/[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 149 by ph10, Mon Apr 16 15:28:08 2007 UTC revision 487 by ph10, Wed Jan 6 10:26:55 2010 UTC
# Line 6  Line 6 
6  and semantics are as close as possible to those of the Perl 5 language.  and semantics are as close as possible to those of the Perl 5 language.
7    
8                         Written by Philip Hazel                         Written by Philip Hazel
9             Copyright (c) 1997-2007 University of Cambridge             Copyright (c) 1997-2010 University of Cambridge
10    
11  -----------------------------------------------------------------------------  -----------------------------------------------------------------------------
12  Redistribution and use in source and binary forms, with or without  Redistribution and use in source and binary forms, with or without
# Line 42  POSSIBILITY OF SUCH DAMAGE. Line 42  POSSIBILITY OF SUCH DAMAGE.
42  supporting internal functions that are not used by other modules. */  supporting internal functions that are not used by other modules. */
43    
44    
45    #ifdef HAVE_CONFIG_H
46    #include "config.h"
47    #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
50  #define PSSTART start_pattern  /* Field containing processed string start */  #define PSSTART start_pattern  /* Field containing processed string start */
51  #define PSEND   end_pattern    /* Field containing processed string end */  #define PSEND   end_pattern    /* Field containing processed string end */
52    
   
53  #include "pcre_internal.h"  #include "pcre_internal.h"
54    
55    
56  /* When DEBUG is defined, we need the pcre_printint() function, which is also  /* When PCRE_DEBUG is defined, we need the pcre_printint() function, which is
57  used by pcretest. DEBUG is not defined when building a production library. */  also used by pcretest. PCRE_DEBUG is not defined when building a production
58    library. */
59    
60  #ifdef DEBUG  #ifdef PCRE_DEBUG
61  #include "pcre_printint.src"  #include "pcre_printint.src"
62  #endif  #endif
63    
64    
65    /* Macro for setting individual bits in class bitmaps. */
66    
67    #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
79  *************************************************/  *************************************************/
# Line 82  are simple data values; negative values Line 98  are simple data values; negative values
98  on. Zero means further processing is needed (for things like \x), or the escape  on. Zero means further processing is needed (for things like \x), or the escape
99  is invalid. */  is invalid. */
100    
101  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
102    
103    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
104    in UTF-8 mode. */
105    
106  static const short int escapes[] = {  static const short int escapes[] = {
107       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
108       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
109     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
110       0,      0,      0,      0,      0,      0,      0,      0,   /* H - O */       0,                       0,
111  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0,      0, -ESC_W,   /* P - W */       0,                       0,
112  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
113     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
114       0,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
115  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0,      0, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
116       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
117         -ESC_D,                  -ESC_E,
118         0,                       -ESC_G,
119         -ESC_H,                  0,
120         0,                       -ESC_K,
121         0,                       0,
122         0,                       0,
123         -ESC_P,                  -ESC_Q,
124         -ESC_R,                  -ESC_S,
125         0,                       0,
126         -ESC_V,                  -ESC_W,
127         -ESC_X,                  0,
128         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
129         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
130         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
131         CHAR_GRAVE_ACCENT,       7,
132         -ESC_b,                  0,
133         -ESC_d,                  ESC_e,
134         ESC_f,                   0,
135         -ESC_h,                  0,
136         0,                       -ESC_k,
137         0,                       0,
138         ESC_n,                   0,
139         -ESC_p,                  0,
140         ESC_r,                   -ESC_s,
141         ESC_tee,                 0,
142         -ESC_v,                  -ESC_w,
143         0,                       0,
144         -ESC_z
145  };  };
146    
147  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
148    
149    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
150    
151  static const short int escapes[] = {  static const short int escapes[] = {
152  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
153  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 106  static const short int escapes[] = { Line 157  static const short int escapes[] = {
157  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  70 */     0,     0,      0,       0,      0,     0,      0,      0,
158  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',  /*  78 */     0,   '`',    ':',     '#',    '@',  '\'',    '=',    '"',
159  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,  /*  80 */     0,     7, -ESC_b,       0, -ESC_d, ESC_e,  ESC_f,      0,
160  /*  88 */     0,     0,      0,     '{',      0,     0,      0,      0,  /*  88 */-ESC_h,     0,      0,     '{',      0,     0,      0,      0,
161  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,  /*  90 */     0,     0, -ESC_k,     'l',      0, ESC_n,      0, -ESC_p,
162  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,  /*  98 */     0, ESC_r,      0,     '}',      0,     0,      0,      0,
163  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,     0, -ESC_w,      0,  /*  A0 */     0,   '~', -ESC_s, ESC_tee,      0,-ESC_v, -ESC_w,      0,
164  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,  /*  A8 */     0,-ESC_z,      0,       0,      0,   '[',      0,      0,
165  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  B0 */     0,     0,      0,       0,      0,     0,      0,      0,
166  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
167  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,  /*  C0 */   '{',-ESC_A, -ESC_B,  -ESC_C, -ESC_D,-ESC_E,      0, -ESC_G,
168  /*  C8 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
169  /*  D0 */   '}',     0,      0,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,
170  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
171  /*  E0 */  '\\',     0, -ESC_S,       0,      0,     0, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
172  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
173  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,  /*  F0 */     0,     0,      0,       0,      0,     0,      0,      0,
174  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0  /*  F8 */     0,     0,      0,       0,      0,     0,      0,      0
# Line 125  static const short int escapes[] = { Line 176  static const short int escapes[] = {
176  #endif  #endif
177    
178    
179  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
180  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
181  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
182    string is built from string macros so that it works in UTF-8 mode on EBCDIC
183  static const char *const posix_names[] = {  platforms. */
184    "alpha", "lower", "upper",  
185    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
186    "print", "punct", "space", "word",  "xdigit" };    int   len;
187      int   op;
188    } verbitem;
189    
190    static const char verbnames[] =
191      STRING_ACCEPT0
192      STRING_COMMIT0
193      STRING_F0
194      STRING_FAIL0
195      STRING_PRUNE0
196      STRING_SKIP0
197      STRING_THEN;
198    
199    static const verbitem verbs[] = {
200      { 6, OP_ACCEPT },
201      { 6, OP_COMMIT },
202      { 1, OP_FAIL },
203      { 4, OP_FAIL },
204      { 5, OP_PRUNE },
205      { 4, OP_SKIP  },
206      { 4, OP_THEN  }
207    };
208    
209    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
210    
211    
212    /* Tables of names of POSIX character classes and their lengths. The names are
213    now all in a single string, to reduce the number of relocations when a shared
214    library is dynamically loaded. The list of lengths is terminated by a zero
215    length entry. The first three must be alpha, lower, upper, as this is assumed
216    for handling case independence. */
217    
218    static const char posix_names[] =
219      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
220      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
221      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
222      STRING_word0  STRING_xdigit;
223    
224  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
225    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 171  static const int posix_class_maps[] = { Line 258  static const int posix_class_maps[] = {
258  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
259  are passed to the outside world. Do not ever re-use any error number, because  are passed to the outside world. Do not ever re-use any error number, because
260  they are documented. Always add a new error instead. Messages marked DEAD below  they are documented. Always add a new error instead. Messages marked DEAD below
261  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
262    the number of relocations needed when a shared library is loaded dynamically,
263  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
264    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
265    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
266    "\\c at end of pattern",  because these strings are used only when there is a compilation error. */
267    "unrecognized character follows \\",  
268    "numbers out of order in {} quantifier",  static const char error_texts[] =
269      "no error\0"
270      "\\ at end of pattern\0"
271      "\\c at end of pattern\0"
272      "unrecognized character follows \\\0"
273      "numbers out of order in {} quantifier\0"
274    /* 5 */    /* 5 */
275    "number too big in {} quantifier",    "number too big in {} quantifier\0"
276    "missing terminating ] for character class",    "missing terminating ] for character class\0"
277    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
278    "range out of order in character class",    "range out of order in character class\0"
279    "nothing to repeat",    "nothing to repeat\0"
280    /* 10 */    /* 10 */
281    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
282    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
283    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
284    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
285    "missing )",    "missing )\0"
286    /* 15 */    /* 15 */
287    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
288    "erroffset passed as NULL",    "erroffset passed as NULL\0"
289    "unknown option bit(s) set",    "unknown option bit(s) set\0"
290    "missing ) after comment",    "missing ) after comment\0"
291    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
292    /* 20 */    /* 20 */
293    "regular expression too large",    "regular expression is too large\0"
294    "failed to get memory",    "failed to get memory\0"
295    "unmatched parentheses",    "unmatched parentheses\0"
296    "internal error: code overflow",    "internal error: code overflow\0"
297    "unrecognized character after (?<",    "unrecognized character after (?<\0"
298    /* 25 */    /* 25 */
299    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
300    "malformed number or name after (?(",    "malformed number or name after (?(\0"
301    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
302    "assertion expected after (?(",    "assertion expected after (?(\0"
303    "(?R or (?digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
304    /* 30 */    /* 30 */
305    "unknown POSIX class name",    "unknown POSIX class name\0"
306    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
307    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
308    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
309    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
310    /* 35 */    /* 35 */
311    "invalid condition (?(0)",    "invalid condition (?(0)\0"
312    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
313    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N, \\U, or \\u\0"
314    "number after (?C is > 255",    "number after (?C is > 255\0"
315    "closing ) for (?C expected",    "closing ) for (?C expected\0"
316    /* 40 */    /* 40 */
317    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
318    "unrecognized character after (?P",    "unrecognized character after (?P\0"
319    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
320    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
321    "invalid UTF-8 string",    "invalid UTF-8 string\0"
322    /* 45 */    /* 45 */
323    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
324    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
325    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
326    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
327    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
328    /* 50 */    /* 50 */
329    "repeated subpattern is too long",    "repeated subpattern is too long\0"    /** DEAD **/
330    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
331    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
332    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
333    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
334    /* 55 */    /* 55 */
335    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
336    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
337    "\\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"
338  };    "a numbered reference must not be zero\0"
339      "(*VERB) with an argument is not supported\0"
340      /* 60 */
341      "(*VERB) not recognized\0"
342      "number is too big\0"
343      "subpattern name expected\0"
344      "digit expected after (?+\0"
345      "] is an invalid data character in JavaScript compatibility mode\0"
346      /* 65 */
347      "different names for subpatterns of the same number are not allowed";
348    
349    
350  /* 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 363  For convenience, we use the same bit def
363    
364  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
365    
366  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
367    
368    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
369    UTF-8 mode. */
370    
371  static const unsigned char digitab[] =  static const unsigned char digitab[] =
372    {    {
373    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 403  static const unsigned char digitab[] =
403    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
404    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
405    
406  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
407    
408    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
409    
410  static const unsigned char digitab[] =  static const unsigned char digitab[] =
411    {    {
412    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 481  static const unsigned char ebcdic_charta
481  /* Definition to allow mutual recursion */  /* Definition to allow mutual recursion */
482    
483  static BOOL  static BOOL
484    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, int, int *,    compile_regex(int, int, uschar **, const uschar **, int *, BOOL, BOOL, int,
485      int *, branch_chain *, compile_data *, int *);      int *, int *, branch_chain *, compile_data *, int *);
486    
487    
488    
489  /*************************************************  /*************************************************
490    *            Find an error text                  *
491    *************************************************/
492    
493    /* The error texts are now all in one long string, to save on relocations. As
494    some of the text is of unknown length, we can't use a table of offsets.
495    Instead, just count through the strings. This is not a performance issue
496    because it happens only when there has been a compilation error.
497    
498    Argument:   the error number
499    Returns:    pointer to the error string
500    */
501    
502    static const char *
503    find_error_text(int n)
504    {
505    const char *s = error_texts;
506    for (; n > 0; n--) while (*s++ != 0) {};
507    return s;
508    }
509    
510    
511    /*************************************************
512  *            Handle escapes                      *  *            Handle escapes                      *
513  *************************************************/  *************************************************/
514    
# Line 399  Arguments: Line 529  Arguments:
529    
530  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
531                   negative => a special escape sequence                   negative => a special escape sequence
532                   on error, errorptr is set                   on error, errorcodeptr is set
533  */  */
534    
535  static int  static int
# Line 417  ptr--; /* Set Line 547  ptr--; /* Set
547    
548  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
549    
550  /* 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
551  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.
552  Otherwise further processing may be required. */  Otherwise further processing may be required. */
553    
554  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
555  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
556  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
557    
558  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
559  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
560  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
561  #endif  #endif
562    
# Line 442  else Line 572  else
572      /* 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
573      error. */      error. */
574    
575      case 'l':      case CHAR_l:
576      case 'L':      case CHAR_L:
577      case 'N':      case CHAR_N:
578      case 'u':      case CHAR_u:
579      case 'U':      case CHAR_U:
580      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
581      break;      break;
582    
583      /* \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:
584      is an absolute backreference. If negative, it is a relative backreference.  
585      This is a Perl 5.10 feature. */      (1) A number, either plain or braced. If positive, it is an absolute
586        backreference. If negative, it is a relative backreference. This is a Perl
587        5.10 feature.
588    
589        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
590        is part of Perl's movement towards a unified syntax for back references. As
591        this is synonymous with \k{name}, we fudge it up by pretending it really
592        was \k.
593    
594        (3) For Oniguruma compatibility we also support \g followed by a name or a
595        number either in angle brackets or in single quotes. However, these are
596        (possibly recursive) subroutine calls, _not_ backreferences. Just return
597        the -ESC_g code (cf \k). */
598    
599        case CHAR_g:
600        if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
601          {
602          c = -ESC_g;
603          break;
604          }
605    
606        /* Handle the Perl-compatible cases */
607    
608      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
609        {        {
610          const uschar *p;
611          for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
612            if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
613          if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
614            {
615            c = -ESC_k;
616            break;
617            }
618        braced = TRUE;        braced = TRUE;
619        ptr++;        ptr++;
620        }        }
621      else braced = FALSE;      else braced = FALSE;
622    
623      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
624        {        {
625        negated = TRUE;        negated = TRUE;
626        ptr++;        ptr++;
# Line 471  else Line 629  else
629    
630      c = 0;      c = 0;
631      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
632        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
633    
634      if (c == 0 || (braced && *(++ptr) != '}'))      if (c < 0)   /* Integer overflow */
635          {
636          *errorcodeptr = ERR61;
637          break;
638          }
639    
640        if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
641        {        {
642        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
643        return 0;        break;
644          }
645    
646        if (c == 0)
647          {
648          *errorcodeptr = ERR58;
649          break;
650        }        }
651    
652      if (negated)      if (negated)
# Line 484  else Line 654  else
654        if (c > bracount)        if (c > bracount)
655          {          {
656          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
657          return 0;          break;
658          }          }
659        c = bracount - (c - 1);        c = bracount - (c - 1);
660        }        }
# Line 504  else Line 674  else
674      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
675      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
676    
677      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:
678      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
679    
680      if (!isclass)      if (!isclass)
681        {        {
682        oldptr = ptr;        oldptr = ptr;
683        c -= '0';        c -= CHAR_0;
684        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
685          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
686          if (c < 0)    /* Integer overflow */
687            {
688            *errorcodeptr = ERR61;
689            break;
690            }
691        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
692          {          {
693          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 525  else Line 700  else
700      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.
701      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
702    
703      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
704        {        {
705        ptr--;        ptr--;
706        c = 0;        c = 0;
# Line 538  else Line 713  else
713      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
714      than 3 octal digits. */      than 3 octal digits. */
715    
716      case '0':      case CHAR_0:
717      c -= '0';      c -= CHAR_0;
718      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
719          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
720      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
721      break;      break;
722    
# Line 549  else Line 724  else
724      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
725      treated as a data character. */      treated as a data character. */
726    
727      case 'x':      case CHAR_x:
728      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
729        {        {
730        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
731        int count = 0;        int count = 0;
# Line 559  else Line 734  else
734        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
735          {          {
736          register int cc = *pt++;          register int cc = *pt++;
737          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
738          count++;          count++;
739    
740  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
741          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
742          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
743  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
744          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
745          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
746  #endif  #endif
747          }          }
748    
749        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
750          {          {
751          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
752          ptr = pt;          ptr = pt;
# Line 587  else Line 762  else
762      c = 0;      c = 0;
763      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
764        {        {
765        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
766        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
767  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
768        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
769        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
770  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
771        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
772        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
773  #endif  #endif
774        }        }
775      break;      break;
# Line 603  else Line 778  else
778      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
779      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
780    
781      case 'c':      case CHAR_c:
782      c = *(++ptr);      c = *(++ptr);
783      if (c == 0)      if (c == 0)
784        {        {
785        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
786        return 0;        break;
787        }        }
788    
789  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
790      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
791      c ^= 0x40;      c ^= 0x40;
792  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
793      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
794      c ^= 0xC0;      c ^= 0xC0;
795  #endif  #endif
796      break;      break;
797    
798      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
799      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
800      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
801      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
802      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
803    
804      default:      default:
805      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 677  if (c == 0) goto ERROR_RETURN; Line 852  if (c == 0) goto ERROR_RETURN;
852  /* \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
853  negation. */  negation. */
854    
855  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
856    {    {
857    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
858      {      {
859      *negptr = TRUE;      *negptr = TRUE;
860      ptr++;      ptr++;
861      }      }
862    for (i = 0; i < sizeof(name) - 1; i++)    for (i = 0; i < (int)sizeof(name) - 1; i++)
863      {      {
864      c = *(++ptr);      c = *(++ptr);
865      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
866      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
867      name[i] = c;      name[i] = c;
868      }      }
869    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
870    name[i] = 0;    name[i] = 0;
871    }    }
872    
# Line 713  top = _pcre_utt_size; Line 888  top = _pcre_utt_size;
888  while (bot < top)  while (bot < top)
889    {    {
890    i = (bot + top) >> 1;    i = (bot + top) >> 1;
891    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
892    if (c == 0)    if (c == 0)
893      {      {
894      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 756  is_counted_repeat(const uschar *p) Line 931  is_counted_repeat(const uschar *p)
931  {  {
932  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
933  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
934  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
935    
936  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
937  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
938    
939  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
940  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
941    
942  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
943  }  }
944    
945    
# Line 797  int max = -1; Line 972  int max = -1;
972  /* 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
973  an integer overflow. */  an integer overflow. */
974    
975  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
976  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
977    {    {
978    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 807  if (min < 0 || min > 65535) Line 982  if (min < 0 || min > 65535)
982  /* 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.
983  Also, max must not be less than min. */  Also, max must not be less than min. */
984    
985  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
986    {    {
987    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
988      {      {
989      max = 0;      max = 0;
990      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
991      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
992        {        {
993        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 837  return p; Line 1012  return p;
1012    
1013    
1014  /*************************************************  /*************************************************
1015  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1016  *************************************************/  *************************************************/
1017    
1018  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1019    top-level call starts at the beginning of the pattern. All other calls must
1020    start at a parenthesis. It scans along a pattern's text looking for capturing
1021  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
1022  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
1023  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
1024  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
1025  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1026    capturing group numbers - the (?| feature.
1027    
1028  Arguments:  Arguments:
1029    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1030    count        current count of capturing parens so far encountered    cd           compile background data
1031    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1032    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1033    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1034      count        pointer to the current capturing subpattern number (updated)
1035    
1036  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1037  */  */
1038    
1039  static int  static int
1040  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,
1041    BOOL xmode)    BOOL xmode, int *count)
1042  {  {
1043  const uschar *thisname;  uschar *ptr = *ptrptr;
1044    int start_count = *count;
1045    int hwm_count = start_count;
1046    BOOL dup_parens = FALSE;
1047    
1048  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1049    dealing with. The very first call may not start with a parenthesis. */
1050    
1051    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1052    {    {
1053    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1054          ptr[2] == CHAR_VERTICAL_LINE)
1055        {
1056        ptr += 3;
1057        dup_parens = TRUE;
1058        }
1059    
1060      /* Handle a normal, unnamed capturing parenthesis */
1061    
1062      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1063        {
1064        *count += 1;
1065        if (name == NULL && *count == lorn) return *count;
1066        ptr++;
1067        }
1068    
1069      /* Handle a condition. If it is an assertion, just carry on so that it
1070      is processed as normal. If not, skip to the closing parenthesis of the
1071      condition (there can't be any nested parens. */
1072    
1073      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1074        {
1075        ptr += 2;
1076        if (ptr[1] != CHAR_QUESTION_MARK)
1077          {
1078          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1079          if (*ptr != 0) ptr++;
1080          }
1081        }
1082    
1083      /* We have either (? or (* and not a condition */
1084    
1085      else
1086        {
1087        ptr += 2;
1088        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1089    
1090        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1091    
1092        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1093            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1094          {
1095          int term;
1096          const uschar *thisname;
1097          *count += 1;
1098          if (name == NULL && *count == lorn) return *count;
1099          term = *ptr++;
1100          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1101          thisname = ptr;
1102          while (*ptr != term) ptr++;
1103          if (name != NULL && lorn == ptr - thisname &&
1104              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1105            return *count;
1106          term++;
1107          }
1108        }
1109      }
1110    
1111    /* Past any initial parenthesis handling, scan for parentheses or vertical
1112    bars. */
1113    
1114    for (; *ptr != 0; ptr++)
1115      {
1116    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1117    
1118    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1119      {      {
1120      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1121      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1122        {        {
1123        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1124        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1125        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1126        }        }
1127      continue;      continue;
1128      }      }
1129    
1130    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1131      are handled for real. If the first character is '^', skip it. Also, if the
1132      first few characters (either before or after ^) are \Q\E or \E we skip them
1133      too. This makes for compatibility with Perl. Note the use of STR macros to
1134      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1135    
1136    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1137      {      {
1138      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1139        for (;;)
1140        {        {
1141        if (*ptr == '\\')        if (ptr[1] == CHAR_BACKSLASH)
1142          {          {
1143          if (*(++ptr) == 0) return -1;          if (ptr[2] == CHAR_E)
1144          if (*ptr == 'Q') for (;;)            ptr+= 2;
1145            else if (strncmp((const char *)ptr+2,
1146                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1147              ptr += 4;
1148            else
1149              break;
1150            }
1151          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1152            {
1153            negate_class = TRUE;
1154            ptr++;
1155            }
1156          else break;
1157          }
1158    
1159        /* If the next character is ']', it is a data character that must be
1160        skipped, except in JavaScript compatibility mode. */
1161    
1162        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1163            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1164          ptr++;
1165    
1166        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1167          {
1168          if (*ptr == 0) return -1;
1169          if (*ptr == CHAR_BACKSLASH)
1170            {
1171            if (*(++ptr) == 0) goto FAIL_EXIT;
1172            if (*ptr == CHAR_Q) for (;;)
1173            {            {
1174            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1175            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1176            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1177            }            }
1178          continue;          continue;
1179          }          }
# Line 904  for (; *ptr != 0; ptr++) Line 1183  for (; *ptr != 0; ptr++)
1183    
1184    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1185    
1186    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1187      {      {
1188      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1189      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1190      continue;      continue;
1191      }      }
1192    
1193    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1194    
1195    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1196      {      {
1197      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1198      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1199      continue;      if (*ptr == 0) goto FAIL_EXIT;
1200        }
1201    
1202      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1203        {
1204        if (dup_parens && *count < hwm_count) *count = hwm_count;
1205        *ptrptr = ptr;
1206        return -1;
1207      }      }
1208    
1209    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1210    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1211        if (*count > hwm_count) hwm_count = *count;
1212        *count = start_count;
1213        }
1214      }
1215    
1216    /* We have to disambiguate (?<! and (?<= from (?<name> */  FAIL_EXIT:
1217    *ptrptr = ptr;
1218    return -1;
1219    }
1220    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1221    
   count++;  
1222    
1223    if (name == NULL && count == lorn) return count;  
1224    term = *ptr++;  /*************************************************
1225    if (term == '<') term = '>';  *       Find forward referenced subpattern       *
1226    thisname = ptr;  *************************************************/
1227    while (*ptr != term) ptr++;  
1228    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1229        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1230      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1231    returns when it reaches a given numbered subpattern. This is used for forward
1232    references to subpatterns. We used to be able to start this scan from the
1233    current compiling point, using the current count value from cd->bracount, and
1234    do it all in a single loop, but the addition of the possibility of duplicate
1235    subpattern numbers means that we have to scan from the very start, in order to
1236    take account of such duplicates, and to use a recursive function to keep track
1237    of the different types of group.
1238    
1239    Arguments:
1240      cd           compile background data
1241      name         name to seek, or NULL if seeking a numbered subpattern
1242      lorn         name length, or subpattern number if name is NULL
1243      xmode        TRUE if we are in /x mode
1244    
1245    Returns:       the number of the found subpattern, or -1 if not found
1246    */
1247    
1248    static int
1249    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1250    {
1251    uschar *ptr = (uschar *)cd->start_pattern;
1252    int count = 0;
1253    int rc;
1254    
1255    /* If the pattern does not start with an opening parenthesis, the first call
1256    to find_parens_sub() will scan right to the end (if necessary). However, if it
1257    does start with a parenthesis, find_parens_sub() will return when it hits the
1258    matching closing parens. That is why we have to have a loop. */
1259    
1260    for (;;)
1261      {
1262      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1263      if (rc > 0 || *ptr++ == 0) break;
1264    }    }
1265    
1266  return -1;  return rc;
1267  }  }
1268    
1269    
1270    
1271    
1272  /*************************************************  /*************************************************
1273  *      Find first significant op code            *  *      Find first significant op code            *
1274  *************************************************/  *************************************************/
# Line 996  for (;;) Line 1318  for (;;)
1318    
1319      case OP_CALLOUT:      case OP_CALLOUT:
1320      case OP_CREF:      case OP_CREF:
1321        case OP_NCREF:
1322      case OP_RREF:      case OP_RREF:
1323        case OP_NRREF:
1324      case OP_DEF:      case OP_DEF:
1325      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1326      break;      break;
# Line 1012  for (;;) Line 1336  for (;;)
1336    
1337    
1338  /*************************************************  /*************************************************
1339  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1340  *************************************************/  *************************************************/
1341    
1342  /* 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,
1343  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.
1344  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
1345    temporarily terminated with OP_END when this function is called.
1346    
1347    This function is called when a backward assertion is encountered, so that if it
1348    fails, the error message can point to the correct place in the pattern.
1349    However, we cannot do this when the assertion contains subroutine calls,
1350    because they can be forward references. We solve this by remembering this case
1351    and doing the check at the end; a flag specifies which mode we are running in.
1352    
1353  Arguments:  Arguments:
1354    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1355    options  the compiling options    options  the compiling options
1356      atend    TRUE if called when the pattern is complete
1357      cd       the "compile data" structure
1358    
1359  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1360                 or -1 if there is no fixed length,
1361               or -2 if \C was encountered               or -2 if \C was encountered
1362                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1363  */  */
1364    
1365  static int  static int
1366  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1367  {  {
1368  int length = -1;  int length = -1;
1369    
# Line 1041  branch, check the length against that of Line 1376  branch, check the length against that of
1376  for (;;)  for (;;)
1377    {    {
1378    int d;    int d;
1379      uschar *ce, *cs;
1380    register int op = *cc;    register int op = *cc;
   
1381    switch (op)    switch (op)
1382      {      {
1383      case OP_CBRA:      case OP_CBRA:
1384      case OP_BRA:      case OP_BRA:
1385      case OP_ONCE:      case OP_ONCE:
1386      case OP_COND:      case OP_COND:
1387      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1388      if (d < 0) return d;      if (d < 0) return d;
1389      branchlength += d;      branchlength += d;
1390      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1072  for (;;) Line 1407  for (;;)
1407      branchlength = 0;      branchlength = 0;
1408      break;      break;
1409    
1410        /* A true recursion implies not fixed length, but a subroutine call may
1411        be OK. If the subroutine is a forward reference, we can't deal with
1412        it until the end of the pattern, so return -3. */
1413    
1414        case OP_RECURSE:
1415        if (!atend) return -3;
1416        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1417        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1418        if (cc > cs && cc < ce) return -1;                /* Recursion */
1419        d = find_fixedlength(cs + 2, options, atend, cd);
1420        if (d < 0) return d;
1421        branchlength += d;
1422        cc += 1 + LINK_SIZE;
1423        break;
1424    
1425      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1426    
1427      case OP_ASSERT:      case OP_ASSERT:
# Line 1085  for (;;) Line 1435  for (;;)
1435    
1436      case OP_REVERSE:      case OP_REVERSE:
1437      case OP_CREF:      case OP_CREF:
1438        case OP_NCREF:
1439      case OP_RREF:      case OP_RREF:
1440        case OP_NRREF:
1441      case OP_DEF:      case OP_DEF:
1442      case OP_OPT:      case OP_OPT:
1443      case OP_CALLOUT:      case OP_CALLOUT:
# Line 1108  for (;;) Line 1460  for (;;)
1460      branchlength++;      branchlength++;
1461      cc += 2;      cc += 2;
1462  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1463      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1464        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1465  #endif  #endif
1466      break;      break;
1467    
# Line 1122  for (;;) Line 1472  for (;;)
1472      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1473      cc += 4;      cc += 4;
1474  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1475      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1476        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1477  #endif  #endif
1478      break;      break;
1479    
1480      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1481      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1482        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1483      cc += 4;      cc += 4;
1484      break;      break;
1485    
# Line 1148  for (;;) Line 1497  for (;;)
1497      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1498      case OP_WORDCHAR:      case OP_WORDCHAR:
1499      case OP_ANY:      case OP_ANY:
1500        case OP_ALLANY:
1501      branchlength++;      branchlength++;
1502      cc++;      cc++;
1503      break;      break;
# Line 1202  for (;;) Line 1552  for (;;)
1552    
1553    
1554  /*************************************************  /*************************************************
1555  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1556  *************************************************/  *************************************************/
1557    
1558  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1559  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1560    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1561    so that it can be called from pcre_study() when finding the minimum matching
1562    length.
1563    
1564  Arguments:  Arguments:
1565    code        points to start of expression    code        points to start of expression
1566    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1567    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1568    
1569  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
1570  */  */
1571    
1572  static const uschar *  const uschar *
1573  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1574  {  {
1575  for (;;)  for (;;)
1576    {    {
# Line 1230  for (;;) Line 1583  for (;;)
1583    
1584    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1585    
1586      /* Handle recursion */
1587    
1588      else if (c == OP_REVERSE)
1589        {
1590        if (number < 0) return (uschar *)code;
1591        code += _pcre_OP_lengths[c];
1592        }
1593    
1594    /* Handle capturing bracket */    /* Handle capturing bracket */
1595    
1596    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1239  for (;;) Line 1600  for (;;)
1600      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1601      }      }
1602    
1603    /* 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
1604    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
1605    arrange to skip the extra bytes. */    two bytes of parameters. */
1606    
1607    else    else
1608      {      {
1609        switch(c)
1610          {
1611          case OP_TYPESTAR:
1612          case OP_TYPEMINSTAR:
1613          case OP_TYPEPLUS:
1614          case OP_TYPEMINPLUS:
1615          case OP_TYPEQUERY:
1616          case OP_TYPEMINQUERY:
1617          case OP_TYPEPOSSTAR:
1618          case OP_TYPEPOSPLUS:
1619          case OP_TYPEPOSQUERY:
1620          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1621          break;
1622    
1623          case OP_TYPEUPTO:
1624          case OP_TYPEMINUPTO:
1625          case OP_TYPEEXACT:
1626          case OP_TYPEPOSUPTO:
1627          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1628          break;
1629          }
1630    
1631        /* Add in the fixed length from the table */
1632    
1633      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1634    
1635      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1636      a multi-byte character. The length in the table is a minimum, so we have to
1637      arrange to skip the extra bytes. */
1638    
1639  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1640      if (utf8) switch(c)      if (utf8) switch(c)
1641        {        {
# Line 1267  for (;;) Line 1657  for (;;)
1657        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1658        break;        break;
1659        }        }
1660    #else
1661        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1662  #endif  #endif
1663      }      }
1664    }    }
# Line 1303  for (;;) Line 1695  for (;;)
1695    
1696    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1697    
1698    /* 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
1699    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
1700    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters. */
   bytes. */  
1701    
1702    else    else
1703      {      {
1704        switch(c)
1705          {
1706          case OP_TYPESTAR:
1707          case OP_TYPEMINSTAR:
1708          case OP_TYPEPLUS:
1709          case OP_TYPEMINPLUS:
1710          case OP_TYPEQUERY:
1711          case OP_TYPEMINQUERY:
1712          case OP_TYPEPOSSTAR:
1713          case OP_TYPEPOSPLUS:
1714          case OP_TYPEPOSQUERY:
1715          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1716          break;
1717    
1718          case OP_TYPEPOSUPTO:
1719          case OP_TYPEUPTO:
1720          case OP_TYPEMINUPTO:
1721          case OP_TYPEEXACT:
1722          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1723          break;
1724          }
1725    
1726        /* Add in the fixed length from the table */
1727    
1728      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1729    
1730        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1731        by a multi-byte character. The length in the table is a minimum, so we have
1732        to arrange to skip the extra bytes. */
1733    
1734  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1735      if (utf8) switch(c)      if (utf8) switch(c)
1736        {        {
# Line 1332  for (;;) Line 1752  for (;;)
1752        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1753        break;        break;
1754        }        }
1755    #else
1756        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1757  #endif  #endif
1758      }      }
1759    }    }
# Line 1347  for (;;) Line 1769  for (;;)
1769  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()
1770  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
1771  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1772  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
1773  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
1774    bracket whose current branch will already have been scanned.
1775    
1776  Arguments:  Arguments:
1777    code        points to start of search    code        points to start of search
# Line 1370  for (code = first_significant_code(code Line 1793  for (code = first_significant_code(code
1793    
1794    c = *code;    c = *code;
1795    
1796    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    /* Skip over forward assertions; the other assertions are skipped by
1797      first_significant_code() with a TRUE final argument. */
1798    
1799      if (c == OP_ASSERT)
1800        {
1801        do code += GET(code, 1); while (*code == OP_ALT);
1802        c = *code;
1803        continue;
1804        }
1805    
1806      /* Groups with zero repeats can of course be empty; skip them. */
1807    
1808      if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1809        {
1810        code += _pcre_OP_lengths[c];
1811        do code += GET(code, 1); while (*code == OP_ALT);
1812        c = *code;
1813        continue;
1814        }
1815    
1816      /* For other groups, scan the branches. */
1817    
1818      if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1819      {      {
1820      BOOL empty_branch;      BOOL empty_branch;
1821      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1822    
1823      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1824        empty branch, so just skip over the conditional, because it could be empty.
1825        Otherwise, scan the individual branches of the group. */
1826    
1827      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;  
1828        code += GET(code, 1);        code += GET(code, 1);
1829        else
1830          {
1831          empty_branch = FALSE;
1832          do
1833            {
1834            if (!empty_branch && could_be_empty_branch(code, endcode, utf8))
1835              empty_branch = TRUE;
1836            code += GET(code, 1);
1837            }
1838          while (*code == OP_ALT);
1839          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1840        }        }
     while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
   
     /* Move past the KET and fudge things so that the increment in the "for"  
     above has no effect. */  
1841    
1842      c = OP_END;      c = *code;
     code += 1 + LINK_SIZE - _pcre_OP_lengths[c];  
1843      continue;      continue;
1844      }      }
1845    
# Line 1399  for (code = first_significant_code(code Line 1847  for (code = first_significant_code(code
1847    
1848    switch (c)    switch (c)
1849      {      {
1850      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1851        cannot be represented just by a bit map. This includes negated single
1852        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1853        actual length is stored in the compiled code, so we must update "code"
1854        here. */
1855    
1856  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1857      case OP_XCLASS:      case OP_XCLASS:
1858      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1859      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1860  #endif  #endif
1861    
# Line 1447  for (code = first_significant_code(code Line 1899  for (code = first_significant_code(code
1899      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1900      case OP_WORDCHAR:      case OP_WORDCHAR:
1901      case OP_ANY:      case OP_ANY:
1902        case OP_ALLANY:
1903      case OP_ANYBYTE:      case OP_ANYBYTE:
1904      case OP_CHAR:      case OP_CHAR:
1905      case OP_CHARNC:      case OP_CHARNC:
# Line 1465  for (code = first_significant_code(code Line 1918  for (code = first_significant_code(code
1918      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1919      return FALSE;      return FALSE;
1920    
1921        /* These are going to continue, as they may be empty, but we have to
1922        fudge the length for the \p and \P cases. */
1923    
1924        case OP_TYPESTAR:
1925        case OP_TYPEMINSTAR:
1926        case OP_TYPEPOSSTAR:
1927        case OP_TYPEQUERY:
1928        case OP_TYPEMINQUERY:
1929        case OP_TYPEPOSQUERY:
1930        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1931        break;
1932    
1933        /* Same for these */
1934    
1935        case OP_TYPEUPTO:
1936        case OP_TYPEMINUPTO:
1937        case OP_TYPEPOSUPTO:
1938        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1939        break;
1940    
1941      /* End of branch */      /* End of branch */
1942    
1943      case OP_KET:      case OP_KET:
# Line 1483  for (code = first_significant_code(code Line 1956  for (code = first_significant_code(code
1956      case OP_QUERY:      case OP_QUERY:
1957      case OP_MINQUERY:      case OP_MINQUERY:
1958      case OP_POSQUERY:      case OP_POSQUERY:
1959        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
1960        break;
1961    
1962      case OP_UPTO:      case OP_UPTO:
1963      case OP_MINUPTO:      case OP_MINUPTO:
1964      case OP_POSUPTO:      case OP_POSUPTO:
1965      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
1966      break;      break;
1967  #endif  #endif
1968      }      }
# Line 1519  static BOOL Line 1995  static BOOL
1995  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
1996    BOOL utf8)    BOOL utf8)
1997  {  {
1998  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
1999    {    {
2000    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8))
2001        return FALSE;
2002    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2003    }    }
2004  return TRUE;  return TRUE;
# Line 1534  return TRUE; Line 2011  return TRUE;
2011  *************************************************/  *************************************************/
2012    
2013  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2014  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
2015  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2016  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2017    
2018    Originally, this function only recognized a sequence of letters between the
2019    terminators, but it seems that Perl recognizes any sequence of characters,
2020    though of course unknown POSIX names are subsequently rejected. Perl gives an
2021    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2022    didn't consider this to be a POSIX class. Likewise for [:1234:].
2023    
2024    The problem in trying to be exactly like Perl is in the handling of escapes. We
2025    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2026    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2027    below handles the special case of \], but does not try to do any other escape
2028    processing. This makes it different from Perl for cases such as [:l\ower:]
2029    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2030    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2031    I think.
2032    
2033  Argument:  Arguments:
2034    ptr      pointer to the initial [    ptr      pointer to the initial [
2035    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2036    
2037  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2038  */  */
2039    
2040  static BOOL  static BOOL
2041  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2042  {  {
2043  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2044  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2045  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2046    {    {
2047    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2048    return TRUE;      {
2049        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2050        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2051          {
2052          *endptr = ptr;
2053          return TRUE;
2054          }
2055        }
2056    }    }
2057  return FALSE;  return FALSE;
2058  }  }
# Line 1581  Returns: a value representing the na Line 2077  Returns: a value representing the na
2077  static int  static int
2078  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2079  {  {
2080    const char *pn = posix_names;
2081  register int yield = 0;  register int yield = 0;
2082  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2083    {    {
2084    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2085      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2086      pn += posix_name_lengths[yield] + 1;
2087    yield++;    yield++;
2088    }    }
2089  return -1;  return -1;
# Line 1600  return -1; Line 2098  return -1;
2098  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2099  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2100  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
2101  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
2102  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
2103  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
2104  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
2105  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2106    OP_END.
2107    
2108  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2109  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 2126  adjust_recurse(uschar *group, int adjust
2126    uschar *save_hwm)    uschar *save_hwm)
2127  {  {
2128  uschar *ptr = group;  uschar *ptr = group;
2129    
2130  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2131    {    {
2132    int offset;    int offset;
# Line 1738  get_othercase_range(unsigned int *cptr, Line 2238  get_othercase_range(unsigned int *cptr,
2238  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2239    
2240  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2241    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2242    
2243  if (c > d) return FALSE;  if (c > d) return FALSE;
2244    
# Line 1747  next = othercase + 1; Line 2247  next = othercase + 1;
2247    
2248  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2249    {    {
2250    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2251    next++;    next++;
2252    }    }
2253    
# Line 1793  if ((options & PCRE_EXTENDED) != 0) Line 2293  if ((options & PCRE_EXTENDED) != 0)
2293    for (;;)    for (;;)
2294      {      {
2295      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2296      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2297        {        {
2298        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2299          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1805  if ((options & PCRE_EXTENDED) != 0) Line 2305  if ((options & PCRE_EXTENDED) != 0)
2305  /* 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
2306  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2307    
2308  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2309    {    {
2310    int temperrorcode = 0;    int temperrorcode = 0;
2311    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 2330  if ((options & PCRE_EXTENDED) != 0)
2330    for (;;)    for (;;)
2331      {      {
2332      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2333      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2334        {        {
2335        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2336          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1841  if ((options & PCRE_EXTENDED) != 0) Line 2341  if ((options & PCRE_EXTENDED) != 0)
2341    
2342  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2343    
2344  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2345    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2346        return FALSE;
2347    
2348  /* 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
2349  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 2358  if (next >= 0) switch(op_code)
2358    case OP_CHAR:    case OP_CHAR:
2359  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2360    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }
2361    #else
2362      (void)(utf8_char);  /* Keep compiler happy by referencing function argument */
2363  #endif  #endif
2364    return item != next;    return item != next;
2365    
# Line 1875  if (next >= 0) switch(op_code) Line 2378  if (next >= 0) switch(op_code)
2378      unsigned int othercase;      unsigned int othercase;
2379      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2380  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2381      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2382  #else  #else
2383      othercase = NOTACHAR;      othercase = NOTACHAR;
2384  #endif  #endif
# Line 1888  if (next >= 0) switch(op_code) Line 2391  if (next >= 0) switch(op_code)
2391    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, "item" must be a single-byte character. */
2392    
2393    case OP_NOT:    case OP_NOT:
   if (next < 0) return FALSE;  /* Not a character */  
2394    if (item == next) return TRUE;    if (item == next) return TRUE;
2395    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2396  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
# Line 1897  if (next >= 0) switch(op_code) Line 2399  if (next >= 0) switch(op_code)
2399      unsigned int othercase;      unsigned int othercase;
2400      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2401  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2402      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2403  #else  #else
2404      othercase = NOTACHAR;      othercase = NOTACHAR;
2405  #endif  #endif
# Line 1925  if (next >= 0) switch(op_code) Line 2427  if (next >= 0) switch(op_code)
2427    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2428    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;    return next <= 127 && (cd->ctypes[next] & ctype_word) != 0;
2429    
2430      case OP_HSPACE:
2431      case OP_NOT_HSPACE:
2432      switch(next)
2433        {
2434        case 0x09:
2435        case 0x20:
2436        case 0xa0:
2437        case 0x1680:
2438        case 0x180e:
2439        case 0x2000:
2440        case 0x2001:
2441        case 0x2002:
2442        case 0x2003:
2443        case 0x2004:
2444        case 0x2005:
2445        case 0x2006:
2446        case 0x2007:
2447        case 0x2008:
2448        case 0x2009:
2449        case 0x200A:
2450        case 0x202f:
2451        case 0x205f:
2452        case 0x3000:
2453        return op_code != OP_HSPACE;
2454        default:
2455        return op_code == OP_HSPACE;
2456        }
2457    
2458      case OP_VSPACE:
2459      case OP_NOT_VSPACE:
2460      switch(next)
2461        {
2462        case 0x0a:
2463        case 0x0b:
2464        case 0x0c:
2465        case 0x0d:
2466        case 0x85:
2467        case 0x2028:
2468        case 0x2029:
2469        return op_code != OP_VSPACE;
2470        default:
2471        return op_code == OP_VSPACE;
2472        }
2473    
2474    default:    default:
2475    return FALSE;    return FALSE;
2476    }    }
# Line 1959  switch(op_code) Line 2505  switch(op_code)
2505      case ESC_W:      case ESC_W:
2506      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;
2507    
2508        case ESC_h:
2509        case ESC_H:
2510        switch(item)
2511          {
2512          case 0x09:
2513          case 0x20:
2514          case 0xa0:
2515          case 0x1680:
2516          case 0x180e:
2517          case 0x2000:
2518          case 0x2001:
2519          case 0x2002:
2520          case 0x2003:
2521          case 0x2004:
2522          case 0x2005:
2523          case 0x2006:
2524          case 0x2007:
2525          case 0x2008:
2526          case 0x2009:
2527          case 0x200A:
2528          case 0x202f:
2529          case 0x205f:
2530          case 0x3000:
2531          return -next != ESC_h;
2532          default:
2533          return -next == ESC_h;
2534          }
2535    
2536        case ESC_v:
2537        case ESC_V:
2538        switch(item)
2539          {
2540          case 0x0a:
2541          case 0x0b:
2542          case 0x0c:
2543          case 0x0d:
2544          case 0x85:
2545          case 0x2028:
2546          case 0x2029:
2547          return -next != ESC_v;
2548          default:
2549          return -next == ESC_v;
2550          }
2551    
2552      default:      default:
2553      return FALSE;      return FALSE;
2554      }      }
2555    
2556    case OP_DIGIT:    case OP_DIGIT:
2557    return next == -ESC_D || next == -ESC_s || next == -ESC_W;    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2558             next == -ESC_h || next == -ESC_v;
2559    
2560    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2561    return next == -ESC_d;    return next == -ESC_d;
# Line 1973  switch(op_code) Line 2564  switch(op_code)
2564    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w;
2565    
2566    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2567    return next == -ESC_s;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2568    
2569      case OP_HSPACE:
2570      return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;
2571    
2572      case OP_NOT_HSPACE:
2573      return next == -ESC_h;
2574    
2575      /* Can't have \S in here because VT matches \S (Perl anomaly) */
2576      case OP_VSPACE:
2577      return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2578    
2579      case OP_NOT_VSPACE:
2580      return next == -ESC_v;
2581    
2582    case OP_WORDCHAR:    case OP_WORDCHAR:
2583    return next == -ESC_W || next == -ESC_s;    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;
2584    
2585    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2586    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2049  uschar classbits[32]; Line 2653  uschar classbits[32];
2653  BOOL class_utf8;  BOOL class_utf8;
2654  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2655  uschar *class_utf8data;  uschar *class_utf8data;
2656    uschar *class_utf8data_base;
2657  uschar utf8_char[6];  uschar utf8_char[6];
2658  #else  #else
2659  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2660  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2661  #endif  #endif
2662    
2663  #ifdef DEBUG  #ifdef PCRE_DEBUG
2664  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2665  #endif  #endif
2666    
# Line 2088  req_caseopt = ((options & PCRE_CASELESS) Line 2693  req_caseopt = ((options & PCRE_CASELESS)
2693  for (;; ptr++)  for (;; ptr++)
2694    {    {
2695    BOOL negate_class;    BOOL negate_class;
2696      BOOL should_flip_negation;
2697    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2698    BOOL is_quantifier;    BOOL is_quantifier;
2699    BOOL is_recurse;    BOOL is_recurse;
2700      BOOL reset_bracount;
2701    int class_charcount;    int class_charcount;
2702    int class_lastchar;    int class_lastchar;
2703    int newoptions;    int newoptions;
2704    int recno;    int recno;
2705      int refsign;
2706    int skipbytes;    int skipbytes;
2707    int subreqbyte;    int subreqbyte;
2708    int subfirstbyte;    int subfirstbyte;
# Line 2111  for (;; ptr++) Line 2719  for (;; ptr++)
2719    
2720    if (lengthptr != NULL)    if (lengthptr != NULL)
2721      {      {
2722  #ifdef DEBUG  #ifdef PCRE_DEBUG
2723      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
2724  #endif  #endif
2725      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */
# Line 2127  for (;; ptr++) Line 2735  for (;; ptr++)
2735      */      */
2736    
2737      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
2738    
2739        /* Paranoid check for integer overflow */
2740    
2741        if (OFLOW_MAX - *lengthptr < code - last_code)
2742          {
2743          *errorcodeptr = ERR20;
2744          goto FAILED;
2745          }
2746    
2747      *lengthptr += code - last_code;      *lengthptr += code - last_code;
2748      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));
2749    
# Line 2164  for (;; ptr++) Line 2781  for (;; ptr++)
2781    
2782    if (inescq && c != 0)    if (inescq && c != 0)
2783      {      {
2784      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
2785        {        {
2786        inescq = FALSE;        inescq = FALSE;
2787        ptr++;        ptr++;
# Line 2190  for (;; ptr++) Line 2807  for (;; ptr++)
2807    /* 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
2808    a quantifier. */    a quantifier. */
2809    
2810    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
2811      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
2812        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
2813    
2814    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
2815         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2206  for (;; ptr++) Line 2824  for (;; ptr++)
2824    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
2825      {      {
2826      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
2827      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
2828        {        {
2829        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2830          {          {
# Line 2231  for (;; ptr++) Line 2849  for (;; ptr++)
2849      {      {
2850      /* ===================================================================*/      /* ===================================================================*/
2851      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
2852      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
2853      case ')':      case CHAR_RIGHT_PARENTHESIS:
2854      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
2855      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
2856      *codeptr = code;      *codeptr = code;
2857      *ptrptr = ptr;      *ptrptr = ptr;
2858      if (lengthptr != NULL)      if (lengthptr != NULL)
2859        {        {
2860          if (OFLOW_MAX - *lengthptr < code - last_code)
2861            {
2862            *errorcodeptr = ERR20;
2863            goto FAILED;
2864            }
2865        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += code - last_code;   /* To include callout length */
2866        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
2867        }        }
# Line 2249  for (;; ptr++) Line 2872  for (;; ptr++)
2872      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
2873      the setting of any following char as a first character. */      the setting of any following char as a first character. */
2874    
2875      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
2876      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
2877        {        {
2878        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2258  for (;; ptr++) Line 2881  for (;; ptr++)
2881      *code++ = OP_CIRC;      *code++ = OP_CIRC;
2882      break;      break;
2883    
2884      case '$':      case CHAR_DOLLAR_SIGN:
2885      previous = NULL;      previous = NULL;
2886      *code++ = OP_DOLL;      *code++ = OP_DOLL;
2887      break;      break;
# Line 2266  for (;; ptr++) Line 2889  for (;; ptr++)
2889      /* 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
2890      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
2891    
2892      case '.':      case CHAR_DOT:
2893      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2894      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
2895      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
2896      previous = code;      previous = code;
2897      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
2898      break;      break;
2899    
2900    
# Line 2286  for (;; ptr++) Line 2909  for (;; ptr++)
2909      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,
2910      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
2911      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.
     */  
2912    
2913      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
2914        default (Perl) mode, it is treated as a data character. */
2915    
2916        case CHAR_RIGHT_SQUARE_BRACKET:
2917        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2918          {
2919          *errorcodeptr = ERR64;
2920          goto FAILED;
2921          }
2922        goto NORMAL_CHAR;
2923    
2924        case CHAR_LEFT_SQUARE_BRACKET:
2925      previous = code;      previous = code;
2926    
2927      /* 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
2928      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. */
2929    
2930      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
2931          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
2932            check_posix_syntax(ptr, &tempptr))
2933        {        {
2934        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
2935        goto FAILED;        goto FAILED;
2936        }        }
2937    
2938      /* 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,
2939        if the first few characters (either before or after ^) are \Q\E or \E we
2940        skip them too. This makes for compatibility with Perl. */
2941    
2942      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
2943        for (;;)
2944        {        {
       negate_class = TRUE;  
2945        c = *(++ptr);        c = *(++ptr);
2946          if (c == CHAR_BACKSLASH)
2947            {
2948            if (ptr[1] == CHAR_E)
2949              ptr++;
2950            else if (strncmp((const char *)ptr+1,
2951                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
2952              ptr += 3;
2953            else
2954              break;
2955            }
2956          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
2957            negate_class = TRUE;
2958          else break;
2959        }        }
2960      else  
2961        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
2962        an initial ']' is taken as a data character -- the code below handles
2963        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
2964        [^] must match any character, so generate OP_ALLANY. */
2965    
2966        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
2967            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
2968        {        {
2969        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
2970          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
2971          zerofirstbyte = firstbyte;
2972          break;
2973        }        }
2974    
2975        /* If a class contains a negative special such as \S, we need to flip the
2976        negation flag at the end, so that support for characters > 255 works
2977        correctly (they are all included in the class). */
2978    
2979        should_flip_negation = FALSE;
2980    
2981      /* 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
2982      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
2983      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 2995  for (;; ptr++)
2995  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2996      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
2997      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
2998        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
2999  #endif  #endif
3000    
3001      /* 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 3011  for (;; ptr++)
3011          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3012          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3013          }          }
3014    
3015          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3016          data and reset the pointer. This is so that very large classes that
3017          contain a zillion UTF-8 characters no longer overwrite the work space
3018          (which is on the stack). */
3019    
3020          if (lengthptr != NULL)
3021            {
3022            *lengthptr += class_utf8data - class_utf8data_base;
3023            class_utf8data = class_utf8data_base;
3024            }
3025    
3026  #endif  #endif
3027    
3028        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3029    
3030        if (inescq)        if (inescq)
3031          {          {
3032          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3033            {            {
3034            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3035            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2366  for (;; ptr++) Line 3044  for (;; ptr++)
3044        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3045        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3046    
3047        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3048            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3049            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3050          {          {
3051          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3052          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3053          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3054          uschar pbits[32];          uschar pbits[32];
3055    
3056          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3057            {            {
3058            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3059            goto FAILED;            goto FAILED;
3060            }            }
3061    
3062          ptr += 2;          ptr += 2;
3063          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3064            {            {
3065            local_negate = TRUE;            local_negate = TRUE;
3066              should_flip_negation = TRUE;  /* Note negative special */
3067            ptr++;            ptr++;
3068            }            }
3069    
# Line 2451  for (;; ptr++) Line 3130  for (;; ptr++)
3130        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
3131        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace.
3132        Elsewhere it marks a word boundary. Other escapes have preset maps ready        Elsewhere it marks a word boundary. Other escapes have preset maps ready
3133        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
3134        character in them, so set class_charcount bigger than one. */        character in them, so set class_charcount bigger than one. */
3135    
3136        if (c == '\\')        if (c == CHAR_BACKSLASH)
3137          {          {
3138          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3139          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3140    
3141          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 */
3142          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 */
3143          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 */
3144          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3145            {            {
3146            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3147              {              {
3148              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3149              }              }
3150            else inescq = TRUE;            else inescq = TRUE;
3151            continue;            continue;
3152            }            }
3153            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3154    
3155          if (c < 0)          if (c < 0)
3156            {            {
# Line 2486  for (;; ptr++) Line 3166  for (;; ptr++)
3166              continue;              continue;
3167    
3168              case ESC_D:              case ESC_D:
3169                should_flip_negation = TRUE;
3170              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3171              continue;              continue;
3172    
# Line 2494  for (;; ptr++) Line 3175  for (;; ptr++)
3175              continue;              continue;
3176    
3177              case ESC_W:              case ESC_W:
3178                should_flip_negation = TRUE;
3179              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3180              continue;              continue;
3181    
# Line 2503  for (;; ptr++) Line 3185  for (;; ptr++)
3185              continue;              continue;
3186    
3187              case ESC_S:              case ESC_S:
3188                should_flip_negation = TRUE;
3189              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3190              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3191              continue;              continue;
3192    
             case ESC_E: /* Perl ignores an orphan \E */  
             continue;  
   
3193              default:    /* Not recognized; fall through */              default:    /* Not recognized; fall through */
3194              break;      /* Need "default" setting to stop compiler warning. */              break;      /* Need "default" setting to stop compiler warning. */
3195              }              }
# Line 2519  for (;; ptr++) Line 3199  for (;; ptr++)
3199            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||            else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||
3200                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;                     c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;
3201    
3202              /* We need to deal with \H, \h, \V, and \v in both phases because
3203              they use extra memory. */
3204    
3205              if (-c == ESC_h)
3206                {
3207                SETBIT(classbits, 0x09); /* VT */
3208                SETBIT(classbits, 0x20); /* SPACE */
3209                SETBIT(classbits, 0xa0); /* NSBP */
3210    #ifdef SUPPORT_UTF8
3211                if (utf8)
3212                  {
3213                  class_utf8 = TRUE;
3214                  *class_utf8data++ = XCL_SINGLE;
3215                  class_utf8data += _pcre_ord2utf8(0x1680, class_utf8data);
3216                  *class_utf8data++ = XCL_SINGLE;
3217                  class_utf8data += _pcre_ord2utf8(0x180e, class_utf8data);
3218                  *class_utf8data++ = XCL_RANGE;
3219                  class_utf8data += _pcre_ord2utf8(0x2000, class_utf8data);
3220                  class_utf8data += _pcre_ord2utf8(0x200A, class_utf8data);
3221                  *class_utf8data++ = XCL_SINGLE;
3222                  class_utf8data += _pcre_ord2utf8(0x202f, class_utf8data);
3223                  *class_utf8data++ = XCL_SINGLE;
3224                  class_utf8data += _pcre_ord2utf8(0x205f, class_utf8data);
3225                  *class_utf8data++ = XCL_SINGLE;
3226                  class_utf8data += _pcre_ord2utf8(0x3000, class_utf8data);
3227                  }
3228    #endif
3229                continue;
3230                }
3231    
3232              if (-c == ESC_H)
3233                {
3234                for (c = 0; c < 32; c++)
3235                  {
3236                  int x = 0xff;
3237                  switch (c)
3238                    {
3239                    case 0x09/8: x ^= 1 << (0x09%8); break;
3240                    case 0x20/8: x ^= 1 << (0x20%8); break;
3241                    case 0xa0/8: x ^= 1 << (0xa0%8); break;
3242                    default: break;
3243                    }
3244                  classbits[c] |= x;
3245                  }
3246    
3247    #ifdef SUPPORT_UTF8
3248                if (utf8)
3249                  {
3250                  class_utf8 = TRUE;
3251                  *class_utf8data++ = XCL_RANGE;
3252                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3253                  class_utf8data += _pcre_ord2utf8(0x167f, class_utf8data);
3254                  *class_utf8data++ = XCL_RANGE;
3255                  class_utf8data += _pcre_ord2utf8(0x1681, class_utf8data);
3256                  class_utf8data += _pcre_ord2utf8(0x180d, class_utf8data);
3257                  *class_utf8data++ = XCL_RANGE;
3258                  class_utf8data += _pcre_ord2utf8(0x180f, class_utf8data);
3259                  class_utf8data += _pcre_ord2utf8(0x1fff, class_utf8data);
3260                  *class_utf8data++ = XCL_RANGE;
3261                  class_utf8data += _pcre_ord2utf8(0x200B, class_utf8data);
3262                  class_utf8data += _pcre_ord2utf8(0x202e, class_utf8data);
3263                  *class_utf8data++ = XCL_RANGE;
3264                  class_utf8data += _pcre_ord2utf8(0x2030, class_utf8data);
3265                  class_utf8data += _pcre_ord2utf8(0x205e, class_utf8data);
3266                  *class_utf8data++ = XCL_RANGE;
3267                  class_utf8data += _pcre_ord2utf8(0x2060, class_utf8data);
3268                  class_utf8data += _pcre_ord2utf8(0x2fff, class_utf8data);
3269                  *class_utf8data++ = XCL_RANGE;
3270                  class_utf8data += _pcre_ord2utf8(0x3001, class_utf8data);
3271                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3272                  }
3273    #endif
3274                continue;
3275                }
3276    
3277              if (-c == ESC_v)
3278                {
3279                SETBIT(classbits, 0x0a); /* LF */
3280                SETBIT(classbits, 0x0b); /* VT */
3281                SETBIT(classbits, 0x0c); /* FF */
3282                SETBIT(classbits, 0x0d); /* CR */
3283                SETBIT(classbits, 0x85); /* NEL */
3284    #ifdef SUPPORT_UTF8
3285                if (utf8)
3286                  {
3287                  class_utf8 = TRUE;
3288                  *class_utf8data++ = XCL_RANGE;
3289                  class_utf8data += _pcre_ord2utf8(0x2028, class_utf8data);
3290                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3291                  }
3292    #endif
3293                continue;
3294                }
3295    
3296              if (-c == ESC_V)
3297                {
3298                for (c = 0; c < 32; c++)
3299                  {
3300                  int x = 0xff;
3301                  switch (c)
3302                    {
3303                    case 0x0a/8: x ^= 1 << (0x0a%8);
3304                                 x ^= 1 << (0x0b%8);
3305                                 x ^= 1 << (0x0c%8);
3306                                 x ^= 1 << (0x0d%8);
3307                                 break;
3308                    case 0x85/8: x ^= 1 << (0x85%8); break;
3309                    default: break;
3310                    }
3311                  classbits[c] |= x;
3312                  }
3313    
3314    #ifdef SUPPORT_UTF8
3315                if (utf8)
3316                  {
3317                  class_utf8 = TRUE;
3318                  *class_utf8data++ = XCL_RANGE;
3319                  class_utf8data += _pcre_ord2utf8(0x0100, class_utf8data);
3320                  class_utf8data += _pcre_ord2utf8(0x2027, class_utf8data);
3321                  *class_utf8data++ = XCL_RANGE;
3322                  class_utf8data += _pcre_ord2utf8(0x2029, class_utf8data);
3323                  class_utf8data += _pcre_ord2utf8(0x7fffffff, class_utf8data);
3324                  }
3325    #endif
3326                continue;
3327                }
3328    
3329            /* We need to deal with \P and \p in both phases. */            /* We need to deal with \P and \p in both phases. */
3330    
3331  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
# Line 2562  for (;; ptr++) Line 3369  for (;; ptr++)
3369        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3370    
3371        CHECK_RANGE:        CHECK_RANGE:
3372        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3373          {          {
3374          inescq = FALSE;          inescq = FALSE;
3375          ptr += 2;          ptr += 2;
# Line 2570  for (;; ptr++) Line 3377  for (;; ptr++)
3377    
3378        oldptr = ptr;        oldptr = ptr;
3379    
3380        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3381    
3382          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3383    
3384          /* Check for range */
3385    
3386          if (!inescq && ptr[1] == CHAR_MINUS)
3387          {          {
3388          int d;          int d;
3389          ptr += 2;          ptr += 2;
3390          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3391    
3392          /* 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
3393          mode. */          mode. */
3394    
3395          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3396            {            {
3397            ptr += 2;            ptr += 2;
3398            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3399                { ptr += 2; continue; }
3400            inescq = TRUE;            inescq = TRUE;
3401            break;            break;
3402            }            }
3403    
3404          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3405            {            {
3406            ptr = oldptr;            ptr = oldptr;
3407            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2606  for (;; ptr++) Line 3420  for (;; ptr++)
3420          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
3421          in such circumstances. */          in such circumstances. */
3422    
3423          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3424            {            {
3425            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3426            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3427    
3428            /* \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
3429            special means the '-' was literal */            special means the '-' was literal */
3430    
3431            if (d < 0)            if (d < 0)
3432              {              {
3433              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS;
3434              else if (d == -ESC_X) d = 'X';              else if (d == -ESC_X) d = CHAR_X;
3435              else if (d == -ESC_R) d = 'R'; else              else if (d == -ESC_R) d = CHAR_R; else
3436                {                {
3437                ptr = oldptr;                ptr = oldptr;
3438                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2637  for (;; ptr++) Line 3451  for (;; ptr++)
3451    
3452          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3453    
3454            /* Remember \r or \n */
3455    
3456            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3457    
3458          /* 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
3459          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3460          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 3477  for (;; ptr++)
3477              unsigned int origd = d;              unsigned int origd = d;
3478              while (get_othercase_range(&cc, origd, &occ, &ocd))              while (get_othercase_range(&cc, origd, &occ, &ocd))
3479                {                {
3480                if (occ >= c && ocd <= d) continue;  /* Skip embedded ranges */                if (occ >= (unsigned int)c &&
3481                      ocd <= (unsigned int)d)
3482                    continue;                          /* Skip embedded ranges */
3483    
3484                if (occ < c  && ocd >= c - 1)        /* Extend the basic range */                if (occ < (unsigned int)c  &&
3485                      ocd >= (unsigned int)c - 1)      /* Extend the basic range */
3486                  {                                  /* if there is overlap,   */                  {                                  /* if there is overlap,   */
3487                  c = occ;                           /* noting that if occ < c */                  c = occ;                           /* noting that if occ < c */
3488                  continue;                          /* we can't have ocd > d  */                  continue;                          /* we can't have ocd > d  */
3489                  }                                  /* because a subrange is  */                  }                                  /* because a subrange is  */
3490                if (ocd > d && occ <= d + 1)         /* always shorter than    */                if (ocd > (unsigned int)d &&
3491                      occ <= (unsigned int)d + 1)      /* always shorter than    */
3492                  {                                  /* the basic range.       */                  {                                  /* the basic range.       */
3493                  d = ocd;                  d = ocd;
3494                  continue;                  continue;
# Line 2751  for (;; ptr++) Line 3573  for (;; ptr++)
3573          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3574            {            {
3575            unsigned int othercase;            unsigned int othercase;
3576            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3577              {              {
3578              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3579              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 2778  for (;; ptr++) Line 3600  for (;; ptr++)
3600    
3601      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" above. */
3602    
3603      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      while ((c = *(++ptr)) != 0 && (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3604    
3605      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)                          /* Missing terminating ']' */
3606        {        {
# Line 2786  for (;; ptr++) Line 3608  for (;; ptr++)
3608        goto FAILED;        goto FAILED;
3609        }        }
3610    
3611    
3612    /* This code has been disabled because it would mean that \s counts as
3613    an explicit \r or \n reference, and that's not really what is wanted. Now
3614    we set the flag only if there is a literal "\r" or "\n" in the class. */
3615    
3616    #if 0
3617        /* Remember whether \r or \n are in this class */
3618    
3619        if (negate_class)
3620          {
3621          if ((classbits[1] & 0x24) != 0x24) cd->external_flags |= PCRE_HASCRORLF;
3622          }
3623        else
3624          {
3625          if ((classbits[1] & 0x24) != 0) cd->external_flags |= PCRE_HASCRORLF;
3626          }
3627    #endif
3628    
3629    
3630      /* 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
3631      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
3632      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
3633      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3634      single-bytes only. This is an historical hangover. Maybe one day we can  
3635      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3636        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3637        operate on single-bytes only. This is an historical hangover. Maybe one day
3638        we can tidy these opcodes to handle multi-byte characters.
3639    
3640      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
3641      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 3645  for (;; ptr++)
3645      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3646    
3647  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3648      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3649            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3650  #else  #else
3651      if (class_charcount == 1)      if (class_charcount == 1)
3652  #endif  #endif
# Line 2847  for (;; ptr++) Line 3689  for (;; ptr++)
3689      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3690    
3691      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3692      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3693      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
3694        the class, so any that were explicitly given as well can be ignored. If
3695        (when there are explicit characters > 255 that must be listed) there are no
3696        characters < 256, we can omit the bitmap in the actual compiled code. */
3697    
3698  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3699      if (class_utf8)      if (class_utf8 && !should_flip_negation)
3700        {        {
3701        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3702        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 2877  for (;; ptr++) Line 3722  for (;; ptr++)
3722        }        }
3723  #endif  #endif
3724    
3725      /* 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
3726      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
3727      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
3728      setting must remain unchanged after any kind of repeat. */      map into the code vector, negating it if necessary. */
3729    
3730        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
3731      if (negate_class)      if (negate_class)
3732        {        {
       *code++ = OP_NCLASS;  
3733        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
3734          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
3735        }        }
3736      else      else
3737        {        {
       *code++ = OP_CLASS;  
3738        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
3739        }        }
3740      code += 32;      code += 32;
# Line 2901  for (;; ptr++) Line 3745  for (;; ptr++)
3745      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
3746      has been tested above. */      has been tested above. */
3747    
3748      case '{':      case CHAR_LEFT_CURLY_BRACKET:
3749      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
3750      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
3751      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
3752      goto REPEAT;      goto REPEAT;
3753    
3754      case '*':      case CHAR_ASTERISK:
3755      repeat_min = 0;      repeat_min = 0;
3756      repeat_max = -1;      repeat_max = -1;
3757      goto REPEAT;      goto REPEAT;
3758    
3759      case '+':      case CHAR_PLUS:
3760      repeat_min = 1;      repeat_min = 1;
3761      repeat_max = -1;      repeat_max = -1;
3762      goto REPEAT;      goto REPEAT;
3763    
3764      case '?':      case CHAR_QUESTION_MARK:
3765      repeat_min = 0;      repeat_min = 0;
3766      repeat_max = 1;      repeat_max = 1;
3767    
# Line 2952  for (;; ptr++) Line 3796  for (;; ptr++)
3796      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
3797      repeat type to the non-default. */      repeat type to the non-default. */
3798    
3799      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
3800        {        {
3801        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
3802        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
3803        ptr++;        ptr++;
3804        }        }
3805      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
3806        {        {
3807        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
3808        ptr++;        ptr++;
# Line 3073  for (;; ptr++) Line 3917  for (;; ptr++)
3917    
3918        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
3919    
3920          /*--------------------------------------------------------------------*/
3921          /* This code is obsolete from release 8.00; the restriction was finally
3922          removed: */
3923    
3924        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
3925        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
3926    
3927        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
3928          /*--------------------------------------------------------------------*/
3929    
3930        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
3931    
# Line 3223  for (;; ptr++) Line 4072  for (;; ptr++)
4072          goto END_REPEAT;          goto END_REPEAT;
4073          }          }
4074    
4075          /*--------------------------------------------------------------------*/
4076          /* This code is obsolete from release 8.00; the restriction was finally
4077          removed: */
4078    
4079        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4080        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4081    
4082        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4083          /*--------------------------------------------------------------------*/
4084    
4085        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4086          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3262  for (;; ptr++) Line 4116  for (;; ptr++)
4116          goto FAILED;          goto FAILED;
4117          }          }
4118    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4119        /* 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
4120        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
4121        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 4138  for (;; ptr++)
4138    
4139        if (repeat_min == 0)        if (repeat_min == 0)
4140          {          {
4141          /* 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
4142          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4143    
4144          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4145          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4146          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4147          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4148          the start of the whole regex. Temporarily terminate the pattern while          **   }
4149          doing this. */  
4150            However, that fails when a group is referenced as a subroutine from
4151            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4152            so that it is skipped on execution. As we don't have a list of which
4153            groups are referenced, we cannot do this selectively.
4154    
4155            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4156            and do no more at this point. However, we do need to adjust any
4157            OP_RECURSE calls inside the group that refer to the group itself or any
4158            internal or forward referenced group, because the offset is from the
4159            start of the whole regex. Temporarily terminate the pattern while doing
4160            this. */
4161    
4162          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4163            {            {
4164            *code = OP_END;            *code = OP_END;
4165            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4166            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4167            code++;            code++;
4168              if (repeat_max == 0)
4169                {
4170                *previous++ = OP_SKIPZERO;
4171                goto END_REPEAT;
4172                }
4173            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4174            }            }
4175    
# Line 3358  for (;; ptr++) Line 4214  for (;; ptr++)
4214          if (repeat_min > 1)          if (repeat_min > 1)
4215            {            {
4216            /* 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
4217            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4218              potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4219              integer type when available, otherwise double. */
4220    
4221            if (lengthptr != NULL)            if (lengthptr != NULL)
4222              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4223                int delta = (repeat_min - 1)*length_prevgroup;
4224                if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4225                      (INT64_OR_DOUBLE)length_prevgroup >
4226                        (INT64_OR_DOUBLE)INT_MAX ||
4227                    OFLOW_MAX - *lengthptr < delta)
4228                  {
4229                  *errorcodeptr = ERR20;
4230                  goto FAILED;
4231                  }
4232                *lengthptr += delta;
4233                }
4234    
4235            /* This is compiling for real */            /* This is compiling for real */
4236    
# Line 3399  for (;; ptr++) Line 4268  for (;; ptr++)
4268          /* 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
4269          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
4270          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
4271          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4272            paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4273            a 64-bit integer type when available, otherwise double. */
4274    
4275          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4276            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4277              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4278                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4279              if ((INT64_OR_DOUBLE)repeat_max *
4280                    (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4281                      > (INT64_OR_DOUBLE)INT_MAX ||
4282                  OFLOW_MAX - *lengthptr < delta)
4283                {
4284                *errorcodeptr = ERR20;
4285                goto FAILED;
4286                }
4287              *lengthptr += delta;
4288              }
4289    
4290          /* This is compiling for real */          /* This is compiling for real */
4291    
# Line 3485  for (;; ptr++) Line 4367  for (;; ptr++)
4367          }          }
4368        }        }
4369    
4370        /* If previous is OP_FAIL, it was generated by an empty class [] in
4371        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4372        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4373        error above. We can just ignore the repeat in JS case. */
4374    
4375        else if (*previous == OP_FAIL) goto END_REPEAT;
4376    
4377      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4378    
4379      else      else
# Line 3509  for (;; ptr++) Line 4398  for (;; ptr++)
4398      if (possessive_quantifier)      if (possessive_quantifier)
4399        {        {
4400        int len;        int len;
4401        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4402            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4403            tempcode += _pcre_OP_lengths[*tempcode] +
4404              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4405    
4406          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4407            {
4408          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4409    #ifdef SUPPORT_UTF8
4410            if (utf8 && tempcode[-1] >= 0xc0)
4411              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4412    #endif
4413            }
4414    
4415        len = code - tempcode;        len = code - tempcode;
4416        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4417          {          {
# Line 3555  for (;; ptr++) Line 4455  for (;; ptr++)
4455      /* ===================================================================*/      /* ===================================================================*/
4456      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4457      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4458      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. */  
4459    
4460      case '(':      case CHAR_LEFT_PARENTHESIS:
4461      newoptions = options;      newoptions = options;
4462      skipbytes = 0;      skipbytes = 0;
4463      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4464      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4465        reset_bracount = FALSE;
4466    
4467        /* First deal with various "verbs" that can be introduced by '*'. */
4468    
4469        if (*(++ptr) == CHAR_ASTERISK && (cd->ctypes[ptr[1]] & ctype_letter) != 0)
4470          {
4471          int i, namelen;
4472          const char *vn = verbnames;
4473          const uschar *name = ++ptr;
4474          previous = NULL;
4475          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4476          if (*ptr == CHAR_COLON)
4477            {
4478            *errorcodeptr = ERR59;   /* Not supported */
4479            goto FAILED;
4480            }
4481          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4482            {
4483            *errorcodeptr = ERR60;
4484            goto FAILED;
4485            }
4486          namelen = ptr - name;
4487          for (i = 0; i < verbcount; i++)
4488            {
4489            if (namelen == verbs[i].len &&
4490                strncmp((char *)name, vn, namelen) == 0)
4491              {
4492              /* Check for open captures before ACCEPT */
4493    
4494              if (verbs[i].op == OP_ACCEPT)
4495                {
4496                open_capitem *oc;
4497                cd->had_accept = TRUE;
4498                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4499                  {
4500                  *code++ = OP_CLOSE;
4501                  PUT2INC(code, 0, oc->number);
4502                  }
4503                }
4504              *code++ = verbs[i].op;
4505              break;
4506              }
4507            vn += verbs[i].len + 1;
4508            }
4509          if (i < verbcount) continue;
4510          *errorcodeptr = ERR60;
4511          goto FAILED;
4512          }
4513    
4514        /* Deal with the extended parentheses; all are introduced by '?', and the
4515        appearance of any of them means that this is not a capturing group. */
4516    
4517      if (*(++ptr) == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4518        {        {
4519        int i, set, unset, namelen;        int i, set, unset, namelen;
4520        int *optset;        int *optset;
# Line 3574  for (;; ptr++) Line 4523  for (;; ptr++)
4523    
4524        switch (*(++ptr))        switch (*(++ptr))
4525          {          {
4526          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4527          ptr++;          ptr++;
4528          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4529          if (*ptr == 0)          if (*ptr == 0)
4530            {            {
4531            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3586  for (;; ptr++) Line 4535  for (;; ptr++)
4535    
4536    
4537          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4538          case ':':                 /* Non-capturing bracket */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4539            reset_bracount = TRUE;
4540            /* Fall through */
4541    
4542            /* ------------------------------------------------------------ */
4543            case CHAR_COLON:          /* Non-capturing bracket */
4544          bravalue = OP_BRA;          bravalue = OP_BRA;
4545          ptr++;          ptr++;
4546          break;          break;
4547    
4548    
4549          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4550          case '(':          case CHAR_LEFT_PARENTHESIS:
4551          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4552    
4553          /* 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 4567  for (;; ptr++)
4567          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4568          including assertions, are processed. */          including assertions, are processed. */
4569    
4570          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4571                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4572            break;            break;
4573    
4574          /* 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 4576  for (;; ptr++)
4576    
4577          code[1+LINK_SIZE] = OP_CREF;          code[1+LINK_SIZE] = OP_CREF;
4578          skipbytes = 3;          skipbytes = 3;
4579            refsign = -1;
4580    
4581          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4582    
4583          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4584            {            {
4585            terminator = -1;            terminator = -1;
4586            ptr += 2;            ptr += 2;
# Line 3634  for (;; ptr++) Line 4590  for (;; ptr++)
4590          /* 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
4591          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4592    
4593          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4594            {            {
4595            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4596            ptr++;            ptr++;
4597            }            }
4598          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4599            {            {
4600            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4601            ptr++;            ptr++;
4602            }            }
4603          else terminator = 0;          else
4604              {
4605              terminator = 0;
4606              if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4607              }
4608    
4609          /* 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 */
4610    
# Line 3663  for (;; ptr++) Line 4623  for (;; ptr++)
4623            {            {
4624            if (recno >= 0)            if (recno >= 0)
4625              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4626                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4627            ptr++;            ptr++;
4628            }            }
4629          namelen = ptr - name;          namelen = ptr - name;
4630    
4631          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4632                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4633            {            {
4634            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4635            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3680  for (;; ptr++) Line 4641  for (;; ptr++)
4641          if (lengthptr != NULL) break;          if (lengthptr != NULL) break;
4642    
4643          /* 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
4644          reference. */          reference. If the string started with "+" or "-" we require the rest to
4645            be digits, in which case recno will be set. */
4646    
4647            if (refsign > 0)
4648              {
4649              if (recno <= 0)
4650                {
4651                *errorcodeptr = ERR58;
4652                goto FAILED;
4653                }
4654              recno = (refsign == CHAR_MINUS)?
4655                cd->bracount - recno + 1 : recno +cd->bracount;
4656              if (recno <= 0 || recno > cd->final_bracount)
4657                {
4658                *errorcodeptr = ERR15;
4659                goto FAILED;
4660                }
4661              PUT2(code, 2+LINK_SIZE, recno);
4662              break;
4663              }
4664    
4665            /* Otherwise (did not start with "+" or "-"), start by looking for the
4666            name. If we find a name, add one to the opcode to change OP_CREF or
4667            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4668            except they record that the reference was originally to a name. The
4669            information is used to check duplicate names. */
4670    
4671          slot = cd->name_table;          slot = cd->name_table;
4672          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 3695  for (;; ptr++) Line 4681  for (;; ptr++)
4681            {            {
4682            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4683            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4684              code[1+LINK_SIZE]++;
4685            }            }
4686    
4687          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
4688    
4689          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
4690                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
4691            {            {
4692            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
4693              code[1+LINK_SIZE]++;
4694            }            }
4695    
4696          /* 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 4708  for (;; ptr++)
4708          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
4709          specific group number. */          specific group number. */
4710    
4711          else if (*name == 'R')          else if (*name == CHAR_R)
4712            {            {
4713            recno = 0;            recno = 0;
4714            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 3730  for (;; ptr++) Line 4718  for (;; ptr++)
4718                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
4719                goto FAILED;                goto FAILED;
4720                }                }
4721              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
4722              }              }
4723            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
4724            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 3740  for (;; ptr++) Line 4728  for (;; ptr++)
4728          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
4729          false. */          false. */
4730    
4731          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
4732            {            {
4733            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
4734            skipbytes = 1;            skipbytes = 1;
4735            }            }
4736    
4737          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
4738            in the second pass here, so final_bracount is set. */
4739    
4740          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
4741            {            {
4742            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4743            }            }
# Line 3764  for (;; ptr++) Line 4753  for (;; ptr++)
4753    
4754    
4755          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4756          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
4757          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
4758          ptr++;          ptr++;
4759          break;          break;
4760    
4761    
4762          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4763          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
4764          ptr++;          ptr++;
4765            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
4766              {
4767              *code++ = OP_FAIL;
4768              previous = NULL;
4769              continue;
4770              }
4771            bravalue = OP_ASSERT_NOT;
4772          break;          break;
4773    
4774    
4775          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4776          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
4777          switch (ptr[1])          switch (ptr[1])
4778            {            {
4779            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
4780            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
4781            ptr += 2;            ptr += 2;
4782            break;            break;
4783    
4784            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
4785            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
4786            ptr += 2;            ptr += 2;
4787            break;            break;
# Line 3801  for (;; ptr++) Line 4796  for (;; ptr++)
4796    
4797    
4798          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4799          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
4800          bravalue = OP_ONCE;          bravalue = OP_ONCE;
4801          ptr++;          ptr++;
4802          break;          break;
4803    
4804    
4805          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4806          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
4807          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
4808          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
4809          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
4810            {            {
4811            int n = 0;            int n = 0;
4812            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
4813              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
4814            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
4815              {              {
4816              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
4817              goto FAILED;              goto FAILED;
# Line 3836  for (;; ptr++) Line 4831  for (;; ptr++)
4831    
4832    
4833          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4834          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
4835          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
4836                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
4837            {            {
4838            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
4839            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
4840            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
4841            }            }
4842          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
4843            {            {
4844            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
4845            goto FAILED;            goto FAILED;
# Line 3853  for (;; ptr++) Line 4849  for (;; ptr++)
4849    
4850          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4851          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
4852          case '\'':          case CHAR_APOSTROPHE:
4853            {            {
4854            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
4855                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
4856            name = ++ptr;            name = ++ptr;
4857    
4858            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
# Line 3886  for (;; ptr++) Line 4883  for (;; ptr++)
4883                }                }
4884              }              }
4885    
4886            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
4887              alphabetical order. Duplicate names for different numbers are
4888              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
4889              number are always OK. (An existing number can be re-used if (?|
4890              appears in the pattern.) In either event, a duplicate name results in
4891              a duplicate entry in the table, even if the number is the same. This
4892              is because the number of names, and hence the table size, is computed
4893              in the pre-compile, and it affects various numbers and pointers which
4894              would all have to be modified, and the compiled code moved down, if
4895              duplicates with the same number were omitted from the table. This
4896              doesn't seem worth the hassle. However, *different* names for the
4897              same number are not permitted. */
4898    
4899            else            else
4900              {              {
4901                BOOL dupname = FALSE;
4902              slot = cd->name_table;              slot = cd->name_table;
4903    
4904              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
4905                {                {
4906                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 3898  for (;; ptr++) Line 4908  for (;; ptr++)
4908                  {                  {
4909                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
4910                    {                    {
4911                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
4912                          (options & PCRE_DUPNAMES) == 0)
4913                      {                      {
4914                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
4915                      goto FAILED;                      goto FAILED;
4916                      }                      }
4917                      else dupname = TRUE;
4918                    }                    }
4919                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
4920                  }                  }
4921    
4922                  /* Make space in the table and break the loop for an earlier
4923                  name. For a duplicate or later name, carry on. We do this for
4924                  duplicates so that in the simple case (when ?(| is not used) they
4925                  are in order of their numbers. */
4926    
4927                if (crc < 0)                if (crc < 0)
4928                  {                  {
4929                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
4930                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
4931                  break;                  break;
4932                  }                  }
4933    
4934                  /* Continue the loop for a later or duplicate name */
4935    
4936                slot += cd->name_entry_size;                slot += cd->name_entry_size;
4937                }                }
4938    
4939                /* For non-duplicate names, check for a duplicate number before
4940                adding the new name. */
4941    
4942                if (!dupname)
4943                  {
4944                  uschar *cslot = cd->name_table;
4945                  for (i = 0; i < cd->names_found; i++)
4946                    {
4947                    if (cslot != slot)
4948                      {
4949                      if (GET2(cslot, 0) == cd->bracount + 1)
4950                        {
4951                        *errorcodeptr = ERR65;
4952                        goto FAILED;
4953                        }
4954                      }
4955                    else i--;
4956                    cslot += cd->name_entry_size;
4957                    }
4958                  }
4959    
4960              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
4961              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
4962              slot[2+namelen] = 0;              slot[2+namelen] = 0;
4963              }              }
4964            }            }
4965    
4966          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
4967            encountered. */
4968    
         ptr++;                    /* Move past > or ' */  
4969          cd->names_found++;          cd->names_found++;
4970            ptr++;                    /* Move past > or ' */
4971          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
4972    
4973    
4974          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4975          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
4976          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
4977          is_recurse = TRUE;          is_recurse = TRUE;
4978          /* Fall through */          /* Fall through */
4979    
4980          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
4981          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
4982          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
4983            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
4984            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
4985    
4986          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
4987          name = ++ptr;          name = ++ptr;
# Line 3948  for (;; ptr++) Line 4993  for (;; ptr++)
4993    
4994          if (lengthptr != NULL)          if (lengthptr != NULL)
4995            {            {
4996              if (namelen == 0)
4997                {
4998                *errorcodeptr = ERR62;
4999                goto FAILED;
5000                }
5001            if (*ptr != terminator)            if (*ptr != terminator)
5002              {              {
5003              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 3961  for (;; ptr++) Line 5011  for (;; ptr++)
5011            recno = 0;            recno = 0;
5012            }            }
5013    
5014          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
5015            first, and then check that we have reached the end of the name in the
5016            table. That way, if the name that is longer than any in the table,
5017            the comparison will fail without reading beyond the table entry. */
5018    
5019          else          else
5020            {            {
5021            slot = cd->name_table;            slot = cd->name_table;
5022            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5023              {              {
5024              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
5025                    slot[2+namelen] == 0)
5026                  break;
5027              slot += cd->name_entry_size;              slot += cd->name_entry_size;
5028              }              }
5029    
# Line 3977  for (;; ptr++) Line 5032  for (;; ptr++)
5032              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5033              }              }
5034            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5035                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5036                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5037              {              {
5038              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 3993  for (;; ptr++) Line 5048  for (;; ptr++)
5048    
5049    
5050          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5051          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5052          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */
5053          /* Fall through */          /* Fall through */
5054    
5055    
5056          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5057          case '0': case '1': case '2': case '3': case '4':   /* Recursion or */          case CHAR_MINUS: case CHAR_PLUS:  /* Recursion or subroutine */
5058          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:
5059            case CHAR_5: case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
5060            {            {
5061            const uschar *called;            const uschar *called;
5062              terminator = CHAR_RIGHT_PARENTHESIS;
5063    
5064              /* Come here from the \g<...> and \g'...' code (Oniguruma
5065              compatibility). However, the syntax has been checked to ensure that
5066              the ... are a (signed) number, so that neither ERR63 nor ERR29 will
5067              be called on this path, nor with the jump to OTHER_CHAR_AFTER_QUERY
5068              ever be taken. */
5069    
5070              HANDLE_NUMERICAL_RECURSION:
5071    
5072              if ((refsign = *ptr) == CHAR_PLUS)
5073                {
5074                ptr++;
5075                if ((digitab[*ptr] & ctype_digit) == 0)
5076                  {
5077                  *errorcodeptr = ERR63;
5078                  goto FAILED;
5079                  }
5080                }
5081              else if (refsign == CHAR_MINUS)
5082                {
5083                if ((digitab[ptr[1]] & ctype_digit) == 0)
5084                  goto OTHER_CHAR_AFTER_QUERY;
5085                ptr++;
5086                }
5087    
5088            recno = 0;            recno = 0;
5089            while((digitab[*ptr] & ctype_digit) != 0)            while((digitab[*ptr] & ctype_digit) != 0)
5090              recno = recno * 10 + *ptr++ - '0';              recno = recno * 10 + *ptr++ - CHAR_0;
5091            if (*ptr != ')')  
5092              if (*ptr != terminator)
5093              {              {
5094              *errorcodeptr = ERR29;              *errorcodeptr = ERR29;
5095              goto FAILED;              goto FAILED;
5096              }              }
5097    
5098              if (refsign == CHAR_MINUS)
5099                {
5100                if (recno == 0)
5101                  {
5102                  *errorcodeptr = ERR58;
5103                  goto FAILED;
5104                  }
5105                recno = cd->bracount - recno + 1;
5106                if (recno <= 0)
5107                  {
5108                  *errorcodeptr = ERR15;
5109                  goto FAILED;
5110                  }
5111                }
5112 &