/[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 205 by ph10, Fri Aug 3 13:18:33 2007 UTC revision 556 by ph10, Tue Oct 26 11:06:44 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 43  supporting internal functions that are n Line 43  supporting internal functions that are n
43    
44    
45  #ifdef HAVE_CONFIG_H  #ifdef HAVE_CONFIG_H
46  #include <config.h>  #include "config.h"
47  #endif  #endif
48    
49  #define NLBLOCK cd             /* Block containing newline information */  #define NLBLOCK cd             /* Block containing newline information */
# Line 53  supporting internal functions that are n Line 53  supporting internal functions that are n
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    
# Line 91  is 4 there is plenty of room. */ Line 92  is 4 there is plenty of room. */
92    
93  #define COMPILE_WORK_SIZE (4096)  #define COMPILE_WORK_SIZE (4096)
94    
95    /* The overrun tests check for a slightly smaller size so that they detect the
96    overrun before it actually does run off the end of the data block. */
97    
98    #define WORK_SIZE_CHECK (COMPILE_WORK_SIZE - 100)
99    
100    
101  /* Table for handling escaped characters in the range '0'-'z'. Positive returns  /* Table for handling escaped characters in the range '0'-'z'. Positive returns
102  are simple data values; negative values are for special things like \d and so  are simple data values; negative values are for special things like \d and so
103  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
104  is invalid. */  is invalid. */
105    
106  #ifndef EBCDIC  /* This is the "normal" table for ASCII systems */  #ifndef EBCDIC
107    
108    /* This is the "normal" table for ASCII systems or for EBCDIC systems running
109    in UTF-8 mode. */
110    
111  static const short int escapes[] = {  static const short int escapes[] = {
112       0,      0,      0,      0,      0,      0,      0,      0,   /* 0 - 7 */       0,                       0,
113       0,      0,    ':',    ';',    '<',    '=',    '>',    '?',   /* 8 - ? */       0,                       0,
114     '@', -ESC_A, -ESC_B, -ESC_C, -ESC_D, -ESC_E,      0, -ESC_G,   /* @ - G */       0,                       0,
115  -ESC_H,      0,      0, -ESC_K,      0,      0,      0,      0,   /* H - O */       0,                       0,
116  -ESC_P, -ESC_Q, -ESC_R, -ESC_S,      0,      0, -ESC_V, -ESC_W,   /* P - W */       0,                       0,
117  -ESC_X,      0, -ESC_Z,    '[',   '\\',    ']',    '^',    '_',   /* X - _ */       CHAR_COLON,              CHAR_SEMICOLON,
118     '`',      7, -ESC_b,      0, -ESC_d,  ESC_e,  ESC_f,      0,   /* ` - g */       CHAR_LESS_THAN_SIGN,     CHAR_EQUALS_SIGN,
119  -ESC_h,      0,      0, -ESC_k,      0,      0,  ESC_n,      0,   /* h - o */       CHAR_GREATER_THAN_SIGN,  CHAR_QUESTION_MARK,
120  -ESC_p,      0,  ESC_r, -ESC_s,  ESC_tee,    0, -ESC_v, -ESC_w,   /* p - w */       CHAR_COMMERCIAL_AT,      -ESC_A,
121       0,      0, -ESC_z                                            /* x - z */       -ESC_B,                  -ESC_C,
122         -ESC_D,                  -ESC_E,
123         0,                       -ESC_G,
124         -ESC_H,                  0,
125         0,                       -ESC_K,
126         0,                       0,
127         -ESC_N,                  0,
128         -ESC_P,                  -ESC_Q,
129         -ESC_R,                  -ESC_S,
130         0,                       0,
131         -ESC_V,                  -ESC_W,
132         -ESC_X,                  0,
133         -ESC_Z,                  CHAR_LEFT_SQUARE_BRACKET,
134         CHAR_BACKSLASH,          CHAR_RIGHT_SQUARE_BRACKET,
135         CHAR_CIRCUMFLEX_ACCENT,  CHAR_UNDERSCORE,
136         CHAR_GRAVE_ACCENT,       7,
137         -ESC_b,                  0,
138         -ESC_d,                  ESC_e,
139         ESC_f,                   0,
140         -ESC_h,                  0,
141         0,                       -ESC_k,
142         0,                       0,
143         ESC_n,                   0,
144         -ESC_p,                  0,
145         ESC_r,                   -ESC_s,
146         ESC_tee,                 0,
147         -ESC_v,                  -ESC_w,
148         0,                       0,
149         -ESC_z
150  };  };
151    
152  #else           /* This is the "abnormal" table for EBCDIC systems */  #else
153    
154    /* This is the "abnormal" table for EBCDIC systems without UTF-8 support. */
155    
156  static const short int escapes[] = {  static const short int escapes[] = {
157  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',  /*  48 */     0,     0,      0,     '.',    '<',   '(',    '+',    '|',
158  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,  /*  50 */   '&',     0,      0,       0,      0,     0,      0,      0,
# Line 130  static const short int escapes[] = { Line 171  static const short int escapes[] = {
171  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',  /*  B8 */     0,     0,      0,       0,      0,   ']',    '=',    '-',
172  /*  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,
173  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,  /*  C8 */-ESC_H,     0,      0,       0,      0,     0,      0,      0,
174  /*  D0 */   '}',     0, -ESC_K,       0,      0,     0,      0, -ESC_P,  /*  D0 */   '}',     0, -ESC_K,       0,      0,-ESC_N,      0, -ESC_P,
175  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,  /*  D8 */-ESC_Q,-ESC_R,      0,       0,      0,     0,      0,      0,
176  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,  /*  E0 */  '\\',     0, -ESC_S,       0,      0,-ESC_V, -ESC_W, -ESC_X,
177  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,  /*  E8 */     0,-ESC_Z,      0,       0,      0,     0,      0,      0,
# Line 140  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* 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
185  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
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    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 179  static const int posix_class_maps[] = { Line 261  static const int posix_class_maps[] = {
261    cbit_xdigit,-1,          0              /* xdigit */    cbit_xdigit,-1,          0              /* xdigit */
262  };  };
263    
264    /* Table of substitutes for \d etc when PCRE_UCP is set. The POSIX class
265    substitutes must be in the order of the names, defined above, and there are
266    both positive and negative cases. NULL means no substitute. */
267    
268    #ifdef SUPPORT_UCP
269    static const uschar *substitutes[] = {
270      (uschar *)"\\P{Nd}",    /* \D */
271      (uschar *)"\\p{Nd}",    /* \d */
272      (uschar *)"\\P{Xsp}",   /* \S */       /* NOTE: Xsp is Perl space */
273      (uschar *)"\\p{Xsp}",   /* \s */
274      (uschar *)"\\P{Xwd}",   /* \W */
275      (uschar *)"\\p{Xwd}"    /* \w */
276    };
277    
278    static const uschar *posix_substitutes[] = {
279      (uschar *)"\\p{L}",     /* alpha */
280      (uschar *)"\\p{Ll}",    /* lower */
281      (uschar *)"\\p{Lu}",    /* upper */
282      (uschar *)"\\p{Xan}",   /* alnum */
283      NULL,                   /* ascii */
284      (uschar *)"\\h",        /* blank */
285      NULL,                   /* cntrl */
286      (uschar *)"\\p{Nd}",    /* digit */
287      NULL,                   /* graph */
288      NULL,                   /* print */
289      NULL,                   /* punct */
290      (uschar *)"\\p{Xps}",   /* space */    /* NOTE: Xps is POSIX space */
291      (uschar *)"\\p{Xwd}",   /* word */
292      NULL,                   /* xdigit */
293      /* Negated cases */
294      (uschar *)"\\P{L}",     /* ^alpha */
295      (uschar *)"\\P{Ll}",    /* ^lower */
296      (uschar *)"\\P{Lu}",    /* ^upper */
297      (uschar *)"\\P{Xan}",   /* ^alnum */
298      NULL,                   /* ^ascii */
299      (uschar *)"\\H",        /* ^blank */
300      NULL,                   /* ^cntrl */
301      (uschar *)"\\P{Nd}",    /* ^digit */
302      NULL,                   /* ^graph */
303      NULL,                   /* ^print */
304      NULL,                   /* ^punct */
305      (uschar *)"\\P{Xps}",   /* ^space */   /* NOTE: Xps is POSIX space */
306      (uschar *)"\\P{Xwd}",   /* ^word */
307      NULL                    /* ^xdigit */
308    };
309    #define POSIX_SUBSIZE (sizeof(posix_substitutes)/sizeof(uschar *))
310    #endif
311    
312  #define STRING(a)  # a  #define STRING(a)  # a
313  #define XSTRING(s) STRING(s)  #define XSTRING(s) STRING(s)
# Line 186  static const int posix_class_maps[] = { Line 315  static const int posix_class_maps[] = {
315  /* The texts of compile-time error messages. These are "char *" because they  /* The texts of compile-time error messages. These are "char *" because they
316  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
317  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
318  are no longer used. */  are no longer used. This used to be a table of strings, but in order to reduce
319    the number of relocations needed when a shared library is loaded dynamically,
320  static const char *error_texts[] = {  it is now one long string. We cannot use a table of offsets, because the
321    "no error",  lengths of inserts such as XSTRING(MAX_NAME_SIZE) are not known. Instead, we
322    "\\ at end of pattern",  simply count through to the one we want - this isn't a performance issue
323    "\\c at end of pattern",  because these strings are used only when there is a compilation error.
324    "unrecognized character follows \\",  
325    "numbers out of order in {} quantifier",  Each substring ends with \0 to insert a null character. This includes the final
326    substring, so that the whole string ends with \0\0, which can be detected when
327    counting through. */
328    
329    static const char error_texts[] =
330      "no error\0"
331      "\\ at end of pattern\0"
332      "\\c at end of pattern\0"
333      "unrecognized character follows \\\0"
334      "numbers out of order in {} quantifier\0"
335    /* 5 */    /* 5 */
336    "number too big in {} quantifier",    "number too big in {} quantifier\0"
337    "missing terminating ] for character class",    "missing terminating ] for character class\0"
338    "invalid escape sequence in character class",    "invalid escape sequence in character class\0"
339    "range out of order in character class",    "range out of order in character class\0"
340    "nothing to repeat",    "nothing to repeat\0"
341    /* 10 */    /* 10 */
342    "operand of unlimited repeat could match the empty string",  /** DEAD **/    "operand of unlimited repeat could match the empty string\0"  /** DEAD **/
343    "internal error: unexpected repeat",    "internal error: unexpected repeat\0"
344    "unrecognized character after (?",    "unrecognized character after (? or (?-\0"
345    "POSIX named classes are supported only within a class",    "POSIX named classes are supported only within a class\0"
346    "missing )",    "missing )\0"
347    /* 15 */    /* 15 */
348    "reference to non-existent subpattern",    "reference to non-existent subpattern\0"
349    "erroffset passed as NULL",    "erroffset passed as NULL\0"
350    "unknown option bit(s) set",    "unknown option bit(s) set\0"
351    "missing ) after comment",    "missing ) after comment\0"
352    "parentheses nested too deeply",  /** DEAD **/    "parentheses nested too deeply\0"  /** DEAD **/
353    /* 20 */    /* 20 */
354    "regular expression is too large",    "regular expression is too large\0"
355    "failed to get memory",    "failed to get memory\0"
356    "unmatched parentheses",    "unmatched parentheses\0"
357    "internal error: code overflow",    "internal error: code overflow\0"
358    "unrecognized character after (?<",    "unrecognized character after (?<\0"
359    /* 25 */    /* 25 */
360    "lookbehind assertion is not fixed length",    "lookbehind assertion is not fixed length\0"
361    "malformed number or name after (?(",    "malformed number or name after (?(\0"
362    "conditional group contains more than two branches",    "conditional group contains more than two branches\0"
363    "assertion expected after (?(",    "assertion expected after (?(\0"
364    "(?R or (?[+-]digits must be followed by )",    "(?R or (?[+-]digits must be followed by )\0"
365    /* 30 */    /* 30 */
366    "unknown POSIX class name",    "unknown POSIX class name\0"
367    "POSIX collating elements are not supported",    "POSIX collating elements are not supported\0"
368    "this version of PCRE is not compiled with PCRE_UTF8 support",    "this version of PCRE is not compiled with PCRE_UTF8 support\0"
369    "spare error",  /** DEAD **/    "spare error\0"  /** DEAD **/
370    "character value in \\x{...} sequence is too large",    "character value in \\x{...} sequence is too large\0"
371    /* 35 */    /* 35 */
372    "invalid condition (?(0)",    "invalid condition (?(0)\0"
373    "\\C not allowed in lookbehind assertion",    "\\C not allowed in lookbehind assertion\0"
374    "PCRE does not support \\L, \\l, \\N, \\U, or \\u",    "PCRE does not support \\L, \\l, \\N{name}, \\U, or \\u\0"
375    "number after (?C is > 255",    "number after (?C is > 255\0"
376    "closing ) for (?C expected",    "closing ) for (?C expected\0"
377    /* 40 */    /* 40 */
378    "recursive call could loop indefinitely",    "recursive call could loop indefinitely\0"
379    "unrecognized character after (?P",    "unrecognized character after (?P\0"
380    "syntax error in subpattern name (missing terminator)",    "syntax error in subpattern name (missing terminator)\0"
381    "two named subpatterns have the same name",    "two named subpatterns have the same name\0"
382    "invalid UTF-8 string",    "invalid UTF-8 string\0"
383    /* 45 */    /* 45 */
384    "support for \\P, \\p, and \\X has not been compiled",    "support for \\P, \\p, and \\X has not been compiled\0"
385    "malformed \\P or \\p sequence",    "malformed \\P or \\p sequence\0"
386    "unknown property name after \\P or \\p",    "unknown property name after \\P or \\p\0"
387    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)",    "subpattern name is too long (maximum " XSTRING(MAX_NAME_SIZE) " characters)\0"
388    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")",    "too many named subpatterns (maximum " XSTRING(MAX_NAME_COUNT) ")\0"
389    /* 50 */    /* 50 */
390    "repeated subpattern is too long",    /** DEAD **/    "repeated subpattern is too long\0"    /** DEAD **/
391    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
392    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
393    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
397    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
400  };    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401      /* 60 */
402      "(*VERB) not recognized\0"
403      "number is too big\0"
404      "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 278  For convenience, we use the same bit def Line 426  For convenience, we use the same bit def
426    
427  Then we can use ctype_digit and ctype_xdigit in the code. */  Then we can use ctype_digit and ctype_xdigit in the code. */
428    
429  #ifndef EBCDIC  /* This is the "normal" case, for ASCII systems */  #ifndef EBCDIC
430    
431    /* This is the "normal" case, for ASCII systems, and EBCDIC systems running in
432    UTF-8 mode. */
433    
434  static const unsigned char digitab[] =  static const unsigned char digitab[] =
435    {    {
436    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7 */
# Line 314  static const unsigned char digitab[] = Line 466  static const unsigned char digitab[] =
466    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /* 240-247 */
467    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00};/* 248-255 */
468    
469  #else           /* This is the "abnormal" case, for EBCDIC systems */  #else
470    
471    /* This is the "abnormal" case, for EBCDIC systems not running in UTF-8 mode. */
472    
473  static const unsigned char digitab[] =  static const unsigned char digitab[] =
474    {    {
475    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */    0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, /*   0-  7  0 */
# Line 395  static BOOL Line 550  static BOOL
550    
551    
552  /*************************************************  /*************************************************
553    *            Find an error text                  *
554    *************************************************/
555    
556    /* The error texts are now all in one long string, to save on relocations. As
557    some of the text is of unknown length, we can't use a table of offsets.
558    Instead, just count through the strings. This is not a performance issue
559    because it happens only when there has been a compilation error.
560    
561    Argument:   the error number
562    Returns:    pointer to the error string
563    */
564    
565    static const char *
566    find_error_text(int n)
567    {
568    const char *s = error_texts;
569    for (; n > 0; n--)
570      {
571      while (*s++ != 0) {};
572      if (*s == 0) return "Error text not found (please report)";
573      }
574    return s;
575    }
576    
577    
578    /*************************************************
579  *            Handle escapes                      *  *            Handle escapes                      *
580  *************************************************/  *************************************************/
581    
# Line 415  Arguments: Line 596  Arguments:
596    
597  Returns:         zero or positive => a data character  Returns:         zero or positive => a data character
598                   negative => a special escape sequence                   negative => a special escape sequence
599                   on error, errorptr is set                   on error, errorcodeptr is set
600  */  */
601    
602  static int  static int
# Line 433  ptr--;                            /* Set Line 614  ptr--;                            /* Set
614    
615  if (c == 0) *errorcodeptr = ERR1;  if (c == 0) *errorcodeptr = ERR1;
616    
617  /* 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
618  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.
619  Otherwise further processing may be required. */  Otherwise further processing may be required. */
620    
621  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
622  else if (c < '0' || c > 'z') {}                           /* Not alphameric */  else if (c < CHAR_0 || c > CHAR_z) {}                     /* Not alphanumeric */
623  else if ((i = escapes[c - '0']) != 0) c = i;  else if ((i = escapes[c - CHAR_0]) != 0) c = i;
624    
625  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
626  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphameric */  else if (c < 'a' || (ebcdic_chartab[c] & 0x0E) == 0) {}   /* Not alphanumeric */
627  else if ((i = escapes[c - 0x48]) != 0)  c = i;  else if ((i = escapes[c - 0x48]) != 0)  c = i;
628  #endif  #endif
629    
# Line 458  else Line 639  else
639      /* 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
640      error. */      error. */
641    
642      case 'l':      case CHAR_l:
643      case 'L':      case CHAR_L:
644      case 'N':      case CHAR_u:
645      case 'u':      case CHAR_U:
     case 'U':  
646      *errorcodeptr = ERR37;      *errorcodeptr = ERR37;
647      break;      break;
648    
649      /* \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:
650      is an absolute backreference. If negative, it is a relative backreference.  
651      This is a Perl 5.10 feature. Perl 5.10 also supports \g{name} as a      (1) A number, either plain or braced. If positive, it is an absolute
652      reference to a named group. This is part of Perl's movement towards a      backreference. If negative, it is a relative backreference. This is a Perl
653      unified syntax for back references. As this is synonymous with \k{name}, we      5.10 feature.
654      fudge it up by pretending it really was \k. */  
655        (2) Perl 5.10 also supports \g{name} as a reference to a named group. This
656        is part of Perl's movement towards a unified syntax for back references. As
657        this is synonymous with \k{name}, we fudge it up by pretending it really
658        was \k.
659    
660        (3) For Oniguruma compatibility we also support \g followed by a name or a
661        number either in angle brackets or in single quotes. However, these are
662        (possibly recursive) subroutine calls, _not_ backreferences. Just return
663        the -ESC_g code (cf \k). */
664    
665      case 'g':      case CHAR_g:
666      if (ptr[1] == '{')      if (ptr[1] == CHAR_LESS_THAN_SIGN || ptr[1] == CHAR_APOSTROPHE)
667          {
668          c = -ESC_g;
669          break;
670          }
671    
672        /* Handle the Perl-compatible cases */
673    
674        if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
675        {        {
676        const uschar *p;        const uschar *p;
677        for (p = ptr+2; *p != 0 && *p != '}'; p++)        for (p = ptr+2; *p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET; p++)
678          if (*p != '-' && (digitab[*p] & ctype_digit) == 0) break;          if (*p != CHAR_MINUS && (digitab[*p] & ctype_digit) == 0) break;
679        if (*p != 0 && *p != '}')        if (*p != 0 && *p != CHAR_RIGHT_CURLY_BRACKET)
680          {          {
681          c = -ESC_k;          c = -ESC_k;
682          break;          break;
# Line 489  else Line 686  else
686        }        }
687      else braced = FALSE;      else braced = FALSE;
688    
689      if (ptr[1] == '-')      if (ptr[1] == CHAR_MINUS)
690        {        {
691        negated = TRUE;        negated = TRUE;
692        ptr++;        ptr++;
# Line 498  else Line 695  else
695    
696      c = 0;      c = 0;
697      while ((digitab[ptr[1]] & ctype_digit) != 0)      while ((digitab[ptr[1]] & ctype_digit) != 0)
698        c = c * 10 + *(++ptr) - '0';        c = c * 10 + *(++ptr) - CHAR_0;
699    
700        if (c < 0)   /* Integer overflow */
701          {
702          *errorcodeptr = ERR61;
703          break;
704          }
705    
706      if (c == 0 || (braced && *(++ptr) != '}'))      if (braced && *(++ptr) != CHAR_RIGHT_CURLY_BRACKET)
707        {        {
708        *errorcodeptr = ERR57;        *errorcodeptr = ERR57;
709        return 0;        break;
710          }
711    
712        if (c == 0)
713          {
714          *errorcodeptr = ERR58;
715          break;
716        }        }
717    
718      if (negated)      if (negated)
# Line 511  else Line 720  else
720        if (c > bracount)        if (c > bracount)
721          {          {
722          *errorcodeptr = ERR15;          *errorcodeptr = ERR15;
723          return 0;          break;
724          }          }
725        c = bracount - (c - 1);        c = bracount - (c - 1);
726        }        }
# Line 531  else Line 740  else
740      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
741      character class, \ followed by a digit is always an octal number. */      character class, \ followed by a digit is always an octal number. */
742    
743      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:
744      case '6': case '7': case '8': case '9':      case CHAR_6: case CHAR_7: case CHAR_8: case CHAR_9:
745    
746      if (!isclass)      if (!isclass)
747        {        {
748        oldptr = ptr;        oldptr = ptr;
749        c -= '0';        c -= CHAR_0;
750        while ((digitab[ptr[1]] & ctype_digit) != 0)        while ((digitab[ptr[1]] & ctype_digit) != 0)
751          c = c * 10 + *(++ptr) - '0';          c = c * 10 + *(++ptr) - CHAR_0;
752          if (c < 0)    /* Integer overflow */
753            {
754            *errorcodeptr = ERR61;
755            break;
756            }
757        if (c < 10 || c <= bracount)        if (c < 10 || c <= bracount)
758          {          {
759          c = -(ESC_REF + c);          c = -(ESC_REF + c);
# Line 552  else Line 766  else
766      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.
767      Thus we have to pull back the pointer by one. */      Thus we have to pull back the pointer by one. */
768    
769      if ((c = *ptr) >= '8')      if ((c = *ptr) >= CHAR_8)
770        {        {
771        ptr--;        ptr--;
772        c = 0;        c = 0;
# Line 565  else Line 779  else
779      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
780      than 3 octal digits. */      than 3 octal digits. */
781    
782      case '0':      case CHAR_0:
783      c -= '0';      c -= CHAR_0;
784      while(i++ < 2 && ptr[1] >= '0' && ptr[1] <= '7')      while(i++ < 2 && ptr[1] >= CHAR_0 && ptr[1] <= CHAR_7)
785          c = c * 8 + *(++ptr) - '0';          c = c * 8 + *(++ptr) - CHAR_0;
786      if (!utf8 && c > 255) *errorcodeptr = ERR51;      if (!utf8 && c > 255) *errorcodeptr = ERR51;
787      break;      break;
788    
# Line 576  else Line 790  else
790      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
791      treated as a data character. */      treated as a data character. */
792    
793      case 'x':      case CHAR_x:
794      if (ptr[1] == '{')      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
795        {        {
796        const uschar *pt = ptr + 2;        const uschar *pt = ptr + 2;
797        int count = 0;        int count = 0;
# Line 586  else Line 800  else
800        while ((digitab[*pt] & ctype_xdigit) != 0)        while ((digitab[*pt] & ctype_xdigit) != 0)
801          {          {
802          register int cc = *pt++;          register int cc = *pt++;
803          if (c == 0 && cc == '0') continue;     /* Leading zeroes */          if (c == 0 && cc == CHAR_0) continue;     /* Leading zeroes */
804          count++;          count++;
805    
806  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
807          if (cc >= 'a') cc -= 32;               /* Convert to upper case */          if (cc >= CHAR_a) cc -= 32;               /* Convert to upper case */
808          c = (c << 4) + cc - ((cc < 'A')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
809  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
810          if (cc >= 'a' && cc <= 'z') cc += 64;  /* Convert to upper case */          if (cc >= CHAR_a && cc <= CHAR_z) cc += 64;  /* Convert to upper case */
811          c = (c << 4) + cc - ((cc >= '0')? '0' : ('A' - 10));          c = (c << 4) + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
812  #endif  #endif
813          }          }
814    
815        if (*pt == '}')        if (*pt == CHAR_RIGHT_CURLY_BRACKET)
816          {          {
817          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;          if (c < 0 || count > (utf8? 8 : 2)) *errorcodeptr = ERR34;
818          ptr = pt;          ptr = pt;
# Line 614  else Line 828  else
828      c = 0;      c = 0;
829      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)      while (i++ < 2 && (digitab[ptr[1]] & ctype_xdigit) != 0)
830        {        {
831        int cc;                               /* Some compilers don't like ++ */        int cc;                                  /* Some compilers don't like */
832        cc = *(++ptr);                        /* in initializers */        cc = *(++ptr);                           /* ++ in initializers */
833  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
834        if (cc >= 'a') cc -= 32;              /* Convert to upper case */        if (cc >= CHAR_a) cc -= 32;              /* Convert to upper case */
835        c = c * 16 + cc - ((cc < 'A')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc < CHAR_A)? CHAR_0 : (CHAR_A - 10));
836  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
837        if (cc <= 'z') cc += 64;              /* Convert to upper case */        if (cc <= CHAR_z) cc += 64;              /* Convert to upper case */
838        c = c * 16 + cc - ((cc >= '0')? '0' : ('A' - 10));        c = c * 16 + cc - ((cc >= CHAR_0)? CHAR_0 : (CHAR_A - 10));
839  #endif  #endif
840        }        }
841      break;      break;
# Line 630  else Line 844  else
844      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
845      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */      ASCII-specific. (However, an EBCDIC equivalent has now been added.) */
846    
847      case 'c':      case CHAR_c:
848      c = *(++ptr);      c = *(++ptr);
849      if (c == 0)      if (c == 0)
850        {        {
851        *errorcodeptr = ERR2;        *errorcodeptr = ERR2;
852        return 0;        break;
853        }        }
854    
855  #ifndef EBCDIC  /* ASCII coding */  #ifndef EBCDIC  /* ASCII/UTF-8 coding */
856      if (c >= 'a' && c <= 'z') c -= 32;      if (c >= CHAR_a && c <= CHAR_z) c -= 32;
857      c ^= 0x40;      c ^= 0x40;
858  #else           /* EBCDIC coding */  #else           /* EBCDIC coding */
859      if (c >= 'a' && c <= 'z') c += 64;      if (c >= CHAR_a && c <= CHAR_z) c += 64;
860      c ^= 0xC0;      c ^= 0xC0;
861  #endif  #endif
862      break;      break;
863    
864      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any      /* PCRE_EXTRA enables extensions to Perl in the matter of escapes. Any
865      other alphameric following \ is an error if PCRE_EXTRA was set; otherwise,      other alphanumeric following \ is an error if PCRE_EXTRA was set;
866      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
867      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
868      in future, so I haven't "optimized" it. */      be again in future, so I haven't "optimized" it. */
869    
870      default:      default:
871      if ((options & PCRE_EXTRA) != 0) switch(c)      if ((options & PCRE_EXTRA) != 0) switch(c)
# Line 664  else Line 878  else
878      }      }
879    }    }
880    
881    /* Perl supports \N{name} for character names, as well as plain \N for "not
882    newline". PCRE does not support \N{name}. */
883    
884    if (c == -ESC_N && ptr[1] == CHAR_LEFT_CURLY_BRACKET)
885      *errorcodeptr = ERR37;
886    
887    /* If PCRE_UCP is set, we change the values for \d etc. */
888    
889    if ((options & PCRE_UCP) != 0 && c <= -ESC_D && c >= -ESC_w)
890      c -= (ESC_DU - ESC_D);
891    
892    /* Set the pointer to the final character before returning. */
893    
894  *ptrptr = ptr;  *ptrptr = ptr;
895  return c;  return c;
896  }  }
# Line 704  if (c == 0) goto ERROR_RETURN; Line 931  if (c == 0) goto ERROR_RETURN;
931  /* \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
932  negation. */  negation. */
933    
934  if (c == '{')  if (c == CHAR_LEFT_CURLY_BRACKET)
935    {    {
936    if (ptr[1] == '^')    if (ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
937      {      {
938      *negptr = TRUE;      *negptr = TRUE;
939      ptr++;      ptr++;
# Line 715  if (c == '{') Line 942  if (c == '{')
942      {      {
943      c = *(++ptr);      c = *(++ptr);
944      if (c == 0) goto ERROR_RETURN;      if (c == 0) goto ERROR_RETURN;
945      if (c == '}') break;      if (c == CHAR_RIGHT_CURLY_BRACKET) break;
946      name[i] = c;      name[i] = c;
947      }      }
948    if (c !='}') goto ERROR_RETURN;    if (c != CHAR_RIGHT_CURLY_BRACKET) goto ERROR_RETURN;
949    name[i] = 0;    name[i] = 0;
950    }    }
951    
# Line 740  top = _pcre_utt_size; Line 967  top = _pcre_utt_size;
967  while (bot < top)  while (bot < top)
968    {    {
969    i = (bot + top) >> 1;    i = (bot + top) >> 1;
970    c = strcmp(name, _pcre_utt[i].name);    c = strcmp(name, _pcre_utt_names + _pcre_utt[i].name_offset);
971    if (c == 0)    if (c == 0)
972      {      {
973      *dptr = _pcre_utt[i].value;      *dptr = _pcre_utt[i].value;
# Line 783  is_counted_repeat(const uschar *p) Line 1010  is_counted_repeat(const uschar *p)
1010  {  {
1011  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1012  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1013  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1014    
1015  if (*p++ != ',') return FALSE;  if (*p++ != CHAR_COMMA) return FALSE;
1016  if (*p == '}') return TRUE;  if (*p == CHAR_RIGHT_CURLY_BRACKET) return TRUE;
1017    
1018  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;  if ((digitab[*p++] & ctype_digit) == 0) return FALSE;
1019  while ((digitab[*p] & ctype_digit) != 0) p++;  while ((digitab[*p] & ctype_digit) != 0) p++;
1020    
1021  return (*p == '}');  return (*p == CHAR_RIGHT_CURLY_BRACKET);
1022  }  }
1023    
1024    
# Line 824  int max = -1; Line 1051  int max = -1;
1051  /* 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
1052  an integer overflow. */  an integer overflow. */
1053    
1054  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - '0';  while ((digitab[*p] & ctype_digit) != 0) min = min * 10 + *p++ - CHAR_0;
1055  if (min < 0 || min > 65535)  if (min < 0 || min > 65535)
1056    {    {
1057    *errorcodeptr = ERR5;    *errorcodeptr = ERR5;
# Line 834  if (min < 0 || min > 65535) Line 1061  if (min < 0 || min > 65535)
1061  /* 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.
1062  Also, max must not be less than min. */  Also, max must not be less than min. */
1063    
1064  if (*p == '}') max = min; else  if (*p == CHAR_RIGHT_CURLY_BRACKET) max = min; else
1065    {    {
1066    if (*(++p) != '}')    if (*(++p) != CHAR_RIGHT_CURLY_BRACKET)
1067      {      {
1068      max = 0;      max = 0;
1069      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - '0';      while((digitab[*p] & ctype_digit) != 0) max = max * 10 + *p++ - CHAR_0;
1070      if (max < 0 || max > 65535)      if (max < 0 || max > 65535)
1071        {        {
1072        *errorcodeptr = ERR5;        *errorcodeptr = ERR5;
# Line 864  return p; Line 1091  return p;
1091    
1092    
1093  /*************************************************  /*************************************************
1094  *       Find forward referenced subpattern       *  *  Subroutine for finding forward reference      *
1095  *************************************************/  *************************************************/
1096    
1097  /* This function scans along a pattern's text looking for capturing  /* This recursive function is called only from find_parens() below. The
1098    top-level call starts at the beginning of the pattern. All other calls must
1099    start at a parenthesis. It scans along a pattern's text looking for capturing
1100  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
1101  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
1102  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
1103  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
1104  be terminated by '>' because that is checked in the first pass.  first pass. Recursion is used to keep track of subpatterns that reset the
1105    capturing group numbers - the (?| feature.
1106    
1107  Arguments:  Arguments:
1108    ptr          current position in the pattern    ptrptr       address of the current character pointer (updated)
1109    count        current count of capturing parens so far encountered    cd           compile background data
1110    name         name to seek, or NULL if seeking a numbered subpattern    name         name to seek, or NULL if seeking a numbered subpattern
1111    lorn         name length, or subpattern number if name is NULL    lorn         name length, or subpattern number if name is NULL
1112    xmode        TRUE if we are in /x mode    xmode        TRUE if we are in /x mode
1113      utf8         TRUE if we are in UTF-8 mode
1114      count        pointer to the current capturing subpattern number (updated)
1115    
1116  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1117  */  */
1118    
1119  static int  static int
1120  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,
1121    BOOL xmode)    BOOL xmode, BOOL utf8, int *count)
1122  {  {
1123  const uschar *thisname;  uschar *ptr = *ptrptr;
1124    int start_count = *count;
1125    int hwm_count = start_count;
1126    BOOL dup_parens = FALSE;
1127    
1128  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1129    dealing with. The very first call may not start with a parenthesis. */
1130    
1131    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1132    {    {
1133    int term;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1134    
1135      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1136    
1137      /* Handle a normal, unnamed capturing parenthesis. */
1138    
1139      else if (ptr[1] != CHAR_QUESTION_MARK)
1140        {
1141        *count += 1;
1142        if (name == NULL && *count == lorn) return *count;
1143        ptr++;
1144        }
1145    
1146      /* All cases now have (? at the start. Remember when we are in a group
1147      where the parenthesis numbers are duplicated. */
1148    
1149      else if (ptr[2] == CHAR_VERTICAL_LINE)
1150        {
1151        ptr += 3;
1152        dup_parens = TRUE;
1153        }
1154    
1155      /* Handle comments; all characters are allowed until a ket is reached. */
1156    
1157      else if (ptr[2] == CHAR_NUMBER_SIGN)
1158        {
1159        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1160        goto FAIL_EXIT;
1161        }
1162    
1163      /* Handle a condition. If it is an assertion, just carry on so that it
1164      is processed as normal. If not, skip to the closing parenthesis of the
1165      condition (there can't be any nested parens). */
1166    
1167      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1168        {
1169        ptr += 2;
1170        if (ptr[1] != CHAR_QUESTION_MARK)
1171          {
1172          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1173          if (*ptr != 0) ptr++;
1174          }
1175        }
1176    
1177      /* Start with (? but not a condition. */
1178    
1179      else
1180        {
1181        ptr += 2;
1182        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1183    
1184        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1185    
1186        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1187            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1188          {
1189          int term;
1190          const uschar *thisname;
1191          *count += 1;
1192          if (name == NULL && *count == lorn) return *count;
1193          term = *ptr++;
1194          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1195          thisname = ptr;
1196          while (*ptr != term) ptr++;
1197          if (name != NULL && lorn == ptr - thisname &&
1198              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1199            return *count;
1200          term++;
1201          }
1202        }
1203      }
1204    
1205    /* Past any initial parenthesis handling, scan for parentheses or vertical
1206    bars. */
1207    
1208    for (; *ptr != 0; ptr++)
1209      {
1210    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1211    
1212    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1213      {      {
1214      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1215      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1216        {        {
1217        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1218        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1219        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1220        }        }
1221      continue;      continue;
1222      }      }
1223    
1224    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1225      are handled for real. If the first character is '^', skip it. Also, if the
1226      first few characters (either before or after ^) are \Q\E or \E we skip them
1227      too. This makes for compatibility with Perl. Note the use of STR macros to
1228      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1229    
1230    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1231      {      {
1232      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1233        for (;;)
1234        {        {
1235        if (*ptr == '\\')        if (ptr[1] == CHAR_BACKSLASH)
1236            {
1237            if (ptr[2] == CHAR_E)
1238              ptr+= 2;
1239            else if (strncmp((const char *)ptr+2,
1240                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1241              ptr += 4;
1242            else
1243              break;
1244            }
1245          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1246          {          {
1247          if (*(++ptr) == 0) return -1;          negate_class = TRUE;
1248          if (*ptr == 'Q') for (;;)          ptr++;
1249            }
1250          else break;
1251          }
1252    
1253        /* If the next character is ']', it is a data character that must be
1254        skipped, except in JavaScript compatibility mode. */
1255    
1256        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1257            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1258          ptr++;
1259    
1260        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1261          {
1262          if (*ptr == 0) return -1;
1263          if (*ptr == CHAR_BACKSLASH)
1264            {
1265            if (*(++ptr) == 0) goto FAIL_EXIT;
1266            if (*ptr == CHAR_Q) for (;;)
1267            {            {
1268            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1269            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1270            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1271            }            }
1272          continue;          continue;
1273          }          }
# Line 931  for (; *ptr != 0; ptr++) Line 1277  for (; *ptr != 0; ptr++)
1277    
1278    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1279    
1280    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1281      {      {
1282      while (*(++ptr) != 0 && *ptr != '\n');      ptr++;
1283      if (*ptr == 0) return -1;      while (*ptr != 0)
1284          {
1285          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
1286          ptr++;
1287    #ifdef SUPPORT_UTF8
1288          if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
1289    #endif
1290          }
1291        if (*ptr == 0) goto FAIL_EXIT;
1292      continue;      continue;
1293      }      }
1294    
1295    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1296    
1297    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1298      {      {
1299      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, count);
1300      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1301      continue;      if (*ptr == 0) goto FAIL_EXIT;
1302      }      }
1303    
1304    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1305    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1306        if (dup_parens && *count < hwm_count) *count = hwm_count;
1307        goto FAIL_EXIT;
1308        }
1309    
1310    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1311        {
1312        if (*count > hwm_count) hwm_count = *count;
1313        *count = start_count;
1314        }
1315      }
1316    
1317    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1318         *ptr != '\'')  *ptrptr = ptr;
1319      continue;  return -1;
1320    }
1321    
   count++;  
1322    
1323    if (name == NULL && count == lorn) return count;  
1324    term = *ptr++;  
1325    if (term == '<') term = '>';  /*************************************************
1326    thisname = ptr;  *       Find forward referenced subpattern       *
1327    while (*ptr != term) ptr++;  *************************************************/
1328    if (name != NULL && lorn == ptr - thisname &&  
1329        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  /* This function scans along a pattern's text looking for capturing
1330      return count;  subpatterns, and counting them. If it finds a named pattern that matches the
1331    name it is given, it returns its number. Alternatively, if the name is NULL, it
1332    returns when it reaches a given numbered subpattern. This is used for forward
1333    references to subpatterns. We used to be able to start this scan from the
1334    current compiling point, using the current count value from cd->bracount, and
1335    do it all in a single loop, but the addition of the possibility of duplicate
1336    subpattern numbers means that we have to scan from the very start, in order to
1337    take account of such duplicates, and to use a recursive function to keep track
1338    of the different types of group.
1339    
1340    Arguments:
1341      cd           compile background data
1342      name         name to seek, or NULL if seeking a numbered subpattern
1343      lorn         name length, or subpattern number if name is NULL
1344      xmode        TRUE if we are in /x mode
1345      utf8         TRUE if we are in UTF-8 mode
1346    
1347    Returns:       the number of the found subpattern, or -1 if not found
1348    */
1349    
1350    static int
1351    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode,
1352      BOOL utf8)
1353    {
1354    uschar *ptr = (uschar *)cd->start_pattern;
1355    int count = 0;
1356    int rc;
1357    
1358    /* If the pattern does not start with an opening parenthesis, the first call
1359    to find_parens_sub() will scan right to the end (if necessary). However, if it
1360    does start with a parenthesis, find_parens_sub() will return when it hits the
1361    matching closing parens. That is why we have to have a loop. */
1362    
1363    for (;;)
1364      {
1365      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, utf8, &count);
1366      if (rc > 0 || *ptr++ == 0) break;
1367    }    }
1368    
1369  return -1;  return rc;
1370  }  }
1371    
1372    
1373    
1374    
1375  /*************************************************  /*************************************************
1376  *      Find first significant op code            *  *      Find first significant op code            *
1377  *************************************************/  *************************************************/
# Line 1023  for (;;) Line 1421  for (;;)
1421    
1422      case OP_CALLOUT:      case OP_CALLOUT:
1423      case OP_CREF:      case OP_CREF:
1424        case OP_NCREF:
1425      case OP_RREF:      case OP_RREF:
1426        case OP_NRREF:
1427      case OP_DEF:      case OP_DEF:
1428      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1429      break;      break;
# Line 1039  for (;;) Line 1439  for (;;)
1439    
1440    
1441  /*************************************************  /*************************************************
1442  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1443  *************************************************/  *************************************************/
1444    
1445  /* 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,
1446  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.
1447  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
1448    temporarily terminated with OP_END when this function is called.
1449    
1450    This function is called when a backward assertion is encountered, so that if it
1451    fails, the error message can point to the correct place in the pattern.
1452    However, we cannot do this when the assertion contains subroutine calls,
1453    because they can be forward references. We solve this by remembering this case
1454    and doing the check at the end; a flag specifies which mode we are running in.
1455    
1456  Arguments:  Arguments:
1457    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1458    options  the compiling options    options  the compiling options
1459      atend    TRUE if called when the pattern is complete
1460      cd       the "compile data" structure
1461    
1462  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1463                 or -1 if there is no fixed length,
1464               or -2 if \C was encountered               or -2 if \C was encountered
1465                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1466  */  */
1467    
1468  static int  static int
1469  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1470  {  {
1471  int length = -1;  int length = -1;
1472    
# Line 1068  branch, check the length against that of Line 1479  branch, check the length against that of
1479  for (;;)  for (;;)
1480    {    {
1481    int d;    int d;
1482      uschar *ce, *cs;
1483    register int op = *cc;    register int op = *cc;
   
1484    switch (op)    switch (op)
1485      {      {
1486      case OP_CBRA:      case OP_CBRA:
1487      case OP_BRA:      case OP_BRA:
1488      case OP_ONCE:      case OP_ONCE:
1489      case OP_COND:      case OP_COND:
1490      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1491      if (d < 0) return d;      if (d < 0) return d;
1492      branchlength += d;      branchlength += d;
1493      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1099  for (;;) Line 1510  for (;;)
1510      branchlength = 0;      branchlength = 0;
1511      break;      break;
1512    
1513        /* A true recursion implies not fixed length, but a subroutine call may
1514        be OK. If the subroutine is a forward reference, we can't deal with
1515        it until the end of the pattern, so return -3. */
1516    
1517        case OP_RECURSE:
1518        if (!atend) return -3;
1519        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1520        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1521        if (cc > cs && cc < ce) return -1;                /* Recursion */
1522        d = find_fixedlength(cs + 2, options, atend, cd);
1523        if (d < 0) return d;
1524        branchlength += d;
1525        cc += 1 + LINK_SIZE;
1526        break;
1527    
1528      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1529    
1530      case OP_ASSERT:      case OP_ASSERT:
# Line 1112  for (;;) Line 1538  for (;;)
1538    
1539      case OP_REVERSE:      case OP_REVERSE:
1540      case OP_CREF:      case OP_CREF:
1541        case OP_NCREF:
1542      case OP_RREF:      case OP_RREF:
1543        case OP_NRREF:
1544      case OP_DEF:      case OP_DEF:
1545      case OP_OPT:      case OP_OPT:
1546      case OP_CALLOUT:      case OP_CALLOUT:
1547      case OP_SOD:      case OP_SOD:
1548      case OP_SOM:      case OP_SOM:
1549        case OP_SET_SOM:
1550      case OP_EOD:      case OP_EOD:
1551      case OP_EODN:      case OP_EODN:
1552      case OP_CIRC:      case OP_CIRC:
# Line 1135  for (;;) Line 1564  for (;;)
1564      branchlength++;      branchlength++;
1565      cc += 2;      cc += 2;
1566  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1567      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1568        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1569  #endif  #endif
1570      break;      break;
1571    
# Line 1149  for (;;) Line 1576  for (;;)
1576      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1577      cc += 4;      cc += 4;
1578  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1579      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1580        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1581  #endif  #endif
1582      break;      break;
1583    
1584      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1585      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1586        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1587      cc += 4;      cc += 4;
1588      break;      break;
1589    
# Line 1175  for (;;) Line 1601  for (;;)
1601      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1602      case OP_WORDCHAR:      case OP_WORDCHAR:
1603      case OP_ANY:      case OP_ANY:
1604        case OP_ALLANY:
1605      branchlength++;      branchlength++;
1606      cc++;      cc++;
1607      break;      break;
# Line 1229  for (;;) Line 1656  for (;;)
1656    
1657    
1658  /*************************************************  /*************************************************
1659  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1660  *************************************************/  *************************************************/
1661    
1662  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1663  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1664    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1665    so that it can be called from pcre_study() when finding the minimum matching
1666    length.
1667    
1668  Arguments:  Arguments:
1669    code        points to start of expression    code        points to start of expression
1670    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1671    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1672    
1673  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
1674  */  */
1675    
1676  static const uschar *  const uschar *
1677  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1678  {  {
1679  for (;;)  for (;;)
1680    {    {
# Line 1257  for (;;) Line 1687  for (;;)
1687    
1688    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1689    
1690      /* Handle recursion */
1691    
1692      else if (c == OP_REVERSE)
1693        {
1694        if (number < 0) return (uschar *)code;
1695        code += _pcre_OP_lengths[c];
1696        }
1697    
1698    /* Handle capturing bracket */    /* Handle capturing bracket */
1699    
1700    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1266  for (;;) Line 1704  for (;;)
1704      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1705      }      }
1706    
1707    /* 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
1708    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
1709    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1710      must add in its length. */
1711    
1712    else    else
1713      {      {
1714        switch(c)
1715          {
1716          case OP_TYPESTAR:
1717          case OP_TYPEMINSTAR:
1718          case OP_TYPEPLUS:
1719          case OP_TYPEMINPLUS:
1720          case OP_TYPEQUERY:
1721          case OP_TYPEMINQUERY:
1722          case OP_TYPEPOSSTAR:
1723          case OP_TYPEPOSPLUS:
1724          case OP_TYPEPOSQUERY:
1725          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1726          break;
1727    
1728          case OP_TYPEUPTO:
1729          case OP_TYPEMINUPTO:
1730          case OP_TYPEEXACT:
1731          case OP_TYPEPOSUPTO:
1732          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1733          break;
1734    
1735          case OP_MARK:
1736          case OP_PRUNE_ARG:
1737          case OP_SKIP_ARG:
1738          code += code[1];
1739          break;
1740    
1741          case OP_THEN_ARG:
1742          code += code[1+LINK_SIZE];
1743          break;
1744          }
1745    
1746        /* Add in the fixed length from the table */
1747    
1748      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1749    
1750      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1751      a multi-byte character. The length in the table is a minimum, so we have to
1752      arrange to skip the extra bytes. */
1753    
1754  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1755      if (utf8) switch(c)      if (utf8) switch(c)
1756        {        {
# Line 1294  for (;;) Line 1772  for (;;)
1772        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1773        break;        break;
1774        }        }
1775    #else
1776        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1777  #endif  #endif
1778      }      }
1779    }    }
# Line 1330  for (;;) Line 1810  for (;;)
1810    
1811    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1812    
1813    /* 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
1814    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
1815    The length in the table is a minimum, so we have to arrange to skip the extra    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1816    bytes. */    must add in its length. */
1817    
1818    else    else
1819      {      {
1820        switch(c)
1821          {
1822          case OP_TYPESTAR:
1823          case OP_TYPEMINSTAR:
1824          case OP_TYPEPLUS:
1825          case OP_TYPEMINPLUS:
1826          case OP_TYPEQUERY:
1827          case OP_TYPEMINQUERY:
1828          case OP_TYPEPOSSTAR:
1829          case OP_TYPEPOSPLUS:
1830          case OP_TYPEPOSQUERY:
1831          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1832          break;
1833    
1834          case OP_TYPEPOSUPTO:
1835          case OP_TYPEUPTO:
1836          case OP_TYPEMINUPTO:
1837          case OP_TYPEEXACT:
1838          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1839          break;
1840    
1841          case OP_MARK:
1842          case OP_PRUNE_ARG:
1843          case OP_SKIP_ARG:
1844          code += code[1];
1845          break;
1846    
1847          case OP_THEN_ARG:
1848          code += code[1+LINK_SIZE];
1849          break;
1850          }
1851    
1852        /* Add in the fixed length from the table */
1853    
1854      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1855    
1856        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1857        by a multi-byte character. The length in the table is a minimum, so we have
1858        to arrange to skip the extra bytes. */
1859    
1860  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1861      if (utf8) switch(c)      if (utf8) switch(c)
1862        {        {
# Line 1359  for (;;) Line 1878  for (;;)
1878        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1879        break;        break;
1880        }        }
1881    #else
1882        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1883  #endif  #endif
1884      }      }
1885    }    }
# Line 1374  for (;;) Line 1895  for (;;)
1895  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()
1896  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
1897  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1898  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
1899  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
1900    bracket whose current branch will already have been scanned.
1901    
1902  Arguments:  Arguments:
1903    code        points to start of search    code        points to start of search
1904    endcode     points to where to stop    endcode     points to where to stop
1905    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1906      cd          contains pointers to tables etc.
1907    
1908  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1909  */  */
1910    
1911  static BOOL  static BOOL
1912  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1913      compile_data *cd)
1914  {  {
1915  register int c;  register int c;
1916  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);  for (code = first_significant_code(code + _pcre_OP_lengths[*code], NULL, 0, TRUE);
# Line 1397  for (code = first_significant_code(code Line 1921  for (code = first_significant_code(code
1921    
1922    c = *code;    c = *code;
1923    
1924      /* Skip over forward assertions; the other assertions are skipped by
1925      first_significant_code() with a TRUE final argument. */
1926    
1927      if (c == OP_ASSERT)
1928        {
1929        do code += GET(code, 1); while (*code == OP_ALT);
1930        c = *code;
1931        continue;
1932        }
1933    
1934    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1935    
1936    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1937      {      {
1938      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1939      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1407  for (code = first_significant_code(code Line 1941  for (code = first_significant_code(code
1941      continue;      continue;
1942      }      }
1943    
1944      /* For a recursion/subroutine call, if its end has been reached, which
1945      implies a subroutine call, we can scan it. */
1946    
1947      if (c == OP_RECURSE)
1948        {
1949        BOOL empty_branch = FALSE;
1950        const uschar *scode = cd->start_code + GET(code, 1);
1951        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1952        do
1953          {
1954          if (could_be_empty_branch(scode, endcode, utf8, cd))
1955            {
1956            empty_branch = TRUE;
1957            break;
1958            }
1959          scode += GET(scode, 1);
1960          }
1961        while (*scode == OP_ALT);
1962        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1963        continue;
1964        }
1965    
1966    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1967    
1968    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1969      {      {
1970      BOOL empty_branch;      BOOL empty_branch;
1971      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1972    
1973      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1974        empty branch, so just skip over the conditional, because it could be empty.
1975        Otherwise, scan the individual branches of the group. */
1976    
1977      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;  
1978        code += GET(code, 1);        code += GET(code, 1);
1979        else
1980          {
1981          empty_branch = FALSE;
1982          do
1983            {
1984            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1985              empty_branch = TRUE;
1986            code += GET(code, 1);
1987            }
1988          while (*code == OP_ALT);
1989          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1990        }        }
1991      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1992      c = *code;      c = *code;
1993      continue;      continue;
1994      }      }
# Line 1433  for (code = first_significant_code(code Line 1997  for (code = first_significant_code(code
1997    
1998    switch (c)    switch (c)
1999      {      {
2000      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
2001        cannot be represented just by a bit map. This includes negated single
2002        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
2003        actual length is stored in the compiled code, so we must update "code"
2004        here. */
2005    
2006  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2007      case OP_XCLASS:      case OP_XCLASS:
2008      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
2009      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
2010  #endif  #endif
2011    
# Line 1481  for (code = first_significant_code(code Line 2049  for (code = first_significant_code(code
2049      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2050      case OP_WORDCHAR:      case OP_WORDCHAR:
2051      case OP_ANY:      case OP_ANY:
2052        case OP_ALLANY:
2053      case OP_ANYBYTE:      case OP_ANYBYTE:
2054      case OP_CHAR:      case OP_CHAR:
2055      case OP_CHARNC:      case OP_CHARNC:
# Line 1499  for (code = first_significant_code(code Line 2068  for (code = first_significant_code(code
2068      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2069      return FALSE;      return FALSE;
2070    
2071        /* These are going to continue, as they may be empty, but we have to
2072        fudge the length for the \p and \P cases. */
2073    
2074        case OP_TYPESTAR:
2075        case OP_TYPEMINSTAR:
2076        case OP_TYPEPOSSTAR:
2077        case OP_TYPEQUERY:
2078        case OP_TYPEMINQUERY:
2079        case OP_TYPEPOSQUERY:
2080        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2081        break;
2082    
2083        /* Same for these */
2084    
2085        case OP_TYPEUPTO:
2086        case OP_TYPEMINUPTO:
2087        case OP_TYPEPOSUPTO:
2088        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2089        break;
2090    
2091      /* End of branch */      /* End of branch */
2092    
2093      case OP_KET:      case OP_KET:
# Line 1517  for (code = first_significant_code(code Line 2106  for (code = first_significant_code(code
2106      case OP_QUERY:      case OP_QUERY:
2107      case OP_MINQUERY:      case OP_MINQUERY:
2108      case OP_POSQUERY:      case OP_POSQUERY:
2109        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2110        break;
2111    
2112      case OP_UPTO:      case OP_UPTO:
2113      case OP_MINUPTO:      case OP_MINUPTO:
2114      case OP_POSUPTO:      case OP_POSUPTO:
2115      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2116      break;      break;
2117  #endif  #endif
2118    
2119        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2120        string. */
2121    
2122        case OP_MARK:
2123        case OP_PRUNE_ARG:
2124        case OP_SKIP_ARG:
2125        code += code[1];
2126        break;
2127    
2128        case OP_THEN_ARG:
2129        code += code[1+LINK_SIZE];
2130        break;
2131    
2132        /* None of the remaining opcodes are required to match a character. */
2133    
2134        default:
2135        break;
2136      }      }
2137    }    }
2138    
# Line 1545  Arguments: Line 2155  Arguments:
2155    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2156    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2157    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2158      cd          pointers to tables etc
2159    
2160  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2161  */  */
2162    
2163  static BOOL  static BOOL
2164  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2165    BOOL utf8)    BOOL utf8, compile_data *cd)
2166  {  {
2167  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2168    {    {
2169    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2170        return FALSE;
2171    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2172    }    }
2173  return TRUE;  return TRUE;
# Line 1568  return TRUE; Line 2180  return TRUE;
2180  *************************************************/  *************************************************/
2181    
2182  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2183  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
2184  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2185  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2186    
2187    Originally, this function only recognized a sequence of letters between the
2188    terminators, but it seems that Perl recognizes any sequence of characters,
2189    though of course unknown POSIX names are subsequently rejected. Perl gives an
2190    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2191    didn't consider this to be a POSIX class. Likewise for [:1234:].
2192    
2193    The problem in trying to be exactly like Perl is in the handling of escapes. We
2194    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2195    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2196    below handles the special case of \], but does not try to do any other escape
2197    processing. This makes it different from Perl for cases such as [:l\ower:]
2198    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2199    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2200    I think.
2201    
2202  Argument:  Arguments:
2203    ptr      pointer to the initial [    ptr      pointer to the initial [
2204    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2205    
2206  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2207  */  */
2208    
2209  static BOOL  static BOOL
2210  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2211  {  {
2212  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2213  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2214  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2215    {    {
2216    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2217    return TRUE;      {
2218        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2219        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2220          {
2221          *endptr = ptr;
2222          return TRUE;
2223          }
2224        }
2225    }    }
2226  return FALSE;  return FALSE;
2227  }  }
# Line 1615  Returns:     a value representing the na Line 2246  Returns:     a value representing the na
2246  static int  static int
2247  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2248  {  {
2249    const char *pn = posix_names;
2250  register int yield = 0;  register int yield = 0;
2251  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2252    {    {
2253    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2254      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2255      pn += posix_name_lengths[yield] + 1;
2256    yield++;    yield++;
2257    }    }
2258  return -1;  return -1;
# Line 1634  return -1; Line 2267  return -1;
2267  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2268  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2269  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
2270  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
2271  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
2272  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
2273  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
2274  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2275    OP_END.
2276    
2277  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2278  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 1661  adjust_recurse(uschar *group, int adjust Line 2295  adjust_recurse(uschar *group, int adjust
2295    uschar *save_hwm)    uschar *save_hwm)
2296  {  {
2297  uschar *ptr = group;  uschar *ptr = group;
2298    
2299  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2300    {    {
2301    int offset;    int offset;
# Line 1714  auto_callout(uschar *code, const uschar Line 2349  auto_callout(uschar *code, const uschar
2349  {  {
2350  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2351  *code++ = 255;  *code++ = 255;
2352  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2353  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2354  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2355  }  }
2356    
# Line 1740  Returns:             nothing Line 2375  Returns:             nothing
2375  static void  static void
2376  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2377  {  {
2378  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2379  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2380  }  }
2381    
# Line 1772  get_othercase_range(unsigned int *cptr, Line 2407  get_othercase_range(unsigned int *cptr,
2407  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2408    
2409  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2410    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2411    
2412  if (c > d) return FALSE;  if (c > d) return FALSE;
2413    
# Line 1781  next = othercase + 1; Line 2416  next = othercase + 1;
2416    
2417  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2418    {    {
2419    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2420    next++;    next++;
2421    }    }
2422    
# Line 1790  for (++c; c <= d; c++) Line 2425  for (++c; c <= d; c++)
2425    
2426  return TRUE;  return TRUE;
2427  }  }
2428    
2429    
2430    
2431    /*************************************************
2432    *        Check a character and a property        *
2433    *************************************************/
2434    
2435    /* This function is called by check_auto_possessive() when a property item
2436    is adjacent to a fixed character.
2437    
2438    Arguments:
2439      c            the character
2440      ptype        the property type
2441      pdata        the data for the type
2442      negated      TRUE if it's a negated property (\P or \p{^)
2443    
2444    Returns:       TRUE if auto-possessifying is OK
2445    */
2446    
2447    static BOOL
2448    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2449    {
2450    const ucd_record *prop = GET_UCD(c);
2451    switch(ptype)
2452      {
2453      case PT_LAMP:
2454      return (prop->chartype == ucp_Lu ||
2455              prop->chartype == ucp_Ll ||
2456              prop->chartype == ucp_Lt) == negated;
2457    
2458      case PT_GC:
2459      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2460    
2461      case PT_PC:
2462      return (pdata == prop->chartype) == negated;
2463    
2464      case PT_SC:
2465      return (pdata == prop->script) == negated;
2466    
2467      /* These are specials */
2468    
2469      case PT_ALNUM:
2470      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2471              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2472    
2473      case PT_SPACE:    /* Perl space */
2474      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2475              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2476              == negated;
2477    
2478      case PT_PXSPACE:  /* POSIX space */
2479      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2480              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2481              c == CHAR_FF || c == CHAR_CR)
2482              == negated;
2483    
2484      case PT_WORD:
2485      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2486              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2487              c == CHAR_UNDERSCORE) == negated;
2488      }
2489    return FALSE;
2490    }
2491  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2492    
2493    
# Line 1803  whether the next thing could possibly ma Line 2501  whether the next thing could possibly ma
2501  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2502    
2503  Arguments:  Arguments:
2504    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2505    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2506    ptr           next character in pattern    ptr           next character in pattern
2507    options       options bits    options       options bits
2508    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1815  Returns:        TRUE if possessifying is Line 2511  Returns:        TRUE if possessifying is
2511  */  */
2512    
2513  static BOOL  static BOOL
2514  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2515    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2516  {  {
2517  int next;  int c, next;
2518    int op_code = *previous++;
2519    
2520  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2521    
# Line 1827  if ((options & PCRE_EXTENDED) != 0) Line 2524  if ((options & PCRE_EXTENDED) != 0)
2524    for (;;)    for (;;)
2525      {      {
2526      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2527      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2528        {        {
2529        while (*(++ptr) != 0)        ptr++;
2530          while (*ptr != 0)
2531            {
2532          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2533            ptr++;
2534    #ifdef SUPPORT_UTF8
2535            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2536    #endif
2537            }
2538        }        }
2539      else break;      else break;
2540      }      }
# Line 1839  if ((options & PCRE_EXTENDED) != 0) Line 2543  if ((options & PCRE_EXTENDED) != 0)
2543  /* 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
2544  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2545    
2546  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2547    {    {
2548    int temperrorcode = 0;    int temperrorcode = 0;
2549    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1864  if ((options & PCRE_EXTENDED) != 0) Line 2568  if ((options & PCRE_EXTENDED) != 0)
2568    for (;;)    for (;;)
2569      {      {
2570      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2571      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2572        {        {
2573        while (*(++ptr) != 0)        ptr++;
2574          while (*ptr != 0)
2575            {
2576          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
2577            ptr++;
2578    #ifdef SUPPORT_UTF8
2579            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
2580    #endif
2581            }
2582        }        }
2583      else break;      else break;
2584      }      }
# Line 1875  if ((options & PCRE_EXTENDED) != 0) Line 2586  if ((options & PCRE_EXTENDED) != 0)
2586    
2587  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2588    
2589  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2590    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2591        return FALSE;
 /* Now compare the next item with the previous opcode. If the previous is a  
 positive single character match, "item" either contains the character or, if  
 "item" is greater than 127 in utf8 mode, the character's bytes are in  
 utf8_char. */  
   
2592    
2593  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2594    the next item is a character. */
2595    
2596  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2597    {    {
2598    case OP_CHAR:    case OP_CHAR:
2599  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2600    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2601    #else
2602      c = *previous;
2603  #endif  #endif
2604    return item != next;    return c != next;
2605    
2606    /* For CHARNC (caseless character) we must check the other case. If we have    /* For CHARNC (caseless character) we must check the other case. If we have
2607    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
# Line 1900  if (next >= 0) switch(op_code) Line 2609  if (next >= 0) switch(op_code)
2609    
2610    case OP_CHARNC:    case OP_CHARNC:
2611  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2612    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2613    #else
2614      c = *previous;
2615  #endif  #endif
2616    if (item == next) return FALSE;    if (c == next) return FALSE;
2617  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2618    if (utf8)    if (utf8)
2619      {      {
2620      unsigned int othercase;      unsigned int othercase;
2621      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2622  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2623      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2624  #else  #else
2625      othercase = NOTACHAR;      othercase = NOTACHAR;
2626  #endif  #endif
2627      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2628      }      }
2629    else    else
2630  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2631    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2632    
2633    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2634    
2635    case OP_NOT:    case OP_NOT:
2636    if (next < 0) return FALSE;  /* Not a character */    if ((c = *previous) == next) return TRUE;
   if (item == next) return TRUE;  
2637    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2638  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2639    if (utf8)    if (utf8)
# Line 1931  if (next >= 0) switch(op_code) Line 2641  if (next >= 0) switch(op_code)
2641      unsigned int othercase;      unsigned int othercase;
2642      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2643  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2644      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2645  #else  #else
2646      othercase = NOTACHAR;      othercase = NOTACHAR;
2647  #endif  #endif
2648      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2649      }      }
2650    else    else
2651  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2652    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2653    
2654      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2655      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2656    
2657    case OP_DIGIT:    case OP_DIGIT:
2658    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 1982  if (next >= 0) switch(op_code) Line 2695  if (next >= 0) switch(op_code)
2695      case 0x202f:      case 0x202f:
2696      case 0x205f:      case 0x205f:
2697      case 0x3000:      case 0x3000:
2698      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2699      default:      default:
2700      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2701      }      }
2702    
2703      case OP_ANYNL:
2704    case OP_VSPACE:    case OP_VSPACE:
2705    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2706    switch(next)    switch(next)
# Line 1998  if (next >= 0) switch(op_code) Line 2712  if (next >= 0) switch(op_code)
2712      case 0x85:      case 0x85:
2713      case 0x2028:      case 0x2028:
2714      case 0x2029:      case 0x2029:
2715      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2716      default:      default:
2717      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2718      }      }
2719    
2720    #ifdef SUPPORT_UCP
2721      case OP_PROP:
2722      return check_char_prop(next, previous[0], previous[1], FALSE);
2723    
2724      case OP_NOTPROP:
2725      return check_char_prop(next, previous[0], previous[1], TRUE);
2726    #endif
2727    
2728    default:    default:
2729    return FALSE;    return FALSE;
2730    }    }
2731    
2732    
2733  /* Handle the case when the next item is \d, \s, etc. */  /* Handle the case when the next item is \d, \s, etc. Note that when PCRE_UCP
2734    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2735    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2736    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2737    replaced by OP_PROP codes when PCRE_UCP is set. */
2738    
2739  switch(op_code)  switch(op_code)
2740    {    {
2741    case OP_CHAR:    case OP_CHAR:
2742    case OP_CHARNC:    case OP_CHARNC:
2743  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2744    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2745    #else
2746      c = *previous;
2747  #endif  #endif
2748    switch(-next)    switch(-next)
2749      {      {
2750      case ESC_d:      case ESC_d:
2751      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2752    
2753      case ESC_D:      case ESC_D:
2754      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2755    
2756      case ESC_s:      case ESC_s:
2757      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2758    
2759      case ESC_S:      case ESC_S:
2760      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2761    
2762      case ESC_w:      case ESC_w:
2763      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2764    
2765      case ESC_W:      case ESC_W:
2766      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2767    
2768      case ESC_h:      case ESC_h:
2769      case ESC_H:      case ESC_H:
2770      switch(item)      switch(c)
2771        {        {
2772        case 0x09:        case 0x09:
2773        case 0x20:        case 0x20:
# Line 2067  switch(op_code) Line 2795  switch(op_code)
2795    
2796      case ESC_v:      case ESC_v:
2797      case ESC_V:      case ESC_V:
2798      switch(item)      switch(c)
2799        {        {
2800        case 0x0a:        case 0x0a:
2801        case 0x0b:        case 0x0b:
# Line 2081  switch(op_code) Line 2809  switch(op_code)
2809        return -next == ESC_v;        return -next == ESC_v;
2810        }        }
2811    
2812        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2813        their substitutions and process them. The result will always be either
2814        -ESC_p or -ESC_P. Then fall through to process those values. */
2815    
2816    #ifdef SUPPORT_UCP
2817        case ESC_du:
2818        case ESC_DU:
2819        case ESC_wu:
2820        case ESC_WU:
2821        case ESC_su:
2822        case ESC_SU:
2823          {
2824          int temperrorcode = 0;
2825          ptr = substitutes[-next - ESC_DU];
2826          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2827          if (temperrorcode != 0) return FALSE;
2828          ptr++;    /* For compatibility */
2829          }
2830        /* Fall through */
2831    
2832        case ESC_p:
2833        case ESC_P:
2834          {
2835          int ptype, pdata, errorcodeptr;
2836          BOOL negated;
2837    
2838          ptr--;      /* Make ptr point at the p or P */
2839          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2840          if (ptype < 0) return FALSE;
2841          ptr++;      /* Point past the final curly ket */
2842    
2843          /* If the property item is optional, we have to give up. (When generated
2844          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2845          to the original \d etc. At this point, ptr will point to a zero byte. */
2846    
2847          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2848            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2849              return FALSE;
2850    
2851          /* Do the property check. */
2852    
2853          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2854          }
2855    #endif
2856    
2857      default:      default:
2858      return FALSE;      return FALSE;
2859      }      }
2860    
2861      /* In principle, support for Unicode properties should be integrated here as
2862      well. It means re-organizing the above code so as to get hold of the property
2863      values before switching on the op-code. However, I wonder how many patterns
2864      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2865      these op-codes are never generated.) */
2866    
2867    case OP_DIGIT:    case OP_DIGIT:
2868    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2869           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2870    
2871    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2872    return next == -ESC_d;    return next == -ESC_d;
2873    
2874    case OP_WHITESPACE:    case OP_WHITESPACE:
2875    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2876    
2877    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2878    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2879    
2880    case OP_HSPACE:    case OP_HSPACE:
2881    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2882             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2883    
2884    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2885    return next == -ESC_h;    return next == -ESC_h;
2886    
2887    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2888      case OP_ANYNL:
2889    case OP_VSPACE:    case OP_VSPACE:
2890    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2891    
2892    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2893    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2894    
2895    case OP_WORDCHAR:    case OP_WORDCHAR:
2896    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2897             next == -ESC_v || next == -ESC_R;
2898    
2899    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2900    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2176  BOOL inescq = FALSE; Line 2958  BOOL inescq = FALSE;
2958  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2959  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2960  const uschar *tempptr;  const uschar *tempptr;
2961    const uschar *nestptr = NULL;
2962  uschar *previous = NULL;  uschar *previous = NULL;
2963  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2964  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2185  uschar classbits[32]; Line 2968  uschar classbits[32];
2968  BOOL class_utf8;  BOOL class_utf8;
2969  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2970  uschar *class_utf8data;  uschar *class_utf8data;
2971    uschar *class_utf8data_base;
2972  uschar utf8_char[6];  uschar utf8_char[6];
2973  #else  #else
2974  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2975  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2976  #endif  #endif
2977    
2978  #ifdef DEBUG  #ifdef PCRE_DEBUG
2979  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2980  #endif  #endif
2981    
# Line 2224  req_caseopt = ((options & PCRE_CASELESS) Line 3008  req_caseopt = ((options & PCRE_CASELESS)
3008  for (;; ptr++)  for (;; ptr++)
3009    {    {
3010    BOOL negate_class;    BOOL negate_class;
3011      BOOL should_flip_negation;
3012    BOOL possessive_quantifier;    BOOL possessive_quantifier;
3013    BOOL is_quantifier;    BOOL is_quantifier;
3014    BOOL is_recurse;    BOOL is_recurse;
# Line 2244  for (;; ptr++) Line 3029  for (;; ptr++)
3029    
3030    c = *ptr;    c = *ptr;
3031    
3032      /* If we are at the end of a nested substitution, revert to the outer level
3033      string. Nesting only happens one level deep. */
3034    
3035      if (c == 0 && nestptr != NULL)
3036        {
3037        ptr = nestptr;
3038        nestptr = NULL;
3039        c = *ptr;
3040        }
3041    
3042    /* If we are in the pre-compile phase, accumulate the length used for the    /* If we are in the pre-compile phase, accumulate the length used for the
3043    previous cycle of this loop. */    previous cycle of this loop. */
3044    
3045    if (lengthptr != NULL)    if (lengthptr != NULL)
3046      {      {
3047  #ifdef DEBUG  #ifdef PCRE_DEBUG
3048      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3049  #endif  #endif
3050      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3051        {        {
3052        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3053        goto FAILED;        goto FAILED;
# Line 2274  for (;; ptr++) Line 3069  for (;; ptr++)
3069        goto FAILED;        goto FAILED;
3070        }        }
3071    
3072      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3073      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));
3074    
3075      /* If "previous" is set and it is not at the start of the work space, move      /* If "previous" is set and it is not at the start of the work space, move
# Line 2301  for (;; ptr++) Line 3096  for (;; ptr++)
3096    /* In the real compile phase, just check the workspace used by the forward    /* In the real compile phase, just check the workspace used by the forward
3097    reference list. */    reference list. */
3098    
3099    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3100      {      {
3101      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3102      goto FAILED;      goto FAILED;
# Line 2311  for (;; ptr++) Line 3106  for (;; ptr++)
3106    
3107    if (inescq && c != 0)    if (inescq && c != 0)
3108      {      {
3109      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3110        {        {
3111        inescq = FALSE;        inescq = FALSE;
3112        ptr++;        ptr++;
# Line 2337  for (;; ptr++) Line 3132  for (;; ptr++)
3132    /* 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
3133    a quantifier. */    a quantifier. */
3134    
3135    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3136      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3137        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3138    
3139    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3140         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2353  for (;; ptr++) Line 3149  for (;; ptr++)
3149    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3150      {      {
3151      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3152      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3153        {        {
3154        while (*(++ptr) != 0)        ptr++;
3155          while (*ptr != 0)
3156          {          {
3157          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen - 1; break; }
3158            ptr++;
3159    #ifdef SUPPORT_UTF8
3160            if (utf8) while ((*ptr & 0xc0) == 0x80) ptr++;
3161    #endif
3162          }          }
3163        if (*ptr != 0) continue;        if (*ptr != 0) continue;
3164    
# Line 2378  for (;; ptr++) Line 3179  for (;; ptr++)
3179      {      {
3180      /* ===================================================================*/      /* ===================================================================*/
3181      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3182      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3183      case ')':      case CHAR_RIGHT_PARENTHESIS:
3184      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3185      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3186      *codeptr = code;      *codeptr = code;
# Line 2391  for (;; ptr++) Line 3192  for (;; ptr++)
3192          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3193          goto FAILED;          goto FAILED;
3194          }          }
3195        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3196        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3197        }        }
3198      return TRUE;      return TRUE;
# Line 2401  for (;; ptr++) Line 3202  for (;; ptr++)
3202      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3203      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3204    
3205      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3206      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3207        {        {
3208        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2410  for (;; ptr++) Line 3211  for (;; ptr++)
3211      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3212      break;      break;
3213    
3214      case '$':      case CHAR_DOLLAR_SIGN:
3215      previous = NULL;      previous = NULL;
3216      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3217      break;      break;
# Line 2418  for (;; ptr++) Line 3219  for (;; ptr++)
3219      /* 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
3220      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3221    
3222      case '.':      case CHAR_DOT:
3223      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3224      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3225      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3226      previous = code;      previous = code;
3227      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3228      break;      break;
3229    
3230    
# Line 2438  for (;; ptr++) Line 3239  for (;; ptr++)
3239      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,
3240      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3241      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.
     */  
3242    
3243      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3244        default (Perl) mode, it is treated as a data character. */
3245    
3246        case CHAR_RIGHT_SQUARE_BRACKET:
3247        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3248          {
3249          *errorcodeptr = ERR64;
3250          goto FAILED;
3251          }
3252        goto NORMAL_CHAR;
3253    
3254        case CHAR_LEFT_SQUARE_BRACKET:
3255      previous = code;      previous = code;
3256    
3257      /* 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
3258      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. */
3259    
3260      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3261          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3262            check_posix_syntax(ptr, &tempptr))
3263        {        {
3264        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3265        goto FAILED;        goto FAILED;
3266        }        }
3267    
3268      /* If the first character is '^', set the negation flag and skip it. Also,      /* If the first character is '^', set the negation flag and skip it. Also,
3269      if the first few characters (either before or after ^) are \Q\E or \E we      if the first few characters (either before or after ^) are \Q\E or \E we
3270      skip them too. This makes for compatibility with Perl. */      skip them too. This makes for compatibility with Perl. */
3271    
3272      negate_class = FALSE;      negate_class = FALSE;
3273      for (;;)      for (;;)
3274        {        {
3275        c = *(++ptr);        c = *(++ptr);
3276        if (c == '\\')        if (c == CHAR_BACKSLASH)
3277          {          {
3278          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3279            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3280              else break;          else if (strncmp((const char *)ptr+1,
3281                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3282              ptr += 3;
3283            else
3284              break;
3285          }          }
3286        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3287          negate_class = TRUE;          negate_class = TRUE;
3288        else break;        else break;
3289        }        }
3290    
3291        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3292        an initial ']' is taken as a data character -- the code below handles
3293        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3294        [^] must match any character, so generate OP_ALLANY. */
3295    
3296        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3297            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3298          {
3299          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3300          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3301          zerofirstbyte = firstbyte;
3302          break;
3303          }
3304    
3305        /* If a class contains a negative special such as \S, we need to flip the
3306        negation flag at the end, so that support for characters > 255 works
3307        correctly (they are all included in the class). */
3308    
3309        should_flip_negation = FALSE;
3310    
3311      /* 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
3312      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
# Line 2489  for (;; ptr++) Line 3325  for (;; ptr++)
3325  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3326      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3327      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3328        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3329  #endif  #endif
3330    
3331      /* 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 2504  for (;; ptr++) Line 3341  for (;; ptr++)
3341          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3342          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3343          }          }
3344    
3345          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3346          data and reset the pointer. This is so that very large classes that
3347          contain a zillion UTF-8 characters no longer overwrite the work space
3348          (which is on the stack). */
3349    
3350          if (lengthptr != NULL)
3351            {
3352            *lengthptr += class_utf8data - class_utf8data_base;
3353            class_utf8data = class_utf8data_base;
3354            }
3355    
3356  #endif  #endif
3357    
3358        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3359    
3360        if (inescq)        if (inescq)
3361          {          {
3362          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3363            {            {
3364            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3365            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2525  for (;; ptr++) Line 3374  for (;; ptr++)
3374        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3375        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3376    
3377        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3378            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3379            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3380          {          {
3381          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3382          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3383          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3384          uschar pbits[32];          uschar pbits[32];
3385    
3386          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3387            {            {
3388            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3389            goto FAILED;            goto FAILED;
3390            }            }
3391    
3392          ptr += 2;          ptr += 2;
3393          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3394            {            {
3395            local_negate = TRUE;            local_negate = TRUE;
3396              should_flip_negation = TRUE;  /* Note negative special */
3397            ptr++;            ptr++;
3398            }            }
3399    
3400          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3401          if (posix_class < 0)          if (posix_class < 0)
3402            {            {
3403            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2561  for (;; ptr++) Line 3411  for (;; ptr++)
3411          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3412            posix_class = 0;            posix_class = 0;
3413    
3414          /* We build the bit map for the POSIX class in a chunk of local store          /* When PCRE_UCP is set, some of the POSIX classes are converted to
3415          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3416          subtract bits that may be in the main map already. At the end we or the  
3417          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3418            if ((options & PCRE_UCP) != 0)
3419              {
3420              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3421              if (posix_substitutes[pc] != NULL)
3422                {
3423                nestptr = tempptr + 1;
3424                ptr = posix_substitutes[pc] - 1;
3425                continue;
3426                }
3427              }
3428    #endif
3429            /* In the non-UCP case, we build the bit map for the POSIX class in a
3430            chunk of local store because we may be adding and subtracting from it,
3431            and we don't want to subtract bits that may be in the main map already.
3432            At the end we or the result into the bit map that is being built. */
3433    
3434          posix_class *= 3;          posix_class *= 3;
3435    
# Line 2608  for (;; ptr++) Line 3473  for (;; ptr++)
3473    
3474        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3475        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
3476        case. Inside a class (and only there) it is treated as backspace.        case. Inside a class (and only there) it is treated as backspace. We
3477        Elsewhere it marks a word boundary. Other escapes have preset maps ready        assume that other escapes have more than one character in them, so set
3478        to 'or' into the one we are building. We assume they have more than one        class_charcount bigger than one. Unrecognized escapes fall through and
3479        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3480          PCRE_EXTRA is set. */
3481    
3482        if (c == '\\')        if (c == CHAR_BACKSLASH)
3483          {          {
3484          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3485          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3486    
3487          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 */
         else if (-c == ESC_X) c = 'X';   /* \X is literal X in a class */  
         else if (-c == ESC_R) c = 'R';   /* \R is literal R in a class */  
3488          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3489            {            {
3490            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3491              {              {
3492              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3493              }              }
3494            else inescq = TRUE;            else inescq = TRUE;
3495            continue;            continue;
3496            }            }
3497            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3498    
3499          if (c < 0)          if (c < 0)
3500            {            {
3501            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3502            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3503    
3504            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3505              {              {
3506    #ifdef SUPPORT_UCP
3507                case ESC_du:     /* These are the values given for \d etc */
3508                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3509                case ESC_wu:     /* escape sequence with an appropriate \p */
3510                case ESC_WU:     /* or \P to test Unicode properties instead */
3511                case ESC_su:     /* of the default ASCII testing. */
3512                case ESC_SU:
3513                nestptr = ptr;
3514                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3515                class_charcount -= 2;                /* Undo! */
3516                continue;
3517    #endif
3518              case ESC_d:              case ESC_d:
3519              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3520              continue;              continue;
3521    
3522              case ESC_D:              case ESC_D:
3523                should_flip_negation = TRUE;
3524              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3525              continue;              continue;
3526    
# Line 2653  for (;; ptr++) Line 3529  for (;; ptr++)
3529              continue;              continue;
3530    
3531              case ESC_W:              case ESC_W:
3532                should_flip_negation = TRUE;
3533              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3534              continue;              continue;
3535    
3536                /* Perl 5.004 onwards omits VT from \s, but we must preserve it
3537                if it was previously set by something earlier in the character
3538                class. */
3539    
3540              case ESC_s:              case ESC_s:
3541              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_space];              classbits[0] |= cbits[cbit_space];
3542              classbits[1] &= ~0x08;   /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= cbits[cbit_space+1] & ~0x08;
3543                for (c = 2; c < 32; c++) classbits[c] |= cbits[c+cbit_space];
3544              continue;              continue;
3545    
3546              case ESC_S:              case ESC_S:
3547                should_flip_negation = TRUE;
3548              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3549              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3550              continue;              continue;
3551    
3552              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
             continue;  
   
             default:    /* Not recognized; fall through */  
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3553              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3554              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3555              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2706  for (;; ptr++) Line 3573  for (;; ptr++)
3573                }                }
3574  #endif  #endif
3575              continue;              continue;
             }  
3576    
3577            if (-c == ESC_H)              case ESC_H:
             {  
3578              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3579                {                {
3580                int x = 0xff;                int x = 0xff;
# Line 2751  for (;; ptr++) Line 3616  for (;; ptr++)
3616                }                }
3617  #endif  #endif
3618              continue;              continue;
             }  
3619    
3620            if (-c == ESC_v)              case ESC_v:
             {  
3621              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3622              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3623              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2770  for (;; ptr++) Line 3633  for (;; ptr++)
3633                }                }
3634  #endif  #endif
3635              continue;              continue;
             }  
3636    
3637            if (-c == ESC_V)              case ESC_V:
             {  
3638              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3639                {                {
3640                int x = 0xff;                int x = 0xff;
# Line 2803  for (;; ptr++) Line 3664  for (;; ptr++)
3664                }                }
3665  #endif  #endif
3666              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3667    
3668  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3669            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3670              {              case ESC_P:
3671              BOOL negated;                {
3672              int pdata;                BOOL negated;
3673              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3674              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3675              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3676              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3677                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3678              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3679              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3680              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3681              continue;                class_charcount -= 2;   /* Not a < 256 character */
3682              }                continue;
3683                  }
3684  #endif  #endif
3685            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3686            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3687            treated as literals. */              treated as literals. */
3688    
3689            if ((options & PCRE_EXTRA) != 0)              default:
3690              {              if ((options & PCRE_EXTRA) != 0)
3691              *errorcodeptr = ERR7;                {
3692              goto FAILED;                *errorcodeptr = ERR7;
3693                  goto FAILED;
3694                  }
3695                class_charcount -= 2;  /* Undo the default count from above */
3696                c = *ptr;              /* Get the final character and fall through */
3697                break;
3698              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3699            }            }
3700    
3701          /* Fall through if we have a single character (c >= 0). This may be          /* Fall through if we have a single character (c >= 0). This may be
# Line 2848  for (;; ptr++) Line 3709  for (;; ptr++)
3709        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3710    
3711        CHECK_RANGE:        CHECK_RANGE:
3712        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3713          {          {
3714          inescq = FALSE;          inescq = FALSE;
3715          ptr += 2;          ptr += 2;
# Line 2856  for (;; ptr++) Line 3717  for (;; ptr++)
3717    
3718        oldptr = ptr;        oldptr = ptr;
3719    
3720        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3721    
3722          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3723    
3724          /* Check for range */
3725    
3726          if (!inescq && ptr[1] == CHAR_MINUS)
3727          {          {
3728          int d;          int d;
3729          ptr += 2;          ptr += 2;
3730          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3731    
3732          /* 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
3733          mode. */          mode. */
3734    
3735          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3736            {            {
3737            ptr += 2;            ptr += 2;
3738            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3739                { ptr += 2; continue; }
3740            inescq = TRUE;            inescq = TRUE;
3741            break;            break;
3742            }            }
3743    
3744          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3745            {            {
3746            ptr = oldptr;            ptr = oldptr;
3747            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2892  for (;; ptr++) Line 3760  for (;; ptr++)
3760          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
3761          in such circumstances. */          in such circumstances. */
3762    
3763          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3764            {            {
3765            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3766            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3767    
3768            /* \b is backslash; \X is literal X; \R is literal R; any other            /* \b is backspace; any other special means the '-' was literal */
           special means the '-' was literal */  
3769    
3770            if (d < 0)            if (d < 0)
3771              {              {
3772              if (d == -ESC_b) d = '\b';              if (d == -ESC_b) d = CHAR_BS; else
             else if (d == -ESC_X) d = 'X';  
             else if (d == -ESC_R) d = 'R'; else  
3773                {                {
3774                ptr = oldptr;                ptr = oldptr;
3775                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2923  for (;; ptr++) Line 3788  for (;; ptr++)
3788    
3789          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3790    
3791            /* Remember \r or \n */
3792    
3793            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3794    
3795          /* 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
3796          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3797          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3041  for (;; ptr++) Line 3910  for (;; ptr++)
3910          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3911            {            {
3912            unsigned int othercase;            unsigned int othercase;
3913            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3914              {              {
3915              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3916              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3066  for (;; ptr++) Line 3935  for (;; ptr++)
3935          }          }
3936        }        }
3937    
3938      /* Loop until ']' reached. This "while" is the end of the "do" above. */      /* Loop until ']' reached. This "while" is the end of the "do" far above.
3939        If we are at the end of an internal nested string, revert to the outer
3940        string. */
3941    
3942        while (((c = *(++ptr)) != 0 ||
3943               (nestptr != NULL &&
3944                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3945               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3946    
3947      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3948    
3949      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3950        {        {
3951        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3952        goto FAILED;        goto FAILED;
3953        }        }
3954    
3955      /* 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
3956      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
3957      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
3958      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3959      single-bytes only. This is an historical hangover. Maybe one day we can  
3960      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3961        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3962        operate on single-bytes only. This is an historical hangover. Maybe one day
3963        we can tidy these opcodes to handle multi-byte characters.
3964    
3965      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
3966      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 3091  for (;; ptr++) Line 3970  for (;; ptr++)
3970      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3971    
3972  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3973      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3974            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3975  #else  #else
3976      if (class_charcount == 1)      if (class_charcount == 1)
3977  #endif  #endif
# Line 3137  for (;; ptr++) Line 4014  for (;; ptr++)
4014      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
4015    
4016      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
4017      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
4018      we can omit the bitmap in the actual compiled code. */      such as \S in the class, and PCRE_UCP is not set, because in that case all
4019        characters > 255 are in the class, so any that were explicitly given as
4020        well can be ignored. If (when there are explicit characters > 255 that must
4021        be listed) there are no characters < 256, we can omit the bitmap in the
4022        actual compiled code. */
4023    
4024  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
4025      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
4026        {        {
4027        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
4028        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3167  for (;; ptr++) Line 4048  for (;; ptr++)
4048        }        }
4049  #endif  #endif
4050    
4051      /* If there are no characters > 255, negate the 32-byte map if necessary,      /* If there are no characters > 255, or they are all to be included or
4052      and copy it into the code vector. If this is the first thing in the branch,      excluded, set the opcode to OP_CLASS or OP_NCLASS, depending on whether the
4053      there can be no first char setting, whatever the repeat count. Any reqbyte      whole class was negated and whether there were negative specials such as \S
4054      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4055        negating it if necessary. */
4056    
4057        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4058      if (negate_class)      if (negate_class)
4059        {        {
       *code++ = OP_NCLASS;  
4060        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4061          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4062        }        }
4063      else      else
4064        {        {
       *code++ = OP_CLASS;  
4065        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4066        }        }
4067      code += 32;      code += 32;
# Line 3191  for (;; ptr++) Line 4072  for (;; ptr++)
4072      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4073      has been tested above. */      has been tested above. */
4074    
4075      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4076      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4077      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4078      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4079      goto REPEAT;      goto REPEAT;
4080    
4081      case '*':      case CHAR_ASTERISK:
4082      repeat_min = 0;      repeat_min = 0;
4083      repeat_max = -1;      repeat_max = -1;
4084      goto REPEAT;      goto REPEAT;
4085    
4086      case '+':      case CHAR_PLUS:
4087      repeat_min = 1;      repeat_min = 1;
4088      repeat_max = -1;      repeat_max = -1;
4089      goto REPEAT;      goto REPEAT;
4090    
4091      case '?':      case CHAR_QUESTION_MARK:
4092      repeat_min = 0;      repeat_min = 0;
4093      repeat_max = 1;      repeat_max = 1;
4094    
# Line 3242  for (;; ptr++) Line 4123  for (;; ptr++)
4123      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
4124      repeat type to the non-default. */      repeat type to the non-default. */
4125    
4126      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4127        {        {
4128        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4129        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4130        ptr++;        ptr++;
4131        }        }
4132      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4133        {        {
4134        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4135        ptr++;        ptr++;
# Line 3295  for (;; ptr++) Line 4176  for (;; ptr++)
4176    
4177        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4178            repeat_max < 0 &&            repeat_max < 0 &&
4179            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4180          {          {
4181          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4182          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3317  for (;; ptr++) Line 4197  for (;; ptr++)
4197        c = previous[1];        c = previous[1];
4198        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4199            repeat_max < 0 &&            repeat_max < 0 &&
4200            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4201          {          {
4202          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4203          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3341  for (;; ptr++) Line 4221  for (;; ptr++)
4221    
4222        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4223            repeat_max < 0 &&            repeat_max < 0 &&
4224            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4225          {          {
4226          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4227          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3363  for (;; ptr++) Line 4243  for (;; ptr++)
4243    
4244        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4245    
4246          /*--------------------------------------------------------------------*/
4247          /* This code is obsolete from release 8.00; the restriction was finally
4248          removed: */
4249    
4250        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4251        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4252    
4253        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4254          /*--------------------------------------------------------------------*/
4255    
4256        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4257    
# Line 3513  for (;; ptr++) Line 4398  for (;; ptr++)
4398          goto END_REPEAT;          goto END_REPEAT;
4399          }          }
4400    
4401          /*--------------------------------------------------------------------*/
4402          /* This code is obsolete from release 8.00; the restriction was finally
4403          removed: */
4404    
4405        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4406        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4407    
4408        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4409          /*--------------------------------------------------------------------*/
4410    
4411        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4412          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3541  for (;; ptr++) Line 4431  for (;; ptr++)
4431        {        {
4432        register int i;        register int i;
4433        int ketoffset = 0;        int ketoffset = 0;
4434        int len = code - previous;        int len = (int)(code - previous);
4435        uschar *bralink = NULL;        uschar *bralink = NULL;
4436    
4437        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3562  for (;; ptr++) Line 4452  for (;; ptr++)
4452          {          {
4453          register uschar *ket = previous;          register uschar *ket = previous;
4454          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4455          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4456          }          }
4457    
4458        /* The case of a zero minimum is special because of the need to stick        /* The case of a zero minimum is special because of the need to stick
# Line 3574  for (;; ptr++) Line 4464  for (;; ptr++)
4464    
4465        if (repeat_min == 0)        if (repeat_min == 0)
4466          {          {
4467          /* 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
4468          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4469    
4470          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4471          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4472          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4473          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4474          the start of the whole regex. Temporarily terminate the pattern while          **   }
4475          doing this. */  
4476            However, that fails when a group is referenced as a subroutine from
4477            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4478            so that it is skipped on execution. As we don't have a list of which
4479            groups are referenced, we cannot do this selectively.
4480    
4481            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4482            and do no more at this point. However, we do need to adjust any
4483            OP_RECURSE calls inside the group that refer to the group itself or any
4484            internal or forward referenced group, because the offset is from the
4485            start of the whole regex. Temporarily terminate the pattern while doing
4486            this. */
4487    
4488          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4489            {            {
4490            *code = OP_END;            *code = OP_END;
4491            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4492            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4493            code++;            code++;
4494              if (repeat_max == 0)
4495                {
4496                *previous++ = OP_SKIPZERO;
4497                goto END_REPEAT;
4498                }
4499            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4500            }            }
4501    
# Line 3620  for (;; ptr++) Line 4520  for (;; ptr++)
4520            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4521            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4522    
4523            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4524            bralink = previous;            bralink = previous;
4525            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4526            }            }
# Line 3641  for (;; ptr++) Line 4541  for (;; ptr++)
4541            {            {
4542            /* 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
4543            just adjust the length as if we had. Do some paranoid checks for            just adjust the length as if we had. Do some paranoid checks for
4544            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4545              integer type when available, otherwise double. */
4546    
4547            if (lengthptr != NULL)            if (lengthptr != NULL)
4548              {              {
4549              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4550              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4551                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4552                        (INT64_OR_DOUBLE)INT_MAX ||
4553                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4554                {                {
4555                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3693  for (;; ptr++) Line 4595  for (;; ptr++)
4595          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
4596          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
4597          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4598          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4599            a 64-bit integer type when available, otherwise double. */
4600    
4601          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4602            {            {
4603            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4604                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4605            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4606                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4607                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4608                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4609              {              {
4610              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3726  for (;; ptr++) Line 4629  for (;; ptr++)
4629              {              {
4630              int offset;              int offset;
4631              *code++ = OP_BRA;              *code++ = OP_BRA;
4632              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4633              bralink = code;              bralink = code;
4634              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4635              }              }
# Line 3747  for (;; ptr++) Line 4650  for (;; ptr++)
4650          while (bralink != NULL)          while (bralink != NULL)
4651            {            {
4652            int oldlinkoffset;            int oldlinkoffset;
4653            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4654            uschar *bra = code - offset;            uschar *bra = code - offset;
4655            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4656            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 3778  for (;; ptr++) Line 4681  for (;; ptr++)
4681            uschar *scode = bracode;            uschar *scode = bracode;
4682            do            do
4683              {              {
4684              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4685                {                {
4686                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4687                break;                break;
# Line 3790  for (;; ptr++) Line 4693  for (;; ptr++)
4693          }          }
4694        }        }
4695    
4696        /* If previous is OP_FAIL, it was generated by an empty class [] in
4697        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4698        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4699        error above. We can just ignore the repeat in JS case. */
4700    
4701        else if (*previous == OP_FAIL) goto END_REPEAT;
4702    
4703      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4704    
4705      else      else
# Line 3814  for (;; ptr++) Line 4724  for (;; ptr++)
4724      if (possessive_quantifier)      if (possessive_quantifier)
4725        {        {
4726        int len;        int len;
4727        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4728            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4729            tempcode += _pcre_OP_lengths[*tempcode] +
4730              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4731    
4732          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4733            {
4734          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4735        len = code - tempcode;  #ifdef SUPPORT_UTF8
4736            if (utf8 && tempcode[-1] >= 0xc0)
4737              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4738    #endif
4739            }
4740    
4741          len = (int)(code - tempcode);
4742        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4743          {          {
4744          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 3835  for (;; ptr++) Line 4756  for (;; ptr++)
4756          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4757          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4758    
4759            /* Because we are moving code along, we must ensure that any
4760            pending recursive references are updated. */
4761    
4762          default:          default:
4763            *code = OP_END;
4764            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4765          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4766          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4767          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3860  for (;; ptr++) Line 4786  for (;; ptr++)
4786      /* ===================================================================*/      /* ===================================================================*/
4787      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4788      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4789      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. */  
4790    
4791      case '(':      case CHAR_LEFT_PARENTHESIS:
4792      newoptions = options;      newoptions = options;
4793      skipbytes = 0;      skipbytes = 0;
4794      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4795      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4796      reset_bracount = FALSE;      reset_bracount = FALSE;
4797    
4798      if (*(++ptr) == '?')      /* First deal with various "verbs" that can be introduced by '*'. */
4799    
4800        if (*(++ptr) == CHAR_ASTERISK &&
4801             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4802          {
4803          int i, namelen;
4804          int arglen = 0;
4805          const char *vn = verbnames;
4806          const uschar *name = ptr + 1;
4807          const uschar *arg = NULL;
4808          previous = NULL;
4809          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4810          namelen = (int)(ptr - name);
4811    
4812          if (*ptr == CHAR_COLON)
4813            {
4814            arg = ++ptr;
4815            while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4816              || *ptr == '_') ptr++;
4817            arglen = (int)(ptr - arg);
4818            }
4819    
4820          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4821            {
4822            *errorcodeptr = ERR60;
4823            goto FAILED;
4824            }
4825    
4826          /* Scan the table of verb names */
4827    
4828          for (i = 0; i < verbcount; i++)
4829            {
4830            if (namelen == verbs[i].len &&
4831                strncmp((char *)name, vn, namelen) == 0)
4832              {
4833              /* Check for open captures before ACCEPT */
4834    
4835              if (verbs[i].op == OP_ACCEPT)
4836                {
4837                open_capitem *oc;
4838                cd->had_accept = TRUE;
4839                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4840                  {
4841                  *code++ = OP_CLOSE;
4842                  PUT2INC(code, 0, oc->number);
4843                  }
4844                }
4845    
4846              /* Handle the cases with/without an argument */
4847    
4848              if (arglen == 0)
4849                {
4850                if (verbs[i].op < 0)   /* Argument is mandatory */
4851                  {
4852                  *errorcodeptr = ERR66;
4853                  goto FAILED;
4854                  }
4855                *code = verbs[i].op;
4856                if (*code++ == OP_THEN)
4857                  {
4858                  PUT(code, 0, code - bcptr->current_branch - 1);
4859                  code += LINK_SIZE;
4860                  }
4861                }
4862    
4863              else
4864                {
4865                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4866                  {
4867                  *errorcodeptr = ERR59;
4868                  goto FAILED;
4869                  }
4870                *code = verbs[i].op_arg;
4871                if (*code++ == OP_THEN_ARG)
4872                  {
4873                  PUT(code, 0, code - bcptr->current_branch - 1);
4874                  code += LINK_SIZE;
4875                  }
4876                *code++ = arglen;
4877                memcpy(code, arg, arglen);
4878                code += arglen;
4879                *code++ = 0;
4880                }
4881    
4882              break;  /* Found verb, exit loop */
4883              }
4884    
4885            vn += verbs[i].len + 1;
4886            }
4887    
4888          if (i < verbcount) continue;    /* Successfully handled a verb */
4889          *errorcodeptr = ERR60;          /* Verb not recognized */
4890          goto FAILED;
4891          }
4892    
4893        /* Deal with the extended parentheses; all are introduced by '?', and the
4894        appearance of any of them means that this is not a capturing group. */
4895    
4896        else if (*ptr == CHAR_QUESTION_MARK)
4897        {        {
4898        int i, set, unset, namelen;        int i, set, unset, namelen;
4899        int *optset;        int *optset;
# Line 3880  for (;; ptr++) Line 4902  for (;; ptr++)
4902    
4903        switch (*(++ptr))        switch (*(++ptr))
4904          {          {
4905          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4906          ptr++;          ptr++;
4907          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4908          if (*ptr == 0)          if (*ptr == 0)
4909            {            {
4910            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3892  for (;; ptr++) Line 4914  for (;; ptr++)
4914    
4915    
4916          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4917          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4918          reset_bracount = TRUE;          reset_bracount = TRUE;
4919          /* Fall through */          /* Fall through */
4920    
4921          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4922          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4923          bravalue = OP_BRA;          bravalue = OP_BRA;
4924          ptr++;          ptr++;
4925          break;          break;
4926    
4927    
4928          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4929          case '(':          case CHAR_LEFT_PARENTHESIS:
4930          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4931    
4932          /* 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 3924  for (;; ptr++) Line 4946  for (;; ptr++)
4946          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4947          including assertions, are processed. */          including assertions, are processed. */
4948    
4949          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4950                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4951            break;            break;
4952    
4953          /* 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 3936  for (;; ptr++) Line 4959  for (;; ptr++)
4959    
4960          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4961    
4962          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4963            {            {
4964            terminator = -1;            terminator = -1;
4965            ptr += 2;            ptr += 2;
# Line 3946  for (;; ptr++) Line 4969  for (;; ptr++)
4969          /* 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
4970          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4971    
4972          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4973            {            {
4974            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4975            ptr++;            ptr++;
4976            }            }
4977          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4978            {            {
4979            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4980            ptr++;            ptr++;
4981            }            }
4982          else          else
4983            {            {
4984            terminator = 0;            terminator = 0;
4985            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4986            }            }
4987    
4988          /* We now expect to read a name; any thing else is an error */          /* We now expect to read a name; any thing else is an error */
# Line 3979  for (;; ptr++) Line 5002  for (;; ptr++)
5002            {            {
5003            if (recno >= 0)            if (recno >= 0)
5004              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
5005                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
5006            ptr++;            ptr++;
5007            }            }
5008          namelen = ptr - name;          namelen = (int)(ptr - name);
5009    
5010          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
5011                *ptr++ != CHAR_RIGHT_PARENTHESIS)
5012            {            {
5013            ptr--;      /* Error offset */            ptr--;      /* Error offset */
5014            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4006  for (;; ptr++) Line 5030  for (;; ptr++)
5030              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
5031              goto FAILED;              goto FAILED;
5032              }              }
5033            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
5034                cd->bracount - recno + 1 : recno +cd->bracount;
5035              if (recno <= 0 || recno > cd->final_bracount)
5036              {              {
5037              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
5038              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
5039              }              }
           else recno += cd->bracount;  
5040            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5041            break;            break;
5042            }            }
5043    
5044          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
5045          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
5046            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
5047            except they record that the reference was originally to a name. The
5048            information is used to check duplicate names. */
5049    
5050          slot = cd->name_table;          slot = cd->name_table;
5051          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4036  for (;; ptr++) Line 5060  for (;; ptr++)
5060            {            {
5061            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5062            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5063              code[1+LINK_SIZE]++;
5064            }            }
5065    
5066          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5067    
5068          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5069                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0, utf8)) > 0)
5070            {            {
5071            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5072              code[1+LINK_SIZE]++;
5073            }            }
5074    
5075          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4061  for (;; ptr++) Line 5087  for (;; ptr++)
5087          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
5088          specific group number. */          specific group number. */
5089    
5090          else if (*name == 'R')          else if (*name == CHAR_R)
5091            {            {
5092            recno = 0;            recno = 0;
5093            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4071  for (;; ptr++) Line 5097  for (;; ptr++)
5097                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5098                goto FAILED;                goto FAILED;
5099                }                }
5100              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
5101              }              }
5102            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
5103            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4081  for (;; ptr++) Line 5107  for (;; ptr++)
5107          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
5108          false. */          false. */
5109    
5110          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
5111            {            {
5112            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
5113            skipbytes = 1;            skipbytes = 1;
5114            }            }
5115    
5116          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
5117            in the second pass here, so final_bracount is set. */
5118    
5119          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
5120            {            {
5121            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5122            }            }
# Line 4105  for (;; ptr++) Line 5132  for (;; ptr++)
5132    
5133    
5134          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5135          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5136          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5137          ptr++;          ptr++;
5138          break;          break;
5139    
5140    
5141          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5142          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
5143          ptr++;          ptr++;
5144            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
5145              {
5146              *code++ = OP_FAIL;
5147              previous = NULL;
5148              continue;
5149              }
5150            bravalue = OP_ASSERT_NOT;
5151          break;          break;
5152    
5153    
5154          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5155          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
5156          switch (ptr[1])          switch (ptr[1])
5157            {            {
5158            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5159            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5160            ptr += 2;            ptr += 2;
5161            break;            break;
5162    
5163            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5164            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5165            ptr += 2;            ptr += 2;
5166            break;            break;
# Line 4142  for (;; ptr++) Line 5175  for (;; ptr++)
5175    
5176    
5177          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */