/[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 116 by ph10, Fri Mar 9 15:23:02 2007 UTC revision 500 by ph10, Sat Mar 6 19:00:29 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) 1997-2006 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,      0,      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 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  };    "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 262  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 298  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 373  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  /*************************************************  /*************************************************
# Line 399  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 417  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 442  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. */      (1) A number, either plain or braced. If positive, it is an absolute
593        backreference. If negative, it is a relative backreference. This is a Perl
594        5.10 feature.
595    
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;
618          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
619            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
620          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
621            {
622            c = -ESC_k;
623            break;
624            }
625        braced = TRUE;        braced = TRUE;
626        ptr++;        ptr++;
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 471  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 484  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 504  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 525  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 538  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 549  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 559  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 587  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 603  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 677  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 713  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 756  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 797  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 807  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 837  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 == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1144      {      {
1145      while (*(++ptr) != ']')      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[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1170            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1171          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 904  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      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1210        {
1211        if (dup_parens && *count < hwm_count) *count = hwm_count;
1212        *ptrptr = ptr;
1213        return -1;
1214      }      }
1215    
1216    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1217    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1218        if (*count > hwm_count) hwm_count = *count;
1219        *count = start_count;
1220        }
1221      }
1222    
1223    FAIL_EXIT:
1224    *ptrptr = ptr;
1225    return -1;
1226    }
1227    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1228    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1229    
   count++;  
1230    
1231    if (name == NULL && count == lorn) return count;  /*************************************************
1232    term = *ptr++;  *       Find forward referenced subpattern       *
1233    if (term == '<') term = '>';  *************************************************/
1234    thisname = ptr;  
1235    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1236    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1237        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1238      return count;  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    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    for (;;)
1268      {
1269      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1270      if (rc > 0 || *ptr++ == 0) break;
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 996  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 1012  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 1041  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 1072  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 1085  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:
1451      case OP_SOD:      case OP_SOD:
1452      case OP_SOM:      case OP_SOM:
1453        case OP_SET_SOM:
1454      case OP_EOD:      case OP_EOD:
1455      case OP_EODN:      case OP_EODN:
1456      case OP_CIRC:      case OP_CIRC:
# Line 1108  for (;;) Line 1468  for (;;)
1468      branchlength++;      branchlength++;
1469      cc += 2;      cc += 2;
1470  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1471      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1472        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1473  #endif  #endif
1474      break;      break;
1475    
# Line 1122  for (;;) Line 1480  for (;;)
1480      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1481      cc += 4;      cc += 4;
1482  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1483      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1484        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1485  #endif  #endif
1486      break;      break;
1487    
1488      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1489      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1490        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1491      cc += 4;      cc += 4;
1492      break;      break;
1493    
# Line 1148  for (;;) Line 1505  for (;;)
1505      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1506      case OP_WORDCHAR:      case OP_WORDCHAR:
1507      case OP_ANY:      case OP_ANY:
1508        case OP_ALLANY:
1509      branchlength++;      branchlength++;
1510      cc++;      cc++;
1511      break;      break;
# Line 1202  for (;;) Line 1560  for (;;)
1560    
1561    
1562  /*************************************************  /*************************************************
1563  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1564  *************************************************/  *************************************************/
1565    
1566  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1567  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1568    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1569    so that it can be called from pcre_study() when finding the minimum matching
1570    length.
1571    
1572  Arguments:  Arguments:
1573    code        points to start of expression    code        points to start of expression
1574    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1575    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1576    
1577  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
1578  */  */
1579    
1580  static const uschar *  const uschar *
1581  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1582  {  {
1583  for (;;)  for (;;)
1584    {    {
# Line 1230  for (;;) Line 1591  for (;;)
1591    
1592    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1593    
1594      /* Handle recursion */
1595    
1596      else if (c == OP_REVERSE)
1597        {
1598        if (number < 0) return (uschar *)code;
1599        code += _pcre_OP_lengths[c];
1600        }
1601    
1602    /* Handle capturing bracket */    /* Handle capturing bracket */
1603    
1604    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1239  for (;;) Line 1608  for (;;)
1608      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1609      }      }
1610    
1611    /* 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
1612    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
1613    arrange to skip the extra bytes. */    two bytes of parameters. */
1614    
1615    else    else
1616      {      {
1617        switch(c)
1618          {
1619          case OP_TYPESTAR:
1620          case OP_TYPEMINSTAR:
1621          case OP_TYPEPLUS:
1622          case OP_TYPEMINPLUS:
1623          case OP_TYPEQUERY:
1624          case OP_TYPEMINQUERY:
1625          case OP_TYPEPOSSTAR:
1626          case OP_TYPEPOSPLUS:
1627          case OP_TYPEPOSQUERY:
1628          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1629          break;
1630    
1631          case OP_TYPEUPTO:
1632          case OP_TYPEMINUPTO:
1633          case OP_TYPEEXACT:
1634          case OP_TYPEPOSUPTO:
1635          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1636          break;
1637          }
1638    
1639        /* Add in the fixed length from the table */
1640    
1641      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1642    
1643      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1644      a multi-byte character. The length in the table is a minimum, so we have to
1645      arrange to skip the extra bytes. */
1646    
1647  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1648      if (utf8) switch(c)      if (utf8) switch(c)
1649        {        {
# Line 1267  for (;;) Line 1665  for (;;)
1665        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1666        break;        break;
1667        }        }
1668    #else
1669        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1670  #endif  #endif
1671      }      }
1672    }    }
# Line 1303  for (;;) Line 1703  for (;;)
1703    
1704    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1705    
1706    /* 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
1707    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
1708    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1709    
1710    else    else
1711      {      {
1712        switch(c)
1713          {
1714          case OP_TYPESTAR:
1715          case OP_TYPEMINSTAR:
1716          case OP_TYPEPLUS:
1717          case OP_TYPEMINPLUS:
1718          case OP_TYPEQUERY:
1719          case OP_TYPEMINQUERY:
1720          case OP_TYPEPOSSTAR:
1721          case OP_TYPEPOSPLUS:
1722          case OP_TYPEPOSQUERY:
1723          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1724          break;
1725    
1726          case OP_TYPEPOSUPTO:
1727          case OP_TYPEUPTO:
1728          case OP_TYPEMINUPTO:
1729          case OP_TYPEEXACT:
1730          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1731          break;
1732          }
1733    
1734        /* Add in the fixed length from the table */
1735    
1736      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1737    
1738        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1739        by a multi-byte character. The length in the table is a minimum, so we have
1740        to arrange to skip the extra bytes. */
1741    
1742  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1743      if (utf8) switch(c)      if (utf8) switch(c)
1744        {        {
# Line 1332  for (;;) Line 1760  for (;;)
1760        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1761        break;        break;
1762        }        }
1763    #else
1764        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1765  #endif  #endif
1766      }      }
1767    }    }
# Line 1347  for (;;) Line 1777  for (;;)
1777  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()
1778  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
1779  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1780  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
1781  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
1782    bracket whose current branch will already have been scanned.
1783    
1784  Arguments:  Arguments:
1785    code        points to start of search    code        points to start of search
# Line 1370  for (code = first_significant_code(code Line 1801  for (code = first_significant_code(code
1801    
1802    c = *code;    c = *code;
1803    
1804    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1805      first_significant_code() with a TRUE final argument. */
1806    
1807      if (c == OP_ASSERT)
1808        {
1809        do code += GET(code, 1); while (*code == OP_ALT);
1810        c = *code;
1811        continue;
1812        }
1813    
1814      /* Groups with zero repeats can of course be empty; skip them. */
1815    
1816      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1817        {
1818        code += _pcre_OP_lengths[c];
1819        do code += GET(code, 1); while (*code == OP_ALT);
1820        c = *code;
1821        continue;
1822        }
1823    
1824      /* For other groups, scan the branches. */
1825    
1826      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1827      {      {
1828      BOOL empty_branch;      BOOL empty_branch;
1829      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1830    
1831      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1832        empty branch, so just skip over the conditional, because it could be empty.
1833        Otherwise, scan the individual branches of the group. */
1834    
1835      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;  
1836        code += GET(code, 1);        code += GET(code, 1);
1837        else
1838          {
1839          empty_branch = FALSE;
1840          do
1841            {
1842            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1843              empty_branch = TRUE;
1844            code += GET(code, 1);
1845            }
1846          while (*code == OP_ALT);
1847          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1848        }        }
     while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
   
     /* Move past the KET and fudge things so that the increment in the "for"  
     above has no effect. */  
1849    
1850      c = OP_END;      c = *code;
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
1851      continue;      continue;
1852      }      }
1853    
# Line 1399  for (code = first_significant_code(code Line 1855  for (code = first_significant_code(code
1855    
1856    switch (c)    switch (c)
1857      {      {
1858      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1859        cannot be represented just by a bit map. This includes negated single
1860        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1861        actual length is stored in the compiled code, so we must update "code"
1862        here. */
1863    
1864  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1865      case OP_XCLASS:      case OP_XCLASS:
1866      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1867      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1868  #endif  #endif
1869    
# Line 1447  for (code = first_significant_code(code Line 1907  for (code = first_significant_code(code
1907      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1908      case OP_WORDCHAR:      case OP_WORDCHAR:
1909      case OP_ANY:      case OP_ANY:
1910        case OP_ALLANY:
1911      case OP_ANYBYTE:      case OP_ANYBYTE:
1912      case OP_CHAR:      case OP_CHAR:
1913      case OP_CHARNC:      case OP_CHARNC:
# Line 1465  for (code = first_significant_code(code Line 1926  for (code = first_significant_code(code
1926      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1927      return FALSE;      return FALSE;
1928    
1929        /* These are going to continue, as they may be empty, but we have to
1930        fudge the length for the \p and \P cases. */
1931    
1932        case OP_TYPESTAR:
1933        case OP_TYPEMINSTAR:
1934        case OP_TYPEPOSSTAR:
1935        case OP_TYPEQUERY:
1936        case OP_TYPEMINQUERY:
1937        case OP_TYPEPOSQUERY:
1938        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1939        break;
1940    
1941        /* Same for these */
1942    
1943        case OP_TYPEUPTO:
1944        case OP_TYPEMINUPTO:
1945        case OP_TYPEPOSUPTO:
1946        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1947        break;
1948    
1949      /* End of branch */      /* End of branch */
1950    
1951      case OP_KET:      case OP_KET:
# Line 1483  for (code = first_significant_code(code Line 1964  for (code = first_significant_code(code
1964      case OP_QUERY:      case OP_QUERY:
1965      case OP_MINQUERY:      case OP_MINQUERY:
1966      case OP_POSQUERY:      case OP_POSQUERY:
1967        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1968        break;
1969    
1970      case OP_UPTO:      case OP_UPTO:
1971      case OP_MINUPTO:      case OP_MINUPTO:
1972      case OP_POSUPTO:      case OP_POSUPTO:
1973      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1974      break;      break;
1975  #endif  #endif
1976      }      }
# Line 1519  static BOOL Line 2003  static BOOL
2003  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2004    BOOL utf8)    BOOL utf8)
2005  {  {
2006  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2007    {    {
2008    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8))
2009        return FALSE;
2010    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2011    }    }
2012  return TRUE;  return TRUE;
# Line 1534  return TRUE; Line 2019  return TRUE;
2019  *************************************************/  *************************************************/
2020    
2021  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2022  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
2023  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2024  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2025    
2026    Originally, this function only recognized a sequence of letters between the
2027    terminators, but it seems that Perl recognizes any sequence of characters,
2028    though of course unknown POSIX names are subsequently rejected. Perl gives an
2029    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2030    didn't consider this to be a POSIX class. Likewise for [:1234:].
2031    
2032    The problem in trying to be exactly like Perl is in the handling of escapes. We
2033    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2034    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2035    below handles the special case of \], but does not try to do any other escape
2036    processing. This makes it different from Perl for cases such as [:l\ower:]
2037    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2038    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2039    I think.
2040    
2041  Argument:  Arguments:
2042    ptr      pointer to the initial [    ptr      pointer to the initial [
2043    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2044    
2045  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2046  */  */
2047    
2048  static BOOL  static BOOL
2049  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2050  {  {
2051  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2052  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2053  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2054    {    {
2055    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2056    return TRUE;      {
2057        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2058        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2059          {
2060          *endptr = ptr;
2061          return TRUE;
2062          }
2063        }
2064    }    }
2065  return FALSE;  return FALSE;
2066  }  }
# Line 1581  Returns:     a value representing the na Line 2085  Returns:     a value representing the na
2085  static int  static int
2086  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2087  {  {
2088    const char *pn = posix_names;
2089  register int yield = 0;  register int yield = 0;
2090  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2091    {    {
2092    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2093      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2094      pn += posix_name_lengths[yield] + 1;
2095    yield++;    yield++;
2096    }    }
2097  return -1;  return -1;
# Line 1600  return -1; Line 2106  return -1;
2106  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2107  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2108  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
2109  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
2110  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
2111  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
2112  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
2113  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2114    OP_END.
2115    
2116  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2117  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1627  adjust_recurse(uschar *group, int adjust Line 2134  adjust_recurse(uschar *group, int adjust
2134    uschar *save_hwm)    uschar *save_hwm)
2135  {  {
2136  uschar *ptr = group;  uschar *ptr = group;
2137    
2138  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2139    {    {
2140    int offset;    int offset;
# Line 1738  get_othercase_range(unsigned int *cptr, Line 2246  get_othercase_range(unsigned int *cptr,
2246  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2247    
2248  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2249    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2250    
2251  if (c > d) return FALSE;  if (c > d) return FALSE;
2252    
# Line 1747  next = othercase + 1; Line 2255  next = othercase + 1;
2255    
2256  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2257    {    {
2258    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2259    next++;    next++;
2260    }    }
2261    
# Line 1793  if ((options & PCRE_EXTENDED) != 0) Line 2301  if ((options & PCRE_EXTENDED) != 0)
2301    for (;;)    for (;;)
2302      {      {
2303      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2304      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2305        {        {
2306        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2307          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1805  if ((options & PCRE_EXTENDED) != 0) Line 2313  if ((options & PCRE_EXTENDED) != 0)
2313  /* 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
2314  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2315    
2316  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2317    {    {
2318    int temperrorcode = 0;    int temperrorcode = 0;
2319    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1830  if ((options & PCRE_EXTENDED) != 0) Line 2338  if ((options & PCRE_EXTENDED) != 0)
2338    for (;;)    for (;;)
2339      {      {
2340      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2341      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2342        {        {
2343        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2344          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1841  if ((options & PCRE_EXTENDED) != 0) Line 2349  if ((options & PCRE_EXTENDED) != 0)
2349    
2350  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2351    
2352  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2353    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2354        return FALSE;
2355    
2356  /* 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
2357  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 1857  if (next >= 0) switch(op_code) Line 2366  if (next >= 0) switch(op_code)
2366    case OP_CHAR:    case OP_CHAR:
2367  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2368    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2369    #else
2370      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2371  #endif  #endif
2372    return item != next;    return item != next;
2373    
# Line 1875  if (next >= 0) switch(op_code) Line 2386  if (next >= 0) switch(op_code)
2386      unsigned int othercase;      unsigned int othercase;
2387      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2388  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2389      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2390  #else  #else
2391      othercase = NOTACHAR;      othercase = NOTACHAR;
2392  #endif  #endif
# Line 1888  if (next >= 0) switch(op_code) Line 2399  if (next >= 0) switch(op_code)
2399    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2400    
2401    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2402    if (item == next) return TRUE;    if (item == next) return TRUE;
2403    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2404  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1897  if (next >= 0) switch(op_code) Line 2407  if (next >= 0) switch(op_code)
2407      unsigned int othercase;      unsigned int othercase;
2408      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2409  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2410      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2411  #else  #else
2412      othercase = NOTACHAR;      othercase = NOTACHAR;
2413  #endif  #endif
# Line 1925  if (next >= 0) switch(op_code) Line 2435  if (next >= 0) switch(op_code)
2435    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2436    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2437    
2438      case OP_HSPACE:
2439      case OP_NOT_HSPACE:
2440      switch(next)
2441        {
2442        case 0x09:
2443        case 0x20:
2444        case 0xa0:
2445        case 0x1680:
2446        case 0x180e:
2447        case 0x2000:
2448        case 0x2001:
2449        case 0x2002:
2450        case 0x2003:
2451        case 0x2004:
2452        case 0x2005:
2453        case 0x2006:
2454        case 0x2007:
2455        case 0x2008:
2456        case 0x2009:
2457        case 0x200A:
2458        case 0x202f:
2459        case 0x205f:
2460        case 0x3000:
2461        return op_code != OP_HSPACE;
2462        default:
2463        return op_code == OP_HSPACE;
2464        }
2465    
2466      case OP_VSPACE:
2467      case OP_NOT_VSPACE:
2468      switch(next)
2469        {
2470        case 0x0a:
2471        case 0x0b:
2472        case 0x0c:
2473        case 0x0d:
2474        case 0x85:
2475        case 0x2028:
2476        case 0x2029:
2477        return op_code != OP_VSPACE;
2478        default:
2479        return op_code == OP_VSPACE;
2480        }
2481    
2482    default:    default:
2483    return FALSE;    return FALSE;
2484    }    }
# Line 1959  switch(op_code) Line 2513  switch(op_code)
2513      case ESC_W:      case ESC_W:
2514      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2515    
2516        case ESC_h:
2517        case ESC_H:
2518        switch(item)
2519          {
2520          case 0x09:
2521          case 0x20:
2522          case 0xa0:
2523          case 0x1680:
2524          case 0x180e:
2525          case 0x2000:
2526          case 0x2001:
2527          case 0x2002:
2528          case 0x2003:
2529          case 0x2004:
2530          case 0x2005:
2531          case 0x2006:
2532          case 0x2007:
2533          case 0x2008:
2534          case 0x2009:
2535          case 0x200A:
2536          case 0x202f:
2537          case 0x205f:
2538          case 0x3000:
2539          return -next != ESC_h;
2540          default:
2541          return -next == ESC_h;
2542          }
2543    
2544        case ESC_v:
2545        case ESC_V:
2546        switch(item)
2547          {
2548          case 0x0a:
2549          case 0x0b:
2550          case 0x0c:
2551          case 0x0d:
2552          case 0x85:
2553          case 0x2028:
2554          case 0x2029:
2555          return -next != ESC_v;
2556          default:
2557          return -next == ESC_v;
2558          }
2559    
2560      default:      default:
2561      return FALSE;      return FALSE;
2562      }      }
2563    
2564    case OP_DIGIT:    case OP_DIGIT:
2565    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2566             next == -ESC_h || next == -ESC_v;
2567    
2568    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2569    return next == -ESC_d;    return next == -ESC_d;
# Line 1973  switch(op_code) Line 2572  switch(op_code)
2572    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2573    
2574    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2575    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2576    
2577      case OP_HSPACE:
2578      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2579    
2580      case OP_NOT_HSPACE:
2581      return next == -ESC_h;
2582    
2583      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2584      case OP_VSPACE:
2585      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2586    
2587      case OP_NOT_VSPACE:
2588      return next == -ESC_v;
2589    
2590    case OP_WORDCHAR:    case OP_WORDCHAR:
2591    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2592    
2593    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2594    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2049  uschar classbits[32]; Line 2661  uschar classbits[32];
2661  BOOL class_utf8;  BOOL class_utf8;
2662  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2663  uschar *class_utf8data;  uschar *class_utf8data;
2664    uschar *class_utf8data_base;
2665  uschar utf8_char[6];  uschar utf8_char[6];
2666  #else  #else
2667  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2668  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2669  #endif  #endif
2670    
2671  #ifdef DEBUG  #ifdef PCRE_DEBUG
2672  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2673  #endif  #endif
2674    
# Line 2088  req_caseopt = ((options & PCRE_CASELESS) Line 2701  req_caseopt = ((options & PCRE_CASELESS)
2701  for (;; ptr++)  for (;; ptr++)
2702    {    {
2703    BOOL negate_class;    BOOL negate_class;
2704      BOOL should_flip_negation;
2705    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2706    BOOL is_quantifier;    BOOL is_quantifier;
2707    BOOL is_recurse;    BOOL is_recurse;
2708      BOOL reset_bracount;
2709    int class_charcount;    int class_charcount;
2710    int class_lastchar;    int class_lastchar;
2711    int newoptions;    int newoptions;
2712    int recno;    int recno;
2713      int refsign;
2714    int skipbytes;    int skipbytes;
2715    int subreqbyte;    int subreqbyte;
2716    int subfirstbyte;    int subfirstbyte;
# Line 2111  for (;; ptr++) Line 2727  for (;; ptr++)
2727    
2728    if (lengthptr != NULL)    if (lengthptr != NULL)
2729      {      {
2730  #ifdef DEBUG  #ifdef PCRE_DEBUG
2731      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2732  #endif  #endif
2733      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
# Line 2127  for (;; ptr++) Line 2743  for (;; ptr++)
2743      */      */
2744    
2745      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2746    
2747        /* Paranoid check for integer overflow */
2748    
2749        if (OFLOW_MAX - *lengthptr < code - last_code)
2750          {
2751          *errorcodeptr = ERR20;
2752          goto FAILED;
2753          }
2754    
2755      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2756      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));
2757    
# Line 2164  for (;; ptr++) Line 2789  for (;; ptr++)
2789    
2790    if (inescq && c != 0)    if (inescq && c != 0)
2791      {      {
2792      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2793        {        {
2794        inescq = FALSE;        inescq = FALSE;
2795        ptr++;        ptr++;
# Line 2190  for (;; ptr++) Line 2815  for (;; ptr++)
2815    /* 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
2816    a quantifier. */    a quantifier. */
2817    
2818    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2819      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2820        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2821    
2822    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2823         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2206  for (;; ptr++) Line 2832  for (;; ptr++)
2832    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2833      {      {
2834      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2835      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2836        {        {
2837        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2838          {          {
# Line 2231  for (;; ptr++) Line 2857  for (;; ptr++)
2857      {      {
2858      /* ===================================================================*/      /* ===================================================================*/
2859      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2860      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2861      case ')':      case CHAR_RIGHT_PARENTHESIS:
2862      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2863      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2864      *codeptr = code;      *codeptr = code;
2865      *ptrptr = ptr;      *ptrptr = ptr;
2866      if (lengthptr != NULL)      if (lengthptr != NULL)
2867        {        {
2868          if (OFLOW_MAX - *lengthptr < code - last_code)
2869            {
2870            *errorcodeptr = ERR20;
2871            goto FAILED;
2872            }
2873        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2874        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2875        }        }
# Line 2249  for (;; ptr++) Line 2880  for (;; ptr++)
2880      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2881      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2882    
2883      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2884      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2885        {        {
2886        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2258  for (;; ptr++) Line 2889  for (;; ptr++)
2889      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2890      break;      break;
2891    
2892      case '$':      case CHAR_DOLLAR_SIGN:
2893      previous = NULL;      previous = NULL;
2894      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2895      break;      break;
# Line 2266  for (;; ptr++) Line 2897  for (;; ptr++)
2897      /* 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
2898      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2899    
2900      case '.':      case CHAR_DOT:
2901      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2902      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2903      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2904      previous = code;      previous = code;
2905      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2906      break;      break;
2907    
2908    
# Line 2286  for (;; ptr++) Line 2917  for (;; ptr++)
2917      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,
2918      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2919      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.
     */  
2920    
2921      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2922      previous = code;      default (Perl) mode, it is treated as a data character. */
2923    
2924      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      case CHAR_RIGHT_SQUARE_BRACKET:
2925      they are encountered at the top level, so we'll do that too. */      if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2926          {
2927          *errorcodeptr = ERR64;
2928          goto FAILED;
2929          }
2930        goto NORMAL_CHAR;
2931    
2932        case CHAR_LEFT_SQUARE_BRACKET:
2933        previous = code;
2934    
2935        /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
2936        they are encountered at the top level, so we'll do that too. */
2937    
2938      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2939          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2940            check_posix_syntax(ptr, &tempptr))
2941        {        {
2942        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2943        goto FAILED;        goto FAILED;
2944        }        }
2945    
2946      /* 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,
2947        if the first few characters (either before or after ^) are \Q\E or \E we
2948        skip them too. This makes for compatibility with Perl. */
2949    
2950      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2951        for (;;)
2952        {        {
       negate_class = TRUE;  
2953        c = *(++ptr);        c = *(++ptr);
2954          if (c == CHAR_BACKSLASH)
2955            {
2956            if (ptr[1] == CHAR_E)
2957              ptr++;
2958            else if (strncmp((const char *)ptr+1,
2959                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2960              ptr += 3;
2961            else
2962              break;
2963            }
2964          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2965            negate_class = TRUE;
2966          else break;
2967        }        }
2968      else  
2969        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2970        an initial ']' is taken as a data character -- the code below handles
2971        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2972        [^] must match any character, so generate OP_ALLANY. */
2973    
2974        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2975            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2976        {        {
2977        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2978          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2979          zerofirstbyte = firstbyte;
2980          break;
2981        }        }
2982    
2983        /* If a class contains a negative special such as \S, we need to flip the
2984        negation flag at the end, so that support for characters > 255 works
2985        correctly (they are all included in the class). */
2986    
2987        should_flip_negation = FALSE;
2988    
2989      /* 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
2990      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
2991      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2330  for (;; ptr++) Line 3003  for (;; ptr++)
3003  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3004      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3005      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3006        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3007  #endif  #endif
3008    
3009      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2345  for (;; ptr++) Line 3019  for (;; ptr++)
3019          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3020          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3021          }          }
3022    
3023          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3024          data and reset the pointer. This is so that very large classes that
3025          contain a zillion UTF-8 characters no longer overwrite the work space
3026          (which is on the stack). */
3027    
3028          if (lengthptr != NULL)
3029            {
3030            *lengthptr += class_utf8data - class_utf8data_base;
3031            class_utf8data = class_utf8data_base;
3032            }
3033    
3034  #endif  #endif
3035    
3036        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3037    
3038        if (inescq)        if (inescq)
3039          {          {
3040          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3041            {            {
3042            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3043            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2366  for (;; ptr++) Line 3052  for (;; ptr++)
3052        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3053        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3054    
3055        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3056            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3057            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3058          {          {
3059          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3060          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3061          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3062          uschar pbits[32];          uschar pbits[32];
3063    
3064          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3065            {            {
3066            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3067            goto FAILED;            goto FAILED;
3068            }            }
3069    
3070          ptr += 2;          ptr += 2;
3071          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3072            {            {
3073            local_negate = TRUE;            local_negate = TRUE;
3074              should_flip_negation = TRUE;  /* Note negative special */
3075            ptr++;            ptr++;
3076            }            }
3077    
# Line 2451  for (;; ptr++) Line 3138  for (;; ptr++)
3138        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
3139        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
3140        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3141        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
3142        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3143    
3144        if (c == '\\')        if (c == CHAR_BACKSLASH)
3145          {          {
3146          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3147          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3148    
3149          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 */
3150          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 */
3151          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 */
3152          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3153            {            {
3154            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3155              {              {
3156              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3157              }              }
3158            else inescq = TRUE;            else inescq = TRUE;
3159            continue;            continue;
3160            }            }
3161            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3162    
3163          if (c < 0)          if (c < 0)
3164            {            {
# Line 2486  for (;; ptr++) Line 3174  for (;; ptr++)
3174              continue;              continue;
3175    
3176              case ESC_D:              case ESC_D:
3177                should_flip_negation = TRUE;
3178              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3179              continue;              continue;
3180    
# Line 2494  for (;; ptr++) Line 3183  for (;; ptr++)
3183              continue;              continue;
3184    
3185              case ESC_W:              case ESC_W:
3186                should_flip_negation = TRUE;
3187              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3188              continue;              continue;
3189    
# Line 2503  for (;; ptr++) Line 3193  for (;; ptr++)
3193              continue;              continue;
3194    
3195              case ESC_S:              case ESC_S:
3196                should_flip_negation = TRUE;
3197              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3198              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3199              continue;              continue;
3200    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
3201              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
3202              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
3203              }              }
# Line 2519  for (;; ptr++) Line 3207  for (;; ptr++)
3207            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3208                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3209    
3210              /* We need to deal with \H, \h, \V, and \v in both phases because
3211              they use extra memory. */
3212    
3213              if (-c == ESC_h)
3214                {
3215                SETBIT(classbits, 0x09); /* VT */
3216                SETBIT(classbits, 0x20); /* SPACE */
3217                SETBIT(classbits, 0xa0); /* NSBP */
3218    #ifdef SUPPORT_UTF8
3219                if (utf8)
3220                  {
3221                  class_utf8 = TRUE;
3222                  *class_utf8data++ = XCL_SINGLE;
3223                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3224                  *class_utf8data++ = XCL_SINGLE;
3225                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3226                  *class_utf8data++ = XCL_RANGE;
3227                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3228                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3229                  *class_utf8data++ = XCL_SINGLE;
3230                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3231                  *class_utf8data++ = XCL_SINGLE;
3232                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3233                  *class_utf8data++ = XCL_SINGLE;
3234                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3235                  }
3236    #endif
3237                continue;
3238                }
3239    
3240              if (-c == ESC_H)
3241                {
3242                for (c = 0; c < 32; c++)
3243                  {
3244                  int x = 0xff;
3245                  switch (c)
3246                    {
3247                    case 0x09/8: x ^= 1 << (0x09%8); break;
3248                    case 0x20/8: x ^= 1 << (0x20%8); break;
3249                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3250                    default: break;
3251                    }
3252                  classbits[c] |= x;
3253                  }
3254    
3255    #ifdef SUPPORT_UTF8
3256                if (utf8)
3257                  {
3258                  class_utf8 = TRUE;
3259                  *class_utf8data++ = XCL_RANGE;
3260                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3261                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3262                  *class_utf8data++ = XCL_RANGE;
3263                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3264                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3265                  *class_utf8data++ = XCL_RANGE;
3266                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3267                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3268                  *class_utf8data++ = XCL_RANGE;
3269                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3270                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3271                  *class_utf8data++ = XCL_RANGE;
3272                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3273                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3274                  *class_utf8data++ = XCL_RANGE;
3275                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3276                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3277                  *class_utf8data++ = XCL_RANGE;
3278                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3279                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3280                  }
3281    #endif
3282                continue;
3283                }
3284    
3285              if (-c == ESC_v)
3286                {
3287                SETBIT(classbits, 0x0a); /* LF */
3288                SETBIT(classbits, 0x0b); /* VT */
3289                SETBIT(classbits, 0x0c); /* FF */
3290                SETBIT(classbits, 0x0d); /* CR */
3291                SETBIT(classbits, 0x85); /* NEL */
3292    #ifdef SUPPORT_UTF8
3293                if (utf8)
3294                  {
3295                  class_utf8 = TRUE;
3296                  *class_utf8data++ = XCL_RANGE;
3297                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3298                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3299                  }
3300    #endif
3301                continue;
3302                }
3303    
3304              if (-c == ESC_V)
3305                {
3306                for (c = 0; c < 32; c++)
3307                  {
3308                  int x = 0xff;
3309                  switch (c)
3310                    {
3311                    case 0x0a/8: x ^= 1 << (0x0a%8);
3312                                 x ^= 1 << (0x0b%8);
3313                                 x ^= 1 << (0x0c%8);
3314                                 x ^= 1 << (0x0d%8);
3315                                 break;
3316                    case 0x85/8: x ^= 1 << (0x85%8); break;
3317                    default: break;
3318                    }
3319                  classbits[c] |= x;
3320                  }
3321    
3322    #ifdef SUPPORT_UTF8
3323                if (utf8)
3324                  {
3325                  class_utf8 = TRUE;
3326                  *class_utf8data++ = XCL_RANGE;
3327                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3328                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3329                  *class_utf8data++ = XCL_RANGE;
3330                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3331                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3332                  }
3333    #endif
3334                continue;
3335                }
3336    
3337            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3338    
3339  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2562  for (;; ptr++) Line 3377  for (;; ptr++)
3377        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3378    
3379        CHECK_RANGE:        CHECK_RANGE:
3380        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3381          {          {
3382          inescq = FALSE;          inescq = FALSE;
3383          ptr += 2;          ptr += 2;
# Line 2570  for (;; ptr++) Line 3385  for (;; ptr++)
3385    
3386        oldptr = ptr;        oldptr = ptr;
3387    
3388        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3389    
3390          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3391    
3392          /* Check for range */
3393    
3394          if (!inescq && ptr[1] == CHAR_MINUS)
3395          {          {
3396          int d;          int d;
3397          ptr += 2;          ptr += 2;
3398          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3399    
3400          /* 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
3401          mode. */          mode. */
3402    
3403          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3404            {            {
3405            ptr += 2;            ptr += 2;
3406            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3407                { ptr += 2; continue; }
3408            inescq = TRUE;            inescq = TRUE;
3409            break;            break;
3410            }            }
3411    
3412          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3413            {            {
3414            ptr = oldptr;            ptr = oldptr;
3415            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2606  for (;; ptr++) Line 3428  for (;; ptr++)
3428          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
3429          in such circumstances. */          in such circumstances. */
3430    
3431          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3432            {            {
3433            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3434            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3435    
3436            /* \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
3437            special means the '-' was literal */            special means the '-' was literal */
3438    
3439            if (d < 0)            if (d < 0)
3440              {              {
3441              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3442              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3443              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3444                {                {
3445                ptr = oldptr;                ptr = oldptr;
3446                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2637  for (;; ptr++) Line 3459  for (;; ptr++)
3459    
3460          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3461    
3462            /* Remember \r or \n */
3463    
3464            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3465    
3466          /* 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
3467          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3468          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2659  for (;; ptr++) Line 3485  for (;; ptr++)
3485              unsigned int origd = d;              unsigned int origd = d;
3486              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3487                {                {
3488                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3489                      ocd <= (unsigned int)d)
3490                    continue;                          /* Skip embedded ranges */
3491    
3492                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3493                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3494                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3495                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3496                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3497                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3498                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3499                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3500                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3501                  d = ocd;                  d = ocd;
3502                  continue;                  continue;
# Line 2751  for (;; ptr++) Line 3581  for (;; ptr++)
3581          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3582            {            {
3583            unsigned int othercase;            unsigned int othercase;
3584            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3585              {              {
3586              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3587              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2778  for (;; ptr++) Line 3608  for (;; ptr++)
3608    
3609      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3610    
3611      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3612    
3613      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3614        {        {
# Line 2786  for (;; ptr++) Line 3616  for (;; ptr++)
3616        goto FAILED;        goto FAILED;
3617        }        }
3618    
3619    
3620    /* This code has been disabled because it would mean that \s counts as
3621    an explicit \r or \n reference, and that's not really what is wanted. Now
3622    we set the flag only if there is a literal "\r" or "\n" in the class. */
3623    
3624    #if 0
3625        /* Remember whether \r or \n are in this class */
3626    
3627        if (negate_class)
3628          {
3629          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3630          }
3631        else
3632          {
3633          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3634          }
3635    #endif
3636    
3637    
3638      /* 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
3639      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
3640      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
3641      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3642      single-bytes only. This is an historical hangover. Maybe one day we can  
3643      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3644        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3645        operate on single-bytes only. This is an historical hangover. Maybe one day
3646        we can tidy these opcodes to handle multi-byte characters.
3647    
3648      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
3649      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2801  for (;; ptr++) Line 3653  for (;; ptr++)
3653      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3654    
3655  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3656      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3657            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3658  #else  #else
3659      if (class_charcount == 1)      if (class_charcount == 1)
3660  #endif  #endif
# Line 2847  for (;; ptr++) Line 3697  for (;; ptr++)
3697      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3698    
3699      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3700      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3701      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
3702        the class, so any that were explicitly given as well can be ignored. If
3703        (when there are explicit characters > 255 that must be listed) there are no
3704        characters < 256, we can omit the bitmap in the actual compiled code. */
3705    
3706  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3707      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3708        {        {
3709        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3710        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2877  for (;; ptr++) Line 3730  for (;; ptr++)
3730        }        }
3731  #endif  #endif
3732    
3733      /* 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
3734      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
3735      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
3736      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3737    
3738        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3739      if (negate_class)      if (negate_class)
3740        {        {
       *code++ = OP_NCLASS;  
3741        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3742          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3743        }        }
3744      else      else
3745        {        {
       *code++ = OP_CLASS;  
3746        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3747        }        }
3748      code += 32;      code += 32;
# Line 2901  for (;; ptr++) Line 3753  for (;; ptr++)
3753      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3754      has been tested above. */      has been tested above. */
3755    
3756      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3757      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3758      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3759      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3760      goto REPEAT;      goto REPEAT;
3761    
3762      case '*':      case CHAR_ASTERISK:
3763      repeat_min = 0;      repeat_min = 0;
3764      repeat_max = -1;      repeat_max = -1;
3765      goto REPEAT;      goto REPEAT;
3766    
3767      case '+':      case CHAR_PLUS:
3768      repeat_min = 1;      repeat_min = 1;
3769      repeat_max = -1;      repeat_max = -1;
3770      goto REPEAT;      goto REPEAT;
3771    
3772      case '?':      case CHAR_QUESTION_MARK:
3773      repeat_min = 0;      repeat_min = 0;
3774      repeat_max = 1;      repeat_max = 1;
3775    
# Line 2952  for (;; ptr++) Line 3804  for (;; ptr++)
3804      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
3805      repeat type to the non-default. */      repeat type to the non-default. */
3806    
3807      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3808        {        {
3809        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3810        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3811        ptr++;        ptr++;
3812        }        }
3813      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3814        {        {
3815        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3816        ptr++;        ptr++;
# Line 3073  for (;; ptr++) Line 3925  for (;; ptr++)
3925    
3926        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3927    
3928          /*--------------------------------------------------------------------*/
3929          /* This code is obsolete from release 8.00; the restriction was finally
3930          removed: */
3931    
3932        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3933        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3934    
3935        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3936          /*--------------------------------------------------------------------*/
3937    
3938        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3939    
# Line 3223  for (;; ptr++) Line 4080  for (;; ptr++)
4080          goto END_REPEAT;          goto END_REPEAT;
4081          }          }
4082    
4083          /*--------------------------------------------------------------------*/
4084          /* This code is obsolete from release 8.00; the restriction was finally
4085          removed: */
4086    
4087        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4088        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4089    
4090        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4091          /*--------------------------------------------------------------------*/
4092    
4093        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4094          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3262  for (;; ptr++) Line 4124  for (;; ptr++)
4124          goto FAILED;          goto FAILED;
4125          }          }
4126    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4127        /* 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
4128        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
4129        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 3292  for (;; ptr++) Line 4146  for (;; ptr++)
4146    
4147        if (repeat_min == 0)        if (repeat_min == 0)
4148          {          {
4149          /* 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
4150          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4151    
4152          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4153          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4154          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4155          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4156          the start of the whole regex. Temporarily terminate the pattern while          **   }
4157          doing this. */  
4158            However, that fails when a group is referenced as a subroutine from
4159            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4160            so that it is skipped on execution. As we don't have a list of which
4161            groups are referenced, we cannot do this selectively.
4162    
4163            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4164            and do no more at this point. However, we do need to adjust any
4165            OP_RECURSE calls inside the group that refer to the group itself or any
4166            internal or forward referenced group, because the offset is from the
4167            start of the whole regex. Temporarily terminate the pattern while doing
4168            this. */
4169    
4170          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4171            {            {
4172            *code = OP_END;            *code = OP_END;
4173            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4174            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4175            code++;            code++;
4176              if (repeat_max == 0)
4177                {
4178                *previous++ = OP_SKIPZERO;
4179                goto END_REPEAT;
4180                }
4181            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4182            }            }
4183    
# Line 3358  for (;; ptr++) Line 4222  for (;; ptr++)
4222          if (repeat_min > 1)          if (repeat_min > 1)
4223            {            {
4224            /* 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
4225            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4226              potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4227              integer type when available, otherwise double. */
4228    
4229            if (lengthptr != NULL)            if (lengthptr != NULL)
4230              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4231                int delta = (repeat_min - 1)*length_prevgroup;
4232                if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4233                      (INT64_OR_DOUBLE)length_prevgroup >
4234                        (INT64_OR_DOUBLE)INT_MAX ||
4235                    OFLOW_MAX - *lengthptr < delta)
4236                  {
4237                  *errorcodeptr = ERR20;
4238                  goto FAILED;
4239                  }
4240                *lengthptr += delta;
4241                }
4242    
4243            /* This is compiling for real */            /* This is compiling for real */
4244    
# Line 3399  for (;; ptr++) Line 4276  for (;; ptr++)
4276          /* 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
4277          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
4278          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
4279          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4280            paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4281            a 64-bit integer type when available, otherwise double. */
4282    
4283          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4284            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4285              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4286                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4287              if ((INT64_OR_DOUBLE)repeat_max *
4288                    (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4289                      > (INT64_OR_DOUBLE)INT_MAX ||
4290                  OFLOW_MAX - *lengthptr < delta)
4291                {
4292                *errorcodeptr = ERR20;
4293                goto FAILED;
4294                }
4295              *lengthptr += delta;
4296              }
4297    
4298          /* This is compiling for real */          /* This is compiling for real */
4299    
# Line 3485  for (;; ptr++) Line 4375  for (;; ptr++)
4375          }          }
4376        }        }
4377    
4378        /* If previous is OP_FAIL, it was generated by an empty class [] in
4379        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4380        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4381        error above. We can just ignore the repeat in JS case. */
4382    
4383        else if (*previous == OP_FAIL) goto END_REPEAT;
4384    
4385      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4386    
4387      else      else
# Line 3509  for (;; ptr++) Line 4406  for (;; ptr++)
4406      if (possessive_quantifier)      if (possessive_quantifier)
4407        {        {
4408        int len;        int len;
4409        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4410            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4411            tempcode += _pcre_OP_lengths[*tempcode] +
4412              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4413    
4414          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4415            {
4416          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4417    #ifdef SUPPORT_UTF8
4418            if (utf8 && tempcode[-1] >= 0xc0)
4419              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4420    #endif
4421            }
4422    
4423        len = code - tempcode;        len = code - tempcode;
4424        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4425          {          {
# Line 3530  for (;; ptr++) Line 4438  for (;; ptr++)
4438          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4439          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4440    
4441            /* Because we are moving code along, we must ensure that any
4442            pending recursive references are updated. */
4443    
4444          default:          default:
4445            *code = OP_END;
4446            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4447          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4448          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4449          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3555  for (;; ptr++) Line 4468  for (;; ptr++)
4468      /* ===================================================================*/      /* ===================================================================*/
4469      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4470      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4471      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. */  
4472    
4473      case '(':      case CHAR_LEFT_PARENTHESIS:
4474      newoptions = options;      newoptions = options;
4475      skipbytes = 0;      skipbytes = 0;
4476      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4477      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4478        reset_bracount = FALSE;
4479    
4480      if (*(++ptr) == '?')      /* First deal with various "verbs" that can be introduced by '*'. */
4481    
4482        if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4483          {
4484          int i, namelen;
4485          const char *vn = verbnames;
4486          const uschar *name = ++ptr;
4487          previous = NULL;
4488          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4489          if (*ptr == CHAR_COLON)
4490            {
4491            *errorcodeptr = ERR59;   /* Not supported */
4492            goto FAILED;
4493            }
4494          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4495            {
4496            *errorcodeptr = ERR60;
4497            goto FAILED;
4498            }
4499          namelen = ptr - name;
4500          for (i = 0; i < verbcount; i++)
4501            {
4502            if (namelen == verbs[i].len &&
4503                strncmp((char *)name, vn, namelen) == 0)
4504              {
4505              /* Check for open captures before ACCEPT */
4506    
4507              if (verbs[i].op == OP_ACCEPT)
4508                {
4509                open_capitem *oc;
4510                cd->had_accept = TRUE;
4511                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4512                  {
4513                  *code++ = OP_CLOSE;
4514                  PUT2INC(code, 0, oc->number);
4515                  }
4516                }
4517              *code++ = verbs[i].op;
4518              break;
4519              }
4520            vn += verbs[i].len + 1;
4521            }
4522          if (i < verbcount) continue;
4523          *errorcodeptr = ERR60;
4524          goto FAILED;
4525          }
4526    
4527        /* Deal with the extended parentheses; all are introduced by '?', and the
4528        appearance of any of them means that this is not a capturing group. */
4529    
4530        else if (*ptr == CHAR_QUESTION_MARK)
4531        {        {
4532        int i, set, unset, namelen;        int i, set, unset, namelen;
4533        int *optset;        int *optset;
# Line 3574  for (;; ptr++) Line 4536  for (;; ptr++)
4536    
4537        switch (*(++ptr))        switch (*(++ptr))
4538          {          {
4539          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4540          ptr++;          ptr++;
4541          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4542          if (*ptr == 0)          if (*ptr == 0)
4543            {            {
4544            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3586  for (;; ptr++) Line 4548  for (;; ptr++)
4548    
4549    
4550          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4551          case ':':                 /* Non-capturing bracket */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4552            reset_bracount = TRUE;
4553            /* Fall through */
4554    
4555            /* ------------------------------------------------------------ */
4556            case CHAR_COLON:          /* Non-capturing bracket */
4557          bravalue = OP_BRA;          bravalue = OP_BRA;
4558          ptr++;          ptr++;
4559          break;          break;
4560    
4561    
4562          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4563          case '(':          case CHAR_LEFT_PARENTHESIS:
4564          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4565    
4566          /* 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 3613  for (;; ptr++) Line 4580  for (;; ptr++)
4580          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4581          including assertions, are processed. */          including assertions, are processed. */
4582    
4583          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4584                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4585            break;            break;
4586    
4587          /* 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 3621  for (;; ptr++) Line 4589  for (;; ptr++)
4589    
4590          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
4591          skipbytes = 3;          skipbytes = 3;
4592            refsign = -1;
4593    
4594          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4595    
4596          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4597            {            {
4598            terminator = -1;            terminator = -1;
4599            ptr += 2;            ptr += 2;
# Line 3634  for (;; ptr++) Line 4603  for (;; ptr++)
4603          /* 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
4604          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4605    
4606          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4607            {            {
4608            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4609            ptr++;            ptr++;
4610            }            }
4611          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4612            {            {
4613            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4614            ptr++;            ptr++;
4615            }            }
4616          else terminator = 0;          else
4617              {
4618              terminator = 0;
4619              if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4620              }
4621    
4622          /* 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 */
4623    
# Line 3663  for (;; ptr++) Line 4636  for (;; ptr++)
4636            {            {
4637            if (recno >= 0)            if (recno >= 0)
4638              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4639                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4640            ptr++;            ptr++;
4641            }            }
4642          namelen = ptr - name;          namelen = ptr - name;
4643    
4644          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4645                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4646            {            {
4647            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4648            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3680  for (;; ptr++) Line 4654  for (;; ptr++)
4654          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
4655    
4656          /* In the real compile we do the work of looking for the actual          /* In the real compile we do the work of looking for the actual
4657          reference. */          reference. If the string started with "+" or "-" we require the rest to
4658            be digits, in which case recno will be set. */
4659    
4660            if (refsign > 0)
4661              {
4662              if (recno <= 0)
4663                {
4664                *errorcodeptr = ERR58;
4665                goto FAILED;
4666                }
4667              recno = (refsign == CHAR_MINUS)?
4668                cd->bracount - recno + 1 : recno +cd->bracount;
4669              if (recno <= 0 || recno > cd->final_bracount)
4670                {
4671                *errorcodeptr = ERR15;
4672                goto FAILED;
4673                }
4674              PUT2(code, 2+LINK_SIZE, recno);
4675              break;
4676              }
4677    
4678            /* Otherwise (did not start with "+" or "-"), start by looking for the
4679            name. If we find a name, add one to the opcode to change OP_CREF or
4680            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4681            except they record that the reference was originally to a name. The
4682            information is used to check duplicate names. */
4683    
4684          slot = cd->name_table;          slot = cd->name_table;
4685          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 3695  for (;; ptr++) Line 4694  for (;; ptr++)
4694            {            {
4695            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4696            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4697              code[1+LINK_SIZE]++;
4698            }            }
4699    
4700          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4701    
4702          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4703                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4704            {            {
4705            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4706              code[1+LINK_SIZE]++;
4707            }            }
4708    
4709          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 3720  for (;; ptr++) Line 4721  for (;; ptr++)
4721          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4722          specific group number. */          specific group number. */
4723    
4724          else if (*name == 'R')          else if (*name == CHAR_R)
4725            {            {
4726            recno = 0;            recno = 0;
4727            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 3730  for (;; ptr++) Line 4731  for (;; ptr++)
4731                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4732                goto FAILED;                goto FAILED;
4733                }                }
4734              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4735              }              }
4736            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4737            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 3740  for (;; ptr++) Line 4741  for (;; ptr++)
4741          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4742          false. */          false. */
4743    
4744          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4745            {            {
4746            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4747            skipbytes = 1;            skipbytes = 1;
4748            }            }
4749    
4750          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4751            in the second pass here, so final_bracount is set. */
4752    
4753          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4754            {            {
4755            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4756            }            }
# Line 3764  for (;; ptr++) Line 4766  for (;; ptr++)
4766    
4767    
4768          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4769          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4770          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4771          ptr++;          ptr++;
4772          break;          break;
4773    
4774    
4775          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4776          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4777          ptr++;          ptr++;
4778            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4779              {
4780              *code++ = OP_FAIL;
4781              previous = NULL;
4782              continue;
4783              }
4784            bravalue = OP_ASSERT_NOT;
4785          break;          break;
4786    
4787    
4788          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4789          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4790          switch (ptr[1])          switch (ptr[1])
4791            {            {
4792            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4793            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4794            ptr += 2;            ptr += 2;
4795            break;            break;
4796    
4797            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4798            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4799            ptr += 2;            ptr += 2;
4800            break;            break;
# Line 3801  for (;; ptr++) Line 4809  for (;; ptr++)
4809    
4810    
4811          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4812          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4813          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4814          ptr++;          ptr++;
4815          break;          break;
4816    
4817    
4818          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4819          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4820          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4821          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4822          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4823            {            {
4824            int n = 0;            int n = 0;
4825            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4826              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4827            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4828              {              {
4829              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4830              goto FAILED;              goto FAILED;
# Line 3836  for (;; ptr++) Line 4844  for (;; ptr++)
4844    
4845    
4846          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4847          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4848          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4849                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4850            {            {
4851            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4852            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4853            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4854            }            }
4855          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4856            {            {
4857            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4858            goto FAILED;            goto FAILED;
# Line 3853  for (;; ptr++) Line 4862  for (;; ptr++)
4862    
4863          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4864          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4865          case '\'':          case CHAR_APOSTROPHE:
4866            {            {
4867            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4868                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4869            name = ++ptr;            name = ++ptr;
4870    
4871            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 3886  for (;; ptr++) Line 4896  for (;; ptr++)
4896                }                }
4897              }              }
4898    
4899            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4900              alphabetical order. Duplicate names for different numbers are
4901              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4902              number are always OK. (An existing number can be re-used if (?|
4903              appears in the pattern.) In either event, a duplicate name results in
4904              a duplicate entry in the table, even if the number is the same. This
4905              is because the number of names, and hence the table size, is computed
4906              in the pre-compile, and it affects various numbers and pointers which
4907              would all have to be modified, and the compiled code moved down, if
4908              duplicates with the same number were omitted from the table. This
4909              doesn't seem worth the hassle. However, *different* names for the
4910              same number are not permitted. */
4911    
4912            else            else
4913              {              {
4914                BOOL dupname = FALSE;
4915              slot = cd->name_table;              slot = cd->name_table;
4916    
4917              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4918                {                {
4919                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 3898  for (;; ptr++) Line 4921  for (;; ptr++)
4921                  {                  {
4922                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4923                    {                    {
4924                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4925                          (options & PCRE_DUPNAMES) == 0)
4926                      {                      {
4927                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4928                      goto FAILED;                      goto FAILED;
4929                      }                      }
4930                      else dupname = TRUE;
4931                    }                    }
4932                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4933                  }                  }
4934    
4935                  /* Make space in the table and break the loop for an earlier
4936                  name. For a duplicate or later name, carry on. We do this for
4937                  duplicates so that in the simple case (when ?(| is not used) they
4938                  are in order of their numbers. */
4939    
4940                if (crc < 0)                if (crc < 0)
4941                  {                  {
4942                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4943                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4944                  break;                  break;
4945                  }                  }
4946    
4947                  /* Continue the loop for a later or duplicate name */
4948    
4949                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4950                }                }
4951    
4952                /* For non-duplicate names, check for a duplicate number before
4953                adding the new name. */
4954    
4955                if (!dupname)
4956                  {
4957                  uschar *cslot = cd->name_table;
4958                  for (i = 0; i < cd->names_found; i++)
4959                    {
4960                    if (cslot != slot)
4961                      {
4962                      if (GET2(cslot, 0) == cd->bracount + 1)
4963                        {
4964                        *errorcodeptr = ERR65;
4965                        goto FAILED;
4966                        }
4967                      }
4968                    else i--;
4969                    cslot += cd->name_entry_size;
4970                    }
4971                  }
4972    
4973              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4974              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
4975              slot[2+namelen] = 0;              slot[2+namelen] = 0;
4976              }              }
4977            }            }
4978    
4979          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
4980            encountered. */
4981    
         ptr++;                    /* Move past > or ' */  
4982          cd->names_found++;          cd->names_found++;
4983            ptr++;                    /* Move past > or ' */
4984          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
4985    
4986    
4987          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4988          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4989          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4990          is_recurse = TRUE;          is_recurse = TRUE;
4991          /* Fall through */          /* Fall through */
4992    
4993          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4994          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4995          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4996            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4997            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4998    
4999          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5000          name = ++ptr;          name = ++ptr;
# Line 3948  for (;; ptr++) Line 5006  for (;; ptr++)
5006    
5007          if (lengthptr != NULL)          if (lengthptr != NULL)
5008            {            {
5009              if (namelen == 0)
5010                {
5011                *errorcodeptr = ERR62;
5012                goto FAILED;
5013                }
5014            if (*ptr != terminator)            if (*ptr != terminator)
5015              {              {
5016              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 3961  for (;; ptr++) Line 5024  for (;; ptr++)
5024            recno = 0;            recno = 0;
5025            }            }
5026    
5027          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
5028            first, and then check that we have reached the end of the name in the
5029            table. That way, if the name that is longer than any in the table,
5030            the comparison will fail without reading beyond the table entry. */
5031    
5032          else          else
5033            {            {
5034            slot = cd->name_table;            slot = cd->name_table;
5035            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5036              {              {
5037              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
5038                    slot[2+namelen] == 0)
5039                  break;
5040              slot += cd->name_entry_size;              slot += cd->name_entry_size;
5041              }              }
5042    
# Line 3977  for (;; ptr++) Line 5045  for (;; ptr++)
5045              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5046              }              }
5047            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5048                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5049                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5050              {              {
5051              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 3993  for (;; ptr++) Line 5061  for (;; ptr++)
5061    
5062    
5063          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5064          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5065          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5066          /* Fall through */          /* Fall through */
5067    
5068    
5069          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5070          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5071          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
5072            case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
5073            {            {
5074            const uschar *called;            const uschar *called;
5075              terminator = CHAR_RIGHT_PARENTHESIS;
5076    
5077              /* Come here from the \g<...> and \g'...' code (Oniguruma
5078              compatibility). However, the syntax has been checked to ensure that
5079              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5080              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5081              ever be taken. */
5082    
5083              HANDLE_NUMERICAL_RECURSION:
5084    
5085              if ((refsign = *ptr) == CHAR_PLUS)
5086                {
5087                ptr++;
5088                if ((digitab[*ptr] & ctype_digit) == 0)
5089                  {
5090                  *errorcodeptr = ERR63;
5091                  goto FAILED;
5092                  }
5093                }
5094              else if (refsign == CHAR_MINUS)
5095                {
5096                if ((digitab[ptr[1]] & ctype_digit) == 0)
5097                  goto OTHER_CHAR_AFTER_QUERY;
5098                ptr++;
5099                }
5100    
5101            recno = 0;            recno = 0;
5102            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5103              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5104            if (*ptr != ')')  
5105              if (*ptr != terminator)
5106              {              {
5107              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5108              goto FAILED;              goto FAILED;
5109              }              }
5110    
5111              if (refsign == CHAR_MINUS)
5112                {
5113                if (recno == 0)
5114                  {
5115                  *errorcodeptr = ERR58;
5116                  goto FAILED;
5117                  }
5118                recno = cd->bracount - recno + 1;
5119                if (recno <= 0)
5120                  {
5121                  *errorcodeptr = ERR15;
5122                  goto FAILED;
5123                  }
5124                }
5125              else if (refsign == CHAR_PLUS)
5126                {
5127                if (recno == 0)
5128                  {
5129                  *errorcodeptr = ERR58;
5130                  goto FAILED;
5131                  }
5132                recno += cd->bracount;
5133                }
5134    
5135            /* Come here from code above that handles a named recursion */            /* Come here from code above that handles a named recursion */
5136    
5137            HANDLE_RECURSION:            HANDLE_RECURSION:
# Line 4029  for (;; ptr++) Line 5149  for (;; ptr++)
5149            if (lengthptr == NULL)            if (lengthptr == NULL)
5150              {              {
5151              *code = OP_END;              *code = OP_END;
5152              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5153                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5154    
5155              /* Forward reference */              /* Forward reference */
5156    
5157              if (called == NULL)              if (called == NULL)
5158                {                {
5159                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5160                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5161                  {                  {
5162                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5163                  goto FAILED;                  goto FAILED;
5164                  }                  }
5165    
5166                  /* Fudge the value of "called" so that when it is inserted as an
5167                  offset below, what it actually inserted is the reference number
5168                  of the group. */
5169    
5170                called = cd->start_code + recno;                called = cd->start_code + recno;
5171                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);                PUTINC(cd->hwm, 0, code + 2 + LINK_SIZE - cd->start_code);
5172                }                }
# Line 4084  for (;; ptr++) Line 5210  for (;; ptr++)
5210    
5211          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5212          default:              /* Other characters: check option setting */          default:              /* Other characters: check option setting */
5213            OTHER_CHAR_AFTER_QUERY:
5214          set = unset = 0;          set = unset = 0;
5215          optset = &set;          optset = &set;
5216    
5217          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5218            {            {
5219            switch (*ptr++)            switch (*ptr++)
5220              {              {
5221              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5222    
5223              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5224              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5225              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5226              break;              break;
5227    
5228              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5229              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5230              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5231              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5232              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5233              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5234    
5235              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5236                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4134  for (;; ptr++) Line 5261  for (;; ptr++)
5261          both phases.          both phases.
5262    
5263          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
5264          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
5265          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. */  
5266    
5267          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5268            {            {
5269            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5270                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5271              {              {
5272              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5273              }              }
5274           else            else
5275              {              {
5276              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))              if ((options & PCRE_IMS) != (newoptions & PCRE_IMS))
5277                {                {
5278                *code++ = OP_OPT;                *code++ = OP_OPT;
5279                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5280                }                }
   
             /* 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;  
5281              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5282              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5283              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5284              }              }
5285    
5286              /* Change options at this level, and pass them back for use
5287              in subsequent branches. When not at the start of the pattern, this
5288              information is also necessary so that a resetting item can be
5289              compiled at the end of a group (if we are in a group). */
5290    
5291              *optionsptr = options = newoptions;
5292            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5293            continue;              /* It is complete */            continue;              /* It is complete */
5294            }            }
# Line 4218  for (;; ptr++) Line 5342  for (;; ptr++)
5342           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
5343           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5344            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5345             reset_bracount,               /* True if (?| group */
5346           skipbytes,                    /* Skip over bracket number */           skipbytes,                    /* Skip over bracket number */
5347           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
5348           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
# Line 4234  for (;; ptr++) Line 5359  for (;; ptr++)
5359      is on the bracket. */      is on the bracket. */
5360    
5361      /* If this is a conditional bracket, check that there are no more than      /* If this is a conditional bracket, check that there are no more than
5362      two branches in the group, or just one if it's a DEFINE group. */      two branches in the group, or just one if it's a DEFINE group. We do this
5363        in the real compile phase, not in the pre-pass, where the whole group may
5364        not be available. */
5365    
5366      if (bravalue == OP_COND)      if (bravalue == OP_COND && lengthptr == NULL)
5367        {        {
5368        uschar *tc = code;        uschar *tc = code;
5369        int condcount = 0;        int condcount = 0;
# Line 4277  for (;; ptr++) Line 5404  for (;; ptr++)
5404    
5405      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5406    
5407      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5408        {        {
5409        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5410        goto FAILED;        goto FAILED;
5411        }        }
5412    
5413      /* 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,
5414      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
5415      just the brackets