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

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1408 - (show annotations)
Tue Dec 3 16:27:00 2013 UTC (6 years, 2 months ago) by ph10
File MIME type: text/plain
File size: 309239 byte(s)
Add support for [[:<:]] and [[:>:]] as a transition aid.
1 /*************************************************
2 * Perl-Compatible Regular Expressions *
3 *************************************************/
4
5 /* PCRE is a library of functions to support regular expressions whose syntax
6 and semantics are as close as possible to those of the Perl 5 language.
7
8 Written by Philip Hazel
9 Copyright (c) 1997-2013 University of Cambridge
10
11 -----------------------------------------------------------------------------
12 Redistribution and use in source and binary forms, with or without
13 modification, are permitted provided that the following conditions are met:
14
15 * Redistributions of source code must retain the above copyright notice,
16 this list of conditions and the following disclaimer.
17
18 * Redistributions in binary form must reproduce the above copyright
19 notice, this list of conditions and the following disclaimer in the
20 documentation and/or other materials provided with the distribution.
21
22 * Neither the name of the University of Cambridge nor the names of its
23 contributors may be used to endorse or promote products derived from
24 this software without specific prior written permission.
25
26 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
27 AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
28 IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
29 ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
30 LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
31 CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
32 SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
33 INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
34 CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
35 ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
36 POSSIBILITY OF SUCH DAMAGE.
37 -----------------------------------------------------------------------------
38 */
39
40
41 /* This module contains the external function pcre_compile(), along with
42 supporting internal functions that are not used by other modules. */
43
44
45 #ifdef HAVE_CONFIG_H
46 #include "config.h"
47 #endif
48
49 #define NLBLOCK cd /* Block containing newline information */
50 #define PSSTART start_pattern /* Field containing processed string start */
51 #define PSEND end_pattern /* Field containing processed string end */
52
53 #include "pcre_internal.h"
54
55
56 /* When PCRE_DEBUG is defined, we need the pcre(16|32)_printint() function, which
57 is also used by pcretest. PCRE_DEBUG is not defined when building a production
58 library. We do not need to select pcre16_printint.c specially, because the
59 COMPILE_PCREx macro will already be appropriately set. */
60
61 #ifdef PCRE_DEBUG
62 /* pcre_printint.c should not include any headers */
63 #define PCRE_INCLUDED
64 #include "pcre_printint.c"
65 #undef PCRE_INCLUDED
66 #endif
67
68
69 /* Macro for setting individual bits in class bitmaps. */
70
71 #define SETBIT(a,b) a[(b)/8] |= (1 << ((b)&7))
72
73 /* Maximum length value to check against when making sure that the integer that
74 holds the compiled pattern length does not overflow. We make it a bit less than
75 INT_MAX to allow for adding in group terminating bytes, so that we don't have
76 to check them every time. */
77
78 #define OFLOW_MAX (INT_MAX - 20)
79
80 /* Definitions to allow mutual recursion */
81
82 static int
83 add_list_to_class(pcre_uint8 *, pcre_uchar **, int, compile_data *,
84 const pcre_uint32 *, unsigned int);
85
86 static BOOL
87 compile_regex(int, pcre_uchar **, const pcre_uchar **, int *, BOOL, BOOL, int, int,
88 pcre_uint32 *, pcre_int32 *, pcre_uint32 *, pcre_int32 *, branch_chain *,
89 compile_data *, int *);
90
91
92
93 /*************************************************
94 * Code parameters and static tables *
95 *************************************************/
96
97 /* This value specifies the size of stack workspace that is used during the
98 first pre-compile phase that determines how much memory is required. The regex
99 is partly compiled into this space, but the compiled parts are discarded as
100 soon as they can be, so that hopefully there will never be an overrun. The code
101 does, however, check for an overrun. The largest amount I've seen used is 218,
102 so this number is very generous.
103
104 The same workspace is used during the second, actual compile phase for
105 remembering forward references to groups so that they can be filled in at the
106 end. Each entry in this list occupies LINK_SIZE bytes, so even when LINK_SIZE
107 is 4 there is plenty of room for most patterns. However, the memory can get
108 filled up by repetitions of forward references, for example patterns like
109 /(?1){0,1999}(b)/, and one user did hit the limit. The code has been changed so
110 that the workspace is expanded using malloc() in this situation. The value
111 below is therefore a minimum, and we put a maximum on it for safety. The
112 minimum is now also defined in terms of LINK_SIZE so that the use of malloc()
113 kicks in at the same number of forward references in all cases. */
114
115 #define COMPILE_WORK_SIZE (2048*LINK_SIZE)
116 #define COMPILE_WORK_SIZE_MAX (100*COMPILE_WORK_SIZE)
117
118 /* This value determines the size of the initial vector that is used for
119 remembering named groups during the pre-compile. It is allocated on the stack,
120 but if it is too small, it is expanded using malloc(), in a similar way to the
121 workspace. The value is the number of slots in the list. */
122
123 #define NAMED_GROUP_LIST_SIZE 20
124
125 /* The overrun tests check for a slightly smaller size so that they detect the
126 overrun before it actually does run off the end of the data block. */
127
128 #define WORK_SIZE_SAFETY_MARGIN (100)
129
130 /* Private flags added to firstchar and reqchar. */
131
132 #define REQ_CASELESS (1 << 0) /* Indicates caselessness */
133 #define REQ_VARY (1 << 1) /* Reqchar followed non-literal item */
134 /* Negative values for the firstchar and reqchar flags */
135 #define REQ_UNSET (-2)
136 #define REQ_NONE (-1)
137
138 /* Repeated character flags. */
139
140 #define UTF_LENGTH 0x10000000l /* The char contains its length. */
141
142 /* Table for handling escaped characters in the range '0'-'z'. Positive returns
143 are simple data values; negative values are for special things like \d and so
144 on. Zero means further processing is needed (for things like \x), or the escape
145 is invalid. */
146
147 #ifndef EBCDIC
148
149 /* This is the "normal" table for ASCII systems or for EBCDIC systems running
150 in UTF-8 mode. */
151
152 static const short int escapes[] = {
153 0, 0,
154 0, 0,
155 0, 0,
156 0, 0,
157 0, 0,
158 CHAR_COLON, CHAR_SEMICOLON,
159 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN,
160 CHAR_GREATER_THAN_SIGN, CHAR_QUESTION_MARK,
161 CHAR_COMMERCIAL_AT, -ESC_A,
162 -ESC_B, -ESC_C,
163 -ESC_D, -ESC_E,
164 0, -ESC_G,
165 -ESC_H, 0,
166 0, -ESC_K,
167 0, 0,
168 -ESC_N, 0,
169 -ESC_P, -ESC_Q,
170 -ESC_R, -ESC_S,
171 0, 0,
172 -ESC_V, -ESC_W,
173 -ESC_X, 0,
174 -ESC_Z, CHAR_LEFT_SQUARE_BRACKET,
175 CHAR_BACKSLASH, CHAR_RIGHT_SQUARE_BRACKET,
176 CHAR_CIRCUMFLEX_ACCENT, CHAR_UNDERSCORE,
177 CHAR_GRAVE_ACCENT, 7,
178 -ESC_b, 0,
179 -ESC_d, ESC_e,
180 ESC_f, 0,
181 -ESC_h, 0,
182 0, -ESC_k,
183 0, 0,
184 ESC_n, 0,
185 -ESC_p, 0,
186 ESC_r, -ESC_s,
187 ESC_tee, 0,
188 -ESC_v, -ESC_w,
189 0, 0,
190 -ESC_z
191 };
192
193 #else
194
195 /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
196
197 static const short int escapes[] = {
198 /* 48 */ 0, 0, 0, '.', '<', '(', '+', '|',
199 /* 50 */ '&', 0, 0, 0, 0, 0, 0, 0,
200 /* 58 */ 0, 0, '!', '$', '*', ')', ';', '~',
201 /* 60 */ '-', '/', 0, 0, 0, 0, 0, 0,
202 /* 68 */ 0, 0, '|', ',', '%', '_', '>', '?',
203 /* 70 */ 0, 0, 0, 0, 0, 0, 0, 0,
204 /* 78 */ 0, '`', ':', '#', '@', '\'', '=', '"',
205 /* 80 */ 0, 7, -ESC_b, 0, -ESC_d, ESC_e, ESC_f, 0,
206 /* 88 */-ESC_h, 0, 0, '{', 0, 0, 0, 0,
207 /* 90 */ 0, 0, -ESC_k, 'l', 0, ESC_n, 0, -ESC_p,
208 /* 98 */ 0, ESC_r, 0, '}', 0, 0, 0, 0,
209 /* A0 */ 0, '~', -ESC_s, ESC_tee, 0,-ESC_v, -ESC_w, 0,
210 /* A8 */ 0,-ESC_z, 0, 0, 0, '[', 0, 0,
211 /* B0 */ 0, 0, 0, 0, 0, 0, 0, 0,
212 /* B8 */ 0, 0, 0, 0, 0, ']', '=', '-',
213 /* C0 */ '{',-ESC_A, -ESC_B, -ESC_C, -ESC_D,-ESC_E, 0, -ESC_G,
214 /* C8 */-ESC_H, 0, 0, 0, 0, 0, 0, 0,
215 /* D0 */ '}', 0, -ESC_K, 0, 0,-ESC_N, 0, -ESC_P,
216 /* D8 */-ESC_Q,-ESC_R, 0, 0, 0, 0, 0, 0,
217 /* E0 */ '\\', 0, -ESC_S, 0, 0,-ESC_V, -ESC_W, -ESC_X,
218 /* E8 */ 0,-ESC_Z, 0, 0, 0, 0, 0, 0,
219 /* F0 */ 0, 0, 0, 0, 0, 0, 0, 0,
220 /* F8 */ 0, 0, 0, 0, 0, 0, 0, 0
221 };
222 #endif
223
224
225 /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
226 searched linearly. Put all the names into a single string, in order to reduce
227 the number of relocations when a shared library is dynamically linked. The
228 string is built from string macros so that it works in UTF-8 mode on EBCDIC
229 platforms. */
230
231 typedef struct verbitem {
232 int len; /* Length of verb name */
233 int op; /* Op when no arg, or -1 if arg mandatory */
234 int op_arg; /* Op when arg present, or -1 if not allowed */
235 } verbitem;
236
237 static const char verbnames[] =
238 "\0" /* Empty name is a shorthand for MARK */
239 STRING_MARK0
240 STRING_ACCEPT0
241 STRING_COMMIT0
242 STRING_F0
243 STRING_FAIL0
244 STRING_PRUNE0
245 STRING_SKIP0
246 STRING_THEN;
247
248 static const verbitem verbs[] = {
249 { 0, -1, OP_MARK },
250 { 4, -1, OP_MARK },
251 { 6, OP_ACCEPT, -1 },
252 { 6, OP_COMMIT, -1 },
253 { 1, OP_FAIL, -1 },
254 { 4, OP_FAIL, -1 },
255 { 5, OP_PRUNE, OP_PRUNE_ARG },
256 { 4, OP_SKIP, OP_SKIP_ARG },
257 { 4, OP_THEN, OP_THEN_ARG }
258 };
259
260 static const int verbcount = sizeof(verbs)/sizeof(verbitem);
261
262
263 /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264 another regex library. */
265
266 static const pcre_uchar sub_start_of_word[] = {
267 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268 CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269
270 static const pcre_uchar sub_end_of_word[] = {
271 CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272 CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273 CHAR_RIGHT_PARENTHESIS, '\0' };
274
275
276 /* Tables of names of POSIX character classes and their lengths. The names are
277 now all in a single string, to reduce the number of relocations when a shared
278 library is dynamically loaded. The list of lengths is terminated by a zero
279 length entry. The first three must be alpha, lower, upper, as this is assumed
280 for handling case independence. The indices for graph, print, and punct are
281 needed, so identify them. */
282
283 static const char posix_names[] =
284 STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
285 STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
286 STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287 STRING_word0 STRING_xdigit;
288
289 static const pcre_uint8 posix_name_lengths[] = {
290 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291
292 #define PC_GRAPH 8
293 #define PC_PRINT 9
294 #define PC_PUNCT 10
295
296
297 /* Table of class bit maps for each POSIX class. Each class is formed from a
298 base map, with an optional addition or removal of another map. Then, for some
299 classes, there is some additional tweaking: for [:blank:] the vertical space
300 characters are removed, and for [:alpha:] and [:alnum:] the underscore
301 character is removed. The triples in the table consist of the base map offset,
302 second map offset or -1 if no second map, and a non-negative value for map
303 addition or a negative value for map subtraction (if there are two maps). The
304 absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
305 remove vertical space characters, 2 => remove underscore. */
306
307 static const int posix_class_maps[] = {
308 cbit_word, cbit_digit, -2, /* alpha */
309 cbit_lower, -1, 0, /* lower */
310 cbit_upper, -1, 0, /* upper */
311 cbit_word, -1, 2, /* alnum - word without underscore */
312 cbit_print, cbit_cntrl, 0, /* ascii */
313 cbit_space, -1, 1, /* blank - a GNU extension */
314 cbit_cntrl, -1, 0, /* cntrl */
315 cbit_digit, -1, 0, /* digit */
316 cbit_graph, -1, 0, /* graph */
317 cbit_print, -1, 0, /* print */
318 cbit_punct, -1, 0, /* punct */
319 cbit_space, -1, 0, /* space */
320 cbit_word, -1, 0, /* word - a Perl extension */
321 cbit_xdigit,-1, 0 /* xdigit */
322 };
323
324 /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325 Unicode property escapes. */
326
327 #ifdef SUPPORT_UCP
328 static const pcre_uchar string_PNd[] = {
329 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331 static const pcre_uchar string_pNd[] = {
332 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334 static const pcre_uchar string_PXsp[] = {
335 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337 static const pcre_uchar string_pXsp[] = {
338 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340 static const pcre_uchar string_PXwd[] = {
341 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343 static const pcre_uchar string_pXwd[] = {
344 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346
347 static const pcre_uchar *substitutes[] = {
348 string_PNd, /* \D */
349 string_pNd, /* \d */
350 string_PXsp, /* \S */ /* Xsp is Perl space, but from 8.34, Perl */
351 string_pXsp, /* \s */ /* space and POSIX space are the same. */
352 string_PXwd, /* \W */
353 string_pXwd /* \w */
354 };
355
356 /* The POSIX class substitutes must be in the order of the POSIX class names,
357 defined above, and there are both positive and negative cases. NULL means no
358 general substitute of a Unicode property escape (\p or \P). However, for some
359 POSIX classes (e.g. graph, print, punct) a special property code is compiled
360 directly. */
361
362 static const pcre_uchar string_pL[] = {
363 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365 static const pcre_uchar string_pLl[] = {
366 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368 static const pcre_uchar string_pLu[] = {
369 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371 static const pcre_uchar string_pXan[] = {
372 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374 static const pcre_uchar string_h[] = {
375 CHAR_BACKSLASH, CHAR_h, '\0' };
376 static const pcre_uchar string_pXps[] = {
377 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379 static const pcre_uchar string_PL[] = {
380 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382 static const pcre_uchar string_PLl[] = {
383 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385 static const pcre_uchar string_PLu[] = {
386 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388 static const pcre_uchar string_PXan[] = {
389 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391 static const pcre_uchar string_H[] = {
392 CHAR_BACKSLASH, CHAR_H, '\0' };
393 static const pcre_uchar string_PXps[] = {
394 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396
397 static const pcre_uchar *posix_substitutes[] = {
398 string_pL, /* alpha */
399 string_pLl, /* lower */
400 string_pLu, /* upper */
401 string_pXan, /* alnum */
402 NULL, /* ascii */
403 string_h, /* blank */
404 NULL, /* cntrl */
405 string_pNd, /* digit */
406 NULL, /* graph */
407 NULL, /* print */
408 NULL, /* punct */
409 string_pXps, /* space */ /* Xps is POSIX space, but from 8.34 */
410 string_pXwd, /* word */ /* Perl and POSIX space are the same */
411 NULL, /* xdigit */
412 /* Negated cases */
413 string_PL, /* ^alpha */
414 string_PLl, /* ^lower */
415 string_PLu, /* ^upper */
416 string_PXan, /* ^alnum */
417 NULL, /* ^ascii */
418 string_H, /* ^blank */
419 NULL, /* ^cntrl */
420 string_PNd, /* ^digit */
421 NULL, /* ^graph */
422 NULL, /* ^print */
423 NULL, /* ^punct */
424 string_PXps, /* ^space */ /* Xps is POSIX space, but from 8.34 */
425 string_PXwd, /* ^word */ /* Perl and POSIX space are the same */
426 NULL /* ^xdigit */
427 };
428 #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429 #endif
430
431 #define STRING(a) # a
432 #define XSTRING(s) STRING(s)
433
434 /* The texts of compile-time error messages. These are "char *" because they
435 are passed to the outside world. Do not ever re-use any error number, because
436 they are documented. Always add a new error instead. Messages marked DEAD below
437 are no longer used. This used to be a table of strings, but in order to reduce
438 the number of relocations needed when a shared library is loaded dynamically,
439 it is now one long string. We cannot use a table of offsets, because the
440 lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
441 simply count through to the one we want - this isn't a performance issue
442 because these strings are used only when there is a compilation error.
443
444 Each substring ends with \0 to insert a null character. This includes the final
445 substring, so that the whole string ends with \0\0, which can be detected when
446 counting through. */
447
448 static const char error_texts[] =
449 "no error\0"
450 "\\ at end of pattern\0"
451 "\\c at end of pattern\0"
452 "unrecognized character follows \\\0"
453 "numbers out of order in {} quantifier\0"
454 /* 5 */
455 "number too big in {} quantifier\0"
456 "missing terminating ] for character class\0"
457 "invalid escape sequence in character class\0"
458 "range out of order in character class\0"
459 "nothing to repeat\0"
460 /* 10 */
461 "operand of unlimited repeat could match the empty string\0" /** DEAD **/
462 "internal error: unexpected repeat\0"
463 "unrecognized character after (? or (?-\0"
464 "POSIX named classes are supported only within a class\0"
465 "missing )\0"
466 /* 15 */
467 "reference to non-existent subpattern\0"
468 "erroffset passed as NULL\0"
469 "unknown option bit(s) set\0"
470 "missing ) after comment\0"
471 "parentheses nested too deeply\0" /** DEAD **/
472 /* 20 */
473 "regular expression is too large\0"
474 "failed to get memory\0"
475 "unmatched parentheses\0"
476 "internal error: code overflow\0"
477 "unrecognized character after (?<\0"
478 /* 25 */
479 "lookbehind assertion is not fixed length\0"
480 "malformed number or name after (?(\0"
481 "conditional group contains more than two branches\0"
482 "assertion expected after (?(\0"
483 "(?R or (?[+-]digits must be followed by )\0"
484 /* 30 */
485 "unknown POSIX class name\0"
486 "POSIX collating elements are not supported\0"
487 "this version of PCRE is compiled without UTF support\0"
488 "spare error\0" /** DEAD **/
489 "character value in \\x{} or \\o{} is too large\0"
490 /* 35 */
491 "invalid condition (?(0)\0"
492 "\\C not allowed in lookbehind assertion\0"
493 "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
494 "number after (?C is > 255\0"
495 "closing ) for (?C expected\0"
496 /* 40 */
497 "recursive call could loop indefinitely\0"
498 "unrecognized character after (?P\0"
499 "syntax error in subpattern name (missing terminator)\0"
500 "two named subpatterns have the same name\0"
501 "invalid UTF-8 string\0"
502 /* 45 */
503 "support for \\P, \\p, and \\X has not been compiled\0"
504 "malformed \\P or \\p sequence\0"
505 "unknown property name after \\P or \\p\0"
506 "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
507 "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508 /* 50 */
509 "repeated subpattern is too long\0" /** DEAD **/
510 "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
511 "internal error: overran compiling workspace\0"
512 "internal error: previously-checked referenced subpattern not found\0"
513 "DEFINE group contains more than one branch\0"
514 /* 55 */
515 "repeating a DEFINE group is not allowed\0" /** DEAD **/
516 "inconsistent NEWLINE options\0"
517 "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
518 "a numbered reference must not be zero\0"
519 "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520 /* 60 */
521 "(*VERB) not recognized or malformed\0"
522 "number is too big\0"
523 "subpattern name expected\0"
524 "digit expected after (?+\0"
525 "] is an invalid data character in JavaScript compatibility mode\0"
526 /* 65 */
527 "different names for subpatterns of the same number are not allowed\0"
528 "(*MARK) must have an argument\0"
529 "this version of PCRE is not compiled with Unicode property support\0"
530 "\\c must be followed by an ASCII character\0"
531 "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532 /* 70 */
533 "internal error: unknown opcode in find_fixedlength()\0"
534 "\\N is not supported in a class\0"
535 "too many forward references\0"
536 "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537 "invalid UTF-16 string\0"
538 /* 75 */
539 "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540 "character value in \\u.... sequence is too large\0"
541 "invalid UTF-32 string\0"
542 "setting UTF is disabled by the application\0"
543 "non-hex character in \\x{} (closing brace missing?)\0"
544 /* 80 */
545 "non-octal character in \\o{} (closing brace missing?)\0"
546 "missing opening brace after \\o\0"
547 "parentheses are too deeply nested\0"
548 "invalid range in character class\0"
549 "group name must start with a non-digit\0"
550 ;
551
552 /* Table to identify digits and hex digits. This is used when compiling
553 patterns. Note that the tables in chartables are dependent on the locale, and
554 may mark arbitrary characters as digits - but the PCRE compiling code expects
555 to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
556 a private table here. It costs 256 bytes, but it is a lot faster than doing
557 character value tests (at least in some simple cases I timed), and in some
558 applications one wants PCRE to compile efficiently as well as match
559 efficiently.
560
561 For convenience, we use the same bit definitions as in chartables:
562
563 0x04 decimal digit
564 0x08 hexadecimal digit
565
566 Then we can use ctype_digit and ctype_xdigit in the code. */
567
568 /* Using a simple comparison for decimal numbers rather than a memory read
569 is much faster, and the resulting code is simpler (the compiler turns it
570 into a subtraction and unsigned comparison). */
571
572 #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
573
574 #ifndef EBCDIC
575
576 /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
577 UTF-8 mode. */
578
579 static const pcre_uint8 digitab[] =
580 {
581 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
582 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
583 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
584 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
585 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
586 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
587 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
588 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
589 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* @ - G */
590 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H - O */
591 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* P - W */
592 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* X - _ */
593 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
594 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
595 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
596 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
597 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
598 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
599 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
600 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
601 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
602 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
603 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
604 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
605 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
606 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
607 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
608 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
609 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
610 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
611 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
612 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
613
614 #else
615
616 /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
617
618 static const pcre_uint8 digitab[] =
619 {
620 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
621 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
622 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
623 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
624 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
625 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
626 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
627 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
628 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
629 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
630 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
631 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
632 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
633 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
634 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
635 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
636 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
637 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
638 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
639 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
640 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
641 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
642 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
643 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
644 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
645 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
646 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
647 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
648 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
649 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
650 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
651 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
652
653 static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
654 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
655 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
656 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
657 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
658 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
659 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
660 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
661 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
662 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
663 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
664 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
665 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
666 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
667 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
668 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
669 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
670 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
671 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
672 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
673 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
674 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
675 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
676 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
677 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
678 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
679 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
680 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
681 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
682 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
683 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
684 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
685 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
686 #endif
687
688
689 /* This table is used to check whether auto-possessification is possible
690 between adjacent character-type opcodes. The left-hand (repeated) opcode is
691 used to select the row, and the right-hand opcode is use to select the column.
692 A value of 1 means that auto-possessification is OK. For example, the second
693 value in the first row means that \D+\d can be turned into \D++\d.
694
695 The Unicode property types (\P and \p) have to be present to fill out the table
696 because of what their opcode values are, but the table values should always be
697 zero because property types are handled separately in the code. The last four
698 columns apply to items that cannot be repeated, so there is no need to have
699 rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
700 *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
701
702 #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
703 #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
704
705 static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
706 /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
707 { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
708 { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
709 { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
710 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
711 { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
712 { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
713 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
714 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
715 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
716 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
717 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
718 { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
719 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
720 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
721 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
722 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
723 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
724 };
725
726
727 /* This table is used to check whether auto-possessification is possible
728 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
729 left-hand (repeated) opcode is used to select the row, and the right-hand
730 opcode is used to select the column. The values are as follows:
731
732 0 Always return FALSE (never auto-possessify)
733 1 Character groups are distinct (possessify if both are OP_PROP)
734 2 Check character categories in the same group (general or particular)
735 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
736
737 4 Check left general category vs right particular category
738 5 Check right general category vs left particular category
739
740 6 Left alphanum vs right general category
741 7 Left space vs right general category
742 8 Left word vs right general category
743
744 9 Right alphanum vs left general category
745 10 Right space vs left general category
746 11 Right word vs left general category
747
748 12 Left alphanum vs right particular category
749 13 Left space vs right particular category
750 14 Left word vs right particular category
751
752 15 Right alphanum vs left particular category
753 16 Right space vs left particular category
754 17 Right word vs left particular category
755 */
756
757 static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
758 /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
759 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
760 { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
761 { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
762 { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
763 { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
764 { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
765 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
766 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
767 { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
768 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
769 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
770 };
771
772 /* This table is used to check whether auto-possessification is possible
773 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
774 specifies a general category and the other specifies a particular category. The
775 row is selected by the general category and the column by the particular
776 category. The value is 1 if the particular category is not part of the general
777 category. */
778
779 static const pcre_uint8 catposstab[7][30] = {
780 /* Cc Cf Cn Co Cs Ll Lm Lo Lt Lu Mc Me Mn Nd Nl No Pc Pd Pe Pf Pi Po Ps Sc Sk Sm So Zl Zp Zs */
781 { 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* C */
782 { 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* L */
783 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* M */
784 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1 }, /* N */
785 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1 }, /* P */
786 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 1 }, /* S */
787 { 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0 } /* Z */
788 };
789
790 /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
791 a general or particular category. The properties in each row are those
792 that apply to the character set in question. Duplication means that a little
793 unnecessary work is done when checking, but this keeps things much simpler
794 because they can all use the same code. For more details see the comment where
795 this table is used.
796
797 Note: SPACE and PXSPACE used to be different because Perl excluded VT from
798 "space", but from Perl 5.18 it's included, so both categories are treated the
799 same here. */
800
801 static const pcre_uint8 posspropstab[3][4] = {
802 { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
803 { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
804 { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
805 };
806
807 /* This table is used when converting repeating opcodes into possessified
808 versions as a result of an explicit possessive quantifier such as ++. A zero
809 value means there is no possessified version - in those cases the item in
810 question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
811 because all relevant opcodes are less than that. */
812
813 static const pcre_uint8 opcode_possessify[] = {
814 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */
815 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */
816
817 0, /* NOTI */
818 OP_POSSTAR, 0, /* STAR, MINSTAR */
819 OP_POSPLUS, 0, /* PLUS, MINPLUS */
820 OP_POSQUERY, 0, /* QUERY, MINQUERY */
821 OP_POSUPTO, 0, /* UPTO, MINUPTO */
822 0, /* EXACT */
823 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */
824
825 OP_POSSTARI, 0, /* STARI, MINSTARI */
826 OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */
827 OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */
828 OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */
829 0, /* EXACTI */
830 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */
831
832 OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */
833 OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */
834 OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */
835 OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */
836 0, /* NOTEXACT */
837 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
838
839 OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */
840 OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */
841 OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */
842 OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */
843 0, /* NOTEXACTI */
844 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
845
846 OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */
847 OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */
848 OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */
849 OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */
850 0, /* TYPEEXACT */
851 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
852
853 OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */
854 OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */
855 OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */
856 OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */
857 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */
858
859 0, 0, 0, /* CLASS, NCLASS, XCLASS */
860 0, 0, /* REF, REFI */
861 0, 0, /* DNREF, DNREFI */
862 0, 0 /* RECURSE, CALLOUT */
863 };
864
865
866
867 /*************************************************
868 * Find an error text *
869 *************************************************/
870
871 /* The error texts are now all in one long string, to save on relocations. As
872 some of the text is of unknown length, we can't use a table of offsets.
873 Instead, just count through the strings. This is not a performance issue
874 because it happens only when there has been a compilation error.
875
876 Argument: the error number
877 Returns: pointer to the error string
878 */
879
880 static const char *
881 find_error_text(int n)
882 {
883 const char *s = error_texts;
884 for (; n > 0; n--)
885 {
886 while (*s++ != CHAR_NULL) {};
887 if (*s == CHAR_NULL) return "Error text not found (please report)";
888 }
889 return s;
890 }
891
892
893
894 /*************************************************
895 * Expand the workspace *
896 *************************************************/
897
898 /* This function is called during the second compiling phase, if the number of
899 forward references fills the existing workspace, which is originally a block on
900 the stack. A larger block is obtained from malloc() unless the ultimate limit
901 has been reached or the increase will be rather small.
902
903 Argument: pointer to the compile data block
904 Returns: 0 if all went well, else an error number
905 */
906
907 static int
908 expand_workspace(compile_data *cd)
909 {
910 pcre_uchar *newspace;
911 int newsize = cd->workspace_size * 2;
912
913 if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
914 if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
915 newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
916 return ERR72;
917
918 newspace = (PUBL(malloc))(IN_UCHARS(newsize));
919 if (newspace == NULL) return ERR21;
920 memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
921 cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
922 if (cd->workspace_size > COMPILE_WORK_SIZE)
923 (PUBL(free))((void *)cd->start_workspace);
924 cd->start_workspace = newspace;
925 cd->workspace_size = newsize;
926 return 0;
927 }
928
929
930
931 /*************************************************
932 * Check for counted repeat *
933 *************************************************/
934
935 /* This function is called when a '{' is encountered in a place where it might
936 start a quantifier. It looks ahead to see if it really is a quantifier or not.
937 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
938 where the ddds are digits.
939
940 Arguments:
941 p pointer to the first char after '{'
942
943 Returns: TRUE or FALSE
944 */
945
946 static BOOL
947 is_counted_repeat(const pcre_uchar *p)
948 {
949 if (!IS_DIGIT(*p)) return FALSE;
950 p++;
951 while (IS_DIGIT(*p)) p++;
952 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
953
954 if (*p++ != CHAR_COMMA) return FALSE;
955 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
956
957 if (!IS_DIGIT(*p)) return FALSE;
958 p++;
959 while (IS_DIGIT(*p)) p++;
960
961 return (*p == CHAR_RIGHT_CURLY_BRACKET);
962 }
963
964
965
966 /*************************************************
967 * Handle escapes *
968 *************************************************/
969
970 /* This function is called when a \ has been encountered. It either returns a
971 positive value for a simple escape such as \n, or 0 for a data character which
972 will be placed in chptr. A backreference to group n is returned as negative n.
973 When UTF-8 is enabled, a positive value greater than 255 may be returned in
974 chptr. On entry, ptr is pointing at the \. On exit, it is on the final
975 character of the escape sequence.
976
977 Arguments:
978 ptrptr points to the pattern position pointer
979 chptr points to a returned data character
980 errorcodeptr points to the errorcode variable
981 bracount number of previous extracting brackets
982 options the options bits
983 isclass TRUE if inside a character class
984
985 Returns: zero => a data character
986 positive => a special escape sequence
987 negative => a back reference
988 on error, errorcodeptr is set
989 */
990
991 static int
992 check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
993 int bracount, int options, BOOL isclass)
994 {
995 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
996 BOOL utf = (options & PCRE_UTF8) != 0;
997 const pcre_uchar *ptr = *ptrptr + 1;
998 pcre_uint32 c;
999 int escape = 0;
1000 int i;
1001
1002 GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
1003 ptr--; /* Set pointer back to the last byte */
1004
1005 /* If backslash is at the end of the pattern, it's an error. */
1006
1007 if (c == CHAR_NULL) *errorcodeptr = ERR1;
1008
1009 /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1010 in a table. A non-zero result is something that can be returned immediately.
1011 Otherwise further processing may be required. */
1012
1013 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1014 /* Not alphanumeric */
1015 else if (c < CHAR_0 || c > CHAR_z) {}
1016 else if ((i = escapes[c - CHAR_0]) != 0)
1017 { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1018
1019 #else /* EBCDIC coding */
1020 /* Not alphanumeric */
1021 else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1022 else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1023 #endif
1024
1025 /* Escapes that need further processing, or are illegal. */
1026
1027 else
1028 {
1029 const pcre_uchar *oldptr;
1030 BOOL braced, negated, overflow;
1031 int s;
1032
1033 switch (c)
1034 {
1035 /* A number of Perl escapes are not handled by PCRE. We give an explicit
1036 error. */
1037
1038 case CHAR_l:
1039 case CHAR_L:
1040 *errorcodeptr = ERR37;
1041 break;
1042
1043 case CHAR_u:
1044 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1045 {
1046 /* In JavaScript, \u must be followed by four hexadecimal numbers.
1047 Otherwise it is a lowercase u letter. */
1048 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1049 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1050 && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1051 && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1052 {
1053 c = 0;
1054 for (i = 0; i < 4; ++i)
1055 {
1056 register pcre_uint32 cc = *(++ptr);
1057 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1058 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1059 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1060 #else /* EBCDIC coding */
1061 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1062 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1063 #endif
1064 }
1065
1066 #if defined COMPILE_PCRE8
1067 if (c > (utf ? 0x10ffffU : 0xffU))
1068 #elif defined COMPILE_PCRE16
1069 if (c > (utf ? 0x10ffffU : 0xffffU))
1070 #elif defined COMPILE_PCRE32
1071 if (utf && c > 0x10ffffU)
1072 #endif
1073 {
1074 *errorcodeptr = ERR76;
1075 }
1076 else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1077 }
1078 }
1079 else
1080 *errorcodeptr = ERR37;
1081 break;
1082
1083 case CHAR_U:
1084 /* In JavaScript, \U is an uppercase U letter. */
1085 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1086 break;
1087
1088 /* In a character class, \g is just a literal "g". Outside a character
1089 class, \g must be followed by one of a number of specific things:
1090
1091 (1) A number, either plain or braced. If positive, it is an absolute
1092 backreference. If negative, it is a relative backreference. This is a Perl
1093 5.10 feature.
1094
1095 (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1096 is part of Perl's movement towards a unified syntax for back references. As
1097 this is synonymous with \k{name}, we fudge it up by pretending it really
1098 was \k.
1099
1100 (3) For Oniguruma compatibility we also support \g followed by a name or a
1101 number either in angle brackets or in single quotes. However, these are
1102 (possibly recursive) subroutine calls, _not_ backreferences. Just return
1103 the ESC_g code (cf \k). */
1104
1105 case CHAR_g:
1106 if (isclass) break;
1107 if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1108 {
1109 escape = ESC_g;
1110 break;
1111 }
1112
1113 /* Handle the Perl-compatible cases */
1114
1115 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1116 {
1117 const pcre_uchar *p;
1118 for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1119 if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1120 if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1121 {
1122 escape = ESC_k;
1123 break;
1124 }
1125 braced = TRUE;
1126 ptr++;
1127 }
1128 else braced = FALSE;
1129
1130 if (ptr[1] == CHAR_MINUS)
1131 {
1132 negated = TRUE;
1133 ptr++;
1134 }
1135 else negated = FALSE;
1136
1137 /* The integer range is limited by the machine's int representation. */
1138 s = 0;
1139 overflow = FALSE;
1140 while (IS_DIGIT(ptr[1]))
1141 {
1142 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1143 {
1144 overflow = TRUE;
1145 break;
1146 }
1147 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1148 }
1149 if (overflow) /* Integer overflow */
1150 {
1151 while (IS_DIGIT(ptr[1]))
1152 ptr++;
1153 *errorcodeptr = ERR61;
1154 break;
1155 }
1156
1157 if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1158 {
1159 *errorcodeptr = ERR57;
1160 break;
1161 }
1162
1163 if (s == 0)
1164 {
1165 *errorcodeptr = ERR58;
1166 break;
1167 }
1168
1169 if (negated)
1170 {
1171 if (s > bracount)
1172 {
1173 *errorcodeptr = ERR15;
1174 break;
1175 }
1176 s = bracount - (s - 1);
1177 }
1178
1179 escape = -s;
1180 break;
1181
1182 /* The handling of escape sequences consisting of a string of digits
1183 starting with one that is not zero is not straightforward. Perl has changed
1184 over the years. Nowadays \g{} for backreferences and \o{} for octal are
1185 recommended to avoid the ambiguities in the old syntax.
1186
1187 Outside a character class, the digits are read as a decimal number. If the
1188 number is less than 8 (used to be 10), or if there are that many previous
1189 extracting left brackets, then it is a back reference. Otherwise, up to
1190 three octal digits are read to form an escaped byte. Thus \123 is likely to
1191 be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1192 the octal value is greater than 377, the least significant 8 bits are
1193 taken. \8 and \9 are treated as the literal characters 8 and 9.
1194
1195 Inside a character class, \ followed by a digit is always either a literal
1196 8 or 9 or an octal number. */
1197
1198 case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1199 case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1200
1201 if (!isclass)
1202 {
1203 oldptr = ptr;
1204 /* The integer range is limited by the machine's int representation. */
1205 s = (int)(c -CHAR_0);
1206 overflow = FALSE;
1207 while (IS_DIGIT(ptr[1]))
1208 {
1209 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1210 {
1211 overflow = TRUE;
1212 break;
1213 }
1214 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1215 }
1216 if (overflow) /* Integer overflow */
1217 {
1218 while (IS_DIGIT(ptr[1]))
1219 ptr++;
1220 *errorcodeptr = ERR61;
1221 break;
1222 }
1223 if (s < 8 || s <= bracount) /* Check for back reference */
1224 {
1225 escape = -s;
1226 break;
1227 }
1228 ptr = oldptr; /* Put the pointer back and fall through */
1229 }
1230
1231 /* Handle a digit following \ when the number is not a back reference. If
1232 the first digit is 8 or 9, Perl used to generate a binary zero byte and
1233 then treat the digit as a following literal. At least by Perl 5.18 this
1234 changed so as not to insert the binary zero. */
1235
1236 if ((c = *ptr) >= CHAR_8) break;
1237
1238 /* Fall through with a digit less than 8 */
1239
1240 /* \0 always starts an octal number, but we may drop through to here with a
1241 larger first octal digit. The original code used just to take the least
1242 significant 8 bits of octal numbers (I think this is what early Perls used
1243 to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1244 but no more than 3 octal digits. */
1245
1246 case CHAR_0:
1247 c -= CHAR_0;
1248 while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1249 c = c * 8 + *(++ptr) - CHAR_0;
1250 #ifdef COMPILE_PCRE8
1251 if (!utf && c > 0xff) *errorcodeptr = ERR51;
1252 #endif
1253 break;
1254
1255 /* \o is a relatively new Perl feature, supporting a more general way of
1256 specifying character codes in octal. The only supported form is \o{ddd}. */
1257
1258 case CHAR_o:
1259 if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1260 {
1261 ptr += 2;
1262 c = 0;
1263 overflow = FALSE;
1264 while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1265 {
1266 register pcre_uint32 cc = *ptr++;
1267 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1268 #ifdef COMPILE_PCRE32
1269 if (c >= 0x20000000l) { overflow = TRUE; break; }
1270 #endif
1271 c = (c << 3) + cc - CHAR_0 ;
1272 #if defined COMPILE_PCRE8
1273 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1274 #elif defined COMPILE_PCRE16
1275 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1276 #elif defined COMPILE_PCRE32
1277 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1278 #endif
1279 }
1280 if (overflow)
1281 {
1282 while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1283 *errorcodeptr = ERR34;
1284 }
1285 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1286 {
1287 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1288 }
1289 else *errorcodeptr = ERR80;
1290 }
1291 break;
1292
1293 /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1294 numbers. Otherwise it is a lowercase x letter. */
1295
1296 case CHAR_x:
1297 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1298 {
1299 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1300 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1301 {
1302 c = 0;
1303 for (i = 0; i < 2; ++i)
1304 {
1305 register pcre_uint32 cc = *(++ptr);
1306 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1307 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1308 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1309 #else /* EBCDIC coding */
1310 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1311 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1312 #endif
1313 }
1314 }
1315 } /* End JavaScript handling */
1316
1317 /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1318 greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1319 digits. If not, { used to be treated as a data character. However, Perl
1320 seems to read hex digits up to the first non-such, and ignore the rest, so
1321 that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1322 now gives an error. */
1323
1324 else
1325 {
1326 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1327 {
1328 ptr += 2;
1329 c = 0;
1330 overflow = FALSE;
1331 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1332 {
1333 register pcre_uint32 cc = *ptr++;
1334 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1335
1336 #ifdef COMPILE_PCRE32
1337 if (c >= 0x10000000l) { overflow = TRUE; break; }
1338 #endif
1339
1340 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1341 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1342 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1343 #else /* EBCDIC coding */
1344 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1345 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1346 #endif
1347
1348 #if defined COMPILE_PCRE8
1349 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1350 #elif defined COMPILE_PCRE16
1351 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1352 #elif defined COMPILE_PCRE32
1353 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1354 #endif
1355 }
1356
1357 if (overflow)
1358 {
1359 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1360 *errorcodeptr = ERR34;
1361 }
1362
1363 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1364 {
1365 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1366 }
1367
1368 /* If the sequence of hex digits does not end with '}', give an error.
1369 We used just to recognize this construct and fall through to the normal
1370 \x handling, but nowadays Perl gives an error, which seems much more
1371 sensible, so we do too. */
1372
1373 else *errorcodeptr = ERR79;
1374 } /* End of \x{} processing */
1375
1376 /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1377
1378 else
1379 {
1380 c = 0;
1381 while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1382 {
1383 pcre_uint32 cc; /* Some compilers don't like */
1384 cc = *(++ptr); /* ++ in initializers */
1385 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1386 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1387 c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1388 #else /* EBCDIC coding */
1389 if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1390 c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1391 #endif
1392 }
1393 } /* End of \xdd handling */
1394 } /* End of Perl-style \x handling */
1395 break;
1396
1397 /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1398 An error is given if the byte following \c is not an ASCII character. This
1399 coding is ASCII-specific, but then the whole concept of \cx is
1400 ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1401
1402 case CHAR_c:
1403 c = *(++ptr);
1404 if (c == CHAR_NULL)
1405 {
1406 *errorcodeptr = ERR2;
1407 break;
1408 }
1409 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1410 if (c > 127) /* Excludes all non-ASCII in either mode */
1411 {
1412 *errorcodeptr = ERR68;
1413 break;
1414 }
1415 if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1416 c ^= 0x40;
1417 #else /* EBCDIC coding */
1418 if (c >= CHAR_a && c <= CHAR_z) c += 64;
1419 c ^= 0xC0;
1420 #endif
1421 break;
1422
1423 /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1424 other alphanumeric following \ is an error if PCRE_EXTRA was set;
1425 otherwise, for Perl compatibility, it is a literal. This code looks a bit
1426 odd, but there used to be some cases other than the default, and there may
1427 be again in future, so I haven't "optimized" it. */
1428
1429 default:
1430 if ((options & PCRE_EXTRA) != 0) switch(c)
1431 {
1432 default:
1433 *errorcodeptr = ERR3;
1434 break;
1435 }
1436 break;
1437 }
1438 }
1439
1440 /* Perl supports \N{name} for character names, as well as plain \N for "not
1441 newline". PCRE does not support \N{name}. However, it does support
1442 quantification such as \N{2,3}. */
1443
1444 if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1445 !is_counted_repeat(ptr+2))
1446 *errorcodeptr = ERR37;
1447
1448 /* If PCRE_UCP is set, we change the values for \d etc. */
1449
1450 if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1451 escape += (ESC_DU - ESC_D);
1452
1453 /* Set the pointer to the final character before returning. */
1454
1455 *ptrptr = ptr;
1456 *chptr = c;
1457 return escape;
1458 }
1459
1460
1461
1462 #ifdef SUPPORT_UCP
1463 /*************************************************
1464 * Handle \P and \p *
1465 *************************************************/
1466
1467 /* This function is called after \P or \p has been encountered, provided that
1468 PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1469 pointing at the P or p. On exit, it is pointing at the final character of the
1470 escape sequence.
1471
1472 Argument:
1473 ptrptr points to the pattern position pointer
1474 negptr points to a boolean that is set TRUE for negation else FALSE
1475 ptypeptr points to an unsigned int that is set to the type value
1476 pdataptr points to an unsigned int that is set to the detailed property value
1477 errorcodeptr points to the error code variable
1478
1479 Returns: TRUE if the type value was found, or FALSE for an invalid type
1480 */
1481
1482 static BOOL
1483 get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1484 unsigned int *pdataptr, int *errorcodeptr)
1485 {
1486 pcre_uchar c;
1487 int i, bot, top;
1488 const pcre_uchar *ptr = *ptrptr;
1489 pcre_uchar name[32];
1490
1491 c = *(++ptr);
1492 if (c == CHAR_NULL) goto ERROR_RETURN;
1493
1494 *negptr = FALSE;
1495
1496 /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1497 negation. */
1498
1499 if (c == CHAR_LEFT_CURLY_BRACKET)
1500 {
1501 if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1502 {
1503 *negptr = TRUE;
1504 ptr++;
1505 }
1506 for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1507 {
1508 c = *(++ptr);
1509 if (c == CHAR_NULL) goto ERROR_RETURN;
1510 if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1511 name[i] = c;
1512 }
1513 if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1514 name[i] = 0;
1515 }
1516
1517 /* Otherwise there is just one following character */
1518
1519 else
1520 {
1521 name[0] = c;
1522 name[1] = 0;
1523 }
1524
1525 *ptrptr = ptr;
1526
1527 /* Search for a recognized property name using binary chop */
1528
1529 bot = 0;
1530 top = PRIV(utt_size);
1531
1532 while (bot < top)
1533 {
1534 int r;
1535 i = (bot + top) >> 1;
1536 r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1537 if (r == 0)
1538 {
1539 *ptypeptr = PRIV(utt)[i].type;
1540 *pdataptr = PRIV(utt)[i].value;
1541 return TRUE;
1542 }
1543 if (r > 0) bot = i + 1; else top = i;
1544 }
1545
1546 *errorcodeptr = ERR47;
1547 *ptrptr = ptr;
1548 return FALSE;
1549
1550 ERROR_RETURN:
1551 *errorcodeptr = ERR46;
1552 *ptrptr = ptr;
1553 return FALSE;
1554 }
1555 #endif
1556
1557
1558
1559 /*************************************************
1560 * Read repeat counts *
1561 *************************************************/
1562
1563 /* Read an item of the form {n,m} and return the values. This is called only
1564 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1565 so the syntax is guaranteed to be correct, but we need to check the values.
1566
1567 Arguments:
1568 p pointer to first char after '{'
1569 minp pointer to int for min
1570 maxp pointer to int for max
1571 returned as -1 if no max
1572 errorcodeptr points to error code variable
1573
1574 Returns: pointer to '}' on success;
1575 current ptr on error, with errorcodeptr set non-zero
1576 */
1577
1578 static const pcre_uchar *
1579 read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1580 {
1581 int min = 0;
1582 int max = -1;
1583
1584 /* Read the minimum value and do a paranoid check: a negative value indicates
1585 an integer overflow. */
1586
1587 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1588 if (min < 0 || min > 65535)
1589 {
1590 *errorcodeptr = ERR5;
1591 return p;
1592 }
1593
1594 /* Read the maximum value if there is one, and again do a paranoid on its size.
1595 Also, max must not be less than min. */
1596
1597 if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1598 {
1599 if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1600 {
1601 max = 0;
1602 while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1603 if (max < 0 || max > 65535)
1604 {
1605 *errorcodeptr = ERR5;
1606 return p;
1607 }
1608 if (max < min)
1609 {
1610 *errorcodeptr = ERR4;
1611 return p;
1612 }
1613 }
1614 }
1615
1616 /* Fill in the required variables, and pass back the pointer to the terminating
1617 '}'. */
1618
1619 *minp = min;
1620 *maxp = max;
1621 return p;
1622 }
1623
1624
1625
1626 /*************************************************
1627 * Find first significant op code *
1628 *************************************************/
1629
1630 /* This is called by several functions that scan a compiled expression looking
1631 for a fixed first character, or an anchoring op code etc. It skips over things
1632 that do not influence this. For some calls, it makes sense to skip negative
1633 forward and all backward assertions, and also the \b assertion; for others it
1634 does not.
1635
1636 Arguments:
1637 code pointer to the start of the group
1638 skipassert TRUE if certain assertions are to be skipped
1639
1640 Returns: pointer to the first significant opcode
1641 */
1642
1643 static const pcre_uchar*
1644 first_significant_code(const pcre_uchar *code, BOOL skipassert)
1645 {
1646 for (;;)
1647 {
1648 switch ((int)*code)
1649 {
1650 case OP_ASSERT_NOT:
1651 case OP_ASSERTBACK:
1652 case OP_ASSERTBACK_NOT:
1653 if (!skipassert) return code;
1654 do code += GET(code, 1); while (*code == OP_ALT);
1655 code += PRIV(OP_lengths)[*code];
1656 break;
1657
1658 case OP_WORD_BOUNDARY:
1659 case OP_NOT_WORD_BOUNDARY:
1660 if (!skipassert) return code;
1661 /* Fall through */
1662
1663 case OP_CALLOUT:
1664 case OP_CREF:
1665 case OP_DNCREF:
1666 case OP_RREF:
1667 case OP_DNRREF:
1668 case OP_DEF:
1669 code += PRIV(OP_lengths)[*code];
1670 break;
1671
1672 default:
1673 return code;
1674 }
1675 }
1676 /* Control never reaches here */
1677 }
1678
1679
1680
1681 /*************************************************
1682 * Find the fixed length of a branch *
1683 *************************************************/
1684
1685 /* Scan a branch and compute the fixed length of subject that will match it,
1686 if the length is fixed. This is needed for dealing with backward assertions.
1687 In UTF8 mode, the result is in characters rather than bytes. The branch is
1688 temporarily terminated with OP_END when this function is called.
1689
1690 This function is called when a backward assertion is encountered, so that if it
1691 fails, the error message can point to the correct place in the pattern.
1692 However, we cannot do this when the assertion contains subroutine calls,
1693 because they can be forward references. We solve this by remembering this case
1694 and doing the check at the end; a flag specifies which mode we are running in.
1695
1696 Arguments:
1697 code points to the start of the pattern (the bracket)
1698 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1699 atend TRUE if called when the pattern is complete
1700 cd the "compile data" structure
1701
1702 Returns: the fixed length,
1703 or -1 if there is no fixed length,
1704 or -2 if \C was encountered (in UTF-8 mode only)
1705 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1706 or -4 if an unknown opcode was encountered (internal error)
1707 */
1708
1709 static int
1710 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1711 {
1712 int length = -1;
1713
1714 register int branchlength = 0;
1715 register pcre_uchar *cc = code + 1 + LINK_SIZE;
1716
1717 /* Scan along the opcodes for this branch. If we get to the end of the
1718 branch, check the length against that of the other branches. */
1719
1720 for (;;)
1721 {
1722 int d;
1723 pcre_uchar *ce, *cs;
1724 register pcre_uchar op = *cc;
1725
1726 switch (op)
1727 {
1728 /* We only need to continue for OP_CBRA (normal capturing bracket) and
1729 OP_BRA (normal non-capturing bracket) because the other variants of these
1730 opcodes are all concerned with unlimited repeated groups, which of course
1731 are not of fixed length. */
1732
1733 case OP_CBRA:
1734 case OP_BRA:
1735 case OP_ONCE:
1736 case OP_ONCE_NC:
1737 case OP_COND:
1738 d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1739 if (d < 0) return d;
1740 branchlength += d;
1741 do cc += GET(cc, 1); while (*cc == OP_ALT);
1742 cc += 1 + LINK_SIZE;
1743 break;
1744
1745 /* Reached end of a branch; if it's a ket it is the end of a nested call.
1746 If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1747 an ALT. If it is END it's the end of the outer call. All can be handled by
1748 the same code. Note that we must not include the OP_KETRxxx opcodes here,
1749 because they all imply an unlimited repeat. */
1750
1751 case OP_ALT:
1752 case OP_KET:
1753 case OP_END:
1754 case OP_ACCEPT:
1755 case OP_ASSERT_ACCEPT:
1756 if (length < 0) length = branchlength;
1757 else if (length != branchlength) return -1;
1758 if (*cc != OP_ALT) return length;
1759 cc += 1 + LINK_SIZE;
1760 branchlength = 0;
1761 break;
1762
1763 /* A true recursion implies not fixed length, but a subroutine call may
1764 be OK. If the subroutine is a forward reference, we can't deal with
1765 it until the end of the pattern, so return -3. */
1766
1767 case OP_RECURSE:
1768 if (!atend) return -3;
1769 cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1770 do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1771 if (cc > cs && cc < ce) return -1; /* Recursion */
1772 d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1773 if (d < 0) return d;
1774 branchlength += d;
1775 cc += 1 + LINK_SIZE;
1776 break;
1777
1778 /* Skip over assertive subpatterns */
1779
1780 case OP_ASSERT:
1781 case OP_ASSERT_NOT:
1782 case OP_ASSERTBACK:
1783 case OP_ASSERTBACK_NOT:
1784 do cc += GET(cc, 1); while (*cc == OP_ALT);
1785 cc += PRIV(OP_lengths)[*cc];
1786 break;
1787
1788 /* Skip over things that don't match chars */
1789
1790 case OP_MARK:
1791 case OP_PRUNE_ARG:
1792 case OP_SKIP_ARG:
1793 case OP_THEN_ARG:
1794 cc += cc[1] + PRIV(OP_lengths)[*cc];
1795 break;
1796
1797 case OP_CALLOUT:
1798 case OP_CIRC:
1799 case OP_CIRCM:
1800 case OP_CLOSE:
1801 case OP_COMMIT:
1802 case OP_CREF:
1803 case OP_DEF:
1804 case OP_DNCREF:
1805 case OP_DNRREF:
1806 case OP_DOLL:
1807 case OP_DOLLM:
1808 case OP_EOD:
1809 case OP_EODN:
1810 case OP_FAIL:
1811 case OP_NOT_WORD_BOUNDARY:
1812 case OP_PRUNE:
1813 case OP_REVERSE:
1814 case OP_RREF:
1815 case OP_SET_SOM:
1816 case OP_SKIP:
1817 case OP_SOD:
1818 case OP_SOM:
1819 case OP_THEN:
1820 case OP_WORD_BOUNDARY:
1821 cc += PRIV(OP_lengths)[*cc];
1822 break;
1823
1824 /* Handle literal characters */
1825
1826 case OP_CHAR:
1827 case OP_CHARI:
1828 case OP_NOT:
1829 case OP_NOTI:
1830 branchlength++;
1831 cc += 2;
1832 #ifdef SUPPORT_UTF
1833 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1834 #endif
1835 break;
1836
1837 /* Handle exact repetitions. The count is already in characters, but we
1838 need to skip over a multibyte character in UTF8 mode. */
1839
1840 case OP_EXACT:
1841 case OP_EXACTI:
1842 case OP_NOTEXACT:
1843 case OP_NOTEXACTI:
1844 branchlength += (int)GET2(cc,1);
1845 cc += 2 + IMM2_SIZE;
1846 #ifdef SUPPORT_UTF
1847 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1848 #endif
1849 break;
1850
1851 case OP_TYPEEXACT:
1852 branchlength += GET2(cc,1);
1853 if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1854 cc += 2;
1855 cc += 1 + IMM2_SIZE + 1;
1856 break;
1857
1858 /* Handle single-char matchers */
1859
1860 case OP_PROP:
1861 case OP_NOTPROP:
1862 cc += 2;
1863 /* Fall through */
1864
1865 case OP_HSPACE:
1866 case OP_VSPACE:
1867 case OP_NOT_HSPACE:
1868 case OP_NOT_VSPACE:
1869 case OP_NOT_DIGIT:
1870 case OP_DIGIT:
1871 case OP_NOT_WHITESPACE:
1872 case OP_WHITESPACE:
1873 case OP_NOT_WORDCHAR:
1874 case OP_WORDCHAR:
1875 case OP_ANY:
1876 case OP_ALLANY:
1877 branchlength++;
1878 cc++;
1879 break;
1880
1881 /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1882 otherwise \C is coded as OP_ALLANY. */
1883
1884 case OP_ANYBYTE:
1885 return -2;
1886
1887 /* Check a class for variable quantification */
1888
1889 case OP_CLASS:
1890 case OP_NCLASS:
1891 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1892 case OP_XCLASS:
1893 /* The original code caused an unsigned overflow in 64 bit systems,
1894 so now we use a conditional statement. */
1895 if (op == OP_XCLASS)
1896 cc += GET(cc, 1);
1897 else
1898 cc += PRIV(OP_lengths)[OP_CLASS];
1899 #else
1900 cc += PRIV(OP_lengths)[OP_CLASS];
1901 #endif
1902
1903 switch (*cc)
1904 {
1905 case OP_CRSTAR:
1906 case OP_CRMINSTAR:
1907 case OP_CRPLUS:
1908 case OP_CRMINPLUS:
1909 case OP_CRQUERY:
1910 case OP_CRMINQUERY:
1911 case OP_CRPOSSTAR:
1912 case OP_CRPOSPLUS:
1913 case OP_CRPOSQUERY:
1914 return -1;
1915
1916 case OP_CRRANGE:
1917 case OP_CRMINRANGE:
1918 case OP_CRPOSRANGE:
1919 if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1920 branchlength += (int)GET2(cc,1);
1921 cc += 1 + 2 * IMM2_SIZE;
1922 break;
1923
1924 default:
1925 branchlength++;
1926 }
1927 break;
1928
1929 /* Anything else is variable length */
1930
1931 case OP_ANYNL:
1932 case OP_BRAMINZERO:
1933 case OP_BRAPOS:
1934 case OP_BRAPOSZERO:
1935 case OP_BRAZERO:
1936 case OP_CBRAPOS:
1937 case OP_EXTUNI:
1938 case OP_KETRMAX:
1939 case OP_KETRMIN:
1940 case OP_KETRPOS:
1941 case OP_MINPLUS:
1942 case OP_MINPLUSI:
1943 case OP_MINQUERY:
1944 case OP_MINQUERYI:
1945 case OP_MINSTAR:
1946 case OP_MINSTARI:
1947 case OP_MINUPTO:
1948 case OP_MINUPTOI:
1949 case OP_NOTMINPLUS:
1950 case OP_NOTMINPLUSI:
1951 case OP_NOTMINQUERY:
1952 case OP_NOTMINQUERYI:
1953 case OP_NOTMINSTAR:
1954 case OP_NOTMINSTARI:
1955 case OP_NOTMINUPTO:
1956 case OP_NOTMINUPTOI:
1957 case OP_NOTPLUS:
1958 case OP_NOTPLUSI:
1959 case OP_NOTPOSPLUS:
1960 case OP_NOTPOSPLUSI:
1961 case OP_NOTPOSQUERY:
1962 case OP_NOTPOSQUERYI:
1963 case OP_NOTPOSSTAR:
1964 case OP_NOTPOSSTARI:
1965 case OP_NOTPOSUPTO:
1966 case OP_NOTPOSUPTOI:
1967 case OP_NOTQUERY:
1968 case OP_NOTQUERYI:
1969 case OP_NOTSTAR:
1970 case OP_NOTSTARI:
1971 case OP_NOTUPTO:
1972 case OP_NOTUPTOI:
1973 case OP_PLUS:
1974 case OP_PLUSI:
1975 case OP_POSPLUS:
1976 case OP_POSPLUSI:
1977 case OP_POSQUERY:
1978 case OP_POSQUERYI:
1979 case OP_POSSTAR:
1980 case OP_POSSTARI:
1981 case OP_POSUPTO:
1982 case OP_POSUPTOI:
1983 case OP_QUERY:
1984 case OP_QUERYI:
1985 case OP_REF:
1986 case OP_REFI:
1987 case OP_DNREF:
1988 case OP_DNREFI:
1989 case OP_SBRA:
1990 case OP_SBRAPOS:
1991 case OP_SCBRA:
1992 case OP_SCBRAPOS:
1993 case OP_SCOND:
1994 case OP_SKIPZERO:
1995 case OP_STAR:
1996 case OP_STARI:
1997 case OP_TYPEMINPLUS:
1998 case OP_TYPEMINQUERY:
1999 case OP_TYPEMINSTAR:
2000 case OP_TYPEMINUPTO:
2001 case OP_TYPEPLUS:
2002 case OP_TYPEPOSPLUS:
2003 case OP_TYPEPOSQUERY:
2004 case OP_TYPEPOSSTAR:
2005 case OP_TYPEPOSUPTO:
2006 case OP_TYPEQUERY:
2007 case OP_TYPESTAR:
2008 case OP_TYPEUPTO:
2009 case OP_UPTO:
2010 case OP_UPTOI:
2011 return -1;
2012
2013 /* Catch unrecognized opcodes so that when new ones are added they
2014 are not forgotten, as has happened in the past. */
2015
2016 default:
2017 return -4;
2018 }
2019 }
2020 /* Control never gets here */
2021 }
2022
2023
2024
2025 /*************************************************
2026 * Scan compiled regex for specific bracket *
2027 *************************************************/
2028
2029 /* This little function scans through a compiled pattern until it finds a
2030 capturing bracket with the given number, or, if the number is negative, an
2031 instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2032 so that it can be called from pcre_study() when finding the minimum matching
2033 length.
2034
2035 Arguments:
2036 code points to start of expression
2037 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2038 number the required bracket number or negative to find a lookbehind
2039
2040 Returns: pointer to the opcode for the bracket, or NULL if not found
2041 */
2042
2043 const pcre_uchar *
2044 PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2045 {
2046 for (;;)
2047 {
2048 register pcre_uchar c = *code;
2049
2050 if (c == OP_END) return NULL;
2051
2052 /* XCLASS is used for classes that cannot be represented just by a bit
2053 map. This includes negated single high-valued characters. The length in
2054 the table is zero; the actual length is stored in the compiled code. */
2055
2056 if (c == OP_XCLASS) code += GET(code, 1);
2057
2058 /* Handle recursion */
2059
2060 else if (c == OP_REVERSE)
2061 {
2062 if (number < 0) return (pcre_uchar *)code;
2063 code += PRIV(OP_lengths)[c];
2064 }
2065
2066 /* Handle capturing bracket */
2067
2068 else if (c == OP_CBRA || c == OP_SCBRA ||
2069 c == OP_CBRAPOS || c == OP_SCBRAPOS)
2070 {
2071 int n = (int)GET2(code, 1+LINK_SIZE);
2072 if (n == number) return (pcre_uchar *)code;
2073 code += PRIV(OP_lengths)[c];
2074 }
2075
2076 /* Otherwise, we can get the item's length from the table, except that for
2077 repeated character types, we have to test for \p and \P, which have an extra
2078 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2079 must add in its length. */
2080
2081 else
2082 {
2083 switch(c)
2084 {
2085 case OP_TYPESTAR:
2086 case OP_TYPEMINSTAR:
2087 case OP_TYPEPLUS:
2088 case OP_TYPEMINPLUS:
2089 case OP_TYPEQUERY:
2090 case OP_TYPEMINQUERY:
2091 case OP_TYPEPOSSTAR:
2092 case OP_TYPEPOSPLUS:
2093 case OP_TYPEPOSQUERY:
2094 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2095 break;
2096
2097 case OP_TYPEUPTO:
2098 case OP_TYPEMINUPTO:
2099 case OP_TYPEEXACT:
2100 case OP_TYPEPOSUPTO:
2101 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2102 code += 2;
2103 break;
2104
2105 case OP_MARK:
2106 case OP_PRUNE_ARG:
2107 case OP_SKIP_ARG:
2108 case OP_THEN_ARG:
2109 code += code[1];
2110 break;
2111 }
2112
2113 /* Add in the fixed length from the table */
2114
2115 code += PRIV(OP_lengths)[c];
2116
2117 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2118 a multi-byte character. The length in the table is a minimum, so we have to
2119 arrange to skip the extra bytes. */
2120
2121 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2122 if (utf) switch(c)
2123 {
2124 case OP_CHAR:
2125 case OP_CHARI:
2126 case OP_EXACT:
2127 case OP_EXACTI:
2128 case OP_UPTO:
2129 case OP_UPTOI:
2130 case OP_MINUPTO:
2131 case OP_MINUPTOI:
2132 case OP_POSUPTO:
2133 case OP_POSUPTOI:
2134 case OP_STAR:
2135 case OP_STARI:
2136 case OP_MINSTAR:
2137 case OP_MINSTARI:
2138 case OP_POSSTAR:
2139 case OP_POSSTARI:
2140 case OP_PLUS:
2141 case OP_PLUSI:
2142 case OP_MINPLUS:
2143 case OP_MINPLUSI:
2144 case OP_POSPLUS:
2145 case OP_POSPLUSI:
2146 case OP_QUERY:
2147 case OP_QUERYI:
2148 case OP_MINQUERY:
2149 case OP_MINQUERYI:
2150 case OP_POSQUERY:
2151 case OP_POSQUERYI:
2152 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2153 break;
2154 }
2155 #else
2156 (void)(utf); /* Keep compiler happy by referencing function argument */
2157 #endif
2158 }
2159 }
2160 }
2161
2162
2163
2164 /*************************************************
2165 * Scan compiled regex for recursion reference *
2166 *************************************************/
2167
2168 /* This little function scans through a compiled pattern until it finds an
2169 instance of OP_RECURSE.
2170
2171 Arguments:
2172 code points to start of expression
2173 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2174
2175 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2176 */
2177
2178 static const pcre_uchar *
2179 find_recurse(const pcre_uchar *code, BOOL utf)
2180 {
2181 for (;;)
2182 {
2183 register pcre_uchar c = *code;
2184 if (c == OP_END) return NULL;
2185 if (c == OP_RECURSE) return code;
2186
2187 /* XCLASS is used for classes that cannot be represented just by a bit
2188 map. This includes negated single high-valued characters. The length in
2189 the table is zero; the actual length is stored in the compiled code. */
2190
2191 if (c == OP_XCLASS) code += GET(code, 1);
2192
2193 /* Otherwise, we can get the item's length from the table, except that for
2194 repeated character types, we have to test for \p and \P, which have an extra
2195 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2196 must add in its length. */
2197
2198 else
2199 {
2200 switch(c)
2201 {
2202 case OP_TYPESTAR:
2203 case OP_TYPEMINSTAR:
2204 case OP_TYPEPLUS:
2205 case OP_TYPEMINPLUS:
2206 case OP_TYPEQUERY:
2207 case OP_TYPEMINQUERY:
2208 case OP_TYPEPOSSTAR:
2209 case OP_TYPEPOSPLUS:
2210 case OP_TYPEPOSQUERY:
2211 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2212 break;
2213
2214 case OP_TYPEPOSUPTO:
2215 case OP_TYPEUPTO:
2216 case OP_TYPEMINUPTO:
2217 case OP_TYPEEXACT:
2218 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2219 code += 2;
2220 break;
2221
2222 case OP_MARK:
2223 case OP_PRUNE_ARG:
2224 case OP_SKIP_ARG:
2225 case OP_THEN_ARG:
2226 code += code[1];
2227 break;
2228 }
2229
2230 /* Add in the fixed length from the table */
2231
2232 code += PRIV(OP_lengths)[c];
2233
2234 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2235 by a multi-byte character. The length in the table is a minimum, so we have
2236 to arrange to skip the extra bytes. */
2237
2238 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2239 if (utf) switch(c)
2240 {
2241 case OP_CHAR:
2242 case OP_CHARI:
2243 case OP_NOT:
2244 case OP_NOTI:
2245 case OP_EXACT:
2246 case OP_EXACTI:
2247 case OP_NOTEXACT:
2248 case OP_NOTEXACTI:
2249 case OP_UPTO:
2250 case OP_UPTOI:
2251 case OP_NOTUPTO:
2252 case OP_NOTUPTOI:
2253 case OP_MINUPTO:
2254 case OP_MINUPTOI:
2255 case OP_NOTMINUPTO:
2256 case OP_NOTMINUPTOI:
2257 case OP_POSUPTO:
2258 case OP_POSUPTOI:
2259 case OP_NOTPOSUPTO:
2260 case OP_NOTPOSUPTOI:
2261 case OP_STAR:
2262 case OP_STARI:
2263 case OP_NOTSTAR:
2264 case OP_NOTSTARI:
2265 case OP_MINSTAR:
2266 case OP_MINSTARI:
2267 case OP_NOTMINSTAR:
2268 case OP_NOTMINSTARI:
2269 case OP_POSSTAR:
2270 case OP_POSSTARI:
2271 case OP_NOTPOSSTAR:
2272 case OP_NOTPOSSTARI:
2273 case OP_PLUS:
2274 case OP_PLUSI:
2275 case OP_NOTPLUS:
2276 case OP_NOTPLUSI:
2277 case OP_MINPLUS:
2278 case OP_MINPLUSI:
2279 case OP_NOTMINPLUS:
2280 case OP_NOTMINPLUSI:
2281 case OP_POSPLUS:
2282 case OP_POSPLUSI:
2283 case OP_NOTPOSPLUS:
2284 case OP_NOTPOSPLUSI:
2285 case OP_QUERY:
2286 case OP_QUERYI:
2287 case OP_NOTQUERY:
2288 case OP_NOTQUERYI:
2289 case OP_MINQUERY:
2290 case OP_MINQUERYI:
2291 case OP_NOTMINQUERY:
2292 case OP_NOTMINQUERYI:
2293 case OP_POSQUERY:
2294 case OP_POSQUERYI:
2295 case OP_NOTPOSQUERY:
2296 case OP_NOTPOSQUERYI:
2297 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2298 break;
2299 }
2300 #else
2301 (void)(utf); /* Keep compiler happy by referencing function argument */
2302 #endif
2303 }
2304 }
2305 }
2306
2307
2308
2309 /*************************************************
2310 * Scan compiled branch for non-emptiness *
2311 *************************************************/
2312
2313 /* This function scans through a branch of a compiled pattern to see whether it
2314 can match the empty string or not. It is called from could_be_empty()
2315 below and from compile_branch() when checking for an unlimited repeat of a
2316 group that can match nothing. Note that first_significant_code() skips over
2317 backward and negative forward assertions when its final argument is TRUE. If we
2318 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2319 bracket whose current branch will already have been scanned.
2320
2321 Arguments:
2322 code points to start of search
2323 endcode points to where to stop
2324 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2325 cd contains pointers to tables etc.
2326 recurses chain of recurse_check to catch mutual recursion
2327
2328 Returns: TRUE if what is matched could be empty
2329 */
2330
2331 typedef struct recurse_check {
2332 struct recurse_check *prev;
2333 const pcre_uchar *group;
2334 } recurse_check;
2335
2336 static BOOL
2337 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2338 BOOL utf, compile_data *cd, recurse_check *recurses)
2339 {
2340 register pcre_uchar c;
2341 recurse_check this_recurse;
2342
2343 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2344 code < endcode;
2345 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2346 {
2347 const pcre_uchar *ccode;
2348
2349 c = *code;
2350
2351 /* Skip over forward assertions; the other assertions are skipped by
2352 first_significant_code() with a TRUE final argument. */
2353
2354 if (c == OP_ASSERT)
2355 {
2356 do code += GET(code, 1); while (*code == OP_ALT);
2357 c = *code;
2358 continue;
2359 }
2360
2361 /* For a recursion/subroutine call, if its end has been reached, which
2362 implies a backward reference subroutine call, we can scan it. If it's a
2363 forward reference subroutine call, we can't. To detect forward reference
2364 we have to scan up the list that is kept in the workspace. This function is
2365 called only when doing the real compile, not during the pre-compile that
2366 measures the size of the compiled pattern. */
2367
2368 if (c == OP_RECURSE)
2369 {
2370 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2371 BOOL empty_branch;
2372
2373 /* Test for forward reference or uncompleted reference. This is disabled
2374 when called to scan a completed pattern by setting cd->start_workspace to
2375 NULL. */
2376
2377 if (cd->start_workspace != NULL)
2378 {
2379 const pcre_uchar *tcode;
2380 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2381 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2382 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2383 }
2384
2385 /* If we are scanning a completed pattern, there are no forward references
2386 and all groups are complete. We need to detect whether this is a recursive
2387 call, as otherwise there will be an infinite loop. If it is a recursion,
2388 just skip over it. Simple recursions are easily detected. For mutual
2389 recursions we keep a chain on the stack. */
2390
2391 else
2392 {
2393 recurse_check *r = recurses;
2394 const pcre_uchar *endgroup = scode;
2395
2396 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2397 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2398
2399 for (r = recurses; r != NULL; r = r->prev)
2400 if (r->group == scode) break;
2401 if (r != NULL) continue; /* Mutual recursion */
2402 }
2403
2404 /* Completed reference; scan the referenced group, remembering it on the
2405 stack chain to detect mutual recursions. */
2406
2407 empty_branch = FALSE;
2408 this_recurse.prev = recurses;
2409 this_recurse.group = scode;
2410
2411 do
2412 {
2413 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2414 {
2415 empty_branch = TRUE;
2416 break;
2417 }
2418 scode += GET(scode, 1);
2419 }
2420 while (*scode == OP_ALT);
2421
2422 if (!empty_branch) return FALSE; /* All branches are non-empty */
2423 continue;
2424 }
2425
2426 /* Groups with zero repeats can of course be empty; skip them. */
2427
2428 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2429 c == OP_BRAPOSZERO)
2430 {
2431 code += PRIV(OP_lengths)[c];
2432 do code += GET(code, 1); while (*code == OP_ALT);
2433 c = *code;
2434 continue;
2435 }
2436
2437 /* A nested group that is already marked as "could be empty" can just be
2438 skipped. */
2439
2440 if (c == OP_SBRA || c == OP_SBRAPOS ||
2441 c == OP_SCBRA || c == OP_SCBRAPOS)
2442 {
2443 do code += GET(code, 1); while (*code == OP_ALT);
2444 c = *code;
2445 continue;
2446 }
2447
2448 /* For other groups, scan the branches. */
2449
2450 if (c == OP_BRA || c == OP_BRAPOS ||
2451 c == OP_CBRA || c == OP_CBRAPOS ||
2452 c == OP_ONCE || c == OP_ONCE_NC ||
2453 c == OP_COND)
2454 {
2455 BOOL empty_branch;
2456 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2457
2458 /* If a conditional group has only one branch, there is a second, implied,
2459 empty branch, so just skip over the conditional, because it could be empty.
2460 Otherwise, scan the individual branches of the group. */
2461
2462 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2463 code += GET(code, 1);
2464 else
2465 {
2466 empty_branch = FALSE;
2467 do
2468 {
2469 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2470 empty_branch = TRUE;
2471 code += GET(code, 1);
2472 }
2473 while (*code == OP_ALT);
2474 if (!empty_branch) return FALSE; /* All branches are non-empty */
2475 }
2476
2477 c = *code;
2478 continue;
2479 }
2480
2481 /* Handle the other opcodes */
2482
2483 switch (c)
2484 {
2485 /* Check for quantifiers after a class. XCLASS is used for classes that
2486 cannot be represented just by a bit map. This includes negated single
2487 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2488 actual length is stored in the compiled code, so we must update "code"
2489 here. */
2490
2491 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2492 case OP_XCLASS:
2493 ccode = code += GET(code, 1);
2494 goto CHECK_CLASS_REPEAT;
2495 #endif
2496
2497 case OP_CLASS:
2498 case OP_NCLASS:
2499 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2500
2501 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2502 CHECK_CLASS_REPEAT:
2503 #endif
2504
2505 switch (*ccode)
2506 {
2507 case OP_CRSTAR: /* These could be empty; continue */
2508 case OP_CRMINSTAR:
2509 case OP_CRQUERY:
2510 case OP_CRMINQUERY:
2511 case OP_CRPOSSTAR:
2512 case OP_CRPOSQUERY:
2513 break;
2514
2515 default: /* Non-repeat => class must match */
2516 case OP_CRPLUS: /* These repeats aren't empty */
2517 case OP_CRMINPLUS:
2518 case OP_CRPOSPLUS:
2519 return FALSE;
2520
2521 case OP_CRRANGE:
2522 case OP_CRMINRANGE:
2523 case OP_CRPOSRANGE:
2524 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2525 break;
2526 }
2527 break;
2528
2529 /* Opcodes that must match a character */
2530
2531 case OP_ANY:
2532 case OP_ALLANY:
2533 case OP_ANYBYTE:
2534
2535 case OP_PROP:
2536 case OP_NOTPROP:
2537 case OP_ANYNL:
2538
2539 case OP_NOT_HSPACE:
2540 case OP_HSPACE:
2541 case OP_NOT_VSPACE:
2542 case OP_VSPACE:
2543 case OP_EXTUNI:
2544
2545 case OP_NOT_DIGIT:
2546 case OP_DIGIT:
2547 case OP_NOT_WHITESPACE:
2548 case OP_WHITESPACE:
2549 case OP_NOT_WORDCHAR:
2550 case OP_WORDCHAR:
2551
2552 case OP_CHAR:
2553 case OP_CHARI:
2554 case OP_NOT:
2555 case OP_NOTI:
2556
2557 case OP_PLUS:
2558 case OP_PLUSI:
2559 case OP_MINPLUS:
2560 case OP_MINPLUSI:
2561
2562 case OP_NOTPLUS:
2563 case OP_NOTPLUSI:
2564 case OP_NOTMINPLUS:
2565 case OP_NOTMINPLUSI:
2566
2567 case OP_POSPLUS:
2568 case OP_POSPLUSI:
2569 case OP_NOTPOSPLUS:
2570 case OP_NOTPOSPLUSI:
2571
2572 case OP_EXACT:
2573 case OP_EXACTI:
2574 case OP_NOTEXACT:
2575 case OP_NOTEXACTI:
2576
2577 case OP_TYPEPLUS:
2578 case OP_TYPEMINPLUS:
2579 case OP_TYPEPOSPLUS:
2580 case OP_TYPEEXACT:
2581
2582 return FALSE;
2583
2584 /* These are going to continue, as they may be empty, but we have to
2585 fudge the length for the \p and \P cases. */
2586
2587 case OP_TYPESTAR:
2588 case OP_TYPEMINSTAR:
2589 case OP_TYPEPOSSTAR:
2590 case OP_TYPEQUERY:
2591 case OP_TYPEMINQUERY:
2592 case OP_TYPEPOSQUERY:
2593 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2594 break;
2595
2596 /* Same for these */
2597
2598 case OP_TYPEUPTO:
2599 case OP_TYPEMINUPTO:
2600 case OP_TYPEPOSUPTO:
2601 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2602 code += 2;
2603 break;
2604
2605 /* End of branch */
2606
2607 case OP_KET:
2608 case OP_KETRMAX:
2609 case OP_KETRMIN:
2610 case OP_KETRPOS:
2611 case OP_ALT:
2612 return TRUE;
2613
2614 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2615 MINUPTO, and POSUPTO and their caseless and negative versions may be
2616 followed by a multibyte character. */
2617
2618 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2619 case OP_STAR:
2620 case OP_STARI:
2621 case OP_NOTSTAR:
2622 case OP_NOTSTARI:
2623
2624 case OP_MINSTAR:
2625 case OP_MINSTARI:
2626 case OP_NOTMINSTAR:
2627 case OP_NOTMINSTARI:
2628
2629 case OP_POSSTAR:
2630 case OP_POSSTARI:
2631 case OP_NOTPOSSTAR:
2632 case OP_NOTPOSSTARI:
2633
2634 case OP_QUERY:
2635 case OP_QUERYI:
2636 case OP_NOTQUERY:
2637 case OP_NOTQUERYI:
2638
2639 case OP_MINQUERY:
2640 case OP_MINQUERYI:
2641 case OP_NOTMINQUERY:
2642 case OP_NOTMINQUERYI:
2643
2644 case OP_POSQUERY:
2645 case OP_POSQUERYI:
2646 case OP_NOTPOSQUERY:
2647 case OP_NOTPOSQUERYI:
2648
2649 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2650 break;
2651
2652 case OP_UPTO:
2653 case OP_UPTOI:
2654 case OP_NOTUPTO:
2655 case OP_NOTUPTOI:
2656
2657 case OP_MINUPTO:
2658 case OP_MINUPTOI:
2659 case OP_NOTMINUPTO:
2660 case OP_NOTMINUPTOI:
2661
2662 case OP_POSUPTO:
2663 case OP_POSUPTOI:
2664 case OP_NOTPOSUPTO:
2665 case OP_NOTPOSUPTOI:
2666
2667 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2668 break;
2669 #endif
2670
2671 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2672 string. */
2673
2674 case OP_MARK:
2675 case OP_PRUNE_ARG:
2676 case OP_SKIP_ARG:
2677 case OP_THEN_ARG:
2678 code += code[1];
2679 break;
2680
2681 /* None of the remaining opcodes are required to match a character. */
2682
2683 default:
2684 break;
2685 }
2686 }
2687
2688 return TRUE;
2689 }
2690
2691
2692
2693 /*************************************************
2694 * Scan compiled regex for non-emptiness *
2695 *************************************************/
2696
2697 /* This function is called to check for left recursive calls. We want to check
2698 the current branch of the current pattern to see if it could match the empty
2699 string. If it could, we must look outwards for branches at other levels,
2700 stopping when we pass beyond the bracket which is the subject of the recursion.
2701 This function is called only during the real compile, not during the
2702 pre-compile.
2703
2704 Arguments:
2705 code points to start of the recursion
2706 endcode points to where to stop (current RECURSE item)
2707 bcptr points to the chain of current (unclosed) branch starts
2708 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2709 cd pointers to tables etc
2710
2711 Returns: TRUE if what is matched could be empty
2712 */
2713
2714 static BOOL
2715 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2716 branch_chain *bcptr, BOOL utf, compile_data *cd)
2717 {
2718 while (bcptr != NULL && bcptr->current_branch >= code)
2719 {
2720 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2721 return FALSE;
2722 bcptr = bcptr->outer;
2723 }
2724 return TRUE;
2725 }
2726
2727
2728
2729 /*************************************************
2730 * Base opcode of repeated opcodes *
2731 *************************************************/
2732
2733 /* Returns the base opcode for repeated single character type opcodes. If the
2734 opcode is not a repeated character type, it returns with the original value.
2735
2736 Arguments: c opcode
2737 Returns: base opcode for the type
2738 */
2739
2740 static pcre_uchar
2741 get_repeat_base(pcre_uchar c)
2742 {
2743 return (c > OP_TYPEPOSUPTO)? c :
2744 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2745 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2746 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2747 (c >= OP_STARI)? OP_STARI :
2748 OP_STAR;
2749 }
2750
2751
2752
2753 #ifdef SUPPORT_UCP
2754 /*************************************************
2755 * Check a character and a property *
2756 *************************************************/
2757
2758 /* This function is called by check_auto_possessive() when a property item
2759 is adjacent to a fixed character.
2760
2761 Arguments:
2762 c the character
2763 ptype the property type
2764 pdata the data for the type
2765 negated TRUE if it's a negated property (\P or \p{^)
2766
2767 Returns: TRUE if auto-possessifying is OK
2768 */
2769
2770 static BOOL
2771 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2772 BOOL negated)
2773 {
2774 const pcre_uint32 *p;
2775 const ucd_record *prop = GET_UCD(c);
2776
2777 switch(ptype)
2778 {
2779 case PT_LAMP:
2780 return (prop->chartype == ucp_Lu ||
2781 prop->chartype == ucp_Ll ||
2782 prop->chartype == ucp_Lt) == negated;
2783
2784 case PT_GC:
2785 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2786
2787 case PT_PC:
2788 return (pdata == prop->chartype) == negated;
2789
2790 case PT_SC:
2791 return (pdata == prop->script) == negated;
2792
2793 /* These are specials */
2794
2795 case PT_ALNUM:
2796 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2797 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2798
2799 /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2800 means that Perl space and POSIX space are now identical. PCRE was changed
2801 at release 8.34. */
2802
2803 case PT_SPACE: /* Perl space */
2804 case PT_PXSPACE: /* POSIX space */
2805 switch(c)
2806 {
2807 HSPACE_CASES:
2808 VSPACE_CASES:
2809 return negated;
2810
2811 default:
2812 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2813 }
2814 break; /* Control never reaches here */
2815
2816 case PT_WORD:
2817 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2818 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2819 c == CHAR_UNDERSCORE) == negated;
2820
2821 case PT_CLIST:
2822 p = PRIV(ucd_caseless_sets) + prop->caseset;
2823 for (;;)
2824 {
2825 if (c < *p) return !negated;
2826 if (c == *p++) return negated;
2827 }
2828 break; /* Control never reaches here */
2829 }
2830
2831 return FALSE;
2832 }
2833 #endif /* SUPPORT_UCP */
2834
2835
2836
2837 /*************************************************
2838 * Fill the character property list *
2839 *************************************************/
2840
2841 /* Checks whether the code points to an opcode that can take part in auto-
2842 possessification, and if so, fills a list with its properties.
2843
2844 Arguments:
2845 code points to start of expression
2846 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2847 fcc points to case-flipping table
2848 list points to output list
2849 list[0] will be filled with the opcode
2850 list[1] will be non-zero if this opcode
2851 can match an empty character string
2852 list[2..7] depends on the opcode
2853
2854 Returns: points to the start of the next opcode if *code is accepted
2855 NULL if *code is not accepted
2856 */
2857
2858 static const pcre_uchar *
2859 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2860 const pcre_uint8 *fcc, pcre_uint32 *list)
2861 {
2862 pcre_uchar c = *code;
2863 const pcre_uchar *end;
2864 const pcre_uint32 *clist_src;
2865 pcre_uint32 *clist_dest;
2866 pcre_uint32 chr;
2867 pcre_uchar base;
2868
2869 list[0] = c;
2870 list[1] = FALSE;
2871 code++;
2872
2873 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2874 {
2875 base = get_repeat_base(c);
2876 c -= (base - OP_STAR);
2877
2878 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2879 code += IMM2_SIZE;
2880
2881 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2882
2883 switch(base)
2884 {
2885 case OP_STAR:
2886 list[0] = OP_CHAR;
2887 break;
2888
2889 case OP_STARI:
2890 list[0] = OP_CHARI;
2891 break;
2892
2893 case OP_NOTSTAR:
2894 list[0] = OP_NOT;
2895 break;
2896
2897 case OP_NOTSTARI:
2898 list[0] = OP_NOTI;
2899 break;
2900
2901 case OP_TYPESTAR:
2902 list[0] = *code;
2903 code++;
2904 break;
2905 }
2906 c = list[0];
2907 }
2908
2909 switch(c)
2910 {
2911 case OP_NOT_DIGIT:
2912 case OP_DIGIT:
2913 case OP_NOT_WHITESPACE:
2914 case OP_WHITESPACE:
2915 case OP_NOT_WORDCHAR:
2916 case OP_WORDCHAR:
2917 case OP_ANY:
2918 case OP_ALLANY:
2919 case OP_ANYNL:
2920 case OP_NOT_HSPACE:
2921 case OP_HSPACE:
2922 case OP_NOT_VSPACE:
2923 case OP_VSPACE:
2924 case OP_EXTUNI:
2925 case OP_EODN:
2926 case OP_EOD:
2927 case OP_DOLL:
2928 case OP_DOLLM:
2929 return code;
2930
2931 case OP_CHAR:
2932 case OP_NOT:
2933 GETCHARINCTEST(chr, code);
2934 list[2] = chr;
2935 list[3] = NOTACHAR;
2936 return code;
2937
2938 case OP_CHARI:
2939 case OP_NOTI:
2940 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2941 GETCHARINCTEST(chr, code);
2942 list[2] = chr;
2943
2944 #ifdef SUPPORT_UCP
2945 if (chr < 128 || (chr < 256 && !utf))
2946 list[3] = fcc[chr];
2947 else
2948 list[3] = UCD_OTHERCASE(chr);
2949 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2950 list[3] = (chr < 256) ? fcc[chr] : chr;
2951 #else
2952 list[3] = fcc[chr];
2953 #endif
2954
2955 /* The othercase might be the same value. */
2956
2957 if (chr == list[3])
2958 list[3] = NOTACHAR;
2959 else
2960 list[4] = NOTACHAR;
2961 return code;
2962
2963 #ifdef SUPPORT_UCP
2964 case OP_PROP:
2965 case OP_NOTPROP:
2966 if (code[0] != PT_CLIST)
2967 {
2968 list[2] = code[0];
2969 list[3] = code[1];
2970 return code + 2;
2971 }
2972
2973 /* Convert only if we have enough space. */
2974
2975 clist_src = PRIV(ucd_caseless_sets) + code[1];
2976 clist_dest = list + 2;
2977 code += 2;
2978
2979 do {
2980 if (clist_dest >= list + 8)
2981 {
2982 /* Early return if there is not enough space. This should never
2983 happen, since all clists are shorter than 5 character now. */
2984 list[2] = code[0];
2985 list[3] = code[1];
2986 return code;
2987 }
2988 *clist_dest++ = *clist_src;
2989 }
2990 while(*clist_src++ != NOTACHAR);
2991
2992 /* All characters are stored. The terminating NOTACHAR
2993 is copied form the clist itself. */
2994
2995 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2996 return code;
2997 #endif
2998
2999 case OP_NCLASS:
3000 case OP_CLASS:
3001 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3002 case OP_XCLASS:
3003 if (c == OP_XCLASS)
3004 end = code + GET(code, 0) - 1;
3005 else
3006 #endif
3007 end = code + 32 / sizeof(pcre_uchar);
3008
3009 switch(*end)
3010 {
3011 case OP_CRSTAR:
3012 case OP_CRMINSTAR:
3013 case OP_CRQUERY:
3014 case OP_CRMINQUERY:
3015 case OP_CRPOSSTAR:
3016 case OP_CRPOSQUERY:
3017 list[1] = TRUE;
3018 end++;
3019 break;
3020
3021 case OP_CRPLUS:
3022 case OP_CRMINPLUS:
3023 case OP_CRPOSPLUS:
3024 end++;
3025 break;
3026
3027 case OP_CRRANGE:
3028 case OP_CRMINRANGE:
3029 case OP_CRPOSRANGE:
3030 list[1] = (GET2(end, 1) == 0);
3031 end += 1 + 2 * IMM2_SIZE;
3032 break;
3033 }
3034 list[2] = end - code;
3035 return end;
3036 }
3037 return NULL; /* Opcode not accepted */
3038 }
3039
3040
3041
3042 /*************************************************
3043 * Scan further character sets for match *
3044 *************************************************/
3045
3046 /* Checks whether the base and the current opcode have a common character, in
3047 which case the base cannot be possessified.
3048
3049 Arguments:
3050 code points to the byte code
3051 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3052 cd static compile data
3053 base_list the data list of the base opcode
3054
3055 Returns: TRUE if the auto-possessification is possible
3056 */
3057
3058 static BOOL
3059 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3060 const pcre_uint32 *base_list, const pcre_uchar *base_end)
3061 {
3062 pcre_uchar c;
3063 pcre_uint32 list[8];
3064 const pcre_uint32 *chr_ptr;
3065 const pcre_uint32 *ochr_ptr;
3066 const pcre_uint32 *list_ptr;
3067 const pcre_uchar *next_code;
3068 const pcre_uint8 *class_bitset;
3069 const pcre_uint32 *set1, *set2, *set_end;
3070 pcre_uint32 chr;
3071 BOOL accepted, invert_bits;
3072
3073 /* Note: the base_list[1] contains whether the current opcode has greedy
3074 (represented by a non-zero value) quantifier. This is a different from
3075 other character type lists, which stores here that the character iterator
3076 matches to an empty string (also represented by a non-zero value). */
3077
3078 for(;;)
3079 {
3080 /* All operations move the code pointer forward.
3081 Therefore infinite recursions are not possible. */
3082
3083 c = *code;
3084
3085 /* Skip over callouts */
3086
3087 if (c == OP_CALLOUT)
3088 {
3089 code += PRIV(OP_lengths)[c];
3090 continue;
3091 }
3092
3093 if (c == OP_ALT)
3094 {
3095 do code += GET(code, 1); while (*code == OP_ALT);
3096 c = *code;
3097 }
3098
3099 switch(c)
3100 {
3101 case OP_END:
3102 case OP_KETRPOS:
3103 /* TRUE only in greedy case. The non-greedy case could be replaced by
3104 an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3105 uses more memory, which we cannot get at this stage.) */
3106
3107 return base_list[1] != 0;
3108
3109 case OP_KET:
3110 /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3111 it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3112 cannot be converted to a possessive form. */
3113
3114 if (base_list[1] == 0) return FALSE;
3115
3116 switch(*(code - GET(code, 1)))
3117 {
3118 case OP_ASSERT:
3119 case OP_ASSERT_NOT:
3120 case OP_ASSERTBACK:
3121 case OP_ASSERTBACK_NOT:
3122 case OP_ONCE:
3123 case OP_ONCE_NC:
3124 /* Atomic sub-patterns and assertions can always auto-possessify their
3125 last iterator. */
3126 return TRUE;
3127 }
3128
3129 code += PRIV(OP_lengths)[c];
3130 continue;
3131
3132 case OP_ONCE:
3133 case OP_ONCE_NC:
3134 case OP_BRA:
3135 case OP_CBRA:
3136 next_code = code + GET(code, 1);
3137 code += PRIV(OP_lengths)[c];
3138
3139 while (*next_code == OP_ALT)
3140 {
3141 if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3142 code = next_code + 1 + LINK_SIZE;
3143 next_code += GET(next_code, 1);
3144 }
3145 continue;
3146
3147 case OP_BRAZERO:
3148 case OP_BRAMINZERO:
3149
3150 next_code = code + 1;
3151 if (*next_code != OP_BRA && *next_code != OP_CBRA
3152 && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3153
3154 do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3155
3156 /* The bracket content will be checked by the
3157 OP_BRA/OP_CBRA case above. */
3158 next_code += 1 + LINK_SIZE;
3159 if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3160 return FALSE;
3161
3162 code += PRIV(OP_lengths)[c];
3163 continue;
3164 }
3165
3166 /* Check for a supported opcode, and load its properties. */
3167
3168 code = get_chr_property_list(code, utf, cd->fcc, list);
3169 if (code == NULL) return FALSE; /* Unsupported */
3170
3171 /* If either opcode is a small character list, set pointers for comparing
3172 characters from that list with another list, or with a property. */
3173
3174 if (base_list[0] == OP_CHAR)
3175 {
3176 chr_ptr = base_list + 2;
3177 list_ptr = list;
3178 }
3179 else if (list[0] == OP_CHAR)
3180 {
3181 chr_ptr = list + 2;
3182 list_ptr = base_list;
3183 }
3184
3185 /* Character bitsets can also be compared to certain opcodes. */
3186
3187 else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3188 #ifdef COMPILE_PCRE8
3189 /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3190 || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3191 #endif
3192 )
3193 {
3194 #ifdef COMPILE_PCRE8
3195 if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3196 #else
3197 if (base_list[0] == OP_CLASS)
3198 #endif
3199 {
3200 set1 = (pcre_uint32 *)(base_end - base_list[2]);
3201 list_ptr = list;
3202 }
3203 else
3204 {
3205 set1 = (pcre_uint32 *)(code - list[2]);
3206 list_ptr = base_list;
3207 }
3208
3209 invert_bits = FALSE;
3210 switch(list_ptr[0])
3211 {
3212 case OP_CLASS:
3213 case OP_NCLASS:
3214 set2 = (pcre_uint32 *)
3215 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3216 break;
3217
3218 /* OP_XCLASS cannot be supported here, because its bitset
3219 is not necessarily complete. E.g: [a-\0x{200}] is stored
3220 as a character range, and the appropriate bits are not set. */
3221
3222 case OP_NOT_DIGIT:
3223 invert_bits = TRUE;
3224 /* Fall through */
3225 case OP_DIGIT:
3226 set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3227 break;
3228
3229 case OP_NOT_WHITESPACE:
3230 invert_bits = TRUE;
3231 /* Fall through */
3232 case OP_WHITESPACE:
3233 set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3234 break;
3235
3236 case OP_NOT_WORDCHAR:
3237 invert_bits = TRUE;
3238 /* Fall through */
3239 case OP_WORDCHAR:
3240 set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3241 break;
3242
3243 default:
3244 return FALSE;
3245 }
3246
3247 /* Compare 4 bytes to improve speed. */
3248 set_end = set1 + (32 / 4);
3249 if (invert_bits)
3250 {
3251 do
3252 {
3253 if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3254 }
3255 while (set1 < set_end);
3256 }
3257 else
3258 {
3259 do
3260 {
3261 if ((*set1++ & *set2++) != 0) return FALSE;
3262 }
3263 while (set1 < set_end);
3264 }
3265
3266 if (list[1] == 0) return TRUE;
3267 /* Might be an empty repeat. */
3268 continue;
3269 }
3270
3271 /* Some property combinations also acceptable. Unicode property opcodes are
3272 processed specially; the rest can be handled with a lookup table. */
3273
3274 else
3275 {
3276 pcre_uint32 leftop, rightop;
3277
3278 leftop = base_list[0];
3279 rightop = list[0];
3280
3281 #ifdef SUPPORT_UCP
3282 accepted = FALSE; /* Always set in non-unicode case. */
3283 if (leftop == OP_PROP || leftop == OP_NOTPROP)
3284 {
3285 if (rightop == OP_EOD)
3286 accepted = TRUE;
3287 else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3288 {
3289 int n;
3290 const pcre_uint8 *p;
3291 BOOL same = leftop == rightop;
3292 BOOL lisprop = leftop == OP_PROP;
3293 BOOL risprop = rightop == OP_PROP;
3294 BOOL bothprop = lisprop && risprop;
3295
3296 /* There's a table that specifies how each combination is to be
3297 processed:
3298 0 Always return FALSE (never auto-possessify)
3299 1 Character groups are distinct (possessify if both are OP_PROP)
3300 2 Check character categories in the same group (general or particular)
3301 3 Return TRUE if the two opcodes are not the same
3302 ... see comments below
3303 */
3304
3305 n = propposstab[base_list[2]][list[2]];
3306 switch(n)
3307 {
3308 case 0: break;
3309 case 1: accepted = bothprop; break;
3310 case 2: accepted = (base_list[3] == list[3]) != same; break;
3311 case 3: accepted = !same; break;
3312
3313 case 4: /* Left general category, right particular category */
3314 accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3315 break;
3316
3317 case 5: /* Right general category, left particular category */
3318 accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3319 break;
3320
3321 /* This code is logically tricky. Think hard before fiddling with it.
3322 The posspropstab table has four entries per row. Each row relates to
3323 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3324 Only WORD actually needs all four entries, but using repeats for the
3325 others means they can all use the same code below.
3326
3327 The first two entries in each row are Unicode general categories, and
3328 apply always, because all the characters they include are part of the
3329 PCRE character set. The third and fourth entries are a general and a
3330 particular category, respectively, that include one or more relevant
3331 characters. One or the other is used, depending on whether the check
3332 is for a general or a particular category. However, in both cases the
3333 category contains more characters than the specials that are defined
3334 for the property being tested against. Therefore, it cannot be used
3335 in a NOTPROP case.
3336
3337 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3338 Underscore is covered by ucp_P or ucp_Po. */
3339
3340 case 6: /* Left alphanum vs right general category */
3341 case 7: /* Left space vs right general category */
3342 case 8: /* Left word vs right general category */
3343 p = posspropstab[n-6];
3344 accepted = risprop && lisprop ==
3345 (list[3] != p[0] &&
3346 list[3] != p[1] &&
3347 (list[3] != p[2] || !lisprop));
3348 break;
3349
3350 case 9: /* Right alphanum vs left general category */
3351 case 10: /* Right space vs left general category */
3352 case 11: /* Right word vs left general category */
3353 p = posspropstab[n-9];
3354 accepted = lisprop && risprop ==
3355 (base_list[3] != p[0] &&
3356 base_list[3] != p[1] &&
3357 (base_list[3] != p[2] || !risprop));
3358 break;
3359
3360 case 12: /* Left alphanum vs right particular category */
3361 case 13: /* Left space vs right particular category */
3362 case 14: /* Left word vs right particular category */
3363 p = posspropstab[n-12];
3364 accepted = risprop && lisprop ==
3365 (catposstab[p[0]][list[3]] &&
3366 catposstab[p[1]][list[3]] &&
3367 (list[3] != p[3] || !lisprop));
3368 break;
3369
3370 case 15: /* Right alphanum vs left particular category */
3371 case 16: /* Right space vs left particular category */
3372 case 17: /* Right word vs left particular category */
3373 p = posspropstab[n-15];
3374 accepted = lisprop && risprop ==
3375 (catposstab[p[0]][base_list[3]] &&
3376 catposstab[p[1]][base_list[3]] &&
3377 (base_list[3] != p[3] || !risprop));
3378 break;
3379 }
3380 }
3381 }
3382
3383 else
3384 #endif /* SUPPORT_UCP */
3385
3386 accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3387 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3388 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3389
3390 if (!accepted)
3391 return FALSE;
3392
3393 if (list[1] == 0) return TRUE;
3394 /* Might be an empty repeat. */
3395 continue;
3396 }
3397
3398 /* Control reaches here only if one of the items is a small character list.
3399 All characters are checked against the other side. */
3400
3401 do
3402 {
3403 chr = *chr_ptr;
3404
3405 switch(list_ptr[0])
3406 {
3407 case OP_CHAR:
3408 ochr_ptr = list_ptr + 2;
3409 do
3410 {
3411 if (chr == *ochr_ptr) return FALSE;
3412 ochr_ptr++;
3413 }
3414 while(*ochr_ptr != NOTACHAR);
3415 break;
3416
3417 case OP_NOT:
3418 ochr_ptr = list_ptr + 2;
3419 do
3420 {
3421 if (chr == *ochr_ptr)
3422 break;
3423 ochr_ptr++;
3424 }
3425 while(*ochr_ptr != NOTACHAR);
3426 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3427 break;
3428
3429 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3430 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3431
3432 case OP_DIGIT:
3433 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3434 break;
3435
3436 case OP_NOT_DIGIT:
3437 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3438 break;
3439
3440 case OP_WHITESPACE:
3441 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3442 break;
3443
3444 case OP_NOT_WHITESPACE:
3445 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3446 break;
3447
3448 case OP_WORDCHAR:
3449 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3450 break;
3451
3452 case OP_NOT_WORDCHAR:
3453 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3454 break;
3455
3456 case OP_HSPACE:
3457 switch(chr)
3458 {
3459 HSPACE_CASES: return FALSE;
3460 default: break;
3461 }
3462 break;
3463
3464 case OP_NOT_HSPACE:
3465 switch(chr)
3466 {
3467 HSPACE_CASES: break;
3468 default: return FALSE;
3469 }
3470 break;
3471
3472 case OP_ANYNL:
3473 case OP_VSPACE:
3474 switch(chr)
3475 {
3476 VSPACE_CASES: return FALSE;
3477 default: break;
3478 }
3479 break;
3480
3481 case OP_NOT_VSPACE:
3482 switch(chr)
3483 {
3484 VSPACE_CASES: break;
3485 default: return FALSE;
3486 }
3487 break;
3488
3489 case OP_DOLL:
3490 case OP_EODN:
3491 switch (chr)
3492 {
3493 case CHAR_CR:
3494 case CHAR_LF:
3495 case CHAR_VT:
3496 case CHAR_FF:
3497 case CHAR_NEL:
3498 #ifndef EBCDIC
3499 case 0x2028:
3500 case 0x2029:
3501 #endif /* Not EBCDIC */
3502 return FALSE;
3503 }
3504 break;
3505
3506 case OP_EOD: /* Can always possessify before \z */
3507 break;
3508
3509 #ifdef SUPPORT_UCP
3510 case OP_PROP:
3511 case OP_NOTPROP:
3512 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3513 list_ptr[0] == OP_NOTPROP))
3514 return FALSE;
3515 break;
3516 #endif
3517
3518 case OP_NCLASS:
3519 if (chr > 255) return FALSE;
3520 /* Fall through */
3521
3522 case OP_CLASS:
3523 if (chr > 255) break;
3524 class_bitset = (pcre_uint8 *)
3525 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3526 if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3527 break;
3528
3529 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3530 case OP_XCLASS:
3531 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3532 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3533 break;
3534 #endif
3535
3536 default:
3537 return FALSE;
3538 }
3539
3540 chr_ptr++;
3541 }
3542 while(*chr_ptr != NOTACHAR);
3543
3544 /* At least one character must be matched from this opcode. */
3545
3546 if (list[1] == 0) return TRUE;
3547 }
3548
3549 return FALSE;
3550 }
3551
3552
3553
3554 /*************************************************
3555 * Scan compiled regex for auto-possession *
3556 *************************************************/
3557
3558 /* Replaces single character iterations with their possessive alternatives
3559 if appropriate. This function modifies the compiled opcode!
3560
3561 Arguments:
3562 code points to start of the byte code
3563 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3564 cd static compile data
3565
3566 Returns: nothing
3567 */
3568
3569 static void
3570 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3571 {
3572 register pcre_uchar c;
3573 const pcre_uchar *end;
3574 pcre_uchar *repeat_opcode;
3575 pcre_uint32 list[8];
3576
3577 for (;;)
3578 {
3579 c = *code;
3580
3581 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3582 {
3583 c -= get_repeat_base(c) - OP_STAR;
3584 end = (c <= OP_MINUPTO) ?
3585 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3586 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3587
3588 if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3589 {
3590 switch(c)
3591 {
3592 case OP_STAR:
3593 *code += OP_POSSTAR - OP_STAR;
3594 break;
3595
3596 case OP_MINSTAR:
3597 *code += OP_POSSTAR - OP_MINSTAR;
3598 break;
3599
3600 case OP_PLUS:
3601 *code += OP_POSPLUS - OP_PLUS;
3602 break;
3603
3604 case OP_MINPLUS:
3605 *code += OP_POSPLUS - OP_MINPLUS;
3606 break;
3607
3608 case OP_QUERY:
3609 *code += OP_POSQUERY - OP_QUERY;
3610 break;
3611
3612 case OP_MINQUERY:
3613 *code += OP_POSQUERY - OP_MINQUERY;
3614 break;
3615
3616 case OP_UPTO:
3617 *code += OP_POSUPTO - OP_UPTO;
3618 break;
3619
3620 case OP_MINUPTO:
3621 *code += OP_MINUPTO - OP_UPTO;
3622 break;
3623 }
3624 }
3625 c = *code;
3626 }
3627 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3628 {
3629 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3630 if (c == OP_XCLASS)
3631 repeat_opcode = code + GET(code, 1);
3632 else
3633 #endif
3634 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3635
3636 c = *repeat_opcode;
3637 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3638 {
3639 /* end must not be NULL. */
3640 end = get_chr_property_list(code, utf, cd->fcc, list);
3641
3642 list[1] = (c & 1) == 0;
3643
3644 if (compare_opcodes(end, utf, cd, list, end))
3645 {
3646 switch (c)
3647 {
3648 case OP_CRSTAR:
3649 case OP_CRMINSTAR:
3650 *repeat_opcode = OP_CRPOSSTAR;
3651 break;
3652
3653 case OP_CRPLUS:
3654 case OP_CRMINPLUS:
3655 *repeat_opcode = OP_CRPOSPLUS;
3656 break;
3657
3658 case OP_CRQUERY:
3659 case OP_CRMINQUERY:
3660 *repeat_opcode = OP_CRPOSQUERY;
3661 break;
3662
3663 case OP_CRRANGE:
3664 case OP_CRMINRANGE:
3665 *repeat_opcode = OP_CRPOSRANGE;
3666 break;
3667 }
3668 }
3669 }
3670 c = *code;
3671 }
3672
3673 switch(c)
3674 {
3675 case OP_END:
3676 return;
3677
3678 case OP_TYPESTAR:
3679 case OP_TYPEMINSTAR:
3680 case OP_TYPEPLUS:
3681 case OP_TYPEMINPLUS:
3682 case OP_TYPEQUERY:
3683 case OP_TYPEMINQUERY:
3684 case OP_TYPEPOSSTAR:
3685 case OP_TYPEPOSPLUS:
3686 case OP_TYPEPOSQUERY:
3687 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3688 break;
3689
3690 case OP_TYPEUPTO:
3691 case OP_TYPEMINUPTO:
3692 case OP_TYPEEXACT:
3693 case OP_TYPEPOSUPTO:
3694 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3695 code += 2;
3696 break;
3697
3698 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3699 case OP_XCLASS:
3700 code += GET(code, 1);
3701 break;
3702 #endif
3703
3704 case OP_MARK:
3705 case OP_PRUNE_ARG:
3706 case OP_SKIP_ARG:
3707 case OP_THEN_ARG:
3708 code += code[1];
3709 break;
3710 }
3711
3712 /* Add in the fixed length from the table */
3713
3714 code += PRIV(OP_lengths)[c];
3715
3716 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3717 a multi-byte character. The length in the table is a minimum, so we have to
3718 arrange to skip the extra bytes. */
3719
3720 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3721 if (utf) switch(c)
3722 {
3723 case OP_CHAR:
3724 case OP_CHARI:
3725 case OP_NOT:
3726 case OP_NOTI:
3727 case OP_STAR:
3728 case OP_MINSTAR:
3729 case OP_PLUS:
3730 case OP_MINPLUS:
3731 case OP_QUERY:
3732 case OP_MINQUERY:
3733 case OP_UPTO:
3734 case OP_MINUPTO:
3735 case OP_EXACT:
3736 case OP_POSSTAR:
3737 case OP_POSPLUS:
3738 case OP_POSQUERY:
3739 case OP_POSUPTO:
3740 case OP_STARI:
3741 case OP_MINSTARI:
3742 case OP_PLUSI:
3743 case OP_MINPLUSI:
3744 case OP_QUERYI:
3745 case OP_MINQUERYI:
3746 case OP_UPTOI:
3747 case OP_MINUPTOI:
3748 case OP_EXACTI:
3749 case OP_POSSTARI:
3750 case OP_POSPLUSI:
3751 case OP_POSQUERYI:
3752 case OP_POSUPTOI:
3753 case OP_NOTSTAR:
3754 case OP_NOTMINSTAR:
3755 case OP_NOTPLUS:
3756 case OP_NOTMINPLUS:
3757 case OP_NOTQUERY:
3758 case OP_NOTMINQUERY:
3759 case OP_NOTUPTO:
3760 case OP_NOTMINUPTO:
3761 case OP_NOTEXACT:
3762 case OP_NOTPOSSTAR:
3763 case OP_NOTPOSPLUS:
3764 case OP_NOTPOSQUERY:
3765 case OP_NOTPOSUPTO:
3766 case OP_NOTSTARI:
3767 case OP_NOTMINSTARI:
3768 case OP_NOTPLUSI:
3769 case OP_NOTMINPLUSI:
3770 case OP_NOTQUERYI:
3771 case OP_NOTMINQUERYI:
3772 case OP_NOTUPTOI:
3773 case OP_NOTMINUPTOI:
3774 case OP_NOTEXACTI:
3775 case OP_NOTPOSSTARI:
3776 case OP_NOTPOSPLUSI:
3777 case OP_NOTPOSQUERYI:
3778 case OP_NOTPOSUPTOI:
3779 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3780 break;
3781 }
3782 #else
3783 (void)(utf); /* Keep compiler happy by referencing function argument */
3784 #endif
3785 }
3786 }
3787
3788
3789
3790 /*************************************************
3791 * Check for POSIX class syntax *
3792 *************************************************/
3793
3794 /* This function is called when the sequence "[:" or "[." or "[=" is
3795 encountered in a character class. It checks whether this is followed by a
3796 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3797 reach an unescaped ']' without the special preceding character, return FALSE.
3798
3799 Originally, this function only recognized a sequence of letters between the
3800 terminators, but it seems that Perl recognizes any sequence of characters,
3801 though of course unknown POSIX names are subsequently rejected. Perl gives an
3802 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3803 didn't consider this to be a POSIX class. Likewise for [:1234:].
3804
3805 The problem in trying to be exactly like Perl is in the handling of escapes. We
3806 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3807 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3808 below handles the special case of \], but does not try to do any other escape
3809 processing. This makes it different from Perl for cases such as [:l\ower:]
3810 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3811 "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3812 I think.
3813
3814 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3815 It seems that the appearance of a nested POSIX class supersedes an apparent
3816 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3817 a digit.
3818
3819 In Perl, unescaped square brackets may also appear as part of class names. For
3820 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3821 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3822 seem right at all. PCRE does not allow closing square brackets in POSIX class
3823 names.
3824
3825 Arguments:
3826 ptr pointer to the initial [
3827 endptr where to return the end pointer
3828
3829 Returns: TRUE or FALSE
3830 */
3831
3832 static BOOL
3833 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3834 {
3835 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3836 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3837 for (++ptr; *ptr != CHAR_NULL; ptr++)
3838 {
3839 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3840 ptr++;
3841 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3842 else
3843 {
3844 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3845 {
3846 *endptr = ptr;
3847 return TRUE;
3848 }
3849 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3850 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3851 ptr[1] == CHAR_EQUALS_SIGN) &&
3852 check_posix_syntax(ptr, endptr))
3853 return FALSE;
3854 }
3855 }
3856 return FALSE;
3857 }
3858
3859
3860
3861
3862 /*************************************************
3863 * Check POSIX class name *
3864 *************************************************/
3865
3866 /* This function is called to check the name given in a POSIX-style class entry
3867 such as [:alnum:].
3868
3869 Arguments:
3870 ptr points to the first letter
3871 len the length of the name
3872
3873 Returns: a value representing the name, or -1 if unknown
3874 */
3875
3876 static int
3877 check_posix_name(const pcre_uchar *ptr, int len)
3878 {
3879 const char *pn = posix_names;
3880 register int yield = 0;
3881 while (posix_name_lengths[yield] != 0)
3882 {
3883 if (len == posix_name_lengths[yield] &&
3884 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3885 pn += posix_name_lengths[yield] + 1;
3886 yield++;
3887 }
3888 return -1;
3889 }
3890
3891
3892 /*************************************************
3893 * Adjust OP_RECURSE items in repeated group *
3894 *************************************************/
3895
3896 /* OP_RECURSE items contain an offset from the start of the regex to the group
3897 that is referenced. This means that groups can be replicated for fixed
3898 repetition simply by copying (because the recursion is allowed to refer to
3899 earlier groups that are outside the current group). However, when a group is
3900 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3901 inserted before it, after it has been compiled. This means that any OP_RECURSE
3902 items within it that refer to the group itself or any contained groups have to
3903 have their offsets adjusted. That one of the jobs of this function. Before it
3904 is called, the partially compiled regex must be temporarily terminated with
3905 OP_END.
3906
3907 This function has been extended with the possibility of forward references for
3908 recursions and subroutine calls. It must also check the list of such references
3909 for the group we are dealing with. If it finds that one of the recursions in
3910 the current group is on this list, it adjusts the offset in the list, not the
3911 value in the reference (which is a group number).
3912
3913 Arguments:
3914 group points to the start of the group
3915 adjust the amount by which the group is to be moved
3916 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3917 cd contains pointers to tables etc.
3918 save_hwm the hwm forward reference pointer at the start of the group
3919
3920 Returns: nothing
3921 */
3922
3923 static void
3924 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3925 pcre_uchar *save_hwm)
3926 {
3927 pcre_uchar *ptr = group;
3928
3929 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3930 {
3931 int offset;
3932 pcre_uchar *hc;
3933
3934 /* See if this recursion is on the forward reference list. If so, adjust the
3935 reference. */
3936
3937 for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3938 {
3939 offset = (int)GET(hc, 0);
3940 if (cd->start_code + offset == ptr + 1)
3941 {
3942 PUT(hc, 0, offset + adjust);
3943 break;
3944 }
3945 }
3946
3947 /* Otherwise, adjust the recursion offset if it's after the start of this
3948 group. */
3949
3950 if (hc >= cd->hwm)
3951 {
3952 offset = (int)GET(ptr, 1);
3953 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3954 }
3955
3956 ptr += 1 + LINK_SIZE;
3957 }
3958 }
3959
3960
3961
3962 /*************************************************
3963 * Insert an automatic callout point *
3964 *************************************************/
3965
3966 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3967 callout points before each pattern item.
3968
3969 Arguments:
3970 code current code pointer
3971 ptr current pattern pointer
3972 cd pointers to tables etc
3973
3974 Returns: new code pointer
3975 */
3976
3977 static pcre_uchar *
3978 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3979 {
3980 *code++ = OP_CALLOUT;
3981 *code++ = 255;
3982 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
3983 PUT(code, LINK_SIZE, 0); /* Default length */
3984 return code + 2 * LINK_SIZE;
3985 }
3986
3987
3988
3989 /*************************************************
3990 * Complete a callout item *
3991 *************************************************/
3992
3993 /* A callout item contains the length of the next item in the pattern, which
3994 we can't fill in till after we have reached the relevant point. This is used
3995 for both automatic and manual callouts.
3996
3997 Arguments:
3998 previous_callout points to previous callout item
3999 ptr current pattern pointer
4000 cd pointers to tables etc
4001
4002 Returns: nothing
4003 */
4004
4005 static void
4006 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4007 {
4008 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4009 PUT(previous_callout, 2 + LINK_SIZE, length);
4010 }
4011
4012
4013
4014 #ifdef SUPPORT_UCP
4015 /*************************************************
4016 * Get othercase range *
4017 *************************************************/
4018
4019 /* This function is passed the start and end of a class range, in UTF-8 mode
4020 with UCP support. It searches up the characters, looking for ranges of
4021 characters in the "other" case. Each call returns the next one, updating the
4022 start address. A character with multiple other cases is returned on its own
4023 with a special return value.
4024
4025 Arguments:
4026 cptr points to starting character value; updated
4027 d end value
4028 ocptr where to put start of othercase range
4029 odptr where to put end of othercase range
4030
4031 Yield: -1 when no more
4032 0 when a range is returned
4033 >0 the CASESET offset for char with multiple other cases
4034 in this case, ocptr contains the original
4035 */
4036
4037 static int
4038 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4039 pcre_uint32 *odptr)
4040 {
4041 pcre_uint32 c, othercase, next;
4042 unsigned int co;
4043
4044 /* Find the first character that has an other case. If it has multiple other
4045 cases, return its case offset value. */
4046
4047 for (c = *cptr; c <= d; c++)
4048 {
4049 if ((co = UCD_CASESET(c)) != 0)
4050 {
4051 *ocptr = c++; /* Character that has the set */
4052 *cptr = c; /* Rest of input range */
4053 return (int)co;
4054 }
4055 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4056 }
4057
4058 if (c > d) return -1; /* Reached end of range */
4059
4060 *ocptr = othercase;
4061 next = othercase + 1;
4062
4063 for (++c; c <= d; c++)
4064 {
4065 if (UCD_OTHERCASE(c) != next) break;
4066 next++;
4067 }
4068
4069 *odptr = next - 1; /* End of othercase range */
4070 *cptr = c; /* Rest of input range */
4071 return 0;
4072 }
4073 #endif /* SUPPORT_UCP */
4074
4075
4076
4077 /*************************************************
4078 * Add a character or range to a class *
4079 *************************************************/
4080
4081 /* This function packages up the logic of adding a character or range of
4082 characters to a class. The character values in the arguments will be within the
4083 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4084 mutually recursive with the function immediately below.
4085
4086 Arguments:
4087 classbits the bit map for characters < 256
4088 uchardptr points to the pointer for extra data
4089 options the options word
4090 cd contains pointers to tables etc.
4091 start start of range character
4092 end end of range character
4093
4094 Returns: the number of < 256 characters added
4095 the pointer to extra data is updated
4096 */
4097
4098 static int
4099 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4100 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4101 {
4102 pcre_uint32 c;
4103 int n8 = 0;
4104
4105 /* If caseless matching is required, scan the range and process alternate
4106 cases. In Unicode, there are 8-bit characters that have alternate cases that
4107 are greater than 255 and vice-versa. Sometimes we can just extend the original
4108 range. */
4109
4110 if ((options & PCRE_CASELESS) != 0)
4111 {
4112 #ifdef SUPPORT_UCP
4113 if ((options & PCRE_UTF8) != 0)
4114 {
4115 int rc;
4116 pcre_uint32 oc, od;
4117
4118 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4119 c = start;
4120
4121 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4122 {
4123 /* Handle a single character that has more than one other case. */
4124
4125 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4126 PRIV(ucd_caseless_sets) + rc, oc);
4127
4128 /* Do nothing if the other case range is within the original range. */
4129
4130 else if (oc >= start && od <= end) continue;
4131
4132 /* Extend the original range if there is overlap, noting that if oc < c, we
4133 can't have od > end because a subrange is always shorter than the basic
4134 range. Otherwise, use a recursive call to add the additional range. */
4135
4136 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4137 else if (od > end && oc <= end + 1) end = od; /* Extend upwards */
4138 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4139 }
4140 }
4141 else
4142 #endif /* SUPPORT_UCP */
4143
4144 /* Not UTF-mode, or no UCP */
4145
4146 for (c = start; c <= end && c < 256; c++)
4147 {
4148 SETBIT(classbits, cd->fcc[c]);
4149 n8++;
4150 }
4151 }
4152
4153 /* Now handle the original range. Adjust the final value according to the bit
4154 length - this means that the same lists of (e.g.) horizontal spaces can be used
4155 in all cases. */
4156
4157 #if defined COMPILE_PCRE8
4158 #ifdef SUPPORT_UTF
4159 if ((options & PCRE_UTF8) == 0)
4160 #endif
4161 if (end > 0xff) end = 0xff;
4162
4163 #elif defined COMPILE_PCRE16
4164 #ifdef SUPPORT_UTF
4165 if ((options & PCRE_UTF16) == 0)
4166 #endif
4167 if (end > 0xffff) end = 0xffff;
4168
4169 #endif /* COMPILE_PCRE[8|16] */
4170
4171 /* If all characters are less than 256, use the bit map. Otherwise use extra
4172 data. */
4173
4174 if (end < 0x100)
4175 {
4176 for (c = start; c <= end; c++)
4177 {
4178 n8++;
4179 SETBIT(classbits, c);
4180 }
4181 }
4182
4183 else
4184 {
4185 pcre_uchar *uchardata = *uchardptr;
4186
4187 #ifdef SUPPORT_UTF
4188 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4189 {
4190 if (start < end)
4191 {
4192 *uchardata++ = XCL_RANGE;
4193 uchardata += PRIV(ord2utf)(start, uchardata);
4194 uchardata += PRIV(ord2utf)(end, uchardata);
4195 }
4196 else if (start == end)
4197 {
4198 *uchardata++ = XCL_SINGLE;
4199 uchardata += PRIV(ord2utf)(start, uchardata);
4200 }
4201 }
4202 else
4203 #endif /* SUPPORT_UTF */
4204
4205 /* Without UTF support, character values are constrained by the bit length,
4206 and can only be > 256 for 16-bit and 32-bit libraries. */
4207
4208 #ifdef COMPILE_PCRE8
4209 {}
4210 #else
4211 if (start < end)
4212 {
4213 *uchardata++ = XCL_RANGE;
4214 *uchardata++ = start;
4215 *uchardata++ = end;
4216 }
4217 else if (start == end)
4218 {
4219 *uchardata++ = XCL_SINGLE;
4220 *uchardata++ = start;
4221 }
4222 #endif
4223
4224 *uchardptr = uchardata; /* Updata extra data pointer */
4225 }
4226
4227 return n8; /* Number of 8-bit characters */
4228 }
4229
4230
4231
4232
4233 /*************************************************
4234 * Add a list of characters to a class *
4235 *************************************************/
4236
4237 /* This function is used for adding a list of case-equivalent characters to a
4238 class, and also for adding a list of horizontal or vertical whitespace. If the
4239 list is in order (which it should be), ranges of characters are detected and
4240 handled appropriately. This function is mutually recursive with the function
4241 above.
4242
4243 Arguments:
4244 classbits the bit map for characters < 256
4245 uchardptr points to the pointer for extra data
4246 options the options word
4247 cd contains pointers to tables etc.
4248 p points to row of 32-bit values, terminated by NOTACHAR
4249 except character to omit; this is used when adding lists of
4250 case-equivalent characters to avoid including the one we
4251 already know about
4252
4253 Returns: the number of < 256 characters added
4254 the pointer to extra data is updated
4255 */
4256
4257 static int
4258 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4259 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4260 {
4261 int n8 = 0;
4262 while (p[0] < NOTACHAR)
4263 {
4264 int n = 0;
4265 if (p[0] != except)
4266 {
4267 while(p[n+1] == p[0] + n + 1) n++;
4268 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4269 }
4270 p += n + 1;
4271 }
4272 return n8;
4273 }
4274
4275
4276
4277 /*************************************************
4278 * Add characters not in a list to a class *
4279 *************************************************/
4280
4281 /* This function is used for adding the complement of a list of horizontal or
4282 vertical whitespace to a class. The list must be in order.
4283
4284 Arguments:
4285 classbits the bit map for characters < 256
4286 uchardptr points to the pointer for extra data
4287 options the options word
4288 cd contains pointers to tables etc.
4289 p points to row of 32-bit values, terminated by NOTACHAR
4290
4291 Returns: the number of < 256 characters added
4292 the pointer to extra data is updated
4293 */
4294
4295 static int
4296 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4297 int options, compile_data *cd, const pcre_uint32 *p)
4298 {
4299 BOOL utf = (options & PCRE_UTF8) != 0;
4300 int n8 = 0;
4301 if (p[0] > 0)
4302 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4303 while (p[0] < NOTACHAR)
4304 {
4305 while (p[1] == p[0] + 1) p++;
4306 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4307 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4308 p++;
4309 }
4310 return n8;
4311 }
4312
4313
4314
4315 /*************************************************
4316 * Compile one branch *
4317 *************************************************/
4318
4319 /* Scan the pattern, compiling it into the a vector. If the options are
4320 changed during the branch, the pointer is used to change the external options
4321 bits. This function is used during the pre-compile phase when we are trying
4322 to find out the amount of memory needed, as well as during the real compile
4323 phase. The value of lengthptr distinguishes the two phases.
4324
4325 Arguments:
4326 optionsptr pointer to the option bits
4327 codeptr points to the pointer to the current code point
4328 ptrptr points to the current pattern pointer
4329 errorcodeptr points to error code variable
4330 firstcharptr place to put the first required character
4331 firstcharflagsptr place to put the first character flags, or a negative number
4332 reqcharptr place to put the last required character
4333 reqcharflagsptr place to put the last required character flags, or a negative number
4334 bcptr points to current branch chain
4335 cond_depth conditional nesting depth
4336 cd contains pointers to tables etc.
4337 lengthptr NULL during the real compile phase
4338 points to length accumulator during pre-compile phase
4339
4340 Returns: TRUE on success
4341 FALSE, with *errorcodeptr set non-zero on error
4342 */
4343
4344 static BOOL
4345 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4346 const pcre_uchar **ptrptr, int *errorcodeptr,
4347 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4348 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4349 branch_chain *bcptr, int cond_depth,
4350 compile_data *cd, int *lengthptr)
4351 {
4352 int repeat_type, op_type;
4353 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4354 int bravalue = 0;
4355 int greedy_default, greedy_non_default;
4356 pcre_uint32 firstchar, reqchar;
4357 pcre_int32 firstcharflags, reqcharflags;
4358 pcre_uint32 zeroreqchar, zerofirstchar;
4359 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4360 pcre_int32 req_caseopt, reqvary, tempreqvary;
4361 int options = *optionsptr; /* May change dynamically */
4362 int after_manual_callout = 0;
4363 int length_prevgroup = 0;
4364 register pcre_uint32 c;
4365 int escape;
4366 register pcre_uchar *code = *codeptr;
4367 pcre_uchar *last_code = code;
4368 pcre_uchar *orig_code = code;
4369 pcre_uchar *tempcode;
4370 BOOL inescq = FALSE;
4371 BOOL groupsetfirstchar = FALSE;
4372 const pcre_uchar *ptr = *ptrptr;
4373 const pcre_uchar *tempptr;
4374 const pcre_uchar *nestptr = NULL;
4375 pcre_uchar *previous = NULL;
4376 pcre_uchar *previous_callout = NULL;
4377 pcre_uchar *save_hwm = NULL;
4378 pcre_uint8 classbits[32];
4379
4380 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4381 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4382 dynamically as we process the pattern. */
4383
4384 #ifdef SUPPORT_UTF
4385 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4386 BOOL utf = (options & PCRE_UTF8) != 0;
4387 #ifndef COMPILE_PCRE32
4388 pcre_uchar utf_chars[6];
4389 #endif
4390 #else
4391 BOOL utf = FALSE;
4392 #endif
4393
4394 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4395 class_uchardata always so that it can be passed to add_to_class() always,
4396 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4397 alternative calls for the different cases. */
4398
4399 pcre_uchar *class_uchardata;
4400 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4401 BOOL xclass;
4402 pcre_uchar *class_uchardata_base;
4403 #endif
4404
4405 #ifdef PCRE_DEBUG
4406 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4407 #endif
4408
4409 /* Set up the default and non-default settings for greediness */
4410
4411 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4412 greedy_non_default = greedy_default ^ 1;
4413
4414 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4415 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4416 matches a non-fixed char first char; reqchar just remains unset if we never
4417 find one.
4418
4419 When we hit a repeat whose minimum is zero, we may have to adjust these values
4420 to take the zero repeat into account. This is implemented by setting them to
4421 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4422 item types that can be repeated set these backoff variables appropriately. */
4423
4424 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4425 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4426
4427 /* The variable req_caseopt contains either the REQ_CASELESS value
4428 or zero, according to the current setting of the caseless flag. The
4429 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4430 firstchar or reqchar variables to record the case status of the
4431 value. This is used only for ASCII characters. */
4432
4433 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4434
4435 /* Switch on next character until the end of the branch */
4436
4437 for (;; ptr++)
4438 {
4439 BOOL negate_class;
4440 BOOL should_flip_negation;
4441 BOOL possessive_quantifier;
4442 BOOL is_quantifier;
4443 BOOL is_recurse;
4444 BOOL reset_bracount;
4445 int class_has_8bitchar;
4446 int class_one_char;
4447 int newoptions;
4448 int recno;
4449 int refsign;
4450 int skipbytes;
4451 pcre_uint32 subreqchar, subfirstchar;
4452 pcre_int32 subreqcharflags, subfirstcharflags;
4453 int terminator;
4454 unsigned int mclength;
4455 unsigned int tempbracount;
4456 pcre_uint32 ec;
4457 pcre_uchar mcbuffer[8];
4458
4459 /* Get next character in the pattern */
4460
4461 c = *ptr;
4462
4463 /* If we are at the end of a nested substitution, revert to the outer level
4464 string. Nesting only happens one level deep. */
4465
4466 if (c == CHAR_NULL && nestptr != NULL)
4467 {
4468 ptr = nestptr;
4469 nestptr = NULL;
4470 c = *ptr;
4471 }
4472
4473 /* If we are in the pre-compile phase, accumulate the length used for the
4474 previous cycle of this loop. */
4475
4476 if (lengthptr != NULL)
4477 {
4478 #ifdef PCRE_DEBUG
4479 if (code > cd->hwm) cd->hwm = code; /* High water info */
4480 #endif
4481 if (code > cd->start_workspace + cd->workspace_size -
4482 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4483 {
4484 *errorcodeptr = ERR52;
4485 goto FAILED;
4486 }
4487
4488 /* There is at least one situation where code goes backwards: this is the
4489 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4490 the class is simply eliminated. However, it is created first, so we have to
4491 allow memory for it. Therefore, don't ever reduce the length at this point.
4492 */
4493
4494 if (code < last_code) code = last_code;
4495
4496 /* Paranoid check for integer overflow */
4497
4498 if (OFLOW_MAX - *lengthptr < code - last_code)
4499 {
4500 *errorcodeptr = ERR20;
4501 goto FAILED;
4502 }
4503
4504 *lengthptr += (int)(code - last_code);
4505 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4506 (int)(code - last_code), c, c));
4507
4508 /* If "previous" is set and it is not at the start of the work space, move
4509 it back to there, in order to avoid filling up the work space. Otherwise,
4510 if "previous" is NULL, reset the current code pointer to the start. */
4511
4512 if (previous != NULL)
4513 {
4514 if (previous > orig_code)
4515 {
4516 memmove(orig_code, previous, IN_UCHARS(code - previous));
4517 code -= previous - orig_code;
4518 previous = orig_code;
4519 }
4520 }
4521 else code = orig_code;
4522
4523 /* Remember where this code item starts so we can pick up the length
4524 next time round. */
4525
4526 last_code = code;
4527 }
4528
4529 /* In the real compile phase, just check the workspace used by the forward
4530 reference list. */
4531
4532 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4533 WORK_SIZE_SAFETY_MARGIN)
4534 {
4535 *errorcodeptr = ERR52;
4536 goto FAILED;
4537 }
4538
4539 /* If in \Q...\E, check for the end; if not, we have a literal */
4540
4541 if (inescq && c != CHAR_NULL)
4542 {
4543 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4544 {
4545 inescq = FALSE;
4546 ptr++;
4547 continue;
4548 }
4549 else
4550 {
4551 if (previous_callout != NULL)
4552 {
4553 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4554 complete_callout(previous_callout, ptr, cd);
4555 previous_callout = NULL;
4556 }
4557 if ((options & PCRE_AUTO_CALLOUT) != 0)
4558 {
4559 previous_callout = code;
4560 code = auto_callout(code, ptr, cd);
4561 }
4562 goto NORMAL_CHAR;
4563 }
4564 /* Control does not reach here. */
4565 }
4566
4567 /* In extended mode, skip white space and comments. We need a loop in order
4568 to check for more white space and more comments after a comment. */
4569
4570 if ((options & PCRE_EXTENDED) != 0)
4571 {
4572 for (;;)
4573 {
4574 while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4575 if (c != CHAR_NUMBER_SIGN) break;
4576 ptr++;
4577 while (*ptr != CHAR_NULL)
4578 {
4579 if (IS_NEWLINE(ptr)) /* For non-fixed-length newline cases, */
4580 { /* IS_NEWLINE sets cd->nllen. */
4581 ptr += cd->nllen;
4582 break;
4583 }
4584 ptr++;
4585 #ifdef SUPPORT_UTF
4586 if (utf) FORWARDCHAR(ptr);
4587 #endif
4588 }
4589 c = *ptr; /* Either NULL or the char after a newline */
4590 }
4591 }
4592
4593 /* See if the next thing is a quantifier. */
4594
4595 is_quantifier =
4596 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4597 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4598
4599 /* Fill in length of a previous callout, except when the next thing is a
4600 quantifier or when processing a property substitution string in UCP mode. */
4601
4602 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4603 after_manual_callout-- <= 0)
4604 {
4605 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4606 complete_callout(previous_callout, ptr, cd);
4607 previous_callout = NULL;
4608 }
4609
4610 /* Create auto callout, except for quantifiers, or while processing property
4611 strings that are substituted for \w etc in UCP mode. */
4612
4613 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4614 {
4615 previous_callout = code;
4616 code = auto_callout(code, ptr, cd);
4617 }
4618
4619 /* Process the next pattern item. */
4620
4621 switch(c)
4622 {
4623 /* ===================================================================*/
4624 case CHAR_NULL: /* The branch terminates at string end */
4625 case CHAR_VERTICAL_LINE: /* or | or ) */
4626 case CHAR_RIGHT_PARENTHESIS:
4627 *firstcharptr = firstchar;
4628 *firstcharflagsptr = firstcharflags;
4629 *reqcharptr = reqchar;
4630 *reqcharflagsptr = reqcharflags;
4631 *codeptr = code;
4632 *ptrptr = ptr;
4633 if (lengthptr != NULL)
4634 {
4635 if (OFLOW_MAX - *lengthptr < code - last_code)
4636 {
4637 *errorcodeptr = ERR20;
4638 goto FAILED;
4639 }
4640 *lengthptr += (int)(code - last_code); /* To include callout length */
4641 DPRINTF((">> end branch\n"));
4642 }
4643 return TRUE;
4644
4645
4646 /* ===================================================================*/
4647 /* Handle single-character metacharacters. In multiline mode, ^ disables
4648 the setting of any following char as a first character. */
4649
4650 case CHAR_CIRCUMFLEX_ACCENT:
4651 previous = NULL;
4652 if ((options & PCRE_MULTILINE) != 0)
4653 {
4654 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4655 *code++ = OP_CIRCM;
4656 }
4657 else *code++ = OP_CIRC;
4658 break;
4659
4660 case CHAR_DOLLAR_SIGN:
4661 previous = NULL;
4662 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4663 break;
4664
4665 /* There can never be a first char if '.' is first, whatever happens about
4666 repeats. The value of reqchar doesn't change either. */
4667
4668 case CHAR_DOT:
4669 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4670 zerofirstchar = firstchar;
4671 zerofirstcharflags = firstcharflags;
4672 zeroreqchar = reqchar;
4673 zeroreqcharflags = reqcharflags;
4674 previous = code;
4675 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4676 break;
4677
4678
4679 /* ===================================================================*/
4680 /* Character classes. If the included characters are all < 256, we build a
4681 32-byte bitmap of the permitted characters, except in the special case
4682 where there is only one such character. For negated classes, we build the
4683 map as usual, then invert it at the end. However, we use a different opcode
4684 so that data characters > 255 can be handled correctly.
4685
4686 If the class contains characters outside the 0-255 range, a different
4687 opcode is compiled. It may optionally have a bit map for characters < 256,
4688 but those above are are explicitly listed afterwards. A flag byte tells
4689 whether the bitmap is present, and whether this is a negated class or not.
4690
4691 In JavaScript compatibility mode, an isolated ']' causes an error. In
4692 default (Perl) mode, it is treated as a data character. */
4693
4694 case CHAR_RIGHT_SQUARE_BRACKET:
4695 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4696 {
4697 *errorcodeptr = ERR64;
4698 goto FAILED;
4699 }
4700 goto NORMAL_CHAR;
4701
4702 /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4703 used for "start of word" and "end of word". As these are otherwise illegal
4704 sequences, we don't break anything by recognizing them. They are replaced
4705 by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4706 erroneous and are handled by the normal code below. */
4707
4708 case CHAR_LEFT_SQUARE_BRACKET:
4709 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4710 {
4711 nestptr = ptr + 7;
4712 ptr = sub_start_of_word - 1;
4713 continue;
4714 }
4715
4716 if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4717 {
4718 nestptr = ptr + 7;
4719 ptr = sub_end_of_word - 1;
4720 continue;
4721 }
4722
4723 /* Handle a real character class. */
4724
4725 previous = code;
4726
4727 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4728 they are encountered at the top level, so we'll do that too. */
4729
4730 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4731 ptr[1] == CHAR_EQUALS_SIGN) &&
4732 check_posix_syntax(ptr, &tempptr))
4733 {
4734 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4735 goto FAILED;
4736 }
4737
4738 /* If the first character is '^', set the negation flag and skip it. Also,
4739 if the first few characters (either before or after ^) are \Q\E or \E we
4740 skip them too. This makes for compatibility with Perl. */
4741
4742 negate_class = FALSE;
4743 for (;;)
4744 {
4745 c = *(++ptr);
4746 if (c == CHAR_BACKSLASH)
4747 {
4748 if (ptr[1] == CHAR_E)
4749 ptr++;
4750 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4751 ptr += 3;
4752 else
4753 break;
4754 }
4755 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4756 negate_class = TRUE;
4757 else break;
4758 }
4759
4760 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4761 an initial ']' is taken as a data character -- the code below handles
4762 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4763 [^] must match any character, so generate OP_ALLANY. */
4764
4765 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4766 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4767 {
4768 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4769 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4770 zerofirstchar = firstchar;
4771 zerofirstcharflags = firstcharflags;
4772 break;
4773 }
4774
4775 /* If a class contains a negative special such as \S, we need to flip the
4776 negation flag at the end, so that support for characters > 255 works
4777 correctly (they are all included in the class). */
4778
4779 should_flip_negation = FALSE;
4780
4781 /* For optimization purposes, we track some properties of the class:
4782 class_has_8bitchar will be non-zero if the class contains at least one <
4783 256 character; class_one_char will be 1 if the class contains just one
4784 character. */
4785
4786 class_has_8bitchar = 0;
4787 class_one_char = 0;
4788
4789 /* Initialize the 32-char bit map to all zeros. We build the map in a
4790 temporary bit of memory, in case the class contains fewer than two
4791 8-bit characters because in that case the compiled code doesn't use the bit
4792 map. */
4793
4794 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4795
4796 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4797 xclass = FALSE;
4798 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4799 class_uchardata_base = class_uchardata; /* Save the start */
4800 #endif
4801
4802 /* Process characters until ] is reached. By writing this as a "do" it
4803 means that an initial ] is taken as a data character. At the start of the
4804 loop, c contains the first byte of the character. */
4805
4806 if (c != CHAR_NULL) do
4807 {
4808 const pcre_uchar *oldptr;
4809
4810 #ifdef SUPPORT_UTF
4811 if (utf && HAS_EXTRALEN(c))
4812 { /* Braces are required because the */
4813 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4814 }
4815 #endif
4816
4817 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4818 /* In the pre-compile phase, accumulate the length of any extra
4819 data and reset the pointer. This is so that very large classes that
4820 contain a zillion > 255 characters no longer overwrite the work space
4821 (which is on the stack). We have to remember that there was XCLASS data,
4822 however. */
4823
4824 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4825 {
4826 xclass = TRUE;
4827 *lengthptr += class_uchardata - class_uchardata_base;
4828 class_uchardata = class_uchardata_base;
4829 }
4830 #endif
4831
4832 /* Inside \Q...\E everything is literal except \E */
4833
4834 if (inescq)
4835 {
4836 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4837 {
4838 inescq = FALSE; /* Reset literal state */
4839 ptr++; /* Skip the 'E' */
4840 continue; /* Carry on with next */
4841 }
4842 goto CHECK_RANGE; /* Could be range if \E follows */
4843 }
4844
4845 /* Handle POSIX class names. Perl allows a negation extension of the
4846 form [:^name:]. A square bracket that doesn't match the syntax is
4847 treated as a literal. We also recognize the POSIX constructions
4848 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4849 5.6 and 5.8 do. */
4850
4851 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4852 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4853 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4854 {
4855 BOOL local_negate = FALSE;
4856 int posix_class, taboffset, tabopt;
4857 register const pcre_uint8 *cbits = cd->cbits;
4858 pcre_uint8 pbits[32];
4859
4860 if (ptr[1] != CHAR_COLON)
4861 {
4862 *errorcodeptr = ERR31;
4863 goto FAILED;
4864 }
4865
4866 ptr += 2;
4867 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4868 {
4869 local_negate = TRUE;
4870 should_flip_negation = TRUE; /* Note negative special */
4871 ptr++;
4872 }
4873
4874 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4875 if (posix_class < 0)
4876 {
4877 *errorcodeptr = ERR30;
4878 goto FAILED;
4879 }
4880
4881 /* If matching is caseless, upper and lower are converted to
4882 alpha. This relies on the fact that the class table starts with
4883 alpha, lower, upper as the first 3 entries. */
4884
4885 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4886 posix_class = 0;
4887
4888 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4889 different escape sequences that use Unicode properties \p or \P. Others
4890 that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4891 directly. */
4892
4893 #ifdef SUPPORT_UCP
4894 if ((options & PCRE_UCP) != 0)
4895 {
4896 unsigned int ptype = 0;
4897 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4898
4899 /* The posix_substitutes table specifies which POSIX classes can be
4900 converted to \p or \P items. */
4901
4902 if (posix_substitutes[pc] != NULL)
4903 {
4904 nestptr = tempptr + 1;
4905 ptr = posix_substitutes[pc] - 1;
4906 continue;
4907 }
4908
4909 /* There are three other classes that generate special property calls
4910 that are recognized only in an XCLASS. */
4911
4912 else switch(posix_class)
4913 {
4914 case PC_GRAPH:
4915 ptype = PT_PXGRAPH;
4916 /* Fall through */
4917 case PC_PRINT:
4918 if (ptype == 0) ptype = PT_PXPRINT;
4919 /* Fall through */
4920 case PC_PUNCT:
4921 if (ptype == 0) ptype = PT_PXPUNCT;
4922 *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4923 *class_uchardata++ = ptype;
4924 *class_uchardata++ = 0;
4925 ptr = tempptr + 1;
4926 continue;
4927
4928 /* For all other POSIX classes, no special action is taken in UCP
4929 mode. Fall through to the non_UCP case. */
4930
4931 default:
4932 break;
4933 }
4934 }
4935 #endif
4936 /* In the non-UCP case, or when UCP makes no difference, we build the
4937 bit map for the POSIX class in a chunk of local store because we may be
4938 adding and subtracting from it, and we don't want to subtract bits that
4939 may be in the main map already. At the end we or the result into the
4940 bit map that is being built. */
4941
4942 posix_class *= 3;
4943
4944 /* Copy in the first table (always present) */
4945
4946 memcpy(pbits, cbits + posix_class_maps[posix_class],
4947 32 * sizeof(pcre_uint8));
4948
4949 /* If there is a second table, add or remove it as required. */
4950
4951 taboffset = posix_class_maps[posix_class + 1];
4952 tabopt = posix_class_maps[posix_class + 2];
4953
4954 if (taboffset >= 0)
4955 {
4956 if (tabopt >= 0)
4957 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
4958 else
4959 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
4960 }
4961
4962 /* Now see if we need to remove any special characters. An option
4963 value of 1 removes vertical space and 2 removes underscore. */
4964
4965 if (tabopt < 0) tabopt = -tabopt;
4966 if (tabopt == 1) pbits[1] &= ~0x3c;
4967 else if (tabopt == 2) pbits[11] &= 0x7f;
4968
4969 /* Add the POSIX table or its complement into the main table that is
4970 being built and we are done. */
4971
4972 if (local_negate)
4973 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
4974 else
4975 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
4976
4977 ptr = tempptr + 1;
4978 /* Every class contains at least one < 256 character. */
4979 class_has_8bitchar = 1;
4980 /* Every class contains at least two characters. */
4981 class_one_char = 2;
4982 continue; /* End of POSIX syntax handling */
4983 }
4984
4985 /* Backslash may introduce a single character, or it may introduce one
4986 of the specials, which just set a flag. The sequence \b is a special
4987 case. Inside a class (and only there) it is treated as backspace. We
4988 assume that other escapes have more than one character in them, so
4989 speculatively set both class_has_8bitchar and class_one_char bigger
4990 than one. Unrecognized escapes fall through and are either treated
4991 as literal characters (by default), or are faulted if
4992 PCRE_EXTRA is set. */
4993
4994 if (c == CHAR_BACKSLASH)
4995 {
4996 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4997 TRUE);
4998 if (*errorcodeptr != 0) goto FAILED;
4999 if (escape == 0) c = ec;
5000 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
5001 else if (escape == ESC_N) /* \N is not supported in a class */
5002 {
5003 *errorcodeptr = ERR71;
5004 goto FAILED;
5005 }
5006 else if (escape == ESC_Q) /* Handle start of quoted string */
5007 {
5008 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5009 {
5010 ptr += 2; /* avoid empty string */
5011 }
5012 else inescq = TRUE;
5013 continue;
5014 }
5015 else if (escape == ESC_E) continue; /* Ignore orphan \E */
5016
5017 else
5018 {
5019 register const pcre_uint8 *cbits = cd->cbits;
5020 /* Every class contains at least two < 256 characters. */
5021 class_has_8bitchar++;
5022 /* Every class contains at least two characters. */
5023 class_one_char += 2;
5024
5025 switch (escape)
5026 {
5027 #ifdef SUPPORT_UCP
5028 case ESC_du: /* These are the values given for \d etc */
5029 case ESC_DU: /* when PCRE_UCP is set. We replace the */
5030 case ESC_wu: /* escape sequence with an appropriate \p */
5031 case ESC_WU: /* or \P to test Unicode properties instead */
5032 case ESC_su: /* of the default ASCII testing. */
5033 case ESC_SU:
5034 nestptr = ptr;
5035 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
5036 class_has_8bitchar--; /* Undo! */
5037 continue;
5038 #endif
5039 case ESC_d:
5040 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
5041 continue;
5042
5043 case ESC_D:
5044 should_flip_negation = TRUE;
5045 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
5046 continue;
5047
5048 case ESC_w:
5049 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
5050 continue;
5051
5052 case ESC_W:
5053 should_flip_negation = TRUE;
5054 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5055 continue;
5056
5057 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5058 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5059 previously set by something earlier in the character class.
5060 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5061 we could just adjust the appropriate bit. From PCRE 8.34 we no
5062 longer treat \s and \S specially. */
5063
5064 case ESC_s:
5065 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
5066 continue;
5067
5068 case ESC_S:
5069 should_flip_negation = TRUE;
5070 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
5071 continue;
5072
5073 /* The rest apply in both UCP and non-UCP cases. */
5074
5075 case ESC_h:
5076 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5077 PRIV(hspace_list), NOTACHAR);
5078 continue;
5079
5080 case ESC_H:
5081 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5082 cd, PRIV(hspace_list));
5083 continue;
5084
5085 case ESC_v:
5086 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5087 PRIV(vspace_list), NOTACHAR);
5088 continue;
5089
5090 case ESC_V:
5091 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5092 cd, PRIV(vspace_list));
5093 continue;
5094
5095 #ifdef SUPPORT_UCP
5096 case ESC_p:
5097 case ESC_P:
5098 {
5099 BOOL negated;
5100 unsigned int ptype = 0, pdata = 0;
5101 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5102 goto FAILED;
5103 *class_uchardata++ = ((escape == ESC_p) != negated)?
5104 XCL_PROP : XCL_NOTPROP;
5105 *class_uchardata++ = ptype;
5106 *class_uchardata++ = pdata;
5107 class_has_8bitchar--; /* Undo! */
5108 continue;
5109 }
5110 #endif
5111 /* Unrecognized escapes are faulted if PCRE is running in its
5112 strict mode. By default, for compatibility with Perl, they are
5113 treated as literals. */
5114
5115 default:
5116 if ((options & PCRE_EXTRA) != 0)
5117 {
5118 *errorcodeptr = ERR7;
5119 goto FAILED;
5120 }
5121 class_has_8bitchar--; /* Undo the speculative increase. */
5122 class_one_char -= 2; /* Undo the speculative increase. */
5123 c = *ptr; /* Get the final character and fall through */
5124 break;
5125 }
5126 }
5127
5128 /* Fall through if the escape just defined a single character (c >= 0).
5129 This may be greater than 256. */
5130
5131 escape = 0;
5132
5133 } /* End of backslash handling */
5134
5135 /* A character may be followed by '-' to form a range. However, Perl does
5136 not permit ']' to be the end of the range. A '-' character at the end is
5137 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5138 code for handling \Q and \E is messy. */
5139
5140 CHECK_RANGE:
5141 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5142 {
5143 inescq = FALSE;
5144 ptr += 2;
5145 }
5146 oldptr = ptr;
5147
5148 /* Remember if \r or \n were explicitly used */
5149
5150 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5151
5152 /* Check for range */
5153
5154 if (!inescq && ptr[1] == CHAR_MINUS)
5155 {
5156 pcre_uint32 d;
5157 ptr += 2;
5158 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5159
5160 /* If we hit \Q (not followed by \E) at this point, go into escaped
5161 mode. */
5162
5163 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5164 {
5165 ptr += 2;
5166 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5167 { ptr += 2; continue; }
5168 inescq = TRUE;
5169 break;
5170 }
5171
5172 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5173 back the pointer and jump to handle the character that preceded it. */
5174
5175 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5176 {
5177 ptr = oldptr;
5178 goto CLASS_SINGLE_CHARACTER;
5179 }
5180
5181 /* Otherwise, we have a potential range; pick up the next character */
5182
5183 #ifdef SUPPORT_UTF
5184 if (utf)
5185 { /* Braces are required because the */
5186 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5187 }
5188 else
5189 #endif
5190 d = *ptr; /* Not UTF-8 mode */
5191
5192 /* The second part of a range can be a single-character escape
5193 sequence, but not any of the other escapes. Perl treats a hyphen as a
5194 literal in such circumstances. However, in Perl's warning mode, a
5195 warning is given, so PCRE now faults it as it is almost certainly a
5196 mistake on the user's part. */
5197
5198 if (!inescq)
5199 {
5200 if (d == CHAR_BACKSLASH)
5201 {
5202 int descape;
5203 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5204 if (*errorcodeptr != 0) goto FAILED;
5205
5206 /* 0 means a character was put into d; \b is backspace; any other
5207 special causes an error. */
5208
5209 if (descape != 0)
5210 {
5211 if (descape == ESC_b) d = CHAR_BS; else
5212 {
5213 *errorcodeptr = ERR83;
5214 goto FAILED;
5215 }
5216 }
5217 }
5218
5219 /* A hyphen followed by a POSIX class is treated in the same way. */
5220
5221 else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5222 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5223 ptr[1] == CHAR_EQUALS_SIGN) &&
5224 check_posix_syntax(ptr, &tempptr))
5225 {
5226 *errorcodeptr = ERR83;
5227 goto FAILED;
5228 }
5229 }
5230
5231 /* Check that the two values are in the correct order. Optimize
5232 one-character ranges. */
5233
5234 if (d < c)
5235 {
5236 *errorcodeptr = ERR8;
5237 goto FAILED;
5238 }
5239 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5240
5241 /* We have found a character range, so single character optimizations
5242 cannot be done anymore. Any value greater than 1 indicates that there
5243 is more than one character. */
5244
5245 class_one_char = 2;
5246
5247 /* Remember an explicit \r or \n, and add the range to the class. */
5248
5249 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5250
5251 class_has_8bitchar +=
5252 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5253
5254 continue; /* Go get the next char in the class */
5255 }
5256
5257 /* Handle a single character - we can get here for a normal non-escape
5258 char, or after \ that introduces a single character or for an apparent
5259 range that isn't. Only the value 1 matters for class_one_char, so don't
5260 increase it if it is already 2 or more ... just in case there's a class
5261 with a zillion characters in it. */
5262
5263 CLASS_SINGLE_CHARACTER:
5264 if (class_one_char < 2) class_one_char++;
5265
5266 /* If class_one_char is 1, we have the first single character in the
5267 class, and there have been no prior ranges, or XCLASS items generated by
5268 escapes. If this is the final character in the class, we can optimize by
5269 turning the item into a 1-character OP_CHAR[I] if it's positive, or
5270 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5271 to be set. Otherwise, there can be no first char if this item is first,
5272 whatever repeat count may follow. In the case of reqchar, save the
5273 previous value for reinstating. */
5274
5275 if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5276 {
5277 ptr++;
5278 zeroreqchar = reqchar;
5279 zeroreqcharflags = reqcharflags;
5280
5281 if (negate_class)
5282 {
5283 #ifdef SUPPORT_UCP
5284 int d;
5285 #endif
5286 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5287 zerofirstchar = firstchar;
5288 zerofirstcharflags = firstcharflags;
5289
5290 /* For caseless UTF-8 mode when UCP support is available, check
5291 whether this character has more than one other case. If so, generate
5292 a special OP_NOTPROP item instead of OP_NOTI. */
5293
5294 #ifdef SUPPORT_UCP
5295 if (utf && (options & PCRE_CASELESS) != 0 &&
5296 (d = UCD_CASESET(c)) != 0)
5297 {
5298 *code++ = OP_NOTPROP;
5299 *code++ = PT_CLIST;
5300 *code++ = d;
5301 }
5302 else
5303 #endif
5304 /* Char has only one other case, or UCP not available */
5305
5306 {
5307 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5308 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5309 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5310 code += PRIV(ord2utf)(c, code);
5311 else
5312 #endif
5313 *code++ = c;
5314 }
5315
5316 /* We are finished with this character class */
5317
5318 goto END_CLASS;
5319 }
5320
5321 /* For a single, positive character, get the value into mcbuffer, and
5322 then we can handle this with the normal one-character code. */
5323
5324 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5325 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5326 mclength = PRIV(ord2utf)(c, mcbuffer);
5327 else
5328 #endif
5329 {
5330 mcbuffer[0] = c;
5331 mclength = 1;
5332 }
5333 goto ONE_CHAR;
5334 } /* End of 1-char optimization */
5335
5336 /* There is more than one character in the class, or an XCLASS item
5337 has been generated. Add this character to the class. */
5338
5339 class_has_8bitchar +=
5340 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5341 }
5342
5343 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5344 If we are at the end of an internal nested string, revert to the outer
5345 string. */
5346
5347 while (((c = *(++ptr)) != CHAR_NULL ||
5348 (nestptr != NULL &&
5349 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5350 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5351
5352 /* Check for missing terminating ']' */
5353
5354 if (c == CHAR_NULL)
5355 {
5356 *errorcodeptr = ERR6;
5357 goto FAILED;
5358 }
5359
5360 /* We will need an XCLASS if data has been placed in class_uchardata. In
5361 the second phase this is a sufficient test. However, in the pre-compile
5362 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5363 only if the very last character in the class needs XCLASS will it contain
5364 anything at this point. For this reason, xclass gets set TRUE above when
5365 uchar_classdata is emptied, and that's why this code is the way it is here
5366 instead of just doing a test on class_uchardata below. */
5367
5368 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5369 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5370 #endif
5371
5372 /* If this is the first thing in the branch, there can be no first char
5373 setting, whatever the repeat count. Any reqchar setting must remain
5374 unchanged after any kind of repeat. */
5375
5376 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5377 zerofirstchar = firstchar;
5378 zerofirstcharflags = firstcharflags;
5379 zeroreqchar = reqchar;
5380 zeroreqcharflags = reqcharflags;
5381
5382 /* If there are characters with values > 255, we have to compile an
5383 extended class, with its own opcode, unless there was a negated special
5384 such as \S in the class, and PCRE_UCP is not set, because in that case all
5385 characters > 255 are in the class, so any that were explicitly given as
5386 well can be ignored. If (when there are explicit characters > 255 that must
5387 be listed) there are no characters < 256, we can omit the bitmap in the
5388 actual compiled code. */
5389
5390 #ifdef SUPPORT_UTF
5391 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5392 #elif !defined COMPILE_PCRE8
5393 if (xclass && !should_flip_negation)
5394 #endif
5395 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5396 {
5397 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5398 *code++ = OP_XCLASS;
5399 code += LINK_SIZE;
5400 *code = negate_class? XCL_NOT:0;
5401
5402 /* If the map is required, move up the extra data to make room for it;
5403 otherwise just move the code pointer to the end of the extra data. */
5404
5405 if (class_has_8bitchar > 0)
5406 {
5407 *code++ |= XCL_MAP;
5408 memmove(code + (32 / sizeof(pcre_uchar)), code,
5409 IN_UCHARS(class_uchardata - code));
5410 memcpy(code, classbits, 32);
5411 code = class_uchardata + (32 / sizeof(pcre_uchar));
5412 }
5413 else code = class_uchardata;
5414
5415 /* Now fill in the complete length of the item */
5416
5417 PUT(previous, 1, (int)(code - previous));
5418 break; /* End of class handling */
5419 }
5420 #endif
5421
5422 /* If there are no characters > 255, or they are all to be included or
5423 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5424 whole class was negated and whether there were negative specials such as \S
5425 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5426 negating it if necessary. */
5427
5428 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5429 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5430 {
5431 if (negate_class)
5432 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5433 memcpy(code, classbits, 32);
5434 }
5435 code += 32 / sizeof(pcre_uchar);
5436
5437 END_CLASS:
5438 break;
5439
5440
5441 /* ===================================================================*/
5442 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5443 has been tested above. */
5444
5445 case CHAR_LEFT_CURLY_BRACKET:
5446 if (!is_quantifier) goto NORMAL_CHAR;
5447 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5448 if (*errorcodeptr != 0) goto FAILED;
5449 goto REPEAT;
5450
5451 case CHAR_ASTERISK:
5452 repeat_min = 0;
5453 repeat_max = -1;
5454 goto REPEAT;
5455
5456 case CHAR_PLUS:
5457 repeat_min = 1;
5458 repeat_max = -1;
5459 goto REPEAT;
5460
5461 case CHAR_QUESTION_MARK:
5462 repeat_min = 0;
5463 repeat_max = 1;
5464
5465 REPEAT:
5466 if (previous == NULL)
5467 {
5468 *errorcodeptr = ERR9;
5469 goto FAILED;
5470 }
5471
5472 if (repeat_min == 0)
5473 {
5474 firstchar = zerofirstchar; /* Adjust for zero repeat */
5475 firstcharflags = zerofirstcharflags;
5476 reqchar = zeroreqchar; /* Ditto */
5477 reqcharflags = zeroreqcharflags;
5478 }
5479
5480 /* Remember whether this is a variable length repeat */
5481
5482 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5483
5484 op_type = 0; /* Default single-char op codes */
5485 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5486
5487 /* Save start of previous item, in case we have to move it up in order to
5488 insert something before it. */
5489
5490 tempcode = previous;
5491
5492 /* Before checking for a possessive quantifier, we must skip over
5493 whitespace and comments in extended mode because Perl allows white space at
5494 this point. */
5495
5496 if ((options & PCRE_EXTENDED) != 0)
5497 {
5498 const pcre_uchar *p = ptr + 1;
5499 for (;;)
5500 {
5501 while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5502 if (*p != CHAR_NUMBER_SIGN) break;
5503 p++;
5504 while (*p != CHAR_NULL)
5505 {
5506 if (IS_NEWLINE(p)) /* For non-fixed-length newline cases, */
5507 { /* IS_NEWLINE sets cd->nllen. */
5508 p += cd->nllen;
5509 break;
5510 }
5511 p++;
5512 #ifdef SUPPORT_UTF
5513 if (utf) FORWARDCHAR(p);
5514 #endif
5515 } /* Loop for comment characters */
5516 } /* Loop for multiple comments */
5517 ptr = p - 1; /* Character before the next significant one. */
5518 }
5519
5520 /* If the next character is '+', we have a possessive quantifier. This
5521 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5522 If the next character is '?' this is a minimizing repeat, by default,
5523 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5524 repeat type to the non-default. */
5525
5526 if (ptr[1] == CHAR_PLUS)
5527 {
5528 repeat_type = 0; /* Force greedy */
5529 possessive_quantifier = TRUE;
5530 ptr++;
5531 }
5532 else if (ptr[1] == CHAR_QUESTION_MARK)
5533 {
5534 repeat_type = greedy_non_default;
5535 ptr++;
5536 }
5537 else repeat_type = greedy_default;
5538
5539 /* If previous was a recursion call, wrap it in atomic brackets so that
5540 previous becomes the atomic group. All recursions were so wrapped in the
5541 past, but it no longer happens for non-repeated recursions. In fact, the
5542 repeated ones could be re-implemented independently so as not to need this,
5543 but for the moment we rely on the code for repeating groups. */
5544
5545 if (*previous == OP_RECURSE)
5546 {
5547 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5548 *previous = OP_ONCE;
5549 PUT(previous, 1, 2 + 2*LINK_SIZE);
5550 previous[2 + 2*LINK_SIZE] = OP_KET;
5551 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5552 code += 2 + 2 * LINK_SIZE;
5553 length_prevgroup = 3 + 3*LINK_SIZE;
5554
5555 /* When actually compiling, we need to check whether this was a forward
5556 reference, and if so, adjust the offset. */
5557
5558 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5559 {
5560 int offset = GET(cd->hwm, -LINK_SIZE);
5561 if (offset == previous + 1 - cd->start_code)
5562 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5563 }
5564 }
5565
5566 /* Now handle repetition for the different types of item. */
5567
5568 /* If previous was a character or negated character match, abolish the item
5569 and generate a repeat item instead. If a char item has a minimum of more
5570 than one, ensure that it is set in reqchar - it might not be if a sequence
5571 such as x{3} is the first thing in a branch because the x will have gone
5572 into firstchar instead. */
5573
5574 if (*previous == OP_CHAR || *previous == OP_CHARI
5575 || *previous == OP_NOT || *previous == OP_NOTI)
5576 {
5577 switch (*previous)
5578 {
5579 default: /* Make compiler happy. */
5580 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5581 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5582 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5583 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5584 }
5585
5586 /* Deal with UTF characters that take up more than one character. It's
5587 easier to write this out separately than try to macrify it. Use c to
5588 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5589 it's a length rather than a small character. */
5590
5591 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5592 if (utf && NOT_FIRSTCHAR(code[-1]))
5593 {
5594 pcre_uchar *lastchar = code - 1;
5595 BACKCHAR(lastchar);
5596 c = (int)(code - lastchar); /* Length of UTF-8 character */
5597 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5598 c |= UTF_LENGTH; /* Flag c as a length */
5599 }
5600 else
5601 #endif /* SUPPORT_UTF */
5602
5603 /* Handle the case of a single charater - either with no UTF support, or
5604 with UTF disabled, or for a single character UTF character. */
5605 {
5606 c = code[-1];
5607 if (*previous <= OP_CHARI && repeat_min > 1)
5608 {
5609 reqchar = c;
5610 reqcharflags = req_caseopt | cd->req_varyopt;
5611 }
5612 }
5613
5614 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5615 }
5616
5617 /* If previous was a character type match (\d or similar), abolish it and
5618 create a suitable repeat item. The code is shared with single-character
5619 repeats by setting op_type to add a suitable offset into repeat_type. Note
5620 the the Unicode property types will be present only when SUPPORT_UCP is
5621 defined, but we don't wrap the little bits of code here because it just
5622 makes it horribly messy. */
5623
5624 else if (*previous < OP_EODN)
5625 {
5626 pcre_uchar *oldcode;
5627 int prop_type, prop_value;
5628 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5629 c = *previous;
5630
5631 OUTPUT_SINGLE_REPEAT:
5632 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5633 {
5634 prop_type = previous[1];
5635 prop_value = previous[2];
5636 }
5637 else prop_type = prop_value = -1;
5638
5639 oldcode = code;
5640 code = previous; /* Usually overwrite previous item */
5641
5642 /* If the maximum is zero then the minimum must also be zero; Perl allows
5643 this case, so we do too - by simply omitting the item altogether. */
5644
5645 if (repeat_max == 0) goto END_REPEAT;
5646
5647 /* Combine the op_type with the repeat_type */
5648
5649 repeat_type += op_type;
5650
5651 /* A minimum of zero is handled either as the special case * or ?, or as
5652 an UPTO, with the maximum given. */
5653
5654 if (repeat_min == 0)
5655 {
5656 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5657 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5658 else
5659 {
5660 *code++ = OP_UPTO + repeat_type;
5661 PUT2INC(code, 0, repeat_max);
5662 }
5663 }
5664
5665 /* A repeat minimum of 1 is optimized into some special cases. If the
5666 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5667 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5668 one less than the maximum. */
5669
5670 else if (repeat_min == 1)
5671 {
5672 if (repeat_max == -1)
5673 *code++ = OP_PLUS + repeat_type;
5674 else
5675 {
5676 code = oldcode; /* leave previous item in place */
5677 if (repeat_max == 1) goto END_REPEAT;
5678 *code++ = OP_UPTO + repeat_type;
5679 PUT2INC(code, 0, repeat_max - 1);
5680 }
5681 }
5682
5683 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5684 handled as an EXACT followed by an UPTO. */
5685
5686 else
5687 {
5688 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5689 PUT2INC(code, 0, repeat_min);
5690
5691 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5692 we have to insert the character for the previous code. For a repeated
5693 Unicode property match, there are two extra bytes that define the
5694 required property. In UTF-8 mode, long characters have their length in
5695 c, with the UTF_LENGTH bit as a flag. */
5696
5697 if (repeat_max < 0)
5698 {
5699 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5700 if (utf && (c & UTF_LENGTH) != 0)
5701 {
5702 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5703 code += c & 7;
5704 }
5705 else
5706 #endif
5707 {
5708 *code++ = c;
5709 if (prop_type >= 0)
5710 {
5711 *code++ = prop_type;
5712 *code++ = prop_value;
5713 }
5714 }
5715 *code++ = OP_STAR + repeat_type;
5716 }
5717
5718 /* Else insert an UPTO if the max is greater than the min, again
5719 preceded by the character, for the previously inserted code. If the
5720 UPTO is just for 1 instance, we can use QUERY instead. */
5721
5722 else if (repeat_max != repeat_min)
5723 {
5724 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5725 if (utf && (c & UTF_LENGTH) != 0)
5726 {
5727 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5728 code += c & 7;
5729 }
5730 else
5731 #endif
5732 *code++ = c;
5733 if (prop_type >= 0)
5734 {
5735 *code++ = prop_type;
5736 *code++ = prop_value;
5737 }
5738 repeat_max -= repeat_min;
5739
5740 if (repeat_max == 1)
5741 {
5742 *code++ = OP_QUERY + repeat_type;
5743 }
5744 else
5745 {
5746 *code++ = OP_UPTO + repeat_type;
5747 PUT2INC(code, 0, repeat_max);
5748 }
5749 }
5750 }
5751
5752 /* The character or character type itself comes last in all cases. */
5753
5754 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5755 if (utf && (c & UTF_LENGTH) != 0)
5756 {
5757 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5758 code += c & 7;
5759 }
5760 else
5761 #endif
5762 *code++ = c;
5763
5764 /* For a repeated Unicode property match, there are two extra bytes that
5765 define the required property. */
5766
5767 #ifdef SUPPORT_UCP
5768 if (prop_type >= 0)
5769 {
5770 *code++ = prop_type;
5771 *code++ = prop_value;
5772 }
5773 #endif
5774 }
5775
5776 /* If previous was a character class or a back reference, we put the repeat
5777 stuff after it, but just skip the item if the repeat was {0,0}. */
5778
5779 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5780 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5781 *previous == OP_XCLASS ||
5782 #endif
5783 *previous == OP_REF || *previous == OP_REFI ||
5784 *previous == OP_DNREF || *previous == OP_DNREFI)
5785 {
5786 if (repeat_max == 0)
5787 {
5788 code = previous;
5789 goto END_REPEAT;
5790 }
5791
5792 if (repeat_min == 0 && repeat_max == -1)
5793 *code++ = OP_CRSTAR + repeat_type;
5794 else if (repeat_min == 1 && repeat_max == -1)
5795 *code++ = OP_CRPLUS + repeat_type;
5796 else if (repeat_min == 0 && repeat_max == 1)
5797 *code++ = OP_CRQUERY + repeat_type;
5798 else
5799 {
5800 *code++ = OP_CRRANGE + repeat_type;
5801 PUT2INC(code, 0, repeat_min);
5802 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5803 PUT2INC(code, 0, repeat_max);
5804 }
5805 }
5806
5807 /* If previous was a bracket group, we may have to replicate it in certain
5808 cases. Note that at this point we can encounter only the "basic" bracket
5809 opcodes such as BRA and CBRA, as this is the place where they get converted
5810 into the more special varieties such as BRAPOS and SBRA. A test for >=
5811 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5812 ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5813 Originally, PCRE did not allow repetition of assertions, but now it does,
5814 for Perl compatibility. */
5815
5816 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5817 {
5818 register int i;
5819 int len = (int)(code - previous);
5820 pcre_uchar *bralink = NULL;
5821 pcre_uchar *brazeroptr = NULL;
5822
5823 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5824 we just ignore the repeat. */
5825
5826 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5827 goto END_REPEAT;
5828
5829 /* There is no sense in actually repeating assertions. The only potential
5830 use of repetition is in cases when the assertion is optional. Therefore,
5831 if the minimum is greater than zero, just ignore the repeat. If the
5832 maximum is not zero or one, set it to 1. */
5833
5834 if (*previous < OP_ONCE) /* Assertion */
5835 {
5836 if (repeat_min > 0) goto END_REPEAT;
5837 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5838 }
5839
5840 /* The case of a zero minimum is special because of the need to stick
5841 OP_BRAZERO in front of it, and because the group appears once in the
5842 data, whereas in other cases it appears the minimum number of times. For
5843 this reason, it is simplest to treat this case separately, as otherwise
5844 the code gets far too messy. There are several special subcases when the
5845 minimum is zero. */
5846
5847 if (repeat_min == 0)
5848 {
5849 /* If the maximum is also zero, we used to just omit the group from the
5850 output altogether, like this:
5851
5852 ** if (repeat_max == 0)
5853 ** {
5854 ** code = previous;
5855 ** goto END_REPEAT;
5856 ** }
5857
5858 However, that fails when a group or a subgroup within it is referenced
5859 as a subroutine from elsewhere in the pattern, so now we stick in
5860 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5861 don't have a list of which groups are referenced, we cannot do this
5862 selectively.
5863
5864 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5865 and do no more at this point. However, we do need to adjust any
5866 OP_RECURSE calls inside the group that refer to the group itself or any
5867 internal or forward referenced group, because the offset is from the
5868 start of the whole regex. Temporarily terminate the pattern while doing
5869 this. */
5870
5871 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5872 {
5873 *code = OP_END;
5874 adjust_recurse(previous, 1, utf, cd, save_hwm);
5875 memmove(previous + 1, previous, IN_UCHARS(len));
5876 code++;
5877 if (repeat_max == 0)
5878 {
5879 *previous++ = OP_SKIPZERO;
5880 goto END_REPEAT;
5881 }
5882 brazeroptr = previous; /* Save for possessive optimizing */
5883 *previous++ = OP_BRAZERO + repeat_type;
5884 }
5885
5886 /* If the maximum is greater than 1 and limited, we have to replicate
5887 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
5888 The first one has to be handled carefully because it's the original
5889 copy, which has to be moved up. The remainder can be handled by code
5890 that is common with the non-zero minimum case below. We have to
5891 adjust the value or repeat_max, since one less copy is required. Once
5892 again, we may have to adjust any OP_RECURSE calls inside the group. */
5893
5894 else
5895 {
5896 int offset;
5897 *code = OP_END;
5898 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5899 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5900 code += 2 + LINK_SIZE;
5901 *previous++ = OP_BRAZERO + repeat_type;
5902 *previous++ = OP_BRA;
5903
5904 /* We chain together the bracket offset fields that have to be
5905 filled in later when the ends of the brackets are reached. */
5906
5907 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
5908 bralink = previous;
5909 PUTINC(previous, 0, offset);
5910 }
5911
5912 repeat_max--;
5913 }
5914
5915 /* If the minimum is greater than zero, replicate the group as many
5916 times as necessary, and adjust the maximum to the number of subsequent
5917 copies that we need. If we set a first char from the group, and didn't
5918 set a required char, copy the latter from the former. If there are any
5919 forward reference subroutine calls in the group, there will be entries on
5920 the workspace list; replicate these with an appropriate increment. */
5921
5922 else
5923 {
5924 if (repeat_min > 1)
5925 {
5926 /* In the pre-compile phase, we don't actually do the replication. We
5927 just adjust the length as if we had. Do some paranoid checks for
5928 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
5929 integer type when available, otherwise double. */
5930
5931 if (lengthptr != NULL)
5932 {
5933 int delta = (repeat_min - 1)*length_prevgroup;
5934 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
5935 (INT64_OR_DOUBLE)length_prevgroup >
5936 (INT64_OR_DOUBLE)INT_MAX ||
5937 OFLOW_MAX - *lengthptr < delta)
5938 {
5939 *errorcodeptr = ERR20;
5940 goto FAILED;
5941 }
5942 *lengthptr += delta;
5943 }
5944
5945 /* This is compiling for real. If there is a set first byte for
5946 the group, and we have not yet set a "required byte", set it. Make
5947 sure there is enough workspace for copying forward references before
5948 doing the copy. */
5949
5950 else
5951 {
5952 if (groupsetfirstchar && reqcharflags < 0)
5953 {
5954 reqchar = firstchar;
5955 reqcharflags = firstcharflags;
5956 }
5957
5958 for (i = 1; i < repeat_min; i++)
5959 {
5960 pcre_uchar *hc;
5961 pcre_uchar *this_hwm = cd->hwm;
5962 memcpy(code, previous, IN_UCHARS(len));
5963
5964 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5965 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5966 {
5967 int save_offset = save_hwm - cd->start_workspace;
5968 int this_offset = this_hwm - cd->start_workspace;
5969 *errorcodeptr = expand_workspace(cd);
5970 if (*errorcodeptr != 0) goto FAILED;
5971 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5972 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5973 }
5974
5975 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5976 {
5977 PUT(cd->hwm, 0, GET(hc, 0) + len);
5978 cd->hwm += LINK_SIZE;
5979 }
5980 save_hwm = this_hwm;
5981 code += len;
5982 }
5983 }
5984 }
5985
5986 if (repeat_max > 0) repeat_max -= repeat_min;
5987 }
5988
5989 /* This code is common to both the zero and non-zero minimum cases. If
5990 the maximum is limited, it replicates the group in a nested fashion,
5991 remembering the bracket starts on a stack. In the case of a zero minimum,
5992 the first one was set up above. In all cases the repeat_max now specifies
5993 the number of additional copies needed. Again, we must remember to
5994 replicate entries on the forward reference list. */
5995
5996 if (repeat_max >= 0)
5997 {
5998 /* In the pre-compile phase, we don't actually do the replication. We
5999 just adjust the length as if we had. For each repetition we must add 1
6000 to the length for BRAZERO and for all but the last repetition we must
6001 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
6002 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
6003 a 64-bit integer type when available, otherwise double. */
6004
6005 if (lengthptr != NULL && repeat_max > 0)
6006 {
6007 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
6008 2 - 2*LINK_SIZE; /* Last one doesn't nest */
6009 if ((INT64_OR_DOUBLE)repeat_max *
6010 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
6011 > (INT64_OR_DOUBLE)INT_MAX ||
6012 OFLOW_MAX - *lengthptr < delta)
6013 {
6014 *errorcodeptr = ERR20;
6015 goto FAILED;
6016 }
6017 *lengthptr += delta;
6018 }
6019
6020 /* This is compiling for real */
6021
6022 else for (i = repeat_max - 1; i >= 0; i--)
6023 {
6024 pcre_uchar *hc;
6025 pcre_uchar *this_hwm = cd->hwm;
6026
6027 *code++ = OP_BRAZERO + repeat_type;
6028
6029 /* All but the final copy start a new nesting, maintaining the
6030 chain of brackets outstanding. */
6031
6032 if (i != 0)
6033 {
6034 int offset;
6035 *code++ = OP_BRA;
6036 offset = (bralink == NULL)? 0 : (int)(code - bralink);
6037 bralink = code;
6038 PUTINC(code, 0, offset);
6039 }
6040
6041 memcpy(code, previous, IN_UCHARS(len));
6042
6043 /* Ensure there is enough workspace for forward references before
6044 copying them. */
6045
6046 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6047 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
6048 {
6049 int save_offset = save_hwm - cd->start_workspace;
6050 int this_offset = this_hwm - cd->start_workspace;
6051 *errorcodeptr = expand_workspace(cd);
6052 if (*errorcodeptr != 0) goto FAILED;
6053 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
6054 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
6055 }
6056
6057 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
6058 {
6059 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
6060 cd->hwm += LINK_SIZE;
6061 }
6062 save_hwm = this_hwm;
6063 code += len;
6064 }
6065
6066 /* Now chain through the pending brackets, and fill in their length
6067 fields (which are holding the chain links pro tem). */
6068
6069 while (bralink != NULL)
6070 {
6071 int oldlinkoffset;
6072 int offset = (int)(code - bralink + 1);
6073 pcre_uchar *bra = code - offset;
6074 oldlinkoffset = GET(bra, 1);
6075 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
6076 *code++ = OP_KET;
6077 PUTINC(code, 0, offset);
6078 PUT(bra, 1, offset);
6079 }
6080 }
6081
6082 /* If the maximum is unlimited, set a repeater in the final copy. For
6083 ONCE brackets, that's all we need to do. However, possessively repeated
6084 ONCE brackets can be converted into non-capturing brackets, as the
6085 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
6086 deal with possessive ONCEs specially.
6087
6088 Otherwise, when we are doing the actual compile phase, check to see
6089 whether this group is one that could match an empty string. If so,
6090 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
6091 that runtime checking can be done. [This check is also applied to ONCE
6092 groups at runtime, but in a different way.]
6093
6094 Then, if the quantifier was possessive and the bracket is not a
6095 conditional, we convert the BRA code to the POS form, and the KET code to
6096 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
6097 subpattern at both the start and at the end.) The use of special opcodes
6098 makes it possible to reduce greatly the stack usage in pcre_exec(). If
6099 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
6100
6101 Then, if the minimum number of matches is 1 or 0, cancel the possessive
6102 flag so that the default action below, of wrapping everything inside
6103 atomic brackets, does not happen. When the minimum is greater than 1,
6104 there will be earlier copies of the group, and so we still have to wrap
6105 the whole thing. */
6106
6107 else
6108 {
6109 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
6110 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
6111
6112 /* Convert possessive ONCE brackets to non-capturing */
6113
6114 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
6115 possessive_quantifier) *bracode = OP_BRA;
6116
6117 /* For non-possessive ONCE brackets, all we need to do is to
6118 set the KET. */
6119
6120 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
6121 *ketcode = OP_KETRMAX + repeat_type;
6122
6123 /* Handle non-ONCE brackets and possessive ONCEs (which have been
6124 converted to non-capturing above). */
6125
6126 else
6127 {
6128 /* In the compile phase, check for empty string matching. */
6129
6130 if (lengthptr == NULL)
6131 {
6132 pcre_uchar *scode = bracode;
6133 do
6134 {
6135 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6136 {
6137 *bracode += OP_SBRA - OP_BRA;
6138 break;
6139 }
6140 scode += GET(scode, 1);
6141 }
6142 while (*scode == OP_ALT);
6143 }
6144
6145 /* Handle possessive quantifiers. */
6146
6147 if (possessive_quantifier)
6148 {
6149 /* For COND brackets, we wrap the whole thing in a possessively
6150 repeated non-capturing bracket, because we have not invented POS
6151 versions of the COND opcodes. Because we are moving code along, we
6152 must ensure that any pending recursive references are updated. */
6153
6154 if (*bracode == OP_COND || *bracode == OP_SCOND)
6155 {
6156 int nlen = (int)(code - bracode);
6157 *code = OP_END;
6158 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
6159 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6160 code += 1 + LINK_SIZE;
6161 nlen += 1 + LINK_SIZE;
6162 *bracode = OP_BRAPOS;
6163 *code++ = OP_KETRPOS;
6164 PUTINC(code, 0, nlen);
6165 PUT(bracode, 1, nlen);
6166 }
6167
6168 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6169
6170 else
6171 {
6172 *bracode += 1; /* Switch to xxxPOS opcodes */
6173 *ketcode = OP_KETRPOS;
6174 }
6175
6176 /* If the minimum is zero, mark it as possessive, then unset the
6177 possessive flag when the minimum is 0 or 1. */
6178
6179 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6180 if (repeat_min < 2) possessive_quantifier = FALSE;
6181 }
6182
6183 /* Non-possessive quantifier */
6184
6185 else *ketcode = OP_KETRMAX + repeat_type;
6186 }
6187 }
6188 }
6189
6190 /* If previous is OP_FAIL, it was generated by an empty class [] in
6191 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6192 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6193 error above. We can just ignore the repeat in JS case. */
6194
6195 else if (*previous == OP_FAIL) goto END_REPEAT;
6196
6197 /* Else there's some kind of shambles */
6198
6199 else
6200 {
6201 *errorcodeptr = ERR11;
6202 goto FAILED;
6203 }
6204
6205 /* If the character following a repeat is '+', possessive_quantifier is
6206 TRUE. For some opcodes, there are special alternative opcodes for this
6207 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6208 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6209 Sun's Java package, but the special opcodes can optimize it.
6210
6211 Some (but not all) possessively repeated subpatterns have already been
6212 completely handled in the code just above. For them, possessive_quantifier
6213 is always FALSE at this stage. Note that the repeated item starts at
6214 tempcode, not at previous, which might be the first part of a string whose
6215 (former) last char we repeated. */
6216
6217 if (possessive_quantifier)
6218 {
6219 int len;
6220
6221 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6222 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6223 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6224 remains is greater than zero, there's a further opcode that can be
6225 handled. If not, do nothing, leaving the EXACT alone. */
6226
6227 switch(*tempcode)
6228 {
6229 case OP_TYPEEXACT:
6230 tempcode += PRIV(OP_lengths)[*tempcode] +
6231 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6232 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6233 break;
6234
6235 /* CHAR opcodes are used for exacts whose count is 1. */
6236
6237 case OP_CHAR:
6238 case OP_CHARI:
6239 case OP_NOT:
6240 case OP_NOTI:
6241 case OP_EXACT:
6242 case OP_EXACTI:
6243 case OP_NOTEXACT:
6244 case OP_NOTEXACTI:
6245 tempcode += PRIV(OP_lengths)[*tempcode];
6246 #ifdef SUPPORT_UTF
6247 if (utf && HAS_EXTRALEN(tempcode[-1]))
6248 tempcode += GET_EXTRALEN(tempcode[-1]);
6249 #endif
6250 break;
6251
6252 /* For the class opcodes, the repeat operator appears at the end;
6253 adjust tempcode to point to it. */
6254
6255 case OP_CLASS:
6256 case OP_NCLASS:
6257 tempcode += 1 + 32/sizeof(pcre_uchar);
6258 break;
6259
6260 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6261 case OP_XCLASS:
6262 tempcode += GET(tempcode, 1);
6263 break;
6264 #endif
6265 }
6266
6267 /* If tempcode is equal to code (which points to the end of the repeated
6268 item), it means we have skipped an EXACT item but there is no following
6269 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6270 all other cases, tempcode will be pointing to the repeat opcode, and will
6271 be less than code, so the value of len will be greater than 0. */
6272
6273 len = (int)(code - tempcode);
6274 if (len > 0)
6275 {
6276 unsigned int repcode = *tempcode;
6277
6278 /* There is a table for possessifying opcodes, all of which are less
6279 than OP_CALLOUT. A zero entry means there is no possessified version.
6280 */
6281
6282 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6283 *tempcode = opcode_possessify[repcode];
6284
6285 /* For opcode without a special possessified version, wrap the item in
6286 ONCE brackets. Because we are moving code along, we must ensure that any
6287 pending recursive references are updated. */
6288
6289 else
6290 {
6291 *code = OP_END;
6292 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6293 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6294 code += 1 + LINK_SIZE;
6295 len += 1 + LINK_SIZE;
6296 tempcode[0] = OP_ONCE;
6297 *code++ = OP_KET;
6298 PUTINC(code, 0, len);
6299 PUT(tempcode, 1, len);
6300 }
6301 }
6302
6303 #ifdef NEVER
6304 if (len > 0) switch (*tempcode)
6305 {
6306 case OP_STAR: *tempcode = OP_POSSTAR; break;
6307 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6308 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6309 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6310
6311 case OP_STARI: *tempcode = OP_POSSTARI; break;
6312 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6313 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6314 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6315
6316 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6317 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6318 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6319 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6320
6321 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6322 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6323 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6324 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6325
6326 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6327 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6328 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6329 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6330
6331 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6332 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6333 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6334 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6335
6336 /* Because we are moving code along, we must ensure that any
6337 pending recursive references are updated. */
6338
6339 default:
6340 *code = OP_END;
6341 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6342 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6343 code += 1 + LINK_SIZE;
6344 len += 1 + LINK_SIZE;
6345 tempcode[0] = OP_ONCE;
6346 *code++ = OP_KET;
6347 PUTINC(code, 0, len);
6348 PUT(tempcode, 1, len);
6349 break;
6350 }
6351 #endif
6352 }
6353
6354 /* In all case we no longer have a previous item. We also set the
6355 "follows varying string" flag for subsequently encountered reqchars if
6356 it isn't already set and we have just passed a varying length item. */
6357
6358 END_REPEAT:
6359 previous = NULL;
6360 cd->req_varyopt |= reqvary;
6361 break;
6362
6363
6364 /* ===================================================================*/
6365 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6366 lookbehind or option setting or condition or all the other extended
6367 parenthesis forms. */
6368
6369 case CHAR_LEFT_PARENTHESIS:
6370 newoptions = options;
6371 skipbytes = 0;
6372 bravalue = OP_CBRA;
6373 save_hwm = cd->hwm;
6374 reset_bracount = FALSE;
6375
6376 /* First deal with various "verbs" that can be introduced by '*'. */
6377
6378 ptr++;
6379 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6380 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6381 {
6382 int i, namelen;
6383 int arglen = 0;
6384 const char *vn = verbnames;
6385 const pcre_uchar *name = ptr + 1;
6386 const pcre_uchar *arg = NULL;
6387 previous = NULL;
6388 ptr++;
6389 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6390 namelen = (int)(ptr - name);
6391
6392 /* It appears that Perl allows any characters whatsoever, other than
6393 a closing parenthesis, to appear in arguments, so we no longer insist on
6394 letters, digits, and underscores. */
6395
6396 if (*ptr == CHAR_COLON)
6397 {
6398 arg = ++ptr;
6399 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6400 arglen = (int)(ptr - arg);
6401 if ((unsigned int)arglen > MAX_MARK)
6402 {
6403 *errorcodeptr = ERR75;
6404 goto FAILED;
6405 }
6406 }
6407
6408 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6409 {
6410 *errorcodeptr = ERR60;
6411 goto FAILED;
6412 }
6413
6414 /* Scan the table of verb names */
6415
6416 for (i = 0; i < verbcount; i++)
6417 {
6418 if (namelen == verbs[i].len &&
6419 STRNCMP_UC_C8(name, vn, namelen) == 0)
6420 {
6421 int setverb;
6422
6423 /* Check for open captures before ACCEPT and convert it to
6424 ASSERT_ACCEPT if in an assertion. */
6425
6426 if (verbs[i].op == OP_ACCEPT)
6427 {
6428 open_capitem *oc;
6429 if (arglen != 0)
6430 {
6431 *errorcodeptr = ERR59;
6432 goto FAILED;
6433 }
6434 cd->had_accept = TRUE;
6435 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6436 {
6437 *code++ = OP_CLOSE;
6438 PUT2INC(code, 0, oc->number);
6439 }
6440 setverb = *code++ =
6441 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6442
6443 /* Do not set firstchar after *ACCEPT */
6444 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6445 }
6446
6447 /* Handle other cases with/without an argument */
6448
6449 else if (arglen == 0)
6450 {
6451 if (verbs[i].op < 0) /* Argument is mandatory */
6452 {
6453 *errorcodeptr = ERR66;
6454 goto FAILED;
6455 }
6456 setverb = *code++ = verbs[i].op;
6457 }
6458
6459 else
6460 {
6461 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6462 {
6463 *errorcodeptr = ERR59;
6464 goto FAILED;
6465 }
6466 setverb = *code++ = verbs[i].op_arg;
6467 *code++ = arglen;
6468 memcpy(code, arg, IN_UCHARS(arglen));
6469 code += arglen;
6470 *code++ = 0;
6471 }
6472
6473 switch (setverb)
6474 {
6475 case OP_THEN:
6476 case OP_THEN_ARG:
6477 cd->external_flags |= PCRE_HASTHEN;
6478 break;
6479
6480 case OP_PRUNE:
6481 case OP_PRUNE_ARG:
6482 case OP_SKIP:
6483 case OP_SKIP_ARG:
6484 cd->had_pruneorskip = TRUE;
6485 break;
6486 }
6487
6488 break; /* Found verb, exit loop */
6489 }
6490
6491 vn += verbs[i].len + 1;
6492 }
6493
6494 if (i < verbcount) continue; /* Successfully handled a verb */
6495 *errorcodeptr = ERR60; /* Verb not recognized */
6496 goto FAILED;
6497 }
6498
6499 /* Deal with the extended parentheses; all are introduced by '?', and the
6500 appearance of any of them means that this is not a capturing group. */
6501
6502 else if (*ptr == CHAR_QUESTION_MARK)
6503 {
6504 int i, set, unset, namelen;
6505 int *optset;
6506 const pcre_uchar *name;
6507 pcre_uchar *slot;
6508
6509 switch (*(++ptr))
6510 {
6511 case CHAR_NUMBER_SIGN: /* Comment; skip to ket */
6512 ptr++;
6513 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6514 if (*ptr == CHAR_NULL)
6515 {
6516 *errorcodeptr = ERR18;
6517 goto FAILED;
6518 }
6519 continue;
6520
6521
6522 /* ------------------------------------------------------------ */
6523 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6524 reset_bracount = TRUE;
6525 /* Fall through */
6526
6527 /* ------------------------------------------------------------ */
6528 case CHAR_COLON: /* Non-capturing bracket */
6529 bravalue = OP_BRA;
6530 ptr++;
6531 break;
6532
6533
6534 /* ------------------------------------------------------------ */
6535 case CHAR_LEFT_PARENTHESIS:
6536 bravalue = OP_COND; /* Conditional group */
6537 tempptr = ptr;
6538
6539 /* A condition can be an assertion, a number (referring to a numbered
6540 group's having been set), a name (referring to a named group), or 'R',
6541 referring to recursion. R<digits> and R&name are also permitted for
6542 recursion tests.
6543
6544 There are ways of testing a named group: (?(name)) is used by Python;
6545 Perl 5.10 onwards uses (?(<name>) or (?('name')).
6546
6547 There is one unfortunate ambiguity, caused by history. 'R' can be the
6548 recursive thing or the name 'R' (and similarly for 'R' followed by
6549 digits). We look for a name first; if not found, we try the other case.
6550
6551 For compatibility with auto-callouts, we allow a callout to be
6552 specified before a condition that is an assertion. First, check for the
6553 syntax of a callout; if found, adjust the temporary pointer that is
6554 used to check for an assertion condition. That's all that is needed! */
6555
6556 if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6557 {
6558 for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6559 if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6560 tempptr += i + 1;
6561 }
6562
6563 /* For conditions that are assertions, check the syntax, and then exit
6564 the switch. This will take control down to where bracketed groups,
6565 including assertions, are processed. */
6566
6567 if (tempptr[1] == CHAR_QUESTION_MARK &&
6568 (tempptr[2] == CHAR_EQUALS_SIGN ||
6569 tempptr[2] == CHAR_EXCLAMATION_MARK ||
6570 tempptr[2] == CHAR_LESS_THAN_SIGN))
6571 break;
6572
6573 /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6574 need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6575
6576 code[1+LINK_SIZE] = OP_CREF;
6577 skipbytes = 1+IMM2_SIZE;
6578 refsign = -1;
6579
6580 /* Check for a test for recursion in a named group. */
6581
6582 ptr++;
6583 if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6584 {
6585 terminator = -1;
6586 ptr += 2;
6587 code[1+LINK_SIZE] = OP_RREF; /* Change the type of test */
6588 }
6589
6590 /* Check for a test for a named group's having been set, using the Perl
6591 syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6592 syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6593
6594 else if (*ptr == CHAR_LESS_THAN_SIGN)
6595 {
6596 terminator = CHAR_GREATER_THAN_SIGN;
6597 ptr++;
6598 }
6599 else if (*ptr == CHAR_APOSTROPHE)
6600 {
6601 terminator = CHAR_APOSTROPHE;
6602 ptr++;
6603 }
6604 else
6605 {
6606 terminator = CHAR_NULL;
6607 if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6608 else if (IS_DIGIT(*ptr)) refsign = 0;
6609 }
6610
6611 /* Handle a number */
6612
6613 if (refsign >= 0)
6614 {
6615 recno = 0;
6616 while (IS_DIGIT(*ptr))
6617 {
6618 recno = recno * 10 + (int)(*ptr - CHAR_0);
6619 ptr++;
6620 }
6621 }
6622
6623 /* Otherwise we expect to read a name; anything else is an error. When
6624 a name is one of a number of duplicates, a different opcode is used and
6625 it needs more memory. Unfortunately we cannot tell whether a name is a
6626 duplicate in the first pass, so we have to allow for more memory. */
6627
6628 else
6629 {
6630 if (IS_DIGIT(*ptr))
6631 {
6632 *errorcodeptr = ERR84;
6633 goto FAILED;
6634 }
6635 if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6636 {
6637 *errorcodeptr = ERR28; /* Assertion expected */
6638 goto FAILED;
6639 }
6640 name = ptr++;
6641 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6642 {
6643 ptr++;
6644 }
6645 namelen = (int)(ptr - name);
6646 if (lengthptr != NULL) *lengthptr += IMM2_SIZE;
6647 }
6648
6649 /* Check the terminator */
6650
6651 if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6652 *ptr++ != CHAR_RIGHT_PARENTHESIS)
6653 {
6654 ptr--; /* Error offset */
6655 *errorcodeptr = ERR26; /* Malformed number or name */
6656 goto FAILED;
6657 }
6658
6659 /* Do no further checking in the pre-compile phase. */
6660
6661 if (lengthptr != NULL) break;