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

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1363 - (show annotations)
Tue Oct 1 16:54:40 2013 UTC (6 years, 1 month ago) by ph10
File MIME type: text/plain
File size: 290075 byte(s)
Refactored auto-possessification code.
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. */
268
269 static const char posix_names[] =
270 STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
271 STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
272 STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
273 STRING_word0 STRING_xdigit;
274
275 static const pcre_uint8 posix_name_lengths[] = {
276 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
277
278 /* Table of class bit maps for each POSIX class. Each class is formed from a
279 base map, with an optional addition or removal of another map. Then, for some
280 classes, there is some additional tweaking: for [:blank:] the vertical space
281 characters are removed, and for [:alpha:] and [:alnum:] the underscore
282 character is removed. The triples in the table consist of the base map offset,
283 second map offset or -1 if no second map, and a non-negative value for map
284 addition or a negative value for map subtraction (if there are two maps). The
285 absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
286 remove vertical space characters, 2 => remove underscore. */
287
288 static const int posix_class_maps[] = {
289 cbit_word, cbit_digit, -2, /* alpha */
290 cbit_lower, -1, 0, /* lower */
291 cbit_upper, -1, 0, /* upper */
292 cbit_word, -1, 2, /* alnum - word without underscore */
293 cbit_print, cbit_cntrl, 0, /* ascii */
294 cbit_space, -1, 1, /* blank - a GNU extension */
295 cbit_cntrl, -1, 0, /* cntrl */
296 cbit_digit, -1, 0, /* digit */
297 cbit_graph, -1, 0, /* graph */
298 cbit_print, -1, 0, /* print */
299 cbit_punct, -1, 0, /* punct */
300 cbit_space, -1, 0, /* space */
301 cbit_word, -1, 0, /* word - a Perl extension */
302 cbit_xdigit,-1, 0 /* xdigit */
303 };
304
305 /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
306 substitutes must be in the order of the names, defined above, and there are
307 both positive and negative cases. NULL means no substitute. */
308
309 #ifdef SUPPORT_UCP
310 static const pcre_uchar string_PNd[] = {
311 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
312 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
313 static const pcre_uchar string_pNd[] = {
314 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
315 CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
316 static const pcre_uchar string_PXsp[] = {
317 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
318 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
319 static const pcre_uchar string_pXsp[] = {
320 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
321 CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
322 static const pcre_uchar string_PXwd[] = {
323 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
324 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
325 static const pcre_uchar string_pXwd[] = {
326 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
327 CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
328
329 static const pcre_uchar *substitutes[] = {
330 string_PNd, /* \D */
331 string_pNd, /* \d */
332 string_PXsp, /* \S */ /* NOTE: Xsp is Perl space */
333 string_pXsp, /* \s */
334 string_PXwd, /* \W */
335 string_pXwd /* \w */
336 };
337
338 static const pcre_uchar string_pL[] = {
339 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
340 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
341 static const pcre_uchar string_pLl[] = {
342 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
343 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
344 static const pcre_uchar string_pLu[] = {
345 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
346 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
347 static const pcre_uchar string_pXan[] = {
348 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
349 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
350 static const pcre_uchar string_h[] = {
351 CHAR_BACKSLASH, CHAR_h, '\0' };
352 static const pcre_uchar string_pXps[] = {
353 CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
354 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
355 static const pcre_uchar string_PL[] = {
356 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
357 CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
358 static const pcre_uchar string_PLl[] = {
359 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
360 CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
361 static const pcre_uchar string_PLu[] = {
362 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
363 CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
364 static const pcre_uchar string_PXan[] = {
365 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
366 CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
367 static const pcre_uchar string_H[] = {
368 CHAR_BACKSLASH, CHAR_H, '\0' };
369 static const pcre_uchar string_PXps[] = {
370 CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
371 CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
372
373 static const pcre_uchar *posix_substitutes[] = {
374 string_pL, /* alpha */
375 string_pLl, /* lower */
376 string_pLu, /* upper */
377 string_pXan, /* alnum */
378 NULL, /* ascii */
379 string_h, /* blank */
380 NULL, /* cntrl */
381 string_pNd, /* digit */
382 NULL, /* graph */
383 NULL, /* print */
384 NULL, /* punct */
385 string_pXps, /* space */ /* NOTE: Xps is POSIX space */
386 string_pXwd, /* word */
387 NULL, /* xdigit */
388 /* Negated cases */
389 string_PL, /* ^alpha */
390 string_PLl, /* ^lower */
391 string_PLu, /* ^upper */
392 string_PXan, /* ^alnum */
393 NULL, /* ^ascii */
394 string_H, /* ^blank */
395 NULL, /* ^cntrl */
396 string_PNd, /* ^digit */
397 NULL, /* ^graph */
398 NULL, /* ^print */
399 NULL, /* ^punct */
400 string_PXps, /* ^space */ /* NOTE: Xps is POSIX space */
401 string_PXwd, /* ^word */
402 NULL /* ^xdigit */
403 };
404 #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
405 #endif
406
407 #define STRING(a) # a
408 #define XSTRING(s) STRING(s)
409
410 /* The texts of compile-time error messages. These are "char *" because they
411 are passed to the outside world. Do not ever re-use any error number, because
412 they are documented. Always add a new error instead. Messages marked DEAD below
413 are no longer used. This used to be a table of strings, but in order to reduce
414 the number of relocations needed when a shared library is loaded dynamically,
415 it is now one long string. We cannot use a table of offsets, because the
416 lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
417 simply count through to the one we want - this isn't a performance issue
418 because these strings are used only when there is a compilation error.
419
420 Each substring ends with \0 to insert a null character. This includes the final
421 substring, so that the whole string ends with \0\0, which can be detected when
422 counting through. */
423
424 static const char error_texts[] =
425 "no error\0"
426 "\\ at end of pattern\0"
427 "\\c at end of pattern\0"
428 "unrecognized character follows \\\0"
429 "numbers out of order in {} quantifier\0"
430 /* 5 */
431 "number too big in {} quantifier\0"
432 "missing terminating ] for character class\0"
433 "invalid escape sequence in character class\0"
434 "range out of order in character class\0"
435 "nothing to repeat\0"
436 /* 10 */
437 "operand of unlimited repeat could match the empty string\0" /** DEAD **/
438 "internal error: unexpected repeat\0"
439 "unrecognized character after (? or (?-\0"
440 "POSIX named classes are supported only within a class\0"
441 "missing )\0"
442 /* 15 */
443 "reference to non-existent subpattern\0"
444 "erroffset passed as NULL\0"
445 "unknown option bit(s) set\0"
446 "missing ) after comment\0"
447 "parentheses nested too deeply\0" /** DEAD **/
448 /* 20 */
449 "regular expression is too large\0"
450 "failed to get memory\0"
451 "unmatched parentheses\0"
452 "internal error: code overflow\0"
453 "unrecognized character after (?<\0"
454 /* 25 */
455 "lookbehind assertion is not fixed length\0"
456 "malformed number or name after (?(\0"
457 "conditional group contains more than two branches\0"
458 "assertion expected after (?(\0"
459 "(?R or (?[+-]digits must be followed by )\0"
460 /* 30 */
461 "unknown POSIX class name\0"
462 "POSIX collating elements are not supported\0"
463 "this version of PCRE is compiled without UTF support\0"
464 "spare error\0" /** DEAD **/
465 "character value in \\x{...} sequence is too large\0"
466 /* 35 */
467 "invalid condition (?(0)\0"
468 "\\C not allowed in lookbehind assertion\0"
469 "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
470 "number after (?C is > 255\0"
471 "closing ) for (?C expected\0"
472 /* 40 */
473 "recursive call could loop indefinitely\0"
474 "unrecognized character after (?P\0"
475 "syntax error in subpattern name (missing terminator)\0"
476 "two named subpatterns have the same name\0"
477 "invalid UTF-8 string\0"
478 /* 45 */
479 "support for \\P, \\p, and \\X has not been compiled\0"
480 "malformed \\P or \\p sequence\0"
481 "unknown property name after \\P or \\p\0"
482 "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
483 "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
484 /* 50 */
485 "repeated subpattern is too long\0" /** DEAD **/
486 "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
487 "internal error: overran compiling workspace\0"
488 "internal error: previously-checked referenced subpattern not found\0"
489 "DEFINE group contains more than one branch\0"
490 /* 55 */
491 "repeating a DEFINE group is not allowed\0" /** DEAD **/
492 "inconsistent NEWLINE options\0"
493 "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
494 "a numbered reference must not be zero\0"
495 "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
496 /* 60 */
497 "(*VERB) not recognized or malformed\0"
498 "number is too big\0"
499 "subpattern name expected\0"
500 "digit expected after (?+\0"
501 "] is an invalid data character in JavaScript compatibility mode\0"
502 /* 65 */
503 "different names for subpatterns of the same number are not allowed\0"
504 "(*MARK) must have an argument\0"
505 "this version of PCRE is not compiled with Unicode property support\0"
506 "\\c must be followed by an ASCII character\0"
507 "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
508 /* 70 */
509 "internal error: unknown opcode in find_fixedlength()\0"
510 "\\N is not supported in a class\0"
511 "too many forward references\0"
512 "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
513 "invalid UTF-16 string\0"
514 /* 75 */
515 "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
516 "character value in \\u.... sequence is too large\0"
517 "invalid UTF-32 string\0"
518 "setting UTF is disabled by the application\0"
519 ;
520
521 /* Table to identify digits and hex digits. This is used when compiling
522 patterns. Note that the tables in chartables are dependent on the locale, and
523 may mark arbitrary characters as digits - but the PCRE compiling code expects
524 to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
525 a private table here. It costs 256 bytes, but it is a lot faster than doing
526 character value tests (at least in some simple cases I timed), and in some
527 applications one wants PCRE to compile efficiently as well as match
528 efficiently.
529
530 For convenience, we use the same bit definitions as in chartables:
531
532 0x04 decimal digit
533 0x08 hexadecimal digit
534
535 Then we can use ctype_digit and ctype_xdigit in the code. */
536
537 /* Using a simple comparison for decimal numbers rather than a memory read
538 is much faster, and the resulting code is simpler (the compiler turns it
539 into a subtraction and unsigned comparison). */
540
541 #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
542
543 #ifndef EBCDIC
544
545 /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
546 UTF-8 mode. */
547
548 static const pcre_uint8 digitab[] =
549 {
550 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
551 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
552 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
553 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
554 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
555 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
556 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
557 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
558 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* @ - G */
559 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H - O */
560 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* P - W */
561 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* X - _ */
562 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
563 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
564 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
565 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
566 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
567 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
568 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
569 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
570 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
571 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
572 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
573 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
574 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
575 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
576 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
577 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
578 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
579 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
580 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
581 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
582
583 #else
584
585 /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
586
587 static const pcre_uint8 digitab[] =
588 {
589 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
590 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
591 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
592 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
593 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
594 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
595 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
596 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
597 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
598 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
599 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
600 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
601 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
602 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
603 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
604 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
605 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
606 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
607 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
608 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
609 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
610 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
611 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
612 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
613 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
614 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
615 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
616 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
617 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
618 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
619 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
620 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
621
622 static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
623 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
624 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
625 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
626 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
627 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
628 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
629 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
630 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
631 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
632 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
633 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
634 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
635 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
636 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
637 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
638 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
639 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
640 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
641 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
642 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
643 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
644 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
645 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
646 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
647 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
648 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
649 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
650 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
651 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
652 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
653 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
654 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
655 #endif
656
657
658 /* This table is used to check whether auto-possessification is possible
659 between adjacent character-type opcodes. The left-hand (repeated) opcode is
660 used to select the row, and the right-hand opcode is use to select the column.
661 A value of 1 means that auto-possessification is OK. For example, the second
662 value in the first row means that \D+\d can be turned into \D++\d.
663
664 The Unicode property types (\P and \p) have to be present to fill out the table
665 because of what their opcode values are, but the table values should always be
666 zero because property types are handled separately in the code. The last four
667 columns apply to items that cannot be repeated, so there is no need to have
668 rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
669 *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
670
671 #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
672 #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
673
674 static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
675 /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
676 { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
677 { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
678 { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
679 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
680 { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
681 { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
682 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
683 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
684 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
685 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
686 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
687 { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
688 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
689 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
690 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
691 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
692 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
693 };
694
695
696 /* This table is used to check whether auto-possessification is possible
697 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
698 left-hand (repeated) opcode is used to select the row, and the right-hand
699 opcode is used to select the column. The values are as follows:
700
701 0 Always return FALSE (never auto-possessify)
702 1 Character groups are distinct (possessify if both are OP_PROP)
703 2 Check character categories in the same group (general or particular)
704 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
705
706 4 Check left general category vs right particular category
707 5 Check right general category vs left particular category
708
709 6 Left alphanum vs right general category
710 7 Left space vs right general category
711 8 Left word vs right general category
712
713 9 Right alphanum vs left general category
714 10 Right space vs left general category
715 11 Right word vs left general category
716
717 12 Left alphanum vs right particular category
718 13 Left space vs right particular category
719 14 Left word vs right particular category
720
721 15 Right alphanum vs left particular category
722 16 Right space vs left particular category
723 17 Right word vs left particular category
724 */
725
726 static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
727 /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
728 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
729 { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
730 { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
731 { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
732 { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
733 { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
734 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
735 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
736 { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
737 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
738 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
739 };
740
741 /* This table is used to check whether auto-possessification is possible
742 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
743 specifies a general category and the other specifies a particular category. The
744 row is selected by the general category and the column by the particular
745 category. The value is 1 if the particular category is not part of the general
746 category. */
747
748 static const pcre_uint8 catposstab[7][30] = {
749 /* 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 */
750 { 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 */
751 { 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 */
752 { 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 */
753 { 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 */
754 { 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 */
755 { 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 */
756 { 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 */
757 };
758
759 /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
760 a general or particular category. The properties in each row are those
761 that apply to the character set in question. Duplication means that a little
762 unnecessary work is done when checking, but this keeps things much simpler
763 because they can all use the same code. For more details see the comment where
764 this table is used.
765
766 Note: SPACE and PXSPACE used to be different because Perl excluded VT from
767 "space", but from Perl 5.18 it's included, so both categories are treated the
768 same here. */
769
770 static const pcre_uint8 posspropstab[3][4] = {
771 { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
772 { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
773 { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
774 };
775
776
777
778 /*************************************************
779 * Find an error text *
780 *************************************************/
781
782 /* The error texts are now all in one long string, to save on relocations. As
783 some of the text is of unknown length, we can't use a table of offsets.
784 Instead, just count through the strings. This is not a performance issue
785 because it happens only when there has been a compilation error.
786
787 Argument: the error number
788 Returns: pointer to the error string
789 */
790
791 static const char *
792 find_error_text(int n)
793 {
794 const char *s = error_texts;
795 for (; n > 0; n--)
796 {
797 while (*s++ != CHAR_NULL) {};
798 if (*s == CHAR_NULL) return "Error text not found (please report)";
799 }
800 return s;
801 }
802
803
804
805 /*************************************************
806 * Expand the workspace *
807 *************************************************/
808
809 /* This function is called during the second compiling phase, if the number of
810 forward references fills the existing workspace, which is originally a block on
811 the stack. A larger block is obtained from malloc() unless the ultimate limit
812 has been reached or the increase will be rather small.
813
814 Argument: pointer to the compile data block
815 Returns: 0 if all went well, else an error number
816 */
817
818 static int
819 expand_workspace(compile_data *cd)
820 {
821 pcre_uchar *newspace;
822 int newsize = cd->workspace_size * 2;
823
824 if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
825 if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
826 newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
827 return ERR72;
828
829 newspace = (PUBL(malloc))(IN_UCHARS(newsize));
830 if (newspace == NULL) return ERR21;
831 memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
832 cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
833 if (cd->workspace_size > COMPILE_WORK_SIZE)
834 (PUBL(free))((void *)cd->start_workspace);
835 cd->start_workspace = newspace;
836 cd->workspace_size = newsize;
837 return 0;
838 }
839
840
841
842 /*************************************************
843 * Check for counted repeat *
844 *************************************************/
845
846 /* This function is called when a '{' is encountered in a place where it might
847 start a quantifier. It looks ahead to see if it really is a quantifier or not.
848 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
849 where the ddds are digits.
850
851 Arguments:
852 p pointer to the first char after '{'
853
854 Returns: TRUE or FALSE
855 */
856
857 static BOOL
858 is_counted_repeat(const pcre_uchar *p)
859 {
860 if (!IS_DIGIT(*p)) return FALSE;
861 p++;
862 while (IS_DIGIT(*p)) p++;
863 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
864
865 if (*p++ != CHAR_COMMA) return FALSE;
866 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
867
868 if (!IS_DIGIT(*p)) return FALSE;
869 p++;
870 while (IS_DIGIT(*p)) p++;
871
872 return (*p == CHAR_RIGHT_CURLY_BRACKET);
873 }
874
875
876
877 /*************************************************
878 * Handle escapes *
879 *************************************************/
880
881 /* This function is called when a \ has been encountered. It either returns a
882 positive value for a simple escape such as \n, or 0 for a data character
883 which will be placed in chptr. A backreference to group n is returned as
884 negative n. When UTF-8 is enabled, a positive value greater than 255 may
885 be returned in chptr.
886 On entry,ptr is pointing at the \. On exit, it is on the final character of the
887 escape sequence.
888
889 Arguments:
890 ptrptr points to the pattern position pointer
891 chptr points to the data character
892 errorcodeptr points to the errorcode variable
893 bracount number of previous extracting brackets
894 options the options bits
895 isclass TRUE if inside a character class
896
897 Returns: zero => a data character
898 positive => a special escape sequence
899 negative => a back reference
900 on error, errorcodeptr is set
901 */
902
903 static int
904 check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
905 int bracount, int options, BOOL isclass)
906 {
907 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
908 BOOL utf = (options & PCRE_UTF8) != 0;
909 const pcre_uchar *ptr = *ptrptr + 1;
910 pcre_uint32 c;
911 int escape = 0;
912 int i;
913
914 GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
915 ptr--; /* Set pointer back to the last byte */
916
917 /* If backslash is at the end of the pattern, it's an error. */
918
919 if (c == CHAR_NULL) *errorcodeptr = ERR1;
920
921 /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
922 in a table. A non-zero result is something that can be returned immediately.
923 Otherwise further processing may be required. */
924
925 #ifndef EBCDIC /* ASCII/UTF-8 coding */
926 /* Not alphanumeric */
927 else if (c < CHAR_0 || c > CHAR_z) {}
928 else if ((i = escapes[c - CHAR_0]) != 0)
929 { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
930
931 #else /* EBCDIC coding */
932 /* Not alphanumeric */
933 else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
934 else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
935 #endif
936
937 /* Escapes that need further processing, or are illegal. */
938
939 else
940 {
941 const pcre_uchar *oldptr;
942 BOOL braced, negated, overflow;
943 int s;
944
945 switch (c)
946 {
947 /* A number of Perl escapes are not handled by PCRE. We give an explicit
948 error. */
949
950 case CHAR_l:
951 case CHAR_L:
952 *errorcodeptr = ERR37;
953 break;
954
955 case CHAR_u:
956 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
957 {
958 /* In JavaScript, \u must be followed by four hexadecimal numbers.
959 Otherwise it is a lowercase u letter. */
960 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
961 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
962 && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
963 && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
964 {
965 c = 0;
966 for (i = 0; i < 4; ++i)
967 {
968 register pcre_uint32 cc = *(++ptr);
969 #ifndef EBCDIC /* ASCII/UTF-8 coding */
970 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
971 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
972 #else /* EBCDIC coding */
973 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
974 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
975 #endif
976 }
977
978 #if defined COMPILE_PCRE8
979 if (c > (utf ? 0x10ffffU : 0xffU))
980 #elif defined COMPILE_PCRE16
981 if (c > (utf ? 0x10ffffU : 0xffffU))
982 #elif defined COMPILE_PCRE32
983 if (utf && c > 0x10ffffU)
984 #endif
985 {
986 *errorcodeptr = ERR76;
987 }
988 else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
989 }
990 }
991 else
992 *errorcodeptr = ERR37;
993 break;
994
995 case CHAR_U:
996 /* In JavaScript, \U is an uppercase U letter. */
997 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
998 break;
999
1000 /* In a character class, \g is just a literal "g". Outside a character
1001 class, \g must be followed by one of a number of specific things:
1002
1003 (1) A number, either plain or braced. If positive, it is an absolute
1004 backreference. If negative, it is a relative backreference. This is a Perl
1005 5.10 feature.
1006
1007 (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1008 is part of Perl's movement towards a unified syntax for back references. As
1009 this is synonymous with \k{name}, we fudge it up by pretending it really
1010 was \k.
1011
1012 (3) For Oniguruma compatibility we also support \g followed by a name or a
1013 number either in angle brackets or in single quotes. However, these are
1014 (possibly recursive) subroutine calls, _not_ backreferences. Just return
1015 the ESC_g code (cf \k). */
1016
1017 case CHAR_g:
1018 if (isclass) break;
1019 if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1020 {
1021 escape = ESC_g;
1022 break;
1023 }
1024
1025 /* Handle the Perl-compatible cases */
1026
1027 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1028 {
1029 const pcre_uchar *p;
1030 for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1031 if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1032 if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1033 {
1034 escape = ESC_k;
1035 break;
1036 }
1037 braced = TRUE;
1038 ptr++;
1039 }
1040 else braced = FALSE;
1041
1042 if (ptr[1] == CHAR_MINUS)
1043 {
1044 negated = TRUE;
1045 ptr++;
1046 }
1047 else negated = FALSE;
1048
1049 /* The integer range is limited by the machine's int representation. */
1050 s = 0;
1051 overflow = FALSE;
1052 while (IS_DIGIT(ptr[1]))
1053 {
1054 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1055 {
1056 overflow = TRUE;
1057 break;
1058 }
1059 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1060 }
1061 if (overflow) /* Integer overflow */
1062 {
1063 while (IS_DIGIT(ptr[1]))
1064 ptr++;
1065 *errorcodeptr = ERR61;
1066 break;
1067 }
1068
1069 if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1070 {
1071 *errorcodeptr = ERR57;
1072 break;
1073 }
1074
1075 if (s == 0)
1076 {
1077 *errorcodeptr = ERR58;
1078 break;
1079 }
1080
1081 if (negated)
1082 {
1083 if (s > bracount)
1084 {
1085 *errorcodeptr = ERR15;
1086 break;
1087 }
1088 s = bracount - (s - 1);
1089 }
1090
1091 escape = -s;
1092 break;
1093
1094 /* The handling of escape sequences consisting of a string of digits
1095 starting with one that is not zero is not straightforward. By experiment,
1096 the way Perl works seems to be as follows:
1097
1098 Outside a character class, the digits are read as a decimal number. If the
1099 number is less than 10, or if there are that many previous extracting
1100 left brackets, then it is a back reference. Otherwise, up to three octal
1101 digits are read to form an escaped byte. Thus \123 is likely to be octal
1102 123 (cf \0123, which is octal 012 followed by the literal 3). If the octal
1103 value is greater than 377, the least significant 8 bits are taken. Inside a
1104 character class, \ followed by a digit is always an octal number. */
1105
1106 case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1107 case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1108
1109 if (!isclass)
1110 {
1111 oldptr = ptr;
1112 /* The integer range is limited by the machine's int representation. */
1113 s = (int)(c -CHAR_0);
1114 overflow = FALSE;
1115 while (IS_DIGIT(ptr[1]))
1116 {
1117 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1118 {
1119 overflow = TRUE;
1120 break;
1121 }
1122 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1123 }
1124 if (overflow) /* Integer overflow */
1125 {
1126 while (IS_DIGIT(ptr[1]))
1127 ptr++;
1128 *errorcodeptr = ERR61;
1129 break;
1130 }
1131 if (s < 10 || s <= bracount)
1132 {
1133 escape = -s;
1134 break;
1135 }
1136 ptr = oldptr; /* Put the pointer back and fall through */
1137 }
1138
1139 /* Handle an octal number following \. If the first digit is 8 or 9, Perl
1140 generates a binary zero byte and treats the digit as a following literal.
1141 Thus we have to pull back the pointer by one. */
1142
1143 if ((c = *ptr) >= CHAR_8)
1144 {
1145 ptr--;
1146 c = 0;
1147 break;
1148 }
1149
1150 /* \0 always starts an octal number, but we may drop through to here with a
1151 larger first octal digit. The original code used just to take the least
1152 significant 8 bits of octal numbers (I think this is what early Perls used
1153 to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1154 but no more than 3 octal digits. */
1155
1156 case CHAR_0:
1157 c -= CHAR_0;
1158 while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1159 c = c * 8 + *(++ptr) - CHAR_0;
1160 #ifdef COMPILE_PCRE8
1161 if (!utf && c > 0xff) *errorcodeptr = ERR51;
1162 #endif
1163 break;
1164
1165 /* \x is complicated. \x{ddd} is a character number which can be greater
1166 than 0xff in utf or non-8bit mode, but only if the ddd are hex digits.
1167 If not, { is treated as a data character. */
1168
1169 case CHAR_x:
1170 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1171 {
1172 /* In JavaScript, \x must be followed by two hexadecimal numbers.
1173 Otherwise it is a lowercase x letter. */
1174 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1175 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1176 {
1177 c = 0;
1178 for (i = 0; i < 2; ++i)
1179 {
1180 register pcre_uint32 cc = *(++ptr);
1181 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1182 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1183 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1184 #else /* EBCDIC coding */
1185 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1186 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1187 #endif
1188 }
1189 }
1190 break;
1191 }
1192
1193 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1194 {
1195 const pcre_uchar *pt = ptr + 2;
1196
1197 c = 0;
1198 overflow = FALSE;
1199 while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0)
1200 {
1201 register pcre_uint32 cc = *pt++;
1202 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1203
1204 #ifdef COMPILE_PCRE32
1205 if (c >= 0x10000000l) { overflow = TRUE; break; }
1206 #endif
1207
1208 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1209 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1210 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1211 #else /* EBCDIC coding */
1212 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1213 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1214 #endif
1215
1216 #if defined COMPILE_PCRE8
1217 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1218 #elif defined COMPILE_PCRE16
1219 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1220 #elif defined COMPILE_PCRE32
1221 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1222 #endif
1223 }
1224
1225 if (overflow)
1226 {
1227 while (MAX_255(*pt) && (digitab[*pt] & ctype_xdigit) != 0) pt++;
1228 *errorcodeptr = ERR34;
1229 }
1230
1231 if (*pt == CHAR_RIGHT_CURLY_BRACKET)
1232 {
1233 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1234 ptr = pt;
1235 break;
1236 }
1237
1238 /* If the sequence of hex digits does not end with '}', then we don't
1239 recognize this construct; fall through to the normal \x handling. */
1240 }
1241
1242 /* Read just a single-byte hex-defined char */
1243
1244 c = 0;
1245 while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1246 {
1247 pcre_uint32 cc; /* Some compilers don't like */
1248 cc = *(++ptr); /* ++ in initializers */
1249 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1250 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1251 c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1252 #else /* EBCDIC coding */
1253 if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1254 c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1255 #endif
1256 }
1257 break;
1258
1259 /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1260 An error is given if the byte following \c is not an ASCII character. This
1261 coding is ASCII-specific, but then the whole concept of \cx is
1262 ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1263
1264 case CHAR_c:
1265 c = *(++ptr);
1266 if (c == CHAR_NULL)
1267 {
1268 *errorcodeptr = ERR2;
1269 break;
1270 }
1271 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1272 if (c > 127) /* Excludes all non-ASCII in either mode */
1273 {
1274 *errorcodeptr = ERR68;
1275 break;
1276 }
1277 if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1278 c ^= 0x40;
1279 #else /* EBCDIC coding */
1280 if (c >= CHAR_a && c <= CHAR_z) c += 64;
1281 c ^= 0xC0;
1282 #endif
1283 break;
1284
1285 /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1286 other alphanumeric following \ is an error if PCRE_EXTRA was set;
1287 otherwise, for Perl compatibility, it is a literal. This code looks a bit
1288 odd, but there used to be some cases other than the default, and there may
1289 be again in future, so I haven't "optimized" it. */
1290
1291 default:
1292 if ((options & PCRE_EXTRA) != 0) switch(c)
1293 {
1294 default:
1295 *errorcodeptr = ERR3;
1296 break;
1297 }
1298 break;
1299 }
1300 }
1301
1302 /* Perl supports \N{name} for character names, as well as plain \N for "not
1303 newline". PCRE does not support \N{name}. However, it does support
1304 quantification such as \N{2,3}. */
1305
1306 if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1307 !is_counted_repeat(ptr+2))
1308 *errorcodeptr = ERR37;
1309
1310 /* If PCRE_UCP is set, we change the values for \d etc. */
1311
1312 if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1313 escape += (ESC_DU - ESC_D);
1314
1315 /* Set the pointer to the final character before returning. */
1316
1317 *ptrptr = ptr;
1318 *chptr = c;
1319 return escape;
1320 }
1321
1322
1323
1324 #ifdef SUPPORT_UCP
1325 /*************************************************
1326 * Handle \P and \p *
1327 *************************************************/
1328
1329 /* This function is called after \P or \p has been encountered, provided that
1330 PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1331 pointing at the P or p. On exit, it is pointing at the final character of the
1332 escape sequence.
1333
1334 Argument:
1335 ptrptr points to the pattern position pointer
1336 negptr points to a boolean that is set TRUE for negation else FALSE
1337 ptypeptr points to an unsigned int that is set to the type value
1338 pdataptr points to an unsigned int that is set to the detailed property value
1339 errorcodeptr points to the error code variable
1340
1341 Returns: TRUE if the type value was found, or FALSE for an invalid type
1342 */
1343
1344 static BOOL
1345 get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1346 unsigned int *pdataptr, int *errorcodeptr)
1347 {
1348 pcre_uchar c;
1349 int i, bot, top;
1350 const pcre_uchar *ptr = *ptrptr;
1351 pcre_uchar name[32];
1352
1353 c = *(++ptr);
1354 if (c == CHAR_NULL) goto ERROR_RETURN;
1355
1356 *negptr = FALSE;
1357
1358 /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1359 negation. */
1360
1361 if (c == CHAR_LEFT_CURLY_BRACKET)
1362 {
1363 if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1364 {
1365 *negptr = TRUE;
1366 ptr++;
1367 }
1368 for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1369 {
1370 c = *(++ptr);
1371 if (c == CHAR_NULL) goto ERROR_RETURN;
1372 if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1373 name[i] = c;
1374 }
1375 if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1376 name[i] = 0;
1377 }
1378
1379 /* Otherwise there is just one following character */
1380
1381 else
1382 {
1383 name[0] = c;
1384 name[1] = 0;
1385 }
1386
1387 *ptrptr = ptr;
1388
1389 /* Search for a recognized property name using binary chop */
1390
1391 bot = 0;
1392 top = PRIV(utt_size);
1393
1394 while (bot < top)
1395 {
1396 int r;
1397 i = (bot + top) >> 1;
1398 r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1399 if (r == 0)
1400 {
1401 *ptypeptr = PRIV(utt)[i].type;
1402 *pdataptr = PRIV(utt)[i].value;
1403 return TRUE;
1404 }
1405 if (r > 0) bot = i + 1; else top = i;
1406 }
1407
1408 *errorcodeptr = ERR47;
1409 *ptrptr = ptr;
1410 return FALSE;
1411
1412 ERROR_RETURN:
1413 *errorcodeptr = ERR46;
1414 *ptrptr = ptr;
1415 return FALSE;
1416 }
1417 #endif
1418
1419
1420
1421 /*************************************************
1422 * Read repeat counts *
1423 *************************************************/
1424
1425 /* Read an item of the form {n,m} and return the values. This is called only
1426 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1427 so the syntax is guaranteed to be correct, but we need to check the values.
1428
1429 Arguments:
1430 p pointer to first char after '{'
1431 minp pointer to int for min
1432 maxp pointer to int for max
1433 returned as -1 if no max
1434 errorcodeptr points to error code variable
1435
1436 Returns: pointer to '}' on success;
1437 current ptr on error, with errorcodeptr set non-zero
1438 */
1439
1440 static const pcre_uchar *
1441 read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1442 {
1443 int min = 0;
1444 int max = -1;
1445
1446 /* Read the minimum value and do a paranoid check: a negative value indicates
1447 an integer overflow. */
1448
1449 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1450 if (min < 0 || min > 65535)
1451 {
1452 *errorcodeptr = ERR5;
1453 return p;
1454 }
1455
1456 /* Read the maximum value if there is one, and again do a paranoid on its size.
1457 Also, max must not be less than min. */
1458
1459 if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1460 {
1461 if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1462 {
1463 max = 0;
1464 while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1465 if (max < 0 || max > 65535)
1466 {
1467 *errorcodeptr = ERR5;
1468 return p;
1469 }
1470 if (max < min)
1471 {
1472 *errorcodeptr = ERR4;
1473 return p;
1474 }
1475 }
1476 }
1477
1478 /* Fill in the required variables, and pass back the pointer to the terminating
1479 '}'. */
1480
1481 *minp = min;
1482 *maxp = max;
1483 return p;
1484 }
1485
1486
1487
1488 /*************************************************
1489 * Find first significant op code *
1490 *************************************************/
1491
1492 /* This is called by several functions that scan a compiled expression looking
1493 for a fixed first character, or an anchoring op code etc. It skips over things
1494 that do not influence this. For some calls, it makes sense to skip negative
1495 forward and all backward assertions, and also the \b assertion; for others it
1496 does not.
1497
1498 Arguments:
1499 code pointer to the start of the group
1500 skipassert TRUE if certain assertions are to be skipped
1501
1502 Returns: pointer to the first significant opcode
1503 */
1504
1505 static const pcre_uchar*
1506 first_significant_code(const pcre_uchar *code, BOOL skipassert)
1507 {
1508 for (;;)
1509 {
1510 switch ((int)*code)
1511 {
1512 case OP_ASSERT_NOT:
1513 case OP_ASSERTBACK:
1514 case OP_ASSERTBACK_NOT:
1515 if (!skipassert) return code;
1516 do code += GET(code, 1); while (*code == OP_ALT);
1517 code += PRIV(OP_lengths)[*code];
1518 break;
1519
1520 case OP_WORD_BOUNDARY:
1521 case OP_NOT_WORD_BOUNDARY:
1522 if (!skipassert) return code;
1523 /* Fall through */
1524
1525 case OP_CALLOUT:
1526 case OP_CREF:
1527 case OP_NCREF:
1528 case OP_RREF:
1529 case OP_NRREF:
1530 case OP_DEF:
1531 code += PRIV(OP_lengths)[*code];
1532 break;
1533
1534 default:
1535 return code;
1536 }
1537 }
1538 /* Control never reaches here */
1539 }
1540
1541
1542
1543 /*************************************************
1544 * Find the fixed length of a branch *
1545 *************************************************/
1546
1547 /* Scan a branch and compute the fixed length of subject that will match it,
1548 if the length is fixed. This is needed for dealing with backward assertions.
1549 In UTF8 mode, the result is in characters rather than bytes. The branch is
1550 temporarily terminated with OP_END when this function is called.
1551
1552 This function is called when a backward assertion is encountered, so that if it
1553 fails, the error message can point to the correct place in the pattern.
1554 However, we cannot do this when the assertion contains subroutine calls,
1555 because they can be forward references. We solve this by remembering this case
1556 and doing the check at the end; a flag specifies which mode we are running in.
1557
1558 Arguments:
1559 code points to the start of the pattern (the bracket)
1560 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1561 atend TRUE if called when the pattern is complete
1562 cd the "compile data" structure
1563
1564 Returns: the fixed length,
1565 or -1 if there is no fixed length,
1566 or -2 if \C was encountered (in UTF-8 mode only)
1567 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1568 or -4 if an unknown opcode was encountered (internal error)
1569 */
1570
1571 static int
1572 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1573 {
1574 int length = -1;
1575
1576 register int branchlength = 0;
1577 register pcre_uchar *cc = code + 1 + LINK_SIZE;
1578
1579 /* Scan along the opcodes for this branch. If we get to the end of the
1580 branch, check the length against that of the other branches. */
1581
1582 for (;;)
1583 {
1584 int d;
1585 pcre_uchar *ce, *cs;
1586 register pcre_uchar op = *cc;
1587
1588 switch (op)
1589 {
1590 /* We only need to continue for OP_CBRA (normal capturing bracket) and
1591 OP_BRA (normal non-capturing bracket) because the other variants of these
1592 opcodes are all concerned with unlimited repeated groups, which of course
1593 are not of fixed length. */
1594
1595 case OP_CBRA:
1596 case OP_BRA:
1597 case OP_ONCE:
1598 case OP_ONCE_NC:
1599 case OP_COND:
1600 d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1601 if (d < 0) return d;
1602 branchlength += d;
1603 do cc += GET(cc, 1); while (*cc == OP_ALT);
1604 cc += 1 + LINK_SIZE;
1605 break;
1606
1607 /* Reached end of a branch; if it's a ket it is the end of a nested call.
1608 If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1609 an ALT. If it is END it's the end of the outer call. All can be handled by
1610 the same code. Note that we must not include the OP_KETRxxx opcodes here,
1611 because they all imply an unlimited repeat. */
1612
1613 case OP_ALT:
1614 case OP_KET:
1615 case OP_END:
1616 case OP_ACCEPT:
1617 case OP_ASSERT_ACCEPT:
1618 if (length < 0) length = branchlength;
1619 else if (length != branchlength) return -1;
1620 if (*cc != OP_ALT) return length;
1621 cc += 1 + LINK_SIZE;
1622 branchlength = 0;
1623 break;
1624
1625 /* A true recursion implies not fixed length, but a subroutine call may
1626 be OK. If the subroutine is a forward reference, we can't deal with
1627 it until the end of the pattern, so return -3. */
1628
1629 case OP_RECURSE:
1630 if (!atend) return -3;
1631 cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1632 do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1633 if (cc > cs && cc < ce) return -1; /* Recursion */
1634 d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1635 if (d < 0) return d;
1636 branchlength += d;
1637 cc += 1 + LINK_SIZE;
1638 break;
1639
1640 /* Skip over assertive subpatterns */
1641
1642 case OP_ASSERT:
1643 case OP_ASSERT_NOT:
1644 case OP_ASSERTBACK:
1645 case OP_ASSERTBACK_NOT:
1646 do cc += GET(cc, 1); while (*cc == OP_ALT);
1647 cc += PRIV(OP_lengths)[*cc];
1648 break;
1649
1650 /* Skip over things that don't match chars */
1651
1652 case OP_MARK:
1653 case OP_PRUNE_ARG:
1654 case OP_SKIP_ARG:
1655 case OP_THEN_ARG:
1656 cc += cc[1] + PRIV(OP_lengths)[*cc];
1657 break;
1658
1659 case OP_CALLOUT:
1660 case OP_CIRC:
1661 case OP_CIRCM:
1662 case OP_CLOSE:
1663 case OP_COMMIT:
1664 case OP_CREF:
1665 case OP_DEF:
1666 case OP_DOLL:
1667 case OP_DOLLM:
1668 case OP_EOD:
1669 case OP_EODN:
1670 case OP_FAIL:
1671 case OP_NCREF:
1672 case OP_NRREF:
1673 case OP_NOT_WORD_BOUNDARY:
1674 case OP_PRUNE:
1675 case OP_REVERSE:
1676 case OP_RREF:
1677 case OP_SET_SOM:
1678 case OP_SKIP:
1679 case OP_SOD:
1680 case OP_SOM:
1681 case OP_THEN:
1682 case OP_WORD_BOUNDARY:
1683 cc += PRIV(OP_lengths)[*cc];
1684 break;
1685
1686 /* Handle literal characters */
1687
1688 case OP_CHAR:
1689 case OP_CHARI:
1690 case OP_NOT:
1691 case OP_NOTI:
1692 branchlength++;
1693 cc += 2;
1694 #ifdef SUPPORT_UTF
1695 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1696 #endif
1697 break;
1698
1699 /* Handle exact repetitions. The count is already in characters, but we
1700 need to skip over a multibyte character in UTF8 mode. */
1701
1702 case OP_EXACT:
1703 case OP_EXACTI:
1704 case OP_NOTEXACT:
1705 case OP_NOTEXACTI:
1706 branchlength += (int)GET2(cc,1);
1707 cc += 2 + IMM2_SIZE;
1708 #ifdef SUPPORT_UTF
1709 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1710 #endif
1711 break;
1712
1713 case OP_TYPEEXACT:
1714 branchlength += GET2(cc,1);
1715 if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1716 cc += 2;
1717 cc += 1 + IMM2_SIZE + 1;
1718 break;
1719
1720 /* Handle single-char matchers */
1721
1722 case OP_PROP:
1723 case OP_NOTPROP:
1724 cc += 2;
1725 /* Fall through */
1726
1727 case OP_HSPACE:
1728 case OP_VSPACE:
1729 case OP_NOT_HSPACE:
1730 case OP_NOT_VSPACE:
1731 case OP_NOT_DIGIT:
1732 case OP_DIGIT:
1733 case OP_NOT_WHITESPACE:
1734 case OP_WHITESPACE:
1735 case OP_NOT_WORDCHAR:
1736 case OP_WORDCHAR:
1737 case OP_ANY:
1738 case OP_ALLANY:
1739 branchlength++;
1740 cc++;
1741 break;
1742
1743 /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1744 otherwise \C is coded as OP_ALLANY. */
1745
1746 case OP_ANYBYTE:
1747 return -2;
1748
1749 /* Check a class for variable quantification */
1750
1751 case OP_CLASS:
1752 case OP_NCLASS:
1753 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1754 case OP_XCLASS:
1755 /* The original code caused an unsigned overflow in 64 bit systems,
1756 so now we use a conditional statement. */
1757 if (op == OP_XCLASS)
1758 cc += GET(cc, 1);
1759 else
1760 cc += PRIV(OP_lengths)[OP_CLASS];
1761 #else
1762 cc += PRIV(OP_lengths)[OP_CLASS];
1763 #endif
1764
1765 switch (*cc)
1766 {
1767 case OP_CRPLUS:
1768 case OP_CRMINPLUS:
1769 case OP_CRSTAR:
1770 case OP_CRMINSTAR:
1771 case OP_CRQUERY:
1772 case OP_CRMINQUERY:
1773 return -1;
1774
1775 case OP_CRRANGE:
1776 case OP_CRMINRANGE:
1777 if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1778 branchlength += (int)GET2(cc,1);
1779 cc += 1 + 2 * IMM2_SIZE;
1780 break;
1781
1782 default:
1783 branchlength++;
1784 }
1785 break;
1786
1787 /* Anything else is variable length */
1788
1789 case OP_ANYNL:
1790 case OP_BRAMINZERO:
1791 case OP_BRAPOS:
1792 case OP_BRAPOSZERO:
1793 case OP_BRAZERO:
1794 case OP_CBRAPOS:
1795 case OP_EXTUNI:
1796 case OP_KETRMAX:
1797 case OP_KETRMIN:
1798 case OP_KETRPOS:
1799 case OP_MINPLUS:
1800 case OP_MINPLUSI:
1801 case OP_MINQUERY:
1802 case OP_MINQUERYI:
1803 case OP_MINSTAR:
1804 case OP_MINSTARI:
1805 case OP_MINUPTO:
1806 case OP_MINUPTOI:
1807 case OP_NOTMINPLUS:
1808 case OP_NOTMINPLUSI:
1809 case OP_NOTMINQUERY:
1810 case OP_NOTMINQUERYI:
1811 case OP_NOTMINSTAR:
1812 case OP_NOTMINSTARI:
1813 case OP_NOTMINUPTO:
1814 case OP_NOTMINUPTOI:
1815 case OP_NOTPLUS:
1816 case OP_NOTPLUSI:
1817 case OP_NOTPOSPLUS:
1818 case OP_NOTPOSPLUSI:
1819 case OP_NOTPOSQUERY:
1820 case OP_NOTPOSQUERYI:
1821 case OP_NOTPOSSTAR:
1822 case OP_NOTPOSSTARI:
1823 case OP_NOTPOSUPTO:
1824 case OP_NOTPOSUPTOI:
1825 case OP_NOTQUERY:
1826 case OP_NOTQUERYI:
1827 case OP_NOTSTAR:
1828 case OP_NOTSTARI:
1829 case OP_NOTUPTO:
1830 case OP_NOTUPTOI:
1831 case OP_PLUS:
1832 case OP_PLUSI:
1833 case OP_POSPLUS:
1834 case OP_POSPLUSI:
1835 case OP_POSQUERY:
1836 case OP_POSQUERYI:
1837 case OP_POSSTAR:
1838 case OP_POSSTARI:
1839 case OP_POSUPTO:
1840 case OP_POSUPTOI:
1841 case OP_QUERY:
1842 case OP_QUERYI:
1843 case OP_REF:
1844 case OP_REFI:
1845 case OP_DNREF:
1846 case OP_DNREFI:
1847 case OP_SBRA:
1848 case OP_SBRAPOS:
1849 case OP_SCBRA:
1850 case OP_SCBRAPOS:
1851 case OP_SCOND:
1852 case OP_SKIPZERO:
1853 case OP_STAR:
1854 case OP_STARI:
1855 case OP_TYPEMINPLUS:
1856 case OP_TYPEMINQUERY:
1857 case OP_TYPEMINSTAR:
1858 case OP_TYPEMINUPTO:
1859 case OP_TYPEPLUS:
1860 case OP_TYPEPOSPLUS:
1861 case OP_TYPEPOSQUERY:
1862 case OP_TYPEPOSSTAR:
1863 case OP_TYPEPOSUPTO:
1864 case OP_TYPEQUERY:
1865 case OP_TYPESTAR:
1866 case OP_TYPEUPTO:
1867 case OP_UPTO:
1868 case OP_UPTOI:
1869 return -1;
1870
1871 /* Catch unrecognized opcodes so that when new ones are added they
1872 are not forgotten, as has happened in the past. */
1873
1874 default:
1875 return -4;
1876 }
1877 }
1878 /* Control never gets here */
1879 }
1880
1881
1882
1883 /*************************************************
1884 * Scan compiled regex for specific bracket *
1885 *************************************************/
1886
1887 /* This little function scans through a compiled pattern until it finds a
1888 capturing bracket with the given number, or, if the number is negative, an
1889 instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1890 so that it can be called from pcre_study() when finding the minimum matching
1891 length.
1892
1893 Arguments:
1894 code points to start of expression
1895 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1896 number the required bracket number or negative to find a lookbehind
1897
1898 Returns: pointer to the opcode for the bracket, or NULL if not found
1899 */
1900
1901 const pcre_uchar *
1902 PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
1903 {
1904 for (;;)
1905 {
1906 register pcre_uchar c = *code;
1907
1908 if (c == OP_END) return NULL;
1909
1910 /* XCLASS is used for classes that cannot be represented just by a bit
1911 map. This includes negated single high-valued characters. The length in
1912 the table is zero; the actual length is stored in the compiled code. */
1913
1914 if (c == OP_XCLASS) code += GET(code, 1);
1915
1916 /* Handle recursion */
1917
1918 else if (c == OP_REVERSE)
1919 {
1920 if (number < 0) return (pcre_uchar *)code;
1921 code += PRIV(OP_lengths)[c];
1922 }
1923
1924 /* Handle capturing bracket */
1925
1926 else if (c == OP_CBRA || c == OP_SCBRA ||
1927 c == OP_CBRAPOS || c == OP_SCBRAPOS)
1928 {
1929 int n = (int)GET2(code, 1+LINK_SIZE);
1930 if (n == number) return (pcre_uchar *)code;
1931 code += PRIV(OP_lengths)[c];
1932 }
1933
1934 /* Otherwise, we can get the item's length from the table, except that for
1935 repeated character types, we have to test for \p and \P, which have an extra
1936 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1937 must add in its length. */
1938
1939 else
1940 {
1941 switch(c)
1942 {
1943 case OP_TYPESTAR:
1944 case OP_TYPEMINSTAR:
1945 case OP_TYPEPLUS:
1946 case OP_TYPEMINPLUS:
1947 case OP_TYPEQUERY:
1948 case OP_TYPEMINQUERY:
1949 case OP_TYPEPOSSTAR:
1950 case OP_TYPEPOSPLUS:
1951 case OP_TYPEPOSQUERY:
1952 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1953 break;
1954
1955 case OP_TYPEUPTO:
1956 case OP_TYPEMINUPTO:
1957 case OP_TYPEEXACT:
1958 case OP_TYPEPOSUPTO:
1959 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
1960 code += 2;
1961 break;
1962
1963 case OP_MARK:
1964 case OP_PRUNE_ARG:
1965 case OP_SKIP_ARG:
1966 case OP_THEN_ARG:
1967 code += code[1];
1968 break;
1969 }
1970
1971 /* Add in the fixed length from the table */
1972
1973 code += PRIV(OP_lengths)[c];
1974
1975 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1976 a multi-byte character. The length in the table is a minimum, so we have to
1977 arrange to skip the extra bytes. */
1978
1979 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
1980 if (utf) switch(c)
1981 {
1982 case OP_CHAR:
1983 case OP_CHARI:
1984 case OP_EXACT:
1985 case OP_EXACTI:
1986 case OP_UPTO:
1987 case OP_UPTOI:
1988 case OP_MINUPTO:
1989 case OP_MINUPTOI:
1990 case OP_POSUPTO:
1991 case OP_POSUPTOI:
1992 case OP_STAR:
1993 case OP_STARI:
1994 case OP_MINSTAR:
1995 case OP_MINSTARI:
1996 case OP_POSSTAR:
1997 case OP_POSSTARI:
1998 case OP_PLUS:
1999 case OP_PLUSI:
2000 case OP_MINPLUS:
2001 case OP_MINPLUSI:
2002 case OP_POSPLUS:
2003 case OP_POSPLUSI:
2004 case OP_QUERY:
2005 case OP_QUERYI:
2006 case OP_MINQUERY:
2007 case OP_MINQUERYI:
2008 case OP_POSQUERY:
2009 case OP_POSQUERYI:
2010 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2011 break;
2012 }
2013 #else
2014 (void)(utf); /* Keep compiler happy by referencing function argument */
2015 #endif
2016 }
2017 }
2018 }
2019
2020
2021
2022 /*************************************************
2023 * Scan compiled regex for recursion reference *
2024 *************************************************/
2025
2026 /* This little function scans through a compiled pattern until it finds an
2027 instance of OP_RECURSE.
2028
2029 Arguments:
2030 code points to start of expression
2031 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2032
2033 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2034 */
2035
2036 static const pcre_uchar *
2037 find_recurse(const pcre_uchar *code, BOOL utf)
2038 {
2039 for (;;)
2040 {
2041 register pcre_uchar c = *code;
2042 if (c == OP_END) return NULL;
2043 if (c == OP_RECURSE) return code;
2044
2045 /* XCLASS is used for classes that cannot be represented just by a bit
2046 map. This includes negated single high-valued characters. The length in
2047 the table is zero; the actual length is stored in the compiled code. */
2048
2049 if (c == OP_XCLASS) code += GET(code, 1);
2050
2051 /* Otherwise, we can get the item's length from the table, except that for
2052 repeated character types, we have to test for \p and \P, which have an extra
2053 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2054 must add in its length. */
2055
2056 else
2057 {
2058 switch(c)
2059 {
2060 case OP_TYPESTAR:
2061 case OP_TYPEMINSTAR:
2062 case OP_TYPEPLUS:
2063 case OP_TYPEMINPLUS:
2064 case OP_TYPEQUERY:
2065 case OP_TYPEMINQUERY:
2066 case OP_TYPEPOSSTAR:
2067 case OP_TYPEPOSPLUS:
2068 case OP_TYPEPOSQUERY:
2069 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2070 break;
2071
2072 case OP_TYPEPOSUPTO:
2073 case OP_TYPEUPTO:
2074 case OP_TYPEMINUPTO:
2075 case OP_TYPEEXACT:
2076 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2077 code += 2;
2078 break;
2079
2080 case OP_MARK:
2081 case OP_PRUNE_ARG:
2082 case OP_SKIP_ARG:
2083 case OP_THEN_ARG:
2084 code += code[1];
2085 break;
2086 }
2087
2088 /* Add in the fixed length from the table */
2089
2090 code += PRIV(OP_lengths)[c];
2091
2092 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2093 by a multi-byte character. The length in the table is a minimum, so we have
2094 to arrange to skip the extra bytes. */
2095
2096 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2097 if (utf) switch(c)
2098 {
2099 case OP_CHAR:
2100 case OP_CHARI:
2101 case OP_NOT:
2102 case OP_NOTI:
2103 case OP_EXACT:
2104 case OP_EXACTI:
2105 case OP_NOTEXACT:
2106 case OP_NOTEXACTI:
2107 case OP_UPTO:
2108 case OP_UPTOI:
2109 case OP_NOTUPTO:
2110 case OP_NOTUPTOI:
2111 case OP_MINUPTO:
2112 case OP_MINUPTOI:
2113 case OP_NOTMINUPTO:
2114 case OP_NOTMINUPTOI:
2115 case OP_POSUPTO:
2116 case OP_POSUPTOI:
2117 case OP_NOTPOSUPTO:
2118 case OP_NOTPOSUPTOI:
2119 case OP_STAR:
2120 case OP_STARI:
2121 case OP_NOTSTAR:
2122 case OP_NOTSTARI:
2123 case OP_MINSTAR:
2124 case OP_MINSTARI:
2125 case OP_NOTMINSTAR:
2126 case OP_NOTMINSTARI:
2127 case OP_POSSTAR:
2128 case OP_POSSTARI:
2129 case OP_NOTPOSSTAR:
2130 case OP_NOTPOSSTARI:
2131 case OP_PLUS:
2132 case OP_PLUSI:
2133 case OP_NOTPLUS:
2134 case OP_NOTPLUSI:
2135 case OP_MINPLUS:
2136 case OP_MINPLUSI:
2137 case OP_NOTMINPLUS:
2138 case OP_NOTMINPLUSI:
2139 case OP_POSPLUS:
2140 case OP_POSPLUSI:
2141 case OP_NOTPOSPLUS:
2142 case OP_NOTPOSPLUSI:
2143 case OP_QUERY:
2144 case OP_QUERYI:
2145 case OP_NOTQUERY:
2146 case OP_NOTQUERYI:
2147 case OP_MINQUERY:
2148 case OP_MINQUERYI:
2149 case OP_NOTMINQUERY:
2150 case OP_NOTMINQUERYI:
2151 case OP_POSQUERY:
2152 case OP_POSQUERYI:
2153 case OP_NOTPOSQUERY:
2154 case OP_NOTPOSQUERYI:
2155 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2156 break;
2157 }
2158 #else
2159 (void)(utf); /* Keep compiler happy by referencing function argument */
2160 #endif
2161 }
2162 }
2163 }
2164
2165
2166
2167 /*************************************************
2168 * Scan compiled branch for non-emptiness *
2169 *************************************************/
2170
2171 /* This function scans through a branch of a compiled pattern to see whether it
2172 can match the empty string or not. It is called from could_be_empty()
2173 below and from compile_branch() when checking for an unlimited repeat of a
2174 group that can match nothing. Note that first_significant_code() skips over
2175 backward and negative forward assertions when its final argument is TRUE. If we
2176 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2177 bracket whose current branch will already have been scanned.
2178
2179 Arguments:
2180 code points to start of search
2181 endcode points to where to stop
2182 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2183 cd contains pointers to tables etc.
2184 recurses chain of recurse_check to catch mutual recursion
2185
2186 Returns: TRUE if what is matched could be empty
2187 */
2188
2189 typedef struct recurse_check {
2190 struct recurse_check *prev;
2191 const pcre_uchar *group;
2192 } recurse_check;
2193
2194 static BOOL
2195 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2196 BOOL utf, compile_data *cd, recurse_check *recurses)
2197 {
2198 register pcre_uchar c;
2199 recurse_check this_recurse;
2200
2201 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2202 code < endcode;
2203 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2204 {
2205 const pcre_uchar *ccode;
2206
2207 c = *code;
2208
2209 /* Skip over forward assertions; the other assertions are skipped by
2210 first_significant_code() with a TRUE final argument. */
2211
2212 if (c == OP_ASSERT)
2213 {
2214 do code += GET(code, 1); while (*code == OP_ALT);
2215 c = *code;
2216 continue;
2217 }
2218
2219 /* For a recursion/subroutine call, if its end has been reached, which
2220 implies a backward reference subroutine call, we can scan it. If it's a
2221 forward reference subroutine call, we can't. To detect forward reference
2222 we have to scan up the list that is kept in the workspace. This function is
2223 called only when doing the real compile, not during the pre-compile that
2224 measures the size of the compiled pattern. */
2225
2226 if (c == OP_RECURSE)
2227 {
2228 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2229 BOOL empty_branch;
2230
2231 /* Test for forward reference or uncompleted reference. This is disabled
2232 when called to scan a completed pattern by setting cd->start_workspace to
2233 NULL. */
2234
2235 if (cd->start_workspace != NULL)
2236 {
2237 const pcre_uchar *tcode;
2238 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2239 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2240 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2241 }
2242
2243 /* If we are scanning a completed pattern, there are no forward references
2244 and all groups are complete. We need to detect whether this is a recursive
2245 call, as otherwise there will be an infinite loop. If it is a recursion,
2246 just skip over it. Simple recursions are easily detected. For mutual
2247 recursions we keep a chain on the stack. */
2248
2249 else
2250 {
2251 recurse_check *r = recurses;
2252 const pcre_uchar *endgroup = scode;
2253
2254 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2255 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2256
2257 for (r = recurses; r != NULL; r = r->prev)
2258 if (r->group == scode) break;
2259 if (r != NULL) continue; /* Mutual recursion */
2260 }
2261
2262 /* Completed reference; scan the referenced group, remembering it on the
2263 stack chain to detect mutual recursions. */
2264
2265 empty_branch = FALSE;
2266 this_recurse.prev = recurses;
2267 this_recurse.group = scode;
2268
2269 do
2270 {
2271 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2272 {
2273 empty_branch = TRUE;
2274 break;
2275 }
2276 scode += GET(scode, 1);
2277 }
2278 while (*scode == OP_ALT);
2279
2280 if (!empty_branch) return FALSE; /* All branches are non-empty */
2281 continue;
2282 }
2283
2284 /* Groups with zero repeats can of course be empty; skip them. */
2285
2286 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2287 c == OP_BRAPOSZERO)
2288 {
2289 code += PRIV(OP_lengths)[c];
2290 do code += GET(code, 1); while (*code == OP_ALT);
2291 c = *code;
2292 continue;
2293 }
2294
2295 /* A nested group that is already marked as "could be empty" can just be
2296 skipped. */
2297
2298 if (c == OP_SBRA || c == OP_SBRAPOS ||
2299 c == OP_SCBRA || c == OP_SCBRAPOS)
2300 {
2301 do code += GET(code, 1); while (*code == OP_ALT);
2302 c = *code;
2303 continue;
2304 }
2305
2306 /* For other groups, scan the branches. */
2307
2308 if (c == OP_BRA || c == OP_BRAPOS ||
2309 c == OP_CBRA || c == OP_CBRAPOS ||
2310 c == OP_ONCE || c == OP_ONCE_NC ||
2311 c == OP_COND)
2312 {
2313 BOOL empty_branch;
2314 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2315
2316 /* If a conditional group has only one branch, there is a second, implied,
2317 empty branch, so just skip over the conditional, because it could be empty.
2318 Otherwise, scan the individual branches of the group. */
2319
2320 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2321 code += GET(code, 1);
2322 else
2323 {
2324 empty_branch = FALSE;
2325 do
2326 {
2327 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2328 empty_branch = TRUE;
2329 code += GET(code, 1);
2330 }
2331 while (*code == OP_ALT);
2332 if (!empty_branch) return FALSE; /* All branches are non-empty */
2333 }
2334
2335 c = *code;
2336 continue;
2337 }
2338
2339 /* Handle the other opcodes */
2340
2341 switch (c)
2342 {
2343 /* Check for quantifiers after a class. XCLASS is used for classes that
2344 cannot be represented just by a bit map. This includes negated single
2345 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2346 actual length is stored in the compiled code, so we must update "code"
2347 here. */
2348
2349 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2350 case OP_XCLASS:
2351 ccode = code += GET(code, 1);
2352 goto CHECK_CLASS_REPEAT;
2353 #endif
2354
2355 case OP_CLASS:
2356 case OP_NCLASS:
2357 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2358
2359 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2360 CHECK_CLASS_REPEAT:
2361 #endif
2362
2363 switch (*ccode)
2364 {
2365 case OP_CRSTAR: /* These could be empty; continue */
2366 case OP_CRMINSTAR:
2367 case OP_CRQUERY:
2368 case OP_CRMINQUERY:
2369 break;
2370
2371 default: /* Non-repeat => class must match */
2372 case OP_CRPLUS: /* These repeats aren't empty */
2373 case OP_CRMINPLUS:
2374 return FALSE;
2375
2376 case OP_CRRANGE:
2377 case OP_CRMINRANGE:
2378 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2379 break;
2380 }
2381 break;
2382
2383 /* Opcodes that must match a character */
2384
2385 case OP_ANY:
2386 case OP_ALLANY:
2387 case OP_ANYBYTE:
2388
2389 case OP_PROP:
2390 case OP_NOTPROP:
2391 case OP_ANYNL:
2392
2393 case OP_NOT_HSPACE:
2394 case OP_HSPACE:
2395 case OP_NOT_VSPACE:
2396 case OP_VSPACE:
2397 case OP_EXTUNI:
2398
2399 case OP_NOT_DIGIT:
2400 case OP_DIGIT:
2401 case OP_NOT_WHITESPACE:
2402 case OP_WHITESPACE:
2403 case OP_NOT_WORDCHAR:
2404 case OP_WORDCHAR:
2405
2406 case OP_CHAR:
2407 case OP_CHARI:
2408 case OP_NOT:
2409 case OP_NOTI:
2410
2411 case OP_PLUS:
2412 case OP_PLUSI:
2413 case OP_MINPLUS:
2414 case OP_MINPLUSI:
2415
2416 case OP_NOTPLUS:
2417 case OP_NOTPLUSI:
2418 case OP_NOTMINPLUS:
2419 case OP_NOTMINPLUSI:
2420
2421 case OP_POSPLUS:
2422 case OP_POSPLUSI:
2423 case OP_NOTPOSPLUS:
2424 case OP_NOTPOSPLUSI:
2425
2426 case OP_EXACT:
2427 case OP_EXACTI:
2428 case OP_NOTEXACT:
2429 case OP_NOTEXACTI:
2430
2431 case OP_TYPEPLUS:
2432 case OP_TYPEMINPLUS:
2433 case OP_TYPEPOSPLUS:
2434 case OP_TYPEEXACT:
2435
2436 return FALSE;
2437
2438 /* These are going to continue, as they may be empty, but we have to
2439 fudge the length for the \p and \P cases. */
2440
2441 case OP_TYPESTAR:
2442 case OP_TYPEMINSTAR:
2443 case OP_TYPEPOSSTAR:
2444 case OP_TYPEQUERY:
2445 case OP_TYPEMINQUERY:
2446 case OP_TYPEPOSQUERY:
2447 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2448 break;
2449
2450 /* Same for these */
2451
2452 case OP_TYPEUPTO:
2453 case OP_TYPEMINUPTO:
2454 case OP_TYPEPOSUPTO:
2455 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2456 code += 2;
2457 break;
2458
2459 /* End of branch */
2460
2461 case OP_KET:
2462 case OP_KETRMAX:
2463 case OP_KETRMIN:
2464 case OP_KETRPOS:
2465 case OP_ALT:
2466 return TRUE;
2467
2468 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2469 MINUPTO, and POSUPTO and their caseless and negative versions may be
2470 followed by a multibyte character. */
2471
2472 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2473 case OP_STAR:
2474 case OP_STARI:
2475 case OP_NOTSTAR:
2476 case OP_NOTSTARI:
2477
2478 case OP_MINSTAR:
2479 case OP_MINSTARI:
2480 case OP_NOTMINSTAR:
2481 case OP_NOTMINSTARI:
2482
2483 case OP_POSSTAR:
2484 case OP_POSSTARI:
2485 case OP_NOTPOSSTAR:
2486 case OP_NOTPOSSTARI:
2487
2488 case OP_QUERY:
2489 case OP_QUERYI:
2490 case OP_NOTQUERY:
2491 case OP_NOTQUERYI:
2492
2493 case OP_MINQUERY:
2494 case OP_MINQUERYI:
2495 case OP_NOTMINQUERY:
2496 case OP_NOTMINQUERYI:
2497
2498 case OP_POSQUERY:
2499 case OP_POSQUERYI:
2500 case OP_NOTPOSQUERY:
2501 case OP_NOTPOSQUERYI:
2502
2503 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2504 break;
2505
2506 case OP_UPTO:
2507 case OP_UPTOI:
2508 case OP_NOTUPTO:
2509 case OP_NOTUPTOI:
2510
2511 case OP_MINUPTO:
2512 case OP_MINUPTOI:
2513 case OP_NOTMINUPTO:
2514 case OP_NOTMINUPTOI:
2515
2516 case OP_POSUPTO:
2517 case OP_POSUPTOI:
2518 case OP_NOTPOSUPTO:
2519 case OP_NOTPOSUPTOI:
2520
2521 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2522 break;
2523 #endif
2524
2525 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2526 string. */
2527
2528 case OP_MARK:
2529 case OP_PRUNE_ARG:
2530 case OP_SKIP_ARG:
2531 case OP_THEN_ARG:
2532 code += code[1];
2533 break;
2534
2535 /* None of the remaining opcodes are required to match a character. */
2536
2537 default:
2538 break;
2539 }
2540 }
2541
2542 return TRUE;
2543 }
2544
2545
2546
2547 /*************************************************
2548 * Scan compiled regex for non-emptiness *
2549 *************************************************/
2550
2551 /* This function is called to check for left recursive calls. We want to check
2552 the current branch of the current pattern to see if it could match the empty
2553 string. If it could, we must look outwards for branches at other levels,
2554 stopping when we pass beyond the bracket which is the subject of the recursion.
2555 This function is called only during the real compile, not during the
2556 pre-compile.
2557
2558 Arguments:
2559 code points to start of the recursion
2560 endcode points to where to stop (current RECURSE item)
2561 bcptr points to the chain of current (unclosed) branch starts
2562 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2563 cd pointers to tables etc
2564
2565 Returns: TRUE if what is matched could be empty
2566 */
2567
2568 static BOOL
2569 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2570 branch_chain *bcptr, BOOL utf, compile_data *cd)
2571 {
2572 while (bcptr != NULL && bcptr->current_branch >= code)
2573 {
2574 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2575 return FALSE;
2576 bcptr = bcptr->outer;
2577 }
2578 return TRUE;
2579 }
2580
2581
2582
2583 /*************************************************
2584 * Base opcode of repeated opcodes *
2585 *************************************************/
2586
2587 /* Returns the base opcode for repeated single character type opcodes. If the
2588 opcode is not a repeated character type, it returns with the original value.
2589
2590 Arguments: c opcode
2591 Returns: base opcode for the type
2592 */
2593
2594 static pcre_uchar
2595 get_repeat_base(pcre_uchar c)
2596 {
2597 return (c > OP_TYPEPOSUPTO)? c :
2598 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2599 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2600 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2601 (c >= OP_STARI)? OP_STARI :
2602 OP_STAR;
2603 }
2604
2605
2606
2607 #ifdef SUPPORT_UCP
2608 /*************************************************
2609 * Check a character and a property *
2610 *************************************************/
2611
2612 /* This function is called by check_auto_possessive() when a property item
2613 is adjacent to a fixed character.
2614
2615 Arguments:
2616 c the character
2617 ptype the property type
2618 pdata the data for the type
2619 negated TRUE if it's a negated property (\P or \p{^)
2620
2621 Returns: TRUE if auto-possessifying is OK
2622 */
2623
2624 static BOOL
2625 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2626 BOOL negated)
2627 {
2628 const pcre_uint32 *p;
2629 const ucd_record *prop = GET_UCD(c);
2630
2631 switch(ptype)
2632 {
2633 case PT_LAMP:
2634 return (prop->chartype == ucp_Lu ||
2635 prop->chartype == ucp_Ll ||
2636 prop->chartype == ucp_Lt) == negated;
2637
2638 case PT_GC:
2639 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2640
2641 case PT_PC:
2642 return (pdata == prop->chartype) == negated;
2643
2644 case PT_SC:
2645 return (pdata == prop->script) == negated;
2646
2647 /* These are specials */
2648
2649 case PT_ALNUM:
2650 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2651 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2652
2653 case PT_SPACE: /* Perl space */
2654 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2655 c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2656 == negated;
2657
2658 case PT_PXSPACE: /* POSIX space */
2659 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z ||
2660 c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2661 c == CHAR_FF || c == CHAR_CR)
2662 == negated;
2663
2664 case PT_WORD:
2665 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2666 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2667 c == CHAR_UNDERSCORE) == negated;
2668
2669 case PT_CLIST:
2670 p = PRIV(ucd_caseless_sets) + prop->caseset;
2671 for (;;)
2672 {
2673 if (c < *p) return !negated;
2674 if (c == *p++) return negated;
2675 }
2676 break; /* Control never reaches here */
2677 }
2678
2679 return FALSE;
2680 }
2681 #endif /* SUPPORT_UCP */
2682
2683
2684
2685 /*************************************************
2686 * Fill the character property list *
2687 *************************************************/
2688
2689 /* Checks whether the code points to an opcode that can take part in auto-
2690 possessification, and if so, fills a list with its properties.
2691
2692 Arguments:
2693 code points to start of expression
2694 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2695 fcc points to case-flipping table
2696 list points to output list
2697 list[0] will be filled with the opcode
2698 list[1] will be non-zero if this opcode
2699 can match an empty character string
2700 list[2..7] depends on the opcode
2701
2702 Returns: points to the start of the next opcode if *code is accepted
2703 NULL if *code is not accepted
2704 */
2705
2706 static const pcre_uchar *
2707 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2708 const pcre_uint8 *fcc, pcre_uint32 *list)
2709 {
2710 pcre_uchar c = *code;
2711 const pcre_uchar *end;
2712 const pcre_uint32 *clist_src;
2713 pcre_uint32 *clist_dest;
2714 pcre_uint32 chr;
2715 pcre_uchar base;
2716
2717 list[0] = c;
2718 list[1] = FALSE;
2719 code++;
2720
2721 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2722 {
2723 base = get_repeat_base(c);
2724 c -= (base - OP_STAR);
2725
2726 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2727 code += IMM2_SIZE;
2728
2729 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2730
2731 switch(base)
2732 {
2733 case OP_STAR:
2734 list[0] = OP_CHAR;
2735 break;
2736
2737 case OP_STARI:
2738 list[0] = OP_CHARI;
2739 break;
2740
2741 case OP_NOTSTAR:
2742 list[0] = OP_NOT;
2743 break;
2744
2745 case OP_NOTSTARI:
2746 list[0] = OP_NOTI;
2747 break;
2748
2749 case OP_TYPESTAR:
2750 list[0] = *code;
2751 code++;
2752 break;
2753 }
2754 c = list[0];
2755 }
2756
2757 switch(c)
2758 {
2759 case OP_NOT_DIGIT:
2760 case OP_DIGIT:
2761 case OP_NOT_WHITESPACE:
2762 case OP_WHITESPACE:
2763 case OP_NOT_WORDCHAR:
2764 case OP_WORDCHAR:
2765 case OP_ANY:
2766 case OP_ALLANY:
2767 case OP_ANYNL:
2768 case OP_NOT_HSPACE:
2769 case OP_HSPACE:
2770 case OP_NOT_VSPACE:
2771 case OP_VSPACE:
2772 case OP_EXTUNI:
2773 case OP_EODN:
2774 case OP_EOD:
2775 case OP_DOLL:
2776 case OP_DOLLM:
2777 return code;
2778
2779 case OP_CHAR:
2780 case OP_NOT:
2781 GETCHARINCTEST(chr, code);
2782 list[2] = chr;
2783 list[3] = NOTACHAR;
2784 return code;
2785
2786 case OP_CHARI:
2787 case OP_NOTI:
2788 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2789 GETCHARINCTEST(chr, code);
2790 list[2] = chr;
2791
2792 #ifdef SUPPORT_UCP
2793 if (chr < 128 || (chr < 256 && !utf))
2794 list[3] = fcc[chr];
2795 else
2796 list[3] = UCD_OTHERCASE(chr);
2797 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2798 list[3] = (chr < 256) ? fcc[chr] : chr;
2799 #else
2800 list[3] = fcc[chr];
2801 #endif
2802
2803 /* The othercase might be the same value. */
2804
2805 if (chr == list[3])
2806 list[3] = NOTACHAR;
2807 else
2808 list[4] = NOTACHAR;
2809 return code;
2810
2811 #ifdef SUPPORT_UCP
2812 case OP_PROP:
2813 case OP_NOTPROP:
2814 if (code[0] != PT_CLIST)
2815 {
2816 list[2] = code[0];
2817 list[3] = code[1];
2818 return code + 2;
2819 }
2820
2821 /* Convert only if we have anough space. */
2822
2823 clist_src = PRIV(ucd_caseless_sets) + code[1];
2824 clist_dest = list + 2;
2825 code += 2;
2826
2827 do {
2828 /* Early return if there is not enough space. */
2829 if (clist_dest >= list + 8)
2830 {
2831 list[2] = code[0];
2832 list[3] = code[1];
2833 return code;
2834 }
2835 *clist_dest++ = *clist_src;
2836 }
2837 while(*clist_src++ != NOTACHAR);
2838
2839 /* Enough space to store all characters. */
2840
2841 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2842 return code;
2843 #endif
2844
2845 case OP_NCLASS:
2846 case OP_CLASS:
2847 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2848 case OP_XCLASS:
2849
2850 if (c == OP_XCLASS)
2851 end = code + GET(code, 0);
2852 else
2853 #endif
2854 end = code + 32 / sizeof(pcre_uchar);
2855
2856 switch(*end)
2857 {
2858 case OP_CRSTAR:
2859 case OP_CRMINSTAR:
2860 case OP_CRQUERY:
2861 case OP_CRMINQUERY:
2862 list[1] = TRUE;
2863 end++;
2864 break;
2865
2866 case OP_CRRANGE:
2867 case OP_CRMINRANGE:
2868 list[1] = (GET2(end, 1) == 0);
2869 end += 1 + 2 * IMM2_SIZE;
2870 break;
2871 }
2872 list[2] = end - code;
2873 return end;
2874 }
2875 return NULL; /* Opcode not accepted */
2876 }
2877
2878
2879
2880 /*************************************************
2881 * Scan further character sets for match *
2882 *************************************************/
2883
2884 /* Checks whether the base and the current opcode have a common character, in
2885 which case the base cannot be possessified.
2886
2887 Arguments:
2888 code points to the byte code
2889 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2890 cd static compile data
2891 base_list the data list of the base opcode
2892
2893 Returns: TRUE if the auto-possessification is possible
2894 */
2895
2896 static BOOL
2897 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
2898 const pcre_uint32* base_list)
2899 {
2900 pcre_uchar c;
2901 pcre_uint32 list[8];
2902 const pcre_uint32* chr_ptr;
2903 const pcre_uint32* ochr_ptr;
2904 const pcre_uint32* list_ptr;
2905 pcre_uint32 chr;
2906
2907 for(;;)
2908 {
2909 c = *code;
2910
2911 /* Skip over callouts */
2912
2913 if (c == OP_CALLOUT)
2914 {
2915 code += PRIV(OP_lengths)[c];
2916 continue;
2917 }
2918
2919 if (c == OP_ALT)
2920 {
2921 do code += GET(code, 1); while (*code == OP_ALT);
2922 c = *code;
2923 }
2924
2925 switch(c)
2926 {
2927 case OP_END:
2928 /* TRUE only in greedy case. The non-greedy case could be replaced by an
2929 OP_EXACT, but it is probably not worth it. (And note that OP_EXACT uses
2930 more memory, which we cannot get at this stage.) */
2931
2932 return base_list[1] != 0;
2933
2934 case OP_KET:
2935 /* If the bracket is capturing, and referenced by an OP_RECURSE, the
2936 non-greedy case cannot be converted to a possessive form. We do not test
2937 the bracket type at the moment, but we might do it in the future to improve
2938 this condition. (But note that recursive calls are always atomic.) */
2939
2940 if (base_list[1] == 0) return FALSE;
2941 code += PRIV(OP_lengths)[c];
2942 continue;
2943 }
2944
2945 /* Check for a supported opcode, and load its properties. */
2946
2947 code = get_chr_property_list(code, utf, cd->fcc, list);
2948 if (code == NULL) return FALSE; /* Unsupported */
2949
2950 /* If either opcode is a small character list, set pointers for comparing
2951 characters from that list with another list, or with a property. */
2952
2953 if (base_list[0] == OP_CHAR)
2954 {
2955 chr_ptr = base_list + 2;
2956 list_ptr = list;
2957 }
2958 else if (list[0] == OP_CHAR)
2959 {
2960 chr_ptr = list + 2;
2961 list_ptr = base_list;
2962 }
2963
2964 /* Some property combinations also acceptable. Unicode property opcodes are
2965 processed specially; the rest can be handled with a lookup table. */
2966
2967 else
2968 {
2969 pcre_uint32 leftop, rightop;
2970
2971 if (list[1] != 0) return FALSE; /* Must match at least one character */
2972 leftop = base_list[0];
2973 rightop = list[0];
2974
2975 #ifdef SUPPORT_UCP
2976 if (leftop == OP_PROP || leftop == OP_NOTPROP)
2977 {
2978 if (rightop == OP_EOD) return TRUE;
2979 if (rightop == OP_PROP || rightop == OP_NOTPROP)
2980 {
2981 int n;
2982 const pcre_uint8 *p;
2983 BOOL same = leftop == rightop;
2984 BOOL lisprop = leftop == OP_PROP;
2985 BOOL risprop = rightop == OP_PROP;
2986 BOOL bothprop = lisprop && risprop;
2987
2988 /* There's a table that specifies how each combination is to be
2989 processed:
2990 0 Always return FALSE (never auto-possessify)
2991 1 Character groups are distinct (possessify if both are OP_PROP)
2992 2 Check character categories in the same group (general or particular)
2993 3 Return TRUE if the two opcodes are not the same
2994 ... see comments below
2995 */
2996
2997 n = propposstab[base_list[2]][list[2]];
2998 switch(n)
2999 {
3000 case 0: return FALSE;
3001 case 1: return bothprop;
3002 case 2: return (base_list[3] == list[3]) != same;
3003 case 3: return !same;
3004
3005 case 4: /* Left general category, right particular category */
3006 return risprop && catposstab[base_list[3]][list[3]] == same;
3007
3008 case 5: /* Right general category, left particular category */
3009 return lisprop && catposstab[list[3]][base_list[3]] == same;
3010
3011 /* This code is logically tricky. Think hard before fiddling with it.
3012 The posspropstab table has four entries per row. Each row relates to
3013 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3014 Only WORD actually needs all four entries, but using repeats for the
3015 others means they can all use the same code below.
3016
3017 The first two entries in each row are Unicode general categories, and
3018 apply always, because all the characters they include are part of the
3019 PCRE character set. The third and fourth entries are a general and a
3020 particular category, respectively, that include one or more relevant
3021 characters. One or the other is used, depending on whether the check
3022 is for a general or a particular category. However, in both cases the
3023 category contains more characters than the specials that are defined
3024 for the property being tested against. Therefore, it cannot be used
3025 in a NOTPROP case.
3026
3027 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3028 Underscore is covered by ucp_P or ucp_Po. */
3029
3030 case 6: /* Left alphanum vs right general category */
3031 case 7: /* Left space vs right general category */
3032 case 8: /* Left word vs right general category */
3033 p = posspropstab[n-6];
3034 return risprop && lisprop ==
3035 (list[3] != p[0] &&
3036 list[3] != p[1] &&
3037 (list[3] != p[2] || !lisprop));
3038
3039 case 9: /* Right alphanum vs left general category */
3040 case 10: /* Right space vs left general category */
3041 case 11: /* Right word vs left general category */
3042 p = posspropstab[n-9];
3043 return lisprop && risprop ==
3044 (base_list[3] != p[0] &&
3045 base_list[3] != p[1] &&
3046 (base_list[3] != p[2] || !risprop));
3047
3048 case 12: /* Left alphanum vs right particular category */
3049 case 13: /* Left space vs right particular category */
3050 case 14: /* Left word vs right particular category */
3051 p = posspropstab[n-12];
3052 return risprop && lisprop ==
3053 (catposstab[p[0]][list[3]] &&
3054 catposstab[p[1]][list[3]] &&
3055 (list[3] != p[3] || !lisprop));
3056
3057 case 15: /* Right alphanum vs left particular category */
3058 case 16: /* Right space vs left particular category */
3059 case 17: /* Right word vs left particular category */
3060 p = posspropstab[n-15];
3061 return lisprop && risprop ==
3062 (catposstab[p[0]][base_list[3]] &&
3063 catposstab[p[1]][base_list[3]] &&
3064 (base_list[3] != p[3] || !risprop));
3065 }
3066 }
3067 return FALSE;
3068 }
3069
3070 else
3071 #endif /* SUPPORT_UCP */
3072
3073 return leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3074 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3075 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3076 }
3077
3078 /* Control reaches here only if one of the items is a small character list.
3079 All characters are checked against the other side. */
3080
3081 do
3082 {
3083 chr = *chr_ptr;
3084
3085 switch(list_ptr[0])
3086 {
3087 case OP_CHAR:
3088 ochr_ptr = list_ptr + 2;
3089 do
3090 {
3091 if (chr == *ochr_ptr) return FALSE;
3092 ochr_ptr++;
3093 }
3094 while(*ochr_ptr != NOTACHAR);
3095 break;
3096
3097 case OP_NOT:
3098 ochr_ptr = list_ptr + 2;
3099 do
3100 {
3101 if (chr == *ochr_ptr)
3102 break;
3103 ochr_ptr++;
3104 }
3105 while(*ochr_ptr != NOTACHAR);
3106 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3107 break;
3108
3109 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3110 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3111
3112 case OP_DIGIT:
3113 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3114 break;
3115
3116 case OP_NOT_DIGIT:
3117 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3118 break;
3119
3120 case OP_WHITESPACE:
3121 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3122 break;
3123
3124 case OP_NOT_WHITESPACE:
3125 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3126 break;
3127
3128 case OP_WORDCHAR:
3129 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3130 break;
3131
3132 case OP_NOT_WORDCHAR:
3133 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3134 break;
3135
3136 case OP_HSPACE:
3137 switch(chr)
3138 {
3139 HSPACE_CASES: return FALSE;
3140 default: break;
3141 }
3142 break;
3143
3144 case OP_NOT_HSPACE:
3145 switch(chr)
3146 {
3147 HSPACE_CASES: break;
3148 default: return FALSE;
3149 }
3150 break;
3151
3152 case OP_ANYNL:
3153 case OP_VSPACE:
3154 switch(chr)
3155 {
3156 VSPACE_CASES: return FALSE;
3157 default: break;
3158 }
3159 break;
3160
3161 case OP_NOT_VSPACE:
3162 switch(chr)
3163 {
3164 VSPACE_CASES: break;
3165 default: return FALSE;
3166 }
3167 break;
3168
3169 case OP_DOLL:
3170 case OP_EODN:
3171 switch (chr)
3172 {
3173 case CHAR_CR:
3174 case CHAR_LF:
3175 case CHAR_VT:
3176 case CHAR_FF:
3177 case CHAR_NEL:
3178 #ifndef EBCDIC
3179 case 0x2028:
3180 case 0x2029:
3181 #endif /* Not EBCDIC */
3182 return FALSE;
3183 }
3184 break;
3185
3186 case OP_EOD: /* Can always possessify before \z */
3187 break;
3188
3189 case OP_PROP:
3190 case OP_NOTPROP:
3191 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3192 list_ptr[0] == OP_NOTPROP))
3193 return FALSE;
3194 break;
3195
3196 /* The class comparisons work only when the class is the second item
3197 of the pair, because there are at present no possessive forms of the
3198 class opcodes. Note also that the "code" variable that is used below
3199 points after the second item, and that the pointer for the first item
3200 is not available, so even if there were possessive forms of the class
3201 opcodes, the correct comparison could not be done. */
3202
3203 case OP_NCLASS:
3204 if (chr > 255) return FALSE;
3205 /* Fall through */
3206
3207 case OP_CLASS:
3208 if (list_ptr != list) return FALSE; /* Class is first opcode */
3209 if (chr > 255) break;
3210 if ((((pcre_uint8 *)(code - list_ptr[2] + 1))[chr >> 3] & (1 << (chr & 7))) != 0)
3211 return FALSE;
3212 break;
3213
3214 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3215 case OP_XCLASS:
3216 if (list_ptr != list) return FALSE; /* Class is first opcode */
3217 if (PRIV(xclass)(chr, code - list_ptr[2] + 1 + LINK_SIZE, utf))
3218 return FALSE;
3219 break;
3220 #endif
3221
3222 default:
3223 return FALSE;
3224 }
3225
3226 chr_ptr++;
3227 }
3228 while(*chr_ptr != NOTACHAR);
3229
3230 /* At least one character must be matched from this opcode. */
3231
3232 if (list[1] == 0) return TRUE;
3233 }
3234
3235 return FALSE;
3236 }
3237
3238
3239
3240 /*************************************************
3241 * Scan compiled regex for auto-possession *
3242 *************************************************/
3243
3244 /* Replaces single character iterations with their possessive alternatives
3245 if appropriate. This function modifies the compiled opcode!
3246
3247 Arguments:
3248 code points to start of the byte code
3249 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3250 cd static compile data
3251
3252 Returns: nothing
3253 */
3254
3255 static void
3256 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3257 {
3258 register pcre_uchar c;
3259 const pcre_uchar *end;
3260 pcre_uint32 list[8];
3261
3262 for (;;)
3263 {
3264 c = *code;
3265
3266 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3267 {
3268 c -= get_repeat_base(c) - OP_STAR;
3269 end = (c <= OP_MINUPTO) ?
3270 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3271 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3272
3273 if (end != NULL && compare_opcodes(end, utf, cd, list))
3274 {
3275 switch(c)
3276 {
3277 case OP_STAR:
3278 *code += OP_POSSTAR - OP_STAR;
3279 break;
3280
3281 case OP_MINSTAR:
3282 *code += OP_POSSTAR - OP_MINSTAR;
3283 break;
3284
3285 case OP_PLUS:
3286 *code += OP_POSPLUS - OP_PLUS;
3287 break;
3288
3289 case OP_MINPLUS:
3290 *code += OP_POSPLUS - OP_MINPLUS;
3291 break;
3292
3293 case OP_QUERY:
3294 *code += OP_POSQUERY - OP_QUERY;
3295 break;
3296
3297 case OP_MINQUERY:
3298 *code += OP_POSQUERY - OP_MINQUERY;
3299 break;
3300
3301 case OP_UPTO:
3302 *code += OP_POSUPTO - OP_UPTO;
3303 break;
3304
3305 case OP_MINUPTO:
3306 *code += OP_MINUPTO - OP_UPTO;
3307 break;
3308 }
3309 }
3310 c = *code;
3311 }
3312
3313 switch(c)
3314 {
3315 case OP_END:
3316 return;
3317
3318 case OP_TYPESTAR:
3319 case OP_TYPEMINSTAR:
3320 case OP_TYPEPLUS:
3321 case OP_TYPEMINPLUS:
3322 case OP_TYPEQUERY:
3323 case OP_TYPEMINQUERY:
3324 case OP_TYPEPOSSTAR:
3325 case OP_TYPEPOSPLUS:
3326 case OP_TYPEPOSQUERY:
3327 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3328 break;
3329
3330 case OP_TYPEUPTO:
3331 case OP_TYPEMINUPTO:
3332 case OP_TYPEEXACT:
3333 case OP_TYPEPOSUPTO:
3334 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3335 code += 2;
3336 break;
3337
3338 case OP_XCLASS:
3339 code += GET(code, 1);
3340 break;
3341
3342 case OP_MARK:
3343 case OP_PRUNE_ARG:
3344 case OP_SKIP_ARG:
3345 case OP_THEN_ARG:
3346 code += code[1];
3347 break;
3348 }
3349
3350 /* Add in the fixed length from the table */
3351
3352 code += PRIV(OP_lengths)[c];
3353
3354 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3355 a multi-byte character. The length in the table is a minimum, so we have to
3356 arrange to skip the extra bytes. */
3357
3358 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3359 if (utf) switch(c)
3360 {
3361 case OP_CHAR:
3362 case OP_CHARI:
3363 case OP_NOT:
3364 case OP_NOTI:
3365 case OP_STAR:
3366 case OP_MINSTAR:
3367 case OP_PLUS:
3368 case OP_MINPLUS:
3369 case OP_QUERY:
3370 case OP_MINQUERY:
3371 case OP_UPTO:
3372 case OP_MINUPTO:
3373 case OP_EXACT:
3374 case OP_POSSTAR:
3375 case OP_POSPLUS:
3376 case OP_POSQUERY:
3377 case OP_POSUPTO:
3378 case OP_STARI:
3379 case OP_MINSTARI:
3380 case OP_PLUSI:
3381 case OP_MINPLUSI:
3382 case OP_QUERYI:
3383 case OP_MINQUERYI:
3384 case OP_UPTOI:
3385 case OP_MINUPTOI:
3386 case OP_EXACTI:
3387 case OP_POSSTARI:
3388 case OP_POSPLUSI:
3389 case OP_POSQUERYI:
3390 case OP_POSUPTOI:
3391 case OP_NOTSTAR:
3392 case OP_NOTMINSTAR:
3393 case OP_NOTPLUS:
3394 case OP_NOTMINPLUS:
3395 case OP_NOTQUERY:
3396 case OP_NOTMINQUERY:
3397 case OP_NOTUPTO:
3398 case OP_NOTMINUPTO:
3399 case OP_NOTEXACT:
3400 case OP_NOTPOSSTAR:
3401 case OP_NOTPOSPLUS:
3402 case OP_NOTPOSQUERY:
3403 case OP_NOTPOSUPTO:
3404 case OP_NOTSTARI:
3405 case OP_NOTMINSTARI:
3406 case OP_NOTPLUSI:
3407 case OP_NOTMINPLUSI:
3408 case OP_NOTQUERYI:
3409 case OP_NOTMINQUERYI:
3410 case OP_NOTUPTOI:
3411 case OP_NOTMINUPTOI:
3412 case OP_NOTEXACTI:
3413 case OP_NOTPOSSTARI:
3414 case OP_NOTPOSPLUSI:
3415 case OP_NOTPOSQUERYI:
3416 case OP_NOTPOSUPTOI:
3417 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3418 break;
3419 }
3420 #else
3421 (void)(utf); /* Keep compiler happy by referencing function argument */
3422 #endif
3423 }
3424 }
3425
3426
3427
3428 /*************************************************
3429 * Check for POSIX class syntax *
3430 *************************************************/
3431
3432 /* This function is called when the sequence "[:" or "[." or "[=" is
3433 encountered in a character class. It checks whether this is followed by a
3434 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3435 reach an unescaped ']' without the special preceding character, return FALSE.
3436
3437 Originally, this function only recognized a sequence of letters between the
3438 terminators, but it seems that Perl recognizes any sequence of characters,
3439 though of course unknown POSIX names are subsequently rejected. Perl gives an
3440 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3441 didn't consider this to be a POSIX class. Likewise for [:1234:].
3442
3443 The problem in trying to be exactly like Perl is in the handling of escapes. We
3444 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3445 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3446 below handles the special case of \], but does not try to do any other escape
3447 processing. This makes it different from Perl for cases such as [:l\ower:]
3448 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3449 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3450 I think.
3451
3452 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3453 It seems that the appearance of a nested POSIX class supersedes an apparent
3454 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3455 a digit.
3456
3457 In Perl, unescaped square brackets may also appear as part of class names. For
3458 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3459 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3460 seem right at all. PCRE does not allow closing square brackets in POSIX class
3461 names.
3462
3463 Arguments:
3464 ptr pointer to the initial [
3465 endptr where to return the end pointer
3466
3467 Returns: TRUE or FALSE
3468 */
3469
3470 static BOOL
3471 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3472 {
3473 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3474 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3475 for (++ptr; *ptr != CHAR_NULL; ptr++)
3476 {
3477 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3478 ptr++;
3479 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3480 else
3481 {
3482 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3483 {
3484 *endptr = ptr;
3485 return TRUE;
3486 }
3487 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3488 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3489 ptr[1] == CHAR_EQUALS_SIGN) &&
3490 check_posix_syntax(ptr, endptr))
3491 return FALSE;
3492 }
3493 }
3494 return FALSE;
3495 }
3496
3497
3498
3499
3500 /*************************************************
3501 * Check POSIX class name *
3502 *************************************************/
3503
3504 /* This function is called to check the name given in a POSIX-style class entry
3505 such as [:alnum:].
3506
3507 Arguments:
3508 ptr points to the first letter
3509 len the length of the name
3510
3511 Returns: a value representing the name, or -1 if unknown
3512 */
3513
3514 static int
3515 check_posix_name(const pcre_uchar *ptr, int len)
3516 {
3517 const char *pn = posix_names;
3518 register int yield = 0;
3519 while (posix_name_lengths[yield] != 0)
3520 {
3521 if (len == posix_name_lengths[yield] &&
3522 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3523 pn += posix_name_lengths[yield] + 1;
3524 yield++;
3525 }
3526 return -1;
3527 }
3528
3529
3530 /*************************************************
3531 * Adjust OP_RECURSE items in repeated group *
3532 *************************************************/
3533
3534 /* OP_RECURSE items contain an offset from the start of the regex to the group
3535 that is referenced. This means that groups can be replicated for fixed
3536 repetition simply by copying (because the recursion is allowed to refer to
3537 earlier groups that are outside the current group). However, when a group is
3538 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3539 inserted before it, after it has been compiled. This means that any OP_RECURSE
3540 items within it that refer to the group itself or any contained groups have to
3541 have their offsets adjusted. That one of the jobs of this function. Before it
3542 is called, the partially compiled regex must be temporarily terminated with
3543 OP_END.
3544
3545 This function has been extended with the possibility of forward references for
3546 recursions and subroutine calls. It must also check the list of such references
3547 for the group we are dealing with. If it finds that one of the recursions in
3548 the current group is on this list, it adjusts the offset in the list, not the
3549 value in the reference (which is a group number).
3550
3551 Arguments:
3552 group points to the start of the group
3553 adjust the amount by which the group is to be moved
3554 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3555 cd contains pointers to tables etc.
3556 save_hwm the hwm forward reference pointer at the start of the group
3557
3558 Returns: nothing
3559 */
3560
3561 static void
3562 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3563 pcre_uchar *save_hwm)
3564 {
3565 pcre_uchar *ptr = group;
3566
3567 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3568 {
3569 int offset;
3570 pcre_uchar *hc;
3571
3572 /* See if this recursion is on the forward reference list. If so, adjust the
3573 reference. */
3574
3575 for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3576 {
3577 offset = (int)GET(hc, 0);
3578 if (cd->start_code + offset == ptr + 1)
3579 {
3580 PUT(hc, 0, offset + adjust);
3581 break;
3582 }
3583 }
3584
3585 /* Otherwise, adjust the recursion offset if it's after the start of this
3586 group. */
3587
3588 if (hc >= cd->hwm)
3589 {
3590 offset = (int)GET(ptr, 1);
3591 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3592 }
3593
3594 ptr += 1 + LINK_SIZE;
3595 }
3596 }
3597
3598
3599
3600 /*************************************************
3601 * Insert an automatic callout point *
3602 *************************************************/
3603
3604 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3605 callout points before each pattern item.
3606
3607 Arguments:
3608 code current code pointer
3609 ptr current pattern pointer
3610 cd pointers to tables etc
3611
3612 Returns: new code pointer
3613 */
3614
3615 static pcre_uchar *
3616 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3617 {
3618 *code++ = OP_CALLOUT;
3619 *code++ = 255;
3620 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
3621 PUT(code, LINK_SIZE, 0); /* Default length */
3622 return code + 2 * LINK_SIZE;
3623 }
3624
3625
3626
3627 /*************************************************
3628 * Complete a callout item *
3629 *************************************************/
3630
3631 /* A callout item contains the length of the next item in the pattern, which
3632 we can't fill in till after we have reached the relevant point. This is used
3633 for both automatic and manual callouts.
3634
3635 Arguments:
3636 previous_callout points to previous callout item
3637 ptr current pattern pointer
3638 cd pointers to tables etc
3639
3640 Returns: nothing
3641 */
3642
3643 static void
3644 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3645 {
3646 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3647 PUT(previous_callout, 2 + LINK_SIZE, length);
3648 }
3649
3650
3651
3652 #ifdef SUPPORT_UCP
3653 /*************************************************
3654 * Get othercase range *
3655 *************************************************/
3656
3657 /* This function is passed the start and end of a class range, in UTF-8 mode
3658 with UCP support. It searches up the characters, looking for ranges of
3659 characters in the "other" case. Each call returns the next one, updating the
3660 start address. A character with multiple other cases is returned on its own
3661 with a special return value.
3662
3663 Arguments:
3664 cptr points to starting character value; updated
3665 d end value
3666 ocptr where to put start of othercase range
3667 odptr where to put end of othercase range
3668
3669 Yield: -1 when no more
3670 0 when a range is returned
3671 >0 the CASESET offset for char with multiple other cases
3672 in this case, ocptr contains the original
3673 */
3674
3675 static int
3676 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
3677 pcre_uint32 *odptr)
3678 {
3679 pcre_uint32 c, othercase, next;
3680 unsigned int co;
3681
3682 /* Find the first character that has an other case. If it has multiple other
3683 cases, return its case offset value. */
3684
3685 for (c = *cptr; c <= d; c++)
3686 {
3687 if ((co = UCD_CASESET(c)) != 0)
3688 {
3689 *ocptr = c++; /* Character that has the set */
3690 *cptr = c; /* Rest of input range */
3691 return (int)co;
3692 }
3693 if ((othercase = UCD_OTHERCASE(c)) != c) break;
3694 }
3695
3696 if (c > d) return -1; /* Reached end of range */
3697
3698 *ocptr = othercase;
3699 next = othercase + 1;
3700
3701 for (++c; c <= d; c++)
3702 {
3703 if (UCD_OTHERCASE(c) != next) break;
3704 next++;
3705 }
3706
3707 *odptr = next - 1; /* End of othercase range */
3708 *cptr = c; /* Rest of input range */
3709 return 0;
3710 }
3711 #endif /* SUPPORT_UCP */
3712
3713
3714
3715 /*************************************************
3716 * Add a character or range to a class *
3717 *************************************************/
3718
3719 /* This function packages up the logic of adding a character or range of
3720 characters to a class. The character values in the arguments will be within the
3721 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
3722 mutually recursive with the function immediately below.
3723
3724 Arguments:
3725 classbits the bit map for characters < 256
3726 uchardptr points to the pointer for extra data
3727 options the options word
3728 cd contains pointers to tables etc.
3729 start start of range character
3730 end end of range character
3731
3732 Returns: the number of < 256 characters added
3733 the pointer to extra data is updated
3734 */
3735
3736 static int
3737 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3738 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
3739 {
3740 pcre_uint32 c;
3741 int n8 = 0;
3742
3743 /* If caseless matching is required, scan the range and process alternate
3744 cases. In Unicode, there are 8-bit characters that have alternate cases that
3745 are greater than 255 and vice-versa. Sometimes we can just extend the original
3746 range. */
3747
3748 if ((options & PCRE_CASELESS) != 0)
3749 {
3750 #ifdef SUPPORT_UCP
3751 if ((options & PCRE_UTF8) != 0)
3752 {
3753 int rc;
3754 pcre_uint32 oc, od;
3755
3756 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
3757 c = start;
3758
3759 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
3760 {
3761 /* Handle a single character that has more than one other case. */
3762
3763 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
3764 PRIV(ucd_caseless_sets) + rc, oc);
3765
3766 /* Do nothing if the other case range is within the original range. */
3767
3768 else if (oc >= start && od <= end) continue;
3769
3770 /* Extend the original range if there is overlap, noting that if oc < c, we
3771 can't have od > end because a subrange is always shorter than the basic
3772 range. Otherwise, use a recursive call to add the additional range. */
3773
3774 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
3775 else if (od > end && oc <= end + 1) end = od; /* Extend upwards */
3776 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
3777 }
3778 }
3779 else
3780 #endif /* SUPPORT_UCP */
3781
3782 /* Not UTF-mode, or no UCP */
3783
3784 for (c = start; c <= end && c < 256; c++)
3785 {
3786 SETBIT(classbits, cd->fcc[c]);
3787 n8++;
3788 }
3789 }
3790
3791 /* Now handle the original range. Adjust the final value according to the bit
3792 length - this means that the same lists of (e.g.) horizontal spaces can be used
3793 in all cases. */
3794
3795 #if defined COMPILE_PCRE8
3796 #ifdef SUPPORT_UTF
3797 if ((options & PCRE_UTF8) == 0)
3798 #endif
3799 if (end > 0xff) end = 0xff;
3800
3801 #elif defined COMPILE_PCRE16
3802 #ifdef SUPPORT_UTF
3803 if ((options & PCRE_UTF16) == 0)
3804 #endif
3805 if (end > 0xffff) end = 0xffff;
3806
3807 #endif /* COMPILE_PCRE[8|16] */
3808
3809 /* If all characters are less than 256, use the bit map. Otherwise use extra
3810 data. */
3811
3812 if (end < 0x100)
3813 {
3814 for (c = start; c <= end; c++)
3815 {
3816 n8++;
3817 SETBIT(classbits, c);
3818 }
3819 }
3820
3821 else
3822 {
3823 pcre_uchar *uchardata = *uchardptr;
3824
3825 #ifdef SUPPORT_UTF
3826 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
3827 {
3828 if (start < end)
3829 {
3830 *uchardata++ = XCL_RANGE;
3831 uchardata += PRIV(ord2utf)(start, uchardata);
3832 uchardata += PRIV(ord2utf)(end, uchardata);
3833 }
3834 else if (start == end)
3835 {
3836 *uchardata++ = XCL_SINGLE;
3837 uchardata += PRIV(ord2utf)(start, uchardata);
3838 }
3839 }
3840 else
3841 #endif /* SUPPORT_UTF */
3842
3843 /* Without UTF support, character values are constrained by the bit length,
3844 and can only be > 256 for 16-bit and 32-bit libraries. */
3845
3846 #ifdef COMPILE_PCRE8
3847 {}
3848 #else
3849 if (start < end)
3850 {
3851 *uchardata++ = XCL_RANGE;
3852 *uchardata++ = start;
3853 *uchardata++ = end;
3854 }
3855 else if (start == end)
3856 {
3857 *uchardata++ = XCL_SINGLE;
3858 *uchardata++ = start;
3859 }
3860 #endif
3861
3862 *uchardptr = uchardata; /* Updata extra data pointer */
3863 }
3864
3865 return n8; /* Number of 8-bit characters */
3866 }
3867
3868
3869
3870
3871 /*************************************************
3872 * Add a list of characters to a class *
3873 *************************************************/
3874
3875 /* This function is used for adding a list of case-equivalent characters to a
3876 class, and also for adding a list of horizontal or vertical whitespace. If the
3877 list is in order (which it should be), ranges of characters are detected and
3878 handled appropriately. This function is mutually recursive with the function
3879 above.
3880
3881 Arguments:
3882 classbits the bit map for characters < 256
3883 uchardptr points to the pointer for extra data
3884 options the options word
3885 cd contains pointers to tables etc.
3886 p points to row of 32-bit values, terminated by NOTACHAR
3887 except character to omit; this is used when adding lists of
3888 case-equivalent characters to avoid including the one we
3889 already know about
3890
3891 Returns: the number of < 256 characters added
3892 the pointer to extra data is updated
3893 */
3894
3895 static int
3896 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
3897 compile_data *cd, const pcre_uint32 *p, unsigned int except)
3898 {
3899 int n8 = 0;
3900 while (p[0] < NOTACHAR)
3901 {
3902 int n = 0;
3903 if (p[0] != except)
3904 {
3905 while(p[n+1] == p[0] + n + 1) n++;
3906 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
3907 }
3908 p += n + 1;
3909 }
3910 return n8;
3911 }
3912
3913
3914
3915 /*************************************************
3916 * Add characters not in a list to a class *
3917 *************************************************/
3918
3919 /* This function is used for adding the complement of a list of horizontal or
3920 vertical whitespace to a class. The list must be in order.
3921
3922 Arguments:
3923 classbits the bit map for characters < 256
3924 uchardptr points to the pointer for extra data
3925 options the options word
3926 cd contains pointers to tables etc.
3927 p points to row of 32-bit values, terminated by NOTACHAR
3928
3929 Returns: the number of < 256 characters added
3930 the pointer to extra data is updated
3931 */
3932
3933 static int
3934 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
3935 int options, compile_data *cd, const pcre_uint32 *p)
3936 {
3937 BOOL utf = (options & PCRE_UTF8) != 0;
3938 int n8 = 0;
3939 if (p[0] > 0)
3940 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
3941 while (p[0] < NOTACHAR)
3942 {
3943 while (p[1] == p[0] + 1) p++;
3944 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
3945 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
3946 p++;
3947 }
3948 return n8;
3949 }
3950
3951
3952
3953 /*************************************************
3954 * Compile one branch *
3955 *************************************************/
3956
3957 /* Scan the pattern, compiling it into the a vector. If the options are
3958 changed during the branch, the pointer is used to change the external options
3959 bits. This function is used during the pre-compile phase when we are trying
3960 to find out the amount of memory needed, as well as during the real compile
3961 phase. The value of lengthptr distinguishes the two phases.
3962
3963 Arguments:
3964 optionsptr pointer to the option bits
3965 codeptr points to the pointer to the current code point
3966 ptrptr points to the current pattern pointer
3967 errorcodeptr points to error code variable
3968 firstcharptr place to put the first required character
3969 firstcharflagsptr place to put the first character flags, or a negative number
3970 reqcharptr place to put the last required character
3971 reqcharflagsptr place to put the last required character flags, or a negative number
3972 bcptr points to current branch chain
3973 cond_depth conditional nesting depth
3974 cd contains pointers to tables etc.
3975 lengthptr NULL during the real compile phase
3976 points to length accumulator during pre-compile phase
3977
3978 Returns: TRUE on success
3979 FALSE, with *errorcodeptr set non-zero on error
3980 */
3981
3982 static BOOL
3983 compile_branch(int *optionsptr, pcre_uchar **codeptr,
3984 const pcre_uchar **ptrptr, int *errorcodeptr,
3985 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
3986 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
3987 branch_chain *bcptr, int cond_depth,
3988 compile_data *cd, int *lengthptr)
3989 {
3990 int repeat_type, op_type;
3991 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
3992 int bravalue = 0;
3993 int greedy_default, greedy_non_default;
3994 pcre_uint32 firstchar, reqchar;
3995 pcre_int32 firstcharflags, reqcharflags;
3996 pcre_uint32 zeroreqchar, zerofirstchar;
3997 pcre_int32 zeroreqcharflags, zerofirstcharflags;
3998 pcre_int32 req_caseopt, reqvary, tempreqvary;
3999 int options = *optionsptr; /* May change dynamically */
4000 int after_manual_callout = 0;
4001 int length_prevgroup = 0;
4002 register pcre_uint32 c;
4003 int escape;
4004 register pcre_uchar *code = *codeptr;
4005 pcre_uchar *last_code = code;
4006 pcre_uchar *orig_code = code;
4007 pcre_uchar *tempcode;
4008 BOOL inescq = FALSE;
4009 BOOL groupsetfirstchar = FALSE;
4010 const pcre_uchar *ptr = *ptrptr;
4011 const pcre_uchar *tempptr;
4012 const pcre_uchar *nestptr = NULL;
4013 pcre_uchar *previous = NULL;
4014 pcre_uchar *previous_callout = NULL;
4015 pcre_uchar *save_hwm = NULL;
4016 pcre_uint8 classbits[32];
4017
4018 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4019 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4020 dynamically as we process the pattern. */
4021
4022 #ifdef SUPPORT_UTF
4023 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4024 BOOL utf = (options & PCRE_UTF8) != 0;
4025 #ifndef COMPILE_PCRE32
4026 pcre_uchar utf_chars[6];
4027 #endif
4028 #else
4029 BOOL utf = FALSE;
4030 #endif
4031
4032 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4033 class_uchardata always so that it can be passed to add_to_class() always,
4034 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4035 alternative calls for the different cases. */
4036
4037 pcre_uchar *class_uchardata;
4038 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4039 BOOL xclass;
4040 pcre_uchar *class_uchardata_base;
4041 #endif
4042
4043 #ifdef PCRE_DEBUG
4044 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4045 #endif
4046
4047 /* Set up the default and non-default settings for greediness */
4048
4049 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4050 greedy_non_default = greedy_default ^ 1;
4051
4052 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4053 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4054 matches a non-fixed char first char; reqchar just remains unset if we never
4055 find one.
4056
4057 When we hit a repeat whose minimum is zero, we may have to adjust these values
4058 to take the zero repeat into account. This is implemented by setting them to
4059 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4060 item types that can be repeated set these backoff variables appropriately. */
4061
4062 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4063 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4064
4065 /* The variable req_caseopt contains either the REQ_CASELESS value
4066 or zero, according to the current setting of the caseless flag. The
4067 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4068 firstchar or reqchar variables to record the case status of the
4069 value. This is used only for ASCII characters. */
4070
4071 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4072
4073 /* Switch on next character until the end of the branch */
4074
4075 for (;; ptr++)
4076 {
4077 BOOL negate_class;
4078 BOOL should_flip_negation;
4079 BOOL possessive_quantifier;
4080 BOOL is_quantifier;
4081 BOOL is_recurse;
4082 BOOL reset_bracount;
4083 int class_has_8bitchar;
4084 int class_one_char;
4085 int newoptions;
4086 int recno;
4087 int refsign;
4088 int skipbytes;
4089 pcre_uint32 subreqchar, subfirstchar;
4090 pcre_int32 subreqcharflags, subfirstcharflags;
4091 int terminator;
4092 unsigned int mclength;
4093 unsigned int tempbracount;
4094 pcre_uint32 ec;
4095 pcre_uchar mcbuffer[8];
4096
4097 /* Get next character in the pattern */
4098
4099 c = *ptr;
4100
4101 /* If we are at the end of a nested substitution, revert to the outer level
4102 string. Nesting only happens one level deep. */
4103
4104 if (c == CHAR_NULL && nestptr != NULL)
4105 {
4106 ptr = nestptr;
4107 nestptr = NULL;
4108 c = *ptr;
4109 }
4110
4111 /* If we are in the pre-compile phase, accumulate the length used for the
4112 previous cycle of this loop. */
4113
4114 if (lengthptr != NULL)
4115 {
4116 #ifdef PCRE_DEBUG
4117 if (code > cd->hwm) cd->hwm = code; /* High water info */
4118 #endif
4119 if (code > cd->start_workspace + cd->workspace_size -
4120 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4121 {
4122 *errorcodeptr = ERR52;
4123 goto FAILED;
4124 }
4125
4126 /* There is at least one situation where code goes backwards: this is the
4127 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4128 the class is simply eliminated. However, it is created first, so we have to
4129 allow memory for it. Therefore, don't ever reduce the length at this point.
4130 */
4131
4132 if (code < last_code) code = last_code;
4133
4134 /* Paranoid check for integer overflow */
4135
4136 if (OFLOW_MAX - *lengthptr < code - last_code)
4137 {
4138 *errorcodeptr = ERR20;
4139 goto FAILED;
4140 }
4141
4142 *lengthptr += (int)(code - last_code);
4143 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4144 (int)(code - last_code), c, c));
4145
4146 /* If "previous" is set and it is not at the start of the work space, move
4147 it back to there, in order to avoid filling up the work space. Otherwise,
4148 if "previous" is NULL, reset the current code pointer to the start. */
4149
4150 if (previous != NULL)
4151 {
4152 if (previous > orig_code)
4153 {
4154 memmove(orig_code, previous, IN_UCHARS(code - previous));
4155 code -= previous - orig_code;
4156 previous = orig_code;
4157 }
4158 }
4159 else code = orig_code;
4160
4161 /* Remember where this code item starts so we can pick up the length
4162 next time round. */
4163
4164 last_code = code;
4165 }
4166
4167 /* In the real compile phase, just check the workspace used by the forward
4168 reference list. */
4169
4170 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4171 WORK_SIZE_SAFETY_MARGIN)
4172 {
4173 *errorcodeptr = ERR52;
4174 goto FAILED;
4175 }
4176
4177 /* If in \Q...\E, check for the end; if not, we have a literal */
4178
4179 if (inescq && c != CHAR_NULL)
4180 {
4181 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4182 {
4183 inescq = FALSE;
4184 ptr++;
4185 continue;
4186 }
4187 else
4188 {
4189 if (previous_callout != NULL)
4190 {
4191 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4192 complete_callout(previous_callout, ptr, cd);
4193 previous_callout = NULL;
4194 }
4195 if ((options & PCRE_AUTO_CALLOUT) != 0)
4196 {
4197 previous_callout = code;
4198 code = auto_callout(code, ptr, cd);
4199 }
4200 goto NORMAL_CHAR;
4201 }
4202 }
4203
4204 /* Fill in length of a previous callout, except when the next thing is
4205 a quantifier. */
4206
4207 is_quantifier =
4208 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4209 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4210
4211 if (!is_quantifier && previous_callout != NULL &&
4212 after_manual_callout-- <= 0)
4213 {
4214 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4215 complete_callout(previous_callout, ptr, cd);
4216 previous_callout = NULL;
4217 }
4218
4219 /* In extended mode, skip white space and comments. */
4220
4221 if ((options & PCRE_EXTENDED) != 0)
4222 {
4223 if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
4224 if (c == CHAR_NUMBER_SIGN)
4225 {
4226 ptr++;
4227 while (*ptr != CHAR_NULL)
4228 {
4229 if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
4230 ptr++;
4231 #ifdef SUPPORT_UTF
4232 if (utf) FORWARDCHAR(ptr);
4233 #endif
4234 }
4235 if (*ptr != CHAR_NULL) continue;
4236
4237 /* Else fall through to handle end of string */
4238 c = 0;
4239 }
4240 }
4241
4242 /* No auto callout for quantifiers. */
4243
4244 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier)
4245 {
4246 previous_callout = code;
4247 code = auto_callout(code, ptr, cd);
4248 }
4249
4250 switch(c)
4251 {
4252 /* ===================================================================*/
4253 case 0: /* The branch terminates at string end */
4254 case CHAR_VERTICAL_LINE: /* or | or ) */
4255 case CHAR_RIGHT_PARENTHESIS:
4256 *firstcharptr = firstchar;
4257 *firstcharflagsptr = firstcharflags;
4258 *reqcharptr = reqchar;
4259 *reqcharflagsptr = reqcharflags;
4260 *codeptr = code;
4261 *ptrptr = ptr;
4262 if (lengthptr != NULL)
4263 {
4264 if (OFLOW_MAX - *lengthptr < code - last_code)
4265 {
4266 *errorcodeptr = ERR20;
4267 goto FAILED;
4268 }
4269 *lengthptr += (int)(code - last_code); /* To include callout length */
4270 DPRINTF((">> end branch\n"));
4271 }
4272 return TRUE;
4273
4274
4275 /* ===================================================================*/
4276 /* Handle single-character metacharacters. In multiline mode, ^ disables
4277 the setting of any following char as a first character. */
4278
4279 case CHAR_CIRCUMFLEX_ACCENT:
4280 previous = NULL;
4281 if ((options & PCRE_MULTILINE) != 0)
4282 {
4283 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4284 *code++ = OP_CIRCM;
4285 }
4286 else *code++ = OP_CIRC;
4287 break;
4288
4289 case CHAR_DOLLAR_SIGN:
4290 previous = NULL;
4291 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4292 break;
4293
4294 /* There can never be a first char if '.' is first, whatever happens about
4295 repeats. The value of reqchar doesn't change either. */
4296
4297 case CHAR_DOT:
4298 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4299 zerofirstchar = firstchar;
4300 zerofirstcharflags = firstcharflags;
4301 zeroreqchar = reqchar;
4302 zeroreqcharflags = reqcharflags;
4303 previous = code;
4304 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4305 break;
4306
4307
4308 /* ===================================================================*/
4309 /* Character classes. If the included characters are all < 256, we build a
4310 32-byte bitmap of the permitted characters, except in the special case
4311 where there is only one such character. For negated classes, we build the
4312 map as usual, then invert it at the end. However, we use a different opcode
4313 so that data characters > 255 can be handled correctly.
4314
4315 If the class contains characters outside the 0-255 range, a different
4316 opcode is compiled. It may optionally have a bit map for characters < 256,
4317 but those above are are explicitly listed afterwards. A flag byte tells
4318 whether the bitmap is present, and whether this is a negated class or not.
4319
4320 In JavaScript compatibility mode, an isolated ']' causes an error. In
4321 default (Perl) mode, it is treated as a data character. */
4322
4323 case CHAR_RIGHT_SQUARE_BRACKET:
4324 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4325 {
4326 *errorcodeptr = ERR64;
4327 goto FAILED;
4328 }
4329 goto NORMAL_CHAR;
4330
4331 case CHAR_LEFT_SQUARE_BRACKET:
4332 previous = code;
4333
4334 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4335 they are encountered at the top level, so we'll do that too. */
4336
4337 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4338 ptr[1] == CHAR_EQUALS_SIGN) &&
4339 check_posix_syntax(ptr, &tempptr))
4340 {
4341 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4342 goto FAILED;
4343 }
4344
4345 /* If the first character is '^', set the negation flag and skip it. Also,
4346 if the first few characters (either before or after ^) are \Q\E or \E we
4347 skip them too. This makes for compatibility with Perl. */
4348
4349 negate_class = FALSE;
4350 for (;;)
4351 {
4352 c = *(++ptr);
4353 if (c == CHAR_BACKSLASH)
4354 {
4355 if (ptr[1] == CHAR_E)
4356 ptr++;
4357 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4358 ptr += 3;
4359 else
4360 break;
4361 }
4362 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4363 negate_class = TRUE;
4364 else break;
4365 }
4366
4367 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4368 an initial ']' is taken as a data character -- the code below handles
4369 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4370 [^] must match any character, so generate OP_ALLANY. */
4371
4372 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4373 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4374 {
4375 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4376 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4377 zerofirstchar = firstchar;
4378 zerofirstcharflags = firstcharflags;
4379 break;
4380 }
4381
4382 /* If a class contains a negative special such as \S, we need to flip the
4383 negation flag at the end, so that support for characters > 255 works
4384 correctly (they are all included in the class). */
4385
4386 should_flip_negation = FALSE;
4387
4388 /* For optimization purposes, we track some properties of the class:
4389 class_has_8bitchar will be non-zero if the class contains at least one <
4390 256 character; class_one_char will be 1 if the class contains just one
4391 character. */
4392
4393 class_has_8bitchar = 0;
4394 class_one_char = 0;
4395
4396 /* Initialize the 32-char bit map to all zeros. We build the map in a
4397 temporary bit of memory, in case the class contains fewer than two
4398 8-bit characters because in that case the compiled code doesn't use the bit
4399 map. */
4400
4401 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4402
4403 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4404 xclass = FALSE;
4405 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4406 class_uchardata_base = class_uchardata; /* Save the start */
4407 #endif
4408
4409 /* Process characters until ] is reached. By writing this as a "do" it
4410 means that an initial ] is taken as a data character. At the start of the
4411 loop, c contains the first byte of the character. */
4412
4413 if (c != CHAR_NULL) do
4414 {
4415 const pcre_uchar *oldptr;
4416
4417 #ifdef SUPPORT_UTF
4418 if (utf && HAS_EXTRALEN(c))
4419 { /* Braces are required because the */
4420 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4421 }
4422 #endif
4423
4424 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4425 /* In the pre-compile phase, accumulate the length of any extra
4426 data and reset the pointer. This is so that very large classes that
4427 contain a zillion > 255 characters no longer overwrite the work space
4428 (which is on the stack). We have to remember that there was XCLASS data,
4429 however. */
4430
4431 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4432 {
4433 xclass = TRUE;
4434 *lengthptr += class_uchardata - class_uchardata_base;
4435 class_uchardata = class_uchardata_base;
4436 }
4437 #endif
4438
4439 /* Inside \Q...\E everything is literal except \E */
4440
4441 if (inescq)
4442 {
4443 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4444 {
4445 inescq = FALSE; /* Reset literal state */
4446 ptr++; /* Skip the 'E' */
4447 continue; /* Carry on with next */
4448 }
4449 goto CHECK_RANGE; /* Could be range if \E follows */
4450 }
4451
4452 /* Handle POSIX class names. Perl allows a negation extension of the
4453 form [:^name:]. A square bracket that doesn't match the syntax is
4454 treated as a literal. We also recognize the POSIX constructions
4455 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4456 5.6 and 5.8 do. */
4457
4458 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4459 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4460 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4461 {
4462 BOOL local_negate = FALSE;
4463 int posix_class, taboffset, tabopt;
4464 register const pcre_uint8 *cbits = cd->cbits;
4465 pcre_uint8 pbits[32];
4466
4467 if (ptr[1] != CHAR_COLON)
4468 {
4469 *errorcodeptr = ERR31;
4470 goto FAILED;
4471 }
4472
4473 ptr += 2;
4474 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4475 {
4476 local_negate = TRUE;
4477 should_flip_negation = TRUE; /* Note negative special */
4478 ptr++;
4479 }
4480
4481 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4482 if (posix_class < 0)
4483 {
4484 *errorcodeptr = ERR30;
4485 goto FAILED;
4486 }
4487
4488 /* If matching is caseless, upper and lower are converted to
4489 alpha. This relies on the fact that the class table starts with
4490 alpha, lower, upper as the first 3 entries. */
4491
4492 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4493 posix_class = 0;
4494
4495 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4496 different escape sequences that use Unicode properties. */
4497
4498 #ifdef SUPPORT_UCP
4499 if ((options & PCRE_UCP) != 0)
4500 {
4501 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4502 if (posix_substitutes[pc] != NULL)
4503 {
4504 nestptr = tempptr + 1;
4505 ptr = posix_substitutes[pc] - 1;
4506 continue;
4507 }
4508 }
4509 #endif
4510 /* In the non-UCP case, we build the bit map for the POSIX class in a
4511 chunk of local store because we may be adding and subtracting from it,
4512 and we don't want to subtract bits that may be in the main map already.
4513 At the end we or the result into the bit map that is being built. */
4514
4515 posix_class *= 3;
4516
4517 /* Copy in the first table (always present) */
4518
4519 memcpy(pbits, cbits + posix_class_maps[posix_class],
4520 32 * sizeof(pcre_uint8));
4521
4522 /* If there is a second table, add or remove it as required. */
4523
4524 taboffset = posix_class_maps[posix_class + 1];
4525 tabopt = posix_class_maps[posix_class + 2];
4526
4527 if (taboffset >= 0)
4528 {
4529 if (tabopt >= 0)
4530 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
4531 else
4532 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
4533 }
4534
4535 /* Now see if we need to remove any special characters. An option
4536 value of 1 removes vertical space and 2 removes underscore. */
4537
4538 if (tabopt < 0) tabopt = -tabopt;
4539 if (tabopt == 1) pbits[1] &= ~0x3c;
4540 else if (tabopt == 2) pbits[11] &= 0x7f;
4541
4542 /* Add the POSIX table or its complement into the main table that is
4543 being built and we are done. */
4544
4545 if (local_negate)
4546 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
4547 else
4548 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
4549
4550 ptr = tempptr + 1;
4551 /* Every class contains at least one < 256 character. */
4552 class_has_8bitchar = 1;
4553 /* Every class contains at least two characters. */
4554 class_one_char = 2;
4555 continue; /* End of POSIX syntax handling */
4556 }
4557
4558 /* Backslash may introduce a single character, or it may introduce one
4559 of the specials, which just set a flag. The sequence \b is a special
4560 case. Inside a class (and only there) it is treated as backspace. We
4561 assume that other escapes have more than one character in them, so
4562 speculatively set both class_has_8bitchar and class_one_char bigger
4563 than one. Unrecognized escapes fall through and are either treated
4564 as literal characters (by default), or are faulted if
4565 PCRE_EXTRA is set. */
4566
4567 if (c == CHAR_BACKSLASH)
4568 {
4569 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4570 TRUE);
4571 if (*errorcodeptr != 0) goto FAILED;
4572 if (escape == 0) c = ec;
4573 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
4574 else if (escape == ESC_N) /* \N is not supported in a class */
4575 {
4576 *errorcodeptr = ERR71;
4577 goto FAILED;
4578 }
4579 else if (escape == ESC_Q) /* Handle start of quoted string */
4580 {
4581 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4582 {
4583 ptr += 2; /* avoid empty string */
4584 }
4585 else inescq = TRUE;
4586 continue;
4587 }
4588 else if (escape == ESC_E) continue; /* Ignore orphan \E */
4589
4590 else
4591 {
4592 register const pcre_uint8 *cbits = cd->cbits;
4593 /* Every class contains at least two < 256 characters. */
4594 class_has_8bitchar++;
4595 /* Every class contains at least two characters. */
4596 class_one_char += 2;
4597
4598 switch (escape)
4599 {
4600 #ifdef SUPPORT_UCP
4601 case ESC_du: /* These are the values given for \d etc */
4602 case ESC_DU: /* when PCRE_UCP is set. We replace the */
4603 case ESC_wu: /* escape sequence with an appropriate \p */
4604 case ESC_WU: /* or \P to test Unicode properties instead */
4605 case ESC_su: /* of the default ASCII testing. */
4606 case ESC_SU:
4607 nestptr = ptr;
4608 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
4609 class_has_8bitchar--; /* Undo! */
4610 continue;
4611 #endif
4612 case ESC_d:
4613 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
4614 continue;
4615
4616 case ESC_D:
4617 should_flip_negation = TRUE;
4618 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
4619 continue;
4620
4621 case ESC_w:
4622 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
4623 continue;
4624
4625 case ESC_W:
4626 should_flip_negation = TRUE;
4627 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4628 continue;
4629
4630 /* Perl 5.004 onwards omits VT from \s, but we must preserve it
4631 if it was previously set by something earlier in the character
4632 class. Luckily, the value of CHAR_VT is 0x0b in both ASCII and
4633 EBCDIC, so we lazily just adjust the appropriate bit. */
4634
4635 case ESC_s:
4636 classbits[0] |= cbits[cbit_space];
4637 classbits[1] |= cbits[cbit_space+1] & ~0x08;
4638 for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
4639 continue;
4640
4641 case ESC_S:
4642 should_flip_negation = TRUE;
4643 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
4644 classbits[1] |= 0x08; /* Perl 5.004 onwards omits VT from \s */
4645 continue;
4646
4647 /* The rest apply in both UCP and non-UCP cases. */
4648
4649 case ESC_h:
4650 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4651 PRIV(hspace_list), NOTACHAR);
4652 continue;
4653
4654 case ESC_H:
4655 (void)add_not_list_to_class(classbits, &class_uchardata, options,
4656 cd, PRIV(hspace_list));
4657 continue;
4658
4659 case ESC_v:
4660 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4661 PRIV(vspace_list), NOTACHAR);
4662 continue;
4663
4664 case ESC_V:
4665 (void)add_not_list_to_class(classbits, &class_uchardata, options,
4666 cd, PRIV(vspace_list));
4667 continue;
4668
4669 #ifdef SUPPORT_UCP
4670 case ESC_p:
4671 case ESC_P:
4672 {
4673 BOOL negated;
4674 unsigned int ptype = 0, pdata = 0;
4675 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
4676 goto FAILED;
4677 *class_uchardata++ = ((escape == ESC_p) != negated)?
4678 XCL_PROP : XCL_NOTPROP;
4679 *class_uchardata++ = ptype;
4680 *class_uchardata++ = pdata;
4681 class_has_8bitchar--; /* Undo! */
4682 continue;
4683 }
4684 #endif
4685 /* Unrecognized escapes are faulted if PCRE is running in its
4686 strict mode. By default, for compatibility with Perl, they are
4687 treated as literals. */
4688
4689 default:
4690 if ((options & PCRE_EXTRA) != 0)
4691 {
4692 *errorcodeptr = ERR7;
4693 goto FAILED;
4694 }
4695 class_has_8bitchar--; /* Undo the speculative increase. */
4696 class_one_char -= 2; /* Undo the speculative increase. */
4697 c = *ptr; /* Get the final character and fall through */
4698 break;
4699 }
4700 }
4701
4702 /* Fall through if the escape just defined a single character (c >= 0).
4703 This may be greater than 256. */
4704
4705 escape = 0;
4706
4707 } /* End of backslash handling */
4708
4709 /* A character may be followed by '-' to form a range. However, Perl does
4710 not permit ']' to be the end of the range. A '-' character at the end is
4711 treated as a literal. Perl ignores orphaned \E sequences entirely. The
4712 code for handling \Q and \E is messy. */
4713
4714 CHECK_RANGE:
4715 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4716 {
4717 inescq = FALSE;
4718 ptr += 2;
4719 }
4720 oldptr = ptr;
4721
4722 /* Remember if \r or \n were explicitly used */
4723
4724 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4725
4726 /* Check for range */
4727
4728 if (!inescq && ptr[1] == CHAR_MINUS)
4729 {
4730 pcre_uint32 d;
4731 ptr += 2;
4732 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
4733
4734 /* If we hit \Q (not followed by \E) at this point, go into escaped
4735 mode. */
4736
4737 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
4738 {
4739 ptr += 2;
4740 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4741 { ptr += 2; continue; }
4742 inescq = TRUE;
4743 break;
4744 }
4745
4746 /* Minus (hyphen) at the end of a class is treated as a literal, so put
4747 back the pointer and jump to handle the character that preceded it. */
4748
4749 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
4750 {
4751 ptr = oldptr;
4752 goto CLASS_SINGLE_CHARACTER;
4753 }
4754
4755 /* Otherwise, we have a potential range; pick up the next character */
4756
4757 #ifdef SUPPORT_UTF
4758 if (utf)
4759 { /* Braces are required because the */
4760 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
4761 }
4762 else
4763 #endif
4764 d = *ptr; /* Not UTF-8 mode */
4765
4766 /* The second part of a range can be a single-character escape, but
4767 not any of the other escapes. Perl 5.6 treats a hyphen as a literal
4768 in such circumstances. */
4769
4770 if (!inescq && d == CHAR_BACKSLASH)
4771 {
4772 int descape;
4773 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
4774 if (*errorcodeptr != 0) goto FAILED;
4775
4776 /* \b is backspace; any other special means the '-' was literal. */
4777
4778 if (descape != 0)
4779 {
4780 if (descape == ESC_b) d = CHAR_BS; else
4781 {
4782 ptr = oldptr;
4783 goto CLASS_SINGLE_CHARACTER; /* A few lines below */
4784 }
4785 }
4786 }
4787
4788 /* Check that the two values are in the correct order. Optimize
4789 one-character ranges. */
4790
4791 if (d < c)
4792 {
4793 *errorcodeptr = ERR8;
4794 goto FAILED;
4795 }
4796 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
4797
4798 /* We have found a character range, so single character optimizations
4799 cannot be done anymore. Any value greater than 1 indicates that there
4800 is more than one character. */
4801
4802 class_one_char = 2;
4803
4804 /* Remember an explicit \r or \n, and add the range to the class. */
4805
4806 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
4807
4808 class_has_8bitchar +=
4809 add_to_class(classbits, &class_uchardata, options, cd, c, d);
4810
4811 continue; /* Go get the next char in the class */
4812 }
4813
4814 /* Handle a single character - we can get here for a normal non-escape
4815 char, or after \ that introduces a single character or for an apparent
4816 range that isn't. Only the value 1 matters for class_one_char, so don't
4817 increase it if it is already 2 or more ... just in case there's a class
4818 with a zillion characters in it. */
4819
4820 CLASS_SINGLE_CHARACTER:
4821 if (class_one_char < 2) class_one_char++;
4822
4823 /* If class_one_char is 1, we have the first single character in the
4824 class, and there have been no prior ranges, or XCLASS items generated by
4825 escapes. If this is the final character in the class, we can optimize by
4826 turning the item into a 1-character OP_CHAR[I] if it's positive, or
4827 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
4828 to be set. Otherwise, there can be no first char if this item is first,
4829 whatever repeat count may follow. In the case of reqchar, save the
4830 previous value for reinstating. */
4831
4832 if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
4833 {
4834 ptr++;
4835 zeroreqchar = reqchar;
4836 zeroreqcharflags = reqcharflags;
4837
4838 if (negate_class)
4839 {
4840 #ifdef SUPPORT_UCP
4841 int d;
4842 #endif
4843 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4844 zerofirstchar = firstchar;
4845 zerofirstcharflags = firstcharflags;
4846
4847 /* For caseless UTF-8 mode when UCP support is available, check
4848 whether this character has more than one other case. If so, generate
4849 a special OP_NOTPROP item instead of OP_NOTI. */
4850
4851 #ifdef SUPPORT_UCP
4852 if (utf && (options & PCRE_CASELESS) != 0 &&
4853 (d = UCD_CASESET(c)) != 0)
4854 {
4855 *code++ = OP_NOTPROP;
4856 *code++ = PT_CLIST;
4857 *code++ = d;
4858 }
4859 else
4860 #endif
4861 /* Char has only one other case, or UCP not available */
4862
4863 {
4864 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
4865 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
4866 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
4867 code += PRIV(ord2utf)(c, code);
4868 else
4869 #endif
4870 *code++ = c;
4871 }
4872
4873 /* We are finished with this character class */
4874
4875 goto END_CLASS;
4876 }
4877
4878 /* For a single, positive character, get the value into mcbuffer, and
4879 then we can handle this with the normal one-character code. */
4880
4881 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
4882 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
4883 mclength = PRIV(ord2utf)(c, mcbuffer);
4884 else
4885 #endif
4886 {
4887 mcbuffer[0] = c;
4888 mclength = 1;
4889 }
4890 goto ONE_CHAR;
4891 } /* End of 1-char optimization */
4892
4893 /* There is more than one character in the class, or an XCLASS item
4894 has been generated. Add this character to the class. */
4895
4896 class_has_8bitchar +=
4897 add_to_class(classbits, &class_uchardata, options, cd, c, c);
4898 }
4899
4900 /* Loop until ']' reached. This "while" is the end of the "do" far above.
4901 If we are at the end of an internal nested string, revert to the outer
4902 string. */
4903
4904 while (((c = *(++ptr)) != CHAR_NULL ||
4905 (nestptr != NULL &&
4906 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
4907 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
4908
4909 /* Check for missing terminating ']' */
4910
4911 if (c == CHAR_NULL)
4912 {
4913 *errorcodeptr = ERR6;
4914 goto FAILED;
4915 }
4916
4917 /* We will need an XCLASS if data has been placed in class_uchardata. In
4918 the second phase this is a sufficient test. However, in the pre-compile
4919 phase, class_uchardata gets emptied to prevent workspace overflow, so it
4920 only if the very last character in the class needs XCLASS will it contain
4921 anything at this point. For this reason, xclass gets set TRUE above when
4922 uchar_classdata is emptied, and that's why this code is the way it is here
4923 instead of just doing a test on class_uchardata below. */
4924
4925 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4926 if (class_uchardata > class_uchardata_base) xclass = TRUE;
4927 #endif
4928
4929 /* If this is the first thing in the branch, there can be no first char
4930 setting, whatever the repeat count. Any reqchar setting must remain
4931 unchanged after any kind of repeat. */
4932
4933 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4934 zerofirstchar = firstchar;
4935 zerofirstcharflags = firstcharflags;
4936 zeroreqchar = reqchar;
4937 zeroreqcharflags = reqcharflags;
4938
4939 /* If there are characters with values > 255, we have to compile an
4940 extended class, with its own opcode, unless there was a negated special
4941 such as \S in the class, and PCRE_UCP is not set, because in that case all
4942 characters > 255 are in the class, so any that were explicitly given as
4943 well can be ignored. If (when there are explicit characters > 255 that must
4944 be listed) there are no characters < 256, we can omit the bitmap in the
4945 actual compiled code. */
4946
4947 #ifdef SUPPORT_UTF
4948 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
4949 #elif !defined COMPILE_PCRE8
4950 if (xclass && !should_flip_negation)
4951 #endif
4952 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4953 {
4954 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
4955 *code++ = OP_XCLASS;
4956 code += LINK_SIZE;
4957 *code = negate_class? XCL_NOT:0;
4958
4959 /* If the map is required, move up the extra data to make room for it;
4960 otherwise just move the code pointer to the end of the extra data. */
4961
4962 if (class_has_8bitchar > 0)
4963 {
4964 *code++ |= XCL_MAP;
4965 memmove(code + (32 / sizeof(pcre_uchar)), code,
4966 IN_UCHARS(class_uchardata - code));
4967 memcpy(code, classbits, 32);
4968 code = class_uchardata + (32 / sizeof(pcre_uchar));
4969 }
4970 else code = class_uchardata;
4971
4972 /* Now fill in the complete length of the item */
4973
4974 PUT(previous, 1, (int)(code - previous));
4975 break; /* End of class handling */
4976 }
4977 #endif
4978
4979 /* If there are no characters > 255, or they are all to be included or
4980 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4981 whole class was negated and whether there were negative specials such as \S
4982 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4983 negating it if necessary. */
4984
4985 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4986 if (lengthptr == NULL) /* Save time in the pre-compile phase */
4987 {
4988 if (negate_class)
4989 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
4990 memcpy(code, classbits, 32);
4991 }
4992 code += 32 / sizeof(pcre_uchar);
4993
4994 END_CLASS:
4995 break;
4996
4997
4998 /* ===================================================================*/
4999 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5000 has been tested above. */
5001
5002 case CHAR_LEFT_CURLY_BRACKET:
5003 if (!is_quantifier) goto NORMAL_CHAR;
5004 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5005 if (*errorcodeptr != 0) goto FAILED;
5006 goto REPEAT;
5007
5008 case CHAR_ASTERISK:
5009 repeat_min = 0;
5010 repeat_max = -1;
5011 goto REPEAT;
5012
5013 case CHAR_PLUS:
5014 repeat_min = 1;
5015 repeat_max = -1;
5016 goto REPEAT;
5017
5018 case CHAR_QUESTION_MARK:
5019 repeat_min = 0;
5020 repeat_max = 1;
5021
5022 REPEAT:
5023 if (previous == NULL)
5024 {
5025 *errorcodeptr = ERR9;
5026 goto FAILED;
5027 }
5028
5029 if (repeat_min == 0)
5030 {
5031 firstchar = zerofirstchar; /* Adjust for zero repeat */
5032 firstcharflags = zerofirstcharflags;
5033 reqchar = zeroreqchar; /* Ditto */
5034 reqcharflags = zeroreqcharflags;
5035 }
5036
5037 /* Remember whether this is a variable length repeat */
5038
5039 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5040
5041 op_type = 0; /* Default single-char op codes */
5042 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5043
5044 /* Save start of previous item, in case we have to move it up in order to
5045 insert something before it. */
5046
5047 tempcode = previous;
5048
5049 /* If the next character is '+', we have a possessive quantifier. This
5050 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5051 If the next character is '?' this is a minimizing repeat, by default,
5052 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5053 repeat type to the non-default. */
5054
5055 if (ptr[1] == CHAR_PLUS)
5056 {
5057 repeat_type = 0; /* Force greedy */
5058 possessive_quantifier = TRUE;
5059 ptr++;
5060 }
5061 else if (ptr[1] == CHAR_QUESTION_MARK)
5062 {
5063 repeat_type = greedy_non_default;
5064 ptr++;
5065 }
5066 else repeat_type = greedy_default;
5067
5068 /* If previous was a recursion call, wrap it in atomic brackets so that
5069 previous becomes the atomic group. All recursions were so wrapped in the
5070 past, but it no longer happens for non-repeated recursions. In fact, the
5071 repeated ones could be re-implemented independently so as not to need this,
5072 but for the moment we rely on the code for repeating groups. */
5073
5074 if (*previous == OP_RECURSE)
5075 {
5076 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5077 *previous = OP_ONCE;
5078 PUT(previous, 1, 2 + 2*LINK_SIZE);
5079 previous[2 + 2*LINK_SIZE] = OP_KET;
5080 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5081 code += 2 + 2 * LINK_SIZE;
5082 length_prevgroup = 3 + 3*LINK_SIZE;
5083
5084 /* When actually compiling, we need to check whether this was a forward
5085 reference, and if so, adjust the offset. */
5086
5087 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5088 {
5089 int offset = GET(cd->hwm, -LINK_SIZE);
5090 if (offset == previous + 1 - cd->start_code)
5091 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5092 }
5093 }
5094
5095 /* Now handle repetition for the different types of item. */
5096
5097 /* If previous was a character or negated character match, abolish the item
5098 and generate a repeat item instead. If a char item has a minimum of more
5099 than one, ensure that it is set in reqchar - it might not be if a sequence
5100 such as x{3} is the first thing in a branch because the x will have gone
5101 into firstchar instead. */
5102
5103 if (*previous == OP_CHAR || *previous == OP_CHARI
5104 || *previous == OP_NOT || *previous == OP_NOTI)
5105 {
5106 switch (*previous)
5107 {
5108 default: /* Make compiler happy. */
5109 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5110 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5111 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5112 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5113 }
5114
5115 /* Deal with UTF characters that take up more than one character. It's
5116 easier to write this out separately than try to macrify it. Use c to
5117 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5118 it's a length rather than a small character. */
5119
5120 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5121 if (utf && NOT_FIRSTCHAR(code[-1]))
5122 {
5123 pcre_uchar *lastchar = code - 1;
5124 BACKCHAR(lastchar);
5125 c = (int)(code - lastchar); /* Length of UTF-8 character */
5126 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5127 c |= UTF_LENGTH; /* Flag c as a length */
5128 }
5129 else
5130 #endif /* SUPPORT_UTF */
5131
5132 /* Handle the case of a single charater - either with no UTF support, or
5133 with UTF disabled, or for a single character UTF character. */
5134 {
5135 c = code[-1];
5136 if (*previous <= OP_CHARI && repeat_min > 1)
5137 {
5138 reqchar = c;
5139 reqcharflags = req_caseopt | cd->req_varyopt;
5140 }
5141 }
5142
5143 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5144 }
5145
5146 /* If previous was a character type match (\d or similar), abolish it and
5147 create a suitable repeat item. The code is shared with single-character
5148 repeats by setting op_type to add a suitable offset into repeat_type. Note
5149 the the Unicode property types will be present only when SUPPORT_UCP is
5150 defined, but we don't wrap the little bits of code here because it just
5151 makes it horribly messy. */
5152
5153 else if (*previous < OP_EODN)
5154 {
5155 pcre_uchar *oldcode;
5156 int prop_type, prop_value;
5157 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5158 c = *previous;
5159
5160 OUTPUT_SINGLE_REPEAT:
5161 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5162 {
5163 prop_type = previous[1];
5164 prop_value = previous[2];
5165 }
5166 else prop_type = prop_value = -1;
5167
5168 oldcode = code;
5169 code = previous; /* Usually overwrite previous item */
5170
5171 /* If the maximum is zero then the minimum must also be zero; Perl allows
5172 this case, so we do too - by simply omitting the item altogether. */
5173
5174 if (repeat_max == 0) goto END_REPEAT;
5175
5176 /* Combine the op_type with the repeat_type */
5177
5178 repeat_type += op_type;
5179
5180 /* A minimum of zero is handled either as the special case * or ?, or as
5181 an UPTO, with the maximum given. */
5182
5183 if (repeat_min == 0)
5184 {
5185 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5186 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5187 else
5188 {
5189 *code++ = OP_UPTO + repeat_type;
5190 PUT2INC(code, 0, repeat_max);
5191 }
5192 }
5193
5194 /* A repeat minimum of 1 is optimized into some special cases. If the
5195 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5196 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5197 one less than the maximum. */
5198
5199 else if (repeat_min == 1)
5200 {
5201 if (repeat_max == -1)
5202 *code++ = OP_PLUS + repeat_type;
5203 else
5204 {
5205 code = oldcode; /* leave previous item in place */
5206 if (repeat_max == 1) goto END_REPEAT;
5207 *code++ = OP_UPTO + repeat_type;
5208 PUT2INC(code, 0, repeat_max - 1);
5209 }
5210 }
5211
5212 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5213 handled as an EXACT followed by an UPTO. */
5214
5215 else
5216 {
5217 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5218 PUT2INC(code, 0, repeat_min);
5219
5220 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5221 we have to insert the character for the previous code. For a repeated
5222 Unicode property match, there are two extra bytes that define the
5223 required property. In UTF-8 mode, long characters have their length in
5224 c, with the UTF_LENGTH bit as a flag. */
5225
5226 if (repeat_max < 0)
5227 {
5228 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5229 if (utf && (c & UTF_LENGTH) != 0)
5230 {
5231 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5232 code += c & 7;
5233 }
5234 else
5235 #endif
5236 {
5237 *code++ = c;
5238 if (prop_type >= 0)
5239 {
5240 *code++ = prop_type;
5241 *code++ = prop_value;
5242 }
5243 }
5244 *code++ = OP_STAR + repeat_type;
5245 }
5246
5247 /* Else insert an UPTO if the max is greater than the min, again
5248 preceded by the character, for the previously inserted code. If the
5249 UPTO is just for 1 instance, we can use QUERY instead. */
5250
5251 else if (repeat_max != repeat_min)
5252 {
5253 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5254 if (utf && (c & UTF_LENGTH) != 0)
5255 {
5256 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5257 code += c & 7;
5258 }
5259 else
5260 #endif
5261 *code++ = c;
5262 if (prop_type >= 0)
5263 {
5264 *code++ = prop_type;
5265 *code++ = prop_value;
5266 }
5267 repeat_max -= repeat_min;
5268
5269 if (repeat_max == 1)
5270 {
5271 *code++ = OP_QUERY + repeat_type;
5272 }
5273 else
5274 {
5275 *code++ = OP_UPTO + repeat_type;
5276 PUT2INC(code, 0, repeat_max);
5277 }
5278 }
5279 }
5280
5281 /* The character or character type itself comes last in all cases. */
5282
5283 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5284 if (utf && (c & UTF_LENGTH) != 0)
5285 {
5286 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5287 code += c & 7;
5288 }
5289 else
5290 #endif
5291 *code++ = c;
5292
5293 /* For a repeated Unicode property match, there are two extra bytes that
5294 define the required property. */
5295
5296 #ifdef SUPPORT_UCP
5297 if (prop_type >= 0)
5298 {
5299 *code++ = prop_type;
5300 *code++ = prop_value;
5301 }
5302 #endif
5303 }
5304
5305 /* If previous was a character class or a back reference, we put the repeat
5306 stuff after it, but just skip the item if the repeat was {0,0}. */
5307
5308 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5309 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5310 *previous == OP_XCLASS ||
5311 #endif
5312 *previous == OP_REF || *previous == OP_REFI ||
5313 *previous == OP_DNREF || *previous == OP_DNREFI)
5314 {
5315 if (repeat_max == 0)
5316 {
5317 code = previous;
5318 goto END_REPEAT;
5319 }
5320
5321 if (repeat_min == 0 && repeat_max == -1)
5322 *code++ = OP_CRSTAR + repeat_type;
5323 else if (repeat_min == 1 && repeat_max == -1)
5324 *code++ = OP_CRPLUS + repeat_type;
5325 else if (repeat_min == 0 && repeat_max == 1)
5326 *code++ = OP_CRQUERY + repeat_type;
5327 else
5328 {
5329 *code++ = OP_CRRANGE + repeat_type;
5330 PUT2INC(code, 0, repeat_min);
5331 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5332 PUT2INC(code, 0, repeat_max);
5333 }
5334 }
5335
5336 /* If previous was a bracket group, we may have to replicate it in certain
5337 cases. Note that at this point we can encounter only the "basic" bracket
5338 opcodes such as BRA and CBRA, as this is the place where they get converted
5339 into the more special varieties such as BRAPOS and SBRA. A test for >=
5340 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5341 ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
5342 repetition of assertions, but now it does, for Perl compatibility. */
5343
5344 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5345 {
5346 register int i;
5347 int len = (int)(code - previous);
5348 pcre_uchar *bralink = NULL;
5349 pcre_uchar *brazeroptr = NULL;
5350
5351 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5352 we just ignore the repeat. */
5353
5354 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5355 goto END_REPEAT;
5356
5357 /* There is no sense in actually repeating assertions. The only potential
5358 use of repetition is in cases when the assertion is optional. Therefore,
5359 if the minimum is greater than zero, just ignore the repeat. If the
5360 maximum is not not zero or one, set it to 1. */
5361
5362 if (*previous < OP_ONCE) /* Assertion */
5363 {
5364 if (repeat_min > 0) goto END_REPEAT;
5365 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5366 }
5367
5368 /* The case of a zero minimum is special because of the need to stick
5369 OP_BRAZERO in front of it, and because the group appears once in the
5370 data, whereas in other cases it appears the minimum number of times. For
5371 this reason, it is simplest to treat this case separately, as otherwise
5372 the code gets far too messy. There are several special subcases when the
5373 minimum is zero. */
5374
5375 if (repeat_min == 0)
5376 {
5377 /* If the maximum is also zero, we used to just omit the group from the
5378 output altogether, like this:
5379
5380 ** if (repeat_max == 0)
5381 ** {
5382 ** code = previous;
5383 ** goto END_REPEAT;
5384 ** }
5385
5386 However, that fails when a group or a subgroup within it is referenced
5387 as a subroutine from elsewhere in the pattern, so now we stick in
5388 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5389 don't have a list of which groups are referenced, we cannot do this
5390 selectively.
5391
5392 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5393 and do no more at this point. However, we do need to adjust any
5394 OP_RECURSE calls inside the group that refer to the group itself or any
5395 internal or forward referenced group, because the offset is from the
5396 start of the whole regex. Temporarily terminate the pattern while doing
5397 this. */
5398
5399 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5400 {
5401 *code = OP_END;
5402 adjust_recurse(previous, 1, utf, cd, save_hwm);
5403 memmove(previous + 1, previous, IN_UCHARS(len));
5404 code++;
5405 if (repeat_max == 0)
5406 {
5407 *previous++ = OP_SKIPZERO;
5408 goto END_REPEAT;
5409 }
5410 brazeroptr = previous; /* Save for possessive optimizing */
5411 *previous++ = OP_BRAZERO + repeat_type;
5412 }
5413
5414 /* If the maximum is greater than 1 and limited, we have to replicate
5415 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
5416 The first one has to be handled carefully because it's the original
5417 copy, which has to be moved up. The remainder can be handled by code
5418 that is common with the non-zero minimum case below. We have to
5419 adjust the value or repeat_max, since one less copy is required. Once
5420 again, we may have to adjust any OP_RECURSE calls inside the group. */
5421
5422 else
5423 {
5424 int offset;
5425 *code = OP_END;
5426 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5427 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5428 code += 2 + LINK_SIZE;
5429 *previous++ = OP_BRAZERO + repeat_type;
5430 *previous++ = OP_BRA;
5431
5432 /* We chain together the bracket offset fields that have to be
5433 filled in later when the ends of the brackets are reached. */
5434
5435 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
5436 bralink = previous;
5437 PUTINC(previous, 0, offset);
5438 }
5439
5440 repeat_max--;
5441 }
5442
5443 /* If the minimum is greater than zero, replicate the group as many
5444 times as necessary, and adjust the maximum to the number of subsequent
5445 copies that we need. If we set a first char from the group, and didn't
5446 set a required char, copy the latter from the former. If there are any
5447 forward reference subroutine calls in the group, there will be entries on
5448 the workspace list; replicate these with an appropriate increment. */
5449
5450 else
5451 {
5452 if (repeat_min > 1)
5453 {
5454 /* In the pre-compile phase, we don't actually do the replication. We
5455 just adjust the length as if we had. Do some paranoid checks for
5456 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
5457 integer type when available, otherwise double. */
5458
5459 if (lengthptr != NULL)
5460 {
5461 int delta = (repeat_min - 1)*length_prevgroup;
5462 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
5463 (INT64_OR_DOUBLE)length_prevgroup >
5464 (INT64_OR_DOUBLE)INT_MAX ||
5465 OFLOW_MAX - *lengthptr < delta)
5466 {
5467 *errorcodeptr = ERR20;
5468 goto FAILED;
5469 }
5470 *lengthptr += delta;
5471 }
5472
5473 /* This is compiling for real. If there is a set first byte for
5474 the group, and we have not yet set a "required byte", set it. Make
5475 sure there is enough workspace for copying forward references before
5476 doing the copy. */
5477
5478 else
5479 {
5480 if (groupsetfirstchar && reqcharflags < 0)
5481 {
5482 reqchar = firstchar;
5483 reqcharflags = firstcharflags;
5484 }
5485
5486 for (i = 1; i < repeat_min; i++)
5487 {
5488 pcre_uchar *hc;
5489 pcre_uchar *this_hwm = cd->hwm;
5490 memcpy(code, previous, IN_UCHARS(len));
5491
5492 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5493 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5494 {
5495 int save_offset = save_hwm - cd->start_workspace;
5496 int this_offset = this_hwm - cd->start_workspace;
5497 *errorcodeptr = expand_workspace(cd);
5498 if (*errorcodeptr != 0) goto FAILED;
5499 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5500 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5501 }
5502
5503 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5504 {
5505 PUT(cd->hwm, 0, GET(hc, 0) + len);
5506 cd->hwm += LINK_SIZE;
5507 }
5508 save_hwm = this_hwm;
5509 code += len;
5510 }
5511 }
5512 }
5513
5514 if (repeat_max > 0) repeat_max -= repeat_min;
5515 }
5516
5517 /* This code is common to both the zero and non-zero minimum cases. If
5518 the maximum is limited, it replicates the group in a nested fashion,
5519 remembering the bracket starts on a stack. In the case of a zero minimum,
5520 the first one was set up above. In all cases the repeat_max now specifies
5521 the number of additional copies needed. Again, we must remember to
5522 replicate entries on the forward reference list. */
5523
5524 if (repeat_max >= 0)
5525 {
5526 /* In the pre-compile phase, we don't actually do the replication. We
5527 just adjust the length as if we had. For each repetition we must add 1
5528 to the length for BRAZERO and for all but the last repetition we must
5529 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
5530 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
5531 a 64-bit integer type when available, otherwise double. */
5532
5533 if (lengthptr != NULL && repeat_max > 0)
5534 {
5535 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
5536 2 - 2*LINK_SIZE; /* Last one doesn't nest */
5537 if ((INT64_OR_DOUBLE)repeat_max *
5538 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
5539 > (INT64_OR_DOUBLE)INT_MAX ||
5540 OFLOW_MAX - *lengthptr < delta)
5541 {
5542 *errorcodeptr = ERR20;
5543 goto FAILED;
5544 }
5545 *lengthptr += delta;
5546 }
5547
5548 /* This is compiling for real */
5549
5550 else for (i = repeat_max - 1; i >= 0; i--)
5551 {
5552 pcre_uchar *hc;
5553 pcre_uchar *this_hwm = cd->hwm;
5554
5555 *code++ = OP_BRAZERO + repeat_type;
5556
5557 /* All but the final copy start a new nesting, maintaining the
5558 chain of brackets outstanding. */
5559
5560 if (i != 0)
5561 {
5562 int offset;
5563 *code++ = OP_BRA;
5564 offset = (bralink == NULL)? 0 : (int)(code - bralink);
5565 bralink = code;
5566 PUTINC(code, 0, offset);
5567 }
5568
5569 memcpy(code, previous, IN_UCHARS(len));
5570
5571 /* Ensure there is enough workspace for forward references before
5572 copying them. */
5573
5574 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5575 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5576 {
5577 int save_offset = save_hwm - cd->start_workspace;
5578 int this_offset = this_hwm - cd->start_workspace;
5579 *errorcodeptr = expand_workspace(cd);
5580 if (*errorcodeptr != 0) goto FAILED;
5581 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5582 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5583 }
5584
5585 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5586 {
5587 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
5588 cd->hwm += LINK_SIZE;
5589 }
5590 save_hwm = this_hwm;
5591 code += len;
5592 }
5593
5594 /* Now chain through the pending brackets, and fill in their length
5595 fields (which are holding the chain links pro tem). */
5596
5597 while (bralink != NULL)
5598 {
5599 int oldlinkoffset;
5600 int offset = (int)(code - bralink + 1);
5601 pcre_uchar *bra = code - offset;
5602 oldlinkoffset = GET(bra, 1);
5603 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
5604 *code++ = OP_KET;
5605 PUTINC(code, 0, offset);
5606 PUT(bra, 1, offset);
5607 }
5608 }
5609
5610 /* If the maximum is unlimited, set a repeater in the final copy. For
5611 ONCE brackets, that's all we need to do. However, possessively repeated
5612 ONCE brackets can be converted into non-capturing brackets, as the
5613 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5614 deal with possessive ONCEs specially.
5615
5616 Otherwise, when we are doing the actual compile phase, check to see
5617 whether this group is one that could match an empty string. If so,
5618 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5619 that runtime checking can be done. [This check is also applied to ONCE
5620 groups at runtime, but in a different way.]
5621
5622 Then, if the quantifier was possessive and the bracket is not a
5623 conditional, we convert the BRA code to the POS form, and the KET code to
5624 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5625 subpattern at both the start and at the end.) The use of special opcodes
5626 makes it possible to reduce greatly the stack usage in pcre_exec(). If
5627 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5628
5629 Then, if the minimum number of matches is 1 or 0, cancel the possessive
5630 flag so that the default action below, of wrapping everything inside
5631 atomic brackets, does not happen. When the minimum is greater than 1,
5632 there will be earlier copies of the group, and so we still have to wrap
5633 the whole thing. */
5634
5635 else
5636 {
5637 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
5638 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
5639
5640 /* Convert possessive ONCE brackets to non-capturing */
5641
5642 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5643 possessive_quantifier) *bracode = OP_BRA;
5644
5645 /* For non-possessive ONCE brackets, all we need to do is to
5646 set the KET. */
5647
5648 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5649 *ketcode = OP_KETRMAX + repeat_type;
5650
5651 /* Handle non-ONCE brackets and possessive ONCEs (which have been
5652 converted to non-capturing above). */
5653
5654 else
5655 {
5656 /* In the compile phase, check for empty string matching. */
5657
5658 if (lengthptr == NULL)
5659 {
5660 pcre_uchar *scode = bracode;
5661 do
5662 {
5663 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
5664 {
5665 *bracode += OP_SBRA - OP_BRA;
5666 break;
5667 }
5668 scode += GET(scode, 1);
5669 }
5670 while (*scode == OP_ALT);
5671 }
5672
5673 /* Handle possessive quantifiers. */
5674
5675 if (possessive_quantifier)
5676 {
5677 /* For COND brackets, we wrap the whole thing in a possessively
5678 repeated non-capturing bracket, because we have not invented POS
5679 versions of the COND opcodes. Because we are moving code along, we
5680 must ensure that any pending recursive references are updated. */
5681
5682 if (*bracode == OP_COND || *bracode == OP_SCOND)
5683 {
5684 int nlen = (int)(code - bracode);
5685 *code = OP_END;
5686 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
5687 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
5688 code += 1 + LINK_SIZE;
5689 nlen += 1 + LINK_SIZE;
5690 *bracode = OP_BRAPOS;
5691 *code++ = OP_KETRPOS;
5692 PUTINC(code, 0, nlen);
5693 PUT(bracode, 1, nlen);
5694 }
5695
5696 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
5697
5698 else
5699 {
5700 *bracode += 1; /* Switch to xxxPOS opcodes */
5701 *ketcode = OP_KETRPOS;
5702 }
5703
5704 /* If the minimum is zero, mark it as possessive, then unset the
5705 possessive flag when the minimum is 0 or 1. */
5706
5707 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
5708 if (repeat_min < 2) possessive_quantifier = FALSE;
5709 }
5710
5711 /* Non-possessive quantifier */
5712
5713 else *ketcode = OP_KETRMAX + repeat_type;
5714 }
5715 }
5716 }
5717
5718 /* If previous is OP_FAIL, it was generated by an empty class [] in
5719 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
5720 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
5721 error above. We can just ignore the repeat in JS case. */
5722
5723 else if (*previous == OP_FAIL) goto END_REPEAT;
5724
5725 /* Else there's some kind of shambles */
5726
5727 else
5728 {
5729 *errorcodeptr = ERR11;
5730 goto FAILED;
5731 }
5732
5733 /* If the character following a repeat is '+', or if certain optimization
5734 tests above succeeded, possessive_quantifier is TRUE. For some opcodes,
5735 there are special alternative opcodes for this case. For anything else, we
5736 wrap the entire repeated item inside OP_ONCE brackets. Logically, the '+'
5737 notation is just syntactic sugar, taken from Sun's Java package, but the
5738 special opcodes can optimize it.
5739
5740 Some (but not all) possessively repeated subpatterns have already been
5741 completely handled in the code just above. For them, possessive_quantifier
5742 is always FALSE at this stage.
5743
5744 Note that the repeated item starts at tempcode, not at previous, which
5745 might be the first part of a string whose (former) last char we repeated.
5746
5747 Possessifying an 'exact' quantifier has no effect, so we can ignore it. But
5748 an 'upto' may follow. We skip over an 'exact' item, and then test the
5749 length of what remains before proceeding. */
5750
5751 if (possessive_quantifier)
5752 {
5753 int len;
5754
5755 if (*tempcode == OP_TYPEEXACT)
5756 tempcode += PRIV(OP_lengths)[*tempcode] +
5757 ((tempcode[1 + IMM2_SIZE] == OP_PROP
5758 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
5759
5760 else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
5761 {
5762 tempcode += PRIV(OP_lengths)[*tempcode];
5763 #ifdef SUPPORT_UTF
5764 if (utf && HAS_EXTRALEN(tempcode[-1]))
5765 tempcode += GET_EXTRALEN(tempcode[-1]);
5766 #endif
5767 }
5768
5769 len = (int)(code - tempcode);
5770 if (len > 0) switch (*tempcode)
5771 {
5772 case OP_STAR: *tempcode = OP_POSSTAR; break;
5773 case OP_PLUS: *tempcode = OP_POSPLUS; break;
5774 case OP_QUERY: *tempcode = OP_POSQUERY; break;
5775 case OP_UPTO: *tempcode = OP_POSUPTO; break;
5776
5777 case OP_STARI: *tempcode = OP_POSSTARI; break;
5778 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
5779 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
5780 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
5781
5782 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
5783 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
5784 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
5785 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
5786
5787 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
5788 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
5789 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
5790 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
5791
5792 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
5793 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
5794 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
5795 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
5796
5797 /* Because we are moving code along, we must ensure that any
5798 pending recursive references are updated. */
5799
5800 default:
5801 *code = OP_END;
5802 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
5803 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
5804 code += 1 + LINK_SIZE;
5805 len += 1 + LINK_SIZE;
5806 tempcode[0] = OP_ONCE;
5807 *code++ = OP_KET;
5808 PUTINC(code, 0, len);
5809 PUT(tempcode, 1, len);
5810 break;
5811 }
5812 }
5813
5814 /* In all case we no longer have a previous item. We also set the
5815 "follows varying string" flag for subsequently encountered reqchars if
5816 it isn't already set and we have just passed a varying length item. */
5817
5818 END_REPEAT:
5819 previous = NULL;
5820 cd->req_varyopt |= reqvary;
5821 break;
5822
5823
5824 /* ===================================================================*/
5825 /* Start of nested parenthesized sub-expression, or comment or lookahead or
5826 lookbehind or option setting or condition or all the other extended
5827 parenthesis forms. */
5828
5829 case CHAR_LEFT_PARENTHESIS:
5830 newoptions = options;
5831 skipbytes = 0;
5832 bravalue = OP_CBRA;
5833 save_hwm = cd->hwm;
5834 reset_bracount = FALSE;
5835
5836 /* First deal with various "verbs" that can be introduced by '*'. */
5837
5838 ptr++;
5839 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
5840 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
5841 {
5842 int i, namelen;
5843 int arglen = 0;
5844 const char *vn = verbnames;
5845 const pcre_uchar *name = ptr + 1;
5846 const pcre_uchar *arg = NULL;
5847 previous = NULL;
5848 ptr++;
5849 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
5850 namelen = (int)(ptr - name);
5851
5852 /* It appears that Perl allows any characters whatsoever, other than
5853 a closing parenthesis, to appear in arguments, so we no longer insist on
5854 letters, digits, and underscores. */
5855
5856 if (*ptr == CHAR_COLON)
5857 {
5858 arg = ++ptr;
5859 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5860 arglen = (int)(ptr - arg);
5861 if ((unsigned int)arglen > MAX_MARK)
5862 {
5863 *errorcodeptr = ERR75;
5864 goto FAILED;
5865 }
5866 }
5867
5868 if (*ptr != CHAR_RIGHT_PARENTHESIS)
5869 {
5870 *errorcodeptr = ERR60;
5871 goto FAILED;
5872 }
5873
5874 /* Scan the table of verb names */
5875
5876 for (i = 0; i < verbcount; i++)
5877 {
5878 if (namelen == verbs[i].len &&
5879 STRNCMP_UC_C8(name, vn, namelen) == 0)
5880 {
5881 int setverb;
5882
5883 /* Check for open captures before ACCEPT and convert it to
5884 ASSERT_ACCEPT if in an assertion. */
5885
5886 if (verbs[i].op == OP_ACCEPT)
5887 {
5888 open_capitem *oc;
5889 if (arglen != 0)
5890 {
5891 *errorcodeptr = ERR59;
5892 goto FAILED;
5893 }
5894 cd->had_accept = TRUE;
5895 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5896 {
5897 *code++ = OP_CLOSE;
5898 PUT2INC(code, 0, oc->number);
5899 }
5900 setverb = *code++ =
5901 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
5902
5903 /* Do not set firstchar after *ACCEPT */
5904 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5905 }
5906
5907 /* Handle other cases with/without an argument */
5908
5909 else if (arglen == 0)
5910 {
5911 if (verbs[i].op < 0) /* Argument is mandatory */
5912 {
5913 *errorcodeptr = ERR66;
5914 goto FAILED;
5915 }
5916 setverb = *code++ = verbs[i].op;
5917 }
5918
5919 else
5920 {
5921 if (verbs[i].op_arg < 0) /* Argument is forbidden */
5922 {
5923 *errorcodeptr = ERR59;
5924 goto FAILED;
5925 }
5926 setverb = *code++ = verbs[i].op_arg;
5927 *code++ = arglen;
5928 memcpy(code, arg, IN_UCHARS(arglen));
5929 code += arglen;
5930 *code++ = 0;
5931 }
5932
5933 switch (setverb)
5934 {
5935 case OP_THEN:
5936 case OP_THEN_ARG:
5937 cd->external_flags |= PCRE_HASTHEN;
5938 break;
5939
5940 case OP_PRUNE:
5941 case OP_PRUNE_ARG:
5942 case OP_SKIP:
5943 case OP_SKIP_ARG:
5944 cd->had_pruneorskip = TRUE;
5945 break;
5946 }
5947
5948 break; /* Found verb, exit loop */
5949 }
5950
5951 vn += verbs[i].len + 1;
5952 }
5953
5954 if (i < verbcount) continue; /* Successfully handled a verb */
5955 *errorcodeptr = ERR60; /* Verb not recognized */
5956 goto FAILED;
5957 }
5958
5959 /* Deal with the extended parentheses; all are introduced by '?', and the
5960 appearance of any of them means that this is not a capturing group. */
5961
5962 else if (*ptr == CHAR_QUESTION_MARK)
5963 {
5964 int i, set, unset, namelen;
5965 int *optset;
5966 const pcre_uchar *name;
5967 pcre_uchar *slot;
5968
5969 switch (*(++ptr))
5970 {
5971 case CHAR_NUMBER_SIGN: /* Comment; skip to ket */
5972 ptr++;
5973 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
5974 if (*ptr == CHAR_NULL)
5975 {
5976 *errorcodeptr = ERR18;
5977 goto FAILED;
5978 }
5979 continue;
5980
5981
5982 /* ------------------------------------------------------------ */
5983 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
5984 reset_bracount = TRUE;
5985 /* Fall through */
5986
5987 /* ------------------------------------------------------------ */
5988 case CHAR_COLON: /* Non-capturing bracket */
5989 bravalue = OP_BRA;
5990 ptr++;
5991 break;
5992
5993
5994 /* ------------------------------------------------------------ */
5995 case CHAR_LEFT_PARENTHESIS:
5996 bravalue = OP_COND; /* Conditional group */
5997 tempptr = ptr;
5998
5999 /* A condition can be an assertion, a number (referring to a numbered
6000 group), a name (referring to a named group), or 'R', referring to
6001 recursion. R<digits> and R&name are also permitted for recursion tests.
6002
6003 There are several syntaxes for testing a named group: (?(name)) is used
6004 by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
6005
6006 There are two unfortunate ambiguities, caused by history. (a) 'R' can
6007 be the recursive thing or the name 'R' (and similarly for 'R' followed
6008 by digits), and (b) a number could be a name that consists of digits.
6009 In both cases, we look for a name first; if not found, we try the other
6010 cases.
6011
6012 For compatibility with auto-callouts, we allow a callout to be
6013 specified before a condition that is an assertion. First, check for the
6014 syntax of a callout; if found, adjust the temporary pointer that is
6015 used to check for an assertion condition. That's all that is needed! */
6016
6017 if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6018 {
6019 for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6020 if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6021 tempptr += i + 1;
6022 }
6023
6024 /* For conditions that are assertions, check the syntax, and then exit
6025 the switch. This will take control down to where bracketed groups,
6026 including assertions, are processed. */
6027
6028 if (tempptr[1] == CHAR_QUESTION_MARK &&
6029 (tempptr[2] == CHAR_EQUALS_SIGN ||
6030 tempptr[2] == CHAR_EXCLAMATION_MARK ||
6031 tempptr[2] == CHAR_LESS_THAN_SIGN))
6032 break;
6033
6034 /* Most other conditions use OP_CREF (a couple change to OP_RREF
6035 below), and all need to skip 1+IMM2_SIZE bytes at the start of the group. */
6036
6037 code[1+LINK_SIZE] = OP_CREF;
6038 skipbytes = 1+IMM2_SIZE;
6039 refsign = -1;
6040
6041 /* Check for a test for recursion in a named group. */
6042
6043 if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
6044 {
6045 terminator = -1;
6046 ptr += 2;
6047 code[1+LINK_SIZE] = OP_RREF; /* Change the type of test */
6048 }
6049
6050 /* Check for a test for a named group's having been set, using the Perl
6051 syntax (?(<name>) or (?('name') */
6052
6053 else if (ptr[1] == CHAR_LESS_THAN_SIGN)
6054 {
6055 terminator = CHAR_GREATER_THAN_SIGN;
6056 ptr++;
6057 }
6058 else if (ptr[1] == CHAR_APOSTROPHE)
6059 {
6060 terminator = CHAR_APOSTROPHE;
6061 ptr++;
6062 }
6063 else
6064 {
6065 terminator = CHAR_NULL;
6066 if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
6067 }
6068
6069 /* We now expect to read a name; any thing else is an error */
6070
6071 if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)
6072 {
6073 ptr += 1; /* To get the right offset */
6074 *errorcodeptr = ERR28;
6075 goto FAILED;
6076 }
6077
6078 /* Read the name, but also get it as a number if it's all digits */
6079
6080 recno = 0;
6081 name = ++ptr;
6082 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6083 {
6084 if (recno >= 0)
6085 recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;
6086 ptr++;
6087 }
6088 namelen = (int)(ptr - name);
6089
6090 if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6091 *ptr++ != CHAR_RIGHT_PARENTHESIS)
6092 {
6093 ptr--; /* Error offset */
6094 *errorcodeptr = ERR26;
6095 goto FAILED;
6096 }
6097
6098 /* Do no further checking in the pre-compile phase. */
6099
6100 if (lengthptr != NULL) break;
6101
6102 /* In the real compile we do the work of looking for the actual
6103 reference. If the string started with "+" or "-" we require the rest to
6104 be digits, in which case recno will be set. */
6105
6106 if (refsign > 0)
6107 {
6108 if (recno <= 0)
6109 {
6110 *errorcodeptr = ERR58;
6111 goto FAILED;
6112 }
6113 recno = (refsign == CHAR_MINUS)?
6114 cd->bracount - recno + 1 : recno +cd->bracount;
6115 if (recno <= 0 || recno > cd->final_bracount)
6116 {
6117 *errorcodeptr = ERR15;
6118 goto FAILED;
6119 }
6120 PUT2(code, 2+LINK_SIZE, recno);
6121 break;
6122 }
6123
6124 /* Otherwise (did not start with "+" or "-"), start by looking for the
6125 name. If we find a name, add one to the opcode to change OP_CREF or
6126 OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
6127 except they record that the reference was originally to a name. The
6128 information is used to check duplicate names. */
6129
6130 slot = cd->name_table;
6131 for (i = 0; i < cd->names_found; i++)
6132 {
6133 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
6134 slot += cd->name_entry_size;
6135 }
6136
6137 /* Found the named subpattern */
6138
6139 if (i < cd->names_found)
6140 {
6141 recno = GET2(slot, 0);
6142 PUT2(code, 2+LINK_SIZE, recno);
6143 code[1+LINK_SIZE]++;
6144 }
6145
6146 /* If terminator == CHAR_NULL it means that the name followed directly
6147 after the opening parenthesis [e.g. (?(abc)...] and in this case there
6148 are some further alternatives to try. For the cases where terminator !=
6149 0 [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
6150 now checked all the possibilities, so give an error. */
6151
6152 else if (terminator != CHAR_NULL)
6153 {
6154 *errorcodeptr = ERR15;
6155 goto FAILED;
6156 }
6157
6158 /* Check for (?(R) for recursion. Allow digits after R to specify a
6159 specific group number. */
6160
6161 else if (*name == CHAR_R)
6162 {
6163 recno = 0;
6164 for (i = 1; i < namelen; i++)
6165 {
6166 if (!IS_DIGIT(name[i]))
6167 {
6168 *errorcodeptr = ERR15;
6169 goto FAILED;
6170 }
6171 recno = recno * 10 + name[i] - CHAR_0;
6172 }
6173 if (recno == 0) recno = RREF_ANY;
6174 code[1+LINK_SIZE] = OP_RREF; /* Change test type */
6175 PUT2(code, 2+LINK_SIZE, recno);
6176 }
6177
6178 /* Similarly, check for the (?(DEFINE) "condition", which is always
6179 false. */
6180
6181 else if (namelen == 6 && STRNCMP_UC_C8(name, STRING_DEFINE, 6) == 0)
6182 {
6183 code[1+LINK_SIZE] = OP_DEF;
6184 skipbytes = 1;
6185 }
6186
6187 /* Check for the "name" actually being a subpattern number. We are
6188 in the second pass here, so final_bracount is set. */
6189
6190 else if (recno > 0 && recno <= cd->final_bracount)
6191 {
6192 PUT2(code, 2+LINK_SIZE, recno);
6193 }
6194
6195 /* Either an unidentified subpattern, or a reference to (?(0) */
6196
6197 else
6198 {
6199 *errorcodeptr = (recno == 0)? ERR35: ERR15;
6200 goto FAILED;
6201 }
6202 break;
6203
6204
6205 /* ------------------------------------------------------------ */
6206 case CHAR_EQUALS_SIGN: /* Positive lookahead */
6207 bravalue = OP_ASSERT;
6208 cd->assert_depth += 1;
6209 ptr++;
6210 break;
6211
6212
6213 /* ------------------------------------------------------------ */
6214 case CHAR_EXCLAMATION_MARK: /* Negative lookahead */
6215 ptr++;
6216 if (*ptr == CHAR_RIGHT_PARENTHESIS) /* Optimize (?!) */
6217 {
6218 *code++ = OP_FAIL;
6219 previous = NULL;
6220 continue;
6221 }
6222 bravalue = OP_ASSERT_NOT;
6223 cd->assert_depth += 1;
6224 break;
6225
6226
6227 /* ------------------------------------------------------------ */
6228 case CHAR_LESS_THAN_SIGN: /* Lookbehind or named define */
6229 switch (ptr[1])
6230 {
6231 case CHAR_EQUALS_SIGN: /* Positive lookbehind */
6232 bravalue = OP_ASSERTBACK;
6233 cd->assert_depth += 1;
6234 ptr += 2;
6235 break;
6236
6237 case CHAR_EXCLAMATION_MARK: /* Negative lookbehind */
6238 bravalue = OP_ASSERTBACK_NOT;
6239 cd->assert_depth += 1;
6240 ptr += 2;
6241 break;
6242
6243 default: /* Could be name define, else bad */
6244 if (MAX_255(ptr[1]) && (cd->ctypes[ptr[1]] & ctype_word) != 0)
6245 goto DEFINE_NAME;
6246 ptr++; /* Correct offset for error */
6247 *errorcodeptr = ERR24;
6248 goto FAILED;
6249 }
6250 break;
6251
6252
6253 /* ------------------------------------------------------------ */
6254 case CHAR_GREATER_THAN_SIGN: /* One-time brackets */
6255 bravalue = OP_ONCE;
6256 ptr++;
6257 break;
6258
6259
6260 /* ------------------------------------------------------------ */
6261 case CHAR_C: /* Callout - may be followed by digits; */
6262 previous_callout = code; /* Save for later completion */
6263 after_manual_callout = 1; /* Skip one item before completing */
6264 *code++ = OP_CALLOUT;
6265 {
6266 int n = 0;
6267 ptr++;
6268 while(IS_DIGIT(*ptr))
6269 n = n * 10 + *ptr++ - CHAR_0;
6270 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6271 {
6272 *errorcodeptr = ERR39;
6273 goto FAILED;
6274 }
6275 if (n > 255)
6276 {
6277 *errorcodeptr = ERR38;
6278 goto FAILED;
6279 }
6280 *code++ = n;
6281 PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
6282 PUT(code, LINK_SIZE, 0); /* Default length */
6283 code += 2 * LINK_SIZE;
6284 }
6285 previous = NULL;
6286 continue;
6287
6288
6289 /* ------------------------------------------------------------ */
6290 case CHAR_P: /* Python-style named subpattern handling */
6291 if (*(++ptr) == CHAR_EQUALS_SIGN ||
6292 *ptr == CHAR_GREATER_THAN_SIGN) /* Reference or recursion */
6293 {
6294 is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
6295 terminator = CHAR_RIGHT_PARENTHESIS;
6296 goto NAMED_REF_OR_RECURSE;
6297 }
6298 else if (*ptr != CHAR_LESS_THAN_SIGN) /* Test for Python-style defn */
6299 {
6300 *errorcodeptr = ERR41;
6301 goto FAILED;
6302 }
6303 /* Fall through to handle (?P< as (?< is handled */
6304
6305
6306 /* ------------------------------------------------------------ */
6307 DEFINE_NAME: /* Come here from (?< handling */
6308 case CHAR_APOSTROPHE:
6309 terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
6310 CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
6311 name = ++ptr;
6312
6313 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
6314 namelen = (int)(ptr - name);
6315
6316 /* In the pre-compile phase, do a syntax check, remember the longest
6317 name, and then remember the group in a vector, expanding it if
6318 necessary. Duplicates for the same number are skipped; other duplicates
6319 are checked for validity. In the actual compile, there is nothing to
6320 do. */
6321
6322 if (lengthptr != NULL)
6323 {
6324 named_group *ng;
6325 pcre_uint32 number = cd->bracount + 1;
6326
6327 if (*ptr != (pcre_uchar)terminator)
6328 {
6329 *errorcodeptr = ERR42;
6330 goto FAILED;
6331 }
6332
6333 if (cd->names_found >= MAX_NAME_COUNT)
6334 {
6335 *errorcodeptr = ERR49;
6336 goto FAILED;
6337 }
6338
6339 if (namelen + IMM2_SIZE + 1 > cd->name_entry_size)
6340 {
6341 cd->name_entry_size = namelen + IMM2_SIZE + 1;
6342 if (namelen > MAX_NAME_SIZE)
6343 {
6344 *errorcodeptr = ERR48;
6345 goto FAILED;
6346 }
6347 }
6348
6349 /* Scan the list to check for duplicates. For duplicate names, if the
6350 number is the same, break the loop, which causes the name to be
6351 discarded; otherwise, if DUPNAMES is not set, give an error.
6352 If it is set, allow the name with a different number, but continue
6353 scanning in case this is a duplicate with the same number. For
6354 non-duplicate names, give an error if the number is duplicated. */
6355
6356 ng = cd->named_groups;
6357 for (i = 0; i < cd->names_found; i++, ng++)
6358 {
6359 if (namelen == ng->length &&
6360 STRNCMP_UC_UC(name, ng->name, namelen) == 0)
6361 {
6362 if (ng->number == number) break;
6363 if ((options & PCRE_DUPNAMES) == 0)
6364 {
6365 *errorcodeptr = ERR43;
6366 goto FAILED;
6367 }
6368 cd->dupnames = TRUE; /* Duplicate names exist */
6369 }
6370 else if (ng->number == number)
6371 {
6372 *errorcodeptr = ERR65;
6373 goto FAILED;
6374 }
6375 }
6376
6377 if (i >= cd->names_found) /* Not a duplicate with same number */
6378 {
6379 /* Increase the list size if necessary */
6380
6381 if (cd->names_found >= cd->named_group_list_size)
6382 {
6383 int newsize = cd->named_group_list_size * 2;
6384 named_group *newspace = (PUBL(malloc))
6385 (newsize * sizeof(named_group));
6386
6387 if (newspace == NULL)
6388 {
6389 *errorcodeptr = ERR21;
6390 goto FAILED;
6391 }
6392
6393 memcpy(newspace, cd->named_groups,
6394 cd->named_group_list_size * sizeof(named_group));
6395 if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
6396 (PUBL(free))((void *)cd->named_groups);
6397 cd->named_groups = newspace;
6398 cd->named_group_list_size = newsize;
6399 }
6400
6401 cd->named_groups[cd->names_found].name = name;
6402 cd->named_groups[cd->names_found].length = namelen;
6403 cd->named_groups[cd->names_found].number = number;
6404 cd->names_found++;
6405 }
6406 }
6407
6408 ptr++; /* Move past > or ' in both passes. */
6409 goto NUMBERED_GROUP;
6410
6411
6412 /* ------------------------------------------------------------ */
6413 case CHAR_AMPERSAND: /* Perl recursion/subroutine syntax */
6414 terminator = CHAR_RIGHT_PARENTHESIS;
6415 is_recurse = TRUE;
6416 /* Fall through */
6417
6418 /* We come here from the Python syntax above that handles both
6419 references (?P=name) and recursion (?P>name), as well as falling
6420 through from the Perl recursion syntax (?&name). We also come here from
6421 the Perl \k<name> or \k'name' back reference syntax and the \k{name}
6422 .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
6423
6424 NAMED_REF_OR_RECURSE:
6425 name = ++ptr;
6426 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
6427 namelen = (int)(ptr - name);
6428
6429 /* In the pre-compile phase, do a syntax check. We used to just set
6430 a dummy reference number, because it was not used in the first pass.
6431 However, with the change of recursive back references to be atomic,
6432 we have to look for the number so that this state can be identified, as
6433 otherwise the incorrect length is computed. If it's not a backwards
6434 reference, the dummy number will do. */
6435
6436 if (lengthptr != NULL)
6437 {
6438 named_group *ng;
6439
6440 if (namelen == 0)
6441 {
6442 *errorcodeptr = ERR62;
6443 goto FAILED;
6444 }
6445 if (*ptr != (pcre_uchar)terminator)
6446 {
6447 *errorcodeptr = ERR42;
6448 goto FAILED;
6449 }
6450 if (namelen > MAX_NAME_SIZE)
6451 {
6452 *errorcodeptr = ERR48;
6453 goto FAILED;
6454 }
6455
6456 /* The name table does not exist in the first pass; instead we must
6457 scan the list of names encountered so far in order to get the
6458 number. If the name is not found, set the value to 0 for a forward
6459 reference. */
6460
6461 ng = cd->named_groups;
6462 for (i = 0; i < cd->names_found; i++, ng++)
6463 {
6464 if (namelen == ng->length &&
6465 STRNCMP_UC_UC(name, ng->name, namelen) == 0)
6466 break;
6467 }
6468 recno = (i < cd->names_found)? ng->number : 0;
6469
6470 /* Count named back references. */
6471
6472 if (!is_recurse) cd->namedrefcount++;
6473 }
6474
6475 /* In the real compile, search the name table. We check the name
6476 first, and then check that we have reached the end of the name in the
6477 table. That way, if the name is longer than any in the table, the
6478 comparison will fail without reading beyond the table entry. */
6479
6480 else
6481 {
6482 slot = cd->name_table;
6483 for (i = 0; i < cd->names_found; i++)
6484 {
6485 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0 &&
6486 slot[IMM2_SIZE+namelen] == 0)
6487 break;
6488 slot += cd->name_entry_size;
6489 }
6490
6491 if (i < cd->names_found)
6492 {
6493 recno = GET2(slot, 0);
6494 }
6495 else
6496 {
6497 *errorcodeptr = ERR15;
6498 goto FAILED;
6499 }
6500 }
6501
6502 /* In both phases, for recursions, we can now go to the code than
6503 handles numerical recursion. */
6504
6505 if (is_recurse) goto HANDLE_RECURSION;
6506
6507 /* In the second pass we must see if the name is duplicated. If so, we
6508 generate a different opcode. */
6509
6510 if (lengthptr == NULL && cd->dupnames)
6511 {
6512 int count = 1;
6513 unsigned int index = i;
6514 pcre_uchar *cslot = slot + cd->name_entry_size;
6515
6516 for (i++; i < cd->names_found; i++)
6517 {
6518 if (STRCMP_UC_UC(slot + IMM2_SIZE, cslot + IMM2_SIZE) != 0) break;
6519 count++;
6520 cslot += cd->name_entry_size;
6521 }
6522
6523 if (count > 1)
6524 {
6525 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6526 previous = code;
6527 *code++ = ((options & PCRE_CASELESS) != 0)? OP_DNREFI : OP_DNREF;
6528 PUT2INC(code, 0, index);
6529 PUT2INC(code, 0, count);
6530
6531 /* Process each potentially referenced group. */
6532
6533 for (; slot < cslot; slot += cd->name_entry_size)
6534 {
6535 open_capitem *oc;
6536 recno = GET2(slot, 0);
6537 cd->backref_map |= (recno < 32)? (1 << recno) : 1;
6538 if (recno > cd->top_backref) cd->top_backref = recno;
6539
6540 /* Check to see if this back reference is recursive, that it, it
6541 is inside the group that it references. A flag is set so that the
6542 group can be made atomic. */
6543
6544 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6545 {
6546 if (oc->number == recno)
6547 {
6548 oc->flag = TRUE;
6549 break;
6550 }
6551 }
6552 }
6553
6554 continue; /* End of back ref handling */
6555 }
6556 }
6557
6558 /* First pass, or a non-duplicated name. */
6559
6560 goto HANDLE_REFERENCE;
6561
6562
6563 /* ------------------------------------------------------------ */
6564 case CHAR_R: /* Recursion */
6565 ptr++; /* Same as (?0) */
6566 /* Fall through */
6567
6568
6569 /* ------------------------------------------------------------ */
6570 case CHAR_MINUS: case CHAR_PLUS: /* Recursion or subroutine */
6571 case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
6572 case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
6573 {
6574 const pcre_uchar *called;
6575 terminator = CHAR_RIGHT_PARENTHESIS;
6576
6577 /* Come here from the \g<...> and \g'...' code (Oniguruma
6578 compatibility). However, the syntax has been checked to ensure that
6579 the ... are a (signed) number, so that neither ERR63 nor ERR29 will
6580 be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
6581 ever be taken. */
6582
6583 HANDLE_NUMERICAL_RECURSION:
6584
6585 if ((refsign = *ptr) == CHAR_PLUS)
6586 {
6587 ptr++;
6588 if (!IS_DIGIT(*ptr))
6589 {
6590 *errorcodeptr = ERR63;
6591 goto FAILED;
6592 }
6593 }
6594 else if (refsign == CHAR_MINUS)
6595 {
6596 if (!IS_DIGIT(ptr[1]))
6597 goto OTHER_CHAR_AFTER_QUERY;
6598 ptr++;
6599 }
6600
6601 recno = 0;
6602 while(IS_DIGIT(*ptr))
6603 recno = recno * 10 + *ptr++ - CHAR_0;
6604
6605 if (*ptr != (pcre_uchar)terminator)
6606 {
6607 *errorcodeptr = ERR29;
6608 goto FAILED;
6609 }
6610
6611 if (refsign == CHAR_MINUS)
6612 {
6613 if (recno == 0)
6614 {
6615 *errorcodeptr = ERR58;
6616 goto FAILED;
6617 }
6618 recno = cd->bracount - recno + 1;
6619 if (recno <= 0)
6620 {
6621 *errorcodeptr = ERR15;
6622 goto FAILED;
6623 }
6624 }
6625 else if (refsign == CHAR_PLUS)
6626 {
6627 if (recno == 0)
6628 {
6629 *errorcodeptr = ERR58;
6630 goto FAILED;
6631 }
6632 recno += cd->bracount;
6633 }
6634
6635 /* Come here from code above that handles a named recursion */
6636
6637 HANDLE_RECURSION:
6638
6639 previous = code;
6640 called = cd->start_code;
6641
6642 /* When we are actually compiling, find the bracket that is being
6643 referenced. Temporarily end the regex in case it doesn't exist before
6644 this point. If we end up with a forward reference, first check that
6645 the bracket does occur later so we can give the error (and position)
6646 now. Then remember this forward reference in the workspace so it can
6647 be filled in at the end. */
6648
6649 if (lengthptr == NULL)
6650 {
6651 *code = OP_END;
6652 if (recno != 0)
6653 called = PRIV(find_bracket)(cd->start_code, utf, recno);
6654
6655 /* Forward reference */
6656
6657 if (called == NULL)
6658 {
6659