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

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1387 - (show annotations)
Sat Nov 2 18:29:05 2013 UTC (6 years, 1 month ago) by ph10
File MIME type: text/plain
File size: 305554 byte(s)
Update POSIX class handling in UCP mode.
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 /* Tables of names of POSIX character classes and their lengths. The names are
264 now all in a single string, to reduce the number of relocations when a shared
265 library is dynamically loaded. The list of lengths is terminated by a zero
266 length entry. The first three must be alpha, lower, upper, as this is assumed
267 for handling case independence. The indices for graph, print, and punct are
268 needed, so identify them. */
269
270 static const char posix_names[] =
271 STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
272 STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
273 STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
274 STRING_word0 STRING_xdigit;
275
276 static const pcre_uint8 posix_name_lengths[] = {
277 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
278
279 #define PC_GRAPH 8
280 #define PC_PRINT 9
281 #define PC_PUNCT 10
282
283
284 /* Table of class bit maps for each POSIX class. Each class is formed from a
285 base map, with an optional addition or removal of another map. Then, for some
286 classes, there is some additional tweaking: for [:blank:] the vertical space
287 characters are removed, and for [:alpha:] and [:alnum:] the underscore
288 character is removed. The triples in the table consist of the base map offset,
289 second map offset or -1 if no second map, and a non-negative value for map
290 addition or a negative value for map subtraction (if there are two maps). The
291 absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
292 remove vertical space characters, 2 => remove underscore. */
293
294 static const int posix_class_maps[] = {
295 cbit_word, cbit_digit, -2, /* alpha */
296 cbit_lower, -1, 0, /* lower */
297 cbit_upper, -1, 0, /* upper */
298 cbit_word, -1, 2, /* alnum - word without underscore */
299 cbit_print, cbit_cntrl, 0, /* ascii */
300 cbit_space, -1, 1, /* blank - a GNU extension */
301 cbit_cntrl, -1, 0, /* cntrl */
302 cbit_digit, -1, 0, /* digit */
303 cbit_graph, -1, 0, /* graph */
304 cbit_print, -1, 0, /* print */
305 cbit_punct, -1, 0, /* punct */
306 cbit_space, -1, 0, /* space */
307 cbit_word, -1, 0, /* word - a Perl extension */
308 cbit_xdigit,-1, 0 /* xdigit */
309 };
310
311 /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
312 Unicode property escapes. */
313
314 #ifdef SUPPORT_UCP
315 static const pcre_uchar string_PNd[] = {
316 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
317 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
318 static const pcre_uchar string_pNd[] = {
319 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
320 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
321 static const pcre_uchar string_PXsp[] = {
322 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
323 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
324 static const pcre_uchar string_pXsp[] = {
325 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
326 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
327 static const pcre_uchar string_PXwd[] = {
328 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
329 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
330 static const pcre_uchar string_pXwd[] = {
331 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
332 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
333
334 static const pcre_uchar *substitutes[] = {
335 string_PNd, /* \D */
336 string_pNd, /* \d */
337 string_PXsp, /* \S */ /* Xsp is Perl space, but from 8.34, Perl */
338 string_pXsp, /* \s */ /* space and POSIX space are the same. */
339 string_PXwd, /* \W */
340 string_pXwd /* \w */
341 };
342
343 /* The POSIX class substitutes must be in the order of the POSIX class names,
344 defined above, and there are both positive and negative cases. NULL means no
345 general substitute of a Unicode property escape (\p or \P). However, for some
346 POSIX classes (e.g. graph, print, punct) a special property code is compiled
347 directly. */
348
349 static const pcre_uchar string_pL[] = {
350 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
351 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
352 static const pcre_uchar string_pLl[] = {
353 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
354 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
355 static const pcre_uchar string_pLu[] = {
356 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
357 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
358 static const pcre_uchar string_pXan[] = {
359 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
360 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
361 static const pcre_uchar string_h[] = {
362 CHAR_BACKSLASH, CHAR_h, '\0' };
363 static const pcre_uchar string_pXps[] = {
364 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
365 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
366 static const pcre_uchar string_PL[] = {
367 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
368 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
369 static const pcre_uchar string_PLl[] = {
370 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
371 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
372 static const pcre_uchar string_PLu[] = {
373 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
374 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
375 static const pcre_uchar string_PXan[] = {
376 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
377 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
378 static const pcre_uchar string_H[] = {
379 CHAR_BACKSLASH, CHAR_H, '\0' };
380 static const pcre_uchar string_PXps[] = {
381 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
382 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
383
384 static const pcre_uchar *posix_substitutes[] = {
385 string_pL, /* alpha */
386 string_pLl, /* lower */
387 string_pLu, /* upper */
388 string_pXan, /* alnum */
389 NULL, /* ascii */
390 string_h, /* blank */
391 NULL, /* cntrl */
392 string_pNd, /* digit */
393 NULL, /* graph */
394 NULL, /* print */
395 NULL, /* punct */
396 string_pXps, /* space */ /* Xps is POSIX space, but from 8.34 */
397 string_pXwd, /* word */ /* Perl and POSIX space are the same */
398 NULL, /* xdigit */
399 /* Negated cases */
400 string_PL, /* ^alpha */
401 string_PLl, /* ^lower */
402 string_PLu, /* ^upper */
403 string_PXan, /* ^alnum */
404 NULL, /* ^ascii */
405 string_H, /* ^blank */
406 NULL, /* ^cntrl */
407 string_PNd, /* ^digit */
408 NULL, /* ^graph */
409 NULL, /* ^print */
410 NULL, /* ^punct */
411 string_PXps, /* ^space */ /* Xps is POSIX space, but from 8.34 */
412 string_PXwd, /* ^word */ /* Perl and POSIX space are the same */
413 NULL /* ^xdigit */
414 };
415 #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
416 #endif
417
418 #define STRING(a) # a
419 #define XSTRING(s) STRING(s)
420
421 /* The texts of compile-time error messages. These are "char *" because they
422 are passed to the outside world. Do not ever re-use any error number, because
423 they are documented. Always add a new error instead. Messages marked DEAD below
424 are no longer used. This used to be a table of strings, but in order to reduce
425 the number of relocations needed when a shared library is loaded dynamically,
426 it is now one long string. We cannot use a table of offsets, because the
427 lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
428 simply count through to the one we want - this isn't a performance issue
429 because these strings are used only when there is a compilation error.
430
431 Each substring ends with \0 to insert a null character. This includes the final
432 substring, so that the whole string ends with \0\0, which can be detected when
433 counting through. */
434
435 static const char error_texts[] =
436 "no error\0"
437 "\\ at end of pattern\0"
438 "\\c at end of pattern\0"
439 "unrecognized character follows \\\0"
440 "numbers out of order in {} quantifier\0"
441 /* 5 */
442 "number too big in {} quantifier\0"
443 "missing terminating ] for character class\0"
444 "invalid escape sequence in character class\0"
445 "range out of order in character class\0"
446 "nothing to repeat\0"
447 /* 10 */
448 "operand of unlimited repeat could match the empty string\0" /** DEAD **/
449 "internal error: unexpected repeat\0"
450 "unrecognized character after (? or (?-\0"
451 "POSIX named classes are supported only within a class\0"
452 "missing )\0"
453 /* 15 */
454 "reference to non-existent subpattern\0"
455 "erroffset passed as NULL\0"
456 "unknown option bit(s) set\0"
457 "missing ) after comment\0"
458 "parentheses nested too deeply\0" /** DEAD **/
459 /* 20 */
460 "regular expression is too large\0"
461 "failed to get memory\0"
462 "unmatched parentheses\0"
463 "internal error: code overflow\0"
464 "unrecognized character after (?<\0"
465 /* 25 */
466 "lookbehind assertion is not fixed length\0"
467 "malformed number or name after (?(\0"
468 "conditional group contains more than two branches\0"
469 "assertion expected after (?(\0"
470 "(?R or (?[+-]digits must be followed by )\0"
471 /* 30 */
472 "unknown POSIX class name\0"
473 "POSIX collating elements are not supported\0"
474 "this version of PCRE is compiled without UTF support\0"
475 "spare error\0" /** DEAD **/
476 "character value in \\x{} or \\o{} is too large\0"
477 /* 35 */
478 "invalid condition (?(0)\0"
479 "\\C not allowed in lookbehind assertion\0"
480 "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
481 "number after (?C is > 255\0"
482 "closing ) for (?C expected\0"
483 /* 40 */
484 "recursive call could loop indefinitely\0"
485 "unrecognized character after (?P\0"
486 "syntax error in subpattern name (missing terminator)\0"
487 "two named subpatterns have the same name\0"
488 "invalid UTF-8 string\0"
489 /* 45 */
490 "support for \\P, \\p, and \\X has not been compiled\0"
491 "malformed \\P or \\p sequence\0"
492 "unknown property name after \\P or \\p\0"
493 "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
494 "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
495 /* 50 */
496 "repeated subpattern is too long\0" /** DEAD **/
497 "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
498 "internal error: overran compiling workspace\0"
499 "internal error: previously-checked referenced subpattern not found\0"
500 "DEFINE group contains more than one branch\0"
501 /* 55 */
502 "repeating a DEFINE group is not allowed\0" /** DEAD **/
503 "inconsistent NEWLINE options\0"
504 "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
505 "a numbered reference must not be zero\0"
506 "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
507 /* 60 */
508 "(*VERB) not recognized or malformed\0"
509 "number is too big\0"
510 "subpattern name expected\0"
511 "digit expected after (?+\0"
512 "] is an invalid data character in JavaScript compatibility mode\0"
513 /* 65 */
514 "different names for subpatterns of the same number are not allowed\0"
515 "(*MARK) must have an argument\0"
516 "this version of PCRE is not compiled with Unicode property support\0"
517 "\\c must be followed by an ASCII character\0"
518 "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
519 /* 70 */
520 "internal error: unknown opcode in find_fixedlength()\0"
521 "\\N is not supported in a class\0"
522 "too many forward references\0"
523 "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
524 "invalid UTF-16 string\0"
525 /* 75 */
526 "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
527 "character value in \\u.... sequence is too large\0"
528 "invalid UTF-32 string\0"
529 "setting UTF is disabled by the application\0"
530 "non-hex character in \\x{} (closing brace missing?)\0"
531 /* 80 */
532 "non-octal character in \\o{} (closing brace missing?)\0"
533 "missing opening brace after \\o\0"
534 ;
535
536 /* Table to identify digits and hex digits. This is used when compiling
537 patterns. Note that the tables in chartables are dependent on the locale, and
538 may mark arbitrary characters as digits - but the PCRE compiling code expects
539 to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
540 a private table here. It costs 256 bytes, but it is a lot faster than doing
541 character value tests (at least in some simple cases I timed), and in some
542 applications one wants PCRE to compile efficiently as well as match
543 efficiently.
544
545 For convenience, we use the same bit definitions as in chartables:
546
547 0x04 decimal digit
548 0x08 hexadecimal digit
549
550 Then we can use ctype_digit and ctype_xdigit in the code. */
551
552 /* Using a simple comparison for decimal numbers rather than a memory read
553 is much faster, and the resulting code is simpler (the compiler turns it
554 into a subtraction and unsigned comparison). */
555
556 #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
557
558 #ifndef EBCDIC
559
560 /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
561 UTF-8 mode. */
562
563 static const pcre_uint8 digitab[] =
564 {
565 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
566 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
567 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
568 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
569 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
570 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
571 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
572 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
573 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* @ - G */
574 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H - O */
575 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* P - W */
576 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* X - _ */
577 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
578 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
579 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
580 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
581 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
582 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
583 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
584 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
585 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
586 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
587 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
588 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
589 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
590 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
591 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
592 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
593 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
594 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
595 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
596 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
597
598 #else
599
600 /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
601
602 static const pcre_uint8 digitab[] =
603 {
604 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
605 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
606 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
607 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
608 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
609 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
610 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
611 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
612 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
613 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
614 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
615 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
616 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
617 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
618 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
619 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
620 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
621 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
622 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
623 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
624 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
625 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
626 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
627 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
628 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
629 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
630 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
631 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
632 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
633 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
634 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
635 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
636
637 static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
638 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
639 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
640 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
641 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
642 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
643 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
644 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
645 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
646 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
647 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
648 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
649 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
650 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
651 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
652 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
653 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
654 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
655 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
656 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
657 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
658 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
659 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
660 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
661 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
662 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
663 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
664 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
665 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
666 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
667 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
668 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
669 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
670 #endif
671
672
673 /* This table is used to check whether auto-possessification is possible
674 between adjacent character-type opcodes. The left-hand (repeated) opcode is
675 used to select the row, and the right-hand opcode is use to select the column.
676 A value of 1 means that auto-possessification is OK. For example, the second
677 value in the first row means that \D+\d can be turned into \D++\d.
678
679 The Unicode property types (\P and \p) have to be present to fill out the table
680 because of what their opcode values are, but the table values should always be
681 zero because property types are handled separately in the code. The last four
682 columns apply to items that cannot be repeated, so there is no need to have
683 rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
684 *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
685
686 #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
687 #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
688
689 static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
690 /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
691 { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
692 { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
693 { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
694 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
695 { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
696 { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
697 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
698 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
699 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
700 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
701 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
702 { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
703 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
704 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
705 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
706 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
707 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
708 };
709
710
711 /* This table is used to check whether auto-possessification is possible
712 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
713 left-hand (repeated) opcode is used to select the row, and the right-hand
714 opcode is used to select the column. The values are as follows:
715
716 0 Always return FALSE (never auto-possessify)
717 1 Character groups are distinct (possessify if both are OP_PROP)
718 2 Check character categories in the same group (general or particular)
719 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
720
721 4 Check left general category vs right particular category
722 5 Check right general category vs left particular category
723
724 6 Left alphanum vs right general category
725 7 Left space vs right general category
726 8 Left word vs right general category
727
728 9 Right alphanum vs left general category
729 10 Right space vs left general category
730 11 Right word vs left general category
731
732 12 Left alphanum vs right particular category
733 13 Left space vs right particular category
734 14 Left word vs right particular category
735
736 15 Right alphanum vs left particular category
737 16 Right space vs left particular category
738 17 Right word vs left particular category
739 */
740
741 static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
742 /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
743 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
744 { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
745 { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
746 { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
747 { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
748 { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
749 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
750 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
751 { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
752 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
753 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
754 };
755
756 /* This table is used to check whether auto-possessification is possible
757 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
758 specifies a general category and the other specifies a particular category. The
759 row is selected by the general category and the column by the particular
760 category. The value is 1 if the particular category is not part of the general
761 category. */
762
763 static const pcre_uint8 catposstab[7][30] = {
764 /* 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 */
765 { 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 */
766 { 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 */
767 { 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 */
768 { 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 */
769 { 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 */
770 { 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 */
771 { 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 */
772 };
773
774 /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
775 a general or particular category. The properties in each row are those
776 that apply to the character set in question. Duplication means that a little
777 unnecessary work is done when checking, but this keeps things much simpler
778 because they can all use the same code. For more details see the comment where
779 this table is used.
780
781 Note: SPACE and PXSPACE used to be different because Perl excluded VT from
782 "space", but from Perl 5.18 it's included, so both categories are treated the
783 same here. */
784
785 static const pcre_uint8 posspropstab[3][4] = {
786 { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
787 { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
788 { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
789 };
790
791 /* This table is used when converting repeating opcodes into possessified
792 versions as a result of an explicit possessive quantifier such as ++. A zero
793 value means there is no possessified version - in those cases the item in
794 question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
795 because all relevant opcodes are less than that. */
796
797 static const pcre_uint8 opcode_possessify[] = {
798 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */
799 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */
800
801 0, /* NOTI */
802 OP_POSSTAR, 0, /* STAR, MINSTAR */
803 OP_POSPLUS, 0, /* PLUS, MINPLUS */
804 OP_POSQUERY, 0, /* QUERY, MINQUERY */
805 OP_POSUPTO, 0, /* UPTO, MINUPTO */
806 0, /* EXACT */
807 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */
808
809 OP_POSSTARI, 0, /* STARI, MINSTARI */
810 OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */
811 OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */
812 OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */
813 0, /* EXACTI */
814 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */
815
816 OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */
817 OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */
818 OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */
819 OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */
820 0, /* NOTEXACT */
821 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
822
823 OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */
824 OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */
825 OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */
826 OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */
827 0, /* NOTEXACTI */
828 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
829
830 OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */
831 OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */
832 OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */
833 OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */
834 0, /* TYPEEXACT */
835 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
836
837 OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */
838 OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */
839 OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */
840 OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */
841 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */
842
843 0, 0, 0, /* CLASS, NCLASS, XCLASS */
844 0, 0, /* REF, REFI */
845 0, 0, /* DNREF, DNREFI */
846 0, 0 /* RECURSE, CALLOUT */
847 };
848
849
850
851 /*************************************************
852 * Find an error text *
853 *************************************************/
854
855 /* The error texts are now all in one long string, to save on relocations. As
856 some of the text is of unknown length, we can't use a table of offsets.
857 Instead, just count through the strings. This is not a performance issue
858 because it happens only when there has been a compilation error.
859
860 Argument: the error number
861 Returns: pointer to the error string
862 */
863
864 static const char *
865 find_error_text(int n)
866 {
867 const char *s = error_texts;
868 for (; n > 0; n--)
869 {
870 while (*s++ != CHAR_NULL) {};
871 if (*s == CHAR_NULL) return "Error text not found (please report)";
872 }
873 return s;
874 }
875
876
877
878 /*************************************************
879 * Expand the workspace *
880 *************************************************/
881
882 /* This function is called during the second compiling phase, if the number of
883 forward references fills the existing workspace, which is originally a block on
884 the stack. A larger block is obtained from malloc() unless the ultimate limit
885 has been reached or the increase will be rather small.
886
887 Argument: pointer to the compile data block
888 Returns: 0 if all went well, else an error number
889 */
890
891 static int
892 expand_workspace(compile_data *cd)
893 {
894 pcre_uchar *newspace;
895 int newsize = cd->workspace_size * 2;
896
897 if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
898 if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
899 newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
900 return ERR72;
901
902 newspace = (PUBL(malloc))(IN_UCHARS(newsize));
903 if (newspace == NULL) return ERR21;
904 memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
905 cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
906 if (cd->workspace_size > COMPILE_WORK_SIZE)
907 (PUBL(free))((void *)cd->start_workspace);
908 cd->start_workspace = newspace;
909 cd->workspace_size = newsize;
910 return 0;
911 }
912
913
914
915 /*************************************************
916 * Check for counted repeat *
917 *************************************************/
918
919 /* This function is called when a '{' is encountered in a place where it might
920 start a quantifier. It looks ahead to see if it really is a quantifier or not.
921 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
922 where the ddds are digits.
923
924 Arguments:
925 p pointer to the first char after '{'
926
927 Returns: TRUE or FALSE
928 */
929
930 static BOOL
931 is_counted_repeat(const pcre_uchar *p)
932 {
933 if (!IS_DIGIT(*p)) return FALSE;
934 p++;
935 while (IS_DIGIT(*p)) p++;
936 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
937
938 if (*p++ != CHAR_COMMA) return FALSE;
939 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
940
941 if (!IS_DIGIT(*p)) return FALSE;
942 p++;
943 while (IS_DIGIT(*p)) p++;
944
945 return (*p == CHAR_RIGHT_CURLY_BRACKET);
946 }
947
948
949
950 /*************************************************
951 * Handle escapes *
952 *************************************************/
953
954 /* This function is called when a \ has been encountered. It either returns a
955 positive value for a simple escape such as \n, or 0 for a data character which
956 will be placed in chptr. A backreference to group n is returned as negative n.
957 When UTF-8 is enabled, a positive value greater than 255 may be returned in
958 chptr. On entry, ptr is pointing at the \. On exit, it is on the final
959 character of the escape sequence.
960
961 Arguments:
962 ptrptr points to the pattern position pointer
963 chptr points to a returned data character
964 errorcodeptr points to the errorcode variable
965 bracount number of previous extracting brackets
966 options the options bits
967 isclass TRUE if inside a character class
968
969 Returns: zero => a data character
970 positive => a special escape sequence
971 negative => a back reference
972 on error, errorcodeptr is set
973 */
974
975 static int
976 check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
977 int bracount, int options, BOOL isclass)
978 {
979 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
980 BOOL utf = (options & PCRE_UTF8) != 0;
981 const pcre_uchar *ptr = *ptrptr + 1;
982 pcre_uint32 c;
983 int escape = 0;
984 int i;
985
986 GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
987 ptr--; /* Set pointer back to the last byte */
988
989 /* If backslash is at the end of the pattern, it's an error. */
990
991 if (c == CHAR_NULL) *errorcodeptr = ERR1;
992
993 /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
994 in a table. A non-zero result is something that can be returned immediately.
995 Otherwise further processing may be required. */
996
997 #ifndef EBCDIC /* ASCII/UTF-8 coding */
998 /* Not alphanumeric */
999 else if (c < CHAR_0 || c > CHAR_z) {}
1000 else if ((i = escapes[c - CHAR_0]) != 0)
1001 { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1002
1003 #else /* EBCDIC coding */
1004 /* Not alphanumeric */
1005 else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1006 else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1007 #endif
1008
1009 /* Escapes that need further processing, or are illegal. */
1010
1011 else
1012 {
1013 const pcre_uchar *oldptr;
1014 BOOL braced, negated, overflow;
1015 int s;
1016
1017 switch (c)
1018 {
1019 /* A number of Perl escapes are not handled by PCRE. We give an explicit
1020 error. */
1021
1022 case CHAR_l:
1023 case CHAR_L:
1024 *errorcodeptr = ERR37;
1025 break;
1026
1027 case CHAR_u:
1028 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1029 {
1030 /* In JavaScript, \u must be followed by four hexadecimal numbers.
1031 Otherwise it is a lowercase u letter. */
1032 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1033 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1034 && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1035 && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1036 {
1037 c = 0;
1038 for (i = 0; i < 4; ++i)
1039 {
1040 register pcre_uint32 cc = *(++ptr);
1041 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1042 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1043 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1044 #else /* EBCDIC coding */
1045 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1046 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1047 #endif
1048 }
1049
1050 #if defined COMPILE_PCRE8
1051 if (c > (utf ? 0x10ffffU : 0xffU))
1052 #elif defined COMPILE_PCRE16
1053 if (c > (utf ? 0x10ffffU : 0xffffU))
1054 #elif defined COMPILE_PCRE32
1055 if (utf && c > 0x10ffffU)
1056 #endif
1057 {
1058 *errorcodeptr = ERR76;
1059 }
1060 else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1061 }
1062 }
1063 else
1064 *errorcodeptr = ERR37;
1065 break;
1066
1067 case CHAR_U:
1068 /* In JavaScript, \U is an uppercase U letter. */
1069 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1070 break;
1071
1072 /* In a character class, \g is just a literal "g". Outside a character
1073 class, \g must be followed by one of a number of specific things:
1074
1075 (1) A number, either plain or braced. If positive, it is an absolute
1076 backreference. If negative, it is a relative backreference. This is a Perl
1077 5.10 feature.
1078
1079 (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1080 is part of Perl's movement towards a unified syntax for back references. As
1081 this is synonymous with \k{name}, we fudge it up by pretending it really
1082 was \k.
1083
1084 (3) For Oniguruma compatibility we also support \g followed by a name or a
1085 number either in angle brackets or in single quotes. However, these are
1086 (possibly recursive) subroutine calls, _not_ backreferences. Just return
1087 the ESC_g code (cf \k). */
1088
1089 case CHAR_g:
1090 if (isclass) break;
1091 if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1092 {
1093 escape = ESC_g;
1094 break;
1095 }
1096
1097 /* Handle the Perl-compatible cases */
1098
1099 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1100 {
1101 const pcre_uchar *p;
1102 for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1103 if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1104 if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1105 {
1106 escape = ESC_k;
1107 break;
1108 }
1109 braced = TRUE;
1110 ptr++;
1111 }
1112 else braced = FALSE;
1113
1114 if (ptr[1] == CHAR_MINUS)
1115 {
1116 negated = TRUE;
1117 ptr++;
1118 }
1119 else negated = FALSE;
1120
1121 /* The integer range is limited by the machine's int representation. */
1122 s = 0;
1123 overflow = FALSE;
1124 while (IS_DIGIT(ptr[1]))
1125 {
1126 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1127 {
1128 overflow = TRUE;
1129 break;
1130 }
1131 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1132 }
1133 if (overflow) /* Integer overflow */
1134 {
1135 while (IS_DIGIT(ptr[1]))
1136 ptr++;
1137 *errorcodeptr = ERR61;
1138 break;
1139 }
1140
1141 if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1142 {
1143 *errorcodeptr = ERR57;
1144 break;
1145 }
1146
1147 if (s == 0)
1148 {
1149 *errorcodeptr = ERR58;
1150 break;
1151 }
1152
1153 if (negated)
1154 {
1155 if (s > bracount)
1156 {
1157 *errorcodeptr = ERR15;
1158 break;
1159 }
1160 s = bracount - (s - 1);
1161 }
1162
1163 escape = -s;
1164 break;
1165
1166 /* The handling of escape sequences consisting of a string of digits
1167 starting with one that is not zero is not straightforward. Perl has changed
1168 over the years. Nowadays \g{} for backreferences and \o{} for octal are
1169 recommended to avoid the ambiguities in the old syntax.
1170
1171 Outside a character class, the digits are read as a decimal number. If the
1172 number is less than 8 (used to be 10), or if there are that many previous
1173 extracting left brackets, then it is a back reference. Otherwise, up to
1174 three octal digits are read to form an escaped byte. Thus \123 is likely to
1175 be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1176 the octal value is greater than 377, the least significant 8 bits are
1177 taken. \8 and \9 are treated as the literal characters 8 and 9.
1178
1179 Inside a character class, \ followed by a digit is always either a literal
1180 8 or 9 or an octal number. */
1181
1182 case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1183 case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1184
1185 if (!isclass)
1186 {
1187 oldptr = ptr;
1188 /* The integer range is limited by the machine's int representation. */
1189 s = (int)(c -CHAR_0);
1190 overflow = FALSE;
1191 while (IS_DIGIT(ptr[1]))
1192 {
1193 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1194 {
1195 overflow = TRUE;
1196 break;
1197 }
1198 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1199 }
1200 if (overflow) /* Integer overflow */
1201 {
1202 while (IS_DIGIT(ptr[1]))
1203 ptr++;
1204 *errorcodeptr = ERR61;
1205 break;
1206 }
1207 if (s < 8 || s <= bracount) /* Check for back reference */
1208 {
1209 escape = -s;
1210 break;
1211 }
1212 ptr = oldptr; /* Put the pointer back and fall through */
1213 }
1214
1215 /* Handle a digit following \ when the number is not a back reference. If
1216 the first digit is 8 or 9, Perl used to generate a binary zero byte and
1217 then treat the digit as a following literal. At least by Perl 5.18 this
1218 changed so as not to insert the binary zero. */
1219
1220 if ((c = *ptr) >= CHAR_8) break;
1221
1222 /* Fall through with a digit less than 8 */
1223
1224 /* \0 always starts an octal number, but we may drop through to here with a
1225 larger first octal digit. The original code used just to take the least
1226 significant 8 bits of octal numbers (I think this is what early Perls used
1227 to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1228 but no more than 3 octal digits. */
1229
1230 case CHAR_0:
1231 c -= CHAR_0;
1232 while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1233 c = c * 8 + *(++ptr) - CHAR_0;
1234 #ifdef COMPILE_PCRE8
1235 if (!utf && c > 0xff) *errorcodeptr = ERR51;
1236 #endif
1237 break;
1238
1239 /* \o is a relatively new Perl feature, supporting a more general way of
1240 specifying character codes in octal. The only supported form is \o{ddd}. */
1241
1242 case CHAR_o:
1243 if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1244 {
1245 ptr += 2;
1246 c = 0;
1247 overflow = FALSE;
1248 while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1249 {
1250 register pcre_uint32 cc = *ptr++;
1251 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1252 #ifdef COMPILE_PCRE32
1253 if (c >= 0x20000000l) { overflow = TRUE; break; }
1254 #endif
1255 c = (c << 3) + cc - CHAR_0 ;
1256 #if defined COMPILE_PCRE8
1257 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1258 #elif defined COMPILE_PCRE16
1259 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1260 #elif defined COMPILE_PCRE32
1261 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1262 #endif
1263 }
1264 if (overflow)
1265 {
1266 while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1267 *errorcodeptr = ERR34;
1268 }
1269 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1270 {
1271 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1272 }
1273 else *errorcodeptr = ERR80;
1274 }
1275 break;
1276
1277 /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1278 numbers. Otherwise it is a lowercase x letter. */
1279
1280 case CHAR_x:
1281 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1282 {
1283 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1284 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1285 {
1286 c = 0;
1287 for (i = 0; i < 2; ++i)
1288 {
1289 register pcre_uint32 cc = *(++ptr);
1290 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1291 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1292 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1293 #else /* EBCDIC coding */
1294 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1295 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1296 #endif
1297 }
1298 }
1299 } /* End JavaScript handling */
1300
1301 /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1302 greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1303 digits. If not, { used to be treated as a data character. However, Perl
1304 seems to read hex digits up to the first non-such, and ignore the rest, so
1305 that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1306 now gives an error. */
1307
1308 else
1309 {
1310 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1311 {
1312 ptr += 2;
1313 c = 0;
1314 overflow = FALSE;
1315 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1316 {
1317 register pcre_uint32 cc = *ptr++;
1318 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1319
1320 #ifdef COMPILE_PCRE32
1321 if (c >= 0x10000000l) { overflow = TRUE; break; }
1322 #endif
1323
1324 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1325 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1326 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1327 #else /* EBCDIC coding */
1328 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1329 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1330 #endif
1331
1332 #if defined COMPILE_PCRE8
1333 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1334 #elif defined COMPILE_PCRE16
1335 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1336 #elif defined COMPILE_PCRE32
1337 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1338 #endif
1339 }
1340
1341 if (overflow)
1342 {
1343 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1344 *errorcodeptr = ERR34;
1345 }
1346
1347 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1348 {
1349 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1350 }
1351
1352 /* If the sequence of hex digits does not end with '}', give an error.
1353 We used just to recognize this construct and fall through to the normal
1354 \x handling, but nowadays Perl gives an error, which seems much more
1355 sensible, so we do too. */
1356
1357 else *errorcodeptr = ERR79;
1358 } /* End of \x{} processing */
1359
1360 /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1361
1362 else
1363 {
1364 c = 0;
1365 while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1366 {
1367 pcre_uint32 cc; /* Some compilers don't like */
1368 cc = *(++ptr); /* ++ in initializers */
1369 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1370 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1371 c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1372 #else /* EBCDIC coding */
1373 if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1374 c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1375 #endif
1376 }
1377 } /* End of \xdd handling */
1378 } /* End of Perl-style \x handling */
1379 break;
1380
1381 /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1382 An error is given if the byte following \c is not an ASCII character. This
1383 coding is ASCII-specific, but then the whole concept of \cx is
1384 ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1385
1386 case CHAR_c:
1387 c = *(++ptr);
1388 if (c == CHAR_NULL)
1389 {
1390 *errorcodeptr = ERR2;
1391 break;
1392 }
1393 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1394 if (c > 127) /* Excludes all non-ASCII in either mode */
1395 {
1396 *errorcodeptr = ERR68;
1397 break;
1398 }
1399 if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1400 c ^= 0x40;
1401 #else /* EBCDIC coding */
1402 if (c >= CHAR_a && c <= CHAR_z) c += 64;
1403 c ^= 0xC0;
1404 #endif
1405 break;
1406
1407 /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1408 other alphanumeric following \ is an error if PCRE_EXTRA was set;
1409 otherwise, for Perl compatibility, it is a literal. This code looks a bit
1410 odd, but there used to be some cases other than the default, and there may
1411 be again in future, so I haven't "optimized" it. */
1412
1413 default:
1414 if ((options & PCRE_EXTRA) != 0) switch(c)
1415 {
1416 default:
1417 *errorcodeptr = ERR3;
1418 break;
1419 }
1420 break;
1421 }
1422 }
1423
1424 /* Perl supports \N{name} for character names, as well as plain \N for "not
1425 newline". PCRE does not support \N{name}. However, it does support
1426 quantification such as \N{2,3}. */
1427
1428 if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1429 !is_counted_repeat(ptr+2))
1430 *errorcodeptr = ERR37;
1431
1432 /* If PCRE_UCP is set, we change the values for \d etc. */
1433
1434 if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1435 escape += (ESC_DU - ESC_D);
1436
1437 /* Set the pointer to the final character before returning. */
1438
1439 *ptrptr = ptr;
1440 *chptr = c;
1441 return escape;
1442 }
1443
1444
1445
1446 #ifdef SUPPORT_UCP
1447 /*************************************************
1448 * Handle \P and \p *
1449 *************************************************/
1450
1451 /* This function is called after \P or \p has been encountered, provided that
1452 PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1453 pointing at the P or p. On exit, it is pointing at the final character of the
1454 escape sequence.
1455
1456 Argument:
1457 ptrptr points to the pattern position pointer
1458 negptr points to a boolean that is set TRUE for negation else FALSE
1459 ptypeptr points to an unsigned int that is set to the type value
1460 pdataptr points to an unsigned int that is set to the detailed property value
1461 errorcodeptr points to the error code variable
1462
1463 Returns: TRUE if the type value was found, or FALSE for an invalid type
1464 */
1465
1466 static BOOL
1467 get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1468 unsigned int *pdataptr, int *errorcodeptr)
1469 {
1470 pcre_uchar c;
1471 int i, bot, top;
1472 const pcre_uchar *ptr = *ptrptr;
1473 pcre_uchar name[32];
1474
1475 c = *(++ptr);
1476 if (c == CHAR_NULL) goto ERROR_RETURN;
1477
1478 *negptr = FALSE;
1479
1480 /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1481 negation. */
1482
1483 if (c == CHAR_LEFT_CURLY_BRACKET)
1484 {
1485 if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1486 {
1487 *negptr = TRUE;
1488 ptr++;
1489 }
1490 for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1491 {
1492 c = *(++ptr);
1493 if (c == CHAR_NULL) goto ERROR_RETURN;
1494 if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1495 name[i] = c;
1496 }
1497 if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1498 name[i] = 0;
1499 }
1500
1501 /* Otherwise there is just one following character */
1502
1503 else
1504 {
1505 name[0] = c;
1506 name[1] = 0;
1507 }
1508
1509 *ptrptr = ptr;
1510
1511 /* Search for a recognized property name using binary chop */
1512
1513 bot = 0;
1514 top = PRIV(utt_size);
1515
1516 while (bot < top)
1517 {
1518 int r;
1519 i = (bot + top) >> 1;
1520 r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1521 if (r == 0)
1522 {
1523 *ptypeptr = PRIV(utt)[i].type;
1524 *pdataptr = PRIV(utt)[i].value;
1525 return TRUE;
1526 }
1527 if (r > 0) bot = i + 1; else top = i;
1528 }
1529
1530 *errorcodeptr = ERR47;
1531 *ptrptr = ptr;
1532 return FALSE;
1533
1534 ERROR_RETURN:
1535 *errorcodeptr = ERR46;
1536 *ptrptr = ptr;
1537 return FALSE;
1538 }
1539 #endif
1540
1541
1542
1543 /*************************************************
1544 * Read repeat counts *
1545 *************************************************/
1546
1547 /* Read an item of the form {n,m} and return the values. This is called only
1548 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1549 so the syntax is guaranteed to be correct, but we need to check the values.
1550
1551 Arguments:
1552 p pointer to first char after '{'
1553 minp pointer to int for min
1554 maxp pointer to int for max
1555 returned as -1 if no max
1556 errorcodeptr points to error code variable
1557
1558 Returns: pointer to '}' on success;
1559 current ptr on error, with errorcodeptr set non-zero
1560 */
1561
1562 static const pcre_uchar *
1563 read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1564 {
1565 int min = 0;
1566 int max = -1;
1567
1568 /* Read the minimum value and do a paranoid check: a negative value indicates
1569 an integer overflow. */
1570
1571 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1572 if (min < 0 || min > 65535)
1573 {
1574 *errorcodeptr = ERR5;
1575 return p;
1576 }
1577
1578 /* Read the maximum value if there is one, and again do a paranoid on its size.
1579 Also, max must not be less than min. */
1580
1581 if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1582 {
1583 if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1584 {
1585 max = 0;
1586 while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1587 if (max < 0 || max > 65535)
1588 {
1589 *errorcodeptr = ERR5;
1590 return p;
1591 }
1592 if (max < min)
1593 {
1594 *errorcodeptr = ERR4;
1595 return p;
1596 }
1597 }
1598 }
1599
1600 /* Fill in the required variables, and pass back the pointer to the terminating
1601 '}'. */
1602
1603 *minp = min;
1604 *maxp = max;
1605 return p;
1606 }
1607
1608
1609
1610 /*************************************************
1611 * Find first significant op code *
1612 *************************************************/
1613
1614 /* This is called by several functions that scan a compiled expression looking
1615 for a fixed first character, or an anchoring op code etc. It skips over things
1616 that do not influence this. For some calls, it makes sense to skip negative
1617 forward and all backward assertions, and also the \b assertion; for others it
1618 does not.
1619
1620 Arguments:
1621 code pointer to the start of the group
1622 skipassert TRUE if certain assertions are to be skipped
1623
1624 Returns: pointer to the first significant opcode
1625 */
1626
1627 static const pcre_uchar*
1628 first_significant_code(const pcre_uchar *code, BOOL skipassert)
1629 {
1630 for (;;)
1631 {
1632 switch ((int)*code)
1633 {
1634 case OP_ASSERT_NOT:
1635 case OP_ASSERTBACK:
1636 case OP_ASSERTBACK_NOT:
1637 if (!skipassert) return code;
1638 do code += GET(code, 1); while (*code == OP_ALT);
1639 code += PRIV(OP_lengths)[*code];
1640 break;
1641
1642 case OP_WORD_BOUNDARY:
1643 case OP_NOT_WORD_BOUNDARY:
1644 if (!skipassert) return code;
1645 /* Fall through */
1646
1647 case OP_CALLOUT:
1648 case OP_CREF:
1649 case OP_DNCREF:
1650 case OP_RREF:
1651 case OP_DNRREF:
1652 case OP_DEF:
1653 code += PRIV(OP_lengths)[*code];
1654 break;
1655
1656 default:
1657 return code;
1658 }
1659 }
1660 /* Control never reaches here */
1661 }
1662
1663
1664
1665 /*************************************************
1666 * Find the fixed length of a branch *
1667 *************************************************/
1668
1669 /* Scan a branch and compute the fixed length of subject that will match it,
1670 if the length is fixed. This is needed for dealing with backward assertions.
1671 In UTF8 mode, the result is in characters rather than bytes. The branch is
1672 temporarily terminated with OP_END when this function is called.
1673
1674 This function is called when a backward assertion is encountered, so that if it
1675 fails, the error message can point to the correct place in the pattern.
1676 However, we cannot do this when the assertion contains subroutine calls,
1677 because they can be forward references. We solve this by remembering this case
1678 and doing the check at the end; a flag specifies which mode we are running in.
1679
1680 Arguments:
1681 code points to the start of the pattern (the bracket)
1682 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1683 atend TRUE if called when the pattern is complete
1684 cd the "compile data" structure
1685
1686 Returns: the fixed length,
1687 or -1 if there is no fixed length,
1688 or -2 if \C was encountered (in UTF-8 mode only)
1689 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1690 or -4 if an unknown opcode was encountered (internal error)
1691 */
1692
1693 static int
1694 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1695 {
1696 int length = -1;
1697
1698 register int branchlength = 0;
1699 register pcre_uchar *cc = code + 1 + LINK_SIZE;
1700
1701 /* Scan along the opcodes for this branch. If we get to the end of the
1702 branch, check the length against that of the other branches. */
1703
1704 for (;;)
1705 {
1706 int d;
1707 pcre_uchar *ce, *cs;
1708 register pcre_uchar op = *cc;
1709
1710 switch (op)
1711 {
1712 /* We only need to continue for OP_CBRA (normal capturing bracket) and
1713 OP_BRA (normal non-capturing bracket) because the other variants of these
1714 opcodes are all concerned with unlimited repeated groups, which of course
1715 are not of fixed length. */
1716
1717 case OP_CBRA:
1718 case OP_BRA:
1719 case OP_ONCE:
1720 case OP_ONCE_NC:
1721 case OP_COND:
1722 d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1723 if (d < 0) return d;
1724 branchlength += d;
1725 do cc += GET(cc, 1); while (*cc == OP_ALT);
1726 cc += 1 + LINK_SIZE;
1727 break;
1728
1729 /* Reached end of a branch; if it's a ket it is the end of a nested call.
1730 If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1731 an ALT. If it is END it's the end of the outer call. All can be handled by
1732 the same code. Note that we must not include the OP_KETRxxx opcodes here,
1733 because they all imply an unlimited repeat. */
1734
1735 case OP_ALT:
1736 case OP_KET:
1737 case OP_END:
1738 case OP_ACCEPT:
1739 case OP_ASSERT_ACCEPT:
1740 if (length < 0) length = branchlength;
1741 else if (length != branchlength) return -1;
1742 if (*cc != OP_ALT) return length;
1743 cc += 1 + LINK_SIZE;
1744 branchlength = 0;
1745 break;
1746
1747 /* A true recursion implies not fixed length, but a subroutine call may
1748 be OK. If the subroutine is a forward reference, we can't deal with
1749 it until the end of the pattern, so return -3. */
1750
1751 case OP_RECURSE:
1752 if (!atend) return -3;
1753 cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1754 do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1755 if (cc > cs && cc < ce) return -1; /* Recursion */
1756 d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1757 if (d < 0) return d;
1758 branchlength += d;
1759 cc += 1 + LINK_SIZE;
1760 break;
1761
1762 /* Skip over assertive subpatterns */
1763
1764 case OP_ASSERT:
1765 case OP_ASSERT_NOT:
1766 case OP_ASSERTBACK:
1767 case OP_ASSERTBACK_NOT:
1768 do cc += GET(cc, 1); while (*cc == OP_ALT);
1769 cc += PRIV(OP_lengths)[*cc];
1770 break;
1771
1772 /* Skip over things that don't match chars */
1773
1774 case OP_MARK:
1775 case OP_PRUNE_ARG:
1776 case OP_SKIP_ARG:
1777 case OP_THEN_ARG:
1778 cc += cc[1] + PRIV(OP_lengths)[*cc];
1779 break;
1780
1781 case OP_CALLOUT:
1782 case OP_CIRC:
1783 case OP_CIRCM:
1784 case OP_CLOSE:
1785 case OP_COMMIT:
1786 case OP_CREF:
1787 case OP_DEF:
1788 case OP_DNCREF:
1789 case OP_DNRREF:
1790 case OP_DOLL:
1791 case OP_DOLLM:
1792 case OP_EOD:
1793 case OP_EODN:
1794 case OP_FAIL:
1795 case OP_NOT_WORD_BOUNDARY:
1796 case OP_PRUNE:
1797 case OP_REVERSE:
1798 case OP_RREF:
1799 case OP_SET_SOM:
1800 case OP_SKIP:
1801 case OP_SOD:
1802 case OP_SOM:
1803 case OP_THEN:
1804 case OP_WORD_BOUNDARY:
1805 cc += PRIV(OP_lengths)[*cc];
1806 break;
1807
1808 /* Handle literal characters */
1809
1810 case OP_CHAR:
1811 case OP_CHARI:
1812 case OP_NOT:
1813 case OP_NOTI:
1814 branchlength++;
1815 cc += 2;
1816 #ifdef SUPPORT_UTF
1817 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1818 #endif
1819 break;
1820
1821 /* Handle exact repetitions. The count is already in characters, but we
1822 need to skip over a multibyte character in UTF8 mode. */
1823
1824 case OP_EXACT:
1825 case OP_EXACTI:
1826 case OP_NOTEXACT:
1827 case OP_NOTEXACTI:
1828 branchlength += (int)GET2(cc,1);
1829 cc += 2 + IMM2_SIZE;
1830 #ifdef SUPPORT_UTF
1831 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1832 #endif
1833 break;
1834
1835 case OP_TYPEEXACT:
1836 branchlength += GET2(cc,1);
1837 if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1838 cc += 2;
1839 cc += 1 + IMM2_SIZE + 1;
1840 break;
1841
1842 /* Handle single-char matchers */
1843
1844 case OP_PROP:
1845 case OP_NOTPROP:
1846 cc += 2;
1847 /* Fall through */
1848
1849 case OP_HSPACE:
1850 case OP_VSPACE:
1851 case OP_NOT_HSPACE:
1852 case OP_NOT_VSPACE:
1853 case OP_NOT_DIGIT:
1854 case OP_DIGIT:
1855 case OP_NOT_WHITESPACE:
1856 case OP_WHITESPACE:
1857 case OP_NOT_WORDCHAR:
1858 case OP_WORDCHAR:
1859 case OP_ANY:
1860 case OP_ALLANY:
1861 branchlength++;
1862 cc++;
1863 break;
1864
1865 /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1866 otherwise \C is coded as OP_ALLANY. */
1867
1868 case OP_ANYBYTE:
1869 return -2;
1870
1871 /* Check a class for variable quantification */
1872
1873 case OP_CLASS:
1874 case OP_NCLASS:
1875 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1876 case OP_XCLASS:
1877 /* The original code caused an unsigned overflow in 64 bit systems,
1878 so now we use a conditional statement. */
1879 if (op == OP_XCLASS)
1880 cc += GET(cc, 1);
1881 else
1882 cc += PRIV(OP_lengths)[OP_CLASS];
1883 #else
1884 cc += PRIV(OP_lengths)[OP_CLASS];
1885 #endif
1886
1887 switch (*cc)
1888 {
1889 case OP_CRSTAR:
1890 case OP_CRMINSTAR:
1891 case OP_CRPLUS:
1892 case OP_CRMINPLUS:
1893 case OP_CRQUERY:
1894 case OP_CRMINQUERY:
1895 case OP_CRPOSSTAR:
1896 case OP_CRPOSPLUS:
1897 case OP_CRPOSQUERY:
1898 return -1;
1899
1900 case OP_CRRANGE:
1901 case OP_CRMINRANGE:
1902 case OP_CRPOSRANGE:
1903 if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1904 branchlength += (int)GET2(cc,1);
1905 cc += 1 + 2 * IMM2_SIZE;
1906 break;
1907
1908 default:
1909 branchlength++;
1910 }
1911 break;
1912
1913 /* Anything else is variable length */
1914
1915 case OP_ANYNL:
1916 case OP_BRAMINZERO:
1917 case OP_BRAPOS:
1918 case OP_BRAPOSZERO:
1919 case OP_BRAZERO:
1920 case OP_CBRAPOS:
1921 case OP_EXTUNI:
1922 case OP_KETRMAX:
1923 case OP_KETRMIN:
1924 case OP_KETRPOS:
1925 case OP_MINPLUS:
1926 case OP_MINPLUSI:
1927 case OP_MINQUERY:
1928 case OP_MINQUERYI:
1929 case OP_MINSTAR:
1930 case OP_MINSTARI:
1931 case OP_MINUPTO:
1932 case OP_MINUPTOI:
1933 case OP_NOTMINPLUS:
1934 case OP_NOTMINPLUSI:
1935 case OP_NOTMINQUERY:
1936 case OP_NOTMINQUERYI:
1937 case OP_NOTMINSTAR:
1938 case OP_NOTMINSTARI:
1939 case OP_NOTMINUPTO:
1940 case OP_NOTMINUPTOI:
1941 case OP_NOTPLUS:
1942 case OP_NOTPLUSI:
1943 case OP_NOTPOSPLUS:
1944 case OP_NOTPOSPLUSI:
1945 case OP_NOTPOSQUERY:
1946 case OP_NOTPOSQUERYI:
1947 case OP_NOTPOSSTAR:
1948 case OP_NOTPOSSTARI:
1949 case OP_NOTPOSUPTO:
1950 case OP_NOTPOSUPTOI:
1951 case OP_NOTQUERY:
1952 case OP_NOTQUERYI:
1953 case OP_NOTSTAR:
1954 case OP_NOTSTARI:
1955 case OP_NOTUPTO:
1956 case OP_NOTUPTOI:
1957 case OP_PLUS:
1958 case OP_PLUSI:
1959 case OP_POSPLUS:
1960 case OP_POSPLUSI:
1961 case OP_POSQUERY:
1962 case OP_POSQUERYI:
1963 case OP_POSSTAR:
1964 case OP_POSSTARI:
1965 case OP_POSUPTO:
1966 case OP_POSUPTOI:
1967 case OP_QUERY:
1968 case OP_QUERYI:
1969 case OP_REF:
1970 case OP_REFI:
1971 case OP_DNREF:
1972 case OP_DNREFI:
1973 case OP_SBRA:
1974 case OP_SBRAPOS:
1975 case OP_SCBRA:
1976 case OP_SCBRAPOS:
1977 case OP_SCOND:
1978 case OP_SKIPZERO:
1979 case OP_STAR:
1980 case OP_STARI:
1981 case OP_TYPEMINPLUS:
1982 case OP_TYPEMINQUERY:
1983 case OP_TYPEMINSTAR:
1984 case OP_TYPEMINUPTO:
1985 case OP_TYPEPLUS:
1986 case OP_TYPEPOSPLUS:
1987 case OP_TYPEPOSQUERY:
1988 case OP_TYPEPOSSTAR:
1989 case OP_TYPEPOSUPTO:
1990 case OP_TYPEQUERY:
1991 case OP_TYPESTAR:
1992 case OP_TYPEUPTO:
1993 case OP_UPTO:
1994 case OP_UPTOI:
1995 return -1;
1996
1997 /* Catch unrecognized opcodes so that when new ones are added they
1998 are not forgotten, as has happened in the past. */
1999
2000 default:
2001 return -4;
2002 }
2003 }
2004 /* Control never gets here */
2005 }
2006
2007
2008
2009 /*************************************************
2010 * Scan compiled regex for specific bracket *
2011 *************************************************/
2012
2013 /* This little function scans through a compiled pattern until it finds a
2014 capturing bracket with the given number, or, if the number is negative, an
2015 instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2016 so that it can be called from pcre_study() when finding the minimum matching
2017 length.
2018
2019 Arguments:
2020 code points to start of expression
2021 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2022 number the required bracket number or negative to find a lookbehind
2023
2024 Returns: pointer to the opcode for the bracket, or NULL if not found
2025 */
2026
2027 const pcre_uchar *
2028 PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2029 {
2030 for (;;)
2031 {
2032 register pcre_uchar c = *code;
2033
2034 if (c == OP_END) return NULL;
2035
2036 /* XCLASS is used for classes that cannot be represented just by a bit
2037 map. This includes negated single high-valued characters. The length in
2038 the table is zero; the actual length is stored in the compiled code. */
2039
2040 if (c == OP_XCLASS) code += GET(code, 1);
2041
2042 /* Handle recursion */
2043
2044 else if (c == OP_REVERSE)
2045 {
2046 if (number < 0) return (pcre_uchar *)code;
2047 code += PRIV(OP_lengths)[c];
2048 }
2049
2050 /* Handle capturing bracket */
2051
2052 else if (c == OP_CBRA || c == OP_SCBRA ||
2053 c == OP_CBRAPOS || c == OP_SCBRAPOS)
2054 {
2055 int n = (int)GET2(code, 1+LINK_SIZE);
2056 if (n == number) return (pcre_uchar *)code;
2057 code += PRIV(OP_lengths)[c];
2058 }
2059
2060 /* Otherwise, we can get the item's length from the table, except that for
2061 repeated character types, we have to test for \p and \P, which have an extra
2062 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2063 must add in its length. */
2064
2065 else
2066 {
2067 switch(c)
2068 {
2069 case OP_TYPESTAR:
2070 case OP_TYPEMINSTAR:
2071 case OP_TYPEPLUS:
2072 case OP_TYPEMINPLUS:
2073 case OP_TYPEQUERY:
2074 case OP_TYPEMINQUERY:
2075 case OP_TYPEPOSSTAR:
2076 case OP_TYPEPOSPLUS:
2077 case OP_TYPEPOSQUERY:
2078 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2079 break;
2080
2081 case OP_TYPEUPTO:
2082 case OP_TYPEMINUPTO:
2083 case OP_TYPEEXACT:
2084 case OP_TYPEPOSUPTO:
2085 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2086 code += 2;
2087 break;
2088
2089 case OP_MARK:
2090 case OP_PRUNE_ARG:
2091 case OP_SKIP_ARG:
2092 case OP_THEN_ARG:
2093 code += code[1];
2094 break;
2095 }
2096
2097 /* Add in the fixed length from the table */
2098
2099 code += PRIV(OP_lengths)[c];
2100
2101 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2102 a multi-byte character. The length in the table is a minimum, so we have to
2103 arrange to skip the extra bytes. */
2104
2105 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2106 if (utf) switch(c)
2107 {
2108 case OP_CHAR:
2109 case OP_CHARI:
2110 case OP_EXACT:
2111 case OP_EXACTI:
2112 case OP_UPTO:
2113 case OP_UPTOI:
2114 case OP_MINUPTO:
2115 case OP_MINUPTOI:
2116 case OP_POSUPTO:
2117 case OP_POSUPTOI:
2118 case OP_STAR:
2119 case OP_STARI:
2120 case OP_MINSTAR:
2121 case OP_MINSTARI:
2122 case OP_POSSTAR:
2123 case OP_POSSTARI:
2124 case OP_PLUS:
2125 case OP_PLUSI:
2126 case OP_MINPLUS:
2127 case OP_MINPLUSI:
2128 case OP_POSPLUS:
2129 case OP_POSPLUSI:
2130 case OP_QUERY:
2131 case OP_QUERYI:
2132 case OP_MINQUERY:
2133 case OP_MINQUERYI:
2134 case OP_POSQUERY:
2135 case OP_POSQUERYI:
2136 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2137 break;
2138 }
2139 #else
2140 (void)(utf); /* Keep compiler happy by referencing function argument */
2141 #endif
2142 }
2143 }
2144 }
2145
2146
2147
2148 /*************************************************
2149 * Scan compiled regex for recursion reference *
2150 *************************************************/
2151
2152 /* This little function scans through a compiled pattern until it finds an
2153 instance of OP_RECURSE.
2154
2155 Arguments:
2156 code points to start of expression
2157 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2158
2159 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2160 */
2161
2162 static const pcre_uchar *
2163 find_recurse(const pcre_uchar *code, BOOL utf)
2164 {
2165 for (;;)
2166 {
2167 register pcre_uchar c = *code;
2168 if (c == OP_END) return NULL;
2169 if (c == OP_RECURSE) return code;
2170
2171 /* XCLASS is used for classes that cannot be represented just by a bit
2172 map. This includes negated single high-valued characters. The length in
2173 the table is zero; the actual length is stored in the compiled code. */
2174
2175 if (c == OP_XCLASS) code += GET(code, 1);
2176
2177 /* Otherwise, we can get the item's length from the table, except that for
2178 repeated character types, we have to test for \p and \P, which have an extra
2179 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2180 must add in its length. */
2181
2182 else
2183 {
2184 switch(c)
2185 {
2186 case OP_TYPESTAR:
2187 case OP_TYPEMINSTAR:
2188 case OP_TYPEPLUS:
2189 case OP_TYPEMINPLUS:
2190 case OP_TYPEQUERY:
2191 case OP_TYPEMINQUERY:
2192 case OP_TYPEPOSSTAR:
2193 case OP_TYPEPOSPLUS:
2194 case OP_TYPEPOSQUERY:
2195 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2196 break;
2197
2198 case OP_TYPEPOSUPTO:
2199 case OP_TYPEUPTO:
2200 case OP_TYPEMINUPTO:
2201 case OP_TYPEEXACT:
2202 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2203 code += 2;
2204 break;
2205
2206 case OP_MARK:
2207 case OP_PRUNE_ARG:
2208 case OP_SKIP_ARG:
2209 case OP_THEN_ARG:
2210 code += code[1];
2211 break;
2212 }
2213
2214 /* Add in the fixed length from the table */
2215
2216 code += PRIV(OP_lengths)[c];
2217
2218 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2219 by a multi-byte character. The length in the table is a minimum, so we have
2220 to arrange to skip the extra bytes. */
2221
2222 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2223 if (utf) switch(c)
2224 {
2225 case OP_CHAR:
2226 case OP_CHARI:
2227 case OP_NOT:
2228 case OP_NOTI:
2229 case OP_EXACT:
2230 case OP_EXACTI:
2231 case OP_NOTEXACT:
2232 case OP_NOTEXACTI:
2233 case OP_UPTO:
2234 case OP_UPTOI:
2235 case OP_NOTUPTO:
2236 case OP_NOTUPTOI:
2237 case OP_MINUPTO:
2238 case OP_MINUPTOI:
2239 case OP_NOTMINUPTO:
2240 case OP_NOTMINUPTOI:
2241 case OP_POSUPTO:
2242 case OP_POSUPTOI:
2243 case OP_NOTPOSUPTO:
2244 case OP_NOTPOSUPTOI:
2245 case OP_STAR:
2246 case OP_STARI:
2247 case OP_NOTSTAR:
2248 case OP_NOTSTARI:
2249 case OP_MINSTAR:
2250 case OP_MINSTARI:
2251 case OP_NOTMINSTAR:
2252 case OP_NOTMINSTARI:
2253 case OP_POSSTAR:
2254 case OP_POSSTARI:
2255 case OP_NOTPOSSTAR:
2256 case OP_NOTPOSSTARI:
2257 case OP_PLUS:
2258 case OP_PLUSI:
2259 case OP_NOTPLUS:
2260 case OP_NOTPLUSI:
2261 case OP_MINPLUS:
2262 case OP_MINPLUSI:
2263 case OP_NOTMINPLUS:
2264 case OP_NOTMINPLUSI:
2265 case OP_POSPLUS:
2266 case OP_POSPLUSI:
2267 case OP_NOTPOSPLUS:
2268 case OP_NOTPOSPLUSI:
2269 case OP_QUERY:
2270 case OP_QUERYI:
2271 case OP_NOTQUERY:
2272 case OP_NOTQUERYI:
2273 case OP_MINQUERY:
2274 case OP_MINQUERYI:
2275 case OP_NOTMINQUERY:
2276 case OP_NOTMINQUERYI:
2277 case OP_POSQUERY:
2278 case OP_POSQUERYI:
2279 case OP_NOTPOSQUERY:
2280 case OP_NOTPOSQUERYI:
2281 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2282 break;
2283 }
2284 #else
2285 (void)(utf); /* Keep compiler happy by referencing function argument */
2286 #endif
2287 }
2288 }
2289 }
2290
2291
2292
2293 /*************************************************
2294 * Scan compiled branch for non-emptiness *
2295 *************************************************/
2296
2297 /* This function scans through a branch of a compiled pattern to see whether it
2298 can match the empty string or not. It is called from could_be_empty()
2299 below and from compile_branch() when checking for an unlimited repeat of a
2300 group that can match nothing. Note that first_significant_code() skips over
2301 backward and negative forward assertions when its final argument is TRUE. If we
2302 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2303 bracket whose current branch will already have been scanned.
2304
2305 Arguments:
2306 code points to start of search
2307 endcode points to where to stop
2308 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2309 cd contains pointers to tables etc.
2310 recurses chain of recurse_check to catch mutual recursion
2311
2312 Returns: TRUE if what is matched could be empty
2313 */
2314
2315 typedef struct recurse_check {
2316 struct recurse_check *prev;
2317 const pcre_uchar *group;
2318 } recurse_check;
2319
2320 static BOOL
2321 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2322 BOOL utf, compile_data *cd, recurse_check *recurses)
2323 {
2324 register pcre_uchar c;
2325 recurse_check this_recurse;
2326
2327 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2328 code < endcode;
2329 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2330 {
2331 const pcre_uchar *ccode;
2332
2333 c = *code;
2334
2335 /* Skip over forward assertions; the other assertions are skipped by
2336 first_significant_code() with a TRUE final argument. */
2337
2338 if (c == OP_ASSERT)
2339 {
2340 do code += GET(code, 1); while (*code == OP_ALT);
2341 c = *code;
2342 continue;
2343 }
2344
2345 /* For a recursion/subroutine call, if its end has been reached, which
2346 implies a backward reference subroutine call, we can scan it. If it's a
2347 forward reference subroutine call, we can't. To detect forward reference
2348 we have to scan up the list that is kept in the workspace. This function is
2349 called only when doing the real compile, not during the pre-compile that
2350 measures the size of the compiled pattern. */
2351
2352 if (c == OP_RECURSE)
2353 {
2354 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2355 BOOL empty_branch;
2356
2357 /* Test for forward reference or uncompleted reference. This is disabled
2358 when called to scan a completed pattern by setting cd->start_workspace to
2359 NULL. */
2360
2361 if (cd->start_workspace != NULL)
2362 {
2363 const pcre_uchar *tcode;
2364 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2365 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2366 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2367 }
2368
2369 /* If we are scanning a completed pattern, there are no forward references
2370 and all groups are complete. We need to detect whether this is a recursive
2371 call, as otherwise there will be an infinite loop. If it is a recursion,
2372 just skip over it. Simple recursions are easily detected. For mutual
2373 recursions we keep a chain on the stack. */
2374
2375 else
2376 {
2377 recurse_check *r = recurses;
2378 const pcre_uchar *endgroup = scode;
2379
2380 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2381 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2382
2383 for (r = recurses; r != NULL; r = r->prev)
2384 if (r->group == scode) break;
2385 if (r != NULL) continue; /* Mutual recursion */
2386 }
2387
2388 /* Completed reference; scan the referenced group, remembering it on the
2389 stack chain to detect mutual recursions. */
2390
2391 empty_branch = FALSE;
2392 this_recurse.prev = recurses;
2393 this_recurse.group = scode;
2394
2395 do
2396 {
2397 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2398 {
2399 empty_branch = TRUE;
2400 break;
2401 }
2402 scode += GET(scode, 1);
2403 }
2404 while (*scode == OP_ALT);
2405
2406 if (!empty_branch) return FALSE; /* All branches are non-empty */
2407 continue;
2408 }
2409
2410 /* Groups with zero repeats can of course be empty; skip them. */
2411
2412 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2413 c == OP_BRAPOSZERO)
2414 {
2415 code += PRIV(OP_lengths)[c];
2416 do code += GET(code, 1); while (*code == OP_ALT);
2417 c = *code;
2418 continue;
2419 }
2420
2421 /* A nested group that is already marked as "could be empty" can just be
2422 skipped. */
2423
2424 if (c == OP_SBRA || c == OP_SBRAPOS ||
2425 c == OP_SCBRA || c == OP_SCBRAPOS)
2426 {
2427 do code += GET(code, 1); while (*code == OP_ALT);
2428 c = *code;
2429 continue;
2430 }
2431
2432 /* For other groups, scan the branches. */
2433
2434 if (c == OP_BRA || c == OP_BRAPOS ||
2435 c == OP_CBRA || c == OP_CBRAPOS ||
2436 c == OP_ONCE || c == OP_ONCE_NC ||
2437 c == OP_COND)
2438 {
2439 BOOL empty_branch;
2440 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2441
2442 /* If a conditional group has only one branch, there is a second, implied,
2443 empty branch, so just skip over the conditional, because it could be empty.
2444 Otherwise, scan the individual branches of the group. */
2445
2446 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2447 code += GET(code, 1);
2448 else
2449 {
2450 empty_branch = FALSE;
2451 do
2452 {
2453 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2454 empty_branch = TRUE;
2455 code += GET(code, 1);
2456 }
2457 while (*code == OP_ALT);
2458 if (!empty_branch) return FALSE; /* All branches are non-empty */
2459 }
2460
2461 c = *code;
2462 continue;
2463 }
2464
2465 /* Handle the other opcodes */
2466
2467 switch (c)
2468 {
2469 /* Check for quantifiers after a class. XCLASS is used for classes that
2470 cannot be represented just by a bit map. This includes negated single
2471 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2472 actual length is stored in the compiled code, so we must update "code"
2473 here. */
2474
2475 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2476 case OP_XCLASS:
2477 ccode = code += GET(code, 1);
2478 goto CHECK_CLASS_REPEAT;
2479 #endif
2480
2481 case OP_CLASS:
2482 case OP_NCLASS:
2483 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2484
2485 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2486 CHECK_CLASS_REPEAT:
2487 #endif
2488
2489 switch (*ccode)
2490 {
2491 case OP_CRSTAR: /* These could be empty; continue */
2492 case OP_CRMINSTAR:
2493 case OP_CRQUERY:
2494 case OP_CRMINQUERY:
2495 case OP_CRPOSSTAR:
2496 case OP_CRPOSQUERY:
2497 break;
2498
2499 default: /* Non-repeat => class must match */
2500 case OP_CRPLUS: /* These repeats aren't empty */
2501 case OP_CRMINPLUS:
2502 case OP_CRPOSPLUS:
2503 return FALSE;
2504
2505 case OP_CRRANGE:
2506 case OP_CRMINRANGE:
2507 case OP_CRPOSRANGE:
2508 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2509 break;
2510 }
2511 break;
2512
2513 /* Opcodes that must match a character */
2514
2515 case OP_ANY:
2516 case OP_ALLANY:
2517 case OP_ANYBYTE:
2518
2519 case OP_PROP:
2520 case OP_NOTPROP:
2521 case OP_ANYNL:
2522
2523 case OP_NOT_HSPACE:
2524 case OP_HSPACE:
2525 case OP_NOT_VSPACE:
2526 case OP_VSPACE:
2527 case OP_EXTUNI:
2528
2529 case OP_NOT_DIGIT:
2530 case OP_DIGIT:
2531 case OP_NOT_WHITESPACE:
2532 case OP_WHITESPACE:
2533 case OP_NOT_WORDCHAR:
2534 case OP_WORDCHAR:
2535
2536 case OP_CHAR:
2537 case OP_CHARI:
2538 case OP_NOT:
2539 case OP_NOTI:
2540
2541 case OP_PLUS:
2542 case OP_PLUSI:
2543 case OP_MINPLUS:
2544 case OP_MINPLUSI:
2545
2546 case OP_NOTPLUS:
2547 case OP_NOTPLUSI:
2548 case OP_NOTMINPLUS:
2549 case OP_NOTMINPLUSI:
2550
2551 case OP_POSPLUS:
2552 case OP_POSPLUSI:
2553 case OP_NOTPOSPLUS:
2554 case OP_NOTPOSPLUSI:
2555
2556 case OP_EXACT:
2557 case OP_EXACTI:
2558 case OP_NOTEXACT:
2559 case OP_NOTEXACTI:
2560
2561 case OP_TYPEPLUS:
2562 case OP_TYPEMINPLUS:
2563 case OP_TYPEPOSPLUS:
2564 case OP_TYPEEXACT:
2565
2566 return FALSE;
2567
2568 /* These are going to continue, as they may be empty, but we have to
2569 fudge the length for the \p and \P cases. */
2570
2571 case OP_TYPESTAR:
2572 case OP_TYPEMINSTAR:
2573 case OP_TYPEPOSSTAR:
2574 case OP_TYPEQUERY:
2575 case OP_TYPEMINQUERY:
2576 case OP_TYPEPOSQUERY:
2577 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2578 break;
2579
2580 /* Same for these */
2581
2582 case OP_TYPEUPTO:
2583 case OP_TYPEMINUPTO:
2584 case OP_TYPEPOSUPTO:
2585 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2586 code += 2;
2587 break;
2588
2589 /* End of branch */
2590
2591 case OP_KET:
2592 case OP_KETRMAX:
2593 case OP_KETRMIN:
2594 case OP_KETRPOS:
2595 case OP_ALT:
2596 return TRUE;
2597
2598 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2599 MINUPTO, and POSUPTO and their caseless and negative versions may be
2600 followed by a multibyte character. */
2601
2602 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2603 case OP_STAR:
2604 case OP_STARI:
2605 case OP_NOTSTAR:
2606 case OP_NOTSTARI:
2607
2608 case OP_MINSTAR:
2609 case OP_MINSTARI:
2610 case OP_NOTMINSTAR:
2611 case OP_NOTMINSTARI:
2612
2613 case OP_POSSTAR:
2614 case OP_POSSTARI:
2615 case OP_NOTPOSSTAR:
2616 case OP_NOTPOSSTARI:
2617
2618 case OP_QUERY:
2619 case OP_QUERYI:
2620 case OP_NOTQUERY:
2621 case OP_NOTQUERYI:
2622
2623 case OP_MINQUERY:
2624 case OP_MINQUERYI:
2625 case OP_NOTMINQUERY:
2626 case OP_NOTMINQUERYI:
2627
2628 case OP_POSQUERY:
2629 case OP_POSQUERYI:
2630 case OP_NOTPOSQUERY:
2631 case OP_NOTPOSQUERYI:
2632
2633 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2634 break;
2635
2636 case OP_UPTO:
2637 case OP_UPTOI:
2638 case OP_NOTUPTO:
2639 case OP_NOTUPTOI:
2640
2641 case OP_MINUPTO:
2642 case OP_MINUPTOI:
2643 case OP_NOTMINUPTO:
2644 case OP_NOTMINUPTOI:
2645
2646 case OP_POSUPTO:
2647 case OP_POSUPTOI:
2648 case OP_NOTPOSUPTO:
2649 case OP_NOTPOSUPTOI:
2650
2651 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2652 break;
2653 #endif
2654
2655 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2656 string. */
2657
2658 case OP_MARK:
2659 case OP_PRUNE_ARG:
2660 case OP_SKIP_ARG:
2661 case OP_THEN_ARG:
2662 code += code[1];
2663 break;
2664
2665 /* None of the remaining opcodes are required to match a character. */
2666
2667 default:
2668 break;
2669 }
2670 }
2671
2672 return TRUE;
2673 }
2674
2675
2676
2677 /*************************************************
2678 * Scan compiled regex for non-emptiness *
2679 *************************************************/
2680
2681 /* This function is called to check for left recursive calls. We want to check
2682 the current branch of the current pattern to see if it could match the empty
2683 string. If it could, we must look outwards for branches at other levels,
2684 stopping when we pass beyond the bracket which is the subject of the recursion.
2685 This function is called only during the real compile, not during the
2686 pre-compile.
2687
2688 Arguments:
2689 code points to start of the recursion
2690 endcode points to where to stop (current RECURSE item)
2691 bcptr points to the chain of current (unclosed) branch starts
2692 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2693 cd pointers to tables etc
2694
2695 Returns: TRUE if what is matched could be empty
2696 */
2697
2698 static BOOL
2699 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2700 branch_chain *bcptr, BOOL utf, compile_data *cd)
2701 {
2702 while (bcptr != NULL && bcptr->current_branch >= code)
2703 {
2704 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2705 return FALSE;
2706 bcptr = bcptr->outer;
2707 }
2708 return TRUE;
2709 }
2710
2711
2712
2713 /*************************************************
2714 * Base opcode of repeated opcodes *
2715 *************************************************/
2716
2717 /* Returns the base opcode for repeated single character type opcodes. If the
2718 opcode is not a repeated character type, it returns with the original value.
2719
2720 Arguments: c opcode
2721 Returns: base opcode for the type
2722 */
2723
2724 static pcre_uchar
2725 get_repeat_base(pcre_uchar c)
2726 {
2727 return (c > OP_TYPEPOSUPTO)? c :
2728 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2729 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2730 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2731 (c >= OP_STARI)? OP_STARI :
2732 OP_STAR;
2733 }
2734
2735
2736
2737 #ifdef SUPPORT_UCP
2738 /*************************************************
2739 * Check a character and a property *
2740 *************************************************/
2741
2742 /* This function is called by check_auto_possessive() when a property item
2743 is adjacent to a fixed character.
2744
2745 Arguments:
2746 c the character
2747 ptype the property type
2748 pdata the data for the type
2749 negated TRUE if it's a negated property (\P or \p{^)
2750
2751 Returns: TRUE if auto-possessifying is OK
2752 */
2753
2754 static BOOL
2755 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2756 BOOL negated)
2757 {
2758 const pcre_uint32 *p;
2759 const ucd_record *prop = GET_UCD(c);
2760
2761 switch(ptype)
2762 {
2763 case PT_LAMP:
2764 return (prop->chartype == ucp_Lu ||
2765 prop->chartype == ucp_Ll ||
2766 prop->chartype == ucp_Lt) == negated;
2767
2768 case PT_GC:
2769 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2770
2771 case PT_PC:
2772 return (pdata == prop->chartype) == negated;
2773
2774 case PT_SC:
2775 return (pdata == prop->script) == negated;
2776
2777 /* These are specials */
2778
2779 case PT_ALNUM:
2780 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2781 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2782
2783 /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2784 means that Perl space and POSIX space are now identical. PCRE was changed
2785 at release 8.34. */
2786
2787 case PT_SPACE: /* Perl space */
2788 case PT_PXSPACE: /* POSIX space */
2789 switch(c)
2790 {
2791 HSPACE_CASES:
2792 VSPACE_CASES:
2793 return negated;
2794
2795 default:
2796 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2797 }
2798 break; /* Control never reaches here */
2799
2800 case PT_WORD:
2801 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2802 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2803 c == CHAR_UNDERSCORE) == negated;
2804
2805 case PT_CLIST:
2806 p = PRIV(ucd_caseless_sets) + prop->caseset;
2807 for (;;)
2808 {
2809 if (c < *p) return !negated;
2810 if (c == *p++) return negated;
2811 }
2812 break; /* Control never reaches here */
2813 }
2814
2815 return FALSE;
2816 }
2817 #endif /* SUPPORT_UCP */
2818
2819
2820
2821 /*************************************************
2822 * Fill the character property list *
2823 *************************************************/
2824
2825 /* Checks whether the code points to an opcode that can take part in auto-
2826 possessification, and if so, fills a list with its properties.
2827
2828 Arguments:
2829 code points to start of expression
2830 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2831 fcc points to case-flipping table
2832 list points to output list
2833 list[0] will be filled with the opcode
2834 list[1] will be non-zero if this opcode
2835 can match an empty character string
2836 list[2..7] depends on the opcode
2837
2838 Returns: points to the start of the next opcode if *code is accepted
2839 NULL if *code is not accepted
2840 */
2841
2842 static const pcre_uchar *
2843 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2844 const pcre_uint8 *fcc, pcre_uint32 *list)
2845 {
2846 pcre_uchar c = *code;
2847 const pcre_uchar *end;
2848 const pcre_uint32 *clist_src;
2849 pcre_uint32 *clist_dest;
2850 pcre_uint32 chr;
2851 pcre_uchar base;
2852
2853 list[0] = c;
2854 list[1] = FALSE;
2855 code++;
2856
2857 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2858 {
2859 base = get_repeat_base(c);
2860 c -= (base - OP_STAR);
2861
2862 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2863 code += IMM2_SIZE;
2864
2865 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2866
2867 switch(base)
2868 {
2869 case OP_STAR:
2870 list[0] = OP_CHAR;
2871 break;
2872
2873 case OP_STARI:
2874 list[0] = OP_CHARI;
2875 break;
2876
2877 case OP_NOTSTAR:
2878 list[0] = OP_NOT;
2879 break;
2880
2881 case OP_NOTSTARI:
2882 list[0] = OP_NOTI;
2883 break;
2884
2885 case OP_TYPESTAR:
2886 list[0] = *code;
2887 code++;
2888 break;
2889 }
2890 c = list[0];
2891 }
2892
2893 switch(c)
2894 {
2895 case OP_NOT_DIGIT:
2896 case OP_DIGIT:
2897 case OP_NOT_WHITESPACE:
2898 case OP_WHITESPACE:
2899 case OP_NOT_WORDCHAR:
2900 case OP_WORDCHAR:
2901 case OP_ANY:
2902 case OP_ALLANY:
2903 case OP_ANYNL:
2904 case OP_NOT_HSPACE:
2905 case OP_HSPACE:
2906 case OP_NOT_VSPACE:
2907 case OP_VSPACE:
2908 case OP_EXTUNI:
2909 case OP_EODN:
2910 case OP_EOD:
2911 case OP_DOLL:
2912 case OP_DOLLM:
2913 return code;
2914
2915 case OP_CHAR:
2916 case OP_NOT:
2917 GETCHARINCTEST(chr, code);
2918 list[2] = chr;
2919 list[3] = NOTACHAR;
2920 return code;
2921
2922 case OP_CHARI:
2923 case OP_NOTI:
2924 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2925 GETCHARINCTEST(chr, code);
2926 list[2] = chr;
2927
2928 #ifdef SUPPORT_UCP
2929 if (chr < 128 || (chr < 256 && !utf))
2930 list[3] = fcc[chr];
2931 else
2932 list[3] = UCD_OTHERCASE(chr);
2933 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2934 list[3] = (chr < 256) ? fcc[chr] : chr;
2935 #else
2936 list[3] = fcc[chr];
2937 #endif
2938
2939 /* The othercase might be the same value. */
2940
2941 if (chr == list[3])
2942 list[3] = NOTACHAR;
2943 else
2944 list[4] = NOTACHAR;
2945 return code;
2946
2947 #ifdef SUPPORT_UCP
2948 case OP_PROP:
2949 case OP_NOTPROP:
2950 if (code[0] != PT_CLIST)
2951 {
2952 list[2] = code[0];
2953 list[3] = code[1];
2954 return code + 2;
2955 }
2956
2957 /* Convert only if we have enough space. */
2958
2959 clist_src = PRIV(ucd_caseless_sets) + code[1];
2960 clist_dest = list + 2;
2961 code += 2;
2962
2963 do {
2964 if (clist_dest >= list + 8)
2965 {
2966 /* Early return if there is not enough space. This should never
2967 happen, since all clists are shorter than 5 character now. */
2968 list[2] = code[0];
2969 list[3] = code[1];
2970 return code;
2971 }
2972 *clist_dest++ = *clist_src;
2973 }
2974 while(*clist_src++ != NOTACHAR);
2975
2976 /* All characters are stored. The terminating NOTACHAR
2977 is copied form the clist itself. */
2978
2979 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2980 return code;
2981 #endif
2982
2983 case OP_NCLASS:
2984 case OP_CLASS:
2985 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2986 case OP_XCLASS:
2987 if (c == OP_XCLASS)
2988 end = code + GET(code, 0) - 1;
2989 else
2990 #endif
2991 end = code + 32 / sizeof(pcre_uchar);
2992
2993 switch(*end)
2994 {
2995 case OP_CRSTAR:
2996 case OP_CRMINSTAR:
2997 case OP_CRQUERY:
2998 case OP_CRMINQUERY:
2999 case OP_CRPOSSTAR:
3000 case OP_CRPOSQUERY:
3001 list[1] = TRUE;
3002 end++;
3003 break;
3004
3005 case OP_CRPLUS:
3006 case OP_CRMINPLUS:
3007 case OP_CRPOSPLUS:
3008 end++;
3009 break;
3010
3011 case OP_CRRANGE:
3012 case OP_CRMINRANGE:
3013 case OP_CRPOSRANGE:
3014 list[1] = (GET2(end, 1) == 0);
3015 end += 1 + 2 * IMM2_SIZE;
3016 break;
3017 }
3018 list[2] = end - code;
3019 return end;
3020 }
3021 return NULL; /* Opcode not accepted */
3022 }
3023
3024
3025
3026 /*************************************************
3027 * Scan further character sets for match *
3028 *************************************************/
3029
3030 /* Checks whether the base and the current opcode have a common character, in
3031 which case the base cannot be possessified.
3032
3033 Arguments:
3034 code points to the byte code
3035 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3036 cd static compile data
3037 base_list the data list of the base opcode
3038
3039 Returns: TRUE if the auto-possessification is possible
3040 */
3041
3042 static BOOL
3043 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3044 const pcre_uint32 *base_list, const pcre_uchar *base_end)
3045 {
3046 pcre_uchar c;
3047 pcre_uint32 list[8];
3048 const pcre_uint32 *chr_ptr;
3049 const pcre_uint32 *ochr_ptr;
3050 const pcre_uint32 *list_ptr;
3051 const pcre_uchar *next_code;
3052 const pcre_uint8 *class_bitset;
3053 const pcre_uint32 *set1, *set2, *set_end;
3054 pcre_uint32 chr;
3055 BOOL accepted, invert_bits;
3056
3057 /* Note: the base_list[1] contains whether the current opcode has greedy
3058 (represented by a non-zero value) quantifier. This is a different from
3059 other character type lists, which stores here that the character iterator
3060 matches to an empty string (also represented by a non-zero value). */
3061
3062 for(;;)
3063 {
3064 /* All operations move the code pointer forward.
3065 Therefore infinite recursions are not possible. */
3066
3067 c = *code;
3068
3069 /* Skip over callouts */
3070
3071 if (c == OP_CALLOUT)
3072 {
3073 code += PRIV(OP_lengths)[c];
3074 continue;
3075 }
3076
3077 if (c == OP_ALT)
3078 {
3079 do code += GET(code, 1); while (*code == OP_ALT);
3080 c = *code;
3081 }
3082
3083 switch(c)
3084 {
3085 case OP_END:
3086 case OP_KETRPOS:
3087 /* TRUE only in greedy case. The non-greedy case could be replaced by
3088 an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3089 uses more memory, which we cannot get at this stage.) */
3090
3091 return base_list[1] != 0;
3092
3093 case OP_KET:
3094 /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3095 it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3096 cannot be converted to a possessive form. */
3097
3098 if (base_list[1] == 0) return FALSE;
3099
3100 switch(*(code - GET(code, 1)))
3101 {
3102 case OP_ASSERT:
3103 case OP_ASSERT_NOT:
3104 case OP_ASSERTBACK:
3105 case OP_ASSERTBACK_NOT:
3106 case OP_ONCE:
3107 case OP_ONCE_NC:
3108 /* Atomic sub-patterns and assertions can always auto-possessify their
3109 last iterator. */
3110 return TRUE;
3111 }
3112
3113 code += PRIV(OP_lengths)[c];
3114 continue;
3115
3116 case OP_ONCE:
3117 case OP_ONCE_NC:
3118 case OP_BRA:
3119 case OP_CBRA:
3120 next_code = code + GET(code, 1);
3121 code += PRIV(OP_lengths)[c];
3122
3123 while (*next_code == OP_ALT)
3124 {
3125 if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3126 code = next_code + 1 + LINK_SIZE;
3127 next_code += GET(next_code, 1);
3128 }
3129 continue;
3130
3131 case OP_BRAZERO:
3132 case OP_BRAMINZERO:
3133
3134 next_code = code + 1;
3135 if (*next_code != OP_BRA && *next_code != OP_CBRA
3136 && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3137
3138 do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3139
3140 /* The bracket content will be checked by the
3141 OP_BRA/OP_CBRA case above. */
3142 next_code += 1 + LINK_SIZE;
3143 if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3144 return FALSE;
3145
3146 code += PRIV(OP_lengths)[c];
3147 continue;
3148 }
3149
3150 /* Check for a supported opcode, and load its properties. */
3151
3152 code = get_chr_property_list(code, utf, cd->fcc, list);
3153 if (code == NULL) return FALSE; /* Unsupported */
3154
3155 /* If either opcode is a small character list, set pointers for comparing
3156 characters from that list with another list, or with a property. */
3157
3158 if (base_list[0] == OP_CHAR)
3159 {
3160 chr_ptr = base_list + 2;
3161 list_ptr = list;
3162 }
3163 else if (list[0] == OP_CHAR)
3164 {
3165 chr_ptr = list + 2;
3166 list_ptr = base_list;
3167 }
3168
3169 /* Character bitsets can also be compared to certain opcodes. */
3170
3171 else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3172 #ifdef COMPILE_PCRE8
3173 /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3174 || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3175 #endif
3176 )
3177 {
3178 #ifdef COMPILE_PCRE8
3179 if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3180 #else
3181 if (base_list[0] == OP_CLASS)
3182 #endif
3183 {
3184 set1 = (pcre_uint32 *)(base_end - base_list[2]);
3185 list_ptr = list;
3186 }
3187 else
3188 {
3189 set1 = (pcre_uint32 *)(code - list[2]);
3190 list_ptr = base_list;
3191 }
3192
3193 invert_bits = FALSE;
3194 switch(list_ptr[0])
3195 {
3196 case OP_CLASS:
3197 case OP_NCLASS:
3198 set2 = (pcre_uint32 *)
3199 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3200 break;
3201
3202 /* OP_XCLASS cannot be supported here, because its bitset
3203 is not necessarily complete. E.g: [a-\0x{200}] is stored
3204 as a character range, and the appropriate bits are not set. */
3205
3206 case OP_NOT_DIGIT:
3207 invert_bits = TRUE;
3208 /* Fall through */
3209 case OP_DIGIT:
3210 set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3211 break;
3212
3213 case OP_NOT_WHITESPACE:
3214 invert_bits = TRUE;
3215 /* Fall through */
3216 case OP_WHITESPACE:
3217 set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3218 break;
3219
3220 case OP_NOT_WORDCHAR:
3221 invert_bits = TRUE;
3222 /* Fall through */
3223 case OP_WORDCHAR:
3224 set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3225 break;
3226
3227 default:
3228 return FALSE;
3229 }
3230
3231 /* Compare 4 bytes to improve speed. */
3232 set_end = set1 + (32 / 4);
3233 if (invert_bits)
3234 {
3235 do
3236 {
3237 if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3238 }
3239 while (set1 < set_end);
3240 }
3241 else
3242 {
3243 do
3244 {
3245 if ((*set1++ & *set2++) != 0) return FALSE;
3246 }
3247 while (set1 < set_end);
3248 }
3249
3250 if (list[1] == 0) return TRUE;
3251 /* Might be an empty repeat. */
3252 continue;
3253 }
3254
3255 /* Some property combinations also acceptable. Unicode property opcodes are
3256 processed specially; the rest can be handled with a lookup table. */
3257
3258 else
3259 {
3260 pcre_uint32 leftop, rightop;
3261
3262 leftop = base_list[0];
3263 rightop = list[0];
3264
3265 #ifdef SUPPORT_UCP
3266 accepted = FALSE; /* Always set in non-unicode case. */
3267 if (leftop == OP_PROP || leftop == OP_NOTPROP)
3268 {
3269 if (rightop == OP_EOD)
3270 accepted = TRUE;
3271 else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3272 {
3273 int n;
3274 const pcre_uint8 *p;
3275 BOOL same = leftop == rightop;
3276 BOOL lisprop = leftop == OP_PROP;
3277 BOOL risprop = rightop == OP_PROP;
3278 BOOL bothprop = lisprop && risprop;
3279
3280 /* There's a table that specifies how each combination is to be
3281 processed:
3282 0 Always return FALSE (never auto-possessify)
3283 1 Character groups are distinct (possessify if both are OP_PROP)
3284 2 Check character categories in the same group (general or particular)
3285 3 Return TRUE if the two opcodes are not the same
3286 ... see comments below
3287 */
3288
3289 n = propposstab[base_list[2]][list[2]];
3290 switch(n)
3291 {
3292 case 0: break;
3293 case 1: accepted = bothprop; break;
3294 case 2: accepted = (base_list[3] == list[3]) != same; break;
3295 case 3: accepted = !same; break;
3296
3297 case 4: /* Left general category, right particular category */
3298 accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3299 break;
3300
3301 case 5: /* Right general category, left particular category */
3302 accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3303 break;
3304
3305 /* This code is logically tricky. Think hard before fiddling with it.
3306 The posspropstab table has four entries per row. Each row relates to
3307 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3308 Only WORD actually needs all four entries, but using repeats for the
3309 others means they can all use the same code below.
3310
3311 The first two entries in each row are Unicode general categories, and
3312 apply always, because all the characters they include are part of the
3313 PCRE character set. The third and fourth entries are a general and a
3314 particular category, respectively, that include one or more relevant
3315 characters. One or the other is used, depending on whether the check
3316 is for a general or a particular category. However, in both cases the
3317 category contains more characters than the specials that are defined
3318 for the property being tested against. Therefore, it cannot be used
3319 in a NOTPROP case.
3320
3321 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3322 Underscore is covered by ucp_P or ucp_Po. */
3323
3324 case 6: /* Left alphanum vs right general category */
3325 case 7: /* Left space vs right general category */
3326 case 8: /* Left word vs right general category */
3327 p = posspropstab[n-6];
3328 accepted = risprop && lisprop ==
3329 (list[3] != p[0] &&
3330 list[3] != p[1] &&
3331 (list[3] != p[2] || !lisprop));
3332 break;
3333
3334 case 9: /* Right alphanum vs left general category */
3335 case 10: /* Right space vs left general category */
3336 case 11: /* Right word vs left general category */
3337 p = posspropstab[n-9];
3338 accepted = lisprop && risprop ==
3339 (base_list[3] != p[0] &&
3340 base_list[3] != p[1] &&
3341 (base_list[3] != p[2] || !risprop));
3342 break;
3343
3344 case 12: /* Left alphanum vs right particular category */
3345 case 13: /* Left space vs right particular category */
3346 case 14: /* Left word vs right particular category */
3347 p = posspropstab[n-12];
3348 accepted = risprop && lisprop ==
3349 (catposstab[p[0]][list[3]] &&
3350 catposstab[p[1]][list[3]] &&
3351 (list[3] != p[3] || !lisprop));
3352 break;
3353
3354 case 15: /* Right alphanum vs left particular category */
3355 case 16: /* Right space vs left particular category */
3356 case 17: /* Right word vs left particular category */
3357 p = posspropstab[n-15];
3358 accepted = lisprop && risprop ==
3359 (catposstab[p[0]][base_list[3]] &&
3360 catposstab[p[1]][base_list[3]] &&
3361 (base_list[3] != p[3] || !risprop));
3362 break;
3363 }
3364 }
3365 }
3366
3367 else
3368 #endif /* SUPPORT_UCP */
3369
3370 accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3371 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3372 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3373
3374 if (!accepted)
3375 return FALSE;
3376
3377 if (list[1] == 0) return TRUE;
3378 /* Might be an empty repeat. */
3379 continue;
3380 }
3381
3382 /* Control reaches here only if one of the items is a small character list.
3383 All characters are checked against the other side. */
3384
3385 do
3386 {
3387 chr = *chr_ptr;
3388
3389 switch(list_ptr[0])
3390 {
3391 case OP_CHAR:
3392 ochr_ptr = list_ptr + 2;
3393 do
3394 {
3395 if (chr == *ochr_ptr) return FALSE;
3396 ochr_ptr++;
3397 }
3398 while(*ochr_ptr != NOTACHAR);
3399 break;
3400
3401 case OP_NOT:
3402 ochr_ptr = list_ptr + 2;
3403 do
3404 {
3405 if (chr == *ochr_ptr)
3406 break;
3407 ochr_ptr++;
3408 }
3409 while(*ochr_ptr != NOTACHAR);
3410 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3411 break;
3412
3413 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3414 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3415
3416 case OP_DIGIT:
3417 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3418 break;
3419
3420 case OP_NOT_DIGIT:
3421 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3422 break;
3423
3424 case OP_WHITESPACE:
3425 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3426 break;
3427
3428 case OP_NOT_WHITESPACE:
3429 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3430 break;
3431
3432 case OP_WORDCHAR:
3433 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3434 break;
3435
3436 case OP_NOT_WORDCHAR:
3437 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3438 break;
3439
3440 case OP_HSPACE:
3441 switch(chr)
3442 {
3443 HSPACE_CASES: return FALSE;
3444 default: break;
3445 }
3446 break;
3447
3448 case OP_NOT_HSPACE:
3449 switch(chr)
3450 {
3451 HSPACE_CASES: break;
3452 default: return FALSE;
3453 }
3454 break;
3455
3456 case OP_ANYNL:
3457 case OP_VSPACE:
3458 switch(chr)
3459 {
3460 VSPACE_CASES: return FALSE;
3461 default: break;
3462 }
3463 break;
3464
3465 case OP_NOT_VSPACE:
3466 switch(chr)
3467 {
3468 VSPACE_CASES: break;
3469 default: return FALSE;
3470 }
3471 break;
3472
3473 case OP_DOLL:
3474 case OP_EODN:
3475 switch (chr)
3476 {
3477 case CHAR_CR:
3478 case CHAR_LF:
3479 case CHAR_VT:
3480 case CHAR_FF:
3481 case CHAR_NEL:
3482 #ifndef EBCDIC
3483 case 0x2028:
3484 case 0x2029:
3485 #endif /* Not EBCDIC */
3486 return FALSE;
3487 }
3488 break;
3489
3490 case OP_EOD: /* Can always possessify before \z */
3491 break;
3492
3493 #ifdef SUPPORT_UCP
3494 case OP_PROP:
3495 case OP_NOTPROP:
3496 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3497 list_ptr[0] == OP_NOTPROP))
3498 return FALSE;
3499 break;
3500 #endif
3501
3502 case OP_NCLASS:
3503 if (chr > 255) return FALSE;
3504 /* Fall through */
3505
3506 case OP_CLASS:
3507 if (chr > 255) break;
3508 class_bitset = (pcre_uint8 *)
3509 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3510 if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3511 break;
3512
3513 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3514 case OP_XCLASS:
3515 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3516 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3517 break;
3518 #endif
3519
3520 default:
3521 return FALSE;
3522 }
3523
3524 chr_ptr++;
3525 }
3526 while(*chr_ptr != NOTACHAR);
3527
3528 /* At least one character must be matched from this opcode. */
3529
3530 if (list[1] == 0) return TRUE;
3531 }
3532
3533 return FALSE;
3534 }
3535
3536
3537
3538 /*************************************************
3539 * Scan compiled regex for auto-possession *
3540 *************************************************/
3541
3542 /* Replaces single character iterations with their possessive alternatives
3543 if appropriate. This function modifies the compiled opcode!
3544
3545 Arguments:
3546 code points to start of the byte code
3547 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3548 cd static compile data
3549
3550 Returns: nothing
3551 */
3552
3553 static void
3554 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3555 {
3556 register pcre_uchar c;
3557 const pcre_uchar *end;
3558 pcre_uchar *repeat_opcode;
3559 pcre_uint32 list[8];
3560
3561 for (;;)
3562 {
3563 c = *code;
3564
3565 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3566 {
3567 c -= get_repeat_base(c) - OP_STAR;
3568 end = (c <= OP_MINUPTO) ?
3569 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3570 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3571
3572 if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3573 {
3574 switch(c)
3575 {
3576 case OP_STAR:
3577 *code += OP_POSSTAR - OP_STAR;
3578 break;
3579
3580 case OP_MINSTAR:
3581 *code += OP_POSSTAR - OP_MINSTAR;
3582 break;
3583
3584 case OP_PLUS:
3585 *code += OP_POSPLUS - OP_PLUS;
3586 break;
3587
3588 case OP_MINPLUS:
3589 *code += OP_POSPLUS - OP_MINPLUS;
3590 break;
3591
3592 case OP_QUERY:
3593 *code += OP_POSQUERY - OP_QUERY;
3594 break;
3595
3596 case OP_MINQUERY:
3597 *code += OP_POSQUERY - OP_MINQUERY;
3598 break;
3599
3600 case OP_UPTO:
3601 *code += OP_POSUPTO - OP_UPTO;
3602 break;
3603
3604 case OP_MINUPTO:
3605 *code += OP_MINUPTO - OP_UPTO;
3606 break;
3607 }
3608 }
3609 c = *code;
3610 }
3611 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3612 {
3613 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3614 if (c == OP_XCLASS)
3615 repeat_opcode = code + GET(code, 1);
3616 else
3617 #endif
3618 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3619
3620 c = *repeat_opcode;
3621 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3622 {
3623 /* end must not be NULL. */
3624 end = get_chr_property_list(code, utf, cd->fcc, list);
3625
3626 list[1] = (c & 1) == 0;
3627
3628 if (compare_opcodes(end, utf, cd, list, end))
3629 {
3630 switch (c)
3631 {
3632 case OP_CRSTAR:
3633 case OP_CRMINSTAR:
3634 *repeat_opcode = OP_CRPOSSTAR;
3635 break;
3636
3637 case OP_CRPLUS:
3638 case OP_CRMINPLUS:
3639 *repeat_opcode = OP_CRPOSPLUS;
3640 break;
3641
3642 case OP_CRQUERY:
3643 case OP_CRMINQUERY:
3644 *repeat_opcode = OP_CRPOSQUERY;
3645 break;
3646
3647 case OP_CRRANGE:
3648 case OP_CRMINRANGE:
3649 *repeat_opcode = OP_CRPOSRANGE;
3650 break;
3651 }
3652 }
3653 }
3654 c = *code;
3655 }
3656
3657 switch(c)
3658 {
3659 case OP_END:
3660 return;
3661
3662 case OP_TYPESTAR:
3663 case OP_TYPEMINSTAR:
3664 case OP_TYPEPLUS:
3665 case OP_TYPEMINPLUS:
3666 case OP_TYPEQUERY:
3667 case OP_TYPEMINQUERY:
3668 case OP_TYPEPOSSTAR:
3669 case OP_TYPEPOSPLUS:
3670 case OP_TYPEPOSQUERY:
3671 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3672 break;
3673
3674 case OP_TYPEUPTO:
3675 case OP_TYPEMINUPTO:
3676 case OP_TYPEEXACT:
3677 case OP_TYPEPOSUPTO:
3678 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3679 code += 2;
3680 break;
3681
3682 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3683 case OP_XCLASS:
3684 code += GET(code, 1);
3685 break;
3686 #endif
3687
3688 case OP_MARK:
3689 case OP_PRUNE_ARG:
3690 case OP_SKIP_ARG:
3691 case OP_THEN_ARG:
3692 code += code[1];
3693 break;
3694 }
3695
3696 /* Add in the fixed length from the table */
3697
3698 code += PRIV(OP_lengths)[c];
3699
3700 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3701 a multi-byte character. The length in the table is a minimum, so we have to
3702 arrange to skip the extra bytes. */
3703
3704 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3705 if (utf) switch(c)
3706 {
3707 case OP_CHAR:
3708 case OP_CHARI:
3709 case OP_NOT:
3710 case OP_NOTI:
3711 case OP_STAR:
3712 case OP_MINSTAR:
3713 case OP_PLUS:
3714 case OP_MINPLUS:
3715 case OP_QUERY:
3716 case OP_MINQUERY:
3717 case OP_UPTO:
3718 case OP_MINUPTO:
3719 case OP_EXACT:
3720 case OP_POSSTAR:
3721 case OP_POSPLUS:
3722 case OP_POSQUERY:
3723 case OP_POSUPTO:
3724 case OP_STARI:
3725 case OP_MINSTARI:
3726 case OP_PLUSI:
3727 case OP_MINPLUSI:
3728 case OP_QUERYI:
3729 case OP_MINQUERYI:
3730 case OP_UPTOI:
3731 case OP_MINUPTOI:
3732 case OP_EXACTI:
3733 case OP_POSSTARI:
3734 case OP_POSPLUSI:
3735 case OP_POSQUERYI:
3736 case OP_POSUPTOI:
3737 case OP_NOTSTAR:
3738 case OP_NOTMINSTAR:
3739 case OP_NOTPLUS:
3740 case OP_NOTMINPLUS:
3741 case OP_NOTQUERY:
3742 case OP_NOTMINQUERY:
3743 case OP_NOTUPTO:
3744 case OP_NOTMINUPTO:
3745 case OP_NOTEXACT:
3746 case OP_NOTPOSSTAR:
3747 case OP_NOTPOSPLUS:
3748 case OP_NOTPOSQUERY:
3749 case OP_NOTPOSUPTO:
3750 case OP_NOTSTARI:
3751 case OP_NOTMINSTARI:
3752 case OP_NOTPLUSI:
3753 case OP_NOTMINPLUSI:
3754 case OP_NOTQUERYI:
3755 case OP_NOTMINQUERYI:
3756 case OP_NOTUPTOI:
3757 case OP_NOTMINUPTOI:
3758 case OP_NOTEXACTI:
3759 case OP_NOTPOSSTARI:
3760 case OP_NOTPOSPLUSI:
3761 case OP_NOTPOSQUERYI:
3762 case OP_NOTPOSUPTOI:
3763 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3764 break;
3765 }
3766 #else
3767 (void)(utf); /* Keep compiler happy by referencing function argument */
3768 #endif
3769 }
3770 }
3771
3772
3773
3774 /*************************************************
3775 * Check for POSIX class syntax *
3776 *************************************************/
3777
3778 /* This function is called when the sequence "[:" or "[." or "[=" is
3779 encountered in a character class. It checks whether this is followed by a
3780 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3781 reach an unescaped ']' without the special preceding character, return FALSE.
3782
3783 Originally, this function only recognized a sequence of letters between the
3784 terminators, but it seems that Perl recognizes any sequence of characters,
3785 though of course unknown POSIX names are subsequently rejected. Perl gives an
3786 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3787 didn't consider this to be a POSIX class. Likewise for [:1234:].
3788
3789 The problem in trying to be exactly like Perl is in the handling of escapes. We
3790 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3791 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3792 below handles the special case of \], but does not try to do any other escape
3793 processing. This makes it different from Perl for cases such as [:l\ower:]
3794 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3795 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3796 I think.
3797
3798 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3799 It seems that the appearance of a nested POSIX class supersedes an apparent
3800 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3801 a digit.
3802
3803 In Perl, unescaped square brackets may also appear as part of class names. For
3804 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3805 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3806 seem right at all. PCRE does not allow closing square brackets in POSIX class
3807 names.
3808
3809 Arguments:
3810 ptr pointer to the initial [
3811 endptr where to return the end pointer
3812
3813 Returns: TRUE or FALSE
3814 */
3815
3816 static BOOL
3817 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3818 {
3819 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3820 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3821 for (++ptr; *ptr != CHAR_NULL; ptr++)
3822 {
3823 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3824 ptr++;
3825 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3826 else
3827 {
3828 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3829 {
3830 *endptr = ptr;
3831 return TRUE;
3832 }
3833 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3834 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3835 ptr[1] == CHAR_EQUALS_SIGN) &&
3836 check_posix_syntax(ptr, endptr))
3837 return FALSE;
3838 }
3839 }
3840 return FALSE;
3841 }
3842
3843
3844
3845
3846 /*************************************************
3847 * Check POSIX class name *
3848 *************************************************/
3849
3850 /* This function is called to check the name given in a POSIX-style class entry
3851 such as [:alnum:].
3852
3853 Arguments:
3854 ptr points to the first letter
3855 len the length of the name
3856
3857 Returns: a value representing the name, or -1 if unknown
3858 */
3859
3860 static int
3861 check_posix_name(const pcre_uchar *ptr, int len)
3862 {
3863 const char *pn = posix_names;
3864 register int yield = 0;
3865 while (posix_name_lengths[yield] != 0)
3866 {
3867 if (len == posix_name_lengths[yield] &&
3868 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3869 pn += posix_name_lengths[yield] + 1;
3870 yield++;
3871 }
3872 return -1;
3873 }
3874
3875
3876 /*************************************************
3877 * Adjust OP_RECURSE items in repeated group *
3878 *************************************************/
3879
3880 /* OP_RECURSE items contain an offset from the start of the regex to the group
3881 that is referenced. This means that groups can be replicated for fixed
3882 repetition simply by copying (because the recursion is allowed to refer to
3883 earlier groups that are outside the current group). However, when a group is
3884 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3885 inserted before it, after it has been compiled. This means that any OP_RECURSE
3886 items within it that refer to the group itself or any contained groups have to
3887 have their offsets adjusted. That one of the jobs of this function. Before it
3888 is called, the partially compiled regex must be temporarily terminated with
3889 OP_END.
3890
3891 This function has been extended with the possibility of forward references for
3892 recursions and subroutine calls. It must also check the list of such references
3893 for the group we are dealing with. If it finds that one of the recursions in
3894 the current group is on this list, it adjusts the offset in the list, not the
3895 value in the reference (which is a group number).
3896
3897 Arguments:
3898 group points to the start of the group
3899 adjust the amount by which the group is to be moved
3900 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3901 cd contains pointers to tables etc.
3902 save_hwm the hwm forward reference pointer at the start of the group
3903
3904 Returns: nothing
3905 */
3906
3907 static void
3908 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3909 pcre_uchar *save_hwm)
3910 {
3911 pcre_uchar *ptr = group;
3912
3913 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3914 {
3915 int offset;
3916 pcre_uchar *hc;
3917
3918 /* See if this recursion is on the forward reference list. If so, adjust the
3919 reference. */
3920
3921 for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3922 {
3923 offset = (int)GET(hc, 0);
3924 if (cd->start_code + offset == ptr + 1)
3925 {
3926 PUT(hc, 0, offset + adjust);
3927 break;
3928 }
3929 }
3930
3931 /* Otherwise, adjust the recursion offset if it's after the start of this
3932 group. */
3933
3934 if (hc >= cd->hwm)
3935 {
3936 offset = (int)GET(ptr, 1);
3937 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3938 }
3939
3940 ptr += 1 + LINK_SIZE;
3941 }
3942 }
3943
3944
3945
3946 /*************************************************
3947 * Insert an automatic callout point *
3948 *************************************************/
3949
3950 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3951 callout points before each pattern item.
3952
3953 Arguments:
3954 code current code pointer
3955 ptr current pattern pointer
3956 cd pointers to tables etc
3957
3958 Returns: new code pointer
3959 */
3960
3961 static pcre_uchar *
3962 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3963 {
3964 *code++ = OP_CALLOUT;
3965 *code++ = 255;
3966 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
3967 PUT(code, LINK_SIZE, 0); /* Default length */
3968 return code + 2 * LINK_SIZE;
3969 }
3970
3971
3972
3973 /*************************************************
3974 * Complete a callout item *
3975 *************************************************/
3976
3977 /* A callout item contains the length of the next item in the pattern, which
3978 we can't fill in till after we have reached the relevant point. This is used
3979 for both automatic and manual callouts.
3980
3981 Arguments:
3982 previous_callout points to previous callout item
3983 ptr current pattern pointer
3984 cd pointers to tables etc
3985
3986 Returns: nothing
3987 */
3988
3989 static void
3990 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3991 {
3992 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3993 PUT(previous_callout, 2 + LINK_SIZE, length);
3994 }
3995
3996
3997
3998 #ifdef SUPPORT_UCP
3999 /*************************************************
4000 * Get othercase range *
4001 *************************************************/
4002
4003 /* This function is passed the start and end of a class range, in UTF-8 mode
4004 with UCP support. It searches up the characters, looking for ranges of
4005 characters in the "other" case. Each call returns the next one, updating the
4006 start address. A character with multiple other cases is returned on its own
4007 with a special return value.
4008
4009 Arguments:
4010 cptr points to starting character value; updated
4011 d end value
4012 ocptr where to put start of othercase range
4013 odptr where to put end of othercase range
4014
4015 Yield: -1 when no more
4016 0 when a range is returned
4017 >0 the CASESET offset for char with multiple other cases
4018 in this case, ocptr contains the original
4019 */
4020
4021 static int
4022 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4023 pcre_uint32 *odptr)
4024 {
4025 pcre_uint32 c, othercase, next;
4026 unsigned int co;
4027
4028 /* Find the first character that has an other case. If it has multiple other
4029 cases, return its case offset value. */
4030
4031 for (c = *cptr; c <= d; c++)
4032 {
4033 if ((co = UCD_CASESET(c)) != 0)
4034 {
4035 *ocptr = c++; /* Character that has the set */
4036 *cptr = c; /* Rest of input range */
4037 return (int)co;
4038 }
4039 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4040 }
4041
4042 if (c > d) return -1; /* Reached end of range */
4043
4044 *ocptr = othercase;
4045 next = othercase + 1;
4046
4047 for (++c; c <= d; c++)
4048 {
4049 if (UCD_OTHERCASE(c) != next) break;
4050 next++;
4051 }
4052
4053 *odptr = next - 1; /* End of othercase range */
4054 *cptr = c; /* Rest of input range */
4055 return 0;
4056 }
4057 #endif /* SUPPORT_UCP */
4058
4059
4060
4061 /*************************************************
4062 * Add a character or range to a class *
4063 *************************************************/
4064
4065 /* This function packages up the logic of adding a character or range of
4066 characters to a class. The character values in the arguments will be within the
4067 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4068 mutually recursive with the function immediately below.
4069
4070 Arguments:
4071 classbits the bit map for characters < 256
4072 uchardptr points to the pointer for extra data
4073 options the options word
4074 cd contains pointers to tables etc.
4075 start start of range character
4076 end end of range character
4077
4078 Returns: the number of < 256 characters added
4079 the pointer to extra data is updated
4080 */
4081
4082 static int
4083 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4084 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4085 {
4086 pcre_uint32 c;
4087 int n8 = 0;
4088
4089 /* If caseless matching is required, scan the range and process alternate
4090 cases. In Unicode, there are 8-bit characters that have alternate cases that
4091 are greater than 255 and vice-versa. Sometimes we can just extend the original
4092 range. */
4093
4094 if ((options & PCRE_CASELESS) != 0)
4095 {
4096 #ifdef SUPPORT_UCP
4097 if ((options & PCRE_UTF8) != 0)
4098 {
4099 int rc;
4100 pcre_uint32 oc, od;
4101
4102 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4103 c = start;
4104
4105 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4106 {
4107 /* Handle a single character that has more than one other case. */
4108
4109 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4110 PRIV(ucd_caseless_sets) + rc, oc);
4111
4112 /* Do nothing if the other case range is within the original range. */
4113
4114 else if (oc >= start && od <= end) continue;
4115
4116 /* Extend the original range if there is overlap, noting that if oc < c, we
4117 can't have od > end because a subrange is always shorter than the basic
4118 range. Otherwise, use a recursive call to add the additional range. */
4119
4120 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4121 else if (od > end && oc <= end + 1) end = od; /* Extend upwards */
4122 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4123 }
4124 }
4125 else
4126 #endif /* SUPPORT_UCP */
4127
4128 /* Not UTF-mode, or no UCP */
4129
4130 for (c = start; c <= end && c < 256; c++)
4131 {
4132 SETBIT(classbits, cd->fcc[c]);
4133 n8++;
4134 }
4135 }
4136
4137 /* Now handle the original range. Adjust the final value according to the bit
4138 length - this means that the same lists of (e.g.) horizontal spaces can be used
4139 in all cases. */
4140
4141 #if defined COMPILE_PCRE8
4142 #ifdef SUPPORT_UTF
4143 if ((options & PCRE_UTF8) == 0)
4144 #endif
4145 if (end > 0xff) end = 0xff;
4146
4147 #elif defined COMPILE_PCRE16
4148 #ifdef SUPPORT_UTF
4149 if ((options & PCRE_UTF16) == 0)
4150 #endif
4151 if (end > 0xffff) end = 0xffff;
4152
4153 #endif /* COMPILE_PCRE[8|16] */
4154
4155 /* If all characters are less than 256, use the bit map. Otherwise use extra
4156 data. */
4157
4158 if (end < 0x100)
4159 {
4160 for (c = start; c <= end; c++)
4161 {
4162 n8++;
4163 SETBIT(classbits, c);
4164 }
4165 }
4166
4167 else
4168 {
4169 pcre_uchar *uchardata = *uchardptr;
4170
4171 #ifdef SUPPORT_UTF
4172 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4173 {
4174 if (start < end)
4175 {
4176 *uchardata++ = XCL_RANGE;
4177 uchardata += PRIV(ord2utf)(start, uchardata);
4178 uchardata += PRIV(ord2utf)(end, uchardata);
4179 }
4180 else if (start == end)
4181 {
4182 *uchardata++ = XCL_SINGLE;
4183 uchardata += PRIV(ord2utf)(start, uchardata);
4184 }
4185 }
4186 else
4187 #endif /* SUPPORT_UTF */
4188
4189 /* Without UTF support, character values are constrained by the bit length,
4190 and can only be > 256 for 16-bit and 32-bit libraries. */
4191
4192 #ifdef COMPILE_PCRE8
4193 {}
4194 #else
4195 if (start < end)
4196 {
4197 *uchardata++ = XCL_RANGE;
4198 *uchardata++ = start;
4199 *uchardata++ = end;
4200 }
4201 else if (start == end)
4202 {
4203 *uchardata++ = XCL_SINGLE;
4204 *uchardata++ = start;
4205 }
4206 #endif
4207
4208 *uchardptr = uchardata; /* Updata extra data pointer */
4209 }
4210
4211 return n8; /* Number of 8-bit characters */
4212 }
4213
4214
4215
4216
4217 /*************************************************
4218 * Add a list of characters to a class *
4219 *************************************************/
4220
4221 /* This function is used for adding a list of case-equivalent characters to a
4222 class, and also for adding a list of horizontal or vertical whitespace. If the
4223 list is in order (which it should be), ranges of characters are detected and
4224 handled appropriately. This function is mutually recursive with the function
4225 above.
4226
4227 Arguments:
4228 classbits the bit map for characters < 256
4229 uchardptr points to the pointer for extra data
4230 options the options word
4231 cd contains pointers to tables etc.
4232 p points to row of 32-bit values, terminated by NOTACHAR
4233 except character to omit; this is used when adding lists of
4234 case-equivalent characters to avoid including the one we
4235 already know about
4236
4237 Returns: the number of < 256 characters added
4238 the pointer to extra data is updated
4239 */
4240
4241 static int
4242 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4243 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4244 {
4245 int n8 = 0;
4246 while (p[0] < NOTACHAR)
4247 {
4248 int n = 0;
4249 if (p[0] != except)
4250 {
4251 while(p[n+1] == p[0] + n + 1) n++;
4252 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4253 }
4254 p += n + 1;
4255 }
4256 return n8;
4257 }
4258
4259
4260
4261 /*************************************************
4262 * Add characters not in a list to a class *
4263 *************************************************/
4264
4265 /* This function is used for adding the complement of a list of horizontal or
4266 vertical whitespace to a class. The list must be in order.
4267
4268 Arguments:
4269 classbits the bit map for characters < 256
4270 uchardptr points to the pointer for extra data
4271 options the options word
4272 cd contains pointers to tables etc.
4273 p points to row of 32-bit values, terminated by NOTACHAR
4274
4275 Returns: the number of < 256 characters added
4276 the pointer to extra data is updated
4277 */
4278
4279 static int
4280 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4281 int options, compile_data *cd, const pcre_uint32 *p)
4282 {
4283 BOOL utf = (options & PCRE_UTF8) != 0;
4284 int n8 = 0;
4285 if (p[0] > 0)
4286 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4287 while (p[0] < NOTACHAR)
4288 {
4289 while (p[1] == p[0] + 1) p++;
4290 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4291 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4292 p++;
4293 }
4294 return n8;
4295 }
4296
4297
4298
4299 /*************************************************
4300 * Compile one branch *
4301 *************************************************/
4302
4303 /* Scan the pattern, compiling it into the a vector. If the options are
4304 changed during the branch, the pointer is used to change the external options
4305 bits. This function is used during the pre-compile phase when we are trying
4306 to find out the amount of memory needed, as well as during the real compile
4307 phase. The value of lengthptr distinguishes the two phases.
4308
4309 Arguments:
4310 optionsptr pointer to the option bits
4311 codeptr points to the pointer to the current code point
4312 ptrptr points to the current pattern pointer
4313 errorcodeptr points to error code variable
4314 firstcharptr place to put the first required character
4315 firstcharflagsptr place to put the first character flags, or a negative number
4316 reqcharptr place to put the last required character
4317 reqcharflagsptr place to put the last required character flags, or a negative number
4318 bcptr points to current branch chain
4319 cond_depth conditional nesting depth
4320 cd contains pointers to tables etc.
4321 lengthptr NULL during the real compile phase
4322 points to length accumulator during pre-compile phase
4323
4324 Returns: TRUE on success
4325 FALSE, with *errorcodeptr set non-zero on error
4326 */
4327
4328 static BOOL
4329 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4330 const pcre_uchar **ptrptr, int *errorcodeptr,
4331 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4332 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4333 branch_chain *bcptr, int cond_depth,
4334 compile_data *cd, int *lengthptr)
4335 {
4336 int repeat_type, op_type;
4337 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4338 int bravalue = 0;
4339 int greedy_default, greedy_non_default;
4340 pcre_uint32 firstchar, reqchar;
4341 pcre_int32 firstcharflags, reqcharflags;
4342 pcre_uint32 zeroreqchar, zerofirstchar;
4343 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4344 pcre_int32 req_caseopt, reqvary, tempreqvary;
4345 int options = *optionsptr; /* May change dynamically */
4346 int after_manual_callout = 0;
4347 int length_prevgroup = 0;
4348 register pcre_uint32 c;
4349 int escape;
4350 register pcre_uchar *code = *codeptr;
4351 pcre_uchar *last_code = code;
4352 pcre_uchar *orig_code = code;
4353 pcre_uchar *tempcode;
4354 BOOL inescq = FALSE;
4355 BOOL groupsetfirstchar = FALSE;
4356 const pcre_uchar *ptr = *ptrptr;
4357 const pcre_uchar *tempptr;
4358 const pcre_uchar *nestptr = NULL;
4359 pcre_uchar *previous = NULL;
4360 pcre_uchar *previous_callout = NULL;
4361 pcre_uchar *save_hwm = NULL;
4362 pcre_uint8 classbits[32];
4363
4364 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4365 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4366 dynamically as we process the pattern. */
4367
4368 #ifdef SUPPORT_UTF
4369 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4370 BOOL utf = (options & PCRE_UTF8) != 0;
4371 #ifndef COMPILE_PCRE32
4372 pcre_uchar utf_chars[6];
4373 #endif
4374 #else
4375 BOOL utf = FALSE;
4376 #endif
4377
4378 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4379 class_uchardata always so that it can be passed to add_to_class() always,
4380 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4381 alternative calls for the different cases. */
4382
4383 pcre_uchar *class_uchardata;
4384 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4385 BOOL xclass;
4386 pcre_uchar *class_uchardata_base;
4387 #endif
4388
4389 #ifdef PCRE_DEBUG
4390 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4391 #endif
4392
4393 /* Set up the default and non-default settings for greediness */
4394
4395 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4396 greedy_non_default = greedy_default ^ 1;
4397
4398 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4399 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4400 matches a non-fixed char first char; reqchar just remains unset if we never
4401 find one.
4402
4403 When we hit a repeat whose minimum is zero, we may have to adjust these values
4404 to take the zero repeat into account. This is implemented by setting them to
4405 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4406 item types that can be repeated set these backoff variables appropriately. */
4407
4408 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4409 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4410
4411 /* The variable req_caseopt contains either the REQ_CASELESS value
4412 or zero, according to the current setting of the caseless flag. The
4413 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4414 firstchar or reqchar variables to record the case status of the
4415 value. This is used only for ASCII characters. */
4416
4417 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4418
4419 /* Switch on next character until the end of the branch */
4420
4421 for (;; ptr++)
4422 {
4423 BOOL negate_class;
4424 BOOL should_flip_negation;
4425 BOOL possessive_quantifier;
4426 BOOL is_quantifier;
4427 BOOL is_recurse;
4428 BOOL reset_bracount;
4429 int class_has_8bitchar;
4430 int class_one_char;
4431 int newoptions;
4432 int recno;
4433 int refsign;
4434 int skipbytes;
4435 pcre_uint32 subreqchar, subfirstchar;
4436 pcre_int32 subreqcharflags, subfirstcharflags;
4437 int terminator;
4438 unsigned int mclength;
4439 unsigned int tempbracount;
4440 pcre_uint32 ec;
4441 pcre_uchar mcbuffer[8];
4442
4443 /* Get next character in the pattern */
4444
4445 c = *ptr;
4446
4447 /* If we are at the end of a nested substitution, revert to the outer level
4448 string. Nesting only happens one level deep. */
4449
4450 if (c == CHAR_NULL && nestptr != NULL)
4451 {
4452 ptr = nestptr;
4453 nestptr = NULL;
4454 c = *ptr;
4455 }
4456
4457 /* If we are in the pre-compile phase, accumulate the length used for the
4458 previous cycle of this loop. */
4459
4460 if (lengthptr != NULL)
4461 {
4462 #ifdef PCRE_DEBUG
4463 if (code > cd->hwm) cd->hwm = code; /* High water info */
4464 #endif
4465 if (code > cd->start_workspace + cd->workspace_size -
4466 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4467 {
4468 *errorcodeptr = ERR52;
4469 goto FAILED;
4470 }
4471
4472 /* There is at least one situation where code goes backwards: this is the
4473 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4474 the class is simply eliminated. However, it is created first, so we have to
4475 allow memory for it. Therefore, don't ever reduce the length at this point.
4476 */
4477
4478 if (code < last_code) code = last_code;
4479
4480 /* Paranoid check for integer overflow */
4481
4482 if (OFLOW_MAX - *lengthptr < code - last_code)
4483 {
4484 *errorcodeptr = ERR20;
4485 goto FAILED;
4486 }
4487
4488 *lengthptr += (int)(code - last_code);
4489 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4490 (int)(code - last_code), c, c));
4491
4492 /* If "previous" is set and it is not at the start of the work space, move
4493 it back to there, in order to avoid filling up the work space. Otherwise,
4494 if "previous" is NULL, reset the current code pointer to the start. */
4495
4496 if (previous != NULL)
4497 {
4498 if (previous > orig_code)
4499 {
4500 memmove(orig_code, previous, IN_UCHARS(code - previous));
4501 code -= previous - orig_code;
4502 previous = orig_code;
4503 }
4504 }
4505 else code = orig_code;
4506
4507 /* Remember where this code item starts so we can pick up the length
4508 next time round. */
4509
4510 last_code = code;
4511 }
4512
4513 /* In the real compile phase, just check the workspace used by the forward
4514 reference list. */
4515
4516 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4517 WORK_SIZE_SAFETY_MARGIN)
4518 {
4519 *errorcodeptr = ERR52;
4520 goto FAILED;
4521 }
4522
4523 /* If in \Q...\E, check for the end; if not, we have a literal */
4524
4525 if (inescq && c != CHAR_NULL)
4526 {
4527 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4528 {
4529 inescq = FALSE;
4530 ptr++;
4531 continue;
4532 }
4533 else
4534 {
4535 if (previous_callout != NULL)
4536 {
4537 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4538 complete_callout(previous_callout, ptr, cd);
4539 previous_callout = NULL;
4540 }
4541 if ((options & PCRE_AUTO_CALLOUT) != 0)
4542 {
4543 previous_callout = code;
4544 code = auto_callout(code, ptr, cd);
4545 }
4546 goto NORMAL_CHAR;
4547 }
4548 }
4549
4550 is_quantifier =
4551 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4552 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4553
4554 /* Fill in length of a previous callout, except when the next thing is a
4555 quantifier or when processing a property substitution string in UCP mode. */
4556
4557 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4558 after_manual_callout-- <= 0)
4559 {
4560 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4561 complete_callout(previous_callout, ptr, cd);
4562 previous_callout = NULL;
4563 }
4564
4565 /* In extended mode, skip white space and comments. */
4566
4567 if ((options & PCRE_EXTENDED) != 0)
4568 {
4569 if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
4570 if (c == CHAR_NUMBER_SIGN)
4571 {
4572 ptr++;
4573 while (*ptr != CHAR_NULL)
4574 {
4575 if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
4576 ptr++;
4577 #ifdef SUPPORT_UTF
4578 if (utf) FORWARDCHAR(ptr);
4579 #endif
4580 }
4581 if (*ptr != CHAR_NULL) continue;
4582
4583 /* Else fall through to handle end of string */
4584 c = 0;
4585 }
4586 }
4587
4588 /* No auto callout for quantifiers, or while processing property strings that
4589 are substituted for \w etc in UCP mode. */
4590
4591 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4592 {
4593 previous_callout = code;
4594 code = auto_callout(code, ptr, cd);
4595 }
4596
4597 switch(c)
4598 {
4599 /* ===================================================================*/
4600 case 0: /* The branch terminates at string end */
4601 case CHAR_VERTICAL_LINE: /* or | or ) */
4602 case CHAR_RIGHT_PARENTHESIS:
4603 *firstcharptr = firstchar;
4604 *firstcharflagsptr = firstcharflags;
4605 *reqcharptr = reqchar;
4606 *reqcharflagsptr = reqcharflags;
4607 *codeptr = code;
4608 *ptrptr = ptr;
4609 if (lengthptr != NULL)
4610 {
4611 if (OFLOW_MAX - *lengthptr < code - last_code)
4612 {
4613 *errorcodeptr = ERR20;
4614 goto FAILED;
4615 }
4616 *lengthptr += (int)(code - last_code); /* To include callout length */
4617 DPRINTF((">> end branch\n"));
4618 }
4619 return TRUE;
4620
4621
4622 /* ===================================================================*/
4623 /* Handle single-character metacharacters. In multiline mode, ^ disables
4624 the setting of any following char as a first character. */
4625
4626 case CHAR_CIRCUMFLEX_ACCENT:
4627 previous = NULL;
4628 if ((options & PCRE_MULTILINE) != 0)
4629 {
4630 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4631 *code++ = OP_CIRCM;
4632 }
4633 else *code++ = OP_CIRC;
4634 break;
4635
4636 case CHAR_DOLLAR_SIGN:
4637 previous = NULL;
4638 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4639 break;
4640
4641 /* There can never be a first char if '.' is first, whatever happens about
4642 repeats. The value of reqchar doesn't change either. */
4643
4644 case CHAR_DOT:
4645 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4646 zerofirstchar = firstchar;
4647 zerofirstcharflags = firstcharflags;
4648 zeroreqchar = reqchar;
4649 zeroreqcharflags = reqcharflags;
4650 previous = code;
4651 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4652 break;
4653
4654
4655 /* ===================================================================*/
4656 /* Character classes. If the included characters are all < 256, we build a
4657 32-byte bitmap of the permitted characters, except in the special case
4658 where there is only one such character. For negated classes, we build the
4659 map as usual, then invert it at the end. However, we use a different opcode
4660 so that data characters > 255 can be handled correctly.
4661
4662 If the class contains characters outside the 0-255 range, a different
4663 opcode is compiled. It may optionally have a bit map for characters < 256,
4664 but those above are are explicitly listed afterwards. A flag byte tells
4665 whether the bitmap is present, and whether this is a negated class or not.
4666
4667 In JavaScript compatibility mode, an isolated ']' causes an error. In
4668 default (Perl) mode, it is treated as a data character. */
4669
4670 case CHAR_RIGHT_SQUARE_BRACKET:
4671 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4672 {
4673 *errorcodeptr = ERR64;
4674 goto FAILED;
4675 }
4676 goto NORMAL_CHAR;
4677
4678 case CHAR_LEFT_SQUARE_BRACKET:
4679 previous = code;
4680
4681 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4682 they are encountered at the top level, so we'll do that too. */
4683
4684 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4685 ptr[1] == CHAR_EQUALS_SIGN) &&
4686 check_posix_syntax(ptr, &tempptr))
4687 {
4688 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4689 goto FAILED;
4690 }
4691
4692 /* If the first character is '^', set the negation flag and skip it. Also,
4693 if the first few characters (either before or after ^) are \Q\E or \E we
4694 skip them too. This makes for compatibility with Perl. */
4695
4696 negate_class = FALSE;
4697 for (;;)
4698 {
4699 c = *(++ptr);
4700 if (c == CHAR_BACKSLASH)
4701 {
4702 if (ptr[1] == CHAR_E)
4703 ptr++;
4704 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4705 ptr += 3;
4706 else
4707 break;
4708 }
4709 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4710 negate_class = TRUE;
4711 else break;
4712 }
4713
4714 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4715 an initial ']' is taken as a data character -- the code below handles
4716 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4717 [^] must match any character, so generate OP_ALLANY. */
4718
4719 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4720 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4721 {
4722 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4723 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4724 zerofirstchar = firstchar;
4725 zerofirstcharflags = firstcharflags;
4726 break;
4727 }
4728
4729 /* If a class contains a negative special such as \S, we need to flip the
4730 negation flag at the end, so that support for characters > 255 works
4731 correctly (they are all included in the class). */
4732
4733 should_flip_negation = FALSE;
4734
4735 /* For optimization purposes, we track some properties of the class:
4736 class_has_8bitchar will be non-zero if the class contains at least one <
4737 256 character; class_one_char will be 1 if the class contains just one
4738 character. */
4739
4740 class_has_8bitchar = 0;
4741 class_one_char = 0;
4742
4743 /* Initialize the 32-char bit map to all zeros. We build the map in a
4744 temporary bit of memory, in case the class contains fewer than two
4745 8-bit characters because in that case the compiled code doesn't use the bit
4746 map. */
4747
4748 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4749
4750 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4751 xclass = FALSE;
4752 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4753 class_uchardata_base = class_uchardata; /* Save the start */
4754 #endif
4755
4756 /* Process characters until ] is reached. By writing this as a "do" it
4757 means that an initial ] is taken as a data character. At the start of the
4758 loop, c contains the first byte of the character. */
4759
4760 if (c != CHAR_NULL) do
4761 {
4762 const pcre_uchar *oldptr;
4763
4764 #ifdef SUPPORT_UTF
4765 if (utf && HAS_EXTRALEN(c))
4766 { /* Braces are required because the */
4767 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4768 }
4769 #endif
4770
4771 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4772 /* In the pre-compile phase, accumulate the length of any extra
4773 data and reset the pointer. This is so that very large classes that
4774 contain a zillion > 255 characters no longer overwrite the work space
4775 (which is on the stack). We have to remember that there was XCLASS data,
4776 however. */
4777
4778 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4779 {
4780 xclass = TRUE;
4781 *lengthptr += class_uchardata - class_uchardata_base;
4782 class_uchardata = class_uchardata_base;
4783 }
4784 #endif
4785
4786 /* Inside \Q...\E everything is literal except \E */
4787
4788 if (inescq)
4789 {
4790 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4791 {
4792 inescq = FALSE; /* Reset literal state */
4793 ptr++; /* Skip the 'E' */
4794 continue; /* Carry on with next */
4795 }
4796 goto CHECK_RANGE; /* Could be range if \E follows */
4797 }
4798
4799 /* Handle POSIX class names. Perl allows a negation extension of the
4800 form [:^name:]. A square bracket that doesn't match the syntax is
4801 treated as a literal. We also recognize the POSIX constructions
4802 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4803 5.6 and 5.8 do. */
4804
4805 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4806 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4807 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4808 {
4809 BOOL local_negate = FALSE;
4810 int posix_class, taboffset, tabopt;
4811 register const pcre_uint8 *cbits = cd->cbits;
4812 pcre_uint8 pbits[32];
4813
4814 if (ptr[1] != CHAR_COLON)
4815 {
4816 *errorcodeptr = ERR31;
4817 goto FAILED;
4818 }
4819
4820 ptr += 2;
4821 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4822 {
4823 local_negate = TRUE;
4824 should_flip_negation = TRUE; /* Note negative special */
4825 ptr++;
4826 }
4827
4828 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4829 if (posix_class < 0)
4830 {
4831 *errorcodeptr = ERR30;
4832 goto FAILED;
4833 }
4834
4835 /* If matching is caseless, upper and lower are converted to
4836 alpha. This relies on the fact that the class table starts with
4837 alpha, lower, upper as the first 3 entries. */
4838
4839 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4840 posix_class = 0;
4841
4842 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4843 different escape sequences that use Unicode properties \p or \P. Others
4844 that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4845 directly. */
4846
4847 #ifdef SUPPORT_UCP
4848 if ((options & PCRE_UCP) != 0)
4849 {
4850 unsigned int ptype = 0;
4851 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4852
4853 /* The posix_substitutes table specifies which POSIX classes can be
4854 converted to \p or \P items. */
4855
4856 if (posix_substitutes[pc] != NULL)
4857 {
4858 nestptr = tempptr + 1;
4859 ptr = posix_substitutes[pc] - 1;
4860 continue;
4861 }
4862
4863 /* There are three other classes that generate special property calls
4864 that are recognized only in an XCLASS. */
4865
4866 else switch(posix_class)
4867 {
4868 case PC_GRAPH:
4869 ptype = PT_PXGRAPH;
4870 /* Fall through */
4871 case PC_PRINT:
4872 if (ptype == 0) ptype = PT_PXPRINT;
4873 /* Fall through */
4874 case PC_PUNCT:
4875 if (ptype == 0) ptype = PT_PXPUNCT;
4876 *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4877 *class_uchardata++ = ptype;
4878 *class_uchardata++ = 0;
4879 ptr = tempptr + 1;
4880 continue;
4881
4882 /* For all other POSIX classes, no special action is taken in UCP
4883 mode. Fall through to the non_UCP case. */
4884
4885 default:
4886 break;
4887 }
4888 }
4889 #endif
4890 /* In the non-UCP case, or when UCP makes no difference, we build the
4891 bit map for the POSIX class in a chunk of local store because we may be
4892 adding and subtracting from it, and we don't want to subtract bits that
4893 may be in the main map already. At the end we or the result into the
4894 bit map that is being built. */
4895
4896 posix_class *= 3;
4897
4898 /* Copy in the first table (always present) */
4899
4900 memcpy(pbits, cbits + posix_class_maps[posix_class],
4901 32 * sizeof(pcre_uint8));
4902
4903 /* If there is a second table, add or remove it as required. */
4904
4905 taboffset = posix_class_maps[posix_class + 1];
4906 tabopt = posix_class_maps[posix_class + 2];
4907
4908 if (taboffset >= 0)
4909 {
4910 if (tabopt >= 0)
4911 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
4912 else
4913 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
4914 }
4915
4916 /* Now see if we need to remove any special characters. An option
4917 value of 1 removes vertical space and 2 removes underscore. */
4918
4919 if (tabopt < 0) tabopt = -tabopt;
4920 if (tabopt == 1) pbits[1] &= ~0x3c;
4921 else if (tabopt == 2) pbits[11] &= 0x7f;
4922
4923 /* Add the POSIX table or its complement into the main table that is
4924 being built and we are done. */
4925
4926 if (local_negate)
4927 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
4928 else
4929 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
4930
4931 ptr = tempptr + 1;
4932 /* Every class contains at least one < 256 character. */
4933 class_has_8bitchar = 1;
4934 /* Every class contains at least two characters. */
4935 class_one_char = 2;
4936 continue; /* End of POSIX syntax handling */
4937 }
4938
4939 /* Backslash may introduce a single character, or it may introduce one
4940 of the specials, which just set a flag. The sequence \b is a special
4941 case. Inside a class (and only there) it is treated as backspace. We
4942 assume that other escapes have more than one character in them, so
4943 speculatively set both class_has_8bitchar and class_one_char bigger
4944 than one. Unrecognized escapes fall through and are either treated
4945 as literal characters (by default), or are faulted if
4946 PCRE_EXTRA is set. */
4947
4948 if (c == CHAR_BACKSLASH)
4949 {
4950 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4951 TRUE);
4952 if (*errorcodeptr != 0) goto FAILED;
4953 if (escape == 0) c = ec;
4954 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
4955 else if (escape == ESC_N) /* \N is not supported in a class */
4956 {
4957 *errorcodeptr = ERR71;
4958 goto FAILED;
4959 }
4960 else if (escape == ESC_Q) /* Handle start of quoted string */
4961 {
4962 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4963 {
4964 ptr += 2; /* avoid empty string */
4965 }
4966 else inescq = TRUE;
4967 continue;
4968 }
4969 else if (escape == ESC_E) continue; /* Ignore orphan \E */
4970
4971 else
4972 {
4973 register const pcre_uint8 *cbits = cd->cbits;
4974 /* Every class contains at least two < 256 characters. */
4975 class_has_8bitchar++;
4976 /* Every class contains at least two characters. */
4977 class_one_char += 2;
4978
4979 switch (escape)
4980 {
4981 #ifdef SUPPORT_UCP
4982 case ESC_du: /* These are the values given for \d etc */
4983 case ESC_DU: /* when PCRE_UCP is set. We replace the */
4984 case ESC_wu: /* escape sequence with an appropriate \p */
4985 case ESC_WU: /* or \P to test Unicode properties instead */
4986 case ESC_su: /* of the default ASCII testing. */
4987 case ESC_SU:
4988 nestptr = ptr;
4989 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
4990 class_has_8bitchar--; /* Undo! */
4991 continue;
4992 #endif
4993 case ESC_d:
4994 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
4995 continue;
4996
4997 case ESC_D:
4998 should_flip_negation = TRUE;
4999 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
5000 continue;
5001
5002 case ESC_w:
5003 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
5004 continue;
5005
5006 case ESC_W:
5007 should_flip_negation = TRUE;
5008 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5009 continue;
5010
5011 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5012 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5013 previously set by something earlier in the character class.
5014 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5015 we could just adjust the appropriate bit. From PCRE 8.34 we no
5016 longer treat \s and \S specially. */
5017
5018 case ESC_s:
5019 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
5020 continue;
5021
5022 case ESC_S:
5023 should_flip_negation = TRUE;
5024 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
5025 continue;
5026
5027 /* The rest apply in both UCP and non-UCP cases. */
5028
5029 case ESC_h:
5030 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5031 PRIV(hspace_list), NOTACHAR);
5032 continue;
5033
5034 case ESC_H:
5035 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5036 cd, PRIV(hspace_list));
5037 continue;
5038
5039 case ESC_v:
5040 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5041 PRIV(vspace_list), NOTACHAR);
5042 continue;
5043
5044 case ESC_V:
5045 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5046 cd, PRIV(vspace_list));
5047 continue;
5048
5049 #ifdef SUPPORT_UCP
5050 case ESC_p:
5051 case ESC_P:
5052 {
5053 BOOL negated;
5054 unsigned int ptype = 0, pdata = 0;
5055 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5056 goto FAILED;
5057 *class_uchardata++ = ((escape == ESC_p) != negated)?
5058 XCL_PROP : XCL_NOTPROP;
5059 *class_uchardata++ = ptype;
5060 *class_uchardata++ = pdata;
5061 class_has_8bitchar--; /* Undo! */
5062 continue;
5063 }
5064 #endif
5065 /* Unrecognized escapes are faulted if PCRE is running in its
5066 strict mode. By default, for compatibility with Perl, they are
5067 treated as literals. */
5068
5069 default:
5070 if ((options & PCRE_EXTRA) != 0)
5071 {
5072 *errorcodeptr = ERR7;
5073 goto FAILED;
5074 }
5075 class_has_8bitchar--; /* Undo the speculative increase. */
5076 class_one_char -= 2; /* Undo the speculative increase. */
5077 c = *ptr; /* Get the final character and fall through */
5078 break;
5079 }
5080 }
5081
5082 /* Fall through if the escape just defined a single character (c >= 0).
5083 This may be greater than 256. */
5084
5085 escape = 0;
5086
5087 } /* End of backslash handling */
5088
5089 /* A character may be followed by '-' to form a range. However, Perl does
5090 not permit ']' to be the end of the range. A '-' character at the end is
5091 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5092 code for handling \Q and \E is messy. */
5093
5094 CHECK_RANGE:
5095 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5096 {
5097 inescq = FALSE;
5098 ptr += 2;
5099 }
5100 oldptr = ptr;
5101
5102 /* Remember if \r or \n were explicitly used */
5103
5104 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5105
5106 /* Check for range */
5107
5108 if (!inescq && ptr[1] == CHAR_MINUS)
5109 {
5110 pcre_uint32 d;
5111 ptr += 2;
5112 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5113
5114 /* If we hit \Q (not followed by \E) at this point, go into escaped
5115 mode. */
5116
5117 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5118 {
5119 ptr += 2;
5120 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5121 { ptr += 2; continue; }
5122 inescq = TRUE;
5123 break;
5124 }
5125
5126 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5127 back the pointer and jump to handle the character that preceded it. */
5128
5129 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5130 {
5131 ptr = oldptr;
5132 goto CLASS_SINGLE_CHARACTER;
5133 }
5134
5135 /* Otherwise, we have a potential range; pick up the next character */
5136
5137 #ifdef SUPPORT_UTF
5138 if (utf)
5139 { /* Braces are required because the */
5140 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5141 }
5142 else
5143 #endif
5144 d = *ptr; /* Not UTF-8 mode */
5145
5146 /* The second part of a range can be a single-character escape, but
5147 not any of the other escapes. Perl 5.6 treats a hyphen as a literal
5148 in such circumstances. */
5149
5150 if (!inescq && d == CHAR_BACKSLASH)
5151 {
5152 int descape;
5153 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5154 if (*errorcodeptr != 0) goto FAILED;
5155
5156 /* \b is backspace; any other special means the '-' was literal. */
5157
5158 if (descape != 0)
5159 {
5160 if (descape == ESC_b) d = CHAR_BS; else
5161 {
5162 ptr = oldptr;
5163 goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5164 }
5165 }
5166 }
5167
5168 /* Check that the two values are in the correct order. Optimize
5169 one-character ranges. */
5170
5171 if (d < c)
5172 {
5173 *errorcodeptr = ERR8;
5174 goto FAILED;
5175 }
5176 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5177
5178 /* We have found a character range, so single character optimizations
5179 cannot be done anymore. Any value greater than 1 indicates that there
5180 is more than one character. */
5181
5182 class_one_char = 2;
5183
5184 /* Remember an explicit \r or \n, and add the range to the class. */
5185
5186 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5187
5188 class_has_8bitchar +=
5189 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5190
5191 continue; /* Go get the next char in the class */
5192 }
5193
5194 /* Handle a single character - we can get here for a normal non-escape
5195 char, or after \ that introduces a single character or for an apparent
5196 range that isn't. Only the value 1 matters for class_one_char, so don't
5197 increase it if it is already 2 or more ... just in case there's a class
5198 with a zillion characters in it. */
5199
5200 CLASS_SINGLE_CHARACTER:
5201 if (class_one_char < 2) class_one_char++;
5202
5203 /* If class_one_char is 1, we have the first single character in the
5204 class, and there have been no prior ranges, or XCLASS items generated by
5205 escapes. If this is the final character in the class, we can optimize by
5206 turning the item into a 1-character OP_CHAR[I] if it's positive, or
5207 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5208 to be set. Otherwise, there can be no first char if this item is first,
5209 whatever repeat count may follow. In the case of reqchar, save the
5210 previous value for reinstating. */
5211
5212 if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5213 {
5214 ptr++;
5215 zeroreqchar = reqchar;
5216 zeroreqcharflags = reqcharflags;
5217
5218 if (negate_class)
5219 {
5220 #ifdef SUPPORT_UCP
5221 int d;
5222 #endif
5223 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5224 zerofirstchar = firstchar;
5225 zerofirstcharflags = firstcharflags;
5226
5227 /* For caseless UTF-8 mode when UCP support is available, check
5228 whether this character has more than one other case. If so, generate
5229 a special OP_NOTPROP item instead of OP_NOTI. */
5230
5231 #ifdef SUPPORT_UCP
5232 if (utf && (options & PCRE_CASELESS) != 0 &&
5233 (d = UCD_CASESET(c)) != 0)
5234 {
5235 *code++ = OP_NOTPROP;
5236 *code++ = PT_CLIST;
5237 *code++ = d;
5238 }
5239 else
5240 #endif
5241 /* Char has only one other case, or UCP not available */
5242
5243 {
5244 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5245 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5246 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5247 code += PRIV(ord2utf)(c, code);
5248 else
5249 #endif
5250 *code++ = c;
5251 }
5252
5253 /* We are finished with this character class */
5254
5255 goto END_CLASS;
5256 }
5257
5258 /* For a single, positive character, get the value into mcbuffer, and
5259 then we can handle this with the normal one-character code. */
5260
5261 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5262 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5263 mclength = PRIV(ord2utf)(c, mcbuffer);
5264 else
5265 #endif
5266 {
5267 mcbuffer[0] = c;
5268 mclength = 1;
5269 }
5270 goto ONE_CHAR;
5271 } /* End of 1-char optimization */
5272
5273 /* There is more than one character in the class, or an XCLASS item
5274 has been generated. Add this character to the class. */
5275
5276 class_has_8bitchar +=
5277 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5278 }
5279
5280 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5281 If we are at the end of an internal nested string, revert to the outer
5282 string. */
5283
5284 while (((c = *(++ptr)) != CHAR_NULL ||
5285 (nestptr != NULL &&
5286 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5287 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5288
5289 /* Check for missing terminating ']' */
5290
5291 if (c == CHAR_NULL)
5292 {
5293 *errorcodeptr = ERR6;
5294 goto FAILED;
5295 }
5296
5297 /* We will need an XCLASS if data has been placed in class_uchardata. In
5298 the second phase this is a sufficient test. However, in the pre-compile
5299 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5300 only if the very last character in the class needs XCLASS will it contain
5301 anything at this point. For this reason, xclass gets set TRUE above when
5302 uchar_classdata is emptied, and that's why this code is the way it is here
5303 instead of just doing a test on class_uchardata below. */
5304
5305 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5306 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5307 #endif
5308
5309 /* If this is the first thing in the branch, there can be no first char
5310 setting, whatever the repeat count. Any reqchar setting must remain
5311 unchanged after any kind of repeat. */
5312
5313 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5314 zerofirstchar = firstchar;
5315 zerofirstcharflags = firstcharflags;
5316 zeroreqchar = reqchar;
5317 zeroreqcharflags = reqcharflags;
5318
5319 /* If there are characters with values > 255, we have to compile an
5320 extended class, with its own opcode, unless there was a negated special
5321 such as \S in the class, and PCRE_UCP is not set, because in that case all
5322 characters > 255 are in the class, so any that were explicitly given as
5323 well can be ignored. If (when there are explicit characters > 255 that must
5324 be listed) there are no characters < 256, we can omit the bitmap in the
5325 actual compiled code. */
5326
5327 #ifdef SUPPORT_UTF
5328 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5329 #elif !defined COMPILE_PCRE8
5330 if (xclass && !should_flip_negation)
5331 #endif
5332 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5333 {
5334 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5335 *code++ = OP_XCLASS;
5336 code += LINK_SIZE;
5337 *code = negate_class? XCL_NOT:0;
5338
5339 /* If the map is required, move up the extra data to make room for it;
5340 otherwise just move the code pointer to the end of the extra data. */
5341
5342 if (class_has_8bitchar > 0)
5343 {
5344 *code++ |= XCL_MAP;
5345 memmove(code + (32 / sizeof(pcre_uchar)), code,
5346 IN_UCHARS(class_uchardata - code));
5347 memcpy(code, classbits, 32);
5348 code = class_uchardata + (32 / sizeof(pcre_uchar));
5349 }
5350 else code = class_uchardata;
5351
5352 /* Now fill in the complete length of the item */
5353
5354 PUT(previous, 1, (int)(code - previous));
5355 break; /* End of class handling */
5356 }
5357 #endif
5358
5359 /* If there are no characters > 255, or they are all to be included or
5360 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5361 whole class was negated and whether there were negative specials such as \S
5362 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5363 negating it if necessary. */
5364
5365 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5366 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5367 {
5368 if (negate_class)
5369 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5370 memcpy(code, classbits, 32);
5371 }
5372 code += 32 / sizeof(pcre_uchar);
5373
5374 END_CLASS:
5375 break;
5376
5377
5378 /* ===================================================================*/
5379 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5380 has been tested above. */
5381
5382 case CHAR_LEFT_CURLY_BRACKET:
5383 if (!is_quantifier) goto NORMAL_CHAR;
5384 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5385 if (*errorcodeptr != 0) goto FAILED;
5386 goto REPEAT;
5387
5388 case CHAR_ASTERISK:
5389 repeat_min = 0;
5390 repeat_max = -1;
5391 goto REPEAT;
5392
5393 case CHAR_PLUS:
5394 repeat_min = 1;
5395 repeat_max = -1;
5396 goto REPEAT;
5397
5398 case CHAR_QUESTION_MARK:
5399 repeat_min = 0;
5400 repeat_max = 1;
5401
5402 REPEAT:
5403 if (previous == NULL)
5404 {
5405 *errorcodeptr = ERR9;
5406 goto FAILED;
5407 }
5408
5409 if (repeat_min == 0)
5410 {
5411 firstchar = zerofirstchar; /* Adjust for zero repeat */
5412 firstcharflags = zerofirstcharflags;
5413 reqchar = zeroreqchar; /* Ditto */
5414 reqcharflags = zeroreqcharflags;
5415 }
5416
5417 /* Remember whether this is a variable length repeat */
5418
5419 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5420
5421 op_type = 0; /* Default single-char op codes */
5422 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5423
5424 /* Save start of previous item, in case we have to move it up in order to
5425 insert something before it. */
5426
5427 tempcode = previous;
5428
5429 /* If the next character is '+', we have a possessive quantifier. This
5430 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5431 If the next character is '?' this is a minimizing repeat, by default,
5432 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5433 repeat type to the non-default. */
5434
5435 if (ptr[1] == CHAR_PLUS)
5436 {
5437 repeat_type = 0; /* Force greedy */
5438 possessive_quantifier = TRUE;
5439 ptr++;
5440 }
5441 else if (ptr[1] == CHAR_QUESTION_MARK)
5442 {
5443 repeat_type = greedy_non_default;
5444 ptr++;
5445 }
5446 else repeat_type = greedy_default;
5447
5448 /* If previous was a recursion call, wrap it in atomic brackets so that
5449 previous becomes the atomic group. All recursions were so wrapped in the
5450 past, but it no longer happens for non-repeated recursions. In fact, the
5451 repeated ones could be re-implemented independently so as not to need this,
5452 but for the moment we rely on the code for repeating groups. */
5453
5454 if (*previous == OP_RECURSE)
5455 {
5456 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5457 *previous = OP_ONCE;
5458 PUT(previous, 1, 2 + 2*LINK_SIZE);
5459 previous[2 + 2*LINK_SIZE] = OP_KET;
5460 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5461 code += 2 + 2 * LINK_SIZE;
5462 length_prevgroup = 3 + 3*LINK_SIZE;
5463
5464 /* When actually compiling, we need to check whether this was a forward
5465 reference, and if so, adjust the offset. */
5466
5467 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5468 {
5469 int offset = GET(cd->hwm, -LINK_SIZE);
5470 if (offset == previous + 1 - cd->start_code)
5471 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5472 }
5473 }
5474
5475 /* Now handle repetition for the different types of item. */
5476
5477 /* If previous was a character or negated character match, abolish the item
5478 and generate a repeat item instead. If a char item has a minimum of more
5479 than one, ensure that it is set in reqchar - it might not be if a sequence
5480 such as x{3} is the first thing in a branch because the x will have gone
5481 into firstchar instead. */
5482
5483 if (*previous == OP_CHAR || *previous == OP_CHARI
5484 || *previous == OP_NOT || *previous == OP_NOTI)
5485 {
5486 switch (*previous)
5487 {
5488 default: /* Make compiler happy. */
5489 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5490 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5491 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5492 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5493 }
5494
5495 /* Deal with UTF characters that take up more than one character. It's
5496 easier to write this out separately than try to macrify it. Use c to
5497 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5498 it's a length rather than a small character. */
5499
5500 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5501 if (utf && NOT_FIRSTCHAR(code[-1]))
5502 {
5503 pcre_uchar *lastchar = code - 1;
5504 BACKCHAR(lastchar);
5505 c = (int)(code - lastchar); /* Length of UTF-8 character */
5506 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5507 c |= UTF_LENGTH; /* Flag c as a length */
5508 }
5509 else
5510 #endif /* SUPPORT_UTF */
5511
5512 /* Handle the case of a single charater - either with no UTF support, or
5513 with UTF disabled, or for a single character UTF character. */
5514 {
5515 c = code[-1];
5516 if (*previous <= OP_CHARI && repeat_min > 1)
5517 {
5518 reqchar = c;
5519 reqcharflags = req_caseopt | cd->req_varyopt;
5520 }
5521 }
5522
5523 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5524 }
5525
5526 /* If previous was a character type match (\d or similar), abolish it and
5527 create a suitable repeat item. The code is shared with single-character
5528 repeats by setting op_type to add a suitable offset into repeat_type. Note
5529 the the Unicode property types will be present only when SUPPORT_UCP is
5530 defined, but we don't wrap the little bits of code here because it just
5531 makes it horribly messy. */
5532
5533 else if (*previous < OP_EODN)
5534 {
5535 pcre_uchar *oldcode;
5536 int prop_type, prop_value;
5537 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5538 c = *previous;
5539
5540 OUTPUT_SINGLE_REPEAT:
5541 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5542 {
5543 prop_type = previous[1];
5544 prop_value = previous[2];
5545 }
5546 else prop_type = prop_value = -1;
5547
5548 oldcode = code;
5549 code = previous; /* Usually overwrite previous item */
5550
5551 /* If the maximum is zero then the minimum must also be zero; Perl allows
5552 this case, so we do too - by simply omitting the item altogether. */
5553
5554 if (repeat_max == 0) goto END_REPEAT;
5555
5556 /* Combine the op_type with the repeat_type */
5557
5558 repeat_type += op_type;
5559
5560 /* A minimum of zero is handled either as the special case * or ?, or as
5561 an UPTO, with the maximum given. */
5562
5563 if (repeat_min == 0)
5564 {
5565 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5566 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5567 else
5568 {
5569 *code++ = OP_UPTO + repeat_type;
5570 PUT2INC(code, 0, repeat_max);
5571 }
5572 }
5573
5574 /* A repeat minimum of 1 is optimized into some special cases. If the
5575 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5576 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5577 one less than the maximum. */
5578
5579 else if (repeat_min == 1)
5580 {
5581 if (repeat_max == -1)
5582 *code++ = OP_PLUS + repeat_type;
5583 else
5584 {
5585 code = oldcode; /* leave previous item in place */
5586 if (repeat_max == 1) goto END_REPEAT;
5587 *code++ = OP_UPTO + repeat_type;
5588 PUT2INC(code, 0, repeat_max - 1);
5589 }
5590 }
5591
5592 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5593 handled as an EXACT followed by an UPTO. */
5594
5595 else
5596 {
5597 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5598 PUT2INC(code, 0, repeat_min);
5599
5600 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5601 we have to insert the character for the previous code. For a repeated
5602 Unicode property match, there are two extra bytes that define the
5603 required property. In UTF-8 mode, long characters have their length in
5604 c, with the UTF_LENGTH bit as a flag. */
5605
5606 if (repeat_max < 0)
5607 {
5608 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5609 if (utf && (c & UTF_LENGTH) != 0)
5610 {
5611 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5612 code += c & 7;
5613 }
5614 else
5615 #endif
5616 {
5617 *code++ = c;
5618 if (prop_type >= 0)
5619 {
5620 *code++ = prop_type;
5621 *code++ = prop_value;
5622 }
5623 }
5624 *code++ = OP_STAR + repeat_type;
5625 }
5626
5627 /* Else insert an UPTO if the max is greater than the min, again
5628 preceded by the character, for the previously inserted code. If the
5629 UPTO is just for 1 instance, we can use QUERY instead. */
5630
5631 else if (repeat_max != repeat_min)
5632 {
5633 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5634 if (utf && (c & UTF_LENGTH) != 0)
5635 {
5636 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5637 code += c & 7;
5638 }
5639 else
5640 #endif
5641 *code++ = c;
5642 if (prop_type >= 0)
5643 {
5644 *code++ = prop_type;
5645 *code++ = prop_value;
5646 }
5647 repeat_max -= repeat_min;
5648
5649 if (repeat_max == 1)
5650 {
5651 *code++ = OP_QUERY + repeat_type;
5652 }
5653 else
5654 {
5655 *code++ = OP_UPTO + repeat_type;
5656 PUT2INC(code, 0, repeat_max);
5657 }
5658 }
5659 }
5660
5661 /* The character or character type itself comes last in all cases. */
5662
5663 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5664 if (utf && (c & UTF_LENGTH) != 0)
5665 {
5666 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5667 code += c & 7;
5668 }
5669 else
5670 #endif
5671 *code++ = c;
5672
5673 /* For a repeated Unicode property match, there are two extra bytes that
5674 define the required property. */
5675
5676 #ifdef SUPPORT_UCP
5677 if (prop_type >= 0)
5678 {
5679 *code++ = prop_type;
5680 *code++ = prop_value;
5681 }
5682 #endif
5683 }
5684
5685 /* If previous was a character class or a back reference, we put the repeat
5686 stuff after it, but just skip the item if the repeat was {0,0}. */
5687
5688 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5689 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5690 *previous == OP_XCLASS ||
5691 #endif
5692 *previous == OP_REF || *previous == OP_REFI ||
5693 *previous == OP_DNREF || *previous == OP_DNREFI)
5694 {
5695 if (repeat_max == 0)
5696 {
5697 code = previous;
5698 goto END_REPEAT;
5699 }
5700
5701 if (repeat_min == 0 && repeat_max == -1)
5702 *code++ = OP_CRSTAR + repeat_type;
5703 else if (repeat_min == 1 && repeat_max == -1)
5704 *code++ = OP_CRPLUS + repeat_type;
5705 else if (repeat_min == 0 && repeat_max == 1)
5706 *code++ = OP_CRQUERY + repeat_type;
5707 else
5708 {
5709 *code++ = OP_CRRANGE + repeat_type;
5710 PUT2INC(code, 0, repeat_min);
5711 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5712 PUT2INC(code, 0, repeat_max);
5713 }
5714 }
5715
5716 /* If previous was a bracket group, we may have to replicate it in certain
5717 cases. Note that at this point we can encounter only the "basic" bracket
5718 opcodes such as BRA and CBRA, as this is the place where they get converted
5719 into the more special varieties such as BRAPOS and SBRA. A test for >=
5720 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5721 ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
5722 repetition of assertions, but now it does, for Perl compatibility. */
5723
5724 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5725 {
5726 register int i;
5727 int len = (int)(code - previous);
5728 pcre_uchar *bralink = NULL;
5729 pcre_uchar *brazeroptr = NULL;
5730
5731 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5732 we just ignore the repeat. */
5733
5734 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5735 goto END_REPEAT;
5736
5737 /* There is no sense in actually repeating assertions. The only potential
5738 use of repetition is in cases when the assertion is optional. Therefore,
5739 if the minimum is greater than zero, just ignore the repeat. If the
5740 maximum is not not zero or one, set it to 1. */
5741
5742 if (*previous < OP_ONCE) /* Assertion */
5743 {
5744 if (repeat_min > 0) goto END_REPEAT;
5745 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5746 }
5747
5748 /* The case of a zero minimum is special because of the need to stick
5749 OP_BRAZERO in front of it, and because the group appears once in the
5750 data, whereas in other cases it appears the minimum number of times. For
5751 this reason, it is simplest to treat this case separately, as otherwise
5752 the code gets far too messy. There are several special subcases when the
5753 minimum is zero. */
5754
5755 if (repeat_min == 0)
5756 {
5757 /* If the maximum is also zero, we used to just omit the group from the
5758 output altogether, like this:
5759
5760 ** if (repeat_max == 0)
5761 ** {
5762 ** code = previous;
5763 ** goto END_REPEAT;
5764 ** }
5765
5766 However, that fails when a group or a subgroup within it is referenced
5767 as a subroutine from elsewhere in the pattern, so now we stick in
5768 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5769 don't have a list of which groups are referenced, we cannot do this
5770 selectively.
5771
5772 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5773 and do no more at this point. However, we do need to adjust any
5774 OP_RECURSE calls inside the group that refer to the group itself or any
5775 internal or forward referenced group, because the offset is from the
5776 start of the whole regex. Temporarily terminate the pattern while doing
5777 this. */
5778
5779 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5780 {
5781 *code = OP_END;
5782 adjust_recurse(previous, 1, utf, cd, save_hwm);
5783 memmove(previous + 1, previous, IN_UCHARS(len));
5784 code++;
5785 if (repeat_max == 0)
5786 {
5787 *previous++ = OP_SKIPZERO;
5788 goto END_REPEAT;
5789 }
5790 brazeroptr = previous; /* Save for possessive optimizing */
5791 *previous++ = OP_BRAZERO + repeat_type;
5792 }
5793
5794 /* If the maximum is greater than 1 and limited, we have to replicate
5795 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
5796 The first one has to be handled carefully because it's the original
5797 copy, which has to be moved up. The remainder can be handled by code
5798 that is common with the non-zero minimum case below. We have to
5799 adjust the value or repeat_max, since one less copy is required. Once
5800 again, we may have to adjust any OP_RECURSE calls inside the group. */
5801
5802 else
5803 {
5804 int offset;
5805 *code = OP_END;
5806 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5807 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5808 code += 2 + LINK_SIZE;
5809 *previous++ = OP_BRAZERO + repeat_type;
5810 *previous++ = OP_BRA;
5811
5812 /* We chain together the bracket offset fields that have to be
5813 filled in later when the ends of the brackets are reached. */
5814
5815 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
5816 bralink = previous;
5817 PUTINC(previous, 0, offset);
5818 }
5819
5820 repeat_max--;
5821 }
5822
5823 /* If the minimum is greater than zero, replicate the group as many
5824 times as necessary, and adjust the maximum to the number of subsequent
5825 copies that we need. If we set a first char from the group, and didn't
5826 set a required char, copy the latter from the former. If there are any
5827 forward reference subroutine calls in the group, there will be entries on
5828 the workspace list; replicate these with an appropriate increment. */
5829
5830 else
5831 {
5832 if (repeat_min > 1)
5833 {
5834 /* In the pre-compile phase, we don't actually do the replication. We
5835 just adjust the length as if we had. Do some paranoid checks for
5836 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
5837 integer type when available, otherwise double. */
5838
5839 if (lengthptr != NULL)
5840 {
5841 int delta = (repeat_min - 1)*length_prevgroup;
5842 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
5843 (INT64_OR_DOUBLE)length_prevgroup >
5844 (INT64_OR_DOUBLE)INT_MAX ||
5845 OFLOW_MAX - *lengthptr < delta)
5846 {
5847 *errorcodeptr = ERR20;
5848 goto FAILED;
5849 }
5850 *lengthptr += delta;
5851 }
5852
5853 /* This is compiling for real. If there is a set first byte for
5854 the group, and we have not yet set a "required byte", set it. Make
5855 sure there is enough workspace for copying forward references before
5856 doing the copy. */
5857
5858 else
5859 {
5860 if (groupsetfirstchar && reqcharflags < 0)
5861 {
5862 reqchar = firstchar;
5863 reqcharflags = firstcharflags;
5864 }
5865
5866 for (i = 1; i < repeat_min; i++)
5867 {
5868 pcre_uchar *hc;
5869 pcre_uchar *this_hwm = cd->hwm;
5870 memcpy(code, previous, IN_UCHARS(len));
5871
5872 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5873 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5874 {
5875 int save_offset = save_hwm - cd->start_workspace;
5876 int this_offset = this_hwm - cd->start_workspace;
5877 *errorcodeptr = expand_workspace(cd);
5878 if (*errorcodeptr != 0) goto FAILED;
5879 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5880 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5881 }
5882
5883 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5884 {
5885 PUT(cd->hwm, 0, GET(hc, 0) + len);
5886 cd->hwm += LINK_SIZE;
5887 }
5888 save_hwm = this_hwm;
5889 code += len;
5890 }
5891 }
5892 }
5893
5894 if (repeat_max > 0) repeat_max -= repeat_min;
5895 }
5896
5897 /* This code is common to both the zero and non-zero minimum cases. If
5898 the maximum is limited, it replicates the group in a nested fashion,
5899 remembering the bracket starts on a stack. In the case of a zero minimum,
5900 the first one was set up above. In all cases the repeat_max now specifies
5901 the number of additional copies needed. Again, we must remember to
5902 replicate entries on the forward reference list. */
5903
5904 if (repeat_max >= 0)
5905 {
5906 /* In the pre-compile phase, we don't actually do the replication. We
5907 just adjust the length as if we had. For each repetition we must add 1
5908 to the length for BRAZERO and for all but the last repetition we must
5909 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
5910 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
5911 a 64-bit integer type when available, otherwise double. */
5912
5913 if (lengthptr != NULL && repeat_max > 0)
5914 {
5915 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
5916 2 - 2*LINK_SIZE; /* Last one doesn't nest */
5917 if ((INT64_OR_DOUBLE)repeat_max *
5918 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
5919 > (INT64_OR_DOUBLE)INT_MAX ||
5920 OFLOW_MAX - *lengthptr < delta)
5921 {
5922 *errorcodeptr = ERR20;
5923 goto FAILED;
5924 }
5925 *lengthptr += delta;
5926 }
5927
5928 /* This is compiling for real */
5929
5930 else for (i = repeat_max - 1; i >= 0; i--)
5931 {
5932 pcre_uchar *hc;
5933 pcre_uchar *this_hwm = cd->hwm;
5934
5935 *code++ = OP_BRAZERO + repeat_type;
5936
5937 /* All but the final copy start a new nesting, maintaining the
5938 chain of brackets outstanding. */
5939
5940 if (i != 0)
5941 {
5942 int offset;
5943 *code++ = OP_BRA;
5944 offset = (bralink == NULL)? 0 : (int)(code - bralink);
5945 bralink = code;
5946 PUTINC(code, 0, offset);
5947 }
5948
5949 memcpy(code, previous, IN_UCHARS(len));
5950
5951 /* Ensure there is enough workspace for forward references before
5952 copying them. */
5953
5954 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5955 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5956 {
5957 int save_offset = save_hwm - cd->start_workspace;
5958 int this_offset = this_hwm - cd->start_workspace;
5959 *errorcodeptr = expand_workspace(cd);
5960 if (*errorcodeptr != 0) goto FAILED;
5961 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5962 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5963 }
5964
5965 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5966 {
5967 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
5968 cd->hwm += LINK_SIZE;
5969 }
5970 save_hwm = this_hwm;
5971 code += len;
5972 }
5973
5974 /* Now chain through the pending brackets, and fill in their length
5975 fields (which are holding the chain links pro tem). */
5976
5977 while (bralink != NULL)
5978 {
5979 int oldlinkoffset;
5980 int offset = (int)(code - bralink + 1);
5981 pcre_uchar *bra = code - offset;
5982 oldlinkoffset = GET(bra, 1);
5983 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
5984 *code++ = OP_KET;
5985 PUTINC(code, 0, offset);
5986 PUT(bra, 1, offset);
5987 }
5988 }
5989
5990 /* If the maximum is unlimited, set a repeater in the final copy. For
5991 ONCE brackets, that's all we need to do. However, possessively repeated
5992 ONCE brackets can be converted into non-capturing brackets, as the
5993 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5994 deal with possessive ONCEs specially.
5995
5996 Otherwise, when we are doing the actual compile phase, check to see
5997 whether this group is one that could match an empty string. If so,
5998 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5999 that runtime checking can be done. [This check is also applied to ONCE
6000 groups at runtime, but in a different way.]
6001
6002 Then, if the quantifier was possessive and the bracket is not a
6003 conditional, we convert the BRA code to the POS form, and the KET code to
6004 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
6005 subpattern at both the start and at the end.) The use of special opcodes
6006 makes it possible to reduce greatly the stack usage in pcre_exec(). If
6007 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
6008
6009 Then, if the minimum number of matches is 1 or 0, cancel the possessive
6010 flag so that the default action below, of wrapping everything inside
6011 atomic brackets, does not happen. When the minimum is greater than 1,
6012 there will be earlier copies of the group, and so we still have to wrap
6013 the whole thing. */
6014
6015 else
6016 {
6017 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
6018 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
6019
6020 /* Convert possessive ONCE brackets to non-capturing */
6021
6022 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
6023 possessive_quantifier) *bracode = OP_BRA;
6024
6025 /* For non-possessive ONCE brackets, all we need to do is to
6026 set the KET. */
6027
6028 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
6029 *ketcode = OP_KETRMAX + repeat_type;
6030
6031 /* Handle non-ONCE brackets and possessive ONCEs (which have been
6032 converted to non-capturing above). */
6033
6034 else
6035 {
6036 /* In the compile phase, check for empty string matching. */
6037
6038 if (lengthptr == NULL)
6039 {
6040 pcre_uchar *scode = bracode;
6041 do
6042 {
6043 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6044 {
6045 *bracode += OP_SBRA - OP_BRA;
6046 break;
6047 }
6048 scode += GET(scode, 1);
6049 }
6050 while (*scode == OP_ALT);
6051 }
6052
6053 /* Handle possessive quantifiers. */
6054
6055 if (possessive_quantifier)
6056 {
6057 /* For COND brackets, we wrap the whole thing in a possessively
6058 repeated non-capturing bracket, because we have not invented POS
6059 versions of the COND opcodes. Because we are moving code along, we
6060 must ensure that any pending recursive references are updated. */
6061
6062 if (*bracode == OP_COND || *bracode == OP_SCOND)
6063 {
6064 int nlen = (int)(code - bracode);
6065 *code = OP_END;
6066 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
6067 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6068 code += 1 + LINK_SIZE;
6069 nlen += 1 + LINK_SIZE;
6070 *bracode = OP_BRAPOS;
6071 *code++ = OP_KETRPOS;
6072 PUTINC(code, 0, nlen);
6073 PUT(bracode, 1, nlen);
6074 }
6075
6076 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6077
6078 else
6079 {
6080 *bracode += 1; /* Switch to xxxPOS opcodes */
6081 *ketcode = OP_KETRPOS;
6082 }
6083
6084 /* If the minimum is zero, mark it as possessive, then unset the
6085 possessive flag when the minimum is 0 or 1. */
6086
6087 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6088 if (repeat_min < 2) possessive_quantifier = FALSE;
6089 }
6090
6091 /* Non-possessive quantifier */
6092
6093 else *ketcode = OP_KETRMAX + repeat_type;
6094 }
6095 }
6096 }
6097
6098 /* If previous is OP_FAIL, it was generated by an empty class [] in
6099 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6100 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6101 error above. We can just ignore the repeat in JS case. */
6102
6103 else if (*previous == OP_FAIL) goto END_REPEAT;
6104
6105 /* Else there's some kind of shambles */
6106
6107 else
6108 {
6109 *errorcodeptr = ERR11;
6110 goto FAILED;
6111 }
6112
6113 /* If the character following a repeat is '+', possessive_quantifier is
6114 TRUE. For some opcodes, there are special alternative opcodes for this
6115 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6116 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6117 Sun's Java package, but the special opcodes can optimize it.
6118
6119 Some (but not all) possessively repeated subpatterns have already been
6120 completely handled in the code just above. For them, possessive_quantifier
6121 is always FALSE at this stage. Note that the repeated item starts at
6122 tempcode, not at previous, which might be the first part of a string whose
6123 (former) last char we repeated. */
6124
6125 if (possessive_quantifier)
6126 {
6127 int len;
6128
6129 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6130 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6131 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6132 remains is greater than zero, there's a further opcode that can be
6133 handled. If not, do nothing, leaving the EXACT alone. */
6134
6135 switch(*tempcode)
6136 {
6137 case OP_TYPEEXACT:
6138 tempcode += PRIV(OP_lengths)[*tempcode] +
6139 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6140 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6141 break;
6142
6143 /* CHAR opcodes are used for exacts whose count is 1. */
6144
6145 case OP_CHAR:
6146 case OP_CHARI:
6147 case OP_NOT:
6148 case OP_NOTI:
6149 case OP_EXACT:
6150 case OP_EXACTI:
6151 case OP_NOTEXACT:
6152 case OP_NOTEXACTI:
6153 tempcode += PRIV(OP_lengths)[*tempcode];
6154 #ifdef SUPPORT_UTF
6155 if (utf && HAS_EXTRALEN(tempcode[-1]))
6156 tempcode += GET_EXTRALEN(tempcode[-1]);
6157 #endif
6158 break;
6159
6160 /* For the class opcodes, the repeat operator appears at the end;
6161 adjust tempcode to point to it. */
6162
6163 case OP_CLASS:
6164 case OP_NCLASS:
6165 tempcode += 1 + 32/sizeof(pcre_uchar);
6166 break;
6167
6168 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6169 case OP_XCLASS:
6170 tempcode += GET(tempcode, 1);
6171 break;
6172 #endif
6173 }
6174
6175 /* If tempcode is equal to code (which points to the end of the repeated
6176 item), it means we have skipped an EXACT item but there is no following
6177 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6178 all other cases, tempcode will be pointing to the repeat opcode, and will
6179 be less than code, so the value of len will be greater than 0. */
6180
6181 len = (int)(code - tempcode);
6182 if (len > 0)
6183 {
6184 unsigned int repcode = *tempcode;
6185
6186 /* There is a table for possessifying opcodes, all of which are less
6187 than OP_CALLOUT. A zero entry means there is no possessified version.
6188 */
6189
6190 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6191 *tempcode = opcode_possessify[repcode];
6192
6193 /* For opcode without a special possessified version, wrap the item in
6194 ONCE brackets. Because we are moving code along, we must ensure that any
6195 pending recursive references are updated. */
6196
6197 else
6198 {
6199 *code = OP_END;
6200 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6201 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6202 code += 1 + LINK_SIZE;
6203 len += 1 + LINK_SIZE;
6204 tempcode[0] = OP_ONCE;
6205 *code++ = OP_KET;
6206 PUTINC(code, 0, len);
6207 PUT(tempcode, 1, len);
6208 }
6209 }
6210
6211 #ifdef NEVER
6212 if (len > 0) switch (*tempcode)
6213 {
6214 case OP_STAR: *tempcode = OP_POSSTAR; break;
6215 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6216 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6217 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6218
6219 case OP_STARI: *tempcode = OP_POSSTARI; break;
6220 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6221 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6222 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6223
6224 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6225 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6226 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6227 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6228
6229 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6230 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6231 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6232 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6233
6234 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6235 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6236 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6237 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6238
6239 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6240 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6241 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6242 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6243
6244 /* Because we are moving code along, we must ensure that any
6245 pending recursive references are updated. */
6246
6247 default:
6248 *code = OP_END;
6249 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6250 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6251 code += 1 + LINK_SIZE;
6252 len += 1 + LINK_SIZE;
6253 tempcode[0] = OP_ONCE;
6254 *code++ = OP_KET;
6255 PUTINC(code, 0, len);
6256 PUT(tempcode, 1, len);
6257 break;
6258 }
6259 #endif
6260 }
6261
6262 /* In all case we no longer have a previous item. We also set the
6263 "follows varying string" flag for subsequently encountered reqchars if
6264 it isn't already set and we have just passed a varying length item. */
6265
6266 END_REPEAT:
6267 previous = NULL;
6268 cd->req_varyopt |= reqvary;
6269 break;
6270
6271
6272 /* ===================================================================*/
6273 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6274 lookbehind or option setting or condition or all the other extended
6275 parenthesis forms. */
6276
6277 case CHAR_LEFT_PARENTHESIS:
6278 newoptions = options;
6279 skipbytes = 0;
6280 bravalue = OP_CBRA;
6281 save_hwm = cd->hwm;
6282 reset_bracount = FALSE;
6283
6284 /* First deal with various "verbs" that can be introduced by '*'. */
6285
6286 ptr++;
6287 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6288 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6289 {
6290 int i, namelen;
6291 int arglen = 0;
6292 const char *vn = verbnames;
6293 const pcre_uchar *name = ptr + 1;
6294 const pcre_uchar *arg = NULL;
6295 previous = NULL;
6296 ptr++;
6297 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6298 namelen = (int)(ptr - name);
6299
6300 /* It appears that Perl allows any characters whatsoever, other than
6301 a closing parenthesis, to appear in arguments, so we no longer insist on
6302 letters, digits, and underscores. */
6303
6304 if (*ptr == CHAR_COLON)
6305 {
6306 arg = ++ptr;
6307 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6308 arglen = (int)(ptr - arg);
6309 if ((unsigned int)arglen > MAX_MARK)
6310 {
6311 *errorcodeptr = ERR75;
6312 goto FAILED;
6313 }
6314 }
6315
6316 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6317 {
6318 *errorcodeptr = ERR60;
6319 goto FAILED;
6320 }
6321
6322 /* Scan the table of verb names */
6323
6324 for (i = 0; i < verbcount; i++)
6325 {
6326 if (namelen == verbs[i].len &&
6327 STRNCMP_UC_C8(name, vn, namelen) == 0)
6328 {
6329 int setverb;
6330
6331 /* Check for open captures before ACCEPT and convert it to
6332 ASSERT_ACCEPT if in an assertion. */
6333
6334 if (verbs[i].op == OP_ACCEPT)
6335 {
6336 open_capitem *oc;
6337 if (arglen != 0)
6338 {
6339 *errorcodeptr = ERR59;
6340 goto FAILED;
6341 }
6342 cd->had_accept = TRUE;
6343 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6344 {
6345 *code++ = OP_CLOSE;
6346 PUT2INC(code, 0, oc->number);
6347 }
6348 setverb = *code++ =
6349 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6350
6351 /* Do not set firstchar after *ACCEPT */
6352 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6353 }
6354
6355 /* Handle other cases with/without an argument */
6356
6357 else if (arglen == 0)
6358 {
6359 if (verbs[i].op < 0) /* Argument is mandatory */
6360 {
6361 *errorcodeptr = ERR66;
6362 goto FAILED;
6363 }
6364 setverb = *code++ = verbs[i].op;
6365 }
6366
6367 else
6368 {
6369 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6370 {
6371 *errorcodeptr = ERR59;
6372 goto FAILED;
6373 }
6374 setverb = *code++ = verbs[i].op_arg;
6375 *code++ = arglen;
6376 memcpy(code, arg, IN_UCHARS(arglen));
6377 code += arglen;
6378 *code++ = 0;
6379 }
6380
6381 switch (setverb)
6382 {
6383 case OP_THEN:
6384 case OP_THEN_ARG:
6385 cd->external_flags |= PCRE_HASTHEN;
6386 break;
6387
6388 case OP_PRUNE:
6389 case OP_PRUNE_ARG:
6390 case OP_SKIP:
6391 case OP_SKIP_ARG:
6392 cd->had_pruneorskip = TRUE;
6393 break;
6394 }
6395
6396 break; /* Found verb, exit loop */
6397 }
6398
6399 vn += verbs[i].len + 1;
6400 }
6401
6402 if (i < verbcount) continue; /* Successfully handled a verb */
6403 *errorcodeptr = ERR60; /* Verb not recognized */
6404 goto FAILED;
6405 }
6406
6407 /* Deal with the extended parentheses; all are introduced by '?', and the
6408 appearance of any of them means that this is not a capturing group. */
6409
6410 else if (*ptr == CHAR_QUESTION_MARK)
6411 {
6412 int i, set, unset, namelen;
6413 int *optset;
6414 const pcre_uchar *name;
6415 pcre_uchar *slot;
6416
6417 switch (*(++ptr))
6418 {
6419 case CHAR_NUMBER_SIGN: /* Comment; skip to ket */
6420 ptr++;
6421 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6422 if (*ptr == CHAR_NULL)
6423 {
6424 *errorcodeptr = ERR18;
6425 goto FAILED;
6426 }
6427 continue;
6428
6429
6430 /* ------------------------------------------------------------ */
6431 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6432 reset_bracount = TRUE;
6433 /* Fall through */
6434
6435 /* ------------------------------------------------------------ */
6436 case CHAR_COLON: /* Non-capturing bracket */
6437 bravalue = OP_BRA;
6438 ptr++;
6439 break;
6440
6441
6442 /* ------------------------------------------------------------ */
6443 case CHAR_LEFT_PARENTHESIS:
6444 bravalue = OP_COND; /* Conditional group */
6445 tempptr = ptr;
6446
6447 /* A condition can be an assertion, a number (referring to a numbered
6448 group), a name (referring to a named group), or 'R', referring to
6449 recursion. R<digits> and R&name are also permitted for recursion tests.
6450
6451 There are several syntaxes for testing a named group: (?(name)) is used
6452 by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
6453
6454 There are two unfortunate ambiguities, caused by history. (a) 'R' can
6455 be the recursive thing or the name 'R' (and similarly for 'R' followed
6456 by digits), and (b) a number could be a name that consists of digits.
6457 In both cases, we look for a name first; if not found, we try the other
6458 cases.
6459
6460 For compatibility with auto-callouts, we allow a callout to be
6461 specified before a condition that is an assertion. First, check for the
6462 syntax of a callout; if found, adjust the temporary pointer that is
6463 used to check for an assertion condition. That's all that is needed! */
6464
6465 if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6466 {
6467 for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6468 if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6469 tempptr += i + 1;
6470 }
6471
6472 /* For conditions that are assertions, check the syntax, and then exit
6473 the switch. This will take control down to where bracketed groups,
6474 including assertions, are processed. */
6475
6476 if (tempptr[1] == CHAR_QUESTION_MARK &&
6477 (tempptr[2] == CHAR_EQUALS_SIGN ||
6478 tempptr[2] == CHAR_EXCLAMATION_MARK ||
6479 tempptr[2] == CHAR_LESS_THAN_SIGN))
6480 break;
6481
6482 /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6483 need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6484
6485 code[1+LINK_SIZE] = OP_CREF;
6486 skipbytes = 1+IMM2_SIZE;
6487 refsign = -1;
6488
6489 /* Check for a test for recursion in a named group. */
6490
6491 if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
6492 {
6493 terminator = -1;
6494 ptr += 2;
6495 code[1+LINK_SIZE] = OP_RREF; /* Change the type of test */
6496 }
6497
6498 /* Check for a test for a named group's having been set, using the Perl
6499 syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6500 syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). As names may
6501 consist entirely of digits, there is scope for ambiguity. */
6502
6503 else if (ptr[1] == CHAR_LESS_THAN_SIGN)
6504 {
6505 terminator = CHAR_GREATER_THAN_SIGN;
6506 ptr++;
6507 }
6508 else if (ptr[1] == CHAR_APOSTROPHE)
6509 {
6510 terminator = CHAR_APOSTROPHE;
6511 ptr++;
6512 }
6513 else
6514 {
6515 terminator = CHAR_NULL;
6516 if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
6517 }
6518
6519 /* When a name is one of a number of duplicates, a different opcode is
6520 used and it needs more memory. Unfortunately we cannot tell whether a
6521 name is a duplicate in the first pass, so we have to allow for more
6522 memory except when we know it is a relative numerical reference. */
6523
6524 if (refsign < 0 && lengthptr != NULL) *lengthptr += IMM2_SIZE;
6525
6526 /* We now expect to read a name (possibly all digits); any thing else
6527 is an error. In the case of all digits, also get it as a number. */
6528
6529 if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)
6530 {
6531 ptr += 1; /* To get the right offset */
6532 *errorcodeptr = ERR28;
6533 goto FAILED;
6534 }
6535
6536 recno = 0;
6537 name = ++ptr;
6538 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6539 {
6540 if (recno >= 0)
6541 recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;
6542 ptr++;
6543 }
6544 namelen = (int)(ptr - name);
6545
6546 /* Check the terminator */
6547
6548 if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6549 *ptr++ != CHAR_RIGHT_PARENTHESIS)
6550 {
6551 ptr--; /* Error offset */
6552 *errorcodeptr = ERR26;
6553 goto FAILED;
6554 }
6555
6556 /* Do no further checking in the pre-compile phase. */
6557
6558 if (lengthptr != NULL) break;
6559
6560 /* In the real compile we do the work of looking for the actual
6561 reference. If the string started with "+" or "-" we require the rest to
6562 be digits, in which case recno will be set. */
6563
6564 if (refsign > 0)
6565 {
6566 if (recno <= 0)
6567 {
6568 *errorcodeptr = ERR58;
6569 goto FAILED;
6570 }
6571 recno = (refsign == CHAR_MINUS)?
6572 cd->bracount - recno + 1 : recno +cd->bracount;
6573 if (recno <= 0 || recno > cd->final_bracount)
6574 {
6575 *errorcodeptr = ERR15;
6576 goto FAILED;
6577 }
6578 PUT2(code, 2+LINK_SIZE, recno);
6579 break;
6580 }
6581
6582 /* Otherwise (did not start with "+" or "-"), start by looking for the
6583 name. */
6584
6585 slot = cd->name_table;
6586 for (i = 0; i < cd->names_found; i++)
6587 {
6588 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
6589 slot += cd->name_entry_size;
6590 }
6591
6592 /* Found the named subpattern. If the name is duplicated, add one to
6593 the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6594 appropriate data values. Otherwise, just insert the unique subpattern
6595 number. */
6596
6597 if (i < cd->names_found)
6598 {
6599 int offset = i++;
6600 int count = 1;
6601 recno = GET2(slot, 0); /* Number from first found */
6602 for (; i < cd->names_found; i++)
6603 {
6604 slot += cd->name_entry_size;
6605 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0) break;
6606 count++;
6607 }
6608 if (count > 1)
6609 {
6610 PUT2(code, 2+LINK_SIZE, offset);
6611 PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
6612 skipbytes += IMM2_SIZE;
6613 code[1+LINK_SIZE]++;
6614 }
6615 else /* Not a duplicated name */
6616 {
6617 PUT2(code, 2+LINK_SIZE, recno);
6618 }
6619 }
6620
6621 /* If terminator == CHAR_NULL it means that the name followed directly
6622 after the opening parenthesis [e.g. (?(abc)...] and in this case there
6623 are some further alternatives to try. For the cases where terminator !=
6624 0 [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
6625 now checked all the possibilities, so give an error. */
6626
6627 else if (terminator != CHAR_NULL)
6628 {
6629 *errorcodeptr = ERR15;
6630 goto FAILED;
6631 }
6632
6633 /* Check for (?(R) for recursion. Allow digits after R to specify a
6634 specific group number. */
6635
6636 else if (*name == CHAR_R)
6637 {
6638 recno = 0;
6639 for (i = 1; i < namelen; i++)
6640 {
6641 if (!IS_DIGIT(name[i]))
6642 {
6643 *errorcodeptr = ERR15;
6644 goto FAILED;
6645 }
6646 recno = recno * 10 + name[i] - CHAR_0;
6647 }
6648 if (recno == 0) recno = RREF_ANY;
6649 code[1+LINK_SIZE] = OP_RREF; /* Change test type */
6650 PUT2(code, 2+LINK_SIZE, recno);
6651 }
6652
6653 /* Similarly, check for the (?(DEFINE) "condition", which is always
6654 false. */
6655
6656 else if (namelen == 6 && STRNCMP_UC_C8(name, STRING_DEFINE, 6) == 0)
6657 {
6658 code[1+LINK_SIZE] = OP_DEF;
6659 skipbytes = 1;
6660 }
6661