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

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1560 - (show annotations)
Tue May 19 16:02:06 2015 UTC (8 days, 14 hours ago) by ph10
File MIME type: text/plain
File size: 316382 byte(s)
Error occurred while calculating annotation data.
Fix buffer overflow for lookbehind within mutually recursive subroutines.
1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2014 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40
41 /* This module contains the external function pcre_compile(), along with
42 supporting internal functions that are not used by other modules. */
43
44
45 #ifdef HAVE_CONFIG_H
46 #include "config.h"
47 #endif
48
49 #define NLBLOCK cd /* Block containing newline information */
50 #define PSSTART start_pattern /* Field containing pattern start */
51 #define PSEND end_pattern /* Field containing pattern end */
52
53 #include "pcre_internal.h"
54
55
56 /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57 is also used by pcretest. PCRE_DEBUG is not defined when building a production
58 library. We do not need to select pcre16_printint.c specially, because the
59 COMPILE_PCREx macro will already be appropriately set. */
60
61 #ifdef PCRE_DEBUG
62 /* pcre_printint.c should not include any headers */
63 #define PCRE_INCLUDED
64 #include "pcre_printint.c"
65 #undef PCRE_INCLUDED
66 #endif
67
68
69 /* Macro for setting individual bits in class bitmaps. */
70
71 #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72
73 /* Maximum length value to check against when making sure that the integer that
74 holds the compiled pattern length does not overflow. We make it a bit less than
75 INT_MAX to allow for adding in group terminating bytes, so that we don't have
76 to check them every time. */
77
78 #define OFLOW_MAX (INT_MAX - 20)
79
80 /* Definitions to allow mutual recursion */
81
82 static int
83 add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84 const pcre_uint32 *, unsigned int);
85
86 static BOOL
87 compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88 pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89 compile_data *, int *);
90
91
92
93 /*************************************************
94 * Code parameters and static tables *
95 *************************************************/
96
97 /* This value specifies the size of stack workspace that is used during the
98 first pre-compile phase that determines how much memory is required. The regex
99 is partly compiled into this space, but the compiled parts are discarded as
100 soon as they can be, so that hopefully there will never be an overrun. The code
101 does, however, check for an overrun. The largest amount I've seen used is 218,
102 so this number is very generous.
103
104 The same workspace is used during the second, actual compile phase for
105 remembering forward references to groups so that they can be filled in at the
106 end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107 is 4 there is plenty of room for most patterns. However, the memory can get
108 filled up by repetitions of forward references, for example patterns like
109 /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110 that the workspace is expanded using malloc() in this situation. The value
111 below is therefore a minimum, and we put a maximum on it for safety. The
112 minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113 kicks in at the same number of forward references in all cases. */
114
115 #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116 #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117
118 /* This value determines the size of the initial vector that is used for
119 remembering named groups during the pre-compile. It is allocated on the stack,
120 but if it is too small, it is expanded using malloc(), in a similar way to the
121 workspace. The value is the number of slots in the list. */
122
123 #define NAMED_GROUP_LIST_SIZE 20
124
125 /* The overrun tests check for a slightly smaller size so that they detect the
126 overrun before it actually does run off the end of the data block. */
127
128 #define WORK_SIZE_SAFETY_MARGIN (100)
129
130 /* Private flags added to firstchar and reqchar. */
131
132 #define REQ_CASELESS (1 << 0) /* Indicates caselessness */
133 #define REQ_VARY (1 << 1) /* Reqchar followed non-literal item */
134 /* Negative values for the firstchar and reqchar flags */
135 #define REQ_UNSET (-2)
136 #define REQ_NONE (-1)
137
138 /* Repeated character flags. */
139
140 #define UTF_LENGTH 0x10000000l /* The char contains its length. */
141
142 /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143 are simple data values; negative values are for special things like \d and so
144 on. Zero means further processing is needed (for things like \x), or the escape
145 is invalid. */
146
147 #ifndef EBCDIC
148
149 /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150 in UTF-8 mode. */
151
152 static const short int escapes[] = {
153 0, 0,
154 0, 0,
155 0, 0,
156 0, 0,
157 0, 0,
158 CHAR_COLON, CHAR_SEMICOLON,
159 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN,
160 CHAR_GREATER_THAN_SIGN, CHAR_QUESTION_MARK,
161 CHAR_COMMERCIAL_AT, -ESC_A,
162 -ESC_B, -ESC_C,
163 -ESC_D, -ESC_E,
164 0, -ESC_G,
165 -ESC_H, 0,
166 0, -ESC_K,
167 0, 0,
168 -ESC_N, 0,
169 -ESC_P, -ESC_Q,
170 -ESC_R, -ESC_S,
171 0, 0,
172 -ESC_V, -ESC_W,
173 -ESC_X, 0,
174 -ESC_Z, CHAR_LEFT_SQUARE_BRACKET,
175 CHAR_BACKSLASH, CHAR_RIGHT_SQUARE_BRACKET,
176 CHAR_CIRCUMFLEX_ACCENT, CHAR_UNDERSCORE,
177 CHAR_GRAVE_ACCENT, 7,
178 -ESC_b, 0,
179 -ESC_d, ESC_e,
180 ESC_f, 0,
181 -ESC_h, 0,
182 0, -ESC_k,
183 0, 0,
184 ESC_n, 0,
185 -ESC_p, 0,
186 ESC_r, -ESC_s,
187 ESC_tee, 0,
188 -ESC_v, -ESC_w,
189 0, 0,
190 -ESC_z
191 };
192
193 #else
194
195 /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
196
197 static const short int escapes[] = {
198 /* 48 */ 0, 0, 0, '.', '<', '(', '+', '|',
199 /* 50 */ '&', 0, 0, 0, 0, 0, 0, 0,
200 /* 58 */ 0, 0, '!', '$', '*', ')', ';', '~',
201 /* 60 */ '-', '/', 0, 0, 0, 0, 0, 0,
202 /* 68 */ 0, 0, '|', ',', '%', '_', '>', '?',
203 /* 70 */ 0, 0, 0, 0, 0, 0, 0, 0,
204 /* 78 */ 0, '`', ':', '#', '@', '\'', '=', '"',
205 /* 80 */ 0, 7, -ESC_b, 0, -ESC_d, ESC_e, ESC_f, 0,
206 /* 88 */-ESC_h, 0, 0, '{', 0, 0, 0, 0,
207 /* 90 */ 0, 0, -ESC_k, 'l', 0, ESC_n, 0, -ESC_p,
208 /* 98 */ 0, ESC_r, 0, '}', 0, 0, 0, 0,
209 /* A0 */ 0, '~', -ESC_s, ESC_tee, 0,-ESC_v, -ESC_w, 0,
210 /* A8 */ 0,-ESC_z, 0, 0, 0, '[', 0, 0,
211 /* B0 */ 0, 0, 0, 0, 0, 0, 0, 0,
212 /* B8 */ 0, 0, 0, 0, 0, ']', '=', '-',
213 /* C0 */ '{',-ESC_A, -ESC_B, -ESC_C, -ESC_D,-ESC_E, 0, -ESC_G,
214 /* C8 */-ESC_H, 0, 0, 0, 0, 0, 0, 0,
215 /* D0 */ '}', 0, -ESC_K, 0, 0,-ESC_N, 0, -ESC_P,
216 /* D8 */-ESC_Q,-ESC_R, 0, 0, 0, 0, 0, 0,
217 /* E0 */ '\\', 0, -ESC_S, 0, 0,-ESC_V, -ESC_W, -ESC_X,
218 /* E8 */ 0,-ESC_Z, 0, 0, 0, 0, 0, 0,
219 /* F0 */ 0, 0, 0, 0, 0, 0, 0, 0,
220 /* F8 */ 0, 0, 0, 0, 0, 0, 0, 0
221 };
222 #endif
223
224
225 /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226 searched linearly. Put all the names into a single string, in order to reduce
227 the number of relocations when a shared library is dynamically linked. The
228 string is built from string macros so that it works in UTF-8 mode on EBCDIC
229 platforms. */
230
231 typedef struct verbitem {
232 int len; /* Length of verb name */
233 int op; /* Op when no arg, or -1 if arg mandatory */
234 int op_arg; /* Op when arg present, or -1 if not allowed */
235 } verbitem;
236
237 static const char verbnames[] =
238 "\0" /* Empty name is a shorthand for MARK */
239 STRING_MARK0
240 STRING_ACCEPT0
241 STRING_COMMIT0
242 STRING_F0
243 STRING_FAIL0
244 STRING_PRUNE0
245 STRING_SKIP0
246 STRING_THEN;
247
248 static const verbitem verbs[] = {
249 { 0, -1, OP_MARK },
250 { 4, -1, OP_MARK },
251 { 6, OP_ACCEPT, -1 },
252 { 6, OP_COMMIT, -1 },
253 { 1, OP_FAIL, -1 },
254 { 4, OP_FAIL, -1 },
255 { 5, OP_PRUNE, OP_PRUNE_ARG },
256 { 4, OP_SKIP, OP_SKIP_ARG },
257 { 4, OP_THEN, OP_THEN_ARG }
258 };
259
260 static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261
262
263 /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264 another regex library. */
265
266 static const pcre_uchar sub_start_of_word[] = {
267 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268 CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269
270 static const pcre_uchar sub_end_of_word[] = {
271 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273 CHAR_RIGHT_PARENTHESIS, '\0' };
274
275
276 /* Tables of names of POSIX character classes and their lengths. The names are
277 now all in a single string, to reduce the number of relocations when a shared
278 library is dynamically loaded. The list of lengths is terminated by a zero
279 length entry. The first three must be alpha, lower, upper, as this is assumed
280 for handling case independence. The indices for graph, print, and punct are
281 needed, so identify them. */
282
283 static const char posix_names[] =
284 STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
285 STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
286 STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287 STRING_word0 STRING_xdigit;
288
289 static const pcre_uint8 posix_name_lengths[] = {
290 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291
292 #define PC_GRAPH 8
293 #define PC_PRINT 9
294 #define PC_PUNCT 10
295
296
297 /* Table of class bit maps for each POSIX class. Each class is formed from a
298 base map, with an optional addition or removal of another map. Then, for some
299 classes, there is some additional tweaking: for [:blank:] the vertical space
300 characters are removed, and for [:alpha:] and [:alnum:] the underscore
301 character is removed. The triples in the table consist of the base map offset,
302 second map offset or -1 if no second map, and a non-negative value for map
303 addition or a negative value for map subtraction (if there are two maps). The
304 absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
305 remove vertical space characters, 2 => remove underscore. */
306
307 static const int posix_class_maps[] = {
308 cbit_word, cbit_digit, -2, /* alpha */
309 cbit_lower, -1, 0, /* lower */
310 cbit_upper, -1, 0, /* upper */
311 cbit_word, -1, 2, /* alnum - word without underscore */
312 cbit_print, cbit_cntrl, 0, /* ascii */
313 cbit_space, -1, 1, /* blank - a GNU extension */
314 cbit_cntrl, -1, 0, /* cntrl */
315 cbit_digit, -1, 0, /* digit */
316 cbit_graph, -1, 0, /* graph */
317 cbit_print, -1, 0, /* print */
318 cbit_punct, -1, 0, /* punct */
319 cbit_space, -1, 0, /* space */
320 cbit_word, -1, 0, /* word - a Perl extension */
321 cbit_xdigit,-1, 0 /* xdigit */
322 };
323
324 /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325 Unicode property escapes. */
326
327 #ifdef SUPPORT_UCP
328 static const pcre_uchar string_PNd[] = {
329 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331 static const pcre_uchar string_pNd[] = {
332 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334 static const pcre_uchar string_PXsp[] = {
335 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337 static const pcre_uchar string_pXsp[] = {
338 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340 static const pcre_uchar string_PXwd[] = {
341 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343 static const pcre_uchar string_pXwd[] = {
344 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346
347 static const pcre_uchar *substitutes[] = {
348 string_PNd, /* \D */
349 string_pNd, /* \d */
350 string_PXsp, /* \S */ /* Xsp is Perl space, but from 8.34, Perl */
351 string_pXsp, /* \s */ /* space and POSIX space are the same. */
352 string_PXwd, /* \W */
353 string_pXwd /* \w */
354 };
355
356 /* The POSIX class substitutes must be in the order of the POSIX class names,
357 defined above, and there are both positive and negative cases. NULL means no
358 general substitute of a Unicode property escape (\p or \P). However, for some
359 POSIX classes (e.g. graph, print, punct) a special property code is compiled
360 directly. */
361
362 static const pcre_uchar string_pL[] = {
363 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365 static const pcre_uchar string_pLl[] = {
366 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368 static const pcre_uchar string_pLu[] = {
369 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371 static const pcre_uchar string_pXan[] = {
372 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374 static const pcre_uchar string_h[] = {
375 CHAR_BACKSLASH, CHAR_h, '\0' };
376 static const pcre_uchar string_pXps[] = {
377 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379 static const pcre_uchar string_PL[] = {
380 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382 static const pcre_uchar string_PLl[] = {
383 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385 static const pcre_uchar string_PLu[] = {
386 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388 static const pcre_uchar string_PXan[] = {
389 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391 static const pcre_uchar string_H[] = {
392 CHAR_BACKSLASH, CHAR_H, '\0' };
393 static const pcre_uchar string_PXps[] = {
394 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396
397 static const pcre_uchar *posix_substitutes[] = {
398 string_pL, /* alpha */
399 string_pLl, /* lower */
400 string_pLu, /* upper */
401 string_pXan, /* alnum */
402 NULL, /* ascii */
403 string_h, /* blank */
404 NULL, /* cntrl */
405 string_pNd, /* digit */
406 NULL, /* graph */
407 NULL, /* print */
408 NULL, /* punct */
409 string_pXps, /* space */ /* Xps is POSIX space, but from 8.34 */
410 string_pXwd, /* word */ /* Perl and POSIX space are the same */
411 NULL, /* xdigit */
412 /* Negated cases */
413 string_PL, /* ^alpha */
414 string_PLl, /* ^lower */
415 string_PLu, /* ^upper */
416 string_PXan, /* ^alnum */
417 NULL, /* ^ascii */
418 string_H, /* ^blank */
419 NULL, /* ^cntrl */
420 string_PNd, /* ^digit */
421 NULL, /* ^graph */
422 NULL, /* ^print */
423 NULL, /* ^punct */
424 string_PXps, /* ^space */ /* Xps is POSIX space, but from 8.34 */
425 string_PXwd, /* ^word */ /* Perl and POSIX space are the same */
426 NULL /* ^xdigit */
427 };
428 #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429 #endif
430
431 #define STRING(a) # a
432 #define XSTRING(s) STRING(s)
433
434 /* The texts of compile-time error messages. These are "char *" because they
435 are passed to the outside world. Do not ever re-use any error number, because
436 they are documented. Always add a new error instead. Messages marked DEAD below
437 are no longer used. This used to be a table of strings, but in order to reduce
438 the number of relocations needed when a shared library is loaded dynamically,
439 it is now one long string. We cannot use a table of offsets, because the
440 lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
441 simply count through to the one we want - this isn't a performance issue
442 because these strings are used only when there is a compilation error.
443
444 Each substring ends with \0 to insert a null character. This includes the final
445 substring, so that the whole string ends with \0\0, which can be detected when
446 counting through. */
447
448 static const char error_texts[] =
449 "no error\0"
450 "\\ at end of pattern\0"
451 "\\c at end of pattern\0"
452 "unrecognized character follows \\\0"
453 "numbers out of order in {} quantifier\0"
454 /* 5 */
455 "number too big in {} quantifier\0"
456 "missing terminating ] for character class\0"
457 "invalid escape sequence in character class\0"
458 "range out of order in character class\0"
459 "nothing to repeat\0"
460 /* 10 */
461 "operand of unlimited repeat could match the empty string\0" /** DEAD **/
462 "internal error: unexpected repeat\0"
463 "unrecognized character after (? or (?-\0"
464 "POSIX named classes are supported only within a class\0"
465 "missing )\0"
466 /* 15 */
467 "reference to non-existent subpattern\0"
468 "erroffset passed as NULL\0"
469 "unknown option bit(s) set\0"
470 "missing ) after comment\0"
471 "parentheses nested too deeply\0" /** DEAD **/
472 /* 20 */
473 "regular expression is too large\0"
474 "failed to get memory\0"
475 "unmatched parentheses\0"
476 "internal error: code overflow\0"
477 "unrecognized character after (?<\0"
478 /* 25 */
479 "lookbehind assertion is not fixed length\0"
480 "malformed number or name after (?(\0"
481 "conditional group contains more than two branches\0"
482 "assertion expected after (?(\0"
483 "(?R or (?[+-]digits must be followed by )\0"
484 /* 30 */
485 "unknown POSIX class name\0"
486 "POSIX collating elements are not supported\0"
487 "this version of PCRE is compiled without UTF support\0"
488 "spare error\0" /** DEAD **/
489 "character value in \\x{} or \\o{} is too large\0"
490 /* 35 */
491 "invalid condition (?(0)\0"
492 "\\C not allowed in lookbehind assertion\0"
493 "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
494 "number after (?C is > 255\0"
495 "closing ) for (?C expected\0"
496 /* 40 */
497 "recursive call could loop indefinitely\0"
498 "unrecognized character after (?P\0"
499 "syntax error in subpattern name (missing terminator)\0"
500 "two named subpatterns have the same name\0"
501 "invalid UTF-8 string\0"
502 /* 45 */
503 "support for \\P, \\p, and \\X has not been compiled\0"
504 "malformed \\P or \\p sequence\0"
505 "unknown property name after \\P or \\p\0"
506 "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
507 "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508 /* 50 */
509 "repeated subpattern is too long\0" /** DEAD **/
510 "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
511 "internal error: overran compiling workspace\0"
512 "internal error: previously-checked referenced subpattern not found\0"
513 "DEFINE group contains more than one branch\0"
514 /* 55 */
515 "repeating a DEFINE group is not allowed\0" /** DEAD **/
516 "inconsistent NEWLINE options\0"
517 "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
518 "a numbered reference must not be zero\0"
519 "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520 /* 60 */
521 "(*VERB) not recognized or malformed\0"
522 "number is too big\0"
523 "subpattern name expected\0"
524 "digit expected after (?+\0"
525 "] is an invalid data character in JavaScript compatibility mode\0"
526 /* 65 */
527 "different names for subpatterns of the same number are not allowed\0"
528 "(*MARK) must have an argument\0"
529 "this version of PCRE is not compiled with Unicode property support\0"
530 "\\c must be followed by an ASCII character\0"
531 "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532 /* 70 */
533 "internal error: unknown opcode in find_fixedlength()\0"
534 "\\N is not supported in a class\0"
535 "too many forward references\0"
536 "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537 "invalid UTF-16 string\0"
538 /* 75 */
539 "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540 "character value in \\u.... sequence is too large\0"
541 "invalid UTF-32 string\0"
542 "setting UTF is disabled by the application\0"
543 "non-hex character in \\x{} (closing brace missing?)\0"
544 /* 80 */
545 "non-octal character in \\o{} (closing brace missing?)\0"
546 "missing opening brace after \\o\0"
547 "parentheses are too deeply nested\0"
548 "invalid range in character class\0"
549 "group name must start with a non-digit\0"
550 /* 85 */
551 "parentheses are too deeply nested (stack check)\0"
552 "digits missing in \\x{} or \\o{}\0"
553 ;
554
555 /* Table to identify digits and hex digits. This is used when compiling
556 patterns. Note that the tables in chartables are dependent on the locale, and
557 may mark arbitrary characters as digits - but the PCRE compiling code expects
558 to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
559 a private table here. It costs 256 bytes, but it is a lot faster than doing
560 character value tests (at least in some simple cases I timed), and in some
561 applications one wants PCRE to compile efficiently as well as match
562 efficiently.
563
564 For convenience, we use the same bit definitions as in chartables:
565
566 0x04 decimal digit
567 0x08 hexadecimal digit
568
569 Then we can use ctype_digit and ctype_xdigit in the code. */
570
571 /* Using a simple comparison for decimal numbers rather than a memory read
572 is much faster, and the resulting code is simpler (the compiler turns it
573 into a subtraction and unsigned comparison). */
574
575 #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
576
577 #ifndef EBCDIC
578
579 /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
580 UTF-8 mode. */
581
582 static const pcre_uint8 digitab[] =
583 {
584 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
585 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
586 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
587 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
588 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
589 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
590 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
591 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
592 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* @ - G */
593 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H - O */
594 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* P - W */
595 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* X - _ */
596 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
597 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
598 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
599 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
600 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
601 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
602 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
603 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
604 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
605 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
606 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
607 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
608 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
609 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
610 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
611 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
612 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
613 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
614 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
615 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
616
617 #else
618
619 /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
620
621 static const pcre_uint8 digitab[] =
622 {
623 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
624 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
625 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
626 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
627 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
628 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
629 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
630 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
631 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
632 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
633 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
634 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
635 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
636 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
637 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
638 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
639 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
640 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
641 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
642 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
643 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
644 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
645 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
646 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
647 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
648 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
649 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
650 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
651 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
652 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
653 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
654 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
655
656 static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
657 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
658 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
659 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
660 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
661 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
662 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
663 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
664 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
665 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
666 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
667 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
668 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
669 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
670 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
671 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
672 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
673 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
674 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
675 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
676 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
677 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
678 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
679 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
680 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
681 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
682 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
683 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
684 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
685 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
686 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
687 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
688 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
689 #endif
690
691
692 /* This table is used to check whether auto-possessification is possible
693 between adjacent character-type opcodes. The left-hand (repeated) opcode is
694 used to select the row, and the right-hand opcode is use to select the column.
695 A value of 1 means that auto-possessification is OK. For example, the second
696 value in the first row means that \D+\d can be turned into \D++\d.
697
698 The Unicode property types (\P and \p) have to be present to fill out the table
699 because of what their opcode values are, but the table values should always be
700 zero because property types are handled separately in the code. The last four
701 columns apply to items that cannot be repeated, so there is no need to have
702 rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
703 *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
704
705 #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
706 #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
707
708 static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
709 /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
710 { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
711 { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
712 { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
713 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
714 { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
715 { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
716 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
717 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
718 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
719 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
720 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
721 { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
722 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
723 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
724 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
725 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
726 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
727 };
728
729
730 /* This table is used to check whether auto-possessification is possible
731 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
732 left-hand (repeated) opcode is used to select the row, and the right-hand
733 opcode is used to select the column. The values are as follows:
734
735 0 Always return FALSE (never auto-possessify)
736 1 Character groups are distinct (possessify if both are OP_PROP)
737 2 Check character categories in the same group (general or particular)
738 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
739
740 4 Check left general category vs right particular category
741 5 Check right general category vs left particular category
742
743 6 Left alphanum vs right general category
744 7 Left space vs right general category
745 8 Left word vs right general category
746
747 9 Right alphanum vs left general category
748 10 Right space vs left general category
749 11 Right word vs left general category
750
751 12 Left alphanum vs right particular category
752 13 Left space vs right particular category
753 14 Left word vs right particular category
754
755 15 Right alphanum vs left particular category
756 16 Right space vs left particular category
757 17 Right word vs left particular category
758 */
759
760 static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
761 /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
762 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
763 { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
764 { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
765 { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
766 { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
767 { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
768 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
769 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
770 { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
771 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
772 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
773 };
774
775 /* This table is used to check whether auto-possessification is possible
776 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
777 specifies a general category and the other specifies a particular category. The
778 row is selected by the general category and the column by the particular
779 category. The value is 1 if the particular category is not part of the general
780 category. */
781
782 static const pcre_uint8 catposstab[7][30] = {
783 /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
784 { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* C */
785 { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* L */
786 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* M */
787 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* N */
788 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 }, /* P */
789 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 }, /* S */
790 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 } /* Z */
791 };
792
793 /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
794 a general or particular category. The properties in each row are those
795 that apply to the character set in question. Duplication means that a little
796 unnecessary work is done when checking, but this keeps things much simpler
797 because they can all use the same code. For more details see the comment where
798 this table is used.
799
800 Note: SPACE and PXSPACE used to be different because Perl excluded VT from
801 "space", but from Perl 5.18 it's included, so both categories are treated the
802 same here. */
803
804 static const pcre_uint8 posspropstab[3][4] = {
805 { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
806 { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
807 { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
808 };
809
810 /* This table is used when converting repeating opcodes into possessified
811 versions as a result of an explicit possessive quantifier such as ++. A zero
812 value means there is no possessified version - in those cases the item in
813 question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
814 because all relevant opcodes are less than that. */
815
816 static const pcre_uint8 opcode_possessify[] = {
817 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */
818 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */
819
820 0, /* NOTI */
821 OP_POSSTAR, 0, /* STAR, MINSTAR */
822 OP_POSPLUS, 0, /* PLUS, MINPLUS */
823 OP_POSQUERY, 0, /* QUERY, MINQUERY */
824 OP_POSUPTO, 0, /* UPTO, MINUPTO */
825 0, /* EXACT */
826 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */
827
828 OP_POSSTARI, 0, /* STARI, MINSTARI */
829 OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */
830 OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */
831 OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */
832 0, /* EXACTI */
833 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */
834
835 OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */
836 OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */
837 OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */
838 OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */
839 0, /* NOTEXACT */
840 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
841
842 OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */
843 OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */
844 OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */
845 OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */
846 0, /* NOTEXACTI */
847 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
848
849 OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */
850 OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */
851 OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */
852 OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */
853 0, /* TYPEEXACT */
854 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
855
856 OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */
857 OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */
858 OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */
859 OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */
860 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */
861
862 0, 0, 0, /* CLASS, NCLASS, XCLASS */
863 0, 0, /* REF, REFI */
864 0, 0, /* DNREF, DNREFI */
865 0, 0 /* RECURSE, CALLOUT */
866 };
867
868
869
870 /*************************************************
871 * Find an error text *
872 *************************************************/
873
874 /* The error texts are now all in one long string, to save on relocations. As
875 some of the text is of unknown length, we can't use a table of offsets.
876 Instead, just count through the strings. This is not a performance issue
877 because it happens only when there has been a compilation error.
878
879 Argument: the error number
880 Returns: pointer to the error string
881 */
882
883 static const char *
884 find_error_text(int n)
885 {
886 const char *s = error_texts;
887 for (; n > 0; n--)
888 {
889 while (*s++ != CHAR_NULL) {};
890 if (*s == CHAR_NULL) return "Error text not found (please report)";
891 }
892 return s;
893 }
894
895
896
897 /*************************************************
898 * Expand the workspace *
899 *************************************************/
900
901 /* This function is called during the second compiling phase, if the number of
902 forward references fills the existing workspace, which is originally a block on
903 the stack. A larger block is obtained from malloc() unless the ultimate limit
904 has been reached or the increase will be rather small.
905
906 Argument: pointer to the compile data block
907 Returns: 0 if all went well, else an error number
908 */
909
910 static int
911 expand_workspace(compile_data *cd)
912 {
913 pcre_uchar *newspace;
914 int newsize = cd->workspace_size * 2;
915
916 if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
917 if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
918 newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
919 return ERR72;
920
921 newspace = (PUBL(malloc))(IN_UCHARS(newsize));
922 if (newspace == NULL) return ERR21;
923 memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
924 cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
925 if (cd->workspace_size > COMPILE_WORK_SIZE)
926 (PUBL(free))((void *)cd->start_workspace);
927 cd->start_workspace = newspace;
928 cd->workspace_size = newsize;
929 return 0;
930 }
931
932
933
934 /*************************************************
935 * Check for counted repeat *
936 *************************************************/
937
938 /* This function is called when a '{' is encountered in a place where it might
939 start a quantifier. It looks ahead to see if it really is a quantifier or not.
940 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
941 where the ddds are digits.
942
943 Arguments:
944 p pointer to the first char after '{'
945
946 Returns: TRUE or FALSE
947 */
948
949 static BOOL
950 is_counted_repeat(const pcre_uchar *p)
951 {
952 if (!IS_DIGIT(*p)) return FALSE;
953 p++;
954 while (IS_DIGIT(*p)) p++;
955 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
956
957 if (*p++ != CHAR_COMMA) return FALSE;
958 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
959
960 if (!IS_DIGIT(*p)) return FALSE;
961 p++;
962 while (IS_DIGIT(*p)) p++;
963
964 return (*p == CHAR_RIGHT_CURLY_BRACKET);
965 }
966
967
968
969 /*************************************************
970 * Handle escapes *
971 *************************************************/
972
973 /* This function is called when a \ has been encountered. It either returns a
974 positive value for a simple escape such as \n, or 0 for a data character which
975 will be placed in chptr. A backreference to group n is returned as negative n.
976 When UTF-8 is enabled, a positive value greater than 255 may be returned in
977 chptr. On entry, ptr is pointing at the \. On exit, it is on the final
978 character of the escape sequence.
979
980 Arguments:
981 ptrptr points to the pattern position pointer
982 chptr points to a returned data character
983 errorcodeptr points to the errorcode variable
984 bracount number of previous extracting brackets
985 options the options bits
986 isclass TRUE if inside a character class
987
988 Returns: zero => a data character
989 positive => a special escape sequence
990 negative => a back reference
991 on error, errorcodeptr is set
992 */
993
994 static int
995 check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
996 int bracount, int options, BOOL isclass)
997 {
998 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
999 BOOL utf = (options & PCRE_UTF8) != 0;
1000 const pcre_uchar *ptr = *ptrptr + 1;
1001 pcre_uint32 c;
1002 int escape = 0;
1003 int i;
1004
1005 GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
1006 ptr--; /* Set pointer back to the last byte */
1007
1008 /* If backslash is at the end of the pattern, it's an error. */
1009
1010 if (c == CHAR_NULL) *errorcodeptr = ERR1;
1011
1012 /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1013 in a table. A non-zero result is something that can be returned immediately.
1014 Otherwise further processing may be required. */
1015
1016 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1017 /* Not alphanumeric */
1018 else if (c < CHAR_0 || c > CHAR_z) {}
1019 else if ((i = escapes[c - CHAR_0]) != 0)
1020 { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1021
1022 #else /* EBCDIC coding */
1023 /* Not alphanumeric */
1024 else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1025 else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1026 #endif
1027
1028 /* Escapes that need further processing, or are illegal. */
1029
1030 else
1031 {
1032 const pcre_uchar *oldptr;
1033 BOOL braced, negated, overflow;
1034 int s;
1035
1036 switch (c)
1037 {
1038 /* A number of Perl escapes are not handled by PCRE. We give an explicit
1039 error. */
1040
1041 case CHAR_l:
1042 case CHAR_L:
1043 *errorcodeptr = ERR37;
1044 break;
1045
1046 case CHAR_u:
1047 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1048 {
1049 /* In JavaScript, \u must be followed by four hexadecimal numbers.
1050 Otherwise it is a lowercase u letter. */
1051 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1052 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1053 && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1054 && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1055 {
1056 c = 0;
1057 for (i = 0; i < 4; ++i)
1058 {
1059 register pcre_uint32 cc = *(++ptr);
1060 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1061 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1062 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1063 #else /* EBCDIC coding */
1064 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1065 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1066 #endif
1067 }
1068
1069 #if defined COMPILE_PCRE8
1070 if (c > (utf ? 0x10ffffU : 0xffU))
1071 #elif defined COMPILE_PCRE16
1072 if (c > (utf ? 0x10ffffU : 0xffffU))
1073 #elif defined COMPILE_PCRE32
1074 if (utf && c > 0x10ffffU)
1075 #endif
1076 {
1077 *errorcodeptr = ERR76;
1078 }
1079 else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1080 }
1081 }
1082 else
1083 *errorcodeptr = ERR37;
1084 break;
1085
1086 case CHAR_U:
1087 /* In JavaScript, \U is an uppercase U letter. */
1088 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1089 break;
1090
1091 /* In a character class, \g is just a literal "g". Outside a character
1092 class, \g must be followed by one of a number of specific things:
1093
1094 (1) A number, either plain or braced. If positive, it is an absolute
1095 backreference. If negative, it is a relative backreference. This is a Perl
1096 5.10 feature.
1097
1098 (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1099 is part of Perl's movement towards a unified syntax for back references. As
1100 this is synonymous with \k{name}, we fudge it up by pretending it really
1101 was \k.
1102
1103 (3) For Oniguruma compatibility we also support \g followed by a name or a
1104 number either in angle brackets or in single quotes. However, these are
1105 (possibly recursive) subroutine calls, _not_ backreferences. Just return
1106 the ESC_g code (cf \k). */
1107
1108 case CHAR_g:
1109 if (isclass) break;
1110 if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1111 {
1112 escape = ESC_g;
1113 break;
1114 }
1115
1116 /* Handle the Perl-compatible cases */
1117
1118 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1119 {
1120 const pcre_uchar *p;
1121 for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1122 if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1123 if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1124 {
1125 escape = ESC_k;
1126 break;
1127 }
1128 braced = TRUE;
1129 ptr++;
1130 }
1131 else braced = FALSE;
1132
1133 if (ptr[1] == CHAR_MINUS)
1134 {
1135 negated = TRUE;
1136 ptr++;
1137 }
1138 else negated = FALSE;
1139
1140 /* The integer range is limited by the machine's int representation. */
1141 s = 0;
1142 overflow = FALSE;
1143 while (IS_DIGIT(ptr[1]))
1144 {
1145 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1146 {
1147 overflow = TRUE;
1148 break;
1149 }
1150 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1151 }
1152 if (overflow) /* Integer overflow */
1153 {
1154 while (IS_DIGIT(ptr[1]))
1155 ptr++;
1156 *errorcodeptr = ERR61;
1157 break;
1158 }
1159
1160 if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1161 {
1162 *errorcodeptr = ERR57;
1163 break;
1164 }
1165
1166 if (s == 0)
1167 {
1168 *errorcodeptr = ERR58;
1169 break;
1170 }
1171
1172 if (negated)
1173 {
1174 if (s > bracount)
1175 {
1176 *errorcodeptr = ERR15;
1177 break;
1178 }
1179 s = bracount - (s - 1);
1180 }
1181
1182 escape = -s;
1183 break;
1184
1185 /* The handling of escape sequences consisting of a string of digits
1186 starting with one that is not zero is not straightforward. Perl has changed
1187 over the years. Nowadays \g{} for backreferences and \o{} for octal are
1188 recommended to avoid the ambiguities in the old syntax.
1189
1190 Outside a character class, the digits are read as a decimal number. If the
1191 number is less than 8 (used to be 10), or if there are that many previous
1192 extracting left brackets, then it is a back reference. Otherwise, up to
1193 three octal digits are read to form an escaped byte. Thus \123 is likely to
1194 be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1195 the octal value is greater than 377, the least significant 8 bits are
1196 taken. \8 and \9 are treated as the literal characters 8 and 9.
1197
1198 Inside a character class, \ followed by a digit is always either a literal
1199 8 or 9 or an octal number. */
1200
1201 case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1202 case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1203
1204 if (!isclass)
1205 {
1206 oldptr = ptr;
1207 /* The integer range is limited by the machine's int representation. */
1208 s = (int)(c -CHAR_0);
1209 overflow = FALSE;
1210 while (IS_DIGIT(ptr[1]))
1211 {
1212 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1213 {
1214 overflow = TRUE;
1215 break;
1216 }
1217 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1218 }
1219 if (overflow) /* Integer overflow */
1220 {
1221 while (IS_DIGIT(ptr[1]))
1222 ptr++;
1223 *errorcodeptr = ERR61;
1224 break;
1225 }
1226 if (s < 8 || s <= bracount) /* Check for back reference */
1227 {
1228 escape = -s;
1229 break;
1230 }
1231 ptr = oldptr; /* Put the pointer back and fall through */
1232 }
1233
1234 /* Handle a digit following \ when the number is not a back reference. If
1235 the first digit is 8 or 9, Perl used to generate a binary zero byte and
1236 then treat the digit as a following literal. At least by Perl 5.18 this
1237 changed so as not to insert the binary zero. */
1238
1239 if ((c = *ptr) >= CHAR_8) break;
1240
1241 /* Fall through with a digit less than 8 */
1242
1243 /* \0 always starts an octal number, but we may drop through to here with a
1244 larger first octal digit. The original code used just to take the least
1245 significant 8 bits of octal numbers (I think this is what early Perls used
1246 to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1247 but no more than 3 octal digits. */
1248
1249 case CHAR_0:
1250 c -= CHAR_0;
1251 while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1252 c = c * 8 + *(++ptr) - CHAR_0;
1253 #ifdef COMPILE_PCRE8
1254 if (!utf && c > 0xff) *errorcodeptr = ERR51;
1255 #endif
1256 break;
1257
1258 /* \o is a relatively new Perl feature, supporting a more general way of
1259 specifying character codes in octal. The only supported form is \o{ddd}. */
1260
1261 case CHAR_o:
1262 if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1263 if (ptr[2] == CHAR_RIGHT_CURLY_BRACKET) *errorcodeptr = ERR86; else
1264 {
1265 ptr += 2;
1266 c = 0;
1267 overflow = FALSE;
1268 while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1269 {
1270 register pcre_uint32 cc = *ptr++;
1271 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1272 #ifdef COMPILE_PCRE32
1273 if (c >= 0x20000000l) { overflow = TRUE; break; }
1274 #endif
1275 c = (c << 3) + cc - CHAR_0 ;
1276 #if defined COMPILE_PCRE8
1277 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1278 #elif defined COMPILE_PCRE16
1279 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1280 #elif defined COMPILE_PCRE32
1281 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1282 #endif
1283 }
1284 if (overflow)
1285 {
1286 while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1287 *errorcodeptr = ERR34;
1288 }
1289 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1290 {
1291 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1292 }
1293 else *errorcodeptr = ERR80;
1294 }
1295 break;
1296
1297 /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1298 numbers. Otherwise it is a lowercase x letter. */
1299
1300 case CHAR_x:
1301 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1302 {
1303 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1304 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1305 {
1306 c = 0;
1307 for (i = 0; i < 2; ++i)
1308 {
1309 register pcre_uint32 cc = *(++ptr);
1310 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1311 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1312 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1313 #else /* EBCDIC coding */
1314 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1315 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1316 #endif
1317 }
1318 }
1319 } /* End JavaScript handling */
1320
1321 /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1322 greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1323 digits. If not, { used to be treated as a data character. However, Perl
1324 seems to read hex digits up to the first non-such, and ignore the rest, so
1325 that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1326 now gives an error. */
1327
1328 else
1329 {
1330 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1331 {
1332 ptr += 2;
1333 if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1334 {
1335 *errorcodeptr = ERR86;
1336 break;
1337 }
1338 c = 0;
1339 overflow = FALSE;
1340 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1341 {
1342 register pcre_uint32 cc = *ptr++;
1343 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1344
1345 #ifdef COMPILE_PCRE32
1346 if (c >= 0x10000000l) { overflow = TRUE; break; }
1347 #endif
1348
1349 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1350 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1351 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1352 #else /* EBCDIC coding */
1353 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1354 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1355 #endif
1356
1357 #if defined COMPILE_PCRE8
1358 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1359 #elif defined COMPILE_PCRE16
1360 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1361 #elif defined COMPILE_PCRE32
1362 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1363 #endif
1364 }
1365
1366 if (overflow)
1367 {
1368 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1369 *errorcodeptr = ERR34;
1370 }
1371
1372 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1373 {
1374 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1375 }
1376
1377 /* If the sequence of hex digits does not end with '}', give an error.
1378 We used just to recognize this construct and fall through to the normal
1379 \x handling, but nowadays Perl gives an error, which seems much more
1380 sensible, so we do too. */
1381
1382 else *errorcodeptr = ERR79;
1383 } /* End of \x{} processing */
1384
1385 /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1386
1387 else
1388 {
1389 c = 0;
1390 while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1391 {
1392 pcre_uint32 cc; /* Some compilers don't like */
1393 cc = *(++ptr); /* ++ in initializers */
1394 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1395 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1396 c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1397 #else /* EBCDIC coding */
1398 if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1399 c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1400 #endif
1401 }
1402 } /* End of \xdd handling */
1403 } /* End of Perl-style \x handling */
1404 break;
1405
1406 /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1407 An error is given if the byte following \c is not an ASCII character. This
1408 coding is ASCII-specific, but then the whole concept of \cx is
1409 ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1410
1411 case CHAR_c:
1412 c = *(++ptr);
1413 if (c == CHAR_NULL)
1414 {
1415 *errorcodeptr = ERR2;
1416 break;
1417 }
1418 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1419 if (c > 127) /* Excludes all non-ASCII in either mode */
1420 {
1421 *errorcodeptr = ERR68;
1422 break;
1423 }
1424 if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1425 c ^= 0x40;
1426 #else /* EBCDIC coding */
1427 if (c >= CHAR_a && c <= CHAR_z) c += 64;
1428 c ^= 0xC0;
1429 #endif
1430 break;
1431
1432 /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1433 other alphanumeric following \ is an error if PCRE_EXTRA was set;
1434 otherwise, for Perl compatibility, it is a literal. This code looks a bit
1435 odd, but there used to be some cases other than the default, and there may
1436 be again in future, so I haven't "optimized" it. */
1437
1438 default:
1439 if ((options & PCRE_EXTRA) != 0) switch(c)
1440 {
1441 default:
1442 *errorcodeptr = ERR3;
1443 break;
1444 }
1445 break;
1446 }
1447 }
1448
1449 /* Perl supports \N{name} for character names, as well as plain \N for "not
1450 newline". PCRE does not support \N{name}. However, it does support
1451 quantification such as \N{2,3}. */
1452
1453 if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1454 !is_counted_repeat(ptr+2))
1455 *errorcodeptr = ERR37;
1456
1457 /* If PCRE_UCP is set, we change the values for \d etc. */
1458
1459 if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1460 escape += (ESC_DU - ESC_D);
1461
1462 /* Set the pointer to the final character before returning. */
1463
1464 *ptrptr = ptr;
1465 *chptr = c;
1466 return escape;
1467 }
1468
1469
1470
1471 #ifdef SUPPORT_UCP
1472 /*************************************************
1473 * Handle \P and \p *
1474 *************************************************/
1475
1476 /* This function is called after \P or \p has been encountered, provided that
1477 PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1478 pointing at the P or p. On exit, it is pointing at the final character of the
1479 escape sequence.
1480
1481 Argument:
1482 ptrptr points to the pattern position pointer
1483 negptr points to a boolean that is set TRUE for negation else FALSE
1484 ptypeptr points to an unsigned int that is set to the type value
1485 pdataptr points to an unsigned int that is set to the detailed property value
1486 errorcodeptr points to the error code variable
1487
1488 Returns: TRUE if the type value was found, or FALSE for an invalid type
1489 */
1490
1491 static BOOL
1492 get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1493 unsigned int *pdataptr, int *errorcodeptr)
1494 {
1495 pcre_uchar c;
1496 int i, bot, top;
1497 const pcre_uchar *ptr = *ptrptr;
1498 pcre_uchar name[32];
1499
1500 c = *(++ptr);
1501 if (c == CHAR_NULL) goto ERROR_RETURN;
1502
1503 *negptr = FALSE;
1504
1505 /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1506 negation. */
1507
1508 if (c == CHAR_LEFT_CURLY_BRACKET)
1509 {
1510 if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1511 {
1512 *negptr = TRUE;
1513 ptr++;
1514 }
1515 for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1516 {
1517 c = *(++ptr);
1518 if (c == CHAR_NULL) goto ERROR_RETURN;
1519 if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1520 name[i] = c;
1521 }
1522 if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1523 name[i] = 0;
1524 }
1525
1526 /* Otherwise there is just one following character */
1527
1528 else
1529 {
1530 name[0] = c;
1531 name[1] = 0;
1532 }
1533
1534 *ptrptr = ptr;
1535
1536 /* Search for a recognized property name using binary chop */
1537
1538 bot = 0;
1539 top = PRIV(utt_size);
1540
1541 while (bot < top)
1542 {
1543 int r;
1544 i = (bot + top) >> 1;
1545 r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1546 if (r == 0)
1547 {
1548 *ptypeptr = PRIV(utt)[i].type;
1549 *pdataptr = PRIV(utt)[i].value;
1550 return TRUE;
1551 }
1552 if (r > 0) bot = i + 1; else top = i;
1553 }
1554
1555 *errorcodeptr = ERR47;
1556 *ptrptr = ptr;
1557 return FALSE;
1558
1559 ERROR_RETURN:
1560 *errorcodeptr = ERR46;
1561 *ptrptr = ptr;
1562 return FALSE;
1563 }
1564 #endif
1565
1566
1567
1568 /*************************************************
1569 * Read repeat counts *
1570 *************************************************/
1571
1572 /* Read an item of the form {n,m} and return the values. This is called only
1573 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1574 so the syntax is guaranteed to be correct, but we need to check the values.
1575
1576 Arguments:
1577 p pointer to first char after '{'
1578 minp pointer to int for min
1579 maxp pointer to int for max
1580 returned as -1 if no max
1581 errorcodeptr points to error code variable
1582
1583 Returns: pointer to '}' on success;
1584 current ptr on error, with errorcodeptr set non-zero
1585 */
1586
1587 static const pcre_uchar *
1588 read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1589 {
1590 int min = 0;
1591 int max = -1;
1592
1593 while (IS_DIGIT(*p))
1594 {
1595 min = min * 10 + (int)(*p++ - CHAR_0);
1596 if (min > 65535)
1597 {
1598 *errorcodeptr = ERR5;
1599 return p;
1600 }
1601 }
1602
1603 if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1604 {
1605 if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1606 {
1607 max = 0;
1608 while(IS_DIGIT(*p))
1609 {
1610 max = max * 10 + (int)(*p++ - CHAR_0);
1611 if (max > 65535)
1612 {
1613 *errorcodeptr = ERR5;
1614 return p;
1615 }
1616 }
1617 if (max < min)
1618 {
1619 *errorcodeptr = ERR4;
1620 return p;
1621 }
1622 }
1623 }
1624
1625 *minp = min;
1626 *maxp = max;
1627 return p;
1628 }
1629
1630
1631
1632 /*************************************************
1633 * Find first significant op code *
1634 *************************************************/
1635
1636 /* This is called by several functions that scan a compiled expression looking
1637 for a fixed first character, or an anchoring op code etc. It skips over things
1638 that do not influence this. For some calls, it makes sense to skip negative
1639 forward and all backward assertions, and also the \b assertion; for others it
1640 does not.
1641
1642 Arguments:
1643 code pointer to the start of the group
1644 skipassert TRUE if certain assertions are to be skipped
1645
1646 Returns: pointer to the first significant opcode
1647 */
1648
1649 static const pcre_uchar*
1650 first_significant_code(const pcre_uchar *code, BOOL skipassert)
1651 {
1652 for (;;)
1653 {
1654 switch ((int)*code)
1655 {
1656 case OP_ASSERT_NOT:
1657 case OP_ASSERTBACK:
1658 case OP_ASSERTBACK_NOT:
1659 if (!skipassert) return code;
1660 do code += GET(code, 1); while (*code == OP_ALT);
1661 code += PRIV(OP_lengths)[*code];
1662 break;
1663
1664 case OP_WORD_BOUNDARY:
1665 case OP_NOT_WORD_BOUNDARY:
1666 if (!skipassert) return code;
1667 /* Fall through */
1668
1669 case OP_CALLOUT:
1670 case OP_CREF:
1671 case OP_DNCREF:
1672 case OP_RREF:
1673 case OP_DNRREF:
1674 case OP_DEF:
1675 code += PRIV(OP_lengths)[*code];
1676 break;
1677
1678 default:
1679 return code;
1680 }
1681 }
1682 /* Control never reaches here */
1683 }
1684
1685
1686
1687 /*************************************************
1688 * Find the fixed length of a branch *
1689 *************************************************/
1690
1691 /* Scan a branch and compute the fixed length of subject that will match it,
1692 if the length is fixed. This is needed for dealing with backward assertions.
1693 In UTF8 mode, the result is in characters rather than bytes. The branch is
1694 temporarily terminated with OP_END when this function is called.
1695
1696 This function is called when a backward assertion is encountered, so that if it
1697 fails, the error message can point to the correct place in the pattern.
1698 However, we cannot do this when the assertion contains subroutine calls,
1699 because they can be forward references. We solve this by remembering this case
1700 and doing the check at the end; a flag specifies which mode we are running in.
1701
1702 Arguments:
1703 code points to the start of the pattern (the bracket)
1704 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1705 atend TRUE if called when the pattern is complete
1706 cd the "compile data" structure
1707 recurses chain of recurse_check to catch mutual recursion
1708
1709 Returns: the fixed length,
1710 or -1 if there is no fixed length,
1711 or -2 if \C was encountered (in UTF-8 mode only)
1712 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1713 or -4 if an unknown opcode was encountered (internal error)
1714 */
1715
1716 static int
1717 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd,
1718 recurse_check *recurses)
1719 {
1720 int length = -1;
1721 recurse_check this_recurse;
1722 register int branchlength = 0;
1723 register pcre_uchar *cc = code + 1 + LINK_SIZE;
1724
1725 /* Scan along the opcodes for this branch. If we get to the end of the
1726 branch, check the length against that of the other branches. */
1727
1728 for (;;)
1729 {
1730 int d;
1731 pcre_uchar *ce, *cs;
1732 register pcre_uchar op = *cc;
1733
1734 switch (op)
1735 {
1736 /* We only need to continue for OP_CBRA (normal capturing bracket) and
1737 OP_BRA (normal non-capturing bracket) because the other variants of these
1738 opcodes are all concerned with unlimited repeated groups, which of course
1739 are not of fixed length. */
1740
1741 case OP_CBRA:
1742 case OP_BRA:
1743 case OP_ONCE:
1744 case OP_ONCE_NC:
1745 case OP_COND:
1746 d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd,
1747 recurses);
1748 if (d < 0) return d;
1749 branchlength += d;
1750 do cc += GET(cc, 1); while (*cc == OP_ALT);
1751 cc += 1 + LINK_SIZE;
1752 break;
1753
1754 /* Reached end of a branch; if it's a ket it is the end of a nested call.
1755 If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1756 an ALT. If it is END it's the end of the outer call. All can be handled by
1757 the same code. Note that we must not include the OP_KETRxxx opcodes here,
1758 because they all imply an unlimited repeat. */
1759
1760 case OP_ALT:
1761 case OP_KET:
1762 case OP_END:
1763 case OP_ACCEPT:
1764 case OP_ASSERT_ACCEPT:
1765 if (length < 0) length = branchlength;
1766 else if (length != branchlength) return -1;
1767 if (*cc != OP_ALT) return length;
1768 cc += 1 + LINK_SIZE;
1769 branchlength = 0;
1770 break;
1771
1772 /* A true recursion implies not fixed length, but a subroutine call may
1773 be OK. If the subroutine is a forward reference, we can't deal with
1774 it until the end of the pattern, so return -3. */
1775
1776 case OP_RECURSE:
1777 if (!atend) return -3;
1778 cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1779 do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1780 if (cc > cs && cc < ce) return -1; /* Recursion */
1781 else /* Check for mutual recursion */
1782 {
1783 recurse_check *r = recurses;
1784 for (r = recurses; r != NULL; r = r->prev) if (r->group == cs) break;
1785 if (r != NULL) return -1; /* Mutual recursion */
1786 }
1787 this_recurse.prev = recurses;
1788 this_recurse.group = cs;
1789 d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd, &this_recurse);
1790 if (d < 0) return d;
1791 branchlength += d;
1792 cc += 1 + LINK_SIZE;
1793 break;
1794
1795 /* Skip over assertive subpatterns */
1796
1797 case OP_ASSERT:
1798 case OP_ASSERT_NOT:
1799 case OP_ASSERTBACK:
1800 case OP_ASSERTBACK_NOT:
1801 do cc += GET(cc, 1); while (*cc == OP_ALT);
1802 cc += 1 + LINK_SIZE;
1803 break;
1804
1805 /* Skip over things that don't match chars */
1806
1807 case OP_MARK:
1808 case OP_PRUNE_ARG:
1809 case OP_SKIP_ARG:
1810 case OP_THEN_ARG:
1811 cc += cc[1] + PRIV(OP_lengths)[*cc];
1812 break;
1813
1814 case OP_CALLOUT:
1815 case OP_CIRC:
1816 case OP_CIRCM:
1817 case OP_CLOSE:
1818 case OP_COMMIT:
1819 case OP_CREF:
1820 case OP_DEF:
1821 case OP_DNCREF:
1822 case OP_DNRREF:
1823 case OP_DOLL:
1824 case OP_DOLLM:
1825 case OP_EOD:
1826 case OP_EODN:
1827 case OP_FAIL:
1828 case OP_NOT_WORD_BOUNDARY:
1829 case OP_PRUNE:
1830 case OP_REVERSE:
1831 case OP_RREF:
1832 case OP_SET_SOM:
1833 case OP_SKIP:
1834 case OP_SOD:
1835 case OP_SOM:
1836 case OP_THEN:
1837 case OP_WORD_BOUNDARY:
1838 cc += PRIV(OP_lengths)[*cc];
1839 break;
1840
1841 /* Handle literal characters */
1842
1843 case OP_CHAR:
1844 case OP_CHARI:
1845 case OP_NOT:
1846 case OP_NOTI:
1847 branchlength++;
1848 cc += 2;
1849 #ifdef SUPPORT_UTF
1850 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1851 #endif
1852 break;
1853
1854 /* Handle exact repetitions. The count is already in characters, but we
1855 need to skip over a multibyte character in UTF8 mode. */
1856
1857 case OP_EXACT:
1858 case OP_EXACTI:
1859 case OP_NOTEXACT:
1860 case OP_NOTEXACTI:
1861 branchlength += (int)GET2(cc,1);
1862 cc += 2 + IMM2_SIZE;
1863 #ifdef SUPPORT_UTF
1864 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1865 #endif
1866 break;
1867
1868 case OP_TYPEEXACT:
1869 branchlength += GET2(cc,1);
1870 if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1871 cc += 2;
1872 cc += 1 + IMM2_SIZE + 1;
1873 break;
1874
1875 /* Handle single-char matchers */
1876
1877 case OP_PROP:
1878 case OP_NOTPROP:
1879 cc += 2;
1880 /* Fall through */
1881
1882 case OP_HSPACE:
1883 case OP_VSPACE:
1884 case OP_NOT_HSPACE:
1885 case OP_NOT_VSPACE:
1886 case OP_NOT_DIGIT:
1887 case OP_DIGIT:
1888 case OP_NOT_WHITESPACE:
1889 case OP_WHITESPACE:
1890 case OP_NOT_WORDCHAR:
1891 case OP_WORDCHAR:
1892 case OP_ANY:
1893 case OP_ALLANY:
1894 branchlength++;
1895 cc++;
1896 break;
1897
1898 /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1899 otherwise \C is coded as OP_ALLANY. */
1900
1901 case OP_ANYBYTE:
1902 return -2;
1903
1904 /* Check a class for variable quantification */
1905
1906 case OP_CLASS:
1907 case OP_NCLASS:
1908 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1909 case OP_XCLASS:
1910 /* The original code caused an unsigned overflow in 64 bit systems,
1911 so now we use a conditional statement. */
1912 if (op == OP_XCLASS)
1913 cc += GET(cc, 1);
1914 else
1915 cc += PRIV(OP_lengths)[OP_CLASS];
1916 #else
1917 cc += PRIV(OP_lengths)[OP_CLASS];
1918 #endif
1919
1920 switch (*cc)
1921 {
1922 case OP_CRSTAR:
1923 case OP_CRMINSTAR:
1924 case OP_CRPLUS:
1925 case OP_CRMINPLUS:
1926 case OP_CRQUERY:
1927 case OP_CRMINQUERY:
1928 case OP_CRPOSSTAR:
1929 case OP_CRPOSPLUS:
1930 case OP_CRPOSQUERY:
1931 return -1;
1932
1933 case OP_CRRANGE:
1934 case OP_CRMINRANGE:
1935 case OP_CRPOSRANGE:
1936 if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1937 branchlength += (int)GET2(cc,1);
1938 cc += 1 + 2 * IMM2_SIZE;
1939 break;
1940
1941 default:
1942 branchlength++;
1943 }
1944 break;
1945
1946 /* Anything else is variable length */
1947
1948 case OP_ANYNL:
1949 case OP_BRAMINZERO:
1950 case OP_BRAPOS:
1951 case OP_BRAPOSZERO:
1952 case OP_BRAZERO:
1953 case OP_CBRAPOS:
1954 case OP_EXTUNI:
1955 case OP_KETRMAX:
1956 case OP_KETRMIN:
1957 case OP_KETRPOS:
1958 case OP_MINPLUS:
1959 case OP_MINPLUSI:
1960 case OP_MINQUERY:
1961 case OP_MINQUERYI:
1962 case OP_MINSTAR:
1963 case OP_MINSTARI:
1964 case OP_MINUPTO:
1965 case OP_MINUPTOI:
1966 case OP_NOTMINPLUS:
1967 case OP_NOTMINPLUSI:
1968 case OP_NOTMINQUERY:
1969 case OP_NOTMINQUERYI:
1970 case OP_NOTMINSTAR:
1971 case OP_NOTMINSTARI:
1972 case OP_NOTMINUPTO:
1973 case OP_NOTMINUPTOI:
1974 case OP_NOTPLUS:
1975 case OP_NOTPLUSI:
1976 case OP_NOTPOSPLUS:
1977 case OP_NOTPOSPLUSI:
1978 case OP_NOTPOSQUERY:
1979 case OP_NOTPOSQUERYI:
1980 case OP_NOTPOSSTAR:
1981 case OP_NOTPOSSTARI:
1982 case OP_NOTPOSUPTO:
1983 case OP_NOTPOSUPTOI:
1984 case OP_NOTQUERY:
1985 case OP_NOTQUERYI:
1986 case OP_NOTSTAR:
1987 case OP_NOTSTARI:
1988 case OP_NOTUPTO:
1989 case OP_NOTUPTOI:
1990 case OP_PLUS:
1991 case OP_PLUSI:
1992 case OP_POSPLUS:
1993 case OP_POSPLUSI:
1994 case OP_POSQUERY:
1995 case OP_POSQUERYI:
1996 case OP_POSSTAR:
1997 case OP_POSSTARI:
1998 case OP_POSUPTO:
1999 case OP_POSUPTOI:
2000 case OP_QUERY:
2001 case OP_QUERYI:
2002 case OP_REF:
2003 case OP_REFI:
2004 case OP_DNREF:
2005 case OP_DNREFI:
2006 case OP_SBRA:
2007 case OP_SBRAPOS:
2008 case OP_SCBRA:
2009 case OP_SCBRAPOS:
2010 case OP_SCOND:
2011 case OP_SKIPZERO:
2012 case OP_STAR:
2013 case OP_STARI:
2014 case OP_TYPEMINPLUS:
2015 case OP_TYPEMINQUERY:
2016 case OP_TYPEMINSTAR:
2017 case OP_TYPEMINUPTO:
2018 case OP_TYPEPLUS:
2019 case OP_TYPEPOSPLUS:
2020 case OP_TYPEPOSQUERY:
2021 case OP_TYPEPOSSTAR:
2022 case OP_TYPEPOSUPTO:
2023 case OP_TYPEQUERY:
2024 case OP_TYPESTAR:
2025 case OP_TYPEUPTO:
2026 case OP_UPTO:
2027 case OP_UPTOI:
2028 return -1;
2029
2030 /* Catch unrecognized opcodes so that when new ones are added they
2031 are not forgotten, as has happened in the past. */
2032
2033 default:
2034 return -4;
2035 }
2036 }
2037 /* Control never gets here */
2038 }
2039
2040
2041
2042 /*************************************************
2043 * Scan compiled regex for specific bracket *
2044 *************************************************/
2045
2046 /* This little function scans through a compiled pattern until it finds a
2047 capturing bracket with the given number, or, if the number is negative, an
2048 instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2049 so that it can be called from pcre_study() when finding the minimum matching
2050 length.
2051
2052 Arguments:
2053 code points to start of expression
2054 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2055 number the required bracket number or negative to find a lookbehind
2056
2057 Returns: pointer to the opcode for the bracket, or NULL if not found
2058 */
2059
2060 const pcre_uchar *
2061 PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2062 {
2063 for (;;)
2064 {
2065 register pcre_uchar c = *code;
2066
2067 if (c == OP_END) return NULL;
2068
2069 /* XCLASS is used for classes that cannot be represented just by a bit
2070 map. This includes negated single high-valued characters. The length in
2071 the table is zero; the actual length is stored in the compiled code. */
2072
2073 if (c == OP_XCLASS) code += GET(code, 1);
2074
2075 /* Handle recursion */
2076
2077 else if (c == OP_REVERSE)
2078 {
2079 if (number < 0) return (pcre_uchar *)code;
2080 code += PRIV(OP_lengths)[c];
2081 }
2082
2083 /* Handle capturing bracket */
2084
2085 else if (c == OP_CBRA || c == OP_SCBRA ||
2086 c == OP_CBRAPOS || c == OP_SCBRAPOS)
2087 {
2088 int n = (int)GET2(code, 1+LINK_SIZE);
2089 if (n == number) return (pcre_uchar *)code;
2090 code += PRIV(OP_lengths)[c];
2091 }
2092
2093 /* Otherwise, we can get the item's length from the table, except that for
2094 repeated character types, we have to test for \p and \P, which have an extra
2095 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2096 must add in its length. */
2097
2098 else
2099 {
2100 switch(c)
2101 {
2102 case OP_TYPESTAR:
2103 case OP_TYPEMINSTAR:
2104 case OP_TYPEPLUS:
2105 case OP_TYPEMINPLUS:
2106 case OP_TYPEQUERY:
2107 case OP_TYPEMINQUERY:
2108 case OP_TYPEPOSSTAR:
2109 case OP_TYPEPOSPLUS:
2110 case OP_TYPEPOSQUERY:
2111 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2112 break;
2113
2114 case OP_TYPEUPTO:
2115 case OP_TYPEMINUPTO:
2116 case OP_TYPEEXACT:
2117 case OP_TYPEPOSUPTO:
2118 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2119 code += 2;
2120 break;
2121
2122 case OP_MARK:
2123 case OP_PRUNE_ARG:
2124 case OP_SKIP_ARG:
2125 case OP_THEN_ARG:
2126 code += code[1];
2127 break;
2128 }
2129
2130 /* Add in the fixed length from the table */
2131
2132 code += PRIV(OP_lengths)[c];
2133
2134 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2135 a multi-byte character. The length in the table is a minimum, so we have to
2136 arrange to skip the extra bytes. */
2137
2138 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2139 if (utf) switch(c)
2140 {
2141 case OP_CHAR:
2142 case OP_CHARI:
2143 case OP_NOT:
2144 case OP_NOTI:
2145 case OP_EXACT:
2146 case OP_EXACTI:
2147 case OP_NOTEXACT:
2148 case OP_NOTEXACTI:
2149 case OP_UPTO:
2150 case OP_UPTOI:
2151 case OP_NOTUPTO:
2152 case OP_NOTUPTOI:
2153 case OP_MINUPTO:
2154 case OP_MINUPTOI:
2155 case OP_NOTMINUPTO:
2156 case OP_NOTMINUPTOI:
2157 case OP_POSUPTO:
2158 case OP_POSUPTOI:
2159 case OP_NOTPOSUPTO:
2160 case OP_NOTPOSUPTOI:
2161 case OP_STAR:
2162 case OP_STARI:
2163 case OP_NOTSTAR:
2164 case OP_NOTSTARI:
2165 case OP_MINSTAR:
2166 case OP_MINSTARI:
2167 case OP_NOTMINSTAR:
2168 case OP_NOTMINSTARI:
2169 case OP_POSSTAR:
2170 case OP_POSSTARI:
2171 case OP_NOTPOSSTAR:
2172 case OP_NOTPOSSTARI:
2173 case OP_PLUS:
2174 case OP_PLUSI:
2175 case OP_NOTPLUS:
2176 case OP_NOTPLUSI:
2177 case OP_MINPLUS:
2178 case OP_MINPLUSI:
2179 case OP_NOTMINPLUS:
2180 case OP_NOTMINPLUSI:
2181 case OP_POSPLUS:
2182 case OP_POSPLUSI:
2183 case OP_NOTPOSPLUS:
2184 case OP_NOTPOSPLUSI:
2185 case OP_QUERY:
2186 case OP_QUERYI:
2187 case OP_NOTQUERY:
2188 case OP_NOTQUERYI:
2189 case OP_MINQUERY:
2190 case OP_MINQUERYI:
2191 case OP_NOTMINQUERY:
2192 case OP_NOTMINQUERYI:
2193 case OP_POSQUERY:
2194 case OP_POSQUERYI:
2195 case OP_NOTPOSQUERY:
2196 case OP_NOTPOSQUERYI:
2197 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2198 break;
2199 }
2200 #else
2201 (void)(utf); /* Keep compiler happy by referencing function argument */
2202 #endif
2203 }
2204 }
2205 }
2206
2207
2208
2209 /*************************************************
2210 * Scan compiled regex for recursion reference *
2211 *************************************************/
2212
2213 /* This little function scans through a compiled pattern until it finds an
2214 instance of OP_RECURSE.
2215
2216 Arguments:
2217 code points to start of expression
2218 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2219
2220 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2221 */
2222
2223 static const pcre_uchar *
2224 find_recurse(const pcre_uchar *code, BOOL utf)
2225 {
2226 for (;;)
2227 {
2228 register pcre_uchar c = *code;
2229 if (c == OP_END) return NULL;
2230 if (c == OP_RECURSE) return code;
2231
2232 /* XCLASS is used for classes that cannot be represented just by a bit
2233 map. This includes negated single high-valued characters. The length in
2234 the table is zero; the actual length is stored in the compiled code. */
2235
2236 if (c == OP_XCLASS) code += GET(code, 1);
2237
2238 /* Otherwise, we can get the item's length from the table, except that for
2239 repeated character types, we have to test for \p and \P, which have an extra
2240 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2241 must add in its length. */
2242
2243 else
2244 {
2245 switch(c)
2246 {
2247 case OP_TYPESTAR:
2248 case OP_TYPEMINSTAR:
2249 case OP_TYPEPLUS:
2250 case OP_TYPEMINPLUS:
2251 case OP_TYPEQUERY:
2252 case OP_TYPEMINQUERY:
2253 case OP_TYPEPOSSTAR:
2254 case OP_TYPEPOSPLUS:
2255 case OP_TYPEPOSQUERY:
2256 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2257 break;
2258
2259 case OP_TYPEPOSUPTO:
2260 case OP_TYPEUPTO:
2261 case OP_TYPEMINUPTO:
2262 case OP_TYPEEXACT:
2263 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2264 code += 2;
2265 break;
2266
2267 case OP_MARK:
2268 case OP_PRUNE_ARG:
2269 case OP_SKIP_ARG:
2270 case OP_THEN_ARG:
2271 code += code[1];
2272 break;
2273 }
2274
2275 /* Add in the fixed length from the table */
2276
2277 code += PRIV(OP_lengths)[c];
2278
2279 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2280 by a multi-byte character. The length in the table is a minimum, so we have
2281 to arrange to skip the extra bytes. */
2282
2283 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2284 if (utf) switch(c)
2285 {
2286 case OP_CHAR:
2287 case OP_CHARI:
2288 case OP_NOT:
2289 case OP_NOTI:
2290 case OP_EXACT:
2291 case OP_EXACTI:
2292 case OP_NOTEXACT:
2293 case OP_NOTEXACTI:
2294 case OP_UPTO:
2295 case OP_UPTOI:
2296 case OP_NOTUPTO:
2297 case OP_NOTUPTOI:
2298 case OP_MINUPTO:
2299 case OP_MINUPTOI:
2300 case OP_NOTMINUPTO:
2301 case OP_NOTMINUPTOI:
2302 case OP_POSUPTO:
2303 case OP_POSUPTOI:
2304 case OP_NOTPOSUPTO:
2305 case OP_NOTPOSUPTOI:
2306 case OP_STAR:
2307 case OP_STARI:
2308 case OP_NOTSTAR:
2309 case OP_NOTSTARI:
2310 case OP_MINSTAR:
2311 case OP_MINSTARI:
2312 case OP_NOTMINSTAR:
2313 case OP_NOTMINSTARI:
2314 case OP_POSSTAR:
2315 case OP_POSSTARI:
2316 case OP_NOTPOSSTAR:
2317 case OP_NOTPOSSTARI:
2318 case OP_PLUS:
2319 case OP_PLUSI:
2320 case OP_NOTPLUS:
2321 case OP_NOTPLUSI:
2322 case OP_MINPLUS:
2323 case OP_MINPLUSI:
2324 case OP_NOTMINPLUS:
2325 case OP_NOTMINPLUSI:
2326 case OP_POSPLUS:
2327 case OP_POSPLUSI:
2328 case OP_NOTPOSPLUS:
2329 case OP_NOTPOSPLUSI:
2330 case OP_QUERY:
2331 case OP_QUERYI:
2332 case OP_NOTQUERY:
2333 case OP_NOTQUERYI:
2334 case OP_MINQUERY:
2335 case OP_MINQUERYI:
2336 case OP_NOTMINQUERY:
2337 case OP_NOTMINQUERYI:
2338 case OP_POSQUERY:
2339 case OP_POSQUERYI:
2340 case OP_NOTPOSQUERY:
2341 case OP_NOTPOSQUERYI:
2342 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2343 break;
2344 }
2345 #else
2346 (void)(utf); /* Keep compiler happy by referencing function argument */
2347 #endif
2348 }
2349 }
2350 }
2351
2352
2353
2354 /*************************************************
2355 * Scan compiled branch for non-emptiness *
2356 *************************************************/
2357
2358 /* This function scans through a branch of a compiled pattern to see whether it
2359 can match the empty string or not. It is called from could_be_empty()
2360 below and from compile_branch() when checking for an unlimited repeat of a
2361 group that can match nothing. Note that first_significant_code() skips over
2362 backward and negative forward assertions when its final argument is TRUE. If we
2363 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2364 bracket whose current branch will already have been scanned.
2365
2366 Arguments:
2367 code points to start of search
2368 endcode points to where to stop
2369 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2370 cd contains pointers to tables etc.
2371 recurses chain of recurse_check to catch mutual recursion
2372
2373 Returns: TRUE if what is matched could be empty
2374 */
2375
2376 static BOOL
2377 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2378 BOOL utf, compile_data *cd, recurse_check *recurses)
2379 {
2380 register pcre_uchar c;
2381 recurse_check this_recurse;
2382
2383 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2384 code < endcode;
2385 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2386 {
2387 const pcre_uchar *ccode;
2388
2389 c = *code;
2390
2391 /* Skip over forward assertions; the other assertions are skipped by
2392 first_significant_code() with a TRUE final argument. */
2393
2394 if (c == OP_ASSERT)
2395 {
2396 do code += GET(code, 1); while (*code == OP_ALT);
2397 c = *code;
2398 continue;
2399 }
2400
2401 /* For a recursion/subroutine call, if its end has been reached, which
2402 implies a backward reference subroutine call, we can scan it. If it's a
2403 forward reference subroutine call, we can't. To detect forward reference
2404 we have to scan up the list that is kept in the workspace. This function is
2405 called only when doing the real compile, not during the pre-compile that
2406 measures the size of the compiled pattern. */
2407
2408 if (c == OP_RECURSE)
2409 {
2410 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2411 const pcre_uchar *endgroup = scode;
2412 BOOL empty_branch;
2413
2414 /* Test for forward reference or uncompleted reference. This is disabled
2415 when called to scan a completed pattern by setting cd->start_workspace to
2416 NULL. */
2417
2418 if (cd->start_workspace != NULL)
2419 {
2420 const pcre_uchar *tcode;
2421 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2422 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2423 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2424 }
2425
2426 /* If the reference is to a completed group, we need to detect whether this
2427 is a recursive call, as otherwise there will be an infinite loop. If it is
2428 a recursion, just skip over it. Simple recursions are easily detected. For
2429 mutual recursions we keep a chain on the stack. */
2430
2431 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2432 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2433 else
2434 {
2435 recurse_check *r = recurses;
2436 for (r = recurses; r != NULL; r = r->prev)
2437 if (r->group == scode) break;
2438 if (r != NULL) continue; /* Mutual recursion */
2439 }
2440
2441 /* Completed reference; scan the referenced group, remembering it on the
2442 stack chain to detect mutual recursions. */
2443
2444 empty_branch = FALSE;
2445 this_recurse.prev = recurses;
2446 this_recurse.group = scode;
2447
2448 do
2449 {
2450 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2451 {
2452 empty_branch = TRUE;
2453 break;
2454 }
2455 scode += GET(scode, 1);
2456 }
2457 while (*scode == OP_ALT);
2458
2459 if (!empty_branch) return FALSE; /* All branches are non-empty */
2460 continue;
2461 }
2462
2463 /* Groups with zero repeats can of course be empty; skip them. */
2464
2465 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2466 c == OP_BRAPOSZERO)
2467 {
2468 code += PRIV(OP_lengths)[c];
2469 do code += GET(code, 1); while (*code == OP_ALT);
2470 c = *code;
2471 continue;
2472 }
2473
2474 /* A nested group that is already marked as "could be empty" can just be
2475 skipped. */
2476
2477 if (c == OP_SBRA || c == OP_SBRAPOS ||
2478 c == OP_SCBRA || c == OP_SCBRAPOS)
2479 {
2480 do code += GET(code, 1); while (*code == OP_ALT);
2481 c = *code;
2482 continue;
2483 }
2484
2485 /* For other groups, scan the branches. */
2486
2487 if (c == OP_BRA || c == OP_BRAPOS ||
2488 c == OP_CBRA || c == OP_CBRAPOS ||
2489 c == OP_ONCE || c == OP_ONCE_NC ||
2490 c == OP_COND)
2491 {
2492 BOOL empty_branch;
2493 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2494
2495 /* If a conditional group has only one branch, there is a second, implied,
2496 empty branch, so just skip over the conditional, because it could be empty.
2497 Otherwise, scan the individual branches of the group. */
2498
2499 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2500 code += GET(code, 1);
2501 else
2502 {
2503 empty_branch = FALSE;
2504 do
2505 {
2506 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd,
2507 recurses)) empty_branch = TRUE;
2508 code += GET(code, 1);
2509 }
2510 while (*code == OP_ALT);
2511 if (!empty_branch) return FALSE; /* All branches are non-empty */
2512 }
2513
2514 c = *code;
2515 continue;
2516 }
2517
2518 /* Handle the other opcodes */
2519
2520 switch (c)
2521 {
2522 /* Check for quantifiers after a class. XCLASS is used for classes that
2523 cannot be represented just by a bit map. This includes negated single
2524 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2525 actual length is stored in the compiled code, so we must update "code"
2526 here. */
2527
2528 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2529 case OP_XCLASS:
2530 ccode = code += GET(code, 1);
2531 goto CHECK_CLASS_REPEAT;
2532 #endif
2533
2534 case OP_CLASS:
2535 case OP_NCLASS:
2536 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2537
2538 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2539 CHECK_CLASS_REPEAT:
2540 #endif
2541
2542 switch (*ccode)
2543 {
2544 case OP_CRSTAR: /* These could be empty; continue */
2545 case OP_CRMINSTAR:
2546 case OP_CRQUERY:
2547 case OP_CRMINQUERY:
2548 case OP_CRPOSSTAR:
2549 case OP_CRPOSQUERY:
2550 break;
2551
2552 default: /* Non-repeat => class must match */
2553 case OP_CRPLUS: /* These repeats aren't empty */
2554 case OP_CRMINPLUS:
2555 case OP_CRPOSPLUS:
2556 return FALSE;
2557
2558 case OP_CRRANGE:
2559 case OP_CRMINRANGE:
2560 case OP_CRPOSRANGE:
2561 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2562 break;
2563 }
2564 break;
2565
2566 /* Opcodes that must match a character */
2567
2568 case OP_ANY:
2569 case OP_ALLANY:
2570 case OP_ANYBYTE:
2571
2572 case OP_PROP:
2573 case OP_NOTPROP:
2574 case OP_ANYNL:
2575
2576 case OP_NOT_HSPACE:
2577 case OP_HSPACE:
2578 case OP_NOT_VSPACE:
2579 case OP_VSPACE:
2580 case OP_EXTUNI:
2581
2582 case OP_NOT_DIGIT:
2583 case OP_DIGIT:
2584 case OP_NOT_WHITESPACE:
2585 case OP_WHITESPACE:
2586 case OP_NOT_WORDCHAR:
2587 case OP_WORDCHAR:
2588
2589 case OP_CHAR:
2590 case OP_CHARI:
2591 case OP_NOT:
2592 case OP_NOTI:
2593
2594 case OP_PLUS:
2595 case OP_PLUSI:
2596 case OP_MINPLUS:
2597 case OP_MINPLUSI:
2598
2599 case OP_NOTPLUS:
2600 case OP_NOTPLUSI:
2601 case OP_NOTMINPLUS:
2602 case OP_NOTMINPLUSI:
2603
2604 case OP_POSPLUS:
2605 case OP_POSPLUSI:
2606 case OP_NOTPOSPLUS:
2607 case OP_NOTPOSPLUSI:
2608
2609 case OP_EXACT:
2610 case OP_EXACTI:
2611 case OP_NOTEXACT:
2612 case OP_NOTEXACTI:
2613
2614 case OP_TYPEPLUS:
2615 case OP_TYPEMINPLUS:
2616 case OP_TYPEPOSPLUS:
2617 case OP_TYPEEXACT:
2618
2619 return FALSE;
2620
2621 /* These are going to continue, as they may be empty, but we have to
2622 fudge the length for the \p and \P cases. */
2623
2624 case OP_TYPESTAR:
2625 case OP_TYPEMINSTAR:
2626 case OP_TYPEPOSSTAR:
2627 case OP_TYPEQUERY:
2628 case OP_TYPEMINQUERY:
2629 case OP_TYPEPOSQUERY:
2630 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2631 break;
2632
2633 /* Same for these */
2634
2635 case OP_TYPEUPTO:
2636 case OP_TYPEMINUPTO:
2637 case OP_TYPEPOSUPTO:
2638 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2639 code += 2;
2640 break;
2641
2642 /* End of branch */
2643
2644 case OP_KET:
2645 case OP_KETRMAX:
2646 case OP_KETRMIN:
2647 case OP_KETRPOS:
2648 case OP_ALT:
2649 return TRUE;
2650
2651 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2652 MINUPTO, and POSUPTO and their caseless and negative versions may be
2653 followed by a multibyte character. */
2654
2655 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2656 case OP_STAR:
2657 case OP_STARI:
2658 case OP_NOTSTAR:
2659 case OP_NOTSTARI:
2660
2661 case OP_MINSTAR:
2662 case OP_MINSTARI:
2663 case OP_NOTMINSTAR:
2664 case OP_NOTMINSTARI:
2665
2666 case OP_POSSTAR:
2667 case OP_POSSTARI:
2668 case OP_NOTPOSSTAR:
2669 case OP_NOTPOSSTARI:
2670
2671 case OP_QUERY:
2672 case OP_QUERYI:
2673 case OP_NOTQUERY:
2674 case OP_NOTQUERYI:
2675
2676 case OP_MINQUERY:
2677 case OP_MINQUERYI:
2678 case OP_NOTMINQUERY:
2679 case OP_NOTMINQUERYI:
2680
2681 case OP_POSQUERY:
2682 case OP_POSQUERYI:
2683 case OP_NOTPOSQUERY:
2684 case OP_NOTPOSQUERYI:
2685
2686 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2687 break;
2688
2689 case OP_UPTO:
2690 case OP_UPTOI:
2691 case OP_NOTUPTO:
2692 case OP_NOTUPTOI:
2693
2694 case OP_MINUPTO:
2695 case OP_MINUPTOI:
2696 case OP_NOTMINUPTO:
2697 case OP_NOTMINUPTOI:
2698
2699 case OP_POSUPTO:
2700 case OP_POSUPTOI:
2701 case OP_NOTPOSUPTO:
2702 case OP_NOTPOSUPTOI:
2703
2704 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2705 break;
2706 #endif
2707
2708 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2709 string. */
2710
2711 case OP_MARK:
2712 case OP_PRUNE_ARG:
2713 case OP_SKIP_ARG:
2714 case OP_THEN_ARG:
2715 code += code[1];
2716 break;
2717
2718 /* None of the remaining opcodes are required to match a character. */
2719
2720 default:
2721 break;
2722 }
2723 }
2724
2725 return TRUE;
2726 }
2727
2728
2729
2730 /*************************************************
2731 * Scan compiled regex for non-emptiness *
2732 *************************************************/
2733
2734 /* This function is called to check for left recursive calls. We want to check
2735 the current branch of the current pattern to see if it could match the empty
2736 string. If it could, we must look outwards for branches at other levels,
2737 stopping when we pass beyond the bracket which is the subject of the recursion.
2738 This function is called only during the real compile, not during the
2739 pre-compile.
2740
2741 Arguments:
2742 code points to start of the recursion
2743 endcode points to where to stop (current RECURSE item)
2744 bcptr points to the chain of current (unclosed) branch starts
2745 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2746 cd pointers to tables etc
2747
2748 Returns: TRUE if what is matched could be empty
2749 */
2750
2751 static BOOL
2752 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2753 branch_chain *bcptr, BOOL utf, compile_data *cd)
2754 {
2755 while (bcptr != NULL && bcptr->current_branch >= code)
2756 {
2757 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2758 return FALSE;
2759 bcptr = bcptr->outer;
2760 }
2761 return TRUE;
2762 }
2763
2764
2765
2766 /*************************************************
2767 * Base opcode of repeated opcodes *
2768 *************************************************/
2769
2770 /* Returns the base opcode for repeated single character type opcodes. If the
2771 opcode is not a repeated character type, it returns with the original value.
2772
2773 Arguments: c opcode
2774 Returns: base opcode for the type
2775 */
2776
2777 static pcre_uchar
2778 get_repeat_base(pcre_uchar c)
2779 {
2780 return (c > OP_TYPEPOSUPTO)? c :
2781 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2782 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2783 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2784 (c >= OP_STARI)? OP_STARI :
2785 OP_STAR;
2786 }
2787
2788
2789
2790 #ifdef SUPPORT_UCP
2791 /*************************************************
2792 * Check a character and a property *
2793 *************************************************/
2794
2795 /* This function is called by check_auto_possessive() when a property item
2796 is adjacent to a fixed character.
2797
2798 Arguments:
2799 c the character
2800 ptype the property type
2801 pdata the data for the type
2802 negated TRUE if it's a negated property (\P or \p{^)
2803
2804 Returns: TRUE if auto-possessifying is OK
2805 */
2806
2807 static BOOL
2808 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2809 BOOL negated)
2810 {
2811 const pcre_uint32 *p;
2812 const ucd_record *prop = GET_UCD(c);
2813
2814 switch(ptype)
2815 {
2816 case PT_LAMP:
2817 return (prop->chartype == ucp_Lu ||
2818 prop->chartype == ucp_Ll ||
2819 prop->chartype == ucp_Lt) == negated;
2820
2821 case PT_GC:
2822 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2823
2824 case PT_PC:
2825 return (pdata == prop->chartype) == negated;
2826
2827 case PT_SC:
2828 return (pdata == prop->script) == negated;
2829
2830 /* These are specials */
2831
2832 case PT_ALNUM:
2833 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2834 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2835
2836 /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2837 means that Perl space and POSIX space are now identical. PCRE was changed
2838 at release 8.34. */
2839
2840 case PT_SPACE: /* Perl space */
2841 case PT_PXSPACE: /* POSIX space */
2842 switch(c)
2843 {
2844 HSPACE_CASES:
2845 VSPACE_CASES:
2846 return negated;
2847
2848 default:
2849 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2850 }
2851 break; /* Control never reaches here */
2852
2853 case PT_WORD:
2854 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2855 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2856 c == CHAR_UNDERSCORE) == negated;
2857
2858 case PT_CLIST:
2859 p = PRIV(ucd_caseless_sets) + prop->caseset;
2860 for (;;)
2861 {
2862 if (c < *p) return !negated;
2863 if (c == *p++) return negated;
2864 }
2865 break; /* Control never reaches here */
2866 }
2867
2868 return FALSE;
2869 }
2870 #endif /* SUPPORT_UCP */
2871
2872
2873
2874 /*************************************************
2875 * Fill the character property list *
2876 *************************************************/
2877
2878 /* Checks whether the code points to an opcode that can take part in auto-
2879 possessification, and if so, fills a list with its properties.
2880
2881 Arguments:
2882 code points to start of expression
2883 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2884 fcc points to case-flipping table
2885 list points to output list
2886 list[0] will be filled with the opcode
2887 list[1] will be non-zero if this opcode
2888 can match an empty character string
2889 list[2..7] depends on the opcode
2890
2891 Returns: points to the start of the next opcode if *code is accepted
2892 NULL if *code is not accepted
2893 */
2894
2895 static const pcre_uchar *
2896 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2897 const pcre_uint8 *fcc, pcre_uint32 *list)
2898 {
2899 pcre_uchar c = *code;
2900 pcre_uchar base;
2901 const pcre_uchar *end;
2902 pcre_uint32 chr;
2903
2904 #ifdef SUPPORT_UCP
2905 pcre_uint32 *clist_dest;
2906 const pcre_uint32 *clist_src;
2907 #else
2908 utf = utf; /* Suppress "unused parameter" compiler warning */
2909 #endif
2910
2911 list[0] = c;
2912 list[1] = FALSE;
2913 code++;
2914
2915 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2916 {
2917 base = get_repeat_base(c);
2918 c -= (base - OP_STAR);
2919
2920 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2921 code += IMM2_SIZE;
2922
2923 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2924
2925 switch(base)
2926 {
2927 case OP_STAR:
2928 list[0] = OP_CHAR;
2929 break;
2930
2931 case OP_STARI:
2932 list[0] = OP_CHARI;
2933 break;
2934
2935 case OP_NOTSTAR:
2936 list[0] = OP_NOT;
2937 break;
2938
2939 case OP_NOTSTARI:
2940 list[0] = OP_NOTI;
2941 break;
2942
2943 case OP_TYPESTAR:
2944 list[0] = *code;
2945 code++;
2946 break;
2947 }
2948 c = list[0];
2949 }
2950
2951 switch(c)
2952 {
2953 case OP_NOT_DIGIT:
2954 case OP_DIGIT:
2955 case OP_NOT_WHITESPACE:
2956 case OP_WHITESPACE:
2957 case OP_NOT_WORDCHAR:
2958 case OP_WORDCHAR:
2959 case OP_ANY:
2960 case OP_ALLANY:
2961 case OP_ANYNL:
2962 case OP_NOT_HSPACE:
2963 case OP_HSPACE:
2964 case OP_NOT_VSPACE:
2965 case OP_VSPACE:
2966 case OP_EXTUNI:
2967 case OP_EODN:
2968 case OP_EOD:
2969 case OP_DOLL:
2970 case OP_DOLLM:
2971 return code;
2972
2973 case OP_CHAR:
2974 case OP_NOT:
2975 GETCHARINCTEST(chr, code);
2976 list[2] = chr;
2977 list[3] = NOTACHAR;
2978 return code;
2979
2980 case OP_CHARI:
2981 case OP_NOTI:
2982 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2983 GETCHARINCTEST(chr, code);
2984 list[2] = chr;
2985
2986 #ifdef SUPPORT_UCP
2987 if (chr < 128 || (chr < 256 && !utf))
2988 list[3] = fcc[chr];
2989 else
2990 list[3] = UCD_OTHERCASE(chr);
2991 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2992 list[3] = (chr < 256) ? fcc[chr] : chr;
2993 #else
2994 list[3] = fcc[chr];
2995 #endif
2996
2997 /* The othercase might be the same value. */
2998
2999 if (chr == list[3])
3000 list[3] = NOTACHAR;
3001 else
3002 list[4] = NOTACHAR;
3003 return code;
3004
3005 #ifdef SUPPORT_UCP
3006 case OP_PROP:
3007 case OP_NOTPROP:
3008 if (code[0] != PT_CLIST)
3009 {
3010 list[2] = code[0];
3011 list[3] = code[1];
3012 return code + 2;
3013 }
3014
3015 /* Convert only if we have enough space. */
3016
3017 clist_src = PRIV(ucd_caseless_sets) + code[1];
3018 clist_dest = list + 2;
3019 code += 2;
3020
3021 do {
3022 if (clist_dest >= list + 8)
3023 {
3024 /* Early return if there is not enough space. This should never
3025 happen, since all clists are shorter than 5 character now. */
3026 list[2] = code[0];
3027 list[3] = code[1];
3028 return code;
3029 }
3030 *clist_dest++ = *clist_src;
3031 }
3032 while(*clist_src++ != NOTACHAR);
3033
3034 /* All characters are stored. The terminating NOTACHAR
3035 is copied form the clist itself. */
3036
3037 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3038 return code;
3039 #endif
3040
3041 case OP_NCLASS:
3042 case OP_CLASS:
3043 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3044 case OP_XCLASS:
3045 if (c == OP_XCLASS)
3046 end = code + GET(code, 0) - 1;
3047 else
3048 #endif
3049 end = code + 32 / sizeof(pcre_uchar);
3050
3051 switch(*end)
3052 {
3053 case OP_CRSTAR:
3054 case OP_CRMINSTAR:
3055 case OP_CRQUERY:
3056 case OP_CRMINQUERY:
3057 case OP_CRPOSSTAR:
3058 case OP_CRPOSQUERY:
3059 list[1] = TRUE;
3060 end++;
3061 break;
3062
3063 case OP_CRPLUS:
3064 case OP_CRMINPLUS:
3065 case OP_CRPOSPLUS:
3066 end++;
3067 break;
3068
3069 case OP_CRRANGE:
3070 case OP_CRMINRANGE:
3071 case OP_CRPOSRANGE:
3072 list[1] = (GET2(end, 1) == 0);
3073 end += 1 + 2 * IMM2_SIZE;
3074 break;
3075 }
3076 list[2] = (pcre_uint32)(end - code);
3077 return end;
3078 }
3079 return NULL; /* Opcode not accepted */
3080 }
3081
3082
3083
3084 /*************************************************
3085 * Scan further character sets for match *
3086 *************************************************/
3087
3088 /* Checks whether the base and the current opcode have a common character, in
3089 which case the base cannot be possessified.
3090
3091 Arguments:
3092 code points to the byte code
3093 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3094 cd static compile data
3095 base_list the data list of the base opcode
3096
3097 Returns: TRUE if the auto-possessification is possible
3098 */
3099
3100 static BOOL
3101 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3102 const pcre_uint32 *base_list, const pcre_uchar *base_end, int *rec_limit)
3103 {
3104 pcre_uchar c;
3105 pcre_uint32 list[8];
3106 const pcre_uint32 *chr_ptr;
3107 const pcre_uint32 *ochr_ptr;
3108 const pcre_uint32 *list_ptr;
3109 const pcre_uchar *next_code;
3110 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3111 const pcre_uchar *xclass_flags;
3112 #endif
3113 const pcre_uint8 *class_bitset;
3114 const pcre_uint8 *set1, *set2, *set_end;
3115 pcre_uint32 chr;
3116 BOOL accepted, invert_bits;
3117 BOOL entered_a_group = FALSE;
3118
3119 if (*rec_limit == 0) return FALSE;
3120 --(*rec_limit);
3121
3122 /* Note: the base_list[1] contains whether the current opcode has greedy
3123 (represented by a non-zero value) quantifier. This is a different from
3124 other character type lists, which stores here that the character iterator
3125 matches to an empty string (also represented by a non-zero value). */
3126
3127 for(;;)
3128 {
3129 /* All operations move the code pointer forward.
3130 Therefore infinite recursions are not possible. */
3131
3132 c = *code;
3133
3134 /* Skip over callouts */
3135
3136 if (c == OP_CALLOUT)
3137 {
3138 code += PRIV(OP_lengths)[c];
3139 continue;
3140 }
3141
3142 if (c == OP_ALT)
3143 {
3144 do code += GET(code, 1); while (*code == OP_ALT);
3145 c = *code;
3146 }
3147
3148 switch(c)
3149 {
3150 case OP_END:
3151 case OP_KETRPOS:
3152 /* TRUE only in greedy case. The non-greedy case could be replaced by
3153 an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3154 uses more memory, which we cannot get at this stage.) */
3155
3156 return base_list[1] != 0;
3157
3158 case OP_KET:
3159 /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3160 it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3161 cannot be converted to a possessive form. */
3162
3163 if (base_list[1] == 0) return FALSE;
3164
3165 switch(*(code - GET(code, 1)))
3166 {
3167 case OP_ASSERT:
3168 case OP_ASSERT_NOT:
3169 case OP_ASSERTBACK:
3170 case OP_ASSERTBACK_NOT:
3171 case OP_ONCE:
3172 case OP_ONCE_NC:
3173 /* Atomic sub-patterns and assertions can always auto-possessify their
3174 last iterator. However, if the group was entered as a result of checking
3175 a previous iterator, this is not possible. */
3176
3177 return !entered_a_group;
3178 }
3179
3180 code += PRIV(OP_lengths)[c];
3181 continue;
3182
3183 case OP_ONCE:
3184 case OP_ONCE_NC:
3185 case OP_BRA:
3186 case OP_CBRA:
3187 next_code = code + GET(code, 1);
3188 code += PRIV(OP_lengths)[c];
3189
3190 while (*next_code == OP_ALT)
3191 {
3192 if (!compare_opcodes(code, utf, cd, base_list, base_end, rec_limit))
3193 return FALSE;
3194 code = next_code + 1 + LINK_SIZE;
3195 next_code += GET(next_code, 1);
3196 }
3197
3198 entered_a_group = TRUE;
3199 continue;
3200
3201 case OP_BRAZERO:
3202 case OP_BRAMINZERO:
3203
3204 next_code = code + 1;
3205 if (*next_code != OP_BRA && *next_code != OP_CBRA
3206 && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3207
3208 do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3209
3210 /* The bracket content will be checked by the
3211 OP_BRA/OP_CBRA case above. */
3212 next_code += 1 + LINK_SIZE;
3213 if (!compare_opcodes(next_code, utf, cd, base_list, base_end, rec_limit))
3214 return FALSE;
3215
3216 code += PRIV(OP_lengths)[c];
3217 continue;
3218
3219 default:
3220 break;
3221 }
3222
3223 /* Check for a supported opcode, and load its properties. */
3224
3225 code = get_chr_property_list(code, utf, cd->fcc, list);
3226 if (code == NULL) return FALSE; /* Unsupported */
3227
3228 /* If either opcode is a small character list, set pointers for comparing
3229 characters from that list with another list, or with a property. */
3230
3231 if (base_list[0] == OP_CHAR)
3232 {
3233 chr_ptr = base_list + 2;
3234 list_ptr = list;
3235 }
3236 else if (list[0] == OP_CHAR)
3237 {
3238 chr_ptr = list + 2;
3239 list_ptr = base_list;
3240 }
3241
3242 /* Character bitsets can also be compared to certain opcodes. */
3243
3244 else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3245 #ifdef COMPILE_PCRE8
3246 /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3247 || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3248 #endif
3249 )
3250 {
3251 #ifdef COMPILE_PCRE8
3252 if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3253 #else
3254 if (base_list[0] == OP_CLASS)
3255 #endif
3256 {
3257 set1 = (pcre_uint8 *)(base_end - base_list[2]);
3258 list_ptr = list;
3259 }
3260 else
3261 {
3262 set1 = (pcre_uint8 *)(code - list[2]);
3263 list_ptr = base_list;
3264 }
3265
3266 invert_bits = FALSE;
3267 switch(list_ptr[0])
3268 {
3269 case OP_CLASS:
3270 case OP_NCLASS:
3271 set2 = (pcre_uint8 *)
3272 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3273 break;
3274
3275 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3276 case OP_XCLASS:
3277 xclass_flags = (list_ptr == list ? code : base_end) - list_ptr[2] + LINK_SIZE;
3278 if ((*xclass_flags & XCL_HASPROP) != 0) return FALSE;
3279 if ((*xclass_flags & XCL_MAP) == 0)
3280 {
3281 /* No bits are set for characters < 256. */
3282 if (list[1] == 0) return TRUE;
3283 /* Might be an empty repeat. */
3284 continue;
3285 }
3286 set2 = (pcre_uint8 *)(xclass_flags + 1);
3287 break;
3288 #endif
3289
3290 case OP_NOT_DIGIT:
3291 invert_bits = TRUE;
3292 /* Fall through */
3293 case OP_DIGIT:
3294 set2 = (pcre_uint8 *)(cd->cbits + cbit_digit);
3295 break;
3296
3297 case OP_NOT_WHITESPACE:
3298 invert_bits = TRUE;
3299 /* Fall through */
3300 case OP_WHITESPACE:
3301 set2 = (pcre_uint8 *)(cd->cbits + cbit_space);
3302 break;
3303
3304 case OP_NOT_WORDCHAR:
3305 invert_bits = TRUE;
3306 /* Fall through */
3307 case OP_WORDCHAR:
3308 set2 = (pcre_uint8 *)(cd->cbits + cbit_word);
3309 break;
3310
3311 default:
3312 return FALSE;
3313 }
3314
3315 /* Because the sets are unaligned, we need
3316 to perform byte comparison here. */
3317 set_end = set1 + 32;
3318 if (invert_bits)
3319 {
3320 do
3321 {
3322 if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3323 }
3324 while (set1 < set_end);
3325 }
3326 else
3327 {
3328 do
3329 {
3330 if ((*set1++ & *set2++) != 0) return FALSE;
3331 }
3332 while (set1 < set_end);
3333 }
3334
3335 if (list[1] == 0) return TRUE;
3336 /* Might be an empty repeat. */
3337 continue;
3338 }
3339
3340 /* Some property combinations also acceptable. Unicode property opcodes are
3341 processed specially; the rest can be handled with a lookup table. */
3342
3343 else
3344 {
3345 pcre_uint32 leftop, rightop;
3346
3347 leftop = base_list[0];
3348 rightop = list[0];
3349
3350 #ifdef SUPPORT_UCP
3351 accepted = FALSE; /* Always set in non-unicode case. */
3352 if (leftop == OP_PROP || leftop == OP_NOTPROP)
3353 {
3354 if (rightop == OP_EOD)
3355 accepted = TRUE;
3356 else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3357 {
3358 int n;
3359 const pcre_uint8 *p;
3360 BOOL same = leftop == rightop;
3361 BOOL lisprop = leftop == OP_PROP;
3362 BOOL risprop = rightop == OP_PROP;
3363 BOOL bothprop = lisprop && risprop;
3364
3365 /* There's a table that specifies how each combination is to be
3366 processed:
3367 0 Always return FALSE (never auto-possessify)
3368 1 Character groups are distinct (possessify if both are OP_PROP)
3369 2 Check character categories in the same group (general or particular)
3370 3 Return TRUE if the two opcodes are not the same
3371 ... see comments below
3372 */
3373
3374 n = propposstab[base_list[2]][list[2]];
3375 switch(n)
3376 {
3377 case 0: break;
3378 case 1: accepted = bothprop; break;
3379 case 2: accepted = (base_list[3] == list[3]) != same; break;
3380 case 3: accepted = !same; break;
3381
3382 case 4: /* Left general category, right particular category */
3383 accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3384 break;
3385
3386 case 5: /* Right general category, left particular category */
3387 accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3388 break;
3389
3390 /* This code is logically tricky. Think hard before fiddling with it.
3391 The posspropstab table has four entries per row. Each row relates to
3392 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3393 Only WORD actually needs all four entries, but using repeats for the
3394 others means they can all use the same code below.
3395
3396 The first two entries in each row are Unicode general categories, and
3397 apply always, because all the characters they include are part of the
3398 PCRE character set. The third and fourth entries are a general and a
3399 particular category, respectively, that include one or more relevant
3400 characters. One or the other is used, depending on whether the check
3401 is for a general or a particular category. However, in both cases the
3402 category contains more characters than the specials that are defined
3403 for the property being tested against. Therefore, it cannot be used
3404 in a NOTPROP case.
3405
3406 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3407 Underscore is covered by ucp_P or ucp_Po. */
3408
3409 case 6: /* Left alphanum vs right general category */
3410 case 7: /* Left space vs right general category */
3411 case 8: /* Left word vs right general category */
3412 p = posspropstab[n-6];
3413 accepted = risprop && lisprop ==
3414 (list[3] != p[0] &&
3415 list[3] != p[1] &&
3416 (list[3] != p[2] || !lisprop));
3417 break;
3418
3419 case 9: /* Right alphanum vs left general category */
3420 case 10: /* Right space vs left general category */
3421 case 11: /* Right word vs left general category */
3422 p = posspropstab[n-9];
3423 accepted = lisprop && risprop ==
3424 (base_list[3] != p[0] &&
3425 base_list[3] != p[1] &&
3426 (base_list[3] != p[2] || !risprop));
3427 break;
3428
3429 case 12: /* Left alphanum vs right particular category */
3430 case 13: /* Left space vs right particular category */
3431 case 14: /* Left word vs right particular category */
3432 p = posspropstab[n-12];
3433 accepted = risprop && lisprop ==
3434 (catposstab[p[0]][list[3]] &&
3435 catposstab[p[1]][list[3]] &&
3436 (list[3] != p[3] || !lisprop));
3437 break;
3438
3439 case 15: /* Right alphanum vs left particular category */
3440 case 16: /* Right space vs left particular category */
3441 case 17: /* Right word vs left particular category */
3442 p = posspropstab[n-15];
3443 accepted = lisprop && risprop ==
3444 (catposstab[p[0]][base_list[3]] &&
3445 catposstab[p[1]][base_list[3]] &&
3446 (base_list[3] != p[3] || !risprop));
3447 break;
3448 }
3449 }
3450 }
3451
3452 else
3453 #endif /* SUPPORT_UCP */
3454
3455 accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3456 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3457 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3458
3459 if (!accepted) return FALSE;
3460
3461 if (list[1] == 0) return TRUE;
3462 /* Might be an empty repeat. */
3463 continue;
3464 }
3465
3466 /* Control reaches here only if one of the items is a small character list.
3467 All characters are checked against the other side. */
3468
3469 do
3470 {
3471 chr = *chr_ptr;
3472
3473 switch(list_ptr[0])
3474 {
3475 case OP_CHAR:
3476 ochr_ptr = list_ptr + 2;
3477 do
3478 {
3479 if (chr == *ochr_ptr) return FALSE;
3480 ochr_ptr++;
3481 }
3482 while(*ochr_ptr != NOTACHAR);
3483 break;
3484
3485 case OP_NOT:
3486 ochr_ptr = list_ptr + 2;
3487 do
3488 {
3489 if (chr == *ochr_ptr)
3490 break;
3491 ochr_ptr++;
3492 }
3493 while(*ochr_ptr != NOTACHAR);
3494 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3495 break;
3496
3497 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3498 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3499
3500 case OP_DIGIT:
3501 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3502 break;
3503
3504 case OP_NOT_DIGIT:
3505 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3506 break;
3507
3508 case OP_WHITESPACE:
3509 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3510 break;
3511
3512 case OP_NOT_WHITESPACE:
3513 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3514 break;
3515
3516 case OP_WORDCHAR:
3517 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3518 break;
3519
3520 case OP_NOT_WORDCHAR:
3521 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3522 break;
3523
3524 case OP_HSPACE:
3525 switch(chr)
3526 {
3527 HSPACE_CASES: return FALSE;
3528 default: break;
3529 }
3530 break;
3531
3532 case OP_NOT_HSPACE:
3533 switch(chr)
3534 {
3535 HSPACE_CASES: break;
3536 default: return FALSE;
3537 }
3538 break;
3539
3540 case OP_ANYNL:
3541 case OP_VSPACE:
3542 switch(chr)
3543 {
3544 VSPACE_CASES: return FALSE;
3545 default: break;
3546 }
3547 break;
3548
3549 case OP_NOT_VSPACE:
3550 switch(chr)
3551 {
3552 VSPACE_CASES: break;
3553 default: return FALSE;
3554 }
3555 break;
3556
3557 case OP_DOLL:
3558 case OP_EODN:
3559 switch (chr)
3560 {
3561 case CHAR_CR:
3562 case CHAR_LF:
3563 case CHAR_VT:
3564 case CHAR_FF:
3565 case CHAR_NEL:
3566 #ifndef EBCDIC
3567 case 0x2028:
3568 case 0x2029:
3569 #endif /* Not EBCDIC */
3570 return FALSE;
3571 }
3572 break;
3573
3574 case OP_EOD: /* Can always possessify before \z */
3575 break;
3576
3577 #ifdef SUPPORT_UCP
3578 case OP_PROP:
3579 case OP_NOTPROP:
3580 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3581 list_ptr[0] == OP_NOTPROP))
3582 return FALSE;
3583 break;
3584 #endif
3585
3586 case OP_NCLASS:
3587 if (chr > 255) return FALSE;
3588 /* Fall through */
3589
3590 case OP_CLASS:
3591 if (chr > 255) break;
3592 class_bitset = (pcre_uint8 *)
3593 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3594 if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3595 break;
3596
3597 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3598 case OP_XCLASS:
3599 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3600 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3601 break;
3602 #endif
3603
3604 default:
3605 return FALSE;
3606 }
3607
3608 chr_ptr++;
3609 }
3610 while(*chr_ptr != NOTACHAR);
3611
3612 /* At least one character must be matched from this opcode. */
3613
3614 if (list[1] == 0) return TRUE;
3615 }
3616
3617 /* Control never reaches here. There used to be a fail-save return FALSE; here,
3618 but some compilers complain about an unreachable statement. */
3619
3620 }
3621
3622
3623
3624 /*************************************************
3625 * Scan compiled regex for auto-possession *
3626 *************************************************/
3627
3628 /* Replaces single character iterations with their possessive alternatives
3629 if appropriate. This function modifies the compiled opcode!
3630
3631 Arguments:
3632 code points to start of the byte code
3633 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3634 cd static compile data
3635
3636 Returns: nothing
3637 */
3638
3639 static void
3640 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3641 {
3642 register pcre_uchar c;
3643 const pcre_uchar *end;
3644 pcre_uchar *repeat_opcode;
3645 pcre_uint32 list[8];
3646 int rec_limit;
3647
3648 for (;;)
3649 {
3650 c = *code;
3651
3652 /* When a pattern with bad UTF-8 encoding is compiled with NO_UTF_CHECK,
3653 it may compile without complaining, but may get into a loop here if the code
3654 pointer points to a bad value. This is, of course a documentated possibility,
3655 when NO_UTF_CHECK is set, so it isn't a bug, but we can detect this case and
3656 just give up on this optimization. */
3657
3658 if (c >= OP_TABLE_LENGTH) return;
3659
3660 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3661 {
3662 c -= get_repeat_base(c) - OP_STAR;
3663 end = (c <= OP_MINUPTO) ?
3664 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3665 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3666
3667 rec_limit = 1000;
3668 if (end != NULL && compare_opcodes(end, utf, cd, list, end, &rec_limit))
3669 {
3670 switch(c)
3671 {
3672 case OP_STAR:
3673 *code += OP_POSSTAR - OP_STAR;
3674 break;
3675
3676 case OP_MINSTAR:
3677 *code += OP_POSSTAR - OP_MINSTAR;
3678 break;
3679
3680 case OP_PLUS:
3681 *code += OP_POSPLUS - OP_PLUS;
3682 break;
3683
3684 case OP_MINPLUS:
3685 *code += OP_POSPLUS - OP_MINPLUS;
3686 break;
3687
3688 case OP_QUERY:
3689 *code += OP_POSQUERY - OP_QUERY;
3690 break;
3691
3692 case OP_MINQUERY:
3693 *code += OP_POSQUERY - OP_MINQUERY;
3694 break;
3695
3696 case OP_UPTO:
3697 *code += OP_POSUPTO - OP_UPTO;
3698 break;
3699
3700 case OP_MINUPTO:
3701 *code += OP_POSUPTO - OP_MINUPTO;
3702 break;
3703 }
3704 }
3705 c = *code;
3706 }
3707 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3708 {
3709 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3710 if (c == OP_XCLASS)
3711 repeat_opcode = code + GET(code, 1);
3712 else
3713 #endif
3714 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3715
3716 c = *repeat_opcode;
3717 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3718 {
3719 /* end must not be NULL. */
3720 end = get_chr_property_list(code, utf, cd->fcc, list);
3721
3722 list[1] = (c & 1) == 0;
3723
3724 rec_limit = 1000;
3725 if (compare_opcodes(end, utf, cd, list, end, &rec_limit))
3726 {
3727 switch (c)
3728 {
3729 case OP_CRSTAR:
3730 case OP_CRMINSTAR:
3731 *repeat_opcode = OP_CRPOSSTAR;
3732 break;
3733
3734 case OP_CRPLUS:
3735 case OP_CRMINPLUS:
3736 *repeat_opcode = OP_CRPOSPLUS;
3737 break;
3738
3739 case OP_CRQUERY:
3740 case OP_CRMINQUERY:
3741 *repeat_opcode = OP_CRPOSQUERY;
3742 break;
3743
3744 case OP_CRRANGE:
3745 case OP_CRMINRANGE:
3746 *repeat_opcode = OP_CRPOSRANGE;
3747 break;
3748 }
3749 }
3750 }
3751 c = *code;
3752 }
3753
3754 switch(c)
3755 {
3756 case OP_END:
3757 return;
3758
3759 case OP_TYPESTAR:
3760 case OP_TYPEMINSTAR:
3761 case OP_TYPEPLUS:
3762 case OP_TYPEMINPLUS:
3763 case OP_TYPEQUERY:
3764 case OP_TYPEMINQUERY:
3765 case OP_TYPEPOSSTAR:
3766 case OP_TYPEPOSPLUS:
3767 case OP_TYPEPOSQUERY:
3768 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3769 break;
3770
3771 case OP_TYPEUPTO:
3772 case OP_TYPEMINUPTO:
3773 case OP_TYPEEXACT:
3774 case OP_TYPEPOSUPTO:
3775 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3776 code += 2;
3777 break;
3778
3779 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3780 case OP_XCLASS:
3781 code += GET(code, 1);
3782 break;
3783 #endif
3784
3785 case OP_MARK:
3786 case OP_PRUNE_ARG:
3787 case OP_SKIP_ARG:
3788 case OP_THEN_ARG:
3789 code += code[1];
3790 break;
3791 }
3792
3793 /* Add in the fixed length from the table */
3794
3795 code += PRIV(OP_lengths)[c];
3796
3797 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3798 a multi-byte character. The length in the table is a minimum, so we have to
3799 arrange to skip the extra bytes. */
3800
3801 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3802 if (utf) switch(c)
3803 {
3804 case OP_CHAR:
3805 case OP_CHARI:
3806 case OP_NOT:
3807 case OP_NOTI:
3808 case OP_STAR:
3809 case OP_MINSTAR:
3810 case OP_PLUS:
3811 case OP_MINPLUS:
3812 case OP_QUERY:
3813 case OP_MINQUERY:
3814 case OP_UPTO:
3815 case OP_MINUPTO:
3816 case OP_EXACT:
3817 case OP_POSSTAR:
3818 case OP_POSPLUS:
3819 case OP_POSQUERY:
3820 case OP_POSUPTO:
3821 case OP_STARI:
3822 case OP_MINSTARI:
3823 case OP_PLUSI:
3824 case OP_MINPLUSI:
3825 case OP_QUERYI:
3826 case OP_MINQUERYI:
3827 case OP_UPTOI:
3828 case OP_MINUPTOI:
3829 case OP_EXACTI:
3830 case OP_POSSTARI:
3831 case OP_POSPLUSI:
3832 case OP_POSQUERYI:
3833 case OP_POSUPTOI:
3834 case OP_NOTSTAR:
3835 case OP_NOTMINSTAR:
3836 case OP_NOTPLUS:
3837 case OP_NOTMINPLUS:
3838 case OP_NOTQUERY:
3839 case OP_NOTMINQUERY:
3840 case OP_NOTUPTO:
3841 case OP_NOTMINUPTO:
3842 case OP_NOTEXACT:
3843 case OP_NOTPOSSTAR:
3844 case OP_NOTPOSPLUS:
3845 case OP_NOTPOSQUERY:
3846 case OP_NOTPOSUPTO:
3847 case OP_NOTSTARI:
3848 case OP_NOTMINSTARI:
3849 case OP_NOTPLUSI:
3850 case OP_NOTMINPLUSI:
3851 case OP_NOTQUERYI:
3852 case OP_NOTMINQUERYI:
3853 case OP_NOTUPTOI:
3854 case OP_NOTMINUPTOI:
3855 case OP_NOTEXACTI:
3856 case OP_NOTPOSSTARI:
3857 case OP_NOTPOSPLUSI:
3858 case OP_NOTPOSQUERYI:
3859 case OP_NOTPOSUPTOI:
3860 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3861 break;
3862 }
3863 #else
3864 (void)(utf); /* Keep compiler happy by referencing function argument */
3865 #endif
3866 }
3867 }
3868
3869
3870
3871 /*************************************************
3872 * Check for POSIX class syntax *
3873 *************************************************/
3874
3875 /* This function is called when the sequence "[:" or "[." or "[=" is
3876 encountered in a character class. It checks whether this is followed by a
3877 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3878 reach an unescaped ']' without the special preceding character, return FALSE.
3879
3880 Originally, this function only recognized a sequence of letters between the
3881 terminators, but it seems that Perl recognizes any sequence of characters,
3882 though of course unknown POSIX names are subsequently rejected. Perl gives an
3883 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3884 didn't consider this to be a POSIX class. Likewise for [:1234:].
3885
3886 The problem in trying to be exactly like Perl is in the handling of escapes. We
3887 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3888 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3889 below handles the special case of \], but does not try to do any other escape
3890 processing. This makes it different from Perl for cases such as [:l\ower:]
3891 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3892 "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3893 I think.
3894
3895 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3896 It seems that the appearance of a nested POSIX class supersedes an apparent
3897 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3898 a digit.
3899
3900 In Perl, unescaped square brackets may also appear as part of class names. For
3901 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3902 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3903 seem right at all. PCRE does not allow closing square brackets in POSIX class
3904 names.
3905
3906 Arguments:
3907 ptr pointer to the initial [
3908 endptr where to return the end pointer
3909
3910 Returns: TRUE or FALSE
3911 */
3912
3913 static BOOL
3914 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3915 {
3916 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3917 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3918 for (++ptr; *ptr != CHAR_NULL; ptr++)
3919 {
3920 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3921 ptr++;
3922 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3923 else
3924 {
3925 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3926 {
3927 *endptr = ptr;
3928 return TRUE;
3929 }
3930 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3931 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3932 ptr[1] == CHAR_EQUALS_SIGN) &&
3933 check_posix_syntax(ptr, endptr))
3934 return FALSE;
3935 }
3936 }
3937 return FALSE;
3938 }
3939
3940
3941
3942
3943 /*************************************************
3944 * Check POSIX class name *
3945 *************************************************/
3946
3947 /* This function is called to check the name given in a POSIX-style class entry
3948 such as [:alnum:].
3949
3950 Arguments:
3951 ptr points to the first letter
3952 len the length of the name
3953
3954 Returns: a value representing the name, or -1 if unknown
3955 */
3956
3957 static int
3958 check_posix_name(const pcre_uchar *ptr, int len)
3959 {
3960 const char *pn = posix_names;
3961 register int yield = 0;
3962 while (posix_name_lengths[yield] != 0)
3963 {
3964 if (len == posix_name_lengths[yield] &&
3965 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3966 pn += posix_name_lengths[yield] + 1;
3967 yield++;
3968 }
3969 return -1;
3970 }
3971
3972
3973 /*************************************************
3974 * Adjust OP_RECURSE items in repeated group *
3975 *************************************************/
3976
3977 /* OP_RECURSE items contain an offset from the start of the regex to the group
3978 that is referenced. This means that groups can be replicated for fixed
3979 repetition simply by copying (because the recursion is allowed to refer to
3980 earlier groups that are outside the current group). However, when a group is
3981 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3982 inserted before it, after it has been compiled. This means that any OP_RECURSE
3983 items within it that refer to the group itself or any contained groups have to
3984 have their offsets adjusted. That one of the jobs of this function. Before it
3985 is called, the partially compiled regex must be temporarily terminated with
3986 OP_END.
3987
3988 This function has been extended to cope with forward references for recursions
3989 and subroutine calls. It must check the list of such references for the
3990 group we are dealing with. If it finds that one of the recursions in the
3991 current group is on this list, it does not adjust the value in the reference
3992 (which is a group number). After the group has been scanned, all the offsets in
3993 the forward reference list for the group are adjusted.
3994
3995 Arguments:
3996 group points to the start of the group
3997 adjust the amount by which the group is to be moved
3998 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3999 cd contains pointers to tables etc.
4000 save_hwm_offset the hwm forward reference offset at the start of the group
4001
4002 Returns: nothing
4003 */
4004
4005 static void
4006 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
4007 size_t save_hwm_offset)
4008 {
4009 int offset;
4010 pcre_uchar *hc;
4011 pcre_uchar *ptr = group;
4012
4013 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
4014 {
4015 for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4016 hc += LINK_SIZE)
4017 {
4018 offset = (int)GET(hc, 0);
4019 if (cd->start_code + offset == ptr + 1) break;
4020 }
4021
4022 /* If we have not found this recursion on the forward reference list, adjust
4023 the recursion's offset if it's after the start of this group. */
4024
4025 if (hc >= cd->hwm)
4026 {
4027 offset = (int)GET(ptr, 1);
4028 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
4029 }
4030
4031 ptr += 1 + LINK_SIZE;
4032 }
4033
4034 /* Now adjust all forward reference offsets for the group. */
4035
4036 for (hc = (pcre_uchar *)cd->start_workspace + save_hwm_offset; hc < cd->hwm;
4037 hc += LINK_SIZE)
4038 {
4039 offset = (int)GET(hc, 0);
4040 PUT(hc, 0, offset + adjust);
4041 }
4042 }
4043
4044
4045
4046 /*************************************************
4047 * Insert an automatic callout point *
4048 *************************************************/
4049
4050 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
4051 callout points before each pattern item.
4052
4053 Arguments:
4054 code current code pointer
4055 ptr current pattern pointer
4056 cd pointers to tables etc
4057
4058 Returns: new code pointer
4059 */
4060
4061 static pcre_uchar *
4062 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
4063 {
4064 *code++ = OP_CALLOUT;
4065 *code++ = 255;
4066 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
4067 PUT(code, LINK_SIZE, 0); /* Default length */
4068 return code + 2 * LINK_SIZE;
4069 }
4070
4071
4072
4073 /*************************************************
4074 * Complete a callout item *
4075 *************************************************/
4076
4077 /* A callout item contains the length of the next item in the pattern, which
4078 we can't fill in till after we have reached the relevant point. This is used
4079 for both automatic and manual callouts.
4080
4081 Arguments:
4082 previous_callout points to previous callout item
4083 ptr current pattern pointer
4084 cd pointers to tables etc
4085
4086 Returns: nothing
4087 */
4088
4089 static void
4090 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4091 {
4092 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4093 PUT(previous_callout, 2 + LINK_SIZE, length);
4094 }
4095
4096
4097
4098 #ifdef SUPPORT_UCP
4099 /*************************************************
4100 * Get othercase range *
4101 *************************************************/
4102
4103 /* This function is passed the start and end of a class range, in UTF-8 mode
4104 with UCP support. It searches up the characters, looking for ranges of
4105 characters in the "other" case. Each call returns the next one, updating the
4106 start address. A character with multiple other cases is returned on its own
4107 with a special return value.
4108
4109 Arguments:
4110 cptr points to starting character value; updated
4111 d end value
4112 ocptr where to put start of othercase range
4113 odptr where to put end of othercase range
4114
4115 Yield: -1 when no more
4116 0 when a range is returned
4117 >0 the CASESET offset for char with multiple other cases
4118 in this case, ocptr contains the original
4119 */
4120
4121 static int
4122 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4123 pcre_uint32 *odptr)
4124 {
4125 pcre_uint32 c, othercase, next;
4126 unsigned int co;
4127
4128 /* Find the first character that has an other case. If it has multiple other
4129 cases, return its case offset value. */
4130
4131 for (c = *cptr; c <= d; c++)
4132 {
4133 if ((co = UCD_CASESET(c)) != 0)
4134 {
4135 *ocptr = c++; /* Character that has the set */
4136 *cptr = c; /* Rest of input range */
4137 return (int)co;
4138 }
4139 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4140 }
4141
4142 if (c > d) return -1; /* Reached end of range */
4143
4144 /* Found a character that has a single other case. Search for the end of the
4145 range, which is either the end of the input range, or a character that has zero
4146 or more than one other cases. */
4147
4148 *ocptr = othercase;
4149 next = othercase + 1;
4150
4151 for (++c; c <= d; c++)
4152 {
4153 if ((co = UCD_CASESET(c)) != 0 || UCD_OTHERCASE(c) != next) break;
4154 next++;
4155 }
4156
4157 *odptr = next - 1; /* End of othercase range */
4158 *cptr = c; /* Rest of input range */
4159 return 0;
4160 }
4161 #endif /* SUPPORT_UCP */
4162
4163
4164
4165 /*************************************************
4166 * Add a character or range to a class *
4167 *************************************************/
4168
4169 /* This function packages up the logic of adding a character or range of
4170 characters to a class. The character values in the arguments will be within the
4171 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4172 mutually recursive with the function immediately below.
4173
4174 Arguments:
4175 classbits the bit map for characters < 256
4176 uchardptr points to the pointer for extra data
4177 options the options word
4178 cd contains pointers to tables etc.
4179 start start of range character
4180 end end of range character
4181
4182 Returns: the number of < 256 characters added
4183 the pointer to extra data is updated
4184 */
4185
4186 static int
4187 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4188 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4189 {
4190 pcre_uint32 c;
4191 pcre_uint32 classbits_end = (end <= 0xff ? end : 0xff);
4192 int n8 = 0;
4193
4194 /* If caseless matching is required, scan the range and process alternate
4195 cases. In Unicode, there are 8-bit characters that have alternate cases that
4196 are greater than 255 and vice-versa. Sometimes we can just extend the original
4197 range. */
4198
4199 if ((options & PCRE_CASELESS) != 0)
4200 {
4201 #ifdef SUPPORT_UCP
4202 if ((options & PCRE_UTF8) != 0)
4203 {
4204 int rc;
4205 pcre_uint32 oc, od;
4206
4207 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4208 c = start;
4209
4210 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4211 {
4212 /* Handle a single character that has more than one other case. */
4213
4214 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4215 PRIV(ucd_caseless_sets) + rc, oc);
4216
4217 /* Do nothing if the other case range is within the original range. */
4218
4219 else if (oc >= start && od <= end) continue;
4220
4221 /* Extend the original range if there is overlap, noting that if oc < c, we
4222 can't have od > end because a subrange is always shorter than the basic
4223 range. Otherwise, use a recursive call to add the additional range. */
4224
4225 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4226 else if (od > end && oc <= end + 1)
4227 {
4228 end = od; /* Extend upwards */
4229 if (end > classbits_end) classbits_end = (end <= 0xff ? end : 0xff);
4230 }
4231 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4232 }
4233 }
4234 else
4235 #endif /* SUPPORT_UCP */
4236
4237 /* Not UTF-mode, or no UCP */
4238
4239 for (c = start; c <= classbits_end; c++)
4240 {
4241 SETBIT(classbits, cd->fcc[c]);
4242 n8++;
4243 }
4244 }
4245
4246 /* Now handle the original range. Adjust the final value according to the bit
4247 length - this means that the same lists of (e.g.) horizontal spaces can be used
4248 in all cases. */
4249
4250 #if defined COMPILE_PCRE8
4251 #ifdef SUPPORT_UTF
4252 if ((options & PCRE_UTF8) == 0)
4253 #endif
4254 if (end > 0xff) end = 0xff;
4255
4256 #elif defined COMPILE_PCRE16
4257 #ifdef SUPPORT_UTF
4258 if ((options & PCRE_UTF16) == 0)
4259 #endif
4260 if (end > 0xffff) end = 0xffff;
4261
4262 #endif /* COMPILE_PCRE[8|16] */
4263
4264 /* Use the bitmap for characters < 256. Otherwise use extra data.*/
4265
4266 for (c = start; c <= classbits_end; c++)
4267 {
4268 /* Regardless of start, c will always be <= 255. */
4269 SETBIT(classbits, c);
4270 n8++;
4271 }
4272
4273 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4274 if (start <= 0xff) start = 0xff + 1;
4275
4276 if (end >= start)
4277 {
4278 pcre_uchar *uchardata = *uchardptr;
4279 #ifdef SUPPORT_UTF
4280 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4281 {
4282 if (start < end)
4283 {
4284 *uchardata++ = XCL_RANGE;
4285 uchardata += PRIV(ord2utf)(start, uchardata);
4286 uchardata += PRIV(ord2utf)(end, uchardata);
4287 }
4288 else if (start == end)
4289 {
4290 *uchardata++ = XCL_SINGLE;
4291 uchardata += PRIV(ord2utf)(start, uchardata);
4292 }
4293 }
4294 else
4295 #endif /* SUPPORT_UTF */
4296
4297 /* Without UTF support, character values are constrained by the bit length,
4298 and can only be > 256 for 16-bit and 32-bit libraries. */
4299
4300 #ifdef COMPILE_PCRE8
4301 {}
4302 #else
4303 if (start < end)
4304 {
4305 *uchardata++ = XCL_RANGE;
4306 *uchardata++ = start;
4307 *uchardata++ = end;
4308 }
4309 else if (start == end)
4310 {
4311 *uchardata++ = XCL_SINGLE;
4312 *uchardata++ = start;
4313 }
4314 #endif
4315
4316 *uchardptr = uchardata; /* Updata extra data pointer */
4317 }
4318 #endif /* SUPPORT_UTF || !COMPILE_PCRE8 */
4319
4320 return n8; /* Number of 8-bit characters */
4321 }
4322
4323
4324
4325
4326 /*************************************************
4327 * Add a list of characters to a class *
4328 *************************************************/
4329
4330 /* This function is used for adding a list of case-equivalent characters to a
4331 class, and also for adding a list of horizontal or vertical whitespace. If the
4332 list is in order (which it should be), ranges of characters are detected and
4333 handled appropriately. This function is mutually recursive with the function
4334 above.
4335
4336 Arguments:
4337 classbits the bit map for characters < 256
4338 uchardptr points to the pointer for extra data
4339 options the options word
4340 cd contains pointers to tables etc.
4341 p points to row of 32-bit values, terminated by NOTACHAR
4342 except character to omit; this is used when adding lists of
4343 case-equivalent characters to avoid including the one we
4344 already know about
4345
4346 Returns: the number of < 256 characters added
4347 the pointer to extra data is updated
4348 */
4349
4350 static int
4351 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4352 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4353 {
4354 int n8 = 0;
4355 while (p[0] < NOTACHAR)
4356 {
4357 int n = 0;
4358 if (p[0] != except)
4359 {
4360 while(p[n+1] == p[0] + n + 1) n++;
4361 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4362 }
4363 p += n + 1;
4364 }
4365 return n8;
4366 }
4367
4368
4369
4370 /*************************************************
4371 * Add characters not in a list to a class *
4372 *************************************************/
4373
4374 /* This function is used for adding the complement of a list of horizontal or
4375 vertical whitespace to a class. The list must be in order.
4376
4377 Arguments:
4378 classbits the bit map for characters < 256
4379 uchardptr points to the pointer for extra data
4380 options the options word
4381 cd contains pointers to tables etc.
4382 p points to row of 32-bit values, terminated by NOTACHAR
4383
4384 Returns: the number of < 256 characters added
4385 the pointer to extra data is updated
4386 */
4387
4388 static int
4389 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4390 int options, compile_data *cd, const pcre_uint32 *p)
4391 {
4392 BOOL utf = (options & PCRE_UTF8) != 0;
4393 int n8 = 0;
4394 if (p[0] > 0)
4395 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4396 while (p[0] < NOTACHAR)
4397 {
4398 while (p[1] == p[0] + 1) p++;
4399 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4400 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4401 p++;
4402 }
4403 return n8;
4404 }
4405
4406
4407
4408 /*************************************************
4409 * Compile one branch *
4410 *************************************************/
4411
4412 /* Scan the pattern, compiling it into the a vector. If the options are
4413 changed during the branch, the pointer is used to change the external options
4414 bits. This function is used during the pre-compile phase when we are trying
4415 to find out the amount of memory needed, as well as during the real compile
4416 phase. The value of lengthptr distinguishes the two phases.
4417
4418 Arguments:
4419 optionsptr pointer to the option bits
4420 codeptr points to the pointer to the current code point
4421 ptrptr points to the current pattern pointer
4422 errorcodeptr points to error code variable
4423 firstcharptr place to put the first required character
4424 firstcharflagsptr place to put the first character flags, or a negative number
4425 reqcharptr place to put the last required character
4426 reqcharflagsptr place to put the last required character flags, or a negative number
4427 bcptr points to current branch chain
4428 cond_depth conditional nesting depth
4429 cd contains pointers to tables etc.
4430 lengthptr NULL during the real compile phase
4431 points to length accumulator during pre-compile phase
4432
4433 Returns: TRUE on success
4434 FALSE, with *errorcodeptr set non-zero on error
4435 */
4436
4437 static BOOL
4438 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4439 const pcre_uchar **ptrptr, int *errorcodeptr,
4440 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4441 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4442 branch_chain *bcptr, int cond_depth,
4443 compile_data *cd, int *lengthptr)
4444 {
4445 int repeat_type, op_type;
4446 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4447 int bravalue = 0;
4448 int greedy_default, greedy_non_default;
4449 pcre_uint32 firstchar, reqchar;
4450 pcre_int32 firstcharflags, reqcharflags;
4451 pcre_uint32 zeroreqchar, zerofirstchar;
4452 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4453 pcre_int32 req_caseopt, reqvary, tempreqvary;
4454 int options = *optionsptr; /* May change dynamically */
4455 int after_manual_callout = 0;
4456 int length_prevgroup = 0;
4457 register pcre_uint32 c;
4458 int escape;
4459 register pcre_uchar *code = *codeptr;
4460 pcre_uchar *last_code = code;
4461 pcre_uchar *orig_code = code;
4462 pcre_uchar *tempcode;
4463 BOOL inescq = FALSE;
4464 BOOL groupsetfirstchar = FALSE;
4465 const pcre_uchar *ptr = *ptrptr;
4466 const pcre_uchar *tempptr;
4467 const pcre_uchar *nestptr = NULL;
4468 pcre_uchar *previous = NULL;
4469 pcre_uchar *previous_callout = NULL;
4470 size_t item_hwm_offset = 0;
4471 pcre_uint8 classbits[32];
4472
4473 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4474 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4475 dynamically as we process the pattern. */
4476
4477 #ifdef SUPPORT_UTF
4478 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4479 BOOL utf = (options & PCRE_UTF8) != 0;
4480 #ifndef COMPILE_PCRE32
4481 pcre_uchar utf_chars[6];
4482 #endif
4483 #else
4484 BOOL utf = FALSE;
4485 #endif
4486
4487 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4488 class_uchardata always so that it can be passed to add_to_class() always,
4489 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4490 alternative calls for the different cases. */
4491
4492 pcre_uchar *class_uchardata;
4493 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4494 BOOL xclass;
4495 pcre_uchar *class_uchardata_base;
4496 #endif
4497
4498 #ifdef PCRE_DEBUG
4499 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4500 #endif
4501
4502 /* Set up the default and non-default settings for greediness */
4503
4504 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4505 greedy_non_default = greedy_default ^ 1;
4506
4507 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4508 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4509 matches a non-fixed char first char; reqchar just remains unset if we never
4510 find one.
4511
4512 When we hit a repeat whose minimum is zero, we may have to adjust these values
4513 to take the zero repeat into account. This is implemented by setting them to
4514 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4515 item types that can be repeated set these backoff variables appropriately. */
4516
4517 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4518 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4519
4520 /* The variable req_caseopt contains either the REQ_CASELESS value
4521 or zero, according to the current setting of the caseless flag. The
4522 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4523 firstchar or reqchar variables to record the case status of the
4524 value. This is used only for ASCII characters. */
4525
4526 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4527
4528 /* Switch on next character until the end of the branch */
4529
4530 for (;; ptr++)
4531 {
4532 BOOL negate_class;
4533 BOOL should_flip_negation;
4534 BOOL possessive_quantifier;
4535 BOOL is_quantifier;
4536 BOOL is_recurse;
4537 BOOL reset_bracount;
4538 int class_has_8bitchar;
4539 int class_one_char;
4540 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4541 BOOL xclass_has_prop;
4542 #endif
4543 int newoptions;
4544 int recno;
4545 int refsign;
4546 int skipbytes;
4547 pcre_uint32 subreqchar, subfirstchar;
4548 pcre_int32 subreqcharflags, subfirstcharflags;
4549 int terminator;
4550 unsigned int mclength;
4551 unsigned int tempbracount;
4552 pcre_uint32 ec;
4553 pcre_uchar mcbuffer[8];
4554
4555 /* Get next character in the pattern */
4556
4557 c = *ptr;
4558
4559 /* If we are at the end of a nested substitution, revert to the outer level
4560 string. Nesting only happens one level deep. */
4561
4562 if (c == CHAR_NULL && nestptr != NULL)
4563 {
4564 ptr = nestptr;
4565 nestptr = NULL;
4566 c = *ptr;
4567 }
4568
4569 /* If we are in the pre-compile phase, accumulate the length used for the
4570 previous cycle of this loop. */
4571
4572 if (lengthptr != NULL)
4573 {
4574 #ifdef PCRE_DEBUG
4575 if (code > cd->hwm) cd->hwm = code; /* High water info */
4576 #endif
4577 if (code > cd->start_workspace + cd->workspace_size -
4578 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4579 {
4580 *errorcodeptr = ERR52;
4581 goto FAILED;
4582 }
4583
4584 /* There is at least one situation where code goes backwards: this is the
4585 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4586 the class is simply eliminated. However, it is created first, so we have to
4587 allow memory for it. Therefore, don't ever reduce the length at this point.
4588 */
4589
4590 if (code < last_code) code = last_code;
4591
4592 /* Paranoid check for integer overflow */
4593
4594 if (OFLOW_MAX - *lengthptr < code - last_code)
4595 {
4596 *errorcodeptr = ERR20;
4597 goto FAILED;
4598 }
4599
4600 *lengthptr += (int)(code - last_code);
4601 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4602 (int)(code - last_code), c, c));
4603
4604 /* If "previous" is set and it is not at the start of the work space, move
4605 it back to there, in order to avoid filling up the work space. Otherwise,
4606 if "previous" is NULL, reset the current code pointer to the start. */
4607
4608 if (previous != NULL)
4609 {
4610 if (previous > orig_code)
4611 {
4612 memmove(orig_code, previous, IN_UCHARS(code - previous));
4613 code -= previous - orig_code;
4614 previous = orig_code;
4615 }
4616 }
4617 else code = orig_code;
4618
4619 /* Remember where this code item starts so we can pick up the length
4620 next time round. */
4621
4622 last_code = code;
4623 }
4624
4625 /* In the real compile phase, just check the workspace used by the forward
4626 reference list. */
4627
4628 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4629 WORK_SIZE_SAFETY_MARGIN)
4630 {
4631 *errorcodeptr = ERR52;
4632 goto FAILED;
4633 }
4634
4635 /* If in \Q...\E, check for the end; if not, we have a literal */
4636
4637 if (inescq && c != CHAR_NULL)
4638 {
4639 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4640 {
4641 inescq = FALSE;
4642 ptr++;
4643 continue;
4644 }
4645 else
4646 {
4647 if (previous_callout != NULL)
4648 {
4649 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4650 complete_callout(previous_callout, ptr, cd);
4651 previous_callout = NULL;
4652 }
4653 if ((options & PCRE_AUTO_CALLOUT) != 0)
4654 {
4655 previous_callout = code;
4656 code = auto_callout(code, ptr, cd);
4657 }
4658 goto NORMAL_CHAR;
4659 }
4660 /* Control does not reach here. */
4661 }
4662
4663 /* In extended mode, skip white space and comments. We need a loop in order
4664 to check for more white space and more comments after a comment. */
4665
4666 if ((options & PCRE_EXTENDED) != 0)
4667 {
4668 for (;;)
4669 {
4670 while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4671 if (c != CHAR_NUMBER_SIGN) break;
4672 ptr++;
4673 while (*ptr != CHAR_NULL)
4674 {
4675 if (IS_NEWLINE(ptr)) /* For non-fixed-length newline cases, */
4676 { /* IS_NEWLINE sets cd->nllen. */
4677 ptr += cd->nllen;
4678 break;
4679 }
4680 ptr++;
4681 #ifdef SUPPORT_UTF
4682 if (utf) FORWARDCHAR(ptr);
4683 #endif
4684 }
4685 c = *ptr; /* Either NULL or the char after a newline */
4686 }
4687 }
4688
4689 /* See if the next thing is a quantifier. */
4690
4691 is_quantifier =
4692 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4693 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4694
4695 /* Fill in length of a previous callout, except when the next thing is a
4696 quantifier or when processing a property substitution string in UCP mode. */
4697
4698 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4699 after_manual_callout-- <= 0)
4700 {
4701 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4702 complete_callout(previous_callout, ptr, cd);
4703 previous_callout = NULL;
4704 }
4705
4706 /* Create auto callout, except for quantifiers, or while processing property
4707 strings that are substituted for \w etc in UCP mode. */
4708
4709 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4710 {
4711 previous_callout = code;
4712 code = auto_callout(code, ptr, cd);
4713 }
4714
4715 /* Process the next pattern item. */
4716
4717 switch(c)
4718 {
4719 /* ===================================================================*/
4720 case CHAR_NULL: /* The branch terminates at string end */
4721 case CHAR_VERTICAL_LINE: /* or | or ) */
4722 case CHAR_RIGHT_PARENTHESIS:
4723 *firstcharptr = firstchar;
4724 *firstcharflagsptr = firstcharflags;
4725 *reqcharptr = reqchar;
4726 *reqcharflagsptr = reqcharflags;
4727 *codeptr = code;
4728 *ptrptr = ptr;
4729 if (lengthptr != NULL)
4730 {
4731 if (OFLOW_MAX - *lengthptr < code - last_code)
4732 {
4733 *errorcodeptr = ERR20;
4734 goto FAILED;
4735 }
4736 *lengthptr += (int)(code - last_code); /* To include callout length */
4737 DPRINTF((">> end branch\n"));
4738 }
4739 return TRUE;
4740
4741
4742 /* ===================================================================*/
4743 /* Handle single-character metacharacters. In multiline mode, ^ disables
4744 the setting of any following char as a first character. */
4745
4746 case CHAR_CIRCUMFLEX_ACCENT:
4747 previous = NULL;
4748 if ((options & PCRE_MULTILINE) != 0)
4749 {
4750 if (firstcharflags == REQ_UNSET)
4751 zerofirstcharflags = firstcharflags = REQ_NONE;
4752 *code++ = OP_CIRCM;
4753 }
4754 else *code++ = OP_CIRC;
4755 break;
4756
4757 case CHAR_DOLLAR_SIGN:
4758 previous = NULL;
4759 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4760 break;
4761
4762 /* There can never be a first char if '.' is first, whatever happens about
4763 repeats. The value of reqchar doesn't change either. */
4764
4765 case CHAR_DOT:
4766 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4767 zerofirstchar = firstchar;
4768 zerofirstcharflags = firstcharflags;
4769 zeroreqchar = reqchar;
4770 zeroreqcharflags = reqcharflags;
4771 previous = code;
4772 item_hwm_offset = cd->hwm - cd->start_workspace;
4773 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4774 break;
4775
4776
4777 /* ===================================================================*/
4778 /* Character classes. If the included characters are all < 256, we build a
4779 32-byte bitmap of the permitted characters, except in the special case
4780 where there is only one such character. For negated classes, we build the
4781 map as usual, then invert it at the end. However, we use a different opcode
4782 so that data characters > 255 can be handled correctly.
4783
4784 If the class contains characters outside the 0-255 range, a different
4785 opcode is compiled. It may optionally have a bit map for characters < 256,
4786 but those above are are explicitly listed afterwards. A flag byte tells
4787 whether the bitmap is present, and whether this is a negated class or not.
4788
4789 In JavaScript compatibility mode, an isolated ']' causes an error. In
4790 default (Perl) mode, it is treated as a data character. */
4791
4792 case CHAR_RIGHT_SQUARE_BRACKET:
4793 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4794 {
4795 *errorcodeptr = ERR64;
4796 goto FAILED;
4797 }
4798 goto NORMAL_CHAR;
4799
4800 /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4801 used for "start of word" and "end of word". As these are otherwise illegal
4802 sequences, we don't break anything by recognizing them. They are replaced
4803 by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4804 erroneous and are handled by the normal code below. */
4805
4806 case CHAR_LEFT_SQUARE_BRACKET:
4807 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4808 {
4809 nestptr = ptr + 7;
4810 ptr = sub_start_of_word - 1;
4811 continue;
4812 }
4813
4814 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4815 {
4816 nestptr = ptr + 7;
4817 ptr = sub_end_of_word - 1;
4818 continue;
4819 }
4820
4821 /* Handle a real character class. */
4822
4823 previous = code;
4824 item_hwm_offset = cd->hwm - cd->start_workspace;
4825
4826 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4827 they are encountered at the top level, so we'll do that too. */
4828
4829 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4830 ptr[1] == CHAR_EQUALS_SIGN) &&
4831 check_posix_syntax(ptr, &tempptr))
4832 {
4833 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4834 goto FAILED;
4835 }
4836
4837 /* If the first character is '^', set the negation flag and skip it. Also,
4838 if the first few characters (either before or after ^) are \Q\E or \E we
4839 skip them too. This makes for compatibility with Perl. */
4840
4841 negate_class = FALSE;
4842 for (;;)
4843 {
4844 c = *(++ptr);
4845 if (c == CHAR_BACKSLASH)
4846 {
4847 if (ptr[1] == CHAR_E)
4848 ptr++;
4849 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4850 ptr += 3;
4851 else
4852 break;
4853 }
4854 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4855 negate_class = TRUE;
4856 else break;
4857 }
4858
4859 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4860 an initial ']' is taken as a data character -- the code below handles
4861 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4862 [^] must match any character, so generate OP_ALLANY. */
4863
4864 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4865 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4866 {
4867 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4868 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4869 zerofirstchar = firstchar;
4870 zerofirstcharflags = firstcharflags;
4871 break;
4872 }
4873
4874 /* If a class contains a negative special such as \S, we need to flip the
4875 negation flag at the end, so that support for characters > 255 works
4876 correctly (they are all included in the class). */
4877
4878 should_flip_negation = FALSE;
4879
4880 /* Extended class (xclass) will be used when characters > 255
4881 might match. */
4882
4883 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4884 xclass = FALSE;
4885 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4886 class_uchardata_base = class_uchardata; /* Save the start */
4887 #endif
4888
4889 /* For optimization purposes, we track some properties of the class:
4890 class_has_8bitchar will be non-zero if the class contains at least one <
4891 256 character; class_one_char will be 1 if the class contains just one
4892 character; xclass_has_prop will be TRUE if unicode property checks
4893 are present in the class. */
4894
4895 class_has_8bitchar = 0;
4896 class_one_char = 0;
4897 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4898 xclass_has_prop = FALSE;
4899 #endif
4900
4901 /* Initialize the 32-char bit map to all zeros. We build the map in a
4902 temporary bit of memory, in case the class contains fewer than two
4903 8-bit characters because in that case the compiled code doesn't use the bit
4904 map. */
4905
4906 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4907
4908 /* Process characters until ] is reached. By writing this as a "do" it
4909 means that an initial ] is taken as a data character. At the start of the
4910 loop, c contains the first byte of the character. */
4911
4912 if (c != CHAR_NULL) do
4913 {
4914 const pcre_uchar *oldptr;
4915
4916 #ifdef SUPPORT_UTF
4917 if (utf && HAS_EXTRALEN(c))
4918 { /* Braces are required because the */
4919 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4920 }
4921 #endif
4922
4923 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4924 /* In the pre-compile phase, accumulate the length of any extra
4925 data and reset the pointer. This is so that very large classes that
4926 contain a zillion > 255 characters no longer overwrite the work space
4927 (which is on the stack). We have to remember that there was XCLASS data,
4928 however. */
4929
4930 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4931 {
4932 xclass = TRUE;
4933 *lengthptr += (int)(class_uchardata - class_uchardata_base);
4934 class_uchardata = class_uchardata_base;
4935 }
4936 #endif
4937
4938 /* Inside \Q...\E everything is literal except \E */
4939
4940 if (inescq)
4941 {
4942 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4943 {
4944 inescq = FALSE; /* Reset literal state */
4945 ptr++; /* Skip the 'E' */
4946 continue; /* Carry on with next */
4947 }
4948 goto CHECK_RANGE; /* Could be range if \E follows */
4949 }
4950
4951 /* Handle POSIX class names. Perl allows a negation extension of the
4952 form [:^name:]. A square bracket that doesn't match the syntax is
4953 treated as a literal. We also recognize the POSIX constructions
4954 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4955 5.6 and 5.8 do. */
4956
4957 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4958 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4959 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4960 {
4961 BOOL local_negate = FALSE;
4962 int posix_class, taboffset, tabopt;
4963 register const pcre_uint8 *cbits = cd->cbits;
4964 pcre_uint8 pbits[32];
4965
4966 if (ptr[1] != CHAR_COLON)
4967 {
4968 *errorcodeptr = ERR31;
4969 goto FAILED;
4970 }
4971
4972 ptr += 2;
4973 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4974 {
4975 local_negate = TRUE;
4976 should_flip_negation = TRUE; /* Note negative special */
4977 ptr++;
4978 }
4979
4980 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4981 if (posix_class < 0)
4982 {
4983 *errorcodeptr = ERR30;
4984 goto FAILED;
4985 }
4986
4987 /* If matching is caseless, upper and lower are converted to
4988 alpha. This relies on the fact that the class table starts with
4989 alpha, lower, upper as the first 3 entries. */
4990
4991 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4992 posix_class = 0;
4993
4994 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4995 different escape sequences that use Unicode properties \p or \P. Others
4996 that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4997 directly. */
4998
4999 #ifdef SUPPORT_UCP
5000 if ((options & PCRE_UCP) != 0)
5001 {
5002 unsigned int ptype = 0;
5003 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
5004
5005 /* The posix_substitutes table specifies which POSIX classes can be
5006 converted to \p or \P items. */
5007
5008 if (posix_substitutes[pc] != NULL)
5009 {
5010 nestptr = tempptr + 1;
5011 ptr = posix_substitutes[pc] - 1;
5012 continue;
5013 }
5014
5015 /* There are three other classes that generate special property calls
5016 that are recognized only in an XCLASS. */
5017
5018 else switch(posix_class)
5019 {
5020 case PC_GRAPH:
5021 ptype = PT_PXGRAPH;
5022 /* Fall through */
5023 case PC_PRINT:
5024 if (ptype == 0) ptype = PT_PXPRINT;
5025 /* Fall through */
5026 case PC_PUNCT:
5027 if (ptype == 0) ptype = PT_PXPUNCT;
5028 *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
5029 *class_uchardata++ = ptype;
5030 *class_uchardata++ = 0;
5031 xclass_has_prop = TRUE;
5032 ptr = tempptr + 1;
5033 continue;
5034
5035 /* For all other POSIX classes, no special action is taken in UCP
5036 mode. Fall through to the non_UCP case. */
5037
5038 default:
5039 break;
5040 }
5041 }
5042 #endif
5043 /* In the non-UCP case, or when UCP makes no difference, we build the
5044 bit map for the POSIX class in a chunk of local store because we may be
5045 adding and subtracting from it, and we don't want to subtract bits that
5046 may be in the main map already. At the end we or the result into the
5047 bit map that is being built. */
5048
5049 posix_class *= 3;
5050
5051 /* Copy in the first table (always present) */
5052
5053 memcpy(pbits, cbits + posix_class_maps[posix_class],
5054 32 * sizeof(pcre_uint8));
5055
5056 /* If there is a second table, add or remove it as required. */
5057
5058 taboffset = posix_class_maps[posix_class + 1];
5059 tabopt = posix_class_maps[posix_class + 2];
5060
5061 if (taboffset >= 0)
5062 {
5063 if (tabopt >= 0)
5064 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
5065 else
5066 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
5067 }
5068
5069 /* Now see if we need to remove any special characters. An option
5070 value of 1 removes vertical space and 2 removes underscore. */
5071
5072 if (tabopt < 0) tabopt = -tabopt;
5073 if (tabopt == 1) pbits[1] &= ~0x3c;
5074 else if (tabopt == 2) pbits[11] &= 0x7f;
5075
5076 /* Add the POSIX table or its complement into the main table that is
5077 being built and we are done. */
5078
5079 if (local_negate)
5080 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
5081 else
5082 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
5083
5084 ptr = tempptr + 1;
5085 /* Every class contains at least one < 256 character. */
5086 class_has_8bitchar = 1;
5087 /* Every class contains at least two characters. */
5088 class_one_char = 2;
5089 continue; /* End of POSIX syntax handling */
5090 }
5091
5092 /* Backslash may introduce a single character, or it may introduce one
5093 of the specials, which just set a flag. The sequence \b is a special
5094 case. Inside a class (and only there) it is treated as backspace. We
5095 assume that other escapes have more than one character in them, so
5096 speculatively set both class_has_8bitchar and class_one_char bigger
5097 than one. Unrecognized escapes fall through and are either treated
5098 as literal characters (by default), or are faulted if
5099 PCRE_EXTRA is set. */
5100
5101 if (c == CHAR_BACKSLASH)
5102 {
5103 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
5104 TRUE);
5105 if (*errorcodeptr != 0) goto FAILED;
5106 if (escape == 0) c = ec;
5107 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
5108 else if (escape == ESC_N) /* \N is not supported in a class */
5109 {
5110 *errorcodeptr = ERR71;
5111 goto FAILED;
5112 }
5113 else if (escape == ESC_Q) /* Handle start of quoted string */
5114 {
5115 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5116 {
5117 ptr += 2; /* avoid empty string */
5118 }
5119 else inescq = TRUE;
5120 continue;
5121 }
5122 else if (escape == ESC_E) continue; /* Ignore orphan \E */
5123
5124 else
5125 {
5126 register const pcre_uint8 *cbits = cd->cbits;
5127 /* Every class contains at least two < 256 characters. */
5128 class_has_8bitchar++;
5129 /* Every class contains at least two characters. */
5130 class_one_char += 2;
5131
5132 switch (escape)
5133 {
5134 #ifdef SUPPORT_UCP
5135 case ESC_du: /* These are the values given for \d etc */
5136 case ESC_DU: /* when PCRE_UCP is set. We replace the */
5137 case ESC_wu: /* escape sequence with an appropriate \p */
5138 case ESC_WU: /* or \P to test Unicode properties instead */
5139 case ESC_su: /* of the default ASCII testing. */
5140 case ESC_SU:
5141 nestptr = ptr;
5142 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
5143 class_has_8bitchar--; /* Undo! */
5144 continue;
5145 #endif
5146 case ESC_d:
5147 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
5148 continue;
5149
5150 case ESC_D:
5151 should_flip_negation = TRUE;
5152 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
5153 continue;
5154
5155 case ESC_w:
5156 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
5157 continue;
5158
5159 case ESC_W:
5160 should_flip_negation = TRUE;
5161 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5162 continue;
5163
5164 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5165 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5166 previously set by something earlier in the character class.
5167 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5168 we could just adjust the appropriate bit. From PCRE 8.34 we no
5169 longer treat \s and \S specially. */
5170
5171 case ESC_s:
5172 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
5173 continue;
5174
5175 case ESC_S:
5176 should_flip_negation = TRUE;
5177 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
5178 continue;
5179
5180 /* The rest apply in both UCP and non-UCP cases. */
5181
5182 case ESC_h:
5183 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5184 PRIV(hspace_list), NOTACHAR);
5185 continue;
5186
5187 case ESC_H:
5188 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5189 cd, PRIV(hspace_list));
5190 continue;
5191
5192 case ESC_v:
5193 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5194 PRIV(vspace_list), NOTACHAR);
5195 continue;
5196
5197 case ESC_V:
5198 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5199 cd, PRIV(vspace_list));
5200 continue;
5201
5202 #ifdef SUPPORT_UCP
5203 case ESC_p:
5204 case ESC_P:
5205 {
5206 BOOL negated;
5207 unsigned int ptype = 0, pdata = 0;
5208 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5209 goto FAILED;
5210 *class_uchardata++ = ((escape == ESC_p) != negated)?
5211 XCL_PROP : XCL_NOTPROP;
5212 *class_uchardata++ = ptype;
5213 *class_uchardata++ = pdata;
5214 xclass_has_prop = TRUE;
5215 class_has_8bitchar--; /* Undo! */
5216 continue;
5217 }
5218 #endif
5219 /* Unrecognized escapes are faulted if PCRE is running in its
5220 strict mode. By default, for compatibility with Perl, they are
5221 treated as literals. */
5222
5223 default:
5224 if ((options & PCRE_EXTRA) != 0)
5225 {
5226 *errorcodeptr = ERR7;
5227 goto FAILED;
5228 }
5229 class_has_8bitchar--; /* Undo the speculative increase. */
5230 class_one_char -= 2; /* Undo the speculative increase. */
5231 c = *ptr; /* Get the final character and fall through */
5232 break;
5233 }
5234 }
5235
5236 /* Fall through if the escape just defined a single character (c >= 0).
5237 This may be greater than 256. */
5238
5239 escape = 0;
5240
5241 } /* End of backslash handling */
5242
5243 /* A character may be followed by '-' to form a range. However, Perl does
5244 not permit ']' to be the end of the range. A '-' character at the end is
5245 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5246 code for handling \Q and \E is messy. */
5247
5248 CHECK_RANGE:
5249 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5250 {
5251 inescq = FALSE;
5252 ptr += 2;
5253 }
5254 oldptr = ptr;
5255
5256 /* Remember if \r or \n were explicitly used */
5257
5258 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5259
5260 /* Check for range */
5261
5262 if (!inescq && ptr[1] == CHAR_MINUS)
5263 {
5264 pcre_uint32 d;
5265 ptr += 2;
5266 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5267
5268 /* If we hit \Q (not followed by \E) at this point, go into escaped
5269 mode. */
5270
5271 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5272 {
5273 ptr += 2;
5274 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5275 { ptr += 2; continue; }
5276 inescq = TRUE;
5277 break;
5278 }
5279
5280 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5281 back the pointer and jump to handle the character that preceded it. */
5282
5283 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5284 {
5285 ptr = oldptr;
5286 goto CLASS_SINGLE_CHARACTER;
5287 }
5288
5289 /* Otherwise, we have a potential range; pick up the next character */
5290
5291 #ifdef SUPPORT_UTF
5292 if (utf)
5293 { /* Braces are required because the */
5294 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5295 }
5296 else
5297 #endif
5298 d = *ptr; /* Not UTF-8 mode */
5299
5300 /* The second part of a range can be a single-character escape
5301 sequence, but not any of the other escapes. Perl treats a hyphen as a
5302 literal in such circumstances. However, in Perl's warning mode, a
5303 warning is given, so PCRE now faults it as it is almost certainly a
5304 mistake on the user's part. */
5305
5306 if (!inescq)
5307 {
5308 if (d == CHAR_BACKSLASH)
5309 {
5310 int descape;
5311 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5312 if (*errorcodeptr != 0) goto FAILED;
5313
5314 /* 0 means a character was put into d; \b is backspace; any other
5315 special causes an error. */
5316
5317 if (descape != 0)
5318 {
5319 if (descape == ESC_b) d = CHAR_BS; else
5320 {
5321 *errorcodeptr = ERR83;
5322 goto FAILED;
5323 }
5324 }
5325 }
5326
5327 /* A hyphen followed by a POSIX class is treated in the same way. */
5328
5329 else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5330 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5331 ptr[1] == CHAR_EQUALS_SIGN) &&
5332 check_posix_syntax(ptr, &tempptr))
5333 {
5334 *errorcodeptr = ERR83;
5335 goto FAILED;
5336 }
5337 }
5338
5339 /* Check that the two values are in the correct order. Optimize
5340 one-character ranges. */
5341
5342 if (d < c)
5343 {
5344 *errorcodeptr = ERR8;
5345 goto FAILED;
5346 }
5347 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5348
5349 /* We have found a character range, so single character optimizations
5350 cannot be done anymore. Any value greater than 1 indicates that there
5351 is more than one character. */
5352
5353 class_one_char = 2;
5354
5355 /* Remember an explicit \r or \n, and add the range to the class. */
5356
5357 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5358
5359 class_has_8bitchar +=
5360 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5361
5362 continue; /* Go get the next char in the class */
5363 }
5364
5365 /* Handle a single character - we can get here for a normal non-escape
5366 char, or after \ that introduces a single character or for an apparent
5367 range that isn't. Only the value 1 matters for class_one_char, so don't
5368 increase it if it is already 2 or more ... just in case there's a class
5369 with a zillion characters in it. */
5370
5371 CLASS_SINGLE_CHARACTER:
5372 if (class_one_char < 2) class_one_char++;
5373
5374 /* If class_one_char is 1, we have the first single character in the
5375 class, and there have been no prior ranges, or XCLASS items generated by
5376 escapes. If this is the final character in the class, we can optimize by
5377 turning the item into a 1-character OP_CHAR[I] if it's positive, or
5378 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5379 to be set. Otherwise, there can be no first char if this item is first,
5380 whatever repeat count may follow. In the case of reqchar, save the
5381 previous value for reinstating. */
5382
5383 if (!inescq && class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5384 {
5385 ptr++;
5386 zeroreqchar = reqchar;
5387 zeroreqcharflags = reqcharflags;
5388
5389 if (negate_class)
5390 {
5391 #ifdef SUPPORT_UCP
5392 int d;
5393 #endif
5394 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5395 zerofirstchar = firstchar;
5396 zerofirstcharflags = firstcharflags;
5397
5398 /* For caseless UTF-8 mode when UCP support is available, check
5399 whether this character has more than one other case. If so, generate
5400 a special OP_NOTPROP item instead of OP_NOTI. */
5401
5402 #ifdef SUPPORT_UCP
5403 if (utf && (options & PCRE_CASELESS) != 0 &&
5404 (d = UCD_CASESET(c)) != 0)
5405 {
5406 *code++ = OP_NOTPROP;
5407 *code++ = PT_CLIST;
5408 *code++ = d;
5409 }
5410 else
5411 #endif
5412 /* Char has only one other case, or UCP not available */
5413
5414 {
5415 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5416 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5417 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5418 code += PRIV(ord2utf)(c, code);
5419 else
5420 #endif
5421 *code++ = c;
5422 }
5423
5424 /* We are finished with this character class */
5425
5426 goto END_CLASS;
5427 }
5428
5429 /* For a single, positive character, get the value into mcbuffer, and
5430 then we can handle this with the normal one-character code. */
5431
5432 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5433 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5434 mclength = PRIV(ord2utf)(c, mcbuffer);
5435 else
5436 #endif
5437 {
5438 mcbuffer[0] = c;
5439 mclength = 1;
5440 }
5441 goto ONE_CHAR;
5442 } /* End of 1-char optimization */
5443
5444 /* There is more than one character in the class, or an XCLASS item
5445 has been generated. Add this character to the class. */
5446
5447 class_has_8bitchar +=
5448 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5449 }
5450
5451 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5452 If we are at the end of an internal nested string, revert to the outer
5453 string. */
5454
5455 while (((c = *(++ptr)) != CHAR_NULL ||
5456 (nestptr != NULL &&
5457 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5458 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5459
5460 /* Check for missing terminating ']' */
5461
5462 if (c == CHAR_NULL)
5463 {
5464 *errorcodeptr = ERR6;
5465 goto FAILED;
5466 }
5467
5468 /* We will need an XCLASS if data has been placed in class_uchardata. In
5469 the second phase this is a sufficient test. However, in the pre-compile
5470 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5471 only if the very last character in the class needs XCLASS will it contain
5472 anything at this point. For this reason, xclass gets set TRUE above when
5473 uchar_classdata is emptied, and that's why this code is the way it is here
5474 instead of just doing a test on class_uchardata below. */
5475
5476 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5477 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5478 #endif
5479
5480 /* If this is the first thing in the branch, there can be no first char
5481 setting, whatever the repeat count. Any reqchar setting must remain
5482 unchanged after any kind of repeat. */
5483
5484 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5485 zerofirstchar = firstchar;
5486 zerofirstcharflags = firstcharflags;
5487 zeroreqchar = reqchar;
5488 zeroreqcharflags = reqcharflags;
5489
5490 /* If there are characters with values > 255, we have to compile an
5491 extended class, with its own opcode, unless there was a negated special
5492 such as \S in the class, and PCRE_UCP is not set, because in that case all
5493 characters > 255 are in the class, so any that were explicitly given as
5494 well can be ignored. If (when there are explicit characters > 255 that must
5495 be listed) there are no characters < 256, we can omit the bitmap in the
5496 actual compiled code. */
5497
5498 #ifdef SUPPORT_UTF
5499 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5500 #elif !defined COMPILE_PCRE8
5501 if (xclass && !should_flip_negation)
5502 #endif
5503 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5504 {
5505 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5506 *code++ = OP_XCLASS;
5507 code += LINK_SIZE;
5508 *code = negate_class? XCL_NOT:0;
5509 if (xclass_has_prop) *code |= XCL_HASPROP;
5510
5511 /* If the map is required, move up the extra data to make room for it;
5512 otherwise just move the code pointer to the end of the extra data. */
5513
5514 if (class_has_8bitchar > 0)
5515 {
5516 *code++ |= XCL_MAP;
5517 memmove(code + (32 / sizeof(pcre_uchar)), code,
5518 IN_UCHARS(class_uchardata - code));
5519 if (negate_class && !xclass_has_prop)
5520 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5521 memcpy(code, classbits, 32);
5522 code = class_uchardata + (32 / sizeof(pcre_uchar));
5523 }
5524 else code = class_uchardata;
5525
5526 /* Now fill in the complete length of the item */
5527
5528 PUT(previous, 1, (int)(code - previous));
5529 break; /* End of class handling */
5530 }
5531
5532 /* Even though any XCLASS list is now discarded, we must allow for
5533 its memory. */
5534
5535 if (lengthptr != NULL)
5536 *lengthptr += (int)(class_uchardata - class_uchardata_base);
5537 #endif
5538
5539 /* If there are no characters > 255, or they are all to be included or
5540 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5541 whole class was negated and whether there were negative specials such as \S
5542 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5543 negating it if necessary. */
5544
5545 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5546 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5547 {
5548 if (negate_class)
5549 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5550 memcpy(code, classbits, 32);
5551 }
5552 code += 32 / sizeof(pcre_uchar);
5553
5554 END_CLASS:
5555 break;
5556
5557
5558 /* ===================================================================*/
5559 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5560 has been tested above. */
5561
5562 case CHAR_LEFT_CURLY_BRACKET:
5563 if (!is_quantifier) goto NORMAL_CHAR;
5564 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5565 if (*errorcodeptr != 0) goto FAILED;
5566 goto REPEAT;
5567
5568 case CHAR_ASTERISK:
5569 repeat_min = 0;
5570 repeat_max = -1;
5571 goto REPEAT;
5572
5573 case CHAR_PLUS:
5574 repeat_min = 1;
5575 repeat_max = -1;
5576 goto REPEAT;
5577
5578 case CHAR_QUESTION_MARK:
5579 repeat_min = 0;
5580 repeat_max = 1;
5581
5582 REPEAT:
5583 if (previous == NULL)
5584 {
5585 *errorcodeptr = ERR9;
5586 goto FAILED;
5587 }
5588
5589 if (repeat_min == 0)
5590 {
5591 firstchar = zerofirstchar; /* Adjust for zero repeat */
5592 firstcharflags = zerofirstcharflags;
5593 reqchar = zeroreqchar; /* Ditto */
5594 reqcharflags = zeroreqcharflags;
5595 }
5596
5597 /* Remember whether this is a variable length repeat */
5598
5599 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5600
5601 op_type = 0; /* Default single-char op codes */
5602 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5603
5604 /* Save start of previous item, in case we have to move it up in order to
5605 insert something before it. */
5606
5607 tempcode = previous;
5608
5609 /* Before checking for a possessive quantifier, we must skip over
5610 whitespace and comments in extended mode because Perl allows white space at
5611 this point. */
5612
5613 if ((options & PCRE_EXTENDED) != 0)
5614 {
5615 const pcre_uchar *p = ptr + 1;
5616 for (;;)
5617 {
5618 while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5619 if (*p != CHAR_NUMBER_SIGN) break;
5620 p++;
5621 while (*p != CHAR_NULL)
5622 {
5623 if (IS_NEWLINE(p)) /* For non-fixed-length newline cases, */
5624 { /* IS_NEWLINE sets cd->nllen. */
5625 p += cd->nllen;
5626 break;
5627 }
5628 p++;
5629 #ifdef SUPPORT_UTF
5630 if (utf) FORWARDCHAR(p);
5631 #endif
5632 } /* Loop for comment characters */
5633 } /* Loop for multiple comments */
5634 ptr = p - 1; /* Character before the next significant one. */
5635 }
5636
5637 /* If the next character is '+', we have a possessive quantifier. This
5638 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5639 If the next character is '?' this is a minimizing repeat, by default,
5640 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5641 repeat type to the non-default. */
5642
5643 if (ptr[1] == CHAR_PLUS)
5644 {
5645 repeat_type = 0; /* Force greedy */
5646 possessive_quantifier = TRUE;
5647 ptr++;
5648 }
5649 else if (ptr[1] == CHAR_QUESTION_MARK)
5650 {
5651 repeat_type = greedy_non_default;
5652 ptr++;
5653 }
5654 else repeat_type = greedy_default;
5655
5656 /* If previous was a recursion call, wrap it in atomic brackets so that
5657 previous becomes the atomic group. All recursions were so wrapped in the
5658 past, but it no longer happens for non-repeated recursions. In fact, the
5659 repeated ones could be re-implemented independently so as not to need this,
5660 but for the moment we rely on the code for repeating groups. */
5661
5662 if (*previous == OP_RECURSE)
5663 {
5664 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5665 *previous = OP_ONCE;
5666 PUT(previous, 1, 2 + 2*LINK_SIZE);
5667 previous[2 + 2*LINK_SIZE] = OP_KET;
5668 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5669 code += 2 + 2 * LINK_SIZE;
5670 length_prevgroup = 3 + 3*LINK_SIZE;
5671
5672 /* When actually compiling, we need to check whether this was a forward
5673 reference, and if so, adjust the offset. */
5674
5675 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5676 {
5677 int offset = GET(cd->hwm, -LINK_SIZE);
5678 if (offset == previous + 1 - cd->start_code)
5679 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5680 }
5681 }
5682
5683 /* Now handle repetition for the different types of item. */
5684
5685 /* If previous was a character or negated character match, abolish the item
5686 and generate a repeat item instead. If a char item has a minimum of more
5687 than one, ensure that it is set in reqchar - it might not be if a sequence
5688 such as x{3} is the first thing in a branch because the x will have gone
5689 into firstchar instead. */
5690
5691 if (*previous == OP_CHAR || *previous == OP_CHARI
5692 || *previous == OP_NOT || *previous == OP_NOTI)
5693 {
5694 switch (*previous)
5695 {
5696 default: /* Make compiler happy. */
5697 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5698 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5699 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5700 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5701 }
5702
5703 /* Deal with UTF characters that take up more than one character. It's
5704 easier to write this out separately than try to macrify it. Use c to
5705 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5706 it's a length rather than a small character. */
5707
5708 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5709 if (utf && NOT_FIRSTCHAR(code[-1]))
5710 {
5711 pcre_uchar *lastchar = code - 1;
5712 BACKCHAR(lastchar);
5713 c = (int)(code - lastchar); /* Length of UTF-8 character */
5714 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5715 c |= UTF_LENGTH; /* Flag c as a length */
5716 }
5717 else
5718 #endif /* SUPPORT_UTF */
5719
5720 /* Handle the case of a single charater - either with no UTF support, or
5721 with UTF disabled, or for a single character UTF character. */
5722 {
5723 c = code[-1];
5724 if (*previous <= OP_CHARI && repeat_min > 1)
5725 {
5726 reqchar = c;
5727 reqcharflags = req_caseopt | cd->req_varyopt;
5728 }
5729 }
5730
5731 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5732 }
5733
5734 /* If previous was a character type match (\d or similar), abolish it and
5735 create a suitable repeat item. The code is shared with single-character
5736 repeats by setting op_type to add a suitable offset into repeat_type. Note
5737 the the Unicode property types will be present only when SUPPORT_UCP is
5738 defined, but we don't wrap the little bits of code here because it just
5739 makes it horribly messy. */
5740
5741 else if (*previous < OP_EODN)
5742 {
5743 pcre_uchar *oldcode;
5744 int prop_type, prop_value;
5745 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5746 c = *previous;
5747
5748 OUTPUT_SINGLE_REPEAT:
5749 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5750 {
5751 prop_type = previous[1];
5752 prop_value = previous[2];
5753 }
5754 else prop_type = prop_value = -1;
5755
5756 oldcode = code;
5757 code = previous; /* Usually overwrite previous item */
5758
5759 /* If the maximum is zero then the minimum must also be zero; Perl allows
5760 this case, so we do too - by simply omitting the item altogether. */
5761
5762 if (repeat_max == 0) goto END_REPEAT;
5763
5764 /* Combine the op_type with the repeat_type */
5765
5766 repeat_type += op_type;
5767
5768 /* A minimum of zero is handled either as the special case * or ?, or as
5769 an UPTO, with the maximum given. */
5770
5771 if (repeat_min == 0)
5772 {
5773 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5774 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5775 else
5776 {
5777 *code++ = OP_UPTO + repeat_type;
5778 PUT2INC(code, 0, repeat_max);
5779 }
5780 }
5781
5782 /* A repeat minimum of 1 is optimized into some special cases. If the
5783 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5784 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5785 one less than the maximum. */
5786
5787 else if (repeat_min == 1)
5788 {
5789 if (repeat_max == -1)
5790 *code++ = OP_PLUS + repeat_type;
5791 else
5792 {
5793 code = oldcode; /* leave previous item in place */
5794 if (repeat_max == 1) goto END_REPEAT;
5795 *code++ = OP_UPTO + repeat_type;
5796 PUT2INC(code, 0, repeat_max - 1);
5797 }
5798 }
5799
5800 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5801 handled as an EXACT followed by an UPTO. */
5802
5803 else
5804 {
5805 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5806 PUT2INC(code, 0, repeat_min);
5807
5808 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5809 we have to insert the character for the previous code. For a repeated
5810 Unicode property match, there are two extra bytes that define the
5811 required property. In UTF-8 mode, long characters have their length in
5812 c, with the UTF_LENGTH bit as a flag. */
5813
5814 if (repeat_max < 0)
5815 {
5816 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5817 if (utf && (c & UTF_LENGTH) != 0)
5818 {
5819 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5820 code += c & 7;
5821 }
5822 else
5823 #endif
5824 {
5825 *code++ = c;
5826 if (prop_type >= 0)
5827 {
5828 *code++ = prop_type;
5829 *code++ = prop_value;
5830 }
5831 }
5832 *code++ = OP_STAR + repeat_type;
5833 }
5834
5835 /* Else insert an UPTO if the max is greater than the min, again
5836 preceded by the character, for the previously inserted code. If the
5837 UPTO is just for 1 instance, we can use QUERY instead. */
5838
5839 else if (repeat_max != repeat_min)
5840 {
5841 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5842 if (utf && (c & UTF_LENGTH) != 0)
5843 {
5844 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5845 code += c & 7;
5846 }
5847 else
5848 #endif
5849 *code++ = c;
5850 if (prop_type >= 0)
5851 {
5852 *code++ = prop_type;
5853 *code++ = prop_value;
5854 }
5855 repeat_max -= repeat_min;
5856
5857 if (repeat_max == 1)
5858 {
5859 *code++ = OP_QUERY + repeat_type;
5860 }
5861 else
5862 {
5863 *code++ = OP_UPTO + repeat_type;
5864 PUT2INC(code, 0, repeat_max);
5865 }
5866 }
5867 }
5868
5869 /* The character or character type itself comes last in all cases. */
5870
5871 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5872 if (utf && (c & UTF_LENGTH) != 0)
5873 {
5874 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5875 code += c & 7;
5876 }
5877 else
5878 #endif
5879 *code++ = c;
5880
5881 /* For a repeated Unicode property match, there are two extra bytes that
5882 define the required property. */
5883
5884 #ifdef SUPPORT_UCP
5885 if (prop_type >= 0)
5886 {
5887 *code++ = prop_type;
5888 *code++ = prop_value;
5889 }
5890 #endif
5891 }
5892
5893 /* If previous was a character class or a back reference, we put the repeat
5894 stuff after it, but just skip the item if the repeat was {0,0}. */
5895
5896 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5897 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5898 *previous == OP_XCLASS ||
5899 #endif
5900 *previous == OP_REF || *previous == OP_REFI ||
5901 *previous == OP_DNREF || *previous == OP_DNREFI)
5902 {
5903 if (repeat_max == 0)
5904 {
5905 code = previous;
5906 goto END_REPEAT;
5907 }
5908
5909 if (repeat_min == 0 && repeat_max == -1)
5910 *code++ = OP_CRSTAR + repeat_type;
5911 else if (repeat_min == 1 && repeat_max == -1)
5912 *code++ = OP_CRPLUS + repeat_type;
5913 else if (repeat_min == 0 && repeat_max == 1)
5914 *code++ = OP_CRQUERY + repeat_type;
5915 else
5916 {
5917 *code++ = OP_CRRANGE + repeat_type;
5918 PUT2INC(code, 0, repeat_min);
5919 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5920 PUT2INC(code, 0, repeat_max);
5921 }
5922 }
5923
5924 /* If previous was a bracket group, we may have to replicate it in certain
5925 cases. Note that at this point we can encounter only the "basic" bracket
5926 opcodes such as BRA and CBRA, as this is the place where they get converted
5927 into the more special varieties such as BRAPOS and SBRA. A test for >=
5928 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5929 ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5930 Originally, PCRE did not allow repetition of assertions, but now it does,
5931 for Perl compatibility. */
5932
5933 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5934 {
5935 register int i;
5936 int len = (int)(code - previous);
5937 size_t base_hwm_offset = item_hwm_offset;
5938 pcre_uchar *bralink = NULL;
5939 pcre_uchar *brazeroptr = NULL;
5940
5941 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5942 we just ignore the repeat. */
5943
5944 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5945 goto END_REPEAT;
5946
5947 /* There is no sense in actually repeating assertions. The only potential
5948 use of repetition is in cases when the assertion is optional. Therefore,
5949 if the minimum is greater than zero, just ignore the repeat. If the
5950 maximum is not zero or one, set it to 1. */
5951
5952 if (*previous < OP_ONCE) /* Assertion */
5953 {
5954 if (repeat_min > 0) goto END_REPEAT;
5955 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5956 }
5957
5958 /* The case of a zero minimum is special because of the need to stick
5959 OP_BRAZERO in front of it, and because the group appears once in the
5960 data, whereas in other cases it appears the minimum number of times. For
5961 this reason, it is simplest to treat this case separately, as otherwise
5962 the code gets far too messy. There are several special subcases when the
5963 minimum is zero. */
5964
5965 if (repeat_min == 0)
5966 {
5967 /* If the maximum is also zero, we used to just omit the group from the
5968 output altogether, like this:
5969
5970 ** if (repeat_max == 0)
5971 ** {
5972 ** code = previous;
5973 ** goto END_REPEAT;
5974 ** }
5975
5976 However, that fails when a group or a subgroup within it is referenced
5977 as a subroutine from elsewhere in the pattern, so now we stick in
5978 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5979 don't have a list of which groups are referenced, we cannot do this
5980 selectively.
5981
5982 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5983 and do no more at this point. However, we do need to adjust any
5984 OP_RECURSE calls inside the group that refer to the group itself or any
5985 internal or forward referenced group, because the offset is from the
5986 start of the whole regex. Temporarily terminate the pattern while doing
5987 this. */
5988
5989 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5990 {
5991 *code = OP_END;
5992 adjust_recurse(previous, 1, utf, cd, item_hwm_offset);
5993 memmove(previous + 1, previous, IN_UCHARS(len));
5994 code++;
5995 if (repeat_max == 0)
5996 {
5997 *previous++ = OP_SKIPZERO;
5998 goto END_REPEAT;
5999 }
6000 brazeroptr = previous; /* Save for possessive optimizing */
6001 *previous++ = OP_BRAZERO + repeat_type;
6002 }
6003
6004 /* If the maximum is greater than 1 and limited, we have to replicate
6005 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
6006 The first one has to be handled carefully because it's the original
6007 copy, which has to be moved up. The remainder can be handled by code
6008 that is common with the non-zero minimum case below. We have to
6009 adjust the value or repeat_max, since one less copy is required. Once
6010 again, we may have to adjust any OP_RECURSE calls inside the group. */
6011
6012 else
6013 {
6014 int offset;
6015 *code = OP_END;
6016 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, item_hwm_offset);
6017 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
6018 code += 2 + LINK_SIZE;
6019 *previous++ = OP_BRAZERO + repeat_type;
6020 *previous++ = OP_BRA;
6021
6022 /* We chain together the bracket offset fields that have to be
6023 filled in later when the ends of the brackets are reached. */
6024
6025 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
6026 bralink = previous;
6027 PUTINC(previous, 0, offset);
6028 }
6029
6030 repeat_max--;
6031 }
6032
6033 /* If the minimum is greater than zero, replicate the group as many
6034 times as necessary, and adjust the maximum to the number of subsequent
6035 copies that we need. If we set a first char from the group, and didn't
6036 set a required char, copy the latter from the former. If there are any
6037 forward reference subroutine calls in the group, there will be entries on
6038 the workspace list; replicate these with an appropriate increment. */
6039
6040 else
6041 {
6042 if (repeat_min > 1)
6043 {
6044 /* In the pre-compile phase, we don't actually do the replication. We
6045 just adjust the length as if we had. Do some paranoid checks for
6046 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
6047 integer type when available, otherwise double. */
6048
6049 if (lengthptr != NULL)
6050 {
6051 int delta = (repeat_min - 1)*length_prevgroup;
6052 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
6053 (INT64_OR_DOUBLE)length_prevgroup >
6054 (INT64_OR_DOUBLE)INT_MAX ||
6055 OFLOW_MAX - *lengthptr < delta)
6056 {
6057 *errorcodeptr = ERR20;
6058 goto FAILED;
6059 }
6060 *lengthptr += delta;
6061 }
6062
6063 /* This is compiling for real. If there is a set first byte for
6064 the group, and we have not yet set a "required byte", set it. Make
6065 sure there is enough workspace for copying forward references before
6066 doing the copy. */
6067
6068 else
6069 {
6070 if (groupsetfirstchar && reqcharflags < 0)
6071 {
6072 reqchar = firstchar;
6073 reqcharflags = firstcharflags;
6074 }
6075
6076 for (i = 1; i < repeat_min; i++)
6077 {
6078 pcre_uchar *hc;
6079 size_t this_hwm_offset = cd->hwm - cd->start_workspace;
6080 memcpy(code, previous, IN_UCHARS(len));
6081
6082 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6083 WORK_SIZE_SAFETY_MARGIN -
6084 (this_hwm_offset - base_hwm_offset))
6085 {
6086 *errorcodeptr = expand_workspace(cd);
6087 if (*errorcodeptr != 0) goto FAILED;
6088 }
6089
6090 for (hc = (pcre_uchar *)cd->start_workspace + base_hwm_offset;
6091 hc < (pcre_uchar *)cd->start_workspace + this_hwm_offset;
6092 hc += LINK_SIZE)
6093 {
6094 PUT(cd->hwm, 0, GET(hc, 0) + len);
6095 cd->hwm += LINK_SIZE;
6096 }
6097 base_hwm_offset = this_hwm_offset;
6098 code += len;
6099 }
6100 }
6101 }
6102
6103 if (repeat_max > 0) repeat_max -= repeat_min;
6104 }
6105
6106 /* This code is common to both the zero and non-zero minimum cases. If
6107 the maximum is limited, it replicates the group in a nested fashion,
6108 remembering the bracket starts on a stack. In the case of a zero minimum,
6109 the first one was set up above. In all cases the repeat_max now specifies
6110 the number of additional copies needed. Again, we must remember to
6111 replicate entries on the forward reference list. */
6112
6113 if (repeat_max >= 0)
6114 {
6115 /* In the pre-compile phase, we don't actually do the replication. We
6116 just adjust the length as if we had. For each repetition we must add 1
6117 to the length for BRAZERO and for all but the last repetition we must
6118 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
6119 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
6120 a 64-bit integer type when available, otherwise double. */
6121
6122 if (lengthptr != NULL && repeat_max > 0)
6123 {
6124 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
6125 2 - 2*LINK_SIZE; /* Last one doesn't nest */
6126 if ((INT64_OR_DOUBLE)repeat_max *
6127 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
6128 > (INT64_OR_DOUBLE)INT_MAX ||
6129 OFLOW_MAX - *lengthptr < delta)
6130 {
6131 *errorcodeptr = ERR20;
6132 goto FAILED;
6133 }
6134 *lengthptr += delta;
6135 }
6136
6137 /* This is compiling for real */
6138
6139 else for (i = repeat_max - 1; i >= 0; i--)
6140 {
6141 pcre_uchar *hc;
6142 size_t this_hwm_offset = cd->hwm - cd->start_workspace;
6143
6144 *code++ = OP_BRAZERO + repeat_type;
6145
6146 /* All but the final copy start a new nesting, maintaining the
6147 chain of brackets outstanding. */
6148
6149 if (i != 0)
6150 {
6151 int offset;
6152 *code++ = OP_BRA;
6153 offset = (bralink == NULL)? 0 : (int)(code - bralink);
6154 bralink = code;
6155 PUTINC(code, 0, offset);
6156 }
6157
6158 memcpy(code, previous, IN_UCHARS(len));
6159
6160 /* Ensure there is enough workspace for forward references before
6161 copying them. */
6162
6163 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6164 WORK_SIZE_SAFETY_MARGIN -
6165 (this_hwm_offset - base_hwm_offset))
6166 {
6167 *errorcodeptr = expand_workspace(cd);
6168 if (*errorcodeptr != 0) goto FAILED;
6169 }
6170
6171 for (hc = (pcre_uchar *)cd->start_workspace + base_hwm_offset;
6172 hc < (pcre_uchar *)cd->start_workspace + this_hwm_offset;
6173 hc += LINK_SIZE)
6174 {
6175 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
6176 cd->hwm += LINK_SIZE;
6177 }
6178 base_hwm_offset = this_hwm_offset;
6179 code += len;
6180 }
6181
6182 /* Now chain through the pending brackets, and fill in their length
6183 fields (which are holding the chain links pro tem). */
6184
6185 while (bralink != NULL)
6186 {
6187 int oldlinkoffset;
6188 int offset = (int)(code - bralink + 1);
6189 pcre_uchar *bra = code - offset;
6190 oldlinkoffset = GET(bra, 1);
6191 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
6192 *code++ = OP_KET;
6193 PUTINC(code, 0, offset);
6194 PUT(bra, 1, offset);
6195 }
6196 }
6197
6198 /* If the maximum is unlimited, set a repeater in the final copy. For
6199 ONCE brackets, that's all we need to do. However, possessively repeated
6200 ONCE brackets can be converted into non-capturing brackets, as the
6201 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
6202 deal with possessive ONCEs specially.
6203
6204 Otherwise, when we are doing the actual compile phase, check to see
6205 whether this group is one that could match an empty string. If so,
6206 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
6207 that runtime checking can be done. [This check is also applied to ONCE
6208 groups at runtime, but in a different way.]
6209
6210 Then, if the quantifier was possessive and the bracket is not a
6211 conditional, we convert the BRA code to the POS form, and the KET code to
6212 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
6213 subpattern at both the start and at the end.) The use of special opcodes
6214 makes it possible to reduce greatly the stack usage in pcre_exec(). If
6215 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
6216
6217 Then, if the minimum number of matches is 1 or 0, cancel the possessive
6218 flag so that the default action below, of wrapping everything inside
6219 atomic brackets, does not happen. When the minimum is greater than 1,
6220 there will be earlier copies of the group, and so we still have to wrap
6221 the whole thing. */
6222
6223 else
6224 {
6225 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
6226 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
6227
6228 /* Convert possessive ONCE brackets to non-capturing */
6229
6230 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
6231 possessive_quantifier) *bracode = OP_BRA;
6232
6233 /* For non-possessive ONCE brackets, all we need to do is to
6234 set the KET. */
6235
6236 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
6237 *ketcode = OP_KETRMAX + repeat_type;
6238
6239 /* Handle non-ONCE brackets and possessive ONCEs (which have been
6240 converted to non-capturing above). */
6241
6242 else
6243 {
6244 /* In the compile phase, check for empty string matching. */
6245
6246 if (lengthptr == NULL)
6247 {
6248 pcre_uchar *scode = bracode;
6249 do
6250 {
6251 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6252 {
6253 *bracode += OP_SBRA - OP_BRA;
6254 break;
6255 }
6256 scode += GET(scode, 1);
6257 }
6258 while (*scode == OP_ALT);
6259 }
6260
6261 /* Handle possessive quantifiers. */
6262
6263 if (possessive_quantifier)
6264 {
6265 /* For COND brackets, we wrap the whole thing in a possessively
6266 repeated non-capturing bracket, because we have not invented POS
6267 versions of the COND opcodes. Because we are moving code along, we
6268 must ensure that any pending recursive references are updated. */
6269
6270 if (*bracode == OP_COND || *bracode == OP_SCOND)
6271 {
6272 int nlen = (int)(code - bracode);
6273 *code = OP_END;
6274 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, item_hwm_offset);
6275 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6276 code += 1 + LINK_SIZE;
6277 nlen += 1 + LINK_SIZE;
6278 *bracode = OP_BRAPOS;
6279 *code++ = OP_KETRPOS;
6280 PUTINC(code, 0, nlen);
6281 PUT(bracode, 1, nlen);
6282 }
6283
6284 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6285
6286 else
6287 {
6288 *bracode += 1; /* Switch to xxxPOS opcodes */
6289 *ketcode = OP_KETRPOS;
6290 }
6291
6292 /* If the minimum is zero, mark it as possessive, then unset the
6293 possessive flag when the minimum is 0 or 1. */
6294
6295 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6296 if (repeat_min < 2) possessive_quantifier = FALSE;
6297 }
6298
6299 /* Non-possessive quantifier */
6300
6301 else *ketcode = OP_KETRMAX + repeat_type;
6302 }
6303 }
6304 }
6305
6306 /* If previous is OP_FAIL, it was generated by an empty class [] in
6307 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6308 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6309 error above. We can just ignore the repeat in JS case. */
6310
6311 else if (*previous == OP_FAIL) goto END_REPEAT;
6312
6313 /* Else there's some kind of shambles */
6314
6315 else
6316 {
6317 *errorcodeptr = ERR11;
6318 goto FAILED;
6319 }
6320
6321 /* If the character following a repeat is '+', possessive_quantifier is
6322 TRUE. For some opcodes, there are special alternative opcodes for this
6323 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6324 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6325 Sun's Java package, but the special opcodes can optimize it.
6326
6327 Some (but not all) possessively repeated subpatterns have already been
6328 completely handled in the code just above. For them, possessive_quantifier
6329 is always FALSE at this stage. Note that the repeated item starts at
6330 tempcode, not at previous, which might be the first part of a string whose
6331 (former) last char we repeated. */
6332
6333 if (possessive_quantifier)
6334 {
6335 int len;
6336
6337 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6338 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6339 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6340 remains is greater than zero, there's a further opcode that can be
6341 handled. If not, do nothing, leaving the EXACT alone. */
6342
6343 switch(*tempcode)
6344 {
6345 case OP_TYPEEXACT:
6346 tempcode += PRIV(OP_lengths)[*tempcode] +
6347 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6348 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6349 break;
6350
6351 /* CHAR opcodes are used for exacts whose count is 1. */
6352
6353 case OP_CHAR:
6354 case OP_CHARI:
6355 case OP_NOT:
6356 case OP_NOTI:
6357 case OP_EXACT:
6358 case OP_EXACTI:
6359 case OP_NOTEXACT:
6360 case OP_NOTEXACTI:
6361 tempcode += PRIV(OP_lengths)[*tempcode];
6362 #ifdef SUPPORT_UTF
6363 if (utf && HAS_EXTRALEN(tempcode[-1]))
6364 tempcode += GET_EXTRALEN(tempcode[-1]);
6365 #endif
6366 break;
6367
6368 /* For the class opcodes, the repeat operator appears at the end;
6369 adjust tempcode to point to it. */
6370
6371 case OP_CLASS:
6372 case OP_NCLASS:
6373 tempcode += 1 + 32/sizeof(pcre_uchar);
6374 break;
6375
6376 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6377 case OP_XCLASS:
6378 tempcode += GET(tempcode, 1);
6379 break;
6380 #endif
6381 }
6382
6383 /* If tempcode is equal to code (which points to the end of the repeated
6384 item), it means we have skipped an EXACT item but there is no following
6385 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6386 all other cases, tempcode will be pointing to the repeat opcode, and will
6387 be less than code, so the value of len will be greater than 0. */
6388
6389 len = (int)(code - tempcode);
6390 if (len > 0)
6391 {
6392 unsigned int repcode = *tempcode;
6393
6394 /* There is a table for possessifying opcodes, all of which are less
6395 than OP_CALLOUT. A zero entry means there is no possessified version.
6396 */
6397
6398 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6399 *tempcode = opcode_possessify[repcode];
6400
6401 /* For opcode without a special possessified version, wrap the item in
6402 ONCE brackets. Because we are moving code along, we must ensure that any
6403 pending recursive references are updated. */
6404
6405 else
6406 {
6407 *code = OP_END;
6408 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, item_hwm_offset);
6409 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6410 code += 1 + LINK_SIZE;
6411 len += 1 + LINK_SIZE;
6412 tempcode[0] = OP_ONCE;
6413 *code++ = OP_KET;
6414 PUTINC(code, 0, len);
6415 PUT(tempcode, 1, len);
6416 }
6417 }
6418
6419 #ifdef NEVER
6420 if (len > 0) switch (*tempcode)
6421 {
6422 case OP_STAR: *tempcode = OP_POSSTAR; break;
6423 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6424 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6425 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6426
6427 case OP_STARI: *tempcode = OP_POSSTARI; break;
6428 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6429 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6430 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6431
6432 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6433 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6434 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6435 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6436
6437 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6438 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6439 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6440 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6441
6442 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6443 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6444 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6445 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6446
6447 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6448 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6449 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6450 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6451
6452 /* Because we are moving code along, we must ensure that any
6453 pending recursive references are updated. */
6454
6455 default:
6456 *code = OP_END;
6457 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, item_hwm_offset);
6458 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6459 code += 1 + LINK_SIZE;
6460 len += 1 + LINK_SIZE;
6461 tempcode[0] = OP_ONCE;
6462 *code++ = OP_KET;
6463 PUTINC(code, 0, len);
6464 PUT(tempcode, 1, len);
6465 break;
6466 }
6467 #endif
6468 }
6469
6470 /* In all case we no longer have a previous item. We also set the
6471 "follows varying string" flag for subsequently encountered reqchars if
6472 it isn't already set and we have just passed a varying length item. */
6473
6474 END_REPEAT:
6475 previous = NULL;
6476 cd->req_varyopt |= reqvary;
6477 break;
6478
6479
6480 /* ===================================================================*/
6481 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6482 lookbehind or option setting or condition or all the other extended
6483 parenthesis forms. */
6484
6485 case CHAR_LEFT_PARENTHESIS:
6486 ptr++;
6487
6488 /* First deal with comments. Putting this code right at the start ensures
6489 that comments have no bad side effects. */
6490
6491 if (ptr[0] == CHAR_QUESTION_MARK && ptr[1] == CHAR_NUMBER_SIGN)
6492 {
6493 ptr += 2;
6494 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6495 if (*ptr == CHAR_NULL)
6496 {
6497 *errorcodeptr = ERR18;
6498 goto FAILED;
6499 }
6500 continue;
6501 }
6502
6503 /* Now deal with various "verbs" that can be introduced by '*'. */
6504
6505 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6506 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6507 {
6508 int i, namelen;
6509 int arglen = 0;
6510 const char *vn = verbnames;
6511 const pcre_uchar *name = ptr + 1;
6512 const pcre_uchar *arg = NULL;
6513 previous = NULL;
6514 ptr++;
6515 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6516 namelen = (int)(ptr - name);
6517
6518 /* It appears that Perl allows any characters whatsoever, other than
6519 a closing parenthesis, to appear in arguments, so we no longer insist on
6520 letters, digits, and underscores. */
6521
6522 if (*ptr == CHAR_COLON)
6523 {
6524 arg = ++ptr;
6525 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6526 arglen = (int)(ptr - arg);
6527 if ((unsigned int)arglen > MAX_MARK)
6528 {
6529 *errorcodeptr = ERR75;
6530 goto FAILED;
6531 }
6532 }
6533
6534 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6535 {
6536 *errorcodeptr = ERR60;
6537 goto FAILED;
6538 }
6539
6540 /* Scan the table of verb names */
6541
6542 for (i = 0; i < verbcount; i++)
6543 {
6544 if (namelen == verbs[i].len &&
6545 STRNCMP_UC_C8(name, vn, namelen) == 0)
6546 {
6547 int setverb;
6548
6549 /* Check for open captures before ACCEPT and convert it to
6550 ASSERT_ACCEPT if in an assertion. */
6551
6552 if (verbs[i].op == OP_ACCEPT)
6553 {
6554 open_capitem *oc;
6555 if (arglen != 0)
6556 {
6557 *errorcodeptr = ERR59;
6558 goto FAILED;
6559 }
6560 cd->had_accept = TRUE;
6561 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6562 {
6563 *code++ = OP_CLOSE;
6564 PUT2INC(code, 0, oc->number);
6565 }
6566 setverb = *code++ =
6567 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6568
6569 /* Do not set firstchar after *ACCEPT */
6570 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6571 }
6572
6573 /* Handle other cases with/without an argument */
6574
6575 else if (arglen == 0)
6576 {
6577 if (verbs[i].op < 0) /* Argument is mandatory */
6578 {
6579 *errorcodeptr = ERR66;
6580 goto FAILED;
6581 }
6582 setverb = *code++ = verbs[i].op;
6583 }
6584
6585 else
6586 {
6587 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6588 {
6589 *errorcodeptr = ERR59;
6590 goto FAILED;
6591 }
6592 setverb = *code++ = verbs[i].op_arg;
6593 *code++ = arglen;
6594 memcpy(code, arg, IN_UCHARS(arglen));
6595 code += arglen;
6596 *code++ = 0;
6597 }
6598
6599 switch (setverb)
6600 {
6601 case OP_THEN:
6602 case OP_THEN_ARG:
6603 cd->external_flags |= PCRE_HASTHEN;
6604 break;
6605
6606 case OP_PRUNE:
6607 case OP_PRUNE_ARG:
6608 case OP_SKIP:
6609 case OP_SKIP_ARG:
6610 cd->had_pruneorskip = TRUE;
6611 break;
6612 }
6613
6614 break; /* Found verb, exit loop */
6615 }
6616
6617 vn += verbs[i].len + 1;
6618 }
6619
6620 if (i < verbcount) continue; /* Successfully handled a verb */
6621 *errorcodeptr = ERR60; /* Verb not recognized */
6622 goto FAILED;
6623 }
6624
6625 /* Initialize for "real" parentheses */
6626
6627 newoptions = options;
6628 skipbytes = 0;
6629 bravalue = OP_CBRA;
6630 item_hwm_offset = cd->hwm - cd->start_workspace;
6631 reset_bracount = FALSE;
6632
6633 /* Deal with the extended parentheses; all are introduced by '?', and the
6634 appearance of any of them means that this is not a capturing group. */
6635
6636 if (*ptr == CHAR_QUESTION_MARK)
6637 {
6638 int i, set, unset, namelen;
6639 int *optset;
6640 const pcre_uchar *name;
6641 pcre_uchar *slot;
6642
6643 switch (*(++ptr))
6644 {
6645 /* ------------------------------------------------------------ */
6646 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6647 reset_bracount = TRUE;
6648 /* Fall through */
6649
6650 /* ------------------------------------------------------------ */
6651 case CHAR_COLON: /* Non-capturing bracket */
6652 bravalue = OP_BRA;
6653 ptr++;
6654 break;
6655
6656
6657 /* ------------------------------------------------------------ */
6658 case CHAR_LEFT_PARENTHESIS: