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: /* Substitutes for [[:<:]] and [[:>:]], which mean start and end of word in
264: another regex library. */
265:
266: static const pcre_uchar sub_start_of_word[] = {
267: CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
268: CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w, CHAR_RIGHT_PARENTHESIS, '\0' };
269:
270: static const pcre_uchar sub_end_of_word[] = {
271: CHAR_BACKSLASH, CHAR_b, CHAR_LEFT_PARENTHESIS, CHAR_QUESTION_MARK,
272: CHAR_LESS_THAN_SIGN, CHAR_EQUALS_SIGN, CHAR_BACKSLASH, CHAR_w,
273: CHAR_RIGHT_PARENTHESIS, '\0' };
274:
275:
276: /* Tables of names of POSIX character classes and their lengths. The names are
277: now all in a single string, to reduce the number of relocations when a shared
278: library is dynamically loaded. The list of lengths is terminated by a zero
279: length entry. The first three must be alpha, lower, upper, as this is assumed
280: for handling case independence. The indices for graph, print, and punct are
281: needed, so identify them. */
282:
283: static const char posix_names[] =
284: STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
285: STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
286: STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
287: STRING_word0 STRING_xdigit;
288:
289: static const pcre_uint8 posix_name_lengths[] = {
290: 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
291:
292: #define PC_GRAPH 8
293: #define PC_PRINT 9
294: #define PC_PUNCT 10
295:
296:
297: /* Table of class bit maps for each POSIX class. Each class is formed from a
298: base map, with an optional addition or removal of another map. Then, for some
299: classes, there is some additional tweaking: for [:blank:] the vertical space
300: characters are removed, and for [:alpha:] and [:alnum:] the underscore
301: character is removed. The triples in the table consist of the base map offset,
302: second map offset or -1 if no second map, and a non-negative value for map
303: addition or a negative value for map subtraction (if there are two maps). The
304: absolute value of the third field has these meanings: 0 => no tweaking, 1 =>
305: remove vertical space characters, 2 => remove underscore. */
306:
307: static const int posix_class_maps[] = {
308: cbit_word, cbit_digit, -2, /* alpha */
309: cbit_lower, -1, 0, /* lower */
310: cbit_upper, -1, 0, /* upper */
311: cbit_word, -1, 2, /* alnum - word without underscore */
312: cbit_print, cbit_cntrl, 0, /* ascii */
313: cbit_space, -1, 1, /* blank - a GNU extension */
314: cbit_cntrl, -1, 0, /* cntrl */
315: cbit_digit, -1, 0, /* digit */
316: cbit_graph, -1, 0, /* graph */
317: cbit_print, -1, 0, /* print */
318: cbit_punct, -1, 0, /* punct */
319: cbit_space, -1, 0, /* space */
320: cbit_word, -1, 0, /* word - a Perl extension */
321: cbit_xdigit,-1, 0 /* xdigit */
322: };
323:
324: /* Table of substitutes for \d etc when PCRE_UCP is set. They are replaced by
325: Unicode property escapes. */
326:
327: #ifdef SUPPORT_UCP
328: static const pcre_uchar string_PNd[] = {
329: CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
330: CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
331: static const pcre_uchar string_pNd[] = {
332: CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
333: CHAR_N, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
334: static const pcre_uchar string_PXsp[] = {
335: CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
336: CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
337: static const pcre_uchar string_pXsp[] = {
338: CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
339: CHAR_X, CHAR_s, CHAR_p, CHAR_RIGHT_CURLY_BRACKET, '\0' };
340: static const pcre_uchar string_PXwd[] = {
341: CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
342: CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
343: static const pcre_uchar string_pXwd[] = {
344: CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
345: CHAR_X, CHAR_w, CHAR_d, CHAR_RIGHT_CURLY_BRACKET, '\0' };
346:
347: static const pcre_uchar *substitutes[] = {
348: string_PNd, /* \D */
349: string_pNd, /* \d */
350: string_PXsp, /* \S */ /* Xsp is Perl space, but from 8.34, Perl */
351: string_pXsp, /* \s */ /* space and POSIX space are the same. */
352: string_PXwd, /* \W */
353: string_pXwd /* \w */
354: };
355:
356: /* The POSIX class substitutes must be in the order of the POSIX class names,
357: defined above, and there are both positive and negative cases. NULL means no
358: general substitute of a Unicode property escape (\p or \P). However, for some
359: POSIX classes (e.g. graph, print, punct) a special property code is compiled
360: directly. */
361:
362: static const pcre_uchar string_pL[] = {
363: CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
364: CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
365: static const pcre_uchar string_pLl[] = {
366: CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
367: CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
368: static const pcre_uchar string_pLu[] = {
369: CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
370: CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
371: static const pcre_uchar string_pXan[] = {
372: CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
373: CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
374: static const pcre_uchar string_h[] = {
375: CHAR_BACKSLASH, CHAR_h, '\0' };
376: static const pcre_uchar string_pXps[] = {
377: CHAR_BACKSLASH, CHAR_p, CHAR_LEFT_CURLY_BRACKET,
378: CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
379: static const pcre_uchar string_PL[] = {
380: CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
381: CHAR_L, CHAR_RIGHT_CURLY_BRACKET, '\0' };
382: static const pcre_uchar string_PLl[] = {
383: CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
384: CHAR_L, CHAR_l, CHAR_RIGHT_CURLY_BRACKET, '\0' };
385: static const pcre_uchar string_PLu[] = {
386: CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
387: CHAR_L, CHAR_u, CHAR_RIGHT_CURLY_BRACKET, '\0' };
388: static const pcre_uchar string_PXan[] = {
389: CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
390: CHAR_X, CHAR_a, CHAR_n, CHAR_RIGHT_CURLY_BRACKET, '\0' };
391: static const pcre_uchar string_H[] = {
392: CHAR_BACKSLASH, CHAR_H, '\0' };
393: static const pcre_uchar string_PXps[] = {
394: CHAR_BACKSLASH, CHAR_P, CHAR_LEFT_CURLY_BRACKET,
395: CHAR_X, CHAR_p, CHAR_s, CHAR_RIGHT_CURLY_BRACKET, '\0' };
396:
397: static const pcre_uchar *posix_substitutes[] = {
398: string_pL, /* alpha */
399: string_pLl, /* lower */
400: string_pLu, /* upper */
401: string_pXan, /* alnum */
402: NULL, /* ascii */
403: string_h, /* blank */
404: NULL, /* cntrl */
405: string_pNd, /* digit */
406: NULL, /* graph */
407: NULL, /* print */
408: NULL, /* punct */
409: string_pXps, /* space */ /* Xps is POSIX space, but from 8.34 */
410: string_pXwd, /* word */ /* Perl and POSIX space are the same */
411: NULL, /* xdigit */
412: /* Negated cases */
413: string_PL, /* ^alpha */
414: string_PLl, /* ^lower */
415: string_PLu, /* ^upper */
416: string_PXan, /* ^alnum */
417: NULL, /* ^ascii */
418: string_H, /* ^blank */
419: NULL, /* ^cntrl */
420: string_PNd, /* ^digit */
421: NULL, /* ^graph */
422: NULL, /* ^print */
423: NULL, /* ^punct */
424: string_PXps, /* ^space */ /* Xps is POSIX space, but from 8.34 */
425: string_PXwd, /* ^word */ /* Perl and POSIX space are the same */
426: NULL /* ^xdigit */
427: };
428: #define POSIX_SUBSIZE (sizeof(posix_substitutes) / sizeof(pcre_uchar *))
429: #endif
430:
431: #define STRING(a) # a
432: #define XSTRING(s) STRING(s)
433:
434: /* The texts of compile-time error messages. These are "char *" because they
435: are passed to the outside world. Do not ever re-use any error number, because
436: they are documented. Always add a new error instead. Messages marked DEAD below
437: are no longer used. This used to be a table of strings, but in order to reduce
438: the number of relocations needed when a shared library is loaded dynamically,
439: it is now one long string. We cannot use a table of offsets, because the
440: lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
441: simply count through to the one we want - this isn't a performance issue
442: because these strings are used only when there is a compilation error.
443:
444: Each substring ends with \0 to insert a null character. This includes the final
445: substring, so that the whole string ends with \0\0, which can be detected when
446: counting through. */
447:
448: static const char error_texts[] =
449: "no error\0"
450: "\\ at end of pattern\0"
451: "\\c at end of pattern\0"
452: "unrecognized character follows \\\0"
453: "numbers out of order in {} quantifier\0"
454: /* 5 */
455: "number too big in {} quantifier\0"
456: "missing terminating ] for character class\0"
457: "invalid escape sequence in character class\0"
458: "range out of order in character class\0"
459: "nothing to repeat\0"
460: /* 10 */
461: "operand of unlimited repeat could match the empty string\0" /** DEAD **/
462: "internal error: unexpected repeat\0"
463: "unrecognized character after (? or (?-\0"
464: "POSIX named classes are supported only within a class\0"
465: "missing )\0"
466: /* 15 */
467: "reference to non-existent subpattern\0"
468: "erroffset passed as NULL\0"
469: "unknown option bit(s) set\0"
470: "missing ) after comment\0"
471: "parentheses nested too deeply\0" /** DEAD **/
472: /* 20 */
473: "regular expression is too large\0"
474: "failed to get memory\0"
475: "unmatched parentheses\0"
476: "internal error: code overflow\0"
477: "unrecognized character after (?<\0"
478: /* 25 */
479: "lookbehind assertion is not fixed length\0"
480: "malformed number or name after (?(\0"
481: "conditional group contains more than two branches\0"
482: "assertion expected after (?(\0"
483: "(?R or (?[+-]digits must be followed by )\0"
484: /* 30 */
485: "unknown POSIX class name\0"
486: "POSIX collating elements are not supported\0"
487: "this version of PCRE is compiled without UTF support\0"
488: "spare error\0" /** DEAD **/
489: "character value in \\x{} or \\o{} is too large\0"
490: /* 35 */
491: "invalid condition (?(0)\0"
492: "\\C not allowed in lookbehind assertion\0"
493: "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
494: "number after (?C is > 255\0"
495: "closing ) for (?C expected\0"
496: /* 40 */
497: "recursive call could loop indefinitely\0"
498: "unrecognized character after (?P\0"
499: "syntax error in subpattern name (missing terminator)\0"
500: "two named subpatterns have the same name\0"
501: "invalid UTF-8 string\0"
502: /* 45 */
503: "support for \\P, \\p, and \\X has not been compiled\0"
504: "malformed \\P or \\p sequence\0"
505: "unknown property name after \\P or \\p\0"
506: "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
507: "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
508: /* 50 */
509: "repeated subpattern is too long\0" /** DEAD **/
510: "octal value is greater than \\377 in 8-bit non-UTF-8 mode\0"
511: "internal error: overran compiling workspace\0"
512: "internal error: previously-checked referenced subpattern not found\0"
513: "DEFINE group contains more than one branch\0"
514: /* 55 */
515: "repeating a DEFINE group is not allowed\0" /** DEAD **/
516: "inconsistent NEWLINE options\0"
517: "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
518: "a numbered reference must not be zero\0"
519: "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
520: /* 60 */
521: "(*VERB) not recognized or malformed\0"
522: "number is too big\0"
523: "subpattern name expected\0"
524: "digit expected after (?+\0"
525: "] is an invalid data character in JavaScript compatibility mode\0"
526: /* 65 */
527: "different names for subpatterns of the same number are not allowed\0"
528: "(*MARK) must have an argument\0"
529: "this version of PCRE is not compiled with Unicode property support\0"
530: "\\c must be followed by an ASCII character\0"
531: "\\k is not followed by a braced, angle-bracketed, or quoted name\0"
532: /* 70 */
533: "internal error: unknown opcode in find_fixedlength()\0"
534: "\\N is not supported in a class\0"
535: "too many forward references\0"
536: "disallowed Unicode code point (>= 0xd800 && <= 0xdfff)\0"
537: "invalid UTF-16 string\0"
538: /* 75 */
539: "name is too long in (*MARK), (*PRUNE), (*SKIP), or (*THEN)\0"
540: "character value in \\u.... sequence is too large\0"
541: "invalid UTF-32 string\0"
542: "setting UTF is disabled by the application\0"
543: "non-hex character in \\x{} (closing brace missing?)\0"
544: /* 80 */
545: "non-octal character in \\o{} (closing brace missing?)\0"
546: "missing opening brace after \\o\0"
547: "parentheses are too deeply nested\0"
548: "invalid range in character class\0"
549: "group name must start with a non-digit\0"
550: ;
551:
552: /* Table to identify digits and hex digits. This is used when compiling
553: patterns. Note that the tables in chartables are dependent on the locale, and
554: may mark arbitrary characters as digits - but the PCRE compiling code expects
555: to handle only 0-9, a-z, and A-Z as digits when compiling. That is why we have
556: a private table here. It costs 256 bytes, but it is a lot faster than doing
557: character value tests (at least in some simple cases I timed), and in some
558: applications one wants PCRE to compile efficiently as well as match
559: efficiently.
560:
561: For convenience, we use the same bit definitions as in chartables:
562:
563: 0x04 decimal digit
564: 0x08 hexadecimal digit
565:
566: Then we can use ctype_digit and ctype_xdigit in the code. */
567:
568: /* Using a simple comparison for decimal numbers rather than a memory read
569: is much faster, and the resulting code is simpler (the compiler turns it
570: into a subtraction and unsigned comparison). */
571:
572: #define IS_DIGIT(x) ((x) >= CHAR_0 && (x) <= CHAR_9)
573:
574: #ifndef EBCDIC
575:
576: /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
577: UTF-8 mode. */
578:
579: static const pcre_uint8 digitab[] =
580: {
581: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 */
582: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
583: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 */
584: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
585: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - ' */
586: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ( - / */
587: 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 */
588: 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00, /* 8 - ? */
589: 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* @ - G */
590: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H - O */
591: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* P - W */
592: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* X - _ */
593: 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* ` - g */
594: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h - o */
595: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* p - w */
596: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* x -127 */
597: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 128-135 */
598: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 136-143 */
599: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144-151 */
600: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 152-159 */
601: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160-167 */
602: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 168-175 */
603: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 176-183 */
604: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
605: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 192-199 */
606: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 200-207 */
607: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 208-215 */
608: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 216-223 */
609: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 224-231 */
610: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 232-239 */
611: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
612: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
613:
614: #else
615:
616: /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
617:
618: static const pcre_uint8 digitab[] =
619: {
620: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 0- 7 0 */
621: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 8- 15 */
622: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 16- 23 10 */
623: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
624: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 32- 39 20 */
625: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
626: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 30 */
627: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
628: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 40 */
629: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 72- | */
630: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 50 */
631: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 88- 95 */
632: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 60 */
633: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 104- ? */
634: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 70 */
635: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
636: 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* 128- g 80 */
637: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
638: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 144- p 90 */
639: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
640: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 160- x A0 */
641: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
642: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 B0 */
643: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
644: 0x00,0x08,0x08,0x08,0x08,0x08,0x08,0x00, /* { - G C0 */
645: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
646: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* } - P D0 */
647: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
648: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* \ - X E0 */
649: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
650: 0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c,0x0c, /* 0 - 7 F0 */
651: 0x0c,0x0c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
652:
653: static const pcre_uint8 ebcdic_chartab[] = { /* chartable partial dup */
654: 0x80,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 0- 7 */
655: 0x00,0x00,0x00,0x00,0x01,0x01,0x00,0x00, /* 8- 15 */
656: 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 16- 23 */
657: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 24- 31 */
658: 0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, /* 32- 39 */
659: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 40- 47 */
660: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 48- 55 */
661: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 56- 63 */
662: 0x01,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - 71 */
663: 0x00,0x00,0x00,0x80,0x00,0x80,0x80,0x80, /* 72- | */
664: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* & - 87 */
665: 0x00,0x00,0x00,0x80,0x80,0x80,0x00,0x00, /* 88- 95 */
666: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* - -103 */
667: 0x00,0x00,0x00,0x00,0x00,0x10,0x00,0x80, /* 104- ? */
668: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 112-119 */
669: 0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 120- " */
670: 0x00,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* 128- g */
671: 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* h -143 */
672: 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* 144- p */
673: 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* q -159 */
674: 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* 160- x */
675: 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* y -175 */
676: 0x80,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* ^ -183 */
677: 0x00,0x00,0x80,0x00,0x00,0x00,0x00,0x00, /* 184-191 */
678: 0x80,0x1a,0x1a,0x1a,0x1a,0x1a,0x1a,0x12, /* { - G */
679: 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* H -207 */
680: 0x00,0x12,0x12,0x12,0x12,0x12,0x12,0x12, /* } - P */
681: 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Q -223 */
682: 0x00,0x00,0x12,0x12,0x12,0x12,0x12,0x12, /* \ - X */
683: 0x12,0x12,0x00,0x00,0x00,0x00,0x00,0x00, /* Y -239 */
684: 0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c,0x1c, /* 0 - 7 */
685: 0x1c,0x1c,0x00,0x00,0x00,0x00,0x00,0x00};/* 8 -255 */
686: #endif
687:
688:
689: /* This table is used to check whether auto-possessification is possible
690: between adjacent character-type opcodes. The left-hand (repeated) opcode is
691: used to select the row, and the right-hand opcode is use to select the column.
692: A value of 1 means that auto-possessification is OK. For example, the second
693: value in the first row means that \D+\d can be turned into \D++\d.
694:
695: The Unicode property types (\P and \p) have to be present to fill out the table
696: because of what their opcode values are, but the table values should always be
697: zero because property types are handled separately in the code. The last four
698: columns apply to items that cannot be repeated, so there is no need to have
699: rows for them. Note that OP_DIGIT etc. are generated only when PCRE_UCP is
700: *not* set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
701:
702: #define APTROWS (LAST_AUTOTAB_LEFT_OP - FIRST_AUTOTAB_OP + 1)
703: #define APTCOLS (LAST_AUTOTAB_RIGHT_OP - FIRST_AUTOTAB_OP + 1)
704:
705: static const pcre_uint8 autoposstab[APTROWS][APTCOLS] = {
706: /* \D \d \S \s \W \w . .+ \C \P \p \R \H \h \V \v \X \Z \z $ $M */
707: { 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \D */
708: { 1, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \d */
709: { 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \S */
710: { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \s */
711: { 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \W */
712: { 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 1, 1, 1 }, /* \w */
713: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* . */
714: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* .+ */
715: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 }, /* \C */
716: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \P */
717: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* \p */
718: { 0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \R */
719: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }, /* \H */
720: { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \h */
721: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 }, /* \V */
722: { 0, 1, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0 }, /* \v */
723: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 } /* \X */
724: };
725:
726:
727: /* This table is used to check whether auto-possessification is possible
728: between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP). The
729: left-hand (repeated) opcode is used to select the row, and the right-hand
730: opcode is used to select the column. The values are as follows:
731:
732: 0 Always return FALSE (never auto-possessify)
733: 1 Character groups are distinct (possessify if both are OP_PROP)
734: 2 Check character categories in the same group (general or particular)
735: 3 TRUE if the two opcodes are not the same (PROP vs NOTPROP)
736:
737: 4 Check left general category vs right particular category
738: 5 Check right general category vs left particular category
739:
740: 6 Left alphanum vs right general category
741: 7 Left space vs right general category
742: 8 Left word vs right general category
743:
744: 9 Right alphanum vs left general category
745: 10 Right space vs left general category
746: 11 Right word vs left general category
747:
748: 12 Left alphanum vs right particular category
749: 13 Left space vs right particular category
750: 14 Left word vs right particular category
751:
752: 15 Right alphanum vs left particular category
753: 16 Right space vs left particular category
754: 17 Right word vs left particular category
755: */
756:
757: static const pcre_uint8 propposstab[PT_TABSIZE][PT_TABSIZE] = {
758: /* ANY LAMP GC PC SC ALNUM SPACE PXSPACE WORD CLIST UCNC */
759: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_ANY */
760: { 0, 3, 0, 0, 0, 3, 1, 1, 0, 0, 0 }, /* PT_LAMP */
761: { 0, 0, 2, 4, 0, 9, 10, 10, 11, 0, 0 }, /* PT_GC */
762: { 0, 0, 5, 2, 0, 15, 16, 16, 17, 0, 0 }, /* PT_PC */
763: { 0, 0, 0, 0, 2, 0, 0, 0, 0, 0, 0 }, /* PT_SC */
764: { 0, 3, 6, 12, 0, 3, 1, 1, 0, 0, 0 }, /* PT_ALNUM */
765: { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_SPACE */
766: { 0, 1, 7, 13, 0, 1, 3, 3, 1, 0, 0 }, /* PT_PXSPACE */
767: { 0, 0, 8, 14, 0, 0, 1, 1, 3, 0, 0 }, /* PT_WORD */
768: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 }, /* PT_CLIST */
769: { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3 } /* PT_UCNC */
770: };
771:
772: /* This table is used to check whether auto-possessification is possible
773: between adjacent Unicode property opcodes (OP_PROP and OP_NOTPROP) when one
774: specifies a general category and the other specifies a particular category. The
775: row is selected by the general category and the column by the particular
776: category. The value is 1 if the particular category is not part of the general
777: category. */
778:
779: static const pcre_uint8 catposstab[7][30] = {
780: /* 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 */
781: { 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 */
782: { 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 */
783: { 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 */
784: { 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 */
785: { 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 */
786: { 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 */
787: { 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 */
788: };
789:
790: /* This table is used when checking ALNUM, (PX)SPACE, SPACE, and WORD against
791: a general or particular category. The properties in each row are those
792: that apply to the character set in question. Duplication means that a little
793: unnecessary work is done when checking, but this keeps things much simpler
794: because they can all use the same code. For more details see the comment where
795: this table is used.
796:
797: Note: SPACE and PXSPACE used to be different because Perl excluded VT from
798: "space", but from Perl 5.18 it's included, so both categories are treated the
799: same here. */
800:
801: static const pcre_uint8 posspropstab[3][4] = {
802: { ucp_L, ucp_N, ucp_N, ucp_Nl }, /* ALNUM, 3rd and 4th values redundant */
803: { ucp_Z, ucp_Z, ucp_C, ucp_Cc }, /* SPACE and PXSPACE, 2nd value redundant */
804: { ucp_L, ucp_N, ucp_P, ucp_Po } /* WORD */
805: };
806:
807: /* This table is used when converting repeating opcodes into possessified
808: versions as a result of an explicit possessive quantifier such as ++. A zero
809: value means there is no possessified version - in those cases the item in
810: question must be wrapped in ONCE brackets. The table is truncated at OP_CALLOUT
811: because all relevant opcodes are less than that. */
812:
813: static const pcre_uint8 opcode_possessify[] = {
814: 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 0 - 15 */
815: 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, /* 16 - 31 */
816:
817: 0, /* NOTI */
818: OP_POSSTAR, 0, /* STAR, MINSTAR */
819: OP_POSPLUS, 0, /* PLUS, MINPLUS */
820: OP_POSQUERY, 0, /* QUERY, MINQUERY */
821: OP_POSUPTO, 0, /* UPTO, MINUPTO */
822: 0, /* EXACT */
823: 0, 0, 0, 0, /* POS{STAR,PLUS,QUERY,UPTO} */
824:
825: OP_POSSTARI, 0, /* STARI, MINSTARI */
826: OP_POSPLUSI, 0, /* PLUSI, MINPLUSI */
827: OP_POSQUERYI, 0, /* QUERYI, MINQUERYI */
828: OP_POSUPTOI, 0, /* UPTOI, MINUPTOI */
829: 0, /* EXACTI */
830: 0, 0, 0, 0, /* POS{STARI,PLUSI,QUERYI,UPTOI} */
831:
832: OP_NOTPOSSTAR, 0, /* NOTSTAR, NOTMINSTAR */
833: OP_NOTPOSPLUS, 0, /* NOTPLUS, NOTMINPLUS */
834: OP_NOTPOSQUERY, 0, /* NOTQUERY, NOTMINQUERY */
835: OP_NOTPOSUPTO, 0, /* NOTUPTO, NOTMINUPTO */
836: 0, /* NOTEXACT */
837: 0, 0, 0, 0, /* NOTPOS{STAR,PLUS,QUERY,UPTO} */
838:
839: OP_NOTPOSSTARI, 0, /* NOTSTARI, NOTMINSTARI */
840: OP_NOTPOSPLUSI, 0, /* NOTPLUSI, NOTMINPLUSI */
841: OP_NOTPOSQUERYI, 0, /* NOTQUERYI, NOTMINQUERYI */
842: OP_NOTPOSUPTOI, 0, /* NOTUPTOI, NOTMINUPTOI */
843: 0, /* NOTEXACTI */
844: 0, 0, 0, 0, /* NOTPOS{STARI,PLUSI,QUERYI,UPTOI} */
845:
846: OP_TYPEPOSSTAR, 0, /* TYPESTAR, TYPEMINSTAR */
847: OP_TYPEPOSPLUS, 0, /* TYPEPLUS, TYPEMINPLUS */
848: OP_TYPEPOSQUERY, 0, /* TYPEQUERY, TYPEMINQUERY */
849: OP_TYPEPOSUPTO, 0, /* TYPEUPTO, TYPEMINUPTO */
850: 0, /* TYPEEXACT */
851: 0, 0, 0, 0, /* TYPEPOS{STAR,PLUS,QUERY,UPTO} */
852:
853: OP_CRPOSSTAR, 0, /* CRSTAR, CRMINSTAR */
854: OP_CRPOSPLUS, 0, /* CRPLUS, CRMINPLUS */
855: OP_CRPOSQUERY, 0, /* CRQUERY, CRMINQUERY */
856: OP_CRPOSRANGE, 0, /* CRRANGE, CRMINRANGE */
857: 0, 0, 0, 0, /* CRPOS{STAR,PLUS,QUERY,RANGE} */
858:
859: 0, 0, 0, /* CLASS, NCLASS, XCLASS */
860: 0, 0, /* REF, REFI */
861: 0, 0, /* DNREF, DNREFI */
862: 0, 0 /* RECURSE, CALLOUT */
863: };
864:
865:
866:
867: /*************************************************
868: * Find an error text *
869: *************************************************/
870:
871: /* The error texts are now all in one long string, to save on relocations. As
872: some of the text is of unknown length, we can't use a table of offsets.
873: Instead, just count through the strings. This is not a performance issue
874: because it happens only when there has been a compilation error.
875:
876: Argument: the error number
877: Returns: pointer to the error string
878: */
879:
880: static const char *
881: find_error_text(int n)
882: {
883: const char *s = error_texts;
884: for (; n > 0; n--)
885: {
886: while (*s++ != CHAR_NULL) {};
887: if (*s == CHAR_NULL) return "Error text not found (please report)";
888: }
889: return s;
890: }
891:
892:
893:
894: /*************************************************
895: * Expand the workspace *
896: *************************************************/
897:
898: /* This function is called during the second compiling phase, if the number of
899: forward references fills the existing workspace, which is originally a block on
900: the stack. A larger block is obtained from malloc() unless the ultimate limit
901: has been reached or the increase will be rather small.
902:
903: Argument: pointer to the compile data block
904: Returns: 0 if all went well, else an error number
905: */
906:
907: static int
908: expand_workspace(compile_data *cd)
909: {
910: pcre_uchar *newspace;
911: int newsize = cd->workspace_size * 2;
912:
913: if (newsize > COMPILE_WORK_SIZE_MAX) newsize = COMPILE_WORK_SIZE_MAX;
914: if (cd->workspace_size >= COMPILE_WORK_SIZE_MAX ||
915: newsize - cd->workspace_size < WORK_SIZE_SAFETY_MARGIN)
916: return ERR72;
917:
918: newspace = (PUBL(malloc))(IN_UCHARS(newsize));
919: if (newspace == NULL) return ERR21;
920: memcpy(newspace, cd->start_workspace, cd->workspace_size * sizeof(pcre_uchar));
921: cd->hwm = (pcre_uchar *)newspace + (cd->hwm - cd->start_workspace);
922: if (cd->workspace_size > COMPILE_WORK_SIZE)
923: (PUBL(free))((void *)cd->start_workspace);
924: cd->start_workspace = newspace;
925: cd->workspace_size = newsize;
926: return 0;
927: }
928:
929:
930:
931: /*************************************************
932: * Check for counted repeat *
933: *************************************************/
934:
935: /* This function is called when a '{' is encountered in a place where it might
936: start a quantifier. It looks ahead to see if it really is a quantifier or not.
937: It is only a quantifier if it is one of the forms {ddd} {ddd,} or {ddd,ddd}
938: where the ddds are digits.
939:
940: Arguments:
941: p pointer to the first char after '{'
942:
943: Returns: TRUE or FALSE
944: */
945:
946: static BOOL
947: is_counted_repeat(const pcre_uchar *p)
948: {
949: if (!IS_DIGIT(*p)) return FALSE;
950: p++;
951: while (IS_DIGIT(*p)) p++;
952: if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
953:
954: if (*p++ != CHAR_COMMA) return FALSE;
955: if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
956:
957: if (!IS_DIGIT(*p)) return FALSE;
958: p++;
959: while (IS_DIGIT(*p)) p++;
960:
961: return (*p == CHAR_RIGHT_CURLY_BRACKET);
962: }
963:
964:
965:
966: /*************************************************
967: * Handle escapes *
968: *************************************************/
969:
970: /* This function is called when a \ has been encountered. It either returns a
971: positive value for a simple escape such as \n, or 0 for a data character which
972: will be placed in chptr. A backreference to group n is returned as negative n.
973: When UTF-8 is enabled, a positive value greater than 255 may be returned in
974: chptr. On entry, ptr is pointing at the \. On exit, it is on the final
975: character of the escape sequence.
976:
977: Arguments:
978: ptrptr points to the pattern position pointer
979: chptr points to a returned data character
980: errorcodeptr points to the errorcode variable
981: bracount number of previous extracting brackets
982: options the options bits
983: isclass TRUE if inside a character class
984:
985: Returns: zero => a data character
986: positive => a special escape sequence
987: negative => a back reference
988: on error, errorcodeptr is set
989: */
990:
991: static int
992: check_escape(const pcre_uchar **ptrptr, pcre_uint32 *chptr, int *errorcodeptr,
993: int bracount, int options, BOOL isclass)
994: {
995: /* PCRE_UTF16 has the same value as PCRE_UTF8. */
996: BOOL utf = (options & PCRE_UTF8) != 0;
997: const pcre_uchar *ptr = *ptrptr + 1;
998: pcre_uint32 c;
999: int escape = 0;
1000: int i;
1001:
1002: GETCHARINCTEST(c, ptr); /* Get character value, increment pointer */
1003: ptr--; /* Set pointer back to the last byte */
1004:
1005: /* If backslash is at the end of the pattern, it's an error. */
1006:
1007: if (c == CHAR_NULL) *errorcodeptr = ERR1;
1008:
1009: /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
1010: in a table. A non-zero result is something that can be returned immediately.
1011: Otherwise further processing may be required. */
1012:
1013: #ifndef EBCDIC /* ASCII/UTF-8 coding */
1014: /* Not alphanumeric */
1015: else if (c < CHAR_0 || c > CHAR_z) {}
1016: else if ((i = escapes[c - CHAR_0]) != 0)
1017: { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1018:
1019: #else /* EBCDIC coding */
1020: /* Not alphanumeric */
1021: else if (c < CHAR_a || (!MAX_255(c) || (ebcdic_chartab[c] & 0x0E) == 0)) {}
1022: else if ((i = escapes[c - 0x48]) != 0) { if (i > 0) c = (pcre_uint32)i; else escape = -i; }
1023: #endif
1024:
1025: /* Escapes that need further processing, or are illegal. */
1026:
1027: else
1028: {
1029: const pcre_uchar *oldptr;
1030: BOOL braced, negated, overflow;
1031: int s;
1032:
1033: switch (c)
1034: {
1035: /* A number of Perl escapes are not handled by PCRE. We give an explicit
1036: error. */
1037:
1038: case CHAR_l:
1039: case CHAR_L:
1040: *errorcodeptr = ERR37;
1041: break;
1042:
1043: case CHAR_u:
1044: if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1045: {
1046: /* In JavaScript, \u must be followed by four hexadecimal numbers.
1047: Otherwise it is a lowercase u letter. */
1048: if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1049: && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0
1050: && MAX_255(ptr[3]) && (digitab[ptr[3]] & ctype_xdigit) != 0
1051: && MAX_255(ptr[4]) && (digitab[ptr[4]] & ctype_xdigit) != 0)
1052: {
1053: c = 0;
1054: for (i = 0; i < 4; ++i)
1055: {
1056: register pcre_uint32 cc = *(++ptr);
1057: #ifndef EBCDIC /* ASCII/UTF-8 coding */
1058: if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1059: c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1060: #else /* EBCDIC coding */
1061: if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1062: c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1063: #endif
1064: }
1065:
1066: #if defined COMPILE_PCRE8
1067: if (c > (utf ? 0x10ffffU : 0xffU))
1068: #elif defined COMPILE_PCRE16
1069: if (c > (utf ? 0x10ffffU : 0xffffU))
1070: #elif defined COMPILE_PCRE32
1071: if (utf && c > 0x10ffffU)
1072: #endif
1073: {
1074: *errorcodeptr = ERR76;
1075: }
1076: else if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1077: }
1078: }
1079: else
1080: *errorcodeptr = ERR37;
1081: break;
1082:
1083: case CHAR_U:
1084: /* In JavaScript, \U is an uppercase U letter. */
1085: if ((options & PCRE_JAVASCRIPT_COMPAT) == 0) *errorcodeptr = ERR37;
1086: break;
1087:
1088: /* In a character class, \g is just a literal "g". Outside a character
1089: class, \g must be followed by one of a number of specific things:
1090:
1091: (1) A number, either plain or braced. If positive, it is an absolute
1092: backreference. If negative, it is a relative backreference. This is a Perl
1093: 5.10 feature.
1094:
1095: (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
1096: is part of Perl's movement towards a unified syntax for back references. As
1097: this is synonymous with \k{name}, we fudge it up by pretending it really
1098: was \k.
1099:
1100: (3) For Oniguruma compatibility we also support \g followed by a name or a
1101: number either in angle brackets or in single quotes. However, these are
1102: (possibly recursive) subroutine calls, _not_ backreferences. Just return
1103: the ESC_g code (cf \k). */
1104:
1105: case CHAR_g:
1106: if (isclass) break;
1107: if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
1108: {
1109: escape = ESC_g;
1110: break;
1111: }
1112:
1113: /* Handle the Perl-compatible cases */
1114:
1115: if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1116: {
1117: const pcre_uchar *p;
1118: for (p = ptr+2; *p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
1119: if (*p != CHAR_MINUS && !IS_DIGIT(*p)) break;
1120: if (*p != CHAR_NULL && *p != CHAR_RIGHT_CURLY_BRACKET)
1121: {
1122: escape = ESC_k;
1123: break;
1124: }
1125: braced = TRUE;
1126: ptr++;
1127: }
1128: else braced = FALSE;
1129:
1130: if (ptr[1] == CHAR_MINUS)
1131: {
1132: negated = TRUE;
1133: ptr++;
1134: }
1135: else negated = FALSE;
1136:
1137: /* The integer range is limited by the machine's int representation. */
1138: s = 0;
1139: overflow = FALSE;
1140: while (IS_DIGIT(ptr[1]))
1141: {
1142: if (s > INT_MAX / 10 - 1) /* Integer overflow */
1143: {
1144: overflow = TRUE;
1145: break;
1146: }
1147: s = s * 10 + (int)(*(++ptr) - CHAR_0);
1148: }
1149: if (overflow) /* Integer overflow */
1150: {
1151: while (IS_DIGIT(ptr[1]))
1152: ptr++;
1153: *errorcodeptr = ERR61;
1154: break;
1155: }
1156:
1157: if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
1158: {
1159: *errorcodeptr = ERR57;
1160: break;
1161: }
1162:
1163: if (s == 0)
1164: {
1165: *errorcodeptr = ERR58;
1166: break;
1167: }
1168:
1169: if (negated)
1170: {
1171: if (s > bracount)
1172: {
1173: *errorcodeptr = ERR15;
1174: break;
1175: }
1176: s = bracount - (s - 1);
1177: }
1178:
1179: escape = -s;
1180: break;
1181:
1182: /* The handling of escape sequences consisting of a string of digits
1183: starting with one that is not zero is not straightforward. Perl has changed
1184: over the years. Nowadays \g{} for backreferences and \o{} for octal are
1185: recommended to avoid the ambiguities in the old syntax.
1186:
1187: Outside a character class, the digits are read as a decimal number. If the
1188: number is less than 8 (used to be 10), or if there are that many previous
1189: extracting left brackets, then it is a back reference. Otherwise, up to
1190: three octal digits are read to form an escaped byte. Thus \123 is likely to
1191: be octal 123 (cf \0123, which is octal 012 followed by the literal 3). If
1192: the octal value is greater than 377, the least significant 8 bits are
1193: taken. \8 and \9 are treated as the literal characters 8 and 9.
1194:
1195: Inside a character class, \ followed by a digit is always either a literal
1196: 8 or 9 or an octal number. */
1197:
1198: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
1199: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
1200:
1201: if (!isclass)
1202: {
1203: oldptr = ptr;
1204: /* The integer range is limited by the machine's int representation. */
1205: s = (int)(c -CHAR_0);
1206: overflow = FALSE;
1207: while (IS_DIGIT(ptr[1]))
1208: {
1209: if (s > INT_MAX / 10 - 1) /* Integer overflow */
1210: {
1211: overflow = TRUE;
1212: break;
1213: }
1214: s = s * 10 + (int)(*(++ptr) - CHAR_0);
1215: }
1216: if (overflow) /* Integer overflow */
1217: {
1218: while (IS_DIGIT(ptr[1]))
1219: ptr++;
1220: *errorcodeptr = ERR61;
1221: break;
1222: }
1223: if (s < 8 || s <= bracount) /* Check for back reference */
1224: {
1225: escape = -s;
1226: break;
1227: }
1228: ptr = oldptr; /* Put the pointer back and fall through */
1229: }
1230:
1231: /* Handle a digit following \ when the number is not a back reference. If
1232: the first digit is 8 or 9, Perl used to generate a binary zero byte and
1233: then treat the digit as a following literal. At least by Perl 5.18 this
1234: changed so as not to insert the binary zero. */
1235:
1236: if ((c = *ptr) >= CHAR_8) break;
1237:
1238: /* Fall through with a digit less than 8 */
1239:
1240: /* \0 always starts an octal number, but we may drop through to here with a
1241: larger first octal digit. The original code used just to take the least
1242: significant 8 bits of octal numbers (I think this is what early Perls used
1243: to do). Nowadays we allow for larger numbers in UTF-8 mode and 16-bit mode,
1244: but no more than 3 octal digits. */
1245:
1246: case CHAR_0:
1247: c -= CHAR_0;
1248: while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
1249: c = c * 8 + *(++ptr) - CHAR_0;
1250: #ifdef COMPILE_PCRE8
1251: if (!utf && c > 0xff) *errorcodeptr = ERR51;
1252: #endif
1253: break;
1254:
1255: /* \o is a relatively new Perl feature, supporting a more general way of
1256: specifying character codes in octal. The only supported form is \o{ddd}. */
1257:
1258: case CHAR_o:
1259: if (ptr[1] != CHAR_LEFT_CURLY_BRACKET) *errorcodeptr = ERR81; else
1260: {
1261: ptr += 2;
1262: c = 0;
1263: overflow = FALSE;
1264: while (*ptr >= CHAR_0 && *ptr <= CHAR_7)
1265: {
1266: register pcre_uint32 cc = *ptr++;
1267: if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1268: #ifdef COMPILE_PCRE32
1269: if (c >= 0x20000000l) { overflow = TRUE; break; }
1270: #endif
1271: c = (c << 3) + cc - CHAR_0 ;
1272: #if defined COMPILE_PCRE8
1273: if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1274: #elif defined COMPILE_PCRE16
1275: if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1276: #elif defined COMPILE_PCRE32
1277: if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1278: #endif
1279: }
1280: if (overflow)
1281: {
1282: while (*ptr >= CHAR_0 && *ptr <= CHAR_7) ptr++;
1283: *errorcodeptr = ERR34;
1284: }
1285: else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1286: {
1287: if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1288: }
1289: else *errorcodeptr = ERR80;
1290: }
1291: break;
1292:
1293: /* \x is complicated. In JavaScript, \x must be followed by two hexadecimal
1294: numbers. Otherwise it is a lowercase x letter. */
1295:
1296: case CHAR_x:
1297: if ((options & PCRE_JAVASCRIPT_COMPAT) != 0)
1298: {
1299: if (MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0
1300: && MAX_255(ptr[2]) && (digitab[ptr[2]] & ctype_xdigit) != 0)
1301: {
1302: c = 0;
1303: for (i = 0; i < 2; ++i)
1304: {
1305: register pcre_uint32 cc = *(++ptr);
1306: #ifndef EBCDIC /* ASCII/UTF-8 coding */
1307: if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1308: c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1309: #else /* EBCDIC coding */
1310: if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1311: c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1312: #endif
1313: }
1314: }
1315: } /* End JavaScript handling */
1316:
1317: /* Handle \x in Perl's style. \x{ddd} is a character number which can be
1318: greater than 0xff in utf or non-8bit mode, but only if the ddd are hex
1319: digits. If not, { used to be treated as a data character. However, Perl
1320: seems to read hex digits up to the first non-such, and ignore the rest, so
1321: that, for example \x{zz} matches a binary zero. This seems crazy, so PCRE
1322: now gives an error. */
1323:
1324: else
1325: {
1326: if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
1327: {
1328: ptr += 2;
1329: c = 0;
1330: overflow = FALSE;
1331: while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0)
1332: {
1333: register pcre_uint32 cc = *ptr++;
1334: if (c == 0 && cc == CHAR_0) continue; /* Leading zeroes */
1335:
1336: #ifdef COMPILE_PCRE32
1337: if (c >= 0x10000000l) { overflow = TRUE; break; }
1338: #endif
1339:
1340: #ifndef EBCDIC /* ASCII/UTF-8 coding */
1341: if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1342: c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1343: #else /* EBCDIC coding */
1344: if (cc >= CHAR_a && cc <= CHAR_z) cc += 64; /* Convert to upper case */
1345: c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1346: #endif
1347:
1348: #if defined COMPILE_PCRE8
1349: if (c > (utf ? 0x10ffffU : 0xffU)) { overflow = TRUE; break; }
1350: #elif defined COMPILE_PCRE16
1351: if (c > (utf ? 0x10ffffU : 0xffffU)) { overflow = TRUE; break; }
1352: #elif defined COMPILE_PCRE32
1353: if (utf && c > 0x10ffffU) { overflow = TRUE; break; }
1354: #endif
1355: }
1356:
1357: if (overflow)
1358: {
1359: while (MAX_255(*ptr) && (digitab[*ptr] & ctype_xdigit) != 0) ptr++;
1360: *errorcodeptr = ERR34;
1361: }
1362:
1363: else if (*ptr == CHAR_RIGHT_CURLY_BRACKET)
1364: {
1365: if (utf && c >= 0xd800 && c <= 0xdfff) *errorcodeptr = ERR73;
1366: }
1367:
1368: /* If the sequence of hex digits does not end with '}', give an error.
1369: We used just to recognize this construct and fall through to the normal
1370: \x handling, but nowadays Perl gives an error, which seems much more
1371: sensible, so we do too. */
1372:
1373: else *errorcodeptr = ERR79;
1374: } /* End of \x{} processing */
1375:
1376: /* Read a single-byte hex-defined char (up to two hex digits after \x) */
1377:
1378: else
1379: {
1380: c = 0;
1381: while (i++ < 2 && MAX_255(ptr[1]) && (digitab[ptr[1]] & ctype_xdigit) != 0)
1382: {
1383: pcre_uint32 cc; /* Some compilers don't like */
1384: cc = *(++ptr); /* ++ in initializers */
1385: #ifndef EBCDIC /* ASCII/UTF-8 coding */
1386: if (cc >= CHAR_a) cc -= 32; /* Convert to upper case */
1387: c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
1388: #else /* EBCDIC coding */
1389: if (cc <= CHAR_z) cc += 64; /* Convert to upper case */
1390: c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
1391: #endif
1392: }
1393: } /* End of \xdd handling */
1394: } /* End of Perl-style \x handling */
1395: break;
1396:
1397: /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
1398: An error is given if the byte following \c is not an ASCII character. This
1399: coding is ASCII-specific, but then the whole concept of \cx is
1400: ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
1401:
1402: case CHAR_c:
1403: c = *(++ptr);
1404: if (c == CHAR_NULL)
1405: {
1406: *errorcodeptr = ERR2;
1407: break;
1408: }
1409: #ifndef EBCDIC /* ASCII/UTF-8 coding */
1410: if (c > 127) /* Excludes all non-ASCII in either mode */
1411: {
1412: *errorcodeptr = ERR68;
1413: break;
1414: }
1415: if (c >= CHAR_a && c <= CHAR_z) c -= 32;
1416: c ^= 0x40;
1417: #else /* EBCDIC coding */
1418: if (c >= CHAR_a && c <= CHAR_z) c += 64;
1419: c ^= 0xC0;
1420: #endif
1421: break;
1422:
1423: /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
1424: other alphanumeric following \ is an error if PCRE_EXTRA was set;
1425: otherwise, for Perl compatibility, it is a literal. This code looks a bit
1426: odd, but there used to be some cases other than the default, and there may
1427: be again in future, so I haven't "optimized" it. */
1428:
1429: default:
1430: if ((options & PCRE_EXTRA) != 0) switch(c)
1431: {
1432: default:
1433: *errorcodeptr = ERR3;
1434: break;
1435: }
1436: break;
1437: }
1438: }
1439:
1440: /* Perl supports \N{name} for character names, as well as plain \N for "not
1441: newline". PCRE does not support \N{name}. However, it does support
1442: quantification such as \N{2,3}. */
1443:
1444: if (escape == ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET &&
1445: !is_counted_repeat(ptr+2))
1446: *errorcodeptr = ERR37;
1447:
1448: /* If PCRE_UCP is set, we change the values for \d etc. */
1449:
1450: if ((options & PCRE_UCP) != 0 && escape >= ESC_D && escape <= ESC_w)
1451: escape += (ESC_DU - ESC_D);
1452:
1453: /* Set the pointer to the final character before returning. */
1454:
1455: *ptrptr = ptr;
1456: *chptr = c;
1457: return escape;
1458: }
1459:
1460:
1461:
1462: #ifdef SUPPORT_UCP
1463: /*************************************************
1464: * Handle \P and \p *
1465: *************************************************/
1466:
1467: /* This function is called after \P or \p has been encountered, provided that
1468: PCRE is compiled with support for Unicode properties. On entry, ptrptr is
1469: pointing at the P or p. On exit, it is pointing at the final character of the
1470: escape sequence.
1471:
1472: Argument:
1473: ptrptr points to the pattern position pointer
1474: negptr points to a boolean that is set TRUE for negation else FALSE
1475: ptypeptr points to an unsigned int that is set to the type value
1476: pdataptr points to an unsigned int that is set to the detailed property value
1477: errorcodeptr points to the error code variable
1478:
1479: Returns: TRUE if the type value was found, or FALSE for an invalid type
1480: */
1481:
1482: static BOOL
1483: get_ucp(const pcre_uchar **ptrptr, BOOL *negptr, unsigned int *ptypeptr,
1484: unsigned int *pdataptr, int *errorcodeptr)
1485: {
1486: pcre_uchar c;
1487: int i, bot, top;
1488: const pcre_uchar *ptr = *ptrptr;
1489: pcre_uchar name[32];
1490:
1491: c = *(++ptr);
1492: if (c == CHAR_NULL) goto ERROR_RETURN;
1493:
1494: *negptr = FALSE;
1495:
1496: /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
1497: negation. */
1498:
1499: if (c == CHAR_LEFT_CURLY_BRACKET)
1500: {
1501: if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1502: {
1503: *negptr = TRUE;
1504: ptr++;
1505: }
1506: for (i = 0; i < (int)(sizeof(name) / sizeof(pcre_uchar)) - 1; i++)
1507: {
1508: c = *(++ptr);
1509: if (c == CHAR_NULL) goto ERROR_RETURN;
1510: if (c == CHAR_RIGHT_CURLY_BRACKET) break;
1511: name[i] = c;
1512: }
1513: if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
1514: name[i] = 0;
1515: }
1516:
1517: /* Otherwise there is just one following character */
1518:
1519: else
1520: {
1521: name[0] = c;
1522: name[1] = 0;
1523: }
1524:
1525: *ptrptr = ptr;
1526:
1527: /* Search for a recognized property name using binary chop */
1528:
1529: bot = 0;
1530: top = PRIV(utt_size);
1531:
1532: while (bot < top)
1533: {
1534: int r;
1535: i = (bot + top) >> 1;
1536: r = STRCMP_UC_C8(name, PRIV(utt_names) + PRIV(utt)[i].name_offset);
1537: if (r == 0)
1538: {
1539: *ptypeptr = PRIV(utt)[i].type;
1540: *pdataptr = PRIV(utt)[i].value;
1541: return TRUE;
1542: }
1543: if (r > 0) bot = i + 1; else top = i;
1544: }
1545:
1546: *errorcodeptr = ERR47;
1547: *ptrptr = ptr;
1548: return FALSE;
1549:
1550: ERROR_RETURN:
1551: *errorcodeptr = ERR46;
1552: *ptrptr = ptr;
1553: return FALSE;
1554: }
1555: #endif
1556:
1557:
1558:
1559: /*************************************************
1560: * Read repeat counts *
1561: *************************************************/
1562:
1563: /* Read an item of the form {n,m} and return the values. This is called only
1564: after is_counted_repeat() has confirmed that a repeat-count quantifier exists,
1565: so the syntax is guaranteed to be correct, but we need to check the values.
1566:
1567: Arguments:
1568: p pointer to first char after '{'
1569: minp pointer to int for min
1570: maxp pointer to int for max
1571: returned as -1 if no max
1572: errorcodeptr points to error code variable
1573:
1574: Returns: pointer to '}' on success;
1575: current ptr on error, with errorcodeptr set non-zero
1576: */
1577:
1578: static const pcre_uchar *
1579: read_repeat_counts(const pcre_uchar *p, int *minp, int *maxp, int *errorcodeptr)
1580: {
1581: int min = 0;
1582: int max = -1;
1583:
1584: /* Read the minimum value and do a paranoid check: a negative value indicates
1585: an integer overflow. */
1586:
1587: while (IS_DIGIT(*p)) min = min * 10 + (int)(*p++ - CHAR_0);
1588: if (min < 0 || min > 65535)
1589: {
1590: *errorcodeptr = ERR5;
1591: return p;
1592: }
1593:
1594: /* Read the maximum value if there is one, and again do a paranoid on its size.
1595: Also, max must not be less than min. */
1596:
1597: if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1598: {
1599: if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1600: {
1601: max = 0;
1602: while(IS_DIGIT(*p)) max = max * 10 + (int)(*p++ - CHAR_0);
1603: if (max < 0 || max > 65535)
1604: {
1605: *errorcodeptr = ERR5;
1606: return p;
1607: }
1608: if (max < min)
1609: {
1610: *errorcodeptr = ERR4;
1611: return p;
1612: }
1613: }
1614: }
1615:
1616: /* Fill in the required variables, and pass back the pointer to the terminating
1617: '}'. */
1618:
1619: *minp = min;
1620: *maxp = max;
1621: return p;
1622: }
1623:
1624:
1625:
1626: /*************************************************
1627: * Find first significant op code *
1628: *************************************************/
1629:
1630: /* This is called by several functions that scan a compiled expression looking
1631: for a fixed first character, or an anchoring op code etc. It skips over things
1632: that do not influence this. For some calls, it makes sense to skip negative
1633: forward and all backward assertions, and also the \b assertion; for others it
1634: does not.
1635:
1636: Arguments:
1637: code pointer to the start of the group
1638: skipassert TRUE if certain assertions are to be skipped
1639:
1640: Returns: pointer to the first significant opcode
1641: */
1642:
1643: static const pcre_uchar*
1644: first_significant_code(const pcre_uchar *code, BOOL skipassert)
1645: {
1646: for (;;)
1647: {
1648: switch ((int)*code)
1649: {
1650: case OP_ASSERT_NOT:
1651: case OP_ASSERTBACK:
1652: case OP_ASSERTBACK_NOT:
1653: if (!skipassert) return code;
1654: do code += GET(code, 1); while (*code == OP_ALT);
1655: code += PRIV(OP_lengths)[*code];
1656: break;
1657:
1658: case OP_WORD_BOUNDARY:
1659: case OP_NOT_WORD_BOUNDARY:
1660: if (!skipassert) return code;
1661: /* Fall through */
1662:
1663: case OP_CALLOUT:
1664: case OP_CREF:
1665: case OP_DNCREF:
1666: case OP_RREF:
1667: case OP_DNRREF:
1668: case OP_DEF:
1669: code += PRIV(OP_lengths)[*code];
1670: break;
1671:
1672: default:
1673: return code;
1674: }
1675: }
1676: /* Control never reaches here */
1677: }
1678:
1679:
1680:
1681: /*************************************************
1682: * Find the fixed length of a branch *
1683: *************************************************/
1684:
1685: /* Scan a branch and compute the fixed length of subject that will match it,
1686: if the length is fixed. This is needed for dealing with backward assertions.
1687: In UTF8 mode, the result is in characters rather than bytes. The branch is
1688: temporarily terminated with OP_END when this function is called.
1689:
1690: This function is called when a backward assertion is encountered, so that if it
1691: fails, the error message can point to the correct place in the pattern.
1692: However, we cannot do this when the assertion contains subroutine calls,
1693: because they can be forward references. We solve this by remembering this case
1694: and doing the check at the end; a flag specifies which mode we are running in.
1695:
1696: Arguments:
1697: code points to the start of the pattern (the bracket)
1698: utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
1699: atend TRUE if called when the pattern is complete
1700: cd the "compile data" structure
1701:
1702: Returns: the fixed length,
1703: or -1 if there is no fixed length,
1704: or -2 if \C was encountered (in UTF-8 mode only)
1705: or -3 if an OP_RECURSE item was encountered and atend is FALSE
1706: or -4 if an unknown opcode was encountered (internal error)
1707: */
1708:
1709: static int
1710: find_fixedlength(pcre_uchar *code, BOOL utf, BOOL atend, compile_data *cd)
1711: {
1712: int length = -1;
1713:
1714: register int branchlength = 0;
1715: register pcre_uchar *cc = code + 1 + LINK_SIZE;
1716:
1717: /* Scan along the opcodes for this branch. If we get to the end of the
1718: branch, check the length against that of the other branches. */
1719:
1720: for (;;)
1721: {
1722: int d;
1723: pcre_uchar *ce, *cs;
1724: register pcre_uchar op = *cc;
1725:
1726: switch (op)
1727: {
1728: /* We only need to continue for OP_CBRA (normal capturing bracket) and
1729: OP_BRA (normal non-capturing bracket) because the other variants of these
1730: opcodes are all concerned with unlimited repeated groups, which of course
1731: are not of fixed length. */
1732:
1733: case OP_CBRA:
1734: case OP_BRA:
1735: case OP_ONCE:
1736: case OP_ONCE_NC:
1737: case OP_COND:
1738: d = find_fixedlength(cc + ((op == OP_CBRA)? IMM2_SIZE : 0), utf, atend, cd);
1739: if (d < 0) return d;
1740: branchlength += d;
1741: do cc += GET(cc, 1); while (*cc == OP_ALT);
1742: cc += 1 + LINK_SIZE;
1743: break;
1744:
1745: /* Reached end of a branch; if it's a ket it is the end of a nested call.
1746: If it's ALT it is an alternation in a nested call. An ACCEPT is effectively
1747: an ALT. If it is END it's the end of the outer call. All can be handled by
1748: the same code. Note that we must not include the OP_KETRxxx opcodes here,
1749: because they all imply an unlimited repeat. */
1750:
1751: case OP_ALT:
1752: case OP_KET:
1753: case OP_END:
1754: case OP_ACCEPT:
1755: case OP_ASSERT_ACCEPT:
1756: if (length < 0) length = branchlength;
1757: else if (length != branchlength) return -1;
1758: if (*cc != OP_ALT) return length;
1759: cc += 1 + LINK_SIZE;
1760: branchlength = 0;
1761: break;
1762:
1763: /* A true recursion implies not fixed length, but a subroutine call may
1764: be OK. If the subroutine is a forward reference, we can't deal with
1765: it until the end of the pattern, so return -3. */
1766:
1767: case OP_RECURSE:
1768: if (!atend) return -3;
1769: cs = ce = (pcre_uchar *)cd->start_code + GET(cc, 1); /* Start subpattern */
1770: do ce += GET(ce, 1); while (*ce == OP_ALT); /* End subpattern */
1771: if (cc > cs && cc < ce) return -1; /* Recursion */
1772: d = find_fixedlength(cs + IMM2_SIZE, utf, atend, cd);
1773: if (d < 0) return d;
1774: branchlength += d;
1775: cc += 1 + LINK_SIZE;
1776: break;
1777:
1778: /* Skip over assertive subpatterns */
1779:
1780: case OP_ASSERT:
1781: case OP_ASSERT_NOT:
1782: case OP_ASSERTBACK:
1783: case OP_ASSERTBACK_NOT:
1784: do cc += GET(cc, 1); while (*cc == OP_ALT);
1785: cc += PRIV(OP_lengths)[*cc];
1786: break;
1787:
1788: /* Skip over things that don't match chars */
1789:
1790: case OP_MARK:
1791: case OP_PRUNE_ARG:
1792: case OP_SKIP_ARG:
1793: case OP_THEN_ARG:
1794: cc += cc[1] + PRIV(OP_lengths)[*cc];
1795: break;
1796:
1797: case OP_CALLOUT:
1798: case OP_CIRC:
1799: case OP_CIRCM:
1800: case OP_CLOSE:
1801: case OP_COMMIT:
1802: case OP_CREF:
1803: case OP_DEF:
1804: case OP_DNCREF:
1805: case OP_DNRREF:
1806: case OP_DOLL:
1807: case OP_DOLLM:
1808: case OP_EOD:
1809: case OP_EODN:
1810: case OP_FAIL:
1811: case OP_NOT_WORD_BOUNDARY:
1812: case OP_PRUNE:
1813: case OP_REVERSE:
1814: case OP_RREF:
1815: case OP_SET_SOM:
1816: case OP_SKIP:
1817: case OP_SOD:
1818: case OP_SOM:
1819: case OP_THEN:
1820: case OP_WORD_BOUNDARY:
1821: cc += PRIV(OP_lengths)[*cc];
1822: break;
1823:
1824: /* Handle literal characters */
1825:
1826: case OP_CHAR:
1827: case OP_CHARI:
1828: case OP_NOT:
1829: case OP_NOTI:
1830: branchlength++;
1831: cc += 2;
1832: #ifdef SUPPORT_UTF
1833: if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1834: #endif
1835: break;
1836:
1837: /* Handle exact repetitions. The count is already in characters, but we
1838: need to skip over a multibyte character in UTF8 mode. */
1839:
1840: case OP_EXACT:
1841: case OP_EXACTI:
1842: case OP_NOTEXACT:
1843: case OP_NOTEXACTI:
1844: branchlength += (int)GET2(cc,1);
1845: cc += 2 + IMM2_SIZE;
1846: #ifdef SUPPORT_UTF
1847: if (utf && HAS_EXTRALEN(cc[-1])) cc += GET_EXTRALEN(cc[-1]);
1848: #endif
1849: break;
1850:
1851: case OP_TYPEEXACT:
1852: branchlength += GET2(cc,1);
1853: if (cc[1 + IMM2_SIZE] == OP_PROP || cc[1 + IMM2_SIZE] == OP_NOTPROP)
1854: cc += 2;
1855: cc += 1 + IMM2_SIZE + 1;
1856: break;
1857:
1858: /* Handle single-char matchers */
1859:
1860: case OP_PROP:
1861: case OP_NOTPROP:
1862: cc += 2;
1863: /* Fall through */
1864:
1865: case OP_HSPACE:
1866: case OP_VSPACE:
1867: case OP_NOT_HSPACE:
1868: case OP_NOT_VSPACE:
1869: case OP_NOT_DIGIT:
1870: case OP_DIGIT:
1871: case OP_NOT_WHITESPACE:
1872: case OP_WHITESPACE:
1873: case OP_NOT_WORDCHAR:
1874: case OP_WORDCHAR:
1875: case OP_ANY:
1876: case OP_ALLANY:
1877: branchlength++;
1878: cc++;
1879: break;
1880:
1881: /* The single-byte matcher isn't allowed. This only happens in UTF-8 mode;
1882: otherwise \C is coded as OP_ALLANY. */
1883:
1884: case OP_ANYBYTE:
1885: return -2;
1886:
1887: /* Check a class for variable quantification */
1888:
1889: case OP_CLASS:
1890: case OP_NCLASS:
1891: #if defined SUPPORT_UTF || defined COMPILE_PCRE16 || defined COMPILE_PCRE32
1892: case OP_XCLASS:
1893: /* The original code caused an unsigned overflow in 64 bit systems,
1894: so now we use a conditional statement. */
1895: if (op == OP_XCLASS)
1896: cc += GET(cc, 1);
1897: else
1898: cc += PRIV(OP_lengths)[OP_CLASS];
1899: #else
1900: cc += PRIV(OP_lengths)[OP_CLASS];
1901: #endif
1902:
1903: switch (*cc)
1904: {
1905: case OP_CRSTAR:
1906: case OP_CRMINSTAR:
1907: case OP_CRPLUS:
1908: case OP_CRMINPLUS:
1909: case OP_CRQUERY:
1910: case OP_CRMINQUERY:
1911: case OP_CRPOSSTAR:
1912: case OP_CRPOSPLUS:
1913: case OP_CRPOSQUERY:
1914: return -1;
1915:
1916: case OP_CRRANGE:
1917: case OP_CRMINRANGE:
1918: case OP_CRPOSRANGE:
1919: if (GET2(cc,1) != GET2(cc,1+IMM2_SIZE)) return -1;
1920: branchlength += (int)GET2(cc,1);
1921: cc += 1 + 2 * IMM2_SIZE;
1922: break;
1923:
1924: default:
1925: branchlength++;
1926: }
1927: break;
1928:
1929: /* Anything else is variable length */
1930:
1931: case OP_ANYNL:
1932: case OP_BRAMINZERO:
1933: case OP_BRAPOS:
1934: case OP_BRAPOSZERO:
1935: case OP_BRAZERO:
1936: case OP_CBRAPOS:
1937: case OP_EXTUNI:
1938: case OP_KETRMAX:
1939: case OP_KETRMIN:
1940: case OP_KETRPOS:
1941: case OP_MINPLUS:
1942: case OP_MINPLUSI:
1943: case OP_MINQUERY:
1944: case OP_MINQUERYI:
1945: case OP_MINSTAR:
1946: case OP_MINSTARI:
1947: case OP_MINUPTO:
1948: case OP_MINUPTOI:
1949: case OP_NOTMINPLUS:
1950: case OP_NOTMINPLUSI:
1951: case OP_NOTMINQUERY:
1952: case OP_NOTMINQUERYI:
1953: case OP_NOTMINSTAR:
1954: case OP_NOTMINSTARI:
1955: case OP_NOTMINUPTO:
1956: case OP_NOTMINUPTOI:
1957: case OP_NOTPLUS:
1958: case OP_NOTPLUSI:
1959: case OP_NOTPOSPLUS:
1960: case OP_NOTPOSPLUSI:
1961: case OP_NOTPOSQUERY:
1962: case OP_NOTPOSQUERYI:
1963: case OP_NOTPOSSTAR:
1964: case OP_NOTPOSSTARI:
1965: case OP_NOTPOSUPTO:
1966: case OP_NOTPOSUPTOI:
1967: case OP_NOTQUERY:
1968: case OP_NOTQUERYI:
1969: case OP_NOTSTAR:
1970: case OP_NOTSTARI:
1971: case OP_NOTUPTO:
1972: case OP_NOTUPTOI:
1973: case OP_PLUS:
1974: case OP_PLUSI:
1975: case OP_POSPLUS:
1976: case OP_POSPLUSI:
1977: case OP_POSQUERY:
1978: case OP_POSQUERYI:
1979: case OP_POSSTAR:
1980: case OP_POSSTARI:
1981: case OP_POSUPTO:
1982: case OP_POSUPTOI:
1983: case OP_QUERY:
1984: case OP_QUERYI:
1985: case OP_REF:
1986: case OP_REFI:
1987: case OP_DNREF:
1988: case OP_DNREFI:
1989: case OP_SBRA:
1990: case OP_SBRAPOS:
1991: case OP_SCBRA:
1992: case OP_SCBRAPOS:
1993: case OP_SCOND:
1994: case OP_SKIPZERO:
1995: case OP_STAR:
1996: case OP_STARI:
1997: case OP_TYPEMINPLUS:
1998: case OP_TYPEMINQUERY:
1999: case OP_TYPEMINSTAR:
2000: case OP_TYPEMINUPTO:
2001: case OP_TYPEPLUS:
2002: case OP_TYPEPOSPLUS:
2003: case OP_TYPEPOSQUERY:
2004: case OP_TYPEPOSSTAR:
2005: case OP_TYPEPOSUPTO:
2006: case OP_TYPEQUERY:
2007: case OP_TYPESTAR:
2008: case OP_TYPEUPTO:
2009: case OP_UPTO:
2010: case OP_UPTOI:
2011: return -1;
2012:
2013: /* Catch unrecognized opcodes so that when new ones are added they
2014: are not forgotten, as has happened in the past. */
2015:
2016: default:
2017: return -4;
2018: }
2019: }
2020: /* Control never gets here */
2021: }
2022:
2023:
2024:
2025: /*************************************************
2026: * Scan compiled regex for specific bracket *
2027: *************************************************/
2028:
2029: /* This little function scans through a compiled pattern until it finds a
2030: capturing bracket with the given number, or, if the number is negative, an
2031: instance of OP_REVERSE for a lookbehind. The function is global in the C sense
2032: so that it can be called from pcre_study() when finding the minimum matching
2033: length.
2034:
2035: Arguments:
2036: code points to start of expression
2037: utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2038: number the required bracket number or negative to find a lookbehind
2039:
2040: Returns: pointer to the opcode for the bracket, or NULL if not found
2041: */
2042:
2043: const pcre_uchar *
2044: PRIV(find_bracket)(const pcre_uchar *code, BOOL utf, int number)
2045: {
2046: for (;;)
2047: {
2048: register pcre_uchar c = *code;
2049:
2050: if (c == OP_END) return NULL;
2051:
2052: /* XCLASS is used for classes that cannot be represented just by a bit
2053: map. This includes negated single high-valued characters. The length in
2054: the table is zero; the actual length is stored in the compiled code. */
2055:
2056: if (c == OP_XCLASS) code += GET(code, 1);
2057:
2058: /* Handle recursion */
2059:
2060: else if (c == OP_REVERSE)
2061: {
2062: if (number < 0) return (pcre_uchar *)code;
2063: code += PRIV(OP_lengths)[c];
2064: }
2065:
2066: /* Handle capturing bracket */
2067:
2068: else if (c == OP_CBRA || c == OP_SCBRA ||
2069: c == OP_CBRAPOS || c == OP_SCBRAPOS)
2070: {
2071: int n = (int)GET2(code, 1+LINK_SIZE);
2072: if (n == number) return (pcre_uchar *)code;
2073: code += PRIV(OP_lengths)[c];
2074: }
2075:
2076: /* Otherwise, we can get the item's length from the table, except that for
2077: repeated character types, we have to test for \p and \P, which have an extra
2078: two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2079: must add in its length. */
2080:
2081: else
2082: {
2083: switch(c)
2084: {
2085: case OP_TYPESTAR:
2086: case OP_TYPEMINSTAR:
2087: case OP_TYPEPLUS:
2088: case OP_TYPEMINPLUS:
2089: case OP_TYPEQUERY:
2090: case OP_TYPEMINQUERY:
2091: case OP_TYPEPOSSTAR:
2092: case OP_TYPEPOSPLUS:
2093: case OP_TYPEPOSQUERY:
2094: if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2095: break;
2096:
2097: case OP_TYPEUPTO:
2098: case OP_TYPEMINUPTO:
2099: case OP_TYPEEXACT:
2100: case OP_TYPEPOSUPTO:
2101: if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2102: code += 2;
2103: break;
2104:
2105: case OP_MARK:
2106: case OP_PRUNE_ARG:
2107: case OP_SKIP_ARG:
2108: case OP_THEN_ARG:
2109: code += code[1];
2110: break;
2111: }
2112:
2113: /* Add in the fixed length from the table */
2114:
2115: code += PRIV(OP_lengths)[c];
2116:
2117: /* In UTF-8 mode, opcodes that are followed by a character may be followed by
2118: a multi-byte character. The length in the table is a minimum, so we have to
2119: arrange to skip the extra bytes. */
2120:
2121: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2122: if (utf) switch(c)
2123: {
2124: case OP_CHAR:
2125: case OP_CHARI:
2126: case OP_EXACT:
2127: case OP_EXACTI:
2128: case OP_UPTO:
2129: case OP_UPTOI:
2130: case OP_MINUPTO:
2131: case OP_MINUPTOI:
2132: case OP_POSUPTO:
2133: case OP_POSUPTOI:
2134: case OP_STAR:
2135: case OP_STARI:
2136: case OP_MINSTAR:
2137: case OP_MINSTARI:
2138: case OP_POSSTAR:
2139: case OP_POSSTARI:
2140: case OP_PLUS:
2141: case OP_PLUSI:
2142: case OP_MINPLUS:
2143: case OP_MINPLUSI:
2144: case OP_POSPLUS:
2145: case OP_POSPLUSI:
2146: case OP_QUERY:
2147: case OP_QUERYI:
2148: case OP_MINQUERY:
2149: case OP_MINQUERYI:
2150: case OP_POSQUERY:
2151: case OP_POSQUERYI:
2152: if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2153: break;
2154: }
2155: #else
2156: (void)(utf); /* Keep compiler happy by referencing function argument */
2157: #endif
2158: }
2159: }
2160: }
2161:
2162:
2163:
2164: /*************************************************
2165: * Scan compiled regex for recursion reference *
2166: *************************************************/
2167:
2168: /* This little function scans through a compiled pattern until it finds an
2169: instance of OP_RECURSE.
2170:
2171: Arguments:
2172: code points to start of expression
2173: utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
2174:
2175: Returns: pointer to the opcode for OP_RECURSE, or NULL if not found
2176: */
2177:
2178: static const pcre_uchar *
2179: find_recurse(const pcre_uchar *code, BOOL utf)
2180: {
2181: for (;;)
2182: {
2183: register pcre_uchar c = *code;
2184: if (c == OP_END) return NULL;
2185: if (c == OP_RECURSE) return code;
2186:
2187: /* XCLASS is used for classes that cannot be represented just by a bit
2188: map. This includes negated single high-valued characters. The length in
2189: the table is zero; the actual length is stored in the compiled code. */
2190:
2191: if (c == OP_XCLASS) code += GET(code, 1);
2192:
2193: /* Otherwise, we can get the item's length from the table, except that for
2194: repeated character types, we have to test for \p and \P, which have an extra
2195: two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
2196: must add in its length. */
2197:
2198: else
2199: {
2200: switch(c)
2201: {
2202: case OP_TYPESTAR:
2203: case OP_TYPEMINSTAR:
2204: case OP_TYPEPLUS:
2205: case OP_TYPEMINPLUS:
2206: case OP_TYPEQUERY:
2207: case OP_TYPEMINQUERY:
2208: case OP_TYPEPOSSTAR:
2209: case OP_TYPEPOSPLUS:
2210: case OP_TYPEPOSQUERY:
2211: if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2212: break;
2213:
2214: case OP_TYPEPOSUPTO:
2215: case OP_TYPEUPTO:
2216: case OP_TYPEMINUPTO:
2217: case OP_TYPEEXACT:
2218: if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2219: code += 2;
2220: break;
2221:
2222: case OP_MARK:
2223: case OP_PRUNE_ARG:
2224: case OP_SKIP_ARG:
2225: case OP_THEN_ARG:
2226: code += code[1];
2227: break;
2228: }
2229:
2230: /* Add in the fixed length from the table */
2231:
2232: code += PRIV(OP_lengths)[c];
2233:
2234: /* In UTF-8 mode, opcodes that are followed by a character may be followed
2235: by a multi-byte character. The length in the table is a minimum, so we have
2236: to arrange to skip the extra bytes. */
2237:
2238: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2239: if (utf) switch(c)
2240: {
2241: case OP_CHAR:
2242: case OP_CHARI:
2243: case OP_NOT:
2244: case OP_NOTI:
2245: case OP_EXACT:
2246: case OP_EXACTI:
2247: case OP_NOTEXACT:
2248: case OP_NOTEXACTI:
2249: case OP_UPTO:
2250: case OP_UPTOI:
2251: case OP_NOTUPTO:
2252: case OP_NOTUPTOI:
2253: case OP_MINUPTO:
2254: case OP_MINUPTOI:
2255: case OP_NOTMINUPTO:
2256: case OP_NOTMINUPTOI:
2257: case OP_POSUPTO:
2258: case OP_POSUPTOI:
2259: case OP_NOTPOSUPTO:
2260: case OP_NOTPOSUPTOI:
2261: case OP_STAR:
2262: case OP_STARI:
2263: case OP_NOTSTAR:
2264: case OP_NOTSTARI:
2265: case OP_MINSTAR:
2266: case OP_MINSTARI:
2267: case OP_NOTMINSTAR:
2268: case OP_NOTMINSTARI:
2269: case OP_POSSTAR:
2270: case OP_POSSTARI:
2271: case OP_NOTPOSSTAR:
2272: case OP_NOTPOSSTARI:
2273: case OP_PLUS:
2274: case OP_PLUSI:
2275: case OP_NOTPLUS:
2276: case OP_NOTPLUSI:
2277: case OP_MINPLUS:
2278: case OP_MINPLUSI:
2279: case OP_NOTMINPLUS:
2280: case OP_NOTMINPLUSI:
2281: case OP_POSPLUS:
2282: case OP_POSPLUSI:
2283: case OP_NOTPOSPLUS:
2284: case OP_NOTPOSPLUSI:
2285: case OP_QUERY:
2286: case OP_QUERYI:
2287: case OP_NOTQUERY:
2288: case OP_NOTQUERYI:
2289: case OP_MINQUERY:
2290: case OP_MINQUERYI:
2291: case OP_NOTMINQUERY:
2292: case OP_NOTMINQUERYI:
2293: case OP_POSQUERY:
2294: case OP_POSQUERYI:
2295: case OP_NOTPOSQUERY:
2296: case OP_NOTPOSQUERYI:
2297: if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
2298: break;
2299: }
2300: #else
2301: (void)(utf); /* Keep compiler happy by referencing function argument */
2302: #endif
2303: }
2304: }
2305: }
2306:
2307:
2308:
2309: /*************************************************
2310: * Scan compiled branch for non-emptiness *
2311: *************************************************/
2312:
2313: /* This function scans through a branch of a compiled pattern to see whether it
2314: can match the empty string or not. It is called from could_be_empty()
2315: below and from compile_branch() when checking for an unlimited repeat of a
2316: group that can match nothing. Note that first_significant_code() skips over
2317: backward and negative forward assertions when its final argument is TRUE. If we
2318: hit an unclosed bracket, we return "empty" - this means we've struck an inner
2319: bracket whose current branch will already have been scanned.
2320:
2321: Arguments:
2322: code points to start of search
2323: endcode points to where to stop
2324: utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2325: cd contains pointers to tables etc.
2326: recurses chain of recurse_check to catch mutual recursion
2327:
2328: Returns: TRUE if what is matched could be empty
2329: */
2330:
2331: typedef struct recurse_check {
2332: struct recurse_check *prev;
2333: const pcre_uchar *group;
2334: } recurse_check;
2335:
2336: static BOOL
2337: could_be_empty_branch(const pcre_uchar *code, const pcre_uchar *endcode,
2338: BOOL utf, compile_data *cd, recurse_check *recurses)
2339: {
2340: register pcre_uchar c;
2341: recurse_check this_recurse;
2342:
2343: for (code = first_significant_code(code + PRIV(OP_lengths)[*code], TRUE);
2344: code < endcode;
2345: code = first_significant_code(code + PRIV(OP_lengths)[c], TRUE))
2346: {
2347: const pcre_uchar *ccode;
2348:
2349: c = *code;
2350:
2351: /* Skip over forward assertions; the other assertions are skipped by
2352: first_significant_code() with a TRUE final argument. */
2353:
2354: if (c == OP_ASSERT)
2355: {
2356: do code += GET(code, 1); while (*code == OP_ALT);
2357: c = *code;
2358: continue;
2359: }
2360:
2361: /* For a recursion/subroutine call, if its end has been reached, which
2362: implies a backward reference subroutine call, we can scan it. If it's a
2363: forward reference subroutine call, we can't. To detect forward reference
2364: we have to scan up the list that is kept in the workspace. This function is
2365: called only when doing the real compile, not during the pre-compile that
2366: measures the size of the compiled pattern. */
2367:
2368: if (c == OP_RECURSE)
2369: {
2370: const pcre_uchar *scode = cd->start_code + GET(code, 1);
2371: BOOL empty_branch;
2372:
2373: /* Test for forward reference or uncompleted reference. This is disabled
2374: when called to scan a completed pattern by setting cd->start_workspace to
2375: NULL. */
2376:
2377: if (cd->start_workspace != NULL)
2378: {
2379: const pcre_uchar *tcode;
2380: for (tcode = cd->start_workspace; tcode < cd->hwm; tcode += LINK_SIZE)
2381: if ((int)GET(tcode, 0) == (int)(code + 1 - cd->start_code)) return TRUE;
2382: if (GET(scode, 1) == 0) return TRUE; /* Unclosed */
2383: }
2384:
2385: /* If we are scanning a completed pattern, there are no forward references
2386: and all groups are complete. We need to detect whether this is a recursive
2387: call, as otherwise there will be an infinite loop. If it is a recursion,
2388: just skip over it. Simple recursions are easily detected. For mutual
2389: recursions we keep a chain on the stack. */
2390:
2391: else
2392: {
2393: recurse_check *r = recurses;
2394: const pcre_uchar *endgroup = scode;
2395:
2396: do endgroup += GET(endgroup, 1); while (*endgroup == OP_ALT);
2397: if (code >= scode && code <= endgroup) continue; /* Simple recursion */
2398:
2399: for (r = recurses; r != NULL; r = r->prev)
2400: if (r->group == scode) break;
2401: if (r != NULL) continue; /* Mutual recursion */
2402: }
2403:
2404: /* Completed reference; scan the referenced group, remembering it on the
2405: stack chain to detect mutual recursions. */
2406:
2407: empty_branch = FALSE;
2408: this_recurse.prev = recurses;
2409: this_recurse.group = scode;
2410:
2411: do
2412: {
2413: if (could_be_empty_branch(scode, endcode, utf, cd, &this_recurse))
2414: {
2415: empty_branch = TRUE;
2416: break;
2417: }
2418: scode += GET(scode, 1);
2419: }
2420: while (*scode == OP_ALT);
2421:
2422: if (!empty_branch) return FALSE; /* All branches are non-empty */
2423: continue;
2424: }
2425:
2426: /* Groups with zero repeats can of course be empty; skip them. */
2427:
2428: if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO ||
2429: c == OP_BRAPOSZERO)
2430: {
2431: code += PRIV(OP_lengths)[c];
2432: do code += GET(code, 1); while (*code == OP_ALT);
2433: c = *code;
2434: continue;
2435: }
2436:
2437: /* A nested group that is already marked as "could be empty" can just be
2438: skipped. */
2439:
2440: if (c == OP_SBRA || c == OP_SBRAPOS ||
2441: c == OP_SCBRA || c == OP_SCBRAPOS)
2442: {
2443: do code += GET(code, 1); while (*code == OP_ALT);
2444: c = *code;
2445: continue;
2446: }
2447:
2448: /* For other groups, scan the branches. */
2449:
2450: if (c == OP_BRA || c == OP_BRAPOS ||
2451: c == OP_CBRA || c == OP_CBRAPOS ||
2452: c == OP_ONCE || c == OP_ONCE_NC ||
2453: c == OP_COND)
2454: {
2455: BOOL empty_branch;
2456: if (GET(code, 1) == 0) return TRUE; /* Hit unclosed bracket */
2457:
2458: /* If a conditional group has only one branch, there is a second, implied,
2459: empty branch, so just skip over the conditional, because it could be empty.
2460: Otherwise, scan the individual branches of the group. */
2461:
2462: if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
2463: code += GET(code, 1);
2464: else
2465: {
2466: empty_branch = FALSE;
2467: do
2468: {
2469: if (!empty_branch && could_be_empty_branch(code, endcode, utf, cd, NULL))
2470: empty_branch = TRUE;
2471: code += GET(code, 1);
2472: }
2473: while (*code == OP_ALT);
2474: if (!empty_branch) return FALSE; /* All branches are non-empty */
2475: }
2476:
2477: c = *code;
2478: continue;
2479: }
2480:
2481: /* Handle the other opcodes */
2482:
2483: switch (c)
2484: {
2485: /* Check for quantifiers after a class. XCLASS is used for classes that
2486: cannot be represented just by a bit map. This includes negated single
2487: high-valued characters. The length in PRIV(OP_lengths)[] is zero; the
2488: actual length is stored in the compiled code, so we must update "code"
2489: here. */
2490:
2491: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2492: case OP_XCLASS:
2493: ccode = code += GET(code, 1);
2494: goto CHECK_CLASS_REPEAT;
2495: #endif
2496:
2497: case OP_CLASS:
2498: case OP_NCLASS:
2499: ccode = code + PRIV(OP_lengths)[OP_CLASS];
2500:
2501: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
2502: CHECK_CLASS_REPEAT:
2503: #endif
2504:
2505: switch (*ccode)
2506: {
2507: case OP_CRSTAR: /* These could be empty; continue */
2508: case OP_CRMINSTAR:
2509: case OP_CRQUERY:
2510: case OP_CRMINQUERY:
2511: case OP_CRPOSSTAR:
2512: case OP_CRPOSQUERY:
2513: break;
2514:
2515: default: /* Non-repeat => class must match */
2516: case OP_CRPLUS: /* These repeats aren't empty */
2517: case OP_CRMINPLUS:
2518: case OP_CRPOSPLUS:
2519: return FALSE;
2520:
2521: case OP_CRRANGE:
2522: case OP_CRMINRANGE:
2523: case OP_CRPOSRANGE:
2524: if (GET2(ccode, 1) > 0) return FALSE; /* Minimum > 0 */
2525: break;
2526: }
2527: break;
2528:
2529: /* Opcodes that must match a character */
2530:
2531: case OP_ANY:
2532: case OP_ALLANY:
2533: case OP_ANYBYTE:
2534:
2535: case OP_PROP:
2536: case OP_NOTPROP:
2537: case OP_ANYNL:
2538:
2539: case OP_NOT_HSPACE:
2540: case OP_HSPACE:
2541: case OP_NOT_VSPACE:
2542: case OP_VSPACE:
2543: case OP_EXTUNI:
2544:
2545: case OP_NOT_DIGIT:
2546: case OP_DIGIT:
2547: case OP_NOT_WHITESPACE:
2548: case OP_WHITESPACE:
2549: case OP_NOT_WORDCHAR:
2550: case OP_WORDCHAR:
2551:
2552: case OP_CHAR:
2553: case OP_CHARI:
2554: case OP_NOT:
2555: case OP_NOTI:
2556:
2557: case OP_PLUS:
2558: case OP_PLUSI:
2559: case OP_MINPLUS:
2560: case OP_MINPLUSI:
2561:
2562: case OP_NOTPLUS:
2563: case OP_NOTPLUSI:
2564: case OP_NOTMINPLUS:
2565: case OP_NOTMINPLUSI:
2566:
2567: case OP_POSPLUS:
2568: case OP_POSPLUSI:
2569: case OP_NOTPOSPLUS:
2570: case OP_NOTPOSPLUSI:
2571:
2572: case OP_EXACT:
2573: case OP_EXACTI:
2574: case OP_NOTEXACT:
2575: case OP_NOTEXACTI:
2576:
2577: case OP_TYPEPLUS:
2578: case OP_TYPEMINPLUS:
2579: case OP_TYPEPOSPLUS:
2580: case OP_TYPEEXACT:
2581:
2582: return FALSE;
2583:
2584: /* These are going to continue, as they may be empty, but we have to
2585: fudge the length for the \p and \P cases. */
2586:
2587: case OP_TYPESTAR:
2588: case OP_TYPEMINSTAR:
2589: case OP_TYPEPOSSTAR:
2590: case OP_TYPEQUERY:
2591: case OP_TYPEMINQUERY:
2592: case OP_TYPEPOSQUERY:
2593: if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2594: break;
2595:
2596: /* Same for these */
2597:
2598: case OP_TYPEUPTO:
2599: case OP_TYPEMINUPTO:
2600: case OP_TYPEPOSUPTO:
2601: if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
2602: code += 2;
2603: break;
2604:
2605: /* End of branch */
2606:
2607: case OP_KET:
2608: case OP_KETRMAX:
2609: case OP_KETRMIN:
2610: case OP_KETRPOS:
2611: case OP_ALT:
2612: return TRUE;
2613:
2614: /* In UTF-8 mode, STAR, MINSTAR, POSSTAR, QUERY, MINQUERY, POSQUERY, UPTO,
2615: MINUPTO, and POSUPTO and their caseless and negative versions may be
2616: followed by a multibyte character. */
2617:
2618: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
2619: case OP_STAR:
2620: case OP_STARI:
2621: case OP_NOTSTAR:
2622: case OP_NOTSTARI:
2623:
2624: case OP_MINSTAR:
2625: case OP_MINSTARI:
2626: case OP_NOTMINSTAR:
2627: case OP_NOTMINSTARI:
2628:
2629: case OP_POSSTAR:
2630: case OP_POSSTARI:
2631: case OP_NOTPOSSTAR:
2632: case OP_NOTPOSSTARI:
2633:
2634: case OP_QUERY:
2635: case OP_QUERYI:
2636: case OP_NOTQUERY:
2637: case OP_NOTQUERYI:
2638:
2639: case OP_MINQUERY:
2640: case OP_MINQUERYI:
2641: case OP_NOTMINQUERY:
2642: case OP_NOTMINQUERYI:
2643:
2644: case OP_POSQUERY:
2645: case OP_POSQUERYI:
2646: case OP_NOTPOSQUERY:
2647: case OP_NOTPOSQUERYI:
2648:
2649: if (utf && HAS_EXTRALEN(code[1])) code += GET_EXTRALEN(code[1]);
2650: break;
2651:
2652: case OP_UPTO:
2653: case OP_UPTOI:
2654: case OP_NOTUPTO:
2655: case OP_NOTUPTOI:
2656:
2657: case OP_MINUPTO:
2658: case OP_MINUPTOI:
2659: case OP_NOTMINUPTO:
2660: case OP_NOTMINUPTOI:
2661:
2662: case OP_POSUPTO:
2663: case OP_POSUPTOI:
2664: case OP_NOTPOSUPTO:
2665: case OP_NOTPOSUPTOI:
2666:
2667: if (utf && HAS_EXTRALEN(code[1 + IMM2_SIZE])) code += GET_EXTRALEN(code[1 + IMM2_SIZE]);
2668: break;
2669: #endif
2670:
2671: /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2672: string. */
2673:
2674: case OP_MARK:
2675: case OP_PRUNE_ARG:
2676: case OP_SKIP_ARG:
2677: case OP_THEN_ARG:
2678: code += code[1];
2679: break;
2680:
2681: /* None of the remaining opcodes are required to match a character. */
2682:
2683: default:
2684: break;
2685: }
2686: }
2687:
2688: return TRUE;
2689: }
2690:
2691:
2692:
2693: /*************************************************
2694: * Scan compiled regex for non-emptiness *
2695: *************************************************/
2696:
2697: /* This function is called to check for left recursive calls. We want to check
2698: the current branch of the current pattern to see if it could match the empty
2699: string. If it could, we must look outwards for branches at other levels,
2700: stopping when we pass beyond the bracket which is the subject of the recursion.
2701: This function is called only during the real compile, not during the
2702: pre-compile.
2703:
2704: Arguments:
2705: code points to start of the recursion
2706: endcode points to where to stop (current RECURSE item)
2707: bcptr points to the chain of current (unclosed) branch starts
2708: utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2709: cd pointers to tables etc
2710:
2711: Returns: TRUE if what is matched could be empty
2712: */
2713:
2714: static BOOL
2715: could_be_empty(const pcre_uchar *code, const pcre_uchar *endcode,
2716: branch_chain *bcptr, BOOL utf, compile_data *cd)
2717: {
2718: while (bcptr != NULL && bcptr->current_branch >= code)
2719: {
2720: if (!could_be_empty_branch(bcptr->current_branch, endcode, utf, cd, NULL))
2721: return FALSE;
2722: bcptr = bcptr->outer;
2723: }
2724: return TRUE;
2725: }
2726:
2727:
2728:
2729: /*************************************************
2730: * Base opcode of repeated opcodes *
2731: *************************************************/
2732:
2733: /* Returns the base opcode for repeated single character type opcodes. If the
2734: opcode is not a repeated character type, it returns with the original value.
2735:
2736: Arguments: c opcode
2737: Returns: base opcode for the type
2738: */
2739:
2740: static pcre_uchar
2741: get_repeat_base(pcre_uchar c)
2742: {
2743: return (c > OP_TYPEPOSUPTO)? c :
2744: (c >= OP_TYPESTAR)? OP_TYPESTAR :
2745: (c >= OP_NOTSTARI)? OP_NOTSTARI :
2746: (c >= OP_NOTSTAR)? OP_NOTSTAR :
2747: (c >= OP_STARI)? OP_STARI :
2748: OP_STAR;
2749: }
2750:
2751:
2752:
2753: #ifdef SUPPORT_UCP
2754: /*************************************************
2755: * Check a character and a property *
2756: *************************************************/
2757:
2758: /* This function is called by check_auto_possessive() when a property item
2759: is adjacent to a fixed character.
2760:
2761: Arguments:
2762: c the character
2763: ptype the property type
2764: pdata the data for the type
2765: negated TRUE if it's a negated property (\P or \p{^)
2766:
2767: Returns: TRUE if auto-possessifying is OK
2768: */
2769:
2770: static BOOL
2771: check_char_prop(pcre_uint32 c, unsigned int ptype, unsigned int pdata,
2772: BOOL negated)
2773: {
2774: const pcre_uint32 *p;
2775: const ucd_record *prop = GET_UCD(c);
2776:
2777: switch(ptype)
2778: {
2779: case PT_LAMP:
2780: return (prop->chartype == ucp_Lu ||
2781: prop->chartype == ucp_Ll ||
2782: prop->chartype == ucp_Lt) == negated;
2783:
2784: case PT_GC:
2785: return (pdata == PRIV(ucp_gentype)[prop->chartype]) == negated;
2786:
2787: case PT_PC:
2788: return (pdata == prop->chartype) == negated;
2789:
2790: case PT_SC:
2791: return (pdata == prop->script) == negated;
2792:
2793: /* These are specials */
2794:
2795: case PT_ALNUM:
2796: return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2797: PRIV(ucp_gentype)[prop->chartype] == ucp_N) == negated;
2798:
2799: /* Perl space used to exclude VT, but from Perl 5.18 it is included, which
2800: means that Perl space and POSIX space are now identical. PCRE was changed
2801: at release 8.34. */
2802:
2803: case PT_SPACE: /* Perl space */
2804: case PT_PXSPACE: /* POSIX space */
2805: switch(c)
2806: {
2807: HSPACE_CASES:
2808: VSPACE_CASES:
2809: return negated;
2810:
2811: default:
2812: return (PRIV(ucp_gentype)[prop->chartype] == ucp_Z) == negated;
2813: }
2814: break; /* Control never reaches here */
2815:
2816: case PT_WORD:
2817: return (PRIV(ucp_gentype)[prop->chartype] == ucp_L ||
2818: PRIV(ucp_gentype)[prop->chartype] == ucp_N ||
2819: c == CHAR_UNDERSCORE) == negated;
2820:
2821: case PT_CLIST:
2822: p = PRIV(ucd_caseless_sets) + prop->caseset;
2823: for (;;)
2824: {
2825: if (c < *p) return !negated;
2826: if (c == *p++) return negated;
2827: }
2828: break; /* Control never reaches here */
2829: }
2830:
2831: return FALSE;
2832: }
2833: #endif /* SUPPORT_UCP */
2834:
2835:
2836:
2837: /*************************************************
2838: * Fill the character property list *
2839: *************************************************/
2840:
2841: /* Checks whether the code points to an opcode that can take part in auto-
2842: possessification, and if so, fills a list with its properties.
2843:
2844: Arguments:
2845: code points to start of expression
2846: utf TRUE if in UTF-8 / UTF-16 / UTF-32 mode
2847: fcc points to case-flipping table
2848: list points to output list
2849: list[0] will be filled with the opcode
2850: list[1] will be non-zero if this opcode
2851: can match an empty character string
2852: list[2..7] depends on the opcode
2853:
2854: Returns: points to the start of the next opcode if *code is accepted
2855: NULL if *code is not accepted
2856: */
2857:
2858: static const pcre_uchar *
2859: get_chr_property_list(const pcre_uchar *code, BOOL utf,
2860: const pcre_uint8 *fcc, pcre_uint32 *list)
2861: {
2862: pcre_uchar c = *code;
2863: pcre_uchar base;
2864: const pcre_uchar *end;
2865: pcre_uint32 chr;
2866:
2867: #ifdef SUPPORT_UCP
2868: pcre_uint32 *clist_dest;
2869: const pcre_uint32 *clist_src;
2870: #else
2871: utf = utf; /* Suppress "unused parameter" compiler warning */
2872: #endif
2873:
2874: list[0] = c;
2875: list[1] = FALSE;
2876: code++;
2877:
2878: if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
2879: {
2880: base = get_repeat_base(c);
2881: c -= (base - OP_STAR);
2882:
2883: if (c == OP_UPTO || c == OP_MINUPTO || c == OP_EXACT || c == OP_POSUPTO)
2884: code += IMM2_SIZE;
2885:
2886: list[1] = (c != OP_PLUS && c != OP_MINPLUS && c != OP_EXACT && c != OP_POSPLUS);
2887:
2888: switch(base)
2889: {
2890: case OP_STAR:
2891: list[0] = OP_CHAR;
2892: break;
2893:
2894: case OP_STARI:
2895: list[0] = OP_CHARI;
2896: break;
2897:
2898: case OP_NOTSTAR:
2899: list[0] = OP_NOT;
2900: break;
2901:
2902: case OP_NOTSTARI:
2903: list[0] = OP_NOTI;
2904: break;
2905:
2906: case OP_TYPESTAR:
2907: list[0] = *code;
2908: code++;
2909: break;
2910: }
2911: c = list[0];
2912: }
2913:
2914: switch(c)
2915: {
2916: case OP_NOT_DIGIT:
2917: case OP_DIGIT:
2918: case OP_NOT_WHITESPACE:
2919: case OP_WHITESPACE:
2920: case OP_NOT_WORDCHAR:
2921: case OP_WORDCHAR:
2922: case OP_ANY:
2923: case OP_ALLANY:
2924: case OP_ANYNL:
2925: case OP_NOT_HSPACE:
2926: case OP_HSPACE:
2927: case OP_NOT_VSPACE:
2928: case OP_VSPACE:
2929: case OP_EXTUNI:
2930: case OP_EODN:
2931: case OP_EOD:
2932: case OP_DOLL:
2933: case OP_DOLLM:
2934: return code;
2935:
2936: case OP_CHAR:
2937: case OP_NOT:
2938: GETCHARINCTEST(chr, code);
2939: list[2] = chr;
2940: list[3] = NOTACHAR;
2941: return code;
2942:
2943: case OP_CHARI:
2944: case OP_NOTI:
2945: list[0] = (c == OP_CHARI) ? OP_CHAR : OP_NOT;
2946: GETCHARINCTEST(chr, code);
2947: list[2] = chr;
2948:
2949: #ifdef SUPPORT_UCP
2950: if (chr < 128 || (chr < 256 && !utf))
2951: list[3] = fcc[chr];
2952: else
2953: list[3] = UCD_OTHERCASE(chr);
2954: #elif defined SUPPORT_UTF || !defined COMPILE_PCRE8
2955: list[3] = (chr < 256) ? fcc[chr] : chr;
2956: #else
2957: list[3] = fcc[chr];
2958: #endif
2959:
2960: /* The othercase might be the same value. */
2961:
2962: if (chr == list[3])
2963: list[3] = NOTACHAR;
2964: else
2965: list[4] = NOTACHAR;
2966: return code;
2967:
2968: #ifdef SUPPORT_UCP
2969: case OP_PROP:
2970: case OP_NOTPROP:
2971: if (code[0] != PT_CLIST)
2972: {
2973: list[2] = code[0];
2974: list[3] = code[1];
2975: return code + 2;
2976: }
2977:
2978: /* Convert only if we have enough space. */
2979:
2980: clist_src = PRIV(ucd_caseless_sets) + code[1];
2981: clist_dest = list + 2;
2982: code += 2;
2983:
2984: do {
2985: if (clist_dest >= list + 8)
2986: {
2987: /* Early return if there is not enough space. This should never
2988: happen, since all clists are shorter than 5 character now. */
2989: list[2] = code[0];
2990: list[3] = code[1];
2991: return code;
2992: }
2993: *clist_dest++ = *clist_src;
2994: }
2995: while(*clist_src++ != NOTACHAR);
2996:
2997: /* All characters are stored. The terminating NOTACHAR
2998: is copied form the clist itself. */
2999:
3000: list[0] = (c == OP_PROP) ? OP_CHAR : OP_NOT;
3001: return code;
3002: #endif
3003:
3004: case OP_NCLASS:
3005: case OP_CLASS:
3006: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3007: case OP_XCLASS:
3008: if (c == OP_XCLASS)
3009: end = code + GET(code, 0) - 1;
3010: else
3011: #endif
3012: end = code + 32 / sizeof(pcre_uchar);
3013:
3014: switch(*end)
3015: {
3016: case OP_CRSTAR:
3017: case OP_CRMINSTAR:
3018: case OP_CRQUERY:
3019: case OP_CRMINQUERY:
3020: case OP_CRPOSSTAR:
3021: case OP_CRPOSQUERY:
3022: list[1] = TRUE;
3023: end++;
3024: break;
3025:
3026: case OP_CRPLUS:
3027: case OP_CRMINPLUS:
3028: case OP_CRPOSPLUS:
3029: end++;
3030: break;
3031:
3032: case OP_CRRANGE:
3033: case OP_CRMINRANGE:
3034: case OP_CRPOSRANGE:
3035: list[1] = (GET2(end, 1) == 0);
3036: end += 1 + 2 * IMM2_SIZE;
3037: break;
3038: }
3039: list[2] = end - code;
3040: return end;
3041: }
3042: return NULL; /* Opcode not accepted */
3043: }
3044:
3045:
3046:
3047: /*************************************************
3048: * Scan further character sets for match *
3049: *************************************************/
3050:
3051: /* Checks whether the base and the current opcode have a common character, in
3052: which case the base cannot be possessified.
3053:
3054: Arguments:
3055: code points to the byte code
3056: utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3057: cd static compile data
3058: base_list the data list of the base opcode
3059:
3060: Returns: TRUE if the auto-possessification is possible
3061: */
3062:
3063: static BOOL
3064: compare_opcodes(const pcre_uchar *code, BOOL utf, const compile_data *cd,
3065: const pcre_uint32 *base_list, const pcre_uchar *base_end)
3066: {
3067: pcre_uchar c;
3068: pcre_uint32 list[8];
3069: const pcre_uint32 *chr_ptr;
3070: const pcre_uint32 *ochr_ptr;
3071: const pcre_uint32 *list_ptr;
3072: const pcre_uchar *next_code;
3073: const pcre_uint8 *class_bitset;
3074: const pcre_uint32 *set1, *set2, *set_end;
3075: pcre_uint32 chr;
3076: BOOL accepted, invert_bits;
3077:
3078: /* Note: the base_list[1] contains whether the current opcode has greedy
3079: (represented by a non-zero value) quantifier. This is a different from
3080: other character type lists, which stores here that the character iterator
3081: matches to an empty string (also represented by a non-zero value). */
3082:
3083: for(;;)
3084: {
3085: /* All operations move the code pointer forward.
3086: Therefore infinite recursions are not possible. */
3087:
3088: c = *code;
3089:
3090: /* Skip over callouts */
3091:
3092: if (c == OP_CALLOUT)
3093: {
3094: code += PRIV(OP_lengths)[c];
3095: continue;
3096: }
3097:
3098: if (c == OP_ALT)
3099: {
3100: do code += GET(code, 1); while (*code == OP_ALT);
3101: c = *code;
3102: }
3103:
3104: switch(c)
3105: {
3106: case OP_END:
3107: case OP_KETRPOS:
3108: /* TRUE only in greedy case. The non-greedy case could be replaced by
3109: an OP_EXACT, but it is probably not worth it. (And note that OP_EXACT
3110: uses more memory, which we cannot get at this stage.) */
3111:
3112: return base_list[1] != 0;
3113:
3114: case OP_KET:
3115: /* If the bracket is capturing, and referenced by an OP_RECURSE, or
3116: it is an atomic sub-pattern (assert, once, etc.) the non-greedy case
3117: cannot be converted to a possessive form. */
3118:
3119: if (base_list[1] == 0) return FALSE;
3120:
3121: switch(*(code - GET(code, 1)))
3122: {
3123: case OP_ASSERT:
3124: case OP_ASSERT_NOT:
3125: case OP_ASSERTBACK:
3126: case OP_ASSERTBACK_NOT:
3127: case OP_ONCE:
3128: case OP_ONCE_NC:
3129: /* Atomic sub-patterns and assertions can always auto-possessify their
3130: last iterator. */
3131: return TRUE;
3132: }
3133:
3134: code += PRIV(OP_lengths)[c];
3135: continue;
3136:
3137: case OP_ONCE:
3138: case OP_ONCE_NC:
3139: case OP_BRA:
3140: case OP_CBRA:
3141: next_code = code + GET(code, 1);
3142: code += PRIV(OP_lengths)[c];
3143:
3144: while (*next_code == OP_ALT)
3145: {
3146: if (!compare_opcodes(code, utf, cd, base_list, base_end)) return FALSE;
3147: code = next_code + 1 + LINK_SIZE;
3148: next_code += GET(next_code, 1);
3149: }
3150: continue;
3151:
3152: case OP_BRAZERO:
3153: case OP_BRAMINZERO:
3154:
3155: next_code = code + 1;
3156: if (*next_code != OP_BRA && *next_code != OP_CBRA
3157: && *next_code != OP_ONCE && *next_code != OP_ONCE_NC) return FALSE;
3158:
3159: do next_code += GET(next_code, 1); while (*next_code == OP_ALT);
3160:
3161: /* The bracket content will be checked by the
3162: OP_BRA/OP_CBRA case above. */
3163: next_code += 1 + LINK_SIZE;
3164: if (!compare_opcodes(next_code, utf, cd, base_list, base_end))
3165: return FALSE;
3166:
3167: code += PRIV(OP_lengths)[c];
3168: continue;
3169: }
3170:
3171: /* Check for a supported opcode, and load its properties. */
3172:
3173: code = get_chr_property_list(code, utf, cd->fcc, list);
3174: if (code == NULL) return FALSE; /* Unsupported */
3175:
3176: /* If either opcode is a small character list, set pointers for comparing
3177: characters from that list with another list, or with a property. */
3178:
3179: if (base_list[0] == OP_CHAR)
3180: {
3181: chr_ptr = base_list + 2;
3182: list_ptr = list;
3183: }
3184: else if (list[0] == OP_CHAR)
3185: {
3186: chr_ptr = list + 2;
3187: list_ptr = base_list;
3188: }
3189:
3190: /* Character bitsets can also be compared to certain opcodes. */
3191:
3192: else if (base_list[0] == OP_CLASS || list[0] == OP_CLASS
3193: #ifdef COMPILE_PCRE8
3194: /* In 8 bit, non-UTF mode, OP_CLASS and OP_NCLASS are the same. */
3195: || (!utf && (base_list[0] == OP_NCLASS || list[0] == OP_NCLASS))
3196: #endif
3197: )
3198: {
3199: #ifdef COMPILE_PCRE8
3200: if (base_list[0] == OP_CLASS || (!utf && base_list[0] == OP_NCLASS))
3201: #else
3202: if (base_list[0] == OP_CLASS)
3203: #endif
3204: {
3205: set1 = (pcre_uint32 *)(base_end - base_list[2]);
3206: list_ptr = list;
3207: }
3208: else
3209: {
3210: set1 = (pcre_uint32 *)(code - list[2]);
3211: list_ptr = base_list;
3212: }
3213:
3214: invert_bits = FALSE;
3215: switch(list_ptr[0])
3216: {
3217: case OP_CLASS:
3218: case OP_NCLASS:
3219: set2 = (pcre_uint32 *)
3220: ((list_ptr == list ? code : base_end) - list_ptr[2]);
3221: break;
3222:
3223: /* OP_XCLASS cannot be supported here, because its bitset
3224: is not necessarily complete. E.g: [a-\0x{200}] is stored
3225: as a character range, and the appropriate bits are not set. */
3226:
3227: case OP_NOT_DIGIT:
3228: invert_bits = TRUE;
3229: /* Fall through */
3230: case OP_DIGIT:
3231: set2 = (pcre_uint32 *)(cd->cbits + cbit_digit);
3232: break;
3233:
3234: case OP_NOT_WHITESPACE:
3235: invert_bits = TRUE;
3236: /* Fall through */
3237: case OP_WHITESPACE:
3238: set2 = (pcre_uint32 *)(cd->cbits + cbit_space);
3239: break;
3240:
3241: case OP_NOT_WORDCHAR:
3242: invert_bits = TRUE;
3243: /* Fall through */
3244: case OP_WORDCHAR:
3245: set2 = (pcre_uint32 *)(cd->cbits + cbit_word);
3246: break;
3247:
3248: default:
3249: return FALSE;
3250: }
3251:
3252: /* Compare 4 bytes to improve speed. */
3253: set_end = set1 + (32 / 4);
3254: if (invert_bits)
3255: {
3256: do
3257: {
3258: if ((*set1++ & ~(*set2++)) != 0) return FALSE;
3259: }
3260: while (set1 < set_end);
3261: }
3262: else
3263: {
3264: do
3265: {
3266: if ((*set1++ & *set2++) != 0) return FALSE;
3267: }
3268: while (set1 < set_end);
3269: }
3270:
3271: if (list[1] == 0) return TRUE;
3272: /* Might be an empty repeat. */
3273: continue;
3274: }
3275:
3276: /* Some property combinations also acceptable. Unicode property opcodes are
3277: processed specially; the rest can be handled with a lookup table. */
3278:
3279: else
3280: {
3281: pcre_uint32 leftop, rightop;
3282:
3283: leftop = base_list[0];
3284: rightop = list[0];
3285:
3286: #ifdef SUPPORT_UCP
3287: accepted = FALSE; /* Always set in non-unicode case. */
3288: if (leftop == OP_PROP || leftop == OP_NOTPROP)
3289: {
3290: if (rightop == OP_EOD)
3291: accepted = TRUE;
3292: else if (rightop == OP_PROP || rightop == OP_NOTPROP)
3293: {
3294: int n;
3295: const pcre_uint8 *p;
3296: BOOL same = leftop == rightop;
3297: BOOL lisprop = leftop == OP_PROP;
3298: BOOL risprop = rightop == OP_PROP;
3299: BOOL bothprop = lisprop && risprop;
3300:
3301: /* There's a table that specifies how each combination is to be
3302: processed:
3303: 0 Always return FALSE (never auto-possessify)
3304: 1 Character groups are distinct (possessify if both are OP_PROP)
3305: 2 Check character categories in the same group (general or particular)
3306: 3 Return TRUE if the two opcodes are not the same
3307: ... see comments below
3308: */
3309:
3310: n = propposstab[base_list[2]][list[2]];
3311: switch(n)
3312: {
3313: case 0: break;
3314: case 1: accepted = bothprop; break;
3315: case 2: accepted = (base_list[3] == list[3]) != same; break;
3316: case 3: accepted = !same; break;
3317:
3318: case 4: /* Left general category, right particular category */
3319: accepted = risprop && catposstab[base_list[3]][list[3]] == same;
3320: break;
3321:
3322: case 5: /* Right general category, left particular category */
3323: accepted = lisprop && catposstab[list[3]][base_list[3]] == same;
3324: break;
3325:
3326: /* This code is logically tricky. Think hard before fiddling with it.
3327: The posspropstab table has four entries per row. Each row relates to
3328: one of PCRE's special properties such as ALNUM or SPACE or WORD.
3329: Only WORD actually needs all four entries, but using repeats for the
3330: others means they can all use the same code below.
3331:
3332: The first two entries in each row are Unicode general categories, and
3333: apply always, because all the characters they include are part of the
3334: PCRE character set. The third and fourth entries are a general and a
3335: particular category, respectively, that include one or more relevant
3336: characters. One or the other is used, depending on whether the check
3337: is for a general or a particular category. However, in both cases the
3338: category contains more characters than the specials that are defined
3339: for the property being tested against. Therefore, it cannot be used
3340: in a NOTPROP case.
3341:
3342: Example: the row for WORD contains ucp_L, ucp_N, ucp_P, ucp_Po.
3343: Underscore is covered by ucp_P or ucp_Po. */
3344:
3345: case 6: /* Left alphanum vs right general category */
3346: case 7: /* Left space vs right general category */
3347: case 8: /* Left word vs right general category */
3348: p = posspropstab[n-6];
3349: accepted = risprop && lisprop ==
3350: (list[3] != p[0] &&
3351: list[3] != p[1] &&
3352: (list[3] != p[2] || !lisprop));
3353: break;
3354:
3355: case 9: /* Right alphanum vs left general category */
3356: case 10: /* Right space vs left general category */
3357: case 11: /* Right word vs left general category */
3358: p = posspropstab[n-9];
3359: accepted = lisprop && risprop ==
3360: (base_list[3] != p[0] &&
3361: base_list[3] != p[1] &&
3362: (base_list[3] != p[2] || !risprop));
3363: break;
3364:
3365: case 12: /* Left alphanum vs right particular category */
3366: case 13: /* Left space vs right particular category */
3367: case 14: /* Left word vs right particular category */
3368: p = posspropstab[n-12];
3369: accepted = risprop && lisprop ==
3370: (catposstab[p[0]][list[3]] &&
3371: catposstab[p[1]][list[3]] &&
3372: (list[3] != p[3] || !lisprop));
3373: break;
3374:
3375: case 15: /* Right alphanum vs left particular category */
3376: case 16: /* Right space vs left particular category */
3377: case 17: /* Right word vs left particular category */
3378: p = posspropstab[n-15];
3379: accepted = lisprop && risprop ==
3380: (catposstab[p[0]][base_list[3]] &&
3381: catposstab[p[1]][base_list[3]] &&
3382: (base_list[3] != p[3] || !risprop));
3383: break;
3384: }
3385: }
3386: }
3387:
3388: else
3389: #endif /* SUPPORT_UCP */
3390:
3391: accepted = leftop >= FIRST_AUTOTAB_OP && leftop <= LAST_AUTOTAB_LEFT_OP &&
3392: rightop >= FIRST_AUTOTAB_OP && rightop <= LAST_AUTOTAB_RIGHT_OP &&
3393: autoposstab[leftop - FIRST_AUTOTAB_OP][rightop - FIRST_AUTOTAB_OP];
3394:
3395: if (!accepted)
3396: return FALSE;
3397:
3398: if (list[1] == 0) return TRUE;
3399: /* Might be an empty repeat. */
3400: continue;
3401: }
3402:
3403: /* Control reaches here only if one of the items is a small character list.
3404: All characters are checked against the other side. */
3405:
3406: do
3407: {
3408: chr = *chr_ptr;
3409:
3410: switch(list_ptr[0])
3411: {
3412: case OP_CHAR:
3413: ochr_ptr = list_ptr + 2;
3414: do
3415: {
3416: if (chr == *ochr_ptr) return FALSE;
3417: ochr_ptr++;
3418: }
3419: while(*ochr_ptr != NOTACHAR);
3420: break;
3421:
3422: case OP_NOT:
3423: ochr_ptr = list_ptr + 2;
3424: do
3425: {
3426: if (chr == *ochr_ptr)
3427: break;
3428: ochr_ptr++;
3429: }
3430: while(*ochr_ptr != NOTACHAR);
3431: if (*ochr_ptr == NOTACHAR) return FALSE; /* Not found */
3432: break;
3433:
3434: /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not*
3435: set. When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
3436:
3437: case OP_DIGIT:
3438: if (chr < 256 && (cd->ctypes[chr] & ctype_digit) != 0) return FALSE;
3439: break;
3440:
3441: case OP_NOT_DIGIT:
3442: if (chr > 255 || (cd->ctypes[chr] & ctype_digit) == 0) return FALSE;
3443: break;
3444:
3445: case OP_WHITESPACE:
3446: if (chr < 256 && (cd->ctypes[chr] & ctype_space) != 0) return FALSE;
3447: break;
3448:
3449: case OP_NOT_WHITESPACE:
3450: if (chr > 255 || (cd->ctypes[chr] & ctype_space) == 0) return FALSE;
3451: break;
3452:
3453: case OP_WORDCHAR:
3454: if (chr < 255 && (cd->ctypes[chr] & ctype_word) != 0) return FALSE;
3455: break;
3456:
3457: case OP_NOT_WORDCHAR:
3458: if (chr > 255 || (cd->ctypes[chr] & ctype_word) == 0) return FALSE;
3459: break;
3460:
3461: case OP_HSPACE:
3462: switch(chr)
3463: {
3464: HSPACE_CASES: return FALSE;
3465: default: break;
3466: }
3467: break;
3468:
3469: case OP_NOT_HSPACE:
3470: switch(chr)
3471: {
3472: HSPACE_CASES: break;
3473: default: return FALSE;
3474: }
3475: break;
3476:
3477: case OP_ANYNL:
3478: case OP_VSPACE:
3479: switch(chr)
3480: {
3481: VSPACE_CASES: return FALSE;
3482: default: break;
3483: }
3484: break;
3485:
3486: case OP_NOT_VSPACE:
3487: switch(chr)
3488: {
3489: VSPACE_CASES: break;
3490: default: return FALSE;
3491: }
3492: break;
3493:
3494: case OP_DOLL:
3495: case OP_EODN:
3496: switch (chr)
3497: {
3498: case CHAR_CR:
3499: case CHAR_LF:
3500: case CHAR_VT:
3501: case CHAR_FF:
3502: case CHAR_NEL:
3503: #ifndef EBCDIC
3504: case 0x2028:
3505: case 0x2029:
3506: #endif /* Not EBCDIC */
3507: return FALSE;
3508: }
3509: break;
3510:
3511: case OP_EOD: /* Can always possessify before \z */
3512: break;
3513:
3514: #ifdef SUPPORT_UCP
3515: case OP_PROP:
3516: case OP_NOTPROP:
3517: if (!check_char_prop(chr, list_ptr[2], list_ptr[3],
3518: list_ptr[0] == OP_NOTPROP))
3519: return FALSE;
3520: break;
3521: #endif
3522:
3523: case OP_NCLASS:
3524: if (chr > 255) return FALSE;
3525: /* Fall through */
3526:
3527: case OP_CLASS:
3528: if (chr > 255) break;
3529: class_bitset = (pcre_uint8 *)
3530: ((list_ptr == list ? code : base_end) - list_ptr[2]);
3531: if ((class_bitset[chr >> 3] & (1 << (chr & 7))) != 0) return FALSE;
3532: break;
3533:
3534: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3535: case OP_XCLASS:
3536: if (PRIV(xclass)(chr, (list_ptr == list ? code : base_end) -
3537: list_ptr[2] + LINK_SIZE, utf)) return FALSE;
3538: break;
3539: #endif
3540:
3541: default:
3542: return FALSE;
3543: }
3544:
3545: chr_ptr++;
3546: }
3547: while(*chr_ptr != NOTACHAR);
3548:
3549: /* At least one character must be matched from this opcode. */
3550:
3551: if (list[1] == 0) return TRUE;
3552: }
3553:
3554: return FALSE;
3555: }
3556:
3557:
3558:
3559: /*************************************************
3560: * Scan compiled regex for auto-possession *
3561: *************************************************/
3562:
3563: /* Replaces single character iterations with their possessive alternatives
3564: if appropriate. This function modifies the compiled opcode!
3565:
3566: Arguments:
3567: code points to start of the byte code
3568: utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3569: cd static compile data
3570:
3571: Returns: nothing
3572: */
3573:
3574: static void
3575: auto_possessify(pcre_uchar *code, BOOL utf, const compile_data *cd)
3576: {
3577: register pcre_uchar c;
3578: const pcre_uchar *end;
3579: pcre_uchar *repeat_opcode;
3580: pcre_uint32 list[8];
3581:
3582: for (;;)
3583: {
3584: c = *code;
3585:
3586: if (c >= OP_STAR && c <= OP_TYPEPOSUPTO)
3587: {
3588: c -= get_repeat_base(c) - OP_STAR;
3589: end = (c <= OP_MINUPTO) ?
3590: get_chr_property_list(code, utf, cd->fcc, list) : NULL;
3591: list[1] = c == OP_STAR || c == OP_PLUS || c == OP_QUERY || c == OP_UPTO;
3592:
3593: if (end != NULL && compare_opcodes(end, utf, cd, list, end))
3594: {
3595: switch(c)
3596: {
3597: case OP_STAR:
3598: *code += OP_POSSTAR - OP_STAR;
3599: break;
3600:
3601: case OP_MINSTAR:
3602: *code += OP_POSSTAR - OP_MINSTAR;
3603: break;
3604:
3605: case OP_PLUS:
3606: *code += OP_POSPLUS - OP_PLUS;
3607: break;
3608:
3609: case OP_MINPLUS:
3610: *code += OP_POSPLUS - OP_MINPLUS;
3611: break;
3612:
3613: case OP_QUERY:
3614: *code += OP_POSQUERY - OP_QUERY;
3615: break;
3616:
3617: case OP_MINQUERY:
3618: *code += OP_POSQUERY - OP_MINQUERY;
3619: break;
3620:
3621: case OP_UPTO:
3622: *code += OP_POSUPTO - OP_UPTO;
3623: break;
3624:
3625: case OP_MINUPTO:
3626: *code += OP_MINUPTO - OP_UPTO;
3627: break;
3628: }
3629: }
3630: c = *code;
3631: }
3632: else if (c == OP_CLASS || c == OP_NCLASS || c == OP_XCLASS)
3633: {
3634: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3635: if (c == OP_XCLASS)
3636: repeat_opcode = code + GET(code, 1);
3637: else
3638: #endif
3639: repeat_opcode = code + 1 + (32 / sizeof(pcre_uchar));
3640:
3641: c = *repeat_opcode;
3642: if (c >= OP_CRSTAR && c <= OP_CRMINRANGE)
3643: {
3644: /* end must not be NULL. */
3645: end = get_chr_property_list(code, utf, cd->fcc, list);
3646:
3647: list[1] = (c & 1) == 0;
3648:
3649: if (compare_opcodes(end, utf, cd, list, end))
3650: {
3651: switch (c)
3652: {
3653: case OP_CRSTAR:
3654: case OP_CRMINSTAR:
3655: *repeat_opcode = OP_CRPOSSTAR;
3656: break;
3657:
3658: case OP_CRPLUS:
3659: case OP_CRMINPLUS:
3660: *repeat_opcode = OP_CRPOSPLUS;
3661: break;
3662:
3663: case OP_CRQUERY:
3664: case OP_CRMINQUERY:
3665: *repeat_opcode = OP_CRPOSQUERY;
3666: break;
3667:
3668: case OP_CRRANGE:
3669: case OP_CRMINRANGE:
3670: *repeat_opcode = OP_CRPOSRANGE;
3671: break;
3672: }
3673: }
3674: }
3675: c = *code;
3676: }
3677:
3678: switch(c)
3679: {
3680: case OP_END:
3681: return;
3682:
3683: case OP_TYPESTAR:
3684: case OP_TYPEMINSTAR:
3685: case OP_TYPEPLUS:
3686: case OP_TYPEMINPLUS:
3687: case OP_TYPEQUERY:
3688: case OP_TYPEMINQUERY:
3689: case OP_TYPEPOSSTAR:
3690: case OP_TYPEPOSPLUS:
3691: case OP_TYPEPOSQUERY:
3692: if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
3693: break;
3694:
3695: case OP_TYPEUPTO:
3696: case OP_TYPEMINUPTO:
3697: case OP_TYPEEXACT:
3698: case OP_TYPEPOSUPTO:
3699: if (code[1 + IMM2_SIZE] == OP_PROP || code[1 + IMM2_SIZE] == OP_NOTPROP)
3700: code += 2;
3701: break;
3702:
3703: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
3704: case OP_XCLASS:
3705: code += GET(code, 1);
3706: break;
3707: #endif
3708:
3709: case OP_MARK:
3710: case OP_PRUNE_ARG:
3711: case OP_SKIP_ARG:
3712: case OP_THEN_ARG:
3713: code += code[1];
3714: break;
3715: }
3716:
3717: /* Add in the fixed length from the table */
3718:
3719: code += PRIV(OP_lengths)[c];
3720:
3721: /* In UTF-8 mode, opcodes that are followed by a character may be followed by
3722: a multi-byte character. The length in the table is a minimum, so we have to
3723: arrange to skip the extra bytes. */
3724:
3725: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
3726: if (utf) switch(c)
3727: {
3728: case OP_CHAR:
3729: case OP_CHARI:
3730: case OP_NOT:
3731: case OP_NOTI:
3732: case OP_STAR:
3733: case OP_MINSTAR:
3734: case OP_PLUS:
3735: case OP_MINPLUS:
3736: case OP_QUERY:
3737: case OP_MINQUERY:
3738: case OP_UPTO:
3739: case OP_MINUPTO:
3740: case OP_EXACT:
3741: case OP_POSSTAR:
3742: case OP_POSPLUS:
3743: case OP_POSQUERY:
3744: case OP_POSUPTO:
3745: case OP_STARI:
3746: case OP_MINSTARI:
3747: case OP_PLUSI:
3748: case OP_MINPLUSI:
3749: case OP_QUERYI:
3750: case OP_MINQUERYI:
3751: case OP_UPTOI:
3752: case OP_MINUPTOI:
3753: case OP_EXACTI:
3754: case OP_POSSTARI:
3755: case OP_POSPLUSI:
3756: case OP_POSQUERYI:
3757: case OP_POSUPTOI:
3758: case OP_NOTSTAR:
3759: case OP_NOTMINSTAR:
3760: case OP_NOTPLUS:
3761: case OP_NOTMINPLUS:
3762: case OP_NOTQUERY:
3763: case OP_NOTMINQUERY:
3764: case OP_NOTUPTO:
3765: case OP_NOTMINUPTO:
3766: case OP_NOTEXACT:
3767: case OP_NOTPOSSTAR:
3768: case OP_NOTPOSPLUS:
3769: case OP_NOTPOSQUERY:
3770: case OP_NOTPOSUPTO:
3771: case OP_NOTSTARI:
3772: case OP_NOTMINSTARI:
3773: case OP_NOTPLUSI:
3774: case OP_NOTMINPLUSI:
3775: case OP_NOTQUERYI:
3776: case OP_NOTMINQUERYI:
3777: case OP_NOTUPTOI:
3778: case OP_NOTMINUPTOI:
3779: case OP_NOTEXACTI:
3780: case OP_NOTPOSSTARI:
3781: case OP_NOTPOSPLUSI:
3782: case OP_NOTPOSQUERYI:
3783: case OP_NOTPOSUPTOI:
3784: if (HAS_EXTRALEN(code[-1])) code += GET_EXTRALEN(code[-1]);
3785: break;
3786: }
3787: #else
3788: (void)(utf); /* Keep compiler happy by referencing function argument */
3789: #endif
3790: }
3791: }
3792:
3793:
3794:
3795: /*************************************************
3796: * Check for POSIX class syntax *
3797: *************************************************/
3798:
3799: /* This function is called when the sequence "[:" or "[." or "[=" is
3800: encountered in a character class. It checks whether this is followed by a
3801: sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
3802: reach an unescaped ']' without the special preceding character, return FALSE.
3803:
3804: Originally, this function only recognized a sequence of letters between the
3805: terminators, but it seems that Perl recognizes any sequence of characters,
3806: though of course unknown POSIX names are subsequently rejected. Perl gives an
3807: "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
3808: didn't consider this to be a POSIX class. Likewise for [:1234:].
3809:
3810: The problem in trying to be exactly like Perl is in the handling of escapes. We
3811: have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
3812: class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
3813: below handles the special case of \], but does not try to do any other escape
3814: processing. This makes it different from Perl for cases such as [:l\ower:]
3815: where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
3816: "l\ower". This is a lesser evil than not diagnosing bad classes when Perl does,
3817: I think.
3818:
3819: A user pointed out that PCRE was rejecting [:a[:digit:]] whereas Perl was not.
3820: It seems that the appearance of a nested POSIX class supersedes an apparent
3821: external class. For example, [:a[:digit:]b:] matches "a", "b", ":", or
3822: a digit.
3823:
3824: In Perl, unescaped square brackets may also appear as part of class names. For
3825: example, [:a[:abc]b:] gives unknown POSIX class "[:abc]b:]". However, for
3826: [:a[:abc]b][b:] it gives unknown POSIX class "[:abc]b][b:]", which does not
3827: seem right at all. PCRE does not allow closing square brackets in POSIX class
3828: names.
3829:
3830: Arguments:
3831: ptr pointer to the initial [
3832: endptr where to return the end pointer
3833:
3834: Returns: TRUE or FALSE
3835: */
3836:
3837: static BOOL
3838: check_posix_syntax(const pcre_uchar *ptr, const pcre_uchar **endptr)
3839: {
3840: pcre_uchar terminator; /* Don't combine these lines; the Solaris cc */
3841: terminator = *(++ptr); /* compiler warns about "non-constant" initializer. */
3842: for (++ptr; *ptr != CHAR_NULL; ptr++)
3843: {
3844: if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3845: ptr++;
3846: else if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
3847: else
3848: {
3849: if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
3850: {
3851: *endptr = ptr;
3852: return TRUE;
3853: }
3854: if (*ptr == CHAR_LEFT_SQUARE_BRACKET &&
3855: (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3856: ptr[1] == CHAR_EQUALS_SIGN) &&
3857: check_posix_syntax(ptr, endptr))
3858: return FALSE;
3859: }
3860: }
3861: return FALSE;
3862: }
3863:
3864:
3865:
3866:
3867: /*************************************************
3868: * Check POSIX class name *
3869: *************************************************/
3870:
3871: /* This function is called to check the name given in a POSIX-style class entry
3872: such as [:alnum:].
3873:
3874: Arguments:
3875: ptr points to the first letter
3876: len the length of the name
3877:
3878: Returns: a value representing the name, or -1 if unknown
3879: */
3880:
3881: static int
3882: check_posix_name(const pcre_uchar *ptr, int len)
3883: {
3884: const char *pn = posix_names;
3885: register int yield = 0;
3886: while (posix_name_lengths[yield] != 0)
3887: {
3888: if (len == posix_name_lengths[yield] &&
3889: STRNCMP_UC_C8(ptr, pn, (unsigned int)len) == 0) return yield;
3890: pn += posix_name_lengths[yield] + 1;
3891: yield++;
3892: }
3893: return -1;
3894: }
3895:
3896:
3897: /*************************************************
3898: * Adjust OP_RECURSE items in repeated group *
3899: *************************************************/
3900:
3901: /* OP_RECURSE items contain an offset from the start of the regex to the group
3902: that is referenced. This means that groups can be replicated for fixed
3903: repetition simply by copying (because the recursion is allowed to refer to
3904: earlier groups that are outside the current group). However, when a group is
3905: optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
3906: inserted before it, after it has been compiled. This means that any OP_RECURSE
3907: items within it that refer to the group itself or any contained groups have to
3908: have their offsets adjusted. That one of the jobs of this function. Before it
3909: is called, the partially compiled regex must be temporarily terminated with
3910: OP_END.
3911:
3912: This function has been extended with the possibility of forward references for
3913: recursions and subroutine calls. It must also check the list of such references
3914: for the group we are dealing with. If it finds that one of the recursions in
3915: the current group is on this list, it adjusts the offset in the list, not the
3916: value in the reference (which is a group number).
3917:
3918: Arguments:
3919: group points to the start of the group
3920: adjust the amount by which the group is to be moved
3921: utf TRUE in UTF-8 / UTF-16 / UTF-32 mode
3922: cd contains pointers to tables etc.
3923: save_hwm the hwm forward reference pointer at the start of the group
3924:
3925: Returns: nothing
3926: */
3927:
3928: static void
3929: adjust_recurse(pcre_uchar *group, int adjust, BOOL utf, compile_data *cd,
3930: pcre_uchar *save_hwm)
3931: {
3932: pcre_uchar *ptr = group;
3933:
3934: while ((ptr = (pcre_uchar *)find_recurse(ptr, utf)) != NULL)
3935: {
3936: int offset;
3937: pcre_uchar *hc;
3938:
3939: /* See if this recursion is on the forward reference list. If so, adjust the
3940: reference. */
3941:
3942: for (hc = save_hwm; hc < cd->hwm; hc += LINK_SIZE)
3943: {
3944: offset = (int)GET(hc, 0);
3945: if (cd->start_code + offset == ptr + 1)
3946: {
3947: PUT(hc, 0, offset + adjust);
3948: break;
3949: }
3950: }
3951:
3952: /* Otherwise, adjust the recursion offset if it's after the start of this
3953: group. */
3954:
3955: if (hc >= cd->hwm)
3956: {
3957: offset = (int)GET(ptr, 1);
3958: if (cd->start_code + offset >= group) PUT(ptr, 1, offset + adjust);
3959: }
3960:
3961: ptr += 1 + LINK_SIZE;
3962: }
3963: }
3964:
3965:
3966:
3967: /*************************************************
3968: * Insert an automatic callout point *
3969: *************************************************/
3970:
3971: /* This function is called when the PCRE_AUTO_CALLOUT option is set, to insert
3972: callout points before each pattern item.
3973:
3974: Arguments:
3975: code current code pointer
3976: ptr current pattern pointer
3977: cd pointers to tables etc
3978:
3979: Returns: new code pointer
3980: */
3981:
3982: static pcre_uchar *
3983: auto_callout(pcre_uchar *code, const pcre_uchar *ptr, compile_data *cd)
3984: {
3985: *code++ = OP_CALLOUT;
3986: *code++ = 255;
3987: PUT(code, 0, (int)(ptr - cd->start_pattern)); /* Pattern offset */
3988: PUT(code, LINK_SIZE, 0); /* Default length */
3989: return code + 2 * LINK_SIZE;
3990: }
3991:
3992:
3993:
3994: /*************************************************
3995: * Complete a callout item *
3996: *************************************************/
3997:
3998: /* A callout item contains the length of the next item in the pattern, which
3999: we can't fill in till after we have reached the relevant point. This is used
4000: for both automatic and manual callouts.
4001:
4002: Arguments:
4003: previous_callout points to previous callout item
4004: ptr current pattern pointer
4005: cd pointers to tables etc
4006:
4007: Returns: nothing
4008: */
4009:
4010: static void
4011: complete_callout(pcre_uchar *previous_callout, const pcre_uchar *ptr, compile_data *cd)
4012: {
4013: int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
4014: PUT(previous_callout, 2 + LINK_SIZE, length);
4015: }
4016:
4017:
4018:
4019: #ifdef SUPPORT_UCP
4020: /*************************************************
4021: * Get othercase range *
4022: *************************************************/
4023:
4024: /* This function is passed the start and end of a class range, in UTF-8 mode
4025: with UCP support. It searches up the characters, looking for ranges of
4026: characters in the "other" case. Each call returns the next one, updating the
4027: start address. A character with multiple other cases is returned on its own
4028: with a special return value.
4029:
4030: Arguments:
4031: cptr points to starting character value; updated
4032: d end value
4033: ocptr where to put start of othercase range
4034: odptr where to put end of othercase range
4035:
4036: Yield: -1 when no more
4037: 0 when a range is returned
4038: >0 the CASESET offset for char with multiple other cases
4039: in this case, ocptr contains the original
4040: */
4041:
4042: static int
4043: get_othercase_range(pcre_uint32 *cptr, pcre_uint32 d, pcre_uint32 *ocptr,
4044: pcre_uint32 *odptr)
4045: {
4046: pcre_uint32 c, othercase, next;
4047: unsigned int co;
4048:
4049: /* Find the first character that has an other case. If it has multiple other
4050: cases, return its case offset value. */
4051:
4052: for (c = *cptr; c <= d; c++)
4053: {
4054: if ((co = UCD_CASESET(c)) != 0)
4055: {
4056: *ocptr = c++; /* Character that has the set */
4057: *cptr = c; /* Rest of input range */
4058: return (int)co;
4059: }
4060: if ((othercase = UCD_OTHERCASE(c)) != c) break;
4061: }
4062:
4063: if (c > d) return -1; /* Reached end of range */
4064:
4065: *ocptr = othercase;
4066: next = othercase + 1;
4067:
4068: for (++c; c <= d; c++)
4069: {
4070: if (UCD_OTHERCASE(c) != next) break;
4071: next++;
4072: }
4073:
4074: *odptr = next - 1; /* End of othercase range */
4075: *cptr = c; /* Rest of input range */
4076: return 0;
4077: }
4078: #endif /* SUPPORT_UCP */
4079:
4080:
4081:
4082: /*************************************************
4083: * Add a character or range to a class *
4084: *************************************************/
4085:
4086: /* This function packages up the logic of adding a character or range of
4087: characters to a class. The character values in the arguments will be within the
4088: valid values for the current mode (8-bit, 16-bit, UTF, etc). This function is
4089: mutually recursive with the function immediately below.
4090:
4091: Arguments:
4092: classbits the bit map for characters < 256
4093: uchardptr points to the pointer for extra data
4094: options the options word
4095: cd contains pointers to tables etc.
4096: start start of range character
4097: end end of range character
4098:
4099: Returns: the number of < 256 characters added
4100: the pointer to extra data is updated
4101: */
4102:
4103: static int
4104: add_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4105: compile_data *cd, pcre_uint32 start, pcre_uint32 end)
4106: {
4107: pcre_uint32 c;
4108: int n8 = 0;
4109:
4110: /* If caseless matching is required, scan the range and process alternate
4111: cases. In Unicode, there are 8-bit characters that have alternate cases that
4112: are greater than 255 and vice-versa. Sometimes we can just extend the original
4113: range. */
4114:
4115: if ((options & PCRE_CASELESS) != 0)
4116: {
4117: #ifdef SUPPORT_UCP
4118: if ((options & PCRE_UTF8) != 0)
4119: {
4120: int rc;
4121: pcre_uint32 oc, od;
4122:
4123: options &= ~PCRE_CASELESS; /* Remove for recursive calls */
4124: c = start;
4125:
4126: while ((rc = get_othercase_range(&c, end, &oc, &od)) >= 0)
4127: {
4128: /* Handle a single character that has more than one other case. */
4129:
4130: if (rc > 0) n8 += add_list_to_class(classbits, uchardptr, options, cd,
4131: PRIV(ucd_caseless_sets) + rc, oc);
4132:
4133: /* Do nothing if the other case range is within the original range. */
4134:
4135: else if (oc >= start && od <= end) continue;
4136:
4137: /* Extend the original range if there is overlap, noting that if oc < c, we
4138: can't have od > end because a subrange is always shorter than the basic
4139: range. Otherwise, use a recursive call to add the additional range. */
4140:
4141: else if (oc < start && od >= start - 1) start = oc; /* Extend downwards */
4142: else if (od > end && oc <= end + 1) end = od; /* Extend upwards */
4143: else n8 += add_to_class(classbits, uchardptr, options, cd, oc, od);
4144: }
4145: }
4146: else
4147: #endif /* SUPPORT_UCP */
4148:
4149: /* Not UTF-mode, or no UCP */
4150:
4151: for (c = start; c <= end && c < 256; c++)
4152: {
4153: SETBIT(classbits, cd->fcc[c]);
4154: n8++;
4155: }
4156: }
4157:
4158: /* Now handle the original range. Adjust the final value according to the bit
4159: length - this means that the same lists of (e.g.) horizontal spaces can be used
4160: in all cases. */
4161:
4162: #if defined COMPILE_PCRE8
4163: #ifdef SUPPORT_UTF
4164: if ((options & PCRE_UTF8) == 0)
4165: #endif
4166: if (end > 0xff) end = 0xff;
4167:
4168: #elif defined COMPILE_PCRE16
4169: #ifdef SUPPORT_UTF
4170: if ((options & PCRE_UTF16) == 0)
4171: #endif
4172: if (end > 0xffff) end = 0xffff;
4173:
4174: #endif /* COMPILE_PCRE[8|16] */
4175:
4176: /* If all characters are less than 256, use the bit map. Otherwise use extra
4177: data. */
4178:
4179: if (end < 0x100)
4180: {
4181: for (c = start; c <= end; c++)
4182: {
4183: n8++;
4184: SETBIT(classbits, c);
4185: }
4186: }
4187:
4188: else
4189: {
4190: pcre_uchar *uchardata = *uchardptr;
4191:
4192: #ifdef SUPPORT_UTF
4193: if ((options & PCRE_UTF8) != 0) /* All UTFs use the same flag bit */
4194: {
4195: if (start < end)
4196: {
4197: *uchardata++ = XCL_RANGE;
4198: uchardata += PRIV(ord2utf)(start, uchardata);
4199: uchardata += PRIV(ord2utf)(end, uchardata);
4200: }
4201: else if (start == end)
4202: {
4203: *uchardata++ = XCL_SINGLE;
4204: uchardata += PRIV(ord2utf)(start, uchardata);
4205: }
4206: }
4207: else
4208: #endif /* SUPPORT_UTF */
4209:
4210: /* Without UTF support, character values are constrained by the bit length,
4211: and can only be > 256 for 16-bit and 32-bit libraries. */
4212:
4213: #ifdef COMPILE_PCRE8
4214: {}
4215: #else
4216: if (start < end)
4217: {
4218: *uchardata++ = XCL_RANGE;
4219: *uchardata++ = start;
4220: *uchardata++ = end;
4221: }
4222: else if (start == end)
4223: {
4224: *uchardata++ = XCL_SINGLE;
4225: *uchardata++ = start;
4226: }
4227: #endif
4228:
4229: *uchardptr = uchardata; /* Updata extra data pointer */
4230: }
4231:
4232: return n8; /* Number of 8-bit characters */
4233: }
4234:
4235:
4236:
4237:
4238: /*************************************************
4239: * Add a list of characters to a class *
4240: *************************************************/
4241:
4242: /* This function is used for adding a list of case-equivalent characters to a
4243: class, and also for adding a list of horizontal or vertical whitespace. If the
4244: list is in order (which it should be), ranges of characters are detected and
4245: handled appropriately. This function is mutually recursive with the function
4246: above.
4247:
4248: Arguments:
4249: classbits the bit map for characters < 256
4250: uchardptr points to the pointer for extra data
4251: options the options word
4252: cd contains pointers to tables etc.
4253: p points to row of 32-bit values, terminated by NOTACHAR
4254: except character to omit; this is used when adding lists of
4255: case-equivalent characters to avoid including the one we
4256: already know about
4257:
4258: Returns: the number of < 256 characters added
4259: the pointer to extra data is updated
4260: */
4261:
4262: static int
4263: add_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr, int options,
4264: compile_data *cd, const pcre_uint32 *p, unsigned int except)
4265: {
4266: int n8 = 0;
4267: while (p[0] < NOTACHAR)
4268: {
4269: int n = 0;
4270: if (p[0] != except)
4271: {
4272: while(p[n+1] == p[0] + n + 1) n++;
4273: n8 += add_to_class(classbits, uchardptr, options, cd, p[0], p[n]);
4274: }
4275: p += n + 1;
4276: }
4277: return n8;
4278: }
4279:
4280:
4281:
4282: /*************************************************
4283: * Add characters not in a list to a class *
4284: *************************************************/
4285:
4286: /* This function is used for adding the complement of a list of horizontal or
4287: vertical whitespace to a class. The list must be in order.
4288:
4289: Arguments:
4290: classbits the bit map for characters < 256
4291: uchardptr points to the pointer for extra data
4292: options the options word
4293: cd contains pointers to tables etc.
4294: p points to row of 32-bit values, terminated by NOTACHAR
4295:
4296: Returns: the number of < 256 characters added
4297: the pointer to extra data is updated
4298: */
4299:
4300: static int
4301: add_not_list_to_class(pcre_uint8 *classbits, pcre_uchar **uchardptr,
4302: int options, compile_data *cd, const pcre_uint32 *p)
4303: {
4304: BOOL utf = (options & PCRE_UTF8) != 0;
4305: int n8 = 0;
4306: if (p[0] > 0)
4307: n8 += add_to_class(classbits, uchardptr, options, cd, 0, p[0] - 1);
4308: while (p[0] < NOTACHAR)
4309: {
4310: while (p[1] == p[0] + 1) p++;
4311: n8 += add_to_class(classbits, uchardptr, options, cd, p[0] + 1,
4312: (p[1] == NOTACHAR) ? (utf ? 0x10ffffu : 0xffffffffu) : p[1] - 1);
4313: p++;
4314: }
4315: return n8;
4316: }
4317:
4318:
4319:
4320: /*************************************************
4321: * Compile one branch *
4322: *************************************************/
4323:
4324: /* Scan the pattern, compiling it into the a vector. If the options are
4325: changed during the branch, the pointer is used to change the external options
4326: bits. This function is used during the pre-compile phase when we are trying
4327: to find out the amount of memory needed, as well as during the real compile
4328: phase. The value of lengthptr distinguishes the two phases.
4329:
4330: Arguments:
4331: optionsptr pointer to the option bits
4332: codeptr points to the pointer to the current code point
4333: ptrptr points to the current pattern pointer
4334: errorcodeptr points to error code variable
4335: firstcharptr place to put the first required character
4336: firstcharflagsptr place to put the first character flags, or a negative number
4337: reqcharptr place to put the last required character
4338: reqcharflagsptr place to put the last required character flags, or a negative number
4339: bcptr points to current branch chain
4340: cond_depth conditional nesting depth
4341: cd contains pointers to tables etc.
4342: lengthptr NULL during the real compile phase
4343: points to length accumulator during pre-compile phase
4344:
4345: Returns: TRUE on success
4346: FALSE, with *errorcodeptr set non-zero on error
4347: */
4348:
4349: static BOOL
4350: compile_branch(int *optionsptr, pcre_uchar **codeptr,
4351: const pcre_uchar **ptrptr, int *errorcodeptr,
4352: pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
4353: pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
4354: branch_chain *bcptr, int cond_depth,
4355: compile_data *cd, int *lengthptr)
4356: {
4357: int repeat_type, op_type;
4358: int repeat_min = 0, repeat_max = 0; /* To please picky compilers */
4359: int bravalue = 0;
4360: int greedy_default, greedy_non_default;
4361: pcre_uint32 firstchar, reqchar;
4362: pcre_int32 firstcharflags, reqcharflags;
4363: pcre_uint32 zeroreqchar, zerofirstchar;
4364: pcre_int32 zeroreqcharflags, zerofirstcharflags;
4365: pcre_int32 req_caseopt, reqvary, tempreqvary;
4366: int options = *optionsptr; /* May change dynamically */
4367: int after_manual_callout = 0;
4368: int length_prevgroup = 0;
4369: register pcre_uint32 c;
4370: int escape;
4371: register pcre_uchar *code = *codeptr;
4372: pcre_uchar *last_code = code;
4373: pcre_uchar *orig_code = code;
4374: pcre_uchar *tempcode;
4375: BOOL inescq = FALSE;
4376: BOOL groupsetfirstchar = FALSE;
4377: const pcre_uchar *ptr = *ptrptr;
4378: const pcre_uchar *tempptr;
4379: const pcre_uchar *nestptr = NULL;
4380: pcre_uchar *previous = NULL;
4381: pcre_uchar *previous_callout = NULL;
4382: pcre_uchar *save_hwm = NULL;
4383: pcre_uint8 classbits[32];
4384:
4385: /* We can fish out the UTF-8 setting once and for all into a BOOL, but we
4386: must not do this for other options (e.g. PCRE_EXTENDED) because they may change
4387: dynamically as we process the pattern. */
4388:
4389: #ifdef SUPPORT_UTF
4390: /* PCRE_UTF[16|32] have the same value as PCRE_UTF8. */
4391: BOOL utf = (options & PCRE_UTF8) != 0;
4392: #ifndef COMPILE_PCRE32
4393: pcre_uchar utf_chars[6];
4394: #endif
4395: #else
4396: BOOL utf = FALSE;
4397: #endif
4398:
4399: /* Helper variables for OP_XCLASS opcode (for characters > 255). We define
4400: class_uchardata always so that it can be passed to add_to_class() always,
4401: though it will not be used in non-UTF 8-bit cases. This avoids having to supply
4402: alternative calls for the different cases. */
4403:
4404: pcre_uchar *class_uchardata;
4405: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4406: BOOL xclass;
4407: pcre_uchar *class_uchardata_base;
4408: #endif
4409:
4410: #ifdef PCRE_DEBUG
4411: if (lengthptr != NULL) DPRINTF((">> start branch\n"));
4412: #endif
4413:
4414: /* Set up the default and non-default settings for greediness */
4415:
4416: greedy_default = ((options & PCRE_UNGREEDY) != 0);
4417: greedy_non_default = greedy_default ^ 1;
4418:
4419: /* Initialize no first byte, no required byte. REQ_UNSET means "no char
4420: matching encountered yet". It gets changed to REQ_NONE if we hit something that
4421: matches a non-fixed char first char; reqchar just remains unset if we never
4422: find one.
4423:
4424: When we hit a repeat whose minimum is zero, we may have to adjust these values
4425: to take the zero repeat into account. This is implemented by setting them to
4426: zerofirstbyte and zeroreqchar when such a repeat is encountered. The individual
4427: item types that can be repeated set these backoff variables appropriately. */
4428:
4429: firstchar = reqchar = zerofirstchar = zeroreqchar = 0;
4430: firstcharflags = reqcharflags = zerofirstcharflags = zeroreqcharflags = REQ_UNSET;
4431:
4432: /* The variable req_caseopt contains either the REQ_CASELESS value
4433: or zero, according to the current setting of the caseless flag. The
4434: REQ_CASELESS leaves the lower 28 bit empty. It is added into the
4435: firstchar or reqchar variables to record the case status of the
4436: value. This is used only for ASCII characters. */
4437:
4438: req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
4439:
4440: /* Switch on next character until the end of the branch */
4441:
4442: for (;; ptr++)
4443: {
4444: BOOL negate_class;
4445: BOOL should_flip_negation;
4446: BOOL possessive_quantifier;
4447: BOOL is_quantifier;
4448: BOOL is_recurse;
4449: BOOL reset_bracount;
4450: int class_has_8bitchar;
4451: int class_one_char;
4452: int newoptions;
4453: int recno;
4454: int refsign;
4455: int skipbytes;
4456: pcre_uint32 subreqchar, subfirstchar;
4457: pcre_int32 subreqcharflags, subfirstcharflags;
4458: int terminator;
4459: unsigned int mclength;
4460: unsigned int tempbracount;
4461: pcre_uint32 ec;
4462: pcre_uchar mcbuffer[8];
4463:
4464: /* Get next character in the pattern */
4465:
4466: c = *ptr;
4467:
4468: /* If we are at the end of a nested substitution, revert to the outer level
4469: string. Nesting only happens one level deep. */
4470:
4471: if (c == CHAR_NULL && nestptr != NULL)
4472: {
4473: ptr = nestptr;
4474: nestptr = NULL;
4475: c = *ptr;
4476: }
4477:
4478: /* If we are in the pre-compile phase, accumulate the length used for the
4479: previous cycle of this loop. */
4480:
4481: if (lengthptr != NULL)
4482: {
4483: #ifdef PCRE_DEBUG
4484: if (code > cd->hwm) cd->hwm = code; /* High water info */
4485: #endif
4486: if (code > cd->start_workspace + cd->workspace_size -
4487: WORK_SIZE_SAFETY_MARGIN) /* Check for overrun */
4488: {
4489: *errorcodeptr = ERR52;
4490: goto FAILED;
4491: }
4492:
4493: /* There is at least one situation where code goes backwards: this is the
4494: case of a zero quantifier after a class (e.g. [ab]{0}). At compile time,
4495: the class is simply eliminated. However, it is created first, so we have to
4496: allow memory for it. Therefore, don't ever reduce the length at this point.
4497: */
4498:
4499: if (code < last_code) code = last_code;
4500:
4501: /* Paranoid check for integer overflow */
4502:
4503: if (OFLOW_MAX - *lengthptr < code - last_code)
4504: {
4505: *errorcodeptr = ERR20;
4506: goto FAILED;
4507: }
4508:
4509: *lengthptr += (int)(code - last_code);
4510: DPRINTF(("length=%d added %d c=%c (0x%x)\n", *lengthptr,
4511: (int)(code - last_code), c, c));
4512:
4513: /* If "previous" is set and it is not at the start of the work space, move
4514: it back to there, in order to avoid filling up the work space. Otherwise,
4515: if "previous" is NULL, reset the current code pointer to the start. */
4516:
4517: if (previous != NULL)
4518: {
4519: if (previous > orig_code)
4520: {
4521: memmove(orig_code, previous, IN_UCHARS(code - previous));
4522: code -= previous - orig_code;
4523: previous = orig_code;
4524: }
4525: }
4526: else code = orig_code;
4527:
4528: /* Remember where this code item starts so we can pick up the length
4529: next time round. */
4530:
4531: last_code = code;
4532: }
4533:
4534: /* In the real compile phase, just check the workspace used by the forward
4535: reference list. */
4536:
4537: else if (cd->hwm > cd->start_workspace + cd->workspace_size -
4538: WORK_SIZE_SAFETY_MARGIN)
4539: {
4540: *errorcodeptr = ERR52;
4541: goto FAILED;
4542: }
4543:
4544: /* If in \Q...\E, check for the end; if not, we have a literal */
4545:
4546: if (inescq && c != CHAR_NULL)
4547: {
4548: if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
4549: {
4550: inescq = FALSE;
4551: ptr++;
4552: continue;
4553: }
4554: else
4555: {
4556: if (previous_callout != NULL)
4557: {
4558: if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4559: complete_callout(previous_callout, ptr, cd);
4560: previous_callout = NULL;
4561: }
4562: if ((options & PCRE_AUTO_CALLOUT) != 0)
4563: {
4564: previous_callout = code;
4565: code = auto_callout(code, ptr, cd);
4566: }
4567: goto NORMAL_CHAR;
4568: }
4569: /* Control does not reach here. */
4570: }
4571:
4572: /* In extended mode, skip white space and comments. We need a loop in order
4573: to check for more white space and more comments after a comment. */
4574:
4575: if ((options & PCRE_EXTENDED) != 0)
4576: {
4577: for (;;)
4578: {
4579: while (MAX_255(c) && (cd->ctypes[c] & ctype_space) != 0) c = *(++ptr);
4580: if (c != CHAR_NUMBER_SIGN) break;
4581: ptr++;
4582: while (*ptr != CHAR_NULL)
4583: {
4584: if (IS_NEWLINE(ptr)) /* For non-fixed-length newline cases, */
4585: { /* IS_NEWLINE sets cd->nllen. */
4586: ptr += cd->nllen;
4587: break;
4588: }
4589: ptr++;
4590: #ifdef SUPPORT_UTF
4591: if (utf) FORWARDCHAR(ptr);
4592: #endif
4593: }
4594: c = *ptr; /* Either NULL or the char after a newline */
4595: }
4596: }
4597:
4598: /* See if the next thing is a quantifier. */
4599:
4600: is_quantifier =
4601: c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
4602: (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
4603:
4604: /* Fill in length of a previous callout, except when the next thing is a
4605: quantifier or when processing a property substitution string in UCP mode. */
4606:
4607: if (!is_quantifier && previous_callout != NULL && nestptr == NULL &&
4608: after_manual_callout-- <= 0)
4609: {
4610: if (lengthptr == NULL) /* Don't attempt in pre-compile phase */
4611: complete_callout(previous_callout, ptr, cd);
4612: previous_callout = NULL;
4613: }
4614:
4615: /* Create auto callout, except for quantifiers, or while processing property
4616: strings that are substituted for \w etc in UCP mode. */
4617:
4618: if ((options & PCRE_AUTO_CALLOUT) != 0 && !is_quantifier && nestptr == NULL)
4619: {
4620: previous_callout = code;
4621: code = auto_callout(code, ptr, cd);
4622: }
4623:
4624: /* Process the next pattern item. */
4625:
4626: switch(c)
4627: {
4628: /* ===================================================================*/
4629: case CHAR_NULL: /* The branch terminates at string end */
4630: case CHAR_VERTICAL_LINE: /* or | or ) */
4631: case CHAR_RIGHT_PARENTHESIS:
4632: *firstcharptr = firstchar;
4633: *firstcharflagsptr = firstcharflags;
4634: *reqcharptr = reqchar;
4635: *reqcharflagsptr = reqcharflags;
4636: *codeptr = code;
4637: *ptrptr = ptr;
4638: if (lengthptr != NULL)
4639: {
4640: if (OFLOW_MAX - *lengthptr < code - last_code)
4641: {
4642: *errorcodeptr = ERR20;
4643: goto FAILED;
4644: }
4645: *lengthptr += (int)(code - last_code); /* To include callout length */
4646: DPRINTF((">> end branch\n"));
4647: }
4648: return TRUE;
4649:
4650:
4651: /* ===================================================================*/
4652: /* Handle single-character metacharacters. In multiline mode, ^ disables
4653: the setting of any following char as a first character. */
4654:
4655: case CHAR_CIRCUMFLEX_ACCENT:
4656: previous = NULL;
4657: if ((options & PCRE_MULTILINE) != 0)
4658: {
4659: if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4660: *code++ = OP_CIRCM;
4661: }
4662: else *code++ = OP_CIRC;
4663: break;
4664:
4665: case CHAR_DOLLAR_SIGN:
4666: previous = NULL;
4667: *code++ = ((options & PCRE_MULTILINE) != 0)? OP_DOLLM : OP_DOLL;
4668: break;
4669:
4670: /* There can never be a first char if '.' is first, whatever happens about
4671: repeats. The value of reqchar doesn't change either. */
4672:
4673: case CHAR_DOT:
4674: if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4675: zerofirstchar = firstchar;
4676: zerofirstcharflags = firstcharflags;
4677: zeroreqchar = reqchar;
4678: zeroreqcharflags = reqcharflags;
4679: previous = code;
4680: *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
4681: break;
4682:
4683:
4684: /* ===================================================================*/
4685: /* Character classes. If the included characters are all < 256, we build a
4686: 32-byte bitmap of the permitted characters, except in the special case
4687: where there is only one such character. For negated classes, we build the
4688: map as usual, then invert it at the end. However, we use a different opcode
4689: so that data characters > 255 can be handled correctly.
4690:
4691: If the class contains characters outside the 0-255 range, a different
4692: opcode is compiled. It may optionally have a bit map for characters < 256,
4693: but those above are are explicitly listed afterwards. A flag byte tells
4694: whether the bitmap is present, and whether this is a negated class or not.
4695:
4696: In JavaScript compatibility mode, an isolated ']' causes an error. In
4697: default (Perl) mode, it is treated as a data character. */
4698:
4699: case CHAR_RIGHT_SQUARE_BRACKET:
4700: if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4701: {
4702: *errorcodeptr = ERR64;
4703: goto FAILED;
4704: }
4705: goto NORMAL_CHAR;
4706:
4707: /* In another (POSIX) regex library, the ugly syntax [[:<:]] and [[:>:]] is
4708: used for "start of word" and "end of word". As these are otherwise illegal
4709: sequences, we don't break anything by recognizing them. They are replaced
4710: by \b(?=\w) and \b(?<=\w) respectively. Sequences like [a[:<:]] are
4711: erroneous and are handled by the normal code below. */
4712:
4713: case CHAR_LEFT_SQUARE_BRACKET:
4714: if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_STARTWORD, 6) == 0)
4715: {
4716: nestptr = ptr + 7;
4717: ptr = sub_start_of_word - 1;
4718: continue;
4719: }
4720:
4721: if (STRNCMP_UC_C8(ptr+1, STRING_WEIRD_ENDWORD, 6) == 0)
4722: {
4723: nestptr = ptr + 7;
4724: ptr = sub_end_of_word - 1;
4725: continue;
4726: }
4727:
4728: /* Handle a real character class. */
4729:
4730: previous = code;
4731:
4732: /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
4733: they are encountered at the top level, so we'll do that too. */
4734:
4735: if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4736: ptr[1] == CHAR_EQUALS_SIGN) &&
4737: check_posix_syntax(ptr, &tempptr))
4738: {
4739: *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
4740: goto FAILED;
4741: }
4742:
4743: /* If the first character is '^', set the negation flag and skip it. Also,
4744: if the first few characters (either before or after ^) are \Q\E or \E we
4745: skip them too. This makes for compatibility with Perl. */
4746:
4747: negate_class = FALSE;
4748: for (;;)
4749: {
4750: c = *(++ptr);
4751: if (c == CHAR_BACKSLASH)
4752: {
4753: if (ptr[1] == CHAR_E)
4754: ptr++;
4755: else if (STRNCMP_UC_C8(ptr + 1, STR_Q STR_BACKSLASH STR_E, 3) == 0)
4756: ptr += 3;
4757: else
4758: break;
4759: }
4760: else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
4761: negate_class = TRUE;
4762: else break;
4763: }
4764:
4765: /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
4766: an initial ']' is taken as a data character -- the code below handles
4767: that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
4768: [^] must match any character, so generate OP_ALLANY. */
4769:
4770: if (c == CHAR_RIGHT_SQUARE_BRACKET &&
4771: (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
4772: {
4773: *code++ = negate_class? OP_ALLANY : OP_FAIL;
4774: if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
4775: zerofirstchar = firstchar;
4776: zerofirstcharflags = firstcharflags;
4777: break;
4778: }
4779:
4780: /* If a class contains a negative special such as \S, we need to flip the
4781: negation flag at the end, so that support for characters > 255 works
4782: correctly (they are all included in the class). */
4783:
4784: should_flip_negation = FALSE;
4785:
4786: /* For optimization purposes, we track some properties of the class:
4787: class_has_8bitchar will be non-zero if the class contains at least one <
4788: 256 character; class_one_char will be 1 if the class contains just one
4789: character. */
4790:
4791: class_has_8bitchar = 0;
4792: class_one_char = 0;
4793:
4794: /* Initialize the 32-char bit map to all zeros. We build the map in a
4795: temporary bit of memory, in case the class contains fewer than two
4796: 8-bit characters because in that case the compiled code doesn't use the bit
4797: map. */
4798:
4799: memset(classbits, 0, 32 * sizeof(pcre_uint8));
4800:
4801: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4802: xclass = FALSE;
4803: class_uchardata = code + LINK_SIZE + 2; /* For XCLASS items */
4804: class_uchardata_base = class_uchardata; /* Save the start */
4805: #endif
4806:
4807: /* Process characters until ] is reached. By writing this as a "do" it
4808: means that an initial ] is taken as a data character. At the start of the
4809: loop, c contains the first byte of the character. */
4810:
4811: if (c != CHAR_NULL) do
4812: {
4813: const pcre_uchar *oldptr;
4814:
4815: #ifdef SUPPORT_UTF
4816: if (utf && HAS_EXTRALEN(c))
4817: { /* Braces are required because the */
4818: GETCHARLEN(c, ptr, ptr); /* macro generates multiple statements */
4819: }
4820: #endif
4821:
4822: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
4823: /* In the pre-compile phase, accumulate the length of any extra
4824: data and reset the pointer. This is so that very large classes that
4825: contain a zillion > 255 characters no longer overwrite the work space
4826: (which is on the stack). We have to remember that there was XCLASS data,
4827: however. */
4828:
4829: if (lengthptr != NULL && class_uchardata > class_uchardata_base)
4830: {
4831: xclass = TRUE;
4832: *lengthptr += class_uchardata - class_uchardata_base;
4833: class_uchardata = class_uchardata_base;
4834: }
4835: #endif
4836:
4837: /* Inside \Q...\E everything is literal except \E */
4838:
4839: if (inescq)
4840: {
4841: if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E) /* If we are at \E */
4842: {
4843: inescq = FALSE; /* Reset literal state */
4844: ptr++; /* Skip the 'E' */
4845: continue; /* Carry on with next */
4846: }
4847: goto CHECK_RANGE; /* Could be range if \E follows */
4848: }
4849:
4850: /* Handle POSIX class names. Perl allows a negation extension of the
4851: form [:^name:]. A square bracket that doesn't match the syntax is
4852: treated as a literal. We also recognize the POSIX constructions
4853: [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
4854: 5.6 and 5.8 do. */
4855:
4856: if (c == CHAR_LEFT_SQUARE_BRACKET &&
4857: (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
4858: ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
4859: {
4860: BOOL local_negate = FALSE;
4861: int posix_class, taboffset, tabopt;
4862: register const pcre_uint8 *cbits = cd->cbits;
4863: pcre_uint8 pbits[32];
4864:
4865: if (ptr[1] != CHAR_COLON)
4866: {
4867: *errorcodeptr = ERR31;
4868: goto FAILED;
4869: }
4870:
4871: ptr += 2;
4872: if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
4873: {
4874: local_negate = TRUE;
4875: should_flip_negation = TRUE; /* Note negative special */
4876: ptr++;
4877: }
4878:
4879: posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
4880: if (posix_class < 0)
4881: {
4882: *errorcodeptr = ERR30;
4883: goto FAILED;
4884: }
4885:
4886: /* If matching is caseless, upper and lower are converted to
4887: alpha. This relies on the fact that the class table starts with
4888: alpha, lower, upper as the first 3 entries. */
4889:
4890: if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
4891: posix_class = 0;
4892:
4893: /* When PCRE_UCP is set, some of the POSIX classes are converted to
4894: different escape sequences that use Unicode properties \p or \P. Others
4895: that are not available via \p or \P generate XCL_PROP/XCL_NOTPROP
4896: directly. */
4897:
4898: #ifdef SUPPORT_UCP
4899: if ((options & PCRE_UCP) != 0)
4900: {
4901: unsigned int ptype = 0;
4902: int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
4903:
4904: /* The posix_substitutes table specifies which POSIX classes can be
4905: converted to \p or \P items. */
4906:
4907: if (posix_substitutes[pc] != NULL)
4908: {
4909: nestptr = tempptr + 1;
4910: ptr = posix_substitutes[pc] - 1;
4911: continue;
4912: }
4913:
4914: /* There are three other classes that generate special property calls
4915: that are recognized only in an XCLASS. */
4916:
4917: else switch(posix_class)
4918: {
4919: case PC_GRAPH:
4920: ptype = PT_PXGRAPH;
4921: /* Fall through */
4922: case PC_PRINT:
4923: if (ptype == 0) ptype = PT_PXPRINT;
4924: /* Fall through */
4925: case PC_PUNCT:
4926: if (ptype == 0) ptype = PT_PXPUNCT;
4927: *class_uchardata++ = local_negate? XCL_NOTPROP : XCL_PROP;
4928: *class_uchardata++ = ptype;
4929: *class_uchardata++ = 0;
4930: ptr = tempptr + 1;
4931: continue;
4932:
4933: /* For all other POSIX classes, no special action is taken in UCP
4934: mode. Fall through to the non_UCP case. */
4935:
4936: default:
4937: break;
4938: }
4939: }
4940: #endif
4941: /* In the non-UCP case, or when UCP makes no difference, we build the
4942: bit map for the POSIX class in a chunk of local store because we may be
4943: adding and subtracting from it, and we don't want to subtract bits that
4944: may be in the main map already. At the end we or the result into the
4945: bit map that is being built. */
4946:
4947: posix_class *= 3;
4948:
4949: /* Copy in the first table (always present) */
4950:
4951: memcpy(pbits, cbits + posix_class_maps[posix_class],
4952: 32 * sizeof(pcre_uint8));
4953:
4954: /* If there is a second table, add or remove it as required. */
4955:
4956: taboffset = posix_class_maps[posix_class + 1];
4957: tabopt = posix_class_maps[posix_class + 2];
4958:
4959: if (taboffset >= 0)
4960: {
4961: if (tabopt >= 0)
4962: for (c = 0; c < 32; c++) pbits[c] |= cbits[c + taboffset];
4963: else
4964: for (c = 0; c < 32; c++) pbits[c] &= ~cbits[c + taboffset];
4965: }
4966:
4967: /* Now see if we need to remove any special characters. An option
4968: value of 1 removes vertical space and 2 removes underscore. */
4969:
4970: if (tabopt < 0) tabopt = -tabopt;
4971: if (tabopt == 1) pbits[1] &= ~0x3c;
4972: else if (tabopt == 2) pbits[11] &= 0x7f;
4973:
4974: /* Add the POSIX table or its complement into the main table that is
4975: being built and we are done. */
4976:
4977: if (local_negate)
4978: for (c = 0; c < 32; c++) classbits[c] |= ~pbits[c];
4979: else
4980: for (c = 0; c < 32; c++) classbits[c] |= pbits[c];
4981:
4982: ptr = tempptr + 1;
4983: /* Every class contains at least one < 256 character. */
4984: class_has_8bitchar = 1;
4985: /* Every class contains at least two characters. */
4986: class_one_char = 2;
4987: continue; /* End of POSIX syntax handling */
4988: }
4989:
4990: /* Backslash may introduce a single character, or it may introduce one
4991: of the specials, which just set a flag. The sequence \b is a special
4992: case. Inside a class (and only there) it is treated as backspace. We
4993: assume that other escapes have more than one character in them, so
4994: speculatively set both class_has_8bitchar and class_one_char bigger
4995: than one. Unrecognized escapes fall through and are either treated
4996: as literal characters (by default), or are faulted if
4997: PCRE_EXTRA is set. */
4998:
4999: if (c == CHAR_BACKSLASH)
5000: {
5001: escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options,
5002: TRUE);
5003: if (*errorcodeptr != 0) goto FAILED;
5004: if (escape == 0) c = ec;
5005: else if (escape == ESC_b) c = CHAR_BS; /* \b is backspace in a class */
5006: else if (escape == ESC_N) /* \N is not supported in a class */
5007: {
5008: *errorcodeptr = ERR71;
5009: goto FAILED;
5010: }
5011: else if (escape == ESC_Q) /* Handle start of quoted string */
5012: {
5013: if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5014: {
5015: ptr += 2; /* avoid empty string */
5016: }
5017: else inescq = TRUE;
5018: continue;
5019: }
5020: else if (escape == ESC_E) continue; /* Ignore orphan \E */
5021:
5022: else
5023: {
5024: register const pcre_uint8 *cbits = cd->cbits;
5025: /* Every class contains at least two < 256 characters. */
5026: class_has_8bitchar++;
5027: /* Every class contains at least two characters. */
5028: class_one_char += 2;
5029:
5030: switch (escape)
5031: {
5032: #ifdef SUPPORT_UCP
5033: case ESC_du: /* These are the values given for \d etc */
5034: case ESC_DU: /* when PCRE_UCP is set. We replace the */
5035: case ESC_wu: /* escape sequence with an appropriate \p */
5036: case ESC_WU: /* or \P to test Unicode properties instead */
5037: case ESC_su: /* of the default ASCII testing. */
5038: case ESC_SU:
5039: nestptr = ptr;
5040: ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
5041: class_has_8bitchar--; /* Undo! */
5042: continue;
5043: #endif
5044: case ESC_d:
5045: for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
5046: continue;
5047:
5048: case ESC_D:
5049: should_flip_negation = TRUE;
5050: for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
5051: continue;
5052:
5053: case ESC_w:
5054: for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_word];
5055: continue;
5056:
5057: case ESC_W:
5058: should_flip_negation = TRUE;
5059: for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
5060: continue;
5061:
5062: /* Perl 5.004 onwards omitted VT from \s, but restored it at Perl
5063: 5.18. Before PCRE 8.34, we had to preserve the VT bit if it was
5064: previously set by something earlier in the character class.
5065: Luckily, the value of CHAR_VT is 0x0b in both ASCII and EBCDIC, so
5066: we could just adjust the appropriate bit. From PCRE 8.34 we no
5067: longer treat \s and \S specially. */
5068:
5069: case ESC_s:
5070: for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
5071: continue;
5072:
5073: case ESC_S:
5074: should_flip_negation = TRUE;
5075: for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
5076: continue;
5077:
5078: /* The rest apply in both UCP and non-UCP cases. */
5079:
5080: case ESC_h:
5081: (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5082: PRIV(hspace_list), NOTACHAR);
5083: continue;
5084:
5085: case ESC_H:
5086: (void)add_not_list_to_class(classbits, &class_uchardata, options,
5087: cd, PRIV(hspace_list));
5088: continue;
5089:
5090: case ESC_v:
5091: (void)add_list_to_class(classbits, &class_uchardata, options, cd,
5092: PRIV(vspace_list), NOTACHAR);
5093: continue;
5094:
5095: case ESC_V:
5096: (void)add_not_list_to_class(classbits, &class_uchardata, options,
5097: cd, PRIV(vspace_list));
5098: continue;
5099:
5100: #ifdef SUPPORT_UCP
5101: case ESC_p:
5102: case ESC_P:
5103: {
5104: BOOL negated;
5105: unsigned int ptype = 0, pdata = 0;
5106: if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
5107: goto FAILED;
5108: *class_uchardata++ = ((escape == ESC_p) != negated)?
5109: XCL_PROP : XCL_NOTPROP;
5110: *class_uchardata++ = ptype;
5111: *class_uchardata++ = pdata;
5112: class_has_8bitchar--; /* Undo! */
5113: continue;
5114: }
5115: #endif
5116: /* Unrecognized escapes are faulted if PCRE is running in its
5117: strict mode. By default, for compatibility with Perl, they are
5118: treated as literals. */
5119:
5120: default:
5121: if ((options & PCRE_EXTRA) != 0)
5122: {
5123: *errorcodeptr = ERR7;
5124: goto FAILED;
5125: }
5126: class_has_8bitchar--; /* Undo the speculative increase. */
5127: class_one_char -= 2; /* Undo the speculative increase. */
5128: c = *ptr; /* Get the final character and fall through */
5129: break;
5130: }
5131: }
5132:
5133: /* Fall through if the escape just defined a single character (c >= 0).
5134: This may be greater than 256. */
5135:
5136: escape = 0;
5137:
5138: } /* End of backslash handling */
5139:
5140: /* A character may be followed by '-' to form a range. However, Perl does
5141: not permit ']' to be the end of the range. A '-' character at the end is
5142: treated as a literal. Perl ignores orphaned \E sequences entirely. The
5143: code for handling \Q and \E is messy. */
5144:
5145: CHECK_RANGE:
5146: while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5147: {
5148: inescq = FALSE;
5149: ptr += 2;
5150: }
5151: oldptr = ptr;
5152:
5153: /* Remember if \r or \n were explicitly used */
5154:
5155: if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5156:
5157: /* Check for range */
5158:
5159: if (!inescq && ptr[1] == CHAR_MINUS)
5160: {
5161: pcre_uint32 d;
5162: ptr += 2;
5163: while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
5164:
5165: /* If we hit \Q (not followed by \E) at this point, go into escaped
5166: mode. */
5167:
5168: while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
5169: {
5170: ptr += 2;
5171: if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
5172: { ptr += 2; continue; }
5173: inescq = TRUE;
5174: break;
5175: }
5176:
5177: /* Minus (hyphen) at the end of a class is treated as a literal, so put
5178: back the pointer and jump to handle the character that preceded it. */
5179:
5180: if (*ptr == CHAR_NULL || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
5181: {
5182: ptr = oldptr;
5183: goto CLASS_SINGLE_CHARACTER;
5184: }
5185:
5186: /* Otherwise, we have a potential range; pick up the next character */
5187:
5188: #ifdef SUPPORT_UTF
5189: if (utf)
5190: { /* Braces are required because the */
5191: GETCHARLEN(d, ptr, ptr); /* macro generates multiple statements */
5192: }
5193: else
5194: #endif
5195: d = *ptr; /* Not UTF-8 mode */
5196:
5197: /* The second part of a range can be a single-character escape
5198: sequence, but not any of the other escapes. Perl treats a hyphen as a
5199: literal in such circumstances. However, in Perl's warning mode, a
5200: warning is given, so PCRE now faults it as it is almost certainly a
5201: mistake on the user's part. */
5202:
5203: if (!inescq)
5204: {
5205: if (d == CHAR_BACKSLASH)
5206: {
5207: int descape;
5208: descape = check_escape(&ptr, &d, errorcodeptr, cd->bracount, options, TRUE);
5209: if (*errorcodeptr != 0) goto FAILED;
5210:
5211: /* 0 means a character was put into d; \b is backspace; any other
5212: special causes an error. */
5213:
5214: if (descape != 0)
5215: {
5216: if (descape == ESC_b) d = CHAR_BS; else
5217: {
5218: *errorcodeptr = ERR83;
5219: goto FAILED;
5220: }
5221: }
5222: }
5223:
5224: /* A hyphen followed by a POSIX class is treated in the same way. */
5225:
5226: else if (d == CHAR_LEFT_SQUARE_BRACKET &&
5227: (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
5228: ptr[1] == CHAR_EQUALS_SIGN) &&
5229: check_posix_syntax(ptr, &tempptr))
5230: {
5231: *errorcodeptr = ERR83;
5232: goto FAILED;
5233: }
5234: }
5235:
5236: /* Check that the two values are in the correct order. Optimize
5237: one-character ranges. */
5238:
5239: if (d < c)
5240: {
5241: *errorcodeptr = ERR8;
5242: goto FAILED;
5243: }
5244: if (d == c) goto CLASS_SINGLE_CHARACTER; /* A few lines below */
5245:
5246: /* We have found a character range, so single character optimizations
5247: cannot be done anymore. Any value greater than 1 indicates that there
5248: is more than one character. */
5249:
5250: class_one_char = 2;
5251:
5252: /* Remember an explicit \r or \n, and add the range to the class. */
5253:
5254: if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
5255:
5256: class_has_8bitchar +=
5257: add_to_class(classbits, &class_uchardata, options, cd, c, d);
5258:
5259: continue; /* Go get the next char in the class */
5260: }
5261:
5262: /* Handle a single character - we can get here for a normal non-escape
5263: char, or after \ that introduces a single character or for an apparent
5264: range that isn't. Only the value 1 matters for class_one_char, so don't
5265: increase it if it is already 2 or more ... just in case there's a class
5266: with a zillion characters in it. */
5267:
5268: CLASS_SINGLE_CHARACTER:
5269: if (class_one_char < 2) class_one_char++;
5270:
5271: /* If class_one_char is 1, we have the first single character in the
5272: class, and there have been no prior ranges, or XCLASS items generated by
5273: escapes. If this is the final character in the class, we can optimize by
5274: turning the item into a 1-character OP_CHAR[I] if it's positive, or
5275: OP_NOT[I] if it's negative. In the positive case, it can cause firstchar
5276: to be set. Otherwise, there can be no first char if this item is first,
5277: whatever repeat count may follow. In the case of reqchar, save the
5278: previous value for reinstating. */
5279:
5280: if (class_one_char == 1 && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
5281: {
5282: ptr++;
5283: zeroreqchar = reqchar;
5284: zeroreqcharflags = reqcharflags;
5285:
5286: if (negate_class)
5287: {
5288: #ifdef SUPPORT_UCP
5289: int d;
5290: #endif
5291: if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5292: zerofirstchar = firstchar;
5293: zerofirstcharflags = firstcharflags;
5294:
5295: /* For caseless UTF-8 mode when UCP support is available, check
5296: whether this character has more than one other case. If so, generate
5297: a special OP_NOTPROP item instead of OP_NOTI. */
5298:
5299: #ifdef SUPPORT_UCP
5300: if (utf && (options & PCRE_CASELESS) != 0 &&
5301: (d = UCD_CASESET(c)) != 0)
5302: {
5303: *code++ = OP_NOTPROP;
5304: *code++ = PT_CLIST;
5305: *code++ = d;
5306: }
5307: else
5308: #endif
5309: /* Char has only one other case, or UCP not available */
5310:
5311: {
5312: *code++ = ((options & PCRE_CASELESS) != 0)? OP_NOTI: OP_NOT;
5313: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5314: if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5315: code += PRIV(ord2utf)(c, code);
5316: else
5317: #endif
5318: *code++ = c;
5319: }
5320:
5321: /* We are finished with this character class */
5322:
5323: goto END_CLASS;
5324: }
5325:
5326: /* For a single, positive character, get the value into mcbuffer, and
5327: then we can handle this with the normal one-character code. */
5328:
5329: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5330: if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
5331: mclength = PRIV(ord2utf)(c, mcbuffer);
5332: else
5333: #endif
5334: {
5335: mcbuffer[0] = c;
5336: mclength = 1;
5337: }
5338: goto ONE_CHAR;
5339: } /* End of 1-char optimization */
5340:
5341: /* There is more than one character in the class, or an XCLASS item
5342: has been generated. Add this character to the class. */
5343:
5344: class_has_8bitchar +=
5345: add_to_class(classbits, &class_uchardata, options, cd, c, c);
5346: }
5347:
5348: /* Loop until ']' reached. This "while" is the end of the "do" far above.
5349: If we are at the end of an internal nested string, revert to the outer
5350: string. */
5351:
5352: while (((c = *(++ptr)) != CHAR_NULL ||
5353: (nestptr != NULL &&
5354: (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != CHAR_NULL)) &&
5355: (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
5356:
5357: /* Check for missing terminating ']' */
5358:
5359: if (c == CHAR_NULL)
5360: {
5361: *errorcodeptr = ERR6;
5362: goto FAILED;
5363: }
5364:
5365: /* We will need an XCLASS if data has been placed in class_uchardata. In
5366: the second phase this is a sufficient test. However, in the pre-compile
5367: phase, class_uchardata gets emptied to prevent workspace overflow, so it
5368: only if the very last character in the class needs XCLASS will it contain
5369: anything at this point. For this reason, xclass gets set TRUE above when
5370: uchar_classdata is emptied, and that's why this code is the way it is here
5371: instead of just doing a test on class_uchardata below. */
5372:
5373: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5374: if (class_uchardata > class_uchardata_base) xclass = TRUE;
5375: #endif
5376:
5377: /* If this is the first thing in the branch, there can be no first char
5378: setting, whatever the repeat count. Any reqchar setting must remain
5379: unchanged after any kind of repeat. */
5380:
5381: if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
5382: zerofirstchar = firstchar;
5383: zerofirstcharflags = firstcharflags;
5384: zeroreqchar = reqchar;
5385: zeroreqcharflags = reqcharflags;
5386:
5387: /* If there are characters with values > 255, we have to compile an
5388: extended class, with its own opcode, unless there was a negated special
5389: such as \S in the class, and PCRE_UCP is not set, because in that case all
5390: characters > 255 are in the class, so any that were explicitly given as
5391: well can be ignored. If (when there are explicit characters > 255 that must
5392: be listed) there are no characters < 256, we can omit the bitmap in the
5393: actual compiled code. */
5394:
5395: #ifdef SUPPORT_UTF
5396: if (xclass && (!should_flip_negation || (options & PCRE_UCP) != 0))
5397: #elif !defined COMPILE_PCRE8
5398: if (xclass && !should_flip_negation)
5399: #endif
5400: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5401: {
5402: *class_uchardata++ = XCL_END; /* Marks the end of extra data */
5403: *code++ = OP_XCLASS;
5404: code += LINK_SIZE;
5405: *code = negate_class? XCL_NOT:0;
5406:
5407: /* If the map is required, move up the extra data to make room for it;
5408: otherwise just move the code pointer to the end of the extra data. */
5409:
5410: if (class_has_8bitchar > 0)
5411: {
5412: *code++ |= XCL_MAP;
5413: memmove(code + (32 / sizeof(pcre_uchar)), code,
5414: IN_UCHARS(class_uchardata - code));
5415: memcpy(code, classbits, 32);
5416: code = class_uchardata + (32 / sizeof(pcre_uchar));
5417: }
5418: else code = class_uchardata;
5419:
5420: /* Now fill in the complete length of the item */
5421:
5422: PUT(previous, 1, (int)(code - previous));
5423: break; /* End of class handling */
5424: }
5425: #endif
5426:
5427: /* If there are no characters > 255, or they are all to be included or
5428: excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
5429: whole class was negated and whether there were negative specials such as \S
5430: (non-UCP) in the class. Then copy the 32-byte map into the code vector,
5431: negating it if necessary. */
5432:
5433: *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
5434: if (lengthptr == NULL) /* Save time in the pre-compile phase */
5435: {
5436: if (negate_class)
5437: for (c = 0; c < 32; c++) classbits[c] = ~classbits[c];
5438: memcpy(code, classbits, 32);
5439: }
5440: code += 32 / sizeof(pcre_uchar);
5441:
5442: END_CLASS:
5443: break;
5444:
5445:
5446: /* ===================================================================*/
5447: /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
5448: has been tested above. */
5449:
5450: case CHAR_LEFT_CURLY_BRACKET:
5451: if (!is_quantifier) goto NORMAL_CHAR;
5452: ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
5453: if (*errorcodeptr != 0) goto FAILED;
5454: goto REPEAT;
5455:
5456: case CHAR_ASTERISK:
5457: repeat_min = 0;
5458: repeat_max = -1;
5459: goto REPEAT;
5460:
5461: case CHAR_PLUS:
5462: repeat_min = 1;
5463: repeat_max = -1;
5464: goto REPEAT;
5465:
5466: case CHAR_QUESTION_MARK:
5467: repeat_min = 0;
5468: repeat_max = 1;
5469:
5470: REPEAT:
5471: if (previous == NULL)
5472: {
5473: *errorcodeptr = ERR9;
5474: goto FAILED;
5475: }
5476:
5477: if (repeat_min == 0)
5478: {
5479: firstchar = zerofirstchar; /* Adjust for zero repeat */
5480: firstcharflags = zerofirstcharflags;
5481: reqchar = zeroreqchar; /* Ditto */
5482: reqcharflags = zeroreqcharflags;
5483: }
5484:
5485: /* Remember whether this is a variable length repeat */
5486:
5487: reqvary = (repeat_min == repeat_max)? 0 : REQ_VARY;
5488:
5489: op_type = 0; /* Default single-char op codes */
5490: possessive_quantifier = FALSE; /* Default not possessive quantifier */
5491:
5492: /* Save start of previous item, in case we have to move it up in order to
5493: insert something before it. */
5494:
5495: tempcode = previous;
5496:
5497: /* Before checking for a possessive quantifier, we must skip over
5498: whitespace and comments in extended mode because Perl allows white space at
5499: this point. */
5500:
5501: if ((options & PCRE_EXTENDED) != 0)
5502: {
5503: const pcre_uchar *p = ptr + 1;
5504: for (;;)
5505: {
5506: while (MAX_255(*p) && (cd->ctypes[*p] & ctype_space) != 0) p++;
5507: if (*p != CHAR_NUMBER_SIGN) break;
5508: p++;
5509: while (*p != CHAR_NULL)
5510: {
5511: if (IS_NEWLINE(p)) /* For non-fixed-length newline cases, */
5512: { /* IS_NEWLINE sets cd->nllen. */
5513: p += cd->nllen;
5514: break;
5515: }
5516: p++;
5517: #ifdef SUPPORT_UTF
5518: if (utf) FORWARDCHAR(p);
5519: #endif
5520: } /* Loop for comment characters */
5521: } /* Loop for multiple comments */
5522: ptr = p - 1; /* Character before the next significant one. */
5523: }
5524:
5525: /* If the next character is '+', we have a possessive quantifier. This
5526: implies greediness, whatever the setting of the PCRE_UNGREEDY option.
5527: If the next character is '?' this is a minimizing repeat, by default,
5528: but if PCRE_UNGREEDY is set, it works the other way round. We change the
5529: repeat type to the non-default. */
5530:
5531: if (ptr[1] == CHAR_PLUS)
5532: {
5533: repeat_type = 0; /* Force greedy */
5534: possessive_quantifier = TRUE;
5535: ptr++;
5536: }
5537: else if (ptr[1] == CHAR_QUESTION_MARK)
5538: {
5539: repeat_type = greedy_non_default;
5540: ptr++;
5541: }
5542: else repeat_type = greedy_default;
5543:
5544: /* If previous was a recursion call, wrap it in atomic brackets so that
5545: previous becomes the atomic group. All recursions were so wrapped in the
5546: past, but it no longer happens for non-repeated recursions. In fact, the
5547: repeated ones could be re-implemented independently so as not to need this,
5548: but for the moment we rely on the code for repeating groups. */
5549:
5550: if (*previous == OP_RECURSE)
5551: {
5552: memmove(previous + 1 + LINK_SIZE, previous, IN_UCHARS(1 + LINK_SIZE));
5553: *previous = OP_ONCE;
5554: PUT(previous, 1, 2 + 2*LINK_SIZE);
5555: previous[2 + 2*LINK_SIZE] = OP_KET;
5556: PUT(previous, 3 + 2*LINK_SIZE, 2 + 2*LINK_SIZE);
5557: code += 2 + 2 * LINK_SIZE;
5558: length_prevgroup = 3 + 3*LINK_SIZE;
5559:
5560: /* When actually compiling, we need to check whether this was a forward
5561: reference, and if so, adjust the offset. */
5562:
5563: if (lengthptr == NULL && cd->hwm >= cd->start_workspace + LINK_SIZE)
5564: {
5565: int offset = GET(cd->hwm, -LINK_SIZE);
5566: if (offset == previous + 1 - cd->start_code)
5567: PUT(cd->hwm, -LINK_SIZE, offset + 1 + LINK_SIZE);
5568: }
5569: }
5570:
5571: /* Now handle repetition for the different types of item. */
5572:
5573: /* If previous was a character or negated character match, abolish the item
5574: and generate a repeat item instead. If a char item has a minimum of more
5575: than one, ensure that it is set in reqchar - it might not be if a sequence
5576: such as x{3} is the first thing in a branch because the x will have gone
5577: into firstchar instead. */
5578:
5579: if (*previous == OP_CHAR || *previous == OP_CHARI
5580: || *previous == OP_NOT || *previous == OP_NOTI)
5581: {
5582: switch (*previous)
5583: {
5584: default: /* Make compiler happy. */
5585: case OP_CHAR: op_type = OP_STAR - OP_STAR; break;
5586: case OP_CHARI: op_type = OP_STARI - OP_STAR; break;
5587: case OP_NOT: op_type = OP_NOTSTAR - OP_STAR; break;
5588: case OP_NOTI: op_type = OP_NOTSTARI - OP_STAR; break;
5589: }
5590:
5591: /* Deal with UTF characters that take up more than one character. It's
5592: easier to write this out separately than try to macrify it. Use c to
5593: hold the length of the character in bytes, plus UTF_LENGTH to flag that
5594: it's a length rather than a small character. */
5595:
5596: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5597: if (utf && NOT_FIRSTCHAR(code[-1]))
5598: {
5599: pcre_uchar *lastchar = code - 1;
5600: BACKCHAR(lastchar);
5601: c = (int)(code - lastchar); /* Length of UTF-8 character */
5602: memcpy(utf_chars, lastchar, IN_UCHARS(c)); /* Save the char */
5603: c |= UTF_LENGTH; /* Flag c as a length */
5604: }
5605: else
5606: #endif /* SUPPORT_UTF */
5607:
5608: /* Handle the case of a single charater - either with no UTF support, or
5609: with UTF disabled, or for a single character UTF character. */
5610: {
5611: c = code[-1];
5612: if (*previous <= OP_CHARI && repeat_min > 1)
5613: {
5614: reqchar = c;
5615: reqcharflags = req_caseopt | cd->req_varyopt;
5616: }
5617: }
5618:
5619: goto OUTPUT_SINGLE_REPEAT; /* Code shared with single character types */
5620: }
5621:
5622: /* If previous was a character type match (\d or similar), abolish it and
5623: create a suitable repeat item. The code is shared with single-character
5624: repeats by setting op_type to add a suitable offset into repeat_type. Note
5625: the the Unicode property types will be present only when SUPPORT_UCP is
5626: defined, but we don't wrap the little bits of code here because it just
5627: makes it horribly messy. */
5628:
5629: else if (*previous < OP_EODN)
5630: {
5631: pcre_uchar *oldcode;
5632: int prop_type, prop_value;
5633: op_type = OP_TYPESTAR - OP_STAR; /* Use type opcodes */
5634: c = *previous;
5635:
5636: OUTPUT_SINGLE_REPEAT:
5637: if (*previous == OP_PROP || *previous == OP_NOTPROP)
5638: {
5639: prop_type = previous[1];
5640: prop_value = previous[2];
5641: }
5642: else prop_type = prop_value = -1;
5643:
5644: oldcode = code;
5645: code = previous; /* Usually overwrite previous item */
5646:
5647: /* If the maximum is zero then the minimum must also be zero; Perl allows
5648: this case, so we do too - by simply omitting the item altogether. */
5649:
5650: if (repeat_max == 0) goto END_REPEAT;
5651:
5652: /* Combine the op_type with the repeat_type */
5653:
5654: repeat_type += op_type;
5655:
5656: /* A minimum of zero is handled either as the special case * or ?, or as
5657: an UPTO, with the maximum given. */
5658:
5659: if (repeat_min == 0)
5660: {
5661: if (repeat_max == -1) *code++ = OP_STAR + repeat_type;
5662: else if (repeat_max == 1) *code++ = OP_QUERY + repeat_type;
5663: else
5664: {
5665: *code++ = OP_UPTO + repeat_type;
5666: PUT2INC(code, 0, repeat_max);
5667: }
5668: }
5669:
5670: /* A repeat minimum of 1 is optimized into some special cases. If the
5671: maximum is unlimited, we use OP_PLUS. Otherwise, the original item is
5672: left in place and, if the maximum is greater than 1, we use OP_UPTO with
5673: one less than the maximum. */
5674:
5675: else if (repeat_min == 1)
5676: {
5677: if (repeat_max == -1)
5678: *code++ = OP_PLUS + repeat_type;
5679: else
5680: {
5681: code = oldcode; /* leave previous item in place */
5682: if (repeat_max == 1) goto END_REPEAT;
5683: *code++ = OP_UPTO + repeat_type;
5684: PUT2INC(code, 0, repeat_max - 1);
5685: }
5686: }
5687:
5688: /* The case {n,n} is just an EXACT, while the general case {n,m} is
5689: handled as an EXACT followed by an UPTO. */
5690:
5691: else
5692: {
5693: *code++ = OP_EXACT + op_type; /* NB EXACT doesn't have repeat_type */
5694: PUT2INC(code, 0, repeat_min);
5695:
5696: /* If the maximum is unlimited, insert an OP_STAR. Before doing so,
5697: we have to insert the character for the previous code. For a repeated
5698: Unicode property match, there are two extra bytes that define the
5699: required property. In UTF-8 mode, long characters have their length in
5700: c, with the UTF_LENGTH bit as a flag. */
5701:
5702: if (repeat_max < 0)
5703: {
5704: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5705: if (utf && (c & UTF_LENGTH) != 0)
5706: {
5707: memcpy(code, utf_chars, IN_UCHARS(c & 7));
5708: code += c & 7;
5709: }
5710: else
5711: #endif
5712: {
5713: *code++ = c;
5714: if (prop_type >= 0)
5715: {
5716: *code++ = prop_type;
5717: *code++ = prop_value;
5718: }
5719: }
5720: *code++ = OP_STAR + repeat_type;
5721: }
5722:
5723: /* Else insert an UPTO if the max is greater than the min, again
5724: preceded by the character, for the previously inserted code. If the
5725: UPTO is just for 1 instance, we can use QUERY instead. */
5726:
5727: else if (repeat_max != repeat_min)
5728: {
5729: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5730: if (utf && (c & UTF_LENGTH) != 0)
5731: {
5732: memcpy(code, utf_chars, IN_UCHARS(c & 7));
5733: code += c & 7;
5734: }
5735: else
5736: #endif
5737: *code++ = c;
5738: if (prop_type >= 0)
5739: {
5740: *code++ = prop_type;
5741: *code++ = prop_value;
5742: }
5743: repeat_max -= repeat_min;
5744:
5745: if (repeat_max == 1)
5746: {
5747: *code++ = OP_QUERY + repeat_type;
5748: }
5749: else
5750: {
5751: *code++ = OP_UPTO + repeat_type;
5752: PUT2INC(code, 0, repeat_max);
5753: }
5754: }
5755: }
5756:
5757: /* The character or character type itself comes last in all cases. */
5758:
5759: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
5760: if (utf && (c & UTF_LENGTH) != 0)
5761: {
5762: memcpy(code, utf_chars, IN_UCHARS(c & 7));
5763: code += c & 7;
5764: }
5765: else
5766: #endif
5767: *code++ = c;
5768:
5769: /* For a repeated Unicode property match, there are two extra bytes that
5770: define the required property. */
5771:
5772: #ifdef SUPPORT_UCP
5773: if (prop_type >= 0)
5774: {
5775: *code++ = prop_type;
5776: *code++ = prop_value;
5777: }
5778: #endif
5779: }
5780:
5781: /* If previous was a character class or a back reference, we put the repeat
5782: stuff after it, but just skip the item if the repeat was {0,0}. */
5783:
5784: else if (*previous == OP_CLASS || *previous == OP_NCLASS ||
5785: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
5786: *previous == OP_XCLASS ||
5787: #endif
5788: *previous == OP_REF || *previous == OP_REFI ||
5789: *previous == OP_DNREF || *previous == OP_DNREFI)
5790: {
5791: if (repeat_max == 0)
5792: {
5793: code = previous;
5794: goto END_REPEAT;
5795: }
5796:
5797: if (repeat_min == 0 && repeat_max == -1)
5798: *code++ = OP_CRSTAR + repeat_type;
5799: else if (repeat_min == 1 && repeat_max == -1)
5800: *code++ = OP_CRPLUS + repeat_type;
5801: else if (repeat_min == 0 && repeat_max == 1)
5802: *code++ = OP_CRQUERY + repeat_type;
5803: else
5804: {
5805: *code++ = OP_CRRANGE + repeat_type;
5806: PUT2INC(code, 0, repeat_min);
5807: if (repeat_max == -1) repeat_max = 0; /* 2-byte encoding for max */
5808: PUT2INC(code, 0, repeat_max);
5809: }
5810: }
5811:
5812: /* If previous was a bracket group, we may have to replicate it in certain
5813: cases. Note that at this point we can encounter only the "basic" bracket
5814: opcodes such as BRA and CBRA, as this is the place where they get converted
5815: into the more special varieties such as BRAPOS and SBRA. A test for >=
5816: OP_ASSERT and <= OP_COND includes ASSERT, ASSERT_NOT, ASSERTBACK,
5817: ASSERTBACK_NOT, ONCE, ONCE_NC, BRA, BRAPOS, CBRA, CBRAPOS, and COND.
5818: Originally, PCRE did not allow repetition of assertions, but now it does,
5819: for Perl compatibility. */
5820:
5821: else if (*previous >= OP_ASSERT && *previous <= OP_COND)
5822: {
5823: register int i;
5824: int len = (int)(code - previous);
5825: pcre_uchar *bralink = NULL;
5826: pcre_uchar *brazeroptr = NULL;
5827:
5828: /* Repeating a DEFINE group is pointless, but Perl allows the syntax, so
5829: we just ignore the repeat. */
5830:
5831: if (*previous == OP_COND && previous[LINK_SIZE+1] == OP_DEF)
5832: goto END_REPEAT;
5833:
5834: /* There is no sense in actually repeating assertions. The only potential
5835: use of repetition is in cases when the assertion is optional. Therefore,
5836: if the minimum is greater than zero, just ignore the repeat. If the
5837: maximum is not zero or one, set it to 1. */
5838:
5839: if (*previous < OP_ONCE) /* Assertion */
5840: {
5841: if (repeat_min > 0) goto END_REPEAT;
5842: if (repeat_max < 0 || repeat_max > 1) repeat_max = 1;
5843: }
5844:
5845: /* The case of a zero minimum is special because of the need to stick
5846: OP_BRAZERO in front of it, and because the group appears once in the
5847: data, whereas in other cases it appears the minimum number of times. For
5848: this reason, it is simplest to treat this case separately, as otherwise
5849: the code gets far too messy. There are several special subcases when the
5850: minimum is zero. */
5851:
5852: if (repeat_min == 0)
5853: {
5854: /* If the maximum is also zero, we used to just omit the group from the
5855: output altogether, like this:
5856:
5857: ** if (repeat_max == 0)
5858: ** {
5859: ** code = previous;
5860: ** goto END_REPEAT;
5861: ** }
5862:
5863: However, that fails when a group or a subgroup within it is referenced
5864: as a subroutine from elsewhere in the pattern, so now we stick in
5865: OP_SKIPZERO in front of it so that it is skipped on execution. As we
5866: don't have a list of which groups are referenced, we cannot do this
5867: selectively.
5868:
5869: If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
5870: and do no more at this point. However, we do need to adjust any
5871: OP_RECURSE calls inside the group that refer to the group itself or any
5872: internal or forward referenced group, because the offset is from the
5873: start of the whole regex. Temporarily terminate the pattern while doing
5874: this. */
5875:
5876: if (repeat_max <= 1) /* Covers 0, 1, and unlimited */
5877: {
5878: *code = OP_END;
5879: adjust_recurse(previous, 1, utf, cd, save_hwm);
5880: memmove(previous + 1, previous, IN_UCHARS(len));
5881: code++;
5882: if (repeat_max == 0)
5883: {
5884: *previous++ = OP_SKIPZERO;
5885: goto END_REPEAT;
5886: }
5887: brazeroptr = previous; /* Save for possessive optimizing */
5888: *previous++ = OP_BRAZERO + repeat_type;
5889: }
5890:
5891: /* If the maximum is greater than 1 and limited, we have to replicate
5892: in a nested fashion, sticking OP_BRAZERO before each set of brackets.
5893: The first one has to be handled carefully because it's the original
5894: copy, which has to be moved up. The remainder can be handled by code
5895: that is common with the non-zero minimum case below. We have to
5896: adjust the value or repeat_max, since one less copy is required. Once
5897: again, we may have to adjust any OP_RECURSE calls inside the group. */
5898:
5899: else
5900: {
5901: int offset;
5902: *code = OP_END;
5903: adjust_recurse(previous, 2 + LINK_SIZE, utf, cd, save_hwm);
5904: memmove(previous + 2 + LINK_SIZE, previous, IN_UCHARS(len));
5905: code += 2 + LINK_SIZE;
5906: *previous++ = OP_BRAZERO + repeat_type;
5907: *previous++ = OP_BRA;
5908:
5909: /* We chain together the bracket offset fields that have to be
5910: filled in later when the ends of the brackets are reached. */
5911:
5912: offset = (bralink == NULL)? 0 : (int)(previous - bralink);
5913: bralink = previous;
5914: PUTINC(previous, 0, offset);
5915: }
5916:
5917: repeat_max--;
5918: }
5919:
5920: /* If the minimum is greater than zero, replicate the group as many
5921: times as necessary, and adjust the maximum to the number of subsequent
5922: copies that we need. If we set a first char from the group, and didn't
5923: set a required char, copy the latter from the former. If there are any
5924: forward reference subroutine calls in the group, there will be entries on
5925: the workspace list; replicate these with an appropriate increment. */
5926:
5927: else
5928: {
5929: if (repeat_min > 1)
5930: {
5931: /* In the pre-compile phase, we don't actually do the replication. We
5932: just adjust the length as if we had. Do some paranoid checks for
5933: potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
5934: integer type when available, otherwise double. */
5935:
5936: if (lengthptr != NULL)
5937: {
5938: int delta = (repeat_min - 1)*length_prevgroup;
5939: if ((INT64_OR_DOUBLE)(repeat_min - 1)*
5940: (INT64_OR_DOUBLE)length_prevgroup >
5941: (INT64_OR_DOUBLE)INT_MAX ||
5942: OFLOW_MAX - *lengthptr < delta)
5943: {
5944: *errorcodeptr = ERR20;
5945: goto FAILED;
5946: }
5947: *lengthptr += delta;
5948: }
5949:
5950: /* This is compiling for real. If there is a set first byte for
5951: the group, and we have not yet set a "required byte", set it. Make
5952: sure there is enough workspace for copying forward references before
5953: doing the copy. */
5954:
5955: else
5956: {
5957: if (groupsetfirstchar && reqcharflags < 0)
5958: {
5959: reqchar = firstchar;
5960: reqcharflags = firstcharflags;
5961: }
5962:
5963: for (i = 1; i < repeat_min; i++)
5964: {
5965: pcre_uchar *hc;
5966: pcre_uchar *this_hwm = cd->hwm;
5967: memcpy(code, previous, IN_UCHARS(len));
5968:
5969: while (cd->hwm > cd->start_workspace + cd->workspace_size -
5970: WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
5971: {
5972: int save_offset = save_hwm - cd->start_workspace;
5973: int this_offset = this_hwm - cd->start_workspace;
5974: *errorcodeptr = expand_workspace(cd);
5975: if (*errorcodeptr != 0) goto FAILED;
5976: save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
5977: this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
5978: }
5979:
5980: for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
5981: {
5982: PUT(cd->hwm, 0, GET(hc, 0) + len);
5983: cd->hwm += LINK_SIZE;
5984: }
5985: save_hwm = this_hwm;
5986: code += len;
5987: }
5988: }
5989: }
5990:
5991: if (repeat_max > 0) repeat_max -= repeat_min;
5992: }
5993:
5994: /* This code is common to both the zero and non-zero minimum cases. If
5995: the maximum is limited, it replicates the group in a nested fashion,
5996: remembering the bracket starts on a stack. In the case of a zero minimum,
5997: the first one was set up above. In all cases the repeat_max now specifies
5998: the number of additional copies needed. Again, we must remember to
5999: replicate entries on the forward reference list. */
6000:
6001: if (repeat_max >= 0)
6002: {
6003: /* In the pre-compile phase, we don't actually do the replication. We
6004: just adjust the length as if we had. For each repetition we must add 1
6005: to the length for BRAZERO and for all but the last repetition we must
6006: add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
6007: paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
6008: a 64-bit integer type when available, otherwise double. */
6009:
6010: if (lengthptr != NULL && repeat_max > 0)
6011: {
6012: int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
6013: 2 - 2*LINK_SIZE; /* Last one doesn't nest */
6014: if ((INT64_OR_DOUBLE)repeat_max *
6015: (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
6016: > (INT64_OR_DOUBLE)INT_MAX ||
6017: OFLOW_MAX - *lengthptr < delta)
6018: {
6019: *errorcodeptr = ERR20;
6020: goto FAILED;
6021: }
6022: *lengthptr += delta;
6023: }
6024:
6025: /* This is compiling for real */
6026:
6027: else for (i = repeat_max - 1; i >= 0; i--)
6028: {
6029: pcre_uchar *hc;
6030: pcre_uchar *this_hwm = cd->hwm;
6031:
6032: *code++ = OP_BRAZERO + repeat_type;
6033:
6034: /* All but the final copy start a new nesting, maintaining the
6035: chain of brackets outstanding. */
6036:
6037: if (i != 0)
6038: {
6039: int offset;
6040: *code++ = OP_BRA;
6041: offset = (bralink == NULL)? 0 : (int)(code - bralink);
6042: bralink = code;
6043: PUTINC(code, 0, offset);
6044: }
6045:
6046: memcpy(code, previous, IN_UCHARS(len));
6047:
6048: /* Ensure there is enough workspace for forward references before
6049: copying them. */
6050:
6051: while (cd->hwm > cd->start_workspace + cd->workspace_size -
6052: WORK_SIZE_SAFETY_MARGIN - (this_hwm - save_hwm))
6053: {
6054: int save_offset = save_hwm - cd->start_workspace;
6055: int this_offset = this_hwm - cd->start_workspace;
6056: *errorcodeptr = expand_workspace(cd);
6057: if (*errorcodeptr != 0) goto FAILED;
6058: save_hwm = (pcre_uchar *)cd->start_workspace + save_offset;
6059: this_hwm = (pcre_uchar *)cd->start_workspace + this_offset;
6060: }
6061:
6062: for (hc = save_hwm; hc < this_hwm; hc += LINK_SIZE)
6063: {
6064: PUT(cd->hwm, 0, GET(hc, 0) + len + ((i != 0)? 2+LINK_SIZE : 1));
6065: cd->hwm += LINK_SIZE;
6066: }
6067: save_hwm = this_hwm;
6068: code += len;
6069: }
6070:
6071: /* Now chain through the pending brackets, and fill in their length
6072: fields (which are holding the chain links pro tem). */
6073:
6074: while (bralink != NULL)
6075: {
6076: int oldlinkoffset;
6077: int offset = (int)(code - bralink + 1);
6078: pcre_uchar *bra = code - offset;
6079: oldlinkoffset = GET(bra, 1);
6080: bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
6081: *code++ = OP_KET;
6082: PUTINC(code, 0, offset);
6083: PUT(bra, 1, offset);
6084: }
6085: }
6086:
6087: /* If the maximum is unlimited, set a repeater in the final copy. For
6088: ONCE brackets, that's all we need to do. However, possessively repeated
6089: ONCE brackets can be converted into non-capturing brackets, as the
6090: behaviour of (?:xx)++ is the same as (?>xx)++ and this saves having to
6091: deal with possessive ONCEs specially.
6092:
6093: Otherwise, when we are doing the actual compile phase, check to see
6094: whether this group is one that could match an empty string. If so,
6095: convert the initial operator to the S form (e.g. OP_BRA -> OP_SBRA) so
6096: that runtime checking can be done. [This check is also applied to ONCE
6097: groups at runtime, but in a different way.]
6098:
6099: Then, if the quantifier was possessive and the bracket is not a
6100: conditional, we convert the BRA code to the POS form, and the KET code to
6101: KETRPOS. (It turns out to be convenient at runtime to detect this kind of
6102: subpattern at both the start and at the end.) The use of special opcodes
6103: makes it possible to reduce greatly the stack usage in pcre_exec(). If
6104: the group is preceded by OP_BRAZERO, convert this to OP_BRAPOSZERO.
6105:
6106: Then, if the minimum number of matches is 1 or 0, cancel the possessive
6107: flag so that the default action below, of wrapping everything inside
6108: atomic brackets, does not happen. When the minimum is greater than 1,
6109: there will be earlier copies of the group, and so we still have to wrap
6110: the whole thing. */
6111:
6112: else
6113: {
6114: pcre_uchar *ketcode = code - 1 - LINK_SIZE;
6115: pcre_uchar *bracode = ketcode - GET(ketcode, 1);
6116:
6117: /* Convert possessive ONCE brackets to non-capturing */
6118:
6119: if ((*bracode == OP_ONCE || *bracode == OP_ONCE_NC) &&
6120: possessive_quantifier) *bracode = OP_BRA;
6121:
6122: /* For non-possessive ONCE brackets, all we need to do is to
6123: set the KET. */
6124:
6125: if (*bracode == OP_ONCE || *bracode == OP_ONCE_NC)
6126: *ketcode = OP_KETRMAX + repeat_type;
6127:
6128: /* Handle non-ONCE brackets and possessive ONCEs (which have been
6129: converted to non-capturing above). */
6130:
6131: else
6132: {
6133: /* In the compile phase, check for empty string matching. */
6134:
6135: if (lengthptr == NULL)
6136: {
6137: pcre_uchar *scode = bracode;
6138: do
6139: {
6140: if (could_be_empty_branch(scode, ketcode, utf, cd, NULL))
6141: {
6142: *bracode += OP_SBRA - OP_BRA;
6143: break;
6144: }
6145: scode += GET(scode, 1);
6146: }
6147: while (*scode == OP_ALT);
6148: }
6149:
6150: /* Handle possessive quantifiers. */
6151:
6152: if (possessive_quantifier)
6153: {
6154: /* For COND brackets, we wrap the whole thing in a possessively
6155: repeated non-capturing bracket, because we have not invented POS
6156: versions of the COND opcodes. Because we are moving code along, we
6157: must ensure that any pending recursive references are updated. */
6158:
6159: if (*bracode == OP_COND || *bracode == OP_SCOND)
6160: {
6161: int nlen = (int)(code - bracode);
6162: *code = OP_END;
6163: adjust_recurse(bracode, 1 + LINK_SIZE, utf, cd, save_hwm);
6164: memmove(bracode + 1 + LINK_SIZE, bracode, IN_UCHARS(nlen));
6165: code += 1 + LINK_SIZE;
6166: nlen += 1 + LINK_SIZE;
6167: *bracode = OP_BRAPOS;
6168: *code++ = OP_KETRPOS;
6169: PUTINC(code, 0, nlen);
6170: PUT(bracode, 1, nlen);
6171: }
6172:
6173: /* For non-COND brackets, we modify the BRA code and use KETRPOS. */
6174:
6175: else
6176: {
6177: *bracode += 1; /* Switch to xxxPOS opcodes */
6178: *ketcode = OP_KETRPOS;
6179: }
6180:
6181: /* If the minimum is zero, mark it as possessive, then unset the
6182: possessive flag when the minimum is 0 or 1. */
6183:
6184: if (brazeroptr != NULL) *brazeroptr = OP_BRAPOSZERO;
6185: if (repeat_min < 2) possessive_quantifier = FALSE;
6186: }
6187:
6188: /* Non-possessive quantifier */
6189:
6190: else *ketcode = OP_KETRMAX + repeat_type;
6191: }
6192: }
6193: }
6194:
6195: /* If previous is OP_FAIL, it was generated by an empty class [] in
6196: JavaScript mode. The other ways in which OP_FAIL can be generated, that is
6197: by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
6198: error above. We can just ignore the repeat in JS case. */
6199:
6200: else if (*previous == OP_FAIL) goto END_REPEAT;
6201:
6202: /* Else there's some kind of shambles */
6203:
6204: else
6205: {
6206: *errorcodeptr = ERR11;
6207: goto FAILED;
6208: }
6209:
6210: /* If the character following a repeat is '+', possessive_quantifier is
6211: TRUE. For some opcodes, there are special alternative opcodes for this
6212: case. For anything else, we wrap the entire repeated item inside OP_ONCE
6213: brackets. Logically, the '+' notation is just syntactic sugar, taken from
6214: Sun's Java package, but the special opcodes can optimize it.
6215:
6216: Some (but not all) possessively repeated subpatterns have already been
6217: completely handled in the code just above. For them, possessive_quantifier
6218: is always FALSE at this stage. Note that the repeated item starts at
6219: tempcode, not at previous, which might be the first part of a string whose
6220: (former) last char we repeated. */
6221:
6222: if (possessive_quantifier)
6223: {
6224: int len;
6225:
6226: /* Possessifying an EXACT quantifier has no effect, so we can ignore it.
6227: However, QUERY, STAR, or UPTO may follow (for quantifiers such as {5,6},
6228: {5,}, or {5,10}). We skip over an EXACT item; if the length of what
6229: remains is greater than zero, there's a further opcode that can be
6230: handled. If not, do nothing, leaving the EXACT alone. */
6231:
6232: switch(*tempcode)
6233: {
6234: case OP_TYPEEXACT:
6235: tempcode += PRIV(OP_lengths)[*tempcode] +
6236: ((tempcode[1 + IMM2_SIZE] == OP_PROP
6237: || tempcode[1 + IMM2_SIZE] == OP_NOTPROP)? 2 : 0);
6238: break;
6239:
6240: /* CHAR opcodes are used for exacts whose count is 1. */
6241:
6242: case OP_CHAR:
6243: case OP_CHARI:
6244: case OP_NOT:
6245: case OP_NOTI:
6246: case OP_EXACT:
6247: case OP_EXACTI:
6248: case OP_NOTEXACT:
6249: case OP_NOTEXACTI:
6250: tempcode += PRIV(OP_lengths)[*tempcode];
6251: #ifdef SUPPORT_UTF
6252: if (utf && HAS_EXTRALEN(tempcode[-1]))
6253: tempcode += GET_EXTRALEN(tempcode[-1]);
6254: #endif
6255: break;
6256:
6257: /* For the class opcodes, the repeat operator appears at the end;
6258: adjust tempcode to point to it. */
6259:
6260: case OP_CLASS:
6261: case OP_NCLASS:
6262: tempcode += 1 + 32/sizeof(pcre_uchar);
6263: break;
6264:
6265: #if defined SUPPORT_UTF || !defined COMPILE_PCRE8
6266: case OP_XCLASS:
6267: tempcode += GET(tempcode, 1);
6268: break;
6269: #endif
6270: }
6271:
6272: /* If tempcode is equal to code (which points to the end of the repeated
6273: item), it means we have skipped an EXACT item but there is no following
6274: QUERY, STAR, or UPTO; the value of len will be 0, and we do nothing. In
6275: all other cases, tempcode will be pointing to the repeat opcode, and will
6276: be less than code, so the value of len will be greater than 0. */
6277:
6278: len = (int)(code - tempcode);
6279: if (len > 0)
6280: {
6281: unsigned int repcode = *tempcode;
6282:
6283: /* There is a table for possessifying opcodes, all of which are less
6284: than OP_CALLOUT. A zero entry means there is no possessified version.
6285: */
6286:
6287: if (repcode < OP_CALLOUT && opcode_possessify[repcode] > 0)
6288: *tempcode = opcode_possessify[repcode];
6289:
6290: /* For opcode without a special possessified version, wrap the item in
6291: ONCE brackets. Because we are moving code along, we must ensure that any
6292: pending recursive references are updated. */
6293:
6294: else
6295: {
6296: *code = OP_END;
6297: adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6298: memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6299: code += 1 + LINK_SIZE;
6300: len += 1 + LINK_SIZE;
6301: tempcode[0] = OP_ONCE;
6302: *code++ = OP_KET;
6303: PUTINC(code, 0, len);
6304: PUT(tempcode, 1, len);
6305: }
6306: }
6307:
6308: #ifdef NEVER
6309: if (len > 0) switch (*tempcode)
6310: {
6311: case OP_STAR: *tempcode = OP_POSSTAR; break;
6312: case OP_PLUS: *tempcode = OP_POSPLUS; break;
6313: case OP_QUERY: *tempcode = OP_POSQUERY; break;
6314: case OP_UPTO: *tempcode = OP_POSUPTO; break;
6315:
6316: case OP_STARI: *tempcode = OP_POSSTARI; break;
6317: case OP_PLUSI: *tempcode = OP_POSPLUSI; break;
6318: case OP_QUERYI: *tempcode = OP_POSQUERYI; break;
6319: case OP_UPTOI: *tempcode = OP_POSUPTOI; break;
6320:
6321: case OP_NOTSTAR: *tempcode = OP_NOTPOSSTAR; break;
6322: case OP_NOTPLUS: *tempcode = OP_NOTPOSPLUS; break;
6323: case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
6324: case OP_NOTUPTO: *tempcode = OP_NOTPOSUPTO; break;
6325:
6326: case OP_NOTSTARI: *tempcode = OP_NOTPOSSTARI; break;
6327: case OP_NOTPLUSI: *tempcode = OP_NOTPOSPLUSI; break;
6328: case OP_NOTQUERYI: *tempcode = OP_NOTPOSQUERYI; break;
6329: case OP_NOTUPTOI: *tempcode = OP_NOTPOSUPTOI; break;
6330:
6331: case OP_TYPESTAR: *tempcode = OP_TYPEPOSSTAR; break;
6332: case OP_TYPEPLUS: *tempcode = OP_TYPEPOSPLUS; break;
6333: case OP_TYPEQUERY: *tempcode = OP_TYPEPOSQUERY; break;
6334: case OP_TYPEUPTO: *tempcode = OP_TYPEPOSUPTO; break;
6335:
6336: case OP_CRSTAR: *tempcode = OP_CRPOSSTAR; break;
6337: case OP_CRPLUS: *tempcode = OP_CRPOSPLUS; break;
6338: case OP_CRQUERY: *tempcode = OP_CRPOSQUERY; break;
6339: case OP_CRRANGE: *tempcode = OP_CRPOSRANGE; break;
6340:
6341: /* Because we are moving code along, we must ensure that any
6342: pending recursive references are updated. */
6343:
6344: default:
6345: *code = OP_END;
6346: adjust_recurse(tempcode, 1 + LINK_SIZE, utf, cd, save_hwm);
6347: memmove(tempcode + 1 + LINK_SIZE, tempcode, IN_UCHARS(len));
6348: code += 1 + LINK_SIZE;
6349: len += 1 + LINK_SIZE;
6350: tempcode[0] = OP_ONCE;
6351: *code++ = OP_KET;
6352: PUTINC(code, 0, len);
6353: PUT(tempcode, 1, len);
6354: break;
6355: }
6356: #endif
6357: }
6358:
6359: /* In all case we no longer have a previous item. We also set the
6360: "follows varying string" flag for subsequently encountered reqchars if
6361: it isn't already set and we have just passed a varying length item. */
6362:
6363: END_REPEAT:
6364: previous = NULL;
6365: cd->req_varyopt |= reqvary;
6366: break;
6367:
6368:
6369: /* ===================================================================*/
6370: /* Start of nested parenthesized sub-expression, or comment or lookahead or
6371: lookbehind or option setting or condition or all the other extended
6372: parenthesis forms. */
6373:
6374: case CHAR_LEFT_PARENTHESIS:
6375: newoptions = options;
6376: skipbytes = 0;
6377: bravalue = OP_CBRA;
6378: save_hwm = cd->hwm;
6379: reset_bracount = FALSE;
6380:
6381: /* First deal with various "verbs" that can be introduced by '*'. */
6382:
6383: ptr++;
6384: if (ptr[0] == CHAR_ASTERISK && (ptr[1] == ':'
6385: || (MAX_255(ptr[1]) && ((cd->ctypes[ptr[1]] & ctype_letter) != 0))))
6386: {
6387: int i, namelen;
6388: int arglen = 0;
6389: const char *vn = verbnames;
6390: const pcre_uchar *name = ptr + 1;
6391: const pcre_uchar *arg = NULL;
6392: previous = NULL;
6393: ptr++;
6394: while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;
6395: namelen = (int)(ptr - name);
6396:
6397: /* It appears that Perl allows any characters whatsoever, other than
6398: a closing parenthesis, to appear in arguments, so we no longer insist on
6399: letters, digits, and underscores. */
6400:
6401: if (*ptr == CHAR_COLON)
6402: {
6403: arg = ++ptr;
6404: while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6405: arglen = (int)(ptr - arg);
6406: if ((unsigned int)arglen > MAX_MARK)
6407: {
6408: *errorcodeptr = ERR75;
6409: goto FAILED;
6410: }
6411: }
6412:
6413: if (*ptr != CHAR_RIGHT_PARENTHESIS)
6414: {
6415: *errorcodeptr = ERR60;
6416: goto FAILED;
6417: }
6418:
6419: /* Scan the table of verb names */
6420:
6421: for (i = 0; i < verbcount; i++)
6422: {
6423: if (namelen == verbs[i].len &&
6424: STRNCMP_UC_C8(name, vn, namelen) == 0)
6425: {
6426: int setverb;
6427:
6428: /* Check for open captures before ACCEPT and convert it to
6429: ASSERT_ACCEPT if in an assertion. */
6430:
6431: if (verbs[i].op == OP_ACCEPT)
6432: {
6433: open_capitem *oc;
6434: if (arglen != 0)
6435: {
6436: *errorcodeptr = ERR59;
6437: goto FAILED;
6438: }
6439: cd->had_accept = TRUE;
6440: for (oc = cd->open_caps; oc != NULL; oc = oc->next)
6441: {
6442: *code++ = OP_CLOSE;
6443: PUT2INC(code, 0, oc->number);
6444: }
6445: setverb = *code++ =
6446: (cd->assert_depth > 0)? OP_ASSERT_ACCEPT : OP_ACCEPT;
6447:
6448: /* Do not set firstchar after *ACCEPT */
6449: if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
6450: }
6451:
6452: /* Handle other cases with/without an argument */
6453:
6454: else if (arglen == 0)
6455: {
6456: if (verbs[i].op < 0) /* Argument is mandatory */
6457: {
6458: *errorcodeptr = ERR66;
6459: goto FAILED;
6460: }
6461: setverb = *code++ = verbs[i].op;
6462: }
6463:
6464: else
6465: {
6466: if (verbs[i].op_arg < 0) /* Argument is forbidden */
6467: {
6468: *errorcodeptr = ERR59;
6469: goto FAILED;
6470: }
6471: setverb = *code++ = verbs[i].op_arg;
6472: *code++ = arglen;
6473: memcpy(code, arg, IN_UCHARS(arglen));
6474: code += arglen;
6475: *code++ = 0;
6476: }
6477:
6478: switch (setverb)
6479: {
6480: case OP_THEN:
6481: case OP_THEN_ARG:
6482: cd->external_flags |= PCRE_HASTHEN;
6483: break;
6484:
6485: case OP_PRUNE:
6486: case OP_PRUNE_ARG:
6487: case OP_SKIP:
6488: case OP_SKIP_ARG:
6489: cd->had_pruneorskip = TRUE;
6490: break;
6491: }
6492:
6493: break; /* Found verb, exit loop */
6494: }
6495:
6496: vn += verbs[i].len + 1;
6497: }
6498:
6499: if (i < verbcount) continue; /* Successfully handled a verb */
6500: *errorcodeptr = ERR60; /* Verb not recognized */
6501: goto FAILED;
6502: }
6503:
6504: /* Deal with the extended parentheses; all are introduced by '?', and the
6505: appearance of any of them means that this is not a capturing group. */
6506:
6507: else if (*ptr == CHAR_QUESTION_MARK)
6508: {
6509: int i, set, unset, namelen;
6510: int *optset;
6511: const pcre_uchar *name;
6512: pcre_uchar *slot;
6513:
6514: switch (*(++ptr))
6515: {
6516: case CHAR_NUMBER_SIGN: /* Comment; skip to ket */
6517: ptr++;
6518: while (*ptr != CHAR_NULL && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
6519: if (*ptr == CHAR_NULL)
6520: {
6521: *errorcodeptr = ERR18;
6522: goto FAILED;
6523: }
6524: continue;
6525:
6526:
6527: /* ------------------------------------------------------------ */
6528: case CHAR_VERTICAL_LINE: /* Reset capture count for each branch */
6529: reset_bracount = TRUE;
6530: /* Fall through */
6531:
6532: /* ------------------------------------------------------------ */
6533: case CHAR_COLON: /* Non-capturing bracket */
6534: bravalue = OP_BRA;
6535: ptr++;
6536: break;
6537:
6538:
6539: /* ------------------------------------------------------------ */
6540: case CHAR_LEFT_PARENTHESIS:
6541: bravalue = OP_COND; /* Conditional group */
6542: tempptr = ptr;
6543:
6544: /* A condition can be an assertion, a number (referring to a numbered
6545: group's having been set), a name (referring to a named group), or 'R',
6546: referring to recursion. R<digits> and R&name are also permitted for
6547: recursion tests.
6548:
6549: There are ways of testing a named group: (?(name)) is used by Python;
6550: Perl 5.10 onwards uses (?(<name>) or (?('name')).
6551:
6552: There is one unfortunate ambiguity, caused by history. 'R' can be the
6553: recursive thing or the name 'R' (and similarly for 'R' followed by
6554: digits). We look for a name first; if not found, we try the other case.
6555:
6556: For compatibility with auto-callouts, we allow a callout to be
6557: specified before a condition that is an assertion. First, check for the
6558: syntax of a callout; if found, adjust the temporary pointer that is
6559: used to check for an assertion condition. That's all that is needed! */
6560:
6561: if (ptr[1] == CHAR_QUESTION_MARK && ptr[2] == CHAR_C)
6562: {
6563: for (i = 3;; i++) if (!IS_DIGIT(ptr[i])) break;
6564: if (ptr[i] == CHAR_RIGHT_PARENTHESIS)
6565: tempptr += i + 1;
6566: }
6567:
6568: /* For conditions that are assertions, check the syntax, and then exit
6569: the switch. This will take control down to where bracketed groups,
6570: including assertions, are processed. */
6571:
6572: if (tempptr[1] == CHAR_QUESTION_MARK &&
6573: (tempptr[2] == CHAR_EQUALS_SIGN ||
6574: tempptr[2] == CHAR_EXCLAMATION_MARK ||
6575: tempptr[2] == CHAR_LESS_THAN_SIGN))
6576: break;
6577:
6578: /* Other conditions use OP_CREF/OP_DNCREF/OP_RREF/OP_DNRREF, and all
6579: need to skip at least 1+IMM2_SIZE bytes at the start of the group. */
6580:
6581: code[1+LINK_SIZE] = OP_CREF;
6582: skipbytes = 1+IMM2_SIZE;
6583: refsign = -1;
6584:
6585: /* Check for a test for recursion in a named group. */
6586:
6587: ptr++;
6588: if (*ptr == CHAR_R && ptr[1] == CHAR_AMPERSAND)
6589: {
6590: terminator = -1;
6591: ptr += 2;
6592: code[1+LINK_SIZE] = OP_RREF; /* Change the type of test */
6593: }
6594:
6595: /* Check for a test for a named group's having been set, using the Perl
6596: syntax (?(<name>) or (?('name'), and also allow for the original PCRE
6597: syntax of (?(name) or for (?(+n), (?(-n), and just (?(n). */
6598:
6599: else if (*ptr == CHAR_LESS_THAN_SIGN)
6600: {
6601: terminator = CHAR_GREATER_THAN_SIGN;
6602: ptr++;
6603: }
6604: else if (*ptr == CHAR_APOSTROPHE)
6605: {
6606: terminator = CHAR_APOSTROPHE;
6607: ptr++;
6608: }
6609: else
6610: {
6611: terminator = CHAR_NULL;
6612: if (*ptr == CHAR_MINUS || *ptr == CHAR_PLUS) refsign = *ptr++;
6613: else if (IS_DIGIT(*ptr)) refsign = 0;
6614: }
6615:
6616: /* Handle a number */
6617:
6618: if (refsign >= 0)
6619: {
6620: recno = 0;
6621: while (IS_DIGIT(*ptr))
6622: {
6623: recno = recno * 10 + (int)(*ptr - CHAR_0);
6624: ptr++;
6625: }
6626: }
6627:
6628: /* Otherwise we expect to read a name; anything else is an error. When
6629: a name is one of a number of duplicates, a different opcode is used and
6630: it needs more memory. Unfortunately we cannot tell whether a name is a
6631: duplicate in the first pass, so we have to allow for more memory. */
6632:
6633: else
6634: {
6635: if (IS_DIGIT(*ptr))
6636: {
6637: *errorcodeptr = ERR84;
6638: goto FAILED;
6639: }
6640: if (!MAX_255(*ptr) || (cd->ctypes[*ptr] & ctype_word) == 0)
6641: {
6642: *errorcodeptr = ERR28; /* Assertion expected */
6643: goto FAILED;
6644: }
6645: name = ptr++;
6646: while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0)
6647: {
6648: ptr++;
6649: }
6650: namelen = (int)(ptr - name);
6651: if (lengthptr != NULL) *lengthptr += IMM2_SIZE;
6652: }
6653:
6654: /* Check the terminator */
6655:
6656: if ((terminator > 0 && *ptr++ != (pcre_uchar)terminator) ||
6657: *ptr++ != CHAR_RIGHT_PARENTHESIS)
6658: {
6659: ptr--; /* Error offset */
6660: *errorcodeptr = ERR26; /* Malformed number or name */
6661: goto FAILED;
6662: }
6663:
6664: /* Do no further checking in the pre-compile phase. */
6665:
6666: if (lengthptr != NULL) break;
6667:
6668: /* In the real compile we do the work of looking for the actual
6669: reference. If refsign is not negative, it means we have a number in
6670: recno. */
6671:
6672: if (refsign >= 0)
6673: {
6674: if (recno <= 0)
6675: {
6676: *errorcodeptr = ERR35;
6677: goto FAILED;
6678: }
6679: if (refsign != 0) recno = (refsign == CHAR_MINUS)?
6680: cd->bracount - recno + 1 : recno + cd->bracount;
6681: if (recno <= 0 || recno > cd->final_bracount)
6682: {
6683: *errorcodeptr = ERR15;
6684: goto FAILED;
6685: }
6686: PUT2(code, 2+LINK_SIZE, recno);
6687: break;
6688: }
6689:
6690: /* Otherwise look for the name. */
6691:
6692: slot = cd->name_table;
6693: for (i = 0; i < cd->names_found; i++)
6694: {
6695: if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0) break;
6696: slot += cd->name_entry_size;
6697: }
6698:
6699: /* Found the named subpattern. If the name is duplicated, add one to
6700: the opcode to change CREF/RREF into DNCREF/DNRREF and insert
6701: appropriate data values. Otherwise, just insert the unique subpattern
6702: number. */
6703:
6704: if (i < cd->names_found)
6705: {
6706: int offset = i++;
6707: int count = 1;
6708: recno = GET2(slot, 0); /* Number from first found */
6709: for (; i < cd->names_found; i++)
6710: {
6711: slot += cd->name_entry_size;
6712: if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) != 0) break;
6713: count++;
6714: }
6715: if (count > 1)
6716: {
6717: PUT2(code, 2+LINK_SIZE, offset);
6718: PUT2(code, 2+LINK_SIZE+IMM2_SIZE, count);
6719: skipbytes += IMM2_SIZE;
6720: code[1+LINK_SIZE]++;
6721: }
6722: else /* Not a duplicated name */
6723: {
6724: PUT2(code, 2+LINK_SIZE, recno);
6725: }
6726: }
6727:
6728: /* If terminator == CHAR_NULL it means that the name followed directly
6729: after the opening parenthesis [e.g. (?(abc)...] and in this case there
6730: are some further alternatives to try. For the cases where terminator !=
6731: CHAR_NULL [things like (?(<name>... or (?('name')... or (?(R&name)... ]
6732: we have now checked all the possibilities, so give an error. */
6733:
6734: else if (terminator != CHAR_NULL)
6735: {
6736: *errorcodeptr = ERR15;
6737: goto FAILED;
6738: }
6739:
6740: /* Check for (?(R) for recursion. Allow digits after R to specify a
6741: specific group number. */
6742:
6743: else if (*name == CHAR_R)
6744: {
6745: recno = 0;
6746: for (i = 1; i < namelen; i++)
6747: {
6748: if (!IS_DIGIT(name[i]))
6749: {
6750: *errorcodeptr = ERR15;
6751: goto FAILED;
6752: }
6753: recno = recno * 10 + name[i] - CHAR_0;
6754: }
6755: if (recno == 0) recno = RREF_ANY;
6756: code[1+LINK_SIZE] = OP_RREF; /* Change test type */
6757: PUT2(code, 2+LINK_SIZE, recno);
6758: }
6759:
6760: /* Similarly, check for the (?(DEFINE) "condition", which is always
6761: false. */
6762:
6763: else if (namelen == 6 && STRNCMP_UC_C8(name, STRING_DEFINE, 6) == 0)
6764: {
6765: code[1+LINK_SIZE] = OP_DEF;
6766: skipbytes = 1;
6767: }
6768:
6769: /* Reference to an unidentified subpattern. */
6770:
6771: else
6772: {
6773: *errorcodeptr = ERR15;
6774: goto FAILED;
6775: }
6776: break;
6777:
6778:
6779: /* ------------------------------------------------------------ */
6780: case CHAR_EQUALS_SIGN: /* Positive lookahead */
6781: bravalue = OP_ASSERT;
6782: cd->assert_depth += 1;
6783: ptr++;
6784: break;
6785:
6786: /* Optimize (?!) to (*FAIL) unless it is quantified - which is a weird
6787: thing to do, but Perl allows all assertions to be quantified, and when
6788: they contain capturing parentheses there may be a potential use for
6789: this feature. Not that that applies to a quantified (?!) but we allow
6790: it for uniformity. */
6791:
6792: /* ------------------------------------------------------------ */
6793: case CHAR_EXCLAMATION_MARK: /* Negative lookahead */
6794: ptr++;
6795: if (*ptr == CHAR_RIGHT_PARENTHESIS && ptr[1] != CHAR_ASTERISK &&
6796: ptr[1] != CHAR_PLUS && ptr[1] != CHAR_QUESTION_MARK &&
6797: (ptr[1] != CHAR_LEFT_CURLY_BRACKET || !is_counted_repeat(ptr+2)))
6798: {
6799: *code++ = OP_FAIL;
6800: previous = NULL;
6801: continue;
6802: }
6803: bravalue = OP_ASSERT_NOT;
6804: cd->assert_depth += 1;
6805: break;
6806:
6807:
6808: /* ------------------------------------------------------------ */
6809: case CHAR_LESS_THAN_SIGN: /* Lookbehind or named define */
6810: switch (ptr[1])
6811: {
6812: case CHAR_EQUALS_SIGN: /* Positive lookbehind */
6813: bravalue = OP_ASSERTBACK;
6814: cd->assert_depth += 1;
6815: ptr += 2;
6816: break;
6817:
6818: case CHAR_EXCLAMATION_MARK: /* Negative lookbehind */
6819: bravalue = OP_ASSERTBACK_NOT;
6820: cd->assert_depth += 1;
6821: ptr += 2;
6822: break;
6823:
6824: default: /* Could be name define, else bad */
6825: if (MAX_255(ptr[1]) && (cd->ctypes[ptr[1]] & ctype_word) != 0)
6826: goto DEFINE_NAME;
6827: ptr++; /* Correct offset for error */
6828: *errorcodeptr = ERR24;
6829: goto FAILED;
6830: }
6831: break;
6832:
6833:
6834: /* ------------------------------------------------------------ */
6835: case CHAR_GREATER_THAN_SIGN: /* One-time brackets */
6836: bravalue = OP_ONCE;
6837: ptr++;
6838: break;
6839:
6840:
6841: /* ------------------------------------------------------------ */
6842: case CHAR_C: /* Callout - may be followed by digits; */
6843: previous_callout = code; /* Save for later completion */
6844: after_manual_callout = 1; /* Skip one item before completing */
6845: *code++ = OP_CALLOUT;
6846: {
6847: int n = 0;
6848: ptr++;
6849: while(IS_DIGIT(*ptr))
6850: n = n * 10 + *ptr++ - CHAR_0;
6851: if (*ptr != CHAR_RIGHT_PARENTHESIS)
6852: {
6853: *errorcodeptr = ERR39;
6854: goto FAILED;
6855: }
6856: if (n > 255)
6857: {
6858: *errorcodeptr = ERR38;
6859: goto FAILED;
6860: }
6861: *code++ = n;
6862: PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
6863: PUT(code, LINK_SIZE, 0); /* Default length */
6864: code += 2 * LINK_SIZE;
6865: }
6866: previous = NULL;
6867: continue;
6868:
6869:
6870: /* ------------------------------------------------------------ */
6871: case CHAR_P: /* Python-style named subpattern handling */
6872: if (*(++ptr) == CHAR_EQUALS_SIGN ||
6873: *ptr == CHAR_GREATER_THAN_SIGN) /* Reference or recursion */
6874: {
6875: is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
6876: terminator = CHAR_RIGHT_PARENTHESIS;
6877: goto NAMED_REF_OR_RECURSE;
6878: }
6879: else if (*ptr != CHAR_LESS_THAN_SIGN) /* Test for Python-style defn */
6880: {
6881: *errorcodeptr = ERR41;
6882: goto FAILED;
6883: }
6884: /* Fall through to handle (?P< as (?< is handled */
6885:
6886:
6887: /* ------------------------------------------------------------ */
6888: DEFINE_NAME: /* Come here from (?< handling */
6889: case CHAR_APOSTROPHE:
6890: terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
6891: CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
6892: name = ++ptr;
6893: if (IS_DIGIT(*ptr))
6894: {
6895: *errorcodeptr = ERR84; /* Group name must start with non-digit */
6896: goto FAILED;
6897: }
6898: while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
6899: namelen = (int)(ptr - name);
6900:
6901: /* In the pre-compile phase, do a syntax check, remember the longest
6902: name, and then remember the group in a vector, expanding it if
6903: necessary. Duplicates for the same number are skipped; other duplicates
6904: are checked for validity. In the actual compile, there is nothing to
6905: do. */
6906:
6907: if (lengthptr != NULL)
6908: {
6909: named_group *ng;
6910: pcre_uint32 number = cd->bracount + 1;
6911:
6912: if (*ptr != (pcre_uchar)terminator)
6913: {
6914: *errorcodeptr = ERR42;
6915: goto FAILED;
6916: }
6917:
6918: if (cd->names_found >= MAX_NAME_COUNT)
6919: {
6920: *errorcodeptr = ERR49;
6921: goto FAILED;
6922: }
6923:
6924: if (namelen + IMM2_SIZE + 1 > cd->name_entry_size)
6925: {
6926: cd->name_entry_size = namelen + IMM2_SIZE + 1;
6927: if (namelen > MAX_NAME_SIZE)
6928: {
6929: *errorcodeptr = ERR48;
6930: goto FAILED;
6931: }
6932: }
6933:
6934: /* Scan the list to check for duplicates. For duplicate names, if the
6935: number is the same, break the loop, which causes the name to be
6936: discarded; otherwise, if DUPNAMES is not set, give an error.
6937: If it is set, allow the name with a different number, but continue
6938: scanning in case this is a duplicate with the same number. For
6939: non-duplicate names, give an error if the number is duplicated. */
6940:
6941: ng = cd->named_groups;
6942: for (i = 0; i < cd->names_found; i++, ng++)
6943: {
6944: if (namelen == ng->length &&
6945: STRNCMP_UC_UC(name, ng->name, namelen) == 0)
6946: {
6947: if (ng->number == number) break;
6948: if ((options & PCRE_DUPNAMES) == 0)
6949: {
6950: *errorcodeptr = ERR43;
6951: goto FAILED;
6952: }
6953: cd->dupnames = TRUE; /* Duplicate names exist */
6954: }
6955: else if (ng->number == number)
6956: {
6957: *errorcodeptr = ERR65;
6958: goto FAILED;
6959: }
6960: }
6961:
6962: if (i >= cd->names_found) /* Not a duplicate with same number */
6963: {
6964: /* Increase the list size if necessary */
6965:
6966: if (cd->names_found >= cd->named_group_list_size)
6967: {
6968: int newsize = cd->named_group_list_size * 2;
6969: named_group *newspace = (PUBL(malloc))
6970: (newsize * sizeof(named_group));
6971:
6972: if (newspace == NULL)
6973: {
6974: *errorcodeptr = ERR21;
6975: goto FAILED;
6976: }
6977:
6978: memcpy(newspace, cd->named_groups,
6979: cd->named_group_list_size * sizeof(named_group));
6980: if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
6981: (PUBL(free))((void *)cd->named_groups);
6982: cd->named_groups = newspace;
6983: cd->named_group_list_size = newsize;
6984: }
6985:
6986: cd->named_groups[cd->names_found].name = name;
6987: cd->named_groups[cd->names_found].length = namelen;
6988: cd->named_groups[cd->names_found].number = number;
6989: cd->names_found++;
6990: }
6991: }
6992:
6993: ptr++; /* Move past > or ' in both passes. */
6994: goto NUMBERED_GROUP;
6995:
6996:
6997: /* ------------------------------------------------------------ */
6998: case CHAR_AMPERSAND: /* Perl recursion/subroutine syntax */
6999: terminator = CHAR_RIGHT_PARENTHESIS;
7000: is_recurse = TRUE;
7001: /* Fall through */
7002:
7003: /* We come here from the Python syntax above that handles both
7004: references (?P=name) and recursion (?P>name), as well as falling
7005: through from the Perl recursion syntax (?&name). We also come here from
7006: the Perl \k<name> or \k'name' back reference syntax and the \k{name}
7007: .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
7008:
7009: NAMED_REF_OR_RECURSE:
7010: name = ++ptr;
7011: if (IS_DIGIT(*ptr))
7012: {
7013: *errorcodeptr = ERR84; /* Group name must start with non-digit */
7014: goto FAILED;
7015: }
7016: while (MAX_255(*ptr) && (cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
7017: namelen = (int)(ptr - name);
7018:
7019: /* In the pre-compile phase, do a syntax check. We used to just set
7020: a dummy reference number, because it was not used in the first pass.
7021: However, with the change of recursive back references to be atomic,
7022: we have to look for the number so that this state can be identified, as
7023: otherwise the incorrect length is computed. If it's not a backwards
7024: reference, the dummy number will do. */
7025:
7026: if (lengthptr != NULL)
7027: {
7028: named_group *ng;
7029:
7030: if (namelen == 0)
7031: {
7032: *errorcodeptr = ERR62;
7033: goto FAILED;
7034: }
7035: if (*ptr != (pcre_uchar)terminator)
7036: {
7037: *errorcodeptr = ERR42;
7038: goto FAILED;
7039: }
7040: if (namelen > MAX_NAME_SIZE)
7041: {
7042: *errorcodeptr = ERR48;
7043: goto FAILED;
7044: }
7045:
7046: /* The name table does not exist in the first pass; instead we must
7047: scan the list of names encountered so far in order to get the
7048: number. If the name is not found, set the value to 0 for a forward
7049: reference. */
7050:
7051: ng = cd->named_groups;
7052: for (i = 0; i < cd->names_found; i++, ng++)
7053: {
7054: if (namelen == ng->length &&
7055: STRNCMP_UC_UC(name, ng->name, namelen) == 0)
7056: break;
7057: }
7058: recno = (i < cd->names_found)? ng->number : 0;
7059:
7060: /* Count named back references. */
7061:
7062: if (!is_recurse) cd->namedrefcount++;
7063: }
7064:
7065: /* In the real compile, search the name table. We check the name
7066: first, and then check that we have reached the end of the name in the
7067: table. That way, if the name is longer than any in the table, the
7068: comparison will fail without reading beyond the table entry. */
7069:
7070: else
7071: {
7072: slot = cd->name_table;
7073: for (i = 0; i < cd->names_found; i++)
7074: {
7075: if (STRNCMP_UC_UC(name, slot+IMM2_SIZE, namelen) == 0 &&
7076: slot[IMM2_SIZE+namelen] == 0)
7077: break;
7078: slot += cd->name_entry_size;
7079: }
7080:
7081: if (i < cd->names_found)
7082: {
7083: recno = GET2(slot, 0);
7084: }
7085: else
7086: {
7087: *errorcodeptr = ERR15;
7088: goto FAILED;
7089: }
7090: }
7091:
7092: /* In both phases, for recursions, we can now go to the code than
7093: handles numerical recursion. */
7094:
7095: if (is_recurse) goto HANDLE_RECURSION;
7096:
7097: /* In the second pass we must see if the name is duplicated. If so, we
7098: generate a different opcode. */
7099:
7100: if (lengthptr == NULL && cd->dupnames)
7101: {
7102: int count = 1;
7103: unsigned int index = i;
7104: pcre_uchar *cslot = slot + cd->name_entry_size;
7105:
7106: for (i++; i < cd->names_found; i++)
7107: {
7108: if (STRCMP_UC_UC(slot + IMM2_SIZE, cslot + IMM2_SIZE) != 0) break;
7109: count++;
7110: cslot += cd->name_entry_size;
7111: }
7112:
7113: if (count > 1)
7114: {
7115: if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
7116: previous = code;
7117: *code++ = ((options & PCRE_CASELESS) != 0)? OP_DNREFI : OP_DNREF;
7118: PUT2INC(code, 0, index);
7119: PUT2INC(code, 0, count);
7120:
7121: /* Process each potentially referenced group. */
7122:
7123: for (; slot < cslot; slot += cd->name_entry_size)
7124: {
7125: open_capitem *oc;
7126: recno = GET2(slot, 0);
7127: cd->backref_map |= (recno < 32)? (1 << recno) : 1;
7128: if (recno > cd->top_backref) cd->top_backref = recno;
7129:
7130: /* Check to see if this back reference is recursive, that it, it
7131: is inside the group that it references. A flag is set so that the
7132: group can be made atomic. */
7133:
7134: for (oc = cd->open_caps; oc != NULL; oc = oc->next)
7135: {
7136: if (oc->number == recno)
7137: {
7138: oc->flag = TRUE;
7139: break;
7140: }
7141: }
7142: }
7143:
7144: continue; /* End of back ref handling */
7145: }
7146: }
7147:
7148: /* First pass, or a non-duplicated name. */
7149:
7150: goto HANDLE_REFERENCE;
7151:
7152:
7153: /* ------------------------------------------------------------ */
7154: case CHAR_R: /* Recursion */
7155: ptr++; /* Same as (?0) */
7156: /* Fall through */
7157:
7158:
7159: /* ------------------------------------------------------------ */
7160: case CHAR_MINUS: case CHAR_PLUS: /* Recursion or subroutine */
7161: case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
7162: case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
7163: {
7164: const pcre_uchar *called;
7165: terminator = CHAR_RIGHT_PARENTHESIS;
7166:
7167: /* Come here from the \g<...> and \g'...' code (Oniguruma
7168: compatibility). However, the syntax has been checked to ensure that
7169: the ... are a (signed) number, so that neither ERR63 nor ERR29 will
7170: be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
7171: ever be taken. */
7172:
7173: HANDLE_NUMERICAL_RECURSION:
7174:
7175: if ((refsign = *ptr) == CHAR_PLUS)
7176: {
7177: ptr++;
7178: if (!IS_DIGIT(*ptr))
7179: {
7180: *errorcodeptr = ERR63;
7181: goto FAILED;
7182: }
7183: }
7184: else if (refsign == CHAR_MINUS)
7185: {
7186: if (!IS_DIGIT(ptr[1]))
7187: goto OTHER_CHAR_AFTER_QUERY;
7188: ptr++;
7189: }
7190:
7191: recno = 0;
7192: while(IS_DIGIT(*ptr))
7193: recno = recno * 10 + *ptr++ - CHAR_0;
7194:
7195: if (*ptr != (pcre_uchar)terminator)
7196: {
7197: *errorcodeptr = ERR29;
7198: goto FAILED;
7199: }
7200:
7201: if (refsign == CHAR_MINUS)
7202: {
7203: if (recno == 0)
7204: {
7205: *errorcodeptr = ERR58;
7206: goto FAILED;
7207: }
7208: recno = cd->bracount - recno + 1;
7209: if (recno <= 0)
7210: {
7211: *errorcodeptr = ERR15;
7212: goto FAILED;
7213: }
7214: }
7215: else if (refsign == CHAR_PLUS)
7216: {
7217: if (recno == 0)
7218: {
7219: *errorcodeptr = ERR58;
7220: goto FAILED;
7221: }
7222: recno += cd->bracount;
7223: }
7224:
7225: /* Come here from code above that handles a named recursion */
7226:
7227: HANDLE_RECURSION:
7228:
7229: previous = code;
7230: called = cd->start_code;
7231:
7232: /* When we are actually compiling, find the bracket that is being
7233: referenced. Temporarily end the regex in case it doesn't exist before
7234: this point. If we end up with a forward reference, first check that
7235: the bracket does occur later so we can give the error (and position)
7236: now. Then remember this forward reference in the workspace so it can
7237: be filled in at the end. */
7238:
7239: if (lengthptr == NULL)
7240: {
7241: *code = OP_END;
7242: if (recno != 0)
7243: called = PRIV(find_bracket)(cd->start_code, utf, recno);
7244:
7245: /* Forward reference */
7246:
7247: if (called == NULL)
7248: {
7249: if (recno > cd->final_bracount)
7250: {
7251: *errorcodeptr = ERR15;
7252: goto FAILED;
7253: }
7254:
7255: /* Fudge the value of "called" so that when it is inserted as an
7256: offset below, what it actually inserted is the reference number
7257: of the group. Then remember the forward reference. */
7258:
7259: called = cd->start_code + recno;
7260: if (cd->hwm >= cd->start_workspace + cd->workspace_size -
7261: WORK_SIZE_SAFETY_MARGIN)
7262: {
7263: *errorcodeptr = expand_workspace(cd);
7264: if (*errorcodeptr != 0) goto FAILED;
7265: }
7266: PUTINC(cd->hwm, 0, (int)(code + 1 - cd->start_code));
7267: }
7268:
7269: /* If not a forward reference, and the subpattern is still open,
7270: this is a recursive call. We check to see if this is a left
7271: recursion that could loop for ever, and diagnose that case. We
7272: must not, however, do this check if we are in a conditional
7273: subpattern because the condition might be testing for recursion in
7274: a pattern such as /(?(R)a+|(?R)b)/, which is perfectly valid.
7275: Forever loops are also detected at runtime, so those that occur in
7276: conditional subpatterns will be picked up then. */
7277:
7278: else if (GET(called, 1) == 0 && cond_depth <= 0 &&
7279: could_be_empty(called, code, bcptr, utf, cd))
7280: {
7281: *errorcodeptr = ERR40;
7282: goto FAILED;
7283: }
7284: }
7285:
7286: /* Insert the recursion/subroutine item. It does not have a set first
7287: character (relevant if it is repeated, because it will then be
7288: wrapped with ONCE brackets). */
7289:
7290: *code = OP_RECURSE;
7291: PUT(code, 1, (int)(called - cd->start_code));
7292: code += 1 + LINK_SIZE;
7293: groupsetfirstchar = FALSE;
7294: }
7295:
7296: /* Can't determine a first byte now */
7297:
7298: if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
7299: continue;
7300:
7301:
7302: /* ------------------------------------------------------------ */
7303: default: /* Other characters: check option setting */
7304: OTHER_CHAR_AFTER_QUERY:
7305: set = unset = 0;
7306: optset = &set;
7307:
7308: while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
7309: {
7310: switch (*ptr++)
7311: {
7312: case CHAR_MINUS: optset = &unset; break;
7313:
7314: case CHAR_J: /* Record that it changed in the external options */
7315: *optset |= PCRE_DUPNAMES;
7316: cd->external_flags |= PCRE_JCHANGED;
7317: break;
7318:
7319: case CHAR_i: *optset |= PCRE_CASELESS; break;
7320: case CHAR_m: *optset |= PCRE_MULTILINE; break;
7321: case CHAR_s: *optset |= PCRE_DOTALL; break;
7322: case CHAR_x: *optset |= PCRE_EXTENDED; break;
7323: case CHAR_U: *optset |= PCRE_UNGREEDY; break;
7324: case CHAR_X: *optset |= PCRE_EXTRA; break;
7325:
7326: default: *errorcodeptr = ERR12;
7327: ptr--; /* Correct the offset */
7328: goto FAILED;
7329: }
7330: }
7331:
7332: /* Set up the changed option bits, but don't change anything yet. */
7333:
7334: newoptions = (options | set) & (~unset);
7335:
7336: /* If the options ended with ')' this is not the start of a nested
7337: group with option changes, so the options change at this level. If this
7338: item is right at the start of the pattern, the options can be
7339: abstracted and made external in the pre-compile phase, and ignored in
7340: the compile phase. This can be helpful when matching -- for instance in
7341: caseless checking of required bytes.
7342:
7343: If the code pointer is not (cd->start_code + 1 + LINK_SIZE), we are
7344: definitely *not* at the start of the pattern because something has been
7345: compiled. In the pre-compile phase, however, the code pointer can have
7346: that value after the start, because it gets reset as code is discarded
7347: during the pre-compile. However, this can happen only at top level - if
7348: we are within parentheses, the starting BRA will still be present. At
7349: any parenthesis level, the length value can be used to test if anything
7350: has been compiled at that level. Thus, a test for both these conditions
7351: is necessary to ensure we correctly detect the start of the pattern in
7352: both phases.
7353:
7354: If we are not at the pattern start, reset the greedy defaults and the
7355: case value for firstchar and reqchar. */
7356:
7357: if (*ptr == CHAR_RIGHT_PARENTHESIS)
7358: {
7359: if (code == cd->start_code + 1 + LINK_SIZE &&
7360: (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
7361: {
7362: cd->external_options = newoptions;
7363: }
7364: else
7365: {
7366: greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
7367: greedy_non_default = greedy_default ^ 1;
7368: req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS:0;
7369: }
7370:
7371: /* Change options at this level, and pass them back for use
7372: in subsequent branches. */
7373:
7374: *optionsptr = options = newoptions;
7375: previous = NULL; /* This item can't be repeated */
7376: continue; /* It is complete */
7377: }
7378:
7379: /* If the options ended with ':' we are heading into a nested group
7380: with possible change of options. Such groups are non-capturing and are
7381: not assertions of any kind. All we need to do is skip over the ':';
7382: the newoptions value is handled below. */
7383:
7384: bravalue = OP_BRA;
7385: ptr++;
7386: } /* End of switch for character following (? */
7387: } /* End of (? handling */
7388:
7389: /* Opening parenthesis not followed by '*' or '?'. If PCRE_NO_AUTO_CAPTURE
7390: is set, all unadorned brackets become non-capturing and behave like (?:...)
7391: brackets. */
7392:
7393: else if ((options & PCRE_NO_AUTO_CAPTURE) != 0)
7394: {
7395: bravalue = OP_BRA;
7396: }
7397:
7398: /* Else we have a capturing group. */
7399:
7400: else
7401: {
7402: NUMBERED_GROUP:
7403: cd->bracount += 1;
7404: PUT2(code, 1+LINK_SIZE, cd->bracount);
7405: skipbytes = IMM2_SIZE;
7406: }
7407:
7408: /* Process nested bracketed regex. First check for parentheses nested too
7409: deeply. */
7410:
7411: if ((cd->parens_depth += 1) > PARENS_NEST_LIMIT)
7412: {
7413: *errorcodeptr = ERR82;
7414: goto FAILED;
7415: }
7416:
7417: /* Assertions used not to be repeatable, but this was changed for Perl
7418: compatibility, so all kinds can now be repeated. We copy code into a
7419: non-register variable (tempcode) in order to be able to pass its address
7420: because some compilers complain otherwise. */
7421:
7422: previous = code; /* For handling repetition */
7423: *code = bravalue;
7424: tempcode = code;
7425: tempreqvary = cd->req_varyopt; /* Save value before bracket */
7426: tempbracount = cd->bracount; /* Save value before bracket */
7427: length_prevgroup = 0; /* Initialize for pre-compile phase */
7428:
7429: if (!compile_regex(
7430: newoptions, /* The complete new option state */
7431: &tempcode, /* Where to put code (updated) */
7432: &ptr, /* Input pointer (updated) */
7433: errorcodeptr, /* Where to put an error message */
7434: (bravalue == OP_ASSERTBACK ||
7435: bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
7436: reset_bracount, /* True if (?| group */
7437: skipbytes, /* Skip over bracket number */
7438: cond_depth +
7439: ((bravalue == OP_COND)?1:0), /* Depth of condition subpatterns */
7440: &subfirstchar, /* For possible first char */
7441: &subfirstcharflags,
7442: &subreqchar, /* For possible last char */
7443: &subreqcharflags,
7444: bcptr, /* Current branch chain */
7445: cd, /* Tables block */
7446: (lengthptr == NULL)? NULL : /* Actual compile phase */
7447: &length_prevgroup /* Pre-compile phase */
7448: ))
7449: goto FAILED;
7450:
7451: cd->parens_depth -= 1;
7452:
7453: /* If this was an atomic group and there are no capturing groups within it,
7454: generate OP_ONCE_NC instead of OP_ONCE. */
7455:
7456: if (bravalue == OP_ONCE && cd->bracount <= tempbracount)
7457: *code = OP_ONCE_NC;
7458:
7459: if (bravalue >= OP_ASSERT && bravalue <= OP_ASSERTBACK_NOT)
7460: cd->assert_depth -= 1;
7461:
7462: /* At the end of compiling, code is still pointing to the start of the
7463: group, while tempcode has been updated to point past the end of the group.
7464: The pattern pointer (ptr) is on the bracket.
7465:
7466: If this is a conditional bracket, check that there are no more than
7467: two branches in the group, or just one if it's a DEFINE group. We do this
7468: in the real compile phase, not in the pre-pass, where the whole group may
7469: not be available. */
7470:
7471: if (bravalue == OP_COND && lengthptr == NULL)
7472: {
7473: pcre_uchar *tc = code;
7474: int condcount = 0;
7475:
7476: do {
7477: condcount++;
7478: tc += GET(tc,1);
7479: }
7480: while (*tc != OP_KET);
7481:
7482: /* A DEFINE group is never obeyed inline (the "condition" is always
7483: false). It must have only one branch. */
7484:
7485: if (code[LINK_SIZE+1] == OP_DEF)
7486: {
7487: if (condcount > 1)
7488: {
7489: *errorcodeptr = ERR54;
7490: goto FAILED;
7491: }
7492: bravalue = OP_DEF; /* Just a flag to suppress char handling below */
7493: }
7494:
7495: /* A "normal" conditional group. If there is just one branch, we must not
7496: make use of its firstchar or reqchar, because this is equivalent to an
7497: empty second branch. */
7498:
7499: else
7500: {
7501: if (condcount > 2)
7502: {
7503: *errorcodeptr = ERR27;
7504: goto FAILED;
7505: }
7506: if (condcount == 1) subfirstcharflags = subreqcharflags = REQ_NONE;
7507: }
7508: }
7509:
7510: /* Error if hit end of pattern */
7511:
7512: if (*ptr != CHAR_RIGHT_PARENTHESIS)
7513: {
7514: *errorcodeptr = ERR14;
7515: goto FAILED;
7516: }
7517:
7518: /* In the pre-compile phase, update the length by the length of the group,
7519: less the brackets at either end. Then reduce the compiled code to just a
7520: set of non-capturing brackets so that it doesn't use much memory if it is
7521: duplicated by a quantifier.*/
7522:
7523: if (lengthptr != NULL)
7524: {
7525: if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
7526: {
7527: *errorcodeptr = ERR20;
7528: goto FAILED;
7529: }
7530: *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
7531: code++; /* This already contains bravalue */
7532: PUTINC(code, 0, 1 + LINK_SIZE);
7533: *code++ = OP_KET;
7534: PUTINC(code, 0, 1 + LINK_SIZE);
7535: break; /* No need to waste time with special character handling */
7536: }
7537:
7538: /* Otherwise update the main code pointer to the end of the group. */
7539:
7540: code = tempcode;
7541:
7542: /* For a DEFINE group, required and first character settings are not
7543: relevant. */
7544:
7545: if (bravalue == OP_DEF) break;
7546:
7547: /* Handle updating of the required and first characters for other types of
7548: group. Update for normal brackets of all kinds, and conditions with two
7549: branches (see code above). If the bracket is followed by a quantifier with
7550: zero repeat, we have to back off. Hence the definition of zeroreqchar and
7551: zerofirstchar outside the main loop so that they can be accessed for the
7552: back off. */
7553:
7554: zeroreqchar = reqchar;
7555: zeroreqcharflags = reqcharflags;
7556: zerofirstchar = firstchar;
7557: zerofirstcharflags = firstcharflags;
7558: groupsetfirstchar = FALSE;
7559:
7560: if (bravalue >= OP_ONCE)
7561: {
7562: /* If we have not yet set a firstchar in this branch, take it from the
7563: subpattern, remembering that it was set here so that a repeat of more
7564: than one can replicate it as reqchar if necessary. If the subpattern has
7565: no firstchar, set "none" for the whole branch. In both cases, a zero
7566: repeat forces firstchar to "none". */
7567:
7568: if (firstcharflags == REQ_UNSET)
7569: {
7570: if (subfirstcharflags >= 0)
7571: {
7572: firstchar = subfirstchar;
7573: firstcharflags = subfirstcharflags;
7574: groupsetfirstchar = TRUE;
7575: }
7576: else firstcharflags = REQ_NONE;
7577: zerofirstcharflags = REQ_NONE;
7578: }
7579:
7580: /* If firstchar was previously set, convert the subpattern's firstchar
7581: into reqchar if there wasn't one, using the vary flag that was in
7582: existence beforehand. */
7583:
7584: else if (subfirstcharflags >= 0 && subreqcharflags < 0)
7585: {
7586: subreqchar = subfirstchar;
7587: subreqcharflags = subfirstcharflags | tempreqvary;
7588: }
7589:
7590: /* If the subpattern set a required byte (or set a first byte that isn't
7591: really the first byte - see above), set it. */
7592:
7593: if (subreqcharflags >= 0)
7594: {
7595: reqchar = subreqchar;
7596: reqcharflags = subreqcharflags;
7597: }
7598: }
7599:
7600: /* For a forward assertion, we take the reqchar, if set. This can be
7601: helpful if the pattern that follows the assertion doesn't set a different
7602: char. For example, it's useful for /(?=abcde).+/. We can't set firstchar
7603: for an assertion, however because it leads to incorrect effect for patterns
7604: such as /(?=a)a.+/ when the "real" "a" would then become a reqchar instead
7605: of a firstchar. This is overcome by a scan at the end if there's no
7606: firstchar, looking for an asserted first char. */
7607:
7608: else if (bravalue == OP_ASSERT && subreqcharflags >= 0)
7609: {
7610: reqchar = subreqchar;
7611: reqcharflags = subreqcharflags;
7612: }
7613: break; /* End of processing '(' */
7614:
7615:
7616: /* ===================================================================*/
7617: /* Handle metasequences introduced by \. For ones like \d, the ESC_ values
7618: are arranged to be the negation of the corresponding OP_values in the
7619: default case when PCRE_UCP is not set. For the back references, the values
7620: are negative the reference number. Only back references and those types
7621: that consume a character may be repeated. We can test for values between
7622: ESC_b and ESC_Z for the latter; this may have to change if any new ones are
7623: ever created. */
7624:
7625: case CHAR_BACKSLASH:
7626: tempptr = ptr;
7627: escape = check_escape(&ptr, &ec, errorcodeptr, cd->bracount, options, FALSE);
7628: if (*errorcodeptr != 0) goto FAILED;
7629:
7630: if (escape == 0) /* The escape coded a single character */
7631: c = ec;
7632: else
7633: {
7634: if (escape == ESC_Q) /* Handle start of quoted string */
7635: {
7636: if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
7637: ptr += 2; /* avoid empty string */
7638: else inescq = TRUE;
7639: continue;
7640: }
7641:
7642: if (escape == ESC_E) continue; /* Perl ignores an orphan \E */
7643:
7644: /* For metasequences that actually match a character, we disable the
7645: setting of a first character if it hasn't already been set. */
7646:
7647: if (firstcharflags == REQ_UNSET && escape > ESC_b && escape < ESC_Z)
7648: firstcharflags = REQ_NONE;
7649:
7650: /* Set values to reset to if this is followed by a zero repeat. */
7651:
7652: zerofirstchar = firstchar;
7653: zerofirstcharflags = firstcharflags;
7654: zeroreqchar = reqchar;
7655: zeroreqcharflags = reqcharflags;
7656:
7657: /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
7658: is a subroutine call by number (Oniguruma syntax). In fact, the value
7659: ESC_g is returned only for these cases. So we don't need to check for <
7660: or ' if the value is ESC_g. For the Perl syntax \g{n} the value is
7661: -n, and for the Perl syntax \g{name} the result is ESC_k (as
7662: that is a synonym for a named back reference). */
7663:
7664: if (escape == ESC_g)
7665: {
7666: const pcre_uchar *p;
7667: pcre_uint32 cf;
7668:
7669: save_hwm = cd->hwm; /* Normally this is set when '(' is read */
7670: terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
7671: CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
7672:
7673: /* These two statements stop the compiler for warning about possibly
7674: unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
7675: fact, because we do the check for a number below, the paths that
7676: would actually be in error are never taken. */
7677:
7678: skipbytes = 0;
7679: reset_bracount = FALSE;
7680:
7681: /* If it's not a signed or unsigned number, treat it as a name. */
7682:
7683: cf = ptr[1];
7684: if (cf != CHAR_PLUS && cf != CHAR_MINUS && !IS_DIGIT(cf))
7685: {
7686: is_recurse = TRUE;
7687: goto NAMED_REF_OR_RECURSE;
7688: }
7689:
7690: /* Signed or unsigned number (cf = ptr[1]) is known to be plus or minus
7691: or a digit. */
7692:
7693: p = ptr + 2;
7694: while (IS_DIGIT(*p)) p++;
7695: if (*p != (pcre_uchar)terminator)
7696: {
7697: *errorcodeptr = ERR57;
7698: break;
7699: }
7700: ptr++;
7701: goto HANDLE_NUMERICAL_RECURSION;
7702: }
7703:
7704: /* \k<name> or \k'name' is a back reference by name (Perl syntax).
7705: We also support \k{name} (.NET syntax). */
7706:
7707: if (escape == ESC_k)
7708: {
7709: if ((ptr[1] != CHAR_LESS_THAN_SIGN &&
7710: ptr[1] != CHAR_APOSTROPHE && ptr[1] != CHAR_LEFT_CURLY_BRACKET))
7711: {
7712: *errorcodeptr = ERR69;
7713: break;
7714: }
7715: is_recurse = FALSE;
7716: terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
7717: CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
7718: CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
7719: goto NAMED_REF_OR_RECURSE;
7720: }
7721:
7722: /* Back references are handled specially; must disable firstchar if
7723: not set to cope with cases like (?=(\w+))\1: which would otherwise set
7724: ':' later. */
7725:
7726: if (escape < 0)
7727: {
7728: open_capitem *oc;
7729: recno = -escape;
7730:
7731: /* Come here from named backref handling when the reference is to a
7732: single group (i.e. not to a duplicated name. */
7733:
7734: HANDLE_REFERENCE:
7735: if (firstcharflags == REQ_UNSET) firstcharflags = REQ_NONE;
7736: previous = code;
7737: *code++ = ((options & PCRE_CASELESS) != 0)? OP_REFI : OP_REF;
7738: PUT2INC(code, 0, recno);
7739: cd->backref_map |= (recno < 32)? (1 << recno) : 1;
7740: if (recno > cd->top_backref) cd->top_backref = recno;
7741:
7742: /* Check to see if this back reference is recursive, that it, it
7743: is inside the group that it references. A flag is set so that the
7744: group can be made atomic. */
7745:
7746: for (oc = cd->open_caps; oc != NULL; oc = oc->next)
7747: {
7748: if (oc->number == recno)
7749: {
7750: oc->flag = TRUE;
7751: break;
7752: }
7753: }
7754: }
7755:
7756: /* So are Unicode property matches, if supported. */
7757:
7758: #ifdef SUPPORT_UCP
7759: else if (escape == ESC_P || escape == ESC_p)
7760: {
7761: BOOL negated;
7762: unsigned int ptype = 0, pdata = 0;
7763: if (!get_ucp(&ptr, &negated, &ptype, &pdata, errorcodeptr))
7764: goto FAILED;
7765: previous = code;
7766: *code++ = ((escape == ESC_p) != negated)? OP_PROP : OP_NOTPROP;
7767: *code++ = ptype;
7768: *code++ = pdata;
7769: }
7770: #else
7771:
7772: /* If Unicode properties are not supported, \X, \P, and \p are not
7773: allowed. */
7774:
7775: else if (escape == ESC_X || escape == ESC_P || escape == ESC_p)
7776: {
7777: *errorcodeptr = ERR45;
7778: goto FAILED;
7779: }
7780: #endif
7781:
7782: /* For the rest (including \X when Unicode properties are supported), we
7783: can obtain the OP value by negating the escape value in the default
7784: situation when PCRE_UCP is not set. When it *is* set, we substitute
7785: Unicode property tests. Note that \b and \B do a one-character
7786: lookbehind, and \A also behaves as if it does. */
7787:
7788: else
7789: {
7790: if ((escape == ESC_b || escape == ESC_B || escape == ESC_A) &&
7791: cd->max_lookbehind == 0)
7792: cd->max_lookbehind = 1;
7793: #ifdef SUPPORT_UCP
7794: if (escape >= ESC_DU && escape <= ESC_wu)
7795: {
7796: nestptr = ptr + 1; /* Where to resume */
7797: ptr = substitutes[escape - ESC_DU] - 1; /* Just before substitute */
7798: }
7799: else
7800: #endif
7801: /* In non-UTF-8 mode, we turn \C into OP_ALLANY instead of OP_ANYBYTE
7802: so that it works in DFA mode and in lookbehinds. */
7803:
7804: {
7805: previous = (escape > ESC_b && escape < ESC_Z)? code : NULL;
7806: *code++ = (!utf && escape == ESC_C)? OP_ALLANY : escape;
7807: }
7808: }
7809: continue;
7810: }
7811:
7812: /* We have a data character whose value is in c. In UTF-8 mode it may have
7813: a value > 127. We set its representation in the length/buffer, and then
7814: handle it as a data character. */
7815:
7816: #if defined SUPPORT_UTF && !defined COMPILE_PCRE32
7817: if (utf && c > MAX_VALUE_FOR_SINGLE_CHAR)
7818: mclength = PRIV(ord2utf)(c, mcbuffer);
7819: else
7820: #endif
7821:
7822: {
7823: mcbuffer[0] = c;
7824: mclength = 1;
7825: }
7826: goto ONE_CHAR;
7827:
7828:
7829: /* ===================================================================*/
7830: /* Handle a literal character. It is guaranteed not to be whitespace or #
7831: when the extended flag is set. If we are in a UTF mode, it may be a
7832: multi-unit literal character. */
7833:
7834: default:
7835: NORMAL_CHAR:
7836: mclength = 1;
7837: mcbuffer[0] = c;
7838:
7839: #ifdef SUPPORT_UTF
7840: if (utf && HAS_EXTRALEN(c))
7841: ACROSSCHAR(TRUE, ptr[1], mcbuffer[mclength++] = *(++ptr));
7842: #endif
7843:
7844: /* At this point we have the character's bytes in mcbuffer, and the length
7845: in mclength. When not in UTF-8 mode, the length is always 1. */
7846:
7847: ONE_CHAR:
7848: previous = code;
7849:
7850: /* For caseless UTF-8 mode when UCP support is available, check whether
7851: this character has more than one other case. If so, generate a special
7852: OP_PROP item instead of OP_CHARI. */
7853:
7854: #ifdef SUPPORT_UCP
7855: if (utf && (options & PCRE_CASELESS) != 0)
7856: {
7857: GETCHAR(c, mcbuffer);
7858: if ((c = UCD_CASESET(c)) != 0)
7859: {
7860: *code++ = OP_PROP;
7861: *code++ = PT_CLIST;
7862: *code++ = c;
7863: if (firstcharflags == REQ_UNSET)
7864: firstcharflags = zerofirstcharflags = REQ_NONE;
7865: break;
7866: }
7867: }
7868: #endif
7869:
7870: /* Caseful matches, or not one of the multicase characters. */
7871:
7872: *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARI : OP_CHAR;
7873: for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
7874:
7875: /* Remember if \r or \n were seen */
7876:
7877: if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
7878: cd->external_flags |= PCRE_HASCRORLF;
7879:
7880: /* Set the first and required bytes appropriately. If no previous first
7881: byte, set it from this character, but revert to none on a zero repeat.
7882: Otherwise, leave the firstchar value alone, and don't change it on a zero
7883: repeat. */
7884:
7885: if (firstcharflags == REQ_UNSET)
7886: {
7887: zerofirstcharflags = REQ_NONE;
7888: zeroreqchar = reqchar;
7889: zeroreqcharflags = reqcharflags;
7890:
7891: /* If the character is more than one byte long, we can set firstchar
7892: only if it is not to be matched caselessly. */
7893:
7894: if (mclength == 1 || req_caseopt == 0)
7895: {
7896: firstchar = mcbuffer[0] | req_caseopt;
7897: firstchar = mcbuffer[0];
7898: firstcharflags = req_caseopt;
7899:
7900: if (mclength != 1)
7901: {
7902: reqchar = code[-1];
7903: reqcharflags = cd->req_varyopt;
7904: }
7905: }
7906: else firstcharflags = reqcharflags = REQ_NONE;
7907: }
7908:
7909: /* firstchar was previously set; we can set reqchar only if the length is
7910: 1 or the matching is caseful. */
7911:
7912: else
7913: {
7914: zerofirstchar = firstchar;
7915: zerofirstcharflags = firstcharflags;
7916: zeroreqchar = reqchar;
7917: zeroreqcharflags = reqcharflags;
7918: if (mclength == 1 || req_caseopt == 0)
7919: {
7920: reqchar = code[-1];
7921: reqcharflags = req_caseopt | cd->req_varyopt;
7922: }
7923: }
7924:
7925: break; /* End of literal character handling */
7926: }
7927: } /* end of big loop */
7928:
7929:
7930: /* Control never reaches here by falling through, only by a goto for all the
7931: error states. Pass back the position in the pattern so that it can be displayed
7932: to the user for diagnosing the error. */
7933:
7934: FAILED:
7935: *ptrptr = ptr;
7936: return FALSE;
7937: }
7938:
7939:
7940:
7941: /*************************************************
7942: * Compile sequence of alternatives *
7943: *************************************************/
7944:
7945: /* On entry, ptr is pointing past the bracket character, but on return it
7946: points to the closing bracket, or vertical bar, or end of string. The code
7947: variable is pointing at the byte into which the BRA operator has been stored.
7948: This function is used during the pre-compile phase when we are trying to find
7949: out the amount of memory needed, as well as during the real compile phase. The
7950: value of lengthptr distinguishes the two phases.
7951:
7952: Arguments:
7953: options option bits, including any changes for this subpattern
7954: codeptr -> the address of the current code pointer
7955: ptrptr -> the address of the current pattern pointer
7956: errorcodeptr -> pointer to error code variable
7957: lookbehind TRUE if this is a lookbehind assertion
7958: reset_bracount TRUE to reset the count for each branch
7959: skipbytes skip this many bytes at start (for brackets and OP_COND)
7960: cond_depth depth of nesting for conditional subpatterns
7961: firstcharptr place to put the first required character
7962: firstcharflagsptr place to put the first character flags, or a negative number
7963: reqcharptr place to put the last required character
7964: reqcharflagsptr place to put the last required character flags, or a negative number
7965: bcptr pointer to the chain of currently open branches
7966: cd points to the data block with tables pointers etc.
7967: lengthptr NULL during the real compile phase
7968: points to length accumulator during pre-compile phase
7969:
7970: Returns: TRUE on success
7971: */
7972:
7973: static BOOL
7974: compile_regex(int options, pcre_uchar **codeptr, const pcre_uchar **ptrptr,
7975: int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
7976: int cond_depth,
7977: pcre_uint32 *firstcharptr, pcre_int32 *firstcharflagsptr,
7978: pcre_uint32 *reqcharptr, pcre_int32 *reqcharflagsptr,
7979: branch_chain *bcptr, compile_data *cd, int *lengthptr)
7980: {
7981: const pcre_uchar *ptr = *ptrptr;
7982: pcre_uchar *code = *codeptr;
7983: pcre_uchar *last_branch = code;
7984: pcre_uchar *start_bracket = code;
7985: pcre_uchar *reverse_count = NULL;
7986: open_capitem capitem;
7987: int capnumber = 0;
7988: pcre_uint32 firstchar, reqchar;
7989: pcre_int32 firstcharflags, reqcharflags;
7990: pcre_uint32 branchfirstchar, branchreqchar;
7991: pcre_int32 branchfirstcharflags, branchreqcharflags;
7992: int length;
7993: unsigned int orig_bracount;
7994: unsigned int max_bracount;
7995: branch_chain bc;
7996:
7997: bc.outer = bcptr;
7998: bc.current_branch = code;
7999:
8000: firstchar = reqchar = 0;
8001: firstcharflags = reqcharflags = REQ_UNSET;
8002:
8003: /* Accumulate the length for use in the pre-compile phase. Start with the
8004: length of the BRA and KET and any extra bytes that are required at the
8005: beginning. We accumulate in a local variable to save frequent testing of
8006: lenthptr for NULL. We cannot do this by looking at the value of code at the
8007: start and end of each alternative, because compiled items are discarded during
8008: the pre-compile phase so that the work space is not exceeded. */
8009:
8010: length = 2 + 2*LINK_SIZE + skipbytes;
8011:
8012: /* WARNING: If the above line is changed for any reason, you must also change
8013: the code that abstracts option settings at the start of the pattern and makes
8014: them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
8015: pre-compile phase to find out whether anything has yet been compiled or not. */
8016:
8017: /* If this is a capturing subpattern, add to the chain of open capturing items
8018: so that we can detect them if (*ACCEPT) is encountered. This is also used to
8019: detect groups that contain recursive back references to themselves. Note that
8020: only OP_CBRA need be tested here; changing this opcode to one of its variants,
8021: e.g. OP_SCBRAPOS, happens later, after the group has been compiled. */
8022:
8023: if (*code == OP_CBRA)
8024: {
8025: capnumber = GET2(code, 1 + LINK_SIZE);
8026: capitem.number = capnumber;
8027: capitem.next = cd->open_caps;
8028: capitem.flag = FALSE;
8029: cd->open_caps = &capitem;
8030: }
8031:
8032: /* Offset is set zero to mark that this bracket is still open */
8033:
8034: PUT(code, 1, 0);
8035: code += 1 + LINK_SIZE + skipbytes;
8036:
8037: /* Loop for each alternative branch */
8038:
8039: orig_bracount = max_bracount = cd->bracount;
8040: for (;;)
8041: {
8042: /* For a (?| group, reset the capturing bracket count so that each branch
8043: uses the same numbers. */
8044:
8045: if (reset_bracount) cd->bracount = orig_bracount;
8046:
8047: /* Set up dummy OP_REVERSE if lookbehind assertion */
8048:
8049: if (lookbehind)
8050: {
8051: *code++ = OP_REVERSE;
8052: reverse_count = code;
8053: PUTINC(code, 0, 0);
8054: length += 1 + LINK_SIZE;
8055: }
8056:
8057: /* Now compile the branch; in the pre-compile phase its length gets added
8058: into the length. */
8059:
8060: if (!compile_branch(&options, &code, &ptr, errorcodeptr, &branchfirstchar,
8061: &branchfirstcharflags, &branchreqchar, &branchreqcharflags, &bc,
8062: cond_depth, cd, (lengthptr == NULL)? NULL : &length))
8063: {
8064: *ptrptr = ptr;
8065: return FALSE;
8066: }
8067:
8068: /* Keep the highest bracket count in case (?| was used and some branch
8069: has fewer than the rest. */
8070:
8071: if (cd->bracount > max_bracount) max_bracount = cd->bracount;
8072:
8073: /* In the real compile phase, there is some post-processing to be done. */
8074:
8075: if (lengthptr == NULL)
8076: {
8077: /* If this is the first branch, the firstchar and reqchar values for the
8078: branch become the values for the regex. */
8079:
8080: if (*last_branch != OP_ALT)
8081: {
8082: firstchar = branchfirstchar;
8083: firstcharflags = branchfirstcharflags;
8084: reqchar = branchreqchar;
8085: reqcharflags = branchreqcharflags;
8086: }
8087:
8088: /* If this is not the first branch, the first char and reqchar have to
8089: match the values from all the previous branches, except that if the
8090: previous value for reqchar didn't have REQ_VARY set, it can still match,
8091: and we set REQ_VARY for the regex. */
8092:
8093: else
8094: {
8095: /* If we previously had a firstchar, but it doesn't match the new branch,
8096: we have to abandon the firstchar for the regex, but if there was
8097: previously no reqchar, it takes on the value of the old firstchar. */
8098:
8099: if (firstcharflags >= 0 &&
8100: (firstcharflags != branchfirstcharflags || firstchar != branchfirstchar))
8101: {
8102: if (reqcharflags < 0)
8103: {
8104: reqchar = firstchar;
8105: reqcharflags = firstcharflags;
8106: }
8107: firstcharflags = REQ_NONE;
8108: }
8109:
8110: /* If we (now or from before) have no firstchar, a firstchar from the
8111: branch becomes a reqchar if there isn't a branch reqchar. */
8112:
8113: if (firstcharflags < 0 && branchfirstcharflags >= 0 && branchreqcharflags < 0)
8114: {
8115: branchreqchar = branchfirstchar;
8116: branchreqcharflags = branchfirstcharflags;
8117: }
8118:
8119: /* Now ensure that the reqchars match */
8120:
8121: if (((reqcharflags & ~REQ_VARY) != (branchreqcharflags & ~REQ_VARY)) ||
8122: reqchar != branchreqchar)
8123: reqcharflags = REQ_NONE;
8124: else
8125: {
8126: reqchar = branchreqchar;
8127: reqcharflags |= branchreqcharflags; /* To "or" REQ_VARY */
8128: }
8129: }
8130:
8131: /* If lookbehind, check that this branch matches a fixed-length string, and
8132: put the length into the OP_REVERSE item. Temporarily mark the end of the
8133: branch with OP_END. If the branch contains OP_RECURSE, the result is -3
8134: because there may be forward references that we can't check here. Set a
8135: flag to cause another lookbehind check at the end. Why not do it all at the
8136: end? Because common, erroneous checks are picked up here and the offset of
8137: the problem can be shown. */
8138:
8139: if (lookbehind)
8140: {
8141: int fixed_length;
8142: *code = OP_END;
8143: fixed_length = find_fixedlength(last_branch, (options & PCRE_UTF8) != 0,
8144: FALSE, cd);
8145: DPRINTF(("fixed length = %d\n", fixed_length));
8146: if (fixed_length == -3)
8147: {
8148: cd->check_lookbehind = TRUE;
8149: }
8150: else if (fixed_length < 0)
8151: {
8152: *errorcodeptr = (fixed_length == -2)? ERR36 :
8153: (fixed_length == -4)? ERR70: ERR25;
8154: *ptrptr = ptr;
8155: return FALSE;
8156: }
8157: else
8158: {
8159: if (fixed_length > cd->max_lookbehind)
8160: cd->max_lookbehind = fixed_length;
8161: PUT(reverse_count, 0, fixed_length);
8162: }
8163: }
8164: }
8165:
8166: /* Reached end of expression, either ')' or end of pattern. In the real
8167: compile phase, go back through the alternative branches and reverse the chain
8168: of offsets, with the field in the BRA item now becoming an offset to the
8169: first alternative. If there are no alternatives, it points to the end of the
8170: group. The length in the terminating ket is always the length of the whole
8171: bracketed item. Return leaving the pointer at the terminating char. */
8172:
8173: if (*ptr != CHAR_VERTICAL_LINE)
8174: {
8175: if (lengthptr == NULL)
8176: {
8177: int branch_length = (int)(code - last_branch);
8178: do
8179: {
8180: int prev_length = GET(last_branch, 1);
8181: PUT(last_branch, 1, branch_length);
8182: branch_length = prev_length;
8183: last_branch -= branch_length;
8184: }
8185: while (branch_length > 0);
8186: }
8187:
8188: /* Fill in the ket */
8189:
8190: *code = OP_KET;
8191: PUT(code, 1, (int)(code - start_bracket));
8192: code += 1 + LINK_SIZE;
8193:
8194: /* If it was a capturing subpattern, check to see if it contained any
8195: recursive back references. If so, we must wrap it in atomic brackets.
8196: In any event, remove the block from the chain. */
8197:
8198: if (capnumber > 0)
8199: {
8200: if (cd->open_caps->flag)
8201: {
8202: memmove(start_bracket + 1 + LINK_SIZE, start_bracket,
8203: IN_UCHARS(code - start_bracket));
8204: *start_bracket = OP_ONCE;
8205: code += 1 + LINK_SIZE;
8206: PUT(start_bracket, 1, (int)(code - start_bracket));
8207: *code = OP_KET;
8208: PUT(code, 1, (int)(code - start_bracket));
8209: code += 1 + LINK_SIZE;
8210: length += 2 + 2*LINK_SIZE;
8211: }
8212: cd->open_caps = cd->open_caps->next;
8213: }
8214:
8215: /* Retain the highest bracket number, in case resetting was used. */
8216:
8217: cd->bracount = max_bracount;
8218:
8219: /* Set values to pass back */
8220:
8221: *codeptr = code;
8222: *ptrptr = ptr;
8223: *firstcharptr = firstchar;
8224: *firstcharflagsptr = firstcharflags;
8225: *reqcharptr = reqchar;
8226: *reqcharflagsptr = reqcharflags;
8227: if (lengthptr != NULL)
8228: {
8229: if (OFLOW_MAX - *lengthptr < length)
8230: {
8231: *errorcodeptr = ERR20;
8232: return FALSE;
8233: }
8234: *lengthptr += length;
8235: }
8236: return TRUE;
8237: }
8238:
8239: /* Another branch follows. In the pre-compile phase, we can move the code
8240: pointer back to where it was for the start of the first branch. (That is,
8241: pretend that each branch is the only one.)
8242:
8243: In the real compile phase, insert an ALT node. Its length field points back
8244: to the previous branch while the bracket remains open. At the end the chain
8245: is reversed. It's done like this so that the start of the bracket has a
8246: zero offset until it is closed, making it possible to detect recursion. */
8247:
8248: if (lengthptr != NULL)
8249: {
8250: code = *codeptr + 1 + LINK_SIZE + skipbytes;
8251: length += 1 + LINK_SIZE;
8252: }
8253: else
8254: {
8255: *code = OP_ALT;
8256: PUT(code, 1, (int)(code - last_branch));
8257: bc.current_branch = last_branch = code;
8258: code += 1 + LINK_SIZE;
8259: }
8260:
8261: ptr++;
8262: }
8263: /* Control never reaches here */
8264: }
8265:
8266:
8267:
8268:
8269: /*************************************************
8270: * Check for anchored expression *
8271: *************************************************/
8272:
8273: /* Try to find out if this is an anchored regular expression. Consider each
8274: alternative branch. If they all start with OP_SOD or OP_CIRC, or with a bracket
8275: all of whose alternatives start with OP_SOD or OP_CIRC (recurse ad lib), then
8276: it's anchored. However, if this is a multiline pattern, then only OP_SOD will
8277: be found, because ^ generates OP_CIRCM in that mode.
8278:
8279: We can also consider a regex to be anchored if OP_SOM starts all its branches.
8280: This is the code for \G, which means "match at start of match position, taking
8281: into account the match offset".
8282:
8283: A branch is also implicitly anchored if it starts with .* and DOTALL is set,
8284: because that will try the rest of the pattern at all possible matching points,
8285: so there is no point trying again.... er ....
8286:
8287: .... except when the .* appears inside capturing parentheses, and there is a
8288: subsequent back reference to those parentheses. We haven't enough information
8289: to catch that case precisely.
8290:
8291: At first, the best we could do was to detect when .* was in capturing brackets
8292: and the highest back reference was greater than or equal to that level.
8293: However, by keeping a bitmap of the first 31 back references, we can catch some
8294: of the more common cases more precisely.
8295:
8296: ... A second exception is when the .* appears inside an atomic group, because
8297: this prevents the number of characters it matches from being adjusted.
8298:
8299: Arguments:
8300: code points to start of expression (the bracket)
8301: bracket_map a bitmap of which brackets we are inside while testing; this
8302: handles up to substring 31; after that we just have to take
8303: the less precise approach
8304: cd points to the compile data block
8305: atomcount atomic group level
8306:
8307: Returns: TRUE or FALSE
8308: */
8309:
8310: static BOOL
8311: is_anchored(register const pcre_uchar *code, unsigned int bracket_map,
8312: compile_data *cd, int atomcount)
8313: {
8314: do {
8315: const pcre_uchar *scode = first_significant_code(
8316: code + PRIV(OP_lengths)[*code], FALSE);
8317: register int op = *scode;
8318:
8319: /* Non-capturing brackets */
8320:
8321: if (op == OP_BRA || op == OP_BRAPOS ||
8322: op == OP_SBRA || op == OP_SBRAPOS)
8323: {
8324: if (!is_anchored(scode, bracket_map, cd, atomcount)) return FALSE;
8325: }
8326:
8327: /* Capturing brackets */
8328:
8329: else if (op == OP_CBRA || op == OP_CBRAPOS ||
8330: op == OP_SCBRA || op == OP_SCBRAPOS)
8331: {
8332: int n = GET2(scode, 1+LINK_SIZE);
8333: int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
8334: if (!is_anchored(scode, new_map, cd, atomcount)) return FALSE;
8335: }
8336:
8337: /* Positive forward assertions and conditions */
8338:
8339: else if (op == OP_ASSERT || op == OP_COND)
8340: {
8341: if (!is_anchored(scode, bracket_map, cd, atomcount)) return FALSE;
8342: }
8343:
8344: /* Atomic groups */
8345:
8346: else if (op == OP_ONCE || op == OP_ONCE_NC)
8347: {
8348: if (!is_anchored(scode, bracket_map, cd, atomcount + 1))
8349: return FALSE;
8350: }
8351:
8352: /* .* is not anchored unless DOTALL is set (which generates OP_ALLANY) and
8353: it isn't in brackets that are or may be referenced or inside an atomic
8354: group. */
8355:
8356: else if ((op == OP_TYPESTAR || op == OP_TYPEMINSTAR ||
8357: op == OP_TYPEPOSSTAR))
8358: {
8359: if (scode[1] != OP_ALLANY || (bracket_map & cd->backref_map) != 0 ||
8360: atomcount > 0 || cd->had_pruneorskip)
8361: return FALSE;
8362: }
8363:
8364: /* Check for explicit anchoring */
8365:
8366: else if (op != OP_SOD && op != OP_SOM && op != OP_CIRC) return FALSE;
8367:
8368: code += GET(code, 1);
8369: }
8370: while (*code == OP_ALT); /* Loop for each alternative */
8371: return TRUE;
8372: }
8373:
8374:
8375:
8376: /*************************************************
8377: * Check for starting with ^ or .* *
8378: *************************************************/
8379:
8380: /* This is called to find out if every branch starts with ^ or .* so that
8381: "first char" processing can be done to speed things up in multiline
8382: matching and for non-DOTALL patterns that start with .* (which must start at
8383: the beginning or after \n). As in the case of is_anchored() (see above), we
8384: have to take account of back references to capturing brackets that contain .*
8385: because in that case we can't make the assumption. Also, the appearance of .*
8386: inside atomic brackets or in a pattern that contains *PRUNE or *SKIP does not
8387: count, because once again the assumption no longer holds.
8388:
8389: Arguments:
8390: code points to start of expression (the bracket)
8391: bracket_map a bitmap of which brackets we are inside while testing; this
8392: handles up to substring 31; after that we just have to take
8393: the less precise approach
8394: cd points to the compile data
8395: atomcount atomic group level
8396:
8397: Returns: TRUE or FALSE
8398: */
8399:
8400: static BOOL
8401: is_startline(const pcre_uchar *code, unsigned int bracket_map,
8402: compile_data *cd, int atomcount)
8403: {
8404: do {
8405: const pcre_uchar *scode = first_significant_code(
8406: code + PRIV(OP_lengths)[*code], FALSE);
8407: register int op = *scode;
8408:
8409: /* If we are at the start of a conditional assertion group, *both* the
8410: conditional assertion *and* what follows the condition must satisfy the test
8411: for start of line. Other kinds of condition fail. Note that there may be an
8412: auto-callout at the start of a condition. */
8413:
8414: if (op == OP_COND)
8415: {
8416: scode += 1 + LINK_SIZE;
8417: if (*scode == OP_CALLOUT) scode += PRIV(OP_lengths)[OP_CALLOUT];
8418: switch (*scode)
8419: {
8420: case OP_CREF:
8421: case OP_DNCREF:
8422: case OP_RREF:
8423: case OP_DNRREF:
8424: case OP_DEF:
8425: return FALSE;
8426:
8427: default: /* Assertion */
8428: if (!is_startline(scode, bracket_map, cd, atomcount)) return FALSE;
8429: do scode += GET(scode, 1); while (*scode == OP_ALT);
8430: scode += 1 + LINK_SIZE;
8431: break;
8432: }
8433: scode = first_significant_code(scode, FALSE);
8434: op = *scode;
8435: }
8436:
8437: /* Non-capturing brackets */
8438:
8439: if (op == OP_BRA || op == OP_BRAPOS ||
8440: op == OP_SBRA || op == OP_SBRAPOS)
8441: {
8442: if (!is_startline(scode, bracket_map, cd, atomcount)) return FALSE;
8443: }
8444:
8445: /* Capturing brackets */
8446:
8447: else if (op == OP_CBRA || op == OP_CBRAPOS ||
8448: op == OP_SCBRA || op == OP_SCBRAPOS)
8449: {
8450: int n = GET2(scode, 1+LINK_SIZE);
8451: int new_map = bracket_map | ((n < 32)? (1 << n) : 1);
8452: if (!is_startline(scode, new_map, cd, atomcount)) return FALSE;
8453: }
8454:
8455: /* Positive forward assertions */
8456:
8457: else if (op == OP_ASSERT)
8458: {
8459: if (!is_startline(scode, bracket_map, cd, atomcount)) return FALSE;
8460: }
8461:
8462: /* Atomic brackets */
8463:
8464: else if (op == OP_ONCE || op == OP_ONCE_NC)
8465: {
8466: if (!is_startline(scode, bracket_map, cd, atomcount + 1)) return FALSE;
8467: }
8468:
8469: /* .* means "start at start or after \n" if it isn't in atomic brackets or
8470: brackets that may be referenced, as long as the pattern does not contain
8471: *PRUNE or *SKIP, because these break the feature. Consider, for example,
8472: /.*?a(*PRUNE)b/ with the subject "aab", which matches "ab", i.e. not at the
8473: start of a line. */
8474:
8475: else if (op == OP_TYPESTAR || op == OP_TYPEMINSTAR || op == OP_TYPEPOSSTAR)
8476: {
8477: if (scode[1] != OP_ANY || (bracket_map & cd->backref_map) != 0 ||
8478: atomcount > 0 || cd->had_pruneorskip)
8479: return FALSE;
8480: }
8481:
8482: /* Check for explicit circumflex; anything else gives a FALSE result. Note
8483: in particular that this includes atomic brackets OP_ONCE and OP_ONCE_NC
8484: because the number of characters matched by .* cannot be adjusted inside
8485: them. */
8486:
8487: else if (op != OP_CIRC && op != OP_CIRCM) return FALSE;
8488:
8489: /* Move on to the next alternative */
8490:
8491: code += GET(code, 1);
8492: }
8493: while (*code == OP_ALT); /* Loop for each alternative */
8494: return TRUE;
8495: }
8496:
8497:
8498:
8499: /*************************************************
8500: * Check for asserted fixed first char *
8501: *************************************************/
8502:
8503: /* During compilation, the "first char" settings from forward assertions are
8504: discarded, because they can cause conflicts with actual literals that follow.
8505: However, if we end up without a first char setting for an unanchored pattern,
8506: it is worth scanning the regex to see if there is an initial asserted first
8507: char. If all branches start with the same asserted char, or with a
8508: non-conditional bracket all of whose alternatives start with the same asserted
8509: char (recurse ad lib), then we return that char, with the flags set to zero or
8510: REQ_CASELESS; otherwise return zero with REQ_NONE in the flags.
8511:
8512: Arguments:
8513: code points to start of expression (the bracket)
8514: flags points to the first char flags, or to REQ_NONE
8515: inassert TRUE if in an assertion
8516:
8517: Returns: the fixed first char, or 0 with REQ_NONE in flags
8518: */
8519:
8520: static pcre_uint32
8521: find_firstassertedchar(const pcre_uchar *code, pcre_int32 *flags,
8522: BOOL inassert)
8523: {
8524: register pcre_uint32 c = 0;
8525: int cflags = REQ_NONE;
8526:
8527: *flags = REQ_NONE;
8528: do {
8529: pcre_uint32 d;
8530: int dflags;
8531: int xl = (*code == OP_CBRA || *code == OP_SCBRA ||
8532: *code == OP_CBRAPOS || *code == OP_SCBRAPOS)? IMM2_SIZE:0;
8533: const pcre_uchar *scode = first_significant_code(code + 1+LINK_SIZE + xl,
8534: TRUE);
8535: register pcre_uchar op = *scode;
8536:
8537: switch(op)
8538: {
8539: default:
8540: return 0;
8541:
8542: case OP_BRA:
8543: case OP_BRAPOS:
8544: case OP_CBRA:
8545: case OP_SCBRA:
8546: case OP_CBRAPOS:
8547: case OP_SCBRAPOS:
8548: case OP_ASSERT:
8549: case OP_ONCE:
8550: case OP_ONCE_NC:
8551: d = find_firstassertedchar(scode, &dflags, op == OP_ASSERT);
8552: if (dflags < 0)
8553: return 0;
8554: if (cflags < 0) { c = d; cflags = dflags; } else if (c != d || cflags != dflags) return 0;
8555: break;
8556:
8557: case OP_EXACT:
8558: scode += IMM2_SIZE;
8559: /* Fall through */
8560:
8561: case OP_CHAR:
8562: case OP_PLUS:
8563: case OP_MINPLUS:
8564: case OP_POSPLUS:
8565: if (!inassert) return 0;
8566: if (cflags < 0) { c = scode[1]; cflags = 0; }
8567: else if (c != scode[1]) return 0;
8568: break;
8569:
8570: case OP_EXACTI:
8571: scode += IMM2_SIZE;
8572: /* Fall through */
8573:
8574: case OP_CHARI:
8575: case OP_PLUSI:
8576: case OP_MINPLUSI:
8577: case OP_POSPLUSI:
8578: if (!inassert) return 0;
8579: if (cflags < 0) { c = scode[1]; cflags = REQ_CASELESS; }
8580: else if (c != scode[1]) return 0;
8581: break;
8582: }
8583:
8584: code += GET(code, 1);
8585: }
8586: while (*code == OP_ALT);
8587:
8588: *flags = cflags;
8589: return c;
8590: }
8591:
8592:
8593:
8594: /*************************************************
8595: * Add an entry to the name/number table *
8596: *************************************************/
8597:
8598: /* This function is called between compiling passes to add an entry to the
8599: name/number table, maintaining alphabetical order. Checking for permitted
8600: and forbidden duplicates has already been done.
8601:
8602: Arguments:
8603: cd the compile data block
8604: name the name to add
8605: length the length of the name
8606: groupno the group number
8607:
8608: Returns: nothing
8609: */
8610:
8611: static void
8612: add_name(compile_data *cd, const pcre_uchar *name, int length,
8613: unsigned int groupno)
8614: {
8615: int i;
8616: pcre_uchar *slot = cd->name_table;
8617:
8618: for (i = 0; i < cd->names_found; i++)
8619: {
8620: int crc = memcmp(name, slot+IMM2_SIZE, IN_UCHARS(length));
8621: if (crc == 0 && slot[IMM2_SIZE+length] != 0)
8622: crc = -1; /* Current name is a substring */
8623:
8624: /* Make space in the table and break the loop for an earlier name. For a
8625: duplicate or later name, carry on. We do this for duplicates so that in the
8626: simple case (when ?(| is not used) they are in order of their numbers. In all
8627: cases they are in the order in which they appear in the pattern. */
8628:
8629: if (crc < 0)
8630: {
8631: memmove(slot + cd->name_entry_size, slot,
8632: IN_UCHARS((cd->names_found - i) * cd->name_entry_size));
8633: break;
8634: }
8635:
8636: /* Continue the loop for a later or duplicate name */
8637:
8638: slot += cd->name_entry_size;
8639: }
8640:
8641: PUT2(slot, 0, groupno);
8642: memcpy(slot + IMM2_SIZE, name, IN_UCHARS(length));
8643: slot[IMM2_SIZE + length] = 0;
8644: cd->names_found++;
8645: }
8646:
8647:
8648:
8649: /*************************************************
8650: * Compile a Regular Expression *
8651: *************************************************/
8652:
8653: /* This function takes a string and returns a pointer to a block of store
8654: holding a compiled version of the expression. The original API for this
8655: function had no error code return variable; it is retained for backwards
8656: compatibility. The new function is given a new name.
8657:
8658: Arguments:
8659: pattern the regular expression
8660: options various option bits
8661: errorcodeptr pointer to error code variable (pcre_compile2() only)
8662: can be NULL if you don't want a code value
8663: errorptr pointer to pointer to error text
8664: erroroffset ptr offset in pattern where error was detected
8665: tables pointer to character tables or NULL
8666:
8667: Returns: pointer to compiled data block, or NULL on error,
8668: with errorptr and erroroffset set
8669: */
8670:
8671: #if defined COMPILE_PCRE8
8672: PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
8673: pcre_compile(const char *pattern, int options, const char **errorptr,
8674: int *erroroffset, const unsigned char *tables)
8675: #elif defined COMPILE_PCRE16
8676: PCRE_EXP_DEFN pcre16 * PCRE_CALL_CONVENTION
8677: pcre16_compile(PCRE_SPTR16 pattern, int options, const char **errorptr,
8678: int *erroroffset, const unsigned char *tables)
8679: #elif defined COMPILE_PCRE32
8680: PCRE_EXP_DEFN pcre32 * PCRE_CALL_CONVENTION
8681: pcre32_compile(PCRE_SPTR32 pattern, int options, const char **errorptr,
8682: int *erroroffset, const unsigned char *tables)
8683: #endif
8684: {
8685: #if defined COMPILE_PCRE8
8686: return pcre_compile2(pattern, options, NULL, errorptr, erroroffset, tables);
8687: #elif defined COMPILE_PCRE16
8688: return pcre16_compile2(pattern, options, NULL, errorptr, erroroffset, tables);
8689: #elif defined COMPILE_PCRE32
8690: return pcre32_compile2(pattern, options, NULL, errorptr, erroroffset, tables);
8691: #endif
8692: }
8693:
8694:
8695: #if defined COMPILE_PCRE8
8696: PCRE_EXP_DEFN pcre * PCRE_CALL_CONVENTION
8697: pcre_compile2(const char *pattern, int options, int *errorcodeptr,
8698: const char **errorptr, int *erroroffset, const unsigned char *tables)
8699: #elif defined COMPILE_PCRE16
8700: PCRE_EXP_DEFN pcre16 * PCRE_CALL_CONVENTION
8701: pcre16_compile2(PCRE_SPTR16 pattern, int options, int *errorcodeptr,
8702: const char **errorptr, int *erroroffset, const unsigned char *tables)
8703: #elif defined COMPILE_PCRE32
8704: PCRE_EXP_DEFN pcre32 * PCRE_CALL_CONVENTION
8705: pcre32_compile2(PCRE_SPTR32 pattern, int options, int *errorcodeptr,
8706: const char **errorptr, int *erroroffset, const unsigned char *tables)
8707: #endif
8708: {
8709: REAL_PCRE *re;
8710: int length = 1; /* For final END opcode */
8711: pcre_int32 firstcharflags, reqcharflags;
8712: pcre_uint32 firstchar, reqchar;
8713: pcre_uint32 limit_match = PCRE_UINT32_MAX;
8714: pcre_uint32 limit_recursion = PCRE_UINT32_MAX;
8715: int newline;
8716: int errorcode = 0;
8717: int skipatstart = 0;
8718: BOOL utf;
8719: BOOL never_utf = FALSE;
8720: size_t size;
8721: pcre_uchar *code;
8722: const pcre_uchar *codestart;
8723: const pcre_uchar *ptr;
8724: compile_data compile_block;
8725: compile_data *cd = &compile_block;
8726:
8727: /* This space is used for "compiling" into during the first phase, when we are
8728: computing the amount of memory that is needed. Compiled items are thrown away
8729: as soon as possible, so that a fairly large buffer should be sufficient for
8730: this purpose. The same space is used in the second phase for remembering where
8731: to fill in forward references to subpatterns. That may overflow, in which case
8732: new memory is obtained from malloc(). */
8733:
8734: pcre_uchar cworkspace[COMPILE_WORK_SIZE];
8735:
8736: /* This vector is used for remembering name groups during the pre-compile. In a
8737: similar way to cworkspace, it can be expanded using malloc() if necessary. */
8738:
8739: named_group named_groups[NAMED_GROUP_LIST_SIZE];
8740:
8741: /* Set this early so that early errors get offset 0. */
8742:
8743: ptr = (const pcre_uchar *)pattern;
8744:
8745: /* We can't pass back an error message if errorptr is NULL; I guess the best we
8746: can do is just return NULL, but we can set a code value if there is a code
8747: pointer. */
8748:
8749: if (errorptr == NULL)
8750: {
8751: if (errorcodeptr != NULL) *errorcodeptr = 99;
8752: return NULL;
8753: }
8754:
8755: *errorptr = NULL;
8756: if (errorcodeptr != NULL) *errorcodeptr = ERR0;
8757:
8758: /* However, we can give a message for this error */
8759:
8760: if (erroroffset == NULL)
8761: {
8762: errorcode = ERR16;
8763: goto PCRE_EARLY_ERROR_RETURN2;
8764: }
8765:
8766: *erroroffset = 0;
8767:
8768: /* Set up pointers to the individual character tables */
8769:
8770: if (tables == NULL) tables = PRIV(default_tables);
8771: cd->lcc = tables + lcc_offset;
8772: cd->fcc = tables + fcc_offset;
8773: cd->cbits = tables + cbits_offset;
8774: cd->ctypes = tables + ctypes_offset;
8775:
8776: /* Check that all undefined public option bits are zero */
8777:
8778: if ((options & ~PUBLIC_COMPILE_OPTIONS) != 0)
8779: {
8780: errorcode = ERR17;
8781: goto PCRE_EARLY_ERROR_RETURN;
8782: }
8783:
8784: /* If PCRE_NEVER_UTF is set, remember it. */
8785:
8786: if ((options & PCRE_NEVER_UTF) != 0) never_utf = TRUE;
8787:
8788: /* Check for global one-time settings at the start of the pattern, and remember
8789: the offset for later. */
8790:
8791: cd->external_flags = 0; /* Initialize here for LIMIT_MATCH/RECURSION */
8792:
8793: while (ptr[skipatstart] == CHAR_LEFT_PARENTHESIS &&
8794: ptr[skipatstart+1] == CHAR_ASTERISK)
8795: {
8796: int newnl = 0;
8797: int newbsr = 0;
8798:
8799: /* For completeness and backward compatibility, (*UTFn) is supported in the
8800: relevant libraries, but (*UTF) is generic and always supported. Note that
8801: PCRE_UTF8 == PCRE_UTF16 == PCRE_UTF32. */
8802:
8803: #ifdef COMPILE_PCRE8
8804: if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UTF8_RIGHTPAR, 5) == 0)
8805: { skipatstart += 7; options |= PCRE_UTF8; continue; }
8806: #endif
8807: #ifdef COMPILE_PCRE16
8808: if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UTF16_RIGHTPAR, 6) == 0)
8809: { skipatstart += 8; options |= PCRE_UTF16; continue; }
8810: #endif
8811: #ifdef COMPILE_PCRE32
8812: if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UTF32_RIGHTPAR, 6) == 0)
8813: { skipatstart += 8; options |= PCRE_UTF32; continue; }
8814: #endif
8815:
8816: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UTF_RIGHTPAR, 4) == 0)
8817: { skipatstart += 6; options |= PCRE_UTF8; continue; }
8818: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_UCP_RIGHTPAR, 4) == 0)
8819: { skipatstart += 6; options |= PCRE_UCP; continue; }
8820: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_NO_AUTO_POSSESS_RIGHTPAR, 16) == 0)
8821: { skipatstart += 18; options |= PCRE_NO_AUTO_POSSESS; continue; }
8822: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_NO_START_OPT_RIGHTPAR, 13) == 0)
8823: { skipatstart += 15; options |= PCRE_NO_START_OPTIMIZE; continue; }
8824:
8825: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_LIMIT_MATCH_EQ, 12) == 0)
8826: {
8827: pcre_uint32 c = 0;
8828: int p = skipatstart + 14;
8829: while (isdigit(ptr[p]))
8830: {
8831: if (c > PCRE_UINT32_MAX / 10 - 1) break; /* Integer overflow */
8832: c = c*10 + ptr[p++] - CHAR_0;
8833: }
8834: if (ptr[p++] != CHAR_RIGHT_PARENTHESIS) break;
8835: if (c < limit_match)
8836: {
8837: limit_match = c;
8838: cd->external_flags |= PCRE_MLSET;
8839: }
8840: skipatstart = p;
8841: continue;
8842: }
8843:
8844: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_LIMIT_RECURSION_EQ, 16) == 0)
8845: {
8846: pcre_uint32 c = 0;
8847: int p = skipatstart + 18;
8848: while (isdigit(ptr[p]))
8849: {
8850: if (c > PCRE_UINT32_MAX / 10 - 1) break; /* Integer overflow check */
8851: c = c*10 + ptr[p++] - CHAR_0;
8852: }
8853: if (ptr[p++] != CHAR_RIGHT_PARENTHESIS) break;
8854: if (c < limit_recursion)
8855: {
8856: limit_recursion = c;
8857: cd->external_flags |= PCRE_RLSET;
8858: }
8859: skipatstart = p;
8860: continue;
8861: }
8862:
8863: if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_CR_RIGHTPAR, 3) == 0)
8864: { skipatstart += 5; newnl = PCRE_NEWLINE_CR; }
8865: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_LF_RIGHTPAR, 3) == 0)
8866: { skipatstart += 5; newnl = PCRE_NEWLINE_LF; }
8867: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_CRLF_RIGHTPAR, 5) == 0)
8868: { skipatstart += 7; newnl = PCRE_NEWLINE_CR + PCRE_NEWLINE_LF; }
8869: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_ANY_RIGHTPAR, 4) == 0)
8870: { skipatstart += 6; newnl = PCRE_NEWLINE_ANY; }
8871: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_ANYCRLF_RIGHTPAR, 8) == 0)
8872: { skipatstart += 10; newnl = PCRE_NEWLINE_ANYCRLF; }
8873:
8874: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_BSR_ANYCRLF_RIGHTPAR, 12) == 0)
8875: { skipatstart += 14; newbsr = PCRE_BSR_ANYCRLF; }
8876: else if (STRNCMP_UC_C8(ptr+skipatstart+2, STRING_BSR_UNICODE_RIGHTPAR, 12) == 0)
8877: { skipatstart += 14; newbsr = PCRE_BSR_UNICODE; }
8878:
8879: if (newnl != 0)
8880: options = (options & ~PCRE_NEWLINE_BITS) | newnl;
8881: else if (newbsr != 0)
8882: options = (options & ~(PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) | newbsr;
8883: else break;
8884: }
8885:
8886: /* PCRE_UTF(16|32) have the same value as PCRE_UTF8. */
8887: utf = (options & PCRE_UTF8) != 0;
8888: if (utf && never_utf)
8889: {
8890: errorcode = ERR78;
8891: goto PCRE_EARLY_ERROR_RETURN2;
8892: }
8893:
8894: /* Can't support UTF unless PCRE has been compiled to include the code. The
8895: return of an error code from PRIV(valid_utf)() is a new feature, introduced in
8896: release 8.13. It is passed back from pcre_[dfa_]exec(), but at the moment is
8897: not used here. */
8898:
8899: #ifdef SUPPORT_UTF
8900: if (utf && (options & PCRE_NO_UTF8_CHECK) == 0 &&
8901: (errorcode = PRIV(valid_utf)((PCRE_PUCHAR)pattern, -1, erroroffset)) != 0)
8902: {
8903: #if defined COMPILE_PCRE8
8904: errorcode = ERR44;
8905: #elif defined COMPILE_PCRE16
8906: errorcode = ERR74;
8907: #elif defined COMPILE_PCRE32
8908: errorcode = ERR77;
8909: #endif
8910: goto PCRE_EARLY_ERROR_RETURN2;
8911: }
8912: #else
8913: if (utf)
8914: {
8915: errorcode = ERR32;
8916: goto PCRE_EARLY_ERROR_RETURN;
8917: }
8918: #endif
8919:
8920: /* Can't support UCP unless PCRE has been compiled to include the code. */
8921:
8922: #ifndef SUPPORT_UCP
8923: if ((options & PCRE_UCP) != 0)
8924: {
8925: errorcode = ERR67;
8926: goto PCRE_EARLY_ERROR_RETURN;
8927: }
8928: #endif
8929:
8930: /* Check validity of \R options. */
8931:
8932: if ((options & (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE)) ==
8933: (PCRE_BSR_ANYCRLF|PCRE_BSR_UNICODE))
8934: {
8935: errorcode = ERR56;
8936: goto PCRE_EARLY_ERROR_RETURN;
8937: }
8938:
8939: /* Handle different types of newline. The three bits give seven cases. The
8940: current code allows for fixed one- or two-byte sequences, plus "any" and
8941: "anycrlf". */
8942:
8943: switch (options & PCRE_NEWLINE_BITS)
8944: {
8945: case 0: newline = NEWLINE; break; /* Build-time default */
8946: case PCRE_NEWLINE_CR: newline = CHAR_CR; break;
8947: case PCRE_NEWLINE_LF: newline = CHAR_NL; break;
8948: case PCRE_NEWLINE_CR+
8949: PCRE_NEWLINE_LF: newline = (CHAR_CR << 8) | CHAR_NL; break;
8950: case PCRE_NEWLINE_ANY: newline = -1; break;
8951: case PCRE_NEWLINE_ANYCRLF: newline = -2; break;
8952: default: errorcode = ERR56; goto PCRE_EARLY_ERROR_RETURN;
8953: }
8954:
8955: if (newline == -2)
8956: {
8957: cd->nltype = NLTYPE_ANYCRLF;
8958: }
8959: else if (newline < 0)
8960: {
8961: cd->nltype = NLTYPE_ANY;
8962: }
8963: else
8964: {
8965: cd->nltype = NLTYPE_FIXED;
8966: if (newline > 255)
8967: {
8968: cd->nllen = 2;
8969: cd->nl[0] = (newline >> 8) & 255;
8970: cd->nl[1] = newline & 255;
8971: }
8972: else
8973: {
8974: cd->nllen = 1;
8975: cd->nl[0] = newline;
8976: }
8977: }
8978:
8979: /* Maximum back reference and backref bitmap. The bitmap records up to 31 back
8980: references to help in deciding whether (.*) can be treated as anchored or not.
8981: */
8982:
8983: cd->top_backref = 0;
8984: cd->backref_map = 0;
8985:
8986: /* Reflect pattern for debugging output */
8987:
8988: DPRINTF(("------------------------------------------------------------------\n"));
8989: #ifdef PCRE_DEBUG
8990: print_puchar(stdout, (PCRE_PUCHAR)pattern);
8991: #endif
8992: DPRINTF(("\n"));
8993:
8994: /* Pretend to compile the pattern while actually just accumulating the length
8995: of memory required. This behaviour is triggered by passing a non-NULL final
8996: argument to compile_regex(). We pass a block of workspace (cworkspace) for it
8997: to compile parts of the pattern into; the compiled code is discarded when it is
8998: no longer needed, so hopefully this workspace will never overflow, though there
8999: is a test for its doing so. */
9000:
9001: cd->bracount = cd->final_bracount = 0;
9002: cd->names_found = 0;
9003: cd->name_entry_size = 0;
9004: cd->name_table = NULL;
9005: cd->dupnames = FALSE;
9006: cd->namedrefcount = 0;
9007: cd->start_code = cworkspace;
9008: cd->hwm = cworkspace;
9009: cd->start_workspace = cworkspace;
9010: cd->workspace_size = COMPILE_WORK_SIZE;
9011: cd->named_groups = named_groups;
9012: cd->named_group_list_size = NAMED_GROUP_LIST_SIZE;
9013: cd->start_pattern = (const pcre_uchar *)pattern;
9014: cd->end_pattern = (const pcre_uchar *)(pattern + STRLEN_UC((const pcre_uchar *)pattern));
9015: cd->req_varyopt = 0;
9016: cd->parens_depth = 0;
9017: cd->assert_depth = 0;
9018: cd->max_lookbehind = 0;
9019: cd->external_options = options;
9020: cd->open_caps = NULL;
9021:
9022: /* Now do the pre-compile. On error, errorcode will be set non-zero, so we
9023: don't need to look at the result of the function here. The initial options have
9024: been put into the cd block so that they can be changed if an option setting is
9025: found within the regex right at the beginning. Bringing initial option settings
9026: outside can help speed up starting point checks. */
9027:
9028: ptr += skipatstart;
9029: code = cworkspace;
9030: *code = OP_BRA;
9031:
9032: (void)compile_regex(cd->external_options, &code, &ptr, &errorcode, FALSE,
9033: FALSE, 0, 0, &firstchar, &firstcharflags, &reqchar, &reqcharflags, NULL,
9034: cd, &length);
9035: if (errorcode != 0) goto PCRE_EARLY_ERROR_RETURN;
9036:
9037: DPRINTF(("end pre-compile: length=%d workspace=%d\n", length,
9038: (int)(cd->hwm - cworkspace)));
9039:
9040: if (length > MAX_PATTERN_SIZE)
9041: {
9042: errorcode = ERR20;
9043: goto PCRE_EARLY_ERROR_RETURN;
9044: }
9045:
9046: /* If there are groups with duplicate names and there are also references by
9047: name, we must allow for the possibility of named references to duplicated
9048: groups. These require an extra data item each. */
9049:
9050: if (cd->dupnames && cd->namedrefcount > 0)
9051: length += cd->namedrefcount * IMM2_SIZE * sizeof(pcre_uchar);
9052:
9053: /* Compute the size of the data block for storing the compiled pattern. Integer
9054: overflow should no longer be possible because nowadays we limit the maximum
9055: value of cd->names_found and cd->name_entry_size. */
9056:
9057: size = sizeof(REAL_PCRE) +
9058: (length + cd->names_found * cd->name_entry_size) * sizeof(pcre_uchar);
9059:
9060: /* Get the memory. */
9061:
9062: re = (REAL_PCRE *)(PUBL(malloc))(size);
9063: if (re == NULL)
9064: {
9065: errorcode = ERR21;
9066: goto PCRE_EARLY_ERROR_RETURN;
9067: }
9068:
9069: /* Put in the magic number, and save the sizes, initial options, internal
9070: flags, and character table pointer. NULL is used for the default character
9071: tables. The nullpad field is at the end; it's there to help in the case when a
9072: regex compiled on a system with 4-byte pointers is run on another with 8-byte
9073: pointers. */
9074:
9075: re->magic_number = MAGIC_NUMBER;
9076: re->size = (int)size;
9077: re->options = cd->external_options;
9078: re->flags = cd->external_flags;
9079: re->limit_match = limit_match;
9080: re->limit_recursion = limit_recursion;
9081: re->first_char = 0;
9082: re->req_char = 0;
9083: re->name_table_offset = sizeof(REAL_PCRE) / sizeof(pcre_uchar);
9084: re->name_entry_size = cd->name_entry_size;
9085: re->name_count = cd->names_found;
9086: re->ref_count = 0;
9087: re->tables = (tables == PRIV(default_tables))? NULL : tables;
9088: re->nullpad = NULL;
9089: #ifdef COMPILE_PCRE32
9090: re->dummy = 0;
9091: #else
9092: re->dummy1 = re->dummy2 = re->dummy3 = 0;
9093: #endif
9094:
9095: /* The starting points of the name/number translation table and of the code are
9096: passed around in the compile data block. The start/end pattern and initial
9097: options are already set from the pre-compile phase, as is the name_entry_size
9098: field. Reset the bracket count and the names_found field. Also reset the hwm
9099: field; this time it's used for remembering forward references to subpatterns.
9100: */
9101:
9102: cd->final_bracount = cd->bracount; /* Save for checking forward references */
9103: cd->parens_depth = 0;
9104: cd->assert_depth = 0;
9105: cd->bracount = 0;
9106: cd->max_lookbehind = 0;
9107: cd->name_table = (pcre_uchar *)re + re->name_table_offset;
9108: codestart = cd->name_table + re->name_entry_size * re->name_count;
9109: cd->start_code = codestart;
9110: cd->hwm = (pcre_uchar *)(cd->start_workspace);
9111: cd->req_varyopt = 0;
9112: cd->had_accept = FALSE;
9113: cd->had_pruneorskip = FALSE;
9114: cd->check_lookbehind = FALSE;
9115: cd->open_caps = NULL;
9116:
9117: /* If any named groups were found, create the name/number table from the list
9118: created in the first pass. */
9119:
9120: if (cd->names_found > 0)
9121: {
9122: int i = cd->names_found;
9123: named_group *ng = cd->named_groups;
9124: cd->names_found = 0;
9125: for (; i > 0; i--, ng++)
9126: add_name(cd, ng->name, ng->length, ng->number);
9127: if (cd->named_group_list_size > NAMED_GROUP_LIST_SIZE)
9128: (PUBL(free))((void *)cd->named_groups);
9129: }
9130:
9131: /* Set up a starting, non-extracting bracket, then compile the expression. On
9132: error, errorcode will be set non-zero, so we don't need to look at the result
9133: of the function here. */
9134:
9135: ptr = (const pcre_uchar *)pattern + skipatstart;
9136: code = (pcre_uchar *)codestart;
9137: *code = OP_BRA;
9138: (void)compile_regex(re->options, &code, &ptr, &errorcode, FALSE, FALSE, 0, 0,
9139: &firstchar, &firstcharflags, &reqchar, &reqcharflags, NULL, cd, NULL);
9140: re->top_bracket = cd->bracount;
9141: re->top_backref = cd->top_backref;
9142: re->max_lookbehind = cd->max_lookbehind;
9143: re->flags = cd->external_flags | PCRE_MODE;
9144:
9145: if (cd->had_accept)
9146: {
9147: reqchar = 0; /* Must disable after (*ACCEPT) */
9148: reqcharflags = REQ_NONE;
9149: }
9150:
9151: /* If not reached end of pattern on success, there's an excess bracket. */
9152:
9153: if (errorcode == 0 && *ptr != CHAR_NULL) errorcode = ERR22;
9154:
9155: /* Fill in the terminating state and check for disastrous overflow, but
9156: if debugging, leave the test till after things are printed out. */
9157:
9158: *code++ = OP_END;
9159:
9160: #ifndef PCRE_DEBUG
9161: if (code - codestart > length) errorcode = ERR23;
9162: #endif
9163:
9164: #ifdef SUPPORT_VALGRIND
9165: /* If the estimated length exceeds the really used length, mark the extra
9166: allocated memory as unaddressable, so that any out-of-bound reads can be
9167: detected. */
9168: VALGRIND_MAKE_MEM_NOACCESS(code, (length - (code - codestart)) * sizeof(pcre_uchar));
9169: #endif
9170:
9171: /* Fill in any forward references that are required. There may be repeated
9172: references; optimize for them, as searching a large regex takes time. */
9173:
9174: if (cd->hwm > cd->start_workspace)
9175: {
9176: int prev_recno = -1;
9177: const pcre_uchar *groupptr = NULL;
9178: while (errorcode == 0 && cd->hwm > cd->start_workspace)
9179: {
9180: int offset, recno;
9181: cd->hwm -= LINK_SIZE;
9182: offset = GET(cd->hwm, 0);
9183: recno = GET(codestart, offset);
9184: if (recno != prev_recno)
9185: {
9186: groupptr = PRIV(find_bracket)(codestart, utf, recno);
9187: prev_recno = recno;
9188: }
9189: if (groupptr == NULL) errorcode = ERR53;
9190: else PUT(((pcre_uchar *)codestart), offset, (int)(groupptr - codestart));
9191: }
9192: }
9193:
9194: /* If the workspace had to be expanded, free the new memory. Set the pointer to
9195: NULL to indicate that forward references have been filled in. */
9196:
9197: if (cd->workspace_size > COMPILE_WORK_SIZE)
9198: (PUBL(free))((void *)cd->start_workspace);
9199: cd->start_workspace = NULL;
9200:
9201: /* Give an error if there's back reference to a non-existent capturing
9202: subpattern. */
9203:
9204: if (errorcode == 0 && re->top_backref > re->top_bracket) errorcode = ERR15;
9205:
9206: /* Unless disabled, check whether single character iterators can be
9207: auto-possessified. The function overwrites the appropriate opcode values. */
9208:
9209: if ((options & PCRE_NO_AUTO_POSSESS) == 0)
9210: auto_possessify((pcre_uchar *)codestart, utf, cd);
9211:
9212: /* If there were any lookbehind assertions that contained OP_RECURSE
9213: (recursions or subroutine calls), a flag is set for them to be checked here,
9214: because they may contain forward references. Actual recursions cannot be fixed
9215: length, but subroutine calls can. It is done like this so that those without
9216: OP_RECURSE that are not fixed length get a diagnosic with a useful offset. The
9217: exceptional ones forgo this. We scan the pattern to check that they are fixed
9218: length, and set their lengths. */
9219:
9220: if (cd->check_lookbehind)
9221: {
9222: pcre_uchar *cc = (pcre_uchar *)codestart;
9223:
9224: /* Loop, searching for OP_REVERSE items, and process those that do not have
9225: their length set. (Actually, it will also re-process any that have a length
9226: of zero, but that is a pathological case, and it does no harm.) When we find
9227: one, we temporarily terminate the branch it is in while we scan it. */
9228:
9229: for (cc = (pcre_uchar *)PRIV(find_bracket)(codestart, utf, -1);
9230: cc != NULL;
9231: cc = (pcre_uchar *)PRIV(find_bracket)(cc, utf, -1))
9232: {
9233: if (GET(cc, 1) == 0)
9234: {
9235: int fixed_length;
9236: pcre_uchar *be = cc - 1 - LINK_SIZE + GET(cc, -LINK_SIZE);
9237: int end_op = *be;
9238: *be = OP_END;
9239: fixed_length = find_fixedlength(cc, (re->options & PCRE_UTF8) != 0, TRUE,
9240: cd);
9241: *be = end_op;
9242: DPRINTF(("fixed length = %d\n", fixed_length));
9243: if (fixed_length < 0)
9244: {
9245: errorcode = (fixed_length == -2)? ERR36 :
9246: (fixed_length == -4)? ERR70 : ERR25;
9247: break;
9248: }
9249: if (fixed_length > cd->max_lookbehind) cd->max_lookbehind = fixed_length;
9250: PUT(cc, 1, fixed_length);
9251: }
9252: cc += 1 + LINK_SIZE;
9253: }
9254: }
9255:
9256: /* Failed to compile, or error while post-processing */
9257:
9258: if (errorcode != 0)
9259: {
9260: (PUBL(free))(re);
9261: PCRE_EARLY_ERROR_RETURN:
9262: *erroroffset = (int)(ptr - (const pcre_uchar *)pattern);
9263: PCRE_EARLY_ERROR_RETURN2:
9264: *errorptr = find_error_text(errorcode);
9265: if (errorcodeptr != NULL) *errorcodeptr = errorcode;
9266: return NULL;
9267: }
9268:
9269: /* If the anchored option was not passed, set the flag if we can determine that
9270: the pattern is anchored by virtue of ^ characters or \A or anything else, such
9271: as starting with non-atomic .* when DOTALL is set and there are no occurrences
9272: of *PRUNE or *SKIP.
9273:
9274: Otherwise, if we know what the first byte has to be, save it, because that
9275: speeds up unanchored matches no end. If not, see if we can set the
9276: PCRE_STARTLINE flag. This is helpful for multiline matches when all branches
9277: start with ^. and also when all branches start with non-atomic .* for
9278: non-DOTALL matches when *PRUNE and SKIP are not present. */
9279:
9280: if ((re->options & PCRE_ANCHORED) == 0)
9281: {
9282: if (is_anchored(codestart, 0, cd, 0)) re->options |= PCRE_ANCHORED;
9283: else
9284: {
9285: if (firstcharflags < 0)
9286: firstchar = find_firstassertedchar(codestart, &firstcharflags, FALSE);
9287: if (firstcharflags >= 0) /* Remove caseless flag for non-caseable chars */
9288: {
9289: #if defined COMPILE_PCRE8
9290: re->first_char = firstchar & 0xff;
9291: #elif defined COMPILE_PCRE16
9292: re->first_char = firstchar & 0xffff;
9293: #elif defined COMPILE_PCRE32
9294: re->first_char = firstchar;
9295: #endif
9296: if ((firstcharflags & REQ_CASELESS) != 0)
9297: {
9298: #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
9299: /* We ignore non-ASCII first chars in 8 bit mode. */
9300: if (utf)
9301: {
9302: if (re->first_char < 128)
9303: {
9304: if (cd->fcc[re->first_char] != re->first_char)
9305: re->flags |= PCRE_FCH_CASELESS;
9306: }
9307: else if (UCD_OTHERCASE(re->first_char) != re->first_char)
9308: re->flags |= PCRE_FCH_CASELESS;
9309: }
9310: else
9311: #endif
9312: if (MAX_255(re->first_char)
9313: && cd->fcc[re->first_char] != re->first_char)
9314: re->flags |= PCRE_FCH_CASELESS;
9315: }
9316:
9317: re->flags |= PCRE_FIRSTSET;
9318: }
9319:
9320: else if (is_startline(codestart, 0, cd, 0)) re->flags |= PCRE_STARTLINE;
9321: }
9322: }
9323:
9324: /* For an anchored pattern, we use the "required byte" only if it follows a
9325: variable length item in the regex. Remove the caseless flag for non-caseable
9326: bytes. */
9327:
9328: if (reqcharflags >= 0 &&
9329: ((re->options & PCRE_ANCHORED) == 0 || (reqcharflags & REQ_VARY) != 0))
9330: {
9331: #if defined COMPILE_PCRE8
9332: re->req_char = reqchar & 0xff;
9333: #elif defined COMPILE_PCRE16
9334: re->req_char = reqchar & 0xffff;
9335: #elif defined COMPILE_PCRE32
9336: re->req_char = reqchar;
9337: #endif
9338: if ((reqcharflags & REQ_CASELESS) != 0)
9339: {
9340: #if defined SUPPORT_UCP && !(defined COMPILE_PCRE8)
9341: /* We ignore non-ASCII first chars in 8 bit mode. */
9342: if (utf)
9343: {
9344: if (re->req_char < 128)
9345: {
9346: if (cd->fcc[re->req_char] != re->req_char)
9347: re->flags |= PCRE_RCH_CASELESS;
9348: }
9349: else if (UCD_OTHERCASE(re->req_char) != re->req_char)
9350: re->flags |= PCRE_RCH_CASELESS;
9351: }
9352: else
9353: #endif
9354: if (MAX_255(re->req_char) && cd->fcc[re->req_char] != re->req_char)
9355: re->flags |= PCRE_RCH_CASELESS;
9356: }
9357:
9358: re->flags |= PCRE_REQCHSET;
9359: }
9360:
9361: /* Print out the compiled data if debugging is enabled. This is never the
9362: case when building a production library. */
9363:
9364: #ifdef PCRE_DEBUG
9365: printf("Length = %d top_bracket = %d top_backref = %d\n",
9366: length, re->top_bracket, re->top_backref);
9367:
9368: printf("Options=%08x\n", re->options);
9369:
9370: if ((re->flags & PCRE_FIRSTSET) != 0)
9371: {
9372: pcre_uchar ch = re->first_char;
9373: const char *caseless =
9374: ((re->flags & PCRE_FCH_CASELESS) == 0)? "" : " (caseless)";
9375: if (PRINTABLE(ch)) printf("First char = %c%s\n", ch, caseless);
9376: else printf("First char = \\x%02x%s\n", ch, caseless);
9377: }
9378:
9379: if ((re->flags & PCRE_REQCHSET) != 0)
9380: {
9381: pcre_uchar ch = re->req_char;
9382: const char *caseless =
9383: ((re->flags & PCRE_RCH_CASELESS) == 0)? "" : " (caseless)";
9384: if (PRINTABLE(ch)) printf("Req char = %c%s\n", ch, caseless);
9385: else printf("Req char = \\x%02x%s\n", ch, caseless);
9386: }
9387:
9388: #if defined COMPILE_PCRE8
9389: pcre_printint((pcre *)re, stdout, TRUE);
9390: #elif defined COMPILE_PCRE16
9391: pcre16_printint((pcre *)re, stdout, TRUE);
9392: #elif defined COMPILE_PCRE32
9393: pcre32_printint((pcre *)re, stdout, TRUE);
9394: #endif
9395:
9396: /* This check is done here in the debugging case so that the code that
9397: was compiled can be seen. */
9398:
9399: if (code - codestart > length)
9400: {
9401: (PUBL(free))(re);
9402: *errorptr = find_error_text(ERR23);
9403: *erroroffset = ptr - (pcre_uchar *)pattern;
9404: if (errorcodeptr != NULL) *errorcodeptr = ERR23;
9405: return NULL;
9406: }
9407: #endif /* PCRE_DEBUG */
9408:
9409: /* Check for a pattern than can match an empty string, so that this information
9410: can be provided to applications. */
9411:
9412: do
9413: {
9414: if (could_be_empty_branch(codestart, code, utf, cd, NULL))
9415: {
9416: re->flags |= PCRE_MATCH_EMPTY;
9417: break;
9418: }
9419: codestart += GET(codestart, 1);
9420: }
9421: while (*codestart == OP_ALT);
9422:
9423: #if defined COMPILE_PCRE8
9424: return (pcre *)re;
9425: #elif defined COMPILE_PCRE16
9426: return (pcre16 *)re;
9427: #elif defined COMPILE_PCRE32
9428: return (pcre32 *)re;
9429: #endif
9430: }
9431:
9432: /* End of pcre_compile.c */
9433:
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>