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

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1482 - (show annotations)
Wed May 28 11:10:58 2014 UTC (5 years, 2 months ago) by ph10
File MIME type: text/plain
File size: 311975 byte(s)
Fix two bugs concerned with duplicate named patterns.
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 processed string start */
51 #define PSEND end_pattern /* Field containing processed string end */
52
53 #include "pcre_internal.h"
54
55
56 /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57 is also used by pcretest. PCRE_DEBUG is not defined when building a production
58 library. We do not need to select pcre16_printint.c specially, because the
59 COMPILE_PCREx macro will already be appropriately set. */
60
61 #ifdef PCRE_DEBUG
62 /* pcre_printint.c should not include any headers */
63 #define PCRE_INCLUDED
64 #include "pcre_printint.c"
65 #undef PCRE_INCLUDED
66 #endif
67
68
69 /* Macro for setting individual bits in class bitmaps. */
70
71 #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72
73 /* Maximum length value to check against when making sure that the integer that
74 holds the compiled pattern length does not overflow. We make it a bit less than
75 INT_MAX to allow for adding in group terminating bytes, so that we don't have
76 to check them every time. */
77
78 #define OFLOW_MAX (INT_MAX - 20)
79
80 /* Definitions to allow mutual recursion */
81
82 static int
83 add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84 const pcre_uint32 *, unsigned int);
85
86 static BOOL
87 compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88 pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89 compile_data *, int *);
90
91
92
93 /*************************************************
94 * Code parameters and static tables *
95 *************************************************/
96
97 /* This value specifies the size of stack workspace that is used during the
98 first pre-compile phase that determines how much memory is required. The regex
99 is partly compiled into this space, but the compiled parts are discarded as
100 soon as they can be, so that hopefully there will never be an overrun. The code
101 does, however, check for an overrun. The largest amount I've seen used is 218,
102 so this number is very generous.
103
104 The same workspace is used during the second, actual compile phase for
105 remembering forward references to groups so that they can be filled in at the
106 end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107 is 4 there is plenty of room for most patterns. However, the memory can get
108 filled up by repetitions of forward references, for example patterns like
109 /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110 that the workspace is expanded using malloc() in this situation. The value
111 below is therefore a minimum, and we put a maximum on it for safety. The
112 minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113 kicks in at the same number of forward references in all cases. */
114
115 #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116 #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117
118 /* This value determines the size of the initial vector that is used for
119 remembering named groups during the pre-compile. It is allocated on the stack,
120 but if it is too small, it is expanded using malloc(), in a similar way to the
121 workspace. The value is the number of slots in the list. */
122
123 #define NAMED_GROUP_LIST_SIZE 20
124
125 /* The overrun tests check for a slightly smaller size so that they detect the
126 overrun before it actually does run off the end of the data block. */
127
128 #define WORK_SIZE_SAFETY_MARGIN (100)
129
130 /* Private flags added to firstchar and reqchar. */
131
132 #define REQ_CASELESS (1 << 0) /* Indicates caselessness */
133 #define REQ_VARY (1 << 1) /* Reqchar followed non-literal item */
134 /* Negative values for the firstchar and reqchar flags */
135 #define REQ_UNSET (-2)
136 #define REQ_NONE (-1)
137
138 /* Repeated character flags. */
139
140 #define UTF_LENGTH 0x10000000l /* The char contains its length. */
141
142 /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143 are simple data values; negative values are for special things like \d and so
144 on. Zero means further processing is needed (for things like \x), or the escape
145 is invalid. */
146
147 #ifndef EBCDIC
148
149 /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150 in UTF-8 mode. */
151
152 static const short int escapes[] = {
153 0, 0,
154 0, 0,
155 0, 0,
156 0, 0,
157 0, 0,
158 CHAR_COLON, CHAR_SEMICOLON,
159 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN,
160 CHAR_GREATER_THAN_SIGN, CHAR_QUESTION_MARK,
161 CHAR_COMMERCIAL_AT, -ESC_A,
162 -ESC_B, -ESC_C,
163 -ESC_D, -ESC_E,
164 0, -ESC_G,
165 -ESC_H, 0,
166 0, -ESC_K,
167 0, 0,
168 -ESC_N, 0,
169 -ESC_P, -ESC_Q,
170 -ESC_R, -ESC_S,
171 0, 0,
172 -ESC_V, -ESC_W,
173 -ESC_X, 0,
174 -ESC_Z, CHAR_LEFT_SQUARE_BRACKET,
175 CHAR_BACKSLASH, CHAR_RIGHT_SQUARE_BRACKET,
176 CHAR_CIRCUMFLEX_ACCENT, CHAR_UNDERSCORE,
177 CHAR_GRAVE_ACCENT, 7,
178 -ESC_b, 0,
179 -ESC_d, ESC_e,
180 ESC_f, 0,
181 -ESC_h, 0,
182 0, -ESC_k,
183 0, 0,
184 ESC_n, 0,
185 -ESC_p, 0,
186 ESC_r, -ESC_s,
187 ESC_tee, 0,
188 -ESC_v, -ESC_w,
189 0, 0,
190 -ESC_z
191 };
192
193 #else
194
195 /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
196
197 static const short int escapes[] = {
198 /* 48 */ 0, 0, 0, '.', '<', '(', '+', '|',
199 /* 50 */ '&', 0, 0, 0, 0, 0, 0, 0,
200 /* 58 */ 0, 0, '!', '$', '*', ')', ';', '~',
201 /* 60 */ '-', '/', 0, 0, 0, 0, 0, 0,
202 /* 68 */ 0, 0, '|', ',', '%', '_', '>', '?',
203 /* 70 */ 0, 0, 0, 0, 0, 0, 0, 0,
204 /* 78 */ 0, '`', ':', '#', '@', '\'', '=', '"',
205 /* 80 */ 0, 7, -ESC_b, 0, -ESC_d, ESC_e, ESC_f, 0,
206 /* 88 */-ESC_h, 0, 0, '{', 0, 0, 0, 0,
207 /* 90 */ 0, 0, -ESC_k, 'l', 0, ESC_n, 0, -ESC_p,
208 /* 98 */ 0, ESC_r, 0, '}', 0, 0, 0, 0,
209 /* A0 */ 0, '~', -ESC_s, ESC_tee, 0,-ESC_v, -ESC_w, 0,
210 /* A8 */ 0,-ESC_z, 0, 0, 0, '[', 0, 0,
211 /* B0 */ 0, 0, 0, 0, 0, 0, 0, 0,
212 /* B8 */ 0, 0, 0, 0, 0, ']', '=', '-',
213 /* C0 */ '{',-ESC_A, -ESC_B, -ESC_C, -ESC_D,-ESC_E, 0, -ESC_G,
214 /* C8 */-ESC_H, 0, 0, 0, 0, 0, 0, 0,
215 /* D0 */ '}', 0, -ESC_K, 0, 0,-ESC_N, 0, -ESC_P,
216 /* D8 */-ESC_Q,-ESC_R, 0, 0, 0, 0, 0, 0,
217 /* E0 */ '\\', 0, -ESC_S, 0, 0,-ESC_V, -ESC_W, -ESC_X,
218 /* E8 */ 0,-ESC_Z, 0, 0, 0, 0, 0, 0,
219 /* F0 */ 0, 0, 0, 0, 0, 0, 0, 0,
220 /* F8 */ 0, 0, 0, 0, 0, 0, 0, 0
221 };
222 #endif
223
224
225 /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226 searched linearly. Put all the names into a single string, in order to reduce
227 the number of relocations when a shared library is dynamically linked. The
228 string is built from string macros so that it works in UTF-8 mode on EBCDIC
229 platforms. */
230
231 typedef struct verbitem {
232 int len; /* Length of verb name */
233 int op; /* Op when no arg, or -1 if arg mandatory */
234 int op_arg; /* Op when arg present, or -1 if not allowed */
235 } verbitem;
236
237 static const char verbnames[] =
238 "\0" /* Empty name is a shorthand for MARK */
239 STRING_MARK0
240 STRING_ACCEPT0
241 STRING_COMMIT0
242 STRING_F0
243 STRING_FAIL0
244 STRING_PRUNE0
245 STRING_SKIP0
246 STRING_THEN;
247
248 static const verbitem verbs[] = {
249 { 0, -1, OP_MARK },
250 { 4, -1, OP_MARK },
251 { 6, OP_ACCEPT, -1 },
252 { 6, OP_COMMIT, -1 },
253 { 1, OP_FAIL, -1 },
254 { 4, OP_FAIL, -1 },
255 { 5, OP_PRUNE, OP_PRUNE_ARG },
256 { 4, OP_SKIP, OP_SKIP_ARG },
257 { 4, OP_THEN, OP_THEN_ARG }
258 };
259
260 static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261
262
263 /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264 another regex library. */
265
266 static const pcre_uchar sub_start_of_word[] = {
267 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268 CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269
270 static const pcre_uchar sub_end_of_word[] = {
271 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273 CHAR_RIGHT_PARENTHESIS, '\0' };
274
275
276 /* Tables of names of POSIX character classes and their lengths. The names are
277 now all in a single string, to reduce the number of relocations when a shared
278 library is dynamically loaded. The list of lengths is terminated by a zero
279 length entry. The first three must be alpha, lower, upper, as this is assumed
280 for handling case independence. The indices for graph, print, and punct are
281 needed, so identify them. */
282
283 static const char posix_names[] =
284 STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
285 STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
286 STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287 STRING_word0 STRING_xdigit;
288
289 static const pcre_uint8 posix_name_lengths[] = {
290 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291
292 #define PC_GRAPH 8
293 #define PC_PRINT 9
294 #define PC_PUNCT 10
295
296
297 /* Table of class bit maps for each POSIX class. Each class is formed from a
298 base map, with an optional addition or removal of another map. Then, for some
299 classes, there is some additional tweaking: for [:blank:] the vertical space
300 characters are removed, and for [:alpha:] and [:alnum:] the underscore
301 character is removed. The triples in the table consist of the base map offset,
302 second map offset or -1 if no second map, and a non-negative value for map
303 addition or a negative value for map subtraction (if there are two maps). The
304 absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
305 remove vertical space characters, 2 => remove underscore. */
306
307 static const int posix_class_maps[] = {
308 cbit_word, cbit_digit, -2, /* alpha */
309 cbit_lower, -1, 0, /* lower */
310 cbit_upper, -1, 0, /* upper */
311 cbit_word, -1, 2, /* alnum - word without underscore */
312 cbit_print, cbit_cntrl, 0, /* ascii */
313 cbit_space, -1, 1, /* blank - a GNU extension */
314 cbit_cntrl, -1, 0, /* cntrl */
315 cbit_digit, -1, 0, /* digit */
316 cbit_graph, -1, 0, /* graph */
317 cbit_print, -1, 0, /* print */
318 cbit_punct, -1, 0, /* punct */
319 cbit_space, -1, 0, /* space */
320 cbit_word, -1, 0, /* word - a Perl extension */
321 cbit_xdigit,-1, 0 /* xdigit */
322 };
323
324 /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325 Unicode property escapes. */
326
327 #ifdef SUPPORT_UCP
328 static const pcre_uchar string_PNd[] = {
329 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331 static const pcre_uchar string_pNd[] = {
332 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334 static const pcre_uchar string_PXsp[] = {
335 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337 static const pcre_uchar string_pXsp[] = {
338 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340 static const pcre_uchar string_PXwd[] = {
341 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343 static const pcre_uchar string_pXwd[] = {
344 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346
347 static const pcre_uchar *substitutes[] = {
348 string_PNd, /* \D */
349 string_pNd, /* \d */
350 string_PXsp, /* \S */ /* Xsp is Perl space, but from 8.34, Perl */
351 string_pXsp, /* \s */ /* space and POSIX space are the same. */
352 string_PXwd, /* \W */
353 string_pXwd /* \w */
354 };
355
356 /* The POSIX class substitutes must be in the order of the POSIX class names,
357 defined above, and there are both positive and negative cases. NULL means no
358 general substitute of a Unicode property escape (\p or \P). However, for some
359 POSIX classes (e.g. graph, print, punct) a special property code is compiled
360 directly. */
361
362 static const pcre_uchar string_pL[] = {
363 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365 static const pcre_uchar string_pLl[] = {
366 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368 static const pcre_uchar string_pLu[] = {
369 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371 static const pcre_uchar string_pXan[] = {
372 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374 static const pcre_uchar string_h[] = {
375 CHAR_BACKSLASH, CHAR_h, '\0' };
376 static const pcre_uchar string_pXps[] = {
377 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379 static const pcre_uchar string_PL[] = {
380 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382 static const pcre_uchar string_PLl[] = {
383 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385 static const pcre_uchar string_PLu[] = {
386 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388 static const pcre_uchar string_PXan[] = {
389 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391 static const pcre_uchar string_H[] = {
392 CHAR_BACKSLASH, CHAR_H, '\0' };
393 static const pcre_uchar string_PXps[] = {
394 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396
397 static const pcre_uchar *posix_substitutes[] = {
398 string_pL, /* alpha */
399 string_pLl, /* lower */
400 string_pLu, /* upper */
401 string_pXan, /* alnum */
402 NULL, /* ascii */
403 string_h, /* blank */
404 NULL, /* cntrl */
405 string_pNd, /* digit */
406 NULL, /* graph */
407 NULL, /* print */
408 NULL, /* punct */
409 string_pXps, /* space */ /* Xps is POSIX space, but from 8.34 */
410 string_pXwd, /* word */ /* Perl and POSIX space are the same */
411 NULL, /* xdigit */
412 /* Negated cases */
413 string_PL, /* ^alpha */
414 string_PLl, /* ^lower */
415 string_PLu, /* ^upper */
416 string_PXan, /* ^alnum */
417 NULL, /* ^ascii */
418 string_H, /* ^blank */
419 NULL, /* ^cntrl */
420 string_PNd, /* ^digit */
421 NULL, /* ^graph */
422 NULL, /* ^print */
423 NULL, /* ^punct */
424 string_PXps, /* ^space */ /* Xps is POSIX space, but from 8.34 */
425 string_PXwd, /* ^word */ /* Perl and POSIX space are the same */
426 NULL /* ^xdigit */
427 };
428 #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429 #endif
430
431 #define STRING(a) # a
432 #define XSTRING(s) STRING(s)
433
434 /* The texts of compile-time error messages. These are "char *" because they
435 are passed to the outside world. Do not ever re-use any error number, because
436 they are documented. Always add a new error instead. Messages marked DEAD below
437 are no longer used. This used to be a table of strings, but in order to reduce
438 the number of relocations needed when a shared library is loaded dynamically,
439 it is now one long string. We cannot use a table of offsets, because the
440 lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
441 simply count through to the one we want - this isn't a performance issue
442 because these strings are used only when there is a compilation error.
443
444 Each substring ends with \0 to insert a null character. This includes the final
445 substring, so that the whole string ends with \0\0, which can be detected when
446 counting through. */
447
448 static const char error_texts[] =
449 "no error\0"
450 "\\ at end of pattern\0"
451 "\\c at end of pattern\0"
452 "unrecognized character follows \\\0"
453 "numbers out of order in {} quantifier\0"
454 /* 5 */
455 "number too big in {} quantifier\0"
456 "missing terminating ] for character class\0"
457 "invalid escape sequence in character class\0"
458 "range out of order in character class\0"
459 "nothing to repeat\0"
460 /* 10 */
461 "operand of unlimited repeat could match the empty string\0" /** DEAD **/
462 "internal error: unexpected repeat\0"
463 "unrecognized character after (? or (?-\0"
464 "POSIX named classes are supported only within a class\0"
465 "missing )\0"
466 /* 15 */
467 "reference to non-existent subpattern\0"
468 "erroffset passed as NULL\0"
469 "unknown option bit(s) set\0"
470 "missing ) after comment\0"
471 "parentheses nested too deeply\0" /** DEAD **/
472 /* 20 */
473 "regular expression is too large\0"
474 "failed to get memory\0"
475 "unmatched parentheses\0"
476 "internal error: code overflow\0"
477 "unrecognized character after (?<\0"
478 /* 25 */
479 "lookbehind assertion is not fixed length\0"
480 "malformed number or name after (?(\0"
481 "conditional group contains more than two branches\0"
482 "assertion expected after (?(\0"
483 "(?R or (?[+-]digits must be followed by )\0"
484 /* 30 */
485 "unknown POSIX class name\0"
486 "POSIX collating elements are not supported\0"
487 "this version of PCRE is compiled without UTF support\0"
488 "spare error\0" /** DEAD **/
489 "character value in \\x{} or \\o{} is too large\0"
490 /* 35 */
491 "invalid condition (?(0)\0"
492 "\\C not allowed in lookbehind assertion\0"
493 "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
494 "number after (?C is > 255\0"
495 "closing ) for (?C expected\0"
496 /* 40 */
497 "recursive call could loop indefinitely\0"
498 "unrecognized character after (?P\0"
499 "syntax error in subpattern name (missing terminator)\0"
500 "two named subpatterns have the same name\0"
501 "invalid UTF-8 string\0"
502 /* 45 */
503 "support for \\P, \\p, and \\X has not been compiled\0"
504 "malformed \\P or \\p sequence\0"
505 "unknown property name after \\P or \\p\0"
506 "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
507 "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508 /* 50 */
509 "repeated subpattern is too long\0" /** DEAD **/
510 "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
511 "internal error: overran compiling workspace\0"
512 "internal error: previously-checked referenced subpattern not found\0"
513 "DEFINE group contains more than one branch\0"
514 /* 55 */
515 "repeating a DEFINE group is not allowed\0" /** DEAD **/
516 "inconsistent NEWLINE options\0"
517 "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
518 "a numbered reference must not be zero\0"
519 "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520 /* 60 */
521 "(*VERB) not recognized or malformed\0"
522 "number is too big\0"
523 "subpattern name expected\0"
524 "digit expected after (?+\0"
525 "] is an invalid data character in JavaScript compatibility mode\0"
526 /* 65 */
527 "different names for subpatterns of the same number are not allowed\0"
528 "(*MARK) must have an argument\0"
529 "this version of PCRE is not compiled with Unicode property support\0"
530 "\\c must be followed by an ASCII character\0"
531 "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532 /* 70 */
533 "internal error: unknown opcode in find_fixedlength()\0"
534 "\\N is not supported in a class\0"
535 "too many forward references\0"
536 "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537 "invalid UTF-16 string\0"
538 /* 75 */
539 "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540 "character value in \\u.... sequence is too large\0"
541 "invalid UTF-32 string\0"
542 "setting UTF is disabled by the application\0"
543 "non-hex character in \\x{} (closing brace missing?)\0"
544 /* 80 */
545 "non-octal character in \\o{} (closing brace missing?)\0"
546 "missing opening brace after \\o\0"
547 "parentheses are too deeply nested\0"
548 "invalid range in character class\0"
549 "group name must start with a non-digit\0"
550 /* 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_EXACT:
2133 case OP_EXACTI:
2134 case OP_UPTO:
2135 case OP_UPTOI:
2136 case OP_MINUPTO:
2137 case OP_MINUPTOI:
2138 case OP_POSUPTO:
2139 case OP_POSUPTOI:
2140 case OP_STAR:
2141 case OP_STARI:
2142 case OP_MINSTAR:
2143 case OP_MINSTARI:
2144 case OP_POSSTAR:
2145 case OP_POSSTARI:
2146 case OP_PLUS:
2147 case OP_PLUSI:
2148 case OP_MINPLUS:
2149 case OP_MINPLUSI:
2150 case OP_POSPLUS:
2151 case OP_POSPLUSI:
2152 case OP_QUERY:
2153 case OP_QUERYI:
2154 case OP_MINQUERY:
2155 case OP_MINQUERYI:
2156 case OP_POSQUERY:
2157 case OP_POSQUERYI:
2158 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2159 break;
2160 }
2161 #else
2162 (void)(utf); /* Keep compiler happy by referencing function argument */
2163 #endif
2164 }
2165 }
2166 }
2167
2168
2169
2170 /*************************************************
2171 * Scan compiled regex for recursion reference *
2172 *************************************************/
2173
2174 /* This little function scans through a compiled pattern until it finds an
2175 instance of OP_RECURSE.
2176
2177 Arguments:
2178 code points to start of expression
2179 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2180
2181 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2182 */
2183
2184 static const pcre_uchar *
2185 find_recurse(const pcre_uchar *code, BOOL utf)
2186 {
2187 for (;;)
2188 {
2189 register pcre_uchar c = *code;
2190 if (c == OP_END) return NULL;
2191 if (c == OP_RECURSE) return code;
2192
2193 /* XCLASS is used for classes that cannot be represented just by a bit
2194 map. This includes negated single high-valued characters. The length in
2195 the table is zero; the actual length is stored in the compiled code. */
2196
2197 if (c == OP_XCLASS) code += GET(code, 1);
2198
2199 /* Otherwise, we can get the item's length from the table, except that for
2200 repeated character types, we have to test for \p and \P, which have an extra
2201 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2202 must add in its length. */
2203
2204 else
2205 {
2206 switch(c)
2207 {
2208 case OP_TYPESTAR:
2209 case OP_TYPEMINSTAR:
2210 case OP_TYPEPLUS:
2211 case OP_TYPEMINPLUS:
2212 case OP_TYPEQUERY:
2213 case OP_TYPEMINQUERY:
2214 case OP_TYPEPOSSTAR:
2215 case OP_TYPEPOSPLUS:
2216 case OP_TYPEPOSQUERY:
2217 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2218 break;
2219
2220 case OP_TYPEPOSUPTO:
2221 case OP_TYPEUPTO:
2222 case OP_TYPEMINUPTO:
2223 case OP_TYPEEXACT:
2224 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2225 code += 2;
2226 break;
2227
2228 case OP_MARK:
2229 case OP_PRUNE_ARG:
2230 case OP_SKIP_ARG:
2231 case OP_THEN_ARG:
2232 code += code[1];
2233 break;
2234 }
2235
2236 /* Add in the fixed length from the table */
2237
2238 code += PRIV(OP_lengths)[c];
2239
2240 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2241 by a multi-byte character. The length in the table is a minimum, so we have
2242 to arrange to skip the extra bytes. */
2243
2244 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2245 if (utf) switch(c)
2246 {
2247 case OP_CHAR:
2248 case OP_CHARI:
2249 case OP_NOT:
2250 case OP_NOTI:
2251 case OP_EXACT:
2252 case OP_EXACTI:
2253 case OP_NOTEXACT:
2254 case OP_NOTEXACTI:
2255 case OP_UPTO:
2256 case OP_UPTOI:
2257 case OP_NOTUPTO:
2258 case OP_NOTUPTOI:
2259 case OP_MINUPTO:
2260 case OP_MINUPTOI:
2261 case OP_NOTMINUPTO:
2262 case OP_NOTMINUPTOI:
2263 case OP_POSUPTO:
2264 case OP_POSUPTOI:
2265 case OP_NOTPOSUPTO:
2266 case OP_NOTPOSUPTOI:
2267 case OP_STAR:
2268 case OP_STARI:
2269 case OP_NOTSTAR:
2270 case OP_NOTSTARI:
2271 case OP_MINSTAR:
2272 case OP_MINSTARI:
2273 case OP_NOTMINSTAR:
2274 case OP_NOTMINSTARI:
2275 case OP_POSSTAR:
2276 case OP_POSSTARI:
2277 case OP_NOTPOSSTAR:
2278 case OP_NOTPOSSTARI:
2279 case OP_PLUS:
2280 case OP_PLUSI:
2281 case OP_NOTPLUS:
2282 case OP_NOTPLUSI:
2283 case OP_MINPLUS:
2284 case OP_MINPLUSI:
2285 case OP_NOTMINPLUS:
2286 case OP_NOTMINPLUSI:
2287 case OP_POSPLUS:
2288 case OP_POSPLUSI:
2289 case OP_NOTPOSPLUS:
2290 case OP_NOTPOSPLUSI:
2291 case OP_QUERY:
2292 case OP_QUERYI:
2293 case OP_NOTQUERY:
2294 case OP_NOTQUERYI:
2295 case OP_MINQUERY:
2296 case OP_MINQUERYI:
2297 case OP_NOTMINQUERY:
2298 case OP_NOTMINQUERYI:
2299 case OP_POSQUERY:
2300 case OP_POSQUERYI:
2301 case OP_NOTPOSQUERY:
2302 case OP_NOTPOSQUERYI:
2303 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2304 break;
2305 }
2306 #else
2307 (void)(utf); /* Keep compiler happy by referencing function argument */
2308 #endif
2309 }
2310 }
2311 }
2312
2313
2314
2315 /*************************************************
2316 * Scan compiled branch for non-emptiness *
2317 *************************************************/
2318
2319 /* This function scans through a branch of a compiled pattern to see whether it
2320 can match the empty string or not. It is called from could_be_empty()
2321 below and from compile_branch() when checking for an unlimited repeat of a
2322 group that can match nothing. Note that first_significant_code() skips over
2323 backward and negative forward assertions when its final argument is TRUE. If we
2324 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2325 bracket whose current branch will already have been scanned.
2326
2327 Arguments:
2328 code points to start of search
2329 endcode points to where to stop
2330 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2331 cd contains pointers to tables etc.
2332 recurses chain of recurse_check to catch mutual recursion
2333
2334 Returns: TRUE if what is matched could be empty
2335 */
2336
2337 typedef struct recurse_check {
2338 struct recurse_check *prev;
2339 const pcre_uchar *group;
2340 } recurse_check;
2341
2342 static BOOL
2343 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2344 BOOL utf, compile_data *cd, recurse_check *recurses)
2345 {
2346 register pcre_uchar c;
2347 recurse_check this_recurse;
2348
2349 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2350 code < endcode;
2351 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2352 {
2353 const pcre_uchar *ccode;
2354
2355 c = *code;
2356
2357 /* Skip over forward assertions; the other assertions are skipped by
2358 first_significant_code() with a TRUE final argument. */
2359
2360 if (c == OP_ASSERT)
2361 {
2362 do code += GET(code, 1); while (*code == OP_ALT);
2363 c = *code;
2364 continue;
2365 }
2366
2367 /* For a recursion/subroutine call, if its end has been reached, which
2368 implies a backward reference subroutine call, we can scan it. If it's a
2369 forward reference subroutine call, we can't. To detect forward reference
2370 we have to scan up the list that is kept in the workspace. This function is
2371 called only when doing the real compile, not during the pre-compile that
2372 measures the size of the compiled pattern. */
2373
2374 if (c == OP_RECURSE)
2375 {
2376 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2377 BOOL empty_branch;
2378
2379 /* Test for forward reference or uncompleted reference. This is disabled
2380 when called to scan a completed pattern by setting cd->start_workspace to
2381 NULL. */
2382
2383 if (cd->start_workspace != NULL)
2384 {
2385 const pcre_uchar *tcode;
2386 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2387 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2388 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2389 }
2390
2391 /* If we are scanning a completed pattern, there are no forward references
2392 and all groups are complete. We need to detect whether this is a recursive
2393 call, as otherwise there will be an infinite loop. If it is a recursion,
2394 just skip over it. Simple recursions are easily detected. For mutual
2395 recursions we keep a chain on the stack. */
2396
2397 else
2398 {
2399 recurse_check *r = recurses;
2400 const pcre_uchar *endgroup = scode;
2401
2402 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2403 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2404
2405 for (r = recurses; r != NULL; r = r->prev)
2406 if (r->group == scode) break;
2407 if (r != NULL) continue; /* Mutual recursion */
2408 }
2409
2410 /* Completed reference; scan the referenced group, remembering it on the
2411 stack chain to detect mutual recursions. */
2412
2413 empty_branch = FALSE;
2414 this_recurse.prev = recurses;
2415 this_recurse.group = scode;
2416
2417 do
2418 {
2419 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2420 {
2421 empty_branch = TRUE;
2422 break;
2423 }
2424 scode += GET(scode, 1);
2425 }
2426 while (*scode == OP_ALT);
2427
2428 if (!empty_branch) return FALSE; /* All branches are non-empty */
2429 continue;
2430 }
2431
2432 /* Groups with zero repeats can of course be empty; skip them. */
2433
2434 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2435 c == OP_BRAPOSZERO)
2436 {
2437 code += PRIV(OP_lengths)[c];
2438 do code += GET(code, 1); while (*code == OP_ALT);
2439 c = *code;
2440 continue;
2441 }
2442
2443 /* A nested group that is already marked as "could be empty" can just be
2444 skipped. */
2445
2446 if (c == OP_SBRA || c == OP_SBRAPOS ||
2447 c == OP_SCBRA || c == OP_SCBRAPOS)
2448 {
2449 do code += GET(code, 1); while (*code == OP_ALT);
2450 c = *code;
2451 continue;
2452 }
2453
2454 /* For other groups, scan the branches. */
2455
2456 if (c == OP_BRA || c == OP_BRAPOS ||
2457 c == OP_CBRA || c == OP_CBRAPOS ||
2458 c == OP_ONCE || c == OP_ONCE_NC ||
2459 c == OP_COND)
2460 {
2461 BOOL empty_branch;
2462 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2463
2464 /* If a conditional group has only one branch, there is a second, implied,
2465 empty branch, so just skip over the conditional, because it could be empty.
2466 Otherwise, scan the individual branches of the group. */
2467
2468 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2469 code += GET(code, 1);
2470 else
2471 {
2472 empty_branch = FALSE;
2473 do
2474 {
2475 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2476 empty_branch = TRUE;
2477 code += GET(code, 1);
2478 }
2479 while (*code == OP_ALT);
2480 if (!empty_branch) return FALSE; /* All branches are non-empty */
2481 }
2482
2483 c = *code;
2484 continue;
2485 }
2486
2487 /* Handle the other opcodes */
2488
2489 switch (c)
2490 {
2491 /* Check for quantifiers after a class. XCLASS is used for classes that
2492 cannot be represented just by a bit map. This includes negated single
2493 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2494 actual length is stored in the compiled code, so we must update "code"
2495 here. */
2496
2497 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2498 case OP_XCLASS:
2499 ccode = code += GET(code, 1);
2500 goto CHECK_CLASS_REPEAT;
2501 #endif
2502
2503 case OP_CLASS:
2504 case OP_NCLASS:
2505 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2506
2507 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2508 CHECK_CLASS_REPEAT:
2509 #endif
2510
2511 switch (*ccode)
2512 {
2513 case OP_CRSTAR: /* These could be empty; continue */
2514 case OP_CRMINSTAR:
2515 case OP_CRQUERY:
2516 case OP_CRMINQUERY:
2517 case OP_CRPOSSTAR:
2518 case OP_CRPOSQUERY:
2519 break;
2520
2521 default: /* Non-repeat => class must match */
2522 case OP_CRPLUS: /* These repeats aren't empty */
2523 case OP_CRMINPLUS:
2524 case OP_CRPOSPLUS:
2525 return FALSE;
2526
2527 case OP_CRRANGE:
2528 case OP_CRMINRANGE:
2529 case OP_CRPOSRANGE:
2530 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2531 break;
2532 }
2533 break;
2534
2535 /* Opcodes that must match a character */
2536
2537 case OP_ANY:
2538 case OP_ALLANY:
2539 case OP_ANYBYTE:
2540
2541 case OP_PROP:
2542 case OP_NOTPROP:
2543 case OP_ANYNL:
2544
2545 case OP_NOT_HSPACE:
2546 case OP_HSPACE:
2547 case OP_NOT_VSPACE:
2548 case OP_VSPACE:
2549 case OP_EXTUNI:
2550
2551 case OP_NOT_DIGIT:
2552 case OP_DIGIT:
2553 case OP_NOT_WHITESPACE:
2554 case OP_WHITESPACE:
2555 case OP_NOT_WORDCHAR:
2556 case OP_WORDCHAR:
2557
2558 case OP_CHAR:
2559 case OP_CHARI:
2560 case OP_NOT:
2561 case OP_NOTI:
2562
2563 case OP_PLUS:
2564 case OP_PLUSI:
2565 case OP_MINPLUS:
2566 case OP_MINPLUSI:
2567
2568 case OP_NOTPLUS:
2569 case OP_NOTPLUSI:
2570 case OP_NOTMINPLUS:
2571 case OP_NOTMINPLUSI:
2572
2573 case OP_POSPLUS:
2574 case OP_POSPLUSI:
2575 case OP_NOTPOSPLUS:
2576 case OP_NOTPOSPLUSI:
2577
2578 case OP_EXACT:
2579 case OP_EXACTI:
2580 case OP_NOTEXACT:
2581 case OP_NOTEXACTI:
2582
2583 case OP_TYPEPLUS:
2584 case OP_TYPEMINPLUS:
2585 case OP_TYPEPOSPLUS:
2586 case OP_TYPEEXACT:
2587
2588 return FALSE;
2589
2590 /* These are going to continue, as they may be empty, but we have to
2591 fudge the length for the \p and \P cases. */
2592
2593 case OP_TYPESTAR:
2594 case OP_TYPEMINSTAR:
2595 case OP_TYPEPOSSTAR:
2596 case OP_TYPEQUERY:
2597 case OP_TYPEMINQUERY:
2598 case OP_TYPEPOSQUERY:
2599 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2600 break;
2601
2602 /* Same for these */
2603
2604 case OP_TYPEUPTO:
2605 case OP_TYPEMINUPTO:
2606 case OP_TYPEPOSUPTO:
2607 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2608 code += 2;
2609 break;
2610
2611 /* End of branch */
2612
2613 case OP_KET:
2614 case OP_KETRMAX:
2615 case OP_KETRMIN:
2616 case OP_KETRPOS:
2617 case OP_ALT:
2618 return TRUE;
2619
2620 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2621 MINUPTO, and POSUPTO and their caseless and negative versions may be
2622 followed by a multibyte character. */
2623
2624 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2625 case OP_STAR:
2626 case OP_STARI:
2627 case OP_NOTSTAR:
2628 case OP_NOTSTARI:
2629
2630 case OP_MINSTAR:
2631 case OP_MINSTARI:
2632 case OP_NOTMINSTAR:
2633 case OP_NOTMINSTARI:
2634
2635 case OP_POSSTAR:
2636 case OP_POSSTARI:
2637 case OP_NOTPOSSTAR:
2638 case OP_NOTPOSSTARI:
2639
2640 case OP_QUERY:
2641 case OP_QUERYI:
2642 case OP_NOTQUERY:
2643 case OP_NOTQUERYI:
2644
2645 case OP_MINQUERY:
2646 case OP_MINQUERYI:
2647 case OP_NOTMINQUERY:
2648 case OP_NOTMINQUERYI:
2649
2650 case OP_POSQUERY:
2651 case OP_POSQUERYI:
2652 case OP_NOTPOSQUERY:
2653 case OP_NOTPOSQUERYI:
2654
2655 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2656 break;
2657
2658 case OP_UPTO:
2659 case OP_UPTOI:
2660 case OP_NOTUPTO:
2661 case OP_NOTUPTOI:
2662
2663 case OP_MINUPTO:
2664 case OP_MINUPTOI:
2665 case OP_NOTMINUPTO:
2666 case OP_NOTMINUPTOI:
2667
2668 case OP_POSUPTO:
2669 case OP_POSUPTOI:
2670 case OP_NOTPOSUPTO:
2671 case OP_NOTPOSUPTOI:
2672
2673 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2674 break;
2675 #endif
2676
2677 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2678 string. */
2679
2680 case OP_MARK:
2681 case OP_PRUNE_ARG:
2682 case OP_SKIP_ARG:
2683 case OP_THEN_ARG:
2684 code += code[1];
2685 break;
2686
2687 /* None of the remaining opcodes are required to match a character. */
2688
2689 default:
2690 break;
2691 }
2692 }
2693
2694 return TRUE;
2695 }
2696
2697
2698
2699 /*************************************************
2700 * Scan compiled regex for non-emptiness *
2701 *************************************************/
2702
2703 /* This function is called to check for left recursive calls. We want to check
2704 the current branch of the current pattern to see if it could match the empty
2705 string. If it could, we must look outwards for branches at other levels,
2706 stopping when we pass beyond the bracket which is the subject of the recursion.
2707 This function is called only during the real compile, not during the
2708 pre-compile.
2709
2710 Arguments:
2711 code points to start of the recursion
2712 endcode points to where to stop (current RECURSE item)
2713 bcptr points to the chain of current (unclosed) branch starts
2714 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2715 cd pointers to tables etc
2716
2717 Returns: TRUE if what is matched could be empty
2718 */
2719
2720 static BOOL
2721 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2722 branch_chain *bcptr, BOOL utf, compile_data *cd)
2723 {
2724 while (bcptr != NULL && bcptr->current_branch >= code)
2725 {
2726 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2727 return FALSE;
2728 bcptr = bcptr->outer;
2729 }
2730 return TRUE;
2731 }
2732
2733
2734
2735 /*************************************************
2736 * Base opcode of repeated opcodes *
2737 *************************************************/
2738
2739 /* Returns the base opcode for repeated single character type opcodes. If the
2740 opcode is not a repeated character type, it returns with the original value.
2741
2742 Arguments: c opcode
2743 Returns: base opcode for the type
2744 */
2745
2746 static pcre_uchar
2747 get_repeat_base(pcre_uchar c)
2748 {
2749 return (c > OP_TYPEPOSUPTO)? c :
2750 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2751 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2752 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2753 (c >= OP_STARI)? OP_STARI :
2754 OP_STAR;
2755 }
2756
2757
2758
2759 #ifdef SUPPORT_UCP
2760 /*************************************************
2761 * Check a character and a property *
2762 *************************************************/
2763
2764 /* This function is called by check_auto_possessive() when a property item
2765 is adjacent to a fixed character.
2766
2767 Arguments:
2768 c the character
2769 ptype the property type
2770 pdata the data for the type
2771 negated TRUE if it's a negated property (\P or \p{^)
2772
2773 Returns: TRUE if auto-possessifying is OK
2774 */
2775
2776 static BOOL
2777 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2778 BOOL negated)
2779 {
2780 const pcre_uint32 *p;
2781 const ucd_record *prop = GET_UCD(c);
2782
2783 switch(ptype)
2784 {
2785 case PT_LAMP:
2786 return (prop->chartype == ucp_Lu ||
2787 prop->chartype == ucp_Ll ||
2788 prop->chartype == ucp_Lt) == negated;
2789
2790 case PT_GC:
2791 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2792
2793 case PT_PC:
2794 return (pdata == prop->chartype) == negated;
2795
2796 case PT_SC:
2797 return (pdata == prop->script) == negated;
2798
2799 /* These are specials */
2800
2801 case PT_ALNUM:
2802 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2803 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2804
2805 /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2806 means that Perl space and POSIX space are now identical. PCRE was changed
2807 at release 8.34. */
2808
2809 case PT_SPACE: /* Perl space */
2810 case PT_PXSPACE: /* POSIX space */
2811 switch(c)
2812 {
2813 HSPACE_CASES:
2814 VSPACE_CASES:
2815 return negated;
2816
2817 default:
2818 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2819 }
2820 break; /* Control never reaches here */
2821
2822 case PT_WORD:
2823 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2824 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2825 c == CHAR_UNDERSCORE) == negated;
2826
2827 case PT_CLIST:
2828 p = PRIV(ucd_caseless_sets) + prop->caseset;
2829 for (;;)
2830 {
2831 if (c < *p) return !negated;
2832 if (c == *p++) return negated;
2833 }
2834 break; /* Control never reaches here */
2835 }
2836
2837 return FALSE;
2838 }
2839 #endif /* SUPPORT_UCP */
2840
2841
2842
2843 /*************************************************
2844 * Fill the character property list *
2845 *************************************************/
2846
2847 /* Checks whether the code points to an opcode that can take part in auto-
2848 possessification, and if so, fills a list with its properties.
2849
2850 Arguments:
2851 code points to start of expression
2852 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2853 fcc points to case-flipping table
2854 list points to output list
2855 list[0] will be filled with the opcode
2856 list[1] will be non-zero if this opcode
2857 can match an empty character string
2858 list[2..7] depends on the opcode
2859
2860 Returns: points to the start of the next opcode if *code is accepted
2861 NULL if *code is not accepted
2862 */
2863
2864 static const pcre_uchar *
2865 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2866 const pcre_uint8 *fcc, pcre_uint32 *list)
2867 {
2868 pcre_uchar c = *code;
2869 pcre_uchar base;
2870 const pcre_uchar *end;
2871 pcre_uint32 chr;
2872
2873 #ifdef SUPPORT_UCP
2874 pcre_uint32 *clist_dest;
2875 const pcre_uint32 *clist_src;
2876 #else
2877 utf = utf; /* Suppress "unused parameter" compiler warning */
2878 #endif
2879
2880 list[0] = c;
2881 list[1] = FALSE;
2882 code++;
2883
2884 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2885 {
2886 base = get_repeat_base(c);
2887 c -= (base - OP_STAR);
2888
2889 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2890 code += IMM2_SIZE;
2891
2892 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2893
2894 switch(base)
2895 {
2896 case OP_STAR:
2897 list[0] = OP_CHAR;
2898 break;
2899
2900 case OP_STARI:
2901 list[0] = OP_CHARI;
2902 break;
2903
2904 case OP_NOTSTAR:
2905 list[0] = OP_NOT;
2906 break;
2907
2908 case OP_NOTSTARI:
2909 list[0] = OP_NOTI;
2910 break;
2911
2912 case OP_TYPESTAR:
2913 list[0] = *code;
2914 code++;
2915 break;
2916 }
2917 c = list[0];
2918 }
2919
2920 switch(c)
2921 {
2922 case OP_NOT_DIGIT:
2923 case OP_DIGIT:
2924 case OP_NOT_WHITESPACE:
2925 case OP_WHITESPACE:
2926 case OP_NOT_WORDCHAR:
2927 case OP_WORDCHAR:
2928 case OP_ANY:
2929 case OP_ALLANY:
2930 case OP_ANYNL:
2931 case OP_NOT_HSPACE:
2932 case OP_HSPACE:
2933 case OP_NOT_VSPACE:
2934 case OP_VSPACE:
2935 case OP_EXTUNI:
2936 case OP_EODN:
2937 case OP_EOD:
2938 case OP_DOLL:
2939 case OP_DOLLM:
2940 return code;
2941
2942 case OP_CHAR:
2943 case OP_NOT:
2944 GETCHARINCTEST(chr, code);
2945 list[2] = chr;
2946 list[3] = NOTACHAR;
2947 return code;
2948
2949 case OP_CHARI:
2950 case OP_NOTI:
2951 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2952 GETCHARINCTEST(chr, code);
2953 list[2] = chr;
2954
2955 #ifdef SUPPORT_UCP
2956 if (chr < 128 || (chr < 256 && !utf))
2957 list[3] = fcc[chr];
2958 else
2959 list[3] = UCD_OTHERCASE(chr);
2960 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2961 list[3] = (chr < 256) ? fcc[chr] : chr;
2962 #else
2963 list[3] = fcc[chr];
2964 #endif
2965
2966 /* The othercase might be the same value. */
2967
2968 if (chr == list[3])
2969 list[3] = NOTACHAR;
2970 else
2971 list[4] = NOTACHAR;
2972 return code;
2973
2974 #ifdef SUPPORT_UCP
2975 case OP_PROP:
2976 case OP_NOTPROP:
2977 if (code[0] != PT_CLIST)
2978 {
2979 list[2] = code[0];
2980 list[3] = code[1];
2981 return code + 2;
2982 }
2983
2984 /* Convert only if we have enough space. */
2985
2986 clist_src = PRIV(ucd_caseless_sets) + code[1];
2987 clist_dest = list + 2;
2988 code += 2;
2989
2990 do {
2991 if (clist_dest >= list + 8)
2992 {
2993 /* Early return if there is not enough space. This should never
2994 happen, since all clists are shorter than 5 character now. */
2995 list[2] = code[0];
2996 list[3] = code[1];
2997 return code;
2998 }
2999 *clist_dest++ = *clist_src;
3000 }
3001 while(*clist_src++ != NOTACHAR);
3002
3003 /* All characters are stored. The terminating NOTACHAR
3004 is copied form the clist itself. */
3005
3006 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3007 return code;
3008 #endif
3009
3010 case OP_NCLASS:
3011 case OP_CLASS:
3012 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3013 case OP_XCLASS:
3014 if (c == OP_XCLASS)
3015 end = code + GET(code, 0) - 1;
3016 else
3017 #endif
3018 end = code + 32 / sizeof(pcre_uchar);
3019
3020 switch(*end)
3021 {
3022 case OP_CRSTAR:
3023 case OP_CRMINSTAR:
3024 case OP_CRQUERY:
3025 case OP_CRMINQUERY:
3026 case OP_CRPOSSTAR:
3027 case OP_CRPOSQUERY:
3028 list[1] = TRUE;
3029 end++;
3030 break;
3031
3032 case OP_CRPLUS:
3033 case OP_CRMINPLUS:
3034 case OP_CRPOSPLUS:
3035 end++;
3036 break;
3037
3038 case OP_CRRANGE:
3039 case OP_CRMINRANGE:
3040 case OP_CRPOSRANGE:
3041 list[1] = (GET2(end, 1) == 0);
3042 end += 1 + 2 * IMM2_SIZE;
3043 break;
3044 }
3045 list[2] = (pcre_uint32)(end - code);
3046 return end;
3047 }
3048 return NULL; /* Opcode not accepted */
3049 }
3050
3051
3052
3053 /*************************************************
3054 * Scan further character sets for match *
3055 *************************************************/
3056
3057 /* Checks whether the base and the current opcode have a common character, in
3058 which case the base cannot be possessified.
3059
3060 Arguments:
3061 code points to the byte code
3062 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3063 cd static compile data
3064 base_list the data list of the base opcode
3065
3066 Returns: TRUE if the auto-possessification is possible
3067 */
3068
3069 static BOOL
3070 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3071 const pcre_uint32 *base_list, const pcre_uchar *base_end)
3072 {
3073 pcre_uchar c;
3074 pcre_uint32 list[8];
3075 const pcre_uint32 *chr_ptr;
3076 const pcre_uint32 *ochr_ptr;
3077 const pcre_uint32 *list_ptr;
3078 const pcre_uchar *next_code;
3079 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3080 const pcre_uchar *xclass_flags;
3081 #endif
3082 const pcre_uint8 *class_bitset;
3083 const pcre_uint8 *set1, *set2, *set_end;
3084 pcre_uint32 chr;
3085 BOOL accepted, invert_bits;
3086 BOOL entered_a_group = FALSE;
3087
3088 /* Note: the base_list[1] contains whether the current opcode has greedy
3089 (represented by a non-zero value) quantifier. This is a different from
3090 other character type lists, which stores here that the character iterator
3091 matches to an empty string (also represented by a non-zero value). */
3092
3093 for(;;)
3094 {
3095 /* All operations move the code pointer forward.
3096 Therefore infinite recursions are not possible. */
3097
3098 c = *code;
3099
3100 /* Skip over callouts */
3101
3102 if (c == OP_CALLOUT)
3103 {
3104 code += PRIV(OP_lengths)[c];
3105 continue;
3106 }
3107
3108 if (c == OP_ALT)
3109 {
3110 do code += GET(code, 1); while (*code == OP_ALT);
3111 c = *code;
3112 }
3113
3114 switch(c)
3115 {
3116 case OP_END:
3117 case OP_KETRPOS:
3118 /* TRUE only in greedy case. The non-greedy case could be replaced by
3119 an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3120 uses more memory, which we cannot get at this stage.) */
3121
3122 return base_list[1] != 0;
3123
3124 case OP_KET:
3125 /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3126 it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3127 cannot be converted to a possessive form. */
3128
3129 if (base_list[1] == 0) return FALSE;
3130
3131 switch(*(code - GET(code, 1)))
3132 {
3133 case OP_ASSERT:
3134 case OP_ASSERT_NOT:
3135 case OP_ASSERTBACK:
3136 case OP_ASSERTBACK_NOT:
3137 case OP_ONCE:
3138 case OP_ONCE_NC:
3139 /* Atomic sub-patterns and assertions can always auto-possessify their
3140 last iterator. However, if the group was entered as a result of checking
3141 a previous iterator, this is not possible. */
3142
3143 return !entered_a_group;
3144 }
3145
3146 code += PRIV(OP_lengths)[c];
3147 continue;
3148
3149 case OP_ONCE:
3150 case OP_ONCE_NC:
3151 case OP_BRA:
3152 case OP_CBRA:
3153 next_code = code + GET(code, 1);
3154 code += PRIV(OP_lengths)[c];
3155
3156 while (*next_code == OP_ALT)
3157 {
3158 if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3159 code = next_code + 1 + LINK_SIZE;
3160 next_code += GET(next_code, 1);
3161 }
3162
3163 entered_a_group = TRUE;
3164 continue;
3165
3166 case OP_BRAZERO:
3167 case OP_BRAMINZERO:
3168
3169 next_code = code + 1;
3170 if (*next_code != OP_BRA && *next_code != OP_CBRA
3171 && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3172
3173 do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3174
3175 /* The bracket content will be checked by the
3176 OP_BRA/OP_CBRA case above. */
3177 next_code += 1 + LINK_SIZE;
3178 if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3179 return FALSE;
3180
3181 code += PRIV(OP_lengths)[c];
3182 continue;
3183
3184 default:
3185 break;
3186 }
3187
3188 /* Check for a supported opcode, and load its properties. */
3189
3190 code = get_chr_property_list(code, utf, cd->fcc, list);
3191 if (code == NULL) return FALSE; /* Unsupported */
3192
3193 /* If either opcode is a small character list, set pointers for comparing
3194 characters from that list with another list, or with a property. */
3195
3196 if (base_list[0] == OP_CHAR)
3197 {
3198 chr_ptr = base_list + 2;
3199 list_ptr = list;
3200 }
3201 else if (list[0] == OP_CHAR)
3202 {
3203 chr_ptr = list + 2;
3204 list_ptr = base_list;
3205 }
3206
3207 /* Character bitsets can also be compared to certain opcodes. */
3208
3209 else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3210 #ifdef COMPILE_PCRE8
3211 /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3212 || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3213 #endif
3214 )
3215 {
3216 #ifdef COMPILE_PCRE8
3217 if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3218 #else
3219 if (base_list[0] == OP_CLASS)
3220 #endif
3221 {
3222 set1 = (pcre_uint8 *)(base_end - base_list[2]);
3223 list_ptr = list;
3224 }
3225 else
3226 {
3227 set1 = (pcre_uint8 *)(code - list[2]);
3228 list_ptr = base_list;
3229 }
3230
3231 invert_bits = FALSE;
3232 switch(list_ptr[0])
3233 {
3234 case OP_CLASS:
3235 case OP_NCLASS:
3236 set2 = (pcre_uint8 *)
3237 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3238 break;
3239
3240 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3241 case OP_XCLASS:
3242 xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3243 if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3244 if ((*xclass_flags & XCL_MAP) == 0)
3245 {
3246 /* No bits are set for characters < 256. */
3247 if (list[1] == 0) return TRUE;
3248 /* Might be an empty repeat. */
3249 continue;
3250 }
3251 set2 = (pcre_uint8 *)(xclass_flags + 1);
3252 break;
3253 #endif
3254
3255 case OP_NOT_DIGIT:
3256 invert_bits = TRUE;
3257 /* Fall through */
3258 case OP_DIGIT:
3259 set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3260 break;
3261
3262 case OP_NOT_WHITESPACE:
3263 invert_bits = TRUE;
3264 /* Fall through */
3265 case OP_WHITESPACE:
3266 set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3267 break;
3268
3269 case OP_NOT_WORDCHAR:
3270 invert_bits = TRUE;
3271 /* Fall through */
3272 case OP_WORDCHAR:
3273 set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3274 break;
3275
3276 default:
3277 return FALSE;
3278 }
3279
3280 /* Because the sets are unaligned, we need
3281 to perform byte comparison here. */
3282 set_end = set1 + 32;
3283 if (invert_bits)
3284 {
3285 do
3286 {
3287 if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3288 }
3289 while (set1 < set_end);
3290 }
3291 else
3292 {
3293 do
3294 {
3295 if ((*set1++ & *set2++) != 0) return FALSE;
3296 }
3297 while (set1 < set_end);
3298 }
3299
3300 if (list[1] == 0) return TRUE;
3301 /* Might be an empty repeat. */
3302 continue;
3303 }
3304
3305 /* Some property combinations also acceptable. Unicode property opcodes are
3306 processed specially; the rest can be handled with a lookup table. */
3307
3308 else
3309 {
3310 pcre_uint32 leftop, rightop;
3311
3312 leftop = base_list[0];
3313 rightop = list[0];
3314
3315 #ifdef SUPPORT_UCP
3316 accepted = FALSE; /* Always set in non-unicode case. */
3317 if (leftop == OP_PROP || leftop == OP_NOTPROP)
3318 {
3319 if (rightop == OP_EOD)
3320 accepted = TRUE;
3321 else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3322 {
3323 int n;
3324 const pcre_uint8 *p;
3325 BOOL same = leftop == rightop;
3326 BOOL lisprop = leftop == OP_PROP;
3327 BOOL risprop = rightop == OP_PROP;
3328 BOOL bothprop = lisprop && risprop;
3329
3330 /* There's a table that specifies how each combination is to be
3331 processed:
3332 0 Always return FALSE (never auto-possessify)
3333 1 Character groups are distinct (possessify if both are OP_PROP)
3334 2 Check character categories in the same group (general or particular)
3335 3 Return TRUE if the two opcodes are not the same
3336 ... see comments below
3337 */
3338
3339 n = propposstab[base_list[2]][list[2]];
3340 switch(n)
3341 {
3342 case 0: break;
3343 case 1: accepted = bothprop; break;
3344 case 2: accepted = (base_list[3] == list[3]) != same; break;
3345 case 3: accepted = !same; break;
3346
3347 case 4: /* Left general category, right particular category */
3348 accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3349 break;
3350
3351 case 5: /* Right general category, left particular category */
3352 accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3353 break;
3354
3355 /* This code is logically tricky. Think hard before fiddling with it.
3356 The posspropstab table has four entries per row. Each row relates to
3357 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3358 Only WORD actually needs all four entries, but using repeats for the
3359 others means they can all use the same code below.
3360
3361 The first two entries in each row are Unicode general categories, and
3362 apply always, because all the characters they include are part of the
3363 PCRE character set. The third and fourth entries are a general and a
3364 particular category, respectively, that include one or more relevant
3365 characters. One or the other is used, depending on whether the check
3366 is for a general or a particular category. However, in both cases the
3367 category contains more characters than the specials that are defined
3368 for the property being tested against. Therefore, it cannot be used
3369 in a NOTPROP case.
3370
3371 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3372 Underscore is covered by ucp_P or ucp_Po. */
3373
3374 case 6: /* Left alphanum vs right general category */
3375 case 7: /* Left space vs right general category */
3376 case 8: /* Left word vs right general category */
3377 p = posspropstab[n-6];
3378 accepted = risprop && lisprop ==
3379 (list[3] != p[0] &&
3380 list[3] != p[1] &&
3381 (list[3] != p[2] || !lisprop));
3382 break;
3383
3384 case 9: /* Right alphanum vs left general category */
3385 case 10: /* Right space vs left general category */
3386 case 11: /* Right word vs left general category */
3387 p = posspropstab[n-9];
3388 accepted = lisprop && risprop ==
3389 (base_list[3] != p[0] &&
3390 base_list[3] != p[1] &&
3391 (base_list[3] != p[2] || !risprop));
3392 break;
3393
3394 case 12: /* Left alphanum vs right particular category */
3395 case 13: /* Left space vs right particular category */
3396 case 14: /* Left word vs right particular category */
3397 p = posspropstab[n-12];
3398 accepted = risprop && lisprop ==
3399 (catposstab[p[0]][list[3]] &&
3400 catposstab[p[1]][list[3]] &&
3401 (list[3] != p[3] || !lisprop));
3402 break;
3403
3404 case 15: /* Right alphanum vs left particular category */
3405 case 16: /* Right space vs left particular category */
3406 case 17: /* Right word vs left particular category */
3407 p = posspropstab[n-15];
3408 accepted = lisprop && risprop ==
3409 (catposstab[p[0]][base_list[3]] &&
3410 catposstab[p[1]][base_list[3]] &&
3411 (base_list[3] != p[3] || !risprop));
3412 break;
3413 }
3414 }
3415 }
3416
3417 else
3418 #endif /* SUPPORT_UCP */
3419
3420 accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3421 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3422 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3423
3424 if (!accepted) return FALSE;
3425
3426 if (list[1] == 0) return TRUE;
3427 /* Might be an empty repeat. */
3428 continue;
3429 }
3430
3431 /* Control reaches here only if one of the items is a small character list.
3432 All characters are checked against the other side. */
3433
3434 do
3435 {
3436 chr = *chr_ptr;
3437
3438 switch(list_ptr[0])
3439 {
3440 case OP_CHAR:
3441 ochr_ptr = list_ptr + 2;
3442 do
3443 {
3444 if (chr == *ochr_ptr) return FALSE;
3445 ochr_ptr++;
3446 }
3447 while(*ochr_ptr != NOTACHAR);
3448 break;
3449
3450 case OP_NOT:
3451 ochr_ptr = list_ptr + 2;
3452 do
3453 {
3454 if (chr == *ochr_ptr)
3455 break;
3456 ochr_ptr++;
3457 }
3458 while(*ochr_ptr != NOTACHAR);
3459 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3460 break;
3461
3462 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3463 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3464
3465 case OP_DIGIT:
3466 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3467 break;
3468
3469 case OP_NOT_DIGIT:
3470 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3471 break;
3472
3473 case OP_WHITESPACE:
3474 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3475 break;
3476
3477 case OP_NOT_WHITESPACE:
3478 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3479 break;
3480
3481 case OP_WORDCHAR:
3482 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3483 break;
3484
3485 case OP_NOT_WORDCHAR:
3486 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3487 break;
3488
3489 case OP_HSPACE:
3490 switch(chr)
3491 {
3492 HSPACE_CASES: return FALSE;
3493 default: break;
3494 }
3495 break;
3496
3497 case OP_NOT_HSPACE:
3498 switch(chr)
3499 {
3500 HSPACE_CASES: break;
3501 default: return FALSE;
3502 }
3503 break;
3504
3505 case OP_ANYNL:
3506 case OP_VSPACE:
3507 switch(chr)
3508 {
3509 VSPACE_CASES: return FALSE;
3510 default: break;
3511 }
3512 break;
3513
3514 case OP_NOT_VSPACE:
3515 switch(chr)
3516 {
3517 VSPACE_CASES: break;
3518 default: return FALSE;
3519 }
3520 break;
3521
3522 case OP_DOLL:
3523 case OP_EODN:
3524 switch (chr)
3525 {
3526 case CHAR_CR:
3527 case CHAR_LF:
3528 case CHAR_VT:
3529 case CHAR_FF:
3530 case CHAR_NEL:
3531 #ifndef EBCDIC
3532 case 0x2028:
3533 case 0x2029:
3534 #endif /* Not EBCDIC */
3535 return FALSE;
3536 }
3537 break;
3538
3539 case OP_EOD: /* Can always possessify before \z */
3540 break;
3541
3542 #ifdef SUPPORT_UCP
3543 case OP_PROP:
3544 case OP_NOTPROP:
3545 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3546 list_ptr[0] == OP_NOTPROP))
3547 return FALSE;
3548 break;
3549 #endif
3550
3551 case OP_NCLASS:
3552 if (chr > 255) return FALSE;
3553 /* Fall through */
3554
3555 case OP_CLASS:
3556 if (chr > 255) break;
3557 class_bitset = (pcre_uint8 *)
3558 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3559 if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3560 break;
3561
3562 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3563 case OP_XCLASS:
3564 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3565 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3566 break;
3567 #endif
3568
3569 default:
3570 return FALSE;
3571 }
3572
3573 chr_ptr++;
3574 }
3575 while(*chr_ptr != NOTACHAR);
3576
3577 /* At least one character must be matched from this opcode. */
3578
3579 if (list[1] == 0) return TRUE;
3580 }
3581
3582 /* Control never reaches here. There used to be a fail-save return FALSE; here,
3583 but some compilers complain about an unreachable statement. */
3584
3585 }
3586
3587
3588
3589 /*************************************************
3590 * Scan compiled regex for auto-possession *
3591 *************************************************/
3592
3593 /* Replaces single character iterations with their possessive alternatives
3594 if appropriate. This function modifies the compiled opcode!
3595
3596 Arguments:
3597 code points to start of the byte code
3598 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3599 cd static compile data
3600
3601 Returns: nothing
3602 */
3603
3604 static void
3605 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3606 {
3607 register pcre_uchar c;
3608 const pcre_uchar *end;
3609 pcre_uchar *repeat_opcode;
3610 pcre_uint32 list[8];
3611
3612 for (;;)
3613 {
3614 c = *code;
3615
3616 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3617 {
3618 c -= get_repeat_base(c) - OP_STAR;
3619 end = (c <= OP_MINUPTO) ?
3620 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3621 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3622
3623 if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3624 {
3625 switch(c)
3626 {
3627 case OP_STAR:
3628 *code += OP_POSSTAR - OP_STAR;
3629 break;
3630
3631 case OP_MINSTAR:
3632 *code += OP_POSSTAR - OP_MINSTAR;
3633 break;
3634
3635 case OP_PLUS:
3636 *code += OP_POSPLUS - OP_PLUS;
3637 break;
3638
3639 case OP_MINPLUS:
3640 *code += OP_POSPLUS - OP_MINPLUS;
3641 break;
3642
3643 case OP_QUERY:
3644 *code += OP_POSQUERY - OP_QUERY;
3645 break;
3646
3647 case OP_MINQUERY:
3648 *code += OP_POSQUERY - OP_MINQUERY;
3649 break;
3650
3651 case OP_UPTO:
3652 *code += OP_POSUPTO - OP_UPTO;
3653 break;
3654
3655 case OP_MINUPTO:
3656 *code += OP_POSUPTO - OP_MINUPTO;
3657 break;
3658 }
3659 }
3660 c = *code;
3661 }
3662 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3663 {
3664 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3665 if (c == OP_XCLASS)
3666 repeat_opcode = code + GET(code, 1);
3667 else
3668 #endif
3669 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3670
3671 c = *repeat_opcode;
3672 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3673 {
3674 /* end must not be NULL. */
3675 end = get_chr_property_list(code, utf, cd->fcc, list);
3676
3677 list[1] = (c & 1) == 0;
3678
3679 if (compare_opcodes(end, utf, cd, list, end))
3680 {
3681 switch (c)
3682 {
3683 case OP_CRSTAR:
3684 case OP_CRMINSTAR:
3685 *repeat_opcode = OP_CRPOSSTAR;
3686 break;
3687
3688 case OP_CRPLUS:
3689 case OP_CRMINPLUS:
3690 *repeat_opcode = OP_CRPOSPLUS;
3691 break;
3692
3693 case OP_CRQUERY:
3694 case OP_CRMINQUERY:
3695 *repeat_opcode = OP_CRPOSQUERY;
3696 break;
3697
3698 case OP_CRRANGE:
3699 case OP_CRMINRANGE:
3700 *repeat_opcode = OP_CRPOSRANGE;
3701 break;
3702 }
3703 }
3704 }
3705 c = *code;
3706 }
3707
3708 switch(c)
3709 {
3710 case OP_END:
3711 return;
3712
3713 case OP_TYPESTAR:
3714 case OP_TYPEMINSTAR:
3715 case OP_TYPEPLUS:
3716 case OP_TYPEMINPLUS:
3717 case OP_TYPEQUERY:
3718 case OP_TYPEMINQUERY:
3719 case OP_TYPEPOSSTAR:
3720 case OP_TYPEPOSPLUS:
3721 case OP_TYPEPOSQUERY:
3722 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3723 break;
3724
3725 case OP_TYPEUPTO:
3726 case OP_TYPEMINUPTO:
3727 case OP_TYPEEXACT:
3728 case OP_TYPEPOSUPTO:
3729 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3730 code += 2;
3731 break;
3732
3733 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3734 case OP_XCLASS:
3735 code += GET(code, 1);
3736 break;
3737 #endif
3738
3739 case OP_MARK:
3740 case OP_PRUNE_ARG:
3741 case OP_SKIP_ARG:
3742 case OP_THEN_ARG:
3743 code += code[1];
3744 break;
3745 }
3746
3747 /* Add in the fixed length from the table */
3748
3749 code += PRIV(OP_lengths)[c];
3750
3751 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3752 a multi-byte character. The length in the table is a minimum, so we have to
3753 arrange to skip the extra bytes. */
3754
3755 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3756 if (utf) switch(c)
3757 {
3758 case OP_CHAR:
3759 case OP_CHARI:
3760 case OP_NOT:
3761 case OP_NOTI:
3762 case OP_STAR:
3763 case OP_MINSTAR:
3764 case OP_PLUS:
3765 case OP_MINPLUS:
3766 case OP_QUERY:
3767 case OP_MINQUERY:
3768 case OP_UPTO:
3769 case OP_MINUPTO:
3770 case OP_EXACT:
3771 case OP_POSSTAR:
3772 case OP_POSPLUS:
3773 case OP_POSQUERY:
3774 case OP_POSUPTO:
3775 case OP_STARI:
3776 case OP_MINSTARI:
3777 case OP_PLUSI:
3778 case OP_MINPLUSI:
3779 case OP_QUERYI:
3780 case OP_MINQUERYI:
3781 case OP_UPTOI:
3782 case OP_MINUPTOI:
3783 case OP_EXACTI:
3784 case OP_POSSTARI:
3785 case OP_POSPLUSI:
3786 case OP_POSQUERYI:
3787 case OP_POSUPTOI:
3788 case OP_NOTSTAR:
3789 case OP_NOTMINSTAR:
3790 case OP_NOTPLUS:
3791 case OP_NOTMINPLUS:
3792 case OP_NOTQUERY:
3793 case OP_NOTMINQUERY:
3794 case OP_NOTUPTO:
3795 case OP_NOTMINUPTO:
3796 case OP_NOTEXACT:
3797 case OP_NOTPOSSTAR:
3798 case OP_NOTPOSPLUS:
3799 case OP_NOTPOSQUERY:
3800 case OP_NOTPOSUPTO:
3801 case OP_NOTSTARI:
3802 case OP_NOTMINSTARI:
3803 case OP_NOTPLUSI:
3804 case OP_NOTMINPLUSI:
3805 case OP_NOTQUERYI:
3806 case OP_NOTMINQUERYI:
3807 case OP_NOTUPTOI:
3808 case OP_NOTMINUPTOI:
3809 case OP_NOTEXACTI:
3810 case OP_NOTPOSSTARI:
3811 case OP_NOTPOSPLUSI:
3812 case OP_NOTPOSQUERYI:
3813 case OP_NOTPOSUPTOI:
3814 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3815 break;
3816 }
3817 #else
3818 (void)(utf); /* Keep compiler happy by referencing function argument */
3819 #endif
3820 }
3821 }
3822
3823
3824
3825 /*************************************************
3826 * Check for POSIX class syntax *
3827 *************************************************/
3828
3829 /* This function is called when the sequence "[:" or "[." or "[=" is
3830 encountered in a character class. It checks whether this is followed by a
3831 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3832 reach an unescaped ']' without the special preceding character, return FALSE.
3833
3834 Originally, this function only recognized a sequence of letters between the
3835 terminators, but it seems that Perl recognizes any sequence of characters,
3836 though of course unknown POSIX names are subsequently rejected. Perl gives an
3837 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3838 didn't consider this to be a POSIX class. Likewise for [:1234:].
3839
3840 The problem in trying to be exactly like Perl is in the handling of escapes. We
3841 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3842 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3843 below handles the special case of \], but does not try to do any other escape
3844 processing. This makes it different from Perl for cases such as [:l\ower:]
3845 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3846 "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3847 I think.
3848
3849 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3850 It seems that the appearance of a nested POSIX class supersedes an apparent
3851 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3852 a digit.
3853
3854 In Perl, unescaped square brackets may also appear as part of class names. For
3855 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3856 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3857 seem right at all. PCRE does not allow closing square brackets in POSIX class
3858 names.
3859
3860 Arguments:
3861 ptr pointer to the initial [
3862 endptr where to return the end pointer
3863
3864 Returns: TRUE or FALSE
3865 */
3866
3867 static BOOL
3868 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3869 {
3870 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3871 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3872 for (++ptr; *ptr != CHAR_NULL; ptr++)
3873 {
3874 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3875 ptr++;
3876 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3877 else
3878 {
3879 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3880 {
3881 *endptr = ptr;
3882 return TRUE;
3883 }
3884 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3885 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3886 ptr[1] == CHAR_EQUALS_SIGN) &&
3887 check_posix_syntax(ptr, endptr))
3888 return FALSE;
3889 }
3890 }
3891 return FALSE;
3892 }
3893
3894
3895
3896
3897 /*************************************************
3898 * Check POSIX class name *
3899 *************************************************/
3900
3901 /* This function is called to check the name given in a POSIX-style class entry
3902 such as [:alnum:].
3903
3904 Arguments:
3905 ptr points to the first letter
3906 len the length of the name
3907
3908 Returns: a value representing the name, or -1 if unknown
3909 */
3910
3911 static int
3912 check_posix_name(const pcre_uchar *ptr, int len)
3913 {
3914 const char *pn = posix_names;
3915 register int yield = 0;
3916 while (posix_name_lengths[yield] != 0)
3917 {
3918 if (len == posix_name_lengths[yield] &&
3919 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3920 pn += posix_name_lengths[yield] + 1;
3921 yield++;
3922 }
3923 return -1;
3924 }
3925
3926
3927 /*************************************************
3928 * Adjust OP_RECURSE items in repeated group *
3929 *************************************************/
3930
3931 /* OP_RECURSE items contain an offset from the start of the regex to the group
3932 that is referenced. This means that groups can be replicated for fixed
3933 repetition simply by copying (because the recursion is allowed to refer to
3934 earlier groups that are outside the current group). However, when a group is
3935 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3936 inserted before it, after it has been compiled. This means that any OP_RECURSE
3937 items within it that refer to the group itself or any contained groups have to
3938 have their offsets adjusted. That one of the jobs of this function. Before it
3939 is called, the partially compiled regex must be temporarily terminated with
3940 OP_END.
3941
3942 This function has been extended with the possibility of forward references for
3943 recursions and subroutine calls. It must also check the list of such references
3944 for the group we are dealing with. If it finds that one of the recursions in
3945 the current group is on this list, it adjusts the offset in the list, not the
3946 value in the reference (which is a group number).
3947
3948 Arguments:
3949 group points to the start of the group
3950 adjust the amount by which the group is to be moved
3951 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3952 cd contains pointers to tables etc.
3953 save_hwm the hwm forward reference pointer at the start of the group
3954
3955 Returns: nothing
3956 */
3957
3958 static void
3959 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3960 pcre_uchar *save_hwm)
3961 {
3962 pcre_uchar *ptr = group;
3963
3964 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3965 {
3966 int offset;
3967 pcre_uchar *hc;
3968
3969 /* See if this recursion is on the forward reference list. If so, adjust the
3970 reference. */
3971
3972 for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3973 {
3974 offset = (int)GET(hc, 0);
3975 if (cd->start_code + offset == ptr + 1)
3976 {
3977 PUT(hc, 0, offset + adjust);
3978 break;
3979 }
3980 }
3981
3982 /* Otherwise, adjust the recursion offset if it's after the start of this
3983 group. */
3984
3985 if (hc >= cd->hwm)
3986 {
3987 offset = (int)GET(ptr, 1);
3988 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3989 }
3990
3991 ptr += 1 + LINK_SIZE;
3992 }
3993 }
3994
3995
3996
3997 /*************************************************
3998 * Insert an automatic callout point *
3999 *************************************************/
4000
4001 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
4002 callout points before each pattern item.
4003
4004 Arguments:
4005 code current code pointer
4006 ptr current pattern pointer
4007 cd pointers to tables etc
4008
4009 Returns: new code pointer
4010 */
4011
4012 static pcre_uchar *
4013 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4014 {
4015 *code++ = OP_CALLOUT;
4016 *code++ = 255;
4017 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
4018 PUT(code, LINK_SIZE, 0); /* Default length */
4019 return code + 2 * LINK_SIZE;
4020 }
4021
4022
4023
4024 /*************************************************
4025 * Complete a callout item *
4026 *************************************************/
4027
4028 /* A callout item contains the length of the next item in the pattern, which
4029 we can't fill in till after we have reached the relevant point. This is used
4030 for both automatic and manual callouts.
4031
4032 Arguments:
4033 previous_callout points to previous callout item
4034 ptr current pattern pointer
4035 cd pointers to tables etc
4036
4037 Returns: nothing
4038 */
4039
4040 static void
4041 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4042 {
4043 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4044 PUT(previous_callout, 2 + LINK_SIZE, length);
4045 }
4046
4047
4048
4049 #ifdef SUPPORT_UCP
4050 /*************************************************
4051 * Get othercase range *
4052 *************************************************/
4053
4054 /* This function is passed the start and end of a class range, in UTF-8 mode
4055 with UCP support. It searches up the characters, looking for ranges of
4056 characters in the "other" case. Each call returns the next one, updating the
4057 start address. A character with multiple other cases is returned on its own
4058 with a special return value.
4059
4060 Arguments:
4061 cptr points to starting character value; updated
4062 d end value
4063 ocptr where to put start of othercase range
4064 odptr where to put end of othercase range
4065
4066 Yield: -1 when no more
4067 0 when a range is returned
4068 >0 the CASESET offset for char with multiple other cases
4069 in this case, ocptr contains the original
4070 */
4071
4072 static int
4073 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4074 pcre_uint32 *odptr)
4075 {
4076 pcre_uint32 c, othercase, next;
4077 unsigned int co;
4078
4079 /* Find the first character that has an other case. If it has multiple other
4080 cases, return its case offset value. */
4081
4082 for (c = *cptr; c <= d; c++)
4083 {
4084 if ((co = UCD_CASESET(c)) != 0)
4085 {
4086 *ocptr = c++; /* Character that has the set */
4087 *cptr = c; /* Rest of input range */
4088 return (int)co;
4089 }
4090 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4091 }
4092
4093 if (c > d) return -1; /* Reached end of range */
4094
4095 /* Found a character that has a single other case. Search for the end of the
4096 range, which is either the end of the input range, or a character that has zero
4097 or more than one other cases. */
4098
4099 *ocptr = othercase;
4100 next = othercase + 1;
4101
4102 for (++c; c <= d; c++)
4103 {
4104 if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4105 next++;
4106 }
4107
4108 *odptr = next - 1; /* End of othercase range */
4109 *cptr = c; /* Rest of input range */
4110 return 0;
4111 }
4112 #endif /* SUPPORT_UCP */
4113
4114
4115
4116 /*************************************************
4117 * Add a character or range to a class *
4118 *************************************************/
4119
4120 /* This function packages up the logic of adding a character or range of
4121 characters to a class. The character values in the arguments will be within the
4122 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4123 mutually recursive with the function immediately below.
4124
4125 Arguments:
4126 classbits the bit map for characters < 256
4127 uchardptr points to the pointer for extra data
4128 options the options word
4129 cd contains pointers to tables etc.
4130 start start of range character
4131 end end of range character
4132
4133 Returns: the number of < 256 characters added
4134 the pointer to extra data is updated
4135 */
4136
4137 static int
4138 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4139 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4140 {
4141 pcre_uint32 c;
4142 pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4143 int n8 = 0;
4144
4145 /* If caseless matching is required, scan the range and process alternate
4146 cases. In Unicode, there are 8-bit characters that have alternate cases that
4147 are greater than 255 and vice-versa. Sometimes we can just extend the original
4148 range. */
4149
4150 if ((options & PCRE_CASELESS) != 0)
4151 {
4152 #ifdef SUPPORT_UCP
4153 if ((options & PCRE_UTF8) != 0)
4154 {
4155 int rc;
4156 pcre_uint32 oc, od;
4157
4158 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4159 c = start;
4160
4161 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4162 {
4163 /* Handle a single character that has more than one other case. */
4164
4165 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4166 PRIV(ucd_caseless_sets) + rc, oc);
4167
4168 /* Do nothing if the other case range is within the original range. */
4169
4170 else if (oc >= start && od <= end) continue;
4171
4172 /* Extend the original range if there is overlap, noting that if oc < c, we
4173 can't have od > end because a subrange is always shorter than the basic
4174 range. Otherwise, use a recursive call to add the additional range. */
4175
4176 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4177 else if (od > end && oc <= end + 1) end = od; /* Extend upwards */
4178 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4179 }
4180 }
4181 else
4182 #endif /* SUPPORT_UCP */
4183
4184 /* Not UTF-mode, or no UCP */
4185
4186 for (c = start; c <= classbits_end; c++)
4187 {
4188 SETBIT(classbits, cd->fcc[c]);
4189 n8++;
4190 }
4191 }
4192
4193 /* Now handle the original range. Adjust the final value according to the bit
4194 length - this means that the same lists of (e.g.) horizontal spaces can be used
4195 in all cases. */
4196
4197 #if defined COMPILE_PCRE8
4198 #ifdef SUPPORT_UTF
4199 if ((options & PCRE_UTF8) == 0)
4200 #endif
4201 if (end > 0xff) end = 0xff;
4202
4203 #elif defined COMPILE_PCRE16
4204 #ifdef SUPPORT_UTF
4205 if ((options & PCRE_UTF16) == 0)
4206 #endif
4207 if (end > 0xffff) end = 0xffff;
4208
4209 #endif /* COMPILE_PCRE[8|16] */
4210
4211 /* Use the bitmap for characters < 256. Otherwise use extra data.*/
4212
4213 for (c = start; c <= classbits_end; c++)
4214 {
4215 /* Regardless of start, c will always be <= 255. */
4216 SETBIT(classbits, c);
4217 n8++;
4218 }
4219
4220 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4221 if (start <= 0xff) start = 0xff + 1;
4222
4223 if (end >= start)
4224 {
4225 pcre_uchar *uchardata = *uchardptr;
4226 #ifdef SUPPORT_UTF
4227 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4228 {
4229 if (start < end)
4230 {
4231 *uchardata++ = XCL_RANGE;
4232 uchardata += PRIV(ord2utf)(start, uchardata);
4233 uchardata += PRIV(ord2utf)(end, uchardata);
4234 }
4235 else if (start == end)
4236 {
4237 *uchardata++ = XCL_SINGLE;
4238 uchardata += PRIV(ord2utf)(start, uchardata);
4239 }
4240 }
4241 else
4242 #endif /* SUPPORT_UTF */
4243
4244 /* Without UTF support, character values are constrained by the bit length,
4245 and can only be > 256 for 16-bit and 32-bit libraries. */
4246
4247 #ifdef COMPILE_PCRE8
4248 {}
4249 #else
4250 if (start < end)
4251 {
4252 *uchardata++ = XCL_RANGE;
4253 *uchardata++ = start;
4254 *uchardata++ = end;
4255 }
4256 else if (start == end)
4257 {
4258 *uchardata++ = XCL_SINGLE;
4259 *uchardata++ = start;
4260 }
4261 #endif
4262
4263 *uchardptr = uchardata; /* Updata extra data pointer */
4264 }
4265 #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4266
4267 return n8; /* Number of 8-bit characters */
4268 }
4269
4270
4271
4272
4273 /*************************************************
4274 * Add a list of characters to a class *
4275 *************************************************/
4276
4277 /* This function is used for adding a list of case-equivalent characters to a
4278 class, and also for adding a list of horizontal or vertical whitespace. If the
4279 list is in order (which it should be), ranges of characters are detected and
4280 handled appropriately. This function is mutually recursive with the function
4281 above.
4282
4283 Arguments:
4284 classbits the bit map for characters < 256
4285 uchardptr points to the pointer for extra data
4286 options the options word
4287 cd contains pointers to tables etc.
4288 p points to row of 32-bit values, terminated by NOTACHAR
4289 except character to omit; this is used when adding lists of
4290 case-equivalent characters to avoid including the one we
4291 already know about
4292
4293 Returns: the number of < 256 characters added
4294 the pointer to extra data is updated
4295 */
4296
4297 static int
4298 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4299 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4300 {
4301 int n8 = 0;
4302 while (p[0] < NOTACHAR)
4303 {
4304 int n = 0;
4305 if (p[0] != except)
4306 {
4307 while(p[n+1] == p[0] + n + 1) n++;
4308 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4309 }
4310 p += n + 1;
4311 }
4312 return n8;
4313 }
4314
4315
4316
4317 /*************************************************
4318 * Add characters not in a list to a class *
4319 *************************************************/
4320
4321 /* This function is used for adding the complement of a list of horizontal or
4322 vertical whitespace to a class. The list must be in order.
4323
4324 Arguments:
4325 classbits the bit map for characters < 256
4326 uchardptr points to the pointer for extra data
4327 options the options word
4328 cd contains pointers to tables etc.
4329 p points to row of 32-bit values, terminated by NOTACHAR
4330
4331 Returns: the number of < 256 characters added
4332 the pointer to extra data is updated
4333 */
4334
4335 static int
4336 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4337 int options, compile_data *cd, const pcre_uint32 *p)
4338 {
4339 BOOL utf = (options & PCRE_UTF8) != 0;
4340 int n8 = 0;
4341 if (p[0] > 0)
4342 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4343 while (p[0] < NOTACHAR)
4344 {
4345 while (p[1] == p[0] + 1) p++;
4346 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4347 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4348 p++;
4349 }
4350 return n8;
4351 }
4352
4353
4354
4355 /*************************************************
4356 * Compile one branch *
4357 *************************************************/
4358
4359 /* Scan the pattern, compiling it into the a vector. If the options are
4360 changed during the branch, the pointer is used to change the external options
4361 bits. This function is used during the pre-compile phase when we are trying
4362 to find out the amount of memory needed, as well as during the real compile
4363 phase. The value of lengthptr distinguishes the two phases.
4364
4365 Arguments:
4366 optionsptr pointer to the option bits
4367 codeptr points to the pointer to the current code point
4368 ptrptr points to the current pattern pointer
4369 errorcodeptr points to error code variable
4370 firstcharptr place to put the first required character
4371 firstcharflagsptr place to put the first character flags, or a negative number
4372 reqcharptr place to put the last required character
4373 reqcharflagsptr place to put the last required character flags, or a negative number
4374 bcptr points to current branch chain
4375 cond_depth conditional nesting depth
4376 cd contains pointers to tables etc.
4377 lengthptr NULL during the real compile phase
4378 points to length accumulator during pre-compile phase
4379
4380 Returns: TRUE on success
4381 FALSE, with *errorcodeptr set non-zero on error
4382 */
4383
4384 static BOOL
4385 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4386 const pcre_uchar **ptrptr, int *errorcodeptr,
4387 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4388 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4389 branch_chain *bcptr, int cond_depth,
4390 compile_data *cd, int *lengthptr)
4391 {
4392 int repeat_type, op_type;
4393 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4394 int bravalue = 0;
4395 int greedy_default, greedy_non_default;
4396 pcre_uint32 firstchar, reqchar;
4397 pcre_int32 firstcharflags, reqcharflags;
4398 pcre_uint32 zeroreqchar, zerofirstchar;
4399 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4400 pcre_int32 req_caseopt, reqvary, tempreqvary;
4401 int options = *optionsptr; /* May change dynamically */
4402 int after_manual_callout = 0;
4403 int length_prevgroup = 0;
4404 register pcre_uint32 c;
4405 int escape;
4406 register pcre_uchar *code = *codeptr;
4407 pcre_uchar *last_code = code;
4408 pcre_uchar *orig_code = code;
4409 pcre_uchar *tempcode;
4410 BOOL inescq = FALSE;
4411 BOOL groupsetfirstchar = FALSE;
4412 const pcre_uchar *ptr = *ptrptr;
4413 const pcre_uchar *tempptr;
4414 const pcre_uchar *nestptr = NULL;
4415 pcre_uchar *previous = NULL;
4416 pcre_uchar *previous_callout = NULL;
4417 pcre_uchar *save_hwm = NULL;
4418 pcre_uint8 classbits[32];
4419
4420 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4421 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4422 dynamically as we process the pattern. */
4423
4424 #ifdef SUPPORT_UTF
4425 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4426 BOOL utf = (options & PCRE_UTF8) != 0;
4427 #ifndef COMPILE_PCRE32
4428 pcre_uchar utf_chars[6];
4429 #endif
4430 #else
4431 BOOL utf = FALSE;
4432 #endif
4433
4434 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4435 class_uchardata always so that it can be passed to add_to_class() always,
4436 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4437 alternative calls for the different cases. */
4438
4439 pcre_uchar *class_uchardata;
4440 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4441 BOOL xclass;
4442 pcre_uchar *class_uchardata_base;
4443 #endif
4444
4445 #ifdef PCRE_DEBUG
4446 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4447 #endif
4448
4449 /* Set up the default and non-default settings for greediness */
4450
4451 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4452 greedy_non_default = greedy_default ^ 1;
4453
4454 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4455 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4456 matches a non-fixed char first char; reqchar just remains unset if we never
4457 find one.
4458
4459 When we hit a repeat whose minimum is zero, we may have to adjust these values
4460 to take the zero repeat into account. This is implemented by setting them to
4461 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4462 item types that can be repeated set these backoff variables appropriately. */
4463
4464 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4465 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4466
4467 /* The variable req_caseopt contains either the REQ_CASELESS value
4468 or zero, according to the current setting of the caseless flag. The
4469 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4470 firstchar or reqchar variables to record the case status of the
4471 value. This is used only for ASCII characters. */
4472
4473 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4474
4475 /* Switch on next character until the end of the branch */
4476
4477 for (;; ptr++)
4478 {
4479 BOOL negate_class;
4480 BOOL should_flip_negation;
4481 BOOL possessive_quantifier;
4482 BOOL is_quantifier;
4483 BOOL is_recurse;
4484 BOOL reset_bracount;
4485 int class_has_8bitchar;
4486 int class_one_char;
4487 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4488 BOOL xclass_has_prop;
4489 #endif
4490 int newoptions;
4491 int recno;
4492 int refsign;
4493 int skipbytes;
4494 pcre_uint32 subreqchar, subfirstchar;
4495 pcre_int32 subreqcharflags, subfirstcharflags;
4496 int terminator;
4497 unsigned int mclength;
4498 unsigned int tempbracount;
4499 pcre_uint32 ec;
4500 pcre_uchar mcbuffer[8];
4501
4502 /* Get next character in the pattern */
4503
4504 c = *ptr;
4505
4506 /* If we are at the end of a nested substitution, revert to the outer level
4507 string. Nesting only happens one level deep. */
4508
4509 if (c == CHAR_NULL && nestptr != NULL)
4510 {
4511 ptr = nestptr;
4512 nestptr = NULL;
4513 c = *ptr;
4514 }
4515
4516 /* If we are in the pre-compile phase, accumulate the length used for the
4517 previous cycle of this loop. */
4518
4519 if (lengthptr != NULL)
4520 {
4521 #ifdef PCRE_DEBUG
4522 if (code > cd->hwm) cd->hwm = code; /* High water info */
4523 #endif
4524 if (code > cd->start_workspace + cd->workspace_size -
4525 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4526 {
4527 *errorcodeptr = ERR52;
4528 goto FAILED;
4529 }
4530
4531 /* There is at least one situation where code goes backwards: this is the
4532 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4533 the class is simply eliminated. However, it is created first, so we have to
4534 allow memory for it. Therefore, don't ever reduce the length at this point.
4535 */
4536
4537 if (code < last_code) code = last_code;
4538
4539 /* Paranoid check for integer overflow */
4540
4541 if (OFLOW_MAX - *lengthptr < code - last_code)
4542 {
4543 *errorcodeptr = ERR20;
4544 goto FAILED;
4545 }
4546
4547 *lengthptr += (int)(code - last_code);
4548 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4549 (int)(code - last_code), c, c));
4550
4551 /* If "previous" is set and it is not at the start of the work space, move
4552 it back to there, in order to avoid filling up the work space. Otherwise,
4553 if "previous" is NULL, reset the current code pointer to the start. */
4554
4555 if (previous != NULL)
4556 {
4557 if (previous > orig_code)
4558 {
4559 memmove(orig_code, previous, IN_UCHARS(code - previous));
4560 code -= previous - orig_code;
4561 previous = orig_code;
4562 }
4563 }
4564 else code = orig_code;
4565
4566 /* Remember where this code item starts so we can pick up the length
4567 next time round. */
4568
4569 last_code = code;
4570 }
4571
4572 /* In the real compile phase, just check the workspace used by the forward
4573 reference list. */
4574
4575 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4576 WORK_SIZE_SAFETY_MARGIN)
4577 {
4578 *errorcodeptr = ERR52;
4579 goto FAILED;
4580 }
4581
4582 /* If in \Q...\E, check for the end; if not, we have a literal */
4583
4584 if (inescq && c != CHAR_NULL)
4585 {
4586 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4587 {
4588 inescq = FALSE;
4589 ptr++;
4590 continue;
4591 }
4592 else
4593 {
4594 if (previous_callout != NULL)
4595 {
4596 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4597 complete_callout(previous_callout, ptr, cd);
4598 previous_callout = NULL;
4599 }
4600 if ((options & PCRE_AUTO_CALLOUT) != 0)
4601 {
4602 previous_callout = code;
4603 code = auto_callout(code, ptr, cd);
4604 }
4605 goto NORMAL_CHAR;
4606 }
4607 /* Control does not reach here. */
4608 }
4609
4610 /* In extended mode, skip white space and comments. We need a loop in order
4611 to check for more white space and more comments after a comment. */
4612
4613 if ((options & PCRE_EXTENDED) != 0)
4614 {
4615 for (;;)
4616 {
4617 while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4618 if (c != CHAR_NUMBER_SIGN) break;
4619 ptr++;
4620 while (*ptr != CHAR_NULL)
4621 {
4622 if (IS_NEWLINE(ptr)) /* For non-fixed-length newline cases, */
4623 { /* IS_NEWLINE sets cd->nllen. */
4624 ptr += cd->nllen;
4625 break;
4626 }
4627 ptr++;
4628 #ifdef SUPPORT_UTF
4629 if (utf) FORWARDCHAR(ptr);
4630 #endif
4631 }
4632 c = *ptr; /* Either NULL or the char after a newline */
4633 }
4634 }
4635
4636 /* See if the next thing is a quantifier. */
4637
4638 is_quantifier =
4639 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4640 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4641
4642 /* Fill in length of a previous callout, except when the next thing is a
4643 quantifier or when processing a property substitution string in UCP mode. */
4644
4645 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4646 after_manual_callout-- <= 0)
4647 {
4648 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4649 complete_callout(previous_callout, ptr, cd);
4650 previous_callout = NULL;
4651 }
4652
4653 /* Create auto callout, except for quantifiers, or while processing property
4654 strings that are substituted for \w etc in UCP mode. */
4655
4656 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4657 {
4658 previous_callout = code;
4659 code = auto_callout(code, ptr, cd);
4660 }
4661
4662 /* Process the next pattern item. */
4663
4664 switch(c)
4665 {
4666 /* ===================================================================*/
4667 case CHAR_NULL: /* The branch terminates at string end */
4668 case CHAR_VERTICAL_LINE: /* or | or ) */
4669 case CHAR_RIGHT_PARENTHESIS:
4670 *firstcharptr = firstchar;
4671 *firstcharflagsptr = firstcharflags;
4672 *reqcharptr = reqchar;
4673 *reqcharflagsptr = reqcharflags;
4674 *codeptr = code;
4675 *ptrptr = ptr;
4676 if (lengthptr != NULL)
4677 {
4678 if (OFLOW_MAX - *lengthptr < code - last_code)
4679 {
4680 *errorcodeptr = ERR20;
4681 goto FAILED;
4682 }
4683 *lengthptr += (int)(code - last_code); /* To include callout length */
4684 DPRINTF((">> end branch\n"));
4685 }
4686 return TRUE;
4687
4688
4689 /* ===================================================================*/
4690 /* Handle single-character metacharacters. In multiline mode, ^ disables
4691 the setting of any following char as a first character. */
4692
4693 case CHAR_CIRCUMFLEX_ACCENT:
4694 previous = NULL;
4695 if ((options & PCRE_MULTILINE) != 0)
4696 {
4697 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4698 *code++ = OP_CIRCM;
4699 }
4700 else *code++ = OP_CIRC;
4701 break;
4702
4703 case CHAR_DOLLAR_SIGN:
4704 previous = NULL;
4705 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4706 break;
4707
4708 /* There can never be a first char if '.' is first, whatever happens about
4709 repeats. The value of reqchar doesn't change either. */
4710
4711 case CHAR_DOT:
4712 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4713 zerofirstchar = firstchar;
4714 zerofirstcharflags = firstcharflags;
4715 zeroreqchar = reqchar;
4716 zeroreqcharflags = reqcharflags;
4717 previous = code;
4718 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4719 break;
4720
4721
4722 /* ===================================================================*/
4723 /* Character classes. If the included characters are all < 256, we build a
4724 32-byte bitmap of the permitted characters, except in the special case
4725 where there is only one such character. For negated classes, we build the
4726 map as usual, then invert it at the end. However, we use a different opcode
4727 so that data characters > 255 can be handled correctly.
4728
4729 If the class contains characters outside the 0-255 range, a different
4730 opcode is compiled. It may optionally have a bit map for characters < 256,
4731 but those above are are explicitly listed afterwards. A flag byte tells
4732 whether the bitmap is present, and whether this is a negated class or not.
4733
4734 In JavaScript compatibility mode, an isolated ']' causes an error. In
4735 default (Perl) mode, it is treated as a data character. */
4736
4737 case CHAR_RIGHT_SQUARE_BRACKET:
4738 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4739 {
4740 *errorcodeptr = ERR64;
4741 goto FAILED;
4742 }
4743 goto NORMAL_CHAR;
4744
4745 /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4746 used for "start of word" and "end of word". As these are otherwise illegal
4747 sequences, we don't break anything by recognizing them. They are replaced
4748 by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4749 erroneous and are handled by the normal code below. */
4750
4751 case CHAR_LEFT_SQUARE_BRACKET:
4752 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4753 {
4754 nestptr = ptr + 7;
4755 ptr = sub_start_of_word - 1;
4756 continue;
4757 }
4758
4759 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4760 {
4761 nestptr = ptr + 7;
4762 ptr = sub_end_of_word - 1;
4763 continue;
4764 }
4765
4766 /* Handle a real character class. */
4767
4768 previous = code;
4769
4770 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4771 they are encountered at the top level, so we'll do that too. */
4772
4773 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4774 ptr[1] == CHAR_EQUALS_SIGN) &&
4775 check_posix_syntax(ptr, &tempptr))
4776 {
4777 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4778 goto FAILED;
4779 }
4780
4781 /* If the first character is '^', set the negation flag and skip it. Also,
4782 if the first few characters (either before or after ^) are \Q\E or \E we
4783 skip them too. This makes for compatibility with Perl. */
4784
4785 negate_class = FALSE;
4786 for (;;)
4787 {
4788 c = *(++ptr);
4789 if (c == CHAR_BACKSLASH)
4790 {
4791 if (ptr[1] == CHAR_E)
4792 ptr++;
4793 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4794 ptr += 3;
4795 else
4796 break;
4797 }
4798 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4799 negate_class = TRUE;
4800 else break;
4801 }
4802
4803 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4804 an initial ']' is taken as a data character -- the code below handles
4805 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4806 [^] must match any character, so generate OP_ALLANY. */
4807
4808 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4809 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4810 {
4811 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4812 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4813 zerofirstchar = firstchar;
4814 zerofirstcharflags = firstcharflags;
4815 break;
4816 }
4817
4818 /* If a class contains a negative special such as \S, we need to flip the
4819 negation flag at the end, so that support for characters > 255 works
4820 correctly (they are all included in the class). */
4821
4822 should_flip_negation = FALSE;
4823
4824 /* Extended class (xclass) will be used when characters > 255
4825 might match. */
4826
4827 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4828 xclass = FALSE;
4829 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4830 class_uchardata_base = class_uchardata; /* Save the start */
4831 #endif
4832
4833 /* For optimization purposes, we track some properties of the class:
4834 class_has_8bitchar will be non-zero if the class contains at least one <
4835 256 character; class_one_char will be 1 if the class contains just one
4836 character; xclass_has_prop will be TRUE if unicode property checks
4837 are present in the class. */
4838
4839 class_has_8bitchar = 0;
4840 class_one_char = 0;
4841 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4842 xclass_has_prop = FALSE;
4843 #endif
4844
4845 /* Initialize the 32-char bit map to all zeros. We build the map in a
4846 temporary bit of memory, in case the class contains fewer than two
4847 8-bit characters because in that case the compiled code doesn't use the bit
4848 map. */
4849
4850 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4851
4852 /* Process characters until ] is reached. By writing this as a "do" it
4853 means that an initial ] is taken as a data character. At the start of the
4854 loop, c contains the first byte of the character. */
4855
4856 if (c != CHAR_NULL) do
4857 {
4858 const pcre_uchar *oldptr;
4859
4860 #ifdef SUPPORT_UTF
4861 if (utf && HAS_EXTRALEN(c))
4862 { /* Braces are required because the */
4863 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4864 }
4865 #endif
4866
4867 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4868 /* In the pre-compile phase, accumulate the length of any extra
4869 data and reset the pointer. This is so that very large classes that
4870 contain a zillion > 255 characters no longer overwrite the work space
4871 (which is on the stack). We have to remember that there was XCLASS data,
4872 however. */
4873
4874 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4875 {
4876 xclass = TRUE;
4877 *lengthptr += (int)(class_uchardata - class_uchardata_base);
4878 class_uchardata = class_uchardata_base;
4879 }
4880 #endif
4881
4882 /* Inside \Q...\E everything is literal except \E */
4883
4884 if (inescq)
4885 {
4886 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4887 {
4888 inescq = FALSE; /* Reset literal state */
4889 ptr++; /* Skip the 'E' */
4890 continue; /* Carry on with next */
4891 }
4892 goto CHECK_RANGE; /* Could be range if \E follows */
4893 }
4894
4895 /* Handle POSIX class names. Perl allows a negation extension of the
4896 form [:^name:]. A square bracket that doesn't match the syntax is
4897 treated as a literal. We also recognize the POSIX constructions
4898 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4899 5.6 and 5.8 do. */
4900
4901 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4902 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4903 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4904 {
4905 BOOL local_negate = FALSE;
4906 int posix_class, taboffset, tabopt;
4907 register const pcre_uint8 *cbits = cd->cbits;
4908 pcre_uint8 pbits[32];
4909
4910 if (ptr[1] != CHAR_COLON)
4911 {
4912 *errorcodeptr = ERR31;
4913 goto FAILED;
4914 }
4915
4916 ptr += 2;
4917 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4918 {
4919 local_negate = TRUE;
4920 should_flip_negation = TRUE; /* Note negative special */
4921 ptr++;
4922 }
4923
4924 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4925 if (posix_class < 0)
4926 {
4927 *errorcodeptr = ERR30;
4928 goto FAILED;
4929 }
4930
4931 /* If matching is caseless, upper and lower are converted to
4932 alpha. This relies on the fact that the class table starts with
4933 alpha, lower, upper as the first 3 entries. */
4934
4935 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4936 posix_class = 0;
4937
4938 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4939 different escape sequences that use Unicode properties \p or \P. Others
4940 that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4941 directly. */
4942
4943 #ifdef SUPPORT_UCP
4944 if ((options & PCRE_UCP) != 0)
4945 {
4946 unsigned int ptype = 0;
4947 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4948
4949 /* The posix_substitutes table specifies which POSIX classes can be
4950 converted to \p or \P items. */
4951
4952 if (posix_substitutes[pc] != NULL)
4953 {
4954 nestptr = tempptr + 1;
4955 ptr = posix_substitutes[pc] - 1;
4956 continue;
4957 }
4958
4959 /* There are three other classes that generate special property calls
4960 that are recognized only in an XCLASS. */
4961
4962 else switch(posix_class)
4963 {
4964 case PC_GRAPH:
4965 ptype = PT_PXGRAPH;
4966 /* Fall through */
4967 case PC_PRINT:
4968 if (ptype == 0) ptype = PT_PXPRINT;
4969 /* Fall through */
4970 case PC_PUNCT:
4971 if (ptype == 0) ptype = PT_PXPUNCT;
4972 *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4973 *class_uchardata++ = ptype;
4974 *class_uchardata++ = 0;
4975 xclass_has_prop = TRUE;
4976 ptr = tempptr + 1;
4977 continue;
4978
4979 /* For all other POSIX classes, no special action is taken in UCP
4980 mode. Fall through to the non_UCP case. */
4981
4982 default:
4983 break;
4984 }
4985 }
4986 #endif
4987 /* In the non-UCP case, or when UCP makes no difference, we build the
4988 bit map for the POSIX class in a chunk of local store because we may be
4989 adding and subtracting from it, and we don't want to subtract bits that
4990 may be in the main map already. At the end we or the result into the
4991 bit map that is being built. */
4992
4993 posix_class *= 3;
4994
4995 /* Copy in the first table (always present) */
4996
4997 memcpy(pbits, cbits + posix_class_maps[posix_class],
4998 32 * sizeof(pcre_uint8));
4999
5000 /* If there is a second table, add or remove it as required. */
5001
5002 taboffset = posix_class_maps[posix_class + 1];
5003 tabopt = posix_class_maps[posix_class + 2];
5004
5005 if (taboffset >= 0)
5006 {
5007 if (tabopt >= 0)
5008 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
5009 else
5010 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
5011 }
5012
5013 /* Now see if we need to remove any special characters. An option
5014 value of 1 removes vertical space and 2 removes underscore. */
5015
5016 if (tabopt < 0) tabopt = -tabopt;
5017 if (tabopt == 1) pbits[1] &= ~0x3c;
5018 else if (tabopt == 2) pbits[11] &= 0x7f;
5019
5020 /* Add the POSIX table or its complement into the main table that is
5021 being built and we are done. */
5022
5023 if (local_negate)
5024 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
5025 else
5026 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
5027
5028 ptr = tempptr + 1;
5029 /* Every class contains at least one < 256 character. */
5030 class_has_8bitchar = 1;
5031 /* Every class contains at least two characters. */
5032 class_one_char = 2;
5033 continue; /* End of POSIX syntax handling */
5034 }
5035
5036 /* Backslash may introduce a single character, or it may introduce one
5037 of the specials, which just set a flag. The sequence \b is a special
5038 case. Inside a class (and only there) it is treated as backspace. We
5039 assume that other escapes have more than one character in them, so
5040 speculatively set both class_has_8bitchar and class_one_char bigger
5041 than one. Unrecognized escapes fall through and are either treated
5042 as literal characters (by default), or are faulted if
5043 PCRE_EXTRA is set. */
5044
5045 if (c == CHAR_BACKSLASH)
5046 {
5047 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
5048 TRUE);
5049 if (*errorcodeptr != 0) goto FAILED;
5050 if (escape == 0) c = ec;
5051 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
5052 else if (escape == ESC_N) /* \N is not supported in a class */
5053 {
5054 *errorcodeptr = ERR71;
5055 goto FAILED;
5056 }
5057 else if (escape == ESC_Q) /* Handle start of quoted string */
5058 {
5059 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5060 {
5061 ptr += 2; /* avoid empty string */
5062 }
5063 else inescq = TRUE;
5064 continue;
5065 }
5066 else if (escape == ESC_E) continue; /* Ignore orphan \E */
5067
5068 else
5069 {
5070 register const pcre_uint8 *cbits = cd->cbits;
5071 /* Every class contains at least two < 256 characters. */
5072 class_has_8bitchar++;
5073 /* Every class contains at least two characters. */
5074 class_one_char += 2;
5075
5076 switch (escape)
5077 {
5078 #ifdef SUPPORT_UCP
5079 case ESC_du: /* These are the values given for \d etc */
5080 case ESC_DU: /* when PCRE_UCP is set. We replace the */
5081 case ESC_wu: /* escape sequence with an appropriate \p */
5082 case ESC_WU: /* or \P to test Unicode properties instead */
5083 case ESC_su: /* of the default ASCII testing. */
5084 case ESC_SU:
5085 nestptr = ptr;
5086 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
5087 class_has_8bitchar--; /* Undo! */
5088 continue;
5089 #endif
5090 case ESC_d:
5091 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
5092 continue;
5093
5094 case ESC_D:
5095 should_flip_negation = TRUE;
5096 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
5097 continue;
5098
5099 case ESC_w:
5100 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
5101 continue;
5102
5103 case ESC_W:
5104 should_flip_negation = TRUE;
5105 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5106 continue;
5107
5108 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5109 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5110 previously set by something earlier in the character class.
5111 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5112 we could just adjust the appropriate bit. From PCRE 8.34 we no
5113 longer treat \s and \S specially. */
5114
5115 case ESC_s:
5116 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
5117 continue;
5118
5119 case ESC_S:
5120 should_flip_negation = TRUE;
5121 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
5122 continue;
5123
5124 /* The rest apply in both UCP and non-UCP cases. */
5125
5126 case ESC_h:
5127 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5128 PRIV(hspace_list), NOTACHAR);
5129 continue;
5130
5131 case ESC_H:
5132 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5133 cd, PRIV(hspace_list));
5134 continue;
5135
5136 case ESC_v:
5137 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5138 PRIV(vspace_list), NOTACHAR);
5139 continue;
5140
5141 case ESC_V:
5142 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5143 cd, PRIV(vspace_list));
5144 continue;
5145
5146 #ifdef SUPPORT_UCP
5147 case ESC_p:
5148 case ESC_P:
5149 {
5150 BOOL negated;
5151 unsigned int ptype = 0, pdata = 0;
5152 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5153 goto FAILED;
5154 *class_uchardata++ = ((escape == ESC_p) != negated)?
5155 XCL_PROP : XCL_NOTPROP;
5156 *class_uchardata++ = ptype;
5157 *class_uchardata++ = pdata;
5158 xclass_has_prop = TRUE;
5159 class_has_8bitchar--; /* Undo! */
5160 continue;
5161 }
5162 #endif
5163 /* Unrecognized escapes are faulted if PCRE is running in its
5164 strict mode. By default, for compatibility with Perl, they are
5165 treated as literals. */
5166
5167 default:
5168 if ((options & PCRE_EXTRA) != 0)
5169 {
5170 *errorcodeptr = ERR7;
5171 goto FAILED;
5172 }
5173 class_has_8bitchar--; /* Undo the speculative increase. */
5174 class_one_char -= 2; /* Undo the speculative increase. */
5175 c = *ptr; /* Get the final character and fall through */
5176 break;
5177 }
5178 }
5179
5180 /* Fall through if the escape just defined a single character (c >= 0).
5181 This may be greater than 256. */
5182
5183 escape = 0;
5184
5185 } /* End of backslash handling */
5186
5187 /* A character may be followed by '-' to form a range. However, Perl does
5188 not permit ']' to be the end of the range. A '-' character at the end is
5189 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5190 code for handling \Q and \E is messy. */
5191
5192 CHECK_RANGE:
5193 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5194 {
5195 inescq = FALSE;
5196 ptr += 2;
5197 }
5198 oldptr = ptr;
5199
5200 /* Remember if \r or \n were explicitly used */
5201
5202 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5203
5204 /* Check for range */
5205
5206 if (!inescq && ptr[1] == CHAR_MINUS)
5207 {
5208 pcre_uint32 d;
5209 ptr += 2;
5210 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5211
5212 /* If we hit \Q (not followed by \E) at this point, go into escaped
5213 mode. */
5214
5215 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5216 {
5217 ptr += 2;
5218 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5219 { ptr += 2; continue; }
5220 inescq = TRUE;
5221 break;
5222 }
5223
5224 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5225 back the pointer and jump to handle the character that preceded it. */
5226
5227 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5228 {
5229 ptr = oldptr;
5230 goto CLASS_SINGLE_CHARACTER;
5231 }
5232
5233 /* Otherwise, we have a potential range; pick up the next character */
5234
5235 #ifdef SUPPORT_UTF
5236 if (utf)
5237 { /* Braces are required because the */
5238 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5239 }
5240 else
5241 #endif
5242 d = *ptr; /* Not UTF-8 mode */
5243
5244 /* The second part of a range can be a single-character escape
5245 sequence, but not any of the other escapes. Perl treats a hyphen as a
5246 literal in such circumstances. However, in Perl's warning mode, a
5247 warning is given, so PCRE now faults it as it is almost certainly a
5248 mistake on the user's part. */
5249
5250 if (!inescq)
5251 {
5252 if (d == CHAR_BACKSLASH)
5253 {
5254 int descape;
5255 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5256 if (*errorcodeptr != 0) goto FAILED;
5257
5258 /* 0 means a character was put into d; \b is backspace; any other
5259 special causes an error. */
5260
5261 if (descape != 0)
5262 {
5263 if (descape == ESC_b) d = CHAR_BS; else
5264 {
5265 *errorcodeptr = ERR83;
5266 goto FAILED;
5267 }
5268 }
5269 }
5270
5271 /* A hyphen followed by a POSIX class is treated in the same way. */
5272
5273 else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5274 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5275 ptr[1] == CHAR_EQUALS_SIGN) &&
5276 check_posix_syntax(ptr, &tempptr))
5277 {
5278 *errorcodeptr = ERR83;
5279 goto FAILED;
5280 }
5281 }
5282
5283 /* Check that the two values are in the correct order. Optimize
5284 one-character ranges. */
5285
5286 if (d < c)
5287 {
5288 *errorcodeptr = ERR8;
5289 goto FAILED;
5290 }
5291 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5292
5293 /* We have found a character range, so single character optimizations
5294 cannot be done anymore. Any value greater than 1 indicates that there
5295 is more than one character. */
5296
5297 class_one_char = 2;
5298
5299 /* Remember an explicit \r or \n, and add the range to the class. */
5300
5301 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5302
5303 class_has_8bitchar +=
5304 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5305
5306 continue; /* Go get the next char in the class */
5307 }
5308
5309 /* Handle a single character - we can get here for a normal non-escape
5310 char, or after \ that introduces a single character or for an apparent
5311 range that isn't. Only the value 1 matters for class_one_char, so don't
5312 increase it if it is already 2 or more ... just in case there's a class
5313 with a zillion characters in it. */
5314
5315 CLASS_SINGLE_CHARACTER:
5316 if (class_one_char < 2) class_one_char++;
5317
5318 /* If class_one_char is 1, we have the first single character in the
5319 class, and there have been no prior ranges, or XCLASS items generated by
5320 escapes. If this is the final character in the class, we can optimize by
5321 turning the item into a 1-character OP_CHAR[I] if it's positive, or
5322 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5323 to be set. Otherwise, there can be no first char if this item is first,
5324 whatever repeat count may follow. In the case of reqchar, save the
5325 previous value for reinstating. */
5326
5327 if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5328 {
5329 ptr++;
5330 zeroreqchar = reqchar;
5331 zeroreqcharflags = reqcharflags;
5332
5333 if (negate_class)
5334 {
5335 #ifdef SUPPORT_UCP
5336 int d;
5337 #endif
5338 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5339 zerofirstchar = firstchar;
5340 zerofirstcharflags = firstcharflags;
5341
5342 /* For caseless UTF-8 mode when UCP support is available, check
5343 whether this character has more than one other case. If so, generate
5344 a special OP_NOTPROP item instead of OP_NOTI. */
5345
5346 #ifdef SUPPORT_UCP
5347 if (utf && (options & PCRE_CASELESS) != 0 &&
5348 (d = UCD_CASESET(c)) != 0)
5349 {
5350 *code++ = OP_NOTPROP;
5351 *code++ = PT_CLIST;
5352 *code++ = d;
5353 }
5354 else
5355 #endif
5356 /* Char has only one other case, or UCP not available */
5357
5358 {
5359 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5360 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5361 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5362 code += PRIV(ord2utf)(c, code);
5363 else
5364 #endif
5365 *code++ = c;
5366 }
5367
5368 /* We are finished with this character class */
5369
5370 goto END_CLASS;
5371 }
5372
5373 /* For a single, positive character, get the value into mcbuffer, and
5374 then we can handle this with the normal one-character code. */
5375
5376 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5377 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5378 mclength = PRIV(ord2utf)(c, mcbuffer);
5379 else
5380 #endif
5381 {
5382 mcbuffer[0] = c;
5383 mclength = 1;
5384 }
5385 goto ONE_CHAR;
5386 } /* End of 1-char optimization */
5387
5388 /* There is more than one character in the class, or an XCLASS item
5389 has been generated. Add this character to the class. */
5390
5391 class_has_8bitchar +=
5392 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5393 }
5394
5395 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5396 If we are at the end of an internal nested string, revert to the outer
5397 string. */
5398
5399 while (((c = *(++ptr)) != CHAR_NULL ||
5400 (nestptr != NULL &&
5401 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5402 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5403
5404 /* Check for missing terminating ']' */
5405
5406 if (c == CHAR_NULL)
5407 {
5408 *errorcodeptr = ERR6;
5409 goto FAILED;
5410 }
5411
5412 /* We will need an XCLASS if data has been placed in class_uchardata. In
5413 the second phase this is a sufficient test. However, in the pre-compile
5414 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5415 only if the very last character in the class needs XCLASS will it contain
5416 anything at this point. For this reason, xclass gets set TRUE above when
5417 uchar_classdata is emptied, and that's why this code is the way it is here
5418 instead of just doing a test on class_uchardata below. */
5419
5420 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5421 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5422 #endif
5423
5424 /* If this is the first thing in the branch, there can be no first char
5425 setting, whatever the repeat count. Any reqchar setting must remain
5426 unchanged after any kind of repeat. */
5427
5428 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5429 zerofirstchar = firstchar;
5430 zerofirstcharflags = firstcharflags;
5431 zeroreqchar = reqchar;
5432 zeroreqcharflags = reqcharflags;
5433
5434 /* If there are characters with values > 255, we have to compile an
5435 extended class, with its own opcode, unless there was a negated special
5436 such as \S in the class, and PCRE_UCP is not set, because in that case all
5437 characters > 255 are in the class, so any that were explicitly given as
5438 well can be ignored. If (when there are explicit characters > 255 that must
5439 be listed) there are no characters < 256, we can omit the bitmap in the
5440 actual compiled code. */
5441
5442 #ifdef SUPPORT_UTF
5443 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5444 #elif !defined COMPILE_PCRE8
5445 if (xclass && !should_flip_negation)
5446 #endif
5447 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5448 {
5449 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5450 *code++ = OP_XCLASS;
5451 code += LINK_SIZE;
5452 *code = negate_class? XCL_NOT:0;
5453 if (xclass_has_prop) *code |= XCL_HASPROP;
5454
5455 /* If the map is required, move up the extra data to make room for it;
5456 otherwise just move the code pointer to the end of the extra data. */
5457
5458 if (class_has_8bitchar > 0)
5459 {
5460 *code++ |= XCL_MAP;
5461 memmove(code + (32 / sizeof(pcre_uchar)), code,
5462 IN_UCHARS(class_uchardata - code));
5463 if (negate_class && !xclass_has_prop)
5464 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5465 memcpy(code, classbits, 32);
5466 code = class_uchardata + (32 / sizeof(pcre_uchar));
5467 }
5468 else code = class_uchardata;
5469
5470 /* Now fill in the complete length of the item */
5471
5472 PUT(previous, 1, (int)(code - previous));
5473 break; /* End of class handling */
5474 }
5475 #endif
5476
5477 /* If there are no characters > 255, or they are all to be included or
5478 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5479 whole class was negated and whether there were negative specials such as \S
5480 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5481 negating it if necessary. */
5482
5483 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5484 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5485 {
5486 if (negate_class)
5487 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5488 memcpy(code, classbits, 32);
5489 }
5490 code += 32 / sizeof(pcre_uchar);
5491
5492 END_CLASS:
5493 break;
5494
5495
5496 /* ===================================================================*/
5497 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5498 has been tested above. */
5499
5500 case CHAR_LEFT_CURLY_BRACKET:
5501 if (!is_quantifier) goto NORMAL_CHAR;
5502 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5503 if (*errorcodeptr != 0) goto FAILED;
5504 goto REPEAT;
5505
5506 case CHAR_ASTERISK:
5507 repeat_min = 0;
5508 repeat_max = -1;
5509 goto REPEAT;
5510
5511 case CHAR_PLUS:
5512 repeat_min = 1;
5513 repeat_max = -1;
5514 goto REPEAT;
5515
5516 case CHAR_QUESTION_MARK:
5517 repeat_min = 0;
5518 repeat_max = 1;
5519
5520 REPEAT:
5521 if (previous == NULL)
5522 {
5523 *errorcodeptr = ERR9;
5524 goto FAILED;
5525 }
5526
5527 if (repeat_min == 0)
5528 {
5529 firstchar = zerofirstchar; /* Adjust for zero repeat */
5530 firstcharflags = zerofirstcharflags;
5531 reqchar = zeroreqchar; /* Ditto */
5532 reqcharflags = zeroreqcharflags;
5533 }
5534
5535 /* Remember whether this is a variable length repeat */
5536
5537 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5538
5539 op_type = 0; /* Default single-char op codes */
5540 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5541
5542 /* Save start of previous item, in case we have to move it up in order to
5543 insert something before it. */
5544
5545 tempcode = previous;
5546
5547 /* Before checking for a possessive quantifier, we must skip over
5548 whitespace and comments in extended mode because Perl allows white space at
5549 this point. */
5550
5551 if ((options & PCRE_EXTENDED) != 0)
5552 {
5553 const pcre_uchar *p = ptr + 1;
5554 for (;;)
5555 {
5556 while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5557 if (*p != CHAR_NUMBER_SIGN) break;
5558 p++;
5559 while (*p != CHAR_NULL)
5560 {
5561 if (IS_NEWLINE(p)) /* For non-fixed-length newline cases, */
5562 { /* IS_NEWLINE sets cd->nllen. */
5563 p += cd->nllen;
5564 break;
5565 }
5566 p++;
5567 #ifdef SUPPORT_UTF
5568 if (utf) FORWARDCHAR(p);
5569 #endif
5570 } /* Loop for comment characters */
5571 } /* Loop for multiple comments */
5572 ptr = p - 1; /* Character before the next significant one. */
5573 }
5574
5575 /* If the next character is '+', we have a possessive quantifier. This
5576 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5577 If the next character is '?' this is a minimizing repeat, by default,
5578 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5579 repeat type to the non-default. */
5580
5581 if (ptr[1] == CHAR_PLUS)
5582 {
5583 repeat_type = 0; /* Force greedy */
5584 possessive_quantifier = TRUE;
5585 ptr++;
5586 }
5587 else if (ptr[1] == CHAR_QUESTION_MARK)
5588 {
5589 repeat_type = greedy_non_default;
5590 ptr++;
5591 }
5592 else repeat_type = greedy_default;
5593
5594 /* If previous was a recursion call, wrap it in atomic brackets so that
5595 previous becomes the atomic group. All recursions were so wrapped in the
5596 past, but it no longer happens for non-repeated recursions. In fact, the
5597 repeated ones could be re-implemented independently so as not to need this,
5598 but for the moment we rely on the code for repeating groups. */
5599
5600 if (*previous == OP_RECURSE)
5601 {
5602 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5603 *previous = OP_ONCE;
5604 PUT(previous, 1, 2 + 2*LINK_SIZE);
5605 previous[2 + 2*LINK_SIZE] = OP_KET;
5606 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5607 code += 2 + 2 * LINK_SIZE;
5608 length_prevgroup = 3 + 3*LINK_SIZE;
5609
5610 /* When actually compiling, we need to check whether this was a forward
5611 reference, and if so, adjust the offset. */
5612
5613 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5614 {
5615 int offset = GET(cd->hwm, -LINK_SIZE);
5616 if (offset == previous + 1 - cd->start_code)
5617 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5618 }
5619 }
5620
5621 /* Now handle repetition for the different types of item. */
5622
5623 /* If previous was a character or negated character match, abolish the item
5624 and generate a repeat item instead. If a char item has a minimum of more
5625 than one, ensure that it is set in reqchar - it might not be if a sequence
5626 such as x{3} is the first thing in a branch because the x will have gone
5627 into firstchar instead. */
5628
5629 if (*previous == OP_CHAR || *previous == OP_CHARI
5630 || *previous == OP_NOT || *previous == OP_NOTI)
5631 {
5632 switch (*previous)
5633 {
5634 default: /* Make compiler happy. */
5635 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5636 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5637 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5638 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5639 }
5640
5641 /* Deal with UTF characters that take up more than one character. It's
5642 easier to write this out separately than try to macrify it. Use c to
5643 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5644 it's a length rather than a small character. */
5645
5646 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5647 if (utf && NOT_FIRSTCHAR(code[-1]))
5648 {
5649 pcre_uchar *lastchar = code - 1;
5650 BACKCHAR(lastchar);
5651 c = (int)(code - lastchar); /* Length of UTF-8 character */
5652 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5653 c |= UTF_LENGTH; /* Flag c as a length */
5654 }
5655 else
5656 #endif /* SUPPORT_UTF */
5657
5658 /* Handle the case of a single charater - either with no UTF support, or
5659 with UTF disabled, or for a single character UTF character. */
5660 {
5661 c = code[-1];
5662 if (*previous <= OP_CHARI && repeat_min > 1)
5663 {
5664 reqchar = c;
5665 reqcharflags = req_caseopt | cd->req_varyopt;
5666 }
5667 }
5668
5669 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5670 }
5671
5672 /* If previous was a character type match (\d or similar), abolish it and
5673 create a suitable repeat item. The code is shared with single-character
5674 repeats by setting op_type to add a suitable offset into repeat_type. Note
5675 the the Unicode property types will be present only when SUPPORT_UCP is
5676 defined, but we don't wrap the little bits of code here because it just
5677 makes it horribly messy. */
5678
5679 else if (*previous < OP_EODN)
5680 {
5681 pcre_uchar *oldcode;
5682 int prop_type, prop_value;
5683 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5684 c = *previous;
5685
5686 OUTPUT_SINGLE_REPEAT:
5687 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5688 {
5689 prop_type = previous[1];
5690 prop_value = previous[2];
5691 }
5692 else prop_type = prop_value = -1;
5693
5694 oldcode = code;
5695 code = previous; /* Usually overwrite previous item */
5696
5697 /* If the maximum is zero then the minimum must also be zero; Perl allows
5698 this case, so we do too - by simply omitting the item altogether. */
5699
5700 if (repeat_max == 0) goto END_REPEAT;
5701
5702 /* Combine the op_type with the repeat_type */
5703
5704 repeat_type += op_type;
5705
5706 /* A minimum of zero is handled either as the special case * or ?, or as
5707 an UPTO, with the maximum given. */
5708
5709 if (repeat_min == 0)
5710 {
5711 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5712 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5713 else
5714 {
5715 *code++ = OP_UPTO + repeat_type;
5716 PUT2INC(code, 0, repeat_max);
5717 }
5718 }
5719
5720 /* A repeat minimum of 1 is optimized into some special cases. If the
5721 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5722 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5723 one less than the maximum. */
5724
5725 else if (repeat_min == 1)
5726 {
5727 if (repeat_max == -1)
5728 *code++ = OP_PLUS + repeat_type;
5729 else
5730 {
5731 code = oldcode; /* leave previous item in place */
5732 if (repeat_max == 1) goto END_REPEAT;
5733 *code++ = OP_UPTO + repeat_type;
5734 PUT2INC(code, 0, repeat_max - 1);
5735 }
5736 }
5737
5738 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5739 handled as an EXACT followed by an UPTO. */
5740
5741 else
5742 {
5743 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5744 PUT2INC(code, 0, repeat_min);
5745
5746 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5747 we have to insert the character for the previous code. For a repeated
5748 Unicode property match, there are two extra bytes that define the
5749 required property. In UTF-8 mode, long characters have their length in
5750 c, with the UTF_LENGTH bit as a flag. */
5751
5752 if (repeat_max < 0)
5753 {
5754 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5755 if (utf && (c & UTF_LENGTH) != 0)
5756 {
5757 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5758 code += c & 7;
5759 }
5760 else
5761 #endif
5762 {
5763 *code++ = c;
5764 if (prop_type >= 0)
5765 {
5766 *code++ = prop_type;
5767 *code++ = prop_value;
5768 }
5769 }
5770 *code++ = OP_STAR + repeat_type;
5771 }
5772
5773 /* Else insert an UPTO if the max is greater than the min, again
5774 preceded by the character, for the previously inserted code. If the
5775 UPTO is just for 1 instance, we can use QUERY instead. */
5776
5777 else if (repeat_max != repeat_min)
5778 {
5779 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5780 if (utf && (c & UTF_LENGTH) != 0)
5781 {
5782 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5783 code += c & 7;
5784 }
5785 else
5786 #endif
5787 *code++ = c;
5788 if (prop_type >= 0)
5789 {
5790 *code++ = prop_type;
5791 *code++ = prop_value;
5792 }
5793 repeat_max -= repeat_min;
5794
5795 if (repeat_max == 1)
5796 {
5797 *code++ = OP_QUERY + repeat_type;
5798 }
5799 else
5800 {
5801 *code++ = OP_UPTO + repeat_type;
5802 PUT2INC(code, 0, repeat_max);
5803 }
5804 }
5805 }
5806
5807 /* The character or character type itself comes last in all cases. */
5808
5809 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5810 if (utf && (c & UTF_LENGTH) != 0)
5811 {
5812 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5813 code += c & 7;
5814 }
5815 else
5816 #endif
5817 *code++ = c;
5818
5819 /* For a repeated Unicode property match, there are two extra bytes that
5820 define the required property. */
5821
5822 #ifdef SUPPORT_UCP
5823 if (prop_type >= 0)
5824 {
5825 *code++ = prop_type;
5826 *code++ = prop_value;
5827 }
5828 #endif
5829 }
5830
5831 /* If previous was a character class or a back reference, we put the repeat
5832 stuff after it, but just skip the item if the repeat was {0,0}. */
5833
5834 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5835 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5836 *previous == OP_XCLASS ||
5837 #endif
5838 *previous == OP_REF || *previous == OP_REFI ||
5839 *previous == OP_DNREF || *previous == OP_DNREFI)
5840 {
5841 if (repeat_max == 0)
5842 {
5843 code = previous;
5844 goto END_REPEAT;
5845 }
5846
5847 if (repeat_min == 0 && repeat_max == -1)
5848 *code++ = OP_CRSTAR + repeat_type;
5849 else if (repeat_min == 1 && repeat_max == -1)
5850 *code++ = OP_CRPLUS + repeat_type;
5851 else if (repeat_min == 0 && repeat_max == 1)
5852 *code++ = OP_CRQUERY + repeat_type;
5853 else
5854 {
5855 *code++ = OP_CRRANGE + repeat_type;
5856 PUT2INC(code, 0, repeat_min);
5857 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5858 PUT2INC(code, 0, repeat_max);
5859 }
5860 }
5861
5862 /* If previous was a bracket group, we may have to replicate it in certain
5863 cases. Note that at this point we can encounter only the "basic" bracket
5864 opcodes such as BRA and CBRA, as this is the place where they get converted
5865 into the more special varieties such as BRAPOS and SBRA. A test for >=
5866 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5867 ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5868 Originally, PCRE did not allow repetition of assertions, but now it does,
5869 for Perl compatibility. */
5870
5871 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5872 {
5873 register int i;
5874 int len = (int)(code - previous);
5875 pcre_uchar *bralink = NULL;
5876 pcre_uchar *brazeroptr = NULL;
5877
5878 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5879 we just ignore the repeat. */
5880
5881 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5882 goto END_REPEAT;
5883
5884 /* There is no sense in actually repeating assertions. The only potential
5885 use of repetition is in cases when the assertion is optional. Therefore,
5886 if the minimum is greater than zero, just ignore the repeat. If the
5887 maximum is not zero or one, set it to 1. */
5888
5889 if (*previous < OP_ONCE) /* Assertion */
5890 {
5891 if (repeat_min > 0) goto END_REPEAT;
5892 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5893 }
5894
5895 /* The case of a zero minimum is special because of the need to stick
5896 OP_BRAZERO in front of it, and because the group appears once in the
5897 data, whereas in other cases it appears the minimum number of times. For
5898 this reason, it is simplest to treat this case separately, as otherwise
5899 the code gets far too messy. There are several special subcases when the
5900 minimum is zero. */
5901
5902 if (repeat_min == 0)
5903 {
5904 /* If the maximum is also zero, we used to just omit the group from the
5905 output altogether, like this:
5906
5907 ** if (repeat_max == 0)
5908 ** {
5909 ** code = previous;
5910 ** goto END_REPEAT;
5911 ** }
5912
5913 However, that fails when a group or a subgroup within it is referenced
5914 as a subroutine from elsewhere in the pattern, so now we stick in
5915 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5916 don't have a list of which groups are referenced, we cannot do this
5917 selectively.
5918
5919 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5920 and do no more at this point. However, we do need to adjust any
5921 OP_RECURSE calls inside the group that refer to the group itself or any
5922 internal or forward referenced group, because the offset is from the
5923 start of the whole regex. Temporarily terminate the pattern while doing
5924 this. */
5925
5926 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5927 {
5928 *code = OP_END;
5929 adjust_recurse(previous, 1, utf, cd, save_hwm);
5930 memmove(previous + 1, previous, IN_UCHARS(len));
5931 code++;
5932 if (repeat_max == 0)
5933 {
5934 *previous++ = OP_SKIPZERO;
5935 goto END_REPEAT;
5936 }
5937 brazeroptr = previous; /* Save for possessive optimizing */
5938 *previous++ = OP_BRAZERO + repeat_type;
5939 }
5940
5941 /* If the maximum is greater than 1 and limited, we have to replicate
5942 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
5943 The first one has to be handled carefully because it's the original
5944 copy, which has to be moved up. The remainder can be handled by code
5945 that is common with the non-zero minimum case below. We have to
5946 adjust the value or repeat_max, since one less copy is required. Once
5947 again, we may have to adjust any OP_RECURSE calls inside the group. */
5948
5949 else
5950 {
5951 int offset;
5952 *code = OP_END;
5953 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5954 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5955 code += 2 + LINK_SIZE;
5956 *previous++ = OP_BRAZERO + repeat_type;
5957 *previous++ = OP_BRA;
5958
5959 /* We chain together the bracket offset fields that have to be
5960 filled in later when the ends of the brackets are reached. */
5961
5962 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
5963 bralink = previous;
5964 PUTINC(previous, 0, offset);
5965 }
5966
5967 repeat_max--;
5968 }
5969
5970 /* If the minimum is greater than zero, replicate the group as many
5971 times as necessary, and adjust the maximum to the number of subsequent
5972 copies that we need. If we set a first char from the group, and didn't
5973 set a required char, copy the latter from the former. If there are any
5974 forward reference subroutine calls in the group, there will be entries on
5975 the workspace list; replicate these with an appropriate increment. */
5976
5977 else
5978 {
5979 if (repeat_min > 1)
5980 {
5981 /* In the pre-compile phase, we don't actually do the replication. We
5982 just adjust the length as if we had. Do some paranoid checks for
5983 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
5984 integer type when available, otherwise double. */
5985
5986 if (lengthptr != NULL)
5987 {
5988 int delta = (repeat_min - 1)*length_prevgroup;
5989 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
5990 (INT64_OR_DOUBLE)length_prevgroup >
5991 (INT64_OR_DOUBLE)INT_MAX ||
5992 OFLOW_MAX - *lengthptr < delta)
5993 {
5994 *errorcodeptr = ERR20;
5995 goto FAILED;
5996 }
5997 *lengthptr += delta;
5998 }
5999
6000 /* This is compiling for real. If there is a set first byte for
6001 the group, and we have not yet set a "required byte", set it. Make
6002 sure there is enough workspace for copying forward references before
6003 doing the copy. */
6004
6005 else
6006 {
6007 if (groupsetfirstchar && reqcharflags < 0)
6008 {
6009 reqchar = firstchar;
6010 reqcharflags = firstcharflags;
6011 }
6012
6013 for (i = 1; i < repeat_min; i++)
6014 {
6015 pcre_uchar *hc;
6016 pcre_uchar *this_hwm = cd->hwm;
6017 memcpy(code, previous, IN_UCHARS(len));
6018
6019 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6020 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
6021 {
6022 size_t save_offset = save_hwm - cd->start_workspace;
6023 size_t this_offset = this_hwm - cd->start_workspace;
6024 *errorcodeptr = expand_workspace(cd);
6025 if (*errorcodeptr != 0) goto FAILED;
6026 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
6027 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
6028 }
6029
6030 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
6031 {
6032 PUT(cd->hwm, 0, GET(hc, 0) + len);
6033 cd->hwm += LINK_SIZE;
6034 }
6035 save_hwm = this_hwm;
6036 code += len;
6037 }
6038 }
6039 }
6040
6041 if (repeat_max > 0) repeat_max -= repeat_min;
6042 }
6043
6044 /* This code is common to both the zero and non-zero minimum cases. If
6045 the maximum is limited, it replicates the group in a nested fashion,
6046 remembering the bracket starts on a stack. In the case of a zero minimum,
6047 the first one was set up above. In all cases the repeat_max now specifies
6048 the number of additional copies needed. Again, we must remember to
6049 replicate entries on the forward reference list. */
6050
6051 if (repeat_max >= 0)
6052 {
6053 /* In the pre-compile phase, we don't actually do the replication. We
6054 just adjust the length as if we had. For each repetition we must add 1
6055 to the length for BRAZERO and for all but the last repetition we must
6056 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
6057 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
6058 a 64-bit integer type when available, otherwise double. */
6059
6060 if (lengthptr != NULL && repeat_max > 0)
6061 {
6062 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
6063 2 - 2*LINK_SIZE; /* Last one doesn't nest */
6064 if ((INT64_OR_DOUBLE)repeat_max *
6065 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
6066 > (INT64_OR_DOUBLE)INT_MAX ||
6067 OFLOW_MAX - *lengthptr < delta)
6068 {
6069 *errorcodeptr = ERR20;
6070 goto FAILED;
6071 }
6072 *lengthptr += delta;
6073 }
6074
6075 /* This is compiling for real */
6076
6077 else for (i = repeat_max - 1; i >= 0; i--)
6078 {
6079 pcre_uchar *hc;
6080 pcre_uchar *this_hwm = cd->hwm;
6081
6082 *code++ = OP_BRAZERO + repeat_type;
6083
6084 /* All but the final copy start a new nesting, maintaining the
6085 chain of brackets outstanding. */
6086
6087 if (i != 0)
6088 {
6089 int offset;
6090 *code++ = OP_BRA;
6091 offset = (bralink == NULL)? 0 : (int)(code - bralink);
6092 bralink = code;
6093 PUTINC(code, 0, offset);
6094 }
6095
6096 memcpy(code, previous, IN_UCHARS(len));
6097
6098 /* Ensure there is enough workspace for forward references before
6099 copying them. */
6100
6101 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6102 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
6103 {
6104 size_t save_offset = save_hwm - cd->start_workspace;
6105 size_t this_offset = this_hwm - cd->start_workspace;
6106 *errorcodeptr = expand_workspace(cd);
6107 if (*errorcodeptr != 0) goto FAILED;
6108 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
6109 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
6110 }
6111
6112 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
6113 {
6114 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
6115 cd->hwm += LINK_SIZE;
6116 }
6117 save_hwm = this_hwm;
6118 code += len;
6119 }
6120
6121 /* Now chain through the pending brackets, and fill in their length
6122 fields (which are holding the chain links pro tem). */
6123
6124 while (bralink != NULL)
6125 {
6126 int oldlinkoffset;
6127 int offset = (int)(code - bralink + 1);
6128 pcre_uchar *bra = code - offset;
6129 oldlinkoffset = GET(bra, 1);
6130 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
6131 *code++ = OP_KET;
6132 PUTINC(code, 0, offset);
6133 PUT(bra, 1, offset);
6134 }
6135 }
6136
6137 /* If the maximum is unlimited, set a repeater in the final copy. For
6138 ONCE brackets, that's all we need to do. However, possessively repeated
6139 ONCE brackets can be converted into non-capturing brackets, as the
6140 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
6141 deal with possessive ONCEs specially.
6142
6143 Otherwise, when we are doing the actual compile phase, check to see
6144 whether this group is one that could match an empty string. If so,
6145 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
6146 that runtime checking can be done. [This check is also applied to ONCE
6147 groups at runtime, but in a different way.]
6148
6149 Then, if the quantifier was possessive and the bracket is not a
6150 conditional, we convert the BRA code to the POS form, and the KET code to
6151 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
6152 subpattern at both the start and at the end.) The use of special opcodes
6153 makes it possible to reduce greatly the stack usage in pcre_exec(). If
6154 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
6155
6156 Then, if the minimum number of matches is 1 or 0, cancel the possessive
6157 flag so that the default action below, of wrapping everything inside
6158 atomic brackets, does not happen. When the minimum is greater than 1,
6159 there will be earlier copies of the group, and so we still have to wrap
6160 the whole thing. */
6161
6162 else
6163 {
6164 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
6165 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
6166
6167 /* Convert possessive ONCE brackets to non-capturing */
6168
6169 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
6170 possessive_quantifier) *bracode = OP_BRA;
6171
6172 /* For non-possessive ONCE brackets, all we need to do is to
6173 set the KET. */
6174
6175 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
6176 *ketcode = OP_KETRMAX + repeat_type;
6177
6178 /* Handle non-ONCE brackets and possessive ONCEs (which have been
6179 converted to non-capturing above). */
6180
6181 else
6182 {
6183 /* In the compile phase, check for empty string matching. */
6184
6185 if (lengthptr == NULL)
6186 {
6187 pcre_uchar *scode = bracode;
6188 do
6189 {
6190 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6191 {
6192 *bracode += OP_SBRA - OP_BRA;
6193 break;
6194 }
6195 scode += GET(scode, 1);
6196 }
6197 while (*scode == OP_ALT);
6198 }
6199
6200 /* Handle possessive quantifiers. */
6201
6202 if (possessive_quantifier)
6203 {
6204 /* For COND brackets, we wrap the whole thing in a possessively
6205 repeated non-capturing bracket, because we have not invented POS
6206 versions of the COND opcodes. Because we are moving code along, we
6207 must ensure that any pending recursive references are updated. */
6208
6209 if (*bracode == OP_COND || *bracode == OP_SCOND)
6210 {
6211 int nlen = (int)(code - bracode);
6212 *code = OP_END;
6213 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
6214 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6215 code += 1 + LINK_SIZE;
6216 nlen += 1 + LINK_SIZE;
6217 *bracode = OP_BRAPOS;
6218 *code++ = OP_KETRPOS;
6219 PUTINC(code, 0, nlen);
6220 PUT(bracode, 1, nlen);
6221 }
6222
6223 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6224
6225 else
6226 {
6227 *bracode += 1; /* Switch to xxxPOS opcodes */
6228 *ketcode = OP_KETRPOS;
6229 }
6230
6231 /* If the minimum is zero, mark it as possessive, then unset the
6232 possessive flag when the minimum is 0 or 1. */
6233
6234 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6235 if (repeat_min < 2) possessive_quantifier = FALSE;
6236 }
6237
6238 /* Non-possessive quantifier */
6239
6240 else *ketcode = OP_KETRMAX + repeat_type;
6241 }
6242 }
6243 }
6244
6245 /* If previous is OP_FAIL, it was generated by an empty class [] in
6246 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6247 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6248 error above. We can just ignore the repeat in JS case. */
6249
6250 else if (*previous == OP_FAIL) goto END_REPEAT;
6251
6252 /* Else there's some kind of shambles */
6253
6254 else
6255 {
6256 *errorcodeptr = ERR11;
6257 goto FAILED;
6258 }
6259
6260 /* If the character following a repeat is '+', possessive_quantifier is
6261 TRUE. For some opcodes, there are special alternative opcodes for this
6262 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6263 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6264 Sun's Java package, but the special opcodes can optimize it.
6265
6266 Some (but not all) possessively repeated subpatterns have already been
6267 completely handled in the code just above. For them, possessive_quantifier
6268 is always FALSE at this stage. Note that the repeated item starts at
6269 tempcode, not at previous, which might be the first part of a string whose
6270 (former) last char we repeated. */
6271
6272 if (possessive_quantifier)
6273 {
6274 int len;
6275
6276 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6277 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6278 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6279 remains is greater than zero, there's a further opcode that can be
6280 handled. If not, do nothing, leaving the EXACT alone. */
6281
6282 switch(*tempcode)
6283 {
6284 case OP_TYPEEXACT:
6285 tempcode += PRIV(OP_lengths)[*tempcode] +
6286 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6287 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6288 break;
6289
6290 /* CHAR opcodes are used for exacts whose count is 1. */
6291
6292 case OP_CHAR:
6293 case OP_CHARI:
6294 case OP_NOT:
6295 case OP_NOTI:
6296 case OP_EXACT:
6297 case OP_EXACTI:
6298 case OP_NOTEXACT:
6299 case OP_NOTEXACTI:
6300 tempcode += PRIV(OP_lengths)[*tempcode];
6301 #ifdef SUPPORT_UTF
6302 if (utf && HAS_EXTRALEN(tempcode[-1]))
6303 tempcode += GET_EXTRALEN(tempcode[-1]);
6304 #endif
6305 break;
6306
6307 /* For the class opcodes, the repeat operator appears at the end;
6308 adjust tempcode to point to it. */
6309
6310 case OP_CLASS:
6311 case OP_NCLASS:
6312 tempcode += 1 + 32/sizeof(pcre_uchar);
6313 break;
6314
6315 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6316 case OP_XCLASS:
6317 tempcode += GET(tempcode, 1);
6318 break;
6319 #endif
6320 }
6321
6322 /* If tempcode is equal to code (which points to the end of the repeated
6323 item), it means we have skipped an EXACT item but there is no following
6324 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6325 all other cases, tempcode will be pointing to the repeat opcode, and will
6326 be less than code, so the value of len will be greater than 0. */
6327
6328 len = (int)(code - tempcode);
6329 if (len > 0)
6330 {
6331 unsigned int repcode = *tempcode;
6332
6333 /* There is a table for possessifying opcodes, all of which are less
6334 than OP_CALLOUT. A zero entry means there is no possessified version.
6335 */
6336
6337 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6338 *tempcode = opcode_possessify[repcode];
6339
6340 /* For opcode without a special possessified version, wrap the item in
6341 ONCE brackets. Because we are moving code along, we must ensure that any
6342 pending recursive references are updated. */
6343
6344 else
6345 {
6346 *code = OP_END;
6347 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6348 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6349 code += 1 + LINK_SIZE;
6350 len += 1 + LINK_SIZE;
6351 tempcode[0] = OP_ONCE;
6352 *code++ = OP_KET;
6353 PUTINC(code, 0, len);
6354 PUT(tempcode, 1, len);
6355 }
6356 }
6357
6358 #ifdef NEVER
6359 if (len > 0) switch (*tempcode)
6360 {
6361 case OP_STAR: *tempcode = OP_POSSTAR; break;
6362 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6363 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6364 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6365
6366 case OP_STARI: *tempcode = OP_POSSTARI; break;
6367 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6368 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6369 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6370
6371 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6372 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6373 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6374 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6375
6376 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6377 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6378 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6379 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6380
6381 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6382 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6383 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6384 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6385
6386 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6387 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6388 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6389 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6390
6391 /* Because we are moving code along, we must ensure that any
6392 pending recursive references are updated. */
6393
6394 default:
6395 *code = OP_END;
6396 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6397 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6398 code += 1 + LINK_SIZE;
6399 len += 1 + LINK_SIZE;
6400 tempcode[0] = OP_ONCE;
6401 *code++ = OP_KET;
6402 PUTINC(code, 0, len);
6403 PUT(tempcode, 1, len);
6404 break;
6405 }
6406 #endif
6407 }
6408
6409 /* In all case we no longer have a previous item. We also set the
6410 "follows varying string" flag for subsequently encountered reqchars if
6411 it isn't already set and we have just passed a varying length item. */
6412
6413 END_REPEAT:
6414 previous = NULL;
6415 cd->req_varyopt |= reqvary;
6416 break;
6417
6418
6419 /* ===================================================================*/
6420 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6421 lookbehind or option setting or condition or all the other extended
6422 parenthesis forms. */
6423
6424 case CHAR_LEFT_PARENTHESIS:
6425 newoptions = options;
6426 skipbytes = 0;
6427 bravalue = OP_CBRA;
6428 save_hwm = cd->hwm;
6429 reset_bracount = FALSE;
6430
6431 /* First deal with various "verbs" that can be introduced by '*'. */
6432
6433 ptr++;
6434 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6435 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6436 {
6437 int i, namelen;
6438 int arglen = 0;
6439 const char *vn = verbnames;
6440 const pcre_uchar *name = ptr + 1;
6441 const pcre_uchar *arg = NULL;
6442 previous = NULL;
6443 ptr++;
6444 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6445 namelen = (int)(ptr - name);
6446
6447 /* It appears that Perl allows any characters whatsoever, other than
6448 a closing parenthesis, to appear in arguments, so we no longer insist on
6449 letters, digits, and underscores. */
6450
6451 if (*ptr == CHAR_COLON)
6452 {
6453 arg = ++ptr;
6454 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6455 arglen = (int)(ptr - arg);
6456 if ((unsigned int)arglen > MAX_MARK)
6457 {
6458 *errorcodeptr = ERR75;
6459 goto FAILED;
6460 }
6461 }
6462
6463 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6464 {
6465 *errorcodeptr = ERR60;
6466 goto FAILED;
6467 }
6468
6469 /* Scan the table of verb names */
6470
6471 for (i = 0; i < verbcount; i++)
6472 {
6473 if (namelen == verbs[i].len &&
6474 STRNCMP_UC_C8(name, vn, namelen) == 0)
6475 {
6476 int setverb;
6477
6478 /* Check for open captures before ACCEPT and convert it to
6479 ASSERT_ACCEPT if in an assertion. */
6480
6481 if (verbs[i].op == OP_ACCEPT)
6482 {
6483 open_capitem *oc;
6484 if (arglen != 0)
6485 {
6486 *errorcodeptr = ERR59;
6487 goto FAILED;
6488 }
6489 cd->had_accept = TRUE;
6490 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6491 {
6492 *code++ = OP_CLOSE;
6493 PUT2INC(code, 0, oc->number);
6494 }
6495 setverb = *code++ =
6496 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6497
6498 /* Do not set firstchar after *ACCEPT */
6499 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6500 }
6501
6502 /* Handle other cases with/without an argument */
6503
6504 else if (arglen == 0)
6505 {
6506 if (verbs[i].op < 0) /* Argument is mandatory */
6507 {
6508 *errorcodeptr = ERR66;
6509 goto FAILED;
6510 }
6511 setverb = *code++ = verbs[i].op;
6512 }
6513
6514 else
6515 {
6516 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6517 {
6518 *errorcodeptr = ERR59;
6519 goto FAILED;
6520 }
6521 setverb = *code++ = verbs[i].op_arg;
6522 *code++ = arglen;
6523 memcpy(code, arg, IN_UCHARS(arglen));
6524 code += arglen;
6525 *code++ = 0;
6526 }
6527
6528 switch (setverb)
6529 {
6530 case OP_THEN:
6531 case OP_THEN_ARG:
6532 cd->external_flags |= PCRE_HASTHEN;
6533 break;
6534
6535 case OP_PRUNE:
6536 case OP_PRUNE_ARG:
6537 case OP_SKIP:
6538 case OP_SKIP_ARG:
6539 cd->had_pruneorskip = TRUE;
6540 break;
6541 }
6542
6543 break; /* Found verb, exit loop */
6544 }
6545
6546 vn += verbs[i].len + 1;
6547 }
6548
6549 if (i < verbcount) continue; /* Successfully handled a verb */
6550 *errorcodeptr = ERR60; /* Verb not recognized */
6551 goto FAILED;
6552 }
6553
6554 /* Deal with the extended parentheses; all are introduced by '?', and the
6555 appearance of any of them means that this is not a capturing group. */
6556
6557 else if (*ptr == CHAR_QUESTION_MARK)
6558 {
6559 int i, set, unset, namelen;
6560 int *optset;
6561 const pcre_uchar *name;
6562 pcre_uchar *slot;
6563
6564 switch (*(++ptr))
6565 {
6566 case CHAR_NUMBER_SIGN: /* Comment; skip to ket */
6567 ptr++;
6568 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6569 if (*ptr == CHAR_NULL)
6570 {
6571 *errorcodeptr = ERR18;
6572 goto FAILED;
6573 }
6574 continue;
6575
6576
6577 /* ------------------------------------------------------------ */
6578 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6579 reset_bracount = TRUE;
6580 /* Fall through */
6581
6582 /* ------------------------------------------------------------ */
6583 case CHAR_COLON: /* Non-capturing bracket */
6584 bravalue = OP_BRA;
6585 ptr++;
6586 break;
6587
6588
6589 /* ------------------------------------------------------------ */
6590 case CHAR_LEFT_PARENTHESIS:
6591 bravalue = OP_COND; /* Conditional group */
6592 tempptr = ptr;
6593
6594 /* A condition can be an assertion, a number (referring to a numbered
6595 group's having been set), a name (referring to a named group), or 'R',
6596 referring to recursion. R<digits> and R&name are also permitted for
6597 recursion tests.
6598
6599 There are ways of testing a named group: (?(name)) is used by Python;
6600 Perl 5.10 onwards uses (?(<name>) or (?('name')).
6601
6602 There is one unfortunate ambiguity, caused by history. 'R' can be the
6603 recursive thing or the name 'R' (and similarly for 'R' followed by
6604 digits). We look for a name first; if not found, we try the other case.
6605
6606 For compatibility with auto-callouts, we allow a callout to be
6607 specified before a condition that is an assertion. First, check for the
6608 syntax of a callout; if found, adjust the temporary pointer that is
6609 used to check for an assertion condition. That's all that is needed! */
6610
6611 if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6612 {
6613 for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6614 if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6615 tempptr += i + 1;
6616 }
6617
6618 /* For conditions that are assertions, check the syntax, and then exit
6619 the switch. This will take control down to where bracketed groups,
6620 including assertions, are processed. */
6621
6622 if (tempptr[1] == CHAR_QUESTION_MARK &&
6623 (tempptr[2] == CHAR_EQUALS_SIGN ||
6624 tempptr[2] == CHAR_EXCLAMATION_MARK ||
6625 tempptr[2] == CHAR_LESS_THAN_SIGN))
6626 break;
6627
6628 /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6629 need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6630
6631 code[1+LINK_SIZE] = OP_CREF;
6632 skipbytes = 1+IMM2_SIZE;
6633 refsign = -1; /* => not a number */
6634 namelen = -1; /* => not a name; must set to avoid warning */
6635 name = NULL; /* Always set to avoid warning */
6636 recno = 0; /* Always set to avoid warning */
6637
6638 /* Check for a test for recursion in a named group. */
6639
6640 ptr++;
6641 if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6642 {
6643 terminator = -1;
6644 ptr += 2;
6645 code[1+LINK_SIZE] = OP_RREF; /* Change the type of test */
6646 }
6647
6648 /* Check for a test for a named group's having been set, using the Perl
6649 syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6650 syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6651
6652 else if (*ptr == CHAR_LESS_THAN_SIGN)
6653 {
6654 terminator = CHAR_GREATER_THAN_SIGN;
6655 ptr++;
6656 }
6657 else if (*ptr == CHAR_APOSTROPHE)
6658 {
6659 terminator = CHAR_APOSTROPHE;
6660<