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

Diff of /code/trunk/pcre_compile.c

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 111 by ph10, Thu Mar 8 16:53:09 2007 UTC revision 598 by ph10, Sat May 7 15:37:31 2011 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2006 University of Cambridge             Copyright (c) 1997-2011 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  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 */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
# Line 76  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 106  static const short int escapes[] = { Line 162  static const short int escapes[] = {
162  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
163  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
164  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
165  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
166  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
167  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
168  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
169  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
170  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
173  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
178  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
179  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 125  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 164  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 171  static const int posix_class_maps[] = { Line 315  static const int posix_class_maps[] = {
315  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
316  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
317  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
318  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
319    the number of relocations needed when a shared library is loaded dynamically,
320  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
321    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
323    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
324    "unrecognized character follows \\",  
325    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329    static const char error_texts[] =
330      "no error\0"
331      "\\ at end of pattern\0"
332      "\\c at end of pattern\0"
333      "unrecognized character follows \\\0"
334      "numbers out of order in {} quantifier\0"
335    /* 5 */    /* 5 */
336    "number too big in {} quantifier",    "number too big in {} quantifier\0"
337    "missing terminating ] for character class",    "missing terminating ] for character class\0"
338    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
339    "range out of order in character class",    "range out of order in character class\0"
340    "nothing to repeat",    "nothing to repeat\0"
341    /* 10 */    /* 10 */
342    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
343    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
344    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
345    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
346    "missing )",    "missing )\0"
347    /* 15 */    /* 15 */
348    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
349    "erroffset passed as NULL",    "erroffset passed as NULL\0"
350    "unknown option bit(s) set",    "unknown option bit(s) set\0"
351    "missing ) after comment",    "missing ) after comment\0"
352    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
353    /* 20 */    /* 20 */
354    "regular expression too large",    "regular expression is too large\0"
355    "failed to get memory",    "failed to get memory\0"
356    "unmatched parentheses",    "unmatched parentheses\0"
357    "internal error: code overflow",    "internal error: code overflow\0"
358    "unrecognized character after (?<",    "unrecognized character after (?<\0"
359    /* 25 */    /* 25 */
360    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
361    "malformed number or name after (?(",    "malformed number or name after (?(\0"
362    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
363    "assertion expected after (?(",    "assertion expected after (?(\0"
364    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
365    /* 30 */    /* 30 */
366    "unknown POSIX class name",    "unknown POSIX class name\0"
367    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
368    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
369    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
370    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)",    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255",    "number after (?C is > 255\0"
376    "closing ) for (?C expected",    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
378    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
379    "unrecognized character after (?P",    "unrecognized character after (?P\0"
380    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
381    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
387    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389    /* 50 */    /* 50 */
390    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
391    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
392    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
393    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
397    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
398    "\\g is not followed by an (optionally braced) non-zero number"    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399  };    "a numbered reference must not be zero\0"
400      "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401      /* 60 */
402      "(*VERB) not recognized\0"
403      "number is too big\0"
404      "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      "\\c must be followed by an ASCII character\0"
412      ;
413    
414  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
415  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 262  For convenience, we use the same bit def Line 427  For convenience, we use the same bit def
427    
428  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
429    
430  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
431    
432    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
433    UTF-8 mode. */
434    
435  static const unsigned char digitab[] =  static const unsigned char digitab[] =
436    {    {
437    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 298  static const unsigned char digitab[] = Line 467  static const unsigned char digitab[] =
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
468    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
469    
470  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
471    
472    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
473    
474  static const unsigned char digitab[] =  static const unsigned char digitab[] =
475    {    {
476    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 373  static const unsigned char ebcdic_charta Line 545  static const unsigned char ebcdic_charta
545  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
546    
547  static BOOL  static BOOL
548    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
549      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
550    
551    
552    
553    /*************************************************
554    *            Find an error text                  *
555    *************************************************/
556    
557    /* The error texts are now all in one long string, to save on relocations. As
558    some of the text is of unknown length, we can't use a table of offsets.
559    Instead, just count through the strings. This is not a performance issue
560    because it happens only when there has been a compilation error.
561    
562    Argument:   the error number
563    Returns:    pointer to the error string
564    */
565    
566    static const char *
567    find_error_text(int n)
568    {
569    const char *s = error_texts;
570    for (; n > 0; n--)
571      {
572      while (*s++ != 0) {};
573      if (*s == 0) return "Error text not found (please report)";
574      }
575    return s;
576    }
577    
578    
579  /*************************************************  /*************************************************
# Line 399  Arguments: Line 597  Arguments:
597    
598  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
599                   negative => a special escape sequence                   negative => a special escape sequence
600                   on error, errorptr is set                   on error, errorcodeptr is set
601  */  */
602    
603  static int  static int
# Line 417  ptr--; /* Set Line 615  ptr--; /* Set
615    
616  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
617    
618  /* Non-alphamerics are literals. For digits or letters, do an initial lookup in  /* Non-alphanumerics are literals. For digits or letters, do an initial lookup
619  a table. A non-zero result is something that can be returned immediately.  in a table. A non-zero result is something that can be returned immediately.
620  Otherwise further processing may be required. */  Otherwise further processing may be required. */
621    
622  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
623  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
624  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
625    
626  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
627  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
628  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
629  #endif  #endif
630    
# Line 442  else Line 640  else
640      /* A number of Perl escapes are not handled by PCRE. We give an explicit      /* A number of Perl escapes are not handled by PCRE. We give an explicit
641      error. */      error. */
642    
643      case 'l':      case CHAR_l:
644      case 'L':      case CHAR_L:
645      case 'N':      case CHAR_u:
646      case 'u':      case CHAR_U:
     case 'U':  
647      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
648      break;      break;
649    
650      /* \g must be followed by a number, either plain or braced. If positive, it      /* \g must be followed by one of a number of specific things:
651      is an absolute backreference. If negative, it is a relative backreference.  
652      This is a Perl 5.10 feature. */      (1) A number, either plain or braced. If positive, it is an absolute
653        backreference. If negative, it is a relative backreference. This is a Perl
654        5.10 feature.
655    
656        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
657        is part of Perl's movement towards a unified syntax for back references. As
658        this is synonymous with \k{name}, we fudge it up by pretending it really
659        was \k.
660    
661        (3) For Oniguruma compatibility we also support \g followed by a name or a
662        number either in angle brackets or in single quotes. However, these are
663        (possibly recursive) subroutine calls, _not_ backreferences. Just return
664        the -ESC_g code (cf \k). */
665    
666      case 'g':      case CHAR_g:
667      if (ptr[1] == '{')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
668        {        {
669          c = -ESC_g;
670          break;
671          }
672    
673        /* Handle the Perl-compatible cases */
674    
675        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
676          {
677          const uschar *p;
678          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
679            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
680          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
681            {
682            c = -ESC_k;
683            break;
684            }
685        braced = TRUE;        braced = TRUE;
686        ptr++;        ptr++;
687        }        }
688      else braced = FALSE;      else braced = FALSE;
689    
690      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
691        {        {
692        negated = TRUE;        negated = TRUE;
693        ptr++;        ptr++;
# Line 471  else Line 696  else
696    
697      c = 0;      c = 0;
698      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
699        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
700    
701        if (c < 0)   /* Integer overflow */
702          {
703          *errorcodeptr = ERR61;
704          break;
705          }
706    
707      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
708        {        {
709        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
710        return 0;        break;
711          }
712    
713        if (c == 0)
714          {
715          *errorcodeptr = ERR58;
716          break;
717        }        }
718    
719      if (negated)      if (negated)
# Line 484  else Line 721  else
721        if (c > bracount)        if (c > bracount)
722          {          {
723          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
724          return 0;          break;
725          }          }
726        c = bracount - (c - 1);        c = bracount - (c - 1);
727        }        }
# Line 504  else Line 741  else
741      value is greater than 377, the least significant 8 bits are taken. Inside a      value is greater than 377, the least significant 8 bits are taken. Inside a
742      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
743    
744      case '1': case '2': case '3': case '4': case '5':      case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4: case CHAR_5:
745      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
746    
747      if (!isclass)      if (!isclass)
748        {        {
749        oldptr = ptr;        oldptr = ptr;
750        c -= '0';        c -= CHAR_0;
751        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
752          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
753          if (c < 0)    /* Integer overflow */
754            {
755            *errorcodeptr = ERR61;
756            break;
757            }
758        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
759          {          {
760          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 525  else Line 767  else
767      generates a binary zero byte and treats the digit as a following literal.      generates a binary zero byte and treats the digit as a following literal.
768      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
769    
770      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
771        {        {
772        ptr--;        ptr--;
773        c = 0;        c = 0;
# Line 538  else Line 780  else
780      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more      to do). Nowadays we allow for larger numbers in UTF-8 mode, but no more
781      than 3 octal digits. */      than 3 octal digits. */
782    
783      case '0':      case CHAR_0:
784      c -= '0';      c -= CHAR_0;
785      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
786          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
787      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
788      break;      break;
789    
# Line 549  else Line 791  else
791      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is      than 0xff in utf8 mode, but only if the ddd are hex digits. If not, { is
792      treated as a data character. */      treated as a data character. */
793    
794      case 'x':      case CHAR_x:
795      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
796        {        {
797        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
798        int count = 0;        int count = 0;
# Line 559  else Line 801  else
801        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
802          {          {
803          register int cc = *pt++;          register int cc = *pt++;
804          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
805          count++;          count++;
806    
807  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
808          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
809          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
810  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
811          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
812          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
813  #endif  #endif
814          }          }
815    
816        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
817          {          {
818          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
819          ptr = pt;          ptr = pt;
# Line 587  else Line 829  else
829      c = 0;      c = 0;
830      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
831        {        {
832        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
833        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
834  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
835        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
836        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
837  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
838        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
839        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
840  #endif  #endif
841        }        }
842      break;      break;
843    
844      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.      /* For \c, a following letter is upper-cased; then the 0x40 bit is flipped.
845      This coding is ASCII-specific, but then the whole concept of \cx is      An error is given if the byte following \c is not an ASCII character. This
846        coding is ASCII-specific, but then the whole concept of \cx is
847      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
848    
849      case 'c':      case CHAR_c:
850      c = *(++ptr);      c = *(++ptr);
851      if (c == 0)      if (c == 0)
852        {        {
853        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
854        return 0;        break;
855        }        }
856    #ifndef EBCDIC    /* ASCII/UTF-8 coding */
857  #ifndef EBCDIC  /* ASCII coding */      if (c > 127)  /* Excludes all non-ASCII in either mode */
858      if (c >= 'a' && c <= 'z') c -= 32;        {
859          *errorcodeptr = ERR68;
860          break;
861          }
862        if (c >= CHAR_a && c <= CHAR_z) c -= 32;
863      c ^= 0x40;      c ^= 0x40;
864  #else           /* EBCDIC coding */  #else             /* EBCDIC coding */
865      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
866      c ^= 0xC0;      c ^= 0xC0;
867  #endif  #endif
868      break;      break;
869    
870      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
871      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
872      for Perl compatibility, it is a literal. This code looks a bit odd, but      otherwise, for Perl compatibility, it is a literal. This code looks a bit
873      there used to be some cases other than the default, and there may be again      odd, but there used to be some cases other than the default, and there may
874      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
875    
876      default:      default:
877      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 637  else Line 884  else
884      }      }
885    }    }
886    
887    /* Perl supports \N{name} for character names, as well as plain \N for "not
888    newline". PCRE does not support \N{name}. */
889    
890    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
891      *errorcodeptr = ERR37;
892    
893    /* If PCRE_UCP is set, we change the values for \d etc. */
894    
895    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
896      c -= (ESC_DU - ESC_D);
897    
898    /* Set the pointer to the final character before returning. */
899    
900  *ptrptr = ptr;  *ptrptr = ptr;
901  return c;  return c;
902  }  }
# Line 677  if (c == 0) goto ERROR_RETURN; Line 937  if (c == 0) goto ERROR_RETURN;
937  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for  /* \P or \p can be followed by a name in {}, optionally preceded by ^ for
938  negation. */  negation. */
939    
940  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
941    {    {
942    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
943      {      {
944      *negptr = TRUE;      *negptr = TRUE;
945      ptr++;      ptr++;
946      }      }
947    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
948      {      {
949      c = *(++ptr);      c = *(++ptr);
950      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
951      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
952      name[i] = c;      name[i] = c;
953      }      }
954    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
955    name[i] = 0;    name[i] = 0;
956    }    }
957    
# Line 713  top = _pcre_utt_size; Line 973  top = _pcre_utt_size;
973  while (bot < top)  while (bot < top)
974    {    {
975    i = (bot + top) >> 1;    i = (bot + top) >> 1;
976    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
977    if (c == 0)    if (c == 0)
978      {      {
979      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 756  is_counted_repeat(const uschar *p) Line 1016  is_counted_repeat(const uschar *p)
1016  {  {
1017  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1018  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1019  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1020    
1021  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1022  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1023    
1024  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1025  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1026    
1027  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1028  }  }
1029    
1030    
# Line 797  int max = -1; Line 1057  int max = -1;
1057  /* Read the minimum value and do a paranoid check: a negative value indicates  /* Read the minimum value and do a paranoid check: a negative value indicates
1058  an integer overflow. */  an integer overflow. */
1059    
1060  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1061  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1062    {    {
1063    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 1067  if (min < 0 || min > 65535)
1067  /* Read the maximum value if there is one, and again do a paranoid on its size.  /* Read the maximum value if there is one, and again do a paranoid on its size.
1068  Also, max must not be less than min. */  Also, max must not be less than min. */
1069    
1070  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1071    {    {
1072    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1073      {      {
1074      max = 0;      max = 0;
1075      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1076      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1077        {        {
1078        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1097  return p;
1097    
1098    
1099  /*************************************************  /*************************************************
1100  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1101  *************************************************/  *************************************************/
1102    
1103  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1104    top-level call starts at the beginning of the pattern. All other calls must
1105    start at a parenthesis. It scans along a pattern's text looking for capturing
1106  subpatterns, and counting them. If it finds a named pattern that matches the  subpatterns, and counting them. If it finds a named pattern that matches the
1107  name it is given, it returns its number. Alternatively, if the name is NULL, it  name it is given, it returns its number. Alternatively, if the name is NULL, it
1108  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. Recursion is used to keep
1109  references to subpatterns. We know that if (?P< is encountered, the name will  track of subpatterns that reset the capturing group numbers - the (?| feature.
1110  be terminated by '>' because that is checked in the first pass.  
1111    This function was originally called only from the second pass, in which we know
1112    that if (?< or (?' or (?P< is encountered, the name will be correctly
1113    terminated because that is checked in the first pass. There is now one call to
1114    this function in the first pass, to check for a recursive back reference by
1115    name (so that we can make the whole group atomic). In this case, we need check
1116    only up to the current position in the pattern, and that is still OK because
1117    and previous occurrences will have been checked. To make this work, the test
1118    for "end of pattern" is a check against cd->end_pattern in the main loop,
1119    instead of looking for a binary zero. This means that the special first-pass
1120    call can adjust cd->end_pattern temporarily. (Checks for binary zero while
1121    processing items within the loop are OK, because afterwards the main loop will
1122    terminate.)
1123    
1124  Arguments:  Arguments:
1125    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1126    count        current count of capturing parens so far encountered    cd           compile background data
1127    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1128    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1129    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1130      utf8         TRUE if we are in UTF-8 mode
1131      count        pointer to the current capturing subpattern number (updated)
1132    
1133  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1134  */  */
1135    
1136  static int  static int
1137  find_parens(const uschar *ptr, int count, const uschar *name, int lorn,  find_parens_sub(uschar **ptrptr, compile_data *cd, const uschar *name, int lorn,
1138    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1139  {  {
1140  const uschar *thisname;  uschar *ptr = *ptrptr;
1141    int start_count = *count;
1142    int hwm_count = start_count;
1143    BOOL dup_parens = FALSE;
1144    
1145    /* If the first character is a parenthesis, check on the type of group we are
1146    dealing with. The very first call may not start with a parenthesis. */
1147    
1148  for (; *ptr != 0; ptr++)  if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1149    {    {
1150    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1151    
1152      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1153    
1154      /* Handle a normal, unnamed capturing parenthesis. */
1155    
1156      else if (ptr[1] != CHAR_QUESTION_MARK)
1157        {
1158        *count += 1;
1159        if (name == NULL && *count == lorn) return *count;
1160        ptr++;
1161        }
1162    
1163      /* All cases now have (? at the start. Remember when we are in a group
1164      where the parenthesis numbers are duplicated. */
1165    
1166      else if (ptr[2] == CHAR_VERTICAL_LINE)
1167        {
1168        ptr += 3;
1169        dup_parens = TRUE;
1170        }
1171    
1172      /* Handle comments; all characters are allowed until a ket is reached. */
1173    
1174      else if (ptr[2] == CHAR_NUMBER_SIGN)
1175        {
1176        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1177        goto FAIL_EXIT;
1178        }
1179    
1180      /* Handle a condition. If it is an assertion, just carry on so that it
1181      is processed as normal. If not, skip to the closing parenthesis of the
1182      condition (there can't be any nested parens). */
1183    
1184      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1185        {
1186        ptr += 2;
1187        if (ptr[1] != CHAR_QUESTION_MARK)
1188          {
1189          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1190          if (*ptr != 0) ptr++;
1191          }
1192        }
1193    
1194      /* Start with (? but not a condition. */
1195    
1196      else
1197        {
1198        ptr += 2;
1199        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1200    
1201        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1202    
1203        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1204            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1205          {
1206          int term;
1207          const uschar *thisname;
1208          *count += 1;
1209          if (name == NULL && *count == lorn) return *count;
1210          term = *ptr++;
1211          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1212          thisname = ptr;
1213          while (*ptr != term) ptr++;
1214          if (name != NULL && lorn == ptr - thisname &&
1215              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1216            return *count;
1217          term++;
1218          }
1219        }
1220      }
1221    
1222    /* Past any initial parenthesis handling, scan for parentheses or vertical
1223    bars. Stop if we get to cd->end_pattern. Note that this is important for the
1224    first-pass call when this value is temporarily adjusted to stop at the current
1225    position. So DO NOT change this to a test for binary zero. */
1226    
1227    for (; ptr < cd->end_pattern; ptr++)
1228      {
1229    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1230    
1231    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1232      {      {
1233      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1234      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1235        {        {
1236        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1237        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1238        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1239        }        }
1240      continue;      continue;
1241      }      }
1242    
1243    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1244      are handled for real. If the first character is '^', skip it. Also, if the
1245      first few characters (either before or after ^) are \Q\E or \E we skip them
1246      too. This makes for compatibility with Perl. Note the use of STR macros to
1247      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1248    
1249    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1250      {      {
1251      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1252        for (;;)
1253        {        {
1254        if (*ptr == '\\')        if (ptr[1] == CHAR_BACKSLASH)
1255          {          {
1256          if (*(++ptr) == 0) return -1;          if (ptr[2] == CHAR_E)
1257          if (*ptr == 'Q') for (;;)            ptr+= 2;
1258            else if (strncmp((const char *)ptr+2,
1259                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1260              ptr += 4;
1261            else
1262              break;
1263            }
1264          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1265            {
1266            negate_class = TRUE;
1267            ptr++;
1268            }
1269          else break;
1270          }
1271    
1272        /* If the next character is ']', it is a data character that must be
1273        skipped, except in JavaScript compatibility mode. */
1274    
1275        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1276            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1277          ptr++;
1278    
1279        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1280          {
1281          if (*ptr == 0) return -1;
1282          if (*ptr == CHAR_BACKSLASH)
1283            {
1284            if (*(++ptr) == 0) goto FAIL_EXIT;
1285            if (*ptr == CHAR_Q) for (;;)
1286            {            {
1287            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1288            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1289            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1290            }            }
1291          continue;          continue;
1292          }          }
# Line 904  for (; *ptr != 0; ptr++) Line 1296  for (; *ptr != 0; ptr++)
1296    
1297    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1298    
1299    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1300      {      {
1301      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1302      if (*ptr == 0) return -1;      while (*ptr != 0)
1303          {
1304          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1305          ptr++;
1306    #ifdef SUPPORT_UTF8
1307          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1308    #endif
1309          }
1310        if (*ptr == 0) goto FAIL_EXIT;
1311      continue;      continue;
1312      }      }
1313    
1314    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1315    
1316    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1317      {      {
1318      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1319      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1320      continue;      if (*ptr == 0) goto FAIL_EXIT;
1321      }      }
1322    
1323    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1324    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1325        if (dup_parens && *count < hwm_count) *count = hwm_count;
1326        goto FAIL_EXIT;
1327        }
1328    
1329      else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1330        {
1331        if (*count > hwm_count) hwm_count = *count;
1332        *count = start_count;
1333        }
1334      }
1335    
1336    FAIL_EXIT:
1337    *ptrptr = ptr;
1338    return -1;
1339    }
1340    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1341    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1342    
   count++;  
1343    
1344    if (name == NULL && count == lorn) return count;  /*************************************************
1345    term = *ptr++;  *       Find forward referenced subpattern       *
1346    if (term == '<') term = '>';  *************************************************/
1347    thisname = ptr;  
1348    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1349    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1350        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1351      return count;  returns when it reaches a given numbered subpattern. This is used for forward
1352    references to subpatterns. We used to be able to start this scan from the
1353    current compiling point, using the current count value from cd->bracount, and
1354    do it all in a single loop, but the addition of the possibility of duplicate
1355    subpattern numbers means that we have to scan from the very start, in order to
1356    take account of such duplicates, and to use a recursive function to keep track
1357    of the different types of group.
1358    
1359    Arguments:
1360      cd           compile background data
1361      name         name to seek, or NULL if seeking a numbered subpattern
1362      lorn         name length, or subpattern number if name is NULL
1363      xmode        TRUE if we are in /x mode
1364      utf8         TRUE if we are in UTF-8 mode
1365    
1366    Returns:       the number of the found subpattern, or -1 if not found
1367    */
1368    
1369    static int
1370    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1371      BOOL utf8)
1372    {
1373    uschar *ptr = (uschar *)cd->start_pattern;
1374    int count = 0;
1375    int rc;
1376    
1377    /* If the pattern does not start with an opening parenthesis, the first call
1378    to find_parens_sub() will scan right to the end (if necessary). However, if it
1379    does start with a parenthesis, find_parens_sub() will return when it hits the
1380    matching closing parens. That is why we have to have a loop. */
1381    
1382    for (;;)
1383      {
1384      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1385      if (rc > 0 || *ptr++ == 0) break;
1386    }    }
1387    
1388  return -1;  return rc;
1389  }  }
1390    
1391    
1392    
1393    
1394  /*************************************************  /*************************************************
1395  *      Find first significant op code            *  *      Find first significant op code            *
1396  *************************************************/  *************************************************/
# Line 996  for (;;) Line 1440  for (;;)
1440    
1441      case OP_CALLOUT:      case OP_CALLOUT:
1442      case OP_CREF:      case OP_CREF:
1443        case OP_NCREF:
1444      case OP_RREF:      case OP_RREF:
1445        case OP_NRREF:
1446      case OP_DEF:      case OP_DEF:
1447      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1448      break;      break;
# Line 1012  for (;;) Line 1458  for (;;)
1458    
1459    
1460  /*************************************************  /*************************************************
1461  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1462  *************************************************/  *************************************************/
1463    
1464  /* Scan a pattern and compute the fixed length of subject that will match it,  /* Scan a branch and compute the fixed length of subject that will match it,
1465  if the length is fixed. This is needed for dealing with backward assertions.  if the length is fixed. This is needed for dealing with backward assertions.
1466  In UTF8 mode, the result is in characters rather than bytes.  In UTF8 mode, the result is in characters rather than bytes. The branch is
1467    temporarily terminated with OP_END when this function is called.
1468    
1469    This function is called when a backward assertion is encountered, so that if it
1470    fails, the error message can point to the correct place in the pattern.
1471    However, we cannot do this when the assertion contains subroutine calls,
1472    because they can be forward references. We solve this by remembering this case
1473    and doing the check at the end; a flag specifies which mode we are running in.
1474    
1475  Arguments:  Arguments:
1476    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1477    options  the compiling options    options  the compiling options
1478      atend    TRUE if called when the pattern is complete
1479      cd       the "compile data" structure
1480    
1481  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1482                 or -1 if there is no fixed length,
1483               or -2 if \C was encountered               or -2 if \C was encountered
1484                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1485  */  */
1486    
1487  static int  static int
1488  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1489  {  {
1490  int length = -1;  int length = -1;
1491    
# Line 1041  branch, check the length against that of Line 1498  branch, check the length against that of
1498  for (;;)  for (;;)
1499    {    {
1500    int d;    int d;
1501      uschar *ce, *cs;
1502    register int op = *cc;    register int op = *cc;
   
1503    switch (op)    switch (op)
1504      {      {
1505      case OP_CBRA:      case OP_CBRA:
1506      case OP_BRA:      case OP_BRA:
1507      case OP_ONCE:      case OP_ONCE:
1508      case OP_COND:      case OP_COND:
1509      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1510      if (d < 0) return d;      if (d < 0) return d;
1511      branchlength += d;      branchlength += d;
1512      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1072  for (;;) Line 1529  for (;;)
1529      branchlength = 0;      branchlength = 0;
1530      break;      break;
1531    
1532        /* A true recursion implies not fixed length, but a subroutine call may
1533        be OK. If the subroutine is a forward reference, we can't deal with
1534        it until the end of the pattern, so return -3. */
1535    
1536        case OP_RECURSE:
1537        if (!atend) return -3;
1538        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1539        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1540        if (cc > cs && cc < ce) return -1;                /* Recursion */
1541        d = find_fixedlength(cs + 2, options, atend, cd);
1542        if (d < 0) return d;
1543        branchlength += d;
1544        cc += 1 + LINK_SIZE;
1545        break;
1546    
1547      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1548    
1549      case OP_ASSERT:      case OP_ASSERT:
# Line 1085  for (;;) Line 1557  for (;;)
1557    
1558      case OP_REVERSE:      case OP_REVERSE:
1559      case OP_CREF:      case OP_CREF:
1560        case OP_NCREF:
1561      case OP_RREF:      case OP_RREF:
1562        case OP_NRREF:
1563      case OP_DEF:      case OP_DEF:
1564      case OP_OPT:      case OP_OPT:
1565      case OP_CALLOUT:      case OP_CALLOUT:
1566      case OP_SOD:      case OP_SOD:
1567      case OP_SOM:      case OP_SOM:
1568        case OP_SET_SOM:
1569      case OP_EOD:      case OP_EOD:
1570      case OP_EODN:      case OP_EODN:
1571      case OP_CIRC:      case OP_CIRC:
# Line 1108  for (;;) Line 1583  for (;;)
1583      branchlength++;      branchlength++;
1584      cc += 2;      cc += 2;
1585  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1586      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1587        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1588  #endif  #endif
1589      break;      break;
1590    
# Line 1122  for (;;) Line 1595  for (;;)
1595      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1596      cc += 4;      cc += 4;
1597  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1598      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1599        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1600  #endif  #endif
1601      break;      break;
1602    
1603      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1604      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1605        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1606      cc += 4;      cc += 4;
1607      break;      break;
1608    
# Line 1148  for (;;) Line 1620  for (;;)
1620      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1621      case OP_WORDCHAR:      case OP_WORDCHAR:
1622      case OP_ANY:      case OP_ANY:
1623        case OP_ALLANY:
1624      branchlength++;      branchlength++;
1625      cc++;      cc++;
1626      break;      break;
# Line 1202  for (;;) Line 1675  for (;;)
1675    
1676    
1677  /*************************************************  /*************************************************
1678  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1679  *************************************************/  *************************************************/
1680    
1681  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1682  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1683    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1684    so that it can be called from pcre_study() when finding the minimum matching
1685    length.
1686    
1687  Arguments:  Arguments:
1688    code        points to start of expression    code        points to start of expression
1689    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1690    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1691    
1692  Returns:      pointer to the opcode for the bracket, or NULL if not found  Returns:      pointer to the opcode for the bracket, or NULL if not found
1693  */  */
1694    
1695  static const uschar *  const uschar *
1696  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1697  {  {
1698  for (;;)  for (;;)
1699    {    {
# Line 1230  for (;;) Line 1706  for (;;)
1706    
1707    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1708    
1709      /* Handle recursion */
1710    
1711      else if (c == OP_REVERSE)
1712        {
1713        if (number < 0) return (uschar *)code;
1714        code += _pcre_OP_lengths[c];
1715        }
1716    
1717    /* Handle capturing bracket */    /* Handle capturing bracket */
1718    
1719    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1239  for (;;) Line 1723  for (;;)
1723      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1724      }      }
1725    
1726    /* In UTF-8 mode, opcodes that are followed by a character may be followed by    /* Otherwise, we can get the item's length from the table, except that for
1727    a multi-byte character. The length in the table is a minimum, so we have to    repeated character types, we have to test for \p and \P, which have an extra
1728    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1729      must add in its length. */
1730    
1731    else    else
1732      {      {
1733        switch(c)
1734          {
1735          case OP_TYPESTAR:
1736          case OP_TYPEMINSTAR:
1737          case OP_TYPEPLUS:
1738          case OP_TYPEMINPLUS:
1739          case OP_TYPEQUERY:
1740          case OP_TYPEMINQUERY:
1741          case OP_TYPEPOSSTAR:
1742          case OP_TYPEPOSPLUS:
1743          case OP_TYPEPOSQUERY:
1744          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1745          break;
1746    
1747          case OP_TYPEUPTO:
1748          case OP_TYPEMINUPTO:
1749          case OP_TYPEEXACT:
1750          case OP_TYPEPOSUPTO:
1751          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1752          break;
1753    
1754          case OP_MARK:
1755          case OP_PRUNE_ARG:
1756          case OP_SKIP_ARG:
1757          code += code[1];
1758          break;
1759    
1760          case OP_THEN_ARG:
1761          code += code[1+LINK_SIZE];
1762          break;
1763          }
1764    
1765        /* Add in the fixed length from the table */
1766    
1767      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1768    
1769      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1770      a multi-byte character. The length in the table is a minimum, so we have to
1771      arrange to skip the extra bytes. */
1772    
1773  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1774      if (utf8) switch(c)      if (utf8) switch(c)
1775        {        {
# Line 1267  for (;;) Line 1791  for (;;)
1791        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1792        break;        break;
1793        }        }
1794    #else
1795        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1796  #endif  #endif
1797      }      }
1798    }    }
# Line 1303  for (;;) Line 1829  for (;;)
1829    
1830    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1831    
1832    /* Otherwise, we get the item's length from the table. In UTF-8 mode, opcodes    /* Otherwise, we can get the item's length from the table, except that for
1833    that are followed by a character may be followed by a multi-byte character.    repeated character types, we have to test for \p and \P, which have an extra
1834    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1835    bytes. */    must add in its length. */
1836    
1837    else    else
1838      {      {
1839        switch(c)
1840          {
1841          case OP_TYPESTAR:
1842          case OP_TYPEMINSTAR:
1843          case OP_TYPEPLUS:
1844          case OP_TYPEMINPLUS:
1845          case OP_TYPEQUERY:
1846          case OP_TYPEMINQUERY:
1847          case OP_TYPEPOSSTAR:
1848          case OP_TYPEPOSPLUS:
1849          case OP_TYPEPOSQUERY:
1850          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1851          break;
1852    
1853          case OP_TYPEPOSUPTO:
1854          case OP_TYPEUPTO:
1855          case OP_TYPEMINUPTO:
1856          case OP_TYPEEXACT:
1857          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1858          break;
1859    
1860          case OP_MARK:
1861          case OP_PRUNE_ARG:
1862          case OP_SKIP_ARG:
1863          code += code[1];
1864          break;
1865    
1866          case OP_THEN_ARG:
1867          code += code[1+LINK_SIZE];
1868          break;
1869          }
1870    
1871        /* Add in the fixed length from the table */
1872    
1873      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1874    
1875        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1876        by a multi-byte character. The length in the table is a minimum, so we have
1877        to arrange to skip the extra bytes. */
1878    
1879  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1880      if (utf8) switch(c)      if (utf8) switch(c)
1881        {        {
# Line 1332  for (;;) Line 1897  for (;;)
1897        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1898        break;        break;
1899        }        }
1900    #else
1901        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1902  #endif  #endif
1903      }      }
1904    }    }
# Line 1347  for (;;) Line 1914  for (;;)
1914  can match the empty string or not. It is called from could_be_empty()  can match the empty string or not. It is called from could_be_empty()
1915  below and from compile_branch() when checking for an unlimited repeat of a  below and from compile_branch() when checking for an unlimited repeat of a
1916  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1917  assertions. If we hit an unclosed bracket, we return "empty" - this means we've  backward and negative forward assertions when its final argument is TRUE. If we
1918  struck an inner bracket whose current branch will already have been scanned.  hit an unclosed bracket, we return "empty" - this means we've struck an inner
1919    bracket whose current branch will already have been scanned.
1920    
1921  Arguments:  Arguments:
1922    code        points to start of search    code        points to start of search
1923    endcode     points to where to stop    endcode     points to where to stop
1924    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1925      cd          contains pointers to tables etc.
1926    
1927  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1928  */  */
1929    
1930  static BOOL  static BOOL
1931  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1932      compile_data *cd)
1933  {  {
1934  register int c;  register int c;
1935  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1370  for (code = first_significant_code(code Line 1940  for (code = first_significant_code(code
1940    
1941    c = *code;    c = *code;
1942    
1943    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1944      first_significant_code() with a TRUE final argument. */
1945    
1946      if (c == OP_ASSERT)
1947      {      {
1948      BOOL empty_branch;      do code += GET(code, 1); while (*code == OP_ALT);
1949      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      c = *code;
1950        continue;
1951        }
1952    
1953      /* Groups with zero repeats can of course be empty; skip them. */
1954    
1955      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1956        {
1957        code += _pcre_OP_lengths[c];
1958        do code += GET(code, 1); while (*code == OP_ALT);
1959        c = *code;
1960        continue;
1961        }
1962    
1963      /* Scan a closed bracket */    /* For a recursion/subroutine call, if its end has been reached, which
1964      implies a subroutine call, we can scan it. */
1965    
1966      empty_branch = FALSE;    if (c == OP_RECURSE)
1967        {
1968        BOOL empty_branch = FALSE;
1969        const uschar *scode = cd->start_code + GET(code, 1);
1970        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1971      do      do
1972        {        {
1973        if (!empty_branch && could_be_empty_branch(code, endcode, utf8))        if (could_be_empty_branch(scode, endcode, utf8, cd))
1974            {
1975          empty_branch = TRUE;          empty_branch = TRUE;
1976        code += GET(code, 1);          break;
1977            }
1978          scode += GET(scode, 1);
1979        }        }
1980      while (*code == OP_ALT);      while (*scode == OP_ALT);
1981      if (!empty_branch) return FALSE;   /* All branches are non-empty */      if (!empty_branch) return FALSE;  /* All branches are non-empty */
1982        continue;
1983        }
1984    
1985      /* Move past the KET and fudge things so that the increment in the "for"    /* For other groups, scan the branches. */
     above has no effect. */  
1986    
1987      c = OP_END;    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1988      code += 1 + LINK_SIZE - _pcre_OP_lengths[c];      {
1989        BOOL empty_branch;
1990        if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1991    
1992        /* If a conditional group has only one branch, there is a second, implied,
1993        empty branch, so just skip over the conditional, because it could be empty.
1994        Otherwise, scan the individual branches of the group. */
1995    
1996        if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
1997          code += GET(code, 1);
1998        else
1999          {
2000          empty_branch = FALSE;
2001          do
2002            {
2003            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
2004              empty_branch = TRUE;
2005            code += GET(code, 1);
2006            }
2007          while (*code == OP_ALT);
2008          if (!empty_branch) return FALSE;   /* All branches are non-empty */
2009          }
2010    
2011        c = *code;
2012      continue;      continue;
2013      }      }
2014    
# Line 1399  for (code = first_significant_code(code Line 2016  for (code = first_significant_code(code
2016    
2017    switch (c)    switch (c)
2018      {      {
2019      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2020        cannot be represented just by a bit map. This includes negated single
2021        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2022        actual length is stored in the compiled code, so we must update "code"
2023        here. */
2024    
2025  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2026      case OP_XCLASS:      case OP_XCLASS:
2027      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2028      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2029  #endif  #endif
2030    
# Line 1447  for (code = first_significant_code(code Line 2068  for (code = first_significant_code(code
2068      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2069      case OP_WORDCHAR:      case OP_WORDCHAR:
2070      case OP_ANY:      case OP_ANY:
2071        case OP_ALLANY:
2072      case OP_ANYBYTE:      case OP_ANYBYTE:
2073      case OP_CHAR:      case OP_CHAR:
2074      case OP_CHARNC:      case OP_CHARNC:
# Line 1465  for (code = first_significant_code(code Line 2087  for (code = first_significant_code(code
2087      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2088      return FALSE;      return FALSE;
2089    
2090        /* These are going to continue, as they may be empty, but we have to
2091        fudge the length for the \p and \P cases. */
2092    
2093        case OP_TYPESTAR:
2094        case OP_TYPEMINSTAR:
2095        case OP_TYPEPOSSTAR:
2096        case OP_TYPEQUERY:
2097        case OP_TYPEMINQUERY:
2098        case OP_TYPEPOSQUERY:
2099        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2100        break;
2101    
2102        /* Same for these */
2103    
2104        case OP_TYPEUPTO:
2105        case OP_TYPEMINUPTO:
2106        case OP_TYPEPOSUPTO:
2107        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2108        break;
2109    
2110      /* End of branch */      /* End of branch */
2111    
2112      case OP_KET:      case OP_KET:
# Line 1483  for (code = first_significant_code(code Line 2125  for (code = first_significant_code(code
2125      case OP_QUERY:      case OP_QUERY:
2126      case OP_MINQUERY:      case OP_MINQUERY:
2127      case OP_POSQUERY:      case OP_POSQUERY:
2128        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2129        break;
2130    
2131      case OP_UPTO:      case OP_UPTO:
2132      case OP_MINUPTO:      case OP_MINUPTO:
2133      case OP_POSUPTO:      case OP_POSUPTO:
2134      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2135      break;      break;
2136  #endif  #endif
2137    
2138        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2139        string. */
2140    
2141        case OP_MARK:
2142        case OP_PRUNE_ARG:
2143        case OP_SKIP_ARG:
2144        code += code[1];
2145        break;
2146    
2147        case OP_THEN_ARG:
2148        code += code[1+LINK_SIZE];
2149        break;
2150    
2151        /* None of the remaining opcodes are required to match a character. */
2152    
2153        default:
2154        break;
2155      }      }
2156    }    }
2157    
# Line 1511  Arguments: Line 2174  Arguments:
2174    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2175    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2176    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2177      cd          pointers to tables etc
2178    
2179  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2180  */  */
2181    
2182  static BOOL  static BOOL
2183  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2184    BOOL utf8)    BOOL utf8, compile_data *cd)
2185  {  {
2186  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2187    {    {
2188    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2189        return FALSE;
2190    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2191    }    }
2192  return TRUE;  return TRUE;
# Line 1534  return TRUE; Line 2199  return TRUE;
2199  *************************************************/  *************************************************/
2200    
2201  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2202  encountered in a character class. It checks whether this is followed by an  encountered in a character class. It checks whether this is followed by a
2203  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2204  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2205    
2206    Originally, this function only recognized a sequence of letters between the
2207    terminators, but it seems that Perl recognizes any sequence of characters,
2208    though of course unknown POSIX names are subsequently rejected. Perl gives an
2209    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2210    didn't consider this to be a POSIX class. Likewise for [:1234:].
2211    
2212    The problem in trying to be exactly like Perl is in the handling of escapes. We
2213    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2214    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2215    below handles the special case of \], but does not try to do any other escape
2216    processing. This makes it different from Perl for cases such as [:l\ower:]
2217    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2218    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2219    I think.
2220    
2221  Argument:  Arguments:
2222    ptr      pointer to the initial [    ptr      pointer to the initial [
2223    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2224    
2225  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2226  */  */
2227    
2228  static BOOL  static BOOL
2229  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2230  {  {
2231  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2232  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2233  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2234    {    {
2235    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2236    return TRUE;      {
2237        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2238        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2239          {
2240          *endptr = ptr;
2241          return TRUE;
2242          }
2243        }
2244    }    }
2245  return FALSE;  return FALSE;
2246  }  }
# Line 1581  Returns: a value representing the na Line 2265  Returns: a value representing the na
2265  static int  static int
2266  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2267  {  {
2268    const char *pn = posix_names;
2269  register int yield = 0;  register int yield = 0;
2270  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2271    {    {
2272    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2273      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2274      pn += posix_name_lengths[yield] + 1;
2275    yield++;    yield++;
2276    }    }
2277  return -1;  return -1;
# Line 1600  return -1; Line 2286  return -1;
2286  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2287  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2288  earlier groups that are outside the current group). However, when a group is  earlier groups that are outside the current group). However, when a group is
2289  optional (i.e. the minimum quantifier is zero), OP_BRAZERO is inserted before  optional (i.e. the minimum quantifier is zero), OP_BRAZERO or OP_SKIPZERO is
2290  it, after it has been compiled. This means that any OP_RECURSE items within it  inserted before it, after it has been compiled. This means that any OP_RECURSE
2291  that refer to the group itself or any contained groups have to have their  items within it that refer to the group itself or any contained groups have to
2292  offsets adjusted. That one of the jobs of this function. Before it is called,  have their offsets adjusted. That one of the jobs of this function. Before it
2293  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2294    OP_END.
2295    
2296  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2297  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1627  adjust_recurse(uschar *group, int adjust Line 2314  adjust_recurse(uschar *group, int adjust
2314    uschar *save_hwm)    uschar *save_hwm)
2315  {  {
2316  uschar *ptr = group;  uschar *ptr = group;
2317    
2318  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2319    {    {
2320    int offset;    int offset;
# Line 1680  auto_callout(uschar *code, const uschar Line 2368  auto_callout(uschar *code, const uschar
2368  {  {
2369  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2370  *code++ = 255;  *code++ = 255;
2371  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2372  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2373  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2374  }  }
2375    
# Line 1706  Returns: nothing Line 2394  Returns: nothing
2394  static void  static void
2395  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2396  {  {
2397  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2398  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2399  }  }
2400    
# Line 1738  get_othercase_range(unsigned int *cptr, Line 2426  get_othercase_range(unsigned int *cptr,
2426  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2427    
2428  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2429    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2430    
2431  if (c > d) return FALSE;  if (c > d) return FALSE;
2432    
# Line 1747  next = othercase + 1; Line 2435  next = othercase + 1;
2435    
2436  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2437    {    {
2438    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2439    next++;    next++;
2440    }    }
2441    
# Line 1756  for (++c; c <= d; c++) Line 2444  for (++c; c <= d; c++)
2444    
2445  return TRUE;  return TRUE;
2446  }  }
2447    
2448    
2449    
2450    /*************************************************
2451    *        Check a character and a property        *
2452    *************************************************/
2453    
2454    /* This function is called by check_auto_possessive() when a property item
2455    is adjacent to a fixed character.
2456    
2457    Arguments:
2458      c            the character
2459      ptype        the property type
2460      pdata        the data for the type
2461      negated      TRUE if it's a negated property (\P or \p{^)
2462    
2463    Returns:       TRUE if auto-possessifying is OK
2464    */
2465    
2466    static BOOL
2467    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2468    {
2469    const ucd_record *prop = GET_UCD(c);
2470    switch(ptype)
2471      {
2472      case PT_LAMP:
2473      return (prop->chartype == ucp_Lu ||
2474              prop->chartype == ucp_Ll ||
2475              prop->chartype == ucp_Lt) == negated;
2476    
2477      case PT_GC:
2478      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2479    
2480      case PT_PC:
2481      return (pdata == prop->chartype) == negated;
2482    
2483      case PT_SC:
2484      return (pdata == prop->script) == negated;
2485    
2486      /* These are specials */
2487    
2488      case PT_ALNUM:
2489      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2490              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2491    
2492      case PT_SPACE:    /* Perl space */
2493      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2494              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2495              == negated;
2496    
2497      case PT_PXSPACE:  /* POSIX space */
2498      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2499              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2500              c == CHAR_FF || c == CHAR_CR)
2501              == negated;
2502    
2503      case PT_WORD:
2504      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2505              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2506              c == CHAR_UNDERSCORE) == negated;
2507      }
2508    return FALSE;
2509    }
2510  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2511    
2512    
# Line 1769  whether the next thing could possibly ma Line 2520  whether the next thing could possibly ma
2520  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2521    
2522  Arguments:  Arguments:
2523    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2524    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2525    ptr           next character in pattern    ptr           next character in pattern
2526    options       options bits    options       options bits
2527    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1781  Returns: TRUE if possessifying is Line 2530  Returns: TRUE if possessifying is
2530  */  */
2531    
2532  static BOOL  static BOOL
2533  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2534    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2535  {  {
2536  int next;  int c, next;
2537    int op_code = *previous++;
2538    
2539  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2540    
# Line 1793  if ((options & PCRE_EXTENDED) != 0) Line 2543  if ((options & PCRE_EXTENDED) != 0)
2543    for (;;)    for (;;)
2544      {      {
2545      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2546      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2547        {        {
2548        while (*(++ptr) != 0)        ptr++;
2549          while (*ptr != 0)
2550            {
2551          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2552            ptr++;
2553    #ifdef SUPPORT_UTF8
2554            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2555    #endif
2556            }
2557        }        }
2558      else break;      else break;
2559      }      }
# Line 1805  if ((options & PCRE_EXTENDED) != 0) Line 2562  if ((options & PCRE_EXTENDED) != 0)
2562  /* If the next item is one that we can handle, get its value. A non-negative  /* If the next item is one that we can handle, get its value. A non-negative
2563  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2564    
2565  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2566    {    {
2567    int temperrorcode = 0;    int temperrorcode = 0;
2568    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1830  if ((options & PCRE_EXTENDED) != 0) Line 2587  if ((options & PCRE_EXTENDED) != 0)
2587    for (;;)    for (;;)
2588      {      {
2589      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2590      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2591        {        {
2592        while (*(++ptr) != 0)        ptr++;
2593          while (*ptr != 0)
2594            {
2595          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2596            ptr++;
2597    #ifdef SUPPORT_UTF8
2598            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2599    #endif
2600            }
2601        }        }
2602      else break;      else break;
2603      }      }
# Line 1841  if ((options & PCRE_EXTENDED) != 0) Line 2605  if ((options & PCRE_EXTENDED) != 0)
2605    
2606  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2607    
2608  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2609    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2610        return FALSE;
 /* Now compare the next item with the previous opcode. If the previous is a  
 positive single character match, "item" either contains the character or, if  
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
2611    
2612  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2613    the next item is a character. */
2614    
2615  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2616    {    {
2617    case OP_CHAR:    case OP_CHAR:
2618  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2619    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2620    #else
2621      c = *previous;
2622  #endif  #endif
2623    return item != next;    return c != next;
2624    
2625    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARNC (caseless character) we must check the other case. If we have
2626    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
# Line 1866  if (next >= 0) switch(op_code) Line 2628  if (next >= 0) switch(op_code)
2628    
2629    case OP_CHARNC:    case OP_CHARNC:
2630  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2631    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2632    #else
2633      c = *previous;
2634  #endif  #endif
2635    if (item == next) return FALSE;    if (c == next) return FALSE;
2636  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2637    if (utf8)    if (utf8)
2638      {      {
2639      unsigned int othercase;      unsigned int othercase;
2640      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2641  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2642      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2643  #else  #else
2644      othercase = NOTACHAR;      othercase = NOTACHAR;
2645  #endif  #endif
2646      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2647      }      }
2648    else    else
2649  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2650    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2651    
2652    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2653    
2654    case OP_NOT:    case OP_NOT:
2655    if (next < 0) return FALSE;  /* Not a character */    if ((c = *previous) == next) return TRUE;
   if (item == next) return TRUE;  
2656    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2657  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2658    if (utf8)    if (utf8)
# Line 1897  if (next >= 0) switch(op_code) Line 2660  if (next >= 0) switch(op_code)
2660      unsigned int othercase;      unsigned int othercase;
2661      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2662  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2663      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2664  #else  #else
2665      othercase = NOTACHAR;      othercase = NOTACHAR;
2666  #endif  #endif
2667      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2668      }      }
2669    else    else
2670  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2671    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2672    
2673      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2674      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2675    
2676    case OP_DIGIT:    case OP_DIGIT:
2677    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 1925  if (next >= 0) switch(op_code) Line 2691  if (next >= 0) switch(op_code)
2691    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2692    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2693    
2694      case OP_HSPACE:
2695      case OP_NOT_HSPACE:
2696      switch(next)
2697        {
2698        case 0x09:
2699        case 0x20:
2700        case 0xa0:
2701        case 0x1680:
2702        case 0x180e:
2703        case 0x2000:
2704        case 0x2001:
2705        case 0x2002:
2706        case 0x2003:
2707        case 0x2004:
2708        case 0x2005:
2709        case 0x2006:
2710        case 0x2007:
2711        case 0x2008:
2712        case 0x2009:
2713        case 0x200A:
2714        case 0x202f:
2715        case 0x205f:
2716        case 0x3000:
2717        return op_code == OP_NOT_HSPACE;
2718        default:
2719        return op_code != OP_NOT_HSPACE;
2720        }
2721    
2722      case OP_ANYNL:
2723      case OP_VSPACE:
2724      case OP_NOT_VSPACE:
2725      switch(next)
2726        {
2727        case 0x0a:
2728        case 0x0b:
2729        case 0x0c:
2730        case 0x0d:
2731        case 0x85:
2732        case 0x2028:
2733        case 0x2029:
2734        return op_code == OP_NOT_VSPACE;
2735        default:
2736        return op_code != OP_NOT_VSPACE;
2737        }
2738    
2739    #ifdef SUPPORT_UCP
2740      case OP_PROP:
2741      return check_char_prop(next, previous[0], previous[1], FALSE);
2742    
2743      case OP_NOTPROP:
2744      return check_char_prop(next, previous[0], previous[1], TRUE);
2745    #endif
2746    
2747    default:    default:
2748    return FALSE;    return FALSE;
2749    }    }
2750    
2751    
2752  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2753    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2754    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2755    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2756    replaced by OP_PROP codes when PCRE_UCP is set. */
2757    
2758  switch(op_code)  switch(op_code)
2759    {    {
2760    case OP_CHAR:    case OP_CHAR:
2761    case OP_CHARNC:    case OP_CHARNC:
2762  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2763    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2764    #else
2765      c = *previous;
2766  #endif  #endif
2767    switch(-next)    switch(-next)
2768      {      {
2769      case ESC_d:      case ESC_d:
2770      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2771    
2772      case ESC_D:      case ESC_D:
2773      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2774    
2775      case ESC_s:      case ESC_s:
2776      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2777    
2778      case ESC_S:      case ESC_S:
2779      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2780    
2781      case ESC_w:      case ESC_w:
2782      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2783    
2784      case ESC_W:      case ESC_W:
2785      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2786    
2787        case ESC_h:
2788        case ESC_H:
2789        switch(c)
2790          {
2791          case 0x09:
2792          case 0x20:
2793          case 0xa0:
2794          case 0x1680:
2795          case 0x180e:
2796          case 0x2000:
2797          case 0x2001:
2798          case 0x2002:
2799          case 0x2003:
2800          case 0x2004:
2801          case 0x2005:
2802          case 0x2006:
2803          case 0x2007:
2804          case 0x2008:
2805          case 0x2009:
2806          case 0x200A:
2807          case 0x202f:
2808          case 0x205f:
2809          case 0x3000:
2810          return -next != ESC_h;
2811          default:
2812          return -next == ESC_h;
2813          }
2814    
2815        case ESC_v:
2816        case ESC_V:
2817        switch(c)
2818          {
2819          case 0x0a:
2820          case 0x0b:
2821          case 0x0c:
2822          case 0x0d:
2823          case 0x85:
2824          case 0x2028:
2825          case 0x2029:
2826          return -next != ESC_v;
2827          default:
2828          return -next == ESC_v;
2829          }
2830    
2831        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2832        their substitutions and process them. The result will always be either
2833        -ESC_p or -ESC_P. Then fall through to process those values. */
2834    
2835    #ifdef SUPPORT_UCP
2836        case ESC_du:
2837        case ESC_DU:
2838        case ESC_wu:
2839        case ESC_WU:
2840        case ESC_su:
2841        case ESC_SU:
2842          {
2843          int temperrorcode = 0;
2844          ptr = substitutes[-next - ESC_DU];
2845          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2846          if (temperrorcode != 0) return FALSE;
2847          ptr++;    /* For compatibility */
2848          }
2849        /* Fall through */
2850    
2851        case ESC_p:
2852        case ESC_P:
2853          {
2854          int ptype, pdata, errorcodeptr;
2855          BOOL negated;
2856    
2857          ptr--;      /* Make ptr point at the p or P */
2858          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2859          if (ptype < 0) return FALSE;
2860          ptr++;      /* Point past the final curly ket */
2861    
2862          /* If the property item is optional, we have to give up. (When generated
2863          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2864          to the original \d etc. At this point, ptr will point to a zero byte. */
2865    
2866          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2867            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2868              return FALSE;
2869    
2870          /* Do the property check. */
2871    
2872          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2873          }
2874    #endif
2875    
2876      default:      default:
2877      return FALSE;      return FALSE;
2878      }      }
2879    
2880      /* In principle, support for Unicode properties should be integrated here as
2881      well. It means re-organizing the above code so as to get hold of the property
2882      values before switching on the op-code. However, I wonder how many patterns
2883      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2884      these op-codes are never generated.) */
2885    
2886    case OP_DIGIT:    case OP_DIGIT:
2887    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2888             next == -ESC_h || next == -ESC_v || next == -ESC_R;
2889    
2890    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2891    return next == -ESC_d;    return next == -ESC_d;
2892    
2893    case OP_WHITESPACE:    case OP_WHITESPACE:
2894    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2895    
2896    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2897    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2898    
2899      case OP_HSPACE:
2900      return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2901             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2902    
2903      case OP_NOT_HSPACE:
2904      return next == -ESC_h;
2905    
2906      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2907      case OP_ANYNL:
2908      case OP_VSPACE:
2909      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2910    
2911      case OP_NOT_VSPACE:
2912      return next == -ESC_v || next == -ESC_R;
2913    
2914    case OP_WORDCHAR:    case OP_WORDCHAR:
2915    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2916             next == -ESC_v || next == -ESC_R;
2917    
2918    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2919    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2040  BOOL inescq = FALSE; Line 2977  BOOL inescq = FALSE;
2977  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2978  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2979  const uschar *tempptr;  const uschar *tempptr;
2980    const uschar *nestptr = NULL;
2981  uschar *previous = NULL;  uschar *previous = NULL;
2982  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2983  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2049  uschar classbits[32]; Line 2987  uschar classbits[32];
2987  BOOL class_utf8;  BOOL class_utf8;
2988  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2989  uschar *class_utf8data;  uschar *class_utf8data;
2990    uschar *class_utf8data_base;
2991  uschar utf8_char[6];  uschar utf8_char[6];
2992  #else  #else
2993  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2994  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2995  #endif  #endif
2996    
2997  #ifdef DEBUG  #ifdef PCRE_DEBUG
2998  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2999  #endif  #endif
3000    
# Line 2088  req_caseopt = ((options & PCRE_CASELESS) Line 3027  req_caseopt = ((options & PCRE_CASELESS)
3027  for (;; ptr++)  for (;; ptr++)
3028    {    {
3029    BOOL negate_class;    BOOL negate_class;
3030      BOOL should_flip_negation;
3031    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3032    BOOL is_quantifier;    BOOL is_quantifier;
3033    BOOL is_recurse;    BOOL is_recurse;
3034      BOOL reset_bracount;
3035    int class_charcount;    int class_charcount;
3036    int class_lastchar;    int class_lastchar;
3037    int newoptions;    int newoptions;
3038    int recno;    int recno;
3039      int refsign;
3040    int skipbytes;    int skipbytes;
3041    int subreqbyte;    int subreqbyte;
3042    int subfirstbyte;    int subfirstbyte;
# Line 2106  for (;; ptr++) Line 3048  for (;; ptr++)
3048    
3049    c = *ptr;    c = *ptr;
3050    
3051      /* If we are at the end of a nested substitution, revert to the outer level
3052      string. Nesting only happens one level deep. */
3053    
3054      if (c == 0 && nestptr != NULL)
3055        {
3056        ptr = nestptr;
3057        nestptr = NULL;
3058        c = *ptr;
3059        }
3060    
3061    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3062    previous cycle of this loop. */    previous cycle of this loop. */
3063    
3064    if (lengthptr != NULL)    if (lengthptr != NULL)
3065      {      {
3066  #ifdef DEBUG  #ifdef PCRE_DEBUG
3067      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3068  #endif  #endif
3069      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3070        {        {
3071        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3072        goto FAILED;        goto FAILED;
# Line 2127  for (;; ptr++) Line 3079  for (;; ptr++)
3079      */      */
3080    
3081      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
3082      *lengthptr += code - last_code;  
3083        /* Paranoid check for integer overflow */
3084    
3085        if (OFLOW_MAX - *lengthptr < code - last_code)
3086          {
3087          *errorcodeptr = ERR20;
3088          goto FAILED;
3089          }
3090    
3091        *lengthptr += (int)(code - last_code);
3092      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));      DPRINTF(("length=%d added %d c=%c\n", *lengthptr, code - last_code, c));
3093    
3094      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
# Line 2154  for (;; ptr++) Line 3115  for (;; ptr++)
3115    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3116    reference list. */    reference list. */
3117    
3118    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3119      {      {
3120      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3121      goto FAILED;      goto FAILED;
# Line 2164  for (;; ptr++) Line 3125  for (;; ptr++)
3125    
3126    if (inescq && c != 0)    if (inescq && c != 0)
3127      {      {
3128      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3129        {        {
3130        inescq = FALSE;        inescq = FALSE;
3131        ptr++;        ptr++;
# Line 2190  for (;; ptr++) Line 3151  for (;; ptr++)
3151    /* Fill in length of a previous callout, except when the next thing is    /* Fill in length of a previous callout, except when the next thing is
3152    a quantifier. */    a quantifier. */
3153    
3154    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3155      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3156        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3157    
3158    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3159         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2206  for (;; ptr++) Line 3168  for (;; ptr++)
3168    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3169      {      {
3170      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3171      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3172        {        {
3173        while (*(++ptr) != 0)        ptr++;
3174          while (*ptr != 0)
3175          {          {
3176          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3177            ptr++;
3178    #ifdef SUPPORT_UTF8
3179            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3180    #endif
3181          }          }
3182        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3183    
# Line 2231  for (;; ptr++) Line 3198  for (;; ptr++)
3198      {      {
3199      /* ===================================================================*/      /* ===================================================================*/
3200      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3201      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3202      case ')':      case CHAR_RIGHT_PARENTHESIS:
3203      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3204      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3205      *codeptr = code;      *codeptr = code;
3206      *ptrptr = ptr;      *ptrptr = ptr;
3207      if (lengthptr != NULL)      if (lengthptr != NULL)
3208        {        {
3209        *lengthptr += code - last_code;   /* To include callout length */        if (OFLOW_MAX - *lengthptr < code - last_code)
3210            {
3211            *errorcodeptr = ERR20;
3212            goto FAILED;
3213            }
3214          *lengthptr += (int)(code - last_code);   /* To include callout length */
3215        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3216        }        }
3217      return TRUE;      return TRUE;
# Line 2249  for (;; ptr++) Line 3221  for (;; ptr++)
3221      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3222      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3223    
3224      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3225      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3226        {        {
3227        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2258  for (;; ptr++) Line 3230  for (;; ptr++)
3230      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3231      break;      break;
3232    
3233      case '$':      case CHAR_DOLLAR_SIGN:
3234      previous = NULL;      previous = NULL;
3235      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3236      break;      break;
# Line 2266  for (;; ptr++) Line 3238  for (;; ptr++)
3238      /* There can never be a first char if '.' is first, whatever happens about      /* There can never be a first char if '.' is first, whatever happens about
3239      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3240    
3241      case '.':      case CHAR_DOT:
3242      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3243      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3244      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3245      previous = code;      previous = code;
3246      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3247      break;      break;
3248    
3249    
# Line 2286  for (;; ptr++) Line 3258  for (;; ptr++)
3258      opcode is compiled. It may optionally have a bit map for characters < 256,      opcode is compiled. It may optionally have a bit map for characters < 256,
3259      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3260      whether the bitmap is present, and whether this is a negated class or not.      whether the bitmap is present, and whether this is a negated class or not.
     */  
3261    
3262      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3263        default (Perl) mode, it is treated as a data character. */
3264    
3265        case CHAR_RIGHT_SQUARE_BRACKET:
3266        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3267          {
3268          *errorcodeptr = ERR64;
3269          goto FAILED;
3270          }
3271        goto NORMAL_CHAR;
3272    
3273        case CHAR_LEFT_SQUARE_BRACKET:
3274      previous = code;      previous = code;
3275    
3276      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
3277      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
3278    
3279      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3280          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3281            check_posix_syntax(ptr, &tempptr))
3282        {        {
3283        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3284        goto FAILED;        goto FAILED;
3285        }        }
3286    
3287      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
3288        if the first few characters (either before or after ^) are \Q\E or \E we
3289        skip them too. This makes for compatibility with Perl. */
3290    
3291      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3292        for (;;)
3293        {        {
       negate_class = TRUE;  
3294        c = *(++ptr);        c = *(++ptr);
3295          if (c == CHAR_BACKSLASH)
3296            {
3297            if (ptr[1] == CHAR_E)
3298              ptr++;
3299            else if (strncmp((const char *)ptr+1,
3300                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3301              ptr += 3;
3302            else
3303              break;
3304            }
3305          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3306            negate_class = TRUE;
3307          else break;
3308        }        }
3309      else  
3310        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3311        an initial ']' is taken as a data character -- the code below handles
3312        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3313        [^] must match any character, so generate OP_ALLANY. */
3314    
3315        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3316            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3317        {        {
3318        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3319          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3320          zerofirstbyte = firstbyte;
3321          break;
3322        }        }
3323    
3324        /* If a class contains a negative special such as \S, we need to flip the
3325        negation flag at the end, so that support for characters > 255 works
3326        correctly (they are all included in the class). */
3327    
3328        should_flip_negation = FALSE;
3329    
3330      /* Keep a count of chars with values < 256 so that we can optimize the case      /* Keep a count of chars with values < 256 so that we can optimize the case
3331      of just a single character (as long as it's < 256). However, For higher      of just a single character (as long as it's < 256). However, For higher
3332      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2330  for (;; ptr++) Line 3344  for (;; ptr++)
3344  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3345      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3346      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3347        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3348  #endif  #endif
3349    
3350      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2345  for (;; ptr++) Line 3360  for (;; ptr++)
3360          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3361          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3362          }          }
3363    
3364          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3365          data and reset the pointer. This is so that very large classes that
3366          contain a zillion UTF-8 characters no longer overwrite the work space
3367          (which is on the stack). */
3368    
3369          if (lengthptr != NULL)
3370            {
3371            *lengthptr += class_utf8data - class_utf8data_base;
3372            class_utf8data = class_utf8data_base;
3373            }
3374    
3375  #endif  #endif
3376    
3377        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3378    
3379        if (inescq)        if (inescq)
3380          {          {
3381          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3382            {            {
3383            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3384            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2366  for (;; ptr++) Line 3393  for (;; ptr++)
3393        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3394        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3395    
3396        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3397            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3398            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3399          {          {
3400          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3401          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3402          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3403          uschar pbits[32];          uschar pbits[32];
3404    
3405          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3406            {            {
3407            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3408            goto FAILED;            goto FAILED;
3409            }            }
3410    
3411          ptr += 2;          ptr += 2;
3412          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3413            {            {
3414            local_negate = TRUE;            local_negate = TRUE;
3415              should_flip_negation = TRUE;  /* Note negative special */
3416            ptr++;            ptr++;
3417            }            }
3418    
3419          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3420          if (posix_class < 0)          if (posix_class < 0)
3421            {            {
3422            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2402  for (;; ptr++) Line 3430  for (;; ptr++)
3430          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3431            posix_class = 0;            posix_class = 0;
3432    
3433          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3434          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3435          subtract bits that may be in the main map already. At the end we or the  
3436          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3437            if ((options & PCRE_UCP) != 0)
3438              {
3439              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3440              if (posix_substitutes[pc] != NULL)
3441                {
3442                nestptr = tempptr + 1;
3443                ptr = posix_substitutes[pc] - 1;
3444                continue;
3445                }
3446              }
3447    #endif
3448            /* In the non-UCP case, we build the bit map for the POSIX class in a
3449            chunk of local store because we may be adding and subtracting from it,
3450            and we don't want to subtract bits that may be in the main map already.
3451            At the end we or the result into the bit map that is being built. */
3452    
3453          posix_class *= 3;          posix_class *= 3;
3454    
# Line 2449  for (;; ptr++) Line 3492  for (;; ptr++)
3492    
3493        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3494        of the specials, which just set a flag. The sequence \b is a special        of the specials, which just set a flag. The sequence \b is a special
3495        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3496        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3497        to or into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3498        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3499          PCRE_EXTRA is set. */
3500    
3501        if (c == '\\')        if (c == CHAR_BACKSLASH)
3502          {          {
3503          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3504          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3505    
3506          if (-c == ESC_b) c = '\b';       /* \b is backslash in a class */          if (-c == ESC_b) c = CHAR_BS;    /* \b is backspace in a class */
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */  
3507          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3508            {            {
3509            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3510              {              {
3511              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3512              }              }
3513            else inescq = TRUE;            else inescq = TRUE;
3514            continue;            continue;
3515            }            }
3516            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3517    
3518          if (c < 0)          if (c < 0)
3519            {            {
3520            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3521            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3522    
3523            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3524              {              {
3525    #ifdef SUPPORT_UCP
3526                case ESC_du:     /* These are the values given for \d etc */
3527                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3528                case ESC_wu:     /* escape sequence with an appropriate \p */
3529                case ESC_WU:     /* or \P to test Unicode properties instead */
3530                case ESC_su:     /* of the default ASCII testing. */
3531                case ESC_SU:
3532                nestptr = ptr;
3533                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3534                class_charcount -= 2;                /* Undo! */
3535                continue;
3536    #endif
3537              case ESC_d:              case ESC_d:
3538              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3539              continue;              continue;
3540    
3541              case ESC_D:              case ESC_D:
3542                should_flip_negation = TRUE;
3543              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3544              continue;              continue;
3545    
# Line 2494  for (;; ptr++) Line 3548  for (;; ptr++)
3548              continue;              continue;
3549    
3550              case ESC_W:              case ESC_W:
3551                should_flip_negation = TRUE;
3552              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3553              continue;              continue;
3554    
3555                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3556                if it was previously set by something earlier in the character
3557                class. */
3558    
3559              case ESC_s:              case ESC_s:
3560              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3561              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3562                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3563              continue;              continue;
3564    
3565              case ESC_S:              case ESC_S:
3566                should_flip_negation = TRUE;
3567              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3568              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3569              continue;              continue;
3570    
3571              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
3572                SETBIT(classbits, 0x09); /* VT */
3573                SETBIT(classbits, 0x20); /* SPACE */
3574                SETBIT(classbits, 0xa0); /* NSBP */
3575    #ifdef SUPPORT_UTF8
3576                if (utf8)
3577                  {
3578                  class_utf8 = TRUE;
3579                  *class_utf8data++ = XCL_SINGLE;
3580                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3581                  *class_utf8data++ = XCL_SINGLE;
3582                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3583                  *class_utf8data++ = XCL_RANGE;
3584                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3585                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3586                  *class_utf8data++ = XCL_SINGLE;
3587                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3588                  *class_utf8data++ = XCL_SINGLE;
3589                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3590                  *class_utf8data++ = XCL_SINGLE;
3591                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3592                  }
3593    #endif
3594              continue;              continue;
3595    
3596              default:    /* Not recognized; fall through */              case ESC_H:
3597              break;      /* Need "default" setting to stop compiler warning. */              for (c = 0; c < 32; c++)
3598              }                {
3599                  int x = 0xff;
3600                  switch (c)
3601                    {
3602                    case 0x09/8: x ^= 1 << (0x09%8); break;
3603                    case 0x20/8: x ^= 1 << (0x20%8); break;
3604                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3605                    default: break;
3606                    }
3607                  classbits[c] |= x;
3608                  }
3609    
3610            /* In the pre-compile phase, just do the recognition. */  #ifdef SUPPORT_UTF8
3611                if (utf8)
3612                  {
3613                  class_utf8 = TRUE;
3614                  *class_utf8data++ = XCL_RANGE;
3615                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3616                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3617                  *class_utf8data++ = XCL_RANGE;
3618                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3619                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3620                  *class_utf8data++ = XCL_RANGE;
3621                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3622                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3623                  *class_utf8data++ = XCL_RANGE;
3624                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3625                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3626                  *class_utf8data++ = XCL_RANGE;
3627                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3628                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3629                  *class_utf8data++ = XCL_RANGE;
3630                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3631                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3632                  *class_utf8data++ = XCL_RANGE;
3633                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3634                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3635                  }
3636    #endif
3637                continue;
3638    
3639            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||              case ESC_v:
3640                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;              SETBIT(classbits, 0x0a); /* LF */
3641                SETBIT(classbits, 0x0b); /* VT */
3642                SETBIT(classbits, 0x0c); /* FF */
3643                SETBIT(classbits, 0x0d); /* CR */
3644                SETBIT(classbits, 0x85); /* NEL */
3645    #ifdef SUPPORT_UTF8
3646                if (utf8)
3647                  {
3648                  class_utf8 = TRUE;
3649                  *class_utf8data++ = XCL_RANGE;
3650                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3651                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3652                  }
3653    #endif
3654                continue;
3655    
3656            /* We need to deal with \P and \p in both phases. */              case ESC_V:
3657                for (c = 0; c < 32; c++)
3658                  {
3659                  int x = 0xff;
3660                  switch (c)
3661                    {
3662                    case 0x0a/8: x ^= 1 << (0x0a%8);
3663                                 x ^= 1 << (0x0b%8);
3664                                 x ^= 1 << (0x0c%8);
3665                                 x ^= 1 << (0x0d%8);
3666                                 break;
3667                    case 0x85/8: x ^= 1 << (0x85%8); break;
3668                    default: break;
3669                    }
3670                  classbits[c] |= x;
3671                  }
3672    
3673  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UTF8
3674            if (-c == ESC_p || -c == ESC_P)              if (utf8)
3675              {                {
3676              BOOL negated;                class_utf8 = TRUE;
3677              int pdata;                *class_utf8data++ = XCL_RANGE;
3678              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3679              if (ptype < 0) goto FAILED;                class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3680              class_utf8 = TRUE;                *class_utf8data++ = XCL_RANGE;
3681              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3682                XCL_PROP : XCL_NOTPROP;                class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3683              *class_utf8data++ = ptype;                }
3684              *class_utf8data++ = pdata;  #endif
             class_charcount -= 2;   /* Not a < 256 character */  
3685              continue;              continue;
3686              }  
3687    #ifdef SUPPORT_UCP
3688                case ESC_p:
3689                case ESC_P:
3690                  {
3691                  BOOL negated;
3692                  int pdata;
3693                  int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3694                  if (ptype < 0) goto FAILED;
3695                  class_utf8 = TRUE;
3696                  *class_utf8data++ = ((-c == ESC_p) != negated)?
3697                    XCL_PROP : XCL_NOTPROP;
3698                  *class_utf8data++ = ptype;
3699                  *class_utf8data++ = pdata;
3700                  class_charcount -= 2;   /* Not a < 256 character */
3701                  continue;
3702                  }
3703  #endif  #endif
3704            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3705            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3706            treated as literals. */              treated as literals. */
3707    
3708            if ((options & PCRE_EXTRA) != 0)              default:
3709              {              if ((options & PCRE_EXTRA) != 0)
3710              *errorcodeptr = ERR7;                {
3711              goto FAILED;                *errorcodeptr = ERR7;
3712                  goto FAILED;
3713                  }
3714                class_charcount -= 2;  /* Undo the default count from above */
3715                c = *ptr;              /* Get the final character and fall through */
3716                break;
3717              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3718            }            }
3719    
3720          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 2562  for (;; ptr++) Line 3728  for (;; ptr++)
3728        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3729    
3730        CHECK_RANGE:        CHECK_RANGE:
3731        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3732          {          {
3733          inescq = FALSE;          inescq = FALSE;
3734          ptr += 2;          ptr += 2;
# Line 2570  for (;; ptr++) Line 3736  for (;; ptr++)
3736    
3737        oldptr = ptr;        oldptr = ptr;
3738    
3739        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3740    
3741          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3742    
3743          /* Check for range */
3744    
3745          if (!inescq && ptr[1] == CHAR_MINUS)
3746          {          {
3747          int d;          int d;
3748          ptr += 2;          ptr += 2;
3749          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3750    
3751          /* If we hit \Q (not followed by \E) at this point, go into escaped          /* If we hit \Q (not followed by \E) at this point, go into escaped
3752          mode. */          mode. */
3753    
3754          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3755            {            {
3756            ptr += 2;            ptr += 2;
3757            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3758                { ptr += 2; continue; }
3759            inescq = TRUE;            inescq = TRUE;
3760            break;            break;
3761            }            }
3762    
3763          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3764            {            {
3765            ptr = oldptr;            ptr = oldptr;
3766            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2606  for (;; ptr++) Line 3779  for (;; ptr++)
3779          not any of the other escapes. Perl 5.6 treats a hyphen as a literal          not any of the other escapes. Perl 5.6 treats a hyphen as a literal
3780          in such circumstances. */          in such circumstances. */
3781    
3782          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3783            {            {
3784            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3785            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3786    
3787            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
3788    
3789            if (d < 0)            if (d < 0)
3790              {              {
3791              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = 'X';  
             else if (d == -ESC_R) d = 'R'; else  
3792                {                {
3793                ptr = oldptr;                ptr = oldptr;
3794                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2637  for (;; ptr++) Line 3807  for (;; ptr++)
3807    
3808          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3809    
3810            /* Remember \r or \n */
3811    
3812            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3813    
3814          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless          /* In UTF-8 mode, if the upper limit is > 255, or > 127 for caseless
3815          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3816          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2659  for (;; ptr++) Line 3833  for (;; ptr++)
3833              unsigned int origd = d;              unsigned int origd = d;
3834              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3835                {                {
3836                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3837                      ocd <= (unsigned int)d)
3838                    continue;                          /* Skip embedded ranges */
3839    
3840                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3841                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3842                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3843                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3844                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3845                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3846                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3847                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3848                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3849                  d = ocd;                  d = ocd;
3850                  continue;                  continue;
# Line 2751  for (;; ptr++) Line 3929  for (;; ptr++)
3929          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3930            {            {
3931            unsigned int othercase;            unsigned int othercase;
3932            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3933              {              {
3934              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3935              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2776  for (;; ptr++) Line 3954  for (;; ptr++)
3954          }          }
3955        }        }
3956    
3957      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" far above.
3958        If we are at the end of an internal nested string, revert to the outer
3959        string. */
3960    
3961        while (((c = *(++ptr)) != 0 ||
3962               (nestptr != NULL &&
3963                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3964               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3965    
3966      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3967    
3968      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3969        {        {
3970        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3971        goto FAILED;        goto FAILED;
3972        }        }
3973    
3974      /* If class_charcount is 1, we saw precisely one character whose value is      /* If class_charcount is 1, we saw precisely one character whose value is
3975      less than 256. In non-UTF-8 mode we can always optimize. In UTF-8 mode, we      less than 256. As long as there were no characters >= 128 and there was no
3976      can optimize the negative case only if there were no characters >= 128      use of \p or \P, in other words, no use of any XCLASS features, we can
3977      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3978      single-bytes only. This is an historical hangover. Maybe one day we can  
3979      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3980        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3981        operate on single-bytes only. This is an historical hangover. Maybe one day
3982        we can tidy these opcodes to handle multi-byte characters.
3983    
3984      The optimization throws away the bit map. We turn the item into a      The optimization throws away the bit map. We turn the item into a
3985      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2801  for (;; ptr++) Line 3989  for (;; ptr++)
3989      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3990    
3991  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3992      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3993            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3994  #else  #else
3995      if (class_charcount == 1)      if (class_charcount == 1)
3996  #endif  #endif
# Line 2847  for (;; ptr++) Line 4033  for (;; ptr++)
4033      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4034    
4035      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4036      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
4037      we can omit the bitmap in the actual compiled code. */      such as \S in the class, and PCRE_UCP is not set, because in that case all
4038        characters > 255 are in the class, so any that were explicitly given as
4039        well can be ignored. If (when there are explicit characters > 255 that must
4040        be listed) there are no characters < 256, we can omit the bitmap in the
4041        actual compiled code. */
4042    
4043  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4044      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4045        {        {
4046        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4047        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2877  for (;; ptr++) Line 4067  for (;; ptr++)
4067        }        }
4068  #endif  #endif
4069    
4070      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, or they are all to be included or
4071      and copy it into the code vector. If this is the first thing in the branch,      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4072      there can be no first char setting, whatever the repeat count. Any reqbyte      whole class was negated and whether there were negative specials such as \S
4073      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4074        negating it if necessary. */
4075    
4076        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4077      if (negate_class)      if (negate_class)
4078        {        {
       *code++ = OP_NCLASS;  
4079        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4080          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4081        }        }
4082      else      else
4083        {        {
       *code++ = OP_CLASS;  
4084        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4085        }        }
4086      code += 32;      code += 32;
# Line 2901  for (;; ptr++) Line 4091  for (;; ptr++)
4091      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4092      has been tested above. */      has been tested above. */
4093    
4094      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4095      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4096      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4097      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4098      goto REPEAT;      goto REPEAT;
4099    
4100      case '*':      case CHAR_ASTERISK:
4101      repeat_min = 0;      repeat_min = 0;
4102      repeat_max = -1;      repeat_max = -1;
4103      goto REPEAT;      goto REPEAT;
4104    
4105      case '+':      case CHAR_PLUS:
4106      repeat_min = 1;      repeat_min = 1;
4107      repeat_max = -1;      repeat_max = -1;
4108      goto REPEAT;      goto REPEAT;
4109    
4110      case '?':      case CHAR_QUESTION_MARK:
4111      repeat_min = 0;      repeat_min = 0;
4112      repeat_max = 1;      repeat_max = 1;
4113    
# Line 2952  for (;; ptr++) Line 4142  for (;; ptr++)
4142      but if PCRE_UNGREEDY is set, it works the other way round. We change the      but if PCRE_UNGREEDY is set, it works the other way round. We change the
4143      repeat type to the non-default. */      repeat type to the non-default. */
4144    
4145      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4146        {        {
4147        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4148        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4149        ptr++;        ptr++;
4150        }        }
4151      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4152        {        {
4153        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4154        ptr++;        ptr++;
# Line 3005  for (;; ptr++) Line 4195  for (;; ptr++)
4195    
4196        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4197            repeat_max < 0 &&            repeat_max < 0 &&
4198            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4199          {          {
4200          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4201          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3027  for (;; ptr++) Line 4216  for (;; ptr++)
4216        c = previous[1];        c = previous[1];
4217        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4218            repeat_max < 0 &&            repeat_max < 0 &&
4219            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4220          {          {
4221          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4222          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3051  for (;; ptr++) Line 4240  for (;; ptr++)
4240    
4241        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4242            repeat_max < 0 &&            repeat_max < 0 &&
4243            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4244          {          {
4245          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4246          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3073  for (;; ptr++) Line 4262  for (;; ptr++)
4262    
4263        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4264    
4265          /*--------------------------------------------------------------------*/
4266          /* This code is obsolete from release 8.00; the restriction was finally
4267          removed: */
4268    
4269        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4270        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4271    
4272        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4273          /*--------------------------------------------------------------------*/
4274    
4275        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4276    
# Line 3223  for (;; ptr++) Line 4417  for (;; ptr++)
4417          goto END_REPEAT;          goto END_REPEAT;
4418          }          }
4419    
4420          /*--------------------------------------------------------------------*/
4421          /* This code is obsolete from release 8.00; the restriction was finally
4422          removed: */
4423    
4424        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4425        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4426    
4427        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4428          /*--------------------------------------------------------------------*/
4429    
4430        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4431          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3251  for (;; ptr++) Line 4450  for (;; ptr++)
4450        {        {
4451        register int i;        register int i;
4452        int ketoffset = 0;        int ketoffset = 0;
4453        int len = code - previous;        int len = (int)(code - previous);
4454        uschar *bralink = NULL;        uschar *bralink = NULL;
4455    
4456        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3262  for (;; ptr++) Line 4461  for (;; ptr++)
4461          goto FAILED;          goto FAILED;
4462          }          }
4463    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4464        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
4465        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
4466        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 3280  for (;; ptr++) Line 4471  for (;; ptr++)
4471          {          {
4472          register uschar *ket = previous;          register uschar *ket = previous;
4473          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4474          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4475          }          }
4476    
4477        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 3292  for (;; ptr++) Line 4483  for (;; ptr++)
4483    
4484        if (repeat_min == 0)        if (repeat_min == 0)
4485          {          {
4486          /* If the maximum is also zero, we just omit the group from the output          /* If the maximum is also zero, we used to just omit the group from the
4487          altogether. */          output altogether, like this:
4488    
4489          if (repeat_max == 0)          ** if (repeat_max == 0)
4490            {          **   {
4491            code = previous;          **   code = previous;
4492            goto END_REPEAT;          **   goto END_REPEAT;
4493            }          **   }
4494    
4495            However, that fails when a group is referenced as a subroutine from
4496            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4497            so that it is skipped on execution. As we don't have a list of which
4498            groups are referenced, we cannot do this selectively.
4499    
4500            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4501            and do no more at this point. However, we do need to adjust any
4502            OP_RECURSE calls inside the group that refer to the group itself or any
4503            internal or forward referenced group, because the offset is from the
4504            start of the whole regex. Temporarily terminate the pattern while doing
4505            this. */
4506    
4507          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4508            {            {
4509            *code = OP_END;            *code = OP_END;
4510            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4511            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4512            code++;            code++;
4513              if (repeat_max == 0)
4514                {
4515                *previous++ = OP_SKIPZERO;
4516                goto END_REPEAT;
4517                }
4518            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4519            }            }
4520    
# Line 3338  for (;; ptr++) Line 4539  for (;; ptr++)
4539            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4540            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4541    
4542            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4543            bralink = previous;            bralink = previous;
4544            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4545            }            }
# Line 3358  for (;; ptr++) Line 4559  for (;; ptr++)
4559          if (repeat_min > 1)          if (repeat_min > 1)
4560            {            {
4561            /* In the pre-compile phase, we don't actually do the replication. We            /* In the pre-compile phase, we don't actually do the replication. We
4562            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4563              potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4564              integer type when available, otherwise double. */
4565    
4566            if (lengthptr != NULL)            if (lengthptr != NULL)
4567              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4568                int delta = (repeat_min - 1)*length_prevgroup;
4569                if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4570                      (INT64_OR_DOUBLE)length_prevgroup >
4571                        (INT64_OR_DOUBLE)INT_MAX ||
4572                    OFLOW_MAX - *lengthptr < delta)
4573                  {
4574                  *errorcodeptr = ERR20;
4575                  goto FAILED;
4576                  }
4577                *lengthptr += delta;
4578                }
4579    
4580            /* This is compiling for real */            /* This is compiling for real */
4581    
# Line 3399  for (;; ptr++) Line 4613  for (;; ptr++)
4613          /* In the pre-compile phase, we don't actually do the replication. We          /* In the pre-compile phase, we don't actually do the replication. We
4614          just adjust the length as if we had. For each repetition we must add 1          just adjust the length as if we had. For each repetition we must add 1
4615          to the length for BRAZERO and for all but the last repetition we must          to the length for BRAZERO and for all but the last repetition we must
4616          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4617            paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4618            a 64-bit integer type when available, otherwise double. */
4619    
4620          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4621            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4622              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4623                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4624              if ((INT64_OR_DOUBLE)repeat_max *
4625                    (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4626                      > (INT64_OR_DOUBLE)INT_MAX ||
4627                  OFLOW_MAX - *lengthptr < delta)
4628                {
4629                *errorcodeptr = ERR20;
4630                goto FAILED;
4631                }
4632              *lengthptr += delta;
4633              }
4634    
4635          /* This is compiling for real */          /* This is compiling for real */
4636    
# Line 3421  for (;; ptr++) Line 4648  for (;; ptr++)
4648              {              {
4649              int offset;              int offset;
4650              *code++ = OP_BRA;              *code++ = OP_BRA;
4651              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4652              bralink = code;              bralink = code;
4653              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4654              }              }
# Line 3442  for (;; ptr++) Line 4669  for (;; ptr++)
4669          while (bralink != NULL)          while (bralink != NULL)
4670            {            {
4671            int oldlinkoffset;            int oldlinkoffset;
4672            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4673            uschar *bra = code - offset;            uschar *bra = code - offset;
4674            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4675            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 3473  for (;; ptr++) Line 4700  for (;; ptr++)
4700            uschar *scode = bracode;            uschar *scode = bracode;
4701            do            do
4702              {              {
4703              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4704                {                {
4705                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4706                break;                break;
# Line 3485  for (;; ptr++) Line 4712  for (;; ptr++)
4712          }          }
4713        }        }
4714    
4715        /* If previous is OP_FAIL, it was generated by an empty class [] in
4716        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4717        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4718        error above. We can just ignore the repeat in JS case. */
4719    
4720        else if (*previous == OP_FAIL) goto END_REPEAT;
4721    
4722      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4723    
4724      else      else
# Line 3509  for (;; ptr++) Line 4743  for (;; ptr++)
4743      if (possessive_quantifier)      if (possessive_quantifier)
4744        {        {
4745        int len;        int len;
4746        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4747            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4748            tempcode += _pcre_OP_lengths[*tempcode] +
4749              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4750    
4751          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4752            {
4753          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4754        len = code - tempcode;  #ifdef SUPPORT_UTF8
4755            if (utf8 && tempcode[-1] >= 0xc0)
4756              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4757    #endif
4758            }
4759    
4760          len = (int)(code - tempcode);
4761        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4762          {          {
4763          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 3530  for (;; ptr++) Line 4775  for (;; ptr++)
4775          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4776          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4777    
4778            /* Because we are moving code along, we must ensure that any
4779            pending recursive references are updated. */
4780    
4781          default:          default:
4782            *code = OP_END;
4783            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4784          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4785          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4786          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3546  for (;; ptr++) Line 4796  for (;; ptr++)
4796      "follows varying string" flag for subsequently encountered reqbytes if      "follows varying string" flag for subsequently encountered reqbytes if
4797      it isn't already set and we have just passed a varying length item. */      it isn't already set and we have just passed a varying length item. */
4798    
4799      END_REPEAT:      END_REPEAT:
4800      previous = NULL;      previous = NULL;
4801      cd->req_varyopt |= reqvary;      cd->req_varyopt |= reqvary;
4802      break;      break;
4803    
4804    
4805        /* ===================================================================*/
4806        /* Start of nested parenthesized sub-expression, or comment or lookahead or
4807        lookbehind or option setting or condition or all the other extended
4808        parenthesis forms.  */
4809    
4810        case CHAR_LEFT_PARENTHESIS:
4811        newoptions = options;
4812        skipbytes = 0;
4813        bravalue = OP_CBRA;
4814        save_hwm = cd->hwm;
4815        reset_bracount = FALSE;
4816