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

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1383 - (show annotations)
Fri Oct 18 17:50:06 2013 UTC (6 years, 1 month ago) by zherczeg
File MIME type: text/plain
File size: 303871 byte(s)
Allow auto-possessify to check more complex bracketed expressions.
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{} or \\o{} 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 "non-hex character in \\x{} (closing brace missing?)\0"
520 /* 80 */
521 "non-octal character in \\o{} (closing brace missing?)\0"
522 "missing opening brace after \\o\0"
523 ;
524
525 /* Table to identify digits and hex digits. This is used when compiling
526 patterns. Note that the tables in chartables are dependent on the locale, and
527 may mark arbitrary characters as digits - but the PCRE compiling code expects
528 to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
529 a private table here. It costs 256 bytes, but it is a lot faster than doing
530 character value tests (at least in some simple cases I timed), and in some
531 applications one wants PCRE to compile efficiently as well as match
532 efficiently.
533
534 For convenience, we use the same bit definitions as in chartables:
535
536 0x04 decimal digit
537 0x08 hexadecimal digit
538
539 Then we can use ctype_digit and ctype_xdigit in the code. */
540
541 /* Using a simple comparison for decimal numbers rather than a memory read
542 is much faster, and the resulting code is simpler (the compiler turns it
543 into a subtraction and unsigned comparison). */
544
545 #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
546
547 #ifndef EBCDIC
548
549 /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
550 UTF-8 mode. */
551
552 static const pcre_uint8 digitab[] =
553 {
554 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
555 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
556 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
557 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
558 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
559 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
560 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
561 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
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 - _ */
566 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
567 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
568 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
569 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
570 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
571 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
572 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
573 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
574 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
575 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
576 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
577 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
578 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
579 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
580 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
581 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
582 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
583 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
584 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
585 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
586
587 #else
588
589 /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
590
591 static const pcre_uint8 digitab[] =
592 {
593 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
594 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
595 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
596 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
597 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
598 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
599 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
600 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
601 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
602 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
603 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
604 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
605 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
606 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
607 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
608 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
609 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
610 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
611 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
612 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
613 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
614 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
615 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
616 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
617 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
618 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
619 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
620 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
621 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
622 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
623 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
624 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
625
626 static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
627 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
628 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
629 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
630 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
631 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
632 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
633 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
634 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
635 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
636 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
637 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
638 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
639 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
640 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
641 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
642 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
643 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
644 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
645 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
646 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
647 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
648 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
649 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
650 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
651 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
652 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
653 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
654 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
655 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
656 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
657 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
658 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
659 #endif
660
661
662 /* This table is used to check whether auto-possessification is possible
663 between adjacent character-type opcodes. The left-hand (repeated) opcode is
664 used to select the row, and the right-hand opcode is use to select the column.
665 A value of 1 means that auto-possessification is OK. For example, the second
666 value in the first row means that \D+\d can be turned into \D++\d.
667
668 The Unicode property types (\P and \p) have to be present to fill out the table
669 because of what their opcode values are, but the table values should always be
670 zero because property types are handled separately in the code. The last four
671 columns apply to items that cannot be repeated, so there is no need to have
672 rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
673 *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
674
675 #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
676 #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
677
678 static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
679 /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
680 { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
681 { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
682 { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
683 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
684 { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
685 { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
686 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
687 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
688 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
689 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
690 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
691 { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
692 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
693 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
694 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
695 { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
696 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
697 };
698
699
700 /* This table is used to check whether auto-possessification is possible
701 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
702 left-hand (repeated) opcode is used to select the row, and the right-hand
703 opcode is used to select the column. The values are as follows:
704
705 0 Always return FALSE (never auto-possessify)
706 1 Character groups are distinct (possessify if both are OP_PROP)
707 2 Check character categories in the same group (general or particular)
708 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
709
710 4 Check left general category vs right particular category
711 5 Check right general category vs left particular category
712
713 6 Left alphanum vs right general category
714 7 Left space vs right general category
715 8 Left word vs right general category
716
717 9 Right alphanum vs left general category
718 10 Right space vs left general category
719 11 Right word vs left general category
720
721 12 Left alphanum vs right particular category
722 13 Left space vs right particular category
723 14 Left word vs right particular category
724
725 15 Right alphanum vs left particular category
726 16 Right space vs left particular category
727 17 Right word vs left particular category
728 */
729
730 static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
731 /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
732 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
733 { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
734 { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
735 { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
736 { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
737 { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
738 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
739 { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
740 { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
741 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
742 { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
743 };
744
745 /* This table is used to check whether auto-possessification is possible
746 between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
747 specifies a general category and the other specifies a particular category. The
748 row is selected by the general category and the column by the particular
749 category. The value is 1 if the particular category is not part of the general
750 category. */
751
752 static const pcre_uint8 catposstab[7][30] = {
753 /* 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 */
754 { 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 */
755 { 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 */
756 { 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 */
757 { 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 */
758 { 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 */
759 { 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 */
760 { 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 */
761 };
762
763 /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
764 a general or particular category. The properties in each row are those
765 that apply to the character set in question. Duplication means that a little
766 unnecessary work is done when checking, but this keeps things much simpler
767 because they can all use the same code. For more details see the comment where
768 this table is used.
769
770 Note: SPACE and PXSPACE used to be different because Perl excluded VT from
771 "space", but from Perl 5.18 it's included, so both categories are treated the
772 same here. */
773
774 static const pcre_uint8 posspropstab[3][4] = {
775 { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
776 { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
777 { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
778 };
779
780 /* This table is used when converting repeating opcodes into possessified
781 versions as a result of an explicit possessive quantifier such as ++. A zero
782 value means there is no possessified version - in those cases the item in
783 question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
784 because all relevant opcodes are less than that. */
785
786 static const pcre_uint8 opcode_possessify[] = {
787 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */
788 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */
789
790 0, /* NOTI */
791 OP_POSSTAR, 0, /* STAR, MINSTAR */
792 OP_POSPLUS, 0, /* PLUS, MINPLUS */
793 OP_POSQUERY, 0, /* QUERY, MINQUERY */
794 OP_POSUPTO, 0, /* UPTO, MINUPTO */
795 0, /* EXACT */
796 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */
797
798 OP_POSSTARI, 0, /* STARI, MINSTARI */
799 OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */
800 OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */
801 OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */
802 0, /* EXACTI */
803 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */
804
805 OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */
806 OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */
807 OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */
808 OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */
809 0, /* NOTEXACT */
810 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
811
812 OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */
813 OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */
814 OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */
815 OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */
816 0, /* NOTEXACTI */
817 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
818
819 OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */
820 OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */
821 OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */
822 OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */
823 0, /* TYPEEXACT */
824 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
825
826 OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */
827 OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */
828 OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */
829 OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */
830 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */
831
832 0, 0, 0, /* CLASS, NCLASS, XCLASS */
833 0, 0, /* REF, REFI */
834 0, 0, /* DNREF, DNREFI */
835 0, 0 /* RECURSE, CALLOUT */
836 };
837
838
839
840 /*************************************************
841 * Find an error text *
842 *************************************************/
843
844 /* The error texts are now all in one long string, to save on relocations. As
845 some of the text is of unknown length, we can't use a table of offsets.
846 Instead, just count through the strings. This is not a performance issue
847 because it happens only when there has been a compilation error.
848
849 Argument: the error number
850 Returns: pointer to the error string
851 */
852
853 static const char *
854 find_error_text(int n)
855 {
856 const char *s = error_texts;
857 for (; n > 0; n--)
858 {
859 while (*s++ != CHAR_NULL) {};
860 if (*s == CHAR_NULL) return "Error text not found (please report)";
861 }
862 return s;
863 }
864
865
866
867 /*************************************************
868 * Expand the workspace *
869 *************************************************/
870
871 /* This function is called during the second compiling phase, if the number of
872 forward references fills the existing workspace, which is originally a block on
873 the stack. A larger block is obtained from malloc() unless the ultimate limit
874 has been reached or the increase will be rather small.
875
876 Argument: pointer to the compile data block
877 Returns: 0 if all went well, else an error number
878 */
879
880 static int
881 expand_workspace(compile_data *cd)
882 {
883 pcre_uchar *newspace;
884 int newsize = cd->workspace_size * 2;
885
886 if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
887 if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
888 newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
889 return ERR72;
890
891 newspace = (PUBL(malloc))(IN_UCHARS(newsize));
892 if (newspace == NULL) return ERR21;
893 memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
894 cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
895 if (cd->workspace_size > COMPILE_WORK_SIZE)
896 (PUBL(free))((void *)cd->start_workspace);
897 cd->start_workspace = newspace;
898 cd->workspace_size = newsize;
899 return 0;
900 }
901
902
903
904 /*************************************************
905 * Check for counted repeat *
906 *************************************************/
907
908 /* This function is called when a '{' is encountered in a place where it might
909 start a quantifier. It looks ahead to see if it really is a quantifier or not.
910 It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
911 where the ddds are digits.
912
913 Arguments:
914 p pointer to the first char after '{'
915
916 Returns: TRUE or FALSE
917 */
918
919 static BOOL
920 is_counted_repeat(const pcre_uchar *p)
921 {
922 if (!IS_DIGIT(*p)) return FALSE;
923 p++;
924 while (IS_DIGIT(*p)) p++;
925 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
926
927 if (*p++ != CHAR_COMMA) return FALSE;
928 if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
929
930 if (!IS_DIGIT(*p)) return FALSE;
931 p++;
932 while (IS_DIGIT(*p)) p++;
933
934 return (*p == CHAR_RIGHT_CURLY_BRACKET);
935 }
936
937
938
939 /*************************************************
940 * Handle escapes *
941 *************************************************/
942
943 /* This function is called when a \ has been encountered. It either returns a
944 positive value for a simple escape such as \n, or 0 for a data character which
945 will be placed in chptr. A backreference to group n is returned as negative n.
946 When UTF-8 is enabled, a positive value greater than 255 may be returned in
947 chptr. On entry, ptr is pointing at the \. On exit, it is on the final
948 character of the escape sequence.
949
950 Arguments:
951 ptrptr points to the pattern position pointer
952 chptr points to a returned data character
953 errorcodeptr points to the errorcode variable
954 bracount number of previous extracting brackets
955 options the options bits
956 isclass TRUE if inside a character class
957
958 Returns: zero => a data character
959 positive => a special escape sequence
960 negative => a back reference
961 on error, errorcodeptr is set
962 */
963
964 static int
965 check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
966 int bracount, int options, BOOL isclass)
967 {
968 /* PCRE_UTF16 has the same value as PCRE_UTF8. */
969 BOOL utf = (options & PCRE_UTF8) != 0;
970 const pcre_uchar *ptr = *ptrptr + 1;
971 pcre_uint32 c;
972 int escape = 0;
973 int i;
974
975 GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
976 ptr--; /* Set pointer back to the last byte */
977
978 /* If backslash is at the end of the pattern, it's an error. */
979
980 if (c == CHAR_NULL) *errorcodeptr = ERR1;
981
982 /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
983 in a table. A non-zero result is something that can be returned immediately.
984 Otherwise further processing may be required. */
985
986 #ifndef EBCDIC /* ASCII/UTF-8 coding */
987 /* Not alphanumeric */
988 else if (c < CHAR_0 || c > CHAR_z) {}
989 else if ((i = escapes[c - CHAR_0]) != 0)
990 { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
991
992 #else /* EBCDIC coding */
993 /* Not alphanumeric */
994 else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
995 else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
996 #endif
997
998 /* Escapes that need further processing, or are illegal. */
999
1000 else
1001 {
1002 const pcre_uchar *oldptr;
1003 BOOL braced, negated, overflow;
1004 int s;
1005
1006 switch (c)
1007 {
1008 /* A number of Perl escapes are not handled by PCRE. We give an explicit
1009 error. */
1010
1011 case CHAR_l:
1012 case CHAR_L:
1013 *errorcodeptr = ERR37;
1014 break;
1015
1016 case CHAR_u:
1017 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1018 {
1019 /* In JavaScript, \u must be followed by four hexadecimal numbers.
1020 Otherwise it is a lowercase u letter. */
1021 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1022 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1023 && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1024 && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1025 {
1026 c = 0;
1027 for (i = 0; i < 4; ++i)
1028 {
1029 register pcre_uint32 cc = *(++ptr);
1030 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1031 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1032 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1033 #else /* EBCDIC coding */
1034 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1035 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1036 #endif
1037 }
1038
1039 #if defined COMPILE_PCRE8
1040 if (c > (utf ? 0x10ffffU : 0xffU))
1041 #elif defined COMPILE_PCRE16
1042 if (c > (utf ? 0x10ffffU : 0xffffU))
1043 #elif defined COMPILE_PCRE32
1044 if (utf && c > 0x10ffffU)
1045 #endif
1046 {
1047 *errorcodeptr = ERR76;
1048 }
1049 else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1050 }
1051 }
1052 else
1053 *errorcodeptr = ERR37;
1054 break;
1055
1056 case CHAR_U:
1057 /* In JavaScript, \U is an uppercase U letter. */
1058 if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1059 break;
1060
1061 /* In a character class, \g is just a literal "g". Outside a character
1062 class, \g must be followed by one of a number of specific things:
1063
1064 (1) A number, either plain or braced. If positive, it is an absolute
1065 backreference. If negative, it is a relative backreference. This is a Perl
1066 5.10 feature.
1067
1068 (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1069 is part of Perl's movement towards a unified syntax for back references. As
1070 this is synonymous with \k{name}, we fudge it up by pretending it really
1071 was \k.
1072
1073 (3) For Oniguruma compatibility we also support \g followed by a name or a
1074 number either in angle brackets or in single quotes. However, these are
1075 (possibly recursive) subroutine calls, _not_ backreferences. Just return
1076 the ESC_g code (cf \k). */
1077
1078 case CHAR_g:
1079 if (isclass) break;
1080 if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1081 {
1082 escape = ESC_g;
1083 break;
1084 }
1085
1086 /* Handle the Perl-compatible cases */
1087
1088 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1089 {
1090 const pcre_uchar *p;
1091 for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1092 if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1093 if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1094 {
1095 escape = ESC_k;
1096 break;
1097 }
1098 braced = TRUE;
1099 ptr++;
1100 }
1101 else braced = FALSE;
1102
1103 if (ptr[1] == CHAR_MINUS)
1104 {
1105 negated = TRUE;
1106 ptr++;
1107 }
1108 else negated = FALSE;
1109
1110 /* The integer range is limited by the machine's int representation. */
1111 s = 0;
1112 overflow = FALSE;
1113 while (IS_DIGIT(ptr[1]))
1114 {
1115 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1116 {
1117 overflow = TRUE;
1118 break;
1119 }
1120 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1121 }
1122 if (overflow) /* Integer overflow */
1123 {
1124 while (IS_DIGIT(ptr[1]))
1125 ptr++;
1126 *errorcodeptr = ERR61;
1127 break;
1128 }
1129
1130 if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1131 {
1132 *errorcodeptr = ERR57;
1133 break;
1134 }
1135
1136 if (s == 0)
1137 {
1138 *errorcodeptr = ERR58;
1139 break;
1140 }
1141
1142 if (negated)
1143 {
1144 if (s > bracount)
1145 {
1146 *errorcodeptr = ERR15;
1147 break;
1148 }
1149 s = bracount - (s - 1);
1150 }
1151
1152 escape = -s;
1153 break;
1154
1155 /* The handling of escape sequences consisting of a string of digits
1156 starting with one that is not zero is not straightforward. Perl has changed
1157 over the years. Nowadays \g{} for backreferences and \o{} for octal are
1158 recommended to avoid the ambiguities in the old syntax.
1159
1160 Outside a character class, the digits are read as a decimal number. If the
1161 number is less than 8 (used to be 10), or if there are that many previous
1162 extracting left brackets, then it is a back reference. Otherwise, up to
1163 three octal digits are read to form an escaped byte. Thus \123 is likely to
1164 be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1165 the octal value is greater than 377, the least significant 8 bits are
1166 taken. \8 and \9 are treated as the literal characters 8 and 9.
1167
1168 Inside a character class, \ followed by a digit is always either a literal
1169 8 or 9 or an octal number. */
1170
1171 case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1172 case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1173
1174 if (!isclass)
1175 {
1176 oldptr = ptr;
1177 /* The integer range is limited by the machine's int representation. */
1178 s = (int)(c -CHAR_0);
1179 overflow = FALSE;
1180 while (IS_DIGIT(ptr[1]))
1181 {
1182 if (s > INT_MAX / 10 - 1) /* Integer overflow */
1183 {
1184 overflow = TRUE;
1185 break;
1186 }
1187 s = s * 10 + (int)(*(++ptr) - CHAR_0);
1188 }
1189 if (overflow) /* Integer overflow */
1190 {
1191 while (IS_DIGIT(ptr[1]))
1192 ptr++;
1193 *errorcodeptr = ERR61;
1194 break;
1195 }
1196 if (s < 8 || s <= bracount) /* Check for back reference */
1197 {
1198 escape = -s;
1199 break;
1200 }
1201 ptr = oldptr; /* Put the pointer back and fall through */
1202 }
1203
1204 /* Handle a digit following \ when the number is not a back reference. If
1205 the first digit is 8 or 9, Perl used to generate a binary zero byte and
1206 then treat the digit as a following literal. At least by Perl 5.18 this
1207 changed so as not to insert the binary zero. */
1208
1209 if ((c = *ptr) >= CHAR_8) break;
1210
1211 /* Fall through with a digit less than 8 */
1212
1213 /* \0 always starts an octal number, but we may drop through to here with a
1214 larger first octal digit. The original code used just to take the least
1215 significant 8 bits of octal numbers (I think this is what early Perls used
1216 to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1217 but no more than 3 octal digits. */
1218
1219 case CHAR_0:
1220 c -= CHAR_0;
1221 while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1222 c = c * 8 + *(++ptr) - CHAR_0;
1223 #ifdef COMPILE_PCRE8
1224 if (!utf && c > 0xff) *errorcodeptr = ERR51;
1225 #endif
1226 break;
1227
1228 /* \o is a relatively new Perl feature, supporting a more general way of
1229 specifying character codes in octal. The only supported form is \o{ddd}. */
1230
1231 case CHAR_o:
1232 if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1233 {
1234 ptr += 2;
1235 c = 0;
1236 overflow = FALSE;
1237 while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1238 {
1239 register pcre_uint32 cc = *ptr++;
1240 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1241 #ifdef COMPILE_PCRE32
1242 if (c >= 0x20000000l) { overflow = TRUE; break; }
1243 #endif
1244 c = (c << 3) + cc - CHAR_0 ;
1245 #if defined COMPILE_PCRE8
1246 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1247 #elif defined COMPILE_PCRE16
1248 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1249 #elif defined COMPILE_PCRE32
1250 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1251 #endif
1252 }
1253 if (overflow)
1254 {
1255 while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1256 *errorcodeptr = ERR34;
1257 }
1258 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1259 {
1260 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1261 }
1262 else *errorcodeptr = ERR80;
1263 }
1264 break;
1265
1266 /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1267 numbers. Otherwise it is a lowercase x letter. */
1268
1269 case CHAR_x:
1270 if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1271 {
1272 if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1273 && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1274 {
1275 c = 0;
1276 for (i = 0; i < 2; ++i)
1277 {
1278 register pcre_uint32 cc = *(++ptr);
1279 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1280 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1281 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1282 #else /* EBCDIC coding */
1283 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1284 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1285 #endif
1286 }
1287 }
1288 } /* End JavaScript handling */
1289
1290 /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1291 greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1292 digits. If not, { used to be treated as a data character. However, Perl
1293 seems to read hex digits up to the first non-such, and ignore the rest, so
1294 that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1295 now gives an error. */
1296
1297 else
1298 {
1299 if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1300 {
1301 ptr += 2;
1302 c = 0;
1303 overflow = FALSE;
1304 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1305 {
1306 register pcre_uint32 cc = *ptr++;
1307 if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1308
1309 #ifdef COMPILE_PCRE32
1310 if (c >= 0x10000000l) { overflow = TRUE; break; }
1311 #endif
1312
1313 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1314 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1315 c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1316 #else /* EBCDIC coding */
1317 if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1318 c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1319 #endif
1320
1321 #if defined COMPILE_PCRE8
1322 if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1323 #elif defined COMPILE_PCRE16
1324 if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1325 #elif defined COMPILE_PCRE32
1326 if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1327 #endif
1328 }
1329
1330 if (overflow)
1331 {
1332 while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1333 *errorcodeptr = ERR34;
1334 }
1335
1336 else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1337 {
1338 if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1339 }
1340
1341 /* If the sequence of hex digits does not end with '}', give an error.
1342 We used just to recognize this construct and fall through to the normal
1343 \x handling, but nowadays Perl gives an error, which seems much more
1344 sensible, so we do too. */
1345
1346 else *errorcodeptr = ERR79;
1347 } /* End of \x{} processing */
1348
1349 /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1350
1351 else
1352 {
1353 c = 0;
1354 while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1355 {
1356 pcre_uint32 cc; /* Some compilers don't like */
1357 cc = *(++ptr); /* ++ in initializers */
1358 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1359 if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1360 c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1361 #else /* EBCDIC coding */
1362 if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1363 c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1364 #endif
1365 }
1366 } /* End of \xdd handling */
1367 } /* End of Perl-style \x handling */
1368 break;
1369
1370 /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1371 An error is given if the byte following \c is not an ASCII character. This
1372 coding is ASCII-specific, but then the whole concept of \cx is
1373 ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1374
1375 case CHAR_c:
1376 c = *(++ptr);
1377 if (c == CHAR_NULL)
1378 {
1379 *errorcodeptr = ERR2;
1380 break;
1381 }
1382 #ifndef EBCDIC /* ASCII/UTF-8 coding */
1383 if (c > 127) /* Excludes all non-ASCII in either mode */
1384 {
1385 *errorcodeptr = ERR68;
1386 break;
1387 }
1388 if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1389 c ^= 0x40;
1390 #else /* EBCDIC coding */
1391 if (c >= CHAR_a && c <= CHAR_z) c += 64;
1392 c ^= 0xC0;
1393 #endif
1394 break;
1395
1396 /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1397 other alphanumeric following \ is an error if PCRE_EXTRA was set;
1398 otherwise, for Perl compatibility, it is a literal. This code looks a bit
1399 odd, but there used to be some cases other than the default, and there may
1400 be again in future, so I haven't "optimized" it. */
1401
1402 default:
1403 if ((options & PCRE_EXTRA) != 0) switch(c)
1404 {
1405 default:
1406 *errorcodeptr = ERR3;
1407 break;
1408 }
1409 break;
1410 }
1411 }
1412
1413 /* Perl supports \N{name} for character names, as well as plain \N for "not
1414 newline". PCRE does not support \N{name}. However, it does support
1415 quantification such as \N{2,3}. */
1416
1417 if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1418 !is_counted_repeat(ptr+2))
1419 *errorcodeptr = ERR37;
1420
1421 /* If PCRE_UCP is set, we change the values for \d etc. */
1422
1423 if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1424 escape += (ESC_DU - ESC_D);
1425
1426 /* Set the pointer to the final character before returning. */
1427
1428 *ptrptr = ptr;
1429 *chptr = c;
1430 return escape;
1431 }
1432
1433
1434
1435 #ifdef SUPPORT_UCP
1436 /*************************************************
1437 * Handle \P and \p *
1438 *************************************************/
1439
1440 /* This function is called after \P or \p has been encountered, provided that
1441 PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1442 pointing at the P or p. On exit, it is pointing at the final character of the
1443 escape sequence.
1444
1445 Argument:
1446 ptrptr points to the pattern position pointer
1447 negptr points to a boolean that is set TRUE for negation else FALSE
1448 ptypeptr points to an unsigned int that is set to the type value
1449 pdataptr points to an unsigned int that is set to the detailed property value
1450 errorcodeptr points to the error code variable
1451
1452 Returns: TRUE if the type value was found, or FALSE for an invalid type
1453 */
1454
1455 static BOOL
1456 get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1457 unsigned int *pdataptr, int *errorcodeptr)
1458 {
1459 pcre_uchar c;
1460 int i, bot, top;
1461 const pcre_uchar *ptr = *ptrptr;
1462 pcre_uchar name[32];
1463
1464 c = *(++ptr);
1465 if (c == CHAR_NULL) goto ERROR_RETURN;
1466
1467 *negptr = FALSE;
1468
1469 /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1470 negation. */
1471
1472 if (c == CHAR_LEFT_CURLY_BRACKET)
1473 {
1474 if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1475 {
1476 *negptr = TRUE;
1477 ptr++;
1478 }
1479 for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1480 {
1481 c = *(++ptr);
1482 if (c == CHAR_NULL) goto ERROR_RETURN;
1483 if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1484 name[i] = c;
1485 }
1486 if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1487 name[i] = 0;
1488 }
1489
1490 /* Otherwise there is just one following character */
1491
1492 else
1493 {
1494 name[0] = c;
1495 name[1] = 0;
1496 }
1497
1498 *ptrptr = ptr;
1499
1500 /* Search for a recognized property name using binary chop */
1501
1502 bot = 0;
1503 top = PRIV(utt_size);
1504
1505 while (bot < top)
1506 {
1507 int r;
1508 i = (bot + top) >> 1;
1509 r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1510 if (r == 0)
1511 {
1512 *ptypeptr = PRIV(utt)[i].type;
1513 *pdataptr = PRIV(utt)[i].value;
1514 return TRUE;
1515 }
1516 if (r > 0) bot = i + 1; else top = i;
1517 }
1518
1519 *errorcodeptr = ERR47;
1520 *ptrptr = ptr;
1521 return FALSE;
1522
1523 ERROR_RETURN:
1524 *errorcodeptr = ERR46;
1525 *ptrptr = ptr;
1526 return FALSE;
1527 }
1528 #endif
1529
1530
1531
1532 /*************************************************
1533 * Read repeat counts *
1534 *************************************************/
1535
1536 /* Read an item of the form {n,m} and return the values. This is called only
1537 after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1538 so the syntax is guaranteed to be correct, but we need to check the values.
1539
1540 Arguments:
1541 p pointer to first char after '{'
1542 minp pointer to int for min
1543 maxp pointer to int for max
1544 returned as -1 if no max
1545 errorcodeptr points to error code variable
1546
1547 Returns: pointer to '}' on success;
1548 current ptr on error, with errorcodeptr set non-zero
1549 */
1550
1551 static const pcre_uchar *
1552 read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1553 {
1554 int min = 0;
1555 int max = -1;
1556
1557 /* Read the minimum value and do a paranoid check: a negative value indicates
1558 an integer overflow. */
1559
1560 while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1561 if (min < 0 || min > 65535)
1562 {
1563 *errorcodeptr = ERR5;
1564 return p;
1565 }
1566
1567 /* Read the maximum value if there is one, and again do a paranoid on its size.
1568 Also, max must not be less than min. */
1569
1570 if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1571 {
1572 if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1573 {
1574 max = 0;
1575 while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1576 if (max < 0 || max > 65535)
1577 {
1578 *errorcodeptr = ERR5;
1579 return p;
1580 }
1581 if (max < min)
1582 {
1583 *errorcodeptr = ERR4;
1584 return p;
1585 }
1586 }
1587 }
1588
1589 /* Fill in the required variables, and pass back the pointer to the terminating
1590 '}'. */
1591
1592 *minp = min;
1593 *maxp = max;
1594 return p;
1595 }
1596
1597
1598
1599 /*************************************************
1600 * Find first significant op code *
1601 *************************************************/
1602
1603 /* This is called by several functions that scan a compiled expression looking
1604 for a fixed first character, or an anchoring op code etc. It skips over things
1605 that do not influence this. For some calls, it makes sense to skip negative
1606 forward and all backward assertions, and also the \b assertion; for others it
1607 does not.
1608
1609 Arguments:
1610 code pointer to the start of the group
1611 skipassert TRUE if certain assertions are to be skipped
1612
1613 Returns: pointer to the first significant opcode
1614 */
1615
1616 static const pcre_uchar*
1617 first_significant_code(const pcre_uchar *code, BOOL skipassert)
1618 {
1619 for (;;)
1620 {
1621 switch ((int)*code)
1622 {
1623 case OP_ASSERT_NOT:
1624 case OP_ASSERTBACK:
1625 case OP_ASSERTBACK_NOT:
1626 if (!skipassert) return code;
1627 do code += GET(code, 1); while (*code == OP_ALT);
1628 code += PRIV(OP_lengths)[*code];
1629 break;
1630
1631 case OP_WORD_BOUNDARY:
1632 case OP_NOT_WORD_BOUNDARY:
1633 if (!skipassert) return code;
1634 /* Fall through */
1635
1636 case OP_CALLOUT:
1637 case OP_CREF:
1638 case OP_DNCREF:
1639 case OP_RREF:
1640 case OP_DNRREF:
1641 case OP_DEF:
1642 code += PRIV(OP_lengths)[*code];
1643 break;
1644
1645 default:
1646 return code;
1647 }
1648 }
1649 /* Control never reaches here */
1650 }
1651
1652
1653
1654 /*************************************************
1655 * Find the fixed length of a branch *
1656 *************************************************/
1657
1658 /* Scan a branch and compute the fixed length of subject that will match it,
1659 if the length is fixed. This is needed for dealing with backward assertions.
1660 In UTF8 mode, the result is in characters rather than bytes. The branch is
1661 temporarily terminated with OP_END when this function is called.
1662
1663 This function is called when a backward assertion is encountered, so that if it
1664 fails, the error message can point to the correct place in the pattern.
1665 However, we cannot do this when the assertion contains subroutine calls,
1666 because they can be forward references. We solve this by remembering this case
1667 and doing the check at the end; a flag specifies which mode we are running in.
1668
1669 Arguments:
1670 code points to the start of the pattern (the bracket)
1671 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1672 atend TRUE if called when the pattern is complete
1673 cd the "compile data" structure
1674
1675 Returns: the fixed length,
1676 or -1 if there is no fixed length,
1677 or -2 if \C was encountered (in UTF-8 mode only)
1678 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1679 or -4 if an unknown opcode was encountered (internal error)
1680 */
1681
1682 static int
1683 find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1684 {
1685 int length = -1;
1686
1687 register int branchlength = 0;
1688 register pcre_uchar *cc = code + 1 + LINK_SIZE;
1689
1690 /* Scan along the opcodes for this branch. If we get to the end of the
1691 branch, check the length against that of the other branches. */
1692
1693 for (;;)
1694 {
1695 int d;
1696 pcre_uchar *ce, *cs;
1697 register pcre_uchar op = *cc;
1698
1699 switch (op)
1700 {
1701 /* We only need to continue for OP_CBRA (normal capturing bracket) and
1702 OP_BRA (normal non-capturing bracket) because the other variants of these
1703 opcodes are all concerned with unlimited repeated groups, which of course
1704 are not of fixed length. */
1705
1706 case OP_CBRA:
1707 case OP_BRA:
1708 case OP_ONCE:
1709 case OP_ONCE_NC:
1710 case OP_COND:
1711 d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1712 if (d < 0) return d;
1713 branchlength += d;
1714 do cc += GET(cc, 1); while (*cc == OP_ALT);
1715 cc += 1 + LINK_SIZE;
1716 break;
1717
1718 /* Reached end of a branch; if it's a ket it is the end of a nested call.
1719 If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1720 an ALT. If it is END it's the end of the outer call. All can be handled by
1721 the same code. Note that we must not include the OP_KETRxxx opcodes here,
1722 because they all imply an unlimited repeat. */
1723
1724 case OP_ALT:
1725 case OP_KET:
1726 case OP_END:
1727 case OP_ACCEPT:
1728 case OP_ASSERT_ACCEPT:
1729 if (length < 0) length = branchlength;
1730 else if (length != branchlength) return -1;
1731 if (*cc != OP_ALT) return length;
1732 cc += 1 + LINK_SIZE;
1733 branchlength = 0;
1734 break;
1735
1736 /* A true recursion implies not fixed length, but a subroutine call may
1737 be OK. If the subroutine is a forward reference, we can't deal with
1738 it until the end of the pattern, so return -3. */
1739
1740 case OP_RECURSE:
1741 if (!atend) return -3;
1742 cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1743 do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1744 if (cc > cs && cc < ce) return -1; /* Recursion */
1745 d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1746 if (d < 0) return d;
1747 branchlength += d;
1748 cc += 1 + LINK_SIZE;
1749 break;
1750
1751 /* Skip over assertive subpatterns */
1752
1753 case OP_ASSERT:
1754 case OP_ASSERT_NOT:
1755 case OP_ASSERTBACK:
1756 case OP_ASSERTBACK_NOT:
1757 do cc += GET(cc, 1); while (*cc == OP_ALT);
1758 cc += PRIV(OP_lengths)[*cc];
1759 break;
1760
1761 /* Skip over things that don't match chars */
1762
1763 case OP_MARK:
1764 case OP_PRUNE_ARG:
1765 case OP_SKIP_ARG:
1766 case OP_THEN_ARG:
1767 cc += cc[1] + PRIV(OP_lengths)[*cc];
1768 break;
1769
1770 case OP_CALLOUT:
1771 case OP_CIRC:
1772 case OP_CIRCM:
1773 case OP_CLOSE:
1774 case OP_COMMIT:
1775 case OP_CREF:
1776 case OP_DEF:
1777 case OP_DNCREF:
1778 case OP_DNRREF:
1779 case OP_DOLL:
1780 case OP_DOLLM:
1781 case OP_EOD:
1782 case OP_EODN:
1783 case OP_FAIL:
1784 case OP_NOT_WORD_BOUNDARY:
1785 case OP_PRUNE:
1786 case OP_REVERSE:
1787 case OP_RREF:
1788 case OP_SET_SOM:
1789 case OP_SKIP:
1790 case OP_SOD:
1791 case OP_SOM:
1792 case OP_THEN:
1793 case OP_WORD_BOUNDARY:
1794 cc += PRIV(OP_lengths)[*cc];
1795 break;
1796
1797 /* Handle literal characters */
1798
1799 case OP_CHAR:
1800 case OP_CHARI:
1801 case OP_NOT:
1802 case OP_NOTI:
1803 branchlength++;
1804 cc += 2;
1805 #ifdef SUPPORT_UTF
1806 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1807 #endif
1808 break;
1809
1810 /* Handle exact repetitions. The count is already in characters, but we
1811 need to skip over a multibyte character in UTF8 mode. */
1812
1813 case OP_EXACT:
1814 case OP_EXACTI:
1815 case OP_NOTEXACT:
1816 case OP_NOTEXACTI:
1817 branchlength += (int)GET2(cc,1);
1818 cc += 2 + IMM2_SIZE;
1819 #ifdef SUPPORT_UTF
1820 if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1821 #endif
1822 break;
1823
1824 case OP_TYPEEXACT:
1825 branchlength += GET2(cc,1);
1826 if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1827 cc += 2;
1828 cc += 1 + IMM2_SIZE + 1;
1829 break;
1830
1831 /* Handle single-char matchers */
1832
1833 case OP_PROP:
1834 case OP_NOTPROP:
1835 cc += 2;
1836 /* Fall through */
1837
1838 case OP_HSPACE:
1839 case OP_VSPACE:
1840 case OP_NOT_HSPACE:
1841 case OP_NOT_VSPACE:
1842 case OP_NOT_DIGIT:
1843 case OP_DIGIT:
1844 case OP_NOT_WHITESPACE:
1845 case OP_WHITESPACE:
1846 case OP_NOT_WORDCHAR:
1847 case OP_WORDCHAR:
1848 case OP_ANY:
1849 case OP_ALLANY:
1850 branchlength++;
1851 cc++;
1852 break;
1853
1854 /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1855 otherwise \C is coded as OP_ALLANY. */
1856
1857 case OP_ANYBYTE:
1858 return -2;
1859
1860 /* Check a class for variable quantification */
1861
1862 case OP_CLASS:
1863 case OP_NCLASS:
1864 #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1865 case OP_XCLASS:
1866 /* The original code caused an unsigned overflow in 64 bit systems,
1867 so now we use a conditional statement. */
1868 if (op == OP_XCLASS)
1869 cc += GET(cc, 1);
1870 else
1871 cc += PRIV(OP_lengths)[OP_CLASS];
1872 #else
1873 cc += PRIV(OP_lengths)[OP_CLASS];
1874 #endif
1875
1876 switch (*cc)
1877 {
1878 case OP_CRSTAR:
1879 case OP_CRMINSTAR:
1880 case OP_CRPLUS:
1881 case OP_CRMINPLUS:
1882 case OP_CRQUERY:
1883 case OP_CRMINQUERY:
1884 case OP_CRPOSSTAR:
1885 case OP_CRPOSPLUS:
1886 case OP_CRPOSQUERY:
1887 return -1;
1888
1889 case OP_CRRANGE:
1890 case OP_CRMINRANGE:
1891 case OP_CRPOSRANGE:
1892 if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1893 branchlength += (int)GET2(cc,1);
1894 cc += 1 + 2 * IMM2_SIZE;
1895 break;
1896
1897 default:
1898 branchlength++;
1899 }
1900 break;
1901
1902 /* Anything else is variable length */
1903
1904 case OP_ANYNL:
1905 case OP_BRAMINZERO:
1906 case OP_BRAPOS:
1907 case OP_BRAPOSZERO:
1908 case OP_BRAZERO:
1909 case OP_CBRAPOS:
1910 case OP_EXTUNI:
1911 case OP_KETRMAX:
1912 case OP_KETRMIN:
1913 case OP_KETRPOS:
1914 case OP_MINPLUS:
1915 case OP_MINPLUSI:
1916 case OP_MINQUERY:
1917 case OP_MINQUERYI:
1918 case OP_MINSTAR:
1919 case OP_MINSTARI:
1920 case OP_MINUPTO:
1921 case OP_MINUPTOI:
1922 case OP_NOTMINPLUS:
1923 case OP_NOTMINPLUSI:
1924 case OP_NOTMINQUERY:
1925 case OP_NOTMINQUERYI:
1926 case OP_NOTMINSTAR:
1927 case OP_NOTMINSTARI:
1928 case OP_NOTMINUPTO:
1929 case OP_NOTMINUPTOI:
1930 case OP_NOTPLUS:
1931 case OP_NOTPLUSI:
1932 case OP_NOTPOSPLUS:
1933 case OP_NOTPOSPLUSI:
1934 case OP_NOTPOSQUERY:
1935 case OP_NOTPOSQUERYI:
1936 case OP_NOTPOSSTAR:
1937 case OP_NOTPOSSTARI:
1938 case OP_NOTPOSUPTO:
1939 case OP_NOTPOSUPTOI:
1940 case OP_NOTQUERY:
1941 case OP_NOTQUERYI:
1942 case OP_NOTSTAR:
1943 case OP_NOTSTARI:
1944 case OP_NOTUPTO:
1945 case OP_NOTUPTOI:
1946 case OP_PLUS:
1947 case OP_PLUSI:
1948 case OP_POSPLUS:
1949 case OP_POSPLUSI:
1950 case OP_POSQUERY:
1951 case OP_POSQUERYI:
1952 case OP_POSSTAR:
1953 case OP_POSSTARI:
1954 case OP_POSUPTO:
1955 case OP_POSUPTOI:
1956 case OP_QUERY:
1957 case OP_QUERYI:
1958 case OP_REF:
1959 case OP_REFI:
1960 case OP_DNREF:
1961 case OP_DNREFI:
1962 case OP_SBRA:
1963 case OP_SBRAPOS:
1964 case OP_SCBRA:
1965 case OP_SCBRAPOS:
1966 case OP_SCOND:
1967 case OP_SKIPZERO:
1968 case OP_STAR:
1969 case OP_STARI:
1970 case OP_TYPEMINPLUS:
1971 case OP_TYPEMINQUERY:
1972 case OP_TYPEMINSTAR:
1973 case OP_TYPEMINUPTO:
1974 case OP_TYPEPLUS:
1975 case OP_TYPEPOSPLUS:
1976 case OP_TYPEPOSQUERY:
1977 case OP_TYPEPOSSTAR:
1978 case OP_TYPEPOSUPTO:
1979 case OP_TYPEQUERY:
1980 case OP_TYPESTAR:
1981 case OP_TYPEUPTO:
1982 case OP_UPTO:
1983 case OP_UPTOI:
1984 return -1;
1985
1986 /* Catch unrecognized opcodes so that when new ones are added they
1987 are not forgotten, as has happened in the past. */
1988
1989 default:
1990 return -4;
1991 }
1992 }
1993 /* Control never gets here */
1994 }
1995
1996
1997
1998 /*************************************************
1999 * Scan compiled regex for specific bracket *
2000 *************************************************/
2001
2002 /* This little function scans through a compiled pattern until it finds a
2003 capturing bracket with the given number, or, if the number is negative, an
2004 instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2005 so that it can be called from pcre_study() when finding the minimum matching
2006 length.
2007
2008 Arguments:
2009 code points to start of expression
2010 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2011 number the required bracket number or negative to find a lookbehind
2012
2013 Returns: pointer to the opcode for the bracket, or NULL if not found
2014 */
2015
2016 const pcre_uchar *
2017 PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2018 {
2019 for (;;)
2020 {
2021 register pcre_uchar c = *code;
2022
2023 if (c == OP_END) return NULL;
2024
2025 /* XCLASS is used for classes that cannot be represented just by a bit
2026 map. This includes negated single high-valued characters. The length in
2027 the table is zero; the actual length is stored in the compiled code. */
2028
2029 if (c == OP_XCLASS) code += GET(code, 1);
2030
2031 /* Handle recursion */
2032
2033 else if (c == OP_REVERSE)
2034 {
2035 if (number < 0) return (pcre_uchar *)code;
2036 code += PRIV(OP_lengths)[c];
2037 }
2038
2039 /* Handle capturing bracket */
2040
2041 else if (c == OP_CBRA || c == OP_SCBRA ||
2042 c == OP_CBRAPOS || c == OP_SCBRAPOS)
2043 {
2044 int n = (int)GET2(code, 1+LINK_SIZE);
2045 if (n == number) return (pcre_uchar *)code;
2046 code += PRIV(OP_lengths)[c];
2047 }
2048
2049 /* Otherwise, we can get the item's length from the table, except that for
2050 repeated character types, we have to test for \p and \P, which have an extra
2051 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2052 must add in its length. */
2053
2054 else
2055 {
2056 switch(c)
2057 {
2058 case OP_TYPESTAR:
2059 case OP_TYPEMINSTAR:
2060 case OP_TYPEPLUS:
2061 case OP_TYPEMINPLUS:
2062 case OP_TYPEQUERY:
2063 case OP_TYPEMINQUERY:
2064 case OP_TYPEPOSSTAR:
2065 case OP_TYPEPOSPLUS:
2066 case OP_TYPEPOSQUERY:
2067 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2068 break;
2069
2070 case OP_TYPEUPTO:
2071 case OP_TYPEMINUPTO:
2072 case OP_TYPEEXACT:
2073 case OP_TYPEPOSUPTO:
2074 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2075 code += 2;
2076 break;
2077
2078 case OP_MARK:
2079 case OP_PRUNE_ARG:
2080 case OP_SKIP_ARG:
2081 case OP_THEN_ARG:
2082 code += code[1];
2083 break;
2084 }
2085
2086 /* Add in the fixed length from the table */
2087
2088 code += PRIV(OP_lengths)[c];
2089
2090 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2091 a multi-byte character. The length in the table is a minimum, so we have to
2092 arrange to skip the extra bytes. */
2093
2094 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2095 if (utf) switch(c)
2096 {
2097 case OP_CHAR:
2098 case OP_CHARI:
2099 case OP_EXACT:
2100 case OP_EXACTI:
2101 case OP_UPTO:
2102 case OP_UPTOI:
2103 case OP_MINUPTO:
2104 case OP_MINUPTOI:
2105 case OP_POSUPTO:
2106 case OP_POSUPTOI:
2107 case OP_STAR:
2108 case OP_STARI:
2109 case OP_MINSTAR:
2110 case OP_MINSTARI:
2111 case OP_POSSTAR:
2112 case OP_POSSTARI:
2113 case OP_PLUS:
2114 case OP_PLUSI:
2115 case OP_MINPLUS:
2116 case OP_MINPLUSI:
2117 case OP_POSPLUS:
2118 case OP_POSPLUSI:
2119 case OP_QUERY:
2120 case OP_QUERYI:
2121 case OP_MINQUERY:
2122 case OP_MINQUERYI:
2123 case OP_POSQUERY:
2124 case OP_POSQUERYI:
2125 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2126 break;
2127 }
2128 #else
2129 (void)(utf); /* Keep compiler happy by referencing function argument */
2130 #endif
2131 }
2132 }
2133 }
2134
2135
2136
2137 /*************************************************
2138 * Scan compiled regex for recursion reference *
2139 *************************************************/
2140
2141 /* This little function scans through a compiled pattern until it finds an
2142 instance of OP_RECURSE.
2143
2144 Arguments:
2145 code points to start of expression
2146 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2147
2148 Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2149 */
2150
2151 static const pcre_uchar *
2152 find_recurse(const pcre_uchar *code, BOOL utf)
2153 {
2154 for (;;)
2155 {
2156 register pcre_uchar c = *code;
2157 if (c == OP_END) return NULL;
2158 if (c == OP_RECURSE) return code;
2159
2160 /* XCLASS is used for classes that cannot be represented just by a bit
2161 map. This includes negated single high-valued characters. The length in
2162 the table is zero; the actual length is stored in the compiled code. */
2163
2164 if (c == OP_XCLASS) code += GET(code, 1);
2165
2166 /* Otherwise, we can get the item's length from the table, except that for
2167 repeated character types, we have to test for \p and \P, which have an extra
2168 two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2169 must add in its length. */
2170
2171 else
2172 {
2173 switch(c)
2174 {
2175 case OP_TYPESTAR:
2176 case OP_TYPEMINSTAR:
2177 case OP_TYPEPLUS:
2178 case OP_TYPEMINPLUS:
2179 case OP_TYPEQUERY:
2180 case OP_TYPEMINQUERY:
2181 case OP_TYPEPOSSTAR:
2182 case OP_TYPEPOSPLUS:
2183 case OP_TYPEPOSQUERY:
2184 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2185 break;
2186
2187 case OP_TYPEPOSUPTO:
2188 case OP_TYPEUPTO:
2189 case OP_TYPEMINUPTO:
2190 case OP_TYPEEXACT:
2191 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2192 code += 2;
2193 break;
2194
2195 case OP_MARK:
2196 case OP_PRUNE_ARG:
2197 case OP_SKIP_ARG:
2198 case OP_THEN_ARG:
2199 code += code[1];
2200 break;
2201 }
2202
2203 /* Add in the fixed length from the table */
2204
2205 code += PRIV(OP_lengths)[c];
2206
2207 /* In UTF-8 mode, opcodes that are followed by a character may be followed
2208 by a multi-byte character. The length in the table is a minimum, so we have
2209 to arrange to skip the extra bytes. */
2210
2211 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2212 if (utf) switch(c)
2213 {
2214 case OP_CHAR:
2215 case OP_CHARI:
2216 case OP_NOT:
2217 case OP_NOTI:
2218 case OP_EXACT:
2219 case OP_EXACTI:
2220 case OP_NOTEXACT:
2221 case OP_NOTEXACTI:
2222 case OP_UPTO:
2223 case OP_UPTOI:
2224 case OP_NOTUPTO:
2225 case OP_NOTUPTOI:
2226 case OP_MINUPTO:
2227 case OP_MINUPTOI:
2228 case OP_NOTMINUPTO:
2229 case OP_NOTMINUPTOI:
2230 case OP_POSUPTO:
2231 case OP_POSUPTOI:
2232 case OP_NOTPOSUPTO:
2233 case OP_NOTPOSUPTOI:
2234 case OP_STAR:
2235 case OP_STARI:
2236 case OP_NOTSTAR:
2237 case OP_NOTSTARI:
2238 case OP_MINSTAR:
2239 case OP_MINSTARI:
2240 case OP_NOTMINSTAR:
2241 case OP_NOTMINSTARI:
2242 case OP_POSSTAR:
2243 case OP_POSSTARI:
2244 case OP_NOTPOSSTAR:
2245 case OP_NOTPOSSTARI:
2246 case OP_PLUS:
2247 case OP_PLUSI:
2248 case OP_NOTPLUS:
2249 case OP_NOTPLUSI:
2250 case OP_MINPLUS:
2251 case OP_MINPLUSI:
2252 case OP_NOTMINPLUS:
2253 case OP_NOTMINPLUSI:
2254 case OP_POSPLUS:
2255 case OP_POSPLUSI:
2256 case OP_NOTPOSPLUS:
2257 case OP_NOTPOSPLUSI:
2258 case OP_QUERY:
2259 case OP_QUERYI:
2260 case OP_NOTQUERY:
2261 case OP_NOTQUERYI:
2262 case OP_MINQUERY:
2263 case OP_MINQUERYI:
2264 case OP_NOTMINQUERY:
2265 case OP_NOTMINQUERYI:
2266 case OP_POSQUERY:
2267 case OP_POSQUERYI:
2268 case OP_NOTPOSQUERY:
2269 case OP_NOTPOSQUERYI:
2270 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2271 break;
2272 }
2273 #else
2274 (void)(utf); /* Keep compiler happy by referencing function argument */
2275 #endif
2276 }
2277 }
2278 }
2279
2280
2281
2282 /*************************************************
2283 * Scan compiled branch for non-emptiness *
2284 *************************************************/
2285
2286 /* This function scans through a branch of a compiled pattern to see whether it
2287 can match the empty string or not. It is called from could_be_empty()
2288 below and from compile_branch() when checking for an unlimited repeat of a
2289 group that can match nothing. Note that first_significant_code() skips over
2290 backward and negative forward assertions when its final argument is TRUE. If we
2291 hit an unclosed bracket, we return "empty" - this means we've struck an inner
2292 bracket whose current branch will already have been scanned.
2293
2294 Arguments:
2295 code points to start of search
2296 endcode points to where to stop
2297 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2298 cd contains pointers to tables etc.
2299 recurses chain of recurse_check to catch mutual recursion
2300
2301 Returns: TRUE if what is matched could be empty
2302 */
2303
2304 typedef struct recurse_check {
2305 struct recurse_check *prev;
2306 const pcre_uchar *group;
2307 } recurse_check;
2308
2309 static BOOL
2310 could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2311 BOOL utf, compile_data *cd, recurse_check *recurses)
2312 {
2313 register pcre_uchar c;
2314 recurse_check this_recurse;
2315
2316 for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2317 code < endcode;
2318 code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2319 {
2320 const pcre_uchar *ccode;
2321
2322 c = *code;
2323
2324 /* Skip over forward assertions; the other assertions are skipped by
2325 first_significant_code() with a TRUE final argument. */
2326
2327 if (c == OP_ASSERT)
2328 {
2329 do code += GET(code, 1); while (*code == OP_ALT);
2330 c = *code;
2331 continue;
2332 }
2333
2334 /* For a recursion/subroutine call, if its end has been reached, which
2335 implies a backward reference subroutine call, we can scan it. If it's a
2336 forward reference subroutine call, we can't. To detect forward reference
2337 we have to scan up the list that is kept in the workspace. This function is
2338 called only when doing the real compile, not during the pre-compile that
2339 measures the size of the compiled pattern. */
2340
2341 if (c == OP_RECURSE)
2342 {
2343 const pcre_uchar *scode = cd->start_code + GET(code, 1);
2344 BOOL empty_branch;
2345
2346 /* Test for forward reference or uncompleted reference. This is disabled
2347 when called to scan a completed pattern by setting cd->start_workspace to
2348 NULL. */
2349
2350 if (cd->start_workspace != NULL)
2351 {
2352 const pcre_uchar *tcode;
2353 for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2354 if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2355 if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2356 }
2357
2358 /* If we are scanning a completed pattern, there are no forward references
2359 and all groups are complete. We need to detect whether this is a recursive
2360 call, as otherwise there will be an infinite loop. If it is a recursion,
2361 just skip over it. Simple recursions are easily detected. For mutual
2362 recursions we keep a chain on the stack. */
2363
2364 else
2365 {
2366 recurse_check *r = recurses;
2367 const pcre_uchar *endgroup = scode;
2368
2369 do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2370 if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2371
2372 for (r = recurses; r != NULL; r = r->prev)
2373 if (r->group == scode) break;
2374 if (r != NULL) continue; /* Mutual recursion */
2375 }
2376
2377 /* Completed reference; scan the referenced group, remembering it on the
2378 stack chain to detect mutual recursions. */
2379
2380 empty_branch = FALSE;
2381 this_recurse.prev = recurses;
2382 this_recurse.group = scode;
2383
2384 do
2385 {
2386 if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2387 {
2388 empty_branch = TRUE;
2389 break;
2390 }
2391 scode += GET(scode, 1);
2392 }
2393 while (*scode == OP_ALT);
2394
2395 if (!empty_branch) return FALSE; /* All branches are non-empty */
2396 continue;
2397 }
2398
2399 /* Groups with zero repeats can of course be empty; skip them. */
2400
2401 if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2402 c == OP_BRAPOSZERO)
2403 {
2404 code += PRIV(OP_lengths)[c];
2405 do code += GET(code, 1); while (*code == OP_ALT);
2406 c = *code;
2407 continue;
2408 }
2409
2410 /* A nested group that is already marked as "could be empty" can just be
2411 skipped. */
2412
2413 if (c == OP_SBRA || c == OP_SBRAPOS ||
2414 c == OP_SCBRA || c == OP_SCBRAPOS)
2415 {
2416 do code += GET(code, 1); while (*code == OP_ALT);
2417 c = *code;
2418 continue;
2419 }
2420
2421 /* For other groups, scan the branches. */
2422
2423 if (c == OP_BRA || c == OP_BRAPOS ||
2424 c == OP_CBRA || c == OP_CBRAPOS ||
2425 c == OP_ONCE || c == OP_ONCE_NC ||
2426 c == OP_COND)
2427 {
2428 BOOL empty_branch;
2429 if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2430
2431 /* If a conditional group has only one branch, there is a second, implied,
2432 empty branch, so just skip over the conditional, because it could be empty.
2433 Otherwise, scan the individual branches of the group. */
2434
2435 if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2436 code += GET(code, 1);
2437 else
2438 {
2439 empty_branch = FALSE;
2440 do
2441 {
2442 if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2443 empty_branch = TRUE;
2444 code += GET(code, 1);
2445 }
2446 while (*code == OP_ALT);
2447 if (!empty_branch) return FALSE; /* All branches are non-empty */
2448 }
2449
2450 c = *code;
2451 continue;
2452 }
2453
2454 /* Handle the other opcodes */
2455
2456 switch (c)
2457 {
2458 /* Check for quantifiers after a class. XCLASS is used for classes that
2459 cannot be represented just by a bit map. This includes negated single
2460 high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2461 actual length is stored in the compiled code, so we must update "code"
2462 here. */
2463
2464 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2465 case OP_XCLASS:
2466 ccode = code += GET(code, 1);
2467 goto CHECK_CLASS_REPEAT;
2468 #endif
2469
2470 case OP_CLASS:
2471 case OP_NCLASS:
2472 ccode = code + PRIV(OP_lengths)[OP_CLASS];
2473
2474 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2475 CHECK_CLASS_REPEAT:
2476 #endif
2477
2478 switch (*ccode)
2479 {
2480 case OP_CRSTAR: /* These could be empty; continue */
2481 case OP_CRMINSTAR:
2482 case OP_CRQUERY:
2483 case OP_CRMINQUERY:
2484 case OP_CRPOSSTAR:
2485 case OP_CRPOSQUERY:
2486 break;
2487
2488 default: /* Non-repeat => class must match */
2489 case OP_CRPLUS: /* These repeats aren't empty */
2490 case OP_CRMINPLUS:
2491 case OP_CRPOSPLUS:
2492 return FALSE;
2493
2494 case OP_CRRANGE:
2495 case OP_CRMINRANGE:
2496 case OP_CRPOSRANGE:
2497 if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2498 break;
2499 }
2500 break;
2501
2502 /* Opcodes that must match a character */
2503
2504 case OP_ANY:
2505 case OP_ALLANY:
2506 case OP_ANYBYTE:
2507
2508 case OP_PROP:
2509 case OP_NOTPROP:
2510 case OP_ANYNL:
2511
2512 case OP_NOT_HSPACE:
2513 case OP_HSPACE:
2514 case OP_NOT_VSPACE:
2515 case OP_VSPACE:
2516 case OP_EXTUNI:
2517
2518 case OP_NOT_DIGIT:
2519 case OP_DIGIT:
2520 case OP_NOT_WHITESPACE:
2521 case OP_WHITESPACE:
2522 case OP_NOT_WORDCHAR:
2523 case OP_WORDCHAR:
2524
2525 case OP_CHAR:
2526 case OP_CHARI:
2527 case OP_NOT:
2528 case OP_NOTI:
2529
2530 case OP_PLUS:
2531 case OP_PLUSI:
2532 case OP_MINPLUS:
2533 case OP_MINPLUSI:
2534
2535 case OP_NOTPLUS:
2536 case OP_NOTPLUSI:
2537 case OP_NOTMINPLUS:
2538 case OP_NOTMINPLUSI:
2539
2540 case OP_POSPLUS:
2541 case OP_POSPLUSI:
2542 case OP_NOTPOSPLUS:
2543 case OP_NOTPOSPLUSI:
2544
2545 case OP_EXACT:
2546 case OP_EXACTI:
2547 case OP_NOTEXACT:
2548 case OP_NOTEXACTI:
2549
2550 case OP_TYPEPLUS:
2551 case OP_TYPEMINPLUS:
2552 case OP_TYPEPOSPLUS:
2553 case OP_TYPEEXACT:
2554
2555 return FALSE;
2556
2557 /* These are going to continue, as they may be empty, but we have to
2558 fudge the length for the \p and \P cases. */
2559
2560 case OP_TYPESTAR:
2561 case OP_TYPEMINSTAR:
2562 case OP_TYPEPOSSTAR:
2563 case OP_TYPEQUERY:
2564 case OP_TYPEMINQUERY:
2565 case OP_TYPEPOSQUERY:
2566 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2567 break;
2568
2569 /* Same for these */
2570
2571 case OP_TYPEUPTO:
2572 case OP_TYPEMINUPTO:
2573 case OP_TYPEPOSUPTO:
2574 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2575 code += 2;
2576 break;
2577
2578 /* End of branch */
2579
2580 case OP_KET:
2581 case OP_KETRMAX:
2582 case OP_KETRMIN:
2583 case OP_KETRPOS:
2584 case OP_ALT:
2585 return TRUE;
2586
2587 /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2588 MINUPTO, and POSUPTO and their caseless and negative versions may be
2589 followed by a multibyte character. */
2590
2591 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2592 case OP_STAR:
2593 case OP_STARI:
2594 case OP_NOTSTAR:
2595 case OP_NOTSTARI:
2596
2597 case OP_MINSTAR:
2598 case OP_MINSTARI:
2599 case OP_NOTMINSTAR:
2600 case OP_NOTMINSTARI:
2601
2602 case OP_POSSTAR:
2603 case OP_POSSTARI:
2604 case OP_NOTPOSSTAR:
2605 case OP_NOTPOSSTARI:
2606
2607 case OP_QUERY:
2608 case OP_QUERYI:
2609 case OP_NOTQUERY:
2610 case OP_NOTQUERYI:
2611
2612 case OP_MINQUERY:
2613 case OP_MINQUERYI:
2614 case OP_NOTMINQUERY:
2615 case OP_NOTMINQUERYI:
2616
2617 case OP_POSQUERY:
2618 case OP_POSQUERYI:
2619 case OP_NOTPOSQUERY:
2620 case OP_NOTPOSQUERYI:
2621
2622 if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2623 break;
2624
2625 case OP_UPTO:
2626 case OP_UPTOI:
2627 case OP_NOTUPTO:
2628 case OP_NOTUPTOI:
2629
2630 case OP_MINUPTO:
2631 case OP_MINUPTOI:
2632 case OP_NOTMINUPTO:
2633 case OP_NOTMINUPTOI:
2634
2635 case OP_POSUPTO:
2636 case OP_POSUPTOI:
2637 case OP_NOTPOSUPTO:
2638 case OP_NOTPOSUPTOI:
2639
2640 if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2641 break;
2642 #endif
2643
2644 /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2645 string. */
2646
2647 case OP_MARK:
2648 case OP_PRUNE_ARG:
2649 case OP_SKIP_ARG:
2650 case OP_THEN_ARG:
2651 code += code[1];
2652 break;
2653
2654 /* None of the remaining opcodes are required to match a character. */
2655
2656 default:
2657 break;
2658 }
2659 }
2660
2661 return TRUE;
2662 }
2663
2664
2665
2666 /*************************************************
2667 * Scan compiled regex for non-emptiness *
2668 *************************************************/
2669
2670 /* This function is called to check for left recursive calls. We want to check
2671 the current branch of the current pattern to see if it could match the empty
2672 string. If it could, we must look outwards for branches at other levels,
2673 stopping when we pass beyond the bracket which is the subject of the recursion.
2674 This function is called only during the real compile, not during the
2675 pre-compile.
2676
2677 Arguments:
2678 code points to start of the recursion
2679 endcode points to where to stop (current RECURSE item)
2680 bcptr points to the chain of current (unclosed) branch starts
2681 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2682 cd pointers to tables etc
2683
2684 Returns: TRUE if what is matched could be empty
2685 */
2686
2687 static BOOL
2688 could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2689 branch_chain *bcptr, BOOL utf, compile_data *cd)
2690 {
2691 while (bcptr != NULL && bcptr->current_branch >= code)
2692 {
2693 if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2694 return FALSE;
2695 bcptr = bcptr->outer;
2696 }
2697 return TRUE;
2698 }
2699
2700
2701
2702 /*************************************************
2703 * Base opcode of repeated opcodes *
2704 *************************************************/
2705
2706 /* Returns the base opcode for repeated single character type opcodes. If the
2707 opcode is not a repeated character type, it returns with the original value.
2708
2709 Arguments: c opcode
2710 Returns: base opcode for the type
2711 */
2712
2713 static pcre_uchar
2714 get_repeat_base(pcre_uchar c)
2715 {
2716 return (c > OP_TYPEPOSUPTO)? c :
2717 (c >= OP_TYPESTAR)? OP_TYPESTAR :
2718 (c >= OP_NOTSTARI)? OP_NOTSTARI :
2719 (c >= OP_NOTSTAR)? OP_NOTSTAR :
2720 (c >= OP_STARI)? OP_STARI :
2721 OP_STAR;
2722 }
2723
2724
2725
2726 #ifdef SUPPORT_UCP
2727 /*************************************************
2728 * Check a character and a property *
2729 *************************************************/
2730
2731 /* This function is called by check_auto_possessive() when a property item
2732 is adjacent to a fixed character.
2733
2734 Arguments:
2735 c the character
2736 ptype the property type
2737 pdata the data for the type
2738 negated TRUE if it's a negated property (\P or \p{^)
2739
2740 Returns: TRUE if auto-possessifying is OK
2741 */
2742
2743 static BOOL
2744 check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2745 BOOL negated)
2746 {
2747 const pcre_uint32 *p;
2748 const ucd_record *prop = GET_UCD(c);
2749
2750 switch(ptype)
2751 {
2752 case PT_LAMP:
2753 return (prop->chartype == ucp_Lu ||
2754 prop->chartype == ucp_Ll ||
2755 prop->chartype == ucp_Lt) == negated;
2756
2757 case PT_GC:
2758 return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2759
2760 case PT_PC:
2761 return (pdata == prop->chartype) == negated;
2762
2763 case PT_SC:
2764 return (pdata == prop->script) == negated;
2765
2766 /* These are specials */
2767
2768 case PT_ALNUM:
2769 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2770 PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2771
2772 /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2773 means that Perl space and POSIX space are now identical. PCRE was changed
2774 at release 8.34. */
2775
2776 case PT_SPACE: /* Perl space */
2777 case PT_PXSPACE: /* POSIX space */
2778 switch(c)
2779 {
2780 HSPACE_CASES:
2781 VSPACE_CASES:
2782 return negated;
2783
2784 default:
2785 return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2786 }
2787 break; /* Control never reaches here */
2788
2789 case PT_WORD:
2790 return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2791 PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2792 c == CHAR_UNDERSCORE) == negated;
2793
2794 case PT_CLIST:
2795 p = PRIV(ucd_caseless_sets) + prop->caseset;
2796 for (;;)
2797 {
2798 if (c < *p) return !negated;
2799 if (c == *p++) return negated;
2800 }
2801 break; /* Control never reaches here */
2802 }
2803
2804 return FALSE;
2805 }
2806 #endif /* SUPPORT_UCP */
2807
2808
2809
2810 /*************************************************
2811 * Fill the character property list *
2812 *************************************************/
2813
2814 /* Checks whether the code points to an opcode that can take part in auto-
2815 possessification, and if so, fills a list with its properties.
2816
2817 Arguments:
2818 code points to start of expression
2819 utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2820 fcc points to case-flipping table
2821 list points to output list
2822 list[0] will be filled with the opcode
2823 list[1] will be non-zero if this opcode
2824 can match an empty character string
2825 list[2..7] depends on the opcode
2826
2827 Returns: points to the start of the next opcode if *code is accepted
2828 NULL if *code is not accepted
2829 */
2830
2831 static const pcre_uchar *
2832 get_chr_property_list(const pcre_uchar *code, BOOL utf,
2833 const pcre_uint8 *fcc, pcre_uint32 *list)
2834 {
2835 pcre_uchar c = *code;
2836 const pcre_uchar *end;
2837 const pcre_uint32 *clist_src;
2838 pcre_uint32 *clist_dest;
2839 pcre_uint32 chr;
2840 pcre_uchar base;
2841
2842 list[0] = c;
2843 list[1] = FALSE;
2844 code++;
2845
2846 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2847 {
2848 base = get_repeat_base(c);
2849 c -= (base - OP_STAR);
2850
2851 if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2852 code += IMM2_SIZE;
2853
2854 list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2855
2856 switch(base)
2857 {
2858 case OP_STAR:
2859 list[0] = OP_CHAR;
2860 break;
2861
2862 case OP_STARI:
2863 list[0] = OP_CHARI;
2864 break;
2865
2866 case OP_NOTSTAR:
2867 list[0] = OP_NOT;
2868 break;
2869
2870 case OP_NOTSTARI:
2871 list[0] = OP_NOTI;
2872 break;
2873
2874 case OP_TYPESTAR:
2875 list[0] = *code;
2876 code++;
2877 break;
2878 }
2879 c = list[0];
2880 }
2881
2882 switch(c)
2883 {
2884 case OP_NOT_DIGIT:
2885 case OP_DIGIT:
2886 case OP_NOT_WHITESPACE:
2887 case OP_WHITESPACE:
2888 case OP_NOT_WORDCHAR:
2889 case OP_WORDCHAR:
2890 case OP_ANY:
2891 case OP_ALLANY:
2892 case OP_ANYNL:
2893 case OP_NOT_HSPACE:
2894 case OP_HSPACE:
2895 case OP_NOT_VSPACE:
2896 case OP_VSPACE:
2897 case OP_EXTUNI:
2898 case OP_EODN:
2899 case OP_EOD:
2900 case OP_DOLL:
2901 case OP_DOLLM:
2902 return code;
2903
2904 case OP_CHAR:
2905 case OP_NOT:
2906 GETCHARINCTEST(chr, code);
2907 list[2] = chr;
2908 list[3] = NOTACHAR;
2909 return code;
2910
2911 case OP_CHARI:
2912 case OP_NOTI:
2913 list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2914 GETCHARINCTEST(chr, code);
2915 list[2] = chr;
2916
2917 #ifdef SUPPORT_UCP
2918 if (chr < 128 || (chr < 256 && !utf))
2919 list[3] = fcc[chr];
2920 else
2921 list[3] = UCD_OTHERCASE(chr);
2922 #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2923 list[3] = (chr < 256) ? fcc[chr] : chr;
2924 #else
2925 list[3] = fcc[chr];
2926 #endif
2927
2928 /* The othercase might be the same value. */
2929
2930 if (chr == list[3])
2931 list[3] = NOTACHAR;
2932 else
2933 list[4] = NOTACHAR;
2934 return code;
2935
2936 #ifdef SUPPORT_UCP
2937 case OP_PROP:
2938 case OP_NOTPROP:
2939 if (code[0] != PT_CLIST)
2940 {
2941 list[2] = code[0];
2942 list[3] = code[1];
2943 return code + 2;
2944 }
2945
2946 /* Convert only if we have enough space. */
2947
2948 clist_src = PRIV(ucd_caseless_sets) + code[1];
2949 clist_dest = list + 2;
2950 code += 2;
2951
2952 do {
2953 if (clist_dest >= list + 8)
2954 {
2955 /* Early return if there is not enough space. This should never
2956 happen, since all clists are shorter than 5 character now. */
2957 list[2] = code[0];
2958 list[3] = code[1];
2959 return code;
2960 }
2961 *clist_dest++ = *clist_src;
2962 }
2963 while(*clist_src++ != NOTACHAR);
2964
2965 /* All characters are stored. The terminating NOTACHAR
2966 is copied form the clist itself. */
2967
2968 list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
2969 return code;
2970 #endif
2971
2972 case OP_NCLASS:
2973 case OP_CLASS:
2974 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2975 case OP_XCLASS:
2976
2977 if (c == OP_XCLASS)
2978 end = code + GET(code, 0) - 1;
2979 else
2980 #endif
2981 end = code + 32 / sizeof(pcre_uchar);
2982
2983 switch(*end)
2984 {
2985 case OP_CRSTAR:
2986 case OP_CRMINSTAR:
2987 case OP_CRQUERY:
2988 case OP_CRMINQUERY:
2989 case OP_CRPOSSTAR:
2990 case OP_CRPOSQUERY:
2991 list[1] = TRUE;
2992 end++;
2993 break;
2994
2995 case OP_CRPLUS:
2996 case OP_CRMINPLUS:
2997 case OP_CRPOSPLUS:
2998 end++;
2999 break;
3000
3001 case OP_CRRANGE:
3002 case OP_CRMINRANGE:
3003 case OP_CRPOSRANGE:
3004 list[1] = (GET2(end, 1) == 0);
3005 end += 1 + 2 * IMM2_SIZE;
3006 break;
3007 }
3008 list[2] = end - code;
3009 return end;
3010 }
3011 return NULL; /* Opcode not accepted */
3012 }
3013
3014
3015
3016 /*************************************************
3017 * Scan further character sets for match *
3018 *************************************************/
3019
3020 /* Checks whether the base and the current opcode have a common character, in
3021 which case the base cannot be possessified.
3022
3023 Arguments:
3024 code points to the byte code
3025 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3026 cd static compile data
3027 base_list the data list of the base opcode
3028
3029 Returns: TRUE if the auto-possessification is possible
3030 */
3031
3032 static BOOL
3033 compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3034 const pcre_uint32 *base_list, const pcre_uchar *base_end)
3035 {
3036 pcre_uchar c;
3037 pcre_uint32 list[8];
3038 const pcre_uint32 *chr_ptr;
3039 const pcre_uint32 *ochr_ptr;
3040 const pcre_uint32 *list_ptr;
3041 const pcre_uchar *next_code;
3042 const pcre_uint8 *class_bitset;
3043 const pcre_uint32 *set1, *set2, *set_end;
3044 pcre_uint32 chr;
3045 BOOL accepted, invert_bits;
3046
3047 /* Note: the base_list[1] contains whether the current opcode has greedy
3048 (represented by a non-zero value) quantifier. This is a different from
3049 other character type lists, which stores here that the character iterator
3050 matches to an empty string (also represented by a non-zero value). */
3051
3052 for(;;)
3053 {
3054 /* All operations move the code pointer forward.
3055 Therefore infinite recursions are not possible. */
3056
3057 c = *code;
3058
3059 /* Skip over callouts */
3060
3061 if (c == OP_CALLOUT)
3062 {
3063 code += PRIV(OP_lengths)[c];
3064 continue;
3065 }
3066
3067 if (c == OP_ALT)
3068 {
3069 do code += GET(code, 1); while (*code == OP_ALT);
3070 c = *code;
3071 }
3072
3073 switch(c)
3074 {
3075 case OP_END:
3076 case OP_KETRPOS:
3077 /* TRUE only in greedy case. The non-greedy case could be replaced by
3078 an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3079 uses more memory, which we cannot get at this stage.) */
3080
3081 return base_list[1] != 0;
3082
3083 case OP_KET:
3084 /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3085 it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3086 cannot be converted to a possessive form. */
3087
3088 if (base_list[1] == 0) return FALSE;
3089
3090 switch(*(code - GET(code, 1)))
3091 {
3092 case OP_ASSERT:
3093 case OP_ASSERT_NOT:
3094 case OP_ASSERTBACK:
3095 case OP_ASSERTBACK_NOT:
3096 case OP_ONCE:
3097 case OP_ONCE_NC:
3098 /* Atomic sub-patterns and assertions can always auto-possessify their
3099 last iterator. */
3100 return TRUE;
3101 }
3102
3103 code += PRIV(OP_lengths)[c];
3104 continue;
3105
3106 case OP_ONCE:
3107 case OP_ONCE_NC:
3108 case OP_BRA:
3109 case OP_CBRA:
3110 next_code = code + GET(code, 1);
3111 code += PRIV(OP_lengths)[c];
3112
3113 while (*next_code == OP_ALT)
3114 {
3115 if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3116 code = next_code + 1 + LINK_SIZE;
3117 next_code += GET(next_code, 1);
3118 }
3119 continue;
3120
3121 case OP_BRAZERO:
3122 case OP_BRAMINZERO:
3123
3124 next_code = code + 1;
3125 if (*next_code != OP_BRA && *next_code != OP_CBRA
3126 && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3127
3128 do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3129
3130 /* The bracket content will be checked by the
3131 OP_BRA/OP_CBRA case above. */
3132 next_code += 1 + LINK_SIZE;
3133 if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3134 return FALSE;
3135
3136 code += PRIV(OP_lengths)[c];
3137 continue;
3138 }
3139
3140 /* Check for a supported opcode, and load its properties. */
3141
3142 code = get_chr_property_list(code, utf, cd->fcc, list);
3143 if (code == NULL) return FALSE; /* Unsupported */
3144
3145 /* If either opcode is a small character list, set pointers for comparing
3146 characters from that list with another list, or with a property. */
3147
3148 if (base_list[0] == OP_CHAR)
3149 {
3150 chr_ptr = base_list + 2;
3151 list_ptr = list;
3152 }
3153 else if (list[0] == OP_CHAR)
3154 {
3155 chr_ptr = list + 2;
3156 list_ptr = base_list;
3157 }
3158
3159 /* Character bitsets can also be compared to certain opcodes. */
3160
3161 else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3162 #ifdef COMPILE_PCRE8
3163 /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3164 || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3165 #endif
3166 )
3167 {
3168 #ifdef COMPILE_PCRE8
3169 if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3170 #else
3171 if (base_list[0] == OP_CLASS)
3172 #endif
3173 {
3174 set1 = (pcre_uint32 *)(base_end - base_list[2]);
3175 list_ptr = list;
3176 }
3177 else
3178 {
3179 set1 = (pcre_uint32 *)(code - list[2]);
3180 list_ptr = base_list;
3181 }
3182
3183 invert_bits = FALSE;
3184 switch(list_ptr[0])
3185 {
3186 case OP_CLASS:
3187 case OP_NCLASS:
3188 set2 = (pcre_uint32 *)
3189 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3190 break;
3191
3192 /* OP_XCLASS cannot be supported here, because its bitset
3193 is not necessarily complete. E.g: [a-\0x{200}] is stored
3194 as a character range, and the appropriate bits are not set. */
3195
3196 case OP_NOT_DIGIT:
3197 invert_bits = TRUE;
3198 /* Fall through */
3199 case OP_DIGIT:
3200 set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3201 break;
3202
3203 case OP_NOT_WHITESPACE:
3204 invert_bits = TRUE;
3205 /* Fall through */
3206 case OP_WHITESPACE:
3207 set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3208 break;
3209
3210 case OP_NOT_WORDCHAR:
3211 invert_bits = TRUE;
3212 /* Fall through */
3213 case OP_WORDCHAR:
3214 set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3215 break;
3216
3217 default:
3218 return FALSE;
3219 }
3220
3221 /* Compare 4 bytes to improve speed. */
3222 set_end = set1 + (32 / 4);
3223 if (invert_bits)
3224 {
3225 do
3226 {
3227 if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3228 }
3229 while (set1 < set_end);
3230 }
3231 else
3232 {
3233 do
3234 {
3235 if ((*set1++ & *set2++) != 0) return FALSE;
3236 }
3237 while (set1 < set_end);
3238 }
3239
3240 if (list[1] == 0) return TRUE;
3241 /* Might be an empty repeat. */
3242 continue;
3243 }
3244
3245 /* Some property combinations also acceptable. Unicode property opcodes are
3246 processed specially; the rest can be handled with a lookup table. */
3247
3248 else
3249 {
3250 pcre_uint32 leftop, rightop;
3251
3252 leftop = base_list[0];
3253 rightop = list[0];
3254
3255 #ifdef SUPPORT_UCP
3256 accepted = FALSE; /* Always set in non-unicode case. */
3257 if (leftop == OP_PROP || leftop == OP_NOTPROP)
3258 {
3259 if (rightop == OP_EOD)
3260 accepted = TRUE;
3261 else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3262 {
3263 int n;
3264 const pcre_uint8 *p;
3265 BOOL same = leftop == rightop;
3266 BOOL lisprop = leftop == OP_PROP;
3267 BOOL risprop = rightop == OP_PROP;
3268 BOOL bothprop = lisprop && risprop;
3269
3270 /* There's a table that specifies how each combination is to be
3271 processed:
3272 0 Always return FALSE (never auto-possessify)
3273 1 Character groups are distinct (possessify if both are OP_PROP)
3274 2 Check character categories in the same group (general or particular)
3275 3 Return TRUE if the two opcodes are not the same
3276 ... see comments below
3277 */
3278
3279 n = propposstab[base_list[2]][list[2]];
3280 switch(n)
3281 {
3282 case 0: break;
3283 case 1: accepted = bothprop; break;
3284 case 2: accepted = (base_list[3] == list[3]) != same; break;
3285 case 3: accepted = !same; break;
3286
3287 case 4: /* Left general category, right particular category */
3288 accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3289 break;
3290
3291 case 5: /* Right general category, left particular category */
3292 accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3293 break;
3294
3295 /* This code is logically tricky. Think hard before fiddling with it.
3296 The posspropstab table has four entries per row. Each row relates to
3297 one of PCRE's special properties such as ALNUM or SPACE or WORD.
3298 Only WORD actually needs all four entries, but using repeats for the
3299 others means they can all use the same code below.
3300
3301 The first two entries in each row are Unicode general categories, and
3302 apply always, because all the characters they include are part of the
3303 PCRE character set. The third and fourth entries are a general and a
3304 particular category, respectively, that include one or more relevant
3305 characters. One or the other is used, depending on whether the check
3306 is for a general or a particular category. However, in both cases the
3307 category contains more characters than the specials that are defined
3308 for the property being tested against. Therefore, it cannot be used
3309 in a NOTPROP case.
3310
3311 Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3312 Underscore is covered by ucp_P or ucp_Po. */
3313
3314 case 6: /* Left alphanum vs right general category */
3315 case 7: /* Left space vs right general category */
3316 case 8: /* Left word vs right general category */
3317 p = posspropstab[n-6];
3318 accepted = risprop && lisprop ==
3319 (list[3] != p[0] &&
3320 list[3] != p[1] &&
3321 (list[3] != p[2] || !lisprop));
3322 break;
3323
3324 case 9: /* Right alphanum vs left general category */
3325 case 10: /* Right space vs left general category */
3326 case 11: /* Right word vs left general category */
3327 p = posspropstab[n-9];
3328 accepted = lisprop && risprop ==
3329 (base_list[3] != p[0] &&
3330 base_list[3] != p[1] &&
3331 (base_list[3] != p[2] || !risprop));
3332 break;
3333
3334 case 12: /* Left alphanum vs right particular category */
3335 case 13: /* Left space vs right particular category */
3336 case 14: /* Left word vs right particular category */
3337 p = posspropstab[n-12];
3338 accepted = risprop && lisprop ==
3339 (catposstab[p[0]][list[3]] &&
3340 catposstab[p[1]][list[3]] &&
3341 (list[3] != p[3] || !lisprop));
3342 break;
3343
3344 case 15: /* Right alphanum vs left particular category */
3345 case 16: /* Right space vs left particular category */
3346 case 17: /* Right word vs left particular category */
3347 p = posspropstab[n-15];
3348 accepted = lisprop && risprop ==
3349 (catposstab[p[0]][base_list[3]] &&
3350 catposstab[p[1]][base_list[3]] &&
3351 (base_list[3] != p[3] || !risprop));
3352 break;
3353 }
3354 }
3355 }
3356
3357 else
3358 #endif /* SUPPORT_UCP */
3359
3360 accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3361 rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3362 autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3363
3364 if (!accepted)
3365 return FALSE;
3366
3367 if (list[1] == 0) return TRUE;
3368 /* Might be an empty repeat. */
3369 continue;
3370 }
3371
3372 /* Control reaches here only if one of the items is a small character list.
3373 All characters are checked against the other side. */
3374
3375 do
3376 {
3377 chr = *chr_ptr;
3378
3379 switch(list_ptr[0])
3380 {
3381 case OP_CHAR:
3382 ochr_ptr = list_ptr + 2;
3383 do
3384 {
3385 if (chr == *ochr_ptr) return FALSE;
3386 ochr_ptr++;
3387 }
3388 while(*ochr_ptr != NOTACHAR);
3389 break;
3390
3391 case OP_NOT:
3392 ochr_ptr = list_ptr + 2;
3393 do
3394 {
3395 if (chr == *ochr_ptr)
3396 break;
3397 ochr_ptr++;
3398 }
3399 while(*ochr_ptr != NOTACHAR);
3400 if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3401 break;
3402
3403 /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3404 set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3405
3406 case OP_DIGIT:
3407 if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3408 break;
3409
3410 case OP_NOT_DIGIT:
3411 if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3412 break;
3413
3414 case OP_WHITESPACE:
3415 if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3416 break;
3417
3418 case OP_NOT_WHITESPACE:
3419 if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3420 break;
3421
3422 case OP_WORDCHAR:
3423 if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3424 break;
3425
3426 case OP_NOT_WORDCHAR:
3427 if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3428 break;
3429
3430 case OP_HSPACE:
3431 switch(chr)
3432 {
3433 HSPACE_CASES: return FALSE;
3434 default: break;
3435 }
3436 break;
3437
3438 case OP_NOT_HSPACE:
3439 switch(chr)
3440 {
3441 HSPACE_CASES: break;
3442 default: return FALSE;
3443 }
3444 break;
3445
3446 case OP_ANYNL:
3447 case OP_VSPACE:
3448 switch(chr)
3449 {
3450 VSPACE_CASES: return FALSE;
3451 default: break;
3452 }
3453 break;
3454
3455 case OP_NOT_VSPACE:
3456 switch(chr)
3457 {
3458 VSPACE_CASES: break;
3459 default: return FALSE;
3460 }
3461 break;
3462
3463 case OP_DOLL:
3464 case OP_EODN:
3465 switch (chr)
3466 {
3467 case CHAR_CR:
3468 case CHAR_LF:
3469 case CHAR_VT:
3470 case CHAR_FF:
3471 case CHAR_NEL:
3472 #ifndef EBCDIC
3473 case 0x2028:
3474 case 0x2029:
3475 #endif /* Not EBCDIC */
3476 return FALSE;
3477 }
3478 break;
3479
3480 case OP_EOD: /* Can always possessify before \z */
3481 break;
3482
3483 case OP_PROP:
3484 case OP_NOTPROP:
3485 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3486 list_ptr[0] == OP_NOTPROP))
3487 return FALSE;
3488 break;
3489
3490 case OP_NCLASS:
3491 if (chr > 255) return FALSE;
3492 /* Fall through */
3493
3494 case OP_CLASS:
3495 if (chr > 255) break;
3496 class_bitset = (pcre_uint8 *)
3497 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3498 if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3499 break;
3500
3501 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3502 case OP_XCLASS:
3503 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3504 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3505 break;
3506 #endif
3507
3508 default:
3509 return FALSE;
3510 }
3511
3512 chr_ptr++;
3513 }
3514 while(*chr_ptr != NOTACHAR);
3515
3516 /* At least one character must be matched from this opcode. */
3517
3518 if (list[1] == 0) return TRUE;
3519 }
3520
3521 return FALSE;
3522 }
3523
3524
3525
3526 /*************************************************
3527 * Scan compiled regex for auto-possession *
3528 *************************************************/
3529
3530 /* Replaces single character iterations with their possessive alternatives
3531 if appropriate. This function modifies the compiled opcode!
3532
3533 Arguments:
3534 code points to start of the byte code
3535 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3536 cd static compile data
3537
3538 Returns: nothing
3539 */
3540
3541 static void
3542 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3543 {
3544 register pcre_uchar c;
3545 const pcre_uchar *end;
3546 pcre_uchar *repeat_opcode;
3547 pcre_uint32 list[8];
3548
3549 for (;;)
3550 {
3551 c = *code;
3552
3553 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3554 {
3555 c -= get_repeat_base(c) - OP_STAR;
3556 end = (c <= OP_MINUPTO) ?
3557 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3558 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3559
3560 if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3561 {
3562 switch(c)
3563 {
3564 case OP_STAR:
3565 *code += OP_POSSTAR - OP_STAR;
3566 break;
3567
3568 case OP_MINSTAR:
3569 *code += OP_POSSTAR - OP_MINSTAR;
3570 break;
3571
3572 case OP_PLUS:
3573 *code += OP_POSPLUS - OP_PLUS;
3574 break;
3575
3576 case OP_MINPLUS:
3577 *code += OP_POSPLUS - OP_MINPLUS;
3578 break;
3579
3580 case OP_QUERY:
3581 *code += OP_POSQUERY - OP_QUERY;
3582 break;
3583
3584 case OP_MINQUERY:
3585 *code += OP_POSQUERY - OP_MINQUERY;
3586 break;
3587
3588 case OP_UPTO:
3589 *code += OP_POSUPTO - OP_UPTO;
3590 break;
3591
3592 case OP_MINUPTO:
3593 *code += OP_MINUPTO - OP_UPTO;
3594 break;
3595 }
3596 }
3597 c = *code;
3598 }
3599 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3600 {
3601 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3602 if (c == OP_XCLASS)
3603 repeat_opcode = code + GET(code, 1);
3604 else
3605 #endif
3606 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3607
3608 c = *repeat_opcode;
3609 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3610 {
3611 /* end must not be NULL. */
3612 end = get_chr_property_list(code, utf, cd->fcc, list);
3613
3614 list[1] = (c & 1) == 0;
3615
3616 if (compare_opcodes(end, utf, cd, list, end))
3617 {
3618 switch (c)
3619 {
3620 case OP_CRSTAR:
3621 case OP_CRMINSTAR:
3622 *repeat_opcode = OP_CRPOSSTAR;
3623 break;
3624
3625 case OP_CRPLUS:
3626 case OP_CRMINPLUS:
3627 *repeat_opcode = OP_CRPOSPLUS;
3628 break;
3629
3630 case OP_CRQUERY:
3631 case OP_CRMINQUERY:
3632 *repeat_opcode = OP_CRPOSQUERY;
3633 break;
3634
3635 case OP_CRRANGE:
3636 case OP_CRMINRANGE:
3637 *repeat_opcode = OP_CRPOSRANGE;
3638 break;
3639 }
3640 }
3641 }
3642 c = *code;
3643 }
3644
3645 switch(c)
3646 {
3647 case OP_END:
3648 return;
3649
3650 case OP_TYPESTAR:
3651 case OP_TYPEMINSTAR:
3652 case OP_TYPEPLUS:
3653 case OP_TYPEMINPLUS:
3654 case OP_TYPEQUERY:
3655 case OP_TYPEMINQUERY:
3656 case OP_TYPEPOSSTAR:
3657 case OP_TYPEPOSPLUS:
3658 case OP_TYPEPOSQUERY:
3659 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3660 break;
3661
3662 case OP_TYPEUPTO:
3663 case OP_TYPEMINUPTO:
3664 case OP_TYPEEXACT:
3665 case OP_TYPEPOSUPTO:
3666 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3667 code += 2;
3668 break;
3669
3670 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3671 case OP_XCLASS:
3672 code += GET(code, 1);
3673 break;
3674 #endif
3675
3676 case OP_MARK:
3677 case OP_PRUNE_ARG:
3678 case OP_SKIP_ARG:
3679 case OP_THEN_ARG:
3680 code += code[1];
3681 break;
3682 }
3683
3684 /* Add in the fixed length from the table */
3685
3686 code += PRIV(OP_lengths)[c];
3687
3688 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3689 a multi-byte character. The length in the table is a minimum, so we have to
3690 arrange to skip the extra bytes. */
3691
3692 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3693 if (utf) switch(c)
3694 {
3695 case OP_CHAR:
3696 case OP_CHARI:
3697 case OP_NOT:
3698 case OP_NOTI:
3699 case OP_STAR:
3700 case OP_MINSTAR:
3701 case OP_PLUS:
3702 case OP_MINPLUS:
3703 case OP_QUERY:
3704 case OP_MINQUERY:
3705 case OP_UPTO:
3706 case OP_MINUPTO:
3707 case OP_EXACT:
3708 case OP_POSSTAR:
3709 case OP_POSPLUS:
3710 case OP_POSQUERY:
3711 case OP_POSUPTO:
3712 case OP_STARI:
3713 case OP_MINSTARI:
3714 case OP_PLUSI:
3715 case OP_MINPLUSI:
3716 case OP_QUERYI:
3717 case OP_MINQUERYI:
3718 case OP_UPTOI:
3719 case OP_MINUPTOI:
3720 case OP_EXACTI:
3721 case OP_POSSTARI:
3722 case OP_POSPLUSI:
3723 case OP_POSQUERYI:
3724 case OP_POSUPTOI:
3725 case OP_NOTSTAR:
3726 case OP_NOTMINSTAR:
3727 case OP_NOTPLUS:
3728 case OP_NOTMINPLUS:
3729 case OP_NOTQUERY:
3730 case OP_NOTMINQUERY:
3731 case OP_NOTUPTO:
3732 case OP_NOTMINUPTO:
3733 case OP_NOTEXACT:
3734 case OP_NOTPOSSTAR:
3735 case OP_NOTPOSPLUS:
3736 case OP_NOTPOSQUERY:
3737 case OP_NOTPOSUPTO:
3738 case OP_NOTSTARI:
3739 case OP_NOTMINSTARI:
3740 case OP_NOTPLUSI:
3741 case OP_NOTMINPLUSI:
3742 case OP_NOTQUERYI:
3743 case OP_NOTMINQUERYI:
3744 case OP_NOTUPTOI:
3745 case OP_NOTMINUPTOI:
3746 case OP_NOTEXACTI:
3747 case OP_NOTPOSSTARI:
3748 case OP_NOTPOSPLUSI:
3749 case OP_NOTPOSQUERYI:
3750 case OP_NOTPOSUPTOI:
3751 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3752 break;
3753 }
3754 #else
3755 (void)(utf); /* Keep compiler happy by referencing function argument */
3756 #endif
3757 }
3758 }
3759
3760
3761
3762 /*************************************************
3763 * Check for POSIX class syntax *
3764 *************************************************/
3765
3766 /* This function is called when the sequence "[:" or "[." or "[=" is
3767 encountered in a character class. It checks whether this is followed by a
3768 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3769 reach an unescaped ']' without the special preceding character, return FALSE.
3770
3771 Originally, this function only recognized a sequence of letters between the
3772 terminators, but it seems that Perl recognizes any sequence of characters,
3773 though of course unknown POSIX names are subsequently rejected. Perl gives an
3774 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3775 didn't consider this to be a POSIX class. Likewise for [:1234:].
3776
3777 The problem in trying to be exactly like Perl is in the handling of escapes. We
3778 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3779 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3780 below handles the special case of \], but does not try to do any other escape
3781 processing. This makes it different from Perl for cases such as [:l\ower:]
3782 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3783 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3784 I think.
3785
3786 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3787 It seems that the appearance of a nested POSIX class supersedes an apparent
3788 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3789 a digit.
3790
3791 In Perl, unescaped square brackets may also appear as part of class names. For
3792 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3793 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3794 seem right at all. PCRE does not allow closing square brackets in POSIX class
3795 names.
3796
3797 Arguments:
3798 ptr pointer to the initial [
3799 endptr where to return the end pointer
3800
3801 Returns: TRUE or FALSE
3802 */
3803
3804 static BOOL
3805 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3806 {
3807 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3808 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3809 for (++ptr; *ptr != CHAR_NULL; ptr++)
3810 {
3811 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3812 ptr++;
3813 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3814 else
3815 {
3816 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3817 {
3818 *endptr = ptr;
3819 return TRUE;
3820 }
3821 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3822 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3823 ptr[1] == CHAR_EQUALS_SIGN) &&
3824 check_posix_syntax(ptr, endptr))
3825 return FALSE;
3826 }
3827 }
3828 return FALSE;
3829 }
3830
3831
3832
3833
3834 /*************************************************
3835 * Check POSIX class name *
3836 *************************************************/
3837
3838 /* This function is called to check the name given in a POSIX-style class entry
3839 such as [:alnum:].
3840
3841 Arguments:
3842 ptr points to the first letter
3843 len the length of the name
3844
3845 Returns: a value representing the name, or -1 if unknown
3846 */
3847
3848 static int
3849 check_posix_name(const pcre_uchar *ptr, int len)
3850 {
3851 const char *pn = posix_names;
3852 register int yield = 0;
3853 while (posix_name_lengths[yield] != 0)
3854 {
3855 if (len == posix_name_lengths[yield] &&
3856 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3857 pn += posix_name_lengths[yield] + 1;
3858 yield++;
3859 }
3860 return -1;
3861 }
3862
3863
3864 /*************************************************
3865 * Adjust OP_RECURSE items in repeated group *
3866 *************************************************/
3867
3868 /* OP_RECURSE items contain an offset from the start of the regex to the group
3869 that is referenced. This means that groups can be replicated for fixed
3870 repetition simply by copying (because the recursion is allowed to refer to
3871 earlier groups that are outside the current group). However, when a group is
3872 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3873 inserted before it, after it has been compiled. This means that any OP_RECURSE
3874 items within it that refer to the group itself or any contained groups have to
3875 have their offsets adjusted. That one of the jobs of this function. Before it
3876 is called, the partially compiled regex must be temporarily terminated with
3877 OP_END.
3878
3879 This function has been extended with the possibility of forward references for
3880 recursions and subroutine calls. It must also check the list of such references
3881 for the group we are dealing with. If it finds that one of the recursions in
3882 the current group is on this list, it adjusts the offset in the list, not the
3883 value in the reference (which is a group number).
3884
3885 Arguments:
3886 group points to the start of the group
3887 adjust the amount by which the group is to be moved
3888 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3889 cd contains pointers to tables etc.
3890 save_hwm the hwm forward reference pointer at the start of the group
3891
3892 Returns: nothing
3893 */
3894
3895 static void
3896 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3897 pcre_uchar *save_hwm)
3898 {
3899 pcre_uchar *ptr = group;
3900
3901 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3902 {
3903 int offset;
3904 pcre_uchar *hc;
3905
3906 /* See if this recursion is on the forward reference list. If so, adjust the
3907 reference. */
3908
3909 for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3910 {
3911 offset = (int)GET(hc, 0);
3912 if (cd->start_code + offset == ptr + 1)
3913 {
3914 PUT(hc, 0, offset + adjust);
3915 break;
3916 }
3917 }
3918
3919 /* Otherwise, adjust the recursion offset if it's after the start of this
3920 group. */
3921
3922 if (hc >= cd->hwm)
3923 {
3924 offset = (int)GET(ptr, 1);
3925 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3926 }
3927
3928 ptr += 1 + LINK_SIZE;
3929 }
3930 }
3931
3932
3933
3934 /*************************************************
3935 * Insert an automatic callout point *
3936 *************************************************/
3937
3938 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3939 callout points before each pattern item.
3940
3941 Arguments:
3942 code current code pointer
3943 ptr current pattern pointer
3944 cd pointers to tables etc
3945
3946 Returns: new code pointer
3947 */
3948
3949 static pcre_uchar *
3950 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3951 {
3952 *code++ = OP_CALLOUT;
3953 *code++ = 255;
3954 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
3955 PUT(code, LINK_SIZE, 0); /* Default length */
3956 return code + 2 * LINK_SIZE;
3957 }
3958
3959
3960
3961 /*************************************************
3962 * Complete a callout item *
3963 *************************************************/
3964
3965 /* A callout item contains the length of the next item in the pattern, which
3966 we can't fill in till after we have reached the relevant point. This is used
3967 for both automatic and manual callouts.
3968
3969 Arguments:
3970 previous_callout points to previous callout item
3971 ptr current pattern pointer
3972 cd pointers to tables etc
3973
3974 Returns: nothing
3975 */
3976
3977 static void
3978 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3979 {
3980 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3981 PUT(previous_callout, 2 + LINK_SIZE, length);
3982 }
3983
3984
3985
3986 #ifdef SUPPORT_UCP
3987 /*************************************************
3988 * Get othercase range *
3989 *************************************************/
3990
3991 /* This function is passed the start and end of a class range, in UTF-8 mode
3992 with UCP support. It searches up the characters, looking for ranges of
3993 characters in the "other" case. Each call returns the next one, updating the
3994 start address. A character with multiple other cases is returned on its own
3995 with a special return value.
3996
3997 Arguments:
3998 cptr points to starting character value; updated
3999 d end value
4000 ocptr where to put start of othercase range
4001 odptr where to put end of othercase range
4002
4003 Yield: -1 when no more
4004 0 when a range is returned
4005 >0 the CASESET offset for char with multiple other cases
4006 in this case, ocptr contains the original
4007 */
4008
4009 static int
4010 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4011 pcre_uint32 *odptr)
4012 {
4013 pcre_uint32 c, othercase, next;
4014 unsigned int co;
4015
4016 /* Find the first character that has an other case. If it has multiple other
4017 cases, return its case offset value. */
4018
4019 for (c = *cptr; c <= d; c++)
4020 {
4021 if ((co = UCD_CASESET(c)) != 0)
4022 {
4023 *ocptr = c++; /* Character that has the set */
4024 *cptr = c; /* Rest of input range */
4025 return (int)co;
4026 }
4027 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4028 }
4029
4030 if (c > d) return -1; /* Reached end of range */
4031
4032 *ocptr = othercase;
4033 next = othercase + 1;
4034
4035 for (++c; c <= d; c++)
4036 {
4037 if (UCD_OTHERCASE(c) != next) break;
4038 next++;
4039 }
4040
4041 *odptr = next - 1; /* End of othercase range */
4042 *cptr = c; /* Rest of input range */
4043 return 0;
4044 }
4045 #endif /* SUPPORT_UCP */
4046
4047
4048
4049 /*************************************************
4050 * Add a character or range to a class *
4051 *************************************************/
4052
4053 /* This function packages up the logic of adding a character or range of
4054 characters to a class. The character values in the arguments will be within the
4055 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4056 mutually recursive with the function immediately below.
4057
4058 Arguments:
4059 classbits the bit map for characters < 256
4060 uchardptr points to the pointer for extra data
4061 options the options word
4062 cd contains pointers to tables etc.
4063 start start of range character
4064 end end of range character
4065
4066 Returns: the number of < 256 characters added
4067 the pointer to extra data is updated
4068 */
4069
4070 static int
4071 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4072 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4073 {
4074 pcre_uint32 c;
4075 int n8 = 0;
4076
4077 /* If caseless matching is required, scan the range and process alternate
4078 cases. In Unicode, there are 8-bit characters that have alternate cases that
4079 are greater than 255 and vice-versa. Sometimes we can just extend the original
4080 range. */
4081
4082 if ((options & PCRE_CASELESS) != 0)
4083 {
4084 #ifdef SUPPORT_UCP
4085 if ((options & PCRE_UTF8) != 0)
4086 {
4087 int rc;
4088 pcre_uint32 oc, od;
4089
4090 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4091 c = start;
4092
4093 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4094 {
4095 /* Handle a single character that has more than one other case. */
4096
4097 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4098 PRIV(ucd_caseless_sets) + rc, oc);
4099
4100 /* Do nothing if the other case range is within the original range. */
4101
4102 else if (oc >= start && od <= end) continue;
4103
4104 /* Extend the original range if there is overlap, noting that if oc < c, we
4105 can't have od > end because a subrange is always shorter than the basic
4106 range. Otherwise, use a recursive call to add the additional range. */
4107
4108 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4109 else if (od > end && oc <= end + 1) end = od; /* Extend upwards */
4110 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4111 }
4112 }
4113 else
4114 #endif /* SUPPORT_UCP */
4115
4116 /* Not UTF-mode, or no UCP */
4117
4118 for (c = start; c <= end && c < 256; c++)
4119 {
4120 SETBIT(classbits, cd->fcc[c]);
4121 n8++;
4122 }
4123 }
4124
4125 /* Now handle the original range. Adjust the final value according to the bit
4126 length - this means that the same lists of (e.g.) horizontal spaces can be used
4127 in all cases. */
4128
4129 #if defined COMPILE_PCRE8
4130 #ifdef SUPPORT_UTF
4131 if ((options & PCRE_UTF8) == 0)
4132 #endif
4133 if (end > 0xff) end = 0xff;
4134
4135 #elif defined COMPILE_PCRE16
4136 #ifdef SUPPORT_UTF
4137 if ((options & PCRE_UTF16) == 0)
4138 #endif
4139 if (end > 0xffff) end = 0xffff;
4140
4141 #endif /* COMPILE_PCRE[8|16] */
4142
4143 /* If all characters are less than 256, use the bit map. Otherwise use extra
4144 data. */
4145
4146 if (end < 0x100)
4147 {
4148 for (c = start; c <= end; c++)
4149 {
4150 n8++;
4151 SETBIT(classbits, c);
4152 }
4153 }
4154
4155 else
4156 {
4157 pcre_uchar *uchardata = *uchardptr;
4158
4159 #ifdef SUPPORT_UTF
4160 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4161 {
4162 if (start < end)
4163 {
4164 *uchardata++ = XCL_RANGE;
4165 uchardata += PRIV(ord2utf)(start, uchardata);
4166 uchardata += PRIV(ord2utf)(end, uchardata);
4167 }
4168 else if (start == end)
4169 {
4170 *uchardata++ = XCL_SINGLE;
4171 uchardata += PRIV(ord2utf)(start, uchardata);
4172 }
4173 }
4174 else
4175 #endif /* SUPPORT_UTF */
4176
4177 /* Without UTF support, character values are constrained by the bit length,
4178 and can only be > 256 for 16-bit and 32-bit libraries. */
4179
4180 #ifdef COMPILE_PCRE8
4181 {}
4182 #else
4183 if (start < end)
4184 {
4185 *uchardata++ = XCL_RANGE;
4186 *uchardata++ = start;
4187 *uchardata++ = end;
4188 }
4189 else if (start == end)
4190 {
4191 *uchardata++ = XCL_SINGLE;
4192 *uchardata++ = start;
4193 }
4194 #endif
4195
4196 *uchardptr = uchardata; /* Updata extra data pointer */
4197 }
4198
4199 return n8; /* Number of 8-bit characters */
4200 }
4201
4202
4203
4204
4205 /*************************************************
4206 * Add a list of characters to a class *
4207 *************************************************/
4208
4209 /* This function is used for adding a list of case-equivalent characters to a
4210 class, and also for adding a list of horizontal or vertical whitespace. If the
4211 list is in order (which it should be), ranges of characters are detected and
4212 handled appropriately. This function is mutually recursive with the function
4213 above.
4214
4215 Arguments:
4216 classbits the bit map for characters < 256
4217 uchardptr points to the pointer for extra data
4218 options the options word
4219 cd contains pointers to tables etc.
4220 p points to row of 32-bit values, terminated by NOTACHAR
4221 except character to omit; this is used when adding lists of
4222 case-equivalent characters to avoid including the one we
4223 already know about
4224
4225 Returns: the number of < 256 characters added
4226 the pointer to extra data is updated
4227 */
4228
4229 static int
4230 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4231 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4232 {
4233 int n8 = 0;
4234 while (p[0] < NOTACHAR)
4235 {
4236 int n = 0;
4237 if (p[0] != except)
4238 {
4239 while(p[n+1] == p[0] + n + 1) n++;
4240 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4241 }
4242 p += n + 1;
4243 }
4244 return n8;
4245 }
4246
4247
4248
4249 /*************************************************
4250 * Add characters not in a list to a class *
4251 *************************************************/
4252
4253 /* This function is used for adding the complement of a list of horizontal or
4254 vertical whitespace to a class. The list must be in order.
4255
4256 Arguments:
4257 classbits the bit map for characters < 256
4258 uchardptr points to the pointer for extra data
4259 options the options word
4260 cd contains pointers to tables etc.
4261 p points to row of 32-bit values, terminated by NOTACHAR
4262
4263 Returns: the number of < 256 characters added
4264 the pointer to extra data is updated
4265 */
4266
4267 static int
4268 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4269 int options, compile_data *cd, const pcre_uint32 *p)
4270 {
4271 BOOL utf = (options & PCRE_UTF8) != 0;
4272 int n8 = 0;
4273 if (p[0] > 0)
4274 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4275 while (p[0] < NOTACHAR)
4276 {
4277 while (p[1] == p[0] + 1) p++;
4278 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4279 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4280 p++;
4281 }
4282 return n8;
4283 }
4284
4285
4286
4287 /*************************************************
4288 * Compile one branch *
4289 *************************************************/
4290
4291 /* Scan the pattern, compiling it into the a vector. If the options are
4292 changed during the branch, the pointer is used to change the external options
4293 bits. This function is used during the pre-compile phase when we are trying
4294 to find out the amount of memory needed, as well as during the real compile
4295 phase. The value of lengthptr distinguishes the two phases.
4296
4297 Arguments:
4298 optionsptr pointer to the option bits
4299 codeptr points to the pointer to the current code point
4300 ptrptr points to the current pattern pointer
4301 errorcodeptr points to error code variable
4302 firstcharptr place to put the first required character
4303 firstcharflagsptr place to put the first character flags, or a negative number
4304 reqcharptr place to put the last required character
4305 reqcharflagsptr place to put the last required character flags, or a negative number
4306 bcptr points to current branch chain
4307 cond_depth conditional nesting depth
4308 cd contains pointers to tables etc.
4309 lengthptr NULL during the real compile phase
4310 points to length accumulator during pre-compile phase
4311
4312 Returns: TRUE on success
4313 FALSE, with *errorcodeptr set non-zero on error
4314 */
4315
4316 static BOOL
4317 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4318 const pcre_uchar **ptrptr, int *errorcodeptr,
4319 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4320 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4321 branch_chain *bcptr, int cond_depth,
4322 compile_data *cd, int *lengthptr)
4323 {
4324 int repeat_type, op_type;
4325 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4326 int bravalue = 0;
4327 int greedy_default, greedy_non_default;
4328 pcre_uint32 firstchar, reqchar;
4329 pcre_int32 firstcharflags, reqcharflags;
4330 pcre_uint32 zeroreqchar, zerofirstchar;
4331 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4332 pcre_int32 req_caseopt, reqvary, tempreqvary;
4333 int options = *optionsptr; /* May change dynamically */
4334 int after_manual_callout = 0;
4335 int length_prevgroup = 0;
4336 register pcre_uint32 c;
4337 int escape;
4338 register pcre_uchar *code = *codeptr;
4339 pcre_uchar *last_code = code;
4340 pcre_uchar *orig_code = code;
4341 pcre_uchar *tempcode;
4342 BOOL inescq = FALSE;
4343 BOOL groupsetfirstchar = FALSE;
4344 const pcre_uchar *ptr = *ptrptr;
4345 const pcre_uchar *tempptr;
4346 const pcre_uchar *nestptr = NULL;
4347 pcre_uchar *previous = NULL;
4348 pcre_uchar *previous_callout = NULL;
4349 pcre_uchar *save_hwm = NULL;
4350 pcre_uint8 classbits[32];
4351
4352 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4353 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4354 dynamically as we process the pattern. */
4355
4356 #ifdef SUPPORT_UTF
4357 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4358 BOOL utf = (options & PCRE_UTF8) != 0;
4359 #ifndef COMPILE_PCRE32
4360 pcre_uchar utf_chars[6];
4361 #endif
4362 #else
4363 BOOL utf = FALSE;
4364 #endif
4365
4366 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4367 class_uchardata always so that it can be passed to add_to_class() always,
4368 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4369 alternative calls for the different cases. */
4370
4371 pcre_uchar *class_uchardata;
4372 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4373 BOOL xclass;
4374 pcre_uchar *class_uchardata_base;
4375 #endif
4376
4377 #ifdef PCRE_DEBUG
4378 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4379 #endif
4380
4381 /* Set up the default and non-default settings for greediness */
4382
4383 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4384 greedy_non_default = greedy_default ^ 1;
4385
4386 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4387 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4388 matches a non-fixed char first char; reqchar just remains unset if we never
4389 find one.
4390
4391 When we hit a repeat whose minimum is zero, we may have to adjust these values
4392 to take the zero repeat into account. This is implemented by setting them to
4393 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4394 item types that can be repeated set these backoff variables appropriately. */
4395
4396 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4397 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4398
4399 /* The variable req_caseopt contains either the REQ_CASELESS value
4400 or zero, according to the current setting of the caseless flag. The
4401 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4402 firstchar or reqchar variables to record the case status of the
4403 value. This is used only for ASCII characters. */
4404
4405 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4406
4407 /* Switch on next character until the end of the branch */
4408
4409 for (;; ptr++)
4410 {
4411 BOOL negate_class;
4412 BOOL should_flip_negation;
4413 BOOL possessive_quantifier;
4414 BOOL is_quantifier;
4415 BOOL is_recurse;
4416 BOOL reset_bracount;
4417 int class_has_8bitchar;
4418 int class_one_char;
4419 int newoptions;
4420 int recno;
4421 int refsign;
4422 int skipbytes;
4423 pcre_uint32 subreqchar, subfirstchar;
4424 pcre_int32 subreqcharflags, subfirstcharflags;
4425 int terminator;
4426 unsigned int mclength;
4427 unsigned int tempbracount;
4428 pcre_uint32 ec;
4429 pcre_uchar mcbuffer[8];
4430
4431 /* Get next character in the pattern */
4432
4433 c = *ptr;
4434
4435 /* If we are at the end of a nested substitution, revert to the outer level
4436 string. Nesting only happens one level deep. */
4437
4438 if (c == CHAR_NULL && nestptr != NULL)
4439 {
4440 ptr = nestptr;
4441 nestptr = NULL;
4442 c = *ptr;
4443 }
4444
4445 /* If we are in the pre-compile phase, accumulate the length used for the
4446 previous cycle of this loop. */
4447
4448 if (lengthptr != NULL)
4449 {
4450 #ifdef PCRE_DEBUG
4451 if (code > cd->hwm) cd->hwm = code; /* High water info */
4452 #endif
4453 if (code > cd->start_workspace + cd->workspace_size -
4454 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4455 {
4456 *errorcodeptr = ERR52;
4457 goto FAILED;
4458 }
4459
4460 /* There is at least one situation where code goes backwards: this is the
4461 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4462 the class is simply eliminated. However, it is created first, so we have to
4463 allow memory for it. Therefore, don't ever reduce the length at this point.
4464 */
4465
4466 if (code < last_code) code = last_code;
4467
4468 /* Paranoid check for integer overflow */
4469
4470 if (OFLOW_MAX - *lengthptr < code - last_code)
4471 {
4472 *errorcodeptr = ERR20;
4473 goto FAILED;
4474 }
4475
4476 *lengthptr += (int)(code - last_code);
4477 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4478 (int)(code - last_code), c, c));
4479
4480 /* If "previous" is set and it is not at the start of the work space, move
4481 it back to there, in order to avoid filling up the work space. Otherwise,
4482 if "previous" is NULL, reset the current code pointer to the start. */
4483
4484 if (previous != NULL)
4485 {
4486 if (previous > orig_code)
4487 {
4488 memmove(orig_code, previous, IN_UCHARS(code - previous));
4489 code -= previous - orig_code;
4490 previous = orig_code;
4491 }
4492 }
4493 else code = orig_code;
4494
4495 /* Remember where this code item starts so we can pick up the length
4496 next time round. */
4497
4498 last_code = code;
4499 }
4500
4501 /* In the real compile phase, just check the workspace used by the forward
4502 reference list. */
4503
4504 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4505 WORK_SIZE_SAFETY_MARGIN)
4506 {
4507 *errorcodeptr = ERR52;
4508 goto FAILED;
4509 }
4510
4511 /* If in \Q...\E, check for the end; if not, we have a literal */
4512
4513 if (inescq && c != CHAR_NULL)
4514 {
4515 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4516 {
4517 inescq = FALSE;
4518 ptr++;
4519 continue;
4520 }
4521 else
4522 {
4523 if (previous_callout != NULL)
4524 {
4525 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4526 complete_callout(previous_callout, ptr, cd);
4527 previous_callout = NULL;
4528 }
4529 if ((options & PCRE_AUTO_CALLOUT) != 0)
4530 {
4531 previous_callout = code;
4532 code = auto_callout(code, ptr, cd);
4533 }
4534 goto NORMAL_CHAR;
4535 }
4536 }
4537
4538 is_quantifier =
4539 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4540 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4541
4542 /* Fill in length of a previous callout, except when the next thing is a
4543 quantifier or when processing a property substitution string in UCP mode. */
4544
4545 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4546 after_manual_callout-- <= 0)
4547 {
4548 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4549 complete_callout(previous_callout, ptr, cd);
4550 previous_callout = NULL;
4551 }
4552
4553 /* In extended mode, skip white space and comments. */
4554
4555 if ((options & PCRE_EXTENDED) != 0)
4556 {
4557 if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
4558 if (c == CHAR_NUMBER_SIGN)
4559 {
4560 ptr++;
4561 while (*ptr != CHAR_NULL)
4562 {
4563 if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
4564 ptr++;
4565 #ifdef SUPPORT_UTF
4566 if (utf) FORWARDCHAR(ptr);
4567 #endif
4568 }
4569 if (*ptr != CHAR_NULL) continue;
4570
4571 /* Else fall through to handle end of string */
4572 c = 0;
4573 }
4574 }
4575
4576 /* No auto callout for quantifiers, or while processing property strings that
4577 are substituted for \w etc in UCP mode. */
4578
4579 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4580 {
4581 previous_callout = code;
4582 code = auto_callout(code, ptr, cd);
4583 }
4584
4585 switch(c)
4586 {
4587 /* ===================================================================*/
4588 case 0: /* The branch terminates at string end */
4589 case CHAR_VERTICAL_LINE: /* or | or ) */
4590 case CHAR_RIGHT_PARENTHESIS:
4591 *firstcharptr = firstchar;
4592 *firstcharflagsptr = firstcharflags;
4593 *reqcharptr = reqchar;
4594 *reqcharflagsptr = reqcharflags;
4595 *codeptr = code;
4596 *ptrptr = ptr;
4597 if (lengthptr != NULL)
4598 {
4599 if (OFLOW_MAX - *lengthptr < code - last_code)
4600 {
4601 *errorcodeptr = ERR20;
4602 goto FAILED;
4603 }
4604 *lengthptr += (int)(code - last_code); /* To include callout length */
4605 DPRINTF((">> end branch\n"));
4606 }
4607 return TRUE;
4608
4609
4610 /* ===================================================================*/
4611 /* Handle single-character metacharacters. In multiline mode, ^ disables
4612 the setting of any following char as a first character. */
4613
4614 case CHAR_CIRCUMFLEX_ACCENT:
4615 previous = NULL;
4616 if ((options & PCRE_MULTILINE) != 0)
4617 {
4618 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4619 *code++ = OP_CIRCM;
4620 }
4621 else *code++ = OP_CIRC;
4622 break;
4623
4624 case CHAR_DOLLAR_SIGN:
4625 previous = NULL;
4626 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4627 break;
4628
4629 /* There can never be a first char if '.' is first, whatever happens about
4630 repeats. The value of reqchar doesn't change either. */
4631
4632 case CHAR_DOT:
4633 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4634 zerofirstchar = firstchar;
4635 zerofirstcharflags = firstcharflags;
4636 zeroreqchar = reqchar;
4637 zeroreqcharflags = reqcharflags;
4638 previous = code;
4639 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4640 break;
4641
4642
4643 /* ===================================================================*/
4644 /* Character classes. If the included characters are all < 256, we build a
4645 32-byte bitmap of the permitted characters, except in the special case
4646 where there is only one such character. For negated classes, we build the
4647 map as usual, then invert it at the end. However, we use a different opcode
4648 so that data characters > 255 can be handled correctly.
4649
4650 If the class contains characters outside the 0-255 range, a different
4651 opcode is compiled. It may optionally have a bit map for characters < 256,
4652 but those above are are explicitly listed afterwards. A flag byte tells
4653 whether the bitmap is present, and whether this is a negated class or not.
4654
4655 In JavaScript compatibility mode, an isolated ']' causes an error. In
4656 default (Perl) mode, it is treated as a data character. */
4657
4658 case CHAR_RIGHT_SQUARE_BRACKET:
4659 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4660 {
4661 *errorcodeptr = ERR64;
4662 goto FAILED;
4663 }
4664 goto NORMAL_CHAR;
4665
4666 case CHAR_LEFT_SQUARE_BRACKET:
4667 previous = code;
4668
4669 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4670 they are encountered at the top level, so we'll do that too. */
4671
4672 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4673 ptr[1] == CHAR_EQUALS_SIGN) &&
4674 check_posix_syntax(ptr, &tempptr))
4675 {
4676 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4677 goto FAILED;
4678 }
4679
4680 /* If the first character is '^', set the negation flag and skip it. Also,
4681 if the first few characters (either before or after ^) are \Q\E or \E we
4682 skip them too. This makes for compatibility with Perl. */
4683
4684 negate_class = FALSE;
4685 for (;;)
4686 {
4687 c = *(++ptr);
4688 if (c == CHAR_BACKSLASH)
4689 {
4690 if (ptr[1] == CHAR_E)
4691 ptr++;
4692 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4693 ptr += 3;
4694 else
4695 break;
4696 }
4697 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4698 negate_class = TRUE;
4699 else break;
4700 }
4701
4702 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4703 an initial ']' is taken as a data character -- the code below handles
4704 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4705 [^] must match any character, so generate OP_ALLANY. */
4706
4707 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4708 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4709 {
4710 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4711 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4712 zerofirstchar = firstchar;
4713 zerofirstcharflags = firstcharflags;
4714 break;
4715 }
4716
4717 /* If a class contains a negative special such as \S, we need to flip the
4718 negation flag at the end, so that support for characters > 255 works
4719 correctly (they are all included in the class). */
4720
4721 should_flip_negation = FALSE;
4722
4723 /* For optimization purposes, we track some properties of the class:
4724 class_has_8bitchar will be non-zero if the class contains at least one <
4725 256 character; class_one_char will be 1 if the class contains just one
4726 character. */
4727
4728 class_has_8bitchar = 0;
4729 class_one_char = 0;
4730
4731 /* Initialize the 32-char bit map to all zeros. We build the map in a
4732 temporary bit of memory, in case the class contains fewer than two
4733 8-bit characters because in that case the compiled code doesn't use the bit
4734 map. */
4735
4736 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4737
4738 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4739 xclass = FALSE;
4740 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4741 class_uchardata_base = class_uchardata; /* Save the start */
4742 #endif
4743
4744 /* Process characters until ] is reached. By writing this as a "do" it
4745 means that an initial ] is taken as a data character. At the start of the
4746 loop, c contains the first byte of the character. */
4747
4748 if (c != CHAR_NULL) do
4749 {
4750 const pcre_uchar *oldptr;
4751
4752 #ifdef SUPPORT_UTF
4753 if (utf && HAS_EXTRALEN(c))
4754 { /* Braces are required because the */
4755 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4756 }
4757 #endif
4758
4759 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4760 /* In the pre-compile phase, accumulate the length of any extra
4761 data and reset the pointer. This is so that very large classes that
4762 contain a zillion > 255 characters no longer overwrite the work space
4763 (which is on the stack). We have to remember that there was XCLASS data,
4764 however. */
4765
4766 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4767 {
4768 xclass = TRUE;
4769 *lengthptr += class_uchardata - class_uchardata_base;
4770 class_uchardata = class_uchardata_base;
4771 }
4772 #endif
4773
4774 /* Inside \Q...\E everything is literal except \E */
4775
4776 if (inescq)
4777 {
4778 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4779 {
4780 inescq = FALSE; /* Reset literal state */
4781 ptr++; /* Skip the 'E' */
4782 continue; /* Carry on with next */
4783 }
4784 goto CHECK_RANGE; /* Could be range if \E follows */
4785 }
4786
4787 /* Handle POSIX class names. Perl allows a negation extension of the
4788 form [:^name:]. A square bracket that doesn't match the syntax is
4789 treated as a literal. We also recognize the POSIX constructions
4790 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4791 5.6 and 5.8 do. */
4792
4793 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4794 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4795 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4796 {
4797 BOOL local_negate = FALSE;
4798 int posix_class, taboffset, tabopt;
4799 register const pcre_uint8 *cbits = cd->cbits;
4800 pcre_uint8 pbits[32];
4801
4802 if (ptr[1] != CHAR_COLON)
4803 {
4804 *errorcodeptr = ERR31;
4805 goto FAILED;
4806 }
4807
4808 ptr += 2;
4809 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4810 {
4811 local_negate = TRUE;
4812 should_flip_negation = TRUE; /* Note negative special */
4813 ptr++;
4814 }
4815
4816 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4817 if (posix_class < 0)
4818 {
4819 *errorcodeptr = ERR30;
4820 goto FAILED;
4821 }
4822
4823 /* If matching is caseless, upper and lower are converted to
4824 alpha. This relies on the fact that the class table starts with
4825 alpha, lower, upper as the first 3 entries. */
4826
4827 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4828 posix_class = 0;
4829
4830 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4831 different escape sequences that use Unicode properties. */
4832
4833 #ifdef SUPPORT_UCP
4834 if ((options & PCRE_UCP) != 0)
4835 {
4836 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4837 if (posix_substitutes[pc] != NULL)
4838 {
4839 nestptr = tempptr + 1;
4840 ptr = posix_substitutes[pc] - 1;
4841 continue;
4842 }
4843 }
4844 #endif
4845 /* In the non-UCP case, we build the bit map for the POSIX class in a
4846 chunk of local store because we may be adding and subtracting from it,
4847 and we don't want to subtract bits that may be in the main map already.
4848 At the end we or the result into the bit map that is being built. */
4849
4850 posix_class *= 3;
4851
4852 /* Copy in the first table (always present) */
4853
4854 memcpy(pbits, cbits + posix_class_maps[posix_class],
4855 32 * sizeof(pcre_uint8));
4856
4857 /* If there is a second table, add or remove it as required. */
4858
4859 taboffset = posix_class_maps[posix_class + 1];
4860 tabopt = posix_class_maps[posix_class + 2];
4861
4862 if (taboffset >= 0)
4863 {
4864 if (tabopt >= 0)
4865 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
4866 else
4867 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
4868 }
4869
4870 /* Now see if we need to remove any special characters. An option
4871 value of 1 removes vertical space and 2 removes underscore. */
4872
4873 if (tabopt < 0) tabopt = -tabopt;
4874 if (tabopt == 1) pbits[1] &= ~0x3c;
4875 else if (tabopt == 2) pbits[11] &= 0x7f;
4876
4877 /* Add the POSIX table or its complement into the main table that is
4878 being built and we are done. */
4879
4880 if (local_negate)
4881 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
4882 else
4883 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
4884
4885 ptr = tempptr + 1;
4886 /* Every class contains at least one < 256 character. */
4887 class_has_8bitchar = 1;
4888 /* Every class contains at least two characters. */
4889 class_one_char = 2;
4890 continue; /* End of POSIX syntax handling */
4891 }
4892
4893 /* Backslash may introduce a single character, or it may introduce one
4894 of the specials, which just set a flag. The sequence \b is a special
4895 case. Inside a class (and only there) it is treated as backspace. We
4896 assume that other escapes have more than one character in them, so
4897 speculatively set both class_has_8bitchar and class_one_char bigger
4898 than one. Unrecognized escapes fall through and are either treated
4899 as literal characters (by default), or are faulted if
4900 PCRE_EXTRA is set. */
4901
4902 if (c == CHAR_BACKSLASH)
4903 {
4904 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4905 TRUE);
4906 if (*errorcodeptr != 0) goto FAILED;
4907 if (escape == 0) c = ec;
4908 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
4909 else if (escape == ESC_N) /* \N is not supported in a class */
4910 {
4911 *errorcodeptr = ERR71;
4912 goto FAILED;
4913 }
4914 else if (escape == ESC_Q) /* Handle start of quoted string */
4915 {
4916 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4917 {
4918 ptr += 2; /* avoid empty string */
4919 }
4920 else inescq = TRUE;
4921 continue;
4922 }
4923 else if (escape == ESC_E) continue; /* Ignore orphan \E */
4924
4925 else
4926 {
4927 register const pcre_uint8 *cbits = cd->cbits;
4928 /* Every class contains at least two < 256 characters. */
4929 class_has_8bitchar++;
4930 /* Every class contains at least two characters. */
4931 class_one_char += 2;
4932
4933 switch (escape)
4934 {
4935 #ifdef SUPPORT_UCP
4936 case ESC_du: /* These are the values given for \d etc */
4937 case ESC_DU: /* when PCRE_UCP is set. We replace the */
4938 case ESC_wu: /* escape sequence with an appropriate \p */
4939 case ESC_WU: /* or \P to test Unicode properties instead */
4940 case ESC_su: /* of the default ASCII testing. */
4941 case ESC_SU:
4942 nestptr = ptr;
4943 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
4944 class_has_8bitchar--; /* Undo! */
4945 continue;
4946 #endif
4947 case ESC_d:
4948 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
4949 continue;
4950
4951 case ESC_D:
4952 should_flip_negation = TRUE;
4953 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
4954 continue;
4955
4956 case ESC_w:
4957 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
4958 continue;
4959
4960 case ESC_W:
4961 should_flip_negation = TRUE;
4962 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4963 continue;
4964
4965 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
4966 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
4967 previously set by something earlier in the character class.
4968 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
4969 we could just adjust the appropriate bit. From PCRE 8.34 we no
4970 longer treat \s and \S specially. */
4971
4972 case ESC_s:
4973 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
4974 continue;
4975
4976 case ESC_S:
4977 should_flip_negation = TRUE;
4978 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
4979 continue;
4980
4981 /* The rest apply in both UCP and non-UCP cases. */
4982
4983 case ESC_h:
4984 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4985 PRIV(hspace_list), NOTACHAR);
4986 continue;
4987
4988 case ESC_H:
4989 (void)add_not_list_to_class(classbits, &class_uchardata, options,
4990 cd, PRIV(hspace_list));
4991 continue;
4992
4993 case ESC_v:
4994 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4995 PRIV(vspace_list), NOTACHAR);
4996 continue;
4997
4998 case ESC_V:
4999 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5000 cd, PRIV(vspace_list));
5001 continue;
5002
5003 #ifdef SUPPORT_UCP
5004 case ESC_p:
5005 case ESC_P:
5006 {
5007 BOOL negated;
5008 unsigned int ptype = 0, pdata = 0;
5009 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5010 goto FAILED;
5011 *class_uchardata++ = ((escape == ESC_p) != negated)?
5012 XCL_PROP : XCL_NOTPROP;
5013 *class_uchardata++ = ptype;
5014 *class_uchardata++ = pdata;
5015 class_has_8bitchar--; /* Undo! */
5016 continue;
5017 }
5018 #endif
5019 /* Unrecognized escapes are faulted if PCRE is running in its
5020 strict mode. By default, for compatibility with Perl, they are
5021 treated as literals. */
5022
5023 default:
5024 if ((options & PCRE_EXTRA) != 0)
5025 {
5026 *errorcodeptr = ERR7;
5027 goto FAILED;
5028 }
5029 class_has_8bitchar--; /* Undo the speculative increase. */
5030 class_one_char -= 2; /* Undo the speculative increase. */
5031 c = *ptr; /* Get the final character and fall through */
5032 break;
5033 }
5034 }
5035
5036 /* Fall through if the escape just defined a single character (c >= 0).
5037 This may be greater than 256. */
5038
5039 escape = 0;
5040
5041 } /* End of backslash handling */
5042
5043 /* A character may be followed by '-' to form a range. However, Perl does
5044 not permit ']' to be the end of the range. A '-' character at the end is
5045 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5046 code for handling \Q and \E is messy. */
5047
5048 CHECK_RANGE:
5049 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5050 {
5051 inescq = FALSE;
5052 ptr += 2;
5053 }
5054 oldptr = ptr;
5055
5056 /* Remember if \r or \n were explicitly used */
5057
5058 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5059
5060 /* Check for range */
5061
5062 if (!inescq && ptr[1] == CHAR_MINUS)
5063 {
5064 pcre_uint32 d;
5065 ptr += 2;
5066 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5067
5068 /* If we hit \Q (not followed by \E) at this point, go into escaped
5069 mode. */
5070
5071 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5072 {
5073 ptr += 2;
5074 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5075 { ptr += 2; continue; }
5076 inescq = TRUE;
5077 break;
5078 }
5079
5080 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5081 back the pointer and jump to handle the character that preceded it. */
5082
5083 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5084 {
5085 ptr = oldptr;
5086 goto CLASS_SINGLE_CHARACTER;
5087 }
5088
5089 /* Otherwise, we have a potential range; pick up the next character */
5090
5091 #ifdef SUPPORT_UTF
5092 if (utf)
5093 { /* Braces are required because the */
5094 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5095 }
5096 else
5097 #endif
5098 d = *ptr; /* Not UTF-8 mode */
5099
5100 /* The second part of a range can be a single-character escape, but
5101 not any of the other escapes. Perl 5.6 treats a hyphen as a literal
5102 in such circumstances. */
5103
5104 if (!inescq && d == CHAR_BACKSLASH)
5105 {
5106 int descape;
5107 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5108 if (*errorcodeptr != 0) goto FAILED;
5109
5110 /* \b is backspace; any other special means the '-' was literal. */
5111
5112 if (descape != 0)
5113 {
5114 if (descape == ESC_b) d = CHAR_BS; else
5115 {
5116 ptr = oldptr;
5117 goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5118 }
5119 }
5120 }
5121
5122 /* Check that the two values are in the correct order. Optimize
5123 one-character ranges. */
5124
5125 if (d < c)
5126 {
5127 *errorcodeptr = ERR8;
5128 goto FAILED;
5129 }
5130 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5131
5132 /* We have found a character range, so single character optimizations
5133 cannot be done anymore. Any value greater than 1 indicates that there
5134 is more than one character. */
5135
5136 class_one_char = 2;
5137
5138 /* Remember an explicit \r or \n, and add the range to the class. */
5139
5140 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5141
5142 class_has_8bitchar +=
5143 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5144
5145 continue; /* Go get the next char in the class */
5146 }
5147
5148 /* Handle a single character - we can get here for a normal non-escape
5149 char, or after \ that introduces a single character or for an apparent
5150 range that isn't. Only the value 1 matters for class_one_char, so don't
5151 increase it if it is already 2 or more ... just in case there's a class
5152 with a zillion characters in it. */
5153
5154 CLASS_SINGLE_CHARACTER:
5155 if (class_one_char < 2) class_one_char++;
5156
5157 /* If class_one_char is 1, we have the first single character in the
5158 class, and there have been no prior ranges, or XCLASS items generated by
5159 escapes. If this is the final character in the class, we can optimize by
5160 turning the item into a 1-character OP_CHAR[I] if it's positive, or
5161 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5162 to be set. Otherwise, there can be no first char if this item is first,
5163 whatever repeat count may follow. In the case of reqchar, save the
5164 previous value for reinstating. */
5165
5166 if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5167 {
5168 ptr++;
5169 zeroreqchar = reqchar;
5170 zeroreqcharflags = reqcharflags;
5171
5172 if (negate_class)
5173 {
5174 #ifdef SUPPORT_UCP
5175 int d;
5176 #endif
5177 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5178 zerofirstchar = firstchar;
5179 zerofirstcharflags = firstcharflags;
5180
5181 /* For caseless UTF-8 mode when UCP support is available, check
5182 whether this character has more than one other case. If so, generate
5183 a special OP_NOTPROP item instead of OP_NOTI. */
5184
5185 #ifdef SUPPORT_UCP
5186 if (utf && (options & PCRE_CASELESS) != 0 &&
5187 (d = UCD_CASESET(c)) != 0)
5188 {
5189 *code++ = OP_NOTPROP;
5190 *code++ = PT_CLIST;
5191 *code++ = d;
5192 }
5193 else
5194 #endif
5195 /* Char has only one other case, or UCP not available */
5196
5197 {
5198 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5199 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5200 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5201 code += PRIV(ord2utf)(c, code);
5202 else
5203 #endif
5204 *code++ = c;
5205 }
5206
5207 /* We are finished with this character class */
5208
5209 goto END_CLASS;
5210 }
5211
5212 /* For a single, positive character, get the value into mcbuffer, and
5213 then we can handle this with the normal one-character code. */
5214
5215 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5216 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5217 mclength = PRIV(ord2utf)(c, mcbuffer);
5218 else
5219 #endif
5220 {
5221 mcbuffer[0] = c;
5222 mclength = 1;
5223 }
5224 goto ONE_CHAR;
5225 } /* End of 1-char optimization */
5226
5227 /* There is more than one character in the class, or an XCLASS item
5228 has been generated. Add this character to the class. */
5229
5230 class_has_8bitchar +=
5231 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5232 }
5233
5234 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5235 If we are at the end of an internal nested string, revert to the outer
5236 string. */
5237
5238 while (((c = *(++ptr)) != CHAR_NULL ||
5239 (nestptr != NULL &&
5240 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5241 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5242
5243 /* Check for missing terminating ']' */
5244
5245 if (c == CHAR_NULL)
5246 {
5247 *errorcodeptr = ERR6;
5248 goto FAILED;
5249 }
5250
5251 /* We will need an XCLASS if data has been placed in class_uchardata. In
5252 the second phase this is a sufficient test. However, in the pre-compile
5253 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5254 only if the very last character in the class needs XCLASS will it contain
5255 anything at this point. For this reason, xclass gets set TRUE above when
5256 uchar_classdata is emptied, and that's why this code is the way it is here
5257 instead of just doing a test on class_uchardata below. */
5258
5259 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5260 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5261 #endif
5262
5263 /* If this is the first thing in the branch, there can be no first char
5264 setting, whatever the repeat count. Any reqchar setting must remain
5265 unchanged after any kind of repeat. */
5266
5267 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5268 zerofirstchar = firstchar;
5269 zerofirstcharflags = firstcharflags;
5270 zeroreqchar = reqchar;
5271 zeroreqcharflags = reqcharflags;
5272
5273 /* If there are characters with values > 255, we have to compile an
5274 extended class, with its own opcode, unless there was a negated special
5275 such as \S in the class, and PCRE_UCP is not set, because in that case all
5276 characters > 255 are in the class, so any that were explicitly given as
5277 well can be ignored. If (when there are explicit characters > 255 that must
5278 be listed) there are no characters < 256, we can omit the bitmap in the
5279 actual compiled code. */
5280
5281 #ifdef SUPPORT_UTF
5282 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5283 #elif !defined COMPILE_PCRE8
5284 if (xclass && !should_flip_negation)
5285 #endif
5286 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5287 {
5288 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5289 *code++ = OP_XCLASS;
5290 code += LINK_SIZE;
5291 *code = negate_class? XCL_NOT:0;
5292
5293 /* If the map is required, move up the extra data to make room for it;
5294 otherwise just move the code pointer to the end of the extra data. */
5295
5296 if (class_has_8bitchar > 0)
5297 {
5298 *code++ |= XCL_MAP;
5299 memmove(code + (32 / sizeof(pcre_uchar)), code,
5300 IN_UCHARS(class_uchardata - code));
5301 memcpy(code, classbits, 32);
5302 code = class_uchardata + (32 / sizeof(pcre_uchar));
5303 }
5304 else code = class_uchardata;
5305
5306 /* Now fill in the complete length of the item */
5307
5308 PUT(previous, 1, (int)(code - previous));
5309 break; /* End of class handling */
5310 }
5311 #endif
5312
5313 /* If there are no characters > 255, or they are all to be included or
5314 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5315 whole class was negated and whether there were negative specials such as \S
5316 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5317 negating it if necessary. */
5318
5319 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5320 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5321 {
5322 if (negate_class)
5323 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5324 memcpy(code, classbits, 32);
5325 }
5326 code += 32 / sizeof(pcre_uchar);
5327
5328 END_CLASS:
5329 break;
5330
5331
5332 /* ===================================================================*/
5333 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5334 has been tested above. */
5335
5336 case CHAR_LEFT_CURLY_BRACKET:
5337 if (!is_quantifier) goto NORMAL_CHAR;
5338 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5339 if (*errorcodeptr != 0) goto FAILED;
5340 goto REPEAT;
5341
5342 case CHAR_ASTERISK:
5343 repeat_min = 0;
5344 repeat_max = -1;
5345 goto REPEAT;
5346
5347 case CHAR_PLUS:
5348 repeat_min = 1;
5349 repeat_max = -1;
5350 goto REPEAT;
5351
5352 case CHAR_QUESTION_MARK:
5353 repeat_min = 0;
5354 repeat_max = 1;
5355
5356 REPEAT:
5357 if (previous == NULL)
5358 {
5359 *errorcodeptr = ERR9;
5360 goto FAILED;
5361 }
5362
5363 if (repeat_min == 0)
5364 {
5365 firstchar = zerofirstchar; /* Adjust for zero repeat */
5366 firstcharflags = zerofirstcharflags;
5367 reqchar = zeroreqchar; /* Ditto */
5368 reqcharflags = zeroreqcharflags;
5369 }
5370
5371 /* Remember whether this is a variable length repeat */
5372
5373 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5374
5375 op_type = 0; /* Default single-char op codes */
5376 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5377
5378 /* Save start of previous item, in case we have to move it up in order to
5379 insert something before it. */
5380
5381 tempcode = previous;
5382
5383 /* If the next character is '+', we have a possessive quantifier. This
5384 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5385 If the next character is '?' this is a minimizing repeat, by default,
5386 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5387 repeat type to the non-default. */
5388
5389 if (ptr[1] == CHAR_PLUS)
5390 {
5391 repeat_type = 0; /* Force greedy */
5392 possessive_quantifier = TRUE;
5393 ptr++;
5394 }
5395 else if (ptr[1] == CHAR_QUESTION_MARK)
5396 {
5397 repeat_type = greedy_non_default;
5398 ptr++;
5399 }
5400 else repeat_type = greedy_default;
5401
5402 /* If previous was a recursion call, wrap it in atomic brackets so that
5403 previous becomes the atomic group. All recursions were so wrapped in the
5404 past, but it no longer happens for non-repeated recursions. In fact, the
5405 repeated ones could be re-implemented independently so as not to need this,
5406 but for the moment we rely on the code for repeating groups. */
5407
5408 if (*previous == OP_RECURSE)
5409 {
5410 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5411 *previous = OP_ONCE;
5412 PUT(previous, 1, 2 + 2*LINK_SIZE);
5413 previous[2 + 2*LINK_SIZE] = OP_KET;
5414 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5415 code += 2 + 2 * LINK_SIZE;
5416 length_prevgroup = 3 + 3*LINK_SIZE;
5417
5418 /* When actually compiling, we need to check whether this was a forward
5419 reference, and if so, adjust the offset. */
5420
5421 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5422 {
5423 int offset = GET(cd->hwm, -LINK_SIZE);
5424 if (offset == previous + 1 - cd->start_code)
5425 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5426 }
5427 }
5428
5429 /* Now handle repetition for the different types of item. */
5430
5431 /* If previous was a character or negated character match, abolish the item
5432 and generate a repeat item instead. If a char item has a minimum of more
5433 than one, ensure that it is set in reqchar - it might not be if a sequence
5434 such as x{3} is the first thing in a branch because the x will have gone
5435 into firstchar instead. */
5436
5437 if (*previous == OP_CHAR || *previous == OP_CHARI
5438 || *previous == OP_NOT || *previous == OP_NOTI)
5439 {
5440 switch (*previous)
5441 {
5442 default: /* Make compiler happy. */
5443 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5444 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5445 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5446 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5447 }
5448
5449 /* Deal with UTF characters that take up more than one character. It's
5450 easier to write this out separately than try to macrify it. Use c to
5451 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5452 it's a length rather than a small character. */
5453
5454 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5455 if (utf && NOT_FIRSTCHAR(code[-1]))
5456 {
5457 pcre_uchar *lastchar = code - 1;
5458 BACKCHAR(lastchar);
5459 c = (int)(code - lastchar); /* Length of UTF-8 character */
5460 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5461 c |= UTF_LENGTH; /* Flag c as a length */
5462 }
5463 else
5464 #endif /* SUPPORT_UTF */
5465
5466 /* Handle the case of a single charater - either with no UTF support, or
5467 with UTF disabled, or for a single character UTF character. */
5468 {
5469 c = code[-1];
5470 if (*previous <= OP_CHARI && repeat_min > 1)
5471 {
5472 reqchar = c;
5473 reqcharflags = req_caseopt | cd->req_varyopt;
5474 }
5475 }
5476
5477 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5478 }
5479
5480 /* If previous was a character type match (\d or similar), abolish it and
5481 create a suitable repeat item. The code is shared with single-character
5482 repeats by setting op_type to add a suitable offset into repeat_type. Note
5483 the the Unicode property types will be present only when SUPPORT_UCP is
5484 defined, but we don't wrap the little bits of code here because it just
5485 makes it horribly messy. */
5486
5487 else if (*previous < OP_EODN)
5488 {
5489 pcre_uchar *oldcode;
5490 int prop_type, prop_value;
5491 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5492 c = *previous;
5493
5494 OUTPUT_SINGLE_REPEAT:
5495 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5496 {
5497 prop_type = previous[1];
5498 prop_value = previous[2];
5499 }
5500 else prop_type = prop_value = -1;
5501
5502 oldcode = code;
5503 code = previous; /* Usually overwrite previous item */
5504
5505 /* If the maximum is zero then the minimum must also be zero; Perl allows
5506 this case, so we do too - by simply omitting the item altogether. */
5507
5508 if (repeat_max == 0) goto END_REPEAT;
5509
5510 /* Combine the op_type with the repeat_type */
5511
5512 repeat_type += op_type;
5513
5514 /* A minimum of zero is handled either as the special case * or ?, or as
5515 an UPTO, with the maximum given. */
5516
5517 if (repeat_min == 0)
5518 {
5519 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5520 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5521 else
5522 {
5523 *code++ = OP_UPTO + repeat_type;
5524 PUT2INC(code, 0, repeat_max);
5525 }
5526 }
5527
5528 /* A repeat minimum of 1 is optimized into some special cases. If the
5529 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5530 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5531 one less than the maximum. */
5532
5533 else if (repeat_min == 1)
5534 {
5535 if (repeat_max == -1)
5536 *code++ = OP_PLUS + repeat_type;
5537 else
5538 {
5539 code = oldcode; /* leave previous item in place */
5540 if (repeat_max == 1) goto END_REPEAT;
5541 *code++ = OP_UPTO + repeat_type;
5542 PUT2INC(code, 0, repeat_max - 1);
5543 }
5544 }
5545
5546 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5547 handled as an EXACT followed by an UPTO. */
5548
5549 else
5550 {
5551 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5552 PUT2INC(code, 0, repeat_min);
5553
5554 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5555 we have to insert the character for the previous code. For a repeated
5556 Unicode property match, there are two extra bytes that define the
5557 required property. In UTF-8 mode, long characters have their length in
5558 c, with the UTF_LENGTH bit as a flag. */
5559
5560 if (repeat_max < 0)
5561 {
5562 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5563 if (utf && (c & UTF_LENGTH) != 0)
5564 {
5565 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5566 code += c & 7;
5567 }
5568 else
5569 #endif
5570 {
5571 *code++ = c;
5572 if (prop_type >= 0)
5573 {
5574 *code++ = prop_type;
5575 *code++ = prop_value;
5576 }
5577 }
5578 *code++ = OP_STAR + repeat_type;
5579 }
5580
5581 /* Else insert an UPTO if the max is greater than the min, again
5582 preceded by the character, for the previously inserted code. If the
5583 UPTO is just for 1 instance, we can use QUERY instead. */
5584
5585 else if (repeat_max != repeat_min)
5586 {
5587 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5588 if (utf && (c & UTF_LENGTH) != 0)
5589 {
5590 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5591 code += c & 7;
5592 }
5593 else
5594 #endif
5595 *code++ = c;
5596 if (prop_type >= 0)
5597 {
5598 *code++ = prop_type;
5599 *code++ = prop_value;
5600 }
5601 repeat_max -= repeat_min;
5602
5603 if (repeat_max == 1)
5604 {
5605 *code++ = OP_QUERY + repeat_type;
5606 }
5607 else
5608 {
5609 *code++ = OP_UPTO + repeat_type;
5610 PUT2INC(code, 0, repeat_max);
5611 }
5612 }
5613 }
5614
5615 /* The character or character type itself comes last in all cases. */
5616
5617 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5618 if (utf && (c & UTF_LENGTH) != 0)
5619 {
5620 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5621 code += c & 7;
5622 }
5623 else
5624 #endif
5625 *code++ = c;
5626
5627 /* For a repeated Unicode property match, there are two extra bytes that
5628 define the required property. */
5629
5630 #ifdef SUPPORT_UCP
5631 if (prop_type >= 0)
5632 {
5633 *code++ = prop_type;
5634 *code++ = prop_value;
5635 }
5636 #endif
5637 }
5638
5639 /* If previous was a character class or a back reference, we put the repeat
5640 stuff after it, but just skip the item if the repeat was {0,0}. */
5641
5642 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5643 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5644 *previous == OP_XCLASS ||
5645 #endif
5646 *previous == OP_REF || *previous == OP_REFI ||
5647 *previous == OP_DNREF || *previous == OP_DNREFI)
5648 {
5649 if (repeat_max == 0)
5650 {
5651 code = previous;
5652 goto END_REPEAT;
5653 }
5654
5655 if (repeat_min == 0 && repeat_max == -1)
5656 *code++ = OP_CRSTAR + repeat_type;
5657 else if (repeat_min == 1 && repeat_max == -1)
5658 *code++ = OP_CRPLUS + repeat_type;
5659 else if (repeat_min == 0 && repeat_max == 1)
5660 *code++ = OP_CRQUERY + repeat_type;
5661 else
5662 {
5663 *code++ = OP_CRRANGE + repeat_type;
5664 PUT2INC(code, 0, repeat_min);
5665 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5666 PUT2INC(code, 0, repeat_max);
5667 }
5668 }
5669
5670 /* If previous was a bracket group, we may have to replicate it in certain
5671 cases. Note that at this point we can encounter only the "basic" bracket
5672 opcodes such as BRA and CBRA, as this is the place where they get converted
5673 into the more special varieties such as BRAPOS and SBRA. A test for >=
5674 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5675 ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
5676 repetition of assertions, but now it does, for Perl compatibility. */
5677
5678 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5679 {
5680 register int i;
5681 int len = (int)(code - previous);
5682 pcre_uchar *bralink = NULL;
5683 pcre_uchar *brazeroptr = NULL;
5684
5685 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5686 we just ignore the repeat. */
5687
5688 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5689 goto END_REPEAT;
5690
5691 /* There is no sense in actually repeating assertions. The only potential
5692 use of repetition is in cases when the assertion is optional. Therefore,
5693 if the minimum is greater than zero, just ignore the repeat. If the
5694 maximum is not not zero or one, set it to 1. */
5695
5696 if (*previous < OP_ONCE) /* Assertion */
5697 {
5698 if (repeat_min > 0) goto END_REPEAT;
5699 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5700 }
5701
5702 /* The case of a zero minimum is special because of the need to stick
5703 OP_BRAZERO in front of it, and because the group appears once in the
5704 data, whereas in other cases it appears the minimum number of times. For
5705 this reason, it is simplest to treat this case separately, as otherwise
5706 the code gets far too messy. There are several special subcases when the
5707 minimum is zero. */
5708
5709 if (repeat_min == 0)
5710 {
5711 /* If the maximum is also zero, we used to just omit the group from the
5712 output altogether, like this:
5713
5714 ** if (repeat_max == 0)
5715 ** {
5716 ** code = previous;
5717 ** goto END_REPEAT;
5718 ** }
5719
5720 However, that fails when a group or a subgroup within it is referenced
5721 as a subroutine from elsewhere in the pattern, so now we stick in
5722 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5723 don't have a list of which groups are referenced, we cannot do this
5724 selectively.
5725
5726 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5727 and do no more at this point. However, we do need to adjust any
5728 OP_RECURSE calls inside the group that refer to the group itself or any
5729 internal or forward referenced group, because the offset is from the
5730 start of the whole regex. Temporarily terminate the pattern while doing
5731 this. */
5732
5733 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5734 {
5735 *code = OP_END;
5736 adjust_recurse(previous, 1, utf, cd, save_hwm);
5737 memmove(previous + 1, previous, IN_UCHARS(len));
5738 code++;
5739 if (repeat_max == 0)
5740 {
5741 *previous++ = OP_SKIPZERO;
5742 goto END_REPEAT;
5743 }
5744 brazeroptr = previous; /* Save for possessive optimizing */
5745 *previous++ = OP_BRAZERO + repeat_type;
5746 }
5747
5748 /* If the maximum is greater than 1 and limited, we have to replicate
5749 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
5750 The first one has to be handled carefully because it's the original
5751 copy, which has to be moved up. The remainder can be handled by code
5752 that is common with the non-zero minimum case below. We have to
5753 adjust the value or repeat_max, since one less copy is required. Once
5754 again, we may have to adjust any OP_RECURSE calls inside the group. */
5755
5756 else
5757 {
5758 int offset;
5759 *code = OP_END;
5760 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5761 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5762 code += 2 + LINK_SIZE;
5763 *previous++ = OP_BRAZERO + repeat_type;
5764 *previous++ = OP_BRA;
5765
5766 /* We chain together the bracket offset fields that have to be
5767 filled in later when the ends of the brackets are reached. */
5768
5769 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
5770 bralink = previous;
5771 PUTINC(previous, 0, offset);
5772 }
5773
5774 repeat_max--;
5775 }
5776
5777 /* If the minimum is greater than zero, replicate the group as many
5778 times as necessary, and adjust the maximum to the number of subsequent
5779 copies that we need. If we set a first char from the group, and didn't
5780 set a required char, copy the latter from the former. If there are any
5781 forward reference subroutine calls in the group, there will be entries on
5782 the workspace list; replicate these with an appropriate increment. */
5783
5784 else
5785 {
5786 if (repeat_min > 1)
5787 {
5788 /* In the pre-compile phase, we don't actually do the replication. We
5789 just adjust the length as if we had. Do some paranoid checks for
5790 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
5791 integer type when available, otherwise double. */
5792
5793 if (lengthptr != NULL)
5794 {
5795 int delta = (repeat_min - 1)*length_prevgroup;
5796 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
5797 (INT64_OR_DOUBLE)length_prevgroup >
5798 (INT64_OR_DOUBLE)INT_MAX ||
5799 OFLOW_MAX - *lengthptr < delta)
5800 {
5801 *errorcodeptr = ERR20;
5802 goto FAILED;
5803 }
5804 *lengthptr += delta;
5805 }
5806
5807 /* This is compiling for real. If there is a set first byte for
5808 the group, and we have not yet set a "required byte", set it. Make
5809 sure there is enough workspace for copying forward references before
5810 doing the copy. */
5811
5812 else
5813 {
5814 if (groupsetfirstchar && reqcharflags < 0)
5815 {
5816 reqchar = firstchar;
5817 reqcharflags = firstcharflags;
5818 }
5819
5820 for (i = 1; i < repeat_min; i++)
5821 {
5822 pcre_uchar *hc;
5823 pcre_uchar *this_hwm = cd->hwm;
5824 memcpy(code, previous, IN_UCHARS(len));
5825
5826 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5827 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5828 {
5829 int save_offset = save_hwm - cd->start_workspace;
5830 int this_offset = this_hwm - cd->start_workspace;
5831 *errorcodeptr = expand_workspace(cd);
5832 if (*errorcodeptr != 0) goto FAILED;
5833 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5834 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5835 }
5836
5837 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5838 {
5839 PUT(cd->hwm, 0, GET(hc, 0) + len);
5840 cd->hwm += LINK_SIZE;
5841 }
5842 save_hwm = this_hwm;
5843 code += len;
5844 }
5845 }
5846 }
5847
5848 if (repeat_max > 0) repeat_max -= repeat_min;
5849 }
5850
5851 /* This code is common to both the zero and non-zero minimum cases. If
5852 the maximum is limited, it replicates the group in a nested fashion,
5853 remembering the bracket starts on a stack. In the case of a zero minimum,
5854 the first one was set up above. In all cases the repeat_max now specifies
5855 the number of additional copies needed. Again, we must remember to
5856 replicate entries on the forward reference list. */
5857
5858 if (repeat_max >= 0)
5859 {
5860 /* In the pre-compile phase, we don't actually do the replication. We
5861 just adjust the length as if we had. For each repetition we must add 1
5862 to the length for BRAZERO and for all but the last repetition we must
5863 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
5864 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
5865 a 64-bit integer type when available, otherwise double. */
5866
5867 if (lengthptr != NULL && repeat_max > 0)
5868 {
5869 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
5870 2 - 2*LINK_SIZE; /* Last one doesn't nest */
5871 if ((INT64_OR_DOUBLE)repeat_max *
5872 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
5873 > (INT64_OR_DOUBLE)INT_MAX ||
5874 OFLOW_MAX - *lengthptr < delta)
5875 {
5876 *errorcodeptr = ERR20;
5877 goto FAILED;
5878 }
5879 *lengthptr += delta;
5880 }
5881
5882 /* This is compiling for real */
5883
5884 else for (i = repeat_max - 1; i >= 0; i--)
5885 {
5886 pcre_uchar *hc;
5887 pcre_uchar *this_hwm = cd->hwm;
5888
5889 *code++ = OP_BRAZERO + repeat_type;
5890
5891 /* All but the final copy start a new nesting, maintaining the
5892 chain of brackets outstanding. */
5893
5894 if (i != 0)
5895 {
5896 int offset;
5897 *code++ = OP_BRA;
5898 offset = (bralink == NULL)? 0 : (int)(code - bralink);
5899 bralink = code;
5900 PUTINC(code, 0, offset);
5901 }
5902
5903 memcpy(code, previous, IN_UCHARS(len));
5904
5905 /* Ensure there is enough workspace for forward references before
5906 copying them. */
5907
5908 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5909 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5910 {
5911 int save_offset = save_hwm - cd->start_workspace;
5912 int this_offset = this_hwm - cd->start_workspace;
5913 *errorcodeptr = expand_workspace(cd);
5914 if (*errorcodeptr != 0) goto FAILED;
5915 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5916 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5917 }
5918
5919 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5920 {
5921 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
5922 cd->hwm += LINK_SIZE;
5923 }
5924 save_hwm = this_hwm;
5925 code += len;
5926 }
5927
5928 /* Now chain through the pending brackets, and fill in their length
5929 fields (which are holding the chain links pro tem). */
5930
5931 while (bralink != NULL)
5932 {
5933 int oldlinkoffset;
5934 int offset = (int)(code - bralink + 1);
5935 pcre_uchar *bra = code - offset;
5936 oldlinkoffset = GET(bra, 1);
5937 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
5938 *code++ = OP_KET;
5939 PUTINC(code, 0, offset);
5940 PUT(bra, 1, offset);
5941 }
5942 }
5943
5944 /* If the maximum is unlimited, set a repeater in the final copy. For
5945 ONCE brackets, that's all we need to do. However, possessively repeated
5946 ONCE brackets can be converted into non-capturing brackets, as the
5947 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5948 deal with possessive ONCEs specially.
5949
5950 Otherwise, when we are doing the actual compile phase, check to see
5951 whether this group is one that could match an empty string. If so,
5952 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5953 that runtime checking can be done. [This check is also applied to ONCE
5954 groups at runtime, but in a different way.]
5955
5956 Then, if the quantifier was possessive and the bracket is not a
5957 conditional, we convert the BRA code to the POS form, and the KET code to
5958 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5959 subpattern at both the start and at the end.) The use of special opcodes
5960 makes it possible to reduce greatly the stack usage in pcre_exec(). If
5961 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5962
5963 Then, if the minimum number of matches is 1 or 0, cancel the possessive
5964 flag so that the default action below, of wrapping everything inside
5965 atomic brackets, does not happen. When the minimum is greater than 1,
5966 there will be earlier copies of the group, and so we still have to wrap
5967 the whole thing. */
5968
5969 else
5970 {
5971 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
5972 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
5973
5974 /* Convert possessive ONCE brackets to non-capturing */
5975
5976 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5977 possessive_quantifier) *bracode = OP_BRA;
5978
5979 /* For non-possessive ONCE brackets, all we need to do is to
5980 set the KET. */
5981
5982 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5983 *ketcode = OP_KETRMAX + repeat_type;
5984
5985 /* Handle non-ONCE brackets and possessive ONCEs (which have been
5986 converted to non-capturing above). */
5987
5988 else
5989 {
5990 /* In the compile phase, check for empty string matching. */
5991
5992 if (lengthptr == NULL)
5993 {
5994 pcre_uchar *scode = bracode;
5995 do
5996 {
5997 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
5998 {
5999 *bracode += OP_SBRA - OP_BRA;
6000 break;
6001 }
6002 scode += GET(scode, 1);
6003 }
6004 while (*scode == OP_ALT);
6005 }
6006
6007 /* Handle possessive quantifiers. */
6008
6009 if (possessive_quantifier)
6010 {
6011 /* For COND brackets, we wrap the whole thing in a possessively
6012 repeated non-capturing bracket, because we have not invented POS
6013 versions of the COND opcodes. Because we are moving code along, we
6014 must ensure that any pending recursive references are updated. */
6015
6016 if (*bracode == OP_COND || *bracode == OP_SCOND)
6017 {
6018 int nlen = (int)(code - bracode);
6019 *code = OP_END;
6020 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
6021 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6022 code += 1 + LINK_SIZE;
6023 nlen += 1 + LINK_SIZE;
6024 *bracode = OP_BRAPOS;
6025 *code++ = OP_KETRPOS;
6026 PUTINC(code, 0, nlen);
6027 PUT(bracode, 1, nlen);
6028 }
6029
6030 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6031
6032 else
6033 {
6034 *bracode += 1; /* Switch to xxxPOS opcodes */
6035 *ketcode = OP_KETRPOS;
6036 }
6037
6038 /* If the minimum is zero, mark it as possessive, then unset the
6039 possessive flag when the minimum is 0 or 1. */
6040
6041 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6042 if (repeat_min < 2) possessive_quantifier = FALSE;
6043 }
6044
6045 /* Non-possessive quantifier */
6046
6047 else *ketcode = OP_KETRMAX + repeat_type;
6048 }
6049 }
6050 }
6051
6052 /* If previous is OP_FAIL, it was generated by an empty class [] in
6053 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6054 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6055 error above. We can just ignore the repeat in JS case. */
6056
6057 else if (*previous == OP_FAIL) goto END_REPEAT;
6058
6059 /* Else there's some kind of shambles */
6060
6061 else
6062 {
6063 *errorcodeptr = ERR11;
6064 goto FAILED;
6065 }
6066
6067 /* If the character following a repeat is '+', possessive_quantifier is
6068 TRUE. For some opcodes, there are special alternative opcodes for this
6069 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6070 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6071 Sun's Java package, but the special opcodes can optimize it.
6072
6073 Some (but not all) possessively repeated subpatterns have already been
6074 completely handled in the code just above. For them, possessive_quantifier
6075 is always FALSE at this stage. Note that the repeated item starts at
6076 tempcode, not at previous, which might be the first part of a string whose
6077 (former) last char we repeated. */
6078
6079 if (possessive_quantifier)
6080 {
6081 int len;
6082
6083 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6084 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6085 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6086 remains is greater than zero, there's a further opcode that can be
6087 handled. If not, do nothing, leaving the EXACT alone. */
6088
6089 switch(*tempcode)
6090 {
6091 case OP_TYPEEXACT:
6092 tempcode += PRIV(OP_lengths)[*tempcode] +
6093 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6094 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6095 break;
6096
6097 /* CHAR opcodes are used for exacts whose count is 1. */
6098
6099 case OP_CHAR:
6100 case OP_CHARI:
6101 case OP_NOT:
6102 case OP_NOTI:
6103 case OP_EXACT:
6104 case OP_EXACTI:
6105 case OP_NOTEXACT:
6106 case OP_NOTEXACTI:
6107 tempcode += PRIV(OP_lengths)[*tempcode];
6108 #ifdef SUPPORT_UTF
6109 if (utf && HAS_EXTRALEN(tempcode[-1]))
6110 tempcode += GET_EXTRALEN(tempcode[-1]);
6111 #endif
6112 break;
6113
6114 /* For the class opcodes, the repeat operator appears at the end;
6115 adjust tempcode to point to it. */
6116
6117 case OP_CLASS:
6118 case OP_NCLASS:
6119 tempcode += 1 + 32/sizeof(pcre_uchar);
6120 break;
6121
6122 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6123 case OP_XCLASS:
6124 tempcode += GET(tempcode, 1);
6125 break;
6126 #endif
6127 }
6128
6129 /* If tempcode is equal to code (which points to the end of the repeated
6130 item), it means we have skipped an EXACT item but there is no following
6131 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6132 all other cases, tempcode will be pointing to the repeat opcode, and will
6133 be less than code, so the value of len will be greater than 0. */
6134
6135 len = (int)(code - tempcode);
6136 if (len > 0)
6137 {
6138 unsigned int repcode = *tempcode;
6139
6140 /* There is a table for possessifying opcodes, all of which are less
6141 than OP_CALLOUT. A zero entry means there is no possessified version.
6142 */
6143
6144 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6145 *tempcode = opcode_possessify[repcode];
6146
6147 /* For opcode without a special possessified version, wrap the item in
6148 ONCE brackets. Because we are moving code along, we must ensure that any
6149 pending recursive references are updated. */
6150
6151 else
6152 {
6153 *code = OP_END;
6154 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6155 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6156 code += 1 + LINK_SIZE;
6157 len += 1 + LINK_SIZE;
6158 tempcode[0] = OP_ONCE;
6159 *code++ = OP_KET;
6160 PUTINC(code, 0, len);
6161 PUT(tempcode, 1, len);
6162 }
6163 }
6164
6165 #ifdef NEVER
6166 if (len > 0) switch (*tempcode)
6167 {
6168 case OP_STAR: *tempcode = OP_POSSTAR; break;
6169 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6170 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6171 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6172
6173 case OP_STARI: *tempcode = OP_POSSTARI; break;
6174 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6175 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6176 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6177
6178 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6179 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6180 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6181 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6182
6183 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6184 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6185 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6186 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6187
6188 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6189 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6190 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6191 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6192
6193 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6194 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6195 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6196 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6197
6198 /* Because we are moving code along, we must ensure that any
6199 pending recursive references are updated. */
6200
6201 default:
6202 *code = OP_END;
6203 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6204 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6205 code += 1 + LINK_SIZE;
6206 len += 1 + LINK_SIZE;
6207 tempcode[0] = OP_ONCE;
6208 *code++ = OP_KET;
6209 PUTINC(code, 0, len);
6210 PUT(tempcode, 1, len);
6211 break;
6212 }
6213 #endif
6214 }
6215
6216 /* In all case we no longer have a previous item. We also set the
6217 "follows varying string" flag for subsequently encountered reqchars if
6218 it isn't already set and we have just passed a varying length item. */
6219
6220 END_REPEAT:
6221 previous = NULL;
6222 cd->req_varyopt |= reqvary;
6223 break;
6224
6225
6226 /* ===================================================================*/
6227 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6228 lookbehind or option setting or condition or all the other extended
6229 parenthesis forms. */
6230
6231 case CHAR_LEFT_PARENTHESIS:
6232 newoptions = options;
6233 skipbytes = 0;
6234 bravalue = OP_CBRA;
6235 save_hwm = cd->hwm;
6236 reset_bracount = FALSE;
6237
6238 /* First deal with various "verbs" that can be introduced by '*'. */
6239
6240 ptr++;
6241 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6242 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6243 {
6244 int i, namelen;
6245 int arglen = 0;
6246 const char *vn = verbnames;
6247 const pcre_uchar *name = ptr + 1;
6248 const pcre_uchar *arg = NULL;
6249 previous = NULL;
6250 ptr++;
6251 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6252 namelen = (int)(ptr - name);
6253
6254 /* It appears that Perl allows any characters whatsoever, other than
6255 a closing parenthesis, to appear in arguments, so we no longer insist on
6256 letters, digits, and underscores. */
6257
6258 if (*ptr == CHAR_COLON)
6259 {
6260 arg = ++ptr;
6261 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6262 arglen = (int)(ptr - arg);
6263 if ((unsigned int)arglen > MAX_MARK)
6264 {
6265 *errorcodeptr = ERR75;
6266 goto FAILED;
6267 }
6268 }
6269
6270 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6271 {
6272 *errorcodeptr = ERR60;
6273 goto FAILED;
6274 }
6275
6276 /* Scan the table of verb names */
6277
6278 for (i = 0; i < verbcount; i++)
6279 {
6280 if (namelen == verbs[i].len &&
6281 STRNCMP_UC_C8(name, vn, namelen) == 0)
6282 {
6283 int setverb;
6284
6285 /* Check for open captures before ACCEPT and convert it to
6286 ASSERT_ACCEPT if in an assertion. */
6287
6288 if (verbs[i].op == OP_ACCEPT)
6289 {
6290 open_capitem *oc;
6291 if (arglen != 0)
6292 {
6293 *errorcodeptr = ERR59;
6294 goto FAILED;
6295 }
6296 cd->had_accept = TRUE;
6297 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6298 {
6299 *code++ = OP_CLOSE;
6300 PUT2INC(code, 0, oc->number);
6301 }
6302 setverb = *code++ =
6303 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6304
6305 /* Do not set firstchar after *ACCEPT */
6306 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6307 }
6308
6309 /* Handle other cases with/without an argument */
6310
6311 else if (arglen == 0)
6312 {
6313 if (verbs[i].op < 0) /* Argument is mandatory */
6314 {
6315 *errorcodeptr = ERR66;
6316 goto FAILED;
6317 }
6318 setverb = *code++ = verbs[i].op;
6319 }
6320
6321 else
6322 {
6323 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6324 {
6325 *errorcodeptr = ERR59;
6326 goto FAILED;
6327 }
6328 setverb = *code++ = verbs[i].op_arg;
6329 *code++ = arglen;
6330 memcpy(code, arg, IN_UCHARS(arglen));
6331 code += arglen;
6332 *code++ = 0;
6333 }
6334
6335 switch (setverb)
6336 {
6337 case OP_THEN:
6338 case OP_THEN_ARG:
6339 cd->external_flags |= PCRE_HASTHEN;
6340 break;
6341
6342 case OP_PRUNE:
6343 case OP_PRUNE_ARG:
6344 case OP_SKIP:
6345 case OP_SKIP_ARG:
6346 cd->had_pruneorskip = TRUE;
6347 break;
6348 }
6349
6350 break; /* Found verb, exit loop */
6351 }
6352
6353 vn += verbs[i].len + 1;
6354 }
6355
6356 if (i < verbcount) continue; /* Successfully handled a verb */
6357 *errorcodeptr = ERR60; /* Verb not recognized */
6358 goto FAILED;
6359 }
6360
6361 /* Deal with the extended parentheses; all are introduced by '?', and the
6362 appearance of any of them means that this is not a capturing group. */
6363
6364 else if (*ptr == CHAR_QUESTION_MARK)
6365 {
6366 int i, set, unset, namelen;
6367 int *optset;
6368 const pcre_uchar *name;
6369 pcre_uchar *slot;
6370
6371 switch (*(++ptr))
6372 {
6373 case CHAR_NUMBER_SIGN: /* Comment; skip to ket */
6374 ptr++;
6375 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6376 if (*ptr == CHAR_NULL)
6377 {
6378 *errorcodeptr = ERR18;
6379 goto FAILED;
6380 }
6381 continue;
6382
6383
6384 /* ------------------------------------------------------------ */
6385 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6386 reset_bracount = TRUE;
6387 /* Fall through */
6388
6389 /* ------------------------------------------------------------ */
6390 case CHAR_COLON: /* Non-capturing bracket */
6391 bravalue = OP_BRA;
6392 ptr++;
6393 break;
6394
6395
6396 /* ------------------------------------------------------------ */
6397 case CHAR_LEFT_PARENTHESIS:
6398 bravalue = OP_COND; /* Conditional group */
6399 tempptr = ptr;
6400
6401 /* A condition can be an assertion, a number (referring to a numbered
6402 group), a name (referring to a named group), or 'R', referring to
6403 recursion. R<digits> and R&name are also permitted for recursion tests.
6404
6405 There are several syntaxes for testing a named group: (?(name)) is used
6406 by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
6407
6408 There are two unfortunate ambiguities, caused by history. (a) 'R' can
6409 be the recursive thing or the name 'R' (and similarly for 'R' followed
6410 by digits), and (b) a number could be a name that consists of digits.
6411 In both cases, we look for a name first; if not found, we try the other
6412 cases.
6413
6414 For compatibility with auto-callouts, we allow a callout to be
6415 specified before a condition that is an assertion. First, check for the
6416 syntax of a callout; if found, adjust the temporary pointer that is
6417 used to check for an assertion condition. That's all that is needed! */
6418
6419 if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6420 {
6421 for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6422 if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6423 tempptr += i + 1;
6424 }
6425
6426 /* For conditions that are assertions, check the syntax, and then exit
6427 the switch. This will take control down to where bracketed groups,
6428 including assertions, are processed. */
6429
6430 if (tempptr[1] == CHAR_QUESTION_MARK &&
6431 (tempptr[2] == CHAR_EQUALS_SIGN ||
6432 tempptr[2] == CHAR_EXCLAMATION_MARK ||
6433 tempptr[2] == CHAR_LESS_THAN_SIGN))
6434 break;
6435
6436 /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6437 need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6438
6439 code[1+LINK_SIZE] = OP_CREF;
6440 skipbytes = 1+IMM2_SIZE;
6441 refsign = -1;
6442
6443 /* Check for a test for recursion in a named group. */
6444
6445 if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
6446 {
6447 terminator = -1;
6448 ptr += 2;
6449 code[1+LINK_SIZE] = OP_RREF; /* Change the type of test */
6450 }
6451
6452 /* Check for a test for a named group's having been set, using the Perl
6453 syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6454 syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). As names may
6455 consist entirely of digits, there is scope for ambiguity. */
6456
6457 else if (ptr[1] == CHAR_LESS_THAN_SIGN)
6458 {
6459 terminator = CHAR_GREATER_THAN_SIGN;
6460 ptr++;
6461 }
6462 else if (ptr[1] == CHAR_APOSTROPHE)
6463 {
6464 terminator = CHAR_APOSTROPHE;
6465 ptr++;
6466 }
6467 else
6468 {
6469 terminator = CHAR_NULL;
6470 if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
6471 }
6472
6473 /* When a name is one of a number of duplicates, a different opcode is
6474 used and it needs more memory. Unfortunately we cannot tell whether a
6475 name is a duplicate in the first pass, so we have to allow for more
6476 memory except when we know it is a relative numerical reference. */
6477
6478 if (refsign < 0 && lengthptr != NULL) *lengthptr += IMM2_SIZE;
6479
6480 /* We now expect to read a name (possibly all digits); any thing else
6481 is an error. In the case of all digits, also get it as a number. */
6482
6483 if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)
6484 {
6485 ptr += 1; /* To get the right offset */
6486 *errorcodeptr = ERR28;
6487 goto FAILED;
6488 }
6489
6490 recno = 0;
6491 name = ++ptr;
6492 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6493 {
6494 if (recno >= 0)
6495 recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;
6496 ptr++;
6497 }
6498 namelen = (int)(ptr - name);
6499
6500 /* Check the terminator */
6501
6502 if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6503 *ptr++ != CHAR_RIGHT_PARENTHESIS)
6504 {
6505 ptr--; /* Error offset */
6506 *errorcodeptr = ERR26;
6507 goto FAILED;
6508 }
6509
6510 /* Do no further checking in the pre-compile phase. */
6511
6512 if (lengthptr != NULL) break;
6513
6514 /* In the real compile we do the work of looking for the actual
6515 reference. If the string started with "+" or "-" we require the rest to
6516 be digits, in which case recno will be set. */
6517
6518 if (refsign > 0)
6519 {
6520 if (recno <= 0)
6521 {
6522 *errorcodeptr = ERR58;
6523 goto FAILED;
6524 }
6525 recno = (refsign == CHAR_MINUS)?
6526 cd->bracount - recno + 1 : recno +cd->bracount;
6527 if (recno <= 0 || recno > cd->final_bracount)
6528 {
6529 *errorcodeptr = ERR15;
6530 goto FAILED;
6531 }
6532 PUT2(code, 2+LINK_SIZE, recno);
6533 break;
6534 }
6535
6536 /* Otherwise (did not start with "+" or "-"), start by looking for the
6537 name. */
6538
6539 slot = cd->name_table;
6540 for (i = 0; i < cd->names_found; i++)
6541 {
6542 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
6543 slot += cd->name_entry_size;
6544 }
6545
6546 /* Found the named subpattern. If the name is duplicated, add one to
6547 the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6548 appropriate data values. Otherwise, just insert the unique subpattern
6549 number. */
6550
6551 if (i < cd->names_found)
6552 {
6553 int offset = i++;
6554 int count = 1;
6555 recno = GET2(slot, 0); /* Number from first found */
6556 for (; i < cd->names_found; i++)
6557 {
6558 slot += cd->name_entry_size;
6559 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0) break;
6560 count++;
6561 }
6562 if (count > 1)
6563 {
6564 PUT2(code, 2+LINK_SIZE, offset);
6565 PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
6566 skipbytes += IMM2_SIZE;
6567 code[1+LINK_SIZE]++;
6568 }
6569 else /* Not a duplicated name */
6570 {
6571 PUT2(code, 2+LINK_SIZE, recno);
6572 }
6573 }
6574
6575 /* If terminator == CHAR_NULL it means that the name followed directly
6576 after the opening parenthesis [e.g. (?(abc)...] and in this case there
6577 are some further alternatives to try. For the cases where terminator !=
6578 0 [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
6579 now checked all the possibilities, so give an error. */
6580
6581 else if (terminator != CHAR_NULL)
6582 {
6583 *errorcodeptr = ERR15;
6584 goto FAILED;
6585 }
6586
6587 /* Check for (?(R) for recursion. Allow digits after R to specify a
6588 specific group number. */
6589
6590 else if (*name == CHAR_R)
6591 {
6592 recno = 0;
6593 for (i = 1; i < namelen; i++)
6594 {
6595 if (!IS_DIGIT(name[i]))
6596 {
6597 *errorcodeptr = ERR15;
6598 goto FAILED;
6599 }
6600 recno = recno * 10 + name[i] - CHAR_0;
6601 }
6602 if (recno == 0) recno = RREF_ANY;
6603 code[1+LINK_SIZE] = OP_RREF; /* Change test type */
6604 PUT2(code, 2+LINK_SIZE, recno);
6605 }
6606
6607 /* Similarly, check for the (?(DEFINE) "condition", which is always
6608 false. */
6609
6610 else if (namelen == 6 && STRNCMP_UC_C8(name, STRING_DEFINE, 6) == 0)
6611 {
6612 code[1+LINK_SIZE] = OP_DEF;
6613 skipbytes = 1;
6614 }
6615
6616 /* Check for the "name" actually being a subpattern number. We are
6617 in the second pass here, so final_bracount is set. */
6618
6619 else if (recno > 0 && recno <= cd->final_bracount)
6620 {
6621 PUT2(code, 2+LINK_SIZE, recno);
6622 }
6623
6624 /* Either an unidentified subpattern, or a reference to (?(0) */
6625
6626 else
6627 {
6628 *errorcodeptr = (recno == 0)? ERR35: ERR15;
6629 goto FAILED;
6630 }
6631 break;
6632
6633
6634 /* ------------------------------------------------------------ */
6635 case CHAR_EQUALS_SIGN: /* Positive lookahead */
6636 bravalue = OP_ASSERT;
6637 cd->assert_depth += 1;
6638 ptr++;
6639 break;
6640
6641
6642 /* ------------------------------------------------------------ */
6643 case CHAR_EXCLAMATION_MARK: /* Negative lookahead */
6644 ptr++;
6645 if (*ptr == CHAR_RIGHT_PARENTHESIS) /* Optimize (?!) */
6646 {
6647 *code++ = OP_FAIL;
6648 previous = NULL;
6649 continue;
6650 }
6651 bravalue = OP_ASSERT_NOT;
6652 cd->assert_depth += 1;
6653 break;
6654
6655
6656 /* ------------------------------------------------------------ */
6657 case CHAR_LESS_THAN_SIGN: /* Lookbehind or named define */
6658 switch (ptr[1])
6659 {
6660 case CHAR_EQUALS_SIGN: /* Positive lookbehind */
6661 bravalue = OP_ASSERTBACK;
6662 cd->assert_depth += 1;
6663 ptr += 2;
6664