/[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 107 by ph10, Wed Mar 7 11:02:28 2007 UTC revision 459 by ph10, Sun Oct 4 09:21:39 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) 1997-2006 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,      0,      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  };    "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 262  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 298  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 373  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 399  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 417  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 442  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 471  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 484  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 504  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 525  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 538  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 549  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 559  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 587  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 603  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 677  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 713  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 756  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 797  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 807  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 837  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[1] == CHAR_BACKSLASH)
1141            {
1142            if (ptr[2] == CHAR_E)
1143              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 == '\\')        if (*ptr == 0) return -1;
1168          if (*ptr == CHAR_BACKSLASH)
1169          {          {
1170          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1171          if (*ptr == 'Q') for (;;)          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 904  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      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1202        {
1203        if (dup_parens && *count < hwm_count) *count = hwm_count;
1204        *ptrptr = ptr;
1205        return -1;
1206        }
1207    
1208      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    ptr += 2;  FAIL_EXIT:
1216    if (*ptr == 'P') ptr++;                      /* Allow optional P */  *ptrptr = ptr;
1217    return -1;
1218    }
1219    
   /* We have to disambiguate (?<! and (?<= from (?<name> */  
1220    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1221    
   count++;  
1222    
1223    if (name == NULL && count == lorn) return count;  /*************************************************
1224    term = *ptr++;  *       Find forward referenced subpattern       *
1225    if (term == '<') term = '>';  *************************************************/
1226    thisname = ptr;  
1227    while (*ptr != term) ptr++;  /* This function scans along a pattern's text looking for capturing
1228    if (name != NULL && lorn == ptr - thisname &&  subpatterns, and counting them. If it finds a named pattern that matches the
1229        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  name it is given, it returns its number. Alternatively, if the name is NULL, it
1230      return count;  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    Returns:       the number of the found subpattern, or -1 if not found
1245    */
1246    
1247    static int
1248    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1249    {
1250    uschar *ptr = (uschar *)cd->start_pattern;
1251    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 996  for (;;) Line 1317  for (;;)
1317    
1318      case OP_CALLOUT:      case OP_CALLOUT:
1319      case OP_CREF:      case OP_CREF:
1320        case OP_NCREF:
1321      case OP_RREF:      case OP_RREF:
1322        case OP_NRREF:
1323      case OP_DEF:      case OP_DEF:
1324      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1325      break;      break;
# Line 1012  for (;;) Line 1335  for (;;)
1335    
1336    
1337  /*************************************************  /*************************************************
1338  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1339  *************************************************/  *************************************************/
1340    
1341  /* 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,
1342  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.
1343  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
1344    temporarily terminated with OP_END when this function is called.
1345    
1346    This function is called when a backward assertion is encountered, so that if it
1347    fails, the error message can point to the correct place in the pattern.
1348    However, we cannot do this when the assertion contains subroutine calls,
1349    because they can be forward references. We solve this by remembering this case
1350    and doing the check at the end; a flag specifies which mode we are running in.
1351    
1352  Arguments:  Arguments:
1353    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1354    options  the compiling options    options  the compiling options
1355      atend    TRUE if called when the pattern is complete
1356      cd       the "compile data" structure
1357    
1358  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1359                 or -1 if there is no fixed length,
1360               or -2 if \C was encountered               or -2 if \C was encountered
1361                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1362  */  */
1363    
1364  static int  static int
1365  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1366  {  {
1367  int length = -1;  int length = -1;
1368    
# Line 1041  branch, check the length against that of Line 1375  branch, check the length against that of
1375  for (;;)  for (;;)
1376    {    {
1377    int d;    int d;
1378      uschar *ce, *cs;
1379    register int op = *cc;    register int op = *cc;
   
1380    switch (op)    switch (op)
1381      {      {
1382      case OP_CBRA:      case OP_CBRA:
1383      case OP_BRA:      case OP_BRA:
1384      case OP_ONCE:      case OP_ONCE:
1385      case OP_COND:      case OP_COND:
1386      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1387      if (d < 0) return d;      if (d < 0) return d;
1388      branchlength += d;      branchlength += d;
1389      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1071  for (;;) Line 1405  for (;;)
1405      cc += 1 + LINK_SIZE;      cc += 1 + LINK_SIZE;
1406      branchlength = 0;      branchlength = 0;
1407      break;      break;
1408    
1409        /* A true recursion implies not fixed length, but a subroutine call may
1410        be OK. If the subroutine is a forward reference, we can't deal with
1411        it until the end of the pattern, so return -3. */
1412    
1413        case OP_RECURSE:
1414        if (!atend) return -3;
1415        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1416        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1417        if (cc > cs && cc < ce) return -1;                /* Recursion */
1418        d = find_fixedlength(cs + 2, options, atend, cd);
1419        if (d < 0) return d;
1420        branchlength += d;
1421        cc += 1 + LINK_SIZE;
1422        break;
1423    
1424      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1425    
# Line 1085  for (;;) Line 1434  for (;;)
1434    
1435      case OP_REVERSE:      case OP_REVERSE:
1436      case OP_CREF:      case OP_CREF:
1437        case OP_NCREF:
1438      case OP_RREF:      case OP_RREF:
1439        case OP_NRREF:
1440      case OP_DEF:      case OP_DEF:
1441      case OP_OPT:      case OP_OPT:
1442      case OP_CALLOUT:      case OP_CALLOUT:
# Line 1108  for (;;) Line 1459  for (;;)
1459      branchlength++;      branchlength++;
1460      cc += 2;      cc += 2;
1461  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1462      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1463        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1464  #endif  #endif
1465      break;      break;
1466    
# Line 1122  for (;;) Line 1471  for (;;)
1471      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1472      cc += 4;      cc += 4;
1473  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1474      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1475        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1476  #endif  #endif
1477      break;      break;
1478    
1479      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1480      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1481        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1482      cc += 4;      cc += 4;
1483      break;      break;
1484    
# Line 1148  for (;;) Line 1496  for (;;)
1496      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1497      case OP_WORDCHAR:      case OP_WORDCHAR:
1498      case OP_ANY:      case OP_ANY:
1499        case OP_ALLANY:
1500      branchlength++;      branchlength++;
1501      cc++;      cc++;
1502      break;      break;
# Line 1202  for (;;) Line 1551  for (;;)
1551    
1552    
1553  /*************************************************  /*************************************************
1554  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1555  *************************************************/  *************************************************/
1556    
1557  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1558  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1559    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1560    so that it can be called from pcre_study() when finding the minimum matching
1561    length.
1562    
1563  Arguments:  Arguments:
1564    code        points to start of expression    code        points to start of expression
1565    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1566    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1567    
1568  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
1569  */  */
1570    
1571  static const uschar *  const uschar *
1572  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1573  {  {
1574  for (;;)  for (;;)
1575    {    {
# Line 1229  for (;;) Line 1581  for (;;)
1581    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. */
1582    
1583    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1584    
1585      /* Handle recursion */
1586    
1587      else if (c == OP_REVERSE)
1588        {
1589        if (number < 0) return (uschar *)code;
1590        code += _pcre_OP_lengths[c];
1591        }
1592    
1593    /* Handle capturing bracket */    /* Handle capturing bracket */
1594    
# Line 1239  for (;;) Line 1599  for (;;)
1599      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1600      }      }
1601    
1602    /* 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
1603    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
1604    arrange to skip the extra bytes. */    two bytes of parameters. */
1605    
1606    else    else
1607      {      {
1608        switch(c)
1609          {
1610          case OP_TYPESTAR:
1611          case OP_TYPEMINSTAR:
1612          case OP_TYPEPLUS:
1613          case OP_TYPEMINPLUS:
1614          case OP_TYPEQUERY:
1615          case OP_TYPEMINQUERY:
1616          case OP_TYPEPOSSTAR:
1617          case OP_TYPEPOSPLUS:
1618          case OP_TYPEPOSQUERY:
1619          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1620          break;
1621    
1622          case OP_TYPEUPTO:
1623          case OP_TYPEMINUPTO:
1624          case OP_TYPEEXACT:
1625          case OP_TYPEPOSUPTO:
1626          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1627          break;
1628          }
1629    
1630        /* Add in the fixed length from the table */
1631    
1632      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1633    
1634      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1635      a multi-byte character. The length in the table is a minimum, so we have to
1636      arrange to skip the extra bytes. */
1637    
1638  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1639      if (utf8) switch(c)      if (utf8) switch(c)
1640        {        {
# Line 1267  for (;;) Line 1656  for (;;)
1656        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1657        break;        break;
1658        }        }
1659  #endif  #else
1660        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1661    #endif
1662      }      }
1663    }    }
1664  }  }
# Line 1303  for (;;) Line 1694  for (;;)
1694    
1695    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1696    
1697    /* 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
1698    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
1699    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1700    
1701    else    else
1702      {      {
1703        switch(c)
1704          {
1705          case OP_TYPESTAR:
1706          case OP_TYPEMINSTAR:
1707          case OP_TYPEPLUS:
1708          case OP_TYPEMINPLUS:
1709          case OP_TYPEQUERY:
1710          case OP_TYPEMINQUERY:
1711          case OP_TYPEPOSSTAR:
1712          case OP_TYPEPOSPLUS:
1713          case OP_TYPEPOSQUERY:
1714          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1715          break;
1716    
1717          case OP_TYPEPOSUPTO:
1718          case OP_TYPEUPTO:
1719          case OP_TYPEMINUPTO:
1720          case OP_TYPEEXACT:
1721          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1722          break;
1723          }
1724    
1725        /* Add in the fixed length from the table */
1726    
1727      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1728    
1729        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1730        by a multi-byte character. The length in the table is a minimum, so we have
1731        to arrange to skip the extra bytes. */
1732    
1733  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1734      if (utf8) switch(c)      if (utf8) switch(c)
1735        {        {
# Line 1332  for (;;) Line 1751  for (;;)
1751        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1752        break;        break;
1753        }        }
1754  #endif  #else
1755        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1756    #endif
1757      }      }
1758    }    }
1759  }  }
# Line 1347  for (;;) Line 1768  for (;;)
1768  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()
1769  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
1770  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1771  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
1772  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
1773    bracket whose current branch will already have been scanned.
1774    
1775  Arguments:  Arguments:
1776    code        points to start of search    code        points to start of search
# Line 1370  for (code = first_significant_code(code Line 1792  for (code = first_significant_code(code
1792    
1793    c = *code;    c = *code;
1794    
1795    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1796      first_significant_code() with a TRUE final argument. */
1797    
1798      if (c == OP_ASSERT)
1799        {
1800        do code += GET(code, 1); while (*code == OP_ALT);
1801        c = *code;
1802        continue;
1803        }
1804    
1805      /* Groups with zero repeats can of course be empty; skip them. */
1806    
1807      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1808        {
1809        code += _pcre_OP_lengths[c];
1810        do code += GET(code, 1); while (*code == OP_ALT);
1811        c = *code;
1812        continue;
1813        }
1814    
1815      /* For other groups, scan the branches. */
1816    
1817      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1818      {      {
1819      BOOL empty_branch;      BOOL empty_branch;
1820      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1821    
1822      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1823        empty branch, so just skip over the conditional, because it could be empty.
1824        Otherwise, scan the individual branches of the group. */
1825    
1826      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;  
1827        code += GET(code, 1);        code += GET(code, 1);
1828        else
1829          {
1830          empty_branch = FALSE;
1831          do
1832            {
1833            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1834              empty_branch = TRUE;
1835            code += GET(code, 1);
1836            }
1837          while (*code == OP_ALT);
1838          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1839        }        }
     while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1840    
1841      /* Move past the KET and fudge things so that the increment in the "for"      c = *code;
     above has no effect. */  
   
     c = OP_END;  
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
1842      continue;      continue;
1843      }      }
1844    
# Line 1399  for (code = first_significant_code(code Line 1846  for (code = first_significant_code(code
1846    
1847    switch (c)    switch (c)
1848      {      {
1849      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1850        cannot be represented just by a bit map. This includes negated single
1851        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1852        actual length is stored in the compiled code, so we must update "code"
1853        here. */
1854    
1855  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1856      case OP_XCLASS:      case OP_XCLASS:
1857      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1858      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1859  #endif  #endif
1860    
# Line 1447  for (code = first_significant_code(code Line 1898  for (code = first_significant_code(code
1898      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1899      case OP_WORDCHAR:      case OP_WORDCHAR:
1900      case OP_ANY:      case OP_ANY:
1901        case OP_ALLANY:
1902      case OP_ANYBYTE:      case OP_ANYBYTE:
1903      case OP_CHAR:      case OP_CHAR:
1904      case OP_CHARNC:      case OP_CHARNC:
# Line 1465  for (code = first_significant_code(code Line 1917  for (code = first_significant_code(code
1917      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1918      return FALSE;      return FALSE;
1919    
1920        /* These are going to continue, as they may be empty, but we have to
1921        fudge the length for the \p and \P cases. */
1922    
1923        case OP_TYPESTAR:
1924        case OP_TYPEMINSTAR:
1925        case OP_TYPEPOSSTAR:
1926        case OP_TYPEQUERY:
1927        case OP_TYPEMINQUERY:
1928        case OP_TYPEPOSQUERY:
1929        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1930        break;
1931    
1932        /* Same for these */
1933    
1934        case OP_TYPEUPTO:
1935        case OP_TYPEMINUPTO:
1936        case OP_TYPEPOSUPTO:
1937        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1938        break;
1939    
1940      /* End of branch */      /* End of branch */
1941    
1942      case OP_KET:      case OP_KET:
# Line 1483  for (code = first_significant_code(code Line 1955  for (code = first_significant_code(code
1955      case OP_QUERY:      case OP_QUERY:
1956      case OP_MINQUERY:      case OP_MINQUERY:
1957      case OP_POSQUERY:      case OP_POSQUERY:
1958        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1959        break;
1960    
1961      case OP_UPTO:      case OP_UPTO:
1962      case OP_MINUPTO:      case OP_MINUPTO:
1963      case OP_POSUPTO:      case OP_POSUPTO:
1964      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1965      break;      break;
1966  #endif  #endif
1967      }      }
# Line 1534  return TRUE; Line 2009  return TRUE;
2009  *************************************************/  *************************************************/
2010    
2011  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2012  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
2013  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2014  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2015    
2016    Originally, this function only recognized a sequence of letters between the
2017    terminators, but it seems that Perl recognizes any sequence of characters,
2018    though of course unknown POSIX names are subsequently rejected. Perl gives an
2019    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2020    didn't consider this to be a POSIX class. Likewise for [:1234:].
2021    
2022    The problem in trying to be exactly like Perl is in the handling of escapes. We
2023    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2024    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2025    below handles the special case of \], but does not try to do any other escape
2026    processing. This makes it different from Perl for cases such as [:l\ower:]
2027    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2028    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2029    I think.
2030    
2031  Argument:  Arguments:
2032    ptr      pointer to the initial [    ptr      pointer to the initial [
2033    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2034    
2035  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2036  */  */
2037    
2038  static BOOL  static BOOL
2039  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2040  {  {
2041  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2042  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2043  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2044    {    {
2045    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2046    return TRUE;      {
2047        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2048        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2049          {
2050          *endptr = ptr;
2051          return TRUE;
2052          }
2053        }
2054    }    }
2055  return FALSE;  return FALSE;
2056  }  }
# Line 1581  Returns:     a value representing the na Line 2075  Returns:     a value representing the na
2075  static int  static int
2076  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2077  {  {
2078    const char *pn = posix_names;
2079  register int yield = 0;  register int yield = 0;
2080  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2081    {    {
2082    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2083      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2084      pn += posix_name_lengths[yield] + 1;
2085    yield++;    yield++;
2086    }    }
2087  return -1;  return -1;
# Line 1600  return -1; Line 2096  return -1;
2096  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2097  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2098  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
2099  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
2100  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
2101  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
2102  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
2103  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2104    OP_END.
2105    
2106  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2107  recursions and subroutine calls. It must also check the list of such references  recursions and subroutine calls. It must also check the list of such references
# Line 1627  adjust_recurse(uschar *group, int adjust Line 2124  adjust_recurse(uschar *group, int adjust
2124    uschar *save_hwm)    uschar *save_hwm)
2125  {  {
2126  uschar *ptr = group;  uschar *ptr = group;
2127    
2128  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2129    {    {
2130    int offset;    int offset;
# Line 1738  get_othercase_range(unsigned int *cptr, Line 2236  get_othercase_range(unsigned int *cptr,
2236  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2237    
2238  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2239    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2240    
2241  if (c > d) return FALSE;  if (c > d) return FALSE;
2242    
# Line 1747  next = othercase + 1; Line 2245  next = othercase + 1;
2245    
2246  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2247    {    {
2248    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2249    next++;    next++;
2250    }    }
2251    
# Line 1793  if ((options & PCRE_EXTENDED) != 0) Line 2291  if ((options & PCRE_EXTENDED) != 0)
2291    for (;;)    for (;;)
2292      {      {
2293      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2294      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2295        {        {
2296        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2297          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1805  if ((options & PCRE_EXTENDED) != 0) Line 2303  if ((options & PCRE_EXTENDED) != 0)
2303  /* 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
2304  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2305    
2306  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2307    {    {
2308    int temperrorcode = 0;    int temperrorcode = 0;
2309    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1830  if ((options & PCRE_EXTENDED) != 0) Line 2328  if ((options & PCRE_EXTENDED) != 0)
2328    for (;;)    for (;;)
2329      {      {
2330      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2331      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2332        {        {
2333        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2334          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1841  if ((options & PCRE_EXTENDED) != 0) Line 2339  if ((options & PCRE_EXTENDED) != 0)
2339    
2340  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2341    
2342  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2343    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2344        return FALSE;
2345    
2346  /* 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
2347  positive single character match, "item" either contains the character or, if  positive single character match, "item" either contains the character or, if
# Line 1857  if (next >= 0) switch(op_code) Line 2356  if (next >= 0) switch(op_code)
2356    case OP_CHAR:    case OP_CHAR:
2357  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2358    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2359    #else
2360      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2361  #endif  #endif
2362    return item != next;    return item != next;
2363    
# Line 1875  if (next >= 0) switch(op_code) Line 2376  if (next >= 0) switch(op_code)
2376      unsigned int othercase;      unsigned int othercase;
2377      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2378  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2379      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2380  #else  #else
2381      othercase = NOTACHAR;      othercase = NOTACHAR;
2382  #endif  #endif
# Line 1888  if (next >= 0) switch(op_code) Line 2389  if (next >= 0) switch(op_code)
2389    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2390    
2391    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2392    if (item == next) return TRUE;    if (item == next) return TRUE;
2393    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2394  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1897  if (next >= 0) switch(op_code) Line 2397  if (next >= 0) switch(op_code)
2397      unsigned int othercase;      unsigned int othercase;
2398      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2399  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2400      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2401  #else  #else
2402      othercase = NOTACHAR;      othercase = NOTACHAR;
2403  #endif  #endif
# Line 1925  if (next >= 0) switch(op_code) Line 2425  if (next >= 0) switch(op_code)
2425    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2426    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2427    
2428      case OP_HSPACE:
2429      case OP_NOT_HSPACE:
2430      switch(next)
2431        {
2432        case 0x09:
2433        case 0x20:
2434        case 0xa0:
2435        case 0x1680:
2436        case 0x180e:
2437        case 0x2000:
2438        case 0x2001:
2439        case 0x2002:
2440        case 0x2003:
2441        case 0x2004:
2442        case 0x2005:
2443        case 0x2006:
2444        case 0x2007:
2445        case 0x2008:
2446        case 0x2009:
2447        case 0x200A:
2448        case 0x202f:
2449        case 0x205f:
2450        case 0x3000:
2451        return op_code != OP_HSPACE;
2452        default:
2453        return op_code == OP_HSPACE;
2454        }
2455    
2456      case OP_VSPACE:
2457      case OP_NOT_VSPACE:
2458      switch(next)
2459        {
2460        case 0x0a:
2461        case 0x0b:
2462        case 0x0c:
2463        case 0x0d:
2464        case 0x85:
2465        case 0x2028:
2466        case 0x2029:
2467        return op_code != OP_VSPACE;
2468        default:
2469        return op_code == OP_VSPACE;
2470        }
2471    
2472    default:    default:
2473    return FALSE;    return FALSE;
2474    }    }
# Line 1959  switch(op_code) Line 2503  switch(op_code)
2503      case ESC_W:      case ESC_W:
2504      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2505    
2506        case ESC_h:
2507        case ESC_H:
2508        switch(item)
2509          {
2510          case 0x09:
2511          case 0x20:
2512          case 0xa0:
2513          case 0x1680:
2514          case 0x180e:
2515          case 0x2000:
2516          case 0x2001:
2517          case 0x2002:
2518          case 0x2003:
2519          case 0x2004:
2520          case 0x2005:
2521          case 0x2006:
2522          case 0x2007:
2523          case 0x2008:
2524          case 0x2009:
2525          case 0x200A:
2526          case 0x202f:
2527          case 0x205f:
2528          case 0x3000:
2529          return -next != ESC_h;
2530          default:
2531          return -next == ESC_h;
2532          }
2533    
2534        case ESC_v:
2535        case ESC_V:
2536        switch(item)
2537          {
2538          case 0x0a:
2539          case 0x0b:
2540          case 0x0c:
2541          case 0x0d:
2542          case 0x85:
2543          case 0x2028:
2544          case 0x2029:
2545          return -next != ESC_v;
2546          default:
2547          return -next == ESC_v;
2548          }
2549    
2550      default:      default:
2551      return FALSE;      return FALSE;
2552      }      }
2553    
2554    case OP_DIGIT:    case OP_DIGIT:
2555    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2556             next == -ESC_h || next == -ESC_v;
2557    
2558    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2559    return next == -ESC_d;    return next == -ESC_d;
# Line 1973  switch(op_code) Line 2562  switch(op_code)
2562    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2563    
2564    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2565    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2566    
2567      case OP_HSPACE:
2568      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2569    
2570      case OP_NOT_HSPACE:
2571      return next == -ESC_h;
2572    
2573      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2574      case OP_VSPACE:
2575      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2576    
2577      case OP_NOT_VSPACE:
2578      return next == -ESC_v;
2579    
2580    case OP_WORDCHAR:    case OP_WORDCHAR:
2581    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2582    
2583    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2584    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2049  uschar classbits[32]; Line 2651  uschar classbits[32];
2651  BOOL class_utf8;  BOOL class_utf8;
2652  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2653  uschar *class_utf8data;  uschar *class_utf8data;
2654    uschar *class_utf8data_base;
2655  uschar utf8_char[6];  uschar utf8_char[6];
2656  #else  #else
2657  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
# Line 2088  req_caseopt = ((options & PCRE_CASELESS) Line 2691  req_caseopt = ((options & PCRE_CASELESS)
2691  for (;; ptr++)  for (;; ptr++)
2692    {    {
2693    BOOL negate_class;    BOOL negate_class;
2694      BOOL should_flip_negation;
2695    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2696    BOOL is_quantifier;    BOOL is_quantifier;
2697    BOOL is_recurse;    BOOL is_recurse;
2698      BOOL reset_bracount;
2699    int class_charcount;    int class_charcount;
2700    int class_lastchar;    int class_lastchar;
2701    int newoptions;    int newoptions;
2702    int recno;    int recno;
2703      int refsign;
2704    int skipbytes;    int skipbytes;
2705    int subreqbyte;    int subreqbyte;
2706    int subfirstbyte;    int subfirstbyte;
# Line 2127  for (;; ptr++) Line 2733  for (;; ptr++)
2733      */      */
2734    
2735      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2736    
2737        /* Paranoid check for integer overflow */
2738    
2739        if (OFLOW_MAX - *lengthptr < code - last_code)
2740          {
2741          *errorcodeptr = ERR20;
2742          goto FAILED;
2743          }
2744    
2745      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2746      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));
2747    
# Line 2164  for (;; ptr++) Line 2779  for (;; ptr++)
2779    
2780    if (inescq && c != 0)    if (inescq && c != 0)
2781      {      {
2782      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2783        {        {
2784        inescq = FALSE;        inescq = FALSE;
2785        ptr++;        ptr++;
# Line 2190  for (;; ptr++) Line 2805  for (;; ptr++)
2805    /* 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
2806    a quantifier. */    a quantifier. */
2807    
2808    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2809      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2810        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2811    
2812    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2813         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2206  for (;; ptr++) Line 2822  for (;; ptr++)
2822    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2823      {      {
2824      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2825      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2826        {        {
2827        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2828          {          {
# Line 2231  for (;; ptr++) Line 2847  for (;; ptr++)
2847      {      {
2848      /* ===================================================================*/      /* ===================================================================*/
2849      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2850      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2851      case ')':      case CHAR_RIGHT_PARENTHESIS:
2852      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2853      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2854      *codeptr = code;      *codeptr = code;
2855      *ptrptr = ptr;      *ptrptr = ptr;
2856      if (lengthptr != NULL)      if (lengthptr != NULL)
2857        {        {
2858          if (OFLOW_MAX - *lengthptr < code - last_code)
2859            {
2860            *errorcodeptr = ERR20;
2861            goto FAILED;
2862            }
2863        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2864        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2865        }        }
# Line 2249  for (;; ptr++) Line 2870  for (;; ptr++)
2870      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2871      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2872    
2873      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2874      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2875        {        {
2876        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2258  for (;; ptr++) Line 2879  for (;; ptr++)
2879      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2880      break;      break;
2881    
2882      case '$':      case CHAR_DOLLAR_SIGN:
2883      previous = NULL;      previous = NULL;
2884      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2885      break;      break;
# Line 2266  for (;; ptr++) Line 2887  for (;; ptr++)
2887      /* 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
2888      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2889    
2890      case '.':      case CHAR_DOT:
2891      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2892      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2893      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2894      previous = code;      previous = code;
2895      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2896      break;      break;
2897    
2898    
# Line 2286  for (;; ptr++) Line 2907  for (;; ptr++)
2907      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,
2908      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2909      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.
     */  
2910    
2911      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2912        default (Perl) mode, it is treated as a data character. */
2913    
2914        case CHAR_RIGHT_SQUARE_BRACKET:
2915        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2916          {
2917          *errorcodeptr = ERR64;
2918          goto FAILED;
2919          }
2920        goto NORMAL_CHAR;
2921    
2922        case CHAR_LEFT_SQUARE_BRACKET:
2923      previous = code;      previous = code;
2924    
2925      /* 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
2926      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. */
2927    
2928      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2929          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2930            check_posix_syntax(ptr, &tempptr))
2931        {        {
2932        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2933        goto FAILED;        goto FAILED;
2934        }        }
2935    
2936      /* 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,
2937        if the first few characters (either before or after ^) are \Q\E or \E we
2938        skip them too. This makes for compatibility with Perl. */
2939    
2940      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2941        for (;;)
2942        {        {
       negate_class = TRUE;  
2943        c = *(++ptr);        c = *(++ptr);
2944          if (c == CHAR_BACKSLASH)
2945            {
2946            if (ptr[1] == CHAR_E)
2947              ptr++;
2948            else if (strncmp((const char *)ptr+1,
2949                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2950              ptr += 3;
2951            else
2952              break;
2953            }
2954          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2955            negate_class = TRUE;
2956          else break;
2957        }        }
2958      else  
2959        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2960        an initial ']' is taken as a data character -- the code below handles
2961        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2962        [^] must match any character, so generate OP_ALLANY. */
2963    
2964        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2965            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2966        {        {
2967        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2968          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2969          zerofirstbyte = firstbyte;
2970          break;
2971        }        }
2972    
2973        /* If a class contains a negative special such as \S, we need to flip the
2974        negation flag at the end, so that support for characters > 255 works
2975        correctly (they are all included in the class). */
2976    
2977        should_flip_negation = FALSE;
2978    
2979      /* 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
2980      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
2981      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2330  for (;; ptr++) Line 2993  for (;; ptr++)
2993  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2994      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2995      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2996        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2997  #endif  #endif
2998    
2999      /* Process characters until ] is reached. By writing this as a "do" it      /* Process characters until ] is reached. By writing this as a "do" it
# Line 2345  for (;; ptr++) Line 3009  for (;; ptr++)
3009          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3010          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3011          }          }
3012    
3013          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3014          data and reset the pointer. This is so that very large classes that
3015          contain a zillion UTF-8 characters no longer overwrite the work space
3016          (which is on the stack). */
3017    
3018          if (lengthptr != NULL)
3019            {
3020            *lengthptr += class_utf8data - class_utf8data_base;
3021            class_utf8data = class_utf8data_base;
3022            }
3023    
3024  #endif  #endif
3025    
3026        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3027    
3028        if (inescq)        if (inescq)
3029          {          {
3030          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3031            {            {
3032            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3033            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2366  for (;; ptr++) Line 3042  for (;; ptr++)
3042        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3043        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3044    
3045        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3046            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3047            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3048          {          {
3049          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3050          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3051          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3052          uschar pbits[32];          uschar pbits[32];
3053    
3054          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3055            {            {
3056            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3057            goto FAILED;            goto FAILED;
3058            }            }
3059    
3060          ptr += 2;          ptr += 2;
3061          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3062            {            {
3063            local_negate = TRUE;            local_negate = TRUE;
3064              should_flip_negation = TRUE;  /* Note negative special */
3065            ptr++;            ptr++;
3066            }            }
3067    
# Line 2451  for (;; ptr++) Line 3128  for (;; ptr++)
3128        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
3129        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
3130        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3131        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
3132        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3133    
3134        if (c == '\\')        if (c == CHAR_BACKSLASH)
3135          {          {
3136          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3137          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3138    
3139          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 */
3140          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 */
3141          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 */
3142          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3143            {            {
3144            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3145              {              {
3146              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3147              }              }
3148            else inescq = TRUE;            else inescq = TRUE;
3149            continue;            continue;
3150            }            }
3151            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3152    
3153          if (c < 0)          if (c < 0)
3154            {            {
# Line 2486  for (;; ptr++) Line 3164  for (;; ptr++)
3164              continue;              continue;
3165    
3166              case ESC_D:              case ESC_D:
3167                should_flip_negation = TRUE;
3168              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3169              continue;              continue;
3170    
# Line 2494  for (;; ptr++) Line 3173  for (;; ptr++)
3173              continue;              continue;
3174    
3175              case ESC_W:              case ESC_W:
3176                should_flip_negation = TRUE;
3177              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3178              continue;              continue;
3179    
# Line 2503  for (;; ptr++) Line 3183  for (;; ptr++)
3183              continue;              continue;
3184    
3185              case ESC_S:              case ESC_S:
3186                should_flip_negation = TRUE;
3187              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3188              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3189              continue;              continue;
3190    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
3191              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
3192              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
3193              }              }
3194    
3195            /* In the pre-compile phase, just do the recognition. */            /* In the pre-compile phase, just do the recognition. */
3196    
3197            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3198                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3199    
3200              /* We need to deal with \H, \h, \V, and \v in both phases because
3201              they use extra memory. */
3202    
3203              if (-c == ESC_h)
3204                {
3205                SETBIT(classbits, 0x09); /* VT */
3206                SETBIT(classbits, 0x20); /* SPACE */
3207                SETBIT(classbits, 0xa0); /* NSBP */
3208    #ifdef SUPPORT_UTF8
3209                if (utf8)
3210                  {
3211                  class_utf8 = TRUE;
3212                  *class_utf8data++ = XCL_SINGLE;
3213                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3214                  *class_utf8data++ = XCL_SINGLE;
3215                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3216                  *class_utf8data++ = XCL_RANGE;
3217                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3218                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3219                  *class_utf8data++ = XCL_SINGLE;
3220                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3221                  *class_utf8data++ = XCL_SINGLE;
3222                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3223                  *class_utf8data++ = XCL_SINGLE;
3224                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3225                  }
3226    #endif
3227                continue;
3228                }
3229    
3230              if (-c == ESC_H)
3231                {
3232                for (c = 0; c < 32; c++)
3233                  {
3234                  int x = 0xff;
3235                  switch (c)
3236                    {
3237                    case 0x09/8: x ^= 1 << (0x09%8); break;
3238                    case 0x20/8: x ^= 1 << (0x20%8); break;
3239                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3240                    default: break;
3241                    }
3242                  classbits[c] |= x;
3243                  }
3244    
3245    #ifdef SUPPORT_UTF8
3246                if (utf8)
3247                  {
3248                  class_utf8 = TRUE;
3249                  *class_utf8data++ = XCL_RANGE;
3250                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3251                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3252                  *class_utf8data++ = XCL_RANGE;
3253                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3254                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3255                  *class_utf8data++ = XCL_RANGE;
3256                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3257                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3258                  *class_utf8data++ = XCL_RANGE;
3259                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3260                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3261                  *class_utf8data++ = XCL_RANGE;
3262                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3263                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3264                  *class_utf8data++ = XCL_RANGE;
3265                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3266                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3267                  *class_utf8data++ = XCL_RANGE;
3268                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3269                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3270                  }
3271    #endif
3272                continue;
3273                }
3274    
3275              if (-c == ESC_v)
3276                {
3277                SETBIT(classbits, 0x0a); /* LF */
3278                SETBIT(classbits, 0x0b); /* VT */
3279                SETBIT(classbits, 0x0c); /* FF */
3280                SETBIT(classbits, 0x0d); /* CR */
3281                SETBIT(classbits, 0x85); /* NEL */
3282    #ifdef SUPPORT_UTF8
3283                if (utf8)
3284                  {
3285                  class_utf8 = TRUE;
3286                  *class_utf8data++ = XCL_RANGE;
3287                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3288                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3289                  }
3290    #endif
3291                continue;
3292                }
3293    
3294              if (-c == ESC_V)
3295                {
3296                for (c = 0; c < 32; c++)
3297                  {
3298                  int x = 0xff;
3299                  switch (c)
3300                    {
3301                    case 0x0a/8: x ^= 1 << (0x0a%8);
3302                                 x ^= 1 << (0x0b%8);
3303                                 x ^= 1 << (0x0c%8);
3304                                 x ^= 1 << (0x0d%8);
3305                                 break;
3306                    case 0x85/8: x ^= 1 << (0x85%8); break;
3307                    default: break;
3308                    }
3309                  classbits[c] |= x;
3310                  }
3311    
3312    #ifdef SUPPORT_UTF8
3313                if (utf8)
3314                  {
3315                  class_utf8 = TRUE;
3316                  *class_utf8data++ = XCL_RANGE;
3317                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3318                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3319                  *class_utf8data++ = XCL_RANGE;
3320                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3321                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3322                  }
3323    #endif
3324                continue;
3325                }
3326    
3327            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3328    
3329  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2562  for (;; ptr++) Line 3367  for (;; ptr++)
3367        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3368    
3369        CHECK_RANGE:        CHECK_RANGE:
3370        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3371          {          {
3372          inescq = FALSE;          inescq = FALSE;
3373          ptr += 2;          ptr += 2;
# Line 2570  for (;; ptr++) Line 3375  for (;; ptr++)
3375    
3376        oldptr = ptr;        oldptr = ptr;
3377    
3378        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3379    
3380          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3381    
3382          /* Check for range */
3383    
3384          if (!inescq && ptr[1] == CHAR_MINUS)
3385          {          {
3386          int d;          int d;
3387          ptr += 2;          ptr += 2;
3388          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3389    
3390          /* 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
3391          mode. */          mode. */
3392    
3393          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3394            {            {
3395            ptr += 2;            ptr += 2;
3396            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3397                { ptr += 2; continue; }
3398            inescq = TRUE;            inescq = TRUE;
3399            break;            break;
3400            }            }
3401    
3402          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3403            {            {
3404            ptr = oldptr;            ptr = oldptr;
3405            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2606  for (;; ptr++) Line 3418  for (;; ptr++)
3418          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
3419          in such circumstances. */          in such circumstances. */
3420    
3421          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3422            {            {
3423            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3424            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3425    
3426            /* \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
3427            special means the '-' was literal */            special means the '-' was literal */
3428    
3429            if (d < 0)            if (d < 0)
3430              {              {
3431              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3432              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3433              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3434                {                {
3435                ptr = oldptr;                ptr = oldptr;
3436                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2637  for (;; ptr++) Line 3449  for (;; ptr++)
3449    
3450          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3451    
3452            /* Remember \r or \n */
3453    
3454            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3455    
3456          /* 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
3457          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3458          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 2659  for (;; ptr++) Line 3475  for (;; ptr++)
3475              unsigned int origd = d;              unsigned int origd = d;
3476              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3477                {                {
3478                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3479                      ocd <= (unsigned int)d)
3480                    continue;                          /* Skip embedded ranges */
3481    
3482                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3483                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3484                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3485                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3486                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3487                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3488                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3489                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3490                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3491                  d = ocd;                  d = ocd;
3492                  continue;                  continue;
# Line 2751  for (;; ptr++) Line 3571  for (;; ptr++)
3571          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3572            {            {
3573            unsigned int othercase;            unsigned int othercase;
3574            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3575              {              {
3576              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3577              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2778  for (;; ptr++) Line 3598  for (;; ptr++)
3598    
3599      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3600    
3601      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3602    
3603      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3604        {        {
# Line 2786  for (;; ptr++) Line 3606  for (;; ptr++)
3606        goto FAILED;        goto FAILED;
3607        }        }
3608    
3609    
3610    /* This code has been disabled because it would mean that \s counts as
3611    an explicit \r or \n reference, and that's not really what is wanted. Now
3612    we set the flag only if there is a literal "\r" or "\n" in the class. */
3613    
3614    #if 0
3615        /* Remember whether \r or \n are in this class */
3616    
3617        if (negate_class)
3618          {
3619          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3620          }
3621        else
3622          {
3623          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3624          }
3625    #endif
3626    
3627    
3628      /* 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
3629      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
3630      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
3631      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3632      single-bytes only. This is an historical hangover. Maybe one day we can  
3633      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3634        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3635        operate on single-bytes only. This is an historical hangover. Maybe one day
3636        we can tidy these opcodes to handle multi-byte characters.
3637    
3638      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
3639      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note      1-character OP_CHAR[NC] if it's positive, or OP_NOT if it's negative. Note
# Line 2801  for (;; ptr++) Line 3643  for (;; ptr++)
3643      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3644    
3645  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3646      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3647            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3648  #else  #else
3649      if (class_charcount == 1)      if (class_charcount == 1)
3650  #endif  #endif
# Line 2847  for (;; ptr++) Line 3687  for (;; ptr++)
3687      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3688    
3689      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3690      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3691      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
3692        the class, so any that were explicitly given as well can be ignored. If
3693        (when there are explicit characters > 255 that must be listed) there are no
3694        characters < 256, we can omit the bitmap in the actual compiled code. */
3695    
3696  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3697      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3698        {        {
3699        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3700        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2877  for (;; ptr++) Line 3720  for (;; ptr++)
3720        }        }
3721  #endif  #endif
3722    
3723      /* 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
3724      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
3725      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
3726      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3727    
3728        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3729      if (negate_class)      if (negate_class)
3730        {        {
       *code++ = OP_NCLASS;  
3731        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3732          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3733        }        }
3734      else      else
3735        {        {
       *code++ = OP_CLASS;  
3736        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3737        }        }
3738      code += 32;      code += 32;
# Line 2901  for (;; ptr++) Line 3743  for (;; ptr++)
3743      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3744      has been tested above. */      has been tested above. */
3745    
3746      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3747      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3748      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3749      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3750      goto REPEAT;      goto REPEAT;
3751    
3752      case '*':      case CHAR_ASTERISK:
3753      repeat_min = 0;      repeat_min = 0;
3754      repeat_max = -1;      repeat_max = -1;
3755      goto REPEAT;      goto REPEAT;
3756    
3757      case '+':      case CHAR_PLUS:
3758      repeat_min = 1;      repeat_min = 1;
3759      repeat_max = -1;      repeat_max = -1;
3760      goto REPEAT;      goto REPEAT;
3761    
3762      case '?':      case CHAR_QUESTION_MARK:
3763      repeat_min = 0;      repeat_min = 0;
3764      repeat_max = 1;      repeat_max = 1;
3765    
# Line 2952  for (;; ptr++) Line 3794  for (;; ptr++)
3794      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
3795      repeat type to the non-default. */      repeat type to the non-default. */
3796    
3797      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3798        {        {
3799        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3800        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3801        ptr++;        ptr++;
3802        }        }
3803      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3804        {        {
3805        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3806        ptr++;        ptr++;
# Line 3073  for (;; ptr++) Line 3915  for (;; ptr++)
3915    
3916        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3917    
3918          /*--------------------------------------------------------------------*/
3919          /* This code is obsolete from release 8.00; the restriction was finally
3920          removed: */
3921    
3922        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3923        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3924    
3925        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3926          /*--------------------------------------------------------------------*/
3927    
3928        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3929    
# Line 3223  for (;; ptr++) Line 4070  for (;; ptr++)
4070          goto END_REPEAT;          goto END_REPEAT;
4071          }          }
4072    
4073          /*--------------------------------------------------------------------*/
4074          /* This code is obsolete from release 8.00; the restriction was finally
4075          removed: */
4076    
4077        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4078        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4079    
4080        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4081          /*--------------------------------------------------------------------*/
4082    
4083        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4084          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3262  for (;; ptr++) Line 4114  for (;; ptr++)
4114          goto FAILED;          goto FAILED;
4115          }          }
4116    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4117        /* 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
4118        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
4119        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 3292  for (;; ptr++) Line 4136  for (;; ptr++)
4136    
4137        if (repeat_min == 0)        if (repeat_min == 0)
4138          {          {
4139          /* 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
4140          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4141    
4142          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4143          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4144          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4145          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4146          the start of the whole regex. Temporarily terminate the pattern while          **   }
4147          doing this. */  
4148            However, that fails when a group is referenced as a subroutine from
4149            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4150            so that it is skipped on execution. As we don't have a list of which
4151            groups are referenced, we cannot do this selectively.
4152    
4153            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4154            and do no more at this point. However, we do need to adjust any
4155            OP_RECURSE calls inside the group that refer to the group itself or any
4156            internal or forward referenced group, because the offset is from the
4157            start of the whole regex. Temporarily terminate the pattern while doing
4158            this. */
4159    
4160          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4161            {            {
4162            *code = OP_END;            *code = OP_END;
4163            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4164            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4165            code++;            code++;
4166              if (repeat_max == 0)
4167                {
4168                *previous++ = OP_SKIPZERO;
4169                goto END_REPEAT;
4170                }
4171            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4172            }            }
4173    
# Line 3358  for (;; ptr++) Line 4212  for (;; ptr++)
4212          if (repeat_min > 1)          if (repeat_min > 1)
4213            {            {
4214            /* 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
4215            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4216              potential integer overflow. */
4217    
4218            if (lengthptr != NULL)            if (lengthptr != NULL)
4219              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4220                int delta = (repeat_min - 1)*length_prevgroup;
4221                if ((double)(repeat_min - 1)*(double)length_prevgroup >
4222                                                                (double)INT_MAX ||
4223                    OFLOW_MAX - *lengthptr < delta)
4224                  {
4225                  *errorcodeptr = ERR20;
4226                  goto FAILED;
4227                  }
4228                *lengthptr += delta;
4229                }
4230    
4231            /* This is compiling for real */            /* This is compiling for real */
4232    
# Line 3399  for (;; ptr++) Line 4264  for (;; ptr++)
4264          /* 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
4265          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
4266          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
4267          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4268            paranoid checks to avoid integer overflow. */
4269    
4270          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4271            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4272              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4273                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4274              if ((double)repeat_max *
4275                    (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4276                      > (double)INT_MAX ||
4277                  OFLOW_MAX - *lengthptr < delta)
4278                {
4279                *errorcodeptr = ERR20;
4280                goto FAILED;
4281                }
4282              *lengthptr += delta;
4283              }
4284    
4285          /* This is compiling for real */          /* This is compiling for real */
4286    
# Line 3485  for (;; ptr++) Line 4362  for (;; ptr++)
4362          }          }
4363        }        }
4364    
4365        /* If previous is OP_FAIL, it was generated by an empty class [] in
4366        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4367        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4368        error above. We can just ignore the repeat in JS case. */
4369    
4370        else if (*previous == OP_FAIL) goto END_REPEAT;
4371    
4372      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4373    
4374      else      else
# Line 3509  for (;; ptr++) Line 4393  for (;; ptr++)
4393      if (possessive_quantifier)      if (possessive_quantifier)
4394        {        {
4395        int len;        int len;
4396        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4397            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4398            tempcode += _pcre_OP_lengths[*tempcode] +
4399              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4400    
4401          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4402            {
4403          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4404    #ifdef SUPPORT_UTF8
4405            if (utf8 && tempcode[-1] >= 0xc0)
4406              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4407    #endif
4408            }
4409    
4410        len = code - tempcode;        len = code - tempcode;
4411        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4412          {          {
# Line 3555  for (;; ptr++) Line 4450  for (;; ptr++)
4450      /* ===================================================================*/      /* ===================================================================*/
4451      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4452      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4453      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. */  
4454    
4455      case '(':      case CHAR_LEFT_PARENTHESIS:
4456      newoptions = options;      newoptions = options;
4457      skipbytes = 0;      skipbytes = 0;
4458      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4459      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4460        reset_bracount = FALSE;
4461    
4462        /* First deal with various "verbs" that can be introduced by '*'. */
4463    
4464        if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4465          {
4466          int i, namelen;
4467          const char *vn = verbnames;
4468          const uschar *name = ++ptr;
4469          previous = NULL;
4470          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4471          if (*ptr == CHAR_COLON)
4472            {
4473            *errorcodeptr = ERR59;   /* Not supported */
4474            goto FAILED;
4475            }
4476          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4477            {
4478            *errorcodeptr = ERR60;
4479            goto FAILED;
4480            }
4481          namelen = ptr - name;
4482          for (i = 0; i < verbcount; i++)
4483            {
4484            if (namelen == verbs[i].len &&
4485                strncmp((char *)name, vn, namelen) == 0)
4486              {
4487              /* Check for open captures before ACCEPT */
4488    
4489              if (verbs[i].op == OP_ACCEPT)
4490                {
4491                open_capitem *oc;
4492                cd->had_accept = TRUE;
4493                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4494                  {
4495                  *code++ = OP_CLOSE;
4496                  PUT2INC(code, 0, oc->number);
4497                  }
4498                }
4499              *code++ = verbs[i].op;
4500              break;
4501              }
4502            vn += verbs[i].len + 1;
4503            }
4504          if (i < verbcount) continue;
4505          *errorcodeptr = ERR60;
4506          goto FAILED;
4507          }
4508    
4509        /* Deal with the extended parentheses; all are introduced by '?', and the
4510        appearance of any of them means that this is not a capturing group. */
4511    
4512      if (*(++ptr) == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4513        {        {
4514        int i, set, unset, namelen;        int i, set, unset, namelen;
4515        int *optset;        int *optset;
# Line 3574  for (;; ptr++) Line 4518  for (;; ptr++)
4518    
4519        switch (*(++ptr))        switch (*(++ptr))
4520          {          {
4521          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4522          ptr++;          ptr++;
4523          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4524          if (*ptr == 0)          if (*ptr == 0)
4525            {            {
4526            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3586  for (;; ptr++) Line 4530  for (;; ptr++)
4530    
4531    
4532          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4533          case ':':                 /* Non-capturing bracket */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4534            reset_bracount = TRUE;
4535            /* Fall through */
4536    
4537            /* ------------------------------------------------------------ */
4538            case CHAR_COLON:          /* Non-capturing bracket */
4539          bravalue = OP_BRA;          bravalue = OP_BRA;
4540          ptr++;          ptr++;
4541          break;          break;
4542    
4543    
4544          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4545          case '(':          case CHAR_LEFT_PARENTHESIS:
4546          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4547    
4548          /* A condition can be an assertion, a number (referring to a numbered          /* A condition can be an assertion, a number (referring to a numbered
# Line 3613  for (;; ptr++) Line 4562  for (;; ptr++)
4562          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4563          including assertions, are processed. */          including assertions, are processed. */
4564    
4565          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4566                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4567            break;            break;
4568    
4569          /* Most other conditions use OP_CREF (a couple change to OP_RREF          /* Most other conditions use OP_CREF (a couple change to OP_RREF
# Line 3621  for (;; ptr++) Line 4571  for (;; ptr++)
4571    
4572          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
4573          skipbytes = 3;          skipbytes = 3;
4574            refsign = -1;
4575    
4576          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4577    
4578          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4579            {            {
4580            terminator = -1;            terminator = -1;
4581            ptr += 2;            ptr += 2;
# Line 3634  for (;; ptr++) Line 4585  for (;; ptr++)
4585          /* 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
4586          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4587    
4588          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4589            {            {
4590            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4591            ptr++;            ptr++;
4592            }            }
4593          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4594            {            {
4595            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4596            ptr++;            ptr++;
4597            }            }
4598          else terminator = 0;          else
4599              {
4600              terminator = 0;
4601              if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4602              }
4603    
4604          /* 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 */
4605    
# Line 3663  for (;; ptr++) Line 4618  for (;; ptr++)
4618            {            {
4619            if (recno >= 0)            if (recno >= 0)
4620              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4621                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4622            ptr++;            ptr++;
4623            }            }
4624          namelen = ptr - name;          namelen = ptr - name;
4625    
4626          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4627                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4628            {            {
4629            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4630            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3680  for (;; ptr++) Line 4636  for (;; ptr++)
4636          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
4637    
4638          /* 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
4639          reference. */          reference. If the string started with "+" or "-" we require the rest to
4640            be digits, in which case recno will be set. */
4641    
4642            if (refsign > 0)
4643              {
4644              if (recno <= 0)
4645                {
4646                *errorcodeptr = ERR58;
4647                goto FAILED;
4648                }
4649              recno = (refsign == CHAR_MINUS)?
4650                cd->bracount - recno + 1 : recno +cd->bracount;
4651              if (recno <= 0 || recno > cd->final_bracount)
4652                {
4653                *errorcodeptr = ERR15;
4654                goto FAILED;
4655                }
4656              PUT2(code, 2+LINK_SIZE, recno);
4657              break;
4658              }
4659    
4660            /* Otherwise (did not start with "+" or "-"), start by looking for the
4661            name. If we find a name, add one to the opcode to change OP_CREF or
4662            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4663            except they record that the reference was originally to a name. The
4664            information is used to check duplicate names. */
4665    
4666          slot = cd->name_table;          slot = cd->name_table;
4667          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 3695  for (;; ptr++) Line 4676  for (;; ptr++)
4676            {            {
4677            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4678            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4679              code[1+LINK_SIZE]++;
4680            }            }
4681    
4682          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4683    
4684          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4685                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4686            {            {
4687            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4688              code[1+LINK_SIZE]++;
4689            }            }
4690    
4691          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 3720  for (;; ptr++) Line 4703  for (;; ptr++)
4703          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4704          specific group number. */          specific group number. */
4705    
4706          else if (*name == 'R')          else if (*name == CHAR_R)
4707            {            {
4708            recno = 0;            recno = 0;
4709            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 3730  for (;; ptr++) Line 4713  for (;; ptr++)
4713                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4714                goto FAILED;                goto FAILED;
4715                }                }
4716              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4717              }              }
4718            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4719            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 3740  for (;; ptr++) Line 4723  for (;; ptr++)
4723          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4724          false. */          false. */
4725    
4726          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4727            {            {
4728            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4729            skipbytes = 1;            skipbytes = 1;
4730            }            }
4731    
4732          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4733            in the second pass here, so final_bracount is set. */
4734    
4735          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4736            {            {
4737            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4738            }            }
# Line 3764  for (;; ptr++) Line 4748  for (;; ptr++)
4748    
4749    
4750          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4751          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4752          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4753          ptr++;          ptr++;
4754          break;          break;
4755    
4756    
4757          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4758          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4759          ptr++;          ptr++;
4760            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4761              {
4762              *code++ = OP_FAIL;
4763              previous = NULL;
4764              continue;
4765              }
4766            bravalue = OP_ASSERT_NOT;
4767          break;          break;
4768    
4769    
4770          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4771          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4772          switch (ptr[1])          switch (ptr[1])
4773            {            {
4774            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4775            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4776            ptr += 2;            ptr += 2;
4777            break;            break;
4778    
4779            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4780            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4781            ptr += 2;            ptr += 2;
4782            break;            break;
# Line 3801  for (;; ptr++) Line 4791  for (;; ptr++)
4791    
4792    
4793          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4794          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4795          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4796          ptr++;          ptr++;
4797          break;          break;
4798    
4799    
4800          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4801          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4802          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4803          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4804          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4805            {            {
4806            int n = 0;            int n = 0;
4807            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4808              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4809            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4810              {              {
4811              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4812              goto FAILED;              goto FAILED;
# Line 3836  for (;; ptr++) Line 4826  for (;; ptr++)
4826    
4827    
4828          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4829          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4830          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4831                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4832            {            {
4833            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4834            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4835            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4836            }            }
4837          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4838            {            {
4839            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4840            goto FAILED;            goto FAILED;
# Line 3853  for (;; ptr++) Line 4844  for (;; ptr++)
4844    
4845          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4846          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4847          case '\'':          case CHAR_APOSTROPHE:
4848            {            {
4849            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4850                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4851            name = ++ptr;            name = ++ptr;
4852    
4853            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 3886  for (;; ptr++) Line 4878  for (;; ptr++)
4878                }                }
4879              }              }
4880    
4881            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4882              alphabetical order. Duplicate names for different numbers are
4883              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4884              number are always OK. (An existing number can be re-used if (?|
4885              appears in the pattern.) In either event, a duplicate name results in
4886              a duplicate entry in the table, even if the number is the same. This
4887              is because the number of names, and hence the table size, is computed
4888              in the pre-compile, and it affects various numbers and pointers which
4889              would all have to be modified, and the compiled code moved down, if
4890              duplicates with the same number were omitted from the table. This
4891              doesn't seem worth the hassle. However, *different* names for the
4892              same number are not permitted. */
4893    
4894            else            else
4895              {              {
4896                BOOL dupname = FALSE;
4897              slot = cd->name_table;              slot = cd->name_table;
4898    
4899              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4900                {                {
4901                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 3898  for (;; ptr++) Line 4903  for (;; ptr++)
4903                  {                  {
4904                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4905                    {                    {
4906                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4907                          (options & PCRE_DUPNAMES) == 0)
4908                      {                      {
4909                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4910                      goto FAILED;                      goto FAILED;
4911                      }                      }
4912                      else dupname = TRUE;
4913                    }                    }
4914                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4915                  }                  }
4916    
4917                  /* Make space in the table and break the loop for an earlier
4918                  name. For a duplicate or later name, carry on. We do this for
4919                  duplicates so that in the simple case (when ?(| is not used) they
4920                  are in order of their numbers. */
4921    
4922                if (crc < 0)                if (crc < 0)
4923                  {                  {
4924                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4925                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4926                  break;                  break;
4927                  }                  }
4928    
4929                  /* Continue the loop for a later or duplicate name */
4930    
4931                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4932                }                }
4933    
4934                /* For non-duplicate names, check for a duplicate number before
4935                adding the new name. */
4936    
4937                if (!dupname)
4938                  {
4939                  uschar *cslot = cd->name_table;
4940                  for (i = 0; i < cd->names_found; i++)
4941                    {
4942                    if (cslot != slot)
4943                      {
4944                      if (GET2(cslot, 0) == cd->bracount + 1)
4945                        {
4946                        *errorcodeptr = ERR65;
4947                        goto FAILED;
4948                        }
4949                      }
4950                    else i--;
4951                    cslot += cd->name_entry_size;
4952                    }
4953                  }
4954    
4955              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4956              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
# Line 3921  for (;; ptr++) Line 4958  for (;; ptr++)
4958              }              }
4959            }            }
4960    
4961          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
4962            encountered. */
4963    
         ptr++;                    /* Move past > or ' */  
4964          cd->names_found++;          cd->names_found++;
4965            ptr++;                    /* Move past > or ' */
4966          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
4967    
4968    
4969          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4970          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4971          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4972          is_recurse = TRUE;          is_recurse = TRUE;
4973          /* Fall through */          /* Fall through */
4974    
4975          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4976          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4977          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4978            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4979            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4980    
4981          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4982          name = ++ptr;          name = ++ptr;
# Line 3948  for (;; ptr++) Line 4988  for (;; ptr++)
4988    
4989          if (lengthptr != NULL)          if (lengthptr != NULL)
4990            {            {
4991              if (namelen == 0)
4992                {
4993                *errorcodeptr = ERR62;
4994                goto FAILED;
4995                }
4996            if (*ptr != terminator)            if (*ptr != terminator)
4997              {              {
4998              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 3961  for (;; ptr++) Line 5006  for (;; ptr++)
5006            recno = 0;            recno = 0;
5007            }            }
5008    
5009          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
5010            first, and then check that we have reached the end of the name in the
5011            table. That way, if the name that is longer than any in the table,
5012            the comparison will fail without reading beyond the table entry. */
5013    
5014          else          else
5015            {            {
5016            slot = cd->name_table;            slot = cd->name_table;
5017            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5018              {              {
5019              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
5020                    slot[2+namelen] == 0)
5021                  break;
5022              slot += cd->name_entry_size;              slot += cd->name_entry_size;
5023              }              }
5024    
# Line 3977  for (;; ptr++) Line 5027  for (;; ptr++)
5027              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5028              }              }
5029            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5030                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5031                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5032              {              {
5033              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 3993  for (;; ptr++) Line 5043  for (;; ptr++)
5043    
5044    
5045          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5046          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5047          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5048          /* Fall through */          /* Fall through */
5049    
5050    
5051          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5052          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5053          case '5': case '6': case '7': case '8': case '9':   /* subroutine */          case CHAR_0: case CHAR_1: case CHAR_2: case CHAR_3: case CHAR_4:
5054            case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
5055            {            {
5056            const uschar *called;            const uschar *called;
5057              terminator = CHAR_RIGHT_PARENTHESIS;
5058    
5059              /* Come here from the \g<...> and \g'...' code (Oniguruma
5060              compatibility). However, the syntax has been checked to ensure that
5061              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5062              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5063              ever be taken. */
5064    
5065              HANDLE_NUMERICAL_RECURSION:
5066    
5067              if ((refsign = *ptr) == CHAR_PLUS)
5068                {
5069                ptr++;
5070                if ((digitab[*ptr] & ctype_digit) == 0)
5071                  {
5072                  *errorcodeptr = ERR63;
5073                  goto FAILED;
5074                  }
5075                }
5076              else if (refsign == CHAR_MINUS)
5077                {
5078                if ((digitab[ptr[1]] & ctype_digit) == 0)
5079                  goto OTHER_CHAR_AFTER_QUERY;
5080                ptr++;
5081                }
5082    
5083            recno = 0;            recno = 0;
5084            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5085              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5086            if (*ptr != ')')  
5087              if (*ptr != terminator)
5088              {              {
5089              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5090              goto FAILED;              goto FAILED;
5091              }              }
5092    
5093              if (refsign == CHAR_MINUS)
5094                {
5095                if (recno == 0)
5096                  {
5097                  *errorcodeptr = ERR58;
5098                  goto FAILED;
5099                  }
5100                recno = cd->bracount - recno + 1;
5101                if (recno <= 0)
5102                  {
5103                  *errorcodeptr = ERR15;
5104                  goto FAILED;
5105                  }
5106                }
5107              else if (refsign == CHAR_PLUS)
5108                {
5109                if (recno == 0)
5110                  {
5111                  *errorcodeptr = ERR58;
5112                  goto FAILED;
5113                  }
5114                recno += cd->bracount;
5115                }
5116    
5117            /* Come here from code above that handles a named recursion */            /* Come here from code above that handles a named recursion */
5118    
5119            HANDLE_RECURSION:            HANDLE_RECURSION:
# Line 4029  for (;; ptr++) Line 5131  for (;; ptr++)
5131            if (lengthptr == NULL)            if (lengthptr == NULL)
5132              {              {
5133              *code = OP_END;              *code = OP_END;
5134              if (recno != 0) called = find_bracket(cd->start_code, utf8, recno);              if (recno != 0)
5135                  called = _pcre_find_bracket(cd->start_code, utf8, recno);
5136    
5137              /* Forward reference */              /* Forward reference */
5138    
5139              if (called == NULL)              if (called == NULL)
5140                {                {
5141                if (find_parens(ptr, cd->bracount, NULL, recno,                if (find_parens(cd, NULL, recno,
5142                     (options & PCRE_EXTENDED) != 0) < 0)                      (options & PCRE_EXTENDED) != 0) < 0)
5143                  {                  {
5144                  *errorcodeptr = ERR15;                  *errorcodeptr = ERR15;
5145                  goto FAILED;                  goto FAILED;
# Line 4084  for (;; ptr++) Line 5187  for (;; ptr++)
5187    
5188          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5189          default:              /* Other characters: check option setting */          default:              /* Other characters: check option setting */
5190            OTHER_CHAR_AFTER_QUERY:
5191          set = unset = 0;          set = unset = 0;
5192          optset = &set;          optset = &set;
5193    
5194          while (*ptr != ')' && *ptr != ':')          while (*ptr != CHAR_RIGHT_PARENTHESIS && *ptr != CHAR_COLON)
5195            {            {
5196            switch (*ptr++)            switch (*ptr++)
5197              {              {
5198              case '-': optset = &unset; break;              case CHAR_MINUS: optset = &unset; break;
5199    
5200              case 'J':    /* Record that it changed in the external options */              case CHAR_J:    /* Record that it changed in the external options */
5201              *optset |= PCRE_DUPNAMES;              *optset |= PCRE_DUPNAMES;
5202              cd->external_options |= PCRE_JCHANGED;              cd->external_flags |= PCRE_JCHANGED;
5203              break;              break;
5204    
5205              case 'i': *optset |= PCRE_CASELESS; break;              case CHAR_i: *optset |= PCRE_CASELESS; break;
5206              case 'm': *optset |= PCRE_MULTILINE; break;              case CHAR_m: *optset |= PCRE_MULTILINE; break;
5207              case 's': *optset |= PCRE_DOTALL; break;              case CHAR_s: *optset |= PCRE_DOTALL; break;
5208              case 'x': *optset |= PCRE_EXTENDED; break;              case CHAR_x: *optset |= PCRE_EXTENDED; break;
5209              case 'U': *optset |= PCRE_UNGREEDY; break;              case CHAR_U: *optset |= PCRE_UNGREEDY; break;
5210              case 'X': *optset |= PCRE_EXTRA; break;              case CHAR_X: *optset |= PCRE_EXTRA; break;
5211    
5212              default:  *errorcodeptr = ERR12;              default:  *errorcodeptr = ERR12;
5213                        ptr--;    /* Correct the offset */                        ptr--;    /* Correct the offset */
# Line 4134  for (;; ptr++) Line 5238  for (;; ptr++)
5238          both phases.          both phases.
5239    
5240          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
5241          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
5242          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. */  
5243    
5244          if (*ptr == ')')          if (*ptr == CHAR_RIGHT_PARENTHESIS)
5245            {            {
5246            if (code == cd->start_code + 1 + LINK_SIZE &&            if (code == cd->start_code + 1 + LINK_SIZE &&
5247                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))                 (lengthptr == NULL || *lengthptr == 2 + 2*LINK_SIZE))
5248              {              {
5249              cd->external_options = newoptions;              cd->external_options = newoptions;
             options = newoptions;  
5250              }              }
5251           else           else
5252              {              {
# Line 4154  for (;; ptr++) Line 5255  for (;; ptr++)
5255                *code++ = OP_OPT;                *code++ = OP_OPT;
5256                *code++ = newoptions & PCRE_IMS;                *code++ = newoptions & PCRE_IMS;
5257                }                }
   
             /* 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;  
5258              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);              greedy_default = ((newoptions & PCRE_UNGREEDY) != 0);
5259              greedy_non_default = greedy_default ^ 1;              greedy_non_default = greedy_default ^ 1;
5260              req_caseopt = ((options & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;              req_caseopt = ((newoptions & PCRE_CASELESS) != 0)? REQ_CASELESS : 0;
5261              }              }
5262    
5263              /* Change options at this level, and pass them back for use
5264              in subsequent branches. When not at the start of the pattern, this
5265              information is also necessary so that a resetting item can be
5266              compiled at the end of a group (if we are in a group). */
5267    
5268              *optionsptr = options = newoptions;
5269            previous = NULL;       /* This item can't be repeated */            previous = NULL;       /* This item can't be repeated */
5270            continue;              /* It is complete */            continue;              /* It is complete */
5271            }            }
# Line 4218  for (;; ptr++) Line 5319  for (;; ptr++)
5319           errorcodeptr,                 /* Where to put an error message */           errorcodeptr,                 /* Where to put an error message */
5320           (bravalue == OP_ASSERTBACK ||           (bravalue == OP_ASSERTBACK ||
5321            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */            bravalue == OP_ASSERTBACK_NOT), /* TRUE if back assert */
5322             reset_bracount,               /* True if (?| group */
5323           skipbytes,                    /* Skip over bracket number */           skipbytes,                    /* Skip over bracket number */
5324           &subfirstbyte,                /* For possible first char */           &subfirstbyte,                /* For possible first char */
5325           &subreqbyte,                  /* For possible last char */           &subreqbyte,                  /* For possible last char */
# Line 4234  for (;; ptr++) Line 5336  for (;; ptr++)
5336      is on the bracket. */      is on the bracket. */
5337    
5338      /* 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
5339      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
5340        in the real compile phase, not in the pre-pass, where the whole group may
5341        not be available. */
5342    
5343      if (bravalue == OP_COND)      if (bravalue == OP_COND && lengthptr == NULL)
5344        {        {
5345        uschar *tc = code;        uschar *tc = code;
5346        int condcount = 0;        int condcount = 0;
# Line 4277  for (;; ptr++) Line 5381  for (;; ptr++)
5381    
5382      /* Error if hit end of pattern */      /* Error if hit end of pattern */
5383    
5384      if (*ptr != ')')      if (*ptr != CHAR_RIGHT_PARENTHESIS)
5385        {        {
5386        *errorcodeptr = ERR14;        *errorcodeptr = ERR14;
5387        goto FAILED;        goto FAILED;
5388        }        }
5389    
5390      /* 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,
5391      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
5392      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
5393      a quantifier. */      duplicated by a quantifier.*/
5394    
5395      if (lengthptr != NULL)      if (lengthptr != NULL)
5396        {        {
5397          if (OFLOW_MAX - *lengthptr < length_prevgroup - 2 - 2*LINK_SIZE)
5398            {
5399            *errorcodeptr = ERR20;
5400            goto FAILED;
5401            }
5402        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;        *lengthptr += length_prevgroup - 2 - 2*LINK_SIZE;
5403        code++;        *code++ = OP_BRA;
5404        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5405        *code++ = OP_KET;        *code++ = OP_KET;
5406        PUTINC(code, 0, 1 + LINK_SIZE);        PUTINC(code, 0, 1 + LINK_SIZE);
5407          break;    /* No need to waste time with special character handling */
5408        }        }
5409    
5410      /* Otherwise update the main code pointer to the end of the group. */      /* Otherwise update the main code pointer to the end of the group. */
5411    
5412      else code = tempcode;      code = tempcode;
5413    
5414      /* For a DEFINE group, required and first character settings are not      /* For a DEFINE group, required and first character settings are not
5415      relevant. */      relevant. */
# Line 4369  for (;; ptr++) Line 5479  for (;; ptr++)
5479      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
5480      have to change if any new ones are ever created. */      have to change if any new ones are ever created. */
5481    
5482      case '\\':      case CHAR_BACKSLASH:
5483      tempptr = ptr;      tempptr = ptr;
5484      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);      c = check_escape(&ptr, errorcodeptr, cd->bracount, options, FALSE);
5485      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
# Line 4378  for (;; ptr++) Line 5488  for (;; ptr++)
5488        {        {
5489        if (-c == ESC_Q)            /* Handle start of quoted string */        if (-c == ESC_Q)            /* Handle start of quoted string */
5490          {          {
5491          if (ptr[1] == '\\' && ptr[2] == 'E') ptr += 2; /* avoid empty string */          if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
5492            else inescq = TRUE;            ptr += 2;               /* avoid empty string */
5493                else inescq = TRUE;
5494          continue;          continue;
5495          }          }
5496    
# Line 4396  for (;; ptr++) Line 5507  for (;; ptr++)
5507        zerofirstbyte = firstbyte;        zerofirstbyte = firstbyte;
5508        zeroreqbyte = reqbyte;        zeroreqbyte = reqbyte;
5509    
5510        /* \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'
5511          is a subroutine call by number (Oniguruma syntax). In fact, the value
5512          -ESC_g is returned only for these cases. So we don't need to check for <
5513          or ' if the value is -ESC_g. For the Perl syntax \g{n} the value is
5514          -ESC_REF+n, and for the Perl syntax \g{name} the result is -ESC_k (as
5515          that is a synonym for a named back reference). */
5516    
5517          if (-c == ESC_g)
5518            {
5519            const uschar *p;
5520            save_hwm = cd->hwm;   /* Normally this is set when '(' is read */
5521            terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5522              CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5523    
5524            /* These two statements stop the compiler for warning about possibly
5525            unset variables caused by the jump to HANDLE_NUMERICAL_RECURSION. In
5526            fact, because we actually check for a number below, the paths that
5527            would actually be in error are never taken. */
5528    
5529            skipbytes = 0;
5530            reset_bracount = FALSE;
5531    
5532            /* Test for a name */
5533    
5534            if (ptr[1] != CHAR_PLUS && ptr[1] != CHAR_MINUS)
5535              {
5536              BOOL isnumber = TRUE;
5537              for (p = ptr + 1; *p != 0 && *p != terminator; p++)
5538                {
5539                if ((cd->ctypes[*p] & ctype_digit) == 0) isnumber = FALSE;
5540                if ((cd->ctypes[*p] & ctype_word) == 0) break;
5541                }
5542              if (*p != terminator)
5543                {
5544                *errorcodeptr = ERR57;
5545                break;
5546                }
5547              if (isnumber)
5548                {
5549                ptr++;
5550                goto HANDLE_NUMERICAL_RECURSION;
5551                }
5552              is_recurse = TRUE;
5553              goto NAMED_REF_OR_RECURSE;
5554              }
5555    
5556            /* Test a signed number in angle brackets or quotes. */
5557    
5558            p = ptr + 2;
5559            while ((digitab[*p] & ctype_digit) != 0) p++;
5560            if (*p != terminator)
5561              {
5562              *errorcodeptr = ERR57;
5563              break;
5564              }
5565            ptr++;
5566            goto HANDLE_NUMERICAL_RECURSION;
5567            }
5568    
5569          /* \k<name> or \k'name' is a back reference by name (Perl syntax).
5570          We also support \k{name} (.NET syntax) */
5571    
5572        if (-c == ESC_k && (ptr[1] == '<' || ptr[1] == '\''))        if (-c == ESC_k && (ptr[1] == CHAR_LESS_THAN_SIGN ||
5573              ptr[1] == CHAR_APOSTROPHE || ptr[1] == CHAR_LEFT_CURLY_BRACKET))
5574          {          {
5575          is_recurse = FALSE;          is_recurse = FALSE;
5576          terminator = (*(++ptr) == '<')? '>' : '\'';          terminator = (*(++ptr) == CHAR_LESS_THAN_SIGN)?
5577              CHAR_GREATER_THAN_SIGN : (*ptr == CHAR_APOSTROPHE)?
5578              CHAR_APOSTROPHE : CHAR_RIGHT_CURLY_BRACKET;
5579          goto NAMED_REF_OR_RECURSE;          goto NAMED_REF_OR_RECURSE;
5580          }          }
5581    
# Line 4502  for (;; ptr++) Line 5676  for (;; ptr++)
5676      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;      *code++ = ((options & PCRE_CASELESS) != 0)? OP_CHARNC : OP_CHAR;
5677      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];      for (c = 0; c < mclength; c++) *code++ = mcbuffer[c];
5678    
5679        /* Remember if \r or \n were seen */
5680    
5681        if (mcbuffer[0] == CHAR_CR || mcbuffer[0] == CHAR_NL)
5682          cd->external_flags |= PCRE_HASCRORLF;
5683    
5684      /* Set the first and required bytes appropriately. If no previous first      /* Set the first and required bytes appropriately. If no previous first
5685      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.
5686      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 4567  This function is used during the pre-com Line 5746  This function is used during the pre-com
5746  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
5747  value of lengthptr distinguishes the two phases.  value of lengthptr distinguishes the two phases.
5748    
5749  Argument:  Arguments:
5750    options        option bits, including any changes for this subpattern    options        option bits, including any changes for this subpattern
5751    oldims         previous settings of ims option bits    oldims         previous settings of ims option bits
5752    codeptr        -> the address of the current code pointer    codeptr        -> the address of the current code pointer
5753    ptrptr         -> the address of the current pattern pointer    ptrptr         -> the address of the current pattern pointer
5754    errorcodeptr   -> pointer to error code variable    errorcodeptr   -> pointer to error code variable
5755    lookbehind     TRUE if this is a lookbehind assertion    lookbehind     TRUE if this is a lookbehind assertion
5756      reset_bracount TRUE to reset the count for each branch
5757    skipbytes      skip this many bytes at start (for brackets and OP_COND)    skipbytes      skip this many bytes at start (for brackets and OP_COND)
5758    firstbyteptr   place to put the first required character, or a negative number    firstbyteptr   place to put the first required character, or a negative number
5759    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 4587  Returns:         TRUE on success Line 5767  Returns:         TRUE on success
5767    
5768  static BOOL  static BOOL
5769  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,  compile_regex(int options, int oldims, uschar **codeptr, const uschar **ptrptr,
5770    int *errorcodeptr, BOOL lookbehind, int skipbytes, int *firstbyteptr,    int *errorcodeptr, BOOL lookbehind, BOOL reset_bracount, int skipbytes,
5771    int *reqbyteptr, branch_chain *bcptr, compile_data *cd, int *lengthptr)    int *firstbyteptr, int *reqbyteptr, branch_chain *bcptr, compile_data *cd,
5772      int *lengthptr)
5773  {  {
5774  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
5775  uschar *code = *codeptr;  uschar *code = *codeptr;
5776  uschar *last_branch = code;  uschar *last_branch = code;
5777  uschar *start_bracket = code;  uschar *start_bracket = code;
5778  uschar *reverse_count = NULL;  uschar *reverse_count = NULL;
5779    open_capitem capitem;
5780    int capnumber = 0;
5781  int firstbyte, reqbyte;  int firstbyte, reqbyte;
5782  int branchfirstbyte, branchreqbyte;  int branchfirstbyte, branchreqbyte;
5783  int length;  int length;
5784    int orig_bracount;
5785    int max_bracount;
5786  branch_chain bc;  branch_chain bc;
5787    
5788  bc.outer = bcptr;  bc.outer = bcptr;
# Line 4619  the code that abstracts option settings Line 5804  the code that abstracts option settings
5804  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
5805  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. */
5806    
5807    /* If this is a capturing subpattern, add to the chain of open capturing items
5808    so that we can detect them if (*ACCEPT) is encountered. */
5809    
5810    if (*code == OP_CBRA)
5811      {
5812      capnumber = GET2(code, 1 + LINK_SIZE);
5813      capitem.number = capnumber;
5814      capitem.next = cd->open_caps;
5815      cd->open_caps = &capitem;
5816      }
5817    
5818  /* Offset is set zero to mark that this bracket is still open */  /* Offset is set zero to mark that this bracket is still open */
5819    
5820  PUT(code, 1, 0);  PUT(code, 1, 0);
# Line 4626  code += 1 + LINK_SIZE + skipbytes; Line 5822  code += 1 + LINK_SIZE + skipbytes;
5822    
5823  /* Loop for each alternative branch */  /* Loop for each alternative branch */
5824    
5825    orig_bracount = max_bracount = cd->bracount;
5826  for (;;)  for (;;)
5827    {    {
5828      /* For a (?| group, reset the capturing bracket count so that each branch
5829      uses the same numbers. */
5830    
5831      if (reset_bracount) cd->bracount = orig_bracount;
5832    
5833    /* Handle a change of ims options at the start of the branch */    /* Handle a change of ims options at the start of the branch */
5834    
5835    if ((options & PCRE_IMS) != oldims)    if ((options & PCRE_IMS) != oldims)
# Line 4657  for (;;) Line 5859  for (;;)
5859      return FALSE;      return FALSE;
5860      }      }
5861    
5862      /* Keep the highest bracket count in case (?| was used and some branch
5863      has fewer than the rest. */
5864    
5865      if (cd->bracount > max_bracount) max_bracount = cd->bracount;
5866    
5867    /* 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. */
5868    
5869    if (lengthptr == NULL)    if (lengthptr == NULL)
# Line 4702  for (;;) Line 5909  for (;;)
5909    
5910      /* If lookbehind, check that this branch matches a fixed-length string, and      /* If lookbehind, check that this branch matches a fixed-length string, and
5911      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
5912      branch with OP_END. */      branch with OP_END. If the branch contains OP_RECURSE, the result is -3
5913        because there may be forward references that we can't check here. Set a
5914        flag to cause another lookbehind check at the end. Why not do it all at the
5915        end? Because common, erroneous checks are picked up here and the offset of
5916        the problem can be shown. */
5917    
5918      if (lookbehind)      if (lookbehind)
5919        {        {
5920        int fixed_length;        int fixed_length;
5921        *code = OP_END;        *code = OP_END;
5922        fixed_length = find_fixedlength(last_branch, options);        fixed_length = find_fixedlength(last_branch, options, FALSE, cd);
5923        DPRINTF(("fixed length = %d\n", fixed_length));        DPRINTF(("fixed length = %d\n", fixed_length));
5924        if (fixed_length < 0)        if (fixed_length == -3)
5925            {
5926            cd->check_lookbehind = TRUE;
5927            }
5928          else if (fixed_length < 0)
5929          {          {
5930          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;          *errorcodeptr = (fixed_length == -2)? ERR36 : ERR25;
5931          *ptrptr = ptr;          *ptrptr = ptr;
5932          return FALSE;          return FALSE;
5933          }          }
5934        PUT(reverse_count, 0, fixed_length);        else { PUT(reverse_count, 0, fixed_length); }
5935        }        }
5936      }      }
5937    
5938    /* Reached end of expression, either ')' or end of pattern. Go back through    /* Reached end of expression, either ')' or end of pattern. In the real
5939    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
5940    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