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

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1398 - (show annotations)
Tue Nov 12 15:20:26 2013 UTC (5 years, 10 months ago) by ph10
File MIME type: text/plain
File size: 308250 byte(s)
Rename the (as yet unreleased) PCRE_NO_AUTO_POSSESSIFY option to 
PCRE_NO_AUTO_POSSESS - no need to be so long - and to match (*NO_AUTO_POSSESS).
</
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 "parentheses are too deeply nested\0"
535 "invalid range in character class\0"
536 "group name must start with a non-digit\0"
537 ;
538
539 /* Table to identify digits and hex digits. This is used when compiling
540 patterns. Note that the tables in chartables are dependent on the locale, and
541 may mark arbitrary characters as digits - but the PCRE compiling code expects
542 to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
543 a private table here. It costs 256 bytes, but it is a lot faster than doing
544 character value tests (at least in some simple cases I timed), and in some
545 applications one wants PCRE to compile efficiently as well as match
546 efficiently.
547
548 For convenience, we use the same bit definitions as in chartables:
549
550 0x04 decimal digit
551 0x08 hexadecimal digit
552
553 Then we can use ctype_digit and ctype_xdigit in the code. */
554
555 /* Using a simple comparison for decimal numbers rather than a memory read
556 is much faster, and the resulting code is simpler (the compiler turns it
557 into a subtraction and unsigned comparison). */
558
559 #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
560
561 #ifndef EBCDIC
562
563 /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
564 UTF-8 mode. */
565
566 static const pcre_uint8 digitab[] =
567 {
568 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
569 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
570 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
571 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
572 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
573 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
574 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
575 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
576 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* @ - G */
577 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H - O */
578 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* P - W */
579 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* X - _ */
580 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
581 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
582 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
583 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
584 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
585 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
586 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
587 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
588 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
589 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
590 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
591 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
592 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
593 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
594 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
595 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
596 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
597 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
598 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
599 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
600
601 #else
602
603 /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
604
605 static const pcre_uint8 digitab[] =
606 {
607 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
608 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
609 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
610 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
611 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
612 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
613 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
614 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
615 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
616 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
617 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
618 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
619 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
620 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
621 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
622 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
623 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
624 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
625 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
626 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
627 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
628 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
629 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
630 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
631 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
632 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
633 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
634 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
635 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
636 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
637 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
638 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
639
640 static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
641 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
642 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
643 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
644 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
645 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
646 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
647 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
648 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
649 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
650 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
651 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
652 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
653 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
654 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
655 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
656 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
657 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
658 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
659 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
660 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
661 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
662 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
663 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
664 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
665 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
666 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
667 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
668 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
669 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
670 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
671 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
672 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
673 #endif
674
675
676 /* This table is used to check whether auto-possessification is possible
677 between adjacent character-type opcodes. The left-hand (repeated) opcode is
678 used to select the row, and the right-hand opcode is use to select the column.
679 A value of 1 means that auto-possessification is OK. For example, the second
680 value in the first row means that \D+\d can be turned into \D++\d.
681
682 The Unicode property types (\P and \p) have to be present to fill out the table
683 because of what their opcode values are, but the table values should always be
684 zero because property types are handled separately in the code. The last four
685 columns apply to items that cannot be repeated, so there is no need to have
686 rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
687 *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
688
689 #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
690 #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
691
692 static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
693 /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
694 { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
695 { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
696 { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
697 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
698 { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
699 { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
700 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
701 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
702 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
703 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
704 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
705 { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
706 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
707 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
708 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
709 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
710 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
711 };
712
713
714 /* This table is used to check whether auto-possessification is possible
715 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
716 left-hand (repeated) opcode is used to select the row, and the right-hand
717 opcode is used to select the column. The values are as follows:
718
719 0 Always return FALSE (never auto-possessify)
720 1 Character groups are distinct (possessify if both are OP_PROP)
721 2 Check character categories in the same group (general or particular)
722 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
723
724 4 Check left general category vs right particular category
725 5 Check right general category vs left particular category
726
727 6 Left alphanum vs right general category
728 7 Left space vs right general category
729 8 Left word vs right general category
730
731 9 Right alphanum vs left general category
732 10 Right space vs left general category
733 11 Right word vs left general category
734
735 12 Left alphanum vs right particular category
736 13 Left space vs right particular category
737 14 Left word vs right particular category
738
739 15 Right alphanum vs left particular category
740 16 Right space vs left particular category
741 17 Right word vs left particular category
742 */
743
744 static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
745 /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
746 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
747 { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
748 { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
749 { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
750 { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
751 { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
752 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
753 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
754 { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
755 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
756 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
757 };
758
759 /* This table is used to check whether auto-possessification is possible
760 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
761 specifies a general category and the other specifies a particular category. The
762 row is selected by the general category and the column by the particular
763 category. The value is 1 if the particular category is not part of the general
764 category. */
765
766 static const pcre_uint8 catposstab[7][30] = {
767 /* 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 */
768 { 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 */
769 { 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 */
770 { 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 */
771 { 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 */
772 { 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 */
773 { 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 */
774 { 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 */
775 };
776
777 /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
778 a general or particular category. The properties in each row are those
779 that apply to the character set in question. Duplication means that a little
780 unnecessary work is done when checking, but this keeps things much simpler
781 because they can all use the same code. For more details see the comment where
782 this table is used.
783
784 Note: SPACE and PXSPACE used to be different because Perl excluded VT from
785 "space", but from Perl 5.18 it's included, so both categories are treated the
786 same here. */
787
788 static const pcre_uint8 posspropstab[3][4] = {
789 { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
790 { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
791 { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
792 };
793
794 /* This table is used when converting repeating opcodes into possessified
795 versions as a result of an explicit possessive quantifier such as ++. A zero
796 value means there is no possessified version - in those cases the item in
797 question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
798 because all relevant opcodes are less than that. */
799
800 static const pcre_uint8 opcode_possessify[] = {
801 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */
802 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */
803
804 0, /* NOTI */
805 OP_POSSTAR, 0, /* STAR, MINSTAR */
806 OP_POSPLUS, 0, /* PLUS, MINPLUS */
807 OP_POSQUERY, 0, /* QUERY, MINQUERY */
808 OP_POSUPTO, 0, /* UPTO, MINUPTO */
809 0, /* EXACT */
810 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */
811
812 OP_POSSTARI, 0, /* STARI, MINSTARI */
813 OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */
814 OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */
815 OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */
816 0, /* EXACTI */
817 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */
818
819 OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */
820 OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */
821 OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */
822 OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */
823 0, /* NOTEXACT */
824 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
825
826 OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */
827 OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */
828 OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */
829 OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */
830 0, /* NOTEXACTI */
831 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
832
833 OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */
834 OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */
835 OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */
836 OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */
837 0, /* TYPEEXACT */
838 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
839
840 OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */
841 OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */
842 OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */
843 OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */
844 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */
845
846 0, 0, 0, /* CLASS, NCLASS, XCLASS */
847 0, 0, /* REF, REFI */
848 0, 0, /* DNREF, DNREFI */
849 0, 0 /* RECURSE, CALLOUT */
850 };
851
852
853
854 /*************************************************
855 * Find an error text *
856 *************************************************/
857
858 /* The error texts are now all in one long string, to save on relocations. As
859 some of the text is of unknown length, we can't use a table of offsets.
860 Instead, just count through the strings. This is not a performance issue
861 because it happens only when there has been a compilation error.
862
863 Argument: the error number
864 Returns: pointer to the error string
865 */
866
867 static const char *
868 find_error_text(int n)
869 {
870 const char *s = error_texts;
871 for (; n > 0; n--)
872 {
873 while (*s++ != CHAR_NULL) {};
874 if (*s == CHAR_NULL) return "Error text not found (please report)";
875 }
876 return s;
877 }
878
879
880
881 /*************************************************
882 * Expand the workspace *
883 *************************************************/
884
885 /* This function is called during the second compiling phase, if the number of
886 forward references fills the existing workspace, which is originally a block on
887 the stack. A larger block is obtained from malloc() unless the ultimate limit
888 has been reached or the increase will be rather small.
889
890 Argument: pointer to the compile data block
891 Returns: 0 if all went well, else an error number
892 */
893
894 static int
895 expand_workspace(compile_data *cd)
896 {
897 pcre_uchar *newspace;
898 int newsize = cd->workspace_size * 2;
899
900 if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
901 if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
902 newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
903 return ERR72;
904
905 newspace = (PUBL(malloc))(IN_UCHARS(newsize));
906 if (newspace == NULL) return ERR21;
907 memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
908 cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
909 if (cd->workspace_size > COMPILE_WORK_SIZE)
910 (PUBL(free))((void *)cd->start_workspace);
911 cd->start_workspace = newspace;
912 cd->workspace_size = newsize;
913 return 0;
914 }
915
916
917
918 /*************************************************
919 * Check for counted repeat *
920 *************************************************/
921
922 /* This function is called when a '{' is encountered in a place where it might
923 start a quantifier. It looks ahead to see if it really is a quantifier or not.
924 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
925 where the ddds are digits.
926
927 Arguments:
928 p pointer to the first char after '{'
929
930 Returns: TRUE or FALSE
931 */
932
933 static BOOL
934 is_counted_repeat(const pcre_uchar *p)
935 {
936 if (!IS_DIGIT(*p)) return FALSE;
937 p++;
938 while (IS_DIGIT(*p)) p++;
939 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
940
941 if (*p++ != CHAR_COMMA) return FALSE;
942 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
943
944 if (!IS_DIGIT(*p)) return FALSE;
945 p++;
946 while (IS_DIGIT(*p)) p++;
947
948 return (*p == CHAR_RIGHT_CURLY_BRACKET);
949 }
950
951
952
953 /*************************************************
954 * Handle escapes *
955 *************************************************/
956
957 /* This function is called when a \ has been encountered. It either returns a
958 positive value for a simple escape such as \n, or 0 for a data character which
959 will be placed in chptr. A backreference to group n is returned as negative n.
960 When UTF-8 is enabled, a positive value greater than 255 may be returned in
961 chptr. On entry, ptr is pointing at the \. On exit, it is on the final
962 character of the escape sequence.
963
964 Arguments:
965 ptrptr points to the pattern position pointer
966 chptr points to a returned data character
967 errorcodeptr points to the errorcode variable
968 bracount number of previous extracting brackets
969 options the options bits
970 isclass TRUE if inside a character class
971
972 Returns: zero => a data character
973 positive => a special escape sequence
974 negative => a back reference
975 on error, errorcodeptr is set
976 */
977
978 static int
979 check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
980 int bracount, int options, BOOL isclass)
981 {
982 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
983 BOOL utf = (options & PCRE_UTF8) != 0;
984 const pcre_uchar *ptr = *ptrptr + 1;
985 pcre_uint32 c;
986 int escape = 0;
987 int i;
988
989 GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
990 ptr--; /* Set pointer back to the last byte */
991
992 /* If backslash is at the end of the pattern, it's an error. */
993
994 if (c == CHAR_NULL) *errorcodeptr = ERR1;
995
996 /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
997 in a table. A non-zero result is something that can be returned immediately.
998 Otherwise further processing may be required. */
999
1000 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1001 /* Not alphanumeric */
1002 else if (c < CHAR_0 || c > CHAR_z) {}
1003 else if ((i = escapes[c - CHAR_0]) != 0)
1004 { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1005
1006 #else /* EBCDIC coding */
1007 /* Not alphanumeric */
1008 else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1009 else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1010 #endif
1011
1012 /* Escapes that need further processing, or are illegal. */
1013
1014 else
1015 {
1016 const pcre_uchar *oldptr;
1017 BOOL braced, negated, overflow;
1018 int s;
1019
1020 switch (c)
1021 {
1022 /* A number of Perl escapes are not handled by PCRE. We give an explicit
1023 error. */
1024
1025 case CHAR_l:
1026 case CHAR_L:
1027 *errorcodeptr = ERR37;
1028 break;
1029
1030 case CHAR_u:
1031 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1032 {
1033 /* In JavaScript, \u must be followed by four hexadecimal numbers.
1034 Otherwise it is a lowercase u letter. */
1035 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1036 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1037 && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1038 && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1039 {
1040 c = 0;
1041 for (i = 0; i < 4; ++i)
1042 {
1043 register pcre_uint32 cc = *(++ptr);
1044 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1045 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1046 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1047 #else /* EBCDIC coding */
1048 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1049 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1050 #endif
1051 }
1052
1053 #if defined COMPILE_PCRE8
1054 if (c > (utf ? 0x10ffffU : 0xffU))
1055 #elif defined COMPILE_PCRE16
1056 if (c > (utf ? 0x10ffffU : 0xffffU))
1057 #elif defined COMPILE_PCRE32
1058 if (utf && c > 0x10ffffU)
1059 #endif
1060 {
1061 *errorcodeptr = ERR76;
1062 }
1063 else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1064 }
1065 }
1066 else
1067 *errorcodeptr = ERR37;
1068 break;
1069
1070 case CHAR_U:
1071 /* In JavaScript, \U is an uppercase U letter. */
1072 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1073 break;
1074
1075 /* In a character class, \g is just a literal "g". Outside a character
1076 class, \g must be followed by one of a number of specific things:
1077
1078 (1) A number, either plain or braced. If positive, it is an absolute
1079 backreference. If negative, it is a relative backreference. This is a Perl
1080 5.10 feature.
1081
1082 (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1083 is part of Perl's movement towards a unified syntax for back references. As
1084 this is synonymous with \k{name}, we fudge it up by pretending it really
1085 was \k.
1086
1087 (3) For Oniguruma compatibility we also support \g followed by a name or a
1088 number either in angle brackets or in single quotes. However, these are
1089 (possibly recursive) subroutine calls, _not_ backreferences. Just return
1090 the ESC_g code (cf \k). */
1091
1092 case CHAR_g:
1093 if (isclass) break;
1094 if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1095 {
1096 escape = ESC_g;
1097 break;
1098 }
1099
1100 /* Handle the Perl-compatible cases */
1101
1102 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1103 {
1104 const pcre_uchar *p;
1105 for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1106 if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1107 if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1108 {
1109 escape = ESC_k;
1110 break;
1111 }
1112 braced = TRUE;
1113 ptr++;
1114 }
1115 else braced = FALSE;
1116
1117 if (ptr[1] == CHAR_MINUS)
1118 {
1119 negated = TRUE;
1120 ptr++;
1121 }
1122 else negated = FALSE;
1123
1124 /* The integer range is limited by the machine's int representation. */
1125 s = 0;
1126 overflow = FALSE;
1127 while (IS_DIGIT(ptr[1]))
1128 {
1129 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1130 {
1131 overflow = TRUE;
1132 break;
1133 }
1134 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1135 }
1136 if (overflow) /* Integer overflow */
1137 {
1138 while (IS_DIGIT(ptr[1]))
1139 ptr++;
1140 *errorcodeptr = ERR61;
1141 break;
1142 }
1143
1144 if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1145 {
1146 *errorcodeptr = ERR57;
1147 break;
1148 }
1149
1150 if (s == 0)
1151 {
1152 *errorcodeptr = ERR58;
1153 break;
1154 }
1155
1156 if (negated)
1157 {
1158 if (s > bracount)
1159 {
1160 *errorcodeptr = ERR15;
1161 break;
1162 }
1163 s = bracount - (s - 1);
1164 }
1165
1166 escape = -s;
1167 break;
1168
1169 /* The handling of escape sequences consisting of a string of digits
1170 starting with one that is not zero is not straightforward. Perl has changed
1171 over the years. Nowadays \g{} for backreferences and \o{} for octal are
1172 recommended to avoid the ambiguities in the old syntax.
1173
1174 Outside a character class, the digits are read as a decimal number. If the
1175 number is less than 8 (used to be 10), or if there are that many previous
1176 extracting left brackets, then it is a back reference. Otherwise, up to
1177 three octal digits are read to form an escaped byte. Thus \123 is likely to
1178 be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1179 the octal value is greater than 377, the least significant 8 bits are
1180 taken. \8 and \9 are treated as the literal characters 8 and 9.
1181
1182 Inside a character class, \ followed by a digit is always either a literal
1183 8 or 9 or an octal number. */
1184
1185 case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1186 case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1187
1188 if (!isclass)
1189 {
1190 oldptr = ptr;
1191 /* The integer range is limited by the machine's int representation. */
1192 s = (int)(c -CHAR_0);
1193 overflow = FALSE;
1194 while (IS_DIGIT(ptr[1]))
1195 {
1196 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1197 {
1198 overflow = TRUE;
1199 break;
1200 }
1201 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1202 }
1203 if (overflow) /* Integer overflow */
1204 {
1205 while (IS_DIGIT(ptr[1]))
1206 ptr++;
1207 *errorcodeptr = ERR61;
1208 break;
1209 }
1210 if (s < 8 || s <= bracount) /* Check for back reference */
1211 {
1212 escape = -s;
1213 break;
1214 }
1215 ptr = oldptr; /* Put the pointer back and fall through */
1216 }
1217
1218 /* Handle a digit following \ when the number is not a back reference. If
1219 the first digit is 8 or 9, Perl used to generate a binary zero byte and
1220 then treat the digit as a following literal. At least by Perl 5.18 this
1221 changed so as not to insert the binary zero. */
1222
1223 if ((c = *ptr) >= CHAR_8) break;
1224
1225 /* Fall through with a digit less than 8 */
1226
1227 /* \0 always starts an octal number, but we may drop through to here with a
1228 larger first octal digit. The original code used just to take the least
1229 significant 8 bits of octal numbers (I think this is what early Perls used
1230 to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1231 but no more than 3 octal digits. */
1232
1233 case CHAR_0:
1234 c -= CHAR_0;
1235 while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1236 c = c * 8 + *(++ptr) - CHAR_0;
1237 #ifdef COMPILE_PCRE8
1238 if (!utf && c > 0xff) *errorcodeptr = ERR51;
1239 #endif
1240 break;
1241
1242 /* \o is a relatively new Perl feature, supporting a more general way of
1243 specifying character codes in octal. The only supported form is \o{ddd}. */
1244
1245 case CHAR_o:
1246 if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1247 {
1248 ptr += 2;
1249 c = 0;
1250 overflow = FALSE;
1251 while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1252 {
1253 register pcre_uint32 cc = *ptr++;
1254 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1255 #ifdef COMPILE_PCRE32
1256 if (c >= 0x20000000l) { overflow = TRUE; break; }
1257 #endif
1258 c = (c << 3) + cc - CHAR_0 ;
1259 #if defined COMPILE_PCRE8
1260 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1261 #elif defined COMPILE_PCRE16
1262 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1263 #elif defined COMPILE_PCRE32
1264 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1265 #endif
1266 }
1267 if (overflow)
1268 {
1269 while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1270 *errorcodeptr = ERR34;
1271 }
1272 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1273 {
1274 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1275 }
1276 else *errorcodeptr = ERR80;
1277 }
1278 break;
1279
1280 /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1281 numbers. Otherwise it is a lowercase x letter. */
1282
1283 case CHAR_x:
1284 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1285 {
1286 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1287 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1288 {
1289 c = 0;
1290 for (i = 0; i < 2; ++i)
1291 {
1292 register pcre_uint32 cc = *(++ptr);
1293 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1294 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1295 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1296 #else /* EBCDIC coding */
1297 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1298 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1299 #endif
1300 }
1301 }
1302 } /* End JavaScript handling */
1303
1304 /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1305 greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1306 digits. If not, { used to be treated as a data character. However, Perl
1307 seems to read hex digits up to the first non-such, and ignore the rest, so
1308 that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1309 now gives an error. */
1310
1311 else
1312 {
1313 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1314 {
1315 ptr += 2;
1316 c = 0;
1317 overflow = FALSE;
1318 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1319 {
1320 register pcre_uint32 cc = *ptr++;
1321 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1322
1323 #ifdef COMPILE_PCRE32
1324 if (c >= 0x10000000l) { overflow = TRUE; break; }
1325 #endif
1326
1327 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1328 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1329 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1330 #else /* EBCDIC coding */
1331 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1332 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1333 #endif
1334
1335 #if defined COMPILE_PCRE8
1336 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1337 #elif defined COMPILE_PCRE16
1338 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1339 #elif defined COMPILE_PCRE32
1340 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1341 #endif
1342 }
1343
1344 if (overflow)
1345 {
1346 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1347 *errorcodeptr = ERR34;
1348 }
1349
1350 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1351 {
1352 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1353 }
1354
1355 /* If the sequence of hex digits does not end with '}', give an error.
1356 We used just to recognize this construct and fall through to the normal
1357 \x handling, but nowadays Perl gives an error, which seems much more
1358 sensible, so we do too. */
1359
1360 else *errorcodeptr = ERR79;
1361 } /* End of \x{} processing */
1362
1363 /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1364
1365 else
1366 {
1367 c = 0;
1368 while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1369 {
1370 pcre_uint32 cc; /* Some compilers don't like */
1371 cc = *(++ptr); /* ++ in initializers */
1372 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1373 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1374 c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1375 #else /* EBCDIC coding */
1376 if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1377 c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1378 #endif
1379 }
1380 } /* End of \xdd handling */
1381 } /* End of Perl-style \x handling */
1382 break;
1383
1384 /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1385 An error is given if the byte following \c is not an ASCII character. This
1386 coding is ASCII-specific, but then the whole concept of \cx is
1387 ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1388
1389 case CHAR_c:
1390 c = *(++ptr);
1391 if (c == CHAR_NULL)
1392 {
1393 *errorcodeptr = ERR2;
1394 break;
1395 }
1396 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1397 if (c > 127) /* Excludes all non-ASCII in either mode */
1398 {
1399 *errorcodeptr = ERR68;
1400 break;
1401 }
1402 if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1403 c ^= 0x40;
1404 #else /* EBCDIC coding */
1405 if (c >= CHAR_a && c <= CHAR_z) c += 64;
1406 c ^= 0xC0;
1407 #endif
1408 break;
1409
1410 /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1411 other alphanumeric following \ is an error if PCRE_EXTRA was set;
1412 otherwise, for Perl compatibility, it is a literal. This code looks a bit
1413 odd, but there used to be some cases other than the default, and there may
1414 be again in future, so I haven't "optimized" it. */
1415
1416 default:
1417 if ((options & PCRE_EXTRA) != 0) switch(c)
1418 {
1419 default:
1420 *errorcodeptr = ERR3;
1421 break;
1422 }
1423 break;
1424 }
1425 }
1426
1427 /* Perl supports \N{name} for character names, as well as plain \N for "not
1428 newline". PCRE does not support \N{name}. However, it does support
1429 quantification such as \N{2,3}. */
1430
1431 if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1432 !is_counted_repeat(ptr+2))
1433 *errorcodeptr = ERR37;
1434
1435 /* If PCRE_UCP is set, we change the values for \d etc. */
1436
1437 if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1438 escape += (ESC_DU - ESC_D);
1439
1440 /* Set the pointer to the final character before returning. */
1441
1442 *ptrptr = ptr;
1443 *chptr = c;
1444 return escape;
1445 }
1446
1447
1448
1449 #ifdef SUPPORT_UCP
1450 /*************************************************
1451 * Handle \P and \p *
1452 *************************************************/
1453
1454 /* This function is called after \P or \p has been encountered, provided that
1455 PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1456 pointing at the P or p. On exit, it is pointing at the final character of the
1457 escape sequence.
1458
1459 Argument:
1460 ptrptr points to the pattern position pointer
1461 negptr points to a boolean that is set TRUE for negation else FALSE
1462 ptypeptr points to an unsigned int that is set to the type value
1463 pdataptr points to an unsigned int that is set to the detailed property value
1464 errorcodeptr points to the error code variable
1465
1466 Returns: TRUE if the type value was found, or FALSE for an invalid type
1467 */
1468
1469 static BOOL
1470 get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1471 unsigned int *pdataptr, int *errorcodeptr)
1472 {
1473 pcre_uchar c;
1474 int i, bot, top;
1475 const pcre_uchar *ptr = *ptrptr;
1476 pcre_uchar name[32];
1477
1478 c = *(++ptr);
1479 if (c == CHAR_NULL) goto ERROR_RETURN;
1480
1481 *negptr = FALSE;
1482
1483 /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1484 negation. */
1485
1486 if (c == CHAR_LEFT_CURLY_BRACKET)
1487 {
1488 if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1489 {
1490 *negptr = TRUE;
1491 ptr++;
1492 }
1493 for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1494 {
1495 c = *(++ptr);
1496 if (c == CHAR_NULL) goto ERROR_RETURN;
1497 if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1498 name[i] = c;
1499 }
1500 if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1501 name[i] = 0;
1502 }
1503
1504 /* Otherwise there is just one following character */
1505
1506 else
1507 {
1508 name[0] = c;
1509 name[1] = 0;
1510 }
1511
1512 *ptrptr = ptr;
1513
1514 /* Search for a recognized property name using binary chop */
1515
1516 bot = 0;
1517 top = PRIV(utt_size);
1518
1519 while (bot < top)
1520 {
1521 int r;
1522 i = (bot + top) >> 1;
1523 r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1524 if (r == 0)
1525 {
1526 *ptypeptr = PRIV(utt)[i].type;
1527 *pdataptr = PRIV(utt)[i].value;
1528 return TRUE;
1529 }
1530 if (r > 0) bot = i + 1; else top = i;
1531 }
1532
1533 *errorcodeptr = ERR47;
1534 *ptrptr = ptr;
1535 return FALSE;
1536
1537 ERROR_RETURN:
1538 *errorcodeptr = ERR46;
1539 *ptrptr = ptr;
1540 return FALSE;
1541 }
1542 #endif
1543
1544
1545
1546 /*************************************************
1547 * Read repeat counts *
1548 *************************************************/
1549
1550 /* Read an item of the form {n,m} and return the values. This is called only
1551 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1552 so the syntax is guaranteed to be correct, but we need to check the values.
1553
1554 Arguments:
1555 p pointer to first char after '{'
1556 minp pointer to int for min
1557 maxp pointer to int for max
1558 returned as -1 if no max
1559 errorcodeptr points to error code variable
1560
1561 Returns: pointer to '}' on success;
1562 current ptr on error, with errorcodeptr set non-zero
1563 */
1564
1565 static const pcre_uchar *
1566 read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1567 {
1568 int min = 0;
1569 int max = -1;
1570
1571 /* Read the minimum value and do a paranoid check: a negative value indicates
1572 an integer overflow. */
1573
1574 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1575 if (min < 0 || min > 65535)
1576 {
1577 *errorcodeptr = ERR5;
1578 return p;
1579 }
1580
1581 /* Read the maximum value if there is one, and again do a paranoid on its size.
1582 Also, max must not be less than min. */
1583
1584 if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1585 {
1586 if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1587 {
1588 max = 0;
1589 while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1590 if (max < 0 || max > 65535)
1591 {
1592 *errorcodeptr = ERR5;
1593 return p;
1594 }
1595 if (max < min)
1596 {
1597 *errorcodeptr = ERR4;
1598 return p;
1599 }
1600 }
1601 }
1602
1603 /* Fill in the required variables, and pass back the pointer to the terminating
1604 '}'. */
1605
1606 *minp = min;
1607 *maxp = max;
1608 return p;
1609 }
1610
1611
1612
1613 /*************************************************
1614 * Find first significant op code *
1615 *************************************************/
1616
1617 /* This is called by several functions that scan a compiled expression looking
1618 for a fixed first character, or an anchoring op code etc. It skips over things
1619 that do not influence this. For some calls, it makes sense to skip negative
1620 forward and all backward assertions, and also the \b assertion; for others it
1621 does not.
1622
1623 Arguments:
1624 code pointer to the start of the group
1625 skipassert TRUE if certain assertions are to be skipped
1626
1627 Returns: pointer to the first significant opcode
1628 */
1629
1630 static const pcre_uchar*
1631 first_significant_code(const pcre_uchar *code, BOOL skipassert)
1632 {
1633 for (;;)
1634 {
1635 switch ((int)*code)
1636 {
1637 case OP_ASSERT_NOT:
1638 case OP_ASSERTBACK:
1639 case OP_ASSERTBACK_NOT:
1640 if (!skipassert) return code;
1641 do code += GET(code, 1); while (*code == OP_ALT);
1642 code += PRIV(OP_lengths)[*code];
1643 break;
1644
1645 case OP_WORD_BOUNDARY:
1646 case OP_NOT_WORD_BOUNDARY:
1647 if (!skipassert) return code;
1648 /* Fall through */
1649
1650 case OP_CALLOUT:
1651 case OP_CREF:
1652 case OP_DNCREF:
1653 case OP_RREF:
1654 case OP_DNRREF:
1655 case OP_DEF:
1656 code += PRIV(OP_lengths)[*code];
1657 break;
1658
1659 default:
1660 return code;
1661 }
1662 }
1663 /* Control never reaches here */
1664 }
1665
1666
1667
1668 /*************************************************
1669 * Find the fixed length of a branch *
1670 *************************************************/
1671
1672 /* Scan a branch and compute the fixed length of subject that will match it,
1673 if the length is fixed. This is needed for dealing with backward assertions.
1674 In UTF8 mode, the result is in characters rather than bytes. The branch is
1675 temporarily terminated with OP_END when this function is called.
1676
1677 This function is called when a backward assertion is encountered, so that if it
1678 fails, the error message can point to the correct place in the pattern.
1679 However, we cannot do this when the assertion contains subroutine calls,
1680 because they can be forward references. We solve this by remembering this case
1681 and doing the check at the end; a flag specifies which mode we are running in.
1682
1683 Arguments:
1684 code points to the start of the pattern (the bracket)
1685 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1686 atend TRUE if called when the pattern is complete
1687 cd the "compile data" structure
1688
1689 Returns: the fixed length,
1690 or -1 if there is no fixed length,
1691 or -2 if \C was encountered (in UTF-8 mode only)
1692 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1693 or -4 if an unknown opcode was encountered (internal error)
1694 */
1695
1696 static int
1697 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1698 {
1699 int length = -1;
1700
1701 register int branchlength = 0;
1702 register pcre_uchar *cc = code + 1 + LINK_SIZE;
1703
1704 /* Scan along the opcodes for this branch. If we get to the end of the
1705 branch, check the length against that of the other branches. */
1706
1707 for (;;)
1708 {
1709 int d;
1710 pcre_uchar *ce, *cs;
1711 register pcre_uchar op = *cc;
1712
1713 switch (op)
1714 {
1715 /* We only need to continue for OP_CBRA (normal capturing bracket) and
1716 OP_BRA (normal non-capturing bracket) because the other variants of these
1717 opcodes are all concerned with unlimited repeated groups, which of course
1718 are not of fixed length. */
1719
1720 case OP_CBRA:
1721 case OP_BRA:
1722 case OP_ONCE:
1723 case OP_ONCE_NC:
1724 case OP_COND:
1725 d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1726 if (d < 0) return d;
1727 branchlength += d;
1728 do cc += GET(cc, 1); while (*cc == OP_ALT);
1729 cc += 1 + LINK_SIZE;
1730 break;
1731
1732 /* Reached end of a branch; if it's a ket it is the end of a nested call.
1733 If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1734 an ALT. If it is END it's the end of the outer call. All can be handled by
1735 the same code. Note that we must not include the OP_KETRxxx opcodes here,
1736 because they all imply an unlimited repeat. */
1737
1738 case OP_ALT:
1739 case OP_KET:
1740 case OP_END:
1741 case OP_ACCEPT:
1742 case OP_ASSERT_ACCEPT:
1743 if (length < 0) length = branchlength;
1744 else if (length != branchlength) return -1;
1745 if (*cc != OP_ALT) return length;
1746 cc += 1 + LINK_SIZE;
1747 branchlength = 0;
1748 break;
1749
1750 /* A true recursion implies not fixed length, but a subroutine call may
1751 be OK. If the subroutine is a forward reference, we can't deal with
1752 it until the end of the pattern, so return -3. */
1753
1754 case OP_RECURSE:
1755 if (!atend) return -3;
1756 cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1757 do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1758 if (cc > cs && cc < ce) return -1; /* Recursion */
1759 d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1760 if (d < 0) return d;
1761 branchlength += d;
1762 cc += 1 + LINK_SIZE;
1763 break;
1764
1765 /* Skip over assertive subpatterns */
1766
1767 case OP_ASSERT:
1768 case OP_ASSERT_NOT:
1769 case OP_ASSERTBACK:
1770 case OP_ASSERTBACK_NOT:
1771 do cc += GET(cc, 1); while (*cc == OP_ALT);
1772 cc += PRIV(OP_lengths)[*cc];
1773 break;
1774
1775 /* Skip over things that don't match chars */
1776
1777 case OP_MARK:
1778 case OP_PRUNE_ARG:
1779 case OP_SKIP_ARG:
1780 case OP_THEN_ARG:
1781 cc += cc[1] + PRIV(OP_lengths)[*cc];
1782 break;
1783
1784 case OP_CALLOUT:
1785 case OP_CIRC:
1786 case OP_CIRCM:
1787 case OP_CLOSE:
1788 case OP_COMMIT:
1789 case OP_CREF:
1790 case OP_DEF:
1791 case OP_DNCREF:
1792 case OP_DNRREF:
1793 case OP_DOLL:
1794 case OP_DOLLM:
1795 case OP_EOD:
1796 case OP_EODN:
1797 case OP_FAIL:
1798 case OP_NOT_WORD_BOUNDARY:
1799 case OP_PRUNE:
1800 case OP_REVERSE:
1801 case OP_RREF:
1802 case OP_SET_SOM:
1803 case OP_SKIP:
1804 case OP_SOD:
1805 case OP_SOM:
1806 case OP_THEN:
1807 case OP_WORD_BOUNDARY:
1808 cc += PRIV(OP_lengths)[*cc];
1809 break;
1810
1811 /* Handle literal characters */
1812
1813 case OP_CHAR:
1814 case OP_CHARI:
1815 case OP_NOT:
1816 case OP_NOTI:
1817 branchlength++;
1818 cc += 2;
1819 #ifdef SUPPORT_UTF
1820 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1821 #endif
1822 break;
1823
1824 /* Handle exact repetitions. The count is already in characters, but we
1825 need to skip over a multibyte character in UTF8 mode. */
1826
1827 case OP_EXACT:
1828 case OP_EXACTI:
1829 case OP_NOTEXACT:
1830 case OP_NOTEXACTI:
1831 branchlength += (int)GET2(cc,1);
1832 cc += 2 + IMM2_SIZE;
1833 #ifdef SUPPORT_UTF
1834 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1835 #endif
1836 break;
1837
1838 case OP_TYPEEXACT:
1839 branchlength += GET2(cc,1);
1840 if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1841 cc += 2;
1842 cc += 1 + IMM2_SIZE + 1;
1843 break;
1844
1845 /* Handle single-char matchers */
1846
1847 case OP_PROP:
1848 case OP_NOTPROP:
1849 cc += 2;
1850 /* Fall through */
1851
1852 case OP_HSPACE:
1853 case OP_VSPACE:
1854 case OP_NOT_HSPACE:
1855 case OP_NOT_VSPACE:
1856 case OP_NOT_DIGIT:
1857 case OP_DIGIT:
1858 case OP_NOT_WHITESPACE:
1859 case OP_WHITESPACE:
1860 case OP_NOT_WORDCHAR:
1861 case OP_WORDCHAR:
1862 case OP_ANY:
1863 case OP_ALLANY:
1864 branchlength++;
1865 cc++;
1866 break;
1867
1868 /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1869 otherwise \C is coded as OP_ALLANY. */
1870
1871 case OP_ANYBYTE:
1872 return -2;
1873
1874 /* Check a class for variable quantification */
1875
1876 case OP_CLASS:
1877 case OP_NCLASS:
1878 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1879 case OP_XCLASS:
1880 /* The original code caused an unsigned overflow in 64 bit systems,
1881 so now we use a conditional statement. */
1882 if (op == OP_XCLASS)
1883 cc += GET(cc, 1);
1884 else
1885 cc += PRIV(OP_lengths)[OP_CLASS];
1886 #else
1887 cc += PRIV(OP_lengths)[OP_CLASS];
1888 #endif
1889
1890 switch (*cc)
1891 {
1892 case OP_CRSTAR:
1893 case OP_CRMINSTAR:
1894 case OP_CRPLUS:
1895 case OP_CRMINPLUS:
1896 case OP_CRQUERY:
1897 case OP_CRMINQUERY:
1898 case OP_CRPOSSTAR:
1899 case OP_CRPOSPLUS:
1900 case OP_CRPOSQUERY:
1901 return -1;
1902
1903 case OP_CRRANGE:
1904 case OP_CRMINRANGE:
1905 case OP_CRPOSRANGE:
1906 if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1907 branchlength += (int)GET2(cc,1);
1908 cc += 1 + 2 * IMM2_SIZE;
1909 break;
1910
1911 default:
1912 branchlength++;
1913 }
1914 break;
1915
1916 /* Anything else is variable length */
1917
1918 case OP_ANYNL:
1919 case OP_BRAMINZERO:
1920 case OP_BRAPOS:
1921 case OP_BRAPOSZERO:
1922 case OP_BRAZERO:
1923 case OP_CBRAPOS:
1924 case OP_EXTUNI:
1925 case OP_KETRMAX:
1926 case OP_KETRMIN:
1927 case OP_KETRPOS:
1928 case OP_MINPLUS:
1929 case OP_MINPLUSI:
1930 case OP_MINQUERY:
1931 case OP_MINQUERYI:
1932 case OP_MINSTAR:
1933 case OP_MINSTARI:
1934 case OP_MINUPTO:
1935 case OP_MINUPTOI:
1936 case OP_NOTMINPLUS:
1937 case OP_NOTMINPLUSI:
1938 case OP_NOTMINQUERY:
1939 case OP_NOTMINQUERYI:
1940 case OP_NOTMINSTAR:
1941 case OP_NOTMINSTARI:
1942 case OP_NOTMINUPTO:
1943 case OP_NOTMINUPTOI:
1944 case OP_NOTPLUS:
1945 case OP_NOTPLUSI:
1946 case OP_NOTPOSPLUS:
1947 case OP_NOTPOSPLUSI:
1948 case OP_NOTPOSQUERY:
1949 case OP_NOTPOSQUERYI:
1950 case OP_NOTPOSSTAR:
1951 case OP_NOTPOSSTARI:
1952 case OP_NOTPOSUPTO:
1953 case OP_NOTPOSUPTOI:
1954 case OP_NOTQUERY:
1955 case OP_NOTQUERYI:
1956 case OP_NOTSTAR:
1957 case OP_NOTSTARI:
1958 case OP_NOTUPTO:
1959 case OP_NOTUPTOI:
1960 case OP_PLUS:
1961 case OP_PLUSI:
1962 case OP_POSPLUS:
1963 case OP_POSPLUSI:
1964 case OP_POSQUERY:
1965 case OP_POSQUERYI:
1966 case OP_POSSTAR:
1967 case OP_POSSTARI:
1968 case OP_POSUPTO:
1969 case OP_POSUPTOI:
1970 case OP_QUERY:
1971 case OP_QUERYI:
1972 case OP_REF:
1973 case OP_REFI:
1974 case OP_DNREF:
1975 case OP_DNREFI:
1976 case OP_SBRA:
1977 case OP_SBRAPOS:
1978 case OP_SCBRA:
1979 case OP_SCBRAPOS:
1980 case OP_SCOND:
1981 case OP_SKIPZERO:
1982 case OP_STAR:
1983 case OP_STARI:
1984 case OP_TYPEMINPLUS:
1985 case OP_TYPEMINQUERY:
1986 case OP_TYPEMINSTAR:
1987 case OP_TYPEMINUPTO:
1988 case OP_TYPEPLUS:
1989 case OP_TYPEPOSPLUS:
1990 case OP_TYPEPOSQUERY:
1991 case OP_TYPEPOSSTAR:
1992 case OP_TYPEPOSUPTO:
1993 case OP_TYPEQUERY:
1994 case OP_TYPESTAR:
1995 case OP_TYPEUPTO:
1996 case OP_UPTO:
1997 case OP_UPTOI:
1998 return -1;
1999
2000 /* Catch unrecognized opcodes so that when new ones are added they
2001 are not forgotten, as has happened in the past. */
2002
2003 default:
2004 return -4;
2005 }
2006 }
2007 /* Control never gets here */
2008 }
2009
2010
2011
2012 /*************************************************
2013 * Scan compiled regex for specific bracket *
2014 *************************************************/
2015
2016 /* This little function scans through a compiled pattern until it finds a
2017 capturing bracket with the given number, or, if the number is negative, an
2018 instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2019 so that it can be called from pcre_study() when finding the minimum matching
2020 length.
2021
2022 Arguments:
2023 code points to start of expression
2024 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2025 number the required bracket number or negative to find a lookbehind
2026
2027 Returns: pointer to the opcode for the bracket, or NULL if not found
2028 */
2029
2030 const pcre_uchar *
2031 PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2032 {
2033 for (;;)
2034 {
2035 register pcre_uchar c = *code;
2036
2037 if (c == OP_END) return NULL;
2038
2039 /* XCLASS is used for classes that cannot be represented just by a bit
2040 map. This includes negated single high-valued characters. The length in
2041 the table is zero; the actual length is stored in the compiled code. */
2042
2043 if (c == OP_XCLASS) code += GET(code, 1);
2044
2045 /* Handle recursion */
2046
2047 else if (c == OP_REVERSE)
2048 {
2049 if (number < 0) return (pcre_uchar *)code;
2050 code += PRIV(OP_lengths)[c];
2051 }
2052
2053 /* Handle capturing bracket */
2054
2055 else if (c == OP_CBRA || c == OP_SCBRA ||
2056 c == OP_CBRAPOS || c == OP_SCBRAPOS)
2057 {
2058 int n = (int)GET2(code, 1+LINK_SIZE);
2059 if (n == number) return (pcre_uchar *)code;
2060 code += PRIV(OP_lengths)[c];
2061 }
2062
2063 /* Otherwise, we can get the item's length from the table, except that for
2064 repeated character types, we have to test for \p and \P, which have an extra
2065 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2066 must add in its length. */
2067
2068 else
2069 {
2070 switch(c)
2071 {
2072 case OP_TYPESTAR:
2073 case OP_TYPEMINSTAR:
2074 case OP_TYPEPLUS:
2075 case OP_TYPEMINPLUS:
2076 case OP_TYPEQUERY:
2077 case OP_TYPEMINQUERY:
2078 case OP_TYPEPOSSTAR:
2079 case OP_TYPEPOSPLUS:
2080 case OP_TYPEPOSQUERY:
2081 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2082 break;
2083
2084 case OP_TYPEUPTO:
2085 case OP_TYPEMINUPTO:
2086 case OP_TYPEEXACT:
2087 case OP_TYPEPOSUPTO:
2088 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2089 code += 2;
2090 break;
2091
2092 case OP_MARK:
2093 case OP_PRUNE_ARG:
2094 case OP_SKIP_ARG:
2095 case OP_THEN_ARG:
2096 code += code[1];
2097 break;
2098 }
2099
2100 /* Add in the fixed length from the table */
2101
2102 code += PRIV(OP_lengths)[c];
2103
2104 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2105 a multi-byte character. The length in the table is a minimum, so we have to
2106 arrange to skip the extra bytes. */
2107
2108 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2109 if (utf) switch(c)
2110 {
2111 case OP_CHAR:
2112 case OP_CHARI:
2113 case OP_EXACT:
2114 case OP_EXACTI:
2115 case OP_UPTO:
2116 case OP_UPTOI:
2117 case OP_MINUPTO:
2118 case OP_MINUPTOI:
2119 case OP_POSUPTO:
2120 case OP_POSUPTOI:
2121 case OP_STAR:
2122 case OP_STARI:
2123 case OP_MINSTAR:
2124 case OP_MINSTARI:
2125 case OP_POSSTAR:
2126 case OP_POSSTARI:
2127 case OP_PLUS:
2128 case OP_PLUSI:
2129 case OP_MINPLUS:
2130 case OP_MINPLUSI:
2131 case OP_POSPLUS:
2132 case OP_POSPLUSI:
2133 case OP_QUERY:
2134 case OP_QUERYI:
2135 case OP_MINQUERY:
2136 case OP_MINQUERYI:
2137 case OP_POSQUERY:
2138 case OP_POSQUERYI:
2139 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2140 break;
2141 }
2142 #else
2143 (void)(utf); /* Keep compiler happy by referencing function argument */
2144 #endif
2145 }
2146 }
2147 }
2148
2149
2150
2151 /*************************************************
2152 * Scan compiled regex for recursion reference *
2153 *************************************************/
2154
2155 /* This little function scans through a compiled pattern until it finds an
2156 instance of OP_RECURSE.
2157
2158 Arguments:
2159 code points to start of expression
2160 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2161
2162 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2163 */
2164
2165 static const pcre_uchar *
2166 find_recurse(const pcre_uchar *code, BOOL utf)
2167 {
2168 for (;;)
2169 {
2170 register pcre_uchar c = *code;
2171 if (c == OP_END) return NULL;
2172 if (c == OP_RECURSE) return code;
2173
2174 /* XCLASS is used for classes that cannot be represented just by a bit
2175 map. This includes negated single high-valued characters. The length in
2176 the table is zero; the actual length is stored in the compiled code. */
2177
2178 if (c == OP_XCLASS) code += GET(code, 1);
2179
2180 /* Otherwise, we can get the item's length from the table, except that for
2181 repeated character types, we have to test for \p and \P, which have an extra
2182 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2183 must add in its length. */
2184
2185 else
2186 {
2187 switch(c)
2188 {
2189 case OP_TYPESTAR:
2190 case OP_TYPEMINSTAR:
2191 case OP_TYPEPLUS:
2192 case OP_TYPEMINPLUS:
2193 case OP_TYPEQUERY:
2194 case OP_TYPEMINQUERY:
2195 case OP_TYPEPOSSTAR:
2196 case OP_TYPEPOSPLUS:
2197 case OP_TYPEPOSQUERY:
2198 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2199 break;
2200
2201 case OP_TYPEPOSUPTO:
2202 case OP_TYPEUPTO:
2203 case OP_TYPEMINUPTO:
2204 case OP_TYPEEXACT:
2205 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2206 code += 2;
2207 break;
2208
2209 case OP_MARK:
2210 case OP_PRUNE_ARG:
2211 case OP_SKIP_ARG:
2212 case OP_THEN_ARG:
2213 code += code[1];
2214 break;
2215 }
2216
2217 /* Add in the fixed length from the table */
2218
2219 code += PRIV(OP_lengths)[c];
2220
2221 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2222 by a multi-byte character. The length in the table is a minimum, so we have
2223 to arrange to skip the extra bytes. */
2224
2225 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2226 if (utf) switch(c)
2227 {
2228 case OP_CHAR:
2229 case OP_CHARI:
2230 case OP_NOT:
2231 case OP_NOTI:
2232 case OP_EXACT:
2233 case OP_EXACTI:
2234 case OP_NOTEXACT:
2235 case OP_NOTEXACTI:
2236 case OP_UPTO:
2237 case OP_UPTOI:
2238 case OP_NOTUPTO:
2239 case OP_NOTUPTOI:
2240 case OP_MINUPTO:
2241 case OP_MINUPTOI:
2242 case OP_NOTMINUPTO:
2243 case OP_NOTMINUPTOI:
2244 case OP_POSUPTO:
2245 case OP_POSUPTOI:
2246 case OP_NOTPOSUPTO:
2247 case OP_NOTPOSUPTOI:
2248 case OP_STAR:
2249 case OP_STARI:
2250 case OP_NOTSTAR:
2251 case OP_NOTSTARI:
2252 case OP_MINSTAR:
2253 case OP_MINSTARI:
2254 case OP_NOTMINSTAR:
2255 case OP_NOTMINSTARI:
2256 case OP_POSSTAR:
2257 case OP_POSSTARI:
2258 case OP_NOTPOSSTAR:
2259 case OP_NOTPOSSTARI:
2260 case OP_PLUS:
2261 case OP_PLUSI:
2262 case OP_NOTPLUS:
2263 case OP_NOTPLUSI:
2264 case OP_MINPLUS:
2265 case OP_MINPLUSI:
2266 case OP_NOTMINPLUS:
2267 case OP_NOTMINPLUSI:
2268 case OP_POSPLUS:
2269 case OP_POSPLUSI:
2270 case OP_NOTPOSPLUS:
2271 case OP_NOTPOSPLUSI:
2272 case OP_QUERY:
2273 case OP_QUERYI:
2274 case OP_NOTQUERY:
2275 case OP_NOTQUERYI:
2276 case OP_MINQUERY:
2277 case OP_MINQUERYI:
2278 case OP_NOTMINQUERY:
2279 case OP_NOTMINQUERYI:
2280 case OP_POSQUERY:
2281 case OP_POSQUERYI:
2282 case OP_NOTPOSQUERY:
2283 case OP_NOTPOSQUERYI:
2284 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2285 break;
2286 }
2287 #else
2288 (void)(utf); /* Keep compiler happy by referencing function argument */
2289 #endif
2290 }
2291 }
2292 }
2293
2294
2295
2296 /*************************************************
2297 * Scan compiled branch for non-emptiness *
2298 *************************************************/
2299
2300 /* This function scans through a branch of a compiled pattern to see whether it
2301 can match the empty string or not. It is called from could_be_empty()
2302 below and from compile_branch() when checking for an unlimited repeat of a
2303 group that can match nothing. Note that first_significant_code() skips over
2304 backward and negative forward assertions when its final argument is TRUE. If we
2305 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2306 bracket whose current branch will already have been scanned.
2307
2308 Arguments:
2309 code points to start of search
2310 endcode points to where to stop
2311 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2312 cd contains pointers to tables etc.
2313 recurses chain of recurse_check to catch mutual recursion
2314
2315 Returns: TRUE if what is matched could be empty
2316 */
2317
2318 typedef struct recurse_check {
2319 struct recurse_check *prev;
2320 const pcre_uchar *group;
2321 } recurse_check;
2322
2323 static BOOL
2324 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2325 BOOL utf, compile_data *cd, recurse_check *recurses)
2326 {
2327 register pcre_uchar c;
2328 recurse_check this_recurse;
2329
2330 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2331 code < endcode;
2332 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2333 {
2334 const pcre_uchar *ccode;
2335
2336 c = *code;
2337
2338 /* Skip over forward assertions; the other assertions are skipped by
2339 first_significant_code() with a TRUE final argument. */
2340
2341 if (c == OP_ASSERT)
2342 {
2343 do code += GET(code, 1); while (*code == OP_ALT);
2344 c = *code;
2345 continue;
2346 }
2347
2348 /* For a recursion/subroutine call, if its end has been reached, which
2349 implies a backward reference subroutine call, we can scan it. If it's a
2350 forward reference subroutine call, we can't. To detect forward reference
2351 we have to scan up the list that is kept in the workspace. This function is
2352 called only when doing the real compile, not during the pre-compile that
2353 measures the size of the compiled pattern. */
2354
2355 if (c == OP_RECURSE)
2356 {
2357 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2358 BOOL empty_branch;
2359
2360 /* Test for forward reference or uncompleted reference. This is disabled
2361 when called to scan a completed pattern by setting cd->start_workspace to
2362 NULL. */
2363
2364 if (cd->start_workspace != NULL)
2365 {
2366 const pcre_uchar *tcode;
2367 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2368 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2369 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2370 }
2371
2372 /* If we are scanning a completed pattern, there are no forward references
2373 and all groups are complete. We need to detect whether this is a recursive
2374 call, as otherwise there will be an infinite loop. If it is a recursion,
2375 just skip over it. Simple recursions are easily detected. For mutual
2376 recursions we keep a chain on the stack. */
2377
2378 else
2379 {
2380 recurse_check *r = recurses;
2381 const pcre_uchar *endgroup = scode;
2382
2383 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2384 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2385
2386 for (r = recurses; r != NULL; r = r->prev)
2387 if (r->group == scode) break;
2388 if (r != NULL) continue; /* Mutual recursion */
2389 }
2390
2391 /* Completed reference; scan the referenced group, remembering it on the
2392 stack chain to detect mutual recursions. */
2393
2394 empty_branch = FALSE;
2395 this_recurse.prev = recurses;
2396 this_recurse.group = scode;
2397
2398 do
2399 {
2400 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2401 {
2402 empty_branch = TRUE;
2403 break;
2404 }
2405 scode += GET(scode, 1);
2406 }
2407 while (*scode == OP_ALT);
2408
2409 if (!empty_branch) return FALSE; /* All branches are non-empty */
2410 continue;
2411 }
2412
2413 /* Groups with zero repeats can of course be empty; skip them. */
2414
2415 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2416 c == OP_BRAPOSZERO)
2417 {
2418 code += PRIV(OP_lengths)[c];
2419 do code += GET(code, 1); while (*code == OP_ALT);
2420 c = *code;
2421 continue;
2422 }
2423
2424 /* A nested group that is already marked as "could be empty" can just be
2425 skipped. */
2426
2427 if (c == OP_SBRA || c == OP_SBRAPOS ||
2428 c == OP_SCBRA || c == OP_SCBRAPOS)
2429 {
2430 do code += GET(code, 1); while (*code == OP_ALT);
2431 c = *code;
2432 continue;
2433 }
2434
2435 /* For other groups, scan the branches. */
2436
2437 if (c == OP_BRA || c == OP_BRAPOS ||
2438 c == OP_CBRA || c == OP_CBRAPOS ||
2439 c == OP_ONCE || c == OP_ONCE_NC ||
2440 c == OP_COND)
2441 {
2442 BOOL empty_branch;
2443 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2444
2445 /* If a conditional group has only one branch, there is a second, implied,
2446 empty branch, so just skip over the conditional, because it could be empty.
2447 Otherwise, scan the individual branches of the group. */
2448
2449 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2450 code += GET(code, 1);
2451 else
2452 {
2453 empty_branch = FALSE;
2454 do
2455 {
2456 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2457 empty_branch = TRUE;
2458 code += GET(code, 1);
2459 }
2460 while (*code == OP_ALT);
2461 if (!empty_branch) return FALSE; /* All branches are non-empty */
2462 }
2463
2464 c = *code;
2465 continue;
2466 }
2467
2468 /* Handle the other opcodes */
2469
2470 switch (c)
2471 {
2472 /* Check for quantifiers after a class. XCLASS is used for classes that
2473 cannot be represented just by a bit map. This includes negated single
2474 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2475 actual length is stored in the compiled code, so we must update "code"
2476 here. */
2477
2478 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2479 case OP_XCLASS:
2480 ccode = code += GET(code, 1);
2481 goto CHECK_CLASS_REPEAT;
2482 #endif
2483
2484 case OP_CLASS:
2485 case OP_NCLASS:
2486 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2487
2488 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2489 CHECK_CLASS_REPEAT:
2490 #endif
2491
2492 switch (*ccode)
2493 {
2494 case OP_CRSTAR: /* These could be empty; continue */
2495 case OP_CRMINSTAR:
2496 case OP_CRQUERY:
2497 case OP_CRMINQUERY:
2498 case OP_CRPOSSTAR:
2499 case OP_CRPOSQUERY:
2500 break;
2501
2502 default: /* Non-repeat => class must match */
2503 case OP_CRPLUS: /* These repeats aren't empty */
2504 case OP_CRMINPLUS:
2505 case OP_CRPOSPLUS:
2506 return FALSE;
2507
2508 case OP_CRRANGE:
2509 case OP_CRMINRANGE:
2510 case OP_CRPOSRANGE:
2511 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2512 break;
2513 }
2514 break;
2515
2516 /* Opcodes that must match a character */
2517
2518 case OP_ANY:
2519 case OP_ALLANY:
2520 case OP_ANYBYTE:
2521
2522 case OP_PROP:
2523 case OP_NOTPROP:
2524 case OP_ANYNL:
2525
2526 case OP_NOT_HSPACE:
2527 case OP_HSPACE:
2528 case OP_NOT_VSPACE:
2529 case OP_VSPACE:
2530 case OP_EXTUNI:
2531
2532 case OP_NOT_DIGIT:
2533 case OP_DIGIT:
2534 case OP_NOT_WHITESPACE:
2535 case OP_WHITESPACE:
2536 case OP_NOT_WORDCHAR:
2537 case OP_WORDCHAR:
2538
2539 case OP_CHAR:
2540 case OP_CHARI:
2541 case OP_NOT:
2542 case OP_NOTI:
2543
2544 case OP_PLUS:
2545 case OP_PLUSI:
2546 case OP_MINPLUS:
2547 case OP_MINPLUSI:
2548
2549 case OP_NOTPLUS:
2550 case OP_NOTPLUSI:
2551 case OP_NOTMINPLUS:
2552 case OP_NOTMINPLUSI:
2553
2554 case OP_POSPLUS:
2555 case OP_POSPLUSI:
2556 case OP_NOTPOSPLUS:
2557 case OP_NOTPOSPLUSI:
2558
2559 case OP_EXACT:
2560 case OP_EXACTI:
2561 case OP_NOTEXACT:
2562 case OP_NOTEXACTI:
2563
2564 case OP_TYPEPLUS:
2565 case OP_TYPEMINPLUS:
2566 case OP_TYPEPOSPLUS:
2567 case OP_TYPEEXACT:
2568
2569 return FALSE;
2570
2571 /* These are going to continue, as they may be empty, but we have to
2572 fudge the length for the \p and \P cases. */
2573
2574 case OP_TYPESTAR:
2575 case OP_TYPEMINSTAR:
2576 case OP_TYPEPOSSTAR:
2577 case OP_TYPEQUERY:
2578 case OP_TYPEMINQUERY:
2579 case OP_TYPEPOSQUERY:
2580 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2581 break;
2582
2583 /* Same for these */
2584
2585 case OP_TYPEUPTO:
2586 case OP_TYPEMINUPTO:
2587 case OP_TYPEPOSUPTO:
2588 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2589 code += 2;
2590 break;
2591
2592 /* End of branch */
2593
2594 case OP_KET:
2595 case OP_KETRMAX:
2596 case OP_KETRMIN:
2597 case OP_KETRPOS:
2598 case OP_ALT:
2599 return TRUE;
2600
2601 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2602 MINUPTO, and POSUPTO and their caseless and negative versions may be
2603 followed by a multibyte character. */
2604
2605 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2606 case OP_STAR:
2607 case OP_STARI:
2608 case OP_NOTSTAR:
2609 case OP_NOTSTARI:
2610
2611 case OP_MINSTAR:
2612 case OP_MINSTARI:
2613 case OP_NOTMINSTAR:
2614 case OP_NOTMINSTARI:
2615
2616 case OP_POSSTAR:
2617 case OP_POSSTARI:
2618 case OP_NOTPOSSTAR:
2619 case OP_NOTPOSSTARI:
2620
2621 case OP_QUERY:
2622 case OP_QUERYI:
2623 case OP_NOTQUERY:
2624 case OP_NOTQUERYI:
2625
2626 case OP_MINQUERY:
2627 case OP_MINQUERYI:
2628 case OP_NOTMINQUERY:
2629 case OP_NOTMINQUERYI:
2630
2631 case OP_POSQUERY:
2632 case OP_POSQUERYI:
2633 case OP_NOTPOSQUERY:
2634 case OP_NOTPOSQUERYI:
2635
2636 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2637 break;
2638
2639 case OP_UPTO:
2640 case OP_UPTOI:
2641 case OP_NOTUPTO:
2642 case OP_NOTUPTOI:
2643
2644 case OP_MINUPTO:
2645 case OP_MINUPTOI:
2646 case OP_NOTMINUPTO:
2647 case OP_NOTMINUPTOI:
2648
2649 case OP_POSUPTO:
2650 case OP_POSUPTOI:
2651 case OP_NOTPOSUPTO:
2652 case OP_NOTPOSUPTOI:
2653
2654 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2655 break;
2656 #endif
2657
2658 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2659 string. */
2660
2661 case OP_MARK:
2662 case OP_PRUNE_ARG:
2663 case OP_SKIP_ARG:
2664 case OP_THEN_ARG:
2665 code += code[1];
2666 break;
2667
2668 /* None of the remaining opcodes are required to match a character. */
2669
2670 default:
2671 break;
2672 }
2673 }
2674
2675 return TRUE;
2676 }
2677
2678
2679
2680 /*************************************************
2681 * Scan compiled regex for non-emptiness *
2682 *************************************************/
2683
2684 /* This function is called to check for left recursive calls. We want to check
2685 the current branch of the current pattern to see if it could match the empty
2686 string. If it could, we must look outwards for branches at other levels,
2687 stopping when we pass beyond the bracket which is the subject of the recursion.
2688 This function is called only during the real compile, not during the
2689 pre-compile.
2690
2691 Arguments:
2692 code points to start of the recursion
2693 endcode points to where to stop (current RECURSE item)
2694 bcptr points to the chain of current (unclosed) branch starts
2695 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2696 cd pointers to tables etc
2697
2698 Returns: TRUE if what is matched could be empty
2699 */
2700
2701 static BOOL
2702 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2703 branch_chain *bcptr, BOOL utf, compile_data *cd)
2704 {
2705 while (bcptr != NULL && bcptr->current_branch >= code)
2706 {
2707 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2708 return FALSE;
2709 bcptr = bcptr->outer;
2710 }
2711 return TRUE;
2712 }
2713
2714
2715
2716 /*************************************************
2717 * Base opcode of repeated opcodes *
2718 *************************************************/
2719
2720 /* Returns the base opcode for repeated single character type opcodes. If the
2721 opcode is not a repeated character type, it returns with the original value.
2722
2723 Arguments: c opcode
2724 Returns: base opcode for the type
2725 */
2726
2727 static pcre_uchar
2728 get_repeat_base(pcre_uchar c)
2729 {
2730 return (c > OP_TYPEPOSUPTO)? c :
2731 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2732 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2733 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2734 (c >= OP_STARI)? OP_STARI :
2735 OP_STAR;
2736 }
2737
2738
2739
2740 #ifdef SUPPORT_UCP
2741 /*************************************************
2742 * Check a character and a property *
2743 *************************************************/
2744
2745 /* This function is called by check_auto_possessive() when a property item
2746 is adjacent to a fixed character.
2747
2748 Arguments:
2749 c the character
2750 ptype the property type
2751 pdata the data for the type
2752 negated TRUE if it's a negated property (\P or \p{^)
2753
2754 Returns: TRUE if auto-possessifying is OK
2755 */
2756
2757 static BOOL
2758 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2759 BOOL negated)
2760 {
2761 const pcre_uint32 *p;
2762 const ucd_record *prop = GET_UCD(c);
2763
2764 switch(ptype)
2765 {
2766 case PT_LAMP:
2767 return (prop->chartype == ucp_Lu ||
2768 prop->chartype == ucp_Ll ||
2769 prop->chartype == ucp_Lt) == negated;
2770
2771 case PT_GC:
2772 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2773
2774 case PT_PC:
2775 return (pdata == prop->chartype) == negated;
2776
2777 case PT_SC:
2778 return (pdata == prop->script) == negated;
2779
2780 /* These are specials */
2781
2782 case PT_ALNUM:
2783 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2784 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2785
2786 /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2787 means that Perl space and POSIX space are now identical. PCRE was changed
2788 at release 8.34. */
2789
2790 case PT_SPACE: /* Perl space */
2791 case PT_PXSPACE: /* POSIX space */
2792 switch(c)
2793 {
2794 HSPACE_CASES:
2795 VSPACE_CASES:
2796 return negated;
2797
2798 default:
2799 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2800 }
2801 break; /* Control never reaches here */
2802
2803 case PT_WORD:
2804 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2805 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2806 c == CHAR_UNDERSCORE) == negated;
2807
2808 case PT_CLIST:
2809 p = PRIV(ucd_caseless_sets) + prop->caseset;
2810 for (;;)
2811 {
2812 if (c < *p) return !negated;
2813 if (c == *p++) return negated;
2814 }
2815 break; /* Control never reaches here */
2816 }
2817
2818 return FALSE;
2819 }
2820 #endif /* SUPPORT_UCP */
2821
2822
2823
2824 /*************************************************
2825 * Fill the character property list *
2826 *************************************************/
2827
2828 /* Checks whether the code points to an opcode that can take part in auto-
2829 possessification, and if so, fills a list with its properties.
2830
2831 Arguments:
2832 code points to start of expression
2833 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2834 fcc points to case-flipping table
2835 list points to output list
2836 list[0] will be filled with the opcode
2837 list[1] will be non-zero if this opcode
2838 can match an empty character string
2839 list[2..7] depends on the opcode
2840
2841 Returns: points to the start of the next opcode if *code is accepted
2842 NULL if *code is not accepted
2843 */
2844
2845 static const pcre_uchar *
2846 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2847 const pcre_uint8 *fcc, pcre_uint32 *list)
2848 {
2849 pcre_uchar c = *code;
2850 const pcre_uchar *end;
2851 const pcre_uint32 *clist_src;
2852 pcre_uint32 *clist_dest;
2853 pcre_uint32 chr;
2854 pcre_uchar base;
2855
2856 list[0] = c;
2857 list[1] = FALSE;
2858 code++;
2859
2860 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2861 {
2862 base = get_repeat_base(c);
2863 c -= (base - OP_STAR);
2864
2865 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2866 code += IMM2_SIZE;
2867
2868 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2869
2870 switch(base)
2871 {
2872 case OP_STAR:
2873 list[0] = OP_CHAR;
2874 break;
2875
2876 case OP_STARI:
2877 list[0] = OP_CHARI;
2878 break;
2879
2880 case OP_NOTSTAR:
2881 list[0] = OP_NOT;
2882 break;
2883
2884 case OP_NOTSTARI:
2885 list[0] = OP_NOTI;
2886 break;
2887
2888 case OP_TYPESTAR:
2889 list[0] = *code;
2890 code++;
2891 break;
2892 }
2893 c = list[0];
2894 }
2895
2896 switch(c)
2897 {
2898 case OP_NOT_DIGIT:
2899 case OP_DIGIT:
2900 case OP_NOT_WHITESPACE:
2901 case OP_WHITESPACE:
2902 case OP_NOT_WORDCHAR:
2903 case OP_WORDCHAR:
2904 case OP_ANY:
2905 case OP_ALLANY:
2906 case OP_ANYNL:
2907 case OP_NOT_HSPACE:
2908 case OP_HSPACE:
2909 case OP_NOT_VSPACE:
2910 case OP_VSPACE:
2911 case OP_EXTUNI:
2912 case OP_EODN:
2913 case OP_EOD:
2914 case OP_DOLL:
2915 case OP_DOLLM:
2916 return code;
2917
2918 case OP_CHAR:
2919 case OP_NOT:
2920 GETCHARINCTEST(chr, code);
2921 list[2] = chr;
2922 list[3] = NOTACHAR;
2923 return code;
2924
2925 case OP_CHARI:
2926 case OP_NOTI:
2927 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2928 GETCHARINCTEST(chr, code);
2929 list[2] = chr;
2930
2931 #ifdef SUPPORT_UCP
2932 if (chr < 128 || (chr < 256 && !utf))
2933 list[3] = fcc[chr];
2934 else
2935 list[3] = UCD_OTHERCASE(chr);
2936 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2937 list[3] = (chr < 256) ? fcc[chr] : chr;
2938 #else
2939 list[3] = fcc[chr];
2940 #endif
2941
2942 /* The othercase might be the same value. */
2943
2944 if (chr == list[3])
2945 list[3] = NOTACHAR;
2946 else
2947 list[4] = NOTACHAR;
2948 return code;
2949
2950 #ifdef SUPPORT_UCP
2951 case OP_PROP:
2952 case OP_NOTPROP:
2953 if (code[0] != PT_CLIST)
2954 {
2955 list[2] = code[0];
2956 list[3] = code[1];
2957 return code + 2;
2958 }
2959
2960 /* Convert only if we have enough space. */
2961
2962 clist_src = PRIV(ucd_caseless_sets) + code[1];
2963 clist_dest = list + 2;
2964 code += 2;
2965
2966 do {
2967 if (clist_dest >= list + 8)
2968 {
2969 /* Early return if there is not enough space. This should never
2970 happen, since all clists are shorter than 5 character now. */
2971 list[2] = code[0];
2972 list[3] = code[1];
2973 return code;
2974 }
2975 *clist_dest++ = *clist_src;
2976 }
2977 while(*clist_src++ != NOTACHAR);
2978
2979 /* All characters are stored. The terminating NOTACHAR
2980 is copied form the clist itself. */
2981
2982 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2983 return code;
2984 #endif
2985
2986 case OP_NCLASS:
2987 case OP_CLASS:
2988 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2989 case OP_XCLASS:
2990 if (c == OP_XCLASS)
2991 end = code + GET(code, 0) - 1;
2992 else
2993 #endif
2994 end = code + 32 / sizeof(pcre_uchar);
2995
2996 switch(*end)
2997 {
2998 case OP_CRSTAR:
2999 case OP_CRMINSTAR:
3000 case OP_CRQUERY:
3001 case OP_CRMINQUERY:
3002 case OP_CRPOSSTAR:
3003 case OP_CRPOSQUERY:
3004 list[1] = TRUE;
3005 end++;
3006 break;
3007
3008 case OP_CRPLUS:
3009 case OP_CRMINPLUS:
3010 case OP_CRPOSPLUS:
3011 end++;
3012 break;
3013
3014 case OP_CRRANGE:
3015 case OP_CRMINRANGE:
3016 case OP_CRPOSRANGE:
3017 list[1] = (GET2(end, 1) == 0);
3018 end += 1 + 2 * IMM2_SIZE;
3019 break;
3020 }
3021 list[2] = end - code;
3022 return end;
3023 }
3024 return NULL; /* Opcode not accepted */
3025 }
3026
3027
3028
3029 /*************************************************
3030 * Scan further character sets for match *
3031 *************************************************/
3032
3033 /* Checks whether the base and the current opcode have a common character, in
3034 which case the base cannot be possessified.
3035
3036 Arguments:
3037 code points to the byte code
3038 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3039 cd static compile data
3040 base_list the data list of the base opcode
3041
3042 Returns: TRUE if the auto-possessification is possible
3043 */
3044
3045 static BOOL
3046 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3047 const pcre_uint32 *base_list, const pcre_uchar *base_end)
3048 {
3049 pcre_uchar c;
3050 pcre_uint32 list[8];
3051 const pcre_uint32 *chr_ptr;
3052 const pcre_uint32 *ochr_ptr;
3053 const pcre_uint32 *list_ptr;
3054 const pcre_uchar *next_code;
3055 const pcre_uint8 *class_bitset;
3056 const pcre_uint32 *set1, *set2, *set_end;
3057 pcre_uint32 chr;
3058 BOOL accepted, invert_bits;
3059
3060 /* Note: the base_list[1] contains whether the current opcode has greedy
3061 (represented by a non-zero value) quantifier. This is a different from
3062 other character type lists, which stores here that the character iterator
3063 matches to an empty string (also represented by a non-zero value). */
3064
3065 for(;;)
3066 {
3067 /* All operations move the code pointer forward.
3068 Therefore infinite recursions are not possible. */
3069
3070 c = *code;
3071
3072 /* Skip over callouts */
3073
3074 if (c == OP_CALLOUT)
3075 {
3076 code += PRIV(OP_lengths)[c];
3077 continue;
3078 }
3079
3080 if (c == OP_ALT)
3081 {
3082 do code += GET(code, 1); while (*code == OP_ALT);
3083 c = *code;
3084 }
3085
3086 switch(c)
3087 {
3088 case OP_END:
3089 case OP_KETRPOS:
3090 /* TRUE only in greedy case. The non-greedy case could be replaced by
3091 an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3092 uses more memory, which we cannot get at this stage.) */
3093
3094 return base_list[1] != 0;
3095
3096 case OP_KET:
3097 /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3098 it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3099 cannot be converted to a possessive form. */
3100
3101 if (base_list[1] == 0) return FALSE;
3102
3103 switch(*(code - GET(code, 1)))
3104 {
3105 case OP_ASSERT:
3106 case OP_ASSERT_NOT:
3107 case OP_ASSERTBACK:
3108 case OP_ASSERTBACK_NOT:
3109 case OP_ONCE:
3110 case OP_ONCE_NC:
3111 /* Atomic sub-patterns and assertions can always auto-possessify their
3112 last iterator. */
3113 return TRUE;
3114 }
3115
3116 code += PRIV(OP_lengths)[c];
3117 continue;
3118
3119 case OP_ONCE:
3120 case OP_ONCE_NC:
3121 case OP_BRA:
3122 case OP_CBRA:
3123 next_code = code + GET(code, 1);
3124 code += PRIV(OP_lengths)[c];
3125
3126 while (*next_code == OP_ALT)
3127 {
3128 if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3129 code = next_code + 1 + LINK_SIZE;
3130 next_code += GET(next_code, 1);
3131 }
3132 continue;
3133
3134 case OP_BRAZERO:
3135 case OP_BRAMINZERO:
3136
3137 next_code = code + 1;
3138 if (*next_code != OP_BRA && *next_code != OP_CBRA
3139 && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3140
3141 do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3142
3143 /* The bracket content will be checked by the
3144 OP_BRA/OP_CBRA case above. */
3145 next_code += 1 + LINK_SIZE;
3146 if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3147 return FALSE;
3148
3149 code += PRIV(OP_lengths)[c];
3150 continue;
3151 }
3152
3153 /* Check for a supported opcode, and load its properties. */
3154
3155 code = get_chr_property_list(code, utf, cd->fcc, list);
3156 if (code == NULL) return FALSE; /* Unsupported */
3157
3158 /* If either opcode is a small character list, set pointers for comparing
3159 characters from that list with another list, or with a property. */
3160
3161 if (base_list[0] == OP_CHAR)
3162 {
3163 chr_ptr = base_list + 2;
3164 list_ptr = list;
3165 }
3166 else if (list[0] == OP_CHAR)
3167 {
3168 chr_ptr = list + 2;
3169 list_ptr = base_list;
3170 }
3171
3172 /* Character bitsets can also be compared to certain opcodes. */
3173
3174 else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3175 #ifdef COMPILE_PCRE8
3176 /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3177 || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3178 #endif
3179 )
3180 {
3181 #ifdef COMPILE_PCRE8
3182 if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3183 #else
3184 if (base_list[0] == OP_CLASS)
3185 #endif
3186 {
3187 set1 = (pcre_uint32 *)(base_end - base_list[2]);
3188 list_ptr = list;
3189 }
3190 else
3191 {
3192 set1 = (pcre_uint32 *)(code - list[2]);
3193 list_ptr = base_list;
3194 }
3195
3196 invert_bits = FALSE;
3197 switch(list_ptr[0])
3198 {
3199 case OP_CLASS:
3200 case OP_NCLASS:
3201 set2 = (pcre_uint32 *)
3202 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3203 break;
3204
3205 /* OP_XCLASS cannot be supported here, because its bitset
3206 is not necessarily complete. E.g: [a-\0x{200}] is stored
3207 as a character range, and the appropriate bits are not set. */
3208
3209 case OP_NOT_DIGIT:
3210 invert_bits = TRUE;
3211 /* Fall through */
3212 case OP_DIGIT:
3213 set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3214 break;
3215
3216 case OP_NOT_WHITESPACE:
3217 invert_bits = TRUE;
3218 /* Fall through */
3219 case OP_WHITESPACE:
3220 set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3221 break;
3222
3223 case OP_NOT_WORDCHAR:
3224 invert_bits = TRUE;
3225 /* Fall through */
3226 case OP_WORDCHAR:
3227 set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3228 break;
3229
3230 default:
3231 return FALSE;
3232 }
3233
3234 /* Compare 4 bytes to improve speed. */
3235 set_end = set1 + (32 / 4);
3236 if (invert_bits)
3237 {
3238 do
3239 {
3240 if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3241 }
3242 while (set1 < set_end);
3243 }
3244 else
3245 {
3246 do
3247 {
3248 if ((*set1++ & *set2++) != 0) return FALSE;
3249 }
3250 while (set1 < set_end);
3251 }
3252
3253 if (list[1] == 0) return TRUE;
3254 /* Might be an empty repeat. */
3255 continue;
3256 }
3257
3258 /* Some property combinations also acceptable. Unicode property opcodes are
3259 processed specially; the rest can be handled with a lookup table. */
3260
3261 else
3262 {
3263 pcre_uint32 leftop, rightop;
3264
3265 leftop = base_list[0];
3266 rightop = list[0];
3267
3268 #ifdef SUPPORT_UCP
3269 accepted = FALSE; /* Always set in non-unicode case. */
3270 if (leftop == OP_PROP || leftop == OP_NOTPROP)
3271 {
3272 if (rightop == OP_EOD)
3273 accepted = TRUE;
3274 else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3275 {
3276 int n;
3277 const pcre_uint8 *p;
3278 BOOL same = leftop == rightop;
3279 BOOL lisprop = leftop == OP_PROP;
3280 BOOL risprop = rightop == OP_PROP;
3281 BOOL bothprop = lisprop && risprop;
3282
3283 /* There's a table that specifies how each combination is to be
3284 processed:
3285 0 Always return FALSE (never auto-possessify)
3286 1 Character groups are distinct (possessify if both are OP_PROP)
3287 2 Check character categories in the same group (general or particular)
3288 3 Return TRUE if the two opcodes are not the same
3289 ... see comments below
3290 */
3291
3292 n = propposstab[base_list[2]][list[2]];
3293 switch(n)
3294 {
3295 case 0: break;
3296 case 1: accepted = bothprop; break;
3297 case 2: accepted = (base_list[3] == list[3]) != same; break;
3298 case 3: accepted = !same; break;
3299
3300 case 4: /* Left general category, right particular category */
3301 accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3302 break;
3303
3304 case 5: /* Right general category, left particular category */
3305 accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3306 break;
3307
3308 /* This code is logically tricky. Think hard before fiddling with it.
3309 The posspropstab table has four entries per row. Each row relates to
3310 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3311 Only WORD actually needs all four entries, but using repeats for the
3312 others means they can all use the same code below.
3313
3314 The first two entries in each row are Unicode general categories, and
3315 apply always, because all the characters they include are part of the
3316 PCRE character set. The third and fourth entries are a general and a
3317 particular category, respectively, that include one or more relevant
3318 characters. One or the other is used, depending on whether the check
3319 is for a general or a particular category. However, in both cases the
3320 category contains more characters than the specials that are defined
3321 for the property being tested against. Therefore, it cannot be used
3322 in a NOTPROP case.
3323
3324 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3325 Underscore is covered by ucp_P or ucp_Po. */
3326
3327 case 6: /* Left alphanum vs right general category */
3328 case 7: /* Left space vs right general category */
3329 case 8: /* Left word vs right general category */
3330 p = posspropstab[n-6];
3331 accepted = risprop && lisprop ==
3332 (list[3] != p[0] &&
3333 list[3] != p[1] &&
3334 (list[3] != p[2] || !lisprop));
3335 break;
3336
3337 case 9: /* Right alphanum vs left general category */
3338 case 10: /* Right space vs left general category */
3339 case 11: /* Right word vs left general category */
3340 p = posspropstab[n-9];
3341 accepted = lisprop && risprop ==
3342 (base_list[3] != p[0] &&
3343 base_list[3] != p[1] &&
3344 (base_list[3] != p[2] || !risprop));
3345 break;
3346
3347 case 12: /* Left alphanum vs right particular category */
3348 case 13: /* Left space vs right particular category */
3349 case 14: /* Left word vs right particular category */
3350 p = posspropstab[n-12];
3351 accepted = risprop && lisprop ==
3352 (catposstab[p[0]][list[3]] &&
3353 catposstab[p[1]][list[3]] &&
3354 (list[3] != p[3] || !lisprop));
3355 break;
3356
3357 case 15: /* Right alphanum vs left particular category */
3358 case 16: /* Right space vs left particular category */
3359 case 17: /* Right word vs left particular category */
3360 p = posspropstab[n-15];
3361 accepted = lisprop && risprop ==
3362 (catposstab[p[0]][base_list[3]] &&
3363 catposstab[p[1]][base_list[3]] &&
3364 (base_list[3] != p[3] || !risprop));
3365 break;
3366 }
3367 }
3368 }
3369
3370 else
3371 #endif /* SUPPORT_UCP */
3372
3373 accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3374 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3375 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3376
3377 if (!accepted)
3378 return FALSE;
3379
3380 if (list[1] == 0) return TRUE;
3381 /* Might be an empty repeat. */
3382 continue;
3383 }
3384
3385 /* Control reaches here only if one of the items is a small character list.
3386 All characters are checked against the other side. */
3387
3388 do
3389 {
3390 chr = *chr_ptr;
3391
3392 switch(list_ptr[0])
3393 {
3394 case OP_CHAR:
3395 ochr_ptr = list_ptr + 2;
3396 do
3397 {
3398 if (chr == *ochr_ptr) return FALSE;
3399 ochr_ptr++;
3400 }
3401 while(*ochr_ptr != NOTACHAR);
3402 break;
3403
3404 case OP_NOT:
3405 ochr_ptr = list_ptr + 2;
3406 do
3407 {
3408 if (chr == *ochr_ptr)
3409 break;
3410 ochr_ptr++;
3411 }
3412 while(*ochr_ptr != NOTACHAR);
3413 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3414 break;
3415
3416 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3417 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3418
3419 case OP_DIGIT:
3420 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3421 break;
3422
3423 case OP_NOT_DIGIT:
3424 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3425 break;
3426
3427 case OP_WHITESPACE:
3428 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3429 break;
3430
3431 case OP_NOT_WHITESPACE:
3432 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3433 break;
3434
3435 case OP_WORDCHAR:
3436 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3437 break;
3438
3439 case OP_NOT_WORDCHAR:
3440 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3441 break;
3442
3443 case OP_HSPACE:
3444 switch(chr)
3445 {
3446 HSPACE_CASES: return FALSE;
3447 default: break;
3448 }
3449 break;
3450
3451 case OP_NOT_HSPACE:
3452 switch(chr)
3453 {
3454 HSPACE_CASES: break;
3455 default: return FALSE;
3456 }
3457 break;
3458
3459 case OP_ANYNL:
3460 case OP_VSPACE:
3461 switch(chr)
3462 {
3463 VSPACE_CASES: return FALSE;
3464 default: break;
3465 }
3466 break;
3467
3468 case OP_NOT_VSPACE:
3469 switch(chr)
3470 {
3471 VSPACE_CASES: break;
3472 default: return FALSE;
3473 }
3474 break;
3475
3476 case OP_DOLL:
3477 case OP_EODN:
3478 switch (chr)
3479 {
3480 case CHAR_CR:
3481 case CHAR_LF:
3482 case CHAR_VT:
3483 case CHAR_FF:
3484 case CHAR_NEL:
3485 #ifndef EBCDIC
3486 case 0x2028:
3487 case 0x2029:
3488 #endif /* Not EBCDIC */
3489 return FALSE;
3490 }
3491 break;
3492
3493 case OP_EOD: /* Can always possessify before \z */
3494 break;
3495
3496 #ifdef SUPPORT_UCP
3497 case OP_PROP:
3498 case OP_NOTPROP:
3499 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3500 list_ptr[0] == OP_NOTPROP))
3501 return FALSE;
3502 break;
3503 #endif
3504
3505 case OP_NCLASS:
3506 if (chr > 255) return FALSE;
3507 /* Fall through */
3508
3509 case OP_CLASS:
3510 if (chr > 255) break;
3511 class_bitset = (pcre_uint8 *)
3512 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3513 if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3514 break;
3515
3516 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3517 case OP_XCLASS:
3518 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3519 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3520 break;
3521 #endif
3522
3523 default:
3524 return FALSE;
3525 }
3526
3527 chr_ptr++;
3528 }
3529 while(*chr_ptr != NOTACHAR);
3530
3531 /* At least one character must be matched from this opcode. */
3532
3533 if (list[1] == 0) return TRUE;
3534 }
3535
3536 return FALSE;
3537 }
3538
3539
3540
3541 /*************************************************
3542 * Scan compiled regex for auto-possession *
3543 *************************************************/
3544
3545 /* Replaces single character iterations with their possessive alternatives
3546 if appropriate. This function modifies the compiled opcode!
3547
3548 Arguments:
3549 code points to start of the byte code
3550 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3551 cd static compile data
3552
3553 Returns: nothing
3554 */
3555
3556 static void
3557 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3558 {
3559 register pcre_uchar c;
3560 const pcre_uchar *end;
3561 pcre_uchar *repeat_opcode;
3562 pcre_uint32 list[8];
3563
3564 for (;;)
3565 {
3566 c = *code;
3567
3568 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3569 {
3570 c -= get_repeat_base(c) - OP_STAR;
3571 end = (c <= OP_MINUPTO) ?
3572 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3573 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3574
3575 if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3576 {
3577 switch(c)
3578 {
3579 case OP_STAR:
3580 *code += OP_POSSTAR - OP_STAR;
3581 break;
3582
3583 case OP_MINSTAR:
3584 *code += OP_POSSTAR - OP_MINSTAR;
3585 break;
3586
3587 case OP_PLUS:
3588 *code += OP_POSPLUS - OP_PLUS;
3589 break;
3590
3591 case OP_MINPLUS:
3592 *code += OP_POSPLUS - OP_MINPLUS;
3593 break;
3594
3595 case OP_QUERY:
3596 *code += OP_POSQUERY - OP_QUERY;
3597 break;
3598
3599 case OP_MINQUERY:
3600 *code += OP_POSQUERY - OP_MINQUERY;
3601 break;
3602
3603 case OP_UPTO:
3604 *code += OP_POSUPTO - OP_UPTO;
3605 break;
3606
3607 case OP_MINUPTO:
3608 *code += OP_MINUPTO - OP_UPTO;
3609 break;
3610 }
3611 }
3612 c = *code;
3613 }
3614 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3615 {
3616 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3617 if (c == OP_XCLASS)
3618 repeat_opcode = code + GET(code, 1);
3619 else
3620 #endif
3621 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3622
3623 c = *repeat_opcode;
3624 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3625 {
3626 /* end must not be NULL. */
3627 end = get_chr_property_list(code, utf, cd->fcc, list);
3628
3629 list[1] = (c & 1) == 0;
3630
3631 if (compare_opcodes(end, utf, cd, list, end))
3632 {
3633 switch (c)
3634 {
3635 case OP_CRSTAR:
3636 case OP_CRMINSTAR:
3637 *repeat_opcode = OP_CRPOSSTAR;
3638 break;
3639
3640 case OP_CRPLUS:
3641 case OP_CRMINPLUS:
3642 *repeat_opcode = OP_CRPOSPLUS;
3643 break;
3644
3645 case OP_CRQUERY:
3646 case OP_CRMINQUERY:
3647 *repeat_opcode = OP_CRPOSQUERY;
3648 break;
3649
3650 case OP_CRRANGE:
3651 case OP_CRMINRANGE:
3652 *repeat_opcode = OP_CRPOSRANGE;
3653 break;
3654 }
3655 }
3656 }
3657 c = *code;
3658 }
3659
3660 switch(c)
3661 {
3662 case OP_END:
3663 return;
3664
3665 case OP_TYPESTAR:
3666 case OP_TYPEMINSTAR:
3667 case OP_TYPEPLUS:
3668 case OP_TYPEMINPLUS:
3669 case OP_TYPEQUERY:
3670 case OP_TYPEMINQUERY:
3671 case OP_TYPEPOSSTAR:
3672 case OP_TYPEPOSPLUS:
3673 case OP_TYPEPOSQUERY:
3674 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3675 break;
3676
3677 case OP_TYPEUPTO:
3678 case OP_TYPEMINUPTO:
3679 case OP_TYPEEXACT:
3680 case OP_TYPEPOSUPTO:
3681 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3682 code += 2;
3683 break;
3684
3685 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3686 case OP_XCLASS:
3687 code += GET(code, 1);
3688 break;
3689 #endif
3690
3691 case OP_MARK:
3692 case OP_PRUNE_ARG:
3693 case OP_SKIP_ARG:
3694 case OP_THEN_ARG:
3695 code += code[1];
3696 break;
3697 }
3698
3699 /* Add in the fixed length from the table */
3700
3701 code += PRIV(OP_lengths)[c];
3702
3703 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3704 a multi-byte character. The length in the table is a minimum, so we have to
3705 arrange to skip the extra bytes. */
3706
3707 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3708 if (utf) switch(c)
3709 {
3710 case OP_CHAR:
3711 case OP_CHARI:
3712 case OP_NOT:
3713 case OP_NOTI:
3714 case OP_STAR:
3715 case OP_MINSTAR:
3716 case OP_PLUS:
3717 case OP_MINPLUS:
3718 case OP_QUERY:
3719 case OP_MINQUERY:
3720 case OP_UPTO:
3721 case OP_MINUPTO:
3722 case OP_EXACT:
3723 case OP_POSSTAR:
3724 case OP_POSPLUS:
3725 case OP_POSQUERY:
3726 case OP_POSUPTO:
3727 case OP_STARI:
3728 case OP_MINSTARI:
3729 case OP_PLUSI:
3730 case OP_MINPLUSI:
3731 case OP_QUERYI:
3732 case OP_MINQUERYI:
3733 case OP_UPTOI:
3734 case OP_MINUPTOI:
3735 case OP_EXACTI:
3736 case OP_POSSTARI:
3737 case OP_POSPLUSI:
3738 case OP_POSQUERYI:
3739 case OP_POSUPTOI:
3740 case OP_NOTSTAR:
3741 case OP_NOTMINSTAR:
3742 case OP_NOTPLUS:
3743 case OP_NOTMINPLUS:
3744 case OP_NOTQUERY:
3745 case OP_NOTMINQUERY:
3746 case OP_NOTUPTO:
3747 case OP_NOTMINUPTO:
3748 case OP_NOTEXACT:
3749 case OP_NOTPOSSTAR:
3750 case OP_NOTPOSPLUS:
3751 case OP_NOTPOSQUERY:
3752 case OP_NOTPOSUPTO:
3753 case OP_NOTSTARI:
3754 case OP_NOTMINSTARI:
3755 case OP_NOTPLUSI:
3756 case OP_NOTMINPLUSI:
3757 case OP_NOTQUERYI:
3758 case OP_NOTMINQUERYI:
3759 case OP_NOTUPTOI:
3760 case OP_NOTMINUPTOI:
3761 case OP_NOTEXACTI:
3762 case OP_NOTPOSSTARI:
3763 case OP_NOTPOSPLUSI:
3764 case OP_NOTPOSQUERYI:
3765 case OP_NOTPOSUPTOI:
3766 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3767 break;
3768 }
3769 #else
3770 (void)(utf); /* Keep compiler happy by referencing function argument */
3771 #endif
3772 }
3773 }
3774
3775
3776
3777 /*************************************************
3778 * Check for POSIX class syntax *
3779 *************************************************/
3780
3781 /* This function is called when the sequence "[:" or "[." or "[=" is
3782 encountered in a character class. It checks whether this is followed by a
3783 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3784 reach an unescaped ']' without the special preceding character, return FALSE.
3785
3786 Originally, this function only recognized a sequence of letters between the
3787 terminators, but it seems that Perl recognizes any sequence of characters,
3788 though of course unknown POSIX names are subsequently rejected. Perl gives an
3789 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3790 didn't consider this to be a POSIX class. Likewise for [:1234:].
3791
3792 The problem in trying to be exactly like Perl is in the handling of escapes. We
3793 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3794 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3795 below handles the special case of \], but does not try to do any other escape
3796 processing. This makes it different from Perl for cases such as [:l\ower:]
3797 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3798 "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3799 I think.
3800
3801 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3802 It seems that the appearance of a nested POSIX class supersedes an apparent
3803 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3804 a digit.
3805
3806 In Perl, unescaped square brackets may also appear as part of class names. For
3807 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3808 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3809 seem right at all. PCRE does not allow closing square brackets in POSIX class
3810 names.
3811
3812 Arguments:
3813 ptr pointer to the initial [
3814 endptr where to return the end pointer
3815
3816 Returns: TRUE or FALSE
3817 */
3818
3819 static BOOL
3820 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3821 {
3822 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3823 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3824 for (++ptr; *ptr != CHAR_NULL; ptr++)
3825 {
3826 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3827 ptr++;
3828 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3829 else
3830 {
3831 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3832 {
3833 *endptr = ptr;
3834 return TRUE;
3835 }
3836 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3837 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3838 ptr[1] == CHAR_EQUALS_SIGN) &&
3839 check_posix_syntax(ptr, endptr))
3840 return FALSE;
3841 }
3842 }
3843 return FALSE;
3844 }
3845
3846
3847
3848
3849 /*************************************************
3850 * Check POSIX class name *
3851 *************************************************/
3852
3853 /* This function is called to check the name given in a POSIX-style class entry
3854 such as [:alnum:].
3855
3856 Arguments:
3857 ptr points to the first letter
3858 len the length of the name
3859
3860 Returns: a value representing the name, or -1 if unknown
3861 */
3862
3863 static int
3864 check_posix_name(const pcre_uchar *ptr, int len)
3865 {
3866 const char *pn = posix_names;
3867 register int yield = 0;
3868 while (posix_name_lengths[yield] != 0)
3869 {
3870 if (len == posix_name_lengths[yield] &&
3871 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3872 pn += posix_name_lengths[yield] + 1;
3873 yield++;
3874 }
3875 return -1;
3876 }
3877
3878
3879 /*************************************************
3880 * Adjust OP_RECURSE items in repeated group *
3881 *************************************************/
3882
3883 /* OP_RECURSE items contain an offset from the start of the regex to the group
3884 that is referenced. This means that groups can be replicated for fixed
3885 repetition simply by copying (because the recursion is allowed to refer to
3886 earlier groups that are outside the current group). However, when a group is
3887 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3888 inserted before it, after it has been compiled. This means that any OP_RECURSE
3889 items within it that refer to the group itself or any contained groups have to
3890 have their offsets adjusted. That one of the jobs of this function. Before it
3891 is called, the partially compiled regex must be temporarily terminated with
3892 OP_END.
3893
3894 This function has been extended with the possibility of forward references for
3895 recursions and subroutine calls. It must also check the list of such references
3896 for the group we are dealing with. If it finds that one of the recursions in
3897 the current group is on this list, it adjusts the offset in the list, not the
3898 value in the reference (which is a group number).
3899
3900 Arguments:
3901 group points to the start of the group
3902 adjust the amount by which the group is to be moved
3903 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3904 cd contains pointers to tables etc.
3905 save_hwm the hwm forward reference pointer at the start of the group
3906
3907 Returns: nothing
3908 */
3909
3910 static void
3911 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3912 pcre_uchar *save_hwm)
3913 {
3914 pcre_uchar *ptr = group;
3915
3916 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3917 {
3918 int offset;
3919 pcre_uchar *hc;
3920
3921 /* See if this recursion is on the forward reference list. If so, adjust the
3922 reference. */
3923
3924 for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3925 {
3926 offset = (int)GET(hc, 0);
3927 if (cd->start_code + offset == ptr + 1)
3928 {
3929 PUT(hc, 0, offset + adjust);
3930 break;
3931 }
3932 }
3933
3934 /* Otherwise, adjust the recursion offset if it's after the start of this
3935 group. */
3936
3937 if (hc >= cd->hwm)
3938 {
3939 offset = (int)GET(ptr, 1);
3940 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3941 }
3942
3943 ptr += 1 + LINK_SIZE;
3944 }
3945 }
3946
3947
3948
3949 /*************************************************
3950 * Insert an automatic callout point *
3951 *************************************************/
3952
3953 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3954 callout points before each pattern item.
3955
3956 Arguments:
3957 code current code pointer
3958 ptr current pattern pointer
3959 cd pointers to tables etc
3960
3961 Returns: new code pointer
3962 */
3963
3964 static pcre_uchar *
3965 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3966 {
3967 *code++ = OP_CALLOUT;
3968 *code++ = 255;
3969 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
3970 PUT(code, LINK_SIZE, 0); /* Default length */
3971 return code + 2 * LINK_SIZE;
3972 }
3973
3974
3975
3976 /*************************************************
3977 * Complete a callout item *
3978 *************************************************/
3979
3980 /* A callout item contains the length of the next item in the pattern, which
3981 we can't fill in till after we have reached the relevant point. This is used
3982 for both automatic and manual callouts.
3983
3984 Arguments:
3985 previous_callout points to previous callout item
3986 ptr current pattern pointer
3987 cd pointers to tables etc
3988
3989 Returns: nothing
3990 */
3991
3992 static void
3993 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3994 {
3995 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3996 PUT(previous_callout, 2 + LINK_SIZE, length);
3997 }
3998
3999
4000
4001 #ifdef SUPPORT_UCP
4002 /*************************************************
4003 * Get othercase range *
4004 *************************************************/
4005
4006 /* This function is passed the start and end of a class range, in UTF-8 mode
4007 with UCP support. It searches up the characters, looking for ranges of
4008 characters in the "other" case. Each call returns the next one, updating the
4009 start address. A character with multiple other cases is returned on its own
4010 with a special return value.
4011
4012 Arguments:
4013 cptr points to starting character value; updated
4014 d end value
4015 ocptr where to put start of othercase range
4016 odptr where to put end of othercase range
4017
4018 Yield: -1 when no more
4019 0 when a range is returned
4020 >0 the CASESET offset for char with multiple other cases
4021 in this case, ocptr contains the original
4022 */
4023
4024 static int
4025 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4026 pcre_uint32 *odptr)
4027 {
4028 pcre_uint32 c, othercase, next;
4029 unsigned int co;
4030
4031 /* Find the first character that has an other case. If it has multiple other
4032 cases, return its case offset value. */
4033
4034 for (c = *cptr; c <= d; c++)
4035 {
4036 if ((co = UCD_CASESET(c)) != 0)
4037 {
4038 *ocptr = c++; /* Character that has the set */
4039 *cptr = c; /* Rest of input range */
4040 return (int)co;
4041 }
4042 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4043 }
4044
4045 if (c > d) return -1; /* Reached end of range */
4046
4047 *ocptr = othercase;
4048 next = othercase + 1;
4049
4050 for (++c; c <= d; c++)
4051 {
4052 if (UCD_OTHERCASE(c) != next) break;
4053 next++;
4054 }
4055
4056 *odptr = next - 1; /* End of othercase range */
4057 *cptr = c; /* Rest of input range */
4058 return 0;
4059 }
4060 #endif /* SUPPORT_UCP */
4061
4062
4063
4064 /*************************************************
4065 * Add a character or range to a class *
4066 *************************************************/
4067
4068 /* This function packages up the logic of adding a character or range of
4069 characters to a class. The character values in the arguments will be within the
4070 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4071 mutually recursive with the function immediately below.
4072
4073 Arguments:
4074 classbits the bit map for characters < 256
4075 uchardptr points to the pointer for extra data
4076 options the options word
4077 cd contains pointers to tables etc.
4078 start start of range character
4079 end end of range character
4080
4081 Returns: the number of < 256 characters added
4082 the pointer to extra data is updated
4083 */
4084
4085 static int
4086 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4087 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4088 {
4089 pcre_uint32 c;
4090 int n8 = 0;
4091
4092 /* If caseless matching is required, scan the range and process alternate
4093 cases. In Unicode, there are 8-bit characters that have alternate cases that
4094 are greater than 255 and vice-versa. Sometimes we can just extend the original
4095 range. */
4096
4097 if ((options & PCRE_CASELESS) != 0)
4098 {
4099 #ifdef SUPPORT_UCP
4100 if ((options & PCRE_UTF8) != 0)
4101 {
4102 int rc;
4103 pcre_uint32 oc, od;
4104
4105 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4106 c = start;
4107
4108 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4109 {
4110 /* Handle a single character that has more than one other case. */
4111
4112 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4113 PRIV(ucd_caseless_sets) + rc, oc);
4114
4115 /* Do nothing if the other case range is within the original range. */
4116
4117 else if (oc >= start && od <= end) continue;
4118
4119 /* Extend the original range if there is overlap, noting that if oc < c, we
4120 can't have od > end because a subrange is always shorter than the basic
4121 range. Otherwise, use a recursive call to add the additional range. */
4122
4123 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4124 else if (od > end && oc <= end + 1) end = od; /* Extend upwards */
4125 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4126 }
4127 }
4128 else
4129 #endif /* SUPPORT_UCP */
4130
4131 /* Not UTF-mode, or no UCP */
4132
4133 for (c = start; c <= end && c < 256; c++)
4134 {
4135 SETBIT(classbits, cd->fcc[c]);
4136 n8++;
4137 }
4138 }
4139
4140 /* Now handle the original range. Adjust the final value according to the bit
4141 length - this means that the same lists of (e.g.) horizontal spaces can be used
4142 in all cases. */
4143
4144 #if defined COMPILE_PCRE8
4145 #ifdef SUPPORT_UTF
4146 if ((options & PCRE_UTF8) == 0)
4147 #endif
4148 if (end > 0xff) end = 0xff;
4149
4150 #elif defined COMPILE_PCRE16
4151 #ifdef SUPPORT_UTF
4152 if ((options & PCRE_UTF16) == 0)
4153 #endif
4154 if (end > 0xffff) end = 0xffff;
4155
4156 #endif /* COMPILE_PCRE[8|16] */
4157
4158 /* If all characters are less than 256, use the bit map. Otherwise use extra
4159 data. */
4160
4161 if (end < 0x100)
4162 {
4163 for (c = start; c <= end; c++)
4164 {
4165 n8++;
4166 SETBIT(classbits, c);
4167 }
4168 }
4169
4170 else
4171 {
4172 pcre_uchar *uchardata = *uchardptr;
4173
4174 #ifdef SUPPORT_UTF
4175 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4176 {
4177 if (start < end)
4178 {
4179 *uchardata++ = XCL_RANGE;
4180 uchardata += PRIV(ord2utf)(start, uchardata);
4181 uchardata += PRIV(ord2utf)(end, uchardata);
4182 }
4183 else if (start == end)
4184 {
4185 *uchardata++ = XCL_SINGLE;
4186 uchardata += PRIV(ord2utf)(start, uchardata);
4187 }
4188 }
4189 else
4190 #endif /* SUPPORT_UTF */
4191
4192 /* Without UTF support, character values are constrained by the bit length,
4193 and can only be > 256 for 16-bit and 32-bit libraries. */
4194
4195 #ifdef COMPILE_PCRE8
4196 {}
4197 #else
4198 if (start < end)
4199 {
4200 *uchardata++ = XCL_RANGE;
4201 *uchardata++ = start;
4202 *uchardata++ = end;
4203 }
4204 else if (start == end)
4205 {
4206 *uchardata++ = XCL_SINGLE;
4207 *uchardata++ = start;
4208 }
4209 #endif
4210
4211 *uchardptr = uchardata; /* Updata extra data pointer */
4212 }
4213
4214 return n8; /* Number of 8-bit characters */
4215 }
4216
4217
4218
4219
4220 /*************************************************
4221 * Add a list of characters to a class *
4222 *************************************************/
4223
4224 /* This function is used for adding a list of case-equivalent characters to a
4225 class, and also for adding a list of horizontal or vertical whitespace. If the
4226 list is in order (which it should be), ranges of characters are detected and
4227 handled appropriately. This function is mutually recursive with the function
4228 above.
4229
4230 Arguments:
4231 classbits the bit map for characters < 256
4232 uchardptr points to the pointer for extra data
4233 options the options word
4234 cd contains pointers to tables etc.
4235 p points to row of 32-bit values, terminated by NOTACHAR
4236 except character to omit; this is used when adding lists of
4237 case-equivalent characters to avoid including the one we
4238 already know about
4239
4240 Returns: the number of < 256 characters added
4241 the pointer to extra data is updated
4242 */
4243
4244 static int
4245 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4246 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4247 {
4248 int n8 = 0;
4249 while (p[0] < NOTACHAR)
4250 {
4251 int n = 0;
4252 if (p[0] != except)
4253 {
4254 while(p[n+1] == p[0] + n + 1) n++;
4255 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4256 }
4257 p += n + 1;
4258 }
4259 return n8;
4260 }
4261
4262
4263
4264 /*************************************************
4265 * Add characters not in a list to a class *
4266 *************************************************/
4267
4268 /* This function is used for adding the complement of a list of horizontal or
4269 vertical whitespace to a class. The list must be in order.
4270
4271 Arguments:
4272 classbits the bit map for characters < 256
4273 uchardptr points to the pointer for extra data
4274 options the options word
4275 cd contains pointers to tables etc.
4276 p points to row of 32-bit values, terminated by NOTACHAR
4277
4278 Returns: the number of < 256 characters added
4279 the pointer to extra data is updated
4280 */
4281
4282 static int
4283 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4284 int options, compile_data *cd, const pcre_uint32 *p)
4285 {
4286 BOOL utf = (options & PCRE_UTF8) != 0;
4287 int n8 = 0;
4288 if (p[0] > 0)
4289 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4290 while (p[0] < NOTACHAR)
4291 {
4292 while (p[1] == p[0] + 1) p++;
4293 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4294 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4295 p++;
4296 }
4297 return n8;
4298 }
4299
4300
4301
4302 /*************************************************
4303 * Compile one branch *
4304 *************************************************/
4305
4306 /* Scan the pattern, compiling it into the a vector. If the options are
4307 changed during the branch, the pointer is used to change the external options
4308 bits. This function is used during the pre-compile phase when we are trying
4309 to find out the amount of memory needed, as well as during the real compile
4310 phase. The value of lengthptr distinguishes the two phases.
4311
4312 Arguments:
4313 optionsptr pointer to the option bits
4314 codeptr points to the pointer to the current code point
4315 ptrptr points to the current pattern pointer
4316 errorcodeptr points to error code variable
4317 firstcharptr place to put the first required character
4318 firstcharflagsptr place to put the first character flags, or a negative number
4319 reqcharptr place to put the last required character
4320 reqcharflagsptr place to put the last required character flags, or a negative number
4321 bcptr points to current branch chain
4322 cond_depth conditional nesting depth
4323 cd contains pointers to tables etc.
4324 lengthptr NULL during the real compile phase
4325 points to length accumulator during pre-compile phase
4326
4327 Returns: TRUE on success
4328 FALSE, with *errorcodeptr set non-zero on error
4329 */
4330
4331 static BOOL
4332 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4333 const pcre_uchar **ptrptr, int *errorcodeptr,
4334 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4335 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4336 branch_chain *bcptr, int cond_depth,
4337 compile_data *cd, int *lengthptr)
4338 {
4339 int repeat_type, op_type;
4340 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4341 int bravalue = 0;
4342 int greedy_default, greedy_non_default;
4343 pcre_uint32 firstchar, reqchar;
4344 pcre_int32 firstcharflags, reqcharflags;
4345 pcre_uint32 zeroreqchar, zerofirstchar;
4346 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4347 pcre_int32 req_caseopt, reqvary, tempreqvary;
4348 int options = *optionsptr; /* May change dynamically */
4349 int after_manual_callout = 0;
4350 int length_prevgroup = 0;
4351 register pcre_uint32 c;
4352 int escape;
4353 register pcre_uchar *code = *codeptr;
4354 pcre_uchar *last_code = code;
4355 pcre_uchar *orig_code = code;
4356 pcre_uchar *tempcode;
4357 BOOL inescq = FALSE;
4358 BOOL groupsetfirstchar = FALSE;
4359 const pcre_uchar *ptr = *ptrptr;
4360 const pcre_uchar *tempptr;
4361 const pcre_uchar *nestptr = NULL;
4362 pcre_uchar *previous = NULL;
4363 pcre_uchar *previous_callout = NULL;
4364 pcre_uchar *save_hwm = NULL;
4365 pcre_uint8 classbits[32];
4366
4367 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4368 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4369 dynamically as we process the pattern. */
4370
4371 #ifdef SUPPORT_UTF
4372 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4373 BOOL utf = (options & PCRE_UTF8) != 0;
4374 #ifndef COMPILE_PCRE32
4375 pcre_uchar utf_chars[6];
4376 #endif
4377 #else
4378 BOOL utf = FALSE;
4379 #endif
4380
4381 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4382 class_uchardata always so that it can be passed to add_to_class() always,
4383 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4384 alternative calls for the different cases. */
4385
4386 pcre_uchar *class_uchardata;
4387 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4388 BOOL xclass;
4389 pcre_uchar *class_uchardata_base;
4390 #endif
4391
4392 #ifdef PCRE_DEBUG
4393 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4394 #endif
4395
4396 /* Set up the default and non-default settings for greediness */
4397
4398 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4399 greedy_non_default = greedy_default ^ 1;
4400
4401 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4402 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4403 matches a non-fixed char first char; reqchar just remains unset if we never
4404 find one.
4405
4406 When we hit a repeat whose minimum is zero, we may have to adjust these values
4407 to take the zero repeat into account. This is implemented by setting them to
4408 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4409 item types that can be repeated set these backoff variables appropriately. */
4410
4411 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4412 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4413
4414 /* The variable req_caseopt contains either the REQ_CASELESS value
4415 or zero, according to the current setting of the caseless flag. The
4416 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4417 firstchar or reqchar variables to record the case status of the
4418 value. This is used only for ASCII characters. */
4419
4420 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4421
4422 /* Switch on next character until the end of the branch */
4423
4424 for (;; ptr++)
4425 {
4426 BOOL negate_class;
4427 BOOL should_flip_negation;
4428 BOOL possessive_quantifier;
4429 BOOL is_quantifier;
4430 BOOL is_recurse;
4431 BOOL reset_bracount;
4432 int class_has_8bitchar;
4433 int class_one_char;
4434 int newoptions;
4435 int recno;
4436 int refsign;
4437 int skipbytes;
4438 pcre_uint32 subreqchar, subfirstchar;
4439 pcre_int32 subreqcharflags, subfirstcharflags;
4440 int terminator;
4441 unsigned int mclength;
4442 unsigned int tempbracount;
4443 pcre_uint32 ec;
4444 pcre_uchar mcbuffer[8];
4445
4446 /* Get next character in the pattern */
4447
4448 c = *ptr;
4449
4450 /* If we are at the end of a nested substitution, revert to the outer level
4451 string. Nesting only happens one level deep. */
4452
4453 if (c == CHAR_NULL && nestptr != NULL)
4454 {
4455 ptr = nestptr;
4456 nestptr = NULL;
4457 c = *ptr;
4458 }
4459
4460 /* If we are in the pre-compile phase, accumulate the length used for the
4461 previous cycle of this loop. */
4462
4463 if (lengthptr != NULL)
4464 {
4465 #ifdef PCRE_DEBUG
4466 if (code > cd->hwm) cd->hwm = code; /* High water info */
4467 #endif
4468 if (code > cd->start_workspace + cd->workspace_size -
4469 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4470 {
4471 *errorcodeptr = ERR52;
4472 goto FAILED;
4473 }
4474
4475 /* There is at least one situation where code goes backwards: this is the
4476 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4477 the class is simply eliminated. However, it is created first, so we have to
4478 allow memory for it. Therefore, don't ever reduce the length at this point.
4479 */
4480
4481 if (code < last_code) code = last_code;
4482
4483 /* Paranoid check for integer overflow */
4484
4485 if (OFLOW_MAX - *lengthptr < code - last_code)
4486 {
4487 *errorcodeptr = ERR20;
4488 goto FAILED;
4489 }
4490
4491 *lengthptr += (int)(code - last_code);
4492 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4493 (int)(code - last_code), c, c));
4494
4495 /* If "previous" is set and it is not at the start of the work space, move
4496 it back to there, in order to avoid filling up the work space. Otherwise,
4497 if "previous" is NULL, reset the current code pointer to the start. */
4498
4499 if (previous != NULL)
4500 {
4501 if (previous > orig_code)
4502 {
4503 memmove(orig_code, previous, IN_UCHARS(code - previous));
4504 code -= previous - orig_code;
4505 previous = orig_code;
4506 }
4507 }
4508 else code = orig_code;
4509
4510 /* Remember where this code item starts so we can pick up the length
4511 next time round. */
4512
4513 last_code = code;
4514 }
4515
4516 /* In the real compile phase, just check the workspace used by the forward
4517 reference list. */
4518
4519 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4520 WORK_SIZE_SAFETY_MARGIN)
4521 {
4522 *errorcodeptr = ERR52;
4523 goto FAILED;
4524 }
4525
4526 /* If in \Q...\E, check for the end; if not, we have a literal */
4527
4528 if (inescq && c != CHAR_NULL)
4529 {
4530 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4531 {
4532 inescq = FALSE;
4533 ptr++;
4534 continue;
4535 }
4536 else
4537 {
4538 if (previous_callout != NULL)
4539 {
4540 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4541 complete_callout(previous_callout, ptr, cd);
4542 previous_callout = NULL;
4543 }
4544 if ((options & PCRE_AUTO_CALLOUT) != 0)
4545 {
4546 previous_callout = code;
4547 code = auto_callout(code, ptr, cd);
4548 }
4549 goto NORMAL_CHAR;
4550 }
4551 /* Control does not reach here. */
4552 }
4553
4554 /* In extended mode, skip white space and comments. We need a loop in order
4555 to check for more white space and more comments after a comment. */
4556
4557 if ((options & PCRE_EXTENDED) != 0)
4558 {
4559 for (;;)
4560 {
4561 while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4562 if (c != CHAR_NUMBER_SIGN) break;
4563 ptr++;
4564 while (*ptr != CHAR_NULL)
4565 {
4566 if (IS_NEWLINE(ptr)) /* For non-fixed-length newline cases, */
4567 { /* IS_NEWLINE sets cd->nllen. */
4568 ptr += cd->nllen;
4569 break;
4570 }
4571 ptr++;
4572 #ifdef SUPPORT_UTF
4573 if (utf) FORWARDCHAR(ptr);
4574 #endif
4575 }
4576 c = *ptr; /* Either NULL or the char after a newline */
4577 }
4578 }
4579
4580 /* See if the next thing is a quantifier. */
4581
4582 is_quantifier =
4583 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4584 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4585
4586 /* Fill in length of a previous callout, except when the next thing is a
4587 quantifier or when processing a property substitution string in UCP mode. */
4588
4589 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4590 after_manual_callout-- <= 0)
4591 {
4592 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4593 complete_callout(previous_callout, ptr, cd);
4594 previous_callout = NULL;
4595 }
4596
4597 /* Create auto callout, except for quantifiers, or while processing property
4598 strings that are substituted for \w etc in UCP mode. */
4599
4600 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4601 {
4602 previous_callout = code;
4603 code = auto_callout(code, ptr, cd);
4604 }
4605
4606 /* Process the next pattern item. */
4607
4608 switch(c)
4609 {
4610 /* ===================================================================*/
4611 case CHAR_NULL: /* The branch terminates at string end */
4612 case CHAR_VERTICAL_LINE: /* or | or ) */
4613 case CHAR_RIGHT_PARENTHESIS:
4614 *firstcharptr = firstchar;
4615 *firstcharflagsptr = firstcharflags;
4616 *reqcharptr = reqchar;
4617 *reqcharflagsptr = reqcharflags;
4618 *codeptr = code;
4619 *ptrptr = ptr;
4620 if (lengthptr != NULL)
4621 {
4622 if (OFLOW_MAX - *lengthptr < code - last_code)
4623 {
4624 *errorcodeptr = ERR20;
4625 goto FAILED;
4626 }
4627 *lengthptr += (int)(code - last_code); /* To include callout length */
4628 DPRINTF((">> end branch\n"));
4629 }
4630 return TRUE;
4631
4632
4633 /* ===================================================================*/
4634 /* Handle single-character metacharacters. In multiline mode, ^ disables
4635 the setting of any following char as a first character. */
4636
4637 case CHAR_CIRCUMFLEX_ACCENT:
4638 previous = NULL;
4639 if ((options & PCRE_MULTILINE) != 0)
4640 {
4641 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4642 *code++ = OP_CIRCM;
4643 }
4644 else *code++ = OP_CIRC;
4645 break;
4646
4647 case CHAR_DOLLAR_SIGN:
4648 previous = NULL;
4649 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4650 break;
4651
4652 /* There can never be a first char if '.' is first, whatever happens about
4653 repeats. The value of reqchar doesn't change either. */
4654
4655 case CHAR_DOT:
4656 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4657 zerofirstchar = firstchar;
4658 zerofirstcharflags = firstcharflags;
4659 zeroreqchar = reqchar;
4660 zeroreqcharflags = reqcharflags;
4661 previous = code;
4662 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4663 break;
4664
4665
4666 /* ===================================================================*/
4667 /* Character classes. If the included characters are all < 256, we build a
4668 32-byte bitmap of the permitted characters, except in the special case
4669 where there is only one such character. For negated classes, we build the
4670 map as usual, then invert it at the end. However, we use a different opcode
4671 so that data characters > 255 can be handled correctly.
4672
4673 If the class contains characters outside the 0-255 range, a different
4674 opcode is compiled. It may optionally have a bit map for characters < 256,
4675 but those above are are explicitly listed afterwards. A flag byte tells
4676 whether the bitmap is present, and whether this is a negated class or not.
4677
4678 In JavaScript compatibility mode, an isolated ']' causes an error. In
4679 default (Perl) mode, it is treated as a data character. */
4680
4681 case CHAR_RIGHT_SQUARE_BRACKET:
4682 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4683 {
4684 *errorcodeptr = ERR64;
4685 goto FAILED;
4686 }
4687 goto NORMAL_CHAR;
4688
4689 case CHAR_LEFT_SQUARE_BRACKET:
4690 previous = code;
4691
4692 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4693 they are encountered at the top level, so we'll do that too. */
4694
4695 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4696 ptr[1] == CHAR_EQUALS_SIGN) &&
4697 check_posix_syntax(ptr, &tempptr))
4698 {
4699 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4700 goto FAILED;
4701 }
4702
4703 /* If the first character is '^', set the negation flag and skip it. Also,
4704 if the first few characters (either before or after ^) are \Q\E or \E we
4705 skip them too. This makes for compatibility with Perl. */
4706
4707 negate_class = FALSE;
4708 for (;;)
4709 {
4710 c = *(++ptr);
4711 if (c == CHAR_BACKSLASH)
4712 {
4713 if (ptr[1] == CHAR_E)
4714 ptr++;
4715 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4716 ptr += 3;
4717 else
4718 break;
4719 }
4720 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4721 negate_class = TRUE;
4722 else break;
4723 }
4724
4725 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4726 an initial ']' is taken as a data character -- the code below handles
4727 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4728 [^] must match any character, so generate OP_ALLANY. */
4729
4730 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4731 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4732 {
4733 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4734 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4735 zerofirstchar = firstchar;
4736 zerofirstcharflags = firstcharflags;
4737 break;
4738 }
4739
4740 /* If a class contains a negative special such as \S, we need to flip the
4741 negation flag at the end, so that support for characters > 255 works
4742 correctly (they are all included in the class). */
4743
4744 should_flip_negation = FALSE;
4745
4746 /* For optimization purposes, we track some properties of the class:
4747 class_has_8bitchar will be non-zero if the class contains at least one <
4748 256 character; class_one_char will be 1 if the class contains just one
4749 character. */
4750
4751 class_has_8bitchar = 0;
4752 class_one_char = 0;
4753
4754 /* Initialize the 32-char bit map to all zeros. We build the map in a
4755 temporary bit of memory, in case the class contains fewer than two
4756 8-bit characters because in that case the compiled code doesn't use the bit
4757 map. */
4758
4759 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4760
4761 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4762 xclass = FALSE;
4763 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4764 class_uchardata_base = class_uchardata; /* Save the start */
4765 #endif
4766
4767 /* Process characters until ] is reached. By writing this as a "do" it
4768 means that an initial ] is taken as a data character. At the start of the
4769 loop, c contains the first byte of the character. */
4770
4771 if (c != CHAR_NULL) do
4772 {
4773 const pcre_uchar *oldptr;
4774
4775 #ifdef SUPPORT_UTF
4776 if (utf && HAS_EXTRALEN(c))
4777 { /* Braces are required because the */
4778 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4779 }
4780 #endif
4781
4782 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4783 /* In the pre-compile phase, accumulate the length of any extra
4784 data and reset the pointer. This is so that very large classes that
4785 contain a zillion > 255 characters no longer overwrite the work space
4786 (which is on the stack). We have to remember that there was XCLASS data,
4787 however. */
4788
4789 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4790 {
4791 xclass = TRUE;
4792 *lengthptr += class_uchardata - class_uchardata_base;
4793 class_uchardata = class_uchardata_base;
4794 }
4795 #endif
4796
4797 /* Inside \Q...\E everything is literal except \E */
4798
4799 if (inescq)
4800 {
4801 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4802 {
4803 inescq = FALSE; /* Reset literal state */
4804 ptr++; /* Skip the 'E' */
4805 continue; /* Carry on with next */
4806 }
4807 goto CHECK_RANGE; /* Could be range if \E follows */
4808 }
4809
4810 /* Handle POSIX class names. Perl allows a negation extension of the
4811 form [:^name:]. A square bracket that doesn't match the syntax is
4812 treated as a literal. We also recognize the POSIX constructions
4813 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4814 5.6 and 5.8 do. */
4815
4816 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4817 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4818 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4819 {
4820 BOOL local_negate = FALSE;
4821 int posix_class, taboffset, tabopt;
4822 register const pcre_uint8 *cbits = cd->cbits;
4823 pcre_uint8 pbits[32];
4824
4825 if (ptr[1] != CHAR_COLON)
4826 {
4827 *errorcodeptr = ERR31;
4828 goto FAILED;
4829 }
4830
4831 ptr += 2;
4832 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4833 {
4834 local_negate = TRUE;
4835 should_flip_negation = TRUE; /* Note negative special */
4836 ptr++;
4837 }
4838
4839 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4840 if (posix_class < 0)
4841 {
4842 *errorcodeptr = ERR30;
4843 goto FAILED;
4844 }
4845
4846 /* If matching is caseless, upper and lower are converted to
4847 alpha. This relies on the fact that the class table starts with
4848 alpha, lower, upper as the first 3 entries. */
4849
4850 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4851 posix_class = 0;
4852
4853 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4854 different escape sequences that use Unicode properties \p or \P. Others
4855 that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4856 directly. */
4857
4858 #ifdef SUPPORT_UCP
4859 if ((options & PCRE_UCP) != 0)
4860 {
4861 unsigned int ptype = 0;
4862 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4863
4864 /* The posix_substitutes table specifies which POSIX classes can be
4865 converted to \p or \P items. */
4866
4867 if (posix_substitutes[pc] != NULL)
4868 {
4869 nestptr = tempptr + 1;
4870 ptr = posix_substitutes[pc] - 1;
4871 continue;
4872 }
4873
4874 /* There are three other classes that generate special property calls
4875 that are recognized only in an XCLASS. */
4876
4877 else switch(posix_class)
4878 {
4879 case PC_GRAPH:
4880 ptype = PT_PXGRAPH;
4881 /* Fall through */
4882 case PC_PRINT:
4883 if (ptype == 0) ptype = PT_PXPRINT;
4884 /* Fall through */
4885 case PC_PUNCT:
4886 if (ptype == 0) ptype = PT_PXPUNCT;
4887 *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4888 *class_uchardata++ = ptype;
4889 *class_uchardata++ = 0;
4890 ptr = tempptr + 1;
4891 continue;
4892
4893 /* For all other POSIX classes, no special action is taken in UCP
4894 mode. Fall through to the non_UCP case. */
4895
4896 default:
4897 break;
4898 }
4899 }
4900 #endif
4901 /* In the non-UCP case, or when UCP makes no difference, we build the
4902 bit map for the POSIX class in a chunk of local store because we may be
4903 adding and subtracting from it, and we don't want to subtract bits that
4904 may be in the main map already. At the end we or the result into the
4905 bit map that is being built. */
4906
4907 posix_class *= 3;
4908
4909 /* Copy in the first table (always present) */
4910
4911 memcpy(pbits, cbits + posix_class_maps[posix_class],
4912 32 * sizeof(pcre_uint8));
4913
4914 /* If there is a second table, add or remove it as required. */
4915
4916 taboffset = posix_class_maps[posix_class + 1];
4917 tabopt = posix_class_maps[posix_class + 2];
4918
4919 if (taboffset >= 0)
4920 {
4921 if (tabopt >= 0)
4922 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
4923 else
4924 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
4925 }
4926
4927 /* Now see if we need to remove any special characters. An option
4928 value of 1 removes vertical space and 2 removes underscore. */
4929
4930 if (tabopt < 0) tabopt = -tabopt;
4931 if (tabopt == 1) pbits[1] &= ~0x3c;
4932 else if (tabopt == 2) pbits[11] &= 0x7f;
4933
4934 /* Add the POSIX table or its complement into the main table that is
4935 being built and we are done. */
4936
4937 if (local_negate)
4938 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
4939 else
4940 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
4941
4942 ptr = tempptr + 1;
4943 /* Every class contains at least one < 256 character. */
4944 class_has_8bitchar = 1;
4945 /* Every class contains at least two characters. */
4946 class_one_char = 2;
4947 continue; /* End of POSIX syntax handling */
4948 }
4949
4950 /* Backslash may introduce a single character, or it may introduce one
4951 of the specials, which just set a flag. The sequence \b is a special
4952 case. Inside a class (and only there) it is treated as backspace. We
4953 assume that other escapes have more than one character in them, so
4954 speculatively set both class_has_8bitchar and class_one_char bigger
4955 than one. Unrecognized escapes fall through and are either treated
4956 as literal characters (by default), or are faulted if
4957 PCRE_EXTRA is set. */
4958
4959 if (c == CHAR_BACKSLASH)
4960 {
4961 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4962 TRUE);
4963 if (*errorcodeptr != 0) goto FAILED;
4964 if (escape == 0) c = ec;
4965 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
4966 else if (escape == ESC_N) /* \N is not supported in a class */
4967 {
4968 *errorcodeptr = ERR71;
4969 goto FAILED;
4970 }
4971 else if (escape == ESC_Q) /* Handle start of quoted string */
4972 {
4973 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4974 {
4975 ptr += 2; /* avoid empty string */
4976 }
4977 else inescq = TRUE;
4978 continue;
4979 }
4980 else if (escape == ESC_E) continue; /* Ignore orphan \E */
4981
4982 else
4983 {
4984 register const pcre_uint8 *cbits = cd->cbits;
4985 /* Every class contains at least two < 256 characters. */
4986 class_has_8bitchar++;
4987 /* Every class contains at least two characters. */
4988 class_one_char += 2;
4989
4990 switch (escape)
4991 {
4992 #ifdef SUPPORT_UCP
4993 case ESC_du: /* These are the values given for \d etc */
4994 case ESC_DU: /* when PCRE_UCP is set. We replace the */
4995 case ESC_wu: /* escape sequence with an appropriate \p */
4996 case ESC_WU: /* or \P to test Unicode properties instead */
4997 case ESC_su: /* of the default ASCII testing. */
4998 case ESC_SU:
4999 nestptr = ptr;
5000 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
5001 class_has_8bitchar--; /* Undo! */
5002 continue;
5003 #endif
5004 case ESC_d:
5005 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
5006 continue;
5007
5008 case ESC_D:
5009 should_flip_negation = TRUE;
5010 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
5011 continue;
5012
5013 case ESC_w:
5014 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
5015 continue;
5016
5017 case ESC_W:
5018 should_flip_negation = TRUE;
5019 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5020 continue;
5021
5022 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5023 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5024 previously set by something earlier in the character class.
5025 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5026 we could just adjust the appropriate bit. From PCRE 8.34 we no
5027 longer treat \s and \S specially. */
5028
5029 case ESC_s:
5030 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
5031 continue;
5032
5033 case ESC_S:
5034 should_flip_negation = TRUE;
5035 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
5036 continue;
5037
5038 /* The rest apply in both UCP and non-UCP cases. */
5039
5040 case ESC_h:
5041 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5042 PRIV(hspace_list), NOTACHAR);
5043 continue;
5044
5045 case ESC_H:
5046 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5047 cd, PRIV(hspace_list));
5048 continue;
5049
5050 case ESC_v:
5051 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5052 PRIV(vspace_list), NOTACHAR);
5053 continue;
5054
5055 case ESC_V:
5056 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5057 cd, PRIV(vspace_list));
5058 continue;
5059
5060 #ifdef SUPPORT_UCP
5061 case ESC_p:
5062 case ESC_P:
5063 {
5064 BOOL negated;
5065 unsigned int ptype = 0, pdata = 0;
5066 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5067 goto FAILED;
5068 *class_uchardata++ = ((escape == ESC_p) != negated)?
5069 XCL_PROP : XCL_NOTPROP;
5070 *class_uchardata++ = ptype;
5071 *class_uchardata++ = pdata;
5072 class_has_8bitchar--; /* Undo! */
5073 continue;
5074 }
5075 #endif
5076 /* Unrecognized escapes are faulted if PCRE is running in its
5077 strict mode. By default, for compatibility with Perl, they are
5078 treated as literals. */
5079
5080 default:
5081 if ((options & PCRE_EXTRA) != 0)
5082 {
5083 *errorcodeptr = ERR7;
5084 goto FAILED;
5085 }
5086 class_has_8bitchar--; /* Undo the speculative increase. */
5087 class_one_char -= 2; /* Undo the speculative increase. */
5088 c = *ptr; /* Get the final character and fall through */
5089 break;
5090 }
5091 }
5092
5093 /* Fall through if the escape just defined a single character (c >= 0).
5094 This may be greater than 256. */
5095
5096 escape = 0;
5097
5098 } /* End of backslash handling */
5099
5100 /* A character may be followed by '-' to form a range. However, Perl does
5101 not permit ']' to be the end of the range. A '-' character at the end is
5102 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5103 code for handling \Q and \E is messy. */
5104
5105 CHECK_RANGE:
5106 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5107 {
5108 inescq = FALSE;
5109 ptr += 2;
5110 }
5111 oldptr = ptr;
5112
5113 /* Remember if \r or \n were explicitly used */
5114
5115 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5116
5117 /* Check for range */
5118
5119 if (!inescq && ptr[1] == CHAR_MINUS)
5120 {
5121 pcre_uint32 d;
5122 ptr += 2;
5123 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5124
5125 /* If we hit \Q (not followed by \E) at this point, go into escaped
5126 mode. */
5127
5128 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5129 {
5130 ptr += 2;
5131 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5132 { ptr += 2; continue; }
5133 inescq = TRUE;
5134 break;
5135 }
5136
5137 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5138 back the pointer and jump to handle the character that preceded it. */
5139
5140 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5141 {
5142 ptr = oldptr;
5143 goto CLASS_SINGLE_CHARACTER;
5144 }
5145
5146 /* Otherwise, we have a potential range; pick up the next character */
5147
5148 #ifdef SUPPORT_UTF
5149 if (utf)
5150 { /* Braces are required because the */
5151 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5152 }
5153 else
5154 #endif
5155 d = *ptr; /* Not UTF-8 mode */
5156
5157 /* The second part of a range can be a single-character escape
5158 sequence, but not any of the other escapes. Perl treats a hyphen as a
5159 literal in such circumstances. However, in Perl's warning mode, a
5160 warning is given, so PCRE now faults it as it is almost certainly a
5161 mistake on the user's part. */
5162
5163 if (!inescq)
5164 {
5165 if (d == CHAR_BACKSLASH)
5166 {
5167 int descape;
5168 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5169 if (*errorcodeptr != 0) goto FAILED;
5170
5171 /* 0 means a character was put into d; \b is backspace; any other
5172 special causes an error. */
5173
5174 if (descape != 0)
5175 {
5176 if (descape == ESC_b) d = CHAR_BS; else
5177 {
5178 *errorcodeptr = ERR83;
5179 goto FAILED;
5180 }
5181 }
5182 }
5183
5184 /* A hyphen followed by a POSIX class is treated in the same way. */
5185
5186 else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5187 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5188 ptr[1] == CHAR_EQUALS_SIGN) &&
5189 check_posix_syntax(ptr, &tempptr))
5190 {
5191 *errorcodeptr = ERR83;
5192 goto FAILED;
5193 }
5194 }
5195
5196 /* Check that the two values are in the correct order. Optimize
5197 one-character ranges. */
5198
5199 if (d < c)
5200 {
5201 *errorcodeptr = ERR8;
5202 goto FAILED;
5203 }
5204 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5205
5206 /* We have found a character range, so single character optimizations
5207 cannot be done anymore. Any value greater than 1 indicates that there
5208 is more than one character. */
5209
5210 class_one_char = 2;
5211
5212 /* Remember an explicit \r or \n, and add the range to the class. */
5213
5214 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5215
5216 class_has_8bitchar +=
5217 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5218
5219 continue; /* Go get the next char in the class */
5220 }
5221
5222 /* Handle a single character - we can get here for a normal non-escape
5223 char, or after \ that introduces a single character or for an apparent
5224 range that isn't. Only the value 1 matters for class_one_char, so don't
5225 increase it if it is already 2 or more ... just in case there's a class
5226 with a zillion characters in it. */
5227
5228 CLASS_SINGLE_CHARACTER:
5229 if (class_one_char < 2) class_one_char++;
5230
5231 /* If class_one_char is 1, we have the first single character in the
5232 class, and there have been no prior ranges, or XCLASS items generated by
5233 escapes. If this is the final character in the class, we can optimize by
5234 turning the item into a 1-character OP_CHAR[I] if it's positive, or
5235 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5236 to be set. Otherwise, there can be no first char if this item is first,
5237 whatever repeat count may follow. In the case of reqchar, save the
5238 previous value for reinstating. */
5239
5240 if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5241 {
5242 ptr++;
5243 zeroreqchar = reqchar;
5244 zeroreqcharflags = reqcharflags;
5245
5246 if (negate_class)
5247 {
5248 #ifdef SUPPORT_UCP
5249 int d;
5250 #endif
5251 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5252 zerofirstchar = firstchar;
5253 zerofirstcharflags = firstcharflags;
5254
5255 /* For caseless UTF-8 mode when UCP support is available, check
5256 whether this character has more than one other case. If so, generate
5257 a special OP_NOTPROP item instead of OP_NOTI. */
5258
5259 #ifdef SUPPORT_UCP
5260 if (utf && (options & PCRE_CASELESS) != 0 &&
5261 (d = UCD_CASESET(c)) != 0)
5262 {
5263 *code++ = OP_NOTPROP;
5264 *code++ = PT_CLIST;
5265 *code++ = d;
5266 }
5267 else
5268 #endif
5269 /* Char has only one other case, or UCP not available */
5270
5271 {
5272 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5273 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5274 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5275 code += PRIV(ord2utf)(c, code);
5276 else
5277 #endif
5278 *code++ = c;
5279 }
5280
5281 /* We are finished with this character class */
5282
5283 goto END_CLASS;
5284 }
5285
5286 /* For a single, positive character, get the value into mcbuffer, and
5287 then we can handle this with the normal one-character code. */
5288
5289 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5290 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5291 mclength = PRIV(ord2utf)(c, mcbuffer);
5292 else
5293 #endif
5294 {
5295 mcbuffer[0] = c;
5296 mclength = 1;
5297 }
5298 goto ONE_CHAR;
5299 } /* End of 1-char optimization */
5300
5301 /* There is more than one character in the class, or an XCLASS item
5302 has been generated. Add this character to the class. */
5303
5304 class_has_8bitchar +=
5305 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5306 }
5307
5308 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5309 If we are at the end of an internal nested string, revert to the outer
5310 string. */
5311
5312 while (((c = *(++ptr)) != CHAR_NULL ||
5313 (nestptr != NULL &&
5314 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5315 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5316
5317 /* Check for missing terminating ']' */
5318
5319 if (c == CHAR_NULL)
5320 {
5321 *errorcodeptr = ERR6;
5322 goto FAILED;
5323 }
5324
5325 /* We will need an XCLASS if data has been placed in class_uchardata. In
5326 the second phase this is a sufficient test. However, in the pre-compile
5327 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5328 only if the very last character in the class needs XCLASS will it contain
5329 anything at this point. For this reason, xclass gets set TRUE above when
5330 uchar_classdata is emptied, and that's why this code is the way it is here
5331 instead of just doing a test on class_uchardata below. */
5332
5333 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5334 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5335 #endif
5336
5337 /* If this is the first thing in the branch, there can be no first char
5338 setting, whatever the repeat count. Any reqchar setting must remain
5339 unchanged after any kind of repeat. */
5340
5341 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5342 zerofirstchar = firstchar;
5343 zerofirstcharflags = firstcharflags;
5344 zeroreqchar = reqchar;
5345 zeroreqcharflags = reqcharflags;
5346
5347 /* If there are characters with values > 255, we have to compile an
5348 extended class, with its own opcode, unless there was a negated special
5349 such as \S in the class, and PCRE_UCP is not set, because in that case all
5350 characters > 255 are in the class, so any that were explicitly given as
5351 well can be ignored. If (when there are explicit characters > 255 that must
5352 be listed) there are no characters < 256, we can omit the bitmap in the
5353 actual compiled code. */
5354
5355 #ifdef SUPPORT_UTF
5356 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5357 #elif !defined COMPILE_PCRE8
5358 if (xclass && !should_flip_negation)
5359 #endif
5360 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5361 {
5362 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5363 *code++ = OP_XCLASS;
5364 code += LINK_SIZE;
5365 *code = negate_class? XCL_NOT:0;
5366
5367 /* If the map is required, move up the extra data to make room for it;
5368 otherwise just move the code pointer to the end of the extra data. */
5369
5370 if (class_has_8bitchar > 0)
5371 {
5372 *code++ |= XCL_MAP;
5373 memmove(code + (32 / sizeof(pcre_uchar)), code,
5374 IN_UCHARS(class_uchardata - code));
5375 memcpy(code, classbits, 32);
5376 code = class_uchardata + (32 / sizeof(pcre_uchar));
5377 }
5378 else code = class_uchardata;
5379
5380 /* Now fill in the complete length of the item */
5381
5382 PUT(previous, 1, (int)(code - previous));
5383 break; /* End of class handling */
5384 }
5385 #endif
5386
5387 /* If there are no characters > 255, or they are all to be included or
5388 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5389 whole class was negated and whether there were negative specials such as \S
5390 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5391 negating it if necessary. */
5392
5393 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5394 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5395 {
5396 if (negate_class)
5397 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5398 memcpy(code, classbits, 32);
5399 }
5400 code += 32 / sizeof(pcre_uchar);
5401
5402 END_CLASS:
5403 break;
5404
5405
5406 /* ===================================================================*/
5407 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5408 has been tested above. */
5409
5410 case CHAR_LEFT_CURLY_BRACKET:
5411 if (!is_quantifier) goto NORMAL_CHAR;
5412 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5413 if (*errorcodeptr != 0) goto FAILED;
5414 goto REPEAT;
5415
5416 case CHAR_ASTERISK:
5417 repeat_min = 0;
5418 repeat_max = -1;
5419 goto REPEAT;
5420
5421 case CHAR_PLUS:
5422 repeat_min = 1;
5423 repeat_max = -1;
5424 goto REPEAT;
5425
5426 case CHAR_QUESTION_MARK:
5427 repeat_min = 0;
5428 repeat_max = 1;
5429
5430 REPEAT:
5431 if (previous == NULL)
5432 {
5433 *errorcodeptr = ERR9;
5434 goto FAILED;
5435 }
5436
5437 if (repeat_min == 0)
5438 {
5439 firstchar = zerofirstchar; /* Adjust for zero repeat */
5440 firstcharflags = zerofirstcharflags;
5441 reqchar = zeroreqchar; /* Ditto */
5442 reqcharflags = zeroreqcharflags;
5443 }
5444
5445 /* Remember whether this is a variable length repeat */
5446
5447 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5448
5449 op_type = 0; /* Default single-char op codes */
5450 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5451
5452 /* Save start of previous item, in case we have to move it up in order to
5453 insert something before it. */
5454
5455 tempcode = previous;
5456
5457 /* Before checking for a possessive quantifier, we must skip over
5458 whitespace and comments in extended mode because Perl allows white space at
5459 this point. */
5460
5461 if ((options & PCRE_EXTENDED) != 0)
5462 {
5463 const pcre_uchar *p = ptr + 1;
5464 for (;;)
5465 {
5466 while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5467 if (*p != CHAR_NUMBER_SIGN) break;
5468 p++;
5469 while (*p != CHAR_NULL)
5470 {
5471 if (IS_NEWLINE(p)) /* For non-fixed-length newline cases, */
5472 { /* IS_NEWLINE sets cd->nllen. */
5473 p += cd->nllen;
5474 break;
5475 }
5476 p++;
5477 #ifdef SUPPORT_UTF
5478 if (utf) FORWARDCHAR(p);
5479 #endif
5480 } /* Loop for comment characters */
5481 } /* Loop for multiple comments */
5482 ptr = p - 1; /* Character before the next significant one. */
5483 }
5484
5485 /* If the next character is '+', we have a possessive quantifier. This
5486 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5487 If the next character is '?' this is a minimizing repeat, by default,
5488 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5489 repeat type to the non-default. */
5490
5491 if (ptr[1] == CHAR_PLUS)
5492 {
5493 repeat_type = 0; /* Force greedy */
5494 possessive_quantifier = TRUE;
5495 ptr++;
5496 }
5497 else if (ptr[1] == CHAR_QUESTION_MARK)
5498 {
5499 repeat_type = greedy_non_default;
5500 ptr++;
5501 }
5502 else repeat_type = greedy_default;
5503
5504 /* If previous was a recursion call, wrap it in atomic brackets so that
5505 previous becomes the atomic group. All recursions were so wrapped in the
5506 past, but it no longer happens for non-repeated recursions. In fact, the
5507 repeated ones could be re-implemented independently so as not to need this,
5508 but for the moment we rely on the code for repeating groups. */
5509
5510 if (*previous == OP_RECURSE)
5511 {
5512 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5513 *previous = OP_ONCE;
5514 PUT(previous, 1, 2 + 2*LINK_SIZE);
5515 previous[2 + 2*LINK_SIZE] = OP_KET;
5516 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5517 code += 2 + 2 * LINK_SIZE;
5518 length_prevgroup = 3 + 3*LINK_SIZE;
5519
5520 /* When actually compiling, we need to check whether this was a forward
5521 reference, and if so, adjust the offset. */
5522
5523 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5524 {
5525 int offset = GET(cd->hwm, -LINK_SIZE);
5526 if (offset == previous + 1 - cd->start_code)
5527 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5528 }
5529 }
5530
5531 /* Now handle repetition for the different types of item. */
5532
5533 /* If previous was a character or negated character match, abolish the item
5534 and generate a repeat item instead. If a char item has a minimum of more
5535 than one, ensure that it is set in reqchar - it might not be if a sequence
5536 such as x{3} is the first thing in a branch because the x will have gone
5537 into firstchar instead. */
5538
5539 if (*previous == OP_CHAR || *previous == OP_CHARI
5540 || *previous == OP_NOT || *previous == OP_NOTI)
5541 {
5542 switch (*previous)
5543 {
5544 default: /* Make compiler happy. */
5545 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5546 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5547 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5548 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5549 }
5550
5551 /* Deal with UTF characters that take up more than one character. It's
5552 easier to write this out separately than try to macrify it. Use c to
5553 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5554 it's a length rather than a small character. */
5555
5556 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5557 if (utf && NOT_FIRSTCHAR(code[-1]))
5558 {
5559 pcre_uchar *lastchar = code - 1;
5560 BACKCHAR(lastchar);
5561 c = (int)(code - lastchar); /* Length of UTF-8 character */
5562 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5563 c |= UTF_LENGTH; /* Flag c as a length */
5564 }
5565 else
5566 #endif /* SUPPORT_UTF */
5567
5568 /* Handle the case of a single charater - either with no UTF support, or
5569 with UTF disabled, or for a single character UTF character. */
5570 {
5571 c = code[-1];
5572 if (*previous <= OP_CHARI && repeat_min > 1)
5573 {
5574 reqchar = c;
5575 reqcharflags = req_caseopt | cd->req_varyopt;
5576 }
5577 }
5578
5579 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5580 }
5581
5582 /* If previous was a character type match (\d or similar), abolish it and
5583 create a suitable repeat item. The code is shared with single-character
5584 repeats by setting op_type to add a suitable offset into repeat_type. Note
5585 the the Unicode property types will be present only when SUPPORT_UCP is
5586 defined, but we don't wrap the little bits of code here because it just
5587 makes it horribly messy. */
5588
5589 else if (*previous < OP_EODN)
5590 {
5591 pcre_uchar *oldcode;
5592 int prop_type, prop_value;
5593 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5594 c = *previous;
5595
5596 OUTPUT_SINGLE_REPEAT:
5597 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5598 {
5599 prop_type = previous[1];
5600 prop_value = previous[2];
5601 }
5602 else prop_type = prop_value = -1;
5603
5604 oldcode = code;
5605 code = previous; /* Usually overwrite previous item */
5606
5607 /* If the maximum is zero then the minimum must also be zero; Perl allows
5608 this case, so we do too - by simply omitting the item altogether. */
5609
5610 if (repeat_max == 0) goto END_REPEAT;
5611
5612 /* Combine the op_type with the repeat_type */
5613
5614 repeat_type += op_type;
5615
5616 /* A minimum of zero is handled either as the special case * or ?, or as
5617 an UPTO, with the maximum given. */
5618
5619 if (repeat_min == 0)
5620 {
5621 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5622 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5623 else
5624 {
5625 *code++ = OP_UPTO + repeat_type;
5626 PUT2INC(code, 0, repeat_max);
5627 }
5628 }
5629
5630 /* A repeat minimum of 1 is optimized into some special cases. If the
5631 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5632 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5633 one less than the maximum. */
5634
5635 else if (repeat_min == 1)
5636 {
5637 if (repeat_max == -1)
5638 *code++ = OP_PLUS + repeat_type;
5639 else
5640 {
5641 code = oldcode; /* leave previous item in place */
5642 if (repeat_max == 1) goto END_REPEAT;
5643 *code++ = OP_UPTO + repeat_type;
5644 PUT2INC(code, 0, repeat_max - 1);
5645 }
5646 }
5647
5648 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5649 handled as an EXACT followed by an UPTO. */
5650
5651 else
5652 {
5653 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5654 PUT2INC(code, 0, repeat_min);
5655
5656 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5657 we have to insert the character for the previous code. For a repeated
5658 Unicode property match, there are two extra bytes that define the
5659 required property. In UTF-8 mode, long characters have their length in
5660 c, with the UTF_LENGTH bit as a flag. */
5661
5662 if (repeat_max < 0)
5663 {
5664 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5665 if (utf && (c & UTF_LENGTH) != 0)
5666 {
5667 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5668 code += c & 7;
5669 }
5670 else
5671 #endif
5672 {
5673 *code++ = c;
5674 if (prop_type >= 0)
5675 {
5676 *code++ = prop_type;
5677 *code++ = prop_value;
5678 }
5679 }
5680 *code++ = OP_STAR + repeat_type;
5681 }
5682
5683 /* Else insert an UPTO if the max is greater than the min, again
5684 preceded by the character, for the previously inserted code. If the
5685 UPTO is just for 1 instance, we can use QUERY instead. */
5686
5687 else if (repeat_max != repeat_min)
5688 {
5689 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5690 if (utf && (c & UTF_LENGTH) != 0)
5691 {
5692 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5693 code += c & 7;
5694 }
5695 else
5696 #endif
5697 *code++ = c;
5698 if (prop_type >= 0)
5699 {
5700 *code++ = prop_type;
5701 *code++ = prop_value;
5702 }
5703 repeat_max -= repeat_min;
5704
5705 if (repeat_max == 1)
5706 {
5707 *code++ = OP_QUERY + repeat_type;
5708 }
5709 else
5710 {
5711 *code++ = OP_UPTO + repeat_type;
5712 PUT2INC(code, 0, repeat_max);
5713 }
5714 }
5715 }
5716
5717 /* The character or character type itself comes last in all cases. */
5718
5719 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5720 if (utf && (c & UTF_LENGTH) != 0)
5721 {
5722 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5723 code += c & 7;
5724 }
5725 else
5726 #endif
5727 *code++ = c;
5728
5729 /* For a repeated Unicode property match, there are two extra bytes that
5730 define the required property. */
5731
5732 #ifdef SUPPORT_UCP
5733 if (prop_type >= 0)
5734 {
5735 *code++ = prop_type;
5736 *code++ = prop_value;
5737 }
5738 #endif
5739 }
5740
5741 /* If previous was a character class or a back reference, we put the repeat
5742 stuff after it, but just skip the item if the repeat was {0,0}. */
5743
5744 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5745 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5746 *previous == OP_XCLASS ||
5747 #endif
5748 *previous == OP_REF || *previous == OP_REFI ||
5749 *previous == OP_DNREF || *previous == OP_DNREFI)
5750 {
5751 if (repeat_max == 0)
5752 {
5753 code = previous;
5754 goto END_REPEAT;
5755 }
5756
5757 if (repeat_min == 0 && repeat_max == -1)
5758 *code++ = OP_CRSTAR + repeat_type;
5759 else if (repeat_min == 1 && repeat_max == -1)
5760 *code++ = OP_CRPLUS + repeat_type;
5761 else if (repeat_min == 0 && repeat_max == 1)
5762 *code++ = OP_CRQUERY + repeat_type;
5763 else
5764 {
5765 *code++ = OP_CRRANGE + repeat_type;
5766 PUT2INC(code, 0, repeat_min);
5767 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5768 PUT2INC(code, 0, repeat_max);
5769 }
5770 }
5771
5772 /* If previous was a bracket group, we may have to replicate it in certain
5773 cases. Note that at this point we can encounter only the "basic" bracket
5774 opcodes such as BRA and CBRA, as this is the place where they get converted
5775 into the more special varieties such as BRAPOS and SBRA. A test for >=
5776 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5777 ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5778 Originally, PCRE did not allow repetition of assertions, but now it does,
5779 for Perl compatibility. */
5780
5781 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5782 {
5783 register int i;
5784 int len = (int)(code - previous);
5785 pcre_uchar *bralink = NULL;
5786 pcre_uchar *brazeroptr = NULL;
5787
5788 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5789 we just ignore the repeat. */
5790
5791 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5792 goto END_REPEAT;
5793
5794 /* There is no sense in actually repeating assertions. The only potential
5795 use of repetition is in cases when the assertion is optional. Therefore,
5796 if the minimum is greater than zero, just ignore the repeat. If the
5797 maximum is not zero or one, set it to 1. */
5798
5799 if (*previous < OP_ONCE) /* Assertion */
5800 {
5801 if (repeat_min > 0) goto END_REPEAT;
5802 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5803 }
5804
5805 /* The case of a zero minimum is special because of the need to stick
5806 OP_BRAZERO in front of it, and because the group appears once in the
5807 data, whereas in other cases it appears the minimum number of times. For
5808 this reason, it is simplest to treat this case separately, as otherwise
5809 the code gets far too messy. There are several special subcases when the
5810 minimum is zero. */
5811
5812 if (repeat_min == 0)
5813 {
5814 /* If the maximum is also zero, we used to just omit the group from the
5815 output altogether, like this:
5816
5817 ** if (repeat_max == 0)
5818 ** {
5819 ** code = previous;
5820 ** goto END_REPEAT;
5821 ** }
5822
5823 However, that fails when a group or a subgroup within it is referenced
5824 as a subroutine from elsewhere in the pattern, so now we stick in
5825 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5826 don't have a list of which groups are referenced, we cannot do this
5827 selectively.
5828
5829 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5830 and do no more at this point. However, we do need to adjust any
5831 OP_RECURSE calls inside the group that refer to the group itself or any
5832 internal or forward referenced group, because the offset is from the
5833 start of the whole regex. Temporarily terminate the pattern while doing
5834 this. */
5835
5836 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5837 {
5838 *code = OP_END;
5839 adjust_recurse(previous, 1, utf, cd, save_hwm);
5840 memmove(previous + 1, previous, IN_UCHARS(len));
5841 code++;
5842 if (repeat_max == 0)
5843 {
5844 *previous++ = OP_SKIPZERO;
5845 goto END_REPEAT;
5846 }
5847 brazeroptr = previous; /* Save for possessive optimizing */
5848 *previous++ = OP_BRAZERO + repeat_type;
5849 }
5850
5851 /* If the maximum is greater than 1 and limited, we have to replicate
5852 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
5853 The first one has to be handled carefully because it's the original
5854 copy, which has to be moved up. The remainder can be handled by code
5855 that is common with the non-zero minimum case below. We have to
5856 adjust the value or repeat_max, since one less copy is required. Once
5857 again, we may have to adjust any OP_RECURSE calls inside the group. */
5858
5859 else
5860 {
5861 int offset;
5862 *code = OP_END;
5863 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5864 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5865 code += 2 + LINK_SIZE;
5866 *previous++ = OP_BRAZERO + repeat_type;
5867 *previous++ = OP_BRA;
5868
5869 /* We chain together the bracket offset fields that have to be
5870 filled in later when the ends of the brackets are reached. */
5871
5872 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
5873 bralink = previous;
5874 PUTINC(previous, 0, offset);
5875 }
5876
5877 repeat_max--;
5878 }
5879
5880 /* If the minimum is greater than zero, replicate the group as many
5881 times as necessary, and adjust the maximum to the number of subsequent
5882 copies that we need. If we set a first char from the group, and didn't
5883 set a required char, copy the latter from the former. If there are any
5884 forward reference subroutine calls in the group, there will be entries on
5885 the workspace list; replicate these with an appropriate increment. */
5886
5887 else
5888 {
5889 if (repeat_min > 1)
5890 {
5891 /* In the pre-compile phase, we don't actually do the replication. We
5892 just adjust the length as if we had. Do some paranoid checks for
5893 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
5894 integer type when available, otherwise double. */
5895
5896 if (lengthptr != NULL)
5897 {
5898 int delta = (repeat_min - 1)*length_prevgroup;
5899 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
5900 (INT64_OR_DOUBLE)length_prevgroup >
5901 (INT64_OR_DOUBLE)INT_MAX ||
5902 OFLOW_MAX - *lengthptr < delta)
5903 {
5904 *errorcodeptr = ERR20;
5905 goto FAILED;
5906 }
5907 *lengthptr += delta;
5908 }
5909
5910 /* This is compiling for real. If there is a set first byte for
5911 the group, and we have not yet set a "required byte", set it. Make
5912 sure there is enough workspace for copying forward references before
5913 doing the copy. */
5914
5915 else
5916 {
5917 if (groupsetfirstchar && reqcharflags < 0)
5918 {
5919 reqchar = firstchar;
5920 reqcharflags = firstcharflags;
5921 }
5922
5923 for (i = 1; i < repeat_min; i++)
5924 {
5925 pcre_uchar *hc;
5926 pcre_uchar *this_hwm = cd->hwm;
5927 memcpy(code, previous, IN_UCHARS(len));
5928
5929 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5930 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5931 {
5932 int save_offset = save_hwm - cd->start_workspace;
5933 int this_offset = this_hwm - cd->start_workspace;
5934 *errorcodeptr = expand_workspace(cd);
5935 if (*errorcodeptr != 0) goto FAILED;
5936 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5937 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5938 }
5939
5940 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5941 {
5942 PUT(cd->hwm, 0, GET(hc, 0) + len);
5943 cd->hwm += LINK_SIZE;
5944 }
5945 save_hwm = this_hwm;
5946 code += len;
5947 }
5948 }
5949 }
5950
5951 if (repeat_max > 0) repeat_max -= repeat_min;
5952 }
5953
5954 /* This code is common to both the zero and non-zero minimum cases. If
5955 the maximum is limited, it replicates the group in a nested fashion,
5956 remembering the bracket starts on a stack. In the case of a zero minimum,
5957 the first one was set up above. In all cases the repeat_max now specifies
5958 the number of additional copies needed. Again, we must remember to
5959 replicate entries on the forward reference list. */
5960
5961 if (repeat_max >= 0)
5962 {
5963 /* In the pre-compile phase, we don't actually do the replication. We
5964 just adjust the length as if we had. For each repetition we must add 1
5965 to the length for BRAZERO and for all but the last repetition we must
5966 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
5967 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
5968 a 64-bit integer type when available, otherwise double. */
5969
5970 if (lengthptr != NULL && repeat_max > 0)
5971 {
5972 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
5973 2 - 2*LINK_SIZE; /* Last one doesn't nest */
5974 if ((INT64_OR_DOUBLE)repeat_max *
5975 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
5976 > (INT64_OR_DOUBLE)INT_MAX ||
5977 OFLOW_MAX - *lengthptr < delta)
5978 {
5979 *errorcodeptr = ERR20;
5980 goto FAILED;
5981 }
5982 *lengthptr += delta;
5983 }
5984
5985 /* This is compiling for real */
5986
5987 else for (i = repeat_max - 1; i >= 0; i--)
5988 {
5989 pcre_uchar *hc;
5990 pcre_uchar *this_hwm = cd->hwm;
5991
5992 *code++ = OP_BRAZERO + repeat_type;
5993
5994 /* All but the final copy start a new nesting, maintaining the
5995 chain of brackets outstanding. */
5996
5997 if (i != 0)
5998 {
5999 int offset;
6000 *code++ = OP_BRA;
6001 offset = (bralink == NULL)? 0 : (int)(code - bralink);
6002 bralink = code;
6003 PUTINC(code, 0, offset);
6004 }
6005
6006 memcpy(code, previous, IN_UCHARS(len));
6007
6008 /* Ensure there is enough workspace for forward references before
6009 copying them. */
6010
6011 while (cd->hwm > cd->start_workspace + cd->workspace_size -
6012 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
6013 {
6014 int save_offset = save_hwm - cd->start_workspace;
6015 int this_offset = this_hwm - cd->start_workspace;
6016 *errorcodeptr = expand_workspace(cd);
6017 if (*errorcodeptr != 0) goto FAILED;
6018 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
6019 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
6020 }
6021
6022 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
6023 {
6024 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
6025 cd->hwm += LINK_SIZE;
6026 }
6027 save_hwm = this_hwm;
6028 code += len;
6029 }
6030
6031 /* Now chain through the pending brackets, and fill in their length
6032 fields (which are holding the chain links pro tem). */
6033
6034 while (bralink != NULL)
6035 {
6036 int oldlinkoffset;
6037 int offset = (int)(code - bralink + 1);
6038 pcre_uchar *bra = code - offset;
6039 oldlinkoffset = GET(bra, 1);
6040 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
6041 *code++ = OP_KET;
6042 PUTINC(code, 0, offset);
6043 PUT(bra, 1, offset);
6044 }
6045 }
6046
6047 /* If the maximum is unlimited, set a repeater in the final copy. For
6048 ONCE brackets, that's all we need to do. However, possessively repeated
6049 ONCE brackets can be converted into non-capturing brackets, as the
6050 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
6051 deal with possessive ONCEs specially.
6052
6053 Otherwise, when we are doing the actual compile phase, check to see
6054 whether this group is one that could match an empty string. If so,
6055 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
6056 that runtime checking can be done. [This check is also applied to ONCE
6057 groups at runtime, but in a different way.]
6058
6059 Then, if the quantifier was possessive and the bracket is not a
6060 conditional, we convert the BRA code to the POS form, and the KET code to
6061 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
6062 subpattern at both the start and at the end.) The use of special opcodes
6063 makes it possible to reduce greatly the stack usage in pcre_exec(). If
6064 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
6065
6066 Then, if the minimum number of matches is 1 or 0, cancel the possessive
6067 flag so that the default action below, of wrapping everything inside
6068 atomic brackets, does not happen. When the minimum is greater than 1,
6069 there will be earlier copies of the group, and so we still have to wrap
6070 the whole thing. */
6071
6072 else
6073 {
6074 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
6075 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
6076
6077 /* Convert possessive ONCE brackets to non-capturing */
6078
6079 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
6080 possessive_quantifier) *bracode = OP_BRA;
6081
6082 /* For non-possessive ONCE brackets, all we need to do is to
6083 set the KET. */
6084
6085 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
6086 *ketcode = OP_KETRMAX + repeat_type;
6087
6088 /* Handle non-ONCE brackets and possessive ONCEs (which have been
6089 converted to non-capturing above). */
6090
6091 else
6092 {
6093 /* In the compile phase, check for empty string matching. */
6094
6095 if (lengthptr == NULL)
6096 {
6097 pcre_uchar *scode = bracode;
6098 do
6099 {
6100 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6101 {
6102 *bracode += OP_SBRA - OP_BRA;
6103 break;
6104 }
6105 scode += GET(scode, 1);
6106 }
6107 while (*scode == OP_ALT);
6108 }
6109
6110 /* Handle possessive quantifiers. */
6111
6112 if (possessive_quantifier)
6113 {
6114 /* For COND brackets, we wrap the whole thing in a possessively
6115 repeated non-capturing bracket, because we have not invented POS
6116 versions of the COND opcodes. Because we are moving code along, we
6117 must ensure that any pending recursive references are updated. */
6118
6119 if (*bracode == OP_COND || *bracode == OP_SCOND)
6120 {
6121 int nlen = (int)(code - bracode);
6122 *code = OP_END;
6123 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
6124 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6125 code += 1 + LINK_SIZE;
6126 nlen += 1 + LINK_SIZE;
6127 *bracode = OP_BRAPOS;
6128 *code++ = OP_KETRPOS;
6129 PUTINC(code, 0, nlen);
6130 PUT(bracode, 1, nlen);
6131 }
6132
6133 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6134
6135 else
6136 {
6137 *bracode += 1; /* Switch to xxxPOS opcodes */
6138 *ketcode = OP_KETRPOS;
6139 }
6140
6141 /* If the minimum is zero, mark it as possessive, then unset the
6142 possessive flag when the minimum is 0 or 1. */
6143
6144 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6145 if (repeat_min < 2) possessive_quantifier = FALSE;
6146 }
6147
6148 /* Non-possessive quantifier */
6149
6150 else *ketcode = OP_KETRMAX + repeat_type;
6151 }
6152 }
6153 }
6154
6155 /* If previous is OP_FAIL, it was generated by an empty class [] in
6156 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6157 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6158 error above. We can just ignore the repeat in JS case. */
6159
6160 else if (*previous == OP_FAIL) goto END_REPEAT;
6161
6162 /* Else there's some kind of shambles */
6163
6164 else
6165 {
6166 *errorcodeptr = ERR11;
6167 goto FAILED;
6168 }
6169
6170 /* If the character following a repeat is '+', possessive_quantifier is
6171 TRUE. For some opcodes, there are special alternative opcodes for this
6172 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6173 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6174 Sun's Java package, but the special opcodes can optimize it.
6175
6176 Some (but not all) possessively repeated subpatterns have already been
6177 completely handled in the code just above. For them, possessive_quantifier
6178 is always FALSE at this stage. Note that the repeated item starts at
6179 tempcode, not at previous, which might be the first part of a string whose
6180 (former) last char we repeated. */
6181
6182 if (possessive_quantifier)
6183 {
6184 int len;
6185
6186 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6187 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6188 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6189 remains is greater than zero, there's a further opcode that can be
6190 handled. If not, do nothing, leaving the EXACT alone. */
6191
6192 switch(*tempcode)
6193 {
6194 case OP_TYPEEXACT:
6195 tempcode += PRIV(OP_lengths)[*tempcode] +
6196 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6197 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6198 break;
6199
6200 /* CHAR opcodes are used for exacts whose count is 1. */
6201
6202 case OP_CHAR:
6203 case OP_CHARI:
6204 case OP_NOT:
6205 case OP_NOTI:
6206 case OP_EXACT:
6207 case OP_EXACTI:
6208 case OP_NOTEXACT:
6209 case OP_NOTEXACTI:
6210 tempcode += PRIV(OP_lengths)[*tempcode];
6211 #ifdef SUPPORT_UTF
6212 if (utf && HAS_EXTRALEN(tempcode[-1]))
6213 tempcode += GET_EXTRALEN(tempcode[-1]);
6214 #endif
6215 break;
6216
6217 /* For the class opcodes, the repeat operator appears at the end;
6218 adjust tempcode to point to it. */
6219
6220 case OP_CLASS:
6221 case OP_NCLASS:
6222 tempcode += 1 + 32/sizeof(pcre_uchar);
6223 break;
6224
6225 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6226 case OP_XCLASS:
6227 tempcode += GET(tempcode, 1);
6228 break;
6229 #endif
6230 }
6231
6232 /* If tempcode is equal to code (which points to the end of the repeated
6233 item), it means we have skipped an EXACT item but there is no following
6234 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6235 all other cases, tempcode will be pointing to the repeat opcode, and will
6236 be less than code, so the value of len will be greater than 0. */
6237
6238 len = (int)(code - tempcode);
6239 if (len > 0)
6240 {
6241 unsigned int repcode = *tempcode;
6242
6243 /* There is a table for possessifying opcodes, all of which are less
6244 than OP_CALLOUT. A zero entry means there is no possessified version.
6245 */
6246
6247 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6248 *tempcode = opcode_possessify[repcode];
6249
6250 /* For opcode without a special possessified version, wrap the item in
6251 ONCE brackets. Because we are moving code along, we must ensure that any
6252 pending recursive references are updated. */
6253
6254 else
6255 {
6256 *code = OP_END;
6257 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6258 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6259 code += 1 + LINK_SIZE;
6260 len += 1 + LINK_SIZE;
6261 tempcode[0] = OP_ONCE;
6262 *code++ = OP_KET;
6263 PUTINC(code, 0, len);
6264 PUT(tempcode, 1, len);
6265 }
6266 }
6267
6268 #ifdef NEVER
6269 if (len > 0) switch (*tempcode)
6270 {
6271 case OP_STAR: *tempcode = OP_POSSTAR; break;
6272 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6273 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6274 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6275
6276 case OP_STARI: *tempcode = OP_POSSTARI; break;
6277 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6278 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6279 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6280
6281 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6282 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6283 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6284 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6285
6286 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6287 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6288 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6289 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6290
6291 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6292 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6293 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6294 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6295
6296 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6297 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6298 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6299 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6300
6301 /* Because we are moving code along, we must ensure that any
6302 pending recursive references are updated. */
6303
6304 default:
6305 *code = OP_END;
6306 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6307 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6308 code += 1 + LINK_SIZE;
6309 len += 1 + LINK_SIZE;
6310 tempcode[0] = OP_ONCE;
6311 *code++ = OP_KET;
6312 PUTINC(code, 0, len);
6313 PUT(tempcode, 1, len);
6314 break;
6315 }
6316 #endif
6317 }
6318
6319 /* In all case we no longer have a previous item. We also set the
6320 "follows varying string" flag for subsequently encountered reqchars if
6321 it isn't already set and we have just passed a varying length item. */
6322
6323 END_REPEAT:
6324 previous = NULL;
6325 cd->req_varyopt |= reqvary;
6326 break;
6327
6328
6329 /* ===================================================================*/
6330 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6331 lookbehind or option setting or condition or all the other extended
6332 parenthesis forms. */
6333
6334 case CHAR_LEFT_PARENTHESIS:
6335 newoptions = options;
6336 skipbytes = 0;
6337 bravalue = OP_CBRA;
6338 save_hwm = cd->hwm;
6339 reset_bracount = FALSE;
6340
6341 /* First deal with various "verbs" that can be introduced by '*'. */
6342
6343 ptr++;
6344 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6345 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6346 {
6347 int i, namelen;
6348 int arglen = 0;
6349 const char *vn = verbnames;
6350 const pcre_uchar *name = ptr + 1;
6351 const pcre_uchar *arg = NULL;
6352 previous = NULL;
6353 ptr++;
6354 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6355 namelen = (int)(ptr - name);
6356
6357 /* It appears that Perl allows any characters whatsoever, other than
6358 a closing parenthesis, to appear in arguments, so we no longer insist on
6359 letters, digits, and underscores. */
6360
6361 if (*ptr == CHAR_COLON)
6362 {
6363 arg = ++ptr;
6364 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6365 arglen = (int)(ptr - arg);
6366 if ((unsigned int)arglen > MAX_MARK)
6367 {
6368 *errorcodeptr = ERR75;
6369 goto FAILED;
6370 }
6371 }
6372
6373 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6374 {
6375 *errorcodeptr = ERR60;
6376 goto FAILED;
6377 }
6378
6379 /* Scan the table of verb names */
6380
6381 for (i = 0; i < verbcount; i++)
6382 {
6383 if (namelen == verbs[i].len &&
6384 STRNCMP_UC_C8(name, vn, namelen) == 0)
6385 {
6386 int setverb;
6387
6388 /* Check for open captures before ACCEPT and convert it to
6389 ASSERT_ACCEPT if in an assertion. */
6390
6391 if (verbs[i].op == OP_ACCEPT)
6392 {
6393 open_capitem *oc;
6394 if (arglen != 0)
6395 {
6396 *errorcodeptr = ERR59;
6397 goto FAILED;
6398 }
6399 cd->had_accept = TRUE;
6400 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6401 {
6402 *code++ = OP_CLOSE;
6403 PUT2INC(code, 0, oc->number);
6404 }
6405 setverb = *code++ =
6406 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6407
6408 /* Do not set firstchar after *ACCEPT */
6409 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6410 }
6411
6412 /* Handle other cases with/without an argument */
6413
6414 else if (arglen == 0)
6415 {
6416 if (verbs[i].op < 0) /* Argument is mandatory */
6417 {
6418 *errorcodeptr = ERR66;
6419 goto FAILED;
6420 }
6421 setverb = *code++ = verbs[i].op;
6422 }
6423
6424 else
6425 {
6426 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6427 {
6428 *errorcodeptr = ERR59;
6429 goto FAILED;
6430 }
6431 setverb = *code++ = verbs[i].op_arg;
6432 *code++ = arglen;
6433 memcpy(code, arg, IN_UCHARS(arglen));
6434 code += arglen;
6435 *code++ = 0;
6436 }
6437
6438 switch (setverb)
6439 {
6440 case OP_THEN:
6441 case OP_THEN_ARG:
6442 cd->external_flags |= PCRE_HASTHEN;
6443 break;
6444
6445 case OP_PRUNE:
6446 case OP_PRUNE_ARG:
6447 case OP_SKIP:
6448 case OP_SKIP_ARG:
6449 cd->had_pruneorskip = TRUE;
6450 break;
6451 }
6452
6453 break; /* Found verb, exit loop */
6454 }
6455
6456 vn += verbs[i].len + 1;
6457 }
6458
6459 if (i < verbcount) continue; /* Successfully handled a verb */
6460 *errorcodeptr = ERR60; /* Verb not recognized */
6461 goto FAILED;
6462 }
6463
6464 /* Deal with the extended parentheses; all are introduced by '?', and the
6465 appearance of any of them means that this is not a capturing group. */
6466
6467 else if (*ptr == CHAR_QUESTION_MARK)
6468 {
6469 int i, set, unset, namelen;
6470 int *optset;
6471 const pcre_uchar *name;
6472 pcre_uchar *slot;
6473
6474 switch (*(++ptr))
6475 {
6476 case CHAR_NUMBER_SIGN: /* Comment; skip to ket */
6477 ptr++;
6478 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6479 if (*ptr == CHAR_NULL)
6480 {
6481 *errorcodeptr = ERR18;
6482 goto FAILED;
6483 }
6484 continue;
6485
6486
6487 /* ------------------------------------------------------------ */
6488 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6489 reset_bracount = TRUE;
6490 /* Fall through */
6491
6492 /* ------------------------------------------------------------ */
6493 case CHAR_COLON: /* Non-capturing bracket */
6494 bravalue = OP_BRA;
6495 ptr++;
6496 break;
6497
6498
6499 /* ------------------------------------------------------------ */
6500 case CHAR_LEFT_PARENTHESIS:
6501 bravalue = OP_COND; /* Conditional group */
6502 tempptr = ptr;
6503
6504 /* A condition can be an assertion, a number (referring to a numbered
6505 group's having been set), a name (referring to a named group), or 'R',
6506 referring to recursion. R<digits> and R&name are also permitted for
6507 recursion tests.
6508
6509 There are ways of testing a named group: (?(name)) is used by Python;
6510 Perl 5.10 onwards uses (?(<name>) or (?('name')).
6511
6512 There is one unfortunate ambiguity, caused by history. 'R' can be the
6513 recursive thing or the name 'R' (and similarly for 'R' followed by
6514 digits). We look for a name first; if not found, we try the other case.
6515
6516 For compatibility with auto-callouts, we allow a callout to be
6517 specified before a condition that is an assertion. First, check for the
6518 syntax of a callout; if found, adjust the temporary pointer that is
6519 used to check for an assertion condition. That's all that is needed! */
6520
6521 if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6522 {
6523 for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6524 if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6525 tempptr += i + 1;
6526 }
6527
6528 /* For conditions that are assertions, check the syntax, and then exit
6529 the switch. This will take control down to where bracketed groups,
6530 including assertions, are processed. */
6531
6532 if (tempptr[1] == CHAR_QUESTION_MARK &&
6533 (tempptr[2] == CHAR_EQUALS_SIGN ||
6534 tempptr[2] == CHAR_EXCLAMATION_MARK ||
6535 tempptr[2] == CHAR_LESS_THAN_SIGN))
6536 break;
6537
6538 /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6539 need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6540
6541 code[1+LINK_SIZE] = OP_CREF;
6542 skipbytes = 1+IMM2_SIZE;
6543 refsign = -1;
6544
6545 /* Check for a test for recursion in a named group. */
6546
6547 ptr++;
6548 if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6549 {
6550 terminator = -1;
6551 ptr += 2;
6552 code[1+LINK_SIZE] = OP_RREF; /* Change the type of test */
6553 }
6554
6555 /* Check for a test for a named group's having been set, using the Perl
6556 syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6557 syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6558
6559 else if (*ptr == CHAR_LESS_THAN_SIGN)
6560 {
6561 terminator = CHAR_GREATER_THAN_SIGN;
6562 ptr++;
6563 }
6564 else if (*ptr == CHAR_APOSTROPHE)
6565 {
6566 terminator = CHAR_APOSTROPHE;
6567 ptr++;
6568 }
6569 else
6570 {
6571 terminator = CHAR_NULL;
6572 if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6573 else if (IS_DIGIT(*ptr)) refsign = 0;
6574 }
6575
6576 /* Handle a number */
6577
6578 if (refsign >= 0)
6579 {
6580 recno = 0;
6581 while (IS_DIGIT(*ptr))
6582 {
6583 recno = recno * 10 + (int)(*ptr - CHAR_0);
6584 ptr++;
6585 }
6586 }
6587
6588 /* Otherwise we expect to read a name; anything else is an error. When
6589 a name is one of a number of duplicates, a different opcode is used and
6590 it needs more memory. Unfortunately we cannot tell whether a name is a
6591 duplicate in the first pass, so we have to allow for more memory. */
6592
6593 else
6594 {
6595 if (IS_DIGIT(*ptr))
6596 {
6597 *errorcodeptr = ERR84;
6598 goto FAILED;
6599 }
6600 if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6601 {
6602 *errorcodeptr = ERR28; /* Assertion expected */
6603 goto FAILED;
6604 }
6605 name = ptr++;
6606 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6607 {
6608 ptr++;
6609 }
6610 namelen = (int)(ptr - name);
6611 if (lengthptr != NULL) *lengthptr += IMM2_SIZE;
6612 }
6613
6614 /* Check the terminator */
6615
6616 if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6617 *ptr++ != CHAR_RIGHT_PARENTHESIS)
6618 {
6619 ptr--; /* Error offset */
6620 *errorcodeptr = ERR26; /* Malformed number or name */
6621 goto FAILED;
6622 }
6623
6624 /* Do no further checking in the pre-compile phase. */
6625
6626 if (lengthptr != NULL) break;
6627
6628 /* In the real compile we do the work of looking for the actual
6629 reference. If refsign is not negative, it means we have a number in
6630 recno. */
6631
6632 if (refsign >= 0)
6633 {
6634 if (recno <= 0)
6635 {
6636 *errorcodeptr = ERR35;
6637 goto FAILED;
6638 }
6639 if (refsign != 0) recno = (refsign == CHAR_MINUS)?
6640 cd->bracount - recno + 1 : recno + cd->bracount;
6641 if (recno <= 0 || recno > cd->final_bracount)
6642 {
6643 *errorcodeptr = ERR15;
6644 goto FAILED;
6645 }
6646 PUT2(code, 2+LINK_SIZE, recno);
6647 break;
6648 }
6649
6650 /* Otherwise look for the name. */
6651
6652 slot = cd->name_table;
6653 for (i = 0; i < cd->names_found; i++)
6654 {
6655 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
6656 slot += cd->name_entry_size;
6657 }
6658
6659 /* Found the named subpattern. If the name is duplicated, add one to
6660 the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6661 appropriate data values. Otherwise, just insert the unique subpattern