_B, -ESC_C,
163 -ESC_D, -ESC_E,
164 0, -ESC_G,
165 -ESC_H, 0,
166 0, -ESC_K,
167 0, 0,
168 -ESC_N, 0,
169 -ESC_P, -ESC_Q,
170 -ESC_R, -ESC_S,
171 0, 0,
172 -ESC_V, -ESC_W,
173 -ESC_X, 0,
174 -ESC_Z, CHAR_LEFT_SQUARE_BRACKET,
175 CHAR_BACKSLASH, CHAR_RIGHT_SQUARE_BRACKET,
176 CHAR_CIRCUMFLEX_ACCENT, CHAR_UNDERSCORE,
177 CHAR_GRAVE_ACCENT, 7,
178 -ESC_b, 0,
179 -ESC_d, ESC_e,
180 ESC_f, 0,
181 -ESC_h, 0,
182 0, -ESC_k,
183 0, 0,
184 ESC_n, 0,
185 -ESC_p, 0,
186 ESC_r, -ESC_s,
187 ESC_tee, 0,
188 -ESC_v, -ESC_w,
189 0, 0,
190 -ESC_z
191 };
192
193 #else
194
195 /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
196
197 static const short int escapes[] = {
198 /* 48 */ 0, 0, 0, '.', '<', '(', '+', '|',
199 /* 50 */ '&', 0, 0, 0, 0, 0, 0, 0,
200 /* 58 */ 0, 0, '!', '$', '*', ')', ';', '~',
201 /* 60 */ '-', '/', 0, 0, 0, 0, 0, 0,
202 /* 68 */ 0, 0, '|', ',', '%', '_', '>', '?',
203 /* 70 */ 0, 0, 0, 0, 0, 0, 0, 0,
204 /* 78 */ 0, '`', ':', '#', '@', '\'', '=', '"',
205 /* 80 */ 0, 7, -ESC_b, 0, -ESC_d, ESC_e, ESC_f, 0,
206 /* 88 */-ESC_h, 0, 0, '{', 0, 0, 0, 0,
207 /* 90 */ 0, 0, -ESC_k, 'l', 0, ESC_n, 0, -ESC_p,
208 /* 98 */ 0, ESC_r, 0, '}', 0, 0, 0, 0,
209 /* A0 */ 0, '~', -ESC_s, ESC_tee, 0,-ESC_v, -ESC_w, 0,
210 /* A8 */ 0,-ESC_z, 0, 0, 0, '[', 0, 0,
211 /* B0 */ 0, 0, 0, 0, 0, 0, 0, 0,
212 /* B8 */ 0, 0, 0, 0, 0, ']', '=', '-',
213 /* C0 */ '{',-ESC_A, -ESC_B, -ESC_C, -ESC_D,-ESC_E, 0, -ESC_G,
214 /* C8 */-ESC_H, 0, 0, 0, 0, 0, 0, 0,
215 /* D0 */ '}', 0, -ESC_K, 0, 0,-ESC_N, 0, -ESC_P,
216 /* D8 */-ESC_Q,-ESC_R, 0, 0, 0, 0, 0, 0,
217 /* E0 */ '\\', 0, -ESC_S, 0, 0,-ESC_V, -ESC_W, -ESC_X,
218 /* E8 */ 0,-ESC_Z, 0, 0, 0, 0, 0, 0,
219 /* F0 */ 0, 0, 0, 0, 0, 0, 0, 0,
220 /* F8 */ 0, 0, 0, 0, 0, 0, 0, 0
221 };
222 #endif
223
224
225 /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226 searched linearly. Put all the names into a single string, in order to reduce
227 the number of relocations when a shared library is dynamically linked. The
228 string is built from string macros so that it works in UTF-8 mode on EBCDIC
229 platforms. */
230
231 typedef struct verbitem {
232 int len; /* Length of verb name */
233 int op; /* Op when no arg, or -1 if arg mandatory */
234 int op_arg; /* Op when arg present, or -1 if not allowed */
235 } verbitem;
236
237 static const char verbnames[] =
238 "\0" /* Empty name is a shorthand for MARK */
239 STRING_MARK0
240 STRING_ACCEPT0
241 STRING_COMMIT0
242 STRING_F0
243 STRING_FAIL0
244 STRING_PRUNE0
245 STRING_SKIP0
246 STRING_THEN;
247
248 static const verbitem verbs[] = {
249 { 0, -1, OP_MARK },
250 { 4, -1, OP_MARK },
251 { 6, OP_ACCEPT, -1 },
252 { 6, OP_COMMIT, -1 },
253 { 1, OP_FAIL, -1 },
254 { 4, OP_FAIL, -1 },
255 { 5, OP_PRUNE, OP_PRUNE_ARG },
256 { 4, OP_SKIP, OP_SKIP_ARG },
257 { 4, OP_THEN, OP_THEN_ARG }
258 };
259
260 static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261
262
263 /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264 another regex library. */
265
266 static const pcre_uchar sub_start_of_word[] = {
267 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268 CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269
270 static const pcre_uchar sub_end_of_word[] = {
271 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273 CHAR_RIGHT_PARENTHESIS, '\0' };
274
275
276 /* Tables of names of POSIX character classes and their lengths. The names are
277 now all in a single string, to reduce the number of relocations when a shared
278 library is dynamically loaded. The list of lengths is terminated by a zero
279 length entry. The first three must be alpha, lower, upper, as this is assumed
280 for handling case independence. The indices for graph, print, and punct are
281 needed, so identify them. */
282
283 static const char posix_names[] =
284 STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
285 STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
286 STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287 STRING_word0 STRING_xdigit;
288
289 static const pcre_uint8 posix_name_lengths[] = {
290 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291
292 #define PC_GRAPH 8
293 #define PC_PRINT 9
294 #define PC_PUNCT 10
295
296
297 /* Table of class bit maps for each POSIX class. Each class is formed from a
298 base map, with an optional addition or removal of another map. Then, for some
299 classes, there is some additional tweaking: for [:blank:] the vertical space
300 characters are removed, and for [:alpha:] and [:alnum:] the underscore
301 character is removed. The triples in the table consist of the base map offset,
302 second map offset or -1 if no second map, and a non-negative value for map
303 addition or a negative value for map subtraction (if there are two maps). The
304 absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
305 remove vertical space characters, 2 => remove underscore. */
306
307 static const int posix_class_maps[] = {
308 cbit_word, cbit_digit, -2, /* alpha */
309 cbit_lower, -1, 0, /* lower */
310 cbit_upper, -1, 0, /* upper */
311 cbit_word, -1, 2, /* alnum - word without underscore */
312 cbit_print, cbit_cntrl, 0, /* ascii */
313 cbit_space, -1, 1, /* blank - a GNU extension */
314 cbit_cntrl, -1, 0, /* cntrl */
315 cbit_digit, -1, 0, /* digit */
316 cbit_graph, -1, 0, /* graph */
317 cbit_print, -1, 0, /* print */
318 cbit_punct, -1, 0, /* punct */
319 cbit_space, -1, 0, /* space */
320 cbit_word, -1, 0, /* word - a Perl extension */
321 cbit_xdigit,-1, 0 /* xdigit */
322 };
323
324 /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325 Unicode property escapes. */
326
327 #ifdef SUPPORT_UCP
328 static const pcre_uchar string_PNd[] = {
329 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331 static const pcre_uchar string_pNd[] = {
332 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334 static const pcre_uchar string_PXsp[] = {
335 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337 static const pcre_uchar string_pXsp[] = {
338 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340 static const pcre_uchar string_PXwd[] = {
341 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343 static const pcre_uchar string_pXwd[] = {
344 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346
347 static const pcre_uchar *substitutes[] = {
348 string_PNd, /* \D */
349 string_pNd, /* \d */
350 string_PXsp, /* \S */ /* Xsp is Perl space, but from 8.34, Perl */
351 string_pXsp, /* \s */ /* space and POSIX space are the same. */
352 string_PXwd, /* \W */
353 string_pXwd /* \w */
354 };
355
356 /* The POSIX class substitutes must be in the order of the POSIX class names,
357 defined above, and there are both positive and negative cases. NULL means no
358 general substitute of a Unicode property escape (\p or \P). However, for some
359 POSIX classes (e.g. graph, print, punct) a special property code is compiled
360 directly. */
361
362 static const pcre_uchar string_pL[] = {
363 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365 static const pcre_uchar string_pLl[] = {
366 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368 static const pcre_uchar string_pLu[] = {
369 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371 static const pcre_uchar string_pXan[] = {
372 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374 static const pcre_uchar string_h[] = {
375 CHAR_BACKSLASH, CHAR_h, '\0' };
376 static const pcre_uchar string_pXps[] = {
377 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379 static const pcre_uchar string_PL[] = {
380 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382 static const pcre_uchar string_PLl[] = {
383 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385 static const pcre_uchar string_PLu[] = {
386 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388 static const pcre_uchar string_PXan[] = {
389 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391 static const pcre_uchar string_H[] = {
392 CHAR_BACKSLASH, CHAR_H, '\0' };
393 static const pcre_uchar string_PXps[] = {
394 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396
397 static const pcre_uchar *posix_substitutes[] = {
398 string_pL, /* alpha */
399 string_pLl, /* lower */
400 string_pLu, /* upper */
401 string_pXan, /* alnum */
402 NULL, /* ascii */
403 string_h, /* blank */
404 NULL, /* cntrl */
405 string_pNd, /* digit */
406 NULL, /* graph */
407 NULL, /* print */
408 NULL, /* punct */
409 string_pXps, /* space */ /* Xps is POSIX space, but from 8.34 */
410 string_pXwd, /* word */ /* Perl and POSIX space are the same */
411 NULL, /* xdigit */
412 /* Negated cases */
413 string_PL, /* ^alpha */
414 string_PLl, /* ^lower */
415 string_PLu, /* ^upper */
416 string_PXan, /* ^alnum */
417 NULL, /* ^ascii */
418 string_H, /* ^blank */
419 NULL, /* ^cntrl */
420 string_PNd, /* ^digit */
421 NULL, /* ^graph */
422 NULL, /* ^print */
423 NULL, /* ^punct */
424 string_PXps, /* ^space */ /* Xps is POSIX space, but from 8.34 */
425 string_PXwd, /* ^word */ /* Perl and POSIX space are the same */
426 NULL /* ^xdigit */
427 };
428 #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429 #endif
430
431 #define STRING(a) # a
432 #define XSTRING(s) STRING(s)
433
434 /* The texts of compile-time error messages. These are "char *" because they
435 are passed to the outside world. Do not ever re-use any error number, because
436 they are documented. Always add a new error instead. Messages marked DEAD below
437 are no longer used. This used to be a table of strings, but in order to reduce
438 the number of relocations needed when a shared library is loaded dynamically,
439 it is now one long string. We cannot use a table of offsets, because the
440 lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
441 simply count through to the one we want - this isn't a performance issue
442 because these strings are used only when there is a compilation error.
443
444 Each substring ends with \0 to insert a null character. This includes the final
445 substring, so that the whole string ends with \0\0, which can be detected when
446 counting through. */
447
448 static const char error_texts[] =
449 "no error\0"
450 "\\ at end of pattern\0"
451 "\\c at end of pattern\0"
452 "unrecognized character follows \\\0"
453 "numbers out of order in {} quantifier\0"
454 /* 5 */
455 "number too big in {} quantifier\0"
456 "missing terminating ] for character class\0"
457 "invalid escape sequence in character class\0"
458 "range out of order in character class\0"
459 "nothing to repeat\0"
460 /* 10 */
461 "operand of unlimited repeat could match the empty string\0" /** DEAD **/
462 "internal error: unexpected repeat\0"
463 "unrecognized character after (? or (?-\0"
464 "POSIX named classes are supported only within a class\0"
465 "missing )\0"
466 /* 15 */
467 "reference to non-existent subpattern\0"
468 "erroffset passed as NULL\0"
469 "unknown option bit(s) set\0"
470 "missing ) after comment\0"
471 "parentheses nested too deeply\0" /** DEAD **/
472 /* 20 */
473 "regular expression is too large\0"
474 "failed to get memory\0"
475 "unmatched parentheses\0"
476 "internal error: code overflow\0"
477 "unrecognized character after (?<\0"
478 /* 25 */
479 "lookbehind assertion is not fixed length\0"
480 "malformed number or name after (?(\0"
481 "conditional group contains more than two branches\0"
482 "assertion expected after (?(\0"
483 "(?R or (?[+-]digits must be followed by )\0"
484 /* 30 */
485 "unknown POSIX class name\0"
486 "POSIX collating elements are not supported\0"
487 "this version of PCRE is compiled without UTF support\0"
488 "spare error\0" /** DEAD **/
489 "character value in \\x{} or \\o{} is too large\0"
490 /* 35 */
491 "invalid condition (?(0)\0"
492 "\\C not allowed in lookbehind assertion\0"
493 "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
494 "number after (?C is > 255\0"
495 "closing ) for (?C expected\0"
496 /* 40 */
497 "recursive call could loop indefinitely\0"
498 "unrecognized character after (?P\0"
499 "syntax error in subpattern name (missing terminator)\0"
500 "two named subpatterns have the same name\0"
501 "invalid UTF-8 string\0"
502 /* 45 */
503 "support for \\P, \\p, and \\X has not been compiled\0"
504 "malformed \\P or \\p sequence\0"
505 "unknown property name after \\P or \\p\0"
506 "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
507 "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508 /* 50 */
509 "repeated subpattern is too long\0" /** DEAD **/
510 "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
511 "internal error: overran compiling workspace\0"
512 "internal error: previously-checked referenced subpattern not found\0"
513 "DEFINE group contains more than one branch\0"
514 /* 55 */
515 "repeating a DEFINE group is not allowed\0" /** DEAD **/
516 "inconsistent NEWLINE options\0"
517 "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
518 "a numbered reference must not be zero\0"
519 "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520 /* 60 */
521 "(*VERB) not recognized or malformed\0"
522 "number is too big\0"
523 "subpattern name expected\0"
524 "digit expected after (?+\0"
525 "] is an invalid data character in JavaScript compatibility mode\0"
526 /* 65 */
527 "different names for subpatterns of the same number are not allowed\0"
528 "(*MARK) must have an argument\0"
529 "this version of PCRE is not compiled with Unicode property support\0"
530 "\\c must be followed by an ASCII character\0"
531 "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532 /* 70 */
533 "internal error: unknown opcode in find_fixedlength()\0"
534 "\\N is not supported in a class\0"
535 "too many forward references\0"
536 "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537 "invalid UTF-16 string\0"
538 /* 75 */
539 "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540 "character value in \\u.... sequence is too large\0"
541 "invalid UTF-32 string\0"
542 "setting UTF is disabled by the application\0"
543 "non-hex character in \\x{} (closing brace missing?)\0"
544 /* 80 */
545 "non-octal character in \\o{} (closing brace missing?)\0"
546 "missing opening brace after \\o\0"
547 "parentheses are too deeply nested\0"
548 "invalid range in character class\0"
549 "group name must start with a non-digit\0"
550 /* 85 */
551 "parentheses are too deeply nested (stack check)\0"
552 "digits missing in \\x{} or \\o{}\0"
553 ;
554
555 /* Table to identify digits and hex digits. This is used when compiling
556 patterns. Note that the tables in chartables are dependent on the locale, and
557 may mark arbitrary characters as digits - but the PCRE compiling code expects
558 to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
559 a private table here. It costs 256 bytes, but it is a lot faster than doing
560 character value tests (at least in some simple cases I timed), and in some
561 applications one wants PCRE to compile efficiently as well as match
562 efficiently.
563
564 For convenience, we use the same bit definitions as in chartables:
565
566 0x04 decimal digit
567 0x08 hexadecimal digit
568
569 Then we can use ctype_digit and ctype_xdigit in the code. */
570
571 /* Using a simple comparison for decimal numbers rather than a memory read
572 is much faster, and the resulting code is simpler (the compiler turns it
573 into a subtraction and unsigned comparison). */
574
575 #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
576
577 #ifndef EBCDIC
578
579 /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
580 UTF-8 mode. */
581
582 static const pcre_uint8 digitab[] =
583 {
584 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
585 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
586 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
587 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
588 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
589 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
590 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
591 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
592 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* @ - G */
593 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H - O */
594 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* P - W */
595 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* X - _ */
596 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
597 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
598 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
599 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
600 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
601 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
602 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
603 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
604 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
605 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
606 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
607 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
608 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
609 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
610 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
611 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
612 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
613 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
614 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
615 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
616
617 #else
618
619 /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
620
621 static const pcre_uint8 digitab[] =
622 {
623 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
624 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
625 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
626 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
627 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
628 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
629 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
630 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
631 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
632 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
633 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
634 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
635 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
636 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
637 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
638 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
639 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
640 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
641 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
642 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
643 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
644 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
645 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
646 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
647 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
648 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
649 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
650 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
651 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
652 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
653 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
654 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
655
656 static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
657 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
658 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
659 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
660 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
661 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
662 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
663 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
664 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
665 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
666 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
667 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
668 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
669 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
670 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
671 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
672 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
673 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
674 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
675 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
676 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
677 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
678 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
679 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
680 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
681 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
682 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
683 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
684 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
685 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
686 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
687 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
688 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
689 #endif
690
691
692 /* This table is used to check whether auto-possessification is possible
693 between adjacent character-type opcodes. The left-hand (repeated) opcode is
694 used to select the row, and the right-hand opcode is use to select the column.
695 A value of 1 means that auto-possessification is OK. For example, the second
696 value in the first row means that \D+\d can be turned into \D++\d.
697
698 The Unicode property types (\P and \p) have to be present to fill out the table
699 because of what their opcode values are, but the table values should always be
700 zero because property types are handled separately in the code. The last four
701 columns apply to items that cannot be repeated, so there is no need to have
702 rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
703 *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
704
705 #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
706 #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
707
708 static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
709 /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
710 { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
711 { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
712 { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
713 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
714 { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
715 { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
716 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
717 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
718 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
719 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
720 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
721 { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
722 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
723 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
724 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
725 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
726 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
727 };
728
729
730 /* This table is used to check whether auto-possessification is possible
731 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
732 left-hand (repeated) opcode is used to select the row, and the right-hand
733 opcode is used to select the column. The values are as follows:
734
735 0 Always return FALSE (never auto-possessify)
736 1 Character groups are distinct (possessify if both are OP_PROP)
737 2 Check character categories in the same group (general or particular)
738 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
739
740 4 Check left general category vs right particular category
741 5 Check right general category vs left particular category
742
743 6 Left alphanum vs right general category
744 7 Left space vs right general category
745 8 Left word vs right general category
746
747 9 Right alphanum vs left general category
748 10 Right space vs left general category
749 11 Right word vs left general category
750
751 12 Left alphanum vs right particular category
752 13 Left space vs right particular category
753 14 Left word vs right particular category
754
755 15 Right alphanum vs left particular category
756 16 Right space vs left particular category
757 17 Right word vs left particular category
758 */
759
760 static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
761 /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
762 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
763 { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
764 { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
765 { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
766 { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
767 { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
768 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
769 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
770 { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
771 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
772 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
773 };
774
775 /* This table is used to check whether auto-possessification is possible
776 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
777 specifies a general category and the other specifies a particular category. The
778 row is selected by the general category and the column by the particular
779 category. The value is 1 if the particular category is not part of the general
780 category. */
781
782 static const pcre_uint8 catposstab[7][30] = {
783 /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
784 { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* C */
785 { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* L */
786 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* M */
787 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* N */
788 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 }, /* P */
789 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 }, /* S */
790 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 } /* Z */
791 };
792
793 /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
794 a general or particular category. The properties in each row are those
795 that apply to the character set in question. Duplication means that a little
796 unnecessary work is done when checking, but this keeps things much simpler
797 because they can all use the same code. For more details see the comment where
798 this table is used.
799
800 Note: SPACE and PXSPACE used to be different because Perl excluded VT from
801 "space", but from Perl 5.18 it's included, so both categories are treated the
802 same here. */
803
804 static const pcre_uint8 posspropstab[3][4] = {
805 { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
806 { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
807 { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
808 };
809
810 /* This table is used when converting repeating opcodes into possessified
811 versions as a result of an explicit possessive quantifier such as ++. A zero
812 value means there is no possessified version - in those cases the item in
813 question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
814 because all relevant opcodes are less than that. */
815
816 static const pcre_uint8 opcode_possessify[] = {
817 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */
818 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */
819
820 0, /* NOTI */
821 OP_POSSTAR, 0, /* STAR, MINSTAR */
822 OP_POSPLUS, 0, /* PLUS, MINPLUS */
823 OP_POSQUERY, 0, /* QUERY, MINQUERY */
824 OP_POSUPTO, 0, /* UPTO, MINUPTO */
825 0, /* EXACT */
826 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */
827
828 OP_POSSTARI, 0, /* STARI, MINSTARI */
829 OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */
830 OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */
831 OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */
832 0, /* EXACTI */
833 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */
834
835 OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */
836 OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */
837 OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */
838 OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */
839 0, /* NOTEXACT */
840 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
841
842 OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */
843 OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */
844 OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */
845 OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */
846 0, /* NOTEXACTI */
847 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
848
849 OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */
850 OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */
851 OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */
852 OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */
853 0, /* TYPEEXACT */
854 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
855
856 OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */
857 OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */
858 OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */
859 OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */
860 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */
861
862 0, 0, 0, /* CLASS, NCLASS, XCLASS */
863 0, 0, /* REF, REFI */
864 0, 0, /* DNREF, DNREFI */
865 0, 0 /* RECURSE, CALLOUT */
866 };
867
868
869
870 /*************************************************
871 * Find an error text *
872 *************************************************/
873
874 /* The error texts are now all in one long string, to save on relocations. As
875 some of the text is of unknown length, we can't use a table of offsets.
876 Instead, just count through the strings. This is not a performance issue
877 because it happens only when there has been a compilation error.
878
879 Argument: the error number
880 Returns: pointer to the error string
881 */
882
883 static const char *
884 find_error_text(int n)
885 {
886 const char *s = error_texts;
887 for (; n > 0; n--)
888 {
889 while (*s++ != CHAR_NULL) {};
890 if (*s == CHAR_NULL) return "Error text not found (please report)";
891 }
892 return s;
893 }
894
895
896
897 /*************************************************
898 * Expand the workspace *
899 *************************************************/
900
901 /* This function is called during the second compiling phase, if the number of
902 forward references fills the existing workspace, which is originally a block on
903 the stack. A larger block is obtained from malloc() unless the ultimate limit
904 has been reached or the increase will be rather small.
905
906 Argument: pointer to the compile data block
907 Returns: 0 if all went well, else an error number
908 */
909
910 static int
911 expand_workspace(compile_data *cd)
912 {
913 pcre_uchar *newspace;
914 int newsize = cd->workspace_size * 2;
915
916 if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
917 if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
918 newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
919 return ERR72;
920
921 newspace = (PUBL(malloc))(IN_UCHARS(newsize));
922 if (newspace == NULL) return ERR21;
923 memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
924 cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
925 if (cd->workspace_size > COMPILE_WORK_SIZE)
926 (PUBL(free))((void *)cd->start_workspace);
927 cd->start_workspace = newspace;
928 cd->workspace_size = newsize;
929 return 0;
930 }
931
932
933
934 /*************************************************
935 * Check for counted repeat *
936 *************************************************/
937
938 /* This function is called when a '{' is encountered in a place where it might
939 start a quantifier. It looks ahead to see if it really is a quantifier or not.
940 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
941 where the ddds are digits.
942
943 Arguments:
944 p pointer to the first char after '{'
945
946 Returns: TRUE or FALSE
947 */
948
949 static BOOL
950 is_counted_repeat(const pcre_uchar *p)
951 {
952 if (!IS_DIGIT(*p)) return FALSE;
953 p++;
954 while (IS_DIGIT(*p)) p++;
955 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
956
957 if (*p++ != CHAR_COMMA) return FALSE;
958 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
959
960 if (!IS_DIGIT(*p)) return FALSE;
961 p++;
962 while (IS_DIGIT(*p)) p++;
963
964 return (*p == CHAR_RIGHT_CURLY_BRACKET);
965 }
966
967
968
969 /*************************************************
970 * Handle escapes *
971 *************************************************/
972
973 /* This function is called when a \ has been encountered. It either returns a
974 positive value for a simple escape such as \n, or 0 for a data character which
975 will be placed in chptr. A backreference to group n is returned as negative n.
976 When UTF-8 is enabled, a positive value greater than 255 may be returned in
977 chptr. On entry, ptr is pointing at the \. On exit, it is on the final
978 character of the escape sequence.
979
980 Arguments:
981 ptrptr points to the pattern position pointer
982 chptr points to a returned data character
983 errorcodeptr points to the errorcode variable
984 bracount number of previous extracting brackets
985 options the options bits
986 isclass TRUE if inside a character class
987
988 Returns: zero => a data character
989 positive => a special escape sequence
990 negative => a back reference
991 on error, errorcodeptr is set
992 */
993
994 static int
995 check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
996 int bracount, int options, BOOL isclass)
997 {
998 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
999 BOOL utf = (options & PCRE_UTF8) != 0;
1000 const pcre_uchar *ptr = *ptrptr + 1;
1001 pcre_uint32 c;
1002 int escape = 0;
1003 int i;
1004
1005 GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
1006 ptr--; /* Set pointer back to the last byte */
1007
1008 /* If backslash is at the end of the pattern, it's an error. */
1009
1010 if (c == CHAR_NULL) *errorcodeptr = ERR1;
1011
1012 /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1013 in a table. A non-zero result is something that can be returned immediately.
1014 Otherwise further processing may be required. */
1015
1016 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1017 /* Not alphanumeric */
1018 else if (c < CHAR_0 || c > CHAR_z) {}
1019 else if ((i = escapes[c - CHAR_0]) != 0)
1020 { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1021
1022 #else /* EBCDIC coding */
1023 /* Not alphanumeric */
1024 else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1025 else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1026 #endif
1027
1028 /* Escapes that need further processing, or are illegal. */
1029
1030 else
1031 {
1032 const pcre_uchar *oldptr;
1033 BOOL braced, negated, overflow;
1034 int s;
1035
1036 switch (c)
1037 {
1038 /* A number of Perl escapes are not handled by PCRE. We give an explicit
1039 error. */
1040
1041 case CHAR_l:
1042 case CHAR_L:
1043 *errorcodeptr = ERR37;
1044 break;
1045
1046 case CHAR_u:
1047 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1048 {
1049 /* In JavaScript, \u must be followed by four hexadecimal numbers.
1050 Otherwise it is a lowercase u letter. */
1051 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1052 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1053 && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1054 && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1055 {
1056 c = 0;
1057 for (i = 0; i < 4; ++i)
1058 {
1059 register pcre_uint32 cc = *(++ptr);
1060 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1061 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1062 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1063 #else /* EBCDIC coding */
1064 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1065 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1066 #endif
1067 }
1068
1069 #if defined COMPILE_PCRE8
1070 if (c > (utf ? 0x10ffffU : 0xffU))
1071 #elif defined COMPILE_PCRE16
1072 if (c > (utf ? 0x10ffffU : 0xffffU))
1073 #elif defined COMPILE_PCRE32
1074 if (utf && c > 0x10ffffU)
1075 #endif
1076 {
1077 *errorcodeptr = ERR76;
1078 }
1079 else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1080 }
1081 }
1082 else
1083 *errorcodeptr = ERR37;
1084 break;
1085
1086 case CHAR_U:
1087 /* In JavaScript, \U is an uppercase U letter. */
1088 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1089 break;
1090
1091 /* In a character class, \g is just a literal "g". Outside a character
1092 class, \g must be followed by one of a number of specific things:
1093
1094 (1) A number, either plain or braced. If positive, it is an absolute
1095 backreference. If negative, it is a relative backreference. This is a Perl
1096 5.10 feature.
1097
1098 (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1099 is part of Perl's movement towards a unified syntax for back references. As
1100 this is synonymous with \k{name}, we fudge it up by pretending it really
1101 was \k.
1102
1103 (3) For Oniguruma compatibility we also support \g followed by a name or a
1104 number either in angle brackets or in single quotes. However, these are
1105 (possibly recursive) subroutine calls, _not_ backreferences. Just return
1106 the ESC_g code (cf \k). */
1107
1108 case CHAR_g:
1109 if (isclass) break;
1110 if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1111 {
1112 escape = ESC_g;
1113 break;
1114 }
1115
1116 /* Handle the Perl-compatible cases */
1117
1118 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1119 {
1120 const pcre_uchar *p;
1121 for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1122 if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1123 if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1124 {
1125 escape = ESC_k;
1126 break;
1127 }
1128 braced = TRUE;
1129 ptr++;
1130 }
1131 else braced = FALSE;
1132
1133 if (ptr[1] == CHAR_MINUS)
1134 {
1135 negated = TRUE;
1136 ptr++;
1137 }
1138 else negated = FALSE;
1139
1140 /* The integer range is limited by the machine's int representation. */
1141 s = 0;
1142 overflow = FALSE;
1143 while (IS_DIGIT(ptr[1]))
1144 {
1145 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1146 {
1147 overflow = TRUE;
1148 break;
1149 }
1150 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1151 }
1152 if (overflow) /* Integer overflow */
1153 {
1154 while (IS_DIGIT(ptr[1]))
1155 ptr++;
1156 *errorcodeptr = ERR61;
1157 break;
1158 }
1159
1160 if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1161 {
1162 *errorcodeptr = ERR57;
1163 break;
1164 }
1165
1166 if (s == 0)
1167 {
1168 *errorcodeptr = ERR58;
1169 break;
1170 }
1171
1172 if (negated)
1173 {
1174 if (s > bracount)
1175 {
1176 *errorcodeptr = ERR15;
1177 break;
1178 }
1179 s = bracount - (s - 1);
1180 }
1181
1182 escape = -s;
1183 break;
1184
1185 /* The handling of escape sequences consisting of a string of digits
1186 starting with one that is not zero is not straightforward. Perl has changed
1187 over the years. Nowadays \g{} for backreferences and \o{} for octal are
1188 recommended to avoid the ambiguities in the old syntax.
1189
1190 Outside a character class, the digits are read as a decimal number. If the
1191 number is less than 8 (used to be 10), or if there are that many previous
1192 extracting left brackets, then it is a back reference. Otherwise, up to
1193 three octal digits are read to form an escaped byte. Thus \123 is likely to
1194 be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1195 the octal value is greater than 377, the least significant 8 bits are
1196 taken. \8 and \9 are treated as the literal characters 8 and 9.
1197
1198 Inside a character class, \ followed by a digit is always either a literal
1199 8 or 9 or an octal number. */
1200
1201 case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1202 case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1203
1204 if (!isclass)
1205 {
1206 oldptr = ptr;
1207 /* The integer range is limited by the machine's int representation. */
1208 s = (int)(c -CHAR_0);
1209 overflow = FALSE;
1210 while (IS_DIGIT(ptr[1]))
1211 {
1212 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1213 {
1214 overflow = TRUE;
1215 break;
1216 }
1217 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1218 }
1219 if (overflow) /* Integer overflow */
1220 {
1221 while (IS_DIGIT(ptr[1]))
1222 ptr++;
1223 *errorcodeptr = ERR61;
1224 break;
1225 }
1226 if (s < 8 || s <= bracount) /* Check for back reference */
1227 {
1228 escape = -s;
1229 break;
1230 }
1231 ptr = oldptr; /* Put the pointer back and fall through */
1232 }
1233
1234 /* Handle a digit following \ when the number is not a back reference. If
1235 the first digit is 8 or 9, Perl used to generate a binary zero byte and
1236 then treat the digit as a following literal. At least by Perl 5.18 this
1237 changed so as not to insert the binary zero. */
1238
1239 if ((c = *ptr) >= CHAR_8) break;
1240
1241 /* Fall through with a digit less than 8 */
1242
1243 /* \0 always starts an octal number, but we may drop through to here with a
1244 larger first octal digit. The original code used just to take the least
1245 significant 8 bits of octal numbers (I think this is what early Perls used
1246 to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1247 but no more than 3 octal digits. */
1248
1249 case CHAR_0:
1250 c -= CHAR_0;
1251 while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1252 c = c * 8 + *(++ptr) - CHAR_0;
1253 #ifdef COMPILE_PCRE8
1254 if (!utf && c > 0xff) *errorcodeptr = ERR51;
1255 #endif
1256 break;
1257
1258 /* \o is a relatively new Perl feature, supporting a more general way of
1259 specifying character codes in octal. The only supported form is \o{ddd}. */
1260
1261 case CHAR_o:
1262 if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1263 if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1264 {
1265 ptr += 2;
1266 c = 0;
1267 overflow = FALSE;
1268 while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1269 {
1270 register pcre_uint32 cc = *ptr++;
1271 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1272 #ifdef COMPILE_PCRE32
1273 if (c >= 0x20000000l) { overflow = TRUE; break; }
1274 #endif
1275 c = (c << 3) + cc - CHAR_0 ;
1276 #if defined COMPILE_PCRE8
1277 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1278 #elif defined COMPILE_PCRE16
1279 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1280 #elif defined COMPILE_PCRE32
1281 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1282 #endif
1283 }
1284 if (overflow)
1285 {
1286 while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1287 *errorcodeptr = ERR34;
1288 }
1289 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1290 {
1291 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1292 }
1293 else *errorcodeptr = ERR80;
1294 }
1295 break;
1296
1297 /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1298 numbers. Otherwise it is a lowercase x letter. */
1299
1300 case CHAR_x:
1301 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1302 {
1303 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1304 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1305 {
1306 c = 0;
1307 for (i = 0; i < 2; ++i)
1308 {
1309 register pcre_uint32 cc = *(++ptr);
1310 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1311 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1312 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1313 #else /* EBCDIC coding */
1314 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1315 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1316 #endif
1317 }
1318 }
1319 } /* End JavaScript handling */
1320
1321 /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1322 greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1323 digits. If not, { used to be treated as a data character. However, Perl
1324 seems to read hex digits up to the first non-such, and ignore the rest, so
1325 that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1326 now gives an error. */
1327
1328 else
1329 {
1330 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1331 {
1332 ptr += 2;
1333 if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1334 {
1335 *errorcodeptr = ERR86;
1336 break;
1337 }
1338 c = 0;
1339 overflow = FALSE;
1340 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1341 {
1342 register pcre_uint32 cc = *ptr++;
1343 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1344
1345 #ifdef COMPILE_PCRE32
1346 if (c >= 0x10000000l) { overflow = TRUE; break; }
1347 #endif
1348
1349 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1350 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1351 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1352 #else /* EBCDIC coding */
1353 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1354 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1355 #endif
1356
1357 #if defined COMPILE_PCRE8
1358 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1359 #elif defined COMPILE_PCRE16
1360 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1361 #elif defined COMPILE_PCRE32
1362 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1363 #endif
1364 }
1365
1366 if (overflow)
1367 {
1368 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1369 *errorcodeptr = ERR34;
1370 }
1371
1372 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1373 {
1374 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1375 }
1376
1377 /* If the sequence of hex digits does not end with '}', give an error.
1378 We used just to recognize this construct and fall through to the normal
1379 \x handling, but nowadays Perl gives an error, which seems much more
1380 sensible, so we do too. */
1381
1382 else *errorcodeptr = ERR79;
1383 } /* End of \x{} processing */
1384
1385 /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1386
1387 else
1388 {
1389 c = 0;
1390 while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1391 {
1392 pcre_uint32 cc; /* Some compilers don't like */
1393 cc = *(++ptr); /* ++ in initializers */
1394 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1395 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1396 c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1397 #else /* EBCDIC coding */
1398 if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1399 c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1400 #endif
1401 }
1402 } /* End of \xdd handling */
1403 } /* End of Perl-style \x handling */
1404 break;
1405
1406 /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1407 An error is given if the byte following \c is not an ASCII character. This
1408 coding is ASCII-specific, but then the whole concept of \cx is
1409 ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1410
1411 case CHAR_c:
1412 c = *(++ptr);
1413 if (c == CHAR_NULL)
1414 {
1415 *errorcodeptr = ERR2;
1416 break;
1417 }
1418 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1419 if (c > 127) /* Excludes all non-ASCII in either mode */
1420 {
1421 *errorcodeptr = ERR68;
1422 break;
1423 }
1424 if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1425 c ^= 0x40;
1426 #else /* EBCDIC coding */
1427 if (c >= CHAR_a && c <= CHAR_z) c += 64;
1428 c ^= 0xC0;
1429 #endif
1430 break;
1431
1432 /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1433 other alphanumeric following \ is an error if PCRE_EXTRA was set;
1434 otherwise, for Perl compatibility, it is a literal. This code looks a bit
1435 odd, but there used to be some cases other than the default, and there may
1436 be again in future, so I haven't "optimized" it. */
1437
1438 default:
1439 if ((options & PCRE_EXTRA) != 0) switch(c)
1440 {
1441 default:
1442 *errorcodeptr = ERR3;
1443 break;
1444 }
1445 break;
1446 }
1447 }
1448
1449 /* Perl supports \N{name} for character names, as well as plain \N for "not
1450 newline". PCRE does not support \N{name}. However, it does support
1451 quantification such as \N{2,3}. */
1452
1453 if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1454 !is_counted_repeat(ptr+2))
1455 *errorcodeptr = ERR37;
1456
1457 /* If PCRE_UCP is set, we change the values for \d etc. */
1458
1459 if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1460 escape += (ESC_DU - ESC_D);
1461
1462 /* Set the pointer to the final character before returning. */
1463
1464 *ptrptr = ptr;
1465 *chptr = c;
1466 return escape;
1467 }
1468
1469
1470
1471 #ifdef SUPPORT_UCP
1472 /*************************************************
1473 * Handle \P and \p *
1474 *************************************************/
1475
1476 /* This function is called after \P or \p has been encountered, provided that
1477 PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1478 pointing at the P or p. On exit, it is pointing at the final character of the
1479 escape sequence.
1480
1481 Argument:
1482 ptrptr points to the pattern position pointer
1483 negptr points to a boolean that is set TRUE for negation else FALSE
1484 ptypeptr points to an unsigned int that is set to the type value
1485 pdataptr points to an unsigned int that is set to the detailed property value
1486 errorcodeptr points to the error code variable
1487
1488 Returns: TRUE if the type value was found, or FALSE for an invalid type
1489 */
1490
1491 static BOOL
1492 get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1493 unsigned int *pdataptr, int *errorcodeptr)
1494 {
1495 pcre_uchar c;
1496 int i, bot, top;
1497 const pcre_uchar *ptr = *ptrptr;
1498 pcre_uchar name[32];
1499
1500 c = *(++ptr);
1501 if (c == CHAR_NULL) goto ERROR_RETURN;
1502
1503 *negptr = FALSE;
1504
1505 /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1506 negation. */
1507
1508 if (c == CHAR_LEFT_CURLY_BRACKET)
1509 {
1510 if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1511 {
1512 *negptr = TRUE;
1513 ptr++;
1514 }
1515 for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1516 {
1517 c = *(++ptr);
1518 if (c == CHAR_NULL) goto ERROR_RETURN;
1519 if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1520 name[i] = c;
1521 }
1522 if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1523 name[i] = 0;
1524 }
1525
1526 /* Otherwise there is just one following character */
1527
1528 else
1529 {
1530 name[0] = c;
1531 name[1] = 0;
1532 }
1533
1534 *ptrptr = ptr;
1535
1536 /* Search for a recognized property name using binary chop */
1537
1538 bot = 0;
1539 top = PRIV(utt_size);
1540
1541 while (bot < top)
1542 {
1543 int r;
1544 i = (bot + top) >> 1;
1545 r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1546 if (r == 0)
1547 {
1548 *ptypeptr = PRIV(utt)[i].type;
1549 *pdataptr = PRIV(utt)[i].value;
1550 return TRUE;
1551 }
1552 if (r > 0) bot = i + 1; else top = i;
1553 }
1554
1555 *errorcodeptr = ERR47;
1556 *ptrptr = ptr;
1557 return FALSE;
1558
1559 ERROR_RETURN:
1560 *errorcodeptr = ERR46;
1561 *ptrptr = ptr;
1562 return FALSE;
1563 }
1564 #endif
1565
1566
1567
1568 /*************************************************
1569 * Read repeat counts *
1570 *************************************************/
1571
1572 /* Read an item of the form {n,m} and return the values. This is called only
1573 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1574 so the syntax is guaranteed to be correct, but we need to check the values.
1575
1576 Arguments:
1577 p pointer to first char after '{'
1578 minp pointer to int for min
1579 maxp pointer to int for max
1580 returned as -1 if no max
1581 errorcodeptr points to error code variable
1582
1583 Returns: pointer to '}' on success;
1584 current ptr on error, with errorcodeptr set non-zero
1585 */
1586
1587 static const pcre_uchar *
1588 read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1589 {
1590 int min = 0;
1591 int max = -1;
1592
1593 while (IS_DIGIT(*p))
1594 {
1595 min = min * 10 + (int)(*p++ - CHAR_0);
1596 if (min > 65535)
1597 {
1598 *errorcodeptr = ERR5;
1599 return p;
1600 }
1601 }
1602
1603 if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1604 {
1605 if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1606 {
1607 max = 0;
1608 while(IS_DIGIT(*p))
1609 {
1610 max = max * 10 + (int)(*p++ - CHAR_0);
1611 if (max > 65535)
1612 {
1613 *errorcodeptr = ERR5;
1614 return p;
1615 }
1616 }
1617 if (max < min)
1618 {
1619 *errorcodeptr = ERR4;
1620 return p;
1621 }
1622 }
1623 }
1624
1625 *minp = min;
1626 *maxp = max;
1627 return p;
1628 }
1629
1630
1631
1632 /*************************************************
1633 * Find first significant op code *
1634 *************************************************/
1635
1636 /* This is called by several functions that scan a compiled expression looking
1637 for a fixed first character, or an anchoring op code etc. It skips over things
1638 that do not influence this. For some calls, it makes sense to skip negative
1639 forward and all backward assertions, and also the \b assertion; for others it
1640 does not.
1641
1642 Arguments:
1643 code pointer to the start of the group
1644 skipassert TRUE if certain assertions are to be skipped
1645
1646 Returns: pointer to the first significant opcode
1647 */
1648
1649 static const pcre_uchar*
1650 first_significant_code(const pcre_uchar *code, BOOL skipassert)
1651 {
1652 for (;;)
1653 {
1654 switch ((int)*code)
1655 {
1656 case OP_ASSERT_NOT:
1657 case OP_ASSERTBACK:
1658 case OP_ASSERTBACK_NOT:
1659 if (!skipassert) return code;
1660 do code += GET(code, 1); while (*code == OP_ALT);
1661 code += PRIV(OP_lengths)[*code];
1662 break;
1663
1664 case OP_WORD_BOUNDARY:
1665 case OP_NOT_WORD_BOUNDARY:
1666 if (!skipassert) return code;
1667 /* Fall through */
1668
1669 case OP_CALLOUT:
1670 case OP_CREF:
1671 case OP_DNCREF:
1672 case OP_RREF:
1673 case OP_DNRREF:
1674 case OP_DEF:
1675 code += PRIV(OP_lengths)[*code];
1676 break;
1677
1678 default:
1679 return code;
1680 }
1681 }
1682 /* Control never reaches here */
1683 }
1684
1685
1686
1687 /*************************************************
1688 * Find the fixed length of a branch *
1689 *************************************************/
1690
1691 /* Scan a branch and compute the fixed length of subject that will match it,
1692 if the length is fixed. This is needed for dealing with backward assertions.
1693 In UTF8 mode, the result is in characters rather than bytes. The branch is
1694 temporarily terminated with OP_END when this function is called.
1695
1696 This function is called when a backward assertion is encountered, so that if it
1697 fails, the error message can point to the correct place in the pattern.
1698 However, we cannot do this when the assertion contains subroutine calls,
1699 because they can be forward references. We solve this by remembering this case
1700 and doing the check at the end; a flag specifies which mode we are running in.
1701
1702 Arguments:
1703 code points to the start of the pattern (the bracket)
1704 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1705 atend TRUE if called when the pattern is complete
1706 cd the "compile data" structure
1707
1708 Returns: the fixed length,
1709 or -1 if there is no fixed length,
1710 or -2 if \C was encountered (in UTF-8 mode only)
1711 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1712 or -4 if an unknown opcode was encountered (internal error)
1713 */
1714
1715 static int
1716 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1717 {
1718 int length = -1;
1719
1720 register int branchlength = 0;
1721 register pcre_uchar *cc = code + 1 + LINK_SIZE;
1722
1723 /* Scan along the opcodes for this branch. If we get to the end of the
1724 branch, check the length against that of the other branches. */
1725
1726 for (;;)
1727 {
1728 int d;
1729 pcre_uchar *ce, *cs;
1730 register pcre_uchar op = *cc;
1731
1732 switch (op)
1733 {
1734 /* We only need to continue for OP_CBRA (normal capturing bracket) and
1735 OP_BRA (normal non-capturing bracket) because the other variants of these
1736 opcodes are all concerned with unlimited repeated groups, which of course
1737 are not of fixed length. */
1738
1739 case OP_CBRA:
1740 case OP_BRA:
1741 case OP_ONCE:
1742 case OP_ONCE_NC:
1743 case OP_COND:
1744 d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1745 if (d < 0) return d;
1746 branchlength += d;
1747 do cc += GET(cc, 1); while (*cc == OP_ALT);
1748 cc += 1 + LINK_SIZE;
1749 break;
1750
1751 /* Reached end of a branch; if it's a ket it is the end of a nested call.
1752 If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1753 an ALT. If it is END it's the end of the outer call. All can be handled by
1754 the same code. Note that we must not include the OP_KETRxxx opcodes here,
1755 because they all imply an unlimited repeat. */
1756
1757 case OP_ALT:
1758 case OP_KET:
1759 case OP_END:
1760 case OP_ACCEPT:
1761 case OP_ASSERT_ACCEPT:
1762 if (length < 0) length = branchlength;
1763 else if (length != branchlength) return -1;
1764 if (*cc != OP_ALT) return length;
1765 cc += 1 + LINK_SIZE;
1766 branchlength = 0;
1767 break;
1768
1769 /* A true recursion implies not fixed length, but a subroutine call may
1770 be OK. If the subroutine is a forward reference, we can't deal with
1771 it until the end of the pattern, so return -3. */
1772
1773 case OP_RECURSE:
1774 if (!atend) return -3;
1775 cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1776 do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1777 if (cc > cs && cc < ce) return -1; /* Recursion */
1778 d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1779 if (d < 0) return d;
1780 branchlength += d;
1781 cc += 1 + LINK_SIZE;
1782 break;
1783
1784 /* Skip over assertive subpatterns */
1785
1786 case OP_ASSERT:
1787 case OP_ASSERT_NOT:
1788 case OP_ASSERTBACK:
1789 case OP_ASSERTBACK_NOT:
1790 do cc += GET(cc, 1); while (*cc == OP_ALT);
1791 cc += PRIV(OP_lengths)[*cc];
1792 break;
1793
1794 /* Skip over things that don't match chars */
1795
1796 case OP_MARK:
1797 case OP_PRUNE_ARG:
1798 case OP_SKIP_ARG:
1799 case OP_THEN_ARG:
1800 cc += cc[1] + PRIV(OP_lengths)[*cc];
1801 break;
1802
1803 case OP_CALLOUT:
1804 case OP_CIRC:
1805 case OP_CIRCM:
1806 case OP_CLOSE:
1807 case OP_COMMIT:
1808 case OP_CREF:
1809 case OP_DEF:
1810 case OP_DNCREF:
1811 case OP_DNRREF:
1812 case OP_DOLL:
1813 case OP_DOLLM:
1814 case OP_EOD:
1815 case OP_EODN:
1816 case OP_FAIL:
1817 case OP_NOT_WORD_BOUNDARY:
1818 case OP_PRUNE:
1819 case OP_REVERSE:
1820 case OP_RREF:
1821 case OP_SET_SOM:
1822 case OP_SKIP:
1823 case OP_SOD:
1824 case OP_SOM:
1825 case OP_THEN:
1826 case OP_WORD_BOUNDARY:
1827 cc += PRIV(OP_lengths)[*cc];
1828 break;
1829
1830 /* Handle literal characters */
1831
1832 case OP_CHAR:
1833 case OP_CHARI:
1834 case OP_NOT:
1835 case OP_NOTI:
1836 branchlength++;
1837 cc += 2;
1838 #ifdef SUPPORT_UTF
1839 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1840 #endif
1841 break;
1842
1843 /* Handle exact repetitions. The count is already in characters, but we
1844 need to skip over a multibyte character in UTF8 mode. */
1845
1846 case OP_EXACT:
1847 case OP_EXACTI:
1848 case OP_NOTEXACT:
1849 case OP_NOTEXACTI:
1850 branchlength += (int)GET2(cc,1);
1851 cc += 2 + IMM2_SIZE;
1852 #ifdef SUPPORT_UTF
1853 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1854 #endif
1855 break;
1856
1857 case OP_TYPEEXACT:
1858 branchlength += GET2(cc,1);
1859 if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1860 cc += 2;
1861 cc += 1 + IMM2_SIZE + 1;
1862 break;
1863
1864 /* Handle single-char matchers */
1865
1866 case OP_PROP:
1867 case OP_NOTPROP:
1868 cc += 2;
1869 /* Fall through */
1870
1871 case OP_HSPACE:
1872 case OP_VSPACE:
1873 case OP_NOT_HSPACE:
1874 case OP_NOT_VSPACE:
1875 case OP_NOT_DIGIT:
1876 case OP_DIGIT:
1877 case OP_NOT_WHITESPACE:
1878 case OP_WHITESPACE:
1879 case OP_NOT_WORDCHAR:
1880 case OP_WORDCHAR:
1881 case OP_ANY:
1882 case OP_ALLANY:
1883 branchlength++;
1884 cc++;
1885 break;
1886
1887 /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1888 otherwise \C is coded as OP_ALLANY. */
1889
1890 case OP_ANYBYTE:
1891 return -2;
1892
1893 /* Check a class for variable quantification */
1894
1895 case OP_CLASS:
1896 case OP_NCLASS:
1897 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1898 case OP_XCLASS:
1899 /* The original code caused an unsigned overflow in 64 bit systems,
1900 so now we use a conditional statement. */
1901 if (op == OP_XCLASS)
1902 cc += GET(cc, 1);
1903 else
1904 cc += PRIV(OP_lengths)[OP_CLASS];
1905 #else
1906 cc += PRIV(OP_lengths)[OP_CLASS];
1907 #endif
1908
1909 switch (*cc)
1910 {
1911 case OP_CRSTAR:
1912 case OP_CRMINSTAR:
1913 case OP_CRPLUS:
1914 case OP_CRMINPLUS:
1915 case OP_CRQUERY:
1916 case OP_CRMINQUERY:
1917 case OP_CRPOSSTAR:
1918 case OP_CRPOSPLUS:
1919 case OP_CRPOSQUERY:
1920 return -1;
1921
1922 case OP_CRRANGE:
1923 case OP_CRMINRANGE:
1924 case OP_CRPOSRANGE:
1925 if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1926 branchlength += (int)GET2(cc,1);
1927 cc += 1 + 2 * IMM2_SIZE;
1928 break;
1929
1930 default:
1931 branchlength++;
1932 }
1933 break;
1934
1935 /* Anything else is variable length */
1936
1937 case OP_ANYNL:
1938 case OP_BRAMINZERO:
1939 case OP_BRAPOS:
1940 case OP_BRAPOSZERO:
1941 case OP_BRAZERO:
1942 case OP_CBRAPOS:
1943 case OP_EXTUNI:
1944 case OP_KETRMAX:
1945 case OP_KETRMIN:
1946 case OP_KETRPOS:
1947 case OP_MINPLUS:
1948 case OP_MINPLUSI:
1949 case OP_MINQUERY:
1950 case OP_MINQUERYI:
1951 case OP_MINSTAR:
1952 case OP_MINSTARI:
1953 case OP_MINUPTO:
1954 case OP_MINUPTOI:
1955 case OP_NOTMINPLUS:
1956 case OP_NOTMINPLUSI:
1957 case OP_NOTMINQUERY:
1958 case OP_NOTMINQUERYI:
1959 case OP_NOTMINSTAR:
1960 case OP_NOTMINSTARI:
1961 case OP_NOTMINUPTO:
1962 case OP_NOTMINUPTOI:
1963 case OP_NOTPLUS:
1964 case OP_NOTPLUSI:
1965 case OP_NOTPOSPLUS:
1966 case OP_NOTPOSPLUSI:
1967 case OP_NOTPOSQUERY:
1968 case OP_NOTPOSQUERYI:
1969 case OP_NOTPOSSTAR:
1970 case OP_NOTPOSSTARI:
1971 case OP_NOTPOSUPTO:
1972 case OP_NOTPOSUPTOI:
1973 case OP_NOTQUERY:
1974 case OP_NOTQUERYI:
1975 case OP_NOTSTAR:
1976 case OP_NOTSTARI:
1977 case OP_NOTUPTO:
1978 case OP_NOTUPTOI:
1979 case OP_PLUS:
1980 case OP_PLUSI:
1981 case OP_POSPLUS:
1982 case OP_POSPLUSI:
1983 case OP_POSQUERY:
1984 case OP_POSQUERYI:
1985 case OP_POSSTAR:
1986 case OP_POSSTARI:
1987 case OP_POSUPTO:
1988 case OP_POSUPTOI:
1989 case OP_QUERY:
1990 case OP_QUERYI:
1991 case OP_REF:
1992 case OP_REFI:
1993 case OP_DNREF:
1994 case OP_DNREFI:
1995 case OP_SBRA:
1996 case OP_SBRAPOS:
1997 case OP_SCBRA:
1998 case OP_SCBRAPOS:
1999 case OP_SCOND:
2000 case OP_SKIPZERO:
2001 case OP_STAR:
2002 case OP_STARI:
2003 case OP_TYPEMINPLUS:
2004 case OP_TYPEMINQUERY:
2005 case OP_TYPEMINSTAR:
2006 case OP_TYPEMINUPTO:
2007 case OP_TYPEPLUS:
2008 case OP_TYPEPOSPLUS:
2009 case OP_TYPEPOSQUERY:
2010 case OP_TYPEPOSSTAR:
2011 case OP_TYPEPOSUPTO:
2012 case OP_TYPEQUERY:
2013 case OP_TYPESTAR:
2014 case OP_TYPEUPTO:
2015 case OP_UPTO:
2016 case OP_UPTOI:
2017 return -1;
2018
2019 /* Catch unrecognized opcodes so that when new ones are added they
2020 are not forgotten, as has happened in the past. */
2021
2022 default:
2023 return -4;
2024 }
2025 }
2026 /* Control never gets here */
2027 }
2028
2029
2030
2031 /*************************************************
2032 * Scan compiled regex for specific bracket *
2033 *************************************************/
2034
2035 /* This little function scans through a compiled pattern until it finds a
2036 capturing bracket with the given number, or, if the number is negative, an
2037 instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2038 so that it can be called from pcre_study() when finding the minimum matching
2039 length.
2040
2041 Arguments:
2042 code points to start of expression
2043 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2044 number the required bracket number or negative to find a lookbehind
2045
2046 Returns: pointer to the opcode for the bracket, or NULL if not found
2047 */
2048
2049 const pcre_uchar *
2050 PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2051 {
2052 for (;;)
2053 {
2054 register pcre_uchar c = *code;
2055
2056 if (c == OP_END) return NULL;
2057
2058 /* XCLASS is used for classes that cannot be represented just by a bit
2059 map. This includes negated single high-valued characters. The length in
2060 the table is zero; the actual length is stored in the compiled code. */
2061
2062 if (c == OP_XCLASS) code += GET(code, 1);
2063
2064 /* Handle recursion */
2065
2066 else if (c == OP_REVERSE)
2067 {
2068 if (number < 0) return (pcre_uchar *)code;
2069 code += PRIV(OP_lengths)[c];
2070 }
2071
2072 /* Handle capturing bracket */
2073
2074 else if (c == OP_CBRA || c == OP_SCBRA ||
2075 c == OP_CBRAPOS || c == OP_SCBRAPOS)
2076 {
2077 int n = (int)GET2(code, 1+LINK_SIZE);
2078 if (n == number) return (pcre_uchar *)code;
2079 code += PRIV(OP_lengths)[c];
2080 }
2081
2082 /* Otherwise, we can get the item's length from the table, except that for
2083 repeated character types, we have to test for \p and \P, which have an extra
2084 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2085 must add in its length. */
2086
2087 else
2088 {
2089 switch(c)
2090 {
2091 case OP_TYPESTAR:
2092 case OP_TYPEMINSTAR:
2093 case OP_TYPEPLUS:
2094 case OP_TYPEMINPLUS:
2095 case OP_TYPEQUERY:
2096 case OP_TYPEMINQUERY:
2097 case OP_TYPEPOSSTAR:
2098 case OP_TYPEPOSPLUS:
2099 case OP_TYPEPOSQUERY:
2100 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2101 break;
2102
2103 case OP_TYPEUPTO:
2104 case OP_TYPEMINUPTO:
2105 case OP_TYPEEXACT:
2106 case OP_TYPEPOSUPTO:
2107 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2108 code += 2;
2109 break;
2110
2111 case OP_MARK:
2112 case OP_PRUNE_ARG:
2113 case OP_SKIP_ARG:
2114 case OP_THEN_ARG:
2115 code += code[1];
2116 break;
2117 }
2118
2119 /* Add in the fixed length from the table */
2120
2121 code += PRIV(OP_lengths)[c];
2122
2123 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2124 a multi-byte character. The length in the table is a minimum, so we have to
2125 arrange to skip the extra bytes. */
2126
2127 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2128 if (utf) switch(c)
2129 {
2130 case OP_CHAR:
2131 case OP_CHARI:
2132 case OP_NOT:
2133 case OP_NOTI:
2134 case OP_EXACT:
2135 case OP_EXACTI:
2136 case OP_NOTEXACT:
2137 case OP_NOTEXACTI:
2138 case OP_UPTO:
2139 case OP_UPTOI:
2140 case OP_NOTUPTO:
2141 case OP_NOTUPTOI:
2142 case OP_MINUPTO:
2143 case OP_MINUPTOI:
2144 case OP_NOTMINUPTO:
2145 case OP_NOTMINUPTOI:
2146 case OP_POSUPTO:
2147 case OP_POSUPTOI:
2148 case OP_NOTPOSUPTO:
2149 case OP_NOTPOSUPTOI:
2150 case OP_STAR:
2151 case OP_STARI:
2152 case OP_NOTSTAR:
2153 case OP_NOTSTARI:
2154 case OP_MINSTAR:
2155 case OP_MINSTARI:
2156 case OP_NOTMINSTAR:
2157 case OP_NOTMINSTARI:
2158 case OP_POSSTAR:
2159 case OP_POSSTARI:
2160 case OP_NOTPOSSTAR:
2161 case OP_NOTPOSSTARI:
2162 case OP_PLUS:
2163 case OP_PLUSI:
2164 case OP_NOTPLUS:
2165 case OP_NOTPLUSI:
2166 case OP_MINPLUS:
2167 case OP_MINPLUSI:
2168 case OP_NOTMINPLUS:
2169 case OP_NOTMINPLUSI:
2170 case OP_POSPLUS:
2171 case OP_POSPLUSI:
2172 case OP_NOTPOSPLUS:
2173 case OP_NOTPOSPLUSI:
2174 case OP_QUERY:
2175 case OP_QUERYI:
2176 case OP_NOTQUERY:
2177 case OP_NOTQUERYI:
2178 case OP_MINQUERY:
2179 case OP_MINQUERYI:
2180 case OP_NOTMINQUERY:
2181 case OP_NOTMINQUERYI:
2182 case OP_POSQUERY:
2183 case OP_POSQUERYI:
2184 case OP_NOTPOSQUERY:
2185 case OP_NOTPOSQUERYI:
2186 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2187 break;
2188 }
2189 #else
2190 (void)(utf); /* Keep compiler happy by referencing function argument */
2191 #endif
2192 }
2193 }
2194 }
2195
2196
2197
2198 /*************************************************
2199 * Scan compiled regex for recursion reference *
2200 *************************************************/
2201
2202 /* This little function scans through a compiled pattern until it finds an
2203 instance of OP_RECURSE.
2204
2205 Arguments:
2206 code points to start of expression
2207 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2208
2209 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2210 */
2211
2212 static const pcre_uchar *
2213 find_recurse(const pcre_uchar *code, BOOL utf)
2214 {
2215 for (;;)
2216 {
2217 register pcre_uchar c = *code;
2218 if (c == OP_END) return NULL;
2219 if (c == OP_RECURSE) return code;
2220
2221 /* XCLASS is used for classes that cannot be represented just by a bit
2222 map. This includes negated single high-valued characters. The length in
2223 the table is zero; the actual length is stored in the compiled code. */
2224
2225 if (c == OP_XCLASS) code += GET(code, 1);
2226
2227 /* Otherwise, we can get the item's length from the table, except that for
2228 repeated character types, we have to test for \p and \P, which have an extra
2229 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2230 must add in its length. */
2231
2232 else
2233 {
2234 switch(c)
2235 {
2236 case OP_TYPESTAR:
2237 case OP_TYPEMINSTAR:
2238 case OP_TYPEPLUS:
2239 case OP_TYPEMINPLUS:
2240 case OP_TYPEQUERY:
2241 case OP_TYPEMINQUERY:
2242 case OP_TYPEPOSSTAR:
2243 case OP_TYPEPOSPLUS:
2244 case OP_TYPEPOSQUERY:
2245 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2246 break;
2247
2248 case OP_TYPEPOSUPTO:
2249 case OP_TYPEUPTO:
2250 case OP_TYPEMINUPTO:
2251 case OP_TYPEEXACT:
2252 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2253 code += 2;
2254 break;
2255
2256 case OP_MARK:
2257 case OP_PRUNE_ARG:
2258 case OP_SKIP_ARG:
2259 case OP_THEN_ARG:
2260 code += code[1];
2261 break;
2262 }
2263
2264 /* Add in the fixed length from the table */
2265
2266 code += PRIV(OP_lengths)[c];
2267
2268 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2269 by a multi-byte character. The length in the table is a minimum, so we have
2270 to arrange to skip the extra bytes. */
2271
2272 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2273 if (utf) switch(c)
2274 {
2275 case OP_CHAR:
2276 case OP_CHARI:
2277 case OP_NOT:
2278 case OP_NOTI:
2279 case OP_EXACT:
2280 case OP_EXACTI:
2281 case OP_NOTEXACT:
2282 case OP_NOTEXACTI:
2283 case OP_UPTO:
2284 case OP_UPTOI:
2285 case OP_NOTUPTO:
2286 case OP_NOTUPTOI:
2287 case OP_MINUPTO:
2288 case OP_MINUPTOI:
2289 case OP_NOTMINUPTO:
2290 case OP_NOTMINUPTOI:
2291 case OP_POSUPTO:
2292 case OP_POSUPTOI:
2293 case OP_NOTPOSUPTO:
2294 case OP_NOTPOSUPTOI:
2295 case OP_STAR:
2296 case OP_STARI:
2297 case OP_NOTSTAR:
2298 case OP_NOTSTARI:
2299 case OP_MINSTAR:
2300 case OP_MINSTARI:
2301 case OP_NOTMINSTAR:
2302 case OP_NOTMINSTARI:
2303 case OP_POSSTAR:
2304 case OP_POSSTARI:
2305 case OP_NOTPOSSTAR:
2306 case OP_NOTPOSSTARI:
2307 case OP_PLUS:
2308 case OP_PLUSI:
2309 case OP_NOTPLUS:
2310 case OP_NOTPLUSI:
2311 case OP_MINPLUS:
2312 case OP_MINPLUSI:
2313 case OP_NOTMINPLUS:
2314 case OP_NOTMINPLUSI:
2315 case OP_POSPLUS:
2316 case OP_POSPLUSI:
2317 case OP_NOTPOSPLUS:
2318 case OP_NOTPOSPLUSI:
2319 case OP_QUERY:
2320 case OP_QUERYI:
2321 case OP_NOTQUERY:
2322 case OP_NOTQUERYI:
2323 case OP_MINQUERY:
2324 case OP_MINQUERYI:
2325 case OP_NOTMINQUERY:
2326 case OP_NOTMINQUERYI:
2327 case OP_POSQUERY:
2328 case OP_POSQUERYI:
2329 case OP_NOTPOSQUERY:
2330 case OP_NOTPOSQUERYI:
2331 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2332 break;
2333 }
2334 #else
2335 (void)(utf); /* Keep compiler happy by referencing function argument */
2336 #endif
2337 }
2338 }
2339 }
2340
2341
2342
2343 /*************************************************
2344 * Scan compiled branch for non-emptiness *
2345 *************************************************/
2346
2347 /* This function scans through a branch of a compiled pattern to see whether it
2348 can match the empty string or not. It is called from could_be_empty()
2349 below and from compile_branch() when checking for an unlimited repeat of a
2350 group that can match nothing. Note that first_significant_code() skips over
2351 backward and negative forward assertions when its final argument is TRUE. If we
2352 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2353 bracket whose current branch will already have been scanned.
2354
2355 Arguments:
2356 code points to start of search
2357 endcode points to where to stop
2358 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2359 cd contains pointers to tables etc.
2360 recurses chain of recurse_check to catch mutual recursion
2361
2362 Returns: TRUE if what is matched could be empty
2363 */
2364
2365 typedef struct recurse_check {
2366 struct recurse_check *prev;
2367 const pcre_uchar *group;
2368 } recurse_check;
2369
2370 static BOOL
2371 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2372 BOOL utf, compile_data *cd, recurse_check *recurses)
2373 {
2374 register pcre_uchar c;
2375 recurse_check this_recurse;
2376
2377 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2378 code < endcode;
2379 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2380 {
2381 const pcre_uchar *ccode;
2382
2383 c = *code;
2384
2385 /* Skip over forward assertions; the other assertions are skipped by
2386 first_significant_code() with a TRUE final argument. */
2387
2388 if (c == OP_ASSERT)
2389 {
2390 do code += GET(code, 1); while (*code == OP_ALT);
2391 c = *code;
2392 continue;
2393 }
2394
2395 /* For a recursion/subroutine call, if its end has been reached, which
2396 implies a backward reference subroutine call, we can scan it. If it's a
2397 forward reference subroutine call, we can't. To detect forward reference
2398 we have to scan up the list that is kept in the workspace. This function is
2399 called only when doing the real compile, not during the pre-compile that
2400 measures the size of the compiled pattern. */
2401
2402 if (c == OP_RECURSE)
2403 {
2404 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2405 const pcre_uchar *endgroup = scode;
2406 BOOL empty_branch;
2407
2408 /* Test for forward reference or uncompleted reference. This is disabled
2409 when called to scan a completed pattern by setting cd->start_workspace to
2410 NULL. */
2411
2412 if (cd->start_workspace != NULL)
2413 {
2414 const pcre_uchar *tcode;
2415 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2416 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2417 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2418 }
2419
2420 /* If the reference is to a completed group, we need to detect whether this
2421 is a recursive call, as otherwise there will be an infinite loop. If it is
2422 a recursion, just skip over it. Simple recursions are easily detected. For
2423 mutual recursions we keep a chain on the stack. */
2424
2425 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2426 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2427 else
2428 {
2429 recurse_check *r = recurses;
2430 for (r = recurses; r != NULL; r = r->prev)
2431 if (r->group == scode) break;
2432 if (r != NULL) continue; /* Mutual recursion */
2433 }
2434
2435 /* Completed reference; scan the referenced group, remembering it on the
2436 stack chain to detect mutual recursions. */
2437
2438 empty_branch = FALSE;
2439 this_recurse.prev = recurses;
2440 this_recurse.group = scode;
2441
2442 do
2443 {
2444 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2445 {
2446 empty_branch = TRUE;
2447 break;
2448 }
2449 scode += GET(scode, 1);
2450 }
2451 while (*scode == OP_ALT);
2452
2453 if (!empty_branch) return FALSE; /* All branches are non-empty */
2454 continue;
2455 }
2456
2457 /* Groups with zero repeats can of course be empty; skip them. */
2458
2459 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2460 c == OP_BRAPOSZERO)
2461 {
2462 code += PRIV(OP_lengths)[c];
2463 do code += GET(code, 1); while (*code == OP_ALT);
2464 c = *code;
2465 continue;
2466 }
2467
2468 /* A nested group that is already marked as "could be empty" can just be
2469 skipped. */
2470
2471 if (c == OP_SBRA || c == OP_SBRAPOS ||
2472 c == OP_SCBRA || c == OP_SCBRAPOS)
2473 {
2474 do code += GET(code, 1); while (*code == OP_ALT);
2475 c = *code;
2476 continue;
2477 }
2478
2479 /* For other groups, scan the branches. */
2480
2481 if (c == OP_BRA || c == OP_BRAPOS ||
2482 c == OP_CBRA || c == OP_CBRAPOS ||
2483 c == OP_ONCE || c == OP_ONCE_NC ||
2484 c == OP_COND)
2485 {
2486 BOOL empty_branch;
2487 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2488
2489 /* If a conditional group has only one branch, there is a second, implied,
2490 empty branch, so just skip over the conditional, because it could be empty.
2491 Otherwise, scan the individual branches of the group. */
2492
2493 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2494 code += GET(code, 1);
2495 else
2496 {
2497 empty_branch = FALSE;
2498 do
2499 {
2500 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2501 empty_branch = TRUE;
2502 code += GET(code, 1);
2503 }
2504 while (*code == OP_ALT);
2505 if (!empty_branch) return FALSE; /* All branches are non-empty */
2506 }
2507
2508 c = *code;
2509 continue;
2510 }
2511
2512 /* Handle the other opcodes */
2513
2514 switch (c)
2515 {
2516 /* Check for quantifiers after a class. XCLASS is used for classes that
2517 cannot be represented just by a bit map. This includes negated single
2518 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2519 actual length is stored in the compiled code, so we must update "code"
2520 here. */
2521
2522 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2523 case OP_XCLASS:
2524 ccode = code += GET(code, 1);
2525 goto CHECK_CLASS_REPEAT;
2526 #endif
2527
2528 case OP_CLASS:
2529 case OP_NCLASS:
2530 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2531
2532 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2533 CHECK_CLASS_REPEAT:
2534 #endif
2535
2536 switch (*ccode)
2537 {
2538 case OP_CRSTAR: /* These could be empty; continue */
2539 case OP_CRMINSTAR:
2540 case OP_CRQUERY:
2541 case OP_CRMINQUERY:
2542 case OP_CRPOSSTAR:
2543 case OP_CRPOSQUERY:
2544 break;
2545
2546 default: /* Non-repeat => class must match */
2547 case OP_CRPLUS: /* These repeats aren't empty */
2548 case OP_CRMINPLUS:
2549 case OP_CRPOSPLUS:
2550 return FALSE;
2551
2552 case OP_CRRANGE:
2553 case OP_CRMINRANGE:
2554 case OP_CRPOSRANGE:
2555 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2556 break;
2557 }
2558 break;
2559
2560 /* Opcodes that must match a character */
2561
2562 case OP_ANY:
2563 case OP_ALLANY:
2564 case OP_ANYBYTE:
2565
2566 case OP_PROP:
2567 case OP_NOTPROP:
2568 case OP_ANYNL:
2569
2570 case OP_NOT_HSPACE:
2571 case OP_HSPACE:
2572 case OP_NOT_VSPACE:
2573 case OP_VSPACE:
2574 case OP_EXTUNI:
2575
2576 case OP_NOT_DIGIT:
2577 case OP_DIGIT:
2578 case OP_NOT_WHITESPACE:
2579 case OP_WHITESPACE:
2580 case OP_NOT_WORDCHAR:
2581 case OP_WORDCHAR:
2582
2583 case OP_CHAR:
2584 case OP_CHARI:
2585 case OP_NOT:
2586 case OP_NOTI:
2587
2588 case OP_PLUS:
2589 case OP_PLUSI:
2590 case OP_MINPLUS:
2591 case OP_MINPLUSI:
2592
2593 case OP_NOTPLUS:
2594 case OP_NOTPLUSI:
2595 case OP_NOTMINPLUS:
2596 case OP_NOTMINPLUSI:
2597
2598 case OP_POSPLUS:
2599 case OP_POSPLUSI:
2600 case OP_NOTPOSPLUS:
2601 case OP_NOTPOSPLUSI:
2602
2603 case OP_EXACT:
2604 case OP_EXACTI:
2605 case OP_NOTEXACT:
2606 case OP_NOTEXACTI:
2607
2608 case OP_TYPEPLUS:
2609 case OP_TYPEMINPLUS:
2610 case OP_TYPEPOSPLUS:
2611 case OP_TYPEEXACT:
2612
2613 return FALSE;
2614
2615 /* These are going to continue, as they may be empty, but we have to
2616 fudge the length for the \p and \P cases. */
2617
2618 case OP_TYPESTAR:
2619 case OP_TYPEMINSTAR:
2620 case OP_TYPEPOSSTAR:
2621 case OP_TYPEQUERY:
2622 case OP_TYPEMINQUERY:
2623 case OP_TYPEPOSQUERY:
2624 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2625 break;
2626
2627 /* Same for these */
2628
2629 case OP_TYPEUPTO:
2630 case OP_TYPEMINUPTO:
2631 case OP_TYPEPOSUPTO:
2632 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2633 code += 2;
2634 break;
2635
2636 /* End of branch */
2637
2638 case OP_KET:
2639 case OP_KETRMAX:
2640 case OP_KETRMIN:
2641 case OP_KETRPOS:
2642 case OP_ALT:
2643 return TRUE;
2644
2645 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2646 MINUPTO, and POSUPTO and their caseless and negative versions may be
2647 followed by a multibyte character. */
2648
2649 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2650 case OP_STAR:
2651 case OP_STARI:
2652 case OP_NOTSTAR:
2653 case OP_NOTSTARI:
2654
2655 case OP_MINSTAR:
2656 case OP_MINSTARI:
2657 case OP_NOTMINSTAR:
2658 case OP_NOTMINSTARI:
2659
2660 case OP_POSSTAR:
2661 case OP_POSSTARI:
2662 case OP_NOTPOSSTAR:
2663 case OP_NOTPOSSTARI:
2664
2665 case OP_QUERY:
2666 case OP_QUERYI:
2667 case OP_NOTQUERY:
2668 case OP_NOTQUERYI:
2669
2670 case OP_MINQUERY:
2671 case OP_MINQUERYI:
2672 case OP_NOTMINQUERY:
2673 case OP_NOTMINQUERYI:
2674
2675 case OP_POSQUERY:
2676 case OP_POSQUERYI:
2677 case OP_NOTPOSQUERY:
2678 case OP_NOTPOSQUERYI:
2679
2680 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2681 break;
2682
2683 case OP_UPTO:
2684 case OP_UPTOI:
2685 case OP_NOTUPTO:
2686 case OP_NOTUPTOI:
2687
2688 case OP_MINUPTO:
2689 case OP_MINUPTOI:
2690 case OP_NOTMINUPTO:
2691 case OP_NOTMINUPTOI:
2692
2693 case OP_POSUPTO:
2694 case OP_POSUPTOI:
2695 case OP_NOTPOSUPTO:
2696 case OP_NOTPOSUPTOI:
2697
2698 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2699 break;
2700 #endif
2701
2702 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2703 string. */
2704
2705 case OP_MARK:
2706 case OP_PRUNE_ARG:
2707 case OP_SKIP_ARG:
2708 case OP_THEN_ARG:
2709 code += code[1];
2710 break;
2711
2712 /* None of the remaining opcodes are required to match a character. */
2713
2714 default:
2715 break;
2716 }
2717 }
2718
2719 return TRUE;
2720 }
2721
2722
2723
2724 /*************************************************
2725 * Scan compiled regex for non-emptiness *
2726 *************************************************/
2727
2728 /* This function is called to check for left recursive calls. We want to check
2729 the current branch of the current pattern to see if it could match the empty
2730 string. If it could, we must look outwards for branches at other levels,
2731 stopping when we pass beyond the bracket which is the subject of the recursion.
2732 This function is called only during the real compile, not during the
2733 pre-compile.
2734
2735 Arguments:
2736 code points to start of the recursion
2737 endcode points to where to stop (current RECURSE item)
2738 bcptr points to the chain of current (unclosed) branch starts
2739 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2740 cd pointers to tables etc
2741
2742 Returns: TRUE if what is matched could be empty
2743 */
2744
2745 static BOOL
2746 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2747 branch_chain *bcptr, BOOL utf, compile_data *cd)
2748 {
2749 while (bcptr != NULL && bcptr->current_branch >= code)
2750 {
2751 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2752 return FALSE;
2753 bcptr = bcptr->outer;
2754 }
2755 return TRUE;
2756 }
2757
2758
2759
2760 /*************************************************
2761 * Base opcode of repeated opcodes *
2762 *************************************************/
2763
2764 /* Returns the base opcode for repeated single character type opcodes. If the
2765 opcode is not a repeated character type, it returns with the original value.
2766
2767 Arguments: c opcode
2768 Returns: base opcode for the type
2769 */
2770
2771 static pcre_uchar
2772 get_repeat_base(pcre_uchar c)
2773 {
2774 return (c > OP_TYPEPOSUPTO)? c :
2775 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2776 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2777 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2778 (c >= OP_STARI)? OP_STARI :
2779 OP_STAR;
2780 }
2781
2782
2783
2784 #ifdef SUPPORT_UCP
2785 /*************************************************
2786 * Check a character and a property *
2787 *************************************************/
2788
2789 /* This function is called by check_auto_possessive() when a property item
2790 is adjacent to a fixed character.
2791
2792 Arguments:
2793 c the character
2794 ptype the property type
2795 pdata the data for the type
2796 negated TRUE if it's a negated property (\P or \p{^)
2797
2798 Returns: TRUE if auto-possessifying is OK
2799 */
2800
2801 static BOOL
2802 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2803 BOOL negated)
2804 {
2805 const pcre_uint32 *p;
2806 const ucd_record *prop = GET_UCD(c);
2807
2808 switch(ptype)
2809 {
2810 case PT_LAMP:
2811 return (prop->chartype == ucp_Lu ||
2812 prop->chartype == ucp_Ll ||
2813 prop->chartype == ucp_Lt) == negated;
2814
2815 case PT_GC:
2816 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2817
2818 case PT_PC:
2819 return (pdata == prop->chartype) == negated;
2820
2821 case PT_SC:
2822 return (pdata == prop->script) == negated;
2823
2824 /* These are specials */
2825
2826 case PT_ALNUM:
2827 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2828 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2829
2830 /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2831 means that Perl space and POSIX space are now identical. PCRE was changed
2832 at release 8.34. */
2833
2834 case PT_SPACE: /* Perl space */
2835 case PT_PXSPACE: /* POSIX space */
2836 switch(c)
2837 {
2838 HSPACE_CASES:
2839 VSPACE_CASES:
2840 return negated;
2841
2842 default:
2843 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2844 }
2845 break; /* Control never reaches here */
2846
2847 case PT_WORD:
2848 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2849 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2850 c == CHAR_UNDERSCORE) == negated;
2851
2852 case PT_CLIST:
2853 p = PRIV(ucd_caseless_sets) + prop->caseset;
2854 for (;;)
2855 {
2856 if (c < *p) return !negated;
2857 if (c == *p++) return negated;
2858 }
2859 break; /* Control never reaches here */
2860 }
2861
2862 return FALSE;
2863 }
2864 #endif /* SUPPORT_UCP */
2865
2866
2867
2868 /*************************************************
2869 * Fill the character property list *
2870 *************************************************/
2871
2872 /* Checks whether the code points to an opcode that can take part in auto-
2873 possessification, and if so, fills a list with its properties.
2874
2875 Arguments:
2876 code points to start of expression
2877 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2878 fcc points to case-flipping table
2879 list points to output list
2880 list[0] will be filled with the opcode
2881 list[1] will be non-zero if this opcode
2882 can match an empty character string
2883 list[2..7] depends on the opcode
2884
2885 Returns: points to the start of the next opcode if *code is accepted
2886 NULL if *code is not accepted
2887 */
2888
2889 static const pcre_uchar *
2890 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2891 const pcre_uint8 *fcc, pcre_uint32 *list)
2892 {
2893 pcre_uchar c = *code;
2894 pcre_uchar base;
2895 const pcre_uchar *end;
2896 pcre_uint32 chr;
2897
2898 #ifdef SUPPORT_UCP
2899 pcre_uint32 *clist_dest;
2900 const pcre_uint32 *clist_src;
2901 #else
2902 utf = utf; /* Suppress "unused parameter" compiler warning */
2903 #endif
2904
2905 list[0] = c;
2906 list[1] = FALSE;
2907 code++;
2908
2909 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2910 {
2911 base = get_repeat_base(c);
2912 c -= (base - OP_STAR);
2913
2914 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2915 code += IMM2_SIZE;
2916
2917 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2918
2919 switch(base)
2920 {
2921 case OP_STAR:
2922 list[0] = OP_CHAR;
2923 break;
2924
2925 case OP_STARI:
2926 list[0] = OP_CHARI;
2927 break;
2928
2929 case OP_NOTSTAR:
2930 list[0] = OP_NOT;
2931 break;
2932
2933 case OP_NOTSTARI:
2934 list[0] = OP_NOTI;
2935 break;
2936
2937 case OP_TYPESTAR:
2938 list[0] = *code;
2939 code++;
2940 break;
2941 }
2942 c = list[0];
2943 }
2944
2945 switch(c)
2946 {
2947 case OP_NOT_DIGIT:
2948 case OP_DIGIT:
2949 case OP_NOT_WHITESPACE:
2950 case OP_WHITESPACE:
2951 case OP_NOT_WORDCHAR:
2952 case OP_WORDCHAR:
2953 case OP_ANY:
2954 case OP_ALLANY:
2955 case OP_ANYNL:
2956 case OP_NOT_HSPACE:
2957 case OP_HSPACE:
2958 case OP_NOT_VSPACE:
2959 case OP_VSPACE:
2960 case OP_EXTUNI:
2961 case OP_EODN:
2962 case OP_EOD:
2963 case OP_DOLL:
2964 case OP_DOLLM:
2965 return code;
2966
2967 case OP_CHAR:
2968 case OP_NOT:
2969 GETCHARINCTEST(chr, code);
2970 list[2] = chr;
2971 list[3] = NOTACHAR;
2972 return code;
2973
2974 case OP_CHARI:
2975 case OP_NOTI:
2976 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2977 GETCHARINCTEST(chr, code);
2978 list[2] = chr;
2979
2980 #ifdef SUPPORT_UCP
2981 if (chr < 128 || (chr < 256 && !utf))
2982 list[3] = fcc[chr];
2983 else
2984 list[3] = UCD_OTHERCASE(chr);
2985 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2986 list[3] = (chr < 256) ? fcc[chr] : chr;
2987 #else
2988 list[3] = fcc[chr];
2989 #endif
2990
2991 /* The othercase might be the same value. */
2992
2993 if (chr == list[3])
2994 list[3] = NOTACHAR;
2995 else
2996 list[4] = NOTACHAR;
2997 return code;
2998
2999 #ifdef SUPPORT_UCP
3000 case OP_PROP:
3001 case OP_NOTPROP:
3002 if (code[0] != PT_CLIST)
3003 {
3004 list[2] = code[0];
3005 list[3] = code[1];
3006 return code + 2;
3007 }
3008
3009 /* Convert only if we have enough space. */
3010
3011 clist_src = PRIV(ucd_caseless_sets) + code[1];
3012 clist_dest = list + 2;
3013 code += 2;
3014
3015 do {
3016 if (clist_dest >= list + 8)
3017 {
3018 /* Early return if there is not enough space. This should never
3019 happen, since all clists are shorter than 5 character now. */
3020 list[2] = code[0];
3021 list[3] = code[1];
3022 return code;
3023 }
3024 *clist_dest++ = *clist_src;
3025 }
3026 while(*clist_src++ != NOTACHAR);
3027
3028 /* All characters are stored. The terminating NOTACHAR
3029 is copied form the clist itself. */
3030
3031 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3032 return code;
3033 #endif
3034
3035 case OP_NCLASS:
3036 case OP_CLASS:
3037 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3038 case OP_XCLASS:
3039 if (c == OP_XCLASS)
3040 end = code + GET(code, 0) - 1;
3041 else
3042 #endif
3043 end = code + 32 / sizeof(pcre_uchar);
3044
3045 switch(*end)
3046 {
3047 case OP_CRSTAR:
3048 case OP_CRMINSTAR:
3049 case OP_CRQUERY:
3050 case OP_CRMINQUERY:
3051 case OP_CRPOSSTAR:
3052 case OP_CRPOSQUERY:
3053 list[1] = TRUE;
3054 end++;
3055 break;
3056
3057 case OP_CRPLUS:
3058 case OP_CRMINPLUS:
3059 case OP_CRPOSPLUS:
3060 end++;
3061 break;
3062
3063 case OP_CRRANGE:
3064 case OP_CRMINRANGE:
3065 case OP_CRPOSRANGE:
3066 list[1] = (GET2(end, 1) == 0);
3067 end += 1 + 2 * IMM2_SIZE;
3068 break;
3069 }
3070 list[2] = (pcre_uint32)(end - code);
3071 return end;
3072 }
3073 return NULL; /* Opcode not accepted */
3074 }
3075
3076
3077
3078 /*************************************************
3079 * Scan further character sets for match *
3080 *************************************************/
3081
3082 /* Checks whether the base and the current opcode have a common character, in
3083 which case the base cannot be possessified.
3084
3085 Arguments:
3086 code points to the byte code
3087 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3088 cd static compile data
3089 base_list the data list of the base opcode
3090
3091 Returns: TRUE if the auto-possessification is possible
3092 */
3093
3094 static BOOL
3095 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3096 const pcre_uint32 *base_list, const pcre_uchar *base_end)
3097 {
3098 pcre_uchar c;
3099 pcre_uint32 list[8];
3100 const pcre_uint32 *chr_ptr;
3101 const pcre_uint32 *ochr_ptr;
3102 const pcre_uint32 *list_ptr;
3103 const pcre_uchar *next_code;
3104 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3105 const pcre_uchar *xclass_flags;
3106 #endif
3107 const pcre_uint8 *class_bitset;
3108 const pcre_uint8 *set1, *set2, *set_end;
3109 pcre_uint32 chr;
3110 BOOL accepted, invert_bits;
3111 BOOL entered_a_group = FALSE;
3112
3113 /* Note: the base_list[1] contains whether the current opcode has greedy
3114 (represented by a non-zero value) quantifier. This is a different from
3115 other character type lists, which stores here that the character iterator
3116 matches to an empty string (also represented by a non-zero value). */
3117
3118 for(;;)
3119 {
3120 /* All operations move the code pointer forward.
3121 Therefore infinite recursions are not possible. */
3122
3123 c = *code;
3124
3125 /* Skip over callouts */
3126
3127 if (c == OP_CALLOUT)
3128 {
3129 code += PRIV(OP_lengths)[c];
3130 continue;
3131 }
3132
3133 if (c == OP_ALT)
3134 {
3135 do code += GET(code, 1); while (*code == OP_ALT);
3136 c = *code;
3137 }
3138
3139 switch(c)
3140 {
3141 case OP_END:
3142 case OP_KETRPOS:
3143 /* TRUE only in greedy case. The non-greedy case could be replaced by
3144 an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3145 uses more memory, which we cannot get at this stage.) */
3146
3147 return base_list[1] != 0;
3148
3149 case OP_KET:
3150 /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3151 it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3152 cannot be converted to a possessive form. */
3153
3154 if (base_list[1] == 0) return FALSE;
3155
3156 switch(*(code - GET(code, 1)))
3157 {
3158 case OP_ASSERT:
3159 case OP_ASSERT_NOT:
3160 case OP_ASSERTBACK:
3161 case OP_ASSERTBACK_NOT:
3162 case OP_ONCE:
3163 case OP_ONCE_NC:
3164 /* Atomic sub-patterns and assertions can always auto-possessify their
3165 last iterator. However, if the group was entered as a result of checking
3166 a previous iterator, this is not possible. */
3167
3168 return !entered_a_group;
3169 }
3170
3171 code += PRIV(OP_lengths)[c];
3172 continue;
3173
3174 case OP_ONCE:
3175 case OP_ONCE_NC:
3176 case OP_BRA:
3177 case OP_CBRA:
3178 next_code = code + GET(code, 1);
3179 code += PRIV(OP_lengths)[c];
3180
3181 while (*next_code == OP_ALT)
3182 {
3183 if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3184 code = next_code + 1 + LINK_SIZE;
3185 next_code += GET(next_code, 1);
3186 }
3187
3188 entered_a_group = TRUE;
3189 continue;
3190
3191 case OP_BRAZERO:
3192 case OP_BRAMINZERO:
3193
3194 next_code = code + 1;
3195 if (*next_code != OP_BRA && *next_code != OP_CBRA
3196 && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3197
3198 do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3199
3200 /* The bracket content will be checked by the
3201 OP_BRA/OP_CBRA case above. */
3202 next_code += 1 + LINK_SIZE;
3203 if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3204 return FALSE;
3205
3206 code += PRIV(OP_lengths)[c];
3207 continue;
3208
3209 default:
3210 break;
3211 }
3212
3213 /* Check for a supported opcode, and load its properties. */
3214
3215 code = get_chr_property_list(code, utf, cd->fcc, list);
3216 if (code == NULL) return FALSE; /* Unsupported */
3217
3218 /* If either opcode is a small character list, set pointers for comparing
3219 characters from that list with another list, or with a property. */
3220
3221 if (base_list[0] == OP_CHAR)
3222 {
3223 chr_ptr = base_list + 2;
3224 list_ptr = list;
3225 }
3226 else if (list[0] == OP_CHAR)
3227 {
3228 chr_ptr = list + 2;
3229 list_ptr = base_list;
3230 }
3231
3232 /* Character bitsets can also be compared to certain opcodes. */
3233
3234 else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3235 #ifdef COMPILE_PCRE8
3236 /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3237 || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3238 #endif
3239 )
3240 {
3241 #ifdef COMPILE_PCRE8
3242 if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3243 #else
3244 if (base_list[0] == OP_CLASS)
3245 #endif
3246 {
3247 set1 = (pcre_uint8 *)(base_end - base_list[2]);
3248 list_ptr = list;
3249 }
3250 else
3251 {
3252 set1 = (pcre_uint8 *)(code - list[2]);
3253 list_ptr = base_list;
3254 }
3255
3256 invert_bits = FALSE;
3257 switch(list_ptr[0])
3258 {
3259 case OP_CLASS:
3260 case OP_NCLASS:
3261 set2 = (pcre_uint8 *)
3262 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3263 break;
3264
3265 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3266 case OP_XCLASS:
3267 xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3268 if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3269 if ((*xclass_flags & XCL_MAP) == 0)
3270 {
3271 /* No bits are set for characters < 256. */
3272 if (list[1] == 0) return TRUE;
3273 /* Might be an empty repeat. */
3274 continue;
3275 }
3276 set2 = (pcre_uint8 *)(xclass_flags + 1);
3277 break;
3278 #endif
3279
3280 case OP_NOT_DIGIT:
3281 invert_bits = TRUE;
3282 /* Fall through */
3283 case OP_DIGIT:
3284 set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3285 break;
3286
3287 case OP_NOT_WHITESPACE:
3288 invert_bits = TRUE;
3289 /* Fall through */
3290 case OP_WHITESPACE:
3291 set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3292 break;
3293
3294 case OP_NOT_WORDCHAR:
3295 invert_bits = TRUE;
3296 /* Fall through */
3297 case OP_WORDCHAR:
3298 set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3299 break;
3300
3301 default:
3302 return FALSE;
3303 }
3304
3305 /* Because the sets are unaligned, we need
3306 to perform byte comparison here. */
3307 set_end = set1 + 32;
3308 if (invert_bits)
3309 {
3310 do
3311 {
3312 if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3313 }
3314 while (set1 < set_end);
3315 }
3316 else
3317 {
3318 do
3319 {
3320 if ((*set1++ & *set2++) != 0) return FALSE;
3321 }
3322 while (set1 < set_end);
3323 }
3324
3325 if (list[1] == 0) return TRUE;
3326 /* Might be an empty repeat. */
3327 continue;
3328 }
3329
3330 /* Some property combinations also acceptable. Unicode property opcodes are
3331 processed specially; the rest can be handled with a lookup table. */
3332
3333 else
3334 {
3335 pcre_uint32 leftop, rightop;
3336
3337 leftop = base_list[0];
3338 rightop = list[0];
3339
3340 #ifdef SUPPORT_UCP
3341 accepted = FALSE; /* Always set in non-unicode case. */
3342 if (leftop == OP_PROP || leftop == OP_NOTPROP)
3343 {
3344 if (rightop == OP_EOD)
3345 accepted = TRUE;
3346 else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3347 {
3348 int n;
3349 const pcre_uint8 *p;
3350 BOOL same = leftop == rightop;
3351 BOOL lisprop = leftop == OP_PROP;
3352 BOOL risprop = rightop == OP_PROP;
3353 BOOL bothprop = lisprop && risprop;
3354
3355 /* There's a table that specifies how each combination is to be
3356 processed:
3357 0 Always return FALSE (never auto-possessify)
3358 1 Character groups are distinct (possessify if both are OP_PROP)
3359 2 Check character categories in the same group (general or particular)
3360 3 Return TRUE if the two opcodes are not the same
3361 ... see comments below
3362 */
3363
3364 n = propposstab[base_list[2]][list[2]];
3365 switch(n)
3366 {
3367 case 0: break;
3368 case 1: accepted = bothprop; break;
3369 case 2: accepted = (base_list[3] == list[3]) != same; break;
3370 case 3: accepted = !same; break;
3371
3372 case 4: /* Left general category, right particular category */
3373 accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3374 break;
3375
3376 case 5: /* Right general category, left particular category */
3377 accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3378 break;
3379
3380 /* This code is logically tricky. Think hard before fiddling with it.
3381 The posspropstab table has four entries per row. Each row relates to
3382 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3383 Only WORD actually needs all four entries, but using repeats for the
3384 others means they can all use the same code below.
3385
3386 The first two entries in each row are Unicode general categories, and
3387 apply always, because all the characters they include are part of the
3388 PCRE character set. The third and fourth entries are a general and a
3389 particular category, respectively, that include one or more relevant
3390 characters. One or the other is used, depending on whether the check
3391 is for a general or a particular category. However, in both cases the
3392 category contains more characters than the specials that are defined
3393 for the property being tested against. Therefore, it cannot be used
3394 in a NOTPROP case.
3395
3396 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3397 Underscore is covered by ucp_P or ucp_Po. */
3398
3399 case 6: /* Left alphanum vs right general category */
3400 case 7: /* Left space vs right general category */
3401 case 8: /* Left word vs right general category */
3402 p = posspropstab[n-6];
3403 accepted = risprop && lisprop ==
3404 (list[3] != p[0] &&
3405 list[3] != p[1] &&
3406 (list[3] != p[2] || !lisprop));
3407 break;
3408
3409 case 9: /* Right alphanum vs left general category */
3410 case 10: /* Right space vs left general category */
3411 case 11: /* Right word vs left general category */
3412 p = posspropstab[n-9];
3413 accepted = lisprop && risprop ==
3414 (base_list[3] != p[0] &&
3415 base_list[3] != p[1] &&
3416 (base_list[3] != p[2] || !risprop));
3417 break;
3418
3419 case 12: /* Left alphanum vs right particular category */
3420 case 13: /* Left space vs right particular category */
3421 case 14: /* Left word vs right particular category */
3422 p = posspropstab[n-12];
3423 accepted = risprop && lisprop ==
3424 (catposstab[p[0]][list[3]] &&
3425 catposstab[p[1]][list[3]] &&
3426 (list[3] != p[3] || !lisprop));
3427 break;
3428
3429 case 15: /* Right alphanum vs left particular category */
3430 case 16: /* Right space vs left particular category */
3431 case 17: /* Right word vs left particular category */
3432 p = posspropstab[n-15];
3433 accepted = lisprop && risprop ==
3434 (catposstab[p[0]][base_list[3]] &&
3435 catposstab[p[1]][base_list[3]] &&
3436 (base_list[3] != p[3] || !risprop));
3437 break;
3438 }
3439 }
3440 }
3441
3442 else
3443 #endif /* SUPPORT_UCP */
3444
3445 accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3446 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3447 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3448
3449 if (!accepted) return FALSE;
3450
3451 if (list[1] == 0) return TRUE;
3452 /* Might be an empty repeat. */
3453 continue;
3454 }
3455
3456 /* Control reaches here only if one of the items is a small character list.
3457 All characters are checked against the other side. */
3458
3459 do
3460 {
3461 chr = *chr_ptr;
3462
3463 switch(list_ptr[0])
3464 {
3465 case OP_CHAR:
3466 ochr_ptr = list_ptr + 2;
3467 do
3468 {
3469 if (chr == *ochr_ptr) return FALSE;
3470 ochr_ptr++;
3471 }
3472 while(*ochr_ptr != NOTACHAR);
3473 break;
3474
3475 case OP_NOT:
3476 ochr_ptr = list_ptr + 2;
3477 do
3478 {
3479 if (chr == *ochr_ptr)
3480 break;
3481 ochr_ptr++;
3482 }
3483 while(*ochr_ptr != NOTACHAR);
3484 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3485 break;
3486
3487 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3488 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3489
3490 case OP_DIGIT:
3491 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3492 break;
3493
3494 case OP_NOT_DIGIT:
3495 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3496 break;
3497
3498 case OP_WHITESPACE:
3499 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3500 break;
3501
3502 case OP_NOT_WHITESPACE:
3503 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3504 break;
3505
3506 case OP_WORDCHAR:
3507 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3508 break;
3509
3510 case OP_NOT_WORDCHAR:
3511 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3512 break;
3513
3514 case OP_HSPACE:
3515 switch(chr)
3516 {
3517 HSPACE_CASES: return FALSE;
3518 default: break;
3519 }
3520 break;
3521
3522 case OP_NOT_HSPACE:
3523 switch(chr)
3524 {
3525 HSPACE_CASES: break;
3526 default: return FALSE;
3527 }
3528 break;
3529
3530 case OP_ANYNL:
3531 case OP_VSPACE:
3532 switch(chr)
3533 {
3534 VSPACE_CASES: return FALSE;
3535 default: break;
3536 }
3537 break;
3538
3539 case OP_NOT_VSPACE:
3540 switch(chr)
3541 {
3542 VSPACE_CASES: break;
3543 default: return FALSE;
3544 }
3545 break;
3546
3547 case OP_DOLL:
3548 case OP_EODN:
3549 switch (chr)
3550 {
3551 case CHAR_CR:
3552 case CHAR_LF:
3553 case CHAR_VT:
3554 case CHAR_FF:
3555 case CHAR_NEL:
3556 #ifndef EBCDIC
3557 case 0x2028:
3558 case 0x2029:
3559 #endif /* Not EBCDIC */
3560 return FALSE;
3561 }
3562 break;
3563
3564 case OP_EOD: /* Can always possessify before \z */
3565 break;
3566
3567 #ifdef SUPPORT_UCP
3568 case OP_PROP:
3569 case OP_NOTPROP:
3570 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3571 list_ptr[0] == OP_NOTPROP))
3572 return FALSE;
3573 break;
3574 #endif
3575
3576 case OP_NCLASS:
3577 if (chr > 255) return FALSE;
3578 /* Fall through */
3579
3580 case OP_CLASS:
3581 if (chr > 255) break;
3582 class_bitset = (pcre_uint8 *)
3583 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3584 if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3585 break;
3586
3587 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3588 case OP_XCLASS:
3589 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3590 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3591 break;
3592 #endif
3593
3594 default:
3595 return FALSE;
3596 }
3597
3598 chr_ptr++;
3599 }
3600 while(*chr_ptr != NOTACHAR);
3601
3602 /* At least one character must be matched from this opcode. */
3603
3604 if (list[1] == 0) return TRUE;
3605 }
3606
3607 /* Control never reaches here. There used to be a fail-save return FALSE; here,
3608 but some compilers complain about an unreachable statement. */
3609
3610 }
3611
3612
3613
3614 /*************************************************
3615 * Scan compiled regex for auto-possession *
3616 *************************************************/
3617
3618 /* Replaces single character iterations with their possessive alternatives
3619 if appropriate. This function modifies the compiled opcode!
3620
3621 Arguments:
3622 code points to start of the byte code
3623 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3624 cd static compile data
3625
3626 Returns: nothing
3627 */
3628
3629 static void
3630 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3631 {
3632 register pcre_uchar c;
3633 const pcre_uchar *end;
3634 pcre_uchar *repeat_opcode;
3635 pcre_uint32 list[8];
3636
3637 for (;;)
3638 {
3639 c = *code;
3640
3641 /* When a pattern with bad UTF-8 encoding is compiled with NO_UTF_CHECK,
3642 it may compile without complaining, but may get into a loop here if the code
3643 pointer points to a bad value. This is, of course a documentated possibility,
3644 when NO_UTF_CHECK is set, so it isn't a bug, but we can detect this case and
3645 just give up on this optimization. */
3646
3647 if (c >= OP_TABLE_LENGTH) return;
3648
3649 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3650 {
3651 c -= get_repeat_base(c) - OP_STAR;
3652 end = (c <= OP_MINUPTO) ?
3653 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3654 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3655
3656 if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3657 {
3658 switch(c)
3659 {
3660 case OP_STAR:
3661 *code += OP_POSSTAR - OP_STAR;
3662 break;
3663
3664 case OP_MINSTAR:
3665 *code += OP_POSSTAR - OP_MINSTAR;
3666 break;
3667
3668 case OP_PLUS:
3669 *code += OP_POSPLUS - OP_PLUS;
3670 break;
3671
3672 case OP_MINPLUS:
3673 *code += OP_POSPLUS - OP_MINPLUS;
3674 break;
3675
3676 case OP_QUERY:
3677 *code += OP_POSQUERY - OP_QUERY;
3678 break;
3679
3680 case OP_MINQUERY:
3681 *code += OP_POSQUERY - OP_MINQUERY;
3682 break;
3683
3684 case OP_UPTO:
3685 *code += OP_POSUPTO - OP_UPTO;
3686 break;
3687
3688 case OP_MINUPTO:
3689 *code += OP_POSUPTO - OP_MINUPTO;
3690 break;
3691 }
3692 }
3693 c = *code;
3694 }
3695 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3696 {
3697 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3698 if (c == OP_XCLASS)
3699 repeat_opcode = code + GET(code, 1);
3700 else
3701 #endif
3702 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3703
3704 c = *repeat_opcode;
3705 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3706 {
3707 /* end must not be NULL. */
3708 end = get_chr_property_list(code, utf, cd->fcc, list);
3709
3710 list[1] = (c & 1) == 0;
3711
3712 if (compare_opcodes(end, utf, cd, list, end))
3713 {
3714 switch (c)
3715 {
3716 case OP_CRSTAR:
3717 case OP_CRMINSTAR:
3718 *repeat_opcode = OP_CRPOSSTAR;
3719 break;
3720
3721 case OP_CRPLUS:
3722 case OP_CRMINPLUS:
3723 *repeat_opcode = OP_CRPOSPLUS;
3724 break;
3725
3726 case OP_CRQUERY:
3727 case OP_CRMINQUERY:
3728 *repeat_opcode = OP_CRPOSQUERY;
3729 break;
3730
3731 case OP_CRRANGE:
3732 case OP_CRMINRANGE:
3733 *repeat_opcode = OP_CRPOSRANGE;
3734 break;
3735 }
3736 }
3737 }
3738 c = *code;
3739 }
3740
3741 switch(c)
3742 {
3743 case OP_END:
3744 return;
3745
3746 case OP_TYPESTAR:
3747 case OP_TYPEMINSTAR:
3748 case OP_TYPEPLUS:
3749 case OP_TYPEMINPLUS:
3750 case OP_TYPEQUERY:
3751 case OP_TYPEMINQUERY:
3752 case OP_TYPEPOSSTAR:
3753 case OP_TYPEPOSPLUS:
3754 case OP_TYPEPOSQUERY:
3755 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3756 break;
3757
3758 case OP_TYPEUPTO:
3759 case OP_TYPEMINUPTO:
3760 case OP_TYPEEXACT:
3761 case OP_TYPEPOSUPTO:
3762 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3763 code += 2;
3764 break;
3765
3766 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3767 case OP_XCLASS:
3768 code += GET(code, 1);
3769 break;
3770 #endif
3771
3772 case OP_MARK:
3773 case OP_PRUNE_ARG:
3774 case OP_SKIP_ARG:
3775 case OP_THEN_ARG:
3776 code += code[1];
3777 break;
3778 }
3779
3780 /* Add in the fixed length from the table */
3781
3782 code += PRIV(OP_lengths)[c];
3783
3784 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3785 a multi-byte character. The length in the table is a minimum, so we have to
3786 arrange to skip the extra bytes. */
3787
3788 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3789 if (utf) switch(c)
3790 {
3791 case OP_CHAR:
3792 case OP_CHARI:
3793 case OP_NOT:
3794 case OP_NOTI:
3795 case OP_STAR:
3796 case OP_MINSTAR:
3797 case OP_PLUS:
3798 case OP_MINPLUS:
3799 case OP_QUERY:
3800 case OP_MINQUERY:
3801 case OP_UPTO:
3802 case OP_MINUPTO:
3803 case OP_EXACT:
3804 case OP_POSSTAR:
3805 case OP_POSPLUS:
3806 case OP_POSQUERY:
3807 case OP_POSUPTO:
3808 case OP_STARI:
3809 case OP_MINSTARI:
3810 case OP_PLUSI:
3811 case OP_MINPLUSI:
3812 case OP_QUERYI:
3813 case OP_MINQUERYI:
3814 case OP_UPTOI:
3815 case OP_MINUPTOI:
3816 case OP_EXACTI:
3817 case OP_POSSTARI:
3818 case OP_POSPLUSI:
3819 case OP_POSQUERYI:
3820 case OP_POSUPTOI:
3821 case OP_NOTSTAR:
3822 case OP_NOTMINSTAR:
3823 case OP_NOTPLUS:
3824 case OP_NOTMINPLUS:
3825 case OP_NOTQUERY:
3826 case OP_NOTMINQUERY:
3827 case OP_NOTUPTO:
3828 case OP_NOTMINUPTO:
3829 case OP_NOTEXACT:
3830 case OP_NOTPOSSTAR:
3831 case OP_NOTPOSPLUS:
3832 case OP_NOTPOSQUERY:
3833 case OP_NOTPOSUPTO:
3834 case OP_NOTSTARI:
3835 case OP_NOTMINSTARI:
3836 case OP_NOTPLUSI:
3837 case OP_NOTMINPLUSI:
3838 case OP_NOTQUERYI:
3839 case OP_NOTMINQUERYI:
3840 case OP_NOTUPTOI:
3841 case OP_NOTMINUPTOI:
3842 case OP_NOTEXACTI:
3843 case OP_NOTPOSSTARI:
3844 case OP_NOTPOSPLUSI:
3845 case OP_NOTPOSQUERYI:
3846 case OP_NOTPOSUPTOI:
3847 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3848 break;
3849 }
3850 #else
3851 (void)(utf); /* Keep compiler happy by referencing function argument */
3852 #endif
3853 }
3854 }
3855
3856
3857
3858 /*************************************************
3859 * Check for POSIX class syntax *
3860 *************************************************/
3861
3862 /* This function is called when the sequence "[:" or "[." or "[=" is
3863 encountered in a character class. It checks whether this is followed by a
3864 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3865 reach an unescaped ']' without the special preceding character, return FALSE.
3866
3867 Originally, this function only recognized a sequence of letters between the
3868 terminators, but it seems that Perl recognizes any sequence of characters,
3869 though of course unknown POSIX names are subsequently rejected. Perl gives an
3870 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3871 didn't consider this to be a POSIX class. Likewise for [:1234:].
3872
3873 The problem in trying to be exactly like Perl is in the handling of escapes. We
3874 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3875 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3876 below handles the special case of \], but does not try to do any other escape
3877 processing. This makes it different from Perl for cases such as [:l\ower:]
3878 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3879 "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3880 I think.
3881
3882 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3883 It seems that the appearance of a nested POSIX class supersedes an apparent
3884 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3885 a digit.
3886
3887 In Perl, unescaped square brackets may also appear as part of class names. For
3888 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3889 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3890 seem right at all. PCRE does not allow closing square brackets in POSIX class
3891 names.
3892
3893 Arguments:
3894 ptr pointer to the initial [
3895 endptr where to return the end pointer
3896
3897 Returns: TRUE or FALSE
3898 */
3899
3900 static BOOL
3901 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3902 {
3903 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3904 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3905 for (++ptr; *ptr != CHAR_NULL; ptr++)
3906 {
3907 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3908 ptr++;
3909 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3910 else
3911 {
3912 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3913 {
3914 *endptr = ptr;
3915 return TRUE;
3916 }
3917 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3918 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3919 ptr[1] == CHAR_EQUALS_SIGN) &&
3920 check_posix_syntax(ptr, endptr))
3921 return FALSE;
3922 }
3923 }
3924 return FALSE;
3925 }
3926
3927
3928
3929
3930 /*************************************************
3931 * Check POSIX class name *
3932 *************************************************/
3933
3934 /* This function is called to check the name given in a POSIX-style class entry
3935 such as [:alnum:].
3936
3937 Arguments:
3938 ptr points to the first letter
3939 len the length of the name
3940
3941 Returns: a value representing the name, or -1 if unknown
3942 */
3943
3944 static int
3945 check_posix_name(const pcre_uchar *ptr, int len)
3946 {
3947 const char *pn = posix_names;
3948 register int yield = 0;
3949 while (posix_name_lengths[yield] != 0)
3950 {
3951 if (len == posix_name_lengths[yield] &&
3952 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3953 pn += posix_name_lengths[yield] + 1;
3954 yield++;
3955 }
3956 return -1;
3957 }
3958
3959
3960 /*************************************************
3961 * Adjust OP_RECURSE items in repeated group *
3962 *************************************************/
3963
3964 /* OP_RECURSE items contain an offset from the start of the regex to the group
3965 that is referenced. This means that groups can be replicated for fixed
3966 repetition simply by copying (because the recursion is allowed to refer to
3967 earlier groups that are outside the current group). However, when a group is
3968 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3969 inserted before it, after it has been compiled. This means that any OP_RECURSE
3970 items within it that refer to the group itself or any contained groups have to
3971 have their offsets adjusted. That one of the jobs of this function. Before it
3972 is called, the partially compiled regex must be temporarily terminated with
3973 OP_END.
3974
3975 This function has been extended with the possibility of forward references for
3976 recursions and subroutine calls. It must also check the list of such references
3977 for the group we are dealing with. If it finds that one of the recursions in
3978 the current group is on this list, it adjusts the offset in the list, not the
3979 value in the reference (which is a group number).
3980
3981 Arguments:
3982 group points to the start of the group
3983 adjust the amount by which the group is to be moved
3984 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3985 cd contains pointers to tables etc.
3986 save_hwm_offset the hwm forward reference offset at the start of the group
3987
3988 Returns: nothing
3989 */
3990
3991 static void
3992 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3993 size_t save_hwm_offset)
3994 {
3995 pcre_uchar *ptr = group;
3996
3997 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3998 {
3999 int offset;
4000 pcre_uchar *hc;
4001
4002 /* See if this recursion is on the forward reference list. If so, adjust the
4003 reference. */
4004
4005 for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4006 hc += LINK_SIZE)
4007 {
4008 offset = (int)GET(hc, 0);
4009 if (cd->start_code + offset == ptr + 1)
4010 {
4011 PUT(hc, 0, offset + adjust);
4012 break;
4013 }
4014 }
4015
4016 /* Otherwise, adjust the recursion offset if it's after the start of this
4017 group. */
4018
4019 if (hc >= cd->hwm)
4020 {
4021 offset = (int)GET(ptr, 1);
4022 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
4023 }
4024
4025 ptr += 1 + LINK_SIZE;
4026 }
4027 }
4028
4029
4030
4031 /*************************************************
4032 * Insert an automatic callout point *
4033 *************************************************/
4034
4035 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
4036 callout points before each pattern item.
4037
4038 Arguments:
4039 code current code pointer
4040 ptr current pattern pointer
4041 cd pointers to tables etc
4042
4043 Returns: new code pointer
4044 */
4045
4046 static pcre_uchar *
4047 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4048 {
4049 *code++ = OP_CALLOUT;
4050 *code++ = 255;
4051 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
4052 PUT(code, LINK_SIZE, 0); /* Default length */
4053 return code + 2 * LINK_SIZE;
4054 }
4055
4056
4057
4058 /*************************************************
4059 * Complete a callout item *
4060 *************************************************/
4061
4062 /* A callout item contains the length of the next item in the pattern, which
4063 we can't fill in till after we have reached the relevant point. This is used
4064 for both automatic and manual callouts.
4065
4066 Arguments:
4067 previous_callout points to previous callout item
4068 ptr current pattern pointer
4069 cd pointers to tables etc
4070
4071 Returns: nothing
4072 */
4073
4074 static void
4075 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4076 {
4077 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4078 PUT(previous_callout, 2 + LINK_SIZE, length);
4079 }
4080
4081
4082
4083 #ifdef SUPPORT_UCP
4084 /*************************************************
4085 * Get othercase range *
4086 *************************************************/
4087
4088 /* This function is passed the start and end of a class range, in UTF-8 mode
4089 with UCP support. It searches up the characters, looking for ranges of
4090 characters in the "other" case. Each call returns the next one, updating the
4091 start address. A character with multiple other cases is returned on its own
4092 with a special return value.
4093
4094 Arguments:
4095 cptr points to starting character value; updated
4096 d end value
4097 ocptr where to put start of othercase range
4098 odptr where to put end of othercase range
4099
4100 Yield: -1 when no more
4101 0 when a range is returned
4102 >0 the CASESET offset for char with multiple other cases
4103 in this case, ocptr contains the original
4104 */
4105
4106 static int
4107 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4108 pcre_uint32 *odptr)
4109 {
4110 pcre_uint32 c, othercase, next;
4111 unsigned int co;
4112
4113 /* Find the first character that has an other case. If it has multiple other
4114 cases, return its case offset value. */
4115
4116 for (c = *cptr; c <= d; c++)
4117 {
4118 if ((co = UCD_CASESET(c)) != 0)
4119 {
4120 *ocptr = c++; /* Character that has the set */
4121 *cptr = c; /* Rest of input range */
4122 return (int)co;
4123 }
4124 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4125 }
4126
4127 if (c > d) return -1; /* Reached end of range */
4128
4129 /* Found a character that has a single other case. Search for the end of the
4130 range, which is either the end of the input range, or a character that has zero
4131 or more than one other cases. */
4132
4133 *ocptr = othercase;
4134 next = othercase + 1;
4135
4136 for (++c; c <= d; c++)
4137 {
4138 if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4139 next++;
4140 }
4141
4142 *odptr = next - 1; /* End of othercase range */
4143 *cptr = c; /* Rest of input range */
4144 return 0;
4145 }
4146 #endif /* SUPPORT_UCP */
4147
4148
4149
4150 /*************************************************
4151 * Add a character or range to a class *
4152 *************************************************/
4153
4154 /* This function packages up the logic of adding a character or range of
4155 characters to a class. The character values in the arguments will be within the
4156 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4157 mutually recursive with the function immediately below.
4158
4159 Arguments:
4160 classbits the bit map for characters < 256
4161 uchardptr points to the pointer for extra data
4162 options the options word
4163 cd contains pointers to tables etc.
4164 start start of range character
4165 end end of range character
4166
4167 Returns: the number of < 256 characters added
4168 the pointer to extra data is updated
4169 */
4170
4171 static int
4172 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4173 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4174 {
4175 pcre_uint32 c;
4176 pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4177 int n8 = 0;
4178
4179 /* If caseless matching is required, scan the range and process alternate
4180 cases. In Unicode, there are 8-bit characters that have alternate cases that
4181 are greater than 255 and vice-versa. Sometimes we can just extend the original
4182 range. */
4183
4184 if ((options & PCRE_CASELESS) != 0)
4185 {
4186 #ifdef SUPPORT_UCP
4187 if ((options & PCRE_UTF8) != 0)
4188 {
4189 int rc;
4190 pcre_uint32 oc, od;
4191
4192 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4193 c = start;
4194
4195 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4196 {
4197 /* Handle a single character that has more than one other case. */
4198
4199 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4200 PRIV(ucd_caseless_sets) + rc, oc);
4201
4202 /* Do nothing if the other case range is within the original range. */
4203
4204 else if (oc >= start && od <= end) continue;
4205
4206 /* Extend the original range if there is overlap, noting that if oc < c, we
4207 can't have od > end because a subrange is always shorter than the basic
4208 range. Otherwise, use a recursive call to add the additional range. */
4209
4210 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4211 else if (od > end && oc <= end + 1)
4212 {
4213 end = od; /* Extend upwards */
4214 if (end > classbits_end) classbits_end = (end <= 0xff ? end : 0xff);
4215 }
4216 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4217 }
4218 }
4219 else
4220 #endif /* SUPPORT_UCP */
4221
4222 /* Not UTF-mode, or no UCP */
4223
4224 for (c = start; c <= classbits_end; c++)
4225 {
4226 SETBIT(classbits, cd->fcc[c]);
4227 n8++;
4228 }
4229 }
4230
4231 /* Now handle the original range. Adjust the final value according to the bit
4232 length - this means that the same lists of (e.g.) horizontal spaces can be used
4233 in all cases. */
4234
4235 #if defined COMPILE_PCRE8
4236 #ifdef SUPPORT_UTF
4237 if ((options & PCRE_UTF8) == 0)
4238 #endif
4239 if (end > 0xff) end = 0xff;
4240
4241 #elif defined COMPILE_PCRE16
4242 #ifdef SUPPORT_UTF
4243 if ((options & PCRE_UTF16) == 0)
4244 #endif
4245 if (end > 0xffff) end = 0xffff;
4246
4247 #endif /* COMPILE_PCRE[8|16] */
4248
4249 /* Use the bitmap for characters < 256. Otherwise use extra data.*/
4250
4251 for (c = start; c <= classbits_end; c++)
4252 {
4253 /* Regardless of start, c will always be <= 255. */
4254 SETBIT(classbits, c);
4255 n8++;
4256 }
4257
4258 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4259 if (start <= 0xff) start = 0xff + 1;
4260
4261 if (end >= start)
4262 {
4263 pcre_uchar *uchardata = *uchardptr;
4264 #ifdef SUPPORT_UTF
4265 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4266 {
4267 if (start < end)
4268 {
4269 *uchardata++ = XCL_RANGE;
4270 uchardata += PRIV(ord2utf)(start, uchardata);
4271 uchardata += PRIV(ord2utf)(end, uchardata);
4272 }
4273 else if (start == end)
4274 {
4275 *uchardata++ = XCL_SINGLE;
4276 uchardata += PRIV(ord2utf)(start, uchardata);
4277 }
4278 }
4279 else
4280 #endif /* SUPPORT_UTF */
4281
4282 /* Without UTF support, character values are constrained by the bit length,
4283 and can only be > 256 for 16-bit and 32-bit libraries. */
4284
4285 #ifdef COMPILE_PCRE8
4286 {}
4287 #else
4288 if (start < end)
4289 {
4290 *uchardata++ = XCL_RANGE;
4291 *uchardata++ = start;
4292 *uchardata++ = end;
4293 }
4294 else if (start == end)
4295 {
4296 *uchardata++ = XCL_SINGLE;
4297 *uchardata++ = start;
4298 }
4299 #endif
4300
4301 *uchardptr = uchardata; /* Updata extra data pointer */
4302 }
4303 #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4304
4305 return n8; /* Number of 8-bit characters */
4306 }
4307
4308
4309
4310
4311 /*************************************************
4312 * Add a list of characters to a class *
4313 *************************************************/
4314
4315 /* This function is used for adding a list of case-equivalent characters to a
4316 class, and also for adding a list of horizontal or vertical whitespace. If the
4317 list is in order (which it should be), ranges of characters are detected and
4318 handled appropriately. This function is mutually recursive with the function
4319 above.
4320
4321 Arguments:
4322 classbits the bit map for characters < 256
4323 uchardptr points to the pointer for extra data
4324 options the options word
4325 cd contains pointers to tables etc.
4326 p points to row of 32-bit values, terminated by NOTACHAR
4327 except character to omit; this is used when adding lists of
4328 case-equivalent characters to avoid including the one we
4329 already know about
4330
4331 Returns: the number of < 256 characters added
4332 the pointer to extra data is updated
4333 */
4334
4335 static int
4336 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4337 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4338 {
4339 int n8 = 0;
4340 while (p[0] < NOTACHAR)
4341 {
4342 int n = 0;
4343 if (p[0] != except)
4344 {
4345 while(p[n+1] == p[0] + n + 1) n++;
4346 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4347 }
4348 p += n + 1;
4349 }
4350 return n8;
4351 }
4352
4353
4354
4355 /*************************************************
4356 * Add characters not in a list to a class *
4357 *************************************************/
4358
4359 /* This function is used for adding the complement of a list of horizontal or
4360 vertical whitespace to a class. The list must be in order.
4361
4362 Arguments:
4363 classbits the bit map for characters < 256
4364 uchardptr points to the pointer for extra data
4365 options the options word
4366 cd contains pointers to tables etc.
4367 p points to row of 32-bit values, terminated by NOTACHAR
4368
4369 Returns: the number of < 256 characters added
4370 the pointer to extra data is updated
4371 */
4372
4373 static int
4374 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4375 int options, compile_data *cd, const pcre_uint32 *p)
4376 {
4377 BOOL utf = (options & PCRE_UTF8) != 0;
4378 int n8 = 0;
4379 if (p[0] > 0)
4380 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4381 while (p[0] < NOTACHAR)
4382 {
4383 while (p[1] == p[0] + 1) p++;
4384 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4385 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4386 p++;
4387 }
4388 return n8;
4389 }
4390
4391
4392
4393 /*************************************************
4394 * Compile one branch *
4395 *************************************************/
4396
4397 /* Scan the pattern, compiling it into the a vector. If the options are
4398 changed during the branch, the pointer is used to change the external options
4399 bits. This function is used during the pre-compile phase when we are trying
4400 to find out the amount of memory needed, as well as during the real compile
4401 phase. The value of lengthptr distinguishes the two phases.
4402
4403 Arguments:
4404 optionsptr pointer to the option bits
4405 codeptr points to the pointer to the current code point
4406 ptrptr points to the current pattern pointer
4407 errorcodeptr points to error code variable
4408 firstcharptr place to put the first required character
4409 firstcharflagsptr place to put the first character flags, or a negative number
4410 reqcharptr place to put the last required character
4411 reqcharflagsptr place to put the last required character flags, or a negative number
4412 bcptr points to current branch chain
4413 cond_depth conditional nesting depth
4414 cd contains pointers to tables etc.
4415 lengthptr NULL during the real compile phase
4416 points to length accumulator during pre-compile phase
4417
4418 Returns: TRUE on success
4419 FALSE, with *errorcodeptr set non-zero on error
4420 */
4421
4422 static BOOL
4423 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4424 const pcre_uchar **ptrptr, int *errorcodeptr,
4425 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4426 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4427 branch_chain *bcptr, int cond_depth,
4428 compile_data *cd, int *lengthptr)
4429 {
4430 int repeat_type, op_type;
4431 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4432 int bravalue = 0;
4433 int greedy_default, greedy_non_default;
4434 pcre_uint32 firstchar, reqchar;
4435 pcre_int32 firstcharflags, reqcharflags;
4436 pcre_uint32 zeroreqchar, zerofirstchar;
4437 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4438 pcre_int32 req_caseopt, reqvary, tempreqvary;
4439 int options = *optionsptr; /* May change dynamically */
4440 int after_manual_callout = 0;
4441 int length_prevgroup = 0;
4442 register pcre_uint32 c;
4443 int escape;
4444 register pcre_uchar *code = *codeptr;
4445 pcre_uchar *last_code = code;
4446 pcre_uchar *orig_code = code;
4447 pcre_uchar *tempcode;
4448 BOOL inescq = FALSE;
4449 BOOL groupsetfirstchar = FALSE;
4450 const pcre_uchar *ptr = *ptrptr;
4451 const pcre_uchar *tempptr;
4452 const pcre_uchar *nestptr = NULL;
4453 pcre_uchar *previous = NULL;
4454 pcre_uchar *previous_callout = NULL;
4455 size_t save_hwm_offset = 0;
4456 pcre_uint8 classbits[32];
4457
4458 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4459 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4460 dynamically as we process the pattern. */
4461
4462 #ifdef SUPPORT_UTF
4463 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4464 BOOL utf = (options & PCRE_UTF8) != 0;
4465 #ifndef COMPILE_PCRE32
4466 pcre_uchar utf_chars[6];
4467 #endif
4468 #else
4469 BOOL utf = FALSE;
4470 #endif
4471
4472 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4473 class_uchardata always so that it can be passed to add_to_class() always,
4474 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4475 alternative calls for the different cases. */
4476
4477 pcre_uchar *class_uchardata;
4478 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4479 BOOL xclass;
4480 pcre_uchar *class_uchardata_base;
4481 #endif
4482
4483 #ifdef PCRE_DEBUG
4484 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4485 #endif
4486
4487 /* Set up the default and non-default settings for greediness */
4488
4489 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4490 greedy_non_default = greedy_default ^ 1;
4491
4492 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4493 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4494 matches a non-fixed char first char; reqchar just remains unset if we never
4495 find one.
4496
4497 When we hit a repeat whose minimum is zero, we may have to adjust these values
4498 to take the zero repeat into account. This is implemented by setting them to
4499 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4500 item types that can be repeated set these backoff variables appropriately. */
4501
4502 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4503 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4504
4505 /* The variable req_caseopt contains either the REQ_CASELESS value
4506 or zero, according to the current setting of the caseless flag. The
4507 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4508 firstchar or reqchar variables to record the case status of the
4509 value. This is used only for ASCII characters. */
4510
4511 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4512
4513 /* Switch on next character until the end of the branch */
4514
4515 for (;; ptr++)
4516 {
4517 BOOL negate_class;
4518 BOOL should_flip_negation;
4519 BOOL possessive_quantifier;
4520 BOOL is_quantifier;
4521 BOOL is_recurse;
4522 BOOL reset_bracount;
4523 int class_has_8bitchar;
4524 int class_one_char;
4525 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4526 BOOL xclass_has_prop;
4527 #endif
4528 int newoptions;
4529 int recno;
4530 int refsign;
4531 int skipbytes;
4532 pcre_uint32 subreqchar, subfirstchar;
4533 pcre_int32 subreqcharflags, subfirstcharflags;
4534 int terminator;
4535 unsigned int mclength;
4536 unsigned int tempbracount;
4537 pcre_uint32 ec;
4538 pcre_uchar mcbuffer[8];
4539
4540 /* Get next character in the pattern */
4541
4542 c = *ptr;
4543
4544 /* If we are at the end of a nested substitution, revert to the outer level
4545 string. Nesting only happens one level deep. */
4546
4547 if (c == CHAR_NULL && nestptr != NULL)
4548 {
4549 ptr = nestptr;
4550 nestptr = NULL;
4551 c = *ptr;
4552 }
4553
4554 /* If we are in the pre-compile phase, accumulate the length used for the
4555 previous cycle of this loop. */
4556
4557 if (lengthptr != NULL)
4558 {
4559 #ifdef PCRE_DEBUG
4560 if (code > cd->hwm) cd->hwm = code; /* High water info */
4561 #endif
4562 if (code > cd->start_workspace + cd->workspace_size -
4563 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4564 {
4565 *errorcodeptr = ERR52;
4566 goto FAILED;
4567 }
4568
4569 /* There is at least one situation where code goes backwards: this is the
4570 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4571 the class is simply eliminated. However, it is created first, so we have to
4572 allow memory for it. Therefore, don't ever reduce the length at this point.
4573 */
4574
4575 if (code < last_code) code = last_code;
4576
4577 /* Paranoid check for integer overflow */
4578
4579 if (OFLOW_MAX - *lengthptr < code - last_code)
4580 {
4581 *errorcodeptr = ERR20;
4582 goto FAILED;
4583 }
4584
4585 *lengthptr += (int)(code - last_code);
4586 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4587 (int)(code - last_code), c, c));
4588
4589 /* If "previous" is set and it is not at the start of the work space, move
4590 it back to there, in order to avoid filling up the work space. Otherwise,
4591 if "previous" is NULL, reset the current code pointer to the start. */
4592
4593 if (previous != NULL)
4594 {
4595 if (previous > orig_code)
4596 {
4597 memmove(orig_code, previous, IN_UCHARS(code - previous));
4598 code -= previous - orig_code;
4599 previous = orig_code;
4600 }
4601 }
4602 else code = orig_code;
4603
4604 /* Remember where this code item starts so we can pick up the length
4605 next time round. */
4606
4607 last_code = code;
4608 }
4609
4610 /* In the real compile phase, just check the workspace used by the forward
4611 reference list. */
4612
4613 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4614 WORK_SIZE_SAFETY_MARGIN)
4615 {
4616 *errorcodeptr = ERR52;
4617 goto FAILED;
4618 }
4619
4620 /* If in \Q...\E, check for the end; if not, we have a literal */
4621
4622 if (inescq && c != CHAR_NULL)
4623 {
4624 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4625 {
4626 inescq = FALSE;
4627 ptr++;
4628 continue;
4629 }
4630 else
4631 {
4632 if (previous_callout != NULL)
4633 {
4634 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4635 complete_callout(previous_callout, ptr, cd);
4636 previous_callout = NULL;
4637 }
4638 if ((options & PCRE_AUTO_CALLOUT) != 0)
4639 {
4640 previous_callout = code;
4641 code = auto_callout(code, ptr, cd);
4642 }
4643 goto NORMAL_CHAR;
4644 }
4645 /* Control does not reach here. */
4646 }
4647
4648 /* In extended mode, skip white space and comments. We need a loop in order
4649 to check for more white space and more comments after a comment. */
4650
4651 if ((options & PCRE_EXTENDED) != 0)
4652 {
4653 for (;;)
4654 {
4655 while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4656 if (c != CHAR_NUMBER_SIGN) break;
4657 ptr++;
4658 while (*ptr != CHAR_NULL)
4659 {
4660 if (IS_NEWLINE(ptr)) /* For non-fixed-length newline cases, */
4661 { /* IS_NEWLINE sets cd->nllen. */
4662 ptr += cd->nllen;
4663 break;
4664 }
4665 ptr++;
4666 #ifdef SUPPORT_UTF
4667 if (utf) FORWARDCHAR(ptr);
4668 #endif
4669 }
4670 c = *ptr; /* Either NULL or the char after a newline */
4671 }
4672 }
4673
4674 /* See if the next thing is a quantifier. */
4675
4676 is_quantifier =
4677 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4678 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4679
4680 /* Fill in length of a previous callout, except when the next thing is a
4681 quantifier or when processing a property substitution string in UCP mode. */
4682
4683 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4684 after_manual_callout-- <= 0)
4685 {
4686 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4687 complete_callout(previous_callout, ptr, cd);
4688 previous_callout = NULL;
4689 }
4690
4691 /* Create auto callout, except for quantifiers, or while processing property
4692 strings that are substituted for \w etc in UCP mode. */
4693
4694 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4695 {
4696 previous_callout = code;
4697 code = auto_callout(code, ptr, cd);
4698 }
4699
4700 /* Process the next pattern item. */
4701
4702 switch(c)
4703 {
4704 /* ===================================================================*/
4705 case CHAR_NULL: /* The branch terminates at string end */
4706 case CHAR_VERTICAL_LINE: /* or | or ) */
4707 case CHAR_RIGHT_PARENTHESIS:
4708 *firstcharptr = firstchar;
4709 *firstcharflagsptr = firstcharflags;
4710 *reqcharptr = reqchar;
4711 *reqcharflagsptr = reqcharflags;
4712 *codeptr = code;
4713 *ptrptr = ptr;
4714 if (lengthptr != NULL)
4715 {
4716 if (OFLOW_MAX - *lengthptr < code - last_code)
4717 {
4718 *errorcodeptr = ERR20;
4719 goto FAILED;
4720 }
4721 *lengthptr += (int)(code - last_code); /* To include callout length */
4722 DPRINTF((">> end branch\n"));
4723 }
4724 return TRUE;
4725
4726
4727 /* ===================================================================*/
4728 /* Handle single-character metacharacters. In multiline mode, ^ disables
4729 the setting of any following char as a first character. */
4730
4731 case CHAR_CIRCUMFLEX_ACCENT:
4732 previous = NULL;
4733 if ((options & PCRE_MULTILINE) != 0)
4734 {
4735 if (firstcharflags == REQ_UNSET)
4736 zerofirstcharflags = firstcharflags = REQ_NONE;
4737 *code++ = OP_CIRCM;
4738 }
4739 else *code++ = OP_CIRC;
4740 break;
4741
4742 case CHAR_DOLLAR_SIGN:
4743 previous = NULL;
4744 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4745 break;
4746
4747 /* There can never be a first char if '.' is first, whatever happens about
4748 repeats. The value of reqchar doesn't change either. */
4749
4750 case CHAR_DOT:
4751 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4752 zerofirstchar = firstchar;
4753 zerofirstcharflags = firstcharflags;
4754 zeroreqchar = reqchar;
4755 zeroreqcharflags = reqcharflags;
4756 previous = code;
4757 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4758 break;
4759
4760
4761 /* ===================================================================*/
4762 /* Character classes. If the included characters are all < 256, we build a
4763 32-byte bitmap of the permitted characters, except in the special case
4764 where there is only one such character. For negated classes, we build the
4765 map as usual, then invert it at the end. However, we use a different opcode
4766 so that data characters > 255 can be handled correctly.
4767
4768 If the class contains characters outside the 0-255 range, a different
4769 opcode is compiled. It may optionally have a bit map for characters < 256,
4770 but those above are are explicitly listed afterwards. A flag byte tells
4771 whether the bitmap is present, and whether this is a negated class or not.
4772
4773 In JavaScript compatibility mode, an isolated ']' causes an error. In
4774 default (Perl) mode, it is treated as a data character. */
4775
4776 case CHAR_RIGHT_SQUARE_BRACKET:
4777 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4778 {
4779 *errorcodeptr = ERR64;
4780 goto FAILED;
4781 }
4782 goto NORMAL_CHAR;
4783
4784 /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4785 used for "start of word" and "end of word". As these are otherwise illegal
4786 sequences, we don't break anything by recognizing them. They are replaced
4787 by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4788 erroneous and are handled by the normal code below. */
4789
4790 case CHAR_LEFT_SQUARE_BRACKET:
4791 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4792 {
4793 nestptr = ptr + 7;
4794 ptr = sub_start_of_word - 1;
4795 continue;
4796 }
4797
4798 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4799 {
4800 nestptr = ptr + 7;
4801 ptr = sub_end_of_word - 1;
4802 continue;
4803 }
4804
4805 /* Handle a real character class. */
4806
4807 previous = code;
4808
4809 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4810 they are encountered at the top level, so we'll do that too. */
4811
4812 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4813 ptr[1] == CHAR_EQUALS_SIGN) &&
4814 check_posix_syntax(ptr, &tempptr))
4815 {
4816 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4817 goto FAILED;
4818 }
4819
4820 /* If the first character is '^', set the negation flag and skip it. Also,
4821 if the first few characters (either before or after ^) are \Q\E or \E we
4822 skip them too. This makes for compatibility with Perl. */
4823
4824 negate_class = FALSE;
4825 for (;;)
4826 {
4827 c = *(++ptr);
4828 if (c == CHAR_BACKSLASH)
4829 {
4830 if (ptr[1] == CHAR_E)
4831 ptr++;
4832 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4833 ptr += 3;
4834 else
4835 break;
4836 }
4837 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4838 negate_class = TRUE;
4839 else break;
4840 }
4841
4842 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4843 an initial ']' is taken as a data character -- the code below handles
4844 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4845 [^] must match any character, so generate OP_ALLANY. */
4846
4847 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4848 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4849 {
4850 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4851 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4852 zerofirstchar = firstchar;
4853 zerofirstcharflags = firstcharflags;
4854 break;
4855 }
4856
4857 /* If a class contains a negative special such as \S, we need to flip the
4858 negation flag at the end, so that support for characters > 255 works
4859 correctly (they are all included in the class). */
4860
4861 should_flip_negation = FALSE;
4862
4863 /* Extended class (xclass) will be used when characters > 255
4864 might match. */
4865
4866 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4867 xclass = FALSE;
4868 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4869 class_uchardata_base = class_uchardata; /* Save the start */
4870 #endif
4871
4872 /* For optimization purposes, we track some properties of the class:
4873 class_has_8bitchar will be non-zero if the class contains at least one <
4874 256 character; class_one_char will be 1 if the class contains just one
4875 character; xclass_has_prop will be TRUE if unicode property checks
4876 are present in the class. */
4877
4878 class_has_8bitchar = 0;
4879 class_one_char = 0;
4880 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4881 xclass_has_prop = FALSE;
4882 #endif
4883
4884 /* Initialize the 32-char bit map to all zeros. We build the map in a
4885 temporary bit of memory, in case the class contains fewer than two
4886 8-bit characters because in that case the compiled code doesn't use the bit
4887 map. */
4888
4889 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4890
4891 /* Process characters until ] is reached. By writing this as a "do" it
4892 means that an initial ] is taken as a data character. At the start of the
4893 loop, c contains the first byte of the character. */
4894
4895 if (c != CHAR_NULL) do
4896 {
4897 const pcre_uchar *oldptr;
4898
4899 #ifdef SUPPORT_UTF
4900 if (utf && HAS_EXTRALEN(c))
4901 { /* Braces are required because the */
4902 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4903 }
4904 #endif
4905
4906 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4907 /* In the pre-compile phase, accumulate the length of any extra
4908 data and reset the pointer. This is so that very large classes that
4909 contain a zillion > 255 characters no longer overwrite the work space
4910 (which is on the stack). We have to remember that there was XCLASS data,
4911 however. */
4912
4913 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4914 {
4915 xclass = TRUE;
4916 *lengthptr += (int)(class_uchardata - class_uchardata_base);
4917 class_uchardata = class_uchardata_base;
4918 }
4919 #endif
4920
4921 /* Inside \Q...\E everything is literal except \E */
4922
4923 if (inescq)
4924 {
4925 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4926 {
4927 inescq = FALSE; /* Reset literal state */
4928 ptr++; /* Skip the 'E' */
4929 continue; /* Carry on with next */
4930 }
4931 goto CHECK_RANGE; /* Could be range if \E follows */
4932 }
4933
4934 /* Handle POSIX class names. Perl allows a negation extension of the
4935 form [:^name:]. A square bracket that doesn't match the syntax is
4936 treated as a literal. We also recognize the POSIX constructions
4937 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4938 5.6 and 5.8 do. */
4939
4940 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4941 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4942 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4943 {
4944 BOOL local_negate = FALSE;
4945 int posix_class, taboffset, tabopt;
4946 register const pcre_uint8 *cbits = cd->cbits;
4947 pcre_uint8 pbits[32];
4948
4949 if (ptr[1] != CHAR_COLON)
4950 {
4951 *errorcodeptr = ERR31;
4952 goto FAILED;
4953 }
4954
4955 ptr += 2;
4956 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4957 {
4958 local_negate = TRUE;
4959 should_flip_negation = TRUE; /* Note negative special */
4960 ptr++;
4961 }
4962
4963 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4964 if (posix_class < 0)
4965 {
4966 *errorcodeptr = ERR30;
4967 goto FAILED;
4968 }
4969
4970 /* If matching is caseless, upper and lower are converted to
4971 alpha. This relies on the fact that the class table starts with
4972 alpha, lower, upper as the first 3 entries. */
4973
4974 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4975 posix_class = 0;
4976
4977 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4978 different escape sequences that use Unicode properties \p or \P. Others
4979 that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4980 directly. */
4981
4982 #ifdef SUPPORT_UCP
4983 if ((options & PCRE_UCP) != 0)
4984 {
4985 unsigned int ptype = 0;
4986 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4987
4988 /* The posix_substitutes table specifies which POSIX classes can be
4989 converted to \p or \P items. */
4990
4991 if (posix_substitutes[pc] != NULL)
4992 {
4993 nestptr = tempptr + 1;
4994 ptr = posix_substitutes[pc] - 1;
4995 continue;
4996 }
4997
4998 /* There are three other classes that generate special property calls
4999 that are recognized only in an XCLASS. */
5000
5001 else switch(posix_class)
5002 {
5003 case PC_GRAPH:
5004 ptype = PT_PXGRAPH;
5005 /* Fall through */
5006 case PC_PRINT:
5007 if (ptype == 0) ptype = PT_PXPRINT;
5008 /* Fall through */
5009 case PC_PUNCT:
5010 if (ptype == 0) ptype = PT_PXPUNCT;
5011 *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
5012 *class_uchardata++ = ptype;
5013 *class_uchardata++ = 0;
5014 xclass_has_prop = TRUE;
5015 ptr = tempptr + 1;
5016 continue;
5017
5018 /* For all other POSIX classes, no special action is taken in UCP
5019 mode. Fall through to the non_UCP case. */
5020
5021 default:
5022 break;
5023 }
5024 }
5025 #endif
5026 /* In the non-UCP case, or when UCP makes no difference, we build the
5027 bit map for the POSIX class in a chunk of local store because we may be
5028 adding and subtracting from it, and we don't want to subtract bits that
5029 may be in the main map already. At the end we or the result into the
5030 bit map that is being built. */
5031
5032 posix_class *= 3;
5033
5034 /* Copy in the first table (always present) */
5035
5036 memcpy(pbits, cbits + posix_class_maps[posix_class],
5037 32 * sizeof(pcre_uint8));
5038
5039 /* If there is a second table, add or remove it as required. */
5040
5041 taboffset = posix_class_maps[posix_class + 1];
5042 tabopt = posix_class_maps[posix_class + 2];
5043
5044 if (taboffset >= 0)
5045 {
5046 if (tabopt >= 0)
5047 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
5048 else
5049 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
5050 }
5051
5052 /* Now see if we need to remove any special characters. An option
5053 value of 1 removes vertical space and 2 removes underscore. */
5054
5055 if (tabopt < 0) tabopt = -tabopt;
5056 if (tabopt == 1) pbits[1] &= ~0x3c;
5057 else if (tabopt == 2) pbits[11] &= 0x7f;
5058
5059 /* Add the POSIX table or its complement into the main table that is
5060 being built and we are done. */
5061
5062 if (local_negate)
5063 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
5064 else
5065 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
5066
5067 ptr = tempptr + 1;
5068 /* Every class contains at least one < 256 character. */
5069 class_has_8bitchar = 1;
5070 /* Every class contains at least two characters. */
5071 class_one_char = 2;
5072 continue; /* End of POSIX syntax handling */
5073 }
5074
5075 /* Backslash may introduce a single character, or it may introduce one
5076 of the specials, which just set a flag. The sequence \b is a special
5077 case. Inside a class (and only there) it is treated as backspace. We
5078 assume that other escapes have more than one character in them, so
5079 speculatively set both class_has_8bitchar and class_one_char bigger
5080 than one. Unrecognized escapes fall through and are either treated
5081 as literal characters (by default), or are faulted if
5082 PCRE_EXTRA is set. */
5083
5084 if (c == CHAR_BACKSLASH)
5085 {
5086 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
5087 TRUE);
5088 if (*errorcodeptr != 0) goto FAILED;
5089 if (escape == 0) c = ec;
5090 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
5091 else if (escape == ESC_N) /* \N is not supported in a class */
5092 {
5093 *errorcodeptr = ERR71;
5094 goto FAILED;
5095 }
5096 else if (escape == ESC_Q) /* Handle start of quoted string */
5097 {
5098 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5099 {
5100 ptr += 2; /* avoid empty string */
5101 }
5102 else inescq = TRUE;
5103 continue;
5104 }
5105 else if (escape == ESC_E) continue; /* Ignore orphan \E */
5106
5107 else
5108 {
5109 register const pcre_uint8 *cbits = cd->cbits;
5110 /* Every class contains at least two < 256 characters. */
5111 class_has_8bitchar++;
5112 /* Every class contains at least two characters. */
5113 class_one_char += 2;
5114
5115 switch (escape)
5116 {
5117 #ifdef SUPPORT_UCP
5118 case ESC_du: /* These are the values given for \d etc */
5119 case ESC_DU: /* when PCRE_UCP is set. We replace the */
5120 case ESC_wu: /* escape sequence with an appropriate \p */
5121 case ESC_WU: /* or \P to test Unicode properties instead */
5122 case ESC_su: /* of the default ASCII testing. */
5123 case ESC_SU:
5124 nestptr = ptr;
5125 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
5126 class_has_8bitchar--; /* Undo! */
5127 continue;
5128 #endif
5129 case ESC_d:
5130 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
5131 continue;
5132
5133 case ESC_D:
5134 should_flip_negation = TRUE;
5135 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
5136 continue;
5137
5138 case ESC_w:
5139 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
5140 continue;
5141
5142 case ESC_W:
5143 should_flip_negation = TRUE;
5144 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5145 continue;
5146
5147 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5148 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5149 previously set by something earlier in the character class.
5150 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5151 we could just adjust the appropriate bit. From PCRE 8.34 we no
5152 longer treat \s and \S specially. */
5153
5154 case ESC_s:
5155 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
5156 continue;
5157
5158 case ESC_S:
5159 should_flip_negation = TRUE;
5160 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
5161 continue;
5162
5163 /* The rest apply in both UCP and non-UCP cases. */
5164
5165 case ESC_h:
5166 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5167 PRIV(hspace_list), NOTACHAR);
5168 continue;
5169
5170 case ESC_H:
5171 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5172 cd, PRIV(hspace_list));
5173 continue;
5174
5175 case ESC_v:
5176 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5177 PRIV(vspace_list), NOTACHAR);
5178 continue;
5179
5180 case ESC_V:
5181 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5182 cd, PRIV(vspace_list));
5183 continue;
5184
5185 #ifdef SUPPORT_UCP
5186 case ESC_p:
5187 case ESC_P:
5188 {
5189 BOOL negated;
5190 unsigned int ptype = 0, pdata = 0;
5191 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5192 goto FAILED;
5193 *class_uchardata++ = ((escape == ESC_p) != negated)?
5194 XCL_PROP : XCL_NOTPROP;
5195 *class_uchardata++ = ptype;
5196 *class_uchardata++ = pdata;
5197 xclass_has_prop = TRUE;
5198 class_has_8bitchar--; /* Undo! */
5199 continue;
5200 }
5201 #endif
5202 /* Unrecognized escapes are faulted if PCRE is running in its
5203 strict mode. By default, for compatibility with Perl, they are
5204 treated as literals. */
5205
5206 default:
5207 if ((options & PCRE_EXTRA) != 0)
5208 {
5209 *errorcodeptr = ERR7;
5210 goto FAILED;
5211 }
5212 class_has_8bitchar--; /* Undo the speculative increase. */
5213 class_one_char -= 2; /* Undo the speculative increase. */
5214 c = *ptr; /* Get the final character and fall through */
5215 break;
5216 }
5217 }
5218
5219 /* Fall through if the escape just defined a single character (c >= 0).
5220 This may be greater than 256. */
5221
5222 escape = 0;
5223
5224 } /* End of backslash handling */
5225
5226 /* A character may be followed by '-' to form a range. However, Perl does
5227 not permit ']' to be the end of the range. A '-' character at the end is
5228 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5229 code for handling \Q and \E is messy. */
5230
5231 CHECK_RANGE:
5232 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5233 {
5234 inescq = FALSE;
5235 ptr += 2;
5236 }
5237 oldptr = ptr;
5238
5239 /* Remember if \r or \n were explicitly used */
5240
5241 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5242
5243 /* Check for range */
5244
5245 if (!inescq && ptr[1] == CHAR_MINUS)
5246 {
5247 pcre_uint32 d;
5248 ptr += 2;
5249 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5250
5251 /* If we hit \Q (not followed by \E) at this point, go into escaped
5252 mode. */
5253
5254 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5255 {
5256 ptr += 2;
5257 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5258 { ptr += 2; continue; }
5259 inescq = TRUE;
5260 break;
5261 }
5262
5263 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5264 back the pointer and jump to handle the character that preceded it. */
5265
5266 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5267 {
5268 ptr = oldptr;
5269 goto CLASS_SINGLE_CHARACTER;
5270 }
5271
5272 /* Otherwise, we have a potential range; pick up the next character */
5273
5274 #ifdef SUPPORT_UTF
5275 if (utf)
5276 { /* Braces are required because the */
5277 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5278 }
5279 else
5280 #endif
5281 d = *ptr; /* Not UTF-8 mode */
5282
5283 /* The second part of a range can be a single-character escape
5284 sequence, but not any of the other escapes. Perl treats a hyphen as a
5285 literal in such circumstances. However, in Perl's warning mode, a
5286 warning is given, so PCRE now faults it as it is almost certainly a
5287 mistake on the user's part. */
5288
5289 if (!inescq)
5290 {
5291 if (d == CHAR_BACKSLASH)
5292 {
5293 int descape;
5294 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5295 if (*errorcodeptr != 0) goto FAILED;
5296
5297 /* 0 means a character was put into d; \b is backspace; any other
5298 special causes an error. */
5299
5300 if (descape != 0)
5301 {
5302 if (descape == ESC_b) d = CHAR_BS; else
5303 {
5304 *errorcodeptr = ERR83;
5305 goto FAILED;
5306 }
5307 }
5308 }
5309
5310 /* A hyphen followed by a POSIX class is treated in the same way. */
5311
5312 else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5313 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5314 ptr[1] == CHAR_EQUALS_SIGN) &&
5315 check_posix_syntax(ptr, &tempptr))
5316 {
5317 *errorcodeptr = ERR83;
5318 goto FAILED;
5319 }
5320 }
5321
5322 /* Check that the two values are in the correct order. Optimize
5323 one-character ranges. */
5324
5325 if (d < c)
5326 {
5327 *errorcodeptr = ERR8;
5328 goto FAILED;
5329 }
5330 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5331
5332 /* We have found a character range, so single character optimizations
5333 cannot be done anymore. Any value greater than 1 indicates that there
5334 is more than one character. */
5335
5336 class_one_char = 2;
5337
5338 /* Remember an explicit \r or \n, and add the range to the class. */
5339
5340 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5341
5342 class_has_8bitchar +=
5343 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5344
5345 continue; /* Go get the next char in the class */
5346 }
5347
5348 /* Handle a single character - we can get here for a normal non-escape
5349 char, or after \ that introduces a single character or for an apparent
5350 range that isn't. Only the value 1 matters for class_one_char, so don't
5351 increase it if it is already 2 or more ... just in case there's a class
5352 with a zillion characters in it. */
5353
5354 CLASS_SINGLE_CHARACTER:
5355 if (class_one_char < 2) class_one_char++;
5356
5357 /* If class_one_char is 1, we have the first single character in the
5358 class, and there have been no prior ranges, or XCLASS items generated by
5359 escapes. If this is the final character in the class, we can optimize by
5360 turning the item into a 1-character OP_CHAR[I] if it's positive, or
5361 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5362 to be set. Otherwise, there can be no first char if this item is first,
5363 whatever repeat count may follow. In the case of reqchar, save the
5364 previous value for reinstating. */
5365
5366 if (!inescq && class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5367 {
5368 ptr++;
5369 zeroreqchar = reqchar;
5370 zeroreqcharflags = reqcharflags;
5371
5372 if (negate_class)
5373 {
5374 #ifdef SUPPORT_UCP
5375 int d;
5376 #endif
5377 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5378 zerofirstchar = firstchar;
5379 zerofirstcharflags = firstcharflags;
5380
5381 /* For caseless UTF-8 mode when UCP support is available, check
5382 whether this character has more than one other case. If so, generate
5383 a special OP_NOTPROP item instead of OP_NOTI. */
5384
5385 #ifdef SUPPORT_UCP
5386 if (utf && (options & PCRE_CASELESS) != 0 &&
5387 (d = UCD_CASESET(c)) != 0)
5388 {
5389 *code++ = OP_NOTPROP;
5390 *code++ = PT_CLIST;
5391 *code++ = d;
5392 }
5393 else
5394 #endif
5395 /* Char has only one other case, or UCP not available */
5396
5397 {
5398 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5399 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5400 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5401 code += PRIV(ord2utf)(c, code);
5402 else
5403 #endif
5404 *code++ = c;
5405 }
5406
5407 /* We are finished with this character class */
5408
5409 goto END_CLASS;
5410 }
5411
5412 /* For a single, positive character, get the value into mcbuffer, and
5413 then we can handle this with the normal one-character code. */
5414
5415 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5416 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5417 mclength = PRIV(ord2utf)(c, mcbuffer);
5418 else
5419 #endif
5420 {
5421 mcbuffer[0] = c;
5422 mclength = 1;
5423 }
5424 goto ONE_CHAR;
5425 } /* End of 1-char optimization */
5426
5427 /* There is more than one character in the class, or an XCLASS item
5428 has been generated. Add this character to the class. */
5429
5430 class_has_8bitchar +=
5431 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5432 }
5433
5434 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5435 If we are at the end of an internal nested string, revert to the outer
5436 string. */
5437
5438 while (((c = *(++ptr)) != CHAR_NULL ||
5439 (nestptr != NULL &&
5440 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5441 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5442
5443 /* Check for missing terminating ']' */
5444
5445 if (c == CHAR_NULL)
5446 {
5447 *errorcodeptr = ERR6;
5448 goto FAILED;
5449 }
5450
5451 /* We will need an XCLASS if data has been placed in class_uchardata. In
5452 the second phase this is a sufficient test. However, in the pre-compile
5453 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5454 only if the very last character in the class needs XCLASS will it contain
5455 anything at this point. For this reason, xclass gets set TRUE above when
5456 uchar_classdata is emptied, and that's why this code is the way it is here
5457 instead of just doing a test on class_uchardata below. */
5458
5459 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5460 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5461 #endif
5462
5463 /* If this is the first thing in the branch, there can be no first char
5464 setting, whatever the repeat count. Any reqchar setting must remain
5465 unchanged after any kind of repeat. */
5466
5467 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5468 zerofirstchar = firstchar;
5469 zerofirstcharflags = firstcharflags;
5470 zeroreqchar = reqchar;
5471 zeroreqcharflags = reqcharflags;
5472
5473 /* If there are characters with values > 255, we have to compile an
5474 extended class, with its own opcode, unless there was a negated special
5475 such as \S in the class, and PCRE_UCP is not set, because in that case all
5476 characters > 255 are in the class, so any that were explicitly given as
5477 well can be ignored. If (when there are explicit characters > 255 that must
5478 be listed) there are no characters < 256, we can omit the bitmap in the
5479 actual compiled code. */
5480
5481 #ifdef SUPPORT_UTF
5482 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5483 #elif !defined COMPILE_PCRE8
5484 if (xclass && !should_flip_negation)
5485 #endif
5486 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5487 {
5488 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5489 *code++ = OP_XCLASS;
5490 code += LINK_SIZE;
5491 *code = negate_class? XCL_NOT:0;
5492 if (xclass_has_prop) *code |= XCL_HASPROP;
5493
5494 /* If the map is required, move up the extra data to make room for it;
5495 otherwise just move the code pointer to the end of the extra data. */
5496
5497 if (class_has_8bitchar > 0)
5498 {
5499 *code++ |= XCL_MAP;
5500 memmove(code + (32 / sizeof(pcre_uchar)), code,
5501 IN_UCHARS(class_uchardata - code));
5502 if (negate_class && !xclass_has_prop)
5503 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5504 memcpy(code, classbits, 32);
5505 code = class_uchardata + (32 / sizeof(pcre_uchar));
5506 }
5507 else code = class_uchardata;
5508
5509 /* Now fill in the complete length of the item */
5510
5511 PUT(previous, 1, (int)(code - previous));
5512 break; /* End of class handling */
5513 }
5514 #endif
5515
5516 /* If there are no characters > 255, or they are all to be included or
5517 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5518 whole class was negated and whether there were negative specials such as \S
5519 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5520 negating it if necessary. */
5521
5522 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5523 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5524 {
5525 if (negate_class)
5526 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5527 memcpy(code, classbits, 32);
5528 }
5529 code += 32 / sizeof(pcre_uchar);
5530
5531 END_CLASS:
5532 break;
5533
5534
5535 /* ===================================================================*/
5536 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5537 has been tested above. */
5538
5539 case CHAR_LEFT_CURLY_BRACKET:
5540 if (!is_quantifier) goto NORMAL_CHAR;
5541 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5542 if (*errorcodeptr != 0) goto FAILED;
5543 goto REPEAT;
5544
5545 case CHAR_ASTERISK:
5546 repeat_min = 0;
5547 repeat_max = -1;
5548 goto REPEAT;
5549
5550 case CHAR_PLUS:
5551 repeat_min = 1;
5552 repeat_max = -1;
5553 goto REPEAT;
5554
5555 case CHAR_QUESTION_MARK:
5556 repeat_min = 0;
5557 repeat_max = 1;
5558
5559 REPEAT:
5560 if (previous == NULL)
5561 {
5562 *errorcodeptr = ERR9;
5563 goto FAILED;
5564 }
5565
5566 if (repeat_min == 0)
5567 {
5568 firstchar = zerofirstchar; /* Adjust for zero repeat */
5569 firstcharflags = zerofirstcharflags;
5570 reqchar = zeroreqchar; /* Ditto */
5571 reqcharflags = zeroreqcharflags;
5572 }
5573
5574 /* Remember whether this is a variable length repeat */
5575
5576 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5577
5578 op_type = 0; /* Default single-char op codes */
5579 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5580
5581 /* Save start of previous item, in case we have to move it up in order to
5582 insert something before it. */
5583
5584 tempcode = previous;
5585
5586 /* Before checking for a possessive quantifier, we must skip over
5587 whitespace and comments in extended mode because Perl allows white space at
5588 this point. */
5589
5590 if ((options & PCRE_EXTENDED) != 0)
5591 {
5592 const pcre_uchar *p = ptr + 1;
5593 for (;;)
5594 {
5595 while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5596 if (*p != CHAR_NUMBER_SIGN) break;
5597 p++;
5598 while (*p != CHAR_NULL)
5599 {
5600 if (IS_NEWLINE(p)) /* For non-fixed-length newline cases, */
5601 { /* IS_NEWLINE sets cd->nllen. */
5602 p += cd->nllen;
5603 break;
5604 }
5605 p++;
5606 #ifdef SUPPORT_UTF
5607 if (utf) FORWARDCHAR(p);
5608 #endif
5609 } /* Loop for comment characters */
5610 } /* Loop for multiple comments */
5611 ptr = p - 1; /* Character before the next significant one. */
5612 }
5613
5614 /* If the next character is '+', we have a possessive quantifier. This
5615 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5616 If the next character is '?' this is a minimizing repeat, by default,
5617 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5618 repeat type to the non-default. */
5619
5620 if (ptr[1] == CHAR_PLUS)
5621 {
5622 repeat_type = 0; /* Force greedy */
5623 possessive_quantifier = TRUE;
5624 ptr++;
5625 }
5626 else if (ptr[1] == CHAR_QUESTION_MARK)
5627 {
5628 repeat_type = greedy_non_default;
5629 ptr++;
5630 }
5631 else repeat_type = greedy_default;
5632
5633 /* If previous was a recursion call, wrap it in atomic brackets so that
5634 previous becomes the atomic group. All recursions were so wrapped in the
5635 past, but it no longer happens for non-repeated recursions. In fact, the
5636 repeated ones could be re-implemented independently so as not to need this,
5637 but for the moment we rely on the code for repeating groups. */
5638
5639 if (*previous == OP_RECURSE)
5640 {
5641 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5642 *previous = OP_ONCE;
5643 PUT(previous, 1, 2 + 2*LINK_SIZE);
5644 previous[2 + 2*LINK_SIZE] = OP_KET;
5645 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5646 code += 2 + 2 * LINK_SIZE;
5647 length_prevgroup = 3 + 3*LINK_SIZE;
5648
5649 /* When actually compiling, we need to check whether this was a forward
5650 reference, and if so, adjust the offset. */
5651
5652 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5653 {
5654 int offset = GET(cd->hwm, -LINK_SIZE);
5655 if (offset == previous + 1 - cd->start_code)
5656 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5657 }
5658 }
5659
5660 /* Now handle repetition for the different types of item. */
5661
5662 /* If previous was a character or negated character match, abolish the item
5663 and generate a repeat item instead. If a char item has a minimum of more
5664 than one, ensure that it is set in reqchar - it might not be if a sequence
5665 such as x{3} is the first thing in a branch because the x will have gone
5666 into firstchar instead. */
5667
5668 if (*previous == OP_CHAR || *previous == OP_CHARI
5669 || *previous == OP_NOT || *previous == OP_NOTI)
5670 {
5671 switch (*previous)
5672 {
5673 default: /* Make compiler happy. */
5674 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5675 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5676 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5677 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5678 }
5679
5680 /* Deal with UTF characters that take up more than one character. It's
5681 easier to write this out separately than try to macrify it. Use c to
5682 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5683 it's a length rather than a small character. */
5684
5685 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5686 if (utf && NOT_FIRSTCHAR(code[-1]))
5687 {
5688 pcre_uchar *lastchar = code - 1;
5689 BACKCHAR(lastchar);
5690 c = (int)(code - lastchar); /* Length of UTF-8 character */
5691 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5692 c |= UTF_LENGTH; /* Flag c as a length */
5693 }
5694 else
5695 #endif /* SUPPORT_UTF */
5696
5697 /* Handle the case of a single charater - either with no UTF support, or
5698 with UTF disabled, or for a single character UTF character. */
5699 {
5700 c = code[-1];
5701 if (*previous <= OP_CHARI && repeat_min > 1)
5702 {
5703 reqchar = c;
5704 reqcharflags = req_caseopt | cd->req_varyopt;
5705 }
5706 }
5707
5708 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5709 }
5710
5711 /* If previous was a character type match (\d or similar), abolish it and
5712 create a suitable repeat item. The code is shared with single-character
5713 repeats by setting op_type to add a suitable offset into repeat_type. Note
5714 the the Unicode property types will be present only when SUPPORT_UCP is
5715 defined, but we don't wrap the little bits of code here because it just
5716 makes it horribly messy. */
5717
5718 else if (*previous < OP_EODN)
5719 {
5720 pcre_uchar *oldcode;
5721 int prop_type, prop_value;
5722 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5723 c = *previous;
5724
5725 OUTPUT_SINGLE_REPEAT:
5726 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5727 {
5728 prop_type = previous[1];
5729 prop_value = previous[2];
5730 }
5731 else prop_type = prop_value = -1;
5732
5733 oldcode = code;
5734 code = previous; /* Usually overwrite previous item */
5735
5736 /* If the maximum is zero then the minimum must also be zero; Perl allows
5737 this case, so we do too - by simply omitting the item altogether. */
5738
5739 if (repeat_max == 0) goto END_REPEAT;
5740
5741 /* Combine the op_type with the repeat_type */
5742
5743 repeat_type += op_type;
5744
5745 /* A minimum of zero is handled either as the special case * or ?, or as
5746 an UPTO, with the maximum given. */
5747
5748 if (repeat_min == 0)
5749 {
5750 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5751 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5752 else
5753 {
5754 *code++ = OP_UPTO + repeat_type;
5755 PUT2INC(code, 0, repeat_max);
5756 }
5757 }
5758
5759 /* A repeat minimum of 1 is optimized into some special cases. If the
5760 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5761 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5762 one less than the maximum. */
5763
5764 else if (repeat_min == 1)
5765 {
5766 if (repeat_max == -1)
5767 *code++ = OP_PLUS + repeat_type;
5768 else
5769 {
5770 code = oldcode; /* leave previous item in place */
5771 if (repeat_max == 1) goto END_REPEAT;
5772 *code++ = OP_UPTO + repeat_type;
5773 PUT2INC(code, 0, repeat_max - 1);
5774 }
5775 }
5776
5777 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5778 handled as an EXACT followed by an UPTO. */
5779
5780 else
5781 {
5782 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5783 PUT2INC(code, 0, repeat_min);
5784
5785 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5786 we have to insert the character for the previous code. For a repeated
5787 Unicode property match, there are two extra bytes that define the
5788 required property. In UTF-8 mode, long characters have their length in
5789 c, with the UTF_LENGTH bit as a flag. */
5790
5791 if (repeat_max < 0)
5792 {
5793 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5794 if (utf && (c & UTF_LENGTH) != 0)
5795 {
5796 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5797 code += c & 7;
5798 }
5799 else
5800 #endif
5801 {
5802 *code++ = c;
5803 if (prop_type >= 0)
5804 {
5805 *code++ = prop_type;
5806 *code++ = prop_value;
5807 }
5808 }
5809 *code++ = OP_STAR + repeat_type;
5810 }
5811
5812 /* Else insert an UPTO if the max is greater than the min, again
5813 preceded by the character, for the previously inserted code. If the
5814 UPTO is just for 1 instance, we can use QUERY instead. */
5815
5816 else if (repeat_max != repeat_min)
5817 {
5818 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5819 if (utf && (c & UTF_LENGTH) != 0)
5820 {
5821 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5822 code += c & 7;
5823 }
5824 else
5825 #endif
5826 *code++ = c;
5827 if (prop_type >= 0)
5828 {
5829 *code++ = prop_type;
5830 *code++ = prop_value;
5831 }
5832 repeat_max -= repeat_min;
5833
5834 if (repeat_max == 1)
5835 {
5836 *code++ = OP_QUERY + repeat_type;
5837 }
5838 else
5839 {
5840 *code++ = OP_UPTO + repeat_type;
5841 PUT2INC(code, 0, repeat_max);
5842 }
5843 }
5844 }
5845
5846 /* The character or character type itself comes last in all cases. */
5847
5848 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5849 if (utf && (c & UTF_LENGTH) != 0)
5850 {
5851 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5852 code += c & 7;
5853 }
5854 else
5855 #endif
5856 *code++ = c;
5857
5858 /* For a repeated Unicode property match, there are two extra bytes that
5859 define the required property. */
5860
5861 #ifdef SUPPORT_UCP
5862 if (prop_type >= 0)
5863 {
5864 *code++ = prop_type;
5865 *code++ = prop_value;
5866 }
5867 #endif
5868 }
5869
5870 /* If previous was a character class or a back reference, we put the repeat
5871 stuff after it, but just skip the item if the repeat was {0,0}. */
5872
5873 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5874 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5875 *previous == OP_XCLASS ||
5876 #endif
5877 *previous == OP_REF || *previous == OP_REFI ||
5878 *previous == OP_DNREF || *previous == OP_DNREFI)
5879 {
5880 if (repeat_max == 0)
5881 {
5882 code = previous;
5883 goto END_REPEAT;
5884 }
5885
5886 if (repeat_min == 0 && repeat_max == -1)
5887 *code++ = OP_CRSTAR + repeat_type;
5888 else if (repeat_min == 1 && repeat_max == -1)
5889 *code++ = OP_CRPLUS + repeat_type;
5890 else if (repeat_min == 0 && repeat_max == 1)
5891 *code++ = OP_CRQUERY + repeat_type;
5892 else
5893 {
5894 *code++ = OP_CRRANGE + repeat_type;
5895 PUT2INC(code, 0, repeat_min);
5896 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5897 PUT2INC(code, 0, repeat_max);
5898 }
5899 }
5900
5901 /* If previous was a bracket group, we may have to replicate it in certain
5902 cases. Note that at this point we can encounter only the "basic" bracket
5903 opcodes such as BRA and CBRA, as this is the place where they get converted
5904 into the more special varieties such as BRAPOS and SBRA. A test for >=
5905 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5906 ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5907 Originally, PCRE did not allow repetition of assertions, but now it does,
5908 for Perl compatibility. */
5909
5910 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5911 {
5912 register int i;
5913 int len = (int)(code - previous);
5914 pcre_uchar *bralink = NULL;
5915 pcre_uchar *brazeroptr = NULL;
5916
5917 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5918 we just ignore the repeat. */
5919
5920 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5921 goto END_REPEAT;
5922
5923 /* There is no sense in actually repeating assertions. The only potential
5924 use of repetition is in cases when the assertion is optional. Therefore,
5925 if the minimum is greater than zero, just ignore the repeat. If the
5926 maximum is not zero or one, set it to 1. */
5927
5928 if (*previous < OP_ONCE) /* Assertion */
5929 {
5930 if (repeat_min > 0) goto END_REPEAT;
5931 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5932 }
5933
5934 /* The case of a zero minimum is special because of the need to stick
5935 OP_BRAZERO in front of it, and because the group appears once in the
5936 data, whereas in other cases it appears the minimum number of times. For
5937 this reason, it is simplest to treat this case separately, as otherwise
5938 the code gets far too messy. There are several special subcases when the
5939 minimum is zero. */
5940
5941 if (repeat_min == 0)
5942 {
5943 /* If the maximum is also zero, we used to just omit the group from the
5944 output altogether, like this:
5945
5946 ** if (repeat_max == 0)
5947 ** {
5948 ** code = previous;
5949 ** goto END_REPEAT;
5950 ** }
5951
5952 However, that fails when a group or a subgroup within it is referenced
5953 as a subroutine from elsewhere in the pattern, so now we stick in
5954 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5955 don't have a list of which groups are referenced, we cannot do this
5956 selectively.
5957
5958 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5959 and do no more at this point. However, we do need to adjust any
5960 OP_RECURSE calls inside the group that refer to the group itself or any
5961 internal or forward referenced group, because the offset is from the
5962 start of the whole regex. Temporarily terminate the pattern while doing
5963 this. */
5964
5965 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5966 {
5967 *code = OP_END;
5968 adjust_recurse(previous, 1, utf, cd, save_hwm_offset);
5969 memmove(previous + 1, previous, IN_UCHARS(len));
5970 code++;
5971 if (repeat_max == 0)
5972 {
5973 *previous++ = OP_SKIPZERO;
5974 goto END_REPEAT;
5975 }
5976 brazeroptr = previous; /* Save for possessive optimizing */
5977 *previous++ = OP_BRAZERO + repeat_type;
5978 }
5979
5980 /* If the maximum is greater than 1 and limited, we have to replicate
5981 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
5982 The first one has to be handled carefully because it's the original
5983 copy, which has to be moved up. The remainder can be handled by code
5984 that is common with the non-zero minimum case below. We have to
5985 adjust the value or repeat_max, since one less copy is required. Once
5986 again, we may have to adjust any OP_RECURSE calls inside the group. */
5987
5988 else
5989 {
5990 int offset;
5991 *code = OP_END;
5992 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm_offset);
5993 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5994 code += 2 + LINK_SIZE;
5995 *previous++ = OP_BRAZERO + repeat_type;
5996 *previous++ = OP_BRA;
5997
5998 /* We chain together the bracket offset fields that have to be
5999 filled in later when the ends of the brackets are reached. */
6000
6001 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
6002 bralink = previous;
6003 PUTINC(previous, 0, offset);
6004 }
6005
6006 repeat_max--;
6007 }
6008
6009 /* If the minimum is greater than zero, replicate the group as many
6010 times as necessary, and adjust the maximum to the number of subsequent
6011 copies that we need. If we set a first char from the group, and didn't
6012 set a required char, copy the latter from the former. If there are any
6013 forward reference subroutine calls in the group, there will be entries on
6014 the workspace list; replicate these with an appropriate increment. */
6015
6016 else
6017 {
6018 if (repeat_min > 1)
6019 {
6020 /* In the pre-compile phase, we don't actually do the replication. We
6021 just adjust the length as if we had. Do some paranoid checks for
6022 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
6023 integer type when available, otherwise double. */
6024
6025 if (lengthptr != NULL)
6026 {
6027 int delta = (repeat_min - 1)*length_prevgroup;
6028 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
6029 (INT64_OR_DOUBLE)length_prevgroup >
6030 (INT64_OR_DOUBLE)INT_MAX ||
6031 OFLOW_MAX - *lengthptr < delta)
6032 {
6033 *errorcodeptr = ERR20;
6034 goto FAILED;
6035 }
6036 *lengthptr += delta;
6037 }
6038
6039 /* This is compiling for real. If there is a set first byte for
6040 the group, and we have not yet set a "required byte", set it. Make
6041 sure there is enough workspace for copying forward references before
6042 doing the copy. */
6043
6044 else
6045 {
6046 if (groupsetfirstchar && reqcharflags < 0)
6047 {
6048 reqchar = firstchar;
6049 reqcharflags = firstcharflags;
6050 }
6051
6052 for (i = 1; i < repeat_min; i++)
6053 {
6054 pcre_uchar *hc;
6055 size_t this_hwm_offset = cd->hwm - cd->start_workspace;
6056 memcpy(code, previous, IN_UCHARS(len));
6057
6058 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6059 WORK_SIZE_SAFETY_MARGIN -
6060 (this_hwm_offset - save_hwm_offset))
6061 {
6062 *errorcodeptr = expand_workspace(cd);
6063 if (*errorcodeptr != 0) goto FAILED;
6064 }
6065
6066 for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset;
6067 hc < (pcre_uchar *)cd->start_workspace + this_hwm_offset;
6068 hc += LINK_SIZE)
6069 {
6070 PUT(cd->hwm, 0, GET(hc, 0) + len);
6071 cd->hwm += LINK_SIZE;
6072 }
6073 save_hwm_offset = this_hwm_offset;
6074 code += len;
6075 }
6076 }
6077 }
6078
6079 if (repeat_max > 0) repeat_max -= repeat_min;
6080 }
6081
6082 /* This code is common to both the zero and non-zero minimum cases. If
6083 the maximum is limited, it replicates the group in a nested fashion,
6084 remembering the bracket starts on a stack. In the case of a zero minimum,
6085 the first one was set up above. In all cases the repeat_max now specifies
6086 the number of additional copies needed. Again, we must remember to
6087 replicate entries on the forward reference list. */
6088
6089 if (repeat_max >= 0)
6090 {
6091 /* In the pre-compile phase, we don't actually do the replication. We
6092 just adjust the length as if we had. For each repetition we must add 1
6093 to the length for BRAZERO and for all but the last repetition we must
6094 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
6095 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
6096 a 64-bit integer type when available, otherwise double. */
6097
6098 if (lengthptr != NULL && repeat_max > 0)
6099 {
6100 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
6101 2 - 2*LINK_SIZE; /* Last one doesn't nest */
6102 if ((INT64_OR_DOUBLE)repeat_max *
6103 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
6104 > (INT64_OR_DOUBLE)INT_MAX ||
6105 OFLOW_MAX - *lengthptr < delta)
6106 {
6107 *errorcodeptr = ERR20;
6108 goto FAILED;
6109 }
6110 *lengthptr += delta;
6111 }
6112
6113 /* This is compiling for real */
6114
6115 else for (i = repeat_max - 1; i >= 0; i--)
6116 {
6117 pcre_uchar *hc;
6118 size_t this_hwm_offset = cd->hwm - cd->start_workspace;
6119
6120 *code++ = OP_BRAZERO + repeat_type;
6121
6122 /* All but the final copy start a new nesting, maintaining the
6123 chain of brackets outstanding. */
6124
6125 if (i != 0)
6126 {
6127 int offset;
6128 *code++ = OP_BRA;
6129 offset = (bralink == NULL)? 0 : (int)(code - bralink);
6130 bralink = code;
6131 PUTINC(code, 0, offset);
6132 }
6133
6134 memcpy(code, previous, IN_UCHARS(len));
6135
6136 /* Ensure there is enough workspace for forward references before
6137 copying them. */
6138
6139 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6140 WORK_SIZE_SAFETY_MARGIN -
6141 (this_hwm_offset - save_hwm_offset))
6142 {
6143 *errorcodeptr = expand_workspace(cd);
6144 if (*errorcodeptr != 0) goto FAILED;
6145 }
6146
6147 for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset;
6148 hc < (pcre_uchar *)cd->start_workspace + this_hwm_offset;
6149 hc += LINK_SIZE)
6150 {
6151 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
6152 cd->hwm += LINK_SIZE;
6153 }
6154 save_hwm_offset = this_hwm_offset;
6155 code += len;
6156 }
6157
6158 /* Now chain through the pending brackets, and fill in their length
6159 fields (which are holding the chain links pro tem). */
6160
6161 while (bralink != NULL)
6162 {
6163 int oldlinkoffset;
6164 int offset = (int)(code - bralink + 1);
6165 pcre_uchar *bra = code - offset;
6166 oldlinkoffset = GET(bra, 1);
6167 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
6168 *code++ = OP_KET;
6169 PUTINC(code, 0, offset);
6170 PUT(bra, 1, offset);
6171 }
6172 }
6173
6174 /* If the maximum is unlimited, set a repeater in the final copy. For
6175 ONCE brackets, that's all we need to do. However, possessively repeated
6176 ONCE brackets can be converted into non-capturing brackets, as the
6177 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
6178 deal with possessive ONCEs specially.
6179
6180 Otherwise, when we are doing the actual compile phase, check to see
6181 whether this group is one that could match an empty string. If so,
6182 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
6183 that runtime checking can be done. [This check is also applied to ONCE
6184 groups at runtime, but in a different way.]
6185
6186 Then, if the quantifier was possessive and the bracket is not a
6187 conditional, we convert the BRA code to the POS form, and the KET code to
6188 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
6189 subpattern at both the start and at the end.) The use of special opcodes
6190 makes it possible to reduce greatly the stack usage in pcre_exec(). If
6191 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
6192
6193 Then, if the minimum number of matches is 1 or 0, cancel the possessive
6194 flag so that the default action below, of wrapping everything inside
6195 atomic brackets, does not happen. When the minimum is greater than 1,
6196 there will be earlier copies of the group, and so we still have to wrap
6197 the whole thing. */
6198
6199 else
6200 {
6201 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
6202 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
6203
6204 /* Convert possessive ONCE brackets to non-capturing */
6205
6206 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
6207 possessive_quantifier) *bracode = OP_BRA;
6208
6209 /* For non-possessive ONCE brackets, all we need to do is to
6210 set the KET. */
6211
6212 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
6213 *ketcode = OP_KETRMAX + repeat_type;
6214
6215 /* Handle non-ONCE brackets and possessive ONCEs (which have been
6216 converted to non-capturing above). */
6217
6218 else
6219 {
6220 /* In the compile phase, check for empty string matching. */
6221
6222 if (lengthptr == NULL)
6223 {
6224 pcre_uchar *scode = bracode;
6225 do
6226 {
6227 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6228 {
6229 *bracode += OP_SBRA - OP_BRA;
6230 break;
6231 }
6232 scode += GET(scode, 1);
6233 }
6234 while (*scode == OP_ALT);
6235 }
6236
6237 /* Handle possessive quantifiers. */
6238
6239 if (possessive_quantifier)
6240 {
6241 /* For COND brackets, we wrap the whole thing in a possessively
6242 repeated non-capturing bracket, because we have not invented POS
6243 versions of the COND opcodes. Because we are moving code along, we
6244 must ensure that any pending recursive references are updated. */
6245
6246 if (*bracode == OP_COND || *bracode == OP_SCOND)
6247 {
6248 int nlen = (int)(code - bracode);
6249 *code = OP_END;
6250 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm_offset);
6251 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6252 code += 1 + LINK_SIZE;
6253 nlen += 1 + LINK_SIZE;
6254 *bracode = OP_BRAPOS;
6255 *code++ = OP_KETRPOS;
6256 PUTINC(code, 0, nlen);
6257 PUT(bracode, 1, nlen);
6258 }
6259
6260 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6261
6262 else
6263 {
6264 *bracode += 1; /* Switch to xxxPOS opcodes */
6265 *ketcode = OP_KETRPOS;
6266 }
6267
6268 /* If the minimum is zero, mark it as possessive, then unset the
6269 possessive flag when the minimum is 0 or 1. */
6270
6271 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6272 if (repeat_min < 2) possessive_quantifier = FALSE;
6273 }
6274
6275 /* Non-possessive quantifier */
6276
6277 else *ketcode = OP_KETRMAX + repeat_type;
6278 }
6279 }
6280 }
6281
6282 /* If previous is OP_FAIL, it was generated by an empty class [] in
6283 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6284 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6285 error above. We can just ignore the repeat in JS case. */
6286
6287 else if (*previous == OP_FAIL) goto END_REPEAT;
6288
6289 /* Else there's some kind of shambles */
6290
6291 else
6292 {
6293 *errorcodeptr = ERR11;
6294 goto FAILED;
6295 }
6296
6297 /* If the character following a repeat is '+', possessive_quantifier is
6298 TRUE. For some opcodes, there are special alternative opcodes for this
6299 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6300 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6301 Sun's Java package, but the special opcodes can optimize it.
6302
6303 Some (but not all) possessively repeated subpatterns have already been
6304 completely handled in the code just above. For them, possessive_quantifier
6305 is always FALSE at this stage. Note that the repeated item starts at
6306 tempcode, not at previous, which might be the first part of a string whose
6307 (former) last char we repeated. */
6308
6309 if (possessive_quantifier)
6310 {
6311 int len;
6312
6313 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6314 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6315 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6316 remains is greater than zero, there's a further opcode that can be
6317 handled. If not, do nothing, leaving the EXACT alone. */
6318
6319 switch(*tempcode)
6320 {
6321 case OP_TYPEEXACT:
6322 tempcode += PRIV(OP_lengths)[*tempcode] +
6323 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6324 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6325 break;
6326
6327 /* CHAR opcodes are used for exacts whose count is 1. */
6328
6329 case OP_CHAR:
6330 case OP_CHARI:
6331 case OP_NOT:
6332 case OP_NOTI:
6333 case OP_EXACT:
6334 case OP_EXACTI:
6335 case OP_NOTEXACT:
6336 case OP_NOTEXACTI:
6337 tempcode += PRIV(OP_lengths)[*tempcode];
6338 #ifdef SUPPORT_UTF
6339 if (utf && HAS_EXTRALEN(tempcode[-1]))
6340 tempcode += GET_EXTRALEN(tempcode[-1]);
6341 #endif
6342 break;
6343
6344 /* For the class opcodes, the repeat operator appears at the end;
6345 adjust tempcode to point to it. */
6346
6347 case OP_CLASS:
6348 case OP_NCLASS:
6349 tempcode += 1 + 32/sizeof(pcre_uchar);
6350 break;
6351
6352 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6353 case OP_XCLASS:
6354 tempcode += GET(tempcode, 1);
6355 break;
6356 #endif
6357 }
6358
6359 /* If tempcode is equal to code (which points to the end of the repeated
6360 item), it means we have skipped an EXACT item but there is no following
6361 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6362 all other cases, tempcode will be pointing to the repeat opcode, and will
6363 be less than code, so the value of len will be greater than 0. */
6364
6365 len = (int)(code - tempcode);
6366 if (len > 0)
6367 {
6368 unsigned int repcode = *tempcode;
6369
6370 /* There is a table for possessifying opcodes, all of which are less
6371 than OP_CALLOUT. A zero entry means there is no possessified version.
6372 */
6373
6374 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6375 *tempcode = opcode_possessify[repcode];
6376
6377 /* For opcode without a special possessified version, wrap the item in
6378 ONCE brackets. Because we are moving code along, we must ensure that any
6379 pending recursive references are updated. */
6380
6381 else
6382 {
6383 *code = OP_END;
6384 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm_offset);
6385 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6386 code += 1 + LINK_SIZE;
6387 len += 1 + LINK_SIZE;
6388 tempcode[0] = OP_ONCE;
6389 *code++ = OP_KET;
6390 PUTINC(code, 0, len);
6391 PUT(tempcode, 1, len);
6392 }
6393 }
6394
6395 #ifdef NEVER
6396 if (len > 0) switch (*tempcode)
6397 {
6398 case OP_STAR: *tempcode = OP_POSSTAR; break;
6399 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6400 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6401 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6402
6403 case OP_STARI: *tempcode = OP_POSSTARI; break;
6404 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6405 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6406 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6407
6408 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6409 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6410 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6411 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6412
6413 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6414 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6415 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6416 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6417
6418 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6419 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6420 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6421 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6422
6423 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6424 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6425 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6426 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6427
6428 /* Because we are moving code along, we must ensure that any
6429 pending recursive references are updated. */
6430
6431 default:
6432 *code = OP_END;
6433 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm_offset);
6434 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6435 code += 1 + LINK_SIZE;
6436 len += 1 + LINK_SIZE;
6437 tempcode[0] = OP_ONCE;
6438 *code++ = OP_KET;
6439 PUTINC(code, 0, len);
6440 PUT(tempcode, 1, len);
6441 break;
6442 }
6443 #endif
6444 }
6445
6446 /* In all case we no longer have a previous item. We also set the
6447 "follows varying string" flag for subsequently encountered reqchars if
6448 it isn't already set and we have just passed a varying length item. */
6449
6450 END_REPEAT:
6451 previous = NULL;
6452 cd->req_varyopt |= reqvary;
6453 break;
6454
6455
6456 /* ===================================================================*/
6457 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6458 lookbehind or option setting or condition or all the other extended
6459 parenthesis forms. */
6460
6461 case CHAR_LEFT_PARENTHESIS:
6462 newoptions = options;
6463 skipbytes = 0;
6464 bravalue = OP_CBRA;
6465 save_hwm_offset = cd->hwm - cd->start_workspace;
6466 reset_bracount = FALSE;
6467
6468 /* First deal with various "verbs" that can be introduced by '*'. */
6469
6470 ptr++;
6471 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6472 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6473 {
6474 int i, namelen;
6475 int arglen = 0;
6476 const char *vn = verbnames;
6477 const pcre_uchar *name = ptr + 1;
6478 const pcre_uchar *arg = NULL;
6479 previous = NULL;
6480 ptr++;
6481 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6482 namelen = (int)(ptr - name);
6483
6484 /* It appears that Perl allows any characters whatsoever, other than
6485 a closing parenthesis, to appear in arguments, so we no longer insist on
6486 letters, digits, and underscores. */
6487
6488 if (*ptr == CHAR_COLON)
6489 {
6490 arg = ++ptr;
6491 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6492 arglen = (int)(ptr - arg);
6493 if ((unsigned int)arglen > MAX_MARK)
6494 {
6495 *errorcodeptr = ERR75;
6496 goto FAILED;
6497 }
6498 }
6499
6500 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6501 {
6502 *errorcodeptr = ERR60;
6503 goto FAILED;
6504 }
6505
6506 /* Scan the table of verb names */
6507
6508 for (i = 0; i < verbcount; i++)
6509 {
6510 if (namelen == verbs[i].len &&
6511 STRNCMP_UC_C8(name, vn, namelen) == 0)
6512 {
6513 int setverb;
6514
6515 /* Check for open captures before ACCEPT and convert it to
6516 ASSERT_ACCEPT if in an assertion. */
6517
6518 if (verbs[i].op == OP_ACCEPT)
6519 {
6520 open_capitem *oc;
6521 if (arglen != 0)
6522 {
6523 *errorcodeptr = ERR59;
6524 goto FAILED;
6525 }
6526 cd->had_accept = TRUE;
6527 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6528 {
6529 *code++ = OP_CLOSE;
6530 PUT2INC(code, 0, oc->number);
6531 }
6532 setverb = *code++ =
6533 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6534
6535 /* Do not set firstchar after *ACCEPT */
6536 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6537 }
6538
6539 /* Handle other cases with/without an argument */
6540
6541 else if (arglen == 0)
6542 {
6543 if (verbs[i].op < 0) /* Argument is mandatory */
6544 {
6545 *errorcodeptr = ERR66;
6546 goto FAILED;
6547 }
6548 setverb = *code++ = verbs[i].op;
6549 }
6550
6551 else
6552 {
6553 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6554 {
6555 *errorcodeptr = ERR59;
6556 goto FAILED;
6557 }
6558 setverb = *code++ = verbs[i].op_arg;
6559 *code++ = arglen;
6560 memcpy(code, arg, IN_UCHARS(arglen));
6561 code += arglen;
6562 *code++ = 0;
6563 }
6564
6565 switch (setverb)
6566 {
6567 case OP_THEN:
6568 case OP_THEN_ARG:
6569 cd->external_flags |= PCRE_HASTHEN;
6570 break;
6571
6572 case OP_PRUNE:
6573 case OP_PRUNE_ARG:
6574 case OP_SKIP:
6575 case OP_SKIP_ARG:
6576 cd->had_pruneorskip = TRUE;
6577 break;
6578 }
6579
6580 break; /* Found verb, exit loop */
6581 }
6582
6583 vn += verbs[i].len + 1;
6584 }
6585
6586 if (i < verbcount) continue; /* Successfully handled a verb */
6587 *errorcodeptr = ERR60; /* Verb not recognized */
6588 goto FAILED;
6589 }
6590
6591 /* Deal with the extended parentheses; all are introduced by '?', and the
6592 appearance of any of them means that this is not a capturing group. */
6593
6594 else if (*ptr == CHAR_QUESTION_MARK)
6595 {
6596 int i, set, unset, namelen;
6597 int *optset;
6598 const pcre_uchar *name;
6599 pcre_uchar *slot;
6600
6601 switch (*(++ptr))
6602 {
6603 case CHAR_NUMBER_SIGN: /* Comment; skip to ket */
6604 ptr++;
6605 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6606 if (*ptr == CHAR_NULL)
6607 {
6608 *errorcodeptr = ERR18;
6609 goto FAILED;
6610 }
6611 continue;
6612
6613
6614 /* ------------------------------------------------------------ */
6615 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6616 reset_bracount = TRUE;
6617 /* Fall through */
6618
6619 /* ------------------------------------------------------------ */
6620 case CHAR_COLON: /* Non-capturing bracket */
6621 bravalue = OP_BRA;
6622 ptr++;
6623 break;
6624
6625
6626 /* ------------------------------------------------------------ */
6627 case CHAR_LEFT_PARENTHESIS:
6628 bravalue = OP_COND; /* Conditional group */
6629 tempptr = ptr;
6630
6631 /* A condition can be an assertion, a number (referring to a numbered
6632 group's having been set), a name (referring to a named group), or 'R',
6633 referring to recursion. R<digits> and R&name are also permitted for
6634 recursion tests.
6635
6636 There are ways of testing a named group: (?(name)) is used by Python;
6637 Perl 5.10 onwards uses (?(<name>) or (?('name')).
6638
6639 There is one unfortunate ambiguity, caused by history. 'R' can be the
6640 recursive thing or the name 'R' (and similarly for 'R' followed by
6641 digits). We look for a name first; if not found, we try the other case.
6642
6643 For compatibility with auto-callouts, we allow a callout to be
6644 specified before a condition that is an assertion. First, check for the
6645 syntax of a callout; if found, adjust the temporary pointer that is
6646 used to check for an assertion condition. That's all that is needed! */
6647
6648 if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6649 {
6650 for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6651 if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6652 tempptr += i + 1;
6653 }
6654
6655 /* For conditions that are assertions, check the syntax, and then exit
6656 the switch. This will take control down to where bracketed groups,
6657 including assertions, are processed. */
6658
6659 if (tempptr[1] == CHAR_QUESTION_MARK &&
6660 (tempptr[2] == CHAR_EQUALS_SIGN ||