/[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 168 by ph10, Tue May 29 15:18:18 2007 UTC revision 457 by ph10, Sat Oct 3 16:24:08 2009 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2009 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    
# Line 58  used by pcretest. DEBUG is not defined w Line 61  used by pcretest. DEBUG is not defined w
61  #endif  #endif
62    
63    
64    /* Macro for setting individual bits in class bitmaps. */
65    
66    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
67    
68    /* Maximum length value to check against when making sure that the integer that
69    holds the compiled pattern length does not overflow. We make it a bit less than
70    INT_MAX to allow for adding in group terminating bytes, so that we don't have
71    to check them every time. */
72    
73    #define OFLOW_MAX (INT_MAX - 20)
74    
75    
76  /*************************************************  /*************************************************
77  *      Code parameters and static tables         *  *      Code parameters and static tables         *
78  *************************************************/  *************************************************/
# Line 82  are simple data values; negative values Line 97  are simple data values; negative values
97  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
98  is invalid. */  is invalid. */
99    
100  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
101    
102    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
103    in UTF-8 mode. */
104    
105  static const short int escapes[] = {  static const short int escapes[] = {
106       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
107       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
108     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
109       0,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
110  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
111  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
112     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
113       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
114  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
115       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
116         -ESC_D,                  -ESC_E,
117         0,                       -ESC_G,
118         -ESC_H,                  0,
119         0,                       -ESC_K,
120         0,                       0,
121         0,                       0,
122         -ESC_P,                  -ESC_Q,
123         -ESC_R,                  -ESC_S,
124         0,                       0,
125         -ESC_V,                  -ESC_W,
126         -ESC_X,                  0,
127         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
128         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
129         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
130         CHAR_GRAVE_ACCENT,       7,
131         -ESC_b,                  0,
132         -ESC_d,                  ESC_e,
133         ESC_f,                   0,
134         -ESC_h,                  0,
135         0,                       -ESC_k,
136         0,                       0,
137         ESC_n,                   0,
138         -ESC_p,                  0,
139         ESC_r,                   -ESC_s,
140         ESC_tee,                 0,
141         -ESC_v,                  -ESC_w,
142         0,                       0,
143         -ESC_z
144  };  };
145    
146  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
147    
148    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
149    
150  static const short int escapes[] = {  static const short int escapes[] = {
151  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
152  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 106  static const short int escapes[] = { Line 156  static const short int escapes[] = {
156  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
157  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
158  /*  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,
159  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
160  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
161  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
162  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
163  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
164  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
165  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
166  /*  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,
167  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
168  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
169  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
170  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
171  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
172  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
173  /*  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 175  static const short int escapes[] = {
175  #endif  #endif
176    
177    
178  /* 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
179  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
180  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
181    string is built from string macros so that it works in UTF-8 mode on EBCDIC
182  static const char *const posix_names[] = {  platforms. */
183    "alpha", "lower", "upper",  
184    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
185    "print", "punct", "space", "word",  "xdigit" };    int   len;
186      int   op;
187    } verbitem;
188    
189    static const char verbnames[] =
190      STRING_ACCEPT0
191      STRING_COMMIT0
192      STRING_F0
193      STRING_FAIL0
194      STRING_PRUNE0
195      STRING_SKIP0
196      STRING_THEN;
197    
198    static const verbitem verbs[] = {
199      { 6, OP_ACCEPT },
200      { 6, OP_COMMIT },
201      { 1, OP_FAIL },
202      { 4, OP_FAIL },
203      { 5, OP_PRUNE },
204      { 4, OP_SKIP  },
205      { 4, OP_THEN  }
206    };
207    
208    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
209    
210    
211    /* Tables of names of POSIX character classes and their lengths. The names are
212    now all in a single string, to reduce the number of relocations when a shared
213    library is dynamically loaded. The list of lengths is terminated by a zero
214    length entry. The first three must be alpha, lower, upper, as this is assumed
215    for handling case independence. */
216    
217    static const char posix_names[] =
218      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
219      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
220      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
221      STRING_word0  STRING_xdigit;
222    
223  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
224    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 257  static const int posix_class_maps[] = {
257  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
258  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
259  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
260  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
261    the number of relocations needed when a shared library is loaded dynamically,
262  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
263    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
264    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
265    "\\c at end of pattern",  because these strings are used only when there is a compilation error. */
266    "unrecognized character follows \\",  
267    "numbers out of order in {} quantifier",  static const char error_texts[] =
268      "no error\0"
269      "\\ at end of pattern\0"
270      "\\c at end of pattern\0"
271      "unrecognized character follows \\\0"
272      "numbers out of order in {} quantifier\0"
273    /* 5 */    /* 5 */
274    "number too big in {} quantifier",    "number too big in {} quantifier\0"
275    "missing terminating ] for character class",    "missing terminating ] for character class\0"
276    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
277    "range out of order in character class",    "range out of order in character class\0"
278    "nothing to repeat",    "nothing to repeat\0"
279    /* 10 */    /* 10 */
280    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
281    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
282    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
283    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
284    "missing )",    "missing )\0"
285    /* 15 */    /* 15 */
286    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
287    "erroffset passed as NULL",    "erroffset passed as NULL\0"
288    "unknown option bit(s) set",    "unknown option bit(s) set\0"
289    "missing ) after comment",    "missing ) after comment\0"
290    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
291    /* 20 */    /* 20 */
292    "regular expression too large",    "regular expression is too large\0"
293    "failed to get memory",    "failed to get memory\0"
294    "unmatched parentheses",    "unmatched parentheses\0"
295    "internal error: code overflow",    "internal error: code overflow\0"
296    "unrecognized character after (?<",    "unrecognized character after (?<\0"
297    /* 25 */    /* 25 */
298    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
299    "malformed number or name after (?(",    "malformed number or name after (?(\0"
300    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
301    "assertion expected after (?(",    "assertion expected after (?(\0"
302    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
303    /* 30 */    /* 30 */
304    "unknown POSIX class name",    "unknown POSIX class name\0"
305    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
306    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
307    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
308    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
309    /* 35 */    /* 35 */
310    "invalid condition (?(0)",    "invalid condition (?(0)\0"
311    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
312    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
313    "number after (?C is > 255",    "number after (?C is > 255\0"
314    "closing ) for (?C expected",    "closing ) for (?C expected\0"
315    /* 40 */    /* 40 */
316    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
317    "unrecognized character after (?P",    "unrecognized character after (?P\0"
318    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
319    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
320    "invalid UTF-8 string",    "invalid UTF-8 string\0"
321    /* 45 */    /* 45 */
322    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
323    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
324    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
325    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
326    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
327    /* 50 */    /* 50 */
328    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
329    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
330    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
331    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
332    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
333    /* 55 */    /* 55 */
334    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
335    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
336    "\\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"
337    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
338  };    "(*VERB) with an argument is not supported\0"
339      /* 60 */
340      "(*VERB) not recognized\0"
341      "number is too big\0"
342      "subpattern name expected\0"
343      "digit expected after (?+\0"
344      "] is an invalid data character in JavaScript compatibility mode\0"
345      /* 65 */
346      "different names for subpatterns of the same number are not allowed";
347    
348    
349  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
# Line 263  For convenience, we use the same bit def Line 362  For convenience, we use the same bit def
362    
363  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
364    
365  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
366    
367    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
368    UTF-8 mode. */
369    
370  static const unsigned char digitab[] =  static const unsigned char digitab[] =
371    {    {
372    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 299  static const unsigned char digitab[] = Line 402  static const unsigned char digitab[] =
402    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
403    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
404    
405  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
406    
407    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
408    
409  static const unsigned char digitab[] =  static const unsigned char digitab[] =
410    {    {
411    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 374  static const unsigned char ebcdic_charta Line 480  static const unsigned char ebcdic_charta
480  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
481    
482  static BOOL  static BOOL
483    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
484      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
485    
486    
487    
488    /*************************************************
489    *            Find an error text                  *
490    *************************************************/
491    
492    /* The error texts are now all in one long string, to save on relocations. As
493    some of the text is of unknown length, we can't use a table of offsets.
494    Instead, just count through the strings. This is not a performance issue
495    because it happens only when there has been a compilation error.
496    
497    Argument:   the error number
498    Returns:    pointer to the error string
499    */
500    
501    static const char *
502    find_error_text(int n)
503    {
504    const char *s = error_texts;
505    for (; n > 0; n--) while (*s++ != 0) {};
506    return s;
507    }
508    
509    
510  /*************************************************  /*************************************************
# Line 400  Arguments: Line 528  Arguments:
528    
529  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
530                   negative => a special escape sequence                   negative => a special escape sequence
531                   on error, errorptr is set                   on error, errorcodeptr is set
532  */  */
533    
534  static int  static int
# Line 418  ptr--;                            /* Set Line 546  ptr--;                            /* Set
546    
547  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
548    
549  /* 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
550  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.
551  Otherwise further processing may be required. */  Otherwise further processing may be required. */
552    
553  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
554  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
555  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
556    
557  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
558  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
559  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
560  #endif  #endif
561    
# Line 443  else Line 571  else
571      /* 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
572      error. */      error. */
573    
574      case 'l':      case CHAR_l:
575      case 'L':      case CHAR_L:
576      case 'N':      case CHAR_N:
577      case 'u':      case CHAR_u:
578      case 'U':      case CHAR_U:
579      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
580      break;      break;
581    
582      /* \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:
583      is an absolute backreference. If negative, it is a relative backreference.  
584      This is a Perl 5.10 feature. */      (1) A number, either plain or braced. If positive, it is an absolute
585        backreference. If negative, it is a relative backreference. This is a Perl
586        5.10 feature.
587    
588        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
589        is part of Perl's movement towards a unified syntax for back references. As
590        this is synonymous with \k{name}, we fudge it up by pretending it really
591        was \k.
592    
593        (3) For Oniguruma compatibility we also support \g followed by a name or a
594        number either in angle brackets or in single quotes. However, these are
595        (possibly recursive) subroutine calls, _not_ backreferences. Just return
596        the -ESC_g code (cf \k). */
597    
598        case CHAR_g:
599        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
600          {
601          c = -ESC_g;
602          break;
603          }
604    
605        /* Handle the Perl-compatible cases */
606    
607      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
608        {        {
609          const uschar *p;
610          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
611            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
612          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
613            {
614            c = -ESC_k;
615            break;
616            }
617        braced = TRUE;        braced = TRUE;
618        ptr++;        ptr++;
619        }        }
620      else braced = FALSE;      else braced = FALSE;
621    
622      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
623        {        {
624        negated = TRUE;        negated = TRUE;
625        ptr++;        ptr++;
# Line 472  else Line 628  else
628    
629      c = 0;      c = 0;
630      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
631        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
632    
633        if (c < 0)   /* Integer overflow */
634          {
635          *errorcodeptr = ERR61;
636          break;
637          }
638    
639      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
640        {        {
641        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
642        return 0;        break;
643          }
644    
645        if (c == 0)
646          {
647          *errorcodeptr = ERR58;
648          break;
649        }        }
650    
651      if (negated)      if (negated)
# Line 485  else Line 653  else
653        if (c > bracount)        if (c > bracount)
654          {          {
655          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
656          return 0;          break;
657          }          }
658        c = bracount - (c - 1);        c = bracount - (c - 1);
659        }        }
# Line 505  else Line 673  else
673      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
674      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
675    
676      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:
677      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
678    
679      if (!isclass)      if (!isclass)
680        {        {
681        oldptr = ptr;        oldptr = ptr;
682        c -= '0';        c -= CHAR_0;
683        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
684          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
685          if (c < 0)    /* Integer overflow */
686            {
687            *errorcodeptr = ERR61;
688            break;
689            }
690        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
691          {          {
692          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 526  else Line 699  else
699      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.
700      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
701    
702      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
703        {        {
704        ptr--;        ptr--;
705        c = 0;        c = 0;
# Line 539  else Line 712  else
712      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
713      than 3 octal digits. */      than 3 octal digits. */
714    
715      case '0':      case CHAR_0:
716      c -= '0';      c -= CHAR_0;
717      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
718          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
719      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
720      break;      break;
721    
# Line 550  else Line 723  else
723      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
724      treated as a data character. */      treated as a data character. */
725    
726      case 'x':      case CHAR_x:
727      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
728        {        {
729        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
730        int count = 0;        int count = 0;
# Line 560  else Line 733  else
733        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
734          {          {
735          register int cc = *pt++;          register int cc = *pt++;
736          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
737          count++;          count++;
738    
739  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
740          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
741          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
742  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
743          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
744          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
745  #endif  #endif
746          }          }
747    
748        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
749          {          {
750          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
751          ptr = pt;          ptr = pt;
# Line 588  else Line 761  else
761      c = 0;      c = 0;
762      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
763        {        {
764        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
765        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
766  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
767        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
768        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
769  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
770        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
771        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
772  #endif  #endif
773        }        }
774      break;      break;
# Line 604  else Line 777  else
777      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
778      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
779    
780      case 'c':      case CHAR_c:
781      c = *(++ptr);      c = *(++ptr);
782      if (c == 0)      if (c == 0)
783        {        {
784        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
785        return 0;        break;
786        }        }
787    
788  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
789      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
790      c ^= 0x40;      c ^= 0x40;
791  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
792      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
793      c ^= 0xC0;      c ^= 0xC0;
794  #endif  #endif
795      break;      break;
796    
797      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
798      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
799      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
800      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
801      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
802    
803      default:      default:
804      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 678  if (c == 0) goto ERROR_RETURN; Line 851  if (c == 0) goto ERROR_RETURN;
851  /* \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
852  negation. */  negation. */
853    
854  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
855    {    {
856    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
857      {      {
858      *negptr = TRUE;      *negptr = TRUE;
859      ptr++;      ptr++;
860      }      }
861    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
862      {      {
863      c = *(++ptr);      c = *(++ptr);
864      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
865      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
866      name[i] = c;      name[i] = c;
867      }      }
868    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
869    name[i] = 0;    name[i] = 0;
870    }    }
871    
# Line 714  top = _pcre_utt_size; Line 887  top = _pcre_utt_size;
887  while (bot < top)  while (bot < top)
888    {    {
889    i = (bot + top) >> 1;    i = (bot + top) >> 1;
890    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
891    if (c == 0)    if (c == 0)
892      {      {
893      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 757  is_counted_repeat(const uschar *p) Line 930  is_counted_repeat(const uschar *p)
930  {  {
931  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
932  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
933  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
934    
935  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
936  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
937    
938  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
939  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
940    
941  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
942  }  }
943    
944    
# Line 798  int max = -1; Line 971  int max = -1;
971  /* 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
972  an integer overflow. */  an integer overflow. */
973    
974  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
975  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
976    {    {
977    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 808  if (min < 0 || min > 65535) Line 981  if (min < 0 || min > 65535)
981  /* 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.
982  Also, max must not be less than min. */  Also, max must not be less than min. */
983    
984  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
985    {    {
986    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
987      {      {
988      max = 0;      max = 0;
989      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
990      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
991        {        {
992        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 838  return p; Line 1011  return p;
1011    
1012    
1013  /*************************************************  /*************************************************
1014  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1015  *************************************************/  *************************************************/
1016    
1017  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1018    top-level call starts at the beginning of the pattern. All other calls must
1019    start at a parenthesis. It scans along a pattern's text looking for capturing
1020  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
1021  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
1022  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
1023  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
1024  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1025    capturing group numbers - the (?| feature.
1026    
1027  Arguments:  Arguments:
1028    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1029    count        current count of capturing parens so far encountered    cd           compile background data
1030    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1031    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1032    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1033      count        pointer to the current capturing subpattern number (updated)
1034    
1035  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1036  */  */
1037    
1038  static int  static int
1039  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,
1040    BOOL xmode)    BOOL xmode, int *count)
1041  {  {
1042  const uschar *thisname;  uschar *ptr = *ptrptr;
1043    int start_count = *count;
1044    int hwm_count = start_count;
1045    BOOL dup_parens = FALSE;
1046    
1047  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1048    dealing with. The very first call may not start with a parenthesis. */
1049    
1050    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1051    {    {
1052    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1053          ptr[2] == CHAR_VERTICAL_LINE)
1054        {
1055        ptr += 3;
1056        dup_parens = TRUE;
1057        }
1058    
1059      /* Handle a normal, unnamed capturing parenthesis */
1060    
1061      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1062        {
1063        *count += 1;
1064        if (name == NULL && *count == lorn) return *count;
1065        ptr++;
1066        }
1067    
1068      /* Handle a condition. If it is an assertion, just carry on so that it
1069      is processed as normal. If not, skip to the closing parenthesis of the
1070      condition (there can't be any nested parens. */
1071    
1072      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1073        {
1074        ptr += 2;
1075        if (ptr[1] != CHAR_QUESTION_MARK)
1076          {
1077          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1078          if (*ptr != 0) ptr++;
1079          }
1080        }
1081    
1082      /* We have either (? or (* and not a condition */
1083    
1084      else
1085        {
1086        ptr += 2;
1087        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1088    
1089        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1090    
1091        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1092            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1093          {
1094          int term;
1095          const uschar *thisname;
1096          *count += 1;
1097          if (name == NULL && *count == lorn) return *count;
1098          term = *ptr++;
1099          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1100          thisname = ptr;
1101          while (*ptr != term) ptr++;
1102          if (name != NULL && lorn == ptr - thisname &&
1103              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1104            return *count;
1105          term++;
1106          }
1107        }
1108      }
1109    
1110    /* Past any initial parenthesis handling, scan for parentheses or vertical
1111    bars. */
1112    
1113    for (; *ptr != 0; ptr++)
1114      {
1115    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1116    
1117    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1118      {      {
1119      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1120      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1121        {        {
1122        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1123        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1124        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1125        }        }
1126      continue;      continue;
1127      }      }
1128    
1129    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1130      are handled for real. If the first character is '^', skip it. Also, if the
1131      first few characters (either before or after ^) are \Q\E or \E we skip them
1132      too. This makes for compatibility with Perl. Note the use of STR macros to
1133      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1134    
1135    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1136      {      {
1137      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1138        for (;;)
1139        {        {
1140        if (*ptr == '\\')        if (ptr[1] == CHAR_BACKSLASH)
1141          {          {
1142          if (*(++ptr) == 0) return -1;          if (ptr[2] == CHAR_E)
1143          if (*ptr == 'Q') for (;;)            ptr+= 2;
1144            else if (strncmp((const char *)ptr+2,
1145                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1146              ptr += 4;
1147            else
1148              break;
1149            }
1150          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1151            {
1152            negate_class = TRUE;
1153            ptr++;
1154            }
1155          else break;
1156          }
1157    
1158        /* If the next character is ']', it is a data character that must be
1159        skipped, except in JavaScript compatibility mode. */
1160    
1161        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1162            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1163          ptr++;
1164    
1165        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1166          {
1167          if (*ptr == 0) return -1;
1168          if (*ptr == CHAR_BACKSLASH)
1169            {
1170            if (*(++ptr) == 0) goto FAIL_EXIT;
1171            if (*ptr == CHAR_Q) for (;;)
1172            {            {
1173            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1174            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1175            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1176            }            }
1177          continue;          continue;
1178          }          }
# Line 905  for (; *ptr != 0; ptr++) Line 1182  for (; *ptr != 0; ptr++)
1182    
1183    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1184    
1185    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1186      {      {
1187      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1188      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1189      continue;      continue;
1190      }      }
1191    
1192    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1193    
1194    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1195      {      {
1196      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1197      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1198      continue;      if (*ptr == 0) goto FAIL_EXIT;
1199      }      }
1200    
1201    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1202    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1203        if (dup_parens && *count < hwm_count) *count = hwm_count;
1204        *ptrptr = ptr;
1205        return -1;
1206        }
1207    
1208    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1209        {
1210        if (*count > hwm_count) hwm_count = *count;
1211        *count = start_count;
1212        }
1213      }
1214    
1215    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1216         *ptr != '\'')  *ptrptr = ptr;
1217      continue;  return -1;
1218    }
1219    
1220    
1221    
1222    
1223    /*************************************************
1224    *       Find forward referenced subpattern       *
1225    *************************************************/
1226    
1227    count++;  /* This function scans along a pattern's text looking for capturing
1228    subpatterns, and counting them. If it finds a named pattern that matches the
1229    name it is given, it returns its number. Alternatively, if the name is NULL, it
1230    returns when it reaches a given numbered subpattern. This is used for forward
1231    references to subpatterns. We used to be able to start this scan from the
1232    current compiling point, using the current count value from cd->bracount, and
1233    do it all in a single loop, but the addition of the possibility of duplicate
1234    subpattern numbers means that we have to scan from the very start, in order to
1235    take account of such duplicates, and to use a recursive function to keep track
1236    of the different types of group.
1237    
1238    Arguments:
1239      cd           compile background data
1240      name         name to seek, or NULL if seeking a numbered subpattern
1241      lorn         name length, or subpattern number if name is NULL
1242      xmode        TRUE if we are in /x mode
1243    
1244    if (name == NULL && count == lorn) return count;  Returns:       the number of the found subpattern, or -1 if not found
1245    term = *ptr++;  */
1246    if (term == '<') term = '>';  
1247    thisname = ptr;  static int
1248    while (*ptr != term) ptr++;  find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1249    if (name != NULL && lorn == ptr - thisname &&  {
1250        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  uschar *ptr = (uschar *)cd->start_pattern;
1251      return count;  int count = 0;
1252    int rc;
1253    
1254    /* If the pattern does not start with an opening parenthesis, the first call
1255    to find_parens_sub() will scan right to the end (if necessary). However, if it
1256    does start with a parenthesis, find_parens_sub() will return when it hits the
1257    matching closing parens. That is why we have to have a loop. */
1258    
1259    for (;;)
1260      {
1261      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1262      if (rc > 0 || *ptr++ == 0) break;
1263    }    }
1264    
1265  return -1;  return rc;
1266  }  }
1267    
1268    
1269    
1270    
1271  /*************************************************  /*************************************************
1272  *      Find first significant op code            *  *      Find first significant op code            *
1273  *************************************************/  *************************************************/
# Line 1013  for (;;) Line 1333  for (;;)
1333    
1334    
1335  /*************************************************  /*************************************************
1336  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1337  *************************************************/  *************************************************/
1338    
1339  /* 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,
1340  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.
1341  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
1342    temporarily terminated with OP_END when this function is called.
1343    
1344    This function is called when a backward assertion is encountered, so that if it
1345    fails, the error message can point to the correct place in the pattern.
1346    However, we cannot do this when the assertion contains subroutine calls,
1347    because they can be forward references. We solve this by remembering this case
1348    and doing the check at the end; a flag specifies which mode we are running in.
1349    
1350  Arguments:  Arguments:
1351    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1352    options  the compiling options    options  the compiling options
1353      atend    TRUE if called when the pattern is complete
1354      cd       the "compile data" structure
1355    
1356  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1357                 or -1 if there is no fixed length,
1358               or -2 if \C was encountered               or -2 if \C was encountered
1359                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1360  */  */
1361    
1362  static int  static int
1363  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1364  {  {
1365  int length = -1;  int length = -1;
1366    
# Line 1042  branch, check the length against that of Line 1373  branch, check the length against that of
1373  for (;;)  for (;;)
1374    {    {
1375    int d;    int d;
1376      uschar *ce, *cs;
1377    register int op = *cc;    register int op = *cc;
   
1378    switch (op)    switch (op)
1379      {      {
1380      case OP_CBRA:      case OP_CBRA:
1381      case OP_BRA:      case OP_BRA:
1382      case OP_ONCE:      case OP_ONCE:
1383      case OP_COND:      case OP_COND:
1384      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1385      if (d < 0) return d;      if (d < 0) return d;
1386      branchlength += d;      branchlength += d;
1387      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1072  for (;;) Line 1403  for (;;)
1403      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1404      branchlength = 0;      branchlength = 0;
1405      break;      break;
1406    
1407        /* A true recursion implies not fixed length, but a subroutine call may
1408        be OK. If the subroutine is a forward reference, we can't deal with
1409        it until the end of the pattern, so return -3. */
1410    
1411        case OP_RECURSE:
1412        if (!atend) return -3;
1413        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1414        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1415        if (cc > cs && cc < ce) return -1;                /* Recursion */
1416        d = find_fixedlength(cs + 2, options, atend, cd);
1417        if (d < 0) return d;
1418        branchlength += d;
1419        cc += 1 + LINK_SIZE;
1420        break;
1421    
1422      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1423    
# Line 1109  for (;;) Line 1455  for (;;)
1455      branchlength++;      branchlength++;
1456      cc += 2;      cc += 2;
1457  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1458      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1459        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1460  #endif  #endif
1461      break;      break;
1462    
# Line 1123  for (;;) Line 1467  for (;;)
1467      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1468      cc += 4;      cc += 4;
1469  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1470      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1471        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1472  #endif  #endif
1473      break;      break;
1474    
1475      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1476      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1477        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1478      cc += 4;      cc += 4;
1479      break;      break;
1480    
# Line 1149  for (;;) Line 1492  for (;;)
1492      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1493      case OP_WORDCHAR:      case OP_WORDCHAR:
1494      case OP_ANY:      case OP_ANY:
1495        case OP_ALLANY:
1496      branchlength++;      branchlength++;
1497      cc++;      cc++;
1498      break;      break;
# Line 1203  for (;;) Line 1547  for (;;)
1547    
1548    
1549  /*************************************************  /*************************************************
1550  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1551  *************************************************/  *************************************************/
1552    
1553  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1554  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1555    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1556    so that it can be called from pcre_study() when finding the minimum matching
1557    length.
1558    
1559  Arguments:  Arguments:
1560    code        points to start of expression    code        points to start of expression
1561    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1562    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1563    
1564  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
1565  */  */
1566    
1567  static const uschar *  const uschar *
1568  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1569  {  {
1570  for (;;)  for (;;)
1571    {    {
# Line 1230  for (;;) Line 1577  for (;;)
1577    the table is zero; the actual length is stored in the compiled code. */    the table is zero; the actual length is stored in the compiled code. */
1578    
1579    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1580    
1581      /* Handle recursion */
1582    
1583      else if (c == OP_REVERSE)
1584        {
1585        if (number < 0) return (uschar *)code;
1586        code += _pcre_OP_lengths[c];
1587        }
1588    
1589    /* Handle capturing bracket */    /* Handle capturing bracket */
1590    
# Line 1240  for (;;) Line 1595  for (;;)
1595      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1596      }      }
1597    
1598    /* 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
1599    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
1600    arrange to skip the extra bytes. */    two bytes of parameters. */
1601    
1602    else    else
1603      {      {
1604        switch(c)
1605          {
1606          case OP_TYPESTAR:
1607          case OP_TYPEMINSTAR:
1608          case OP_TYPEPLUS:
1609          case OP_TYPEMINPLUS:
1610          case OP_TYPEQUERY:
1611          case OP_TYPEMINQUERY:
1612          case OP_TYPEPOSSTAR:
1613          case OP_TYPEPOSPLUS:
1614          case OP_TYPEPOSQUERY:
1615          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1616          break;
1617    
1618          case OP_TYPEUPTO:
1619          case OP_TYPEMINUPTO:
1620          case OP_TYPEEXACT:
1621          case OP_TYPEPOSUPTO:
1622          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1623          break;
1624          }
1625    
1626        /* Add in the fixed length from the table */
1627    
1628      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1629    
1630      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1631      a multi-byte character. The length in the table is a minimum, so we have to
1632      arrange to skip the extra bytes. */
1633    
1634  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1635      if (utf8) switch(c)      if (utf8) switch(c)
1636        {        {
# Line 1268  for (;;) Line 1652  for (;;)
1652        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1653        break;        break;
1654        }        }
1655    #else
1656        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1657  #endif  #endif
1658      }      }
1659    }    }
# Line 1304  for (;;) Line 1690  for (;;)
1690    
1691    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1692    
1693    /* 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
1694    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
1695    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1696    
1697    else    else
1698      {      {
1699        switch(c)
1700          {
1701          case OP_TYPESTAR:
1702          case OP_TYPEMINSTAR:
1703          case OP_TYPEPLUS:
1704          case OP_TYPEMINPLUS:
1705          case OP_TYPEQUERY:
1706          case OP_TYPEMINQUERY:
1707          case OP_TYPEPOSSTAR:
1708          case OP_TYPEPOSPLUS:
1709          case OP_TYPEPOSQUERY:
1710          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1711          break;
1712    
1713          case OP_TYPEPOSUPTO:
1714          case OP_TYPEUPTO:
1715          case OP_TYPEMINUPTO:
1716          case OP_TYPEEXACT:
1717          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1718          break;
1719          }
1720    
1721        /* Add in the fixed length from the table */
1722    
1723      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1724    
1725        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1726        by a multi-byte character. The length in the table is a minimum, so we have
1727        to arrange to skip the extra bytes. */
1728    
1729  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1730      if (utf8) switch(c)      if (utf8) switch(c)
1731        {        {
# Line 1333  for (;;) Line 1747  for (;;)
1747        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1748        break;        break;
1749        }        }
1750    #else
1751        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1752  #endif  #endif
1753      }      }
1754    }    }
# Line 1348  for (;;) Line 1764  for (;;)
1764  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()
1765  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
1766  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1767  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
1768  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
1769    bracket whose current branch will already have been scanned.
1770    
1771  Arguments:  Arguments:
1772    code        points to start of search    code        points to start of search
# Line 1371  for (code = first_significant_code(code Line 1788  for (code = first_significant_code(code
1788    
1789    c = *code;    c = *code;
1790    
1791    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1792      first_significant_code() with a TRUE final argument. */
1793    
1794      if (c == OP_ASSERT)
1795        {
1796        do code += GET(code, 1); while (*code == OP_ALT);
1797        c = *code;
1798        continue;
1799        }
1800    
1801      /* Groups with zero repeats can of course be empty; skip them. */
1802    
1803      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1804        {
1805        code += _pcre_OP_lengths[c];
1806        do code += GET(code, 1); while (*code == OP_ALT);
1807        c = *code;
1808        continue;
1809        }
1810    
1811      /* For other groups, scan the branches. */
1812    
1813      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1814      {      {
1815      BOOL empty_branch;      BOOL empty_branch;
1816      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1817    
1818      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1819        empty branch, so just skip over the conditional, because it could be empty.
1820        Otherwise, scan the individual branches of the group. */
1821    
1822      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;  
1823        code += GET(code, 1);        code += GET(code, 1);
1824        else
1825          {
1826          empty_branch = FALSE;
1827          do
1828            {
1829            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1830              empty_branch = TRUE;
1831            code += GET(code, 1);
1832            }
1833          while (*code == OP_ALT);
1834          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1835        }        }
     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. */  
1836    
1837      c = OP_END;      c = *code;
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
1838      continue;      continue;
1839      }      }
1840    
# Line 1400  for (code = first_significant_code(code Line 1842  for (code = first_significant_code(code
1842    
1843    switch (c)    switch (c)
1844      {      {
1845      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1846        cannot be represented just by a bit map. This includes negated single
1847        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1848        actual length is stored in the compiled code, so we must update "code"
1849        here. */
1850    
1851  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1852      case OP_XCLASS:      case OP_XCLASS:
1853      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1854      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1855  #endif  #endif
1856    
# Line 1448  for (code = first_significant_code(code Line 1894  for (code = first_significant_code(code
1894      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1895      case OP_WORDCHAR:      case OP_WORDCHAR:
1896      case OP_ANY:      case OP_ANY:
1897        case OP_ALLANY:
1898      case OP_ANYBYTE:      case OP_ANYBYTE:
1899      case OP_CHAR:      case OP_CHAR:
1900      case OP_CHARNC:      case OP_CHARNC:
# Line 1466  for (code = first_significant_code(code Line 1913  for (code = first_significant_code(code
1913      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1914      return FALSE;      return FALSE;
1915    
1916        /* These are going to continue, as they may be empty, but we have to
1917        fudge the length for the \p and \P cases. */
1918    
1919        case OP_TYPESTAR:
1920        case OP_TYPEMINSTAR:
1921        case OP_TYPEPOSSTAR:
1922        case OP_TYPEQUERY:
1923        case OP_TYPEMINQUERY:
1924        case OP_TYPEPOSQUERY:
1925        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1926        break;
1927    
1928        /* Same for these */
1929    
1930        case OP_TYPEUPTO:
1931        case OP_TYPEMINUPTO:
1932        case OP_TYPEPOSUPTO:
1933        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1934        break;
1935    
1936      /* End of branch */      /* End of branch */
1937    
1938      case OP_KET:      case OP_KET:
# Line 1484  for (code = first_significant_code(code Line 1951  for (code = first_significant_code(code
1951      case OP_QUERY:      case OP_QUERY:
1952      case OP_MINQUERY:      case OP_MINQUERY:
1953      case OP_POSQUERY:      case OP_POSQUERY:
1954        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1955        break;
1956    
1957      case OP_UPTO:      case OP_UPTO:
1958      case OP_MINUPTO:      case OP_MINUPTO:
1959      case OP_POSUPTO:      case OP_POSUPTO:
1960      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1961      break;      break;
1962  #endif  #endif
1963      }      }
# Line 1535  return TRUE; Line 2005  return TRUE;
2005  *************************************************/  *************************************************/
2006    
2007  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2008  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
2009  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2010  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2011    
2012    Originally, this function only recognized a sequence of letters between the
2013    terminators, but it seems that Perl recognizes any sequence of characters,
2014    though of course unknown POSIX names are subsequently rejected. Perl gives an
2015    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2016    didn't consider this to be a POSIX class. Likewise for [:1234:].
2017    
2018    The problem in trying to be exactly like Perl is in the handling of escapes. We
2019    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2020    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2021    below handles the special case of \], but does not try to do any other escape
2022    processing. This makes it different from Perl for cases such as [:l\ower:]
2023    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2024    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2025    I think.
2026    
2027  Argument:  Arguments:
2028    ptr      pointer to the initial [    ptr      pointer to the initial [
2029    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2030    
2031  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2032  */  */
2033    
2034  static BOOL  static BOOL
2035  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2036  {  {
2037  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2038  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2039  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2040    {    {
2041    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2042    return TRUE;      {
2043        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2044        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2045          {
2046          *endptr = ptr;
2047          return TRUE;
2048          }
2049        }
2050    }    }
2051  return FALSE;  return FALSE;
2052  }  }
# Line 1582  Returns:     a value representing the na Line 2071  Returns:     a value representing the na
2071  static int  static int
2072  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2073  {  {
2074    const char *pn = posix_names;
2075  register int yield = 0;  register int yield = 0;
2076  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2077    {    {
2078    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2079      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2080      pn += posix_name_lengths[yield] + 1;
2081    yield++;    yield++;
2082    }    }
2083  return -1;  return -1;
# Line 1601  return -1; Line 2092  return -1;
2092  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2093  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2094  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
2095  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
2096  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
2097  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
2098  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
2099  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2100    OP_END.
2101    
2102  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2103  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 1628  adjust_recurse(uschar *group, int adjust Line 2120  adjust_recurse(uschar *group, int adjust
2120    uschar *save_hwm)    uschar *save_hwm)
2121  {  {
2122  uschar *ptr = group;  uschar *ptr = group;
2123    
2124  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2125    {    {
2126    int offset;    int offset;
# Line 1739  get_othercase_range(unsigned int *cptr, Line 2232  get_othercase_range(unsigned int *cptr,
2232  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2233    
2234  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2235    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2236    
2237  if (c > d) return FALSE;  if (c > d) return FALSE;
2238    
# Line 1748  next = othercase + 1; Line 2241  next = othercase + 1;
2241    
2242  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2243    {    {
2244    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2245    next++;    next++;
2246    }    }
2247    
# Line 1794  if ((options & PCRE_EXTENDED) != 0) Line 2287  if ((options & PCRE_EXTENDED) != 0)
2287    for (;;)    for (;;)
2288      {      {
2289      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2290      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2291        {        {
2292        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2293          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1806  if ((options & PCRE_EXTENDED) != 0) Line 2299  if ((options & PCRE_EXTENDED) != 0)
2299  /* 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
2300  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2301    
2302  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2303    {    {
2304    int temperrorcode = 0;    int temperrorcode = 0;
2305    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1831  if ((options & PCRE_EXTENDED) != 0) Line 2324  if ((options & PCRE_EXTENDED) != 0)
2324    for (;;)    for (;;)
2325      {      {
2326      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2327      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2328        {        {
2329        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2330          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1842  if ((options & PCRE_EXTENDED) != 0) Line 2335  if ((options & PCRE_EXTENDED) != 0)
2335    
2336  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2337    
2338  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2339    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2340        return FALSE;
2341    
2342  /* 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
2343  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 1858  if (next >= 0) switch(op_code) Line 2352  if (next >= 0) switch(op_code)
2352    case OP_CHAR:    case OP_CHAR:
2353  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2354    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2355    #else
2356      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2357  #endif  #endif
2358    return item != next;    return item != next;
2359    
# Line 1876  if (next >= 0) switch(op_code) Line 2372  if (next >= 0) switch(op_code)
2372      unsigned int othercase;      unsigned int othercase;
2373      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2374  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2375      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2376  #else  #else
2377      othercase = NOTACHAR;      othercase = NOTACHAR;
2378  #endif  #endif
# Line 1889  if (next >= 0) switch(op_code) Line 2385  if (next >= 0) switch(op_code)
2385    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2386    
2387    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2388    if (item == next) return TRUE;    if (item == next) return TRUE;
2389    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2390  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1898  if (next >= 0) switch(op_code) Line 2393  if (next >= 0) switch(op_code)
2393      unsigned int othercase;      unsigned int othercase;
2394      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2395  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2396      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2397  #else  #else
2398      othercase = NOTACHAR;      othercase = NOTACHAR;
2399  #endif  #endif
# Line 1926  if (next >= 0) switch(op_code) Line 2421  if (next >= 0) switch(op_code)
2421    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2422    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2423    
2424      case OP_HSPACE:
2425      case OP_NOT_HSPACE:
2426      switch(next)
2427        {
2428        case 0x09:
2429        case 0x20:
2430        case 0xa0:
2431        case 0x1680:
2432        case 0x180e:
2433        case 0x2000:
2434        case 0x2001:
2435        case 0x2002:
2436        case 0x2003:
2437        case 0x2004:
2438        case 0x2005:
2439        case 0x2006:
2440        case 0x2007:
2441        case 0x2008:
2442        case 0x2009:
2443        case 0x200A:
2444        case 0x202f:
2445        case 0x205f:
2446        case 0x3000:
2447        return op_code != OP_HSPACE;
2448        default:
2449        return op_code == OP_HSPACE;
2450        }
2451    
2452      case OP_VSPACE:
2453      case OP_NOT_VSPACE:
2454      switch(next)
2455        {
2456        case 0x0a:
2457        case 0x0b:
2458        case 0x0c:
2459        case 0x0d:
2460        case 0x85:
2461        case 0x2028:
2462        case 0x2029:
2463        return op_code != OP_VSPACE;
2464        default:
2465        return op_code == OP_VSPACE;
2466        }
2467    
2468    default:    default:
2469    return FALSE;    return FALSE;
2470    }    }
# Line 1960  switch(op_code) Line 2499  switch(op_code)
2499      case ESC_W:      case ESC_W:
2500      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2501    
2502        case ESC_h:
2503        case ESC_H:
2504        switch(item)
2505          {
2506          case 0x09:
2507          case 0x20:
2508          case 0xa0:
2509          case 0x1680:
2510          case 0x180e:
2511          case 0x2000:
2512          case 0x2001:
2513          case 0x2002:
2514          case 0x2003:
2515          case 0x2004:
2516          case 0x2005:
2517          case 0x2006:
2518          case 0x2007:
2519          case 0x2008:
2520          case 0x2009:
2521          case 0x200A:
2522          case 0x202f:
2523          case 0x205f:
2524          case 0x3000:
2525          return -next != ESC_h;
2526          default:
2527          return -next == ESC_h;
2528          }
2529    
2530        case ESC_v:
2531        case ESC_V:
2532        switch(item)
2533          {
2534          case 0x0a:
2535          case 0x0b:
2536          case 0x0c:
2537          case 0x0d:
2538          case 0x85:
2539          case 0x2028:
2540          case 0x2029:
2541          return -next != ESC_v;
2542          default:
2543          return -next == ESC_v;
2544          }
2545    
2546      default:      default:
2547      return FALSE;      return FALSE;
2548      }      }
2549    
2550    case OP_DIGIT:    case OP_DIGIT:
2551    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2552             next == -ESC_h || next == -ESC_v;
2553    
2554    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2555    return next == -ESC_d;    return next == -ESC_d;
# Line 1974  switch(op_code) Line 2558  switch(op_code)
2558    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2559    
2560    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2561    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2562    
2563      case OP_HSPACE:
2564      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2565    
2566      case OP_NOT_HSPACE:
2567      return next == -ESC_h;
2568    
2569      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2570      case OP_VSPACE:
2571      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2572    
2573      case OP_NOT_VSPACE:
2574      return next == -ESC_v;
2575    
2576    case OP_WORDCHAR:    case OP_WORDCHAR:
2577    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2578    
2579    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2580    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2050  uschar classbits[32]; Line 2647  uschar classbits[32];
2647  BOOL class_utf8;  BOOL class_utf8;
2648  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2649  uschar *class_utf8data;  uschar *class_utf8data;
2650    uschar *class_utf8data_base;
2651  uschar utf8_char[6];  uschar utf8_char[6];
2652  #else  #else
2653  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2089  req_caseopt = ((options & PCRE_CASELESS) Line 2687  req_caseopt = ((options & PCRE_CASELESS)
2687  for (;; ptr++)  for (;; ptr++)
2688    {    {
2689    BOOL negate_class;    BOOL negate_class;
2690      BOOL should_flip_negation;
2691    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2692    BOOL is_quantifier;    BOOL is_quantifier;
2693    BOOL is_recurse;    BOOL is_recurse;
2694      BOOL reset_bracount;
2695    int class_charcount;    int class_charcount;
2696    int class_lastchar;    int class_lastchar;
2697    int newoptions;    int newoptions;
2698    int recno;    int recno;
2699    int refsign;    int refsign;
2700    int skipbytes;    int skipbytes;
2701    int subreqbyte;    int subreqbyte;
2702    int subfirstbyte;    int subfirstbyte;
# Line 2129  for (;; ptr++) Line 2729  for (;; ptr++)
2729      */      */
2730    
2731      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2732    
2733        /* Paranoid check for integer overflow */
2734    
2735        if (OFLOW_MAX - *lengthptr < code - last_code)
2736          {
2737          *errorcodeptr = ERR20;
2738          goto FAILED;
2739          }
2740    
2741      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2742      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));
2743    
# Line 2166  for (;; ptr++) Line 2775  for (;; ptr++)
2775    
2776    if (inescq && c != 0)    if (inescq && c != 0)
2777      {      {
2778      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2779        {        {
2780        inescq = FALSE;        inescq = FALSE;
2781        ptr++;        ptr++;
# Line 2192  for (;; ptr++) Line 2801  for (;; ptr++)
2801    /* 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
2802    a quantifier. */    a quantifier. */
2803    
2804    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2805      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2806        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2807    
2808    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2809         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2208  for (;; ptr++) Line 2818  for (;; ptr++)
2818    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2819      {      {
2820      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2821      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2822        {        {
2823        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2824          {          {
# Line 2233  for (;; ptr++) Line 2843  for (;; ptr++)
2843      {      {
2844      /* ===================================================================*/      /* ===================================================================*/
2845      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2846      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2847      case ')':      case CHAR_RIGHT_PARENTHESIS:
2848      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2849      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2850      *codeptr = code;      *codeptr = code;
2851      *ptrptr = ptr;      *ptrptr = ptr;
2852      if (lengthptr != NULL)      if (lengthptr != NULL)
2853        {        {
2854          if (OFLOW_MAX - *lengthptr < code - last_code)
2855            {
2856            *errorcodeptr = ERR20;
2857            goto FAILED;
2858            }
2859        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2860        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2861        }        }
# Line 2251  for (;; ptr++) Line 2866  for (;; ptr++)
2866      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2867      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2868    
2869      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2870      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2871        {        {
2872        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2260  for (;; ptr++) Line 2875  for (;; ptr++)
2875      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2876      break;      break;
2877    
2878      case '$':      case CHAR_DOLLAR_SIGN:
2879      previous = NULL;      previous = NULL;
2880      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2881      break;      break;
# Line 2268  for (;; ptr++) Line 2883  for (;; ptr++)
2883      /* 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
2884      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2885    
2886      case '.':      case CHAR_DOT:
2887      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2888      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2889      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2890      previous = code;      previous = code;
2891      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2892      break;      break;
2893    
2894    
# Line 2288  for (;; ptr++) Line 2903  for (;; ptr++)
2903      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,
2904      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2905      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.
     */  
2906    
2907      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2908        default (Perl) mode, it is treated as a data character. */
2909    
2910        case CHAR_RIGHT_SQUARE_BRACKET:
2911        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2912          {
2913          *errorcodeptr = ERR64;
2914          goto FAILED;
2915          }
2916        goto NORMAL_CHAR;
2917    
2918        case CHAR_LEFT_SQUARE_BRACKET:
2919      previous = code;      previous = code;
2920    
2921      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if      /* PCRE supports POSIX class stuff inside a class. Perl gives an error if
2922      they are encountered at the top level, so we'll do that too. */      they are encountered at the top level, so we'll do that too. */
2923    
2924      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2925          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2926            check_posix_syntax(ptr, &tempptr))
2927        {        {
2928        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2929        goto FAILED;        goto FAILED;
2930        }        }
2931    
2932      /* 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,
2933        if the first few characters (either before or after ^) are \Q\E or \E we
2934        skip them too. This makes for compatibility with Perl. */
2935    
2936      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2937        for (;;)
2938        {        {
       negate_class = TRUE;  
2939        c = *(++ptr);        c = *(++ptr);
2940          if (c == CHAR_BACKSLASH)
2941            {
2942            if (ptr[1] == CHAR_E)
2943              ptr++;
2944            else if (strncmp((const char *)ptr+1,
2945                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2946              ptr += 3;
2947            else
2948              break;
2949            }
2950          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2951            negate_class = TRUE;
2952          else break;
2953        }        }
2954      else  
2955        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2956        an initial ']' is taken as a data character -- the code below handles
2957        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2958        [^] must match any character, so generate OP_ALLANY. */
2959    
2960        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2961            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2962        {        {
2963        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2964          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2965          zerofirstbyte = firstbyte;
2966          break;
2967        }        }
2968    
2969        /* If a class contains a negative special such as \S, we need to flip the
2970        negation flag at the end, so that support for characters > 255 works
2971        correctly (they are all included in the class). */
2972    
2973        should_flip_negation = FALSE;
2974    
2975      /* 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
2976      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
2977      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2332  for (;; ptr++) Line 2989  for (;; ptr++)
2989  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2990      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2991      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2992        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2993  #endif  #endif
2994    
2995      /* 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 2347  for (;; ptr++) Line 3005  for (;; ptr++)
3005          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3006          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3007          }          }
3008    
3009          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3010          data and reset the pointer. This is so that very large classes that
3011          contain a zillion UTF-8 characters no longer overwrite the work space
3012          (which is on the stack). */
3013    
3014          if (lengthptr != NULL)
3015            {
3016            *lengthptr += class_utf8data - class_utf8data_base;
3017            class_utf8data = class_utf8data_base;
3018            }
3019    
3020  #endif  #endif
3021    
3022        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3023    
3024        if (inescq)        if (inescq)
3025          {          {
3026          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3027            {            {
3028            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3029            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2368  for (;; ptr++) Line 3038  for (;; ptr++)
3038        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3039        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3040    
3041        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3042            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3043            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3044          {          {
3045          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3046          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3047          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3048          uschar pbits[32];          uschar pbits[32];
3049    
3050          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3051            {            {
3052            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3053            goto FAILED;            goto FAILED;
3054            }            }
3055    
3056          ptr += 2;          ptr += 2;
3057          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3058            {            {
3059            local_negate = TRUE;            local_negate = TRUE;
3060              should_flip_negation = TRUE;  /* Note negative special */
3061            ptr++;            ptr++;
3062            }            }
3063    
# Line 2453  for (;; ptr++) Line 3124  for (;; ptr++)
3124        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
3125        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
3126        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3127        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
3128        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3129    
3130        if (c == '\\')        if (c == CHAR_BACKSLASH)
3131          {          {
3132          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3133          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3134    
3135          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 */
3136          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 */
3137          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 */
3138          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3139            {            {
3140            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3141              {              {
3142              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3143              }              }
3144            else inescq = TRUE;            else inescq = TRUE;
3145            continue;            continue;
3146            }            }
3147            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3148    
3149          if (c < 0)          if (c < 0)
3150            {            {
# Line 2488  for (;; ptr++) Line 3160  for (;; ptr++)
3160              continue;              continue;
3161    
3162              case ESC_D:              case ESC_D:
3163                should_flip_negation = TRUE;
3164              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3165              continue;              continue;
3166    
# Line 2496  for (;; ptr++) Line 3169  for (;; ptr++)
3169              continue;              continue;
3170    
3171              case ESC_W:              case ESC_W:
3172                should_flip_negation = TRUE;
3173              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3174              continue;              continue;
3175    
# Line 2505  for (;; ptr++) Line 3179  for (;; ptr++)
3179              continue;              continue;
3180    
3181              case ESC_S:              case ESC_S:
3182                should_flip_negation = TRUE;
3183              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3184              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3185              continue;              continue;
3186    
3187              case ESC_E: /* Perl ignores an orphan \E */              default:    /* Not recognized; fall through */
3188                break;      /* Need "default" setting to stop compiler warning. */
3189                }
3190    
3191              /* In the pre-compile phase, just do the recognition. */
3192    
3193              else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3194                       c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3195    
3196              /* We need to deal with \H, \h, \V, and \v in both phases because
3197              they use extra memory. */
3198    
3199              if (-c == ESC_h)
3200                {
3201                SETBIT(classbits, 0x09); /* VT */
3202                SETBIT(classbits, 0x20); /* SPACE */
3203                SETBIT(classbits, 0xa0); /* NSBP */
3204    #ifdef SUPPORT_UTF8
3205                if (utf8)
3206                  {
3207                  class_utf8 = TRUE;
3208                  *class_utf8data++ = XCL_SINGLE;
3209                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3210                  *class_utf8data++ = XCL_SINGLE;
3211                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3212                  *class_utf8data++ = XCL_RANGE;
3213                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3214                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3215                  *class_utf8data++ = XCL_SINGLE;
3216                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3217                  *class_utf8data++ = XCL_SINGLE;
3218                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3219                  *class_utf8data++ = XCL_SINGLE;
3220                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3221                  }
3222    #endif
3223                continue;
3224                }
3225    
3226              if (-c == ESC_H)
3227                {
3228                for (c = 0; c < 32; c++)
3229                  {
3230                  int x = 0xff;
3231                  switch (c)
3232                    {
3233                    case 0x09/8: x ^= 1 << (0x09%8); break;
3234                    case 0x20/8: x ^= 1 << (0x20%8); break;
3235                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3236                    default: break;
3237                    }
3238                  classbits[c] |= x;
3239                  }
3240    
3241    #ifdef SUPPORT_UTF8
3242                if (utf8)
3243                  {
3244                  class_utf8 = TRUE;
3245                  *class_utf8data++ = XCL_RANGE;
3246                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3247                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3248                  *class_utf8data++ = XCL_RANGE;
3249                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3250                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3251                  *class_utf8data++ = XCL_RANGE;
3252                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3253                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3254                  *class_utf8data++ = XCL_RANGE;
3255                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3256                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3257                  *class_utf8data++ = XCL_RANGE;
3258                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3259                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3260                  *class_utf8data++ = XCL_RANGE;
3261                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3262                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3263                  *class_utf8data++ = XCL_RANGE;
3264                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3265                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3266                  }
3267    #endif
3268                continue;
3269                }
3270    
3271              if (-c == ESC_v)
3272                {
3273                SETBIT(classbits, 0x0a); /* LF */
3274                SETBIT(classbits, 0x0b); /* VT */
3275                SETBIT(classbits, 0x0c); /* FF */
3276                SETBIT(classbits, 0x0d); /* CR */
3277                SETBIT(classbits, 0x85); /* NEL */
3278    #ifdef SUPPORT_UTF8
3279                if (utf8)
3280                  {
3281                  class_utf8 = TRUE;
3282                  *class_utf8data++ = XCL_RANGE;
3283                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3284                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3285                  }
3286    #endif
3287              continue;              continue;
   
             default:    /* Not recognized; fall through */  
             break;      /* Need "default" setting to stop compiler warning. */  
3288              }              }
3289    
3290            /* In the pre-compile phase, just do the recognition. */            if (-c == ESC_V)
3291                {
3292                for (c = 0; c < 32; c++)
3293                  {
3294                  int x = 0xff;
3295                  switch (c)
3296                    {
3297                    case 0x0a/8: x ^= 1 << (0x0a%8);
3298                                 x ^= 1 << (0x0b%8);
3299                                 x ^= 1 << (0x0c%8);
3300                                 x ^= 1 << (0x0d%8);
3301                                 break;
3302                    case 0x85/8: x ^= 1 << (0x85%8); break;
3303                    default: break;
3304                    }
3305                  classbits[c] |= x;
3306                  }
3307    
3308            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  #ifdef SUPPORT_UTF8
3309                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;              if (utf8)
3310                  {
3311                  class_utf8 = TRUE;
3312                  *class_utf8data++ = XCL_RANGE;
3313                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3314                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3315                  *class_utf8data++ = XCL_RANGE;
3316                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3317                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3318                  }
3319    #endif
3320                continue;
3321                }
3322    
3323            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3324    
# Line 2564  for (;; ptr++) Line 3363  for (;; ptr++)
3363        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3364    
3365        CHECK_RANGE:        CHECK_RANGE:
3366        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3367          {          {
3368          inescq = FALSE;          inescq = FALSE;
3369          ptr += 2;          ptr += 2;
# Line 2572  for (;; ptr++) Line 3371  for (;; ptr++)
3371    
3372        oldptr = ptr;        oldptr = ptr;
3373    
3374        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3375    
3376          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3377    
3378          /* Check for range */
3379    
3380          if (!inescq && ptr[1] == CHAR_MINUS)
3381          {          {
3382          int d;          int d;
3383          ptr += 2;          ptr += 2;
3384          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3385    
3386          /* 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
3387          mode. */          mode. */
3388    
3389          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3390            {            {
3391            ptr += 2;            ptr += 2;
3392            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3393                { ptr += 2; continue; }
3394            inescq = TRUE;            inescq = TRUE;
3395            break;            break;
3396            }            }
3397    
3398          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3399            {            {
3400            ptr = oldptr;            ptr = oldptr;
3401            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2608  for (;; ptr++) Line 3414  for (;; ptr++)
3414          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
3415          in such circumstances. */          in such circumstances. */
3416    
3417          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3418            {            {
3419            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3420            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3421    
3422            /* \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
3423            special means the '-' was literal */            special means the '-' was literal */
3424    
3425            if (d < 0)            if (d < 0)
3426              {              {
3427              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3428              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3429              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3430                {                {
3431                ptr = oldptr;                ptr = oldptr;
3432                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2639  for (;; ptr++) Line 3445  for (;; ptr++)
3445    
3446          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3447    
3448            /* Remember \r or \n */
3449    
3450            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3451    
3452          /* 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
3453          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3454          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2661  for (;; ptr++) Line 3471  for (;; ptr++)
3471              unsigned int origd = d;              unsigned int origd = d;
3472              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3473                {                {
3474                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3475                      ocd <= (unsigned int)d)
3476                    continue;                          /* Skip embedded ranges */
3477    
3478                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3479                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3480                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3481                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3482                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3483                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3484                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3485                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3486                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3487                  d = ocd;                  d = ocd;
3488                  continue;                  continue;
# Line 2753  for (;; ptr++) Line 3567  for (;; ptr++)
3567          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3568            {            {
3569            unsigned int othercase;            unsigned int othercase;
3570            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3571              {              {
3572              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3573              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2780  for (;; ptr++) Line 3594  for (;; ptr++)
3594    
3595      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3596    
3597      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3598    
3599      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3600        {        {
# Line 2788  for (;; ptr++) Line 3602  for (;; ptr++)
3602        goto FAILED;        goto FAILED;
3603        }        }
3604    
3605    
3606    /* This code has been disabled because it would mean that \s counts as
3607    an explicit \r or \n reference, and that's not really what is wanted. Now
3608    we set the flag only if there is a literal "\r" or "\n" in the class. */
3609    
3610    #if 0
3611        /* Remember whether \r or \n are in this class */
3612    
3613        if (negate_class)
3614          {
3615          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3616          }
3617        else
3618          {
3619          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3620          }
3621    #endif
3622    
3623    
3624      /* 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
3625      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
3626      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
3627      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3628      single-bytes only. This is an historical hangover. Maybe one day we can  
3629      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3630        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3631        operate on single-bytes only. This is an historical hangover. Maybe one day
3632        we can tidy these opcodes to handle multi-byte characters.
3633    
3634      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
3635      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 2803  for (;; ptr++) Line 3639  for (;; ptr++)
3639      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3640    
3641  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3642      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3643            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3644  #else  #else
3645      if (class_charcount == 1)      if (class_charcount == 1)
3646  #endif  #endif
# Line 2849  for (;; ptr++) Line 3683  for (;; ptr++)
3683      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3684    
3685      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3686      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3687      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
3688        the class, so any that were explicitly given as well can be ignored. If
3689        (when there are explicit characters > 255 that must be listed) there are no
3690        characters < 256, we can omit the bitmap in the actual compiled code. */
3691    
3692  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3693      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3694        {        {
3695        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3696        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2879  for (;; ptr++) Line 3716  for (;; ptr++)
3716        }        }
3717  #endif  #endif
3718    
3719      /* 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
3720      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
3721      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
3722      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3723    
3724        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3725      if (negate_class)      if (negate_class)
3726        {        {
       *code++ = OP_NCLASS;  
3727        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3728          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3729        }        }
3730      else      else
3731        {        {
       *code++ = OP_CLASS;  
3732        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3733        }        }
3734      code += 32;      code += 32;
# Line 2903  for (;; ptr++) Line 3739  for (;; ptr++)
3739      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3740      has been tested above. */      has been tested above. */
3741    
3742      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3743      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3744      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3745      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3746      goto REPEAT;      goto REPEAT;
3747    
3748      case '*':      case CHAR_ASTERISK:
3749      repeat_min = 0;      repeat_min = 0;
3750      repeat_max = -1;      repeat_max = -1;
3751      goto REPEAT;      goto REPEAT;
3752    
3753      case '+':      case CHAR_PLUS:
3754      repeat_min = 1;      repeat_min = 1;
3755      repeat_max = -1;      repeat_max = -1;
3756      goto REPEAT;      goto REPEAT;
3757    
3758      case '?':      case CHAR_QUESTION_MARK:
3759      repeat_min = 0;      repeat_min = 0;
3760      repeat_max = 1;      repeat_max = 1;
3761    
# Line 2954  for (;; ptr++) Line 3790  for (;; ptr++)
3790      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
3791      repeat type to the non-default. */      repeat type to the non-default. */
3792    
3793      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3794        {        {
3795        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3796        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3797        ptr++;        ptr++;
3798        }        }
3799      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3800        {        {
3801        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3802        ptr++;        ptr++;
# Line 3075  for (;; ptr++) Line 3911  for (;; ptr++)
3911    
3912        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3913    
3914          /*--------------------------------------------------------------------*/
3915          /* This code is obsolete from release 8.00; the restriction was finally
3916          removed: */
3917    
3918        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3919        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3920    
3921        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3922          /*--------------------------------------------------------------------*/
3923    
3924        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3925    
# Line 3225  for (;; ptr++) Line 4066  for (;; ptr++)
4066          goto END_REPEAT;          goto END_REPEAT;
4067          }          }
4068    
4069          /*--------------------------------------------------------------------*/
4070          /* This code is obsolete from release 8.00; the restriction was finally
4071          removed: */
4072    
4073        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4074        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4075    
4076        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4077          /*--------------------------------------------------------------------*/
4078    
4079        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4080          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3264  for (;; ptr++) Line 4110  for (;; ptr++)
4110          goto FAILED;          goto FAILED;
4111          }          }
4112    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4113        /* 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
4114        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
4115        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 3294  for (;; ptr++) Line 4132  for (;; ptr++)
4132    
4133        if (repeat_min == 0)        if (repeat_min == 0)
4134          {          {
4135          /* 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
4136          altogether. */          output altogether, like this:
4137    
4138          if (repeat_max == 0)          ** if (repeat_max == 0)
4139            {          **   {
4140            code = previous;          **   code = previous;
4141            goto END_REPEAT;          **   goto END_REPEAT;
4142            }          **   }
4143    
4144            However, that fails when a group is referenced as a subroutine from
4145            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4146            so that it is skipped on execution. As we don't have a list of which
4147            groups are referenced, we cannot do this selectively.
4148    
4149            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4150            and do no more at this point. However, we do need to adjust any
4151            OP_RECURSE calls inside the group that refer to the group itself or any
4152            internal or forward referenced group, because the offset is from the
4153            start of the whole regex. Temporarily terminate the pattern while doing
4154            this. */
4155    
4156          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4157            {            {
4158            *code = OP_END;            *code = OP_END;
4159            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4160            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4161            code++;            code++;
4162              if (repeat_max == 0)
4163                {
4164                *previous++ = OP_SKIPZERO;
4165                goto END_REPEAT;
4166                }
4167            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4168            }            }
4169    
# Line 3360  for (;; ptr++) Line 4208  for (;; ptr++)
4208          if (repeat_min > 1)          if (repeat_min > 1)
4209            {            {
4210            /* 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
4211            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4212              potential integer overflow. */
4213    
4214            if (lengthptr != NULL)            if (lengthptr != NULL)
4215              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4216                int delta = (repeat_min - 1)*length_prevgroup;
4217                if ((double)(repeat_min - 1)*(double)length_prevgroup >
4218                                                                (double)INT_MAX ||
4219                    OFLOW_MAX - *lengthptr < delta)
4220                  {
4221                  *errorcodeptr = ERR20;
4222                  goto FAILED;
4223                  }
4224                *lengthptr += delta;
4225                }
4226    
4227            /* This is compiling for real */            /* This is compiling for real */
4228    
# Line 3401  for (;; ptr++) Line 4260  for (;; ptr++)
4260          /* 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
4261          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
4262          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
4263          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4264            paranoid checks to avoid integer overflow. */
4265    
4266          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4267            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4268              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4269                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4270              if ((double)repeat_max *
4271                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4272                      > (double)INT_MAX ||
4273                  OFLOW_MAX - *lengthptr < delta)
4274                {
4275                *errorcodeptr = ERR20;
4276                goto FAILED;
4277                }
4278              *lengthptr += delta;
4279              }
4280    
4281          /* This is compiling for real */          /* This is compiling for real */
4282    
# Line 3487  for (;; ptr++) Line 4358  for (;; ptr++)
4358          }          }
4359        }        }
4360    
4361        /* If previous is OP_FAIL, it was generated by an empty class [] in
4362        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4363        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4364        error above. We can just ignore the repeat in JS case. */
4365    
4366        else if (*previous == OP_FAIL) goto END_REPEAT;
4367    
4368      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4369    
4370      else      else
# Line 3511  for (;; ptr++) Line 4389  for (;; ptr++)
4389      if (possessive_quantifier)      if (possessive_quantifier)
4390        {        {
4391        int len;        int len;
4392        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4393            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4394            tempcode += _pcre_OP_lengths[*tempcode] +
4395              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4396    
4397          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4398            {
4399          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4400    #ifdef SUPPORT_UTF8
4401            if (utf8 && tempcode[-1] >= 0xc0)
4402              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4403    #endif
4404            }
4405    
4406        len = code - tempcode;        len = code - tempcode;
4407        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4408          {          {
# Line 3557  for (;; ptr++) Line 4446  for (;; ptr++)
4446      /* ===================================================================*/      /* ===================================================================*/
4447      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4448      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4449      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. */  
4450    
4451      case '(':      case CHAR_LEFT_PARENTHESIS:
4452      newoptions = options;      newoptions = options;
4453      skipbytes = 0;      skipbytes = 0;
4454      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4455      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4456        reset_bracount = FALSE;
4457    
4458        /* First deal with various "verbs" that can be introduced by '*'. */
4459    
4460        if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4461          {
4462          int i, namelen;
4463          const char *vn = verbnames;
4464          const uschar *name = ++ptr;
4465          previous = NULL;
4466          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4467          if (*ptr == CHAR_COLON)
4468            {
4469            *errorcodeptr = ERR59;   /* Not supported */
4470            goto FAILED;
4471            }
4472          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4473            {
4474            *errorcodeptr = ERR60;
4475            goto FAILED;
4476            }
4477          namelen = ptr - name;
4478          for (i = 0; i < verbcount; i++)
4479            {
4480            if (namelen == verbs[i].len &&
4481                strncmp((char *)name, vn, namelen) == 0)
4482              {
4483              /* Check for open captures before ACCEPT */
4484    
4485              if (verbs[i].op == OP_ACCEPT)
4486                {
4487                open_capitem *oc;
4488                cd->had_accept = TRUE;
4489                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4490                  {
4491                  *code++ = OP_CLOSE;
4492                  PUT2INC(code, 0, oc->number);
4493                  }
4494                }
4495              *code++ = verbs[i].op;
4496              break;
4497              }
4498            vn += verbs[i].len + 1;
4499            }
4500          if (i < verbcount) continue;
4501          *errorcodeptr = ERR60;
4502          goto FAILED;
4503          }
4504    
4505      if (*(++ptr) == '?')      /* Deal with the extended parentheses; all are introduced by '?', and the
4506        appearance of any of them means that this is not a capturing group. */
4507    
4508        else if (*ptr == CHAR_QUESTION_MARK)
4509        {        {
4510        int i, set, unset, namelen;        int i, set, unset, namelen;
4511        int *optset;        int *optset;
# Line 3576  for (;; ptr++) Line 4514  for (;; ptr++)
4514    
4515        switch (*(++ptr))        switch (*(++ptr))
4516          {          {
4517          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4518          ptr++;          ptr++;
4519          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4520          if (*ptr == 0)          if (*ptr == 0)
4521            {            {
4522            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3588  for (;; ptr++) Line 4526  for (;; ptr++)
4526    
4527    
4528          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4529          case ':':                 /* Non-capturing bracket */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4530            reset_bracount = TRUE;
4531            /* Fall through */
4532    
4533            /* ------------------------------------------------------------ */
4534            case CHAR_COLON:          /* Non-capturing bracket */
4535          bravalue = OP_BRA;          bravalue = OP_BRA;
4536          ptr++;          ptr++;
4537          break;          break;
4538    
4539    
4540          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4541          case '(':          case CHAR_LEFT_PARENTHESIS:
4542          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4543    
4544          /* 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 3615  for (;; ptr++) Line 4558  for (;; ptr++)
4558          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4559          including assertions, are processed. */          including assertions, are processed. */
4560    
4561          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4562                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4563            break;            break;
4564    
4565          /* 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 3623  for (;; ptr++) Line 4567  for (;; ptr++)
4567    
4568          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
4569          skipbytes = 3;          skipbytes = 3;
4570          refsign = -1;          refsign = -1;
4571    
4572          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4573    
4574          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4575            {            {
4576            terminator = -1;            terminator = -1;
4577            ptr += 2;            ptr += 2;
# Line 3637  for (;; ptr++) Line 4581  for (;; ptr++)
4581          /* 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
4582          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4583    
4584          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4585            {            {
4586            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4587            ptr++;            ptr++;
4588            }            }
4589          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4590            {            {
4591            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4592            ptr++;            ptr++;
4593            }            }
4594          else          else
4595            {            {
4596            terminator = 0;            terminator = 0;
4597            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4598            }            }
4599    
4600          /* 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 */
4601    
# Line 3670  for (;; ptr++) Line 4614  for (;; ptr++)
4614            {            {
4615            if (recno >= 0)            if (recno >= 0)
4616              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4617                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4618            ptr++;            ptr++;
4619            }            }
4620          namelen = ptr - name;          namelen = ptr - name;
4621    
4622          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4623                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4624            {            {
4625            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4626            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3689  for (;; ptr++) Line 4634  for (;; ptr++)
4634          /* 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
4635          reference. If the string started with "+" or "-" we require the rest to          reference. If the string started with "+" or "-" we require the rest to
4636          be digits, in which case recno will be set. */          be digits, in which case recno will be set. */
4637    
4638          if (refsign > 0)          if (refsign > 0)
4639            {            {
4640            if (recno <= 0)            if (recno <= 0)
4641              {              {
4642              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4643              goto FAILED;              goto FAILED;
4644              }              }
4645            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4646                cd->bracount - recno + 1 : recno +cd->bracount;
4647              if (recno <= 0 || recno > cd->final_bracount)
4648              {              {
4649              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4650              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4651              }              }
           else recno += cd->bracount;  
4652            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4653            break;            break;
4654            }            }
4655    
4656          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4657          name. */          name. */
4658    
4659          slot = cd->name_table;          slot = cd->name_table;
4660          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
4661            {            {
# Line 3731  for (;; ptr++) Line 4673  for (;; ptr++)
4673    
4674          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4675    
4676          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4677                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4678            {            {
4679            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
# Line 3752  for (;; ptr++) Line 4694  for (;; ptr++)
4694          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4695          specific group number. */          specific group number. */
4696    
4697          else if (*name == 'R')          else if (*name == CHAR_R)
4698            {            {
4699            recno = 0;            recno = 0;
4700            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 3762  for (;; ptr++) Line 4704  for (;; ptr++)
4704                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4705                goto FAILED;                goto FAILED;
4706                }                }
4707              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4708              }              }
4709            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4710            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 3772  for (;; ptr++) Line 4714  for (;; ptr++)
4714          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4715          false. */          false. */
4716    
4717          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4718            {            {
4719            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4720            skipbytes = 1;            skipbytes = 1;
4721            }            }
4722    
4723          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4724            in the second pass here, so final_bracount is set. */
4725    
4726          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4727            {            {
4728            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4729            }            }
# Line 3796  for (;; ptr++) Line 4739  for (;; ptr++)
4739    
4740    
4741          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4742          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4743          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4744          ptr++;          ptr++;
4745          break;          break;
4746    
4747    
4748          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4749          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4750          ptr++;          ptr++;
4751            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4752              {
4753              *code++ = OP_FAIL;
4754              previous = NULL;
4755              continue;
4756              }
4757            bravalue = OP_ASSERT_NOT;
4758          break;          break;
4759    
4760    
4761          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4762          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4763          switch (ptr[1])          switch (ptr[1])
4764            {            {
4765            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4766            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4767            ptr += 2;            ptr += 2;
4768            break;            break;
4769    
4770            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4771            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4772            ptr += 2;            ptr += 2;
4773            break;            break;
# Line 3833  for (;; ptr++) Line 4782  for (;; ptr++)
4782    
4783    
4784          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4785          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4786          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4787          ptr++;          ptr++;
4788          break;          break;
4789    
4790    
4791          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4792          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4793          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4794          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4795          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4796            {            {
4797            int n = 0;            int n = 0;
4798            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4799              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4800            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4801              {              {
4802              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4803              goto FAILED;              goto FAILED;
# Line 3868  for (;; ptr++) Line 4817  for (;; ptr++)
4817    
4818    
4819          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4820          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4821          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4822                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4823            {            {
4824            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4825            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4826            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4827            }            }
4828          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4829            {            {
4830            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4831            goto FAILED;            goto FAILED;
# Line 3885  for (;; ptr++) Line 4835  for (;; ptr++)
4835    
4836          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4837          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4838          case '\'':          case CHAR_APOSTROPHE:
4839            {            {
4840            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4841                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4842            name = ++ptr;            name = ++ptr;
4843    
4844            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 3918  for (;; ptr++) Line 4869  for (;; ptr++)
4869                }                }
4870              }              }
4871    
4872            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4873              alphabetical order. Duplicate names for different numbers are
4874              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4875              number are always OK. (An existing number can be re-used if (?|
4876              appears in the pattern.) In either event, a duplicate name results in
4877              a duplicate entry in the table, even if the number is the same. This
4878              is because the number of names, and hence the table size, is computed
4879              in the pre-compile, and it affects various numbers and pointers which
4880              would all have to be modified, and the compiled code moved down, if
4881              duplicates with the same number were omitted from the table. This
4882              doesn't seem worth the hassle. However, *different* names for the
4883              same number are not permitted. */
4884    
4885            else            else
4886              {              {
4887                BOOL dupname = FALSE;
4888              slot = cd->name_table;              slot = cd->name_table;
4889    
4890              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4891                {                {
4892                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 3930  for (;; ptr++) Line 4894  for (;; ptr++)
4894                  {                  {
4895                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4896                    {                    {
4897                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4898                          (options & PCRE_DUPNAMES) == 0)
4899                      {                      {
4900                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4901                      goto FAILED;                      goto FAILED;
4902                      }                      }
4903                      else dupname = TRUE;
4904                    }                    }
4905                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4906                  }                  }
4907    
4908                  /* Make space in the table and break the loop for an earlier
4909                  name. For a duplicate or later name, carry on. We do this for
4910                  duplicates so that in the simple case (when ?(| is not used) they
4911                  are in order of their numbers. */
4912    
4913                if (crc < 0)                if (crc < 0)
4914                  {                  {
4915                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4916                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4917                  break;                  break;
4918                  }                  }
4919    
4920                  /* Continue the loop for a later or duplicate name */
4921    
4922                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4923                }                }
4924    
4925                /* For non-duplicate names, check for a duplicate number before
4926                adding the new name. */
4927    
4928                if (!dupname)
4929                  {
4930                  uschar *cslot = cd->name_table;
4931                  for (i = 0; i < cd->names_found; i++)
4932                    {
4933                    if (cslot != slot)
4934                      {
4935                      if (GET2(cslot, 0) == cd->bracount + 1)
4936                        {
4937                        *errorcodeptr = ERR65;
4938                        goto FAILED;
4939                        }
4940                      }
4941                    else i--;
4942                    cslot += cd->name_entry_size;
4943                    }
4944                  }
4945    
4946              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4947              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
# Line 3953  for (;; ptr++) Line 4949  for (;; ptr++)
4949              }              }
4950            }            }
4951    
4952          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
4953            encountered. */
4954    
         ptr++;                    /* Move past > or ' */  
4955          cd->names_found++;          cd->names_found++;
4956            ptr++;                    /* Move past > or ' */
4957          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
4958    
4959    
4960          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4961          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4962          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4963          is_recurse = TRUE;          is_recurse = TRUE;
4964          /* Fall through */          /* Fall through */
4965    
4966          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4967          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4968          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4969            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4970            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4971    
4972          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4973          name = ++ptr;          name = ++ptr;
# Line 3980  for (;; ptr++) Line 4979  for (;; ptr++)
4979    
4980          if (lengthptr != NULL)          if (lengthptr != NULL)
4981            {            {
4982              if (namelen == 0)
4983                {
4984                *errorcodeptr = ERR62;
4985                goto FAILED;
4986                }
4987            if (*ptr != terminator)            if (*ptr != terminator)
4988              {              {
4989              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 3993  for (;; ptr++) Line 4997  for (;; ptr++)
4997            recno = 0;            recno = 0;
4998            }            }
4999    
5000          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
5001            first, and then check that we have reached the end of the name in the
5002            table. That way, if the name that is longer than any in the table,
5003            the comparison will fail without reading beyond the table entry. */
5004    
5005          else          else
5006            {            {
5007            slot = cd->name_table;            slot = cd->name_table;
5008            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5009              {              {
5010              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
5011                    slot[2+namelen] == 0)
5012                  break;
5013              slot += cd->name_entry_size;              slot += cd->name_entry_size;
5014              }              }
5015    
# Line 4009  for (;; ptr++) Line 5018  for (;; ptr++)
5018              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5019              }              }
5020            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5021                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5022                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5023              {              {
5024              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4025  for (;; ptr++) Line 5034  for (;; ptr++)
5034    
5035    
5036          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5037          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5038          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5039          /* Fall through */          /* Fall through */
5040    
5041    
5042          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5043          case '-': case '+':          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5044          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
5045          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
5046            {            {
5047            const uschar *called;            const uschar *called;
5048              terminator = CHAR_RIGHT_PARENTHESIS;
5049    
5050              /* Come here from the \g<...> and \g'...' code (Oniguruma
5051              compatibility). However, the syntax has been checked to ensure that
5052              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5053              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5054              ever be taken. */
5055    
5056              HANDLE_NUMERICAL_RECURSION:
5057    
5058            if ((refsign = *ptr) == '+') ptr++;            if ((refsign = *ptr) == CHAR_PLUS)
5059            else if (refsign == '-')              {
5060                ptr++;
5061                if ((digitab[*ptr] & ctype_digit) == 0)
5062                  {
5063                  *errorcodeptr = ERR63;
5064                  goto FAILED;
5065                  }
5066                }
5067              else if (refsign == CHAR_MINUS)
5068              {              {
5069              if ((digitab[ptr[1]] & ctype_digit) == 0)              if ((digitab[ptr[1]] & ctype_digit) == 0)
5070                goto OTHER_CHAR_AFTER_QUERY;                goto OTHER_CHAR_AFTER_QUERY;
5071              ptr++;              ptr++;
5072              }              }
5073    
5074            recno = 0;            recno = 0;
5075            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5076              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5077    
5078            if (*ptr != ')')            if (*ptr != terminator)
5079              {              {
5080              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5081              goto FAILED;              goto FAILED;
5082              }              }
5083    
5084            if (refsign == '-')            if (refsign == CHAR_MINUS)
5085              {              {
5086              if (recno == 0)              if (recno == 0)
5087                {                {
5088                *errorcodeptr = ERR58;                *errorcodeptr = ERR58;
5089                goto FAILED;                goto FAILED;
5090                }                }
5091              recno = cd->bracount - recno + 1;              recno = cd->bracount - recno + 1;
5092              if (recno <= 0)              if (recno <= 0)
5093                {                {
5094                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5095                goto FAILED;                goto FAILED;
5096                }                }
5097              }              }
5098            else if (refsign == '+')            else if (refsign == CHAR_PLUS)
5099              {              {
5100              if (recno == 0)              if (recno == 0)
5101                {                {
5102                *errorcodeptr = ERR58;                *errorcodeptr = ERR58;
5103                goto FAILED;                goto FAILED;
5104                }                }
5105              recno += cd->bracount;              recno += cd->bracount;
5106              }              }
5107    
5108            /* Come here from code above that handles a named recursion */            /* Come here from code above that handles a named recursion */
5109    
# Line 4096  for (;; ptr++) Line 5122  for (;; ptr++)
5122            if (lengthptr == NULL)            if (lengthptr == NULL)
5123              {              {
5124              *code = OP_END;              *code = OP_END;
5125              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5126                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5127    
5128              /* Forward reference */              /* Forward reference */
5129    
5130              if (called == NULL)              if (called == NULL)
5131                {                {
5132                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5133                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5134                  {                  {
5135                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5136                  goto FAILED;                  goto FAILED;
# Line 4151  for (;; ptr++) Line 5178  for (;; ptr++)
5178    
5179          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5180          default:              /* Other characters: check option setting */          default:              /* Other characters: check option setting */
5181          OTHER_CHAR_AFTER_QUERY:          OTHER_CHAR_AFTER_QUERY:
5182          set = unset = 0;          set = unset = 0;
5183          optset = &set;          optset = &set;
5184    
5185          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5186            {            {
5187            switch (*ptr++)            switch (*ptr++)
5188              {              {
5189              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5190    
5191              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5192              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5193              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5194              break;              break;
5195    
5196              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5197              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5198              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5199              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5200              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5201              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5202    
5203              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5204                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4202  for (;; ptr++) Line 5229  for (;; ptr++)
5229          both phases.          both phases.
5230    
5231          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
5232          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
5233          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. */  
5234    
5235          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5236            {            {
5237            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5238                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5239              {              {
5240              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5241              }              }
5242           else           else
5243              {              {
# Line 4222  for (;; ptr++) Line 5246  for (;; ptr++)
5246                *code++ = OP_OPT;                *code++ = OP_OPT;
5247                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5248                }                }
   
             /* 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;  
5249              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5250              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5251              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5252              }              }
5253    
5254              /* Change options at this level, and pass them back for use
5255              in subsequent branches. When not at the start of the pattern, this
5256              information is also necessary so that a resetting item can be
5257              compiled at the end of a group (if we are in a group). */
5258    
5259              *optionsptr = options = newoptions;
5260            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5261            continue;              /* It is complete */            continue;              /* It is complete */
5262            }            }
# Line 4286  for (;; ptr++) Line 5310  for (;; ptr++)
5310           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
5311           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5312            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5313             reset_bracount,               /* True if (?| group */
5314           skipbytes,                    /* Skip over bracket number */           skipbytes,                    /* Skip over bracket number */
5315           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
5316           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
# Line 4302  for (;; ptr++) Line 5327  for (;; ptr++)
5327      is on the bracket. */      is on the bracket. */
5328    
5329      /* 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
5330      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
5331        in the real compile phase, not in the pre-pass, where the whole group may
5332        not be available. */
5333    
5334      if (bravalue == OP_COND)      if (bravalue == OP_COND && lengthptr == NULL)
5335        {        {
5336        uschar *tc = code;        uschar *tc = code;
5337        int condcount = 0;        int condcount = 0;
# Line 4345  for (;; ptr++) Line 5372  for (;; ptr++)
5372    
5373      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5374    
5375      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5376        {        {
5377        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5378        goto FAILED;        goto FAILED;
5379        }        }
5380    
5381      /* 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,
5382      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
5383      just the brackets so that it doesn't use much memory if it is duplicated by      set of non-capturing brackets so that it doesn't use much memory if it is
5384      a quantifier. */      duplicated by a quantifier.*/
5385    
5386      if (lengthptr != NULL)      if (lengthptr != NULL)
5387        {        {
5388          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
5389            {
5390            *errorcodeptr = ERR20;
5391            goto FAILED;
5392            }
5393        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5394        code++;        *code++ = OP_BRA;
5395        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5396        *code++ = OP_KET;        *code++ = OP_KET;
5397        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5398          break;    /* No need to waste time with special character handling */
5399        }        }
5400    
5401      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
5402    
5403      else code = tempcode;      code = tempcode;
5404    
5405      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
5406      relevant. */      relevant. */
# Line 4437  for (;; ptr++) Line 5470  for (;; ptr++)
5470      We can test for values between ESC_b and ESC_Z for the latter; this may      We can test for values between ESC_b and ESC_Z for the latter; this may
5471      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5472    
5473      case '\\':      case CHAR_BACKSLASH:
5474      tempptr = ptr;      tempptr = ptr;
5475      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5476      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 4446  for (;; ptr++) Line 5479  for (;; ptr++)
5479        {        {
5480        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5481          {          {
5482          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5483            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5484                else inescq = TRUE;
5485          continue;          continue;
5486          }          }
5487    
# Line 4464  for (;; ptr++) Line 5498  for (;; ptr++)
5498        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5499        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5500    
5501        /* \k<name> or \k'name' is a back reference by name (Perl syntax) */        /* \g<name> or \g'name' is a subroutine call by name and \g<n> or \g'n'
5502          is a subroutine call by number (Oniguruma syntax). In fact, the value
5503          -ESC_g is returned only for these cases. So we don't need to check for <
5504          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5505          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5506          that is a synonym for a named back reference). */
5507    
5508          if (-c == ESC_g)
5509            {
5510            const uschar *p;
5511            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5512            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5513              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5514    
5515            /* These two statements stop the compiler for warning about possibly
5516            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5517            fact, because we actually check for a number below, the paths that
5518            would actually be in error are never taken. */
5519    
5520            skipbytes = 0;
5521            reset_bracount = FALSE;
5522    
5523            /* Test for a name */
5524    
5525            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5526              {
5527              BOOL isnumber = TRUE;
5528              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5529                {
5530                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5531                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5532                }
5533              if (*p != terminator)
5534                {
5535                *errorcodeptr = ERR57;
5536                break;
5537                }
5538              if (isnumber)
5539                {
5540                ptr++;
5541                goto HANDLE_NUMERICAL_RECURSION;
5542                }
5543              is_recurse = TRUE;
5544              goto NAMED_REF_OR_RECURSE;
5545              }
5546    
5547            /* Test a signed number in angle brackets or quotes. */
5548    
5549            p = ptr + 2;
5550            while ((digitab[*p] & ctype_digit) != 0) p++;
5551            if (*p != terminator)
5552              {
5553              *errorcodeptr = ERR57;
5554              break;
5555              }
5556            ptr++;
5557            goto HANDLE_NUMERICAL_RECURSION;
5558            }
5559    
5560          /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5561          We also support \k{name} (.NET syntax) */
5562    
5563        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\''))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5564              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5565          {          {
5566          is_recurse = FALSE;          is_recurse = FALSE;
5567          terminator = (*(++ptr) == '<')? '>' : '\'';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5568              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5569              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5570          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5571          }          }
5572    
# Line 4570  for (;; ptr++) Line 5667  for (;; ptr++)
5667      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5668      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5669    
5670        /* Remember if \r or \n were seen */
5671    
5672        if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5673          cd->external_flags |= PCRE_HASCRORLF;
5674    
5675      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5676      byte, set it from this character, but revert to none on a zero repeat.      byte, set it from this character, but revert to none on a zero repeat.
5677      Otherwise, leave the firstbyte value alone, and don't change it on a zero      Otherwise, leave the firstbyte value alone, and don't change it on a zero
# Line 4635  This function is used during the pre-com Line 5737  This function is used during the pre-com
5737  out the amount of memory needed, as well as during the real compile phase. The  out the amount of memory needed, as well as during the real compile phase. The
5738  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
5739    
5740  Argument:  Arguments:
5741    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
5742    oldims         previous settings of ims option bits    oldims         previous settings of ims option bits
5743    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
5744    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
5745    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
5746    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
5747      reset_bracount TRUE to reset the count for each branch
5748    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
5749    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
5750    reqbyteptr     place to put the last required character, or a negative number    reqbyteptr     place to put the last required character, or a negative number
# Line 4655  Returns:         TRUE on success Line 5758  Returns:         TRUE on success
5758    
5759  static BOOL  static BOOL
5760  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,
5761    int *errorcodeptr, BOOL lookbehind, int skipbytes, int *firstbyteptr,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
5762    int *reqbyteptr, branch_chain *bcptr, compile_data *cd, int *lengthptr)    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
5763      int *lengthptr)
5764  {  {
5765  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
5766  uschar *code = *codeptr;  uschar *code = *codeptr;
5767  uschar *last_branch = code;  uschar *last_branch = code;
5768  uschar *start_bracket = code;  uschar *start_bracket = code;
5769  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5770    open_capitem capitem;
5771    int capnumber = 0;
5772  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5773  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5774  int length;  int length;
5775    int orig_bracount;
5776    int max_bracount;
5777  branch_chain bc;  branch_chain bc;
5778    
5779  bc.outer = bcptr;  bc.outer = bcptr;
# Line 4687  the code that abstracts option settings Line 5795  the code that abstracts option settings
5795  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the  them global. It tests the value of length for (2 + 2*LINK_SIZE) in the
5796  pre-compile phase to find out whether anything has yet been compiled or not. */  pre-compile phase to find out whether anything has yet been compiled or not. */
5797    
5798    /* If this is a capturing subpattern, add to the chain of open capturing items
5799    so that we can detect them if (*ACCEPT) is encountered. */
5800    
5801    if (*code == OP_CBRA)
5802      {
5803      capnumber = GET2(code, 1 + LINK_SIZE);
5804      capitem.number = capnumber;
5805      capitem.next = cd->open_caps;
5806      cd->open_caps = &capitem;
5807      }
5808    
5809  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5810    
5811  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 4694  code += 1 + LINK_SIZE + skipbytes; Line 5813  code += 1 + LINK_SIZE + skipbytes;
5813    
5814  /* Loop for each alternative branch */  /* Loop for each alternative branch */
5815    
5816    orig_bracount = max_bracount = cd->bracount;
5817  for (;;)  for (;;)
5818    {    {
5819      /* For a (?| group, reset the capturing bracket count so that each branch
5820      uses the same numbers. */
5821    
5822      if (reset_bracount) cd->bracount = orig_bracount;
5823    
5824    /* Handle a change of ims options at the start of the branch */    /* Handle a change of ims options at the start of the branch */
5825    
5826    if ((options & PCRE_IMS) != oldims)    if ((options & PCRE_IMS) != oldims)
# Line 4725  for (;;) Line 5850  for (;;)
5850      return FALSE;      return FALSE;
5851      }      }
5852    
5853      /* Keep the highest bracket count in case (?| was used and some branch
5854      has fewer than the rest. */
5855    
5856      if (cd->bracount > max_bracount) max_bracount = cd->bracount;
5857    
5858    /* In the real compile phase, there is some post-processing to be done. */    /* In the real compile phase, there is some post-processing to be done. */
5859    
5860    if (lengthptr == NULL)    if (lengthptr == NULL)
# Line 4770  for (;;) Line 5900  for (;;)
5900    
5901      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5902      put the length into the OP_REVERSE item. Temporarily mark the end of the      put the length into the OP_REVERSE item. Temporarily mark the end of the
5903      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5904        because there may be forward references that we can't check here. Set a
5905        flag to cause another lookbehind check at the end. Why not do it all at the
5906        end? Because common, erroneous checks are picked up here and the offset of
5907        the problem can be shown. */
5908    
5909      if (lookbehind)      if (lookbehind)
5910        {        {
5911        int fixed_length;        int fixed_length;
5912        *code = OP_END;        *code = OP_END;
5913        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
5914        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
5915        if (fixed_length < 0)        if (fixed_length == -3)
5916            {
5917            cd->check_lookbehind = TRUE;
5918            }
5919          else if (fixed_length < 0)
5920          {          {
5921          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
5922          *ptrptr = ptr;          *ptrptr = ptr;
5923          return FALSE;          return FALSE;
5924          }          }
5925        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
5926        }        }
5927      }      }
5928    
5929    /* Reached end of expression, either ')' or end of pattern. Go back through    /* Reached end of expression, either ')' or end of pattern. In the real
5930    the alternative branches and reverse the chain of offsets, with the field in    compile phase, go back through the alternative branches and reverse the chain
5931    the BRA item now becoming an offset to the first alternative. If there are    of offsets, with the field in the BRA item now becoming an offset to the
5932    no alternatives, it points to the end of the group. The length in the    first alternative. If there are no alternatives, it points to the end of the
5933    terminating ket is always the length of the whole bracketed item. If any of    group. The length in the terminating ket is always the length of the whole
5934    the ims options were changed inside the group, compile a resetting op-code    bracketed item. If any of the ims options were changed inside the group,
5935    following, except at the very end of the pattern. Return leaving the pointer    compile a resetting op-code following, except at the very end of the pattern.
5936    at the terminating char. */    Return leaving the pointer at the terminating char. */
5937    
5938    if (*ptr != '|')    if (*ptr != CHAR_VERTICAL_LINE)
5939      {      {
5940      int branch_length = code - last_branch;      if (lengthptr == NULL)
     do  
5941        {        {
5942        int prev_length = GET(last_branch, 1);        int branch_length = code - last_branch;
5943        PUT(last_branch, 1, branch_length);        do
5944        branch_length = prev_length;          {
5945        last_branch -= branch_length;          int prev_length = GET(last_branch, 1);
5946        }          PUT(last_branch, 1, branch_length);
5947      while (branch_length > 0);          branch_length = prev_length;
5948            last_branch -= branch_length;
5949            }
5950          while (branch_length > 0);
5951          }
5952    
5953        /* If it was a capturing subpattern, remove it from the chain. */
5954