/[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 172 by ph10, Tue Jun 5 10:40:13 2007 UTC revision 499 by ph10, Wed Mar 3 20:09:39 2010 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) 1997-2007 University of Cambridge             Copyright (c) 1997-2010 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 82  are simple data values; negative values Line 98  are simple data values; negative values
98  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
99  is invalid. */  is invalid. */
100    
101  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
102    
103    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
104    in UTF-8 mode. */
105    
106  static const short int escapes[] = {  static const short int escapes[] = {
107       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
108       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
109     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
110       0,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
111  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
112  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
113     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
114       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
115  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
116       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
117         -ESC_D,                  -ESC_E,
118         0,                       -ESC_G,
119         -ESC_H,                  0,
120         0,                       -ESC_K,
121         0,                       0,
122         0,                       0,
123         -ESC_P,                  -ESC_Q,
124         -ESC_R,                  -ESC_S,
125         0,                       0,
126         -ESC_V,                  -ESC_W,
127         -ESC_X,                  0,
128         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
129         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
130         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
131         CHAR_GRAVE_ACCENT,       7,
132         -ESC_b,                  0,
133         -ESC_d,                  ESC_e,
134         ESC_f,                   0,
135         -ESC_h,                  0,
136         0,                       -ESC_k,
137         0,                       0,
138         ESC_n,                   0,
139         -ESC_p,                  0,
140         ESC_r,                   -ESC_s,
141         ESC_tee,                 0,
142         -ESC_v,                  -ESC_w,
143         0,                       0,
144         -ESC_z
145  };  };
146    
147  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
148    
149    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
150    
151  static const short int escapes[] = {  static const short int escapes[] = {
152  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
153  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 106  static const short int escapes[] = { Line 157  static const short int escapes[] = {
157  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
158  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
159  /*  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,
160  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
161  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
162  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
163  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
164  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
165  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
166  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
167  /*  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,
168  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
169  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
170  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
171  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
172  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
173  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
174  /*  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 176  static const short int escapes[] = {
176  #endif  #endif
177    
178    
179  /* 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
180  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
181  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
182    string is built from string macros so that it works in UTF-8 mode on EBCDIC
183  static const char *const posix_names[] = {  platforms. */
184    "alpha", "lower", "upper",  
185    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
186    "print", "punct", "space", "word",  "xdigit" };    int   len;
187      int   op;
188    } verbitem;
189    
190    static const char verbnames[] =
191      STRING_ACCEPT0
192      STRING_COMMIT0
193      STRING_F0
194      STRING_FAIL0
195      STRING_PRUNE0
196      STRING_SKIP0
197      STRING_THEN;
198    
199    static const verbitem verbs[] = {
200      { 6, OP_ACCEPT },
201      { 6, OP_COMMIT },
202      { 1, OP_FAIL },
203      { 4, OP_FAIL },
204      { 5, OP_PRUNE },
205      { 4, OP_SKIP  },
206      { 4, OP_THEN  }
207    };
208    
209    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
210    
211    
212    /* Tables of names of POSIX character classes and their lengths. The names are
213    now all in a single string, to reduce the number of relocations when a shared
214    library is dynamically loaded. The list of lengths is terminated by a zero
215    length entry. The first three must be alpha, lower, upper, as this is assumed
216    for handling case independence. */
217    
218    static const char posix_names[] =
219      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
220      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
221      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
222      STRING_word0  STRING_xdigit;
223    
224  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
225    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 171  static const int posix_class_maps[] = { Line 258  static const int posix_class_maps[] = {
258  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
259  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
260  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
261  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
262    the number of relocations needed when a shared library is loaded dynamically,
263  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
264    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
265    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
266    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
267    "unrecognized character follows \\",  
268    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
269    substring, so that the whole string ends with \0\0, which can be detected when
270    counting through. */
271    
272    static const char error_texts[] =
273      "no error\0"
274      "\\ at end of pattern\0"
275      "\\c at end of pattern\0"
276      "unrecognized character follows \\\0"
277      "numbers out of order in {} quantifier\0"
278    /* 5 */    /* 5 */
279    "number too big in {} quantifier",    "number too big in {} quantifier\0"
280    "missing terminating ] for character class",    "missing terminating ] for character class\0"
281    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
282    "range out of order in character class",    "range out of order in character class\0"
283    "nothing to repeat",    "nothing to repeat\0"
284    /* 10 */    /* 10 */
285    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
286    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
287    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
288    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
289    "missing )",    "missing )\0"
290    /* 15 */    /* 15 */
291    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
292    "erroffset passed as NULL",    "erroffset passed as NULL\0"
293    "unknown option bit(s) set",    "unknown option bit(s) set\0"
294    "missing ) after comment",    "missing ) after comment\0"
295    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
296    /* 20 */    /* 20 */
297    "regular expression too large",    "regular expression is too large\0"
298    "failed to get memory",    "failed to get memory\0"
299    "unmatched parentheses",    "unmatched parentheses\0"
300    "internal error: code overflow",    "internal error: code overflow\0"
301    "unrecognized character after (?<",    "unrecognized character after (?<\0"
302    /* 25 */    /* 25 */
303    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
304    "malformed number or name after (?(",    "malformed number or name after (?(\0"
305    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
306    "assertion expected after (?(",    "assertion expected after (?(\0"
307    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
308    /* 30 */    /* 30 */
309    "unknown POSIX class name",    "unknown POSIX class name\0"
310    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
311    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
312    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
313    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
314    /* 35 */    /* 35 */
315    "invalid condition (?(0)",    "invalid condition (?(0)\0"
316    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
317    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
318    "number after (?C is > 255",    "number after (?C is > 255\0"
319    "closing ) for (?C expected",    "closing ) for (?C expected\0"
320    /* 40 */    /* 40 */
321    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
322    "unrecognized character after (?P",    "unrecognized character after (?P\0"
323    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
324    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
325    "invalid UTF-8 string",    "invalid UTF-8 string\0"
326    /* 45 */    /* 45 */
327    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
328    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
329    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
330    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
331    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
332    /* 50 */    /* 50 */
333    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
334    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
335    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
336    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
337    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
338    /* 55 */    /* 55 */
339    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
340    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
341    "\\g is not followed by a braced name or 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"
342    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
343  };    "(*VERB) with an argument is not supported\0"
344      /* 60 */
345      "(*VERB) not recognized\0"
346      "number is too big\0"
347      "subpattern name expected\0"
348      "digit expected after (?+\0"
349      "] is an invalid data character in JavaScript compatibility mode\0"
350      /* 65 */
351      "different names for subpatterns of the same number are not allowed\0";
352    
353  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
354  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 263  For convenience, we use the same bit def Line 366  For convenience, we use the same bit def
366    
367  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
368    
369  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
370    
371    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
372    UTF-8 mode. */
373    
374  static const unsigned char digitab[] =  static const unsigned char digitab[] =
375    {    {
376    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 299  static const unsigned char digitab[] = Line 406  static const unsigned char digitab[] =
406    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
407    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
408    
409  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
410    
411    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
412    
413  static const unsigned char digitab[] =  static const unsigned char digitab[] =
414    {    {
415    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 374  static const unsigned char ebcdic_charta Line 484  static const unsigned char ebcdic_charta
484  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
485    
486  static BOOL  static BOOL
487    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
488      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
489    
490    
491    
492  /*************************************************  /*************************************************
493    *            Find an error text                  *
494    *************************************************/
495    
496    /* The error texts are now all in one long string, to save on relocations. As
497    some of the text is of unknown length, we can't use a table of offsets.
498    Instead, just count through the strings. This is not a performance issue
499    because it happens only when there has been a compilation error.
500    
501    Argument:   the error number
502    Returns:    pointer to the error string
503    */
504    
505    static const char *
506    find_error_text(int n)
507    {
508    const char *s = error_texts;
509    for (; n > 0; n--)
510      {
511      while (*s++ != 0) {};
512      if (*s == 0) return "Error text not found (please report)";
513      }
514    return s;
515    }
516    
517    
518    /*************************************************
519  *            Handle escapes                      *  *            Handle escapes                      *
520  *************************************************/  *************************************************/
521    
# Line 400  Arguments: Line 536  Arguments:
536    
537  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
538                   negative => a special escape sequence                   negative => a special escape sequence
539                   on error, errorptr is set                   on error, errorcodeptr is set
540  */  */
541    
542  static int  static int
# Line 418  ptr--;                            /* Set Line 554  ptr--;                            /* Set
554    
555  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
556    
557  /* 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
558  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.
559  Otherwise further processing may be required. */  Otherwise further processing may be required. */
560    
561  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
562  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
563  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
564    
565  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
566  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
567  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
568  #endif  #endif
569    
# Line 443  else Line 579  else
579      /* 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
580      error. */      error. */
581    
582      case 'l':      case CHAR_l:
583      case 'L':      case CHAR_L:
584      case 'N':      case CHAR_N:
585      case 'u':      case CHAR_u:
586      case 'U':      case CHAR_U:
587      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
588      break;      break;
589    
590      /* \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:
591      is an absolute backreference. If negative, it is a relative backreference.  
592      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
593      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
594      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
595      fudge it up by pretending it really was \k. */  
596        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
597        is part of Perl's movement towards a unified syntax for back references. As
598        this is synonymous with \k{name}, we fudge it up by pretending it really
599        was \k.
600    
601        (3) For Oniguruma compatibility we also support \g followed by a name or a
602        number either in angle brackets or in single quotes. However, these are
603        (possibly recursive) subroutine calls, _not_ backreferences. Just return
604        the -ESC_g code (cf \k). */
605    
606        case CHAR_g:
607        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
608          {
609          c = -ESC_g;
610          break;
611          }
612    
613        /* Handle the Perl-compatible cases */
614    
615      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
616        {        {
617        const uschar *p;        const uschar *p;
618        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
619          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
620        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
621          {          {
622          c = -ESC_k;          c = -ESC_k;
623          break;          break;
# Line 474  else Line 627  else
627        }        }
628      else braced = FALSE;      else braced = FALSE;
629    
630      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
631        {        {
632        negated = TRUE;        negated = TRUE;
633        ptr++;        ptr++;
# Line 483  else Line 636  else
636    
637      c = 0;      c = 0;
638      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
639        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
640    
641        if (c < 0)   /* Integer overflow */
642          {
643          *errorcodeptr = ERR61;
644          break;
645          }
646    
647      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
648        {        {
649        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
650        return 0;        break;
651          }
652    
653        if (c == 0)
654          {
655          *errorcodeptr = ERR58;
656          break;
657        }        }
658    
659      if (negated)      if (negated)
# Line 496  else Line 661  else
661        if (c > bracount)        if (c > bracount)
662          {          {
663          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
664          return 0;          break;
665          }          }
666        c = bracount - (c - 1);        c = bracount - (c - 1);
667        }        }
# Line 516  else Line 681  else
681      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
682      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
683    
684      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:
685      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
686    
687      if (!isclass)      if (!isclass)
688        {        {
689        oldptr = ptr;        oldptr = ptr;
690        c -= '0';        c -= CHAR_0;
691        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
692          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
693          if (c < 0)    /* Integer overflow */
694            {
695            *errorcodeptr = ERR61;
696            break;
697            }
698        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
699          {          {
700          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 537  else Line 707  else
707      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.
708      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
709    
710      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
711        {        {
712        ptr--;        ptr--;
713        c = 0;        c = 0;
# Line 550  else Line 720  else
720      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
721      than 3 octal digits. */      than 3 octal digits. */
722    
723      case '0':      case CHAR_0:
724      c -= '0';      c -= CHAR_0;
725      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
726          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
727      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
728      break;      break;
729    
# Line 561  else Line 731  else
731      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
732      treated as a data character. */      treated as a data character. */
733    
734      case 'x':      case CHAR_x:
735      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
736        {        {
737        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
738        int count = 0;        int count = 0;
# Line 571  else Line 741  else
741        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
742          {          {
743          register int cc = *pt++;          register int cc = *pt++;
744          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
745          count++;          count++;
746    
747  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
748          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
749          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
750  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
751          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
752          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
753  #endif  #endif
754          }          }
755    
756        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
757          {          {
758          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
759          ptr = pt;          ptr = pt;
# Line 599  else Line 769  else
769      c = 0;      c = 0;
770      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
771        {        {
772        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
773        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
774  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
775        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
776        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
777  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
778        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
779        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
780  #endif  #endif
781        }        }
782      break;      break;
# Line 615  else Line 785  else
785      This coding is ASCII-specific, but then the whole concept of \cx is      This coding is ASCII-specific, but then the whole concept of \cx is
786      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
787    
788      case 'c':      case CHAR_c:
789      c = *(++ptr);      c = *(++ptr);
790      if (c == 0)      if (c == 0)
791        {        {
792        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
793        return 0;        break;
794        }        }
795    
796  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
797      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
798      c ^= 0x40;      c ^= 0x40;
799  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
800      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
801      c ^= 0xC0;      c ^= 0xC0;
802  #endif  #endif
803      break;      break;
804    
805      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
806      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
807      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
808      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
809      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
810    
811      default:      default:
812      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 689  if (c == 0) goto ERROR_RETURN; Line 859  if (c == 0) goto ERROR_RETURN;
859  /* \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
860  negation. */  negation. */
861    
862  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
863    {    {
864    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
865      {      {
866      *negptr = TRUE;      *negptr = TRUE;
867      ptr++;      ptr++;
868      }      }
869    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
870      {      {
871      c = *(++ptr);      c = *(++ptr);
872      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
873      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
874      name[i] = c;      name[i] = c;
875      }      }
876    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
877    name[i] = 0;    name[i] = 0;
878    }    }
879    
# Line 725  top = _pcre_utt_size; Line 895  top = _pcre_utt_size;
895  while (bot < top)  while (bot < top)
896    {    {
897    i = (bot + top) >> 1;    i = (bot + top) >> 1;
898    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
899    if (c == 0)    if (c == 0)
900      {      {
901      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 768  is_counted_repeat(const uschar *p) Line 938  is_counted_repeat(const uschar *p)
938  {  {
939  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
940  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
941  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
942    
943  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
944  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
945    
946  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
947  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
948    
949  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
950  }  }
951    
952    
# Line 809  int max = -1; Line 979  int max = -1;
979  /* 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
980  an integer overflow. */  an integer overflow. */
981    
982  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
983  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
984    {    {
985    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 819  if (min < 0 || min > 65535) Line 989  if (min < 0 || min > 65535)
989  /* 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.
990  Also, max must not be less than min. */  Also, max must not be less than min. */
991    
992  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
993    {    {
994    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
995      {      {
996      max = 0;      max = 0;
997      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
998      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
999        {        {
1000        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 849  return p; Line 1019  return p;
1019    
1020    
1021  /*************************************************  /*************************************************
1022  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1023  *************************************************/  *************************************************/
1024    
1025  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1026    top-level call starts at the beginning of the pattern. All other calls must
1027    start at a parenthesis. It scans along a pattern's text looking for capturing
1028  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
1029  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
1030  returns when it reaches a given numbered subpattern. This is used for forward  returns when it reaches a given numbered subpattern. We know that if (?P< is
1031  references to subpatterns. We know that if (?P< is encountered, the name will  encountered, the name will be terminated by '>' because that is checked in the
1032  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1033    capturing group numbers - the (?| feature.
1034    
1035  Arguments:  Arguments:
1036    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1037    count        current count of capturing parens so far encountered    cd           compile background data
1038    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1039    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1040    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1041      count        pointer to the current capturing subpattern number (updated)
1042    
1043  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1044  */  */
1045    
1046  static int  static int
1047  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,
1048    BOOL xmode)    BOOL xmode, int *count)
1049  {  {
1050  const uschar *thisname;  uschar *ptr = *ptrptr;
1051    int start_count = *count;
1052    int hwm_count = start_count;
1053    BOOL dup_parens = FALSE;
1054    
1055  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1056    dealing with. The very first call may not start with a parenthesis. */
1057    
1058    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1059    {    {
1060    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1061          ptr[2] == CHAR_VERTICAL_LINE)
1062        {
1063        ptr += 3;
1064        dup_parens = TRUE;
1065        }
1066    
1067      /* Handle a normal, unnamed capturing parenthesis */
1068    
1069      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1070        {
1071        *count += 1;
1072        if (name == NULL && *count == lorn) return *count;
1073        ptr++;
1074        }
1075    
1076      /* Handle a condition. If it is an assertion, just carry on so that it
1077      is processed as normal. If not, skip to the closing parenthesis of the
1078      condition (there can't be any nested parens. */
1079    
1080      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1081        {
1082        ptr += 2;
1083        if (ptr[1] != CHAR_QUESTION_MARK)
1084          {
1085          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1086          if (*ptr != 0) ptr++;
1087          }
1088        }
1089    
1090      /* We have either (? or (* and not a condition */
1091    
1092      else
1093        {
1094        ptr += 2;
1095        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1096    
1097        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1098    
1099        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1100            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1101          {
1102          int term;
1103          const uschar *thisname;
1104          *count += 1;
1105          if (name == NULL && *count == lorn) return *count;
1106          term = *ptr++;
1107          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1108          thisname = ptr;
1109          while (*ptr != term) ptr++;
1110          if (name != NULL && lorn == ptr - thisname &&
1111              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1112            return *count;
1113          term++;
1114          }
1115        }
1116      }
1117    
1118    /* Past any initial parenthesis handling, scan for parentheses or vertical
1119    bars. */
1120    
1121    for (; *ptr != 0; ptr++)
1122      {
1123    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1124    
1125    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1126      {      {
1127      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1128      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1129        {        {
1130        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1131        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1132        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1133        }        }
1134      continue;      continue;
1135      }      }
1136    
1137    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1138      are handled for real. If the first character is '^', skip it. Also, if the
1139      first few characters (either before or after ^) are \Q\E or \E we skip them
1140      too. This makes for compatibility with Perl. Note the use of STR macros to
1141      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1142    
1143      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1144        {
1145        BOOL negate_class = FALSE;
1146        for (;;)
1147          {
1148          if (ptr[1] == CHAR_BACKSLASH)
1149            {
1150            if (ptr[2] == CHAR_E)
1151              ptr+= 2;
1152            else if (strncmp((const char *)ptr+2,
1153                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1154              ptr += 4;
1155            else
1156              break;
1157            }
1158          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1159            {
1160            negate_class = TRUE;
1161            ptr++;
1162            }
1163          else break;
1164          }
1165    
1166        /* If the next character is ']', it is a data character that must be
1167        skipped, except in JavaScript compatibility mode. */
1168    
1169    if (*ptr == '[')      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1170      {          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1171      while (*(++ptr) != ']')        ptr++;
1172    
1173        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1174        {        {
1175        if (*ptr == '\\')        if (*ptr == 0) return -1;
1176          if (*ptr == CHAR_BACKSLASH)
1177          {          {
1178          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1179          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1180            {            {
1181            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1182            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1183            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1184            }            }
1185          continue;          continue;
1186          }          }
# Line 916  for (; *ptr != 0; ptr++) Line 1190  for (; *ptr != 0; ptr++)
1190    
1191    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1192    
1193    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1194      {      {
1195      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1196      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1197      continue;      continue;
1198      }      }
1199    
1200    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1201    
1202    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1203      {      {
1204      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1205      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1206      continue;      if (*ptr == 0) goto FAIL_EXIT;
1207      }      }
1208    
1209    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1210    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1211        if (dup_parens && *count < hwm_count) *count = hwm_count;
1212        *ptrptr = ptr;
1213        return -1;
1214        }
1215    
1216    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1217        {
1218        if (*count > hwm_count) hwm_count = *count;
1219        *count = start_count;
1220        }
1221      }
1222    
1223    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1224         *ptr != '\'')  *ptrptr = ptr;
1225      continue;  return -1;
1226    }
1227    
1228    
1229    
1230    
1231    /*************************************************
1232    *       Find forward referenced subpattern       *
1233    *************************************************/
1234    
1235    /* This function scans along a pattern's text looking for capturing
1236    subpatterns, and counting them. If it finds a named pattern that matches the
1237    name it is given, it returns its number. Alternatively, if the name is NULL, it
1238    returns when it reaches a given numbered subpattern. This is used for forward
1239    references to subpatterns. We used to be able to start this scan from the
1240    current compiling point, using the current count value from cd->bracount, and
1241    do it all in a single loop, but the addition of the possibility of duplicate
1242    subpattern numbers means that we have to scan from the very start, in order to
1243    take account of such duplicates, and to use a recursive function to keep track
1244    of the different types of group.
1245    
1246    Arguments:
1247      cd           compile background data
1248      name         name to seek, or NULL if seeking a numbered subpattern
1249      lorn         name length, or subpattern number if name is NULL
1250      xmode        TRUE if we are in /x mode
1251    
1252    Returns:       the number of the found subpattern, or -1 if not found
1253    */
1254    
1255    count++;  static int
1256    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1257    {
1258    uschar *ptr = (uschar *)cd->start_pattern;
1259    int count = 0;
1260    int rc;
1261    
1262    /* If the pattern does not start with an opening parenthesis, the first call
1263    to find_parens_sub() will scan right to the end (if necessary). However, if it
1264    does start with a parenthesis, find_parens_sub() will return when it hits the
1265    matching closing parens. That is why we have to have a loop. */
1266    
1267    if (name == NULL && count == lorn) return count;  for (;;)
1268    term = *ptr++;    {
1269    if (term == '<') term = '>';    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1270    thisname = ptr;    if (rc > 0 || *ptr++ == 0) break;
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1271    }    }
1272    
1273  return -1;  return rc;
1274  }  }
1275    
1276    
1277    
1278    
1279  /*************************************************  /*************************************************
1280  *      Find first significant op code            *  *      Find first significant op code            *
1281  *************************************************/  *************************************************/
# Line 1008  for (;;) Line 1325  for (;;)
1325    
1326      case OP_CALLOUT:      case OP_CALLOUT:
1327      case OP_CREF:      case OP_CREF:
1328        case OP_NCREF:
1329      case OP_RREF:      case OP_RREF:
1330        case OP_NRREF:
1331      case OP_DEF:      case OP_DEF:
1332      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1333      break;      break;
# Line 1024  for (;;) Line 1343  for (;;)
1343    
1344    
1345  /*************************************************  /*************************************************
1346  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1347  *************************************************/  *************************************************/
1348    
1349  /* 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,
1350  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.
1351  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
1352    temporarily terminated with OP_END when this function is called.
1353    
1354    This function is called when a backward assertion is encountered, so that if it
1355    fails, the error message can point to the correct place in the pattern.
1356    However, we cannot do this when the assertion contains subroutine calls,
1357    because they can be forward references. We solve this by remembering this case
1358    and doing the check at the end; a flag specifies which mode we are running in.
1359    
1360  Arguments:  Arguments:
1361    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1362    options  the compiling options    options  the compiling options
1363      atend    TRUE if called when the pattern is complete
1364      cd       the "compile data" structure
1365    
1366  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1367                 or -1 if there is no fixed length,
1368               or -2 if \C was encountered               or -2 if \C was encountered
1369                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1370  */  */
1371    
1372  static int  static int
1373  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1374  {  {
1375  int length = -1;  int length = -1;
1376    
# Line 1053  branch, check the length against that of Line 1383  branch, check the length against that of
1383  for (;;)  for (;;)
1384    {    {
1385    int d;    int d;
1386      uschar *ce, *cs;
1387    register int op = *cc;    register int op = *cc;
   
1388    switch (op)    switch (op)
1389      {      {
1390      case OP_CBRA:      case OP_CBRA:
1391      case OP_BRA:      case OP_BRA:
1392      case OP_ONCE:      case OP_ONCE:
1393      case OP_COND:      case OP_COND:
1394      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1395      if (d < 0) return d;      if (d < 0) return d;
1396      branchlength += d;      branchlength += d;
1397      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1084  for (;;) Line 1414  for (;;)
1414      branchlength = 0;      branchlength = 0;
1415      break;      break;
1416    
1417        /* A true recursion implies not fixed length, but a subroutine call may
1418        be OK. If the subroutine is a forward reference, we can't deal with
1419        it until the end of the pattern, so return -3. */
1420    
1421        case OP_RECURSE:
1422        if (!atend) return -3;
1423        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1424        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1425        if (cc > cs && cc < ce) return -1;                /* Recursion */
1426        d = find_fixedlength(cs + 2, options, atend, cd);
1427        if (d < 0) return d;
1428        branchlength += d;
1429        cc += 1 + LINK_SIZE;
1430        break;
1431    
1432      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1433    
1434      case OP_ASSERT:      case OP_ASSERT:
# Line 1097  for (;;) Line 1442  for (;;)
1442    
1443      case OP_REVERSE:      case OP_REVERSE:
1444      case OP_CREF:      case OP_CREF:
1445        case OP_NCREF:
1446      case OP_RREF:      case OP_RREF:
1447        case OP_NRREF:
1448      case OP_DEF:      case OP_DEF:
1449      case OP_OPT:      case OP_OPT:
1450      case OP_CALLOUT:      case OP_CALLOUT:
# Line 1120  for (;;) Line 1467  for (;;)
1467      branchlength++;      branchlength++;
1468      cc += 2;      cc += 2;
1469  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1470      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1471        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1472  #endif  #endif
1473      break;      break;
1474    
# Line 1134  for (;;) Line 1479  for (;;)
1479      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1480      cc += 4;      cc += 4;
1481  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1482      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1483        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1484  #endif  #endif
1485      break;      break;
1486    
1487      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1488      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1489        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1490      cc += 4;      cc += 4;
1491      break;      break;
1492    
# Line 1160  for (;;) Line 1504  for (;;)
1504      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1505      case OP_WORDCHAR:      case OP_WORDCHAR:
1506      case OP_ANY:      case OP_ANY:
1507        case OP_ALLANY:
1508      branchlength++;      branchlength++;
1509      cc++;      cc++;
1510      break;      break;
# Line 1214  for (;;) Line 1559  for (;;)
1559    
1560    
1561  /*************************************************  /*************************************************
1562  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1563  *************************************************/  *************************************************/
1564    
1565  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1566  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1567    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1568    so that it can be called from pcre_study() when finding the minimum matching
1569    length.
1570    
1571  Arguments:  Arguments:
1572    code        points to start of expression    code        points to start of expression
1573    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1574    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1575    
1576  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
1577  */  */
1578    
1579  static const uschar *  const uschar *
1580  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1581  {  {
1582  for (;;)  for (;;)
1583    {    {
# Line 1242  for (;;) Line 1590  for (;;)
1590    
1591    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1592    
1593      /* Handle recursion */
1594    
1595      else if (c == OP_REVERSE)
1596        {
1597        if (number < 0) return (uschar *)code;
1598        code += _pcre_OP_lengths[c];
1599        }
1600    
1601    /* Handle capturing bracket */    /* Handle capturing bracket */
1602    
1603    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1251  for (;;) Line 1607  for (;;)
1607      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1608      }      }
1609    
1610    /* 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
1611    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
1612    arrange to skip the extra bytes. */    two bytes of parameters. */
1613    
1614    else    else
1615      {      {
1616        switch(c)
1617          {
1618          case OP_TYPESTAR:
1619          case OP_TYPEMINSTAR:
1620          case OP_TYPEPLUS:
1621          case OP_TYPEMINPLUS:
1622          case OP_TYPEQUERY:
1623          case OP_TYPEMINQUERY:
1624          case OP_TYPEPOSSTAR:
1625          case OP_TYPEPOSPLUS:
1626          case OP_TYPEPOSQUERY:
1627          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1628          break;
1629    
1630          case OP_TYPEUPTO:
1631          case OP_TYPEMINUPTO:
1632          case OP_TYPEEXACT:
1633          case OP_TYPEPOSUPTO:
1634          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1635          break;
1636          }
1637    
1638        /* Add in the fixed length from the table */
1639    
1640      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1641    
1642      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1643      a multi-byte character. The length in the table is a minimum, so we have to
1644      arrange to skip the extra bytes. */
1645    
1646  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1647      if (utf8) switch(c)      if (utf8) switch(c)
1648        {        {
# Line 1279  for (;;) Line 1664  for (;;)
1664        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1665        break;        break;
1666        }        }
1667    #else
1668        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1669  #endif  #endif
1670      }      }
1671    }    }
# Line 1315  for (;;) Line 1702  for (;;)
1702    
1703    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1704    
1705    /* 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
1706    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
1707    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1708    
1709    else    else
1710      {      {
1711        switch(c)
1712          {
1713          case OP_TYPESTAR:
1714          case OP_TYPEMINSTAR:
1715          case OP_TYPEPLUS:
1716          case OP_TYPEMINPLUS:
1717          case OP_TYPEQUERY:
1718          case OP_TYPEMINQUERY:
1719          case OP_TYPEPOSSTAR:
1720          case OP_TYPEPOSPLUS:
1721          case OP_TYPEPOSQUERY:
1722          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1723          break;
1724    
1725          case OP_TYPEPOSUPTO:
1726          case OP_TYPEUPTO:
1727          case OP_TYPEMINUPTO:
1728          case OP_TYPEEXACT:
1729          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1730          break;
1731          }
1732    
1733        /* Add in the fixed length from the table */
1734    
1735      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1736    
1737        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1738        by a multi-byte character. The length in the table is a minimum, so we have
1739        to arrange to skip the extra bytes. */
1740    
1741  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1742      if (utf8) switch(c)      if (utf8) switch(c)
1743        {        {
# Line 1344  for (;;) Line 1759  for (;;)
1759        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1760        break;        break;
1761        }        }
1762    #else
1763        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1764  #endif  #endif
1765      }      }
1766    }    }
# Line 1359  for (;;) Line 1776  for (;;)
1776  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()
1777  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
1778  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1779  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
1780  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
1781    bracket whose current branch will already have been scanned.
1782    
1783  Arguments:  Arguments:
1784    code        points to start of search    code        points to start of search
# Line 1382  for (code = first_significant_code(code Line 1800  for (code = first_significant_code(code
1800    
1801    c = *code;    c = *code;
1802    
1803      /* Skip over forward assertions; the other assertions are skipped by
1804      first_significant_code() with a TRUE final argument. */
1805    
1806      if (c == OP_ASSERT)
1807        {
1808        do code += GET(code, 1); while (*code == OP_ALT);
1809        c = *code;
1810        continue;
1811        }
1812    
1813    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1814    
1815    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1816      {      {
1817      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1818      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1394  for (code = first_significant_code(code Line 1822  for (code = first_significant_code(code
1822    
1823    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1824    
1825    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1826      {      {
1827      BOOL empty_branch;      BOOL empty_branch;
1828      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1829    
1830      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1831        empty branch, so just skip over the conditional, because it could be empty.
1832        Otherwise, scan the individual branches of the group. */
1833    
1834      empty_branch = FALSE;      if (c == OP_COND && code[GET(code, 1)] != OP_ALT)
     do  
       {  
       if (!empty_branch && could_be_empty_branch(code, endcode, utf8))  
         empty_branch = TRUE;  
1835        code += GET(code, 1);        code += GET(code, 1);
1836        else
1837          {
1838          empty_branch = FALSE;
1839          do
1840            {
1841            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1842              empty_branch = TRUE;
1843            code += GET(code, 1);
1844            }
1845          while (*code == OP_ALT);
1846          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1847        }        }
1848      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1849      c = *code;      c = *code;
1850      continue;      continue;
1851      }      }
# Line 1418  for (code = first_significant_code(code Line 1854  for (code = first_significant_code(code
1854    
1855    switch (c)    switch (c)
1856      {      {
1857      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1858        cannot be represented just by a bit map. This includes negated single
1859        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1860        actual length is stored in the compiled code, so we must update "code"
1861        here. */
1862    
1863  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1864      case OP_XCLASS:      case OP_XCLASS:
1865      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1866      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1867  #endif  #endif
1868    
# Line 1466  for (code = first_significant_code(code Line 1906  for (code = first_significant_code(code
1906      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1907      case OP_WORDCHAR:      case OP_WORDCHAR:
1908      case OP_ANY:      case OP_ANY:
1909        case OP_ALLANY:
1910      case OP_ANYBYTE:      case OP_ANYBYTE:
1911      case OP_CHAR:      case OP_CHAR:
1912      case OP_CHARNC:      case OP_CHARNC:
# Line 1484  for (code = first_significant_code(code Line 1925  for (code = first_significant_code(code
1925      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1926      return FALSE;      return FALSE;
1927    
1928        /* These are going to continue, as they may be empty, but we have to
1929        fudge the length for the \p and \P cases. */
1930    
1931        case OP_TYPESTAR:
1932        case OP_TYPEMINSTAR:
1933        case OP_TYPEPOSSTAR:
1934        case OP_TYPEQUERY:
1935        case OP_TYPEMINQUERY:
1936        case OP_TYPEPOSQUERY:
1937        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1938        break;
1939    
1940        /* Same for these */
1941    
1942        case OP_TYPEUPTO:
1943        case OP_TYPEMINUPTO:
1944        case OP_TYPEPOSUPTO:
1945        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1946        break;
1947    
1948      /* End of branch */      /* End of branch */
1949    
1950      case OP_KET:      case OP_KET:
# Line 1502  for (code = first_significant_code(code Line 1963  for (code = first_significant_code(code
1963      case OP_QUERY:      case OP_QUERY:
1964      case OP_MINQUERY:      case OP_MINQUERY:
1965      case OP_POSQUERY:      case OP_POSQUERY:
1966        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1967        break;
1968    
1969      case OP_UPTO:      case OP_UPTO:
1970      case OP_MINUPTO:      case OP_MINUPTO:
1971      case OP_POSUPTO:      case OP_POSUPTO:
1972      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1973      break;      break;
1974  #endif  #endif
1975      }      }
# Line 1538  static BOOL Line 2002  static BOOL
2002  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2003    BOOL utf8)    BOOL utf8)
2004  {  {
2005  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2006    {    {
2007    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8))
2008        return FALSE;
2009    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2010    }    }
2011  return TRUE;  return TRUE;
# Line 1553  return TRUE; Line 2018  return TRUE;
2018  *************************************************/  *************************************************/
2019    
2020  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2021  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
2022  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2023  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2024    
2025    Originally, this function only recognized a sequence of letters between the
2026    terminators, but it seems that Perl recognizes any sequence of characters,
2027    though of course unknown POSIX names are subsequently rejected. Perl gives an
2028    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2029    didn't consider this to be a POSIX class. Likewise for [:1234:].
2030    
2031    The problem in trying to be exactly like Perl is in the handling of escapes. We
2032    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2033    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2034    below handles the special case of \], but does not try to do any other escape
2035    processing. This makes it different from Perl for cases such as [:l\ower:]
2036    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2037    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2038    I think.
2039    
2040  Argument:  Arguments:
2041    ptr      pointer to the initial [    ptr      pointer to the initial [
2042    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2043    
2044  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2045  */  */
2046    
2047  static BOOL  static BOOL
2048  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2049  {  {
2050  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2051  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2052  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2053    {    {
2054    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2055    return TRUE;      {
2056        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2057        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2058          {
2059          *endptr = ptr;
2060          return TRUE;
2061          }
2062        }
2063    }    }
2064  return FALSE;  return FALSE;
2065  }  }
# Line 1600  Returns:     a value representing the na Line 2084  Returns:     a value representing the na
2084  static int  static int
2085  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2086  {  {
2087    const char *pn = posix_names;
2088  register int yield = 0;  register int yield = 0;
2089  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2090    {    {
2091    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2092      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2093      pn += posix_name_lengths[yield] + 1;
2094    yield++;    yield++;
2095    }    }
2096  return -1;  return -1;
# Line 1619  return -1; Line 2105  return -1;
2105  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2106  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2107  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
2108  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
2109  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
2110  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
2111  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
2112  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2113    OP_END.
2114    
2115  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2116  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 1646  adjust_recurse(uschar *group, int adjust Line 2133  adjust_recurse(uschar *group, int adjust
2133    uschar *save_hwm)    uschar *save_hwm)
2134  {  {
2135  uschar *ptr = group;  uschar *ptr = group;
2136    
2137  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2138    {    {
2139    int offset;    int offset;
# Line 1757  get_othercase_range(unsigned int *cptr, Line 2245  get_othercase_range(unsigned int *cptr,
2245  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2246    
2247  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2248    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2249    
2250  if (c > d) return FALSE;  if (c > d) return FALSE;
2251    
# Line 1766  next = othercase + 1; Line 2254  next = othercase + 1;
2254    
2255  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2256    {    {
2257    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2258    next++;    next++;
2259    }    }
2260    
# Line 1812  if ((options & PCRE_EXTENDED) != 0) Line 2300  if ((options & PCRE_EXTENDED) != 0)
2300    for (;;)    for (;;)
2301      {      {
2302      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2303      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2304        {        {
2305        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2306          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1824  if ((options & PCRE_EXTENDED) != 0) Line 2312  if ((options & PCRE_EXTENDED) != 0)
2312  /* 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
2313  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2314    
2315  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2316    {    {
2317    int temperrorcode = 0;    int temperrorcode = 0;
2318    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1849  if ((options & PCRE_EXTENDED) != 0) Line 2337  if ((options & PCRE_EXTENDED) != 0)
2337    for (;;)    for (;;)
2338      {      {
2339      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2340      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2341        {        {
2342        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2343          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1860  if ((options & PCRE_EXTENDED) != 0) Line 2348  if ((options & PCRE_EXTENDED) != 0)
2348    
2349  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2350    
2351  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2352    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2353        return FALSE;
2354    
2355  /* Now compare the next item with the previous opcode. If the previous is a  /* Now compare the next item with the previous opcode. If the previous is a
2356  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 1876  if (next >= 0) switch(op_code) Line 2365  if (next >= 0) switch(op_code)
2365    case OP_CHAR:    case OP_CHAR:
2366  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2367    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2368    #else
2369      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2370  #endif  #endif
2371    return item != next;    return item != next;
2372    
# Line 1894  if (next >= 0) switch(op_code) Line 2385  if (next >= 0) switch(op_code)
2385      unsigned int othercase;      unsigned int othercase;
2386      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2387  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2388      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2389  #else  #else
2390      othercase = NOTACHAR;      othercase = NOTACHAR;
2391  #endif  #endif
# Line 1907  if (next >= 0) switch(op_code) Line 2398  if (next >= 0) switch(op_code)
2398    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2399    
2400    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2401    if (item == next) return TRUE;    if (item == next) return TRUE;
2402    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2403  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1916  if (next >= 0) switch(op_code) Line 2406  if (next >= 0) switch(op_code)
2406      unsigned int othercase;      unsigned int othercase;
2407      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2408  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2409      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2410  #else  #else
2411      othercase = NOTACHAR;      othercase = NOTACHAR;
2412  #endif  #endif
# Line 1944  if (next >= 0) switch(op_code) Line 2434  if (next >= 0) switch(op_code)
2434    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2435    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2436    
2437      case OP_HSPACE:
2438      case OP_NOT_HSPACE:
2439      switch(next)
2440        {
2441        case 0x09:
2442        case 0x20:
2443        case 0xa0:
2444        case 0x1680:
2445        case 0x180e:
2446        case 0x2000:
2447        case 0x2001:
2448        case 0x2002:
2449        case 0x2003:
2450        case 0x2004:
2451        case 0x2005:
2452        case 0x2006:
2453        case 0x2007:
2454        case 0x2008:
2455        case 0x2009:
2456        case 0x200A:
2457        case 0x202f:
2458        case 0x205f:
2459        case 0x3000:
2460        return op_code != OP_HSPACE;
2461        default:
2462        return op_code == OP_HSPACE;
2463        }
2464    
2465      case OP_VSPACE:
2466      case OP_NOT_VSPACE:
2467      switch(next)
2468        {
2469        case 0x0a:
2470        case 0x0b:
2471        case 0x0c:
2472        case 0x0d:
2473        case 0x85:
2474        case 0x2028:
2475        case 0x2029:
2476        return op_code != OP_VSPACE;
2477        default:
2478        return op_code == OP_VSPACE;
2479        }
2480    
2481    default:    default:
2482    return FALSE;    return FALSE;
2483    }    }
# Line 1978  switch(op_code) Line 2512  switch(op_code)
2512      case ESC_W:      case ESC_W:
2513      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2514    
2515        case ESC_h:
2516        case ESC_H:
2517        switch(item)
2518          {
2519          case 0x09:
2520          case 0x20:
2521          case 0xa0:
2522          case 0x1680:
2523          case 0x180e:
2524          case 0x2000:
2525          case 0x2001:
2526          case 0x2002:
2527          case 0x2003:
2528          case 0x2004:
2529          case 0x2005:
2530          case 0x2006:
2531          case 0x2007:
2532          case 0x2008:
2533          case 0x2009:
2534          case 0x200A:
2535          case 0x202f:
2536          case 0x205f:
2537          case 0x3000:
2538          return -next != ESC_h;
2539          default:
2540          return -next == ESC_h;
2541          }
2542    
2543        case ESC_v:
2544        case ESC_V:
2545        switch(item)
2546          {
2547          case 0x0a:
2548          case 0x0b:
2549          case 0x0c:
2550          case 0x0d:
2551          case 0x85:
2552          case 0x2028:
2553          case 0x2029:
2554          return -next != ESC_v;
2555          default:
2556          return -next == ESC_v;
2557          }
2558    
2559      default:      default:
2560      return FALSE;      return FALSE;
2561      }      }
2562    
2563    case OP_DIGIT:    case OP_DIGIT:
2564    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2565             next == -ESC_h || next == -ESC_v;
2566    
2567    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2568    return next == -ESC_d;    return next == -ESC_d;
# Line 1992  switch(op_code) Line 2571  switch(op_code)
2571    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2572    
2573    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2574    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2575    
2576      case OP_HSPACE:
2577      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2578    
2579      case OP_NOT_HSPACE:
2580      return next == -ESC_h;
2581    
2582      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2583      case OP_VSPACE:
2584      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2585    
2586      case OP_NOT_VSPACE:
2587      return next == -ESC_v;
2588    
2589    case OP_WORDCHAR:    case OP_WORDCHAR:
2590    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2591    
2592    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2593    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2068  uschar classbits[32]; Line 2660  uschar classbits[32];
2660  BOOL class_utf8;  BOOL class_utf8;
2661  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2662  uschar *class_utf8data;  uschar *class_utf8data;
2663    uschar *class_utf8data_base;
2664  uschar utf8_char[6];  uschar utf8_char[6];
2665  #else  #else
2666  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2667  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2668  #endif  #endif
2669    
2670  #ifdef DEBUG  #ifdef PCRE_DEBUG
2671  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2672  #endif  #endif
2673    
# Line 2107  req_caseopt = ((options & PCRE_CASELESS) Line 2700  req_caseopt = ((options & PCRE_CASELESS)
2700  for (;; ptr++)  for (;; ptr++)
2701    {    {
2702    BOOL negate_class;    BOOL negate_class;
2703      BOOL should_flip_negation;
2704    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2705    BOOL is_quantifier;    BOOL is_quantifier;
2706    BOOL is_recurse;    BOOL is_recurse;
2707      BOOL reset_bracount;
2708    int class_charcount;    int class_charcount;
2709    int class_lastchar;    int class_lastchar;
2710    int newoptions;    int newoptions;
# Line 2131  for (;; ptr++) Line 2726  for (;; ptr++)
2726    
2727    if (lengthptr != NULL)    if (lengthptr != NULL)
2728      {      {
2729  #ifdef DEBUG  #ifdef PCRE_DEBUG
2730      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2731  #endif  #endif
2732      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
# Line 2147  for (;; ptr++) Line 2742  for (;; ptr++)
2742      */      */
2743    
2744      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2745    
2746        /* Paranoid check for integer overflow */
2747    
2748        if (OFLOW_MAX - *lengthptr < code - last_code)
2749          {
2750          *errorcodeptr = ERR20;
2751          goto FAILED;
2752          }
2753    
2754      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2755      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));
2756    
# Line 2184  for (;; ptr++) Line 2788  for (;; ptr++)
2788    
2789    if (inescq && c != 0)    if (inescq && c != 0)
2790      {      {
2791      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2792        {        {
2793        inescq = FALSE;        inescq = FALSE;
2794        ptr++;        ptr++;
# Line 2210  for (;; ptr++) Line 2814  for (;; ptr++)
2814    /* 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
2815    a quantifier. */    a quantifier. */
2816    
2817    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2818      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2819        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2820    
2821    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2822         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2226  for (;; ptr++) Line 2831  for (;; ptr++)
2831    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2832      {      {
2833      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2834      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2835        {        {
2836        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2837          {          {
# Line 2251  for (;; ptr++) Line 2856  for (;; ptr++)
2856      {      {
2857      /* ===================================================================*/      /* ===================================================================*/
2858      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2859      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2860      case ')':      case CHAR_RIGHT_PARENTHESIS:
2861      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2862      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2863      *codeptr = code;      *codeptr = code;
2864      *ptrptr = ptr;      *ptrptr = ptr;
2865      if (lengthptr != NULL)      if (lengthptr != NULL)
2866        {        {
2867          if (OFLOW_MAX - *lengthptr < code - last_code)
2868            {
2869            *errorcodeptr = ERR20;
2870            goto FAILED;
2871            }
2872        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2873        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2874        }        }
# Line 2269  for (;; ptr++) Line 2879  for (;; ptr++)
2879      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2880      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2881    
2882      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2883      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2884        {        {
2885        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2278  for (;; ptr++) Line 2888  for (;; ptr++)
2888      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2889      break;      break;
2890    
2891      case '$':      case CHAR_DOLLAR_SIGN:
2892      previous = NULL;      previous = NULL;
2893      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2894      break;      break;
# Line 2286  for (;; ptr++) Line 2896  for (;; ptr++)
2896      /* 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
2897      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2898    
2899      case '.':      case CHAR_DOT:
2900      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2901      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2902      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2903      previous = code;      previous = code;
2904      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2905      break;      break;
2906    
2907    
# Line 2306  for (;; ptr++) Line 2916  for (;; ptr++)
2916      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,
2917      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2918      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.
     */  
2919    
2920      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2921        default (Perl) mode, it is treated as a data character. */
2922    
2923        case CHAR_RIGHT_SQUARE_BRACKET:
2924        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2925          {
2926          *errorcodeptr = ERR64;
2927          goto FAILED;
2928          }
2929        goto NORMAL_CHAR;
2930    
2931        case CHAR_LEFT_SQUARE_BRACKET:
2932      previous = code;      previous = code;
2933    
2934      /* 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
2935      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. */
2936    
2937      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2938          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2939            check_posix_syntax(ptr, &tempptr))
2940        {        {
2941        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2942        goto FAILED;        goto FAILED;
2943        }        }
2944    
2945      /* 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,
2946        if the first few characters (either before or after ^) are \Q\E or \E we
2947        skip them too. This makes for compatibility with Perl. */
2948    
2949      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2950        for (;;)
2951        {        {
       negate_class = TRUE;  
2952        c = *(++ptr);        c = *(++ptr);
2953          if (c == CHAR_BACKSLASH)
2954            {
2955            if (ptr[1] == CHAR_E)
2956              ptr++;
2957            else if (strncmp((const char *)ptr+1,
2958                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2959              ptr += 3;
2960            else
2961              break;
2962            }
2963          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2964            negate_class = TRUE;
2965          else break;
2966        }        }
2967      else  
2968        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2969        an initial ']' is taken as a data character -- the code below handles
2970        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2971        [^] must match any character, so generate OP_ALLANY. */
2972    
2973        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2974            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2975        {        {
2976        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2977          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2978          zerofirstbyte = firstbyte;
2979          break;
2980        }        }
2981    
2982        /* If a class contains a negative special such as \S, we need to flip the
2983        negation flag at the end, so that support for characters > 255 works
2984        correctly (they are all included in the class). */
2985    
2986        should_flip_negation = FALSE;
2987    
2988      /* 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
2989      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
2990      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2350  for (;; ptr++) Line 3002  for (;; ptr++)
3002  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3003      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3004      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3005        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3006  #endif  #endif
3007    
3008      /* 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 2365  for (;; ptr++) Line 3018  for (;; ptr++)
3018          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3019          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3020          }          }
3021    
3022          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3023          data and reset the pointer. This is so that very large classes that
3024          contain a zillion UTF-8 characters no longer overwrite the work space
3025          (which is on the stack). */
3026    
3027          if (lengthptr != NULL)
3028            {
3029            *lengthptr += class_utf8data - class_utf8data_base;
3030            class_utf8data = class_utf8data_base;
3031            }
3032    
3033  #endif  #endif
3034    
3035        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3036    
3037        if (inescq)        if (inescq)
3038          {          {
3039          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3040            {            {
3041            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3042            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2386  for (;; ptr++) Line 3051  for (;; ptr++)
3051        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3052        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3053    
3054        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3055            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3056            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3057          {          {
3058          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3059          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3060          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3061          uschar pbits[32];          uschar pbits[32];
3062    
3063          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3064            {            {
3065            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3066            goto FAILED;            goto FAILED;
3067            }            }
3068    
3069          ptr += 2;          ptr += 2;
3070          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3071            {            {
3072            local_negate = TRUE;            local_negate = TRUE;
3073              should_flip_negation = TRUE;  /* Note negative special */
3074            ptr++;            ptr++;
3075            }            }
3076    
# Line 2471  for (;; ptr++) Line 3137  for (;; ptr++)
3137        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
3138        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
3139        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3140        to or into the one we are building. We assume they have more than one        to 'or' into the one we are building. We assume they have more than one
3141        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3142    
3143        if (c == '\\')        if (c == CHAR_BACKSLASH)
3144          {          {
3145          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3146          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3147    
3148          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 */
3149          else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */          else if (-c == ESC_X) c = CHAR_X;   /* \X is literal X in a class */
3150          else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */          else if (-c == ESC_R) c = CHAR_R;   /* \R is literal R in a class */
3151          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3152            {            {
3153            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3154              {              {
3155              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3156              }              }
3157            else inescq = TRUE;            else inescq = TRUE;
3158            continue;            continue;
3159            }            }
3160            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3161    
3162          if (c < 0)          if (c < 0)
3163            {            {
# Line 2506  for (;; ptr++) Line 3173  for (;; ptr++)
3173              continue;              continue;
3174    
3175              case ESC_D:              case ESC_D:
3176                should_flip_negation = TRUE;
3177              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3178              continue;              continue;
3179    
# Line 2514  for (;; ptr++) Line 3182  for (;; ptr++)
3182              continue;              continue;
3183    
3184              case ESC_W:              case ESC_W:
3185                should_flip_negation = TRUE;
3186              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3187              continue;              continue;
3188    
# Line 2523  for (;; ptr++) Line 3192  for (;; ptr++)
3192              continue;              continue;
3193    
3194              case ESC_S:              case ESC_S:
3195                should_flip_negation = TRUE;
3196              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3197              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3198              continue;              continue;
3199    
3200              case ESC_E: /* Perl ignores an orphan \E */              default:    /* Not recognized; fall through */
3201                break;      /* Need "default" setting to stop compiler warning. */
3202                }
3203    
3204              /* In the pre-compile phase, just do the recognition. */
3205    
3206              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3207                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3208    
3209              /* We need to deal with \H, \h, \V, and \v in both phases because
3210              they use extra memory. */
3211    
3212              if (-c == ESC_h)
3213                {
3214                SETBIT(classbits, 0x09); /* VT */
3215                SETBIT(classbits, 0x20); /* SPACE */
3216                SETBIT(classbits, 0xa0); /* NSBP */
3217    #ifdef SUPPORT_UTF8
3218                if (utf8)
3219                  {
3220                  class_utf8 = TRUE;
3221                  *class_utf8data++ = XCL_SINGLE;
3222                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3223                  *class_utf8data++ = XCL_SINGLE;
3224                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3225                  *class_utf8data++ = XCL_RANGE;
3226                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3227                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3228                  *class_utf8data++ = XCL_SINGLE;
3229                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3230                  *class_utf8data++ = XCL_SINGLE;
3231                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3232                  *class_utf8data++ = XCL_SINGLE;
3233                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3234                  }
3235    #endif
3236                continue;
3237                }
3238    
3239              if (-c == ESC_H)
3240                {
3241                for (c = 0; c < 32; c++)
3242                  {
3243                  int x = 0xff;
3244                  switch (c)
3245                    {
3246                    case 0x09/8: x ^= 1 << (0x09%8); break;
3247                    case 0x20/8: x ^= 1 << (0x20%8); break;
3248                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3249                    default: break;
3250                    }
3251                  classbits[c] |= x;
3252                  }
3253    
3254    #ifdef SUPPORT_UTF8
3255                if (utf8)
3256                  {
3257                  class_utf8 = TRUE;
3258                  *class_utf8data++ = XCL_RANGE;
3259                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3260                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3261                  *class_utf8data++ = XCL_RANGE;
3262                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3263                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3264                  *class_utf8data++ = XCL_RANGE;
3265                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3266                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3267                  *class_utf8data++ = XCL_RANGE;
3268                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3269                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3270                  *class_utf8data++ = XCL_RANGE;
3271                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3272                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3273                  *class_utf8data++ = XCL_RANGE;
3274                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3275                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3276                  *class_utf8data++ = XCL_RANGE;
3277                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3278                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3279                  }
3280    #endif
3281                continue;
3282                }
3283    
3284              if (-c == ESC_v)
3285                {
3286                SETBIT(classbits, 0x0a); /* LF */
3287                SETBIT(classbits, 0x0b); /* VT */
3288                SETBIT(classbits, 0x0c); /* FF */
3289                SETBIT(classbits, 0x0d); /* CR */
3290                SETBIT(classbits, 0x85); /* NEL */
3291    #ifdef SUPPORT_UTF8
3292                if (utf8)
3293                  {
3294                  class_utf8 = TRUE;
3295                  *class_utf8data++ = XCL_RANGE;
3296                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3297                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3298                  }
3299    #endif
3300                continue;
3301                }
3302    
3303              if (-c == ESC_V)
3304                {
3305                for (c = 0; c < 32; c++)
3306                  {
3307                  int x = 0xff;
3308                  switch (c)
3309                    {
3310                    case 0x0a/8: x ^= 1 << (0x0a%8);
3311                                 x ^= 1 << (0x0b%8);
3312                                 x ^= 1 << (0x0c%8);
3313                                 x ^= 1 << (0x0d%8);
3314                                 break;
3315                    case 0x85/8: x ^= 1 << (0x85%8); break;
3316                    default: break;
3317                    }
3318                  classbits[c] |= x;
3319                  }
3320    
3321    #ifdef SUPPORT_UTF8
3322                if (utf8)
3323                  {
3324                  class_utf8 = TRUE;
3325                  *class_utf8data++ = XCL_RANGE;
3326                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3327                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3328                  *class_utf8data++ = XCL_RANGE;
3329                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3330                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3331                  }
3332    #endif
3333              continue;              continue;
   
             default:    /* Not recognized; fall through */  
             break;      /* Need "default" setting to stop compiler warning. */  
3334              }              }
3335    
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
3336            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3337    
3338  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2582  for (;; ptr++) Line 3376  for (;; ptr++)
3376        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3377    
3378        CHECK_RANGE:        CHECK_RANGE:
3379        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3380          {          {
3381          inescq = FALSE;          inescq = FALSE;
3382          ptr += 2;          ptr += 2;
# Line 2590  for (;; ptr++) Line 3384  for (;; ptr++)
3384    
3385        oldptr = ptr;        oldptr = ptr;
3386    
3387        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3388    
3389          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3390    
3391          /* Check for range */
3392    
3393          if (!inescq && ptr[1] == CHAR_MINUS)
3394          {          {
3395          int d;          int d;
3396          ptr += 2;          ptr += 2;
3397          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3398    
3399          /* 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
3400          mode. */          mode. */
3401    
3402          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3403            {            {
3404            ptr += 2;            ptr += 2;
3405            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3406                { ptr += 2; continue; }
3407            inescq = TRUE;            inescq = TRUE;
3408            break;            break;
3409            }            }
3410    
3411          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3412            {            {
3413            ptr = oldptr;            ptr = oldptr;
3414            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2626  for (;; ptr++) Line 3427  for (;; ptr++)
3427          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
3428          in such circumstances. */          in such circumstances. */
3429    
3430          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3431            {            {
3432            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3433            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3434    
3435            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; \X is literal X; \R is literal R; any other
3436            special means the '-' was literal */            special means the '-' was literal */
3437    
3438            if (d < 0)            if (d < 0)
3439              {              {
3440              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3441              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3442              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3443                {                {
3444                ptr = oldptr;                ptr = oldptr;
3445                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2657  for (;; ptr++) Line 3458  for (;; ptr++)
3458    
3459          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3460    
3461            /* Remember \r or \n */
3462    
3463            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3464    
3465          /* 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
3466          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3467          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2679  for (;; ptr++) Line 3484  for (;; ptr++)
3484              unsigned int origd = d;              unsigned int origd = d;
3485              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3486                {                {
3487                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3488                      ocd <= (unsigned int)d)
3489                    continue;                          /* Skip embedded ranges */
3490    
3491                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3492                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3493                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3494                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3495                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3496                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3497                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3498                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3499                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3500                  d = ocd;                  d = ocd;
3501                  continue;                  continue;
# Line 2771  for (;; ptr++) Line 3580  for (;; ptr++)
3580          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3581            {            {
3582            unsigned int othercase;            unsigned int othercase;
3583            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3584              {              {
3585              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3586              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2798  for (;; ptr++) Line 3607  for (;; ptr++)
3607    
3608      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3609    
3610      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3611    
3612      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3613        {        {
# Line 2806  for (;; ptr++) Line 3615  for (;; ptr++)
3615        goto FAILED;        goto FAILED;
3616        }        }
3617    
3618    
3619    /* This code has been disabled because it would mean that \s counts as
3620    an explicit \r or \n reference, and that's not really what is wanted. Now
3621    we set the flag only if there is a literal "\r" or "\n" in the class. */
3622    
3623    #if 0
3624        /* Remember whether \r or \n are in this class */
3625    
3626        if (negate_class)
3627          {
3628          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3629          }
3630        else
3631          {
3632          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3633          }
3634    #endif
3635    
3636    
3637      /* 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
3638      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
3639      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
3640      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3641      single-bytes only. This is an historical hangover. Maybe one day we can  
3642      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3643        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3644        operate on single-bytes only. This is an historical hangover. Maybe one day
3645        we can tidy these opcodes to handle multi-byte characters.
3646    
3647      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
3648      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 2821  for (;; ptr++) Line 3652  for (;; ptr++)
3652      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3653    
3654  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3655      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3656            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3657  #else  #else
3658      if (class_charcount == 1)      if (class_charcount == 1)
3659  #endif  #endif
# Line 2867  for (;; ptr++) Line 3696  for (;; ptr++)
3696      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3697    
3698      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3699      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3700      we can omit the bitmap in the actual compiled code. */      such as \S in the class, because in that case all characters > 255 are in
3701        the class, so any that were explicitly given as well can be ignored. If
3702        (when there are explicit characters > 255 that must be listed) there are no
3703        characters < 256, we can omit the bitmap in the actual compiled code. */
3704    
3705  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3706      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3707        {        {
3708        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3709        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2897  for (;; ptr++) Line 3729  for (;; ptr++)
3729        }        }
3730  #endif  #endif
3731    
3732      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, set the opcode to OP_CLASS or
3733      and copy it into the code vector. If this is the first thing in the branch,      OP_NCLASS, depending on whether the whole class was negated and whether
3734      there can be no first char setting, whatever the repeat count. Any reqbyte      there were negative specials such as \S in the class. Then copy the 32-byte
3735      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3736    
3737        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3738      if (negate_class)      if (negate_class)
3739        {        {
       *code++ = OP_NCLASS;  
3740        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3741          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3742        }        }
3743      else      else
3744        {        {
       *code++ = OP_CLASS;  
3745        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3746        }        }
3747      code += 32;      code += 32;
# Line 2921  for (;; ptr++) Line 3752  for (;; ptr++)
3752      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3753      has been tested above. */      has been tested above. */
3754    
3755      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3756      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3757      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3758      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3759      goto REPEAT;      goto REPEAT;
3760    
3761      case '*':      case CHAR_ASTERISK:
3762      repeat_min = 0;      repeat_min = 0;
3763      repeat_max = -1;      repeat_max = -1;
3764      goto REPEAT;      goto REPEAT;
3765    
3766      case '+':      case CHAR_PLUS:
3767      repeat_min = 1;      repeat_min = 1;
3768      repeat_max = -1;      repeat_max = -1;
3769      goto REPEAT;      goto REPEAT;
3770    
3771      case '?':      case CHAR_QUESTION_MARK:
3772      repeat_min = 0;      repeat_min = 0;
3773      repeat_max = 1;      repeat_max = 1;
3774    
# Line 2972  for (;; ptr++) Line 3803  for (;; ptr++)
3803      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
3804      repeat type to the non-default. */      repeat type to the non-default. */
3805    
3806      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3807        {        {
3808        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3809        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3810        ptr++;        ptr++;
3811        }        }
3812      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3813        {        {
3814        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3815        ptr++;        ptr++;
# Line 3093  for (;; ptr++) Line 3924  for (;; ptr++)
3924    
3925        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3926    
3927          /*--------------------------------------------------------------------*/
3928          /* This code is obsolete from release 8.00; the restriction was finally
3929          removed: */
3930    
3931        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3932        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3933    
3934        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3935          /*--------------------------------------------------------------------*/
3936    
3937        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3938    
# Line 3243  for (;; ptr++) Line 4079  for (;; ptr++)
4079          goto END_REPEAT;          goto END_REPEAT;
4080          }          }
4081    
4082          /*--------------------------------------------------------------------*/
4083          /* This code is obsolete from release 8.00; the restriction was finally
4084          removed: */
4085    
4086        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4087        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4088    
4089        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4090          /*--------------------------------------------------------------------*/
4091    
4092        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4093          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3282  for (;; ptr++) Line 4123  for (;; ptr++)
4123          goto FAILED;          goto FAILED;
4124          }          }
4125    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4126        /* 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
4127        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
4128        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 3312  for (;; ptr++) Line 4145  for (;; ptr++)
4145    
4146        if (repeat_min == 0)        if (repeat_min == 0)
4147          {          {
4148          /* 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
4149          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4150    
4151          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4152          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4153          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4154          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4155          the start of the whole regex. Temporarily terminate the pattern while          **   }
4156          doing this. */  
4157            However, that fails when a group is referenced as a subroutine from
4158            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4159            so that it is skipped on execution. As we don't have a list of which
4160            groups are referenced, we cannot do this selectively.
4161    
4162            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4163            and do no more at this point. However, we do need to adjust any
4164            OP_RECURSE calls inside the group that refer to the group itself or any
4165            internal or forward referenced group, because the offset is from the
4166            start of the whole regex. Temporarily terminate the pattern while doing
4167            this. */
4168    
4169          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4170            {            {
4171            *code = OP_END;            *code = OP_END;
4172            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4173            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4174            code++;            code++;
4175              if (repeat_max == 0)
4176                {
4177                *previous++ = OP_SKIPZERO;
4178                goto END_REPEAT;
4179                }
4180            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4181            }            }
4182    
# Line 3378  for (;; ptr++) Line 4221  for (;; ptr++)
4221          if (repeat_min > 1)          if (repeat_min > 1)
4222            {            {
4223            /* 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
4224            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4225              potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4226              integer type when available, otherwise double. */
4227    
4228            if (lengthptr != NULL)            if (lengthptr != NULL)
4229              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4230                int delta = (repeat_min - 1)*length_prevgroup;
4231                if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4232                      (INT64_OR_DOUBLE)length_prevgroup >
4233                        (INT64_OR_DOUBLE)INT_MAX ||
4234                    OFLOW_MAX - *lengthptr < delta)
4235                  {
4236                  *errorcodeptr = ERR20;
4237                  goto FAILED;
4238                  }
4239                *lengthptr += delta;
4240                }
4241    
4242            /* This is compiling for real */            /* This is compiling for real */
4243    
# Line 3419  for (;; ptr++) Line 4275  for (;; ptr++)
4275          /* 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
4276          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
4277          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
4278          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4279            paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4280            a 64-bit integer type when available, otherwise double. */
4281    
4282          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4283            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4284              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4285                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4286              if ((INT64_OR_DOUBLE)repeat_max *
4287                    (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4288                      > (INT64_OR_DOUBLE)INT_MAX ||
4289                  OFLOW_MAX - *lengthptr < delta)
4290                {
4291                *errorcodeptr = ERR20;
4292                goto FAILED;
4293                }
4294              *lengthptr += delta;
4295              }
4296    
4297          /* This is compiling for real */          /* This is compiling for real */
4298    
# Line 3505  for (;; ptr++) Line 4374  for (;; ptr++)
4374          }          }
4375        }        }
4376    
4377        /* If previous is OP_FAIL, it was generated by an empty class [] in
4378        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4379        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4380        error above. We can just ignore the repeat in JS case. */
4381    
4382        else if (*previous == OP_FAIL) goto END_REPEAT;
4383    
4384      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4385    
4386      else      else
# Line 3529  for (;; ptr++) Line 4405  for (;; ptr++)
4405      if (possessive_quantifier)      if (possessive_quantifier)
4406        {        {
4407        int len;        int len;
4408        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4409            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4410            tempcode += _pcre_OP_lengths[*tempcode] +
4411              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4412    
4413          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4414            {
4415          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4416    #ifdef SUPPORT_UTF8
4417            if (utf8 && tempcode[-1] >= 0xc0)
4418              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4419    #endif
4420            }
4421    
4422        len = code - tempcode;        len = code - tempcode;
4423        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4424          {          {
# Line 3550  for (;; ptr++) Line 4437  for (;; ptr++)
4437          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4438          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4439    
4440            /* Because we are moving code along, we must ensure that any
4441            pending recursive references are updated. */
4442    
4443          default:          default:
4444            *code = OP_END;
4445            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4446          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4447          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4448          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3575  for (;; ptr++) Line 4467  for (;; ptr++)
4467      /* ===================================================================*/      /* ===================================================================*/
4468      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4469      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4470      parenthesis forms. First deal with the specials; all are introduced by ?,      parenthesis forms.  */
     and the appearance of any of them means that this is not a capturing  
     group. */  
4471    
4472      case '(':      case CHAR_LEFT_PARENTHESIS:
4473      newoptions = options;      newoptions = options;
4474      skipbytes = 0;      skipbytes = 0;
4475      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4476      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4477        reset_bracount = FALSE;
4478    
4479        /* First deal with various "verbs" that can be introduced by '*'. */
4480    
4481        if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4482          {
4483          int i, namelen;
4484          const char *vn = verbnames;
4485          const uschar *name = ++ptr;
4486          previous = NULL;
4487          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4488          if (*ptr == CHAR_COLON)
4489            {
4490            *errorcodeptr = ERR59;   /* Not supported */
4491            goto FAILED;
4492            }
4493          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4494            {
4495            *errorcodeptr = ERR60;
4496            goto FAILED;
4497            }
4498          namelen = ptr - name;
4499          for (i = 0; i < verbcount; i++)
4500            {
4501            if (namelen == verbs[i].len &&
4502                strncmp((char *)name, vn, namelen) == 0)
4503              {
4504              /* Check for open captures before ACCEPT */
4505    
4506              if (verbs[i].op == OP_ACCEPT)
4507                {
4508                open_capitem *oc;
4509                cd->had_accept = TRUE;
4510                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4511                  {
4512                  *code++ = OP_CLOSE;
4513                  PUT2INC(code, 0, oc->number);
4514                  }
4515                }
4516              *code++ = verbs[i].op;
4517              break;
4518              }
4519            vn += verbs[i].len + 1;
4520            }
4521          if (i < verbcount) continue;
4522          *errorcodeptr = ERR60;
4523          goto FAILED;
4524          }
4525    
4526        /* Deal with the extended parentheses; all are introduced by '?', and the
4527        appearance of any of them means that this is not a capturing group. */
4528    
4529      if (*(++ptr) == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4530        {        {
4531        int i, set, unset, namelen;        int i, set, unset, namelen;
4532        int *optset;        int *optset;
# Line 3594  for (;; ptr++) Line 4535  for (;; ptr++)
4535    
4536        switch (*(++ptr))        switch (*(++ptr))
4537          {          {
4538          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4539          ptr++;          ptr++;
4540          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4541          if (*ptr == 0)          if (*ptr == 0)
4542            {            {
4543            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3606  for (;; ptr++) Line 4547  for (;; ptr++)
4547    
4548    
4549          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4550          case ':':                 /* Non-capturing bracket */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4551            reset_bracount = TRUE;
4552            /* Fall through */
4553    
4554            /* ------------------------------------------------------------ */
4555            case CHAR_COLON:          /* Non-capturing bracket */
4556          bravalue = OP_BRA;          bravalue = OP_BRA;
4557          ptr++;          ptr++;
4558          break;          break;
4559    
4560    
4561          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4562          case '(':          case CHAR_LEFT_PARENTHESIS:
4563          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4564    
4565          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 3633  for (;; ptr++) Line 4579  for (;; ptr++)
4579          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4580          including assertions, are processed. */          including assertions, are processed. */
4581    
4582          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4583                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4584            break;            break;
4585    
4586          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
# Line 3645  for (;; ptr++) Line 4592  for (;; ptr++)
4592    
4593          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4594    
4595          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4596            {            {
4597            terminator = -1;            terminator = -1;
4598            ptr += 2;            ptr += 2;
# Line 3655  for (;; ptr++) Line 4602  for (;; ptr++)
4602          /* Check for a test for a named group's having been set, using the Perl          /* Check for a test for a named group's having been set, using the Perl
4603          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4604    
4605          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4606            {            {
4607            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4608            ptr++;            ptr++;
4609            }            }
4610          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4611            {            {
4612            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4613            ptr++;            ptr++;
4614            }            }
4615          else          else
4616            {            {
4617            terminator = 0;            terminator = 0;
4618            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4619            }            }
4620    
4621          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
# Line 3688  for (;; ptr++) Line 4635  for (;; ptr++)
4635            {            {
4636            if (recno >= 0)            if (recno >= 0)
4637              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4638                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4639            ptr++;            ptr++;
4640            }            }
4641          namelen = ptr - name;          namelen = ptr - name;
4642    
4643          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4644                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4645            {            {
4646            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4647            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3715  for (;; ptr++) Line 4663  for (;; ptr++)
4663              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4664              goto FAILED;              goto FAILED;
4665              }              }
4666            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4667                cd->bracount - recno + 1 : recno +cd->bracount;
4668              if (recno <= 0 || recno > cd->final_bracount)
4669              {              {
4670              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4671              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4672              }              }
           else recno += cd->bracount;  
4673            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4674            break;            break;
4675            }            }
4676    
4677          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4678          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4679            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4680            except they record that the reference was originally to a name. The
4681            information is used to check duplicate names. */
4682    
4683          slot = cd->name_table;          slot = cd->name_table;
4684          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 3745  for (;; ptr++) Line 4693  for (;; ptr++)
4693            {            {
4694            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4695            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4696              code[1+LINK_SIZE]++;
4697            }            }
4698    
4699          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4700    
4701          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4702                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4703            {            {
4704            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4705              code[1+LINK_SIZE]++;
4706            }            }
4707    
4708          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 3770  for (;; ptr++) Line 4720  for (;; ptr++)
4720          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4721          specific group number. */          specific group number. */
4722    
4723          else if (*name == 'R')          else if (*name == CHAR_R)
4724            {            {
4725            recno = 0;            recno = 0;
4726            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 3780  for (;; ptr++) Line 4730  for (;; ptr++)
4730                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4731                goto FAILED;                goto FAILED;
4732                }                }
4733              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4734              }              }
4735            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4736            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 3790  for (;; ptr++) Line 4740  for (;; ptr++)
4740          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4741          false. */          false. */
4742    
4743          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4744            {            {
4745            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4746            skipbytes = 1;            skipbytes = 1;
4747            }            }
4748    
4749          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4750            in the second pass here, so final_bracount is set. */
4751    
4752          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4753            {            {
4754            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4755            }            }
# Line 3814  for (;; ptr++) Line 4765  for (;; ptr++)
4765    
4766    
4767          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4768          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4769          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4770          ptr++;          ptr++;
4771          break;          break;
4772    
4773    
4774          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4775          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4776          ptr++;          ptr++;
4777            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4778              {
4779              *code++ = OP_FAIL;
4780              previous = NULL;
4781              continue;
4782              }
4783            bravalue = OP_ASSERT_NOT;
4784          break;          break;
4785    
4786    
4787          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4788          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4789          switch (ptr[1])          switch (ptr[1])
4790            {            {
4791            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4792            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4793            ptr += 2;            ptr += 2;
4794            break;            break;
4795    
4796            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4797            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4798            ptr += 2;            ptr += 2;
4799            break;            break;
# Line 3851  for (;; ptr++) Line 4808  for (;; ptr++)
4808    
4809    
4810          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4811          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4812          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4813          ptr++;          ptr++;
4814          break;          break;
4815    
4816    
4817          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4818          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4819          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4820          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4821          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4822            {            {
4823            int n = 0;            int n = 0;
4824            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4825              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4826            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4827              {              {
4828              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4829              goto FAILED;              goto FAILED;
# Line 3886  for (;; ptr++) Line 4843  for (;; ptr++)
4843    
4844    
4845          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4846          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4847          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4848                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4849            {            {
4850            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4851            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4852            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4853            }            }
4854          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4855            {            {
4856            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4857            goto FAILED;            goto FAILED;
# Line 3903  for (;; ptr++) Line 4861  for (;; ptr++)
4861    
4862          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4863          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4864          case '\'':          case CHAR_APOSTROPHE:
4865            {            {
4866            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4867                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4868            name = ++ptr;            name = ++ptr;
4869    
4870            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 3936  for (;; ptr++) Line 4895  for (;; ptr++)
4895                }                }
4896              }              }
4897    
4898            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4899              alphabetical order. Duplicate names for different numbers are
4900              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4901              number are always OK. (An existing number can be re-used if (?|
4902              appears in the pattern.) In either event, a duplicate name results in
4903              a duplicate entry in the table, even if the number is the same. This
4904              is because the number of names, and hence the table size, is computed
4905              in the pre-compile, and it affects various numbers and pointers which
4906              would all have to be modified, and the compiled code moved down, if
4907              duplicates with the same number were omitted from the table. This
4908              doesn't seem worth the hassle. However, *different* names for the
4909              same number are not permitted. */
4910    
4911            else            else
4912              {              {
4913                BOOL dupname = FALSE;
4914              slot = cd->name_table;              slot = cd->name_table;
4915    
4916              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4917                {                {
4918                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 3948  for (;; ptr++) Line 4920  for (;; ptr++)
4920                  {                  {
4921                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4922                    {                    {
4923                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4924                          (options & PCRE_DUPNAMES) == 0)
4925                      {                      {
4926                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4927                      goto FAILED;                      goto FAILED;
4928                      }                      }
4929                      else dupname = TRUE;
4930                    }                    }
4931                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4932                  }                  }
4933    
4934                  /* Make space in the table and break the loop for an earlier
4935                  name. For a duplicate or later name, carry on. We do this for
4936                  duplicates so that in the simple case (when ?(| is not used) they
4937                  are in order of their numbers. */
4938    
4939                if (crc < 0)                if (crc < 0)
4940                  {                  {
4941                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4942                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4943                  break;                  break;
4944                  }                  }
4945    
4946                  /* Continue the loop for a later or duplicate name */
4947    
4948                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4949                }                }
4950    
4951                /* For non-duplicate names, check for a duplicate number before
4952                adding the new name. */
4953    
4954                if (!dupname)
4955                  {
4956                  uschar *cslot = cd->name_table;
4957                  for (i = 0; i < cd->names_found; i++)
4958                    {
4959                    if (cslot != slot)
4960                      {
4961                      if (GET2(cslot, 0) == cd->bracount + 1)
4962                        {
4963                        *errorcodeptr = ERR65;
4964                        goto FAILED;
4965                        }
4966                      }
4967                    else i--;
4968                    cslot += cd->name_entry_size;
4969                    }
4970                  }
4971    
4972              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4973              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
4974              slot[2+namelen] = 0;              slot[2+namelen] = 0;
4975              }              }
4976            }            }
4977    
4978          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
4979            encountered. */
4980    
         ptr++;                    /* Move past > or ' */  
4981          cd->names_found++;          cd->names_found++;
4982            ptr++;                    /* Move past > or ' */
4983          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
4984    
4985    
4986          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4987          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4988          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4989          is_recurse = TRUE;          is_recurse = TRUE;
4990          /* Fall through */          /* Fall through */
4991    
4992          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4993          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4994          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4995            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4996            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4997    
4998          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4999          name = ++ptr;          name = ++ptr;
# Line 3998  for (;; ptr++) Line 5005  for (;; ptr++)
5005    
5006          if (lengthptr != NULL)          if (lengthptr != NULL)
5007            {            {
5008              if (namelen == 0)
5009                {
5010                *errorcodeptr = ERR62;
5011                goto FAILED;
5012                }
5013            if (*ptr != terminator)            if (*ptr != terminator)
5014              {              {
5015              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4011  for (;; ptr++) Line 5023  for (;; ptr++)
5023            recno = 0;            recno = 0;
5024            }            }
5025    
5026          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
5027            first, and then check that we have reached the end of the name in the
5028            table. That way, if the name that is longer than any in the table,
5029            the comparison will fail without reading beyond the table entry. */
5030    
5031          else          else
5032            {            {
5033            slot = cd->name_table;            slot = cd->name_table;
5034            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5035              {              {
5036              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
5037                    slot[2+namelen] == 0)
5038                  break;
5039              slot += cd->name_entry_size;              slot += cd->name_entry_size;
5040              }              }
5041    
# Line 4027  for (;; ptr++) Line 5044  for (;; ptr++)
5044              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5045              }              }
5046            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5047                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5048                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5049              {              {
5050              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4043  for (;; ptr++) Line 5060  for (;; ptr++)
5060    
5061    
5062          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5063          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5064          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5065          /* Fall through */          /* Fall through */
5066    
5067    
5068          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5069          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5070          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
5071          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
5072            {            {
5073            const uschar *called;            const uschar *called;
5074              terminator = CHAR_RIGHT_PARENTHESIS;
5075    
5076              /* Come here from the \g<...> and \g'...' code (Oniguruma
5077              compatibility). However, the syntax has been checked to ensure that
5078              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5079              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5080              ever be taken. */
5081    
5082              HANDLE_NUMERICAL_RECURSION:
5083    
5084            if ((refsign = *ptr) == '+') ptr++;            if ((refsign = *ptr) == CHAR_PLUS)
5085            else if (refsign == '-')              {
5086                ptr++;
5087                if ((digitab[*ptr] & ctype_digit) == 0)
5088                  {
5089                  *errorcodeptr = ERR63;
5090                  goto FAILED;
5091                  }
5092                }
5093              else if (refsign == CHAR_MINUS)
5094              {              {
5095              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5096                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
# Line 4065  for (;; ptr++) Line 5099  for (;; ptr++)
5099    
5100            recno = 0;            recno = 0;
5101            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5102              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5103    
5104            if (*ptr != ')')            if (*ptr != terminator)
5105              {              {
5106              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5107              goto FAILED;              goto FAILED;
5108              }              }
5109    
5110            if (refsign == '-')            if (refsign == CHAR_MINUS)
5111              {              {
5112              if (recno == 0)              if (recno == 0)
5113                {                {
# Line 4087  for (;; ptr++) Line 5121  for (;; ptr++)
5121                goto FAILED;                goto FAILED;
5122                }                }
5123              }              }
5124            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5125              {              {
5126              if (recno == 0)              if (recno == 0)
5127                {                {
# Line 4114  for (;; ptr++) Line 5148  for (;; ptr++)
5148            if (lengthptr == NULL)            if (lengthptr == NULL)
5149              {              {
5150              *code = OP_END;              *code = OP_END;
5151              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5152                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5153    
5154              /* Forward reference */              /* Forward reference */
5155    
5156              if (called == NULL)              if (called == NULL)
5157                {                {
5158                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5159                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5160                  {                  {
5161                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5162                  goto FAILED;                  goto FAILED;
5163                  }                  }
5164    
5165                  /* Fudge the value of "called" so that when it is inserted as an
5166                  offset below, what it actually inserted is the reference number
5167                  of the group. */
5168    
5169                called = cd->start_code + recno;                called = cd->start_code + recno;
5170                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5171                }                }
# Line 4173  for (;; ptr++) Line 5213  for (;; ptr++)
5213          set = unset = 0;          set = unset = 0;
5214          optset = &set;          optset = &set;
5215    
5216          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5217            {            {
5218            switch (*ptr++)            switch (*ptr++)
5219              {              {
5220              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5221    
5222              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5223              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5224              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5225              break;              break;
5226    
5227              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5228              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5229              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5230              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5231              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5232              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5233    
5234              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5235                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4220  for (;; ptr++) Line 5260  for (;; ptr++)
5260          both phases.          both phases.
5261    
5262          If we are not at the pattern start, compile code to change the ims          If we are not at the pattern start, compile code to change the ims
5263          options if this setting actually changes any of them. We also pass the          options if this setting actually changes any of them, and reset the
5264          new setting back so that it can be put at the start of any following          greedy defaults and the case value for firstbyte and reqbyte. */
         branches, and when this group ends (if we are in a group), a resetting  
         item can be compiled. */  
5265    
5266          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5267            {            {
5268            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5269                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5270              {              {
5271              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5272              }              }
5273           else            else
5274              {              {
5275              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5276                {                {
5277                *code++ = OP_OPT;                *code++ = OP_OPT;
5278                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5279                }                }
   
             /* Change options at this level, and pass them back for use  
             in subsequent branches. Reset the greedy defaults and the case  
             value for firstbyte and reqbyte. */  
   
             *optionsptr = options = newoptions;  
5280              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5281              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5282              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5283              }              }
5284    
5285              /* Change options at this level, and pass them back for use
5286              in subsequent branches. When not at the start of the pattern, this
5287              information is also necessary so that a resetting item can be
5288              compiled at the end of a group (if we are in a group). */
5289    
5290              *optionsptr = options = newoptions;
5291            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5292            continue;              /* It is complete */            continue;              /* It is complete */
5293            }            }
# Line 4304  for (;; ptr++) Line 5341  for (;; ptr++)
5341           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
5342           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5343            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5344             reset_bracount,               /* True if (?| group */
5345           skipbytes,                    /* Skip over bracket number */           skipbytes,                    /* Skip over bracket number */
5346           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
5347           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
# Line 4365  for (;; ptr++) Line 5403  for (;; ptr++)
5403    
5404      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5405    
5406      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5407        {        {
5408        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5409        goto FAILED;        goto FAILED;
5410        }        }
5411    
5412      /* In the pre-compile phase, update the length by the length of the nested      /* In the pre-compile phase, update the length by the length of the group,
5413      group, less the brackets at either end. Then reduce the compiled code to      less the brackets at either end. Then reduce the compiled code to just a
5414      just the brackets so that it doesn't use much memory if it is duplicated by      set of non-capturing brackets so that it doesn't use much memory if it is
5415      a quantifier. */      duplicated by a quantifier.*/
5416    
5417      if (lengthptr != NULL)      if (lengthptr != NULL)
5418        {        {
5419          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
5420            {
5421            *errorcodeptr = ERR20;
5422            goto FAILED;
5423            }
5424        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5425        code++;        *code++ = OP_BRA;
5426        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5427        *code++ = OP_KET;        *code++ = OP_KET;
5428        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5429          break;    /* No need to waste time with special character handling */
5430        }        }
5431    
5432      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
5433    
5434      else code = tempcode;      code = tempcode;
5435    
5436      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
5437      relevant. */      relevant. */
# Line 4457  for (;; ptr++) Line 5501  for (;; ptr++)
5501      We can test for values between ESC_b and ESC_Z for the latter; this may      We can test for values between ESC_b and ESC_Z for the latter; this may
5502      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5503    
5504      case '\\':      case CHAR_BACKSLASH:
5505      tempptr = ptr;      tempptr = ptr;
5506      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5507      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 4466  for (;; ptr++) Line 5510  for (;; ptr++)
5510        {        {
5511        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5512          {          {
5513          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5514            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5515                else inescq = TRUE;
5516          continue;          continue;
5517          }          }
5518    
# Line 4484  for (;; ptr++) Line 5529  for (;; ptr++)
5529        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5530        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5531    
5532          /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5533          is a subroutine call by number (Oniguruma syntax). In fact, the value
5534          -ESC_g is returned only for these cases. So we don't need to check for <
5535          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5536          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5537          that is a synonym for a named back reference). */
5538    
5539          if (-c == ESC_g)
5540            {
5541            const uschar *p;
5542            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5543            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5544              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5545    
5546            /* These two statements stop the compiler for warning about possibly
5547            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5548            fact, because we actually check for a number below, the paths that
5549            would actually be in error are never taken. */
5550    
5551            skipbytes = 0;
5552            reset_bracount = FALSE;
5553    
5554            /* Test for a name */
5555    
5556            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5557              {
5558              BOOL isnumber = TRUE;
5559              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5560                {
5561                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5562                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5563                }
5564              if (*p != terminator)
5565                {
5566                *errorcodeptr = ERR57;
5567                break;
5568                }
5569              if (isnumber)
5570                {
5571                ptr++;
5572                goto HANDLE_NUMERICAL_RECURSION;
5573                }
5574              is_recurse = TRUE;
5575              goto NAMED_REF_OR_RECURSE;
5576              }
5577    
5578            /* Test a signed number in angle brackets or quotes. */
5579    
5580            p = ptr + 2;
5581            while ((digitab[*p] & ctype_digit) != 0) p++;
5582            if (*p != terminator)
5583              {
5584              *errorcodeptr = ERR57;
5585              break;
5586              }
5587            ptr++;
5588            goto HANDLE_NUMERICAL_RECURSION;
5589            }
5590    
5591        /* \k<name> or \k'name' is a back reference by name (Perl syntax).        /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5592        We also support \k{name} (.NET syntax) */        We also support \k{name} (.NET syntax) */
5593    
5594        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\'' || ptr[1] == '{'))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5595              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5596          {          {
5597          is_recurse = FALSE;          is_recurse = FALSE;
5598          terminator = (*(++ptr) == '<')? '>' : (*ptr == '\'')? '\'' : '}';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5599              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5600              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5601          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5602          }          }
5603    
# Line 4500  for (;; ptr++) Line 5607  for (;; ptr++)
5607    
5608        if (-c >= ESC_REF)        if (-c >= ESC_REF)
5609          {          {
5610            open_capitem *oc;
5611          recno = -c - ESC_REF;          recno = -c - ESC_REF;
5612    
5613          HANDLE_REFERENCE:    /* Come here from named backref handling */          HANDLE_REFERENCE:    /* Come here from named backref handling */
# Line 4509  for (;; ptr++) Line 5617  for (;; ptr++)
5617          PUT2INC(code, 0, recno);          PUT2INC(code, 0, recno);
5618          cd->backref_map |= (recno < 32)? (1 << recno) : 1;          cd->backref_map |= (recno < 32)? (1 << recno) : 1;
5619          if (recno > cd->top_backref) cd->top_backref = recno;          if (recno > cd->top_backref) cd->top_backref = recno;
5620    
5621            /* Check to see if this back reference is recursive, that it, it
5622            is inside the group that it references. A flag is set so that the
5623            group can be made atomic. */
5624    
5625            for (oc = cd->open_caps; oc != NULL; oc = oc->next)
5626              {
5627              if (oc->number == recno)
5628                {
5629                oc->flag = TRUE;
5630                break;
5631                }
5632              }
5633          }          }
5634    
5635        /* So are Unicode property matches, if supported. */        /* So are Unicode property matches, if supported. */
# Line 4591  for (;; ptr++) Line 5712  for (;; ptr++)
5712      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5713      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5714    
5715        /* Remember if \r or \n were seen */
5716    
5717        if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5718          cd->external_flags |= PCRE_HASCRORLF;
5719    
5720      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5721      byte, set it from this character, but revert to none on a zero repeat.      byte, set it from this character, but revert to none on a zero repeat.
5722      Otherwise, leave the firstbyte value alone, and don't change it on a zero      Otherwise, leave the firstbyte value alone, and don't change it on a zero
# Line 4663  Arguments: Line 5789  Arguments:
5789    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
5790    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
5791    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
5792      reset_bracount TRUE to reset the count for each branch
5793    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
5794    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
5795    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
# Line 4676  Returns:         TRUE on success Line 5803  Returns:         TRUE on success
5803    
5804  static BOOL  static BOOL
5805  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,
5806    int *errorcodeptr, BOOL lookbehind, int skipbytes, int *firstbyteptr,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
5807    int *reqbyteptr, branch_chain *bcptr, compile_data *cd, int *lengthptr)    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
5808      int *lengthptr)
5809  {  {
5810  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
5811  uschar *code = *codeptr;  uschar *code = *codeptr;
5812  uschar *last_branch = code;  uschar *last_branch = code;
5813  uschar *start_bracket = code;  uschar *start_bracket = code;
5814  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5815    open_capitem capitem;
5816    int capnumber = 0;
5817  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5818  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5819  int length;  int length;
5820    int orig_bracount;
5821    int max_bracount;
5822    int old_external_options = cd->external_options;
5823  branch_chain bc;  branch_chain bc;
5824    
5825  bc.outer = bcptr;  bc.outer = bcptr;
5826  bc.current = code;  bc.current_branch = code;
5827    
5828  firstbyte = reqbyte = REQ_UNSET;  firstbyte = reqbyte = REQ_UNSET;
5829    
# Line 4708  the code that abstracts option settings Line 5841  the code that abstracts option settings
5841  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
5842  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
5843    
5844    /* If this is a capturing subpattern, add to the chain of open capturing items
5845    so that we can detect them if (*ACCEPT) is encountered. This is also used to
5846    detect groups that contain recursive back references to themselves. */
5847    
5848    if (*code == OP_CBRA)
5849      {
5850      capnumber = GET2(code, 1 + LINK_SIZE);
5851      capitem.number = capnumber;
5852      capitem.next = cd->open_caps;
5853      capitem.flag = FALSE;
5854      cd->open_caps = &capitem;
5855      }
5856    
5857  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5858    
5859  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 4715  code += 1 + LINK_SIZE + skipbytes; Line 5861  code += 1 + LINK_SIZE + skipbytes;
5861    
5862  /* Loop for each alternative branch */  /* Loop for each alternative branch */
5863    
5864    orig_bracount = max_bracount = cd->bracount;
5865  for (;;)  for (;;)
5866    {    {
5867      /* For a (?| group, reset the capturing bracket count so that each branch
5868      uses the same numbers. */
5869    
5870      if (reset_bracount) cd->bracount = orig_bracount;
5871    
5872    /* Handle a change of ims options at the start of the branch */    /* Handle a change of ims options at the start of the branch */
5873    
5874    if ((options & PCRE_IMS) != oldims)    if ((options & PCRE_IMS) != oldims)
# Line