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

Contents of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1384 - (show annotations)
Fri Oct 25 17:37:50 2013 UTC (5 years, 10 months ago) by zherczeg
File MIME type: text/plain
File size: 303897 byte(s)
Guard the call of check_char_prop since it does not exists when UCP is disabled.
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 #ifdef SUPPORT_UCP
3484 case OP_PROP:
3485 case OP_NOTPROP:
3486 if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3487 list_ptr[0] == OP_NOTPROP))
3488 return FALSE;
3489 break;
3490 #endif
3491
3492 case OP_NCLASS:
3493 if (chr > 255) return FALSE;
3494 /* Fall through */
3495
3496 case OP_CLASS:
3497 if (chr > 255) break;
3498 class_bitset = (pcre_uint8 *)
3499 ((list_ptr == list ? code : base_end) - list_ptr[2]);
3500 if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3501 break;
3502
3503 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3504 case OP_XCLASS:
3505 if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3506 list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3507 break;
3508 #endif
3509
3510 default:
3511 return FALSE;
3512 }
3513
3514 chr_ptr++;
3515 }
3516 while(*chr_ptr != NOTACHAR);
3517
3518 /* At least one character must be matched from this opcode. */
3519
3520 if (list[1] == 0) return TRUE;
3521 }
3522
3523 return FALSE;
3524 }
3525
3526
3527
3528 /*************************************************
3529 * Scan compiled regex for auto-possession *
3530 *************************************************/
3531
3532 /* Replaces single character iterations with their possessive alternatives
3533 if appropriate. This function modifies the compiled opcode!
3534
3535 Arguments:
3536 code points to start of the byte code
3537 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3538 cd static compile data
3539
3540 Returns: nothing
3541 */
3542
3543 static void
3544 auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3545 {
3546 register pcre_uchar c;
3547 const pcre_uchar *end;
3548 pcre_uchar *repeat_opcode;
3549 pcre_uint32 list[8];
3550
3551 for (;;)
3552 {
3553 c = *code;
3554
3555 if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3556 {
3557 c -= get_repeat_base(c) - OP_STAR;
3558 end = (c <= OP_MINUPTO) ?
3559 get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3560 list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3561
3562 if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3563 {
3564 switch(c)
3565 {
3566 case OP_STAR:
3567 *code += OP_POSSTAR - OP_STAR;
3568 break;
3569
3570 case OP_MINSTAR:
3571 *code += OP_POSSTAR - OP_MINSTAR;
3572 break;
3573
3574 case OP_PLUS:
3575 *code += OP_POSPLUS - OP_PLUS;
3576 break;
3577
3578 case OP_MINPLUS:
3579 *code += OP_POSPLUS - OP_MINPLUS;
3580 break;
3581
3582 case OP_QUERY:
3583 *code += OP_POSQUERY - OP_QUERY;
3584 break;
3585
3586 case OP_MINQUERY:
3587 *code += OP_POSQUERY - OP_MINQUERY;
3588 break;
3589
3590 case OP_UPTO:
3591 *code += OP_POSUPTO - OP_UPTO;
3592 break;
3593
3594 case OP_MINUPTO:
3595 *code += OP_MINUPTO - OP_UPTO;
3596 break;
3597 }
3598 }
3599 c = *code;
3600 }
3601 else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3602 {
3603 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3604 if (c == OP_XCLASS)
3605 repeat_opcode = code + GET(code, 1);
3606 else
3607 #endif
3608 repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3609
3610 c = *repeat_opcode;
3611 if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3612 {
3613 /* end must not be NULL. */
3614 end = get_chr_property_list(code, utf, cd->fcc, list);
3615
3616 list[1] = (c & 1) == 0;
3617
3618 if (compare_opcodes(end, utf, cd, list, end))
3619 {
3620 switch (c)
3621 {
3622 case OP_CRSTAR:
3623 case OP_CRMINSTAR:
3624 *repeat_opcode = OP_CRPOSSTAR;
3625 break;
3626
3627 case OP_CRPLUS:
3628 case OP_CRMINPLUS:
3629 *repeat_opcode = OP_CRPOSPLUS;
3630 break;
3631
3632 case OP_CRQUERY:
3633 case OP_CRMINQUERY:
3634 *repeat_opcode = OP_CRPOSQUERY;
3635 break;
3636
3637 case OP_CRRANGE:
3638 case OP_CRMINRANGE:
3639 *repeat_opcode = OP_CRPOSRANGE;
3640 break;
3641 }
3642 }
3643 }
3644 c = *code;
3645 }
3646
3647 switch(c)
3648 {
3649 case OP_END:
3650 return;
3651
3652 case OP_TYPESTAR:
3653 case OP_TYPEMINSTAR:
3654 case OP_TYPEPLUS:
3655 case OP_TYPEMINPLUS:
3656 case OP_TYPEQUERY:
3657 case OP_TYPEMINQUERY:
3658 case OP_TYPEPOSSTAR:
3659 case OP_TYPEPOSPLUS:
3660 case OP_TYPEPOSQUERY:
3661 if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3662 break;
3663
3664 case OP_TYPEUPTO:
3665 case OP_TYPEMINUPTO:
3666 case OP_TYPEEXACT:
3667 case OP_TYPEPOSUPTO:
3668 if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3669 code += 2;
3670 break;
3671
3672 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3673 case OP_XCLASS:
3674 code += GET(code, 1);
3675 break;
3676 #endif
3677
3678 case OP_MARK:
3679 case OP_PRUNE_ARG:
3680 case OP_SKIP_ARG:
3681 case OP_THEN_ARG:
3682 code += code[1];
3683 break;
3684 }
3685
3686 /* Add in the fixed length from the table */
3687
3688 code += PRIV(OP_lengths)[c];
3689
3690 /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3691 a multi-byte character. The length in the table is a minimum, so we have to
3692 arrange to skip the extra bytes. */
3693
3694 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3695 if (utf) switch(c)
3696 {
3697 case OP_CHAR:
3698 case OP_CHARI:
3699 case OP_NOT:
3700 case OP_NOTI:
3701 case OP_STAR:
3702 case OP_MINSTAR:
3703 case OP_PLUS:
3704 case OP_MINPLUS:
3705 case OP_QUERY:
3706 case OP_MINQUERY:
3707 case OP_UPTO:
3708 case OP_MINUPTO:
3709 case OP_EXACT:
3710 case OP_POSSTAR:
3711 case OP_POSPLUS:
3712 case OP_POSQUERY:
3713 case OP_POSUPTO:
3714 case OP_STARI:
3715 case OP_MINSTARI:
3716 case OP_PLUSI:
3717 case OP_MINPLUSI:
3718 case OP_QUERYI:
3719 case OP_MINQUERYI:
3720 case OP_UPTOI:
3721 case OP_MINUPTOI:
3722 case OP_EXACTI:
3723 case OP_POSSTARI:
3724 case OP_POSPLUSI:
3725 case OP_POSQUERYI:
3726 case OP_POSUPTOI:
3727 case OP_NOTSTAR:
3728 case OP_NOTMINSTAR:
3729 case OP_NOTPLUS:
3730 case OP_NOTMINPLUS:
3731 case OP_NOTQUERY:
3732 case OP_NOTMINQUERY:
3733 case OP_NOTUPTO:
3734 case OP_NOTMINUPTO:
3735 case OP_NOTEXACT:
3736 case OP_NOTPOSSTAR:
3737 case OP_NOTPOSPLUS:
3738 case OP_NOTPOSQUERY:
3739 case OP_NOTPOSUPTO:
3740 case OP_NOTSTARI:
3741 case OP_NOTMINSTARI:
3742 case OP_NOTPLUSI:
3743 case OP_NOTMINPLUSI:
3744 case OP_NOTQUERYI:
3745 case OP_NOTMINQUERYI:
3746 case OP_NOTUPTOI:
3747 case OP_NOTMINUPTOI:
3748 case OP_NOTEXACTI:
3749 case OP_NOTPOSSTARI:
3750 case OP_NOTPOSPLUSI:
3751 case OP_NOTPOSQUERYI:
3752 case OP_NOTPOSUPTOI:
3753 if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3754 break;
3755 }
3756 #else
3757 (void)(utf); /* Keep compiler happy by referencing function argument */
3758 #endif
3759 }
3760 }
3761
3762
3763
3764 /*************************************************
3765 * Check for POSIX class syntax *
3766 *************************************************/
3767
3768 /* This function is called when the sequence "[:" or "[." or "[=" is
3769 encountered in a character class. It checks whether this is followed by a
3770 sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3771 reach an unescaped ']' without the special preceding character, return FALSE.
3772
3773 Originally, this function only recognized a sequence of letters between the
3774 terminators, but it seems that Perl recognizes any sequence of characters,
3775 though of course unknown POSIX names are subsequently rejected. Perl gives an
3776 "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3777 didn't consider this to be a POSIX class. Likewise for [:1234:].
3778
3779 The problem in trying to be exactly like Perl is in the handling of escapes. We
3780 have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3781 class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3782 below handles the special case of \], but does not try to do any other escape
3783 processing. This makes it different from Perl for cases such as [:l\ower:]
3784 where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3785 "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
3786 I think.
3787
3788 A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3789 It seems that the appearance of a nested POSIX class supersedes an apparent
3790 external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3791 a digit.
3792
3793 In Perl, unescaped square brackets may also appear as part of class names. For
3794 example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3795 [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3796 seem right at all. PCRE does not allow closing square brackets in POSIX class
3797 names.
3798
3799 Arguments:
3800 ptr pointer to the initial [
3801 endptr where to return the end pointer
3802
3803 Returns: TRUE or FALSE
3804 */
3805
3806 static BOOL
3807 check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3808 {
3809 pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3810 terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3811 for (++ptr; *ptr != CHAR_NULL; ptr++)
3812 {
3813 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3814 ptr++;
3815 else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3816 else
3817 {
3818 if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3819 {
3820 *endptr = ptr;
3821 return TRUE;
3822 }
3823 if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3824 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3825 ptr[1] == CHAR_EQUALS_SIGN) &&
3826 check_posix_syntax(ptr, endptr))
3827 return FALSE;
3828 }
3829 }
3830 return FALSE;
3831 }
3832
3833
3834
3835
3836 /*************************************************
3837 * Check POSIX class name *
3838 *************************************************/
3839
3840 /* This function is called to check the name given in a POSIX-style class entry
3841 such as [:alnum:].
3842
3843 Arguments:
3844 ptr points to the first letter
3845 len the length of the name
3846
3847 Returns: a value representing the name, or -1 if unknown
3848 */
3849
3850 static int
3851 check_posix_name(const pcre_uchar *ptr, int len)
3852 {
3853 const char *pn = posix_names;
3854 register int yield = 0;
3855 while (posix_name_lengths[yield] != 0)
3856 {
3857 if (len == posix_name_lengths[yield] &&
3858 STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3859 pn += posix_name_lengths[yield] + 1;
3860 yield++;
3861 }
3862 return -1;
3863 }
3864
3865
3866 /*************************************************
3867 * Adjust OP_RECURSE items in repeated group *
3868 *************************************************/
3869
3870 /* OP_RECURSE items contain an offset from the start of the regex to the group
3871 that is referenced. This means that groups can be replicated for fixed
3872 repetition simply by copying (because the recursion is allowed to refer to
3873 earlier groups that are outside the current group). However, when a group is
3874 optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3875 inserted before it, after it has been compiled. This means that any OP_RECURSE
3876 items within it that refer to the group itself or any contained groups have to
3877 have their offsets adjusted. That one of the jobs of this function. Before it
3878 is called, the partially compiled regex must be temporarily terminated with
3879 OP_END.
3880
3881 This function has been extended with the possibility of forward references for
3882 recursions and subroutine calls. It must also check the list of such references
3883 for the group we are dealing with. If it finds that one of the recursions in
3884 the current group is on this list, it adjusts the offset in the list, not the
3885 value in the reference (which is a group number).
3886
3887 Arguments:
3888 group points to the start of the group
3889 adjust the amount by which the group is to be moved
3890 utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3891 cd contains pointers to tables etc.
3892 save_hwm the hwm forward reference pointer at the start of the group
3893
3894 Returns: nothing
3895 */
3896
3897 static void
3898 adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3899 pcre_uchar *save_hwm)
3900 {
3901 pcre_uchar *ptr = group;
3902
3903 while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3904 {
3905 int offset;
3906 pcre_uchar *hc;
3907
3908 /* See if this recursion is on the forward reference list. If so, adjust the
3909 reference. */
3910
3911 for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3912 {
3913 offset = (int)GET(hc, 0);
3914 if (cd->start_code + offset == ptr + 1)
3915 {
3916 PUT(hc, 0, offset + adjust);
3917 break;
3918 }
3919 }
3920
3921 /* Otherwise, adjust the recursion offset if it's after the start of this
3922 group. */
3923
3924 if (hc >= cd->hwm)
3925 {
3926 offset = (int)GET(ptr, 1);
3927 if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3928 }
3929
3930 ptr += 1 + LINK_SIZE;
3931 }
3932 }
3933
3934
3935
3936 /*************************************************
3937 * Insert an automatic callout point *
3938 *************************************************/
3939
3940 /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3941 callout points before each pattern item.
3942
3943 Arguments:
3944 code current code pointer
3945 ptr current pattern pointer
3946 cd pointers to tables etc
3947
3948 Returns: new code pointer
3949 */
3950
3951 static pcre_uchar *
3952 auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3953 {
3954 *code++ = OP_CALLOUT;
3955 *code++ = 255;
3956 PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
3957 PUT(code, LINK_SIZE, 0); /* Default length */
3958 return code + 2 * LINK_SIZE;
3959 }
3960
3961
3962
3963 /*************************************************
3964 * Complete a callout item *
3965 *************************************************/
3966
3967 /* A callout item contains the length of the next item in the pattern, which
3968 we can't fill in till after we have reached the relevant point. This is used
3969 for both automatic and manual callouts.
3970
3971 Arguments:
3972 previous_callout points to previous callout item
3973 ptr current pattern pointer
3974 cd pointers to tables etc
3975
3976 Returns: nothing
3977 */
3978
3979 static void
3980 complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
3981 {
3982 int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
3983 PUT(previous_callout, 2 + LINK_SIZE, length);
3984 }
3985
3986
3987
3988 #ifdef SUPPORT_UCP
3989 /*************************************************
3990 * Get othercase range *
3991 *************************************************/
3992
3993 /* This function is passed the start and end of a class range, in UTF-8 mode
3994 with UCP support. It searches up the characters, looking for ranges of
3995 characters in the "other" case. Each call returns the next one, updating the
3996 start address. A character with multiple other cases is returned on its own
3997 with a special return value.
3998
3999 Arguments:
4000 cptr points to starting character value; updated
4001 d end value
4002 ocptr where to put start of othercase range
4003 odptr where to put end of othercase range
4004
4005 Yield: -1 when no more
4006 0 when a range is returned
4007 >0 the CASESET offset for char with multiple other cases
4008 in this case, ocptr contains the original
4009 */
4010
4011 static int
4012 get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4013 pcre_uint32 *odptr)
4014 {
4015 pcre_uint32 c, othercase, next;
4016 unsigned int co;
4017
4018 /* Find the first character that has an other case. If it has multiple other
4019 cases, return its case offset value. */
4020
4021 for (c = *cptr; c <= d; c++)
4022 {
4023 if ((co = UCD_CASESET(c)) != 0)
4024 {
4025 *ocptr = c++; /* Character that has the set */
4026 *cptr = c; /* Rest of input range */
4027 return (int)co;
4028 }
4029 if ((othercase = UCD_OTHERCASE(c)) != c) break;
4030 }
4031
4032 if (c > d) return -1; /* Reached end of range */
4033
4034 *ocptr = othercase;
4035 next = othercase + 1;
4036
4037 for (++c; c <= d; c++)
4038 {
4039 if (UCD_OTHERCASE(c) != next) break;
4040 next++;
4041 }
4042
4043 *odptr = next - 1; /* End of othercase range */
4044 *cptr = c; /* Rest of input range */
4045 return 0;
4046 }
4047 #endif /* SUPPORT_UCP */
4048
4049
4050
4051 /*************************************************
4052 * Add a character or range to a class *
4053 *************************************************/
4054
4055 /* This function packages up the logic of adding a character or range of
4056 characters to a class. The character values in the arguments will be within the
4057 valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4058 mutually recursive with the function immediately below.
4059
4060 Arguments:
4061 classbits the bit map for characters < 256
4062 uchardptr points to the pointer for extra data
4063 options the options word
4064 cd contains pointers to tables etc.
4065 start start of range character
4066 end end of range character
4067
4068 Returns: the number of < 256 characters added
4069 the pointer to extra data is updated
4070 */
4071
4072 static int
4073 add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4074 compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4075 {
4076 pcre_uint32 c;
4077 int n8 = 0;
4078
4079 /* If caseless matching is required, scan the range and process alternate
4080 cases. In Unicode, there are 8-bit characters that have alternate cases that
4081 are greater than 255 and vice-versa. Sometimes we can just extend the original
4082 range. */
4083
4084 if ((options & PCRE_CASELESS) != 0)
4085 {
4086 #ifdef SUPPORT_UCP
4087 if ((options & PCRE_UTF8) != 0)
4088 {
4089 int rc;
4090 pcre_uint32 oc, od;
4091
4092 options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4093 c = start;
4094
4095 while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4096 {
4097 /* Handle a single character that has more than one other case. */
4098
4099 if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4100 PRIV(ucd_caseless_sets) + rc, oc);
4101
4102 /* Do nothing if the other case range is within the original range. */
4103
4104 else if (oc >= start && od <= end) continue;
4105
4106 /* Extend the original range if there is overlap, noting that if oc < c, we
4107 can't have od > end because a subrange is always shorter than the basic
4108 range. Otherwise, use a recursive call to add the additional range. */
4109
4110 else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4111 else if (od > end && oc <= end + 1) end = od; /* Extend upwards */
4112 else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4113 }
4114 }
4115 else
4116 #endif /* SUPPORT_UCP */
4117
4118 /* Not UTF-mode, or no UCP */
4119
4120 for (c = start; c <= end && c < 256; c++)
4121 {
4122 SETBIT(classbits, cd->fcc[c]);
4123 n8++;
4124 }
4125 }
4126
4127 /* Now handle the original range. Adjust the final value according to the bit
4128 length - this means that the same lists of (e.g.) horizontal spaces can be used
4129 in all cases. */
4130
4131 #if defined COMPILE_PCRE8
4132 #ifdef SUPPORT_UTF
4133 if ((options & PCRE_UTF8) == 0)
4134 #endif
4135 if (end > 0xff) end = 0xff;
4136
4137 #elif defined COMPILE_PCRE16
4138 #ifdef SUPPORT_UTF
4139 if ((options & PCRE_UTF16) == 0)
4140 #endif
4141 if (end > 0xffff) end = 0xffff;
4142
4143 #endif /* COMPILE_PCRE[8|16] */
4144
4145 /* If all characters are less than 256, use the bit map. Otherwise use extra
4146 data. */
4147
4148 if (end < 0x100)
4149 {
4150 for (c = start; c <= end; c++)
4151 {
4152 n8++;
4153 SETBIT(classbits, c);
4154 }
4155 }
4156
4157 else
4158 {
4159 pcre_uchar *uchardata = *uchardptr;
4160
4161 #ifdef SUPPORT_UTF
4162 if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4163 {
4164 if (start < end)
4165 {
4166 *uchardata++ = XCL_RANGE;
4167 uchardata += PRIV(ord2utf)(start, uchardata);
4168 uchardata += PRIV(ord2utf)(end, uchardata);
4169 }
4170 else if (start == end)
4171 {
4172 *uchardata++ = XCL_SINGLE;
4173 uchardata += PRIV(ord2utf)(start, uchardata);
4174 }
4175 }
4176 else
4177 #endif /* SUPPORT_UTF */
4178
4179 /* Without UTF support, character values are constrained by the bit length,
4180 and can only be > 256 for 16-bit and 32-bit libraries. */
4181
4182 #ifdef COMPILE_PCRE8
4183 {}
4184 #else
4185 if (start < end)
4186 {
4187 *uchardata++ = XCL_RANGE;
4188 *uchardata++ = start;
4189 *uchardata++ = end;
4190 }
4191 else if (start == end)
4192 {
4193 *uchardata++ = XCL_SINGLE;
4194 *uchardata++ = start;
4195 }
4196 #endif
4197
4198 *uchardptr = uchardata; /* Updata extra data pointer */
4199 }
4200
4201 return n8; /* Number of 8-bit characters */
4202 }
4203
4204
4205
4206
4207 /*************************************************
4208 * Add a list of characters to a class *
4209 *************************************************/
4210
4211 /* This function is used for adding a list of case-equivalent characters to a
4212 class, and also for adding a list of horizontal or vertical whitespace. If the
4213 list is in order (which it should be), ranges of characters are detected and
4214 handled appropriately. This function is mutually recursive with the function
4215 above.
4216
4217 Arguments:
4218 classbits the bit map for characters < 256
4219 uchardptr points to the pointer for extra data
4220 options the options word
4221 cd contains pointers to tables etc.
4222 p points to row of 32-bit values, terminated by NOTACHAR
4223 except character to omit; this is used when adding lists of
4224 case-equivalent characters to avoid including the one we
4225 already know about
4226
4227 Returns: the number of < 256 characters added
4228 the pointer to extra data is updated
4229 */
4230
4231 static int
4232 add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4233 compile_data *cd, const pcre_uint32 *p, unsigned int except)
4234 {
4235 int n8 = 0;
4236 while (p[0] < NOTACHAR)
4237 {
4238 int n = 0;
4239 if (p[0] != except)
4240 {
4241 while(p[n+1] == p[0] + n + 1) n++;
4242 n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4243 }
4244 p += n + 1;
4245 }
4246 return n8;
4247 }
4248
4249
4250
4251 /*************************************************
4252 * Add characters not in a list to a class *
4253 *************************************************/
4254
4255 /* This function is used for adding the complement of a list of horizontal or
4256 vertical whitespace to a class. The list must be in order.
4257
4258 Arguments:
4259 classbits the bit map for characters < 256
4260 uchardptr points to the pointer for extra data
4261 options the options word
4262 cd contains pointers to tables etc.
4263 p points to row of 32-bit values, terminated by NOTACHAR
4264
4265 Returns: the number of < 256 characters added
4266 the pointer to extra data is updated
4267 */
4268
4269 static int
4270 add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4271 int options, compile_data *cd, const pcre_uint32 *p)
4272 {
4273 BOOL utf = (options & PCRE_UTF8) != 0;
4274 int n8 = 0;
4275 if (p[0] > 0)
4276 n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4277 while (p[0] < NOTACHAR)
4278 {
4279 while (p[1] == p[0] + 1) p++;
4280 n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4281 (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4282 p++;
4283 }
4284 return n8;
4285 }
4286
4287
4288
4289 /*************************************************
4290 * Compile one branch *
4291 *************************************************/
4292
4293 /* Scan the pattern, compiling it into the a vector. If the options are
4294 changed during the branch, the pointer is used to change the external options
4295 bits. This function is used during the pre-compile phase when we are trying
4296 to find out the amount of memory needed, as well as during the real compile
4297 phase. The value of lengthptr distinguishes the two phases.
4298
4299 Arguments:
4300 optionsptr pointer to the option bits
4301 codeptr points to the pointer to the current code point
4302 ptrptr points to the current pattern pointer
4303 errorcodeptr points to error code variable
4304 firstcharptr place to put the first required character
4305 firstcharflagsptr place to put the first character flags, or a negative number
4306 reqcharptr place to put the last required character
4307 reqcharflagsptr place to put the last required character flags, or a negative number
4308 bcptr points to current branch chain
4309 cond_depth conditional nesting depth
4310 cd contains pointers to tables etc.
4311 lengthptr NULL during the real compile phase
4312 points to length accumulator during pre-compile phase
4313
4314 Returns: TRUE on success
4315 FALSE, with *errorcodeptr set non-zero on error
4316 */
4317
4318 static BOOL
4319 compile_branch(int *optionsptr, pcre_uchar **codeptr,
4320 const pcre_uchar **ptrptr, int *errorcodeptr,
4321 pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4322 pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4323 branch_chain *bcptr, int cond_depth,
4324 compile_data *cd, int *lengthptr)
4325 {
4326 int repeat_type, op_type;
4327 int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4328 int bravalue = 0;
4329 int greedy_default, greedy_non_default;
4330 pcre_uint32 firstchar, reqchar;
4331 pcre_int32 firstcharflags, reqcharflags;
4332 pcre_uint32 zeroreqchar, zerofirstchar;
4333 pcre_int32 zeroreqcharflags, zerofirstcharflags;
4334 pcre_int32 req_caseopt, reqvary, tempreqvary;
4335 int options = *optionsptr; /* May change dynamically */
4336 int after_manual_callout = 0;
4337 int length_prevgroup = 0;
4338 register pcre_uint32 c;
4339 int escape;
4340 register pcre_uchar *code = *codeptr;
4341 pcre_uchar *last_code = code;
4342 pcre_uchar *orig_code = code;
4343 pcre_uchar *tempcode;
4344 BOOL inescq = FALSE;
4345 BOOL groupsetfirstchar = FALSE;
4346 const pcre_uchar *ptr = *ptrptr;
4347 const pcre_uchar *tempptr;
4348 const pcre_uchar *nestptr = NULL;
4349 pcre_uchar *previous = NULL;
4350 pcre_uchar *previous_callout = NULL;
4351 pcre_uchar *save_hwm = NULL;
4352 pcre_uint8 classbits[32];
4353
4354 /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4355 must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4356 dynamically as we process the pattern. */
4357
4358 #ifdef SUPPORT_UTF
4359 /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4360 BOOL utf = (options & PCRE_UTF8) != 0;
4361 #ifndef COMPILE_PCRE32
4362 pcre_uchar utf_chars[6];
4363 #endif
4364 #else
4365 BOOL utf = FALSE;
4366 #endif
4367
4368 /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4369 class_uchardata always so that it can be passed to add_to_class() always,
4370 though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4371 alternative calls for the different cases. */
4372
4373 pcre_uchar *class_uchardata;
4374 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4375 BOOL xclass;
4376 pcre_uchar *class_uchardata_base;
4377 #endif
4378
4379 #ifdef PCRE_DEBUG
4380 if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4381 #endif
4382
4383 /* Set up the default and non-default settings for greediness */
4384
4385 greedy_default = ((options & PCRE_UNGREEDY) != 0);
4386 greedy_non_default = greedy_default ^ 1;
4387
4388 /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4389 matching encountered yet". It gets changed to REQ_NONE if we hit something that
4390 matches a non-fixed char first char; reqchar just remains unset if we never
4391 find one.
4392
4393 When we hit a repeat whose minimum is zero, we may have to adjust these values
4394 to take the zero repeat into account. This is implemented by setting them to
4395 zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4396 item types that can be repeated set these backoff variables appropriately. */
4397
4398 firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4399 firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4400
4401 /* The variable req_caseopt contains either the REQ_CASELESS value
4402 or zero, according to the current setting of the caseless flag. The
4403 REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4404 firstchar or reqchar variables to record the case status of the
4405 value. This is used only for ASCII characters. */
4406
4407 req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4408
4409 /* Switch on next character until the end of the branch */
4410
4411 for (;; ptr++)
4412 {
4413 BOOL negate_class;
4414 BOOL should_flip_negation;
4415 BOOL possessive_quantifier;
4416 BOOL is_quantifier;
4417 BOOL is_recurse;
4418 BOOL reset_bracount;
4419 int class_has_8bitchar;
4420 int class_one_char;
4421 int newoptions;
4422 int recno;
4423 int refsign;
4424 int skipbytes;
4425 pcre_uint32 subreqchar, subfirstchar;
4426 pcre_int32 subreqcharflags, subfirstcharflags;
4427 int terminator;
4428 unsigned int mclength;
4429 unsigned int tempbracount;
4430 pcre_uint32 ec;
4431 pcre_uchar mcbuffer[8];
4432
4433 /* Get next character in the pattern */
4434
4435 c = *ptr;
4436
4437 /* If we are at the end of a nested substitution, revert to the outer level
4438 string. Nesting only happens one level deep. */
4439
4440 if (c == CHAR_NULL && nestptr != NULL)
4441 {
4442 ptr = nestptr;
4443 nestptr = NULL;
4444 c = *ptr;
4445 }
4446
4447 /* If we are in the pre-compile phase, accumulate the length used for the
4448 previous cycle of this loop. */
4449
4450 if (lengthptr != NULL)
4451 {
4452 #ifdef PCRE_DEBUG
4453 if (code > cd->hwm) cd->hwm = code; /* High water info */
4454 #endif
4455 if (code > cd->start_workspace + cd->workspace_size -
4456 WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4457 {
4458 *errorcodeptr = ERR52;
4459 goto FAILED;
4460 }
4461
4462 /* There is at least one situation where code goes backwards: this is the
4463 case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4464 the class is simply eliminated. However, it is created first, so we have to
4465 allow memory for it. Therefore, don't ever reduce the length at this point.
4466 */
4467
4468 if (code < last_code) code = last_code;
4469
4470 /* Paranoid check for integer overflow */
4471
4472 if (OFLOW_MAX - *lengthptr < code - last_code)
4473 {
4474 *errorcodeptr = ERR20;
4475 goto FAILED;
4476 }
4477
4478 *lengthptr += (int)(code - last_code);
4479 DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4480 (int)(code - last_code), c, c));
4481
4482 /* If "previous" is set and it is not at the start of the work space, move
4483 it back to there, in order to avoid filling up the work space. Otherwise,
4484 if "previous" is NULL, reset the current code pointer to the start. */
4485
4486 if (previous != NULL)
4487 {
4488 if (previous > orig_code)
4489 {
4490 memmove(orig_code, previous, IN_UCHARS(code - previous));
4491 code -= previous - orig_code;
4492 previous = orig_code;
4493 }
4494 }
4495 else code = orig_code;
4496
4497 /* Remember where this code item starts so we can pick up the length
4498 next time round. */
4499
4500 last_code = code;
4501 }
4502
4503 /* In the real compile phase, just check the workspace used by the forward
4504 reference list. */
4505
4506 else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4507 WORK_SIZE_SAFETY_MARGIN)
4508 {
4509 *errorcodeptr = ERR52;
4510 goto FAILED;
4511 }
4512
4513 /* If in \Q...\E, check for the end; if not, we have a literal */
4514
4515 if (inescq && c != CHAR_NULL)
4516 {
4517 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4518 {
4519 inescq = FALSE;
4520 ptr++;
4521 continue;
4522 }
4523 else
4524 {
4525 if (previous_callout != NULL)
4526 {
4527 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4528 complete_callout(previous_callout, ptr, cd);
4529 previous_callout = NULL;
4530 }
4531 if ((options & PCRE_AUTO_CALLOUT) != 0)
4532 {
4533 previous_callout = code;
4534 code = auto_callout(code, ptr, cd);
4535 }
4536 goto NORMAL_CHAR;
4537 }
4538 }
4539
4540 is_quantifier =
4541 c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4542 (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4543
4544 /* Fill in length of a previous callout, except when the next thing is a
4545 quantifier or when processing a property substitution string in UCP mode. */
4546
4547 if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4548 after_manual_callout-- <= 0)
4549 {
4550 if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4551 complete_callout(previous_callout, ptr, cd);
4552 previous_callout = NULL;
4553 }
4554
4555 /* In extended mode, skip white space and comments. */
4556
4557 if ((options & PCRE_EXTENDED) != 0)
4558 {
4559 if (MAX_255(*ptr) && (cd->ctypes[c] & ctype_space) != 0) continue;
4560 if (c == CHAR_NUMBER_SIGN)
4561 {
4562 ptr++;
4563 while (*ptr != CHAR_NULL)
4564 {
4565 if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
4566 ptr++;
4567 #ifdef SUPPORT_UTF
4568 if (utf) FORWARDCHAR(ptr);
4569 #endif
4570 }
4571 if (*ptr != CHAR_NULL) continue;
4572
4573 /* Else fall through to handle end of string */
4574 c = 0;
4575 }
4576 }
4577
4578 /* No auto callout for quantifiers, or while processing property strings that
4579 are substituted for \w etc in UCP mode. */
4580
4581 if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4582 {
4583 previous_callout = code;
4584 code = auto_callout(code, ptr, cd);
4585 }
4586
4587 switch(c)
4588 {
4589 /* ===================================================================*/
4590 case 0: /* The branch terminates at string end */
4591 case CHAR_VERTICAL_LINE: /* or | or ) */
4592 case CHAR_RIGHT_PARENTHESIS:
4593 *firstcharptr = firstchar;
4594 *firstcharflagsptr = firstcharflags;
4595 *reqcharptr = reqchar;
4596 *reqcharflagsptr = reqcharflags;
4597 *codeptr = code;
4598 *ptrptr = ptr;
4599 if (lengthptr != NULL)
4600 {
4601 if (OFLOW_MAX - *lengthptr < code - last_code)
4602 {
4603 *errorcodeptr = ERR20;
4604 goto FAILED;
4605 }
4606 *lengthptr += (int)(code - last_code); /* To include callout length */
4607 DPRINTF((">> end branch\n"));
4608 }
4609 return TRUE;
4610
4611
4612 /* ===================================================================*/
4613 /* Handle single-character metacharacters. In multiline mode, ^ disables
4614 the setting of any following char as a first character. */
4615
4616 case CHAR_CIRCUMFLEX_ACCENT:
4617 previous = NULL;
4618 if ((options & PCRE_MULTILINE) != 0)
4619 {
4620 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4621 *code++ = OP_CIRCM;
4622 }
4623 else *code++ = OP_CIRC;
4624 break;
4625
4626 case CHAR_DOLLAR_SIGN:
4627 previous = NULL;
4628 *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4629 break;
4630
4631 /* There can never be a first char if '.' is first, whatever happens about
4632 repeats. The value of reqchar doesn't change either. */
4633
4634 case CHAR_DOT:
4635 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4636 zerofirstchar = firstchar;
4637 zerofirstcharflags = firstcharflags;
4638 zeroreqchar = reqchar;
4639 zeroreqcharflags = reqcharflags;
4640 previous = code;
4641 *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4642 break;
4643
4644
4645 /* ===================================================================*/
4646 /* Character classes. If the included characters are all < 256, we build a
4647 32-byte bitmap of the permitted characters, except in the special case
4648 where there is only one such character. For negated classes, we build the
4649 map as usual, then invert it at the end. However, we use a different opcode
4650 so that data characters > 255 can be handled correctly.
4651
4652 If the class contains characters outside the 0-255 range, a different
4653 opcode is compiled. It may optionally have a bit map for characters < 256,
4654 but those above are are explicitly listed afterwards. A flag byte tells
4655 whether the bitmap is present, and whether this is a negated class or not.
4656
4657 In JavaScript compatibility mode, an isolated ']' causes an error. In
4658 default (Perl) mode, it is treated as a data character. */
4659
4660 case CHAR_RIGHT_SQUARE_BRACKET:
4661 if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4662 {
4663 *errorcodeptr = ERR64;
4664 goto FAILED;
4665 }
4666 goto NORMAL_CHAR;
4667
4668 case CHAR_LEFT_SQUARE_BRACKET:
4669 previous = code;
4670
4671 /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4672 they are encountered at the top level, so we'll do that too. */
4673
4674 if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4675 ptr[1] == CHAR_EQUALS_SIGN) &&
4676 check_posix_syntax(ptr, &tempptr))
4677 {
4678 *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4679 goto FAILED;
4680 }
4681
4682 /* If the first character is '^', set the negation flag and skip it. Also,
4683 if the first few characters (either before or after ^) are \Q\E or \E we
4684 skip them too. This makes for compatibility with Perl. */
4685
4686 negate_class = FALSE;
4687 for (;;)
4688 {
4689 c = *(++ptr);
4690 if (c == CHAR_BACKSLASH)
4691 {
4692 if (ptr[1] == CHAR_E)
4693 ptr++;
4694 else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4695 ptr += 3;
4696 else
4697 break;
4698 }
4699 else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4700 negate_class = TRUE;
4701 else break;
4702 }
4703
4704 /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4705 an initial ']' is taken as a data character -- the code below handles
4706 that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4707 [^] must match any character, so generate OP_ALLANY. */
4708
4709 if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4710 (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4711 {
4712 *code++ = negate_class? OP_ALLANY : OP_FAIL;
4713 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4714 zerofirstchar = firstchar;
4715 zerofirstcharflags = firstcharflags;
4716 break;
4717 }
4718
4719 /* If a class contains a negative special such as \S, we need to flip the
4720 negation flag at the end, so that support for characters > 255 works
4721 correctly (they are all included in the class). */
4722
4723 should_flip_negation = FALSE;
4724
4725 /* For optimization purposes, we track some properties of the class:
4726 class_has_8bitchar will be non-zero if the class contains at least one <
4727 256 character; class_one_char will be 1 if the class contains just one
4728 character. */
4729
4730 class_has_8bitchar = 0;
4731 class_one_char = 0;
4732
4733 /* Initialize the 32-char bit map to all zeros. We build the map in a
4734 temporary bit of memory, in case the class contains fewer than two
4735 8-bit characters because in that case the compiled code doesn't use the bit
4736 map. */
4737
4738 memset(classbits, 0, 32 * sizeof(pcre_uint8));
4739
4740 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4741 xclass = FALSE;
4742 class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4743 class_uchardata_base = class_uchardata; /* Save the start */
4744 #endif
4745
4746 /* Process characters until ] is reached. By writing this as a "do" it
4747 means that an initial ] is taken as a data character. At the start of the
4748 loop, c contains the first byte of the character. */
4749
4750 if (c != CHAR_NULL) do
4751 {
4752 const pcre_uchar *oldptr;
4753
4754 #ifdef SUPPORT_UTF
4755 if (utf && HAS_EXTRALEN(c))
4756 { /* Braces are required because the */
4757 GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4758 }
4759 #endif
4760
4761 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4762 /* In the pre-compile phase, accumulate the length of any extra
4763 data and reset the pointer. This is so that very large classes that
4764 contain a zillion > 255 characters no longer overwrite the work space
4765 (which is on the stack). We have to remember that there was XCLASS data,
4766 however. */
4767
4768 if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4769 {
4770 xclass = TRUE;
4771 *lengthptr += class_uchardata - class_uchardata_base;
4772 class_uchardata = class_uchardata_base;
4773 }
4774 #endif
4775
4776 /* Inside \Q...\E everything is literal except \E */
4777
4778 if (inescq)
4779 {
4780 if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4781 {
4782 inescq = FALSE; /* Reset literal state */
4783 ptr++; /* Skip the 'E' */
4784 continue; /* Carry on with next */
4785 }
4786 goto CHECK_RANGE; /* Could be range if \E follows */
4787 }
4788
4789 /* Handle POSIX class names. Perl allows a negation extension of the
4790 form [:^name:]. A square bracket that doesn't match the syntax is
4791 treated as a literal. We also recognize the POSIX constructions
4792 [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4793 5.6 and 5.8 do. */
4794
4795 if (c == CHAR_LEFT_SQUARE_BRACKET &&
4796 (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4797 ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4798 {
4799 BOOL local_negate = FALSE;
4800 int posix_class, taboffset, tabopt;
4801 register const pcre_uint8 *cbits = cd->cbits;
4802 pcre_uint8 pbits[32];
4803
4804 if (ptr[1] != CHAR_COLON)
4805 {
4806 *errorcodeptr = ERR31;
4807 goto FAILED;
4808 }
4809
4810 ptr += 2;
4811 if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4812 {
4813 local_negate = TRUE;
4814 should_flip_negation = TRUE; /* Note negative special */
4815 ptr++;
4816 }
4817
4818 posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4819 if (posix_class < 0)
4820 {
4821 *errorcodeptr = ERR30;
4822 goto FAILED;
4823 }
4824
4825 /* If matching is caseless, upper and lower are converted to
4826 alpha. This relies on the fact that the class table starts with
4827 alpha, lower, upper as the first 3 entries. */
4828
4829 if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4830 posix_class = 0;
4831
4832 /* When PCRE_UCP is set, some of the POSIX classes are converted to
4833 different escape sequences that use Unicode properties. */
4834
4835 #ifdef SUPPORT_UCP
4836 if ((options & PCRE_UCP) != 0)
4837 {
4838 int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4839 if (posix_substitutes[pc] != NULL)
4840 {
4841 nestptr = tempptr + 1;
4842 ptr = posix_substitutes[pc] - 1;
4843 continue;
4844 }
4845 }
4846 #endif
4847 /* In the non-UCP case, we build the bit map for the POSIX class in a
4848 chunk of local store because we may be adding and subtracting from it,
4849 and we don't want to subtract bits that may be in the main map already.
4850 At the end we or the result into the bit map that is being built. */
4851
4852 posix_class *= 3;
4853
4854 /* Copy in the first table (always present) */
4855
4856 memcpy(pbits, cbits + posix_class_maps[posix_class],
4857 32 * sizeof(pcre_uint8));
4858
4859 /* If there is a second table, add or remove it as required. */
4860
4861 taboffset = posix_class_maps[posix_class + 1];
4862 tabopt = posix_class_maps[posix_class + 2];
4863
4864 if (taboffset >= 0)
4865 {
4866 if (tabopt >= 0)
4867 for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
4868 else
4869 for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
4870 }
4871
4872 /* Now see if we need to remove any special characters. An option
4873 value of 1 removes vertical space and 2 removes underscore. */
4874
4875 if (tabopt < 0) tabopt = -tabopt;
4876 if (tabopt == 1) pbits[1] &= ~0x3c;
4877 else if (tabopt == 2) pbits[11] &= 0x7f;
4878
4879 /* Add the POSIX table or its complement into the main table that is
4880 being built and we are done. */
4881
4882 if (local_negate)
4883 for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
4884 else
4885 for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
4886
4887 ptr = tempptr + 1;
4888 /* Every class contains at least one < 256 character. */
4889 class_has_8bitchar = 1;
4890 /* Every class contains at least two characters. */
4891 class_one_char = 2;
4892 continue; /* End of POSIX syntax handling */
4893 }
4894
4895 /* Backslash may introduce a single character, or it may introduce one
4896 of the specials, which just set a flag. The sequence \b is a special
4897 case. Inside a class (and only there) it is treated as backspace. We
4898 assume that other escapes have more than one character in them, so
4899 speculatively set both class_has_8bitchar and class_one_char bigger
4900 than one. Unrecognized escapes fall through and are either treated
4901 as literal characters (by default), or are faulted if
4902 PCRE_EXTRA is set. */
4903
4904 if (c == CHAR_BACKSLASH)
4905 {
4906 escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
4907 TRUE);
4908 if (*errorcodeptr != 0) goto FAILED;
4909 if (escape == 0) c = ec;
4910 else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
4911 else if (escape == ESC_N) /* \N is not supported in a class */
4912 {
4913 *errorcodeptr = ERR71;
4914 goto FAILED;
4915 }
4916 else if (escape == ESC_Q) /* Handle start of quoted string */
4917 {
4918 if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
4919 {
4920 ptr += 2; /* avoid empty string */
4921 }
4922 else inescq = TRUE;
4923 continue;
4924 }
4925 else if (escape == ESC_E) continue; /* Ignore orphan \E */
4926
4927 else
4928 {
4929 register const pcre_uint8 *cbits = cd->cbits;
4930 /* Every class contains at least two < 256 characters. */
4931 class_has_8bitchar++;
4932 /* Every class contains at least two characters. */
4933 class_one_char += 2;
4934
4935 switch (escape)
4936 {
4937 #ifdef SUPPORT_UCP
4938 case ESC_du: /* These are the values given for \d etc */
4939 case ESC_DU: /* when PCRE_UCP is set. We replace the */
4940 case ESC_wu: /* escape sequence with an appropriate \p */
4941 case ESC_WU: /* or \P to test Unicode properties instead */
4942 case ESC_su: /* of the default ASCII testing. */
4943 case ESC_SU:
4944 nestptr = ptr;
4945 ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
4946 class_has_8bitchar--; /* Undo! */
4947 continue;
4948 #endif
4949 case ESC_d:
4950 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
4951 continue;
4952
4953 case ESC_D:
4954 should_flip_negation = TRUE;
4955 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
4956 continue;
4957
4958 case ESC_w:
4959 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
4960 continue;
4961
4962 case ESC_W:
4963 should_flip_negation = TRUE;
4964 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
4965 continue;
4966
4967 /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
4968 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
4969 previously set by something earlier in the character class.
4970 Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
4971 we could just adjust the appropriate bit. From PCRE 8.34 we no
4972 longer treat \s and \S specially. */
4973
4974 case ESC_s:
4975 for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
4976 continue;
4977
4978 case ESC_S:
4979 should_flip_negation = TRUE;
4980 for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
4981 continue;
4982
4983 /* The rest apply in both UCP and non-UCP cases. */
4984
4985 case ESC_h:
4986 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4987 PRIV(hspace_list), NOTACHAR);
4988 continue;
4989
4990 case ESC_H:
4991 (void)add_not_list_to_class(classbits, &class_uchardata, options,
4992 cd, PRIV(hspace_list));
4993 continue;
4994
4995 case ESC_v:
4996 (void)add_list_to_class(classbits, &class_uchardata, options, cd,
4997 PRIV(vspace_list), NOTACHAR);
4998 continue;
4999
5000 case ESC_V:
5001 (void)add_not_list_to_class(classbits, &class_uchardata, options,
5002 cd, PRIV(vspace_list));
5003 continue;
5004
5005 #ifdef SUPPORT_UCP
5006 case ESC_p:
5007 case ESC_P:
5008 {
5009 BOOL negated;
5010 unsigned int ptype = 0, pdata = 0;
5011 if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5012 goto FAILED;
5013 *class_uchardata++ = ((escape == ESC_p) != negated)?
5014 XCL_PROP : XCL_NOTPROP;
5015 *class_uchardata++ = ptype;
5016 *class_uchardata++ = pdata;
5017 class_has_8bitchar--; /* Undo! */
5018 continue;
5019 }
5020 #endif
5021 /* Unrecognized escapes are faulted if PCRE is running in its
5022 strict mode. By default, for compatibility with Perl, they are
5023 treated as literals. */
5024
5025 default:
5026 if ((options & PCRE_EXTRA) != 0)
5027 {
5028 *errorcodeptr = ERR7;
5029 goto FAILED;
5030 }
5031 class_has_8bitchar--; /* Undo the speculative increase. */
5032 class_one_char -= 2; /* Undo the speculative increase. */
5033 c = *ptr; /* Get the final character and fall through */
5034 break;
5035 }
5036 }
5037
5038 /* Fall through if the escape just defined a single character (c >= 0).
5039 This may be greater than 256. */
5040
5041 escape = 0;
5042
5043 } /* End of backslash handling */
5044
5045 /* A character may be followed by '-' to form a range. However, Perl does
5046 not permit ']' to be the end of the range. A '-' character at the end is
5047 treated as a literal. Perl ignores orphaned \E sequences entirely. The
5048 code for handling \Q and \E is messy. */
5049
5050 CHECK_RANGE:
5051 while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5052 {
5053 inescq = FALSE;
5054 ptr += 2;
5055 }
5056 oldptr = ptr;
5057
5058 /* Remember if \r or \n were explicitly used */
5059
5060 if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5061
5062 /* Check for range */
5063
5064 if (!inescq && ptr[1] == CHAR_MINUS)
5065 {
5066 pcre_uint32 d;
5067 ptr += 2;
5068 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5069
5070 /* If we hit \Q (not followed by \E) at this point, go into escaped
5071 mode. */
5072
5073 while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5074 {
5075 ptr += 2;
5076 if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5077 { ptr += 2; continue; }
5078 inescq = TRUE;
5079 break;
5080 }
5081
5082 /* Minus (hyphen) at the end of a class is treated as a literal, so put
5083 back the pointer and jump to handle the character that preceded it. */
5084
5085 if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5086 {
5087 ptr = oldptr;
5088 goto CLASS_SINGLE_CHARACTER;
5089 }
5090
5091 /* Otherwise, we have a potential range; pick up the next character */
5092
5093 #ifdef SUPPORT_UTF
5094 if (utf)
5095 { /* Braces are required because the */
5096 GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5097 }
5098 else
5099 #endif
5100 d = *ptr; /* Not UTF-8 mode */
5101
5102 /* The second part of a range can be a single-character escape, but
5103 not any of the other escapes. Perl 5.6 treats a hyphen as a literal
5104 in such circumstances. */
5105
5106 if (!inescq && d == CHAR_BACKSLASH)
5107 {
5108 int descape;
5109 descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5110 if (*errorcodeptr != 0) goto FAILED;
5111
5112 /* \b is backspace; any other special means the '-' was literal. */
5113
5114 if (descape != 0)
5115 {
5116 if (descape == ESC_b) d = CHAR_BS; else
5117 {
5118 ptr = oldptr;
5119 goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5120 }
5121 }
5122 }
5123
5124 /* Check that the two values are in the correct order. Optimize
5125 one-character ranges. */
5126
5127 if (d < c)
5128 {
5129 *errorcodeptr = ERR8;
5130 goto FAILED;
5131 }
5132 if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5133
5134 /* We have found a character range, so single character optimizations
5135 cannot be done anymore. Any value greater than 1 indicates that there
5136 is more than one character. */
5137
5138 class_one_char = 2;
5139
5140 /* Remember an explicit \r or \n, and add the range to the class. */
5141
5142 if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5143
5144 class_has_8bitchar +=
5145 add_to_class(classbits, &class_uchardata, options, cd, c, d);
5146
5147 continue; /* Go get the next char in the class */
5148 }
5149
5150 /* Handle a single character - we can get here for a normal non-escape
5151 char, or after \ that introduces a single character or for an apparent
5152 range that isn't. Only the value 1 matters for class_one_char, so don't
5153 increase it if it is already 2 or more ... just in case there's a class
5154 with a zillion characters in it. */
5155
5156 CLASS_SINGLE_CHARACTER:
5157 if (class_one_char < 2) class_one_char++;
5158
5159 /* If class_one_char is 1, we have the first single character in the
5160 class, and there have been no prior ranges, or XCLASS items generated by
5161 escapes. If this is the final character in the class, we can optimize by
5162 turning the item into a 1-character OP_CHAR[I] if it's positive, or
5163 OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5164 to be set. Otherwise, there can be no first char if this item is first,
5165 whatever repeat count may follow. In the case of reqchar, save the
5166 previous value for reinstating. */
5167
5168 if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5169 {
5170 ptr++;
5171 zeroreqchar = reqchar;
5172 zeroreqcharflags = reqcharflags;
5173
5174 if (negate_class)
5175 {
5176 #ifdef SUPPORT_UCP
5177 int d;
5178 #endif
5179 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5180 zerofirstchar = firstchar;
5181 zerofirstcharflags = firstcharflags;
5182
5183 /* For caseless UTF-8 mode when UCP support is available, check
5184 whether this character has more than one other case. If so, generate
5185 a special OP_NOTPROP item instead of OP_NOTI. */
5186
5187 #ifdef SUPPORT_UCP
5188 if (utf && (options & PCRE_CASELESS) != 0 &&
5189 (d = UCD_CASESET(c)) != 0)
5190 {
5191 *code++ = OP_NOTPROP;
5192 *code++ = PT_CLIST;
5193 *code++ = d;
5194 }
5195 else
5196 #endif
5197 /* Char has only one other case, or UCP not available */
5198
5199 {
5200 *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5201 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5202 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5203 code += PRIV(ord2utf)(c, code);
5204 else
5205 #endif
5206 *code++ = c;
5207 }
5208
5209 /* We are finished with this character class */
5210
5211 goto END_CLASS;
5212 }
5213
5214 /* For a single, positive character, get the value into mcbuffer, and
5215 then we can handle this with the normal one-character code. */
5216
5217 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5218 if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5219 mclength = PRIV(ord2utf)(c, mcbuffer);
5220 else
5221 #endif
5222 {
5223 mcbuffer[0] = c;
5224 mclength = 1;
5225 }
5226 goto ONE_CHAR;
5227 } /* End of 1-char optimization */
5228
5229 /* There is more than one character in the class, or an XCLASS item
5230 has been generated. Add this character to the class. */
5231
5232 class_has_8bitchar +=
5233 add_to_class(classbits, &class_uchardata, options, cd, c, c);
5234 }
5235
5236 /* Loop until ']' reached. This "while" is the end of the "do" far above.
5237 If we are at the end of an internal nested string, revert to the outer
5238 string. */
5239
5240 while (((c = *(++ptr)) != CHAR_NULL ||
5241 (nestptr != NULL &&
5242 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5243 (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5244
5245 /* Check for missing terminating ']' */
5246
5247 if (c == CHAR_NULL)
5248 {
5249 *errorcodeptr = ERR6;
5250 goto FAILED;
5251 }
5252
5253 /* We will need an XCLASS if data has been placed in class_uchardata. In
5254 the second phase this is a sufficient test. However, in the pre-compile
5255 phase, class_uchardata gets emptied to prevent workspace overflow, so it
5256 only if the very last character in the class needs XCLASS will it contain
5257 anything at this point. For this reason, xclass gets set TRUE above when
5258 uchar_classdata is emptied, and that's why this code is the way it is here
5259 instead of just doing a test on class_uchardata below. */
5260
5261 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5262 if (class_uchardata > class_uchardata_base) xclass = TRUE;
5263 #endif
5264
5265 /* If this is the first thing in the branch, there can be no first char
5266 setting, whatever the repeat count. Any reqchar setting must remain
5267 unchanged after any kind of repeat. */
5268
5269 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5270 zerofirstchar = firstchar;
5271 zerofirstcharflags = firstcharflags;
5272 zeroreqchar = reqchar;
5273 zeroreqcharflags = reqcharflags;
5274
5275 /* If there are characters with values > 255, we have to compile an
5276 extended class, with its own opcode, unless there was a negated special
5277 such as \S in the class, and PCRE_UCP is not set, because in that case all
5278 characters > 255 are in the class, so any that were explicitly given as
5279 well can be ignored. If (when there are explicit characters > 255 that must
5280 be listed) there are no characters < 256, we can omit the bitmap in the
5281 actual compiled code. */
5282
5283 #ifdef SUPPORT_UTF
5284 if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5285 #elif !defined COMPILE_PCRE8
5286 if (xclass && !should_flip_negation)
5287 #endif
5288 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5289 {
5290 *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5291 *code++ = OP_XCLASS;
5292 code += LINK_SIZE;
5293 *code = negate_class? XCL_NOT:0;
5294
5295 /* If the map is required, move up the extra data to make room for it;
5296 otherwise just move the code pointer to the end of the extra data. */
5297
5298 if (class_has_8bitchar > 0)
5299 {
5300 *code++ |= XCL_MAP;
5301 memmove(code + (32 / sizeof(pcre_uchar)), code,
5302 IN_UCHARS(class_uchardata - code));
5303 memcpy(code, classbits, 32);
5304 code = class_uchardata + (32 / sizeof(pcre_uchar));
5305 }
5306 else code = class_uchardata;
5307
5308 /* Now fill in the complete length of the item */
5309
5310 PUT(previous, 1, (int)(code - previous));
5311 break; /* End of class handling */
5312 }
5313 #endif
5314
5315 /* If there are no characters > 255, or they are all to be included or
5316 excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5317 whole class was negated and whether there were negative specials such as \S
5318 (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5319 negating it if necessary. */
5320
5321 *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5322 if (lengthptr == NULL) /* Save time in the pre-compile phase */
5323 {
5324 if (negate_class)
5325 for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5326 memcpy(code, classbits, 32);
5327 }
5328 code += 32 / sizeof(pcre_uchar);
5329
5330 END_CLASS:
5331 break;
5332
5333
5334 /* ===================================================================*/
5335 /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5336 has been tested above. */
5337
5338 case CHAR_LEFT_CURLY_BRACKET:
5339 if (!is_quantifier) goto NORMAL_CHAR;
5340 ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5341 if (*errorcodeptr != 0) goto FAILED;
5342 goto REPEAT;
5343
5344 case CHAR_ASTERISK:
5345 repeat_min = 0;
5346 repeat_max = -1;
5347 goto REPEAT;
5348
5349 case CHAR_PLUS:
5350 repeat_min = 1;
5351 repeat_max = -1;
5352 goto REPEAT;
5353
5354 case CHAR_QUESTION_MARK:
5355 repeat_min = 0;
5356 repeat_max = 1;
5357
5358 REPEAT:
5359 if (previous == NULL)
5360 {
5361 *errorcodeptr = ERR9;
5362 goto FAILED;
5363 }
5364
5365 if (repeat_min == 0)
5366 {
5367 firstchar = zerofirstchar; /* Adjust for zero repeat */
5368 firstcharflags = zerofirstcharflags;
5369 reqchar = zeroreqchar; /* Ditto */
5370 reqcharflags = zeroreqcharflags;
5371 }
5372
5373 /* Remember whether this is a variable length repeat */
5374
5375 reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5376
5377 op_type = 0; /* Default single-char op codes */
5378 possessive_quantifier = FALSE; /* Default not possessive quantifier */
5379
5380 /* Save start of previous item, in case we have to move it up in order to
5381 insert something before it. */
5382
5383 tempcode = previous;
5384
5385 /* If the next character is '+', we have a possessive quantifier. This
5386 implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5387 If the next character is '?' this is a minimizing repeat, by default,
5388 but if PCRE_UNGREEDY is set, it works the other way round. We change the
5389 repeat type to the non-default. */
5390
5391 if (ptr[1] == CHAR_PLUS)
5392 {
5393 repeat_type = 0; /* Force greedy */
5394 possessive_quantifier = TRUE;
5395 ptr++;
5396 }
5397 else if (ptr[1] == CHAR_QUESTION_MARK)
5398 {
5399 repeat_type = greedy_non_default;
5400 ptr++;
5401 }
5402 else repeat_type = greedy_default;
5403
5404 /* If previous was a recursion call, wrap it in atomic brackets so that
5405 previous becomes the atomic group. All recursions were so wrapped in the
5406 past, but it no longer happens for non-repeated recursions. In fact, the
5407 repeated ones could be re-implemented independently so as not to need this,
5408 but for the moment we rely on the code for repeating groups. */
5409
5410 if (*previous == OP_RECURSE)
5411 {
5412 memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5413 *previous = OP_ONCE;
5414 PUT(previous, 1, 2 + 2*LINK_SIZE);
5415 previous[2 + 2*LINK_SIZE] = OP_KET;
5416 PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5417 code += 2 + 2 * LINK_SIZE;
5418 length_prevgroup = 3 + 3*LINK_SIZE;
5419
5420 /* When actually compiling, we need to check whether this was a forward
5421 reference, and if so, adjust the offset. */
5422
5423 if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5424 {
5425 int offset = GET(cd->hwm, -LINK_SIZE);
5426 if (offset == previous + 1 - cd->start_code)
5427 PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5428 }
5429 }
5430
5431 /* Now handle repetition for the different types of item. */
5432
5433 /* If previous was a character or negated character match, abolish the item
5434 and generate a repeat item instead. If a char item has a minimum of more
5435 than one, ensure that it is set in reqchar - it might not be if a sequence
5436 such as x{3} is the first thing in a branch because the x will have gone
5437 into firstchar instead. */
5438
5439 if (*previous == OP_CHAR || *previous == OP_CHARI
5440 || *previous == OP_NOT || *previous == OP_NOTI)
5441 {
5442 switch (*previous)
5443 {
5444 default: /* Make compiler happy. */
5445 case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5446 case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5447 case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5448 case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5449 }
5450
5451 /* Deal with UTF characters that take up more than one character. It's
5452 easier to write this out separately than try to macrify it. Use c to
5453 hold the length of the character in bytes, plus UTF_LENGTH to flag that
5454 it's a length rather than a small character. */
5455
5456 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5457 if (utf && NOT_FIRSTCHAR(code[-1]))
5458 {
5459 pcre_uchar *lastchar = code - 1;
5460 BACKCHAR(lastchar);
5461 c = (int)(code - lastchar); /* Length of UTF-8 character */
5462 memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5463 c |= UTF_LENGTH; /* Flag c as a length */
5464 }
5465 else
5466 #endif /* SUPPORT_UTF */
5467
5468 /* Handle the case of a single charater - either with no UTF support, or
5469 with UTF disabled, or for a single character UTF character. */
5470 {
5471 c = code[-1];
5472 if (*previous <= OP_CHARI && repeat_min > 1)
5473 {
5474 reqchar = c;
5475 reqcharflags = req_caseopt | cd->req_varyopt;
5476 }
5477 }
5478
5479 goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5480 }
5481
5482 /* If previous was a character type match (\d or similar), abolish it and
5483 create a suitable repeat item. The code is shared with single-character
5484 repeats by setting op_type to add a suitable offset into repeat_type. Note
5485 the the Unicode property types will be present only when SUPPORT_UCP is
5486 defined, but we don't wrap the little bits of code here because it just
5487 makes it horribly messy. */
5488
5489 else if (*previous < OP_EODN)
5490 {
5491 pcre_uchar *oldcode;
5492 int prop_type, prop_value;
5493 op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5494 c = *previous;
5495
5496 OUTPUT_SINGLE_REPEAT:
5497 if (*previous == OP_PROP || *previous == OP_NOTPROP)
5498 {
5499 prop_type = previous[1];
5500 prop_value = previous[2];
5501 }
5502 else prop_type = prop_value = -1;
5503
5504 oldcode = code;
5505 code = previous; /* Usually overwrite previous item */
5506
5507 /* If the maximum is zero then the minimum must also be zero; Perl allows
5508 this case, so we do too - by simply omitting the item altogether. */
5509
5510 if (repeat_max == 0) goto END_REPEAT;
5511
5512 /* Combine the op_type with the repeat_type */
5513
5514 repeat_type += op_type;
5515
5516 /* A minimum of zero is handled either as the special case * or ?, or as
5517 an UPTO, with the maximum given. */
5518
5519 if (repeat_min == 0)
5520 {
5521 if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5522 else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5523 else
5524 {
5525 *code++ = OP_UPTO + repeat_type;
5526 PUT2INC(code, 0, repeat_max);
5527 }
5528 }
5529
5530 /* A repeat minimum of 1 is optimized into some special cases. If the
5531 maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5532 left in place and, if the maximum is greater than 1, we use OP_UPTO with
5533 one less than the maximum. */
5534
5535 else if (repeat_min == 1)
5536 {
5537 if (repeat_max == -1)
5538 *code++ = OP_PLUS + repeat_type;
5539 else
5540 {
5541 code = oldcode; /* leave previous item in place */
5542 if (repeat_max == 1) goto END_REPEAT;
5543 *code++ = OP_UPTO + repeat_type;
5544 PUT2INC(code, 0, repeat_max - 1);
5545 }
5546 }
5547
5548 /* The case {n,n} is just an EXACT, while the general case {n,m} is
5549 handled as an EXACT followed by an UPTO. */
5550
5551 else
5552 {
5553 *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5554 PUT2INC(code, 0, repeat_min);
5555
5556 /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5557 we have to insert the character for the previous code. For a repeated
5558 Unicode property match, there are two extra bytes that define the
5559 required property. In UTF-8 mode, long characters have their length in
5560 c, with the UTF_LENGTH bit as a flag. */
5561
5562 if (repeat_max < 0)
5563 {
5564 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5565 if (utf && (c & UTF_LENGTH) != 0)
5566 {
5567 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5568 code += c & 7;
5569 }
5570 else
5571 #endif
5572 {
5573 *code++ = c;
5574 if (prop_type >= 0)
5575 {
5576 *code++ = prop_type;
5577 *code++ = prop_value;
5578 }
5579 }
5580 *code++ = OP_STAR + repeat_type;
5581 }
5582
5583 /* Else insert an UPTO if the max is greater than the min, again
5584 preceded by the character, for the previously inserted code. If the
5585 UPTO is just for 1 instance, we can use QUERY instead. */
5586
5587 else if (repeat_max != repeat_min)
5588 {
5589 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5590 if (utf && (c & UTF_LENGTH) != 0)
5591 {
5592 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5593 code += c & 7;
5594 }
5595 else
5596 #endif
5597 *code++ = c;
5598 if (prop_type >= 0)
5599 {
5600 *code++ = prop_type;
5601 *code++ = prop_value;
5602 }
5603 repeat_max -= repeat_min;
5604
5605 if (repeat_max == 1)
5606 {
5607 *code++ = OP_QUERY + repeat_type;
5608 }
5609 else
5610 {
5611 *code++ = OP_UPTO + repeat_type;
5612 PUT2INC(code, 0, repeat_max);
5613 }
5614 }
5615 }
5616
5617 /* The character or character type itself comes last in all cases. */
5618
5619 #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5620 if (utf && (c & UTF_LENGTH) != 0)
5621 {
5622 memcpy(code, utf_chars, IN_UCHARS(c & 7));
5623 code += c & 7;
5624 }
5625 else
5626 #endif
5627 *code++ = c;
5628
5629 /* For a repeated Unicode property match, there are two extra bytes that
5630 define the required property. */
5631
5632 #ifdef SUPPORT_UCP
5633 if (prop_type >= 0)
5634 {
5635 *code++ = prop_type;
5636 *code++ = prop_value;
5637 }
5638 #endif
5639 }
5640
5641 /* If previous was a character class or a back reference, we put the repeat
5642 stuff after it, but just skip the item if the repeat was {0,0}. */
5643
5644 else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5645 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5646 *previous == OP_XCLASS ||
5647 #endif
5648 *previous == OP_REF || *previous == OP_REFI ||
5649 *previous == OP_DNREF || *previous == OP_DNREFI)
5650 {
5651 if (repeat_max == 0)
5652 {
5653 code = previous;
5654 goto END_REPEAT;
5655 }
5656
5657 if (repeat_min == 0 && repeat_max == -1)
5658 *code++ = OP_CRSTAR + repeat_type;
5659 else if (repeat_min == 1 && repeat_max == -1)
5660 *code++ = OP_CRPLUS + repeat_type;
5661 else if (repeat_min == 0 && repeat_max == 1)
5662 *code++ = OP_CRQUERY + repeat_type;
5663 else
5664 {
5665 *code++ = OP_CRRANGE + repeat_type;
5666 PUT2INC(code, 0, repeat_min);
5667 if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5668 PUT2INC(code, 0, repeat_max);
5669 }
5670 }
5671
5672 /* If previous was a bracket group, we may have to replicate it in certain
5673 cases. Note that at this point we can encounter only the "basic" bracket
5674 opcodes such as BRA and CBRA, as this is the place where they get converted
5675 into the more special varieties such as BRAPOS and SBRA. A test for >=
5676 OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5677 ASSERTBACK_NOT, ONCE, BRA, CBRA, and COND. Originally, PCRE did not allow
5678 repetition of assertions, but now it does, for Perl compatibility. */
5679
5680 else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5681 {
5682 register int i;
5683 int len = (int)(code - previous);
5684 pcre_uchar *bralink = NULL;
5685 pcre_uchar *brazeroptr = NULL;
5686
5687 /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5688 we just ignore the repeat. */
5689
5690 if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5691 goto END_REPEAT;
5692
5693 /* There is no sense in actually repeating assertions. The only potential
5694 use of repetition is in cases when the assertion is optional. Therefore,
5695 if the minimum is greater than zero, just ignore the repeat. If the
5696 maximum is not not zero or one, set it to 1. */
5697
5698 if (*previous < OP_ONCE) /* Assertion */
5699 {
5700 if (repeat_min > 0) goto END_REPEAT;
5701 if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5702 }
5703
5704 /* The case of a zero minimum is special because of the need to stick
5705 OP_BRAZERO in front of it, and because the group appears once in the
5706 data, whereas in other cases it appears the minimum number of times. For
5707 this reason, it is simplest to treat this case separately, as otherwise
5708 the code gets far too messy. There are several special subcases when the
5709 minimum is zero. */
5710
5711 if (repeat_min == 0)
5712 {
5713 /* If the maximum is also zero, we used to just omit the group from the
5714 output altogether, like this:
5715
5716 ** if (repeat_max == 0)
5717 ** {
5718 ** code = previous;
5719 ** goto END_REPEAT;
5720 ** }
5721
5722 However, that fails when a group or a subgroup within it is referenced
5723 as a subroutine from elsewhere in the pattern, so now we stick in
5724 OP_SKIPZERO in front of it so that it is skipped on execution. As we
5725 don't have a list of which groups are referenced, we cannot do this
5726 selectively.
5727
5728 If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5729 and do no more at this point. However, we do need to adjust any
5730 OP_RECURSE calls inside the group that refer to the group itself or any
5731 internal or forward referenced group, because the offset is from the
5732 start of the whole regex. Temporarily terminate the pattern while doing
5733 this. */
5734
5735 if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5736 {
5737 *code = OP_END;
5738 adjust_recurse(previous, 1, utf, cd, save_hwm);
5739 memmove(previous + 1, previous, IN_UCHARS(len));
5740 code++;
5741 if (repeat_max == 0)
5742 {
5743 *previous++ = OP_SKIPZERO;
5744 goto END_REPEAT;
5745 }
5746 brazeroptr = previous; /* Save for possessive optimizing */
5747 *previous++ = OP_BRAZERO + repeat_type;
5748 }
5749
5750 /* If the maximum is greater than 1 and limited, we have to replicate
5751 in a nested fashion, sticking OP_BRAZERO before each set of brackets.
5752 The first one has to be handled carefully because it's the original
5753 copy, which has to be moved up. The remainder can be handled by code
5754 that is common with the non-zero minimum case below. We have to
5755 adjust the value or repeat_max, since one less copy is required. Once
5756 again, we may have to adjust any OP_RECURSE calls inside the group. */
5757
5758 else
5759 {
5760 int offset;
5761 *code = OP_END;
5762 adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5763 memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5764 code += 2 + LINK_SIZE;
5765 *previous++ = OP_BRAZERO + repeat_type;
5766 *previous++ = OP_BRA;
5767
5768 /* We chain together the bracket offset fields that have to be
5769 filled in later when the ends of the brackets are reached. */
5770
5771 offset = (bralink == NULL)? 0 : (int)(previous - bralink);
5772 bralink = previous;
5773 PUTINC(previous, 0, offset);
5774 }
5775
5776 repeat_max--;
5777 }
5778
5779 /* If the minimum is greater than zero, replicate the group as many
5780 times as necessary, and adjust the maximum to the number of subsequent
5781 copies that we need. If we set a first char from the group, and didn't
5782 set a required char, copy the latter from the former. If there are any
5783 forward reference subroutine calls in the group, there will be entries on
5784 the workspace list; replicate these with an appropriate increment. */
5785
5786 else
5787 {
5788 if (repeat_min > 1)
5789 {
5790 /* In the pre-compile phase, we don't actually do the replication. We
5791 just adjust the length as if we had. Do some paranoid checks for
5792 potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
5793 integer type when available, otherwise double. */
5794
5795 if (lengthptr != NULL)
5796 {
5797 int delta = (repeat_min - 1)*length_prevgroup;
5798 if ((INT64_OR_DOUBLE)(repeat_min - 1)*
5799 (INT64_OR_DOUBLE)length_prevgroup >
5800 (INT64_OR_DOUBLE)INT_MAX ||
5801 OFLOW_MAX - *lengthptr < delta)
5802 {
5803 *errorcodeptr = ERR20;
5804 goto FAILED;
5805 }
5806 *lengthptr += delta;
5807 }
5808
5809 /* This is compiling for real. If there is a set first byte for
5810 the group, and we have not yet set a "required byte", set it. Make
5811 sure there is enough workspace for copying forward references before
5812 doing the copy. */
5813
5814 else
5815 {
5816 if (groupsetfirstchar && reqcharflags < 0)
5817 {
5818 reqchar = firstchar;
5819 reqcharflags = firstcharflags;
5820 }
5821
5822 for (i = 1; i < repeat_min; i++)
5823 {
5824 pcre_uchar *hc;
5825 pcre_uchar *this_hwm = cd->hwm;
5826 memcpy(code, previous, IN_UCHARS(len));
5827
5828 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5829 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5830 {
5831 int save_offset = save_hwm - cd->start_workspace;
5832 int this_offset = this_hwm - cd->start_workspace;
5833 *errorcodeptr = expand_workspace(cd);
5834 if (*errorcodeptr != 0) goto FAILED;
5835 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5836 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5837 }
5838
5839 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5840 {
5841 PUT(cd->hwm, 0, GET(hc, 0) + len);
5842 cd->hwm += LINK_SIZE;
5843 }
5844 save_hwm = this_hwm;
5845 code += len;
5846 }
5847 }
5848 }
5849
5850 if (repeat_max > 0) repeat_max -= repeat_min;
5851 }
5852
5853 /* This code is common to both the zero and non-zero minimum cases. If
5854 the maximum is limited, it replicates the group in a nested fashion,
5855 remembering the bracket starts on a stack. In the case of a zero minimum,
5856 the first one was set up above. In all cases the repeat_max now specifies
5857 the number of additional copies needed. Again, we must remember to
5858 replicate entries on the forward reference list. */
5859
5860 if (repeat_max >= 0)
5861 {
5862 /* In the pre-compile phase, we don't actually do the replication. We
5863 just adjust the length as if we had. For each repetition we must add 1
5864 to the length for BRAZERO and for all but the last repetition we must
5865 add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
5866 paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
5867 a 64-bit integer type when available, otherwise double. */
5868
5869 if (lengthptr != NULL && repeat_max > 0)
5870 {
5871 int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
5872 2 - 2*LINK_SIZE; /* Last one doesn't nest */
5873 if ((INT64_OR_DOUBLE)repeat_max *
5874 (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
5875 > (INT64_OR_DOUBLE)INT_MAX ||
5876 OFLOW_MAX - *lengthptr < delta)
5877 {
5878 *errorcodeptr = ERR20;
5879 goto FAILED;
5880 }
5881 *lengthptr += delta;
5882 }
5883
5884 /* This is compiling for real */
5885
5886 else for (i = repeat_max - 1; i >= 0; i--)
5887 {
5888 pcre_uchar *hc;
5889 pcre_uchar *this_hwm = cd->hwm;
5890
5891 *code++ = OP_BRAZERO + repeat_type;
5892
5893 /* All but the final copy start a new nesting, maintaining the
5894 chain of brackets outstanding. */
5895
5896 if (i != 0)
5897 {
5898 int offset;
5899 *code++ = OP_BRA;
5900 offset = (bralink == NULL)? 0 : (int)(code - bralink);
5901 bralink = code;
5902 PUTINC(code, 0, offset);
5903 }
5904
5905 memcpy(code, previous, IN_UCHARS(len));
5906
5907 /* Ensure there is enough workspace for forward references before
5908 copying them. */
5909
5910 while (cd->hwm > cd->start_workspace + cd->workspace_size -
5911 WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5912 {
5913 int save_offset = save_hwm - cd->start_workspace;
5914 int this_offset = this_hwm - cd->start_workspace;
5915 *errorcodeptr = expand_workspace(cd);
5916 if (*errorcodeptr != 0) goto FAILED;
5917 save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5918 this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5919 }
5920
5921 for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5922 {
5923 PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
5924 cd->hwm += LINK_SIZE;
5925 }
5926 save_hwm = this_hwm;
5927 code += len;
5928 }
5929
5930 /* Now chain through the pending brackets, and fill in their length
5931 fields (which are holding the chain links pro tem). */
5932
5933 while (bralink != NULL)
5934 {
5935 int oldlinkoffset;
5936 int offset = (int)(code - bralink + 1);
5937 pcre_uchar *bra = code - offset;
5938 oldlinkoffset = GET(bra, 1);
5939 bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
5940 *code++ = OP_KET;
5941 PUTINC(code, 0, offset);
5942 PUT(bra, 1, offset);
5943 }
5944 }
5945
5946 /* If the maximum is unlimited, set a repeater in the final copy. For
5947 ONCE brackets, that's all we need to do. However, possessively repeated
5948 ONCE brackets can be converted into non-capturing brackets, as the
5949 behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
5950 deal with possessive ONCEs specially.
5951
5952 Otherwise, when we are doing the actual compile phase, check to see
5953 whether this group is one that could match an empty string. If so,
5954 convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
5955 that runtime checking can be done. [This check is also applied to ONCE
5956 groups at runtime, but in a different way.]
5957
5958 Then, if the quantifier was possessive and the bracket is not a
5959 conditional, we convert the BRA code to the POS form, and the KET code to
5960 KETRPOS. (It turns out to be convenient at runtime to detect this kind of
5961 subpattern at both the start and at the end.) The use of special opcodes
5962 makes it possible to reduce greatly the stack usage in pcre_exec(). If
5963 the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
5964
5965 Then, if the minimum number of matches is 1 or 0, cancel the possessive
5966 flag so that the default action below, of wrapping everything inside
5967 atomic brackets, does not happen. When the minimum is greater than 1,
5968 there will be earlier copies of the group, and so we still have to wrap
5969 the whole thing. */
5970
5971 else
5972 {
5973 pcre_uchar *ketcode = code - 1 - LINK_SIZE;
5974 pcre_uchar *bracode = ketcode - GET(ketcode, 1);
5975
5976 /* Convert possessive ONCE brackets to non-capturing */
5977
5978 if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
5979 possessive_quantifier) *bracode = OP_BRA;
5980
5981 /* For non-possessive ONCE brackets, all we need to do is to
5982 set the KET. */
5983
5984 if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
5985 *ketcode = OP_KETRMAX + repeat_type;
5986
5987 /* Handle non-ONCE brackets and possessive ONCEs (which have been
5988 converted to non-capturing above). */
5989
5990 else
5991 {
5992 /* In the compile phase, check for empty string matching. */
5993
5994 if (lengthptr == NULL)
5995 {
5996 pcre_uchar *scode = bracode;
5997 do
5998 {
5999 if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6000 {
6001 *bracode += OP_SBRA - OP_BRA;
6002 break;
6003 }
6004 scode += GET(scode, 1);
6005 }
6006 while (*scode == OP_ALT);
6007 }
6008
6009 /* Handle possessive quantifiers. */
6010
6011 if (possessive_quantifier)
6012 {
6013 /* For COND brackets, we wrap the whole thing in a possessively
6014 repeated non-capturing bracket, because we have not invented POS
6015 versions of the COND opcodes. Because we are moving code along, we
6016 must ensure that any pending recursive references are updated. */
6017
6018 if (*bracode == OP_COND || *bracode == OP_SCOND)
6019 {
6020 int nlen = (int)(code - bracode);
6021 *code = OP_END;
6022 adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
6023 memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6024 code += 1 + LINK_SIZE;
6025 nlen += 1 + LINK_SIZE;
6026 *bracode = OP_BRAPOS;
6027 *code++ = OP_KETRPOS;
6028 PUTINC(code, 0, nlen);
6029 PUT(bracode, 1, nlen);
6030 }
6031
6032 /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6033
6034 else
6035 {
6036 *bracode += 1; /* Switch to xxxPOS opcodes */
6037 *ketcode = OP_KETRPOS;
6038 }
6039
6040 /* If the minimum is zero, mark it as possessive, then unset the
6041 possessive flag when the minimum is 0 or 1. */
6042
6043 if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6044 if (repeat_min < 2) possessive_quantifier = FALSE;
6045 }
6046
6047 /* Non-possessive quantifier */
6048
6049 else *ketcode = OP_KETRMAX + repeat_type;
6050 }
6051 }
6052 }
6053
6054 /* If previous is OP_FAIL, it was generated by an empty class [] in
6055 JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6056 by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6057 error above. We can just ignore the repeat in JS case. */
6058
6059 else if (*previous == OP_FAIL) goto END_REPEAT;
6060
6061 /* Else there's some kind of shambles */
6062
6063 else
6064 {
6065 *errorcodeptr = ERR11;
6066 goto FAILED;
6067 }
6068
6069 /* If the character following a repeat is '+', possessive_quantifier is
6070 TRUE. For some opcodes, there are special alternative opcodes for this
6071 case. For anything else, we wrap the entire repeated item inside OP_ONCE
6072 brackets. Logically, the '+' notation is just syntactic sugar, taken from
6073 Sun's Java package, but the special opcodes can optimize it.
6074
6075 Some (but not all) possessively repeated subpatterns have already been
6076 completely handled in the code just above. For them, possessive_quantifier
6077 is always FALSE at this stage. Note that the repeated item starts at
6078 tempcode, not at previous, which might be the first part of a string whose
6079 (former) last char we repeated. */
6080
6081 if (possessive_quantifier)
6082 {
6083 int len;
6084
6085 /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6086 However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6087 {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6088 remains is greater than zero, there's a further opcode that can be
6089 handled. If not, do nothing, leaving the EXACT alone. */
6090
6091 switch(*tempcode)
6092 {
6093 case OP_TYPEEXACT:
6094 tempcode += PRIV(OP_lengths)[*tempcode] +
6095 ((tempcode[1 + IMM2_SIZE] == OP_PROP
6096 || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6097 break;
6098
6099 /* CHAR opcodes are used for exacts whose count is 1. */
6100
6101 case OP_CHAR:
6102 case OP_CHARI:
6103 case OP_NOT:
6104 case OP_NOTI:
6105 case OP_EXACT:
6106 case OP_EXACTI:
6107 case OP_NOTEXACT:
6108 case OP_NOTEXACTI:
6109 tempcode += PRIV(OP_lengths)[*tempcode];
6110 #ifdef SUPPORT_UTF
6111 if (utf && HAS_EXTRALEN(tempcode[-1]))
6112 tempcode += GET_EXTRALEN(tempcode[-1]);
6113 #endif
6114 break;
6115
6116 /* For the class opcodes, the repeat operator appears at the end;
6117 adjust tempcode to point to it. */
6118
6119 case OP_CLASS:
6120 case OP_NCLASS:
6121 tempcode += 1 + 32/sizeof(pcre_uchar);
6122 break;
6123
6124 #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6125 case OP_XCLASS:
6126 tempcode += GET(tempcode, 1);
6127 break;
6128 #endif
6129 }
6130
6131 /* If tempcode is equal to code (which points to the end of the repeated
6132 item), it means we have skipped an EXACT item but there is no following
6133 QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6134 all other cases, tempcode will be pointing to the repeat opcode, and will
6135 be less than code, so the value of len will be greater than 0. */
6136
6137 len = (int)(code - tempcode);
6138 if (len > 0)
6139 {
6140 unsigned int repcode = *tempcode;
6141
6142 /* There is a table for possessifying opcodes, all of which are less
6143 than OP_CALLOUT. A zero entry means there is no possessified version.
6144 */
6145
6146 if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6147 *tempcode = opcode_possessify[repcode];
6148
6149 /* For opcode without a special possessified version, wrap the item in
6150 ONCE brackets. Because we are moving code along, we must ensure that any
6151 pending recursive references are updated. */
6152
6153 else
6154 {
6155 *code = OP_END;
6156 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6157 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6158 code += 1 + LINK_SIZE;
6159 len += 1 + LINK_SIZE;
6160 tempcode[0] = OP_ONCE;
6161 *code++ = OP_KET;
6162 PUTINC(code, 0, len);
6163 PUT(tempcode, 1, len);
6164 }
6165 }
6166
6167 #ifdef NEVER
6168 if (len > 0) switch (*tempcode)
6169 {
6170 case OP_STAR: *tempcode = OP_POSSTAR; break;
6171 case OP_PLUS: *tempcode = OP_POSPLUS; break;
6172 case OP_QUERY: *tempcode = OP_POSQUERY; break;
6173 case OP_UPTO: *tempcode = OP_POSUPTO; break;
6174
6175 case OP_STARI: *tempcode = OP_POSSTARI; break;
6176 case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6177 case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6178 case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6179
6180 case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6181 case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6182 case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6183 case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6184
6185 case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6186 case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6187 case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6188 case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6189
6190 case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6191 case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6192 case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6193 case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6194
6195 case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6196 case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6197 case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6198 case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6199
6200 /* Because we are moving code along, we must ensure that any
6201 pending recursive references are updated. */
6202
6203 default:
6204 *code = OP_END;
6205 adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6206 memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6207 code += 1 + LINK_SIZE;
6208 len += 1 + LINK_SIZE;
6209 tempcode[0] = OP_ONCE;
6210 *code++ = OP_KET;
6211 PUTINC(code, 0, len);
6212 PUT(tempcode, 1, len);
6213 break;
6214 }
6215 #endif
6216 }
6217
6218 /* In all case we no longer have a previous item. We also set the
6219 "follows varying string" flag for subsequently encountered reqchars if
6220 it isn't already set and we have just passed a varying length item. */
6221
6222 END_REPEAT:
6223 previous = NULL;
6224 cd->req_varyopt |= reqvary;
6225 break;
6226
6227
6228 /* ===================================================================*/
6229 /* Start of nested parenthesized sub-expression, or comment or lookahead or
6230 lookbehind or option setting or condition or all the other extended
6231 parenthesis forms. */
6232
6233 case CHAR_LEFT_PARENTHESIS:
6234 newoptions = options;
6235 skipbytes = 0;
6236 bravalue = OP_CBRA;
6237 save_hwm = cd->hwm;
6238 reset_bracount = FALSE;
6239
6240 /* First deal with various "verbs" that can be introduced by '*'. */
6241
6242 ptr++;
6243 if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6244 || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6245 {
6246 int i, namelen;
6247 int arglen = 0;
6248 const char *vn = verbnames;
6249 const pcre_uchar *name = ptr + 1;
6250 const pcre_uchar *arg = NULL;
6251 previous = NULL;
6252 ptr++;
6253 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6254 namelen = (int)(ptr - name);
6255
6256 /* It appears that Perl allows any characters whatsoever, other than
6257 a closing parenthesis, to appear in arguments, so we no longer insist on
6258 letters, digits, and underscores. */
6259
6260 if (*ptr == CHAR_COLON)
6261 {
6262 arg = ++ptr;
6263 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6264 arglen = (int)(ptr - arg);
6265 if ((unsigned int)arglen > MAX_MARK)
6266 {
6267 *errorcodeptr = ERR75;
6268 goto FAILED;
6269 }
6270 }
6271
6272 if (*ptr != CHAR_RIGHT_PARENTHESIS)
6273 {
6274 *errorcodeptr = ERR60;
6275 goto FAILED;
6276 }
6277
6278 /* Scan the table of verb names */
6279
6280 for (i = 0; i < verbcount; i++)
6281 {
6282 if (namelen == verbs[i].len &&
6283 STRNCMP_UC_C8(name, vn, namelen) == 0)
6284 {
6285 int setverb;
6286
6287 /* Check for open captures before ACCEPT and convert it to
6288 ASSERT_ACCEPT if in an assertion. */
6289
6290 if (verbs[i].op == OP_ACCEPT)
6291 {
6292 open_capitem *oc;
6293 if (arglen != 0)
6294 {
6295 *errorcodeptr = ERR59;
6296 goto FAILED;
6297 }
6298 cd->had_accept = TRUE;
6299 for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6300 {
6301 *code++ = OP_CLOSE;
6302 PUT2INC(code, 0, oc->number);
6303 }
6304 setverb = *code++ =
6305 (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6306
6307 /* Do not set firstchar after *ACCEPT */
6308 if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6309 }
6310
6311 /* Handle other cases with/without an argument */
6312
6313 else if (arglen == 0)
6314 {
6315 if (verbs[i].op < 0) /* Argument is mandatory */
6316 {
6317 *errorcodeptr = ERR66;
6318 goto FAILED;
6319 }
6320 setverb = *code++ = verbs[i].op;
6321 }
6322
6323 else
6324 {
6325 if (verbs[i].op_arg < 0) /* Argument is forbidden */
6326 {
6327 *errorcodeptr = ERR59;
6328 goto FAILED;
6329 }
6330 setverb = *code++ = verbs[i].op_arg;
6331 *code++ = arglen;
6332 memcpy(code, arg, IN_UCHARS(arglen));
6333 code += arglen;
6334 *code++ = 0;
6335 }
6336
6337 switch (setverb)
6338 {
6339 case OP_THEN:
6340 case OP_THEN_ARG:
6341 cd->external_flags |= PCRE_HASTHEN;
6342 break;
6343
6344 case OP_PRUNE:
6345 case OP_PRUNE_ARG:
6346 case OP_SKIP:
6347 case OP_SKIP_ARG:
6348 cd->had_pruneorskip = TRUE;
6349 break;
6350 }
6351
6352 break; /* Found verb, exit loop */
6353 }
6354
6355 vn += verbs[i].len + 1;
6356 }
6357
6358 if (i < verbcount) continue; /* Successfully handled a verb */
6359 *errorcodeptr = ERR60; /* Verb not recognized */
6360 goto FAILED;
6361 }
6362
6363 /* Deal with the extended parentheses; all are introduced by '?', and the
6364 appearance of any of them means that this is not a capturing group. */
6365
6366 else if (*ptr == CHAR_QUESTION_MARK)
6367 {
6368 int i, set, unset, namelen;
6369 int *optset;
6370 const pcre_uchar *name;
6371 pcre_uchar *slot;
6372
6373 switch (*(++ptr))
6374 {
6375 case CHAR_NUMBER_SIGN: /* Comment; skip to ket */
6376 ptr++;
6377 while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6378 if (*ptr == CHAR_NULL)
6379 {
6380 *errorcodeptr = ERR18;
6381 goto FAILED;
6382 }
6383 continue;
6384
6385
6386 /* ------------------------------------------------------------ */
6387 case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6388 reset_bracount = TRUE;
6389 /* Fall through */
6390
6391 /* ------------------------------------------------------------ */
6392 case CHAR_COLON: /* Non-capturing bracket */
6393 bravalue = OP_BRA;
6394 ptr++;
6395 break;
6396
6397
6398 /* ------------------------------------------------------------ */
6399 case CHAR_LEFT_PARENTHESIS:
6400 bravalue = OP_COND; /* Conditional group */
6401 tempptr = ptr;
6402
6403 /* A condition can be an assertion, a number (referring to a numbered
6404 group), a name (referring to a named group), or 'R', referring to
6405 recursion. R<digits> and R&name are also permitted for recursion tests.
6406
6407 There are several syntaxes for testing a named group: (?(name)) is used
6408 by Python; Perl 5.10 onwards uses (?(<name>) or (?('name')).
6409
6410 There are two unfortunate ambiguities, caused by history. (a) 'R' can
6411 be the recursive thing or the name 'R' (and similarly for 'R' followed
6412 by digits), and (b) a number could be a name that consists of digits.
6413 In both cases, we look for a name first; if not found, we try the other
6414 cases.
6415
6416 For compatibility with auto-callouts, we allow a callout to be
6417 specified before a condition that is an assertion. First, check for the
6418 syntax of a callout; if found, adjust the temporary pointer that is
6419 used to check for an assertion condition. That's all that is needed! */
6420
6421 if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6422 {
6423 for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6424 if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6425 tempptr += i + 1;
6426 }
6427
6428 /* For conditions that are assertions, check the syntax, and then exit
6429 the switch. This will take control down to where bracketed groups,
6430 including assertions, are processed. */
6431
6432 if (tempptr[1] == CHAR_QUESTION_MARK &&
6433 (tempptr[2] == CHAR_EQUALS_SIGN ||
6434 tempptr[2] == CHAR_EXCLAMATION_MARK ||
6435 tempptr[2] == CHAR_LESS_THAN_SIGN))
6436 break;
6437
6438 /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6439 need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6440
6441 code[1+LINK_SIZE] = OP_CREF;
6442 skipbytes = 1+IMM2_SIZE;
6443 refsign = -1;
6444
6445 /* Check for a test for recursion in a named group. */
6446
6447 if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
6448 {
6449 terminator = -1;
6450 ptr += 2;
6451 code[1+LINK_SIZE] = OP_RREF; /* Change the type of test */
6452 }
6453
6454 /* Check for a test for a named group's having been set, using the Perl
6455 syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6456 syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). As names may
6457 consist entirely of digits, there is scope for ambiguity. */
6458
6459 else if (ptr[1] == CHAR_LESS_THAN_SIGN)
6460 {
6461 terminator = CHAR_GREATER_THAN_SIGN;
6462 ptr++;
6463 }
6464 else if (ptr[1] == CHAR_APOSTROPHE)
6465 {
6466 terminator = CHAR_APOSTROPHE;
6467 ptr++;
6468 }
6469 else
6470 {
6471 terminator = CHAR_NULL;
6472 if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
6473 }
6474
6475 /* When a name is one of a number of duplicates, a different opcode is
6476 used and it needs more memory. Unfortunately we cannot tell whether a
6477 name is a duplicate in the first pass, so we have to allow for more
6478 memory except when we know it is a relative numerical reference. */
6479
6480 if (refsign < 0 && lengthptr != NULL) *lengthptr += IMM2_SIZE;
6481
6482 /* We now expect to read a name (possibly all digits); any thing else
6483 is an error. In the case of all digits, also get it as a number. */
6484
6485 if (!MAX_255(ptr[1]) || (cd->ctypes[ptr[1]] & ctype_word) == 0)
6486 {
6487 ptr += 1; /* To get the right offset */
6488 *errorcodeptr = ERR28;
6489 goto FAILED;
6490 }
6491
6492 recno = 0;
6493 name = ++ptr;
6494 while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6495 {
6496 if (recno >= 0)
6497 recno = (IS_DIGIT(*ptr))? recno * 10 + (int)(*ptr - CHAR_0) : -1;
6498 ptr++;
6499 }
6500 namelen = (int)(ptr - name);
6501
6502 /* Check the terminator */
6503
6504 if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6505 *ptr++ != CHAR_RIGHT_PARENTHESIS)
6506 {
6507 ptr--; /* Error offset */
6508 *errorcodeptr = ERR26;
6509 goto FAILED;
6510 }
6511
6512 /* Do no further checking in the pre-compile phase. */
6513
6514 if (lengthptr != NULL) break;
6515
6516 /* In the real compile we do the work of looking for the actual
6517 reference. If the string started with "+" or "-" we require the rest to
6518 be digits, in which case recno will be set. */
6519
6520 if (refsign > 0)
6521 {
6522 if (recno <= 0)
6523 {
6524 *errorcodeptr = ERR58;
6525 goto FAILED;
6526 }
6527 recno = (refsign == CHAR_MINUS)?
6528 cd->bracount - recno + 1 : recno +cd->bracount;
6529 if (recno <= 0 || recno > cd->final_bracount)
6530 {
6531 *errorcodeptr = ERR15;
6532 goto FAILED;
6533 }
6534 PUT2(code, 2+LINK_SIZE, recno);
6535 break;
6536 }
6537
6538 /* Otherwise (did not start with "+" or "-"), start by looking for the
6539 name. */
6540
6541 slot = cd->name_table;
6542 for (i = 0; i < cd->names_found; i++)
6543 {
6544 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
6545 slot += cd->name_entry_size;
6546 }
6547
6548 /* Found the named subpattern. If the name is duplicated, add one to
6549 the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6550 appropriate data values. Otherwise, just insert the unique subpattern
6551 number. */
6552
6553 if (i < cd->names_found)
6554 {
6555 int offset = i++;
6556 int count = 1;
6557 recno = GET2(slot, 0); /* Number from first found */
6558 for (; i < cd->names_found; i++)
6559 {
6560 slot += cd->name_entry_size;
6561 if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0) break;
6562 count++;
6563 }
6564 if (count > 1)
6565 {
6566 PUT2(code, 2+LINK_SIZE, offset);
6567 PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
6568 skipbytes += IMM2_SIZE;
6569 code[1+LINK_SIZE]++;
6570 }
6571 else /* Not a duplicated name */
6572 {
6573 PUT2(code, 2+LINK_SIZE, recno);
6574 }
6575 }
6576
6577 /* If terminator == CHAR_NULL it means that the name followed directly
6578 after the opening parenthesis [e.g. (?(abc)...] and in this case there
6579 are some further alternatives to try. For the cases where terminator !=
6580 0 [things like (?(<name>... or (?('name')... or (?(R&name)... ] we have
6581 now checked all the possibilities, so give an error. */
6582
6583 else if (terminator != CHAR_NULL)
6584 {
6585 *errorcodeptr = ERR15;
6586 goto FAILED;
6587 }
6588
6589 /* Check for (?(R) for recursion. Allow digits after R to specify a
6590 specific group number. */
6591
6592 else if (*name == CHAR_R)
6593 {
6594 recno = 0;
6595 for (i = 1; i < namelen; i++)
6596 {
6597 if (!IS_DIGIT(name[i]))
6598 {
6599 *errorcodeptr = ERR15;
6600 goto FAILED;
6601 }
6602 recno = recno * 10 + name[i] - CHAR_0;
6603 }
6604 if (recno == 0) recno = RREF_ANY;
6605 code[1+LINK_SIZE] = OP_RREF; /* Change test type */
6606 PUT2(code, 2+LINK_SIZE, recno);
6607 }
6608
6609 /* Similarly, check for the (?(DEFINE) "condition", which is always
6610 false. */
6611
6612 else if (namelen == 6 && STRNCMP_UC_C8(name, STRING_DEFINE, 6) == 0)
6613 {
6614 code[1+LINK_SIZE] = OP_DEF;
6615 skipbytes = 1;
6616 }
6617
6618 /* Check for the "name" actually being a subpattern number. We are
6619 in the second pass here, so final_bracount is set. */
6620
6621 else if (recno > 0 && recno <= cd->final_bracount)
6622 {
6623 PUT2(code, 2+LINK_SIZE, recno);
6624 }
6625
6626 /* Either an unidentified subpattern, or a reference to (?(0) */
6627
6628 else
6629 {
6630 *errorcodeptr = (recno == 0)? ERR35: ERR15;
6631 goto FAILED;
6632 }
6633 break;
6634
6635
6636 /* ------------------------------------------------------------ */
6637 case CHAR_EQUALS_SIGN: /* Positive lookahead */
6638 bravalue = OP_ASSERT;
6639 cd->assert_depth += 1;
6640 ptr++;
6641 break;
6642
6643
6644 /* ------------------------------------------------------------ */
6645 case CHAR_EXCLAMATION_MARK: /* Negative lookahead */
6646 ptr++;
6647 if (*ptr == CHAR_RIGHT_PARENTHESIS) /* Optimize (?!) */
6648 {
6649 *code++ = OP_FAIL;
6650 previous = NULL;
6651 continue;
6652 }
6653 bravalue = OP_ASSERT_NOT;
6654 cd->assert_depth += 1;
6655 break;
6656
6657
6658 /* ------------------------------------------------------------ */
6659 case CHAR_LESS_THAN_SIGN: /* Lookbehind or named define */
6660 switch (ptr[1])
6661 {
6662 case CHAR_EQUALS_SIGN: /* Positive lookbehind */
6663 bravalue = OP_ASSERTBACK;
6664 cd->assert_depth += 1;
66