/[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 210 by ph10, Wed Aug 8 14:24:50 2007 UTC revision 533 by ph10, Wed Jun 2 19:02:41 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  /* Table of special "verbs" like (*PRUNE) */  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185    searched linearly. Put all the names into a single string, in order to reduce
186    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    platforms. */
189    
190  typedef struct verbitem {  typedef struct verbitem {
191    const char *name;    int   len;                 /* Length of verb name */
192    int   len;    int   op;                  /* Op when no arg, or -1 if arg mandatory */
193    int   op;    int   op_arg;              /* Op when arg present, or -1 if not allowed */
194  } verbitem;  } verbitem;
195    
196  static verbitem verbs[] = {  static const char verbnames[] =
197    { "ACCEPT", 6, OP_ACCEPT },    "\0"                       /* Empty name is a shorthand for MARK */
198    { "COMMIT", 6, OP_COMMIT },    STRING_MARK0
199    { "F",      1, OP_FAIL },    STRING_ACCEPT0
200    { "FAIL",   4, OP_FAIL },    STRING_COMMIT0
201    { "PRUNE",  5, OP_PRUNE },    STRING_F0
202    { "SKIP",   4, OP_SKIP  },    STRING_FAIL0
203    { "THEN",   4, OP_THEN  }    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 int verbcount = sizeof(verbs)/sizeof(verbitem);  static const int verbcount = sizeof(verbs)/sizeof(verbitem);
   
220    
 /* Tables of names of POSIX character classes and their lengths. The list is  
 terminated by a zero length entry. The first three must be alpha, lower, upper,  
 as this is assumed for handling case independence. */  
221    
222  static const char *const posix_names[] = {  /* Tables of names of POSIX character classes and their lengths. The names are
223    "alpha", "lower", "upper",  now all in a single string, to reduce the number of relocations when a shared
224    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  library is dynamically loaded. The list of lengths is terminated by a zero
225    "print", "punct", "space", "word",  "xdigit" };  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 200  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 207  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    "(*VERB) with an argument is not supported",    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401    /* 60 */    /* 60 */
402    "(*VERB) not recognized"    "(*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 302  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 338  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 419  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 439  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 457  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 482  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 CHAR_g:
666        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      case 'g':      if (ptr[1] == CHAR_LEFT_CURLY_BRACKET)
     if (ptr[1] == '{')  
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 513  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 522  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 535  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 555  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 576  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 589  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 600  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 610  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 638  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 654  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 688  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 728  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 739  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 764  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 807  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 848  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 858  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 888  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      count        pointer to the current capturing subpattern number (updated)
1114    
1115  Returns:       the number of the named subpattern, or -1 if not found  Returns:       the number of the named subpattern, or -1 if not found
1116  */  */
1117    
1118  static int  static int
1119  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,
1120    BOOL xmode)    BOOL xmode, int *count)
1121  {  {
1122  const uschar *thisname;  uschar *ptr = *ptrptr;
1123    int start_count = *count;
1124    int hwm_count = start_count;
1125    BOOL dup_parens = FALSE;
1126    
1127  for (; *ptr != 0; ptr++)  /* If the first character is a parenthesis, check on the type of group we are
1128    dealing with. The very first call may not start with a parenthesis. */
1129    
1130    if (ptr[0] == CHAR_LEFT_PARENTHESIS)
1131    {    {
1132    int term;    if (ptr[1] == CHAR_QUESTION_MARK &&
1133          ptr[2] == CHAR_VERTICAL_LINE)
1134        {
1135        ptr += 3;
1136        dup_parens = TRUE;
1137        }
1138    
1139      /* Handle a normal, unnamed capturing parenthesis */
1140    
1141      else if (ptr[1] != CHAR_QUESTION_MARK && ptr[1] != CHAR_ASTERISK)
1142        {
1143        *count += 1;
1144        if (name == NULL && *count == lorn) return *count;
1145        ptr++;
1146        }
1147    
1148      /* Handle a condition. If it is an assertion, just carry on so that it
1149      is processed as normal. If not, skip to the closing parenthesis of the
1150      condition (there can't be any nested parens. */
1151    
1152      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1153        {
1154        ptr += 2;
1155        if (ptr[1] != CHAR_QUESTION_MARK)
1156          {
1157          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1158          if (*ptr != 0) ptr++;
1159          }
1160        }
1161    
1162      /* We have either (? or (* and not a condition */
1163    
1164      else
1165        {
1166        ptr += 2;
1167        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1168    
1169        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1170    
1171        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1172            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1173          {
1174          int term;
1175          const uschar *thisname;
1176          *count += 1;
1177          if (name == NULL && *count == lorn) return *count;
1178          term = *ptr++;
1179          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1180          thisname = ptr;
1181          while (*ptr != term) ptr++;
1182          if (name != NULL && lorn == ptr - thisname &&
1183              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1184            return *count;
1185          term++;
1186          }
1187        }
1188      }
1189    
1190    /* Past any initial parenthesis handling, scan for parentheses or vertical
1191    bars. */
1192    
1193    for (; *ptr != 0; ptr++)
1194      {
1195    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1196    
1197    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1198      {      {
1199      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1200      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1201        {        {
1202        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1203        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1204        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1205        }        }
1206      continue;      continue;
1207      }      }
1208    
1209    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1210      are handled for real. If the first character is '^', skip it. Also, if the
1211      first few characters (either before or after ^) are \Q\E or \E we skip them
1212      too. This makes for compatibility with Perl. Note the use of STR macros to
1213      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1214    
1215    if (*ptr == '[')    if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1216      {      {
1217      while (*(++ptr) != ']')      BOOL negate_class = FALSE;
1218        for (;;)
1219          {
1220          if (ptr[1] == CHAR_BACKSLASH)
1221            {
1222            if (ptr[2] == CHAR_E)
1223              ptr+= 2;
1224            else if (strncmp((const char *)ptr+2,
1225                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1226              ptr += 4;
1227            else
1228              break;
1229            }
1230          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1231            {
1232            negate_class = TRUE;
1233            ptr++;
1234            }
1235          else break;
1236          }
1237    
1238        /* If the next character is ']', it is a data character that must be
1239        skipped, except in JavaScript compatibility mode. */
1240    
1241        if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1242            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1243          ptr++;
1244    
1245        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1246        {        {
1247        if (*ptr == '\\')        if (*ptr == 0) return -1;
1248          if (*ptr == CHAR_BACKSLASH)
1249          {          {
1250          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1251          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1252            {            {
1253            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1254            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1255            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1256            }            }
1257          continue;          continue;
1258          }          }
# Line 955  for (; *ptr != 0; ptr++) Line 1262  for (; *ptr != 0; ptr++)
1262    
1263    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1264    
1265    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1266      {      {
1267      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1268      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1269      continue;      continue;
1270      }      }
1271    
1272    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1273    
1274    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?' && ptr[1] != '*')  
1275      {      {
1276      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1277      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1278      continue;      if (*ptr == 0) goto FAIL_EXIT;
1279        }
1280    
1281      else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1282        {
1283        if (dup_parens && *count < hwm_count) *count = hwm_count;
1284        *ptrptr = ptr;
1285        return -1;
1286      }      }
1287    
1288    ptr += 2;    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1289    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1290        if (*count > hwm_count) hwm_count = *count;
1291        *count = start_count;
1292        }
1293      }
1294    
1295    /* We have to disambiguate (?<! and (?<= from (?<name> */  FAIL_EXIT:
1296    *ptrptr = ptr;
1297    return -1;
1298    }
1299    
   if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  
        *ptr != '\'')  
     continue;  
1300    
   count++;  
1301    
1302    if (name == NULL && count == lorn) return count;  
1303    term = *ptr++;  /*************************************************
1304    if (term == '<') term = '>';  *       Find forward referenced subpattern       *
1305    thisname = ptr;  *************************************************/
1306    while (*ptr != term) ptr++;  
1307    if (name != NULL && lorn == ptr - thisname &&  /* This function scans along a pattern's text looking for capturing
1308        strncmp((const char *)name, (const char *)thisname, lorn) == 0)  subpatterns, and counting them. If it finds a named pattern that matches the
1309      return count;  name it is given, it returns its number. Alternatively, if the name is NULL, it
1310    returns when it reaches a given numbered subpattern. This is used for forward
1311    references to subpatterns. We used to be able to start this scan from the
1312    current compiling point, using the current count value from cd->bracount, and
1313    do it all in a single loop, but the addition of the possibility of duplicate
1314    subpattern numbers means that we have to scan from the very start, in order to
1315    take account of such duplicates, and to use a recursive function to keep track
1316    of the different types of group.
1317    
1318    Arguments:
1319      cd           compile background data
1320      name         name to seek, or NULL if seeking a numbered subpattern
1321      lorn         name length, or subpattern number if name is NULL
1322      xmode        TRUE if we are in /x mode
1323    
1324    Returns:       the number of the found subpattern, or -1 if not found
1325    */
1326    
1327    static int
1328    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1329    {
1330    uschar *ptr = (uschar *)cd->start_pattern;
1331    int count = 0;
1332    int rc;
1333    
1334    /* If the pattern does not start with an opening parenthesis, the first call
1335    to find_parens_sub() will scan right to the end (if necessary). However, if it
1336    does start with a parenthesis, find_parens_sub() will return when it hits the
1337    matching closing parens. That is why we have to have a loop. */
1338    
1339    for (;;)
1340      {
1341      rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1342      if (rc > 0 || *ptr++ == 0) break;
1343    }    }
1344    
1345  return -1;  return rc;
1346  }  }
1347    
1348    
1349    
1350    
1351  /*************************************************  /*************************************************
1352  *      Find first significant op code            *  *      Find first significant op code            *
1353  *************************************************/  *************************************************/
# Line 1047  for (;;) Line 1397  for (;;)
1397    
1398      case OP_CALLOUT:      case OP_CALLOUT:
1399      case OP_CREF:      case OP_CREF:
1400        case OP_NCREF:
1401      case OP_RREF:      case OP_RREF:
1402        case OP_NRREF:
1403      case OP_DEF:      case OP_DEF:
1404      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1405      break;      break;
# Line 1063  for (;;) Line 1415  for (;;)
1415    
1416    
1417  /*************************************************  /*************************************************
1418  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1419  *************************************************/  *************************************************/
1420    
1421  /* 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,
1422  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.
1423  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
1424    temporarily terminated with OP_END when this function is called.
1425    
1426    This function is called when a backward assertion is encountered, so that if it
1427    fails, the error message can point to the correct place in the pattern.
1428    However, we cannot do this when the assertion contains subroutine calls,
1429    because they can be forward references. We solve this by remembering this case
1430    and doing the check at the end; a flag specifies which mode we are running in.
1431    
1432  Arguments:  Arguments:
1433    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1434    options  the compiling options    options  the compiling options
1435      atend    TRUE if called when the pattern is complete
1436      cd       the "compile data" structure
1437    
1438  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1439                 or -1 if there is no fixed length,
1440               or -2 if \C was encountered               or -2 if \C was encountered
1441                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1442  */  */
1443    
1444  static int  static int
1445  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1446  {  {
1447  int length = -1;  int length = -1;
1448    
# Line 1092  branch, check the length against that of Line 1455  branch, check the length against that of
1455  for (;;)  for (;;)
1456    {    {
1457    int d;    int d;
1458      uschar *ce, *cs;
1459    register int op = *cc;    register int op = *cc;
   
1460    switch (op)    switch (op)
1461      {      {
1462      case OP_CBRA:      case OP_CBRA:
1463      case OP_BRA:      case OP_BRA:
1464      case OP_ONCE:      case OP_ONCE:
1465      case OP_COND:      case OP_COND:
1466      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1467      if (d < 0) return d;      if (d < 0) return d;
1468      branchlength += d;      branchlength += d;
1469      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1123  for (;;) Line 1486  for (;;)
1486      branchlength = 0;      branchlength = 0;
1487      break;      break;
1488    
1489        /* A true recursion implies not fixed length, but a subroutine call may
1490        be OK. If the subroutine is a forward reference, we can't deal with
1491        it until the end of the pattern, so return -3. */
1492    
1493        case OP_RECURSE:
1494        if (!atend) return -3;
1495        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1496        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1497        if (cc > cs && cc < ce) return -1;                /* Recursion */
1498        d = find_fixedlength(cs + 2, options, atend, cd);
1499        if (d < 0) return d;
1500        branchlength += d;
1501        cc += 1 + LINK_SIZE;
1502        break;
1503    
1504      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1505    
1506      case OP_ASSERT:      case OP_ASSERT:
# Line 1136  for (;;) Line 1514  for (;;)
1514    
1515      case OP_REVERSE:      case OP_REVERSE:
1516      case OP_CREF:      case OP_CREF:
1517        case OP_NCREF:
1518      case OP_RREF:      case OP_RREF:
1519        case OP_NRREF:
1520      case OP_DEF:      case OP_DEF:
1521      case OP_OPT:      case OP_OPT:
1522      case OP_CALLOUT:      case OP_CALLOUT:
1523      case OP_SOD:      case OP_SOD:
1524      case OP_SOM:      case OP_SOM:
1525        case OP_SET_SOM:
1526      case OP_EOD:      case OP_EOD:
1527      case OP_EODN:      case OP_EODN:
1528      case OP_CIRC:      case OP_CIRC:
# Line 1159  for (;;) Line 1540  for (;;)
1540      branchlength++;      branchlength++;
1541      cc += 2;      cc += 2;
1542  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1543      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1544        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1545  #endif  #endif
1546      break;      break;
1547    
# Line 1173  for (;;) Line 1552  for (;;)
1552      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1553      cc += 4;      cc += 4;
1554  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1555      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1556        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1557  #endif  #endif
1558      break;      break;
1559    
1560      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1561      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1562        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1563      cc += 4;      cc += 4;
1564      break;      break;
1565    
# Line 1199  for (;;) Line 1577  for (;;)
1577      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1578      case OP_WORDCHAR:      case OP_WORDCHAR:
1579      case OP_ANY:      case OP_ANY:
1580        case OP_ALLANY:
1581      branchlength++;      branchlength++;
1582      cc++;      cc++;
1583      break;      break;
# Line 1253  for (;;) Line 1632  for (;;)
1632    
1633    
1634  /*************************************************  /*************************************************
1635  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1636  *************************************************/  *************************************************/
1637    
1638  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1639  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1640    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1641    so that it can be called from pcre_study() when finding the minimum matching
1642    length.
1643    
1644  Arguments:  Arguments:
1645    code        points to start of expression    code        points to start of expression
1646    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1647    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1648    
1649  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
1650  */  */
1651    
1652  static const uschar *  const uschar *
1653  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1654  {  {
1655  for (;;)  for (;;)
1656    {    {
# Line 1281  for (;;) Line 1663  for (;;)
1663    
1664    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1665    
1666      /* Handle recursion */
1667    
1668      else if (c == OP_REVERSE)
1669        {
1670        if (number < 0) return (uschar *)code;
1671        code += _pcre_OP_lengths[c];
1672        }
1673    
1674    /* Handle capturing bracket */    /* Handle capturing bracket */
1675    
1676    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1290  for (;;) Line 1680  for (;;)
1680      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1681      }      }
1682    
1683    /* 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
1684    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
1685    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1686      must add in its length. */
1687    
1688    else    else
1689      {      {
1690        switch(c)
1691          {
1692          case OP_TYPESTAR:
1693          case OP_TYPEMINSTAR:
1694          case OP_TYPEPLUS:
1695          case OP_TYPEMINPLUS:
1696          case OP_TYPEQUERY:
1697          case OP_TYPEMINQUERY:
1698          case OP_TYPEPOSSTAR:
1699          case OP_TYPEPOSPLUS:
1700          case OP_TYPEPOSQUERY:
1701          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1702          break;
1703    
1704          case OP_TYPEUPTO:
1705          case OP_TYPEMINUPTO:
1706          case OP_TYPEEXACT:
1707          case OP_TYPEPOSUPTO:
1708          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1709          break;
1710    
1711          case OP_MARK:
1712          case OP_PRUNE_ARG:
1713          case OP_SKIP_ARG:
1714          case OP_THEN_ARG:
1715          code += code[1];
1716          break;
1717          }
1718    
1719        /* Add in the fixed length from the table */
1720    
1721      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1722    
1723      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1724      a multi-byte character. The length in the table is a minimum, so we have to
1725      arrange to skip the extra bytes. */
1726    
1727  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1728      if (utf8) switch(c)      if (utf8) switch(c)
1729        {        {
# Line 1318  for (;;) Line 1745  for (;;)
1745        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1746        break;        break;
1747        }        }
1748    #else
1749        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1750  #endif  #endif
1751      }      }
1752    }    }
# Line 1354  for (;;) Line 1783  for (;;)
1783    
1784    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1785    
1786    /* 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
1787    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
1788    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
1789    bytes. */    must add in its length. */
1790    
1791    else    else
1792      {      {
1793        switch(c)
1794          {
1795          case OP_TYPESTAR:
1796          case OP_TYPEMINSTAR:
1797          case OP_TYPEPLUS:
1798          case OP_TYPEMINPLUS:
1799          case OP_TYPEQUERY:
1800          case OP_TYPEMINQUERY:
1801          case OP_TYPEPOSSTAR:
1802          case OP_TYPEPOSPLUS:
1803          case OP_TYPEPOSQUERY:
1804          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1805          break;
1806    
1807          case OP_TYPEPOSUPTO:
1808          case OP_TYPEUPTO:
1809          case OP_TYPEMINUPTO:
1810          case OP_TYPEEXACT:
1811          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1812          break;
1813    
1814          case OP_MARK:
1815          case OP_PRUNE_ARG:
1816          case OP_SKIP_ARG:
1817          case OP_THEN_ARG:
1818          code += code[1];
1819          break;
1820          }
1821    
1822        /* Add in the fixed length from the table */
1823    
1824      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1825    
1826        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1827        by a multi-byte character. The length in the table is a minimum, so we have
1828        to arrange to skip the extra bytes. */
1829    
1830  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1831      if (utf8) switch(c)      if (utf8) switch(c)
1832        {        {
# Line 1383  for (;;) Line 1848  for (;;)
1848        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1849        break;        break;
1850        }        }
1851    #else
1852        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1853  #endif  #endif
1854      }      }
1855    }    }
# Line 1398  for (;;) Line 1865  for (;;)
1865  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()
1866  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
1867  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1868  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
1869  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
1870    bracket whose current branch will already have been scanned.
1871    
1872  Arguments:  Arguments:
1873    code        points to start of search    code        points to start of search
1874    endcode     points to where to stop    endcode     points to where to stop
1875    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1876      cd          contains pointers to tables etc.
1877    
1878  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1879  */  */
1880    
1881  static BOOL  static BOOL
1882  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1883      compile_data *cd)
1884  {  {
1885  register int c;  register int c;
1886  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 1421  for (code = first_significant_code(code Line 1891  for (code = first_significant_code(code
1891    
1892    c = *code;    c = *code;
1893    
1894      /* Skip over forward assertions; the other assertions are skipped by
1895      first_significant_code() with a TRUE final argument. */
1896    
1897      if (c == OP_ASSERT)
1898        {
1899        do code += GET(code, 1); while (*code == OP_ALT);
1900        c = *code;
1901        continue;
1902        }
1903    
1904    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1905    
1906    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1907      {      {
1908      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1909      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1431  for (code = first_significant_code(code Line 1911  for (code = first_significant_code(code
1911      continue;      continue;
1912      }      }
1913    
1914      /* For a recursion/subroutine call, if its end has been reached, which
1915      implies a subroutine call, we can scan it. */
1916    
1917      if (c == OP_RECURSE)
1918        {
1919        BOOL empty_branch = FALSE;
1920        const uschar *scode = cd->start_code + GET(code, 1);
1921        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1922        do
1923          {
1924          if (could_be_empty_branch(scode, endcode, utf8, cd))
1925            {
1926            empty_branch = TRUE;
1927            break;
1928            }
1929          scode += GET(scode, 1);
1930          }
1931        while (*scode == OP_ALT);
1932        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1933        continue;
1934        }
1935    
1936    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1937    
1938    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
# Line 1438  for (code = first_significant_code(code Line 1940  for (code = first_significant_code(code
1940      BOOL empty_branch;      BOOL empty_branch;
1941      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1942    
1943      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1944        empty branch, so just skip over the conditional, because it could be empty.
1945        Otherwise, scan the individual branches of the group. */
1946    
1947      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;  
1948        code += GET(code, 1);        code += GET(code, 1);
1949        else
1950          {
1951          empty_branch = FALSE;
1952          do
1953            {
1954            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1955              empty_branch = TRUE;
1956            code += GET(code, 1);
1957            }
1958          while (*code == OP_ALT);
1959          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1960        }        }
1961      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1962      c = *code;      c = *code;
1963      continue;      continue;
1964      }      }
# Line 1457  for (code = first_significant_code(code Line 1967  for (code = first_significant_code(code
1967    
1968    switch (c)    switch (c)
1969      {      {
1970      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1971        cannot be represented just by a bit map. This includes negated single
1972        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1973        actual length is stored in the compiled code, so we must update "code"
1974        here. */
1975    
1976  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1977      case OP_XCLASS:      case OP_XCLASS:
1978      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1979      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1980  #endif  #endif
1981    
# Line 1505  for (code = first_significant_code(code Line 2019  for (code = first_significant_code(code
2019      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2020      case OP_WORDCHAR:      case OP_WORDCHAR:
2021      case OP_ANY:      case OP_ANY:
2022        case OP_ALLANY:
2023      case OP_ANYBYTE:      case OP_ANYBYTE:
2024      case OP_CHAR:      case OP_CHAR:
2025      case OP_CHARNC:      case OP_CHARNC:
# Line 1523  for (code = first_significant_code(code Line 2038  for (code = first_significant_code(code
2038      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2039      return FALSE;      return FALSE;
2040    
2041        /* These are going to continue, as they may be empty, but we have to
2042        fudge the length for the \p and \P cases. */
2043    
2044        case OP_TYPESTAR:
2045        case OP_TYPEMINSTAR:
2046        case OP_TYPEPOSSTAR:
2047        case OP_TYPEQUERY:
2048        case OP_TYPEMINQUERY:
2049        case OP_TYPEPOSQUERY:
2050        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2051        break;
2052    
2053        /* Same for these */
2054    
2055        case OP_TYPEUPTO:
2056        case OP_TYPEMINUPTO:
2057        case OP_TYPEPOSUPTO:
2058        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2059        break;
2060    
2061      /* End of branch */      /* End of branch */
2062    
2063      case OP_KET:      case OP_KET:
# Line 1541  for (code = first_significant_code(code Line 2076  for (code = first_significant_code(code
2076      case OP_QUERY:      case OP_QUERY:
2077      case OP_MINQUERY:      case OP_MINQUERY:
2078      case OP_POSQUERY:      case OP_POSQUERY:
2079        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2080        break;
2081    
2082      case OP_UPTO:      case OP_UPTO:
2083      case OP_MINUPTO:      case OP_MINUPTO:
2084      case OP_POSUPTO:      case OP_POSUPTO:
2085      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2086      break;      break;
2087  #endif  #endif
2088    
2089        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2090        string. */
2091    
2092        case OP_MARK:
2093        case OP_PRUNE_ARG:
2094        case OP_SKIP_ARG:
2095        case OP_THEN_ARG:
2096        code += code[1];
2097        break;
2098    
2099        /* None of the remaining opcodes are required to match a character. */
2100    
2101        default:
2102        break;
2103      }      }
2104    }    }
2105    
# Line 1569  Arguments: Line 2122  Arguments:
2122    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2123    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2124    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2125      cd          pointers to tables etc
2126    
2127  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2128  */  */
2129    
2130  static BOOL  static BOOL
2131  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2132    BOOL utf8)    BOOL utf8, compile_data *cd)
2133  {  {
2134  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2135    {    {
2136    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2137        return FALSE;
2138    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2139    }    }
2140  return TRUE;  return TRUE;
# Line 1592  return TRUE; Line 2147  return TRUE;
2147  *************************************************/  *************************************************/
2148    
2149  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2150  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
2151  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2152  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2153    
2154    Originally, this function only recognized a sequence of letters between the
2155    terminators, but it seems that Perl recognizes any sequence of characters,
2156    though of course unknown POSIX names are subsequently rejected. Perl gives an
2157    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2158    didn't consider this to be a POSIX class. Likewise for [:1234:].
2159    
2160    The problem in trying to be exactly like Perl is in the handling of escapes. We
2161    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2162    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2163    below handles the special case of \], but does not try to do any other escape
2164    processing. This makes it different from Perl for cases such as [:l\ower:]
2165    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2166    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2167    I think.
2168    
2169  Argument:  Arguments:
2170    ptr      pointer to the initial [    ptr      pointer to the initial [
2171    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2172    
2173  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2174  */  */
2175    
2176  static BOOL  static BOOL
2177  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2178  {  {
2179  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2180  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2181  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2182    {    {
2183    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2184    return TRUE;      {
2185        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2186        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2187          {
2188          *endptr = ptr;
2189          return TRUE;
2190          }
2191        }
2192    }    }
2193  return FALSE;  return FALSE;
2194  }  }
# Line 1639  Returns:     a value representing the na Line 2213  Returns:     a value representing the na
2213  static int  static int
2214  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2215  {  {
2216    const char *pn = posix_names;
2217  register int yield = 0;  register int yield = 0;
2218  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2219    {    {
2220    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2221      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2222      pn += posix_name_lengths[yield] + 1;
2223    yield++;    yield++;
2224    }    }
2225  return -1;  return -1;
# Line 1658  return -1; Line 2234  return -1;
2234  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2235  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2236  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
2237  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
2238  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
2239  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
2240  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
2241  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2242    OP_END.
2243    
2244  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2245  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 1685  adjust_recurse(uschar *group, int adjust Line 2262  adjust_recurse(uschar *group, int adjust
2262    uschar *save_hwm)    uschar *save_hwm)
2263  {  {
2264  uschar *ptr = group;  uschar *ptr = group;
2265    
2266  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2267    {    {
2268    int offset;    int offset;
# Line 1738  auto_callout(uschar *code, const uschar Line 2316  auto_callout(uschar *code, const uschar
2316  {  {
2317  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2318  *code++ = 255;  *code++ = 255;
2319  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2320  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2321  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2322  }  }
2323    
# Line 1764  Returns:             nothing Line 2342  Returns:             nothing
2342  static void  static void
2343  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2344  {  {
2345  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2346  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2347  }  }
2348    
# Line 1796  get_othercase_range(unsigned int *cptr, Line 2374  get_othercase_range(unsigned int *cptr,
2374  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2375    
2376  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2377    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2378    
2379  if (c > d) return FALSE;  if (c > d) return FALSE;
2380    
# Line 1805  next = othercase + 1; Line 2383  next = othercase + 1;
2383    
2384  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2385    {    {
2386    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2387    next++;    next++;
2388    }    }
2389    
# Line 1814  for (++c; c <= d; c++) Line 2392  for (++c; c <= d; c++)
2392    
2393  return TRUE;  return TRUE;
2394  }  }
2395    
2396    
2397    
2398    /*************************************************
2399    *        Check a character and a property        *
2400    *************************************************/
2401    
2402    /* This function is called by check_auto_possessive() when a property item
2403    is adjacent to a fixed character.
2404    
2405    Arguments:
2406      c            the character
2407      ptype        the property type
2408      pdata        the data for the type
2409      negated      TRUE if it's a negated property (\P or \p{^)
2410    
2411    Returns:       TRUE if auto-possessifying is OK
2412    */
2413    
2414    static BOOL
2415    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2416    {
2417    const ucd_record *prop = GET_UCD(c);
2418    switch(ptype)
2419      {
2420      case PT_LAMP:
2421      return (prop->chartype == ucp_Lu ||
2422              prop->chartype == ucp_Ll ||
2423              prop->chartype == ucp_Lt) == negated;
2424    
2425      case PT_GC:
2426      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2427    
2428      case PT_PC:
2429      return (pdata == prop->chartype) == negated;
2430    
2431      case PT_SC:
2432      return (pdata == prop->script) == negated;
2433    
2434      /* These are specials */
2435    
2436      case PT_ALNUM:
2437      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2438              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2439    
2440      case PT_SPACE:    /* Perl space */
2441      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2442              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2443              == negated;
2444    
2445      case PT_PXSPACE:  /* POSIX space */
2446      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2447              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2448              c == CHAR_FF || c == CHAR_CR)
2449              == negated;
2450    
2451      case PT_WORD:
2452      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2453              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2454              c == CHAR_UNDERSCORE) == negated;
2455      }
2456    return FALSE;
2457    }
2458  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2459    
2460    
# Line 1827  whether the next thing could possibly ma Line 2468  whether the next thing could possibly ma
2468  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2469    
2470  Arguments:  Arguments:
2471    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2472    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2473    ptr           next character in pattern    ptr           next character in pattern
2474    options       options bits    options       options bits
2475    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1839  Returns:        TRUE if possessifying is Line 2478  Returns:        TRUE if possessifying is
2478  */  */
2479    
2480  static BOOL  static BOOL
2481  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2482    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2483  {  {
2484  int next;  int c, next;
2485    int op_code = *previous++;
2486    
2487  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2488    
# Line 1851  if ((options & PCRE_EXTENDED) != 0) Line 2491  if ((options & PCRE_EXTENDED) != 0)
2491    for (;;)    for (;;)
2492      {      {
2493      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2494      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2495        {        {
2496        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2497          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1863  if ((options & PCRE_EXTENDED) != 0) Line 2503  if ((options & PCRE_EXTENDED) != 0)
2503  /* 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
2504  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2505    
2506  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2507    {    {
2508    int temperrorcode = 0;    int temperrorcode = 0;
2509    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1888  if ((options & PCRE_EXTENDED) != 0) Line 2528  if ((options & PCRE_EXTENDED) != 0)
2528    for (;;)    for (;;)
2529      {      {
2530      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2531      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2532        {        {
2533        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2534          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1899  if ((options & PCRE_EXTENDED) != 0) Line 2539  if ((options & PCRE_EXTENDED) != 0)
2539    
2540  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2541    
2542  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2543    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2544        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. */  
   
2545    
2546  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2547    the next item is a character. */
2548    
2549  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2550    {    {
2551    case OP_CHAR:    case OP_CHAR:
2552  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2553    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2554  #endif  #else
2555    return item != next;    c = *previous;
2556    #endif
2557      return c != next;
2558    
2559    /* 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
2560    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
2561    high-valued characters. */    high-valued characters. */
2562    
2563    case OP_CHARNC:    case OP_CHARNC:
2564  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2565    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2566  #endif  #else
2567    if (item == next) return FALSE;    c = *previous;
2568    #endif
2569      if (c == next) return FALSE;
2570  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2571    if (utf8)    if (utf8)
2572      {      {
2573      unsigned int othercase;      unsigned int othercase;
2574      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2575  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2576      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2577  #else  #else
2578      othercase = NOTACHAR;      othercase = NOTACHAR;
2579  #endif  #endif
2580      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2581      }      }
2582    else    else
2583  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2584    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2585    
2586    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2587    
2588    case OP_NOT:    case OP_NOT:
2589    if (next < 0) return FALSE;  /* Not a character */    if ((c = *previous) == next) return TRUE;
   if (item == next) return TRUE;  
2590    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2591  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2592    if (utf8)    if (utf8)
# Line 1955  if (next >= 0) switch(op_code) Line 2594  if (next >= 0) switch(op_code)
2594      unsigned int othercase;      unsigned int othercase;
2595      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2596  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2597      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2598  #else  #else
2599      othercase = NOTACHAR;      othercase = NOTACHAR;
2600  #endif  #endif
2601      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2602      }      }
2603    else    else
2604  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2605    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2606    
2607      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2608      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2609    
2610    case OP_DIGIT:    case OP_DIGIT:
2611    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 2006  if (next >= 0) switch(op_code) Line 2648  if (next >= 0) switch(op_code)
2648      case 0x202f:      case 0x202f:
2649      case 0x205f:      case 0x205f:
2650      case 0x3000:      case 0x3000:
2651      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2652      default:      default:
2653      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2654      }      }
2655    
2656      case OP_ANYNL:
2657    case OP_VSPACE:    case OP_VSPACE:
2658    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2659    switch(next)    switch(next)
# Line 2022  if (next >= 0) switch(op_code) Line 2665  if (next >= 0) switch(op_code)
2665      case 0x85:      case 0x85:
2666      case 0x2028:      case 0x2028:
2667      case 0x2029:      case 0x2029:
2668      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2669      default:      default:
2670      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2671      }      }
2672    
2673    #ifdef SUPPORT_UCP
2674      case OP_PROP:
2675      return check_char_prop(next, previous[0], previous[1], FALSE);
2676    
2677      case OP_NOTPROP:
2678      return check_char_prop(next, previous[0], previous[1], TRUE);
2679    #endif
2680    
2681    default:    default:
2682    return FALSE;    return FALSE;
2683    }    }
2684    
2685    
2686  /* 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
2687    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2688    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2689    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2690    replaced by OP_PROP codes when PCRE_UCP is set. */
2691    
2692  switch(op_code)  switch(op_code)
2693    {    {
2694    case OP_CHAR:    case OP_CHAR:
2695    case OP_CHARNC:    case OP_CHARNC:
2696  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2697    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2698  #endif  #else
2699      c = *previous;
2700    #endif
2701    switch(-next)    switch(-next)
2702      {      {
2703      case ESC_d:      case ESC_d:
2704      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2705    
2706      case ESC_D:      case ESC_D:
2707      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2708    
2709      case ESC_s:      case ESC_s:
2710      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2711    
2712      case ESC_S:      case ESC_S:
2713      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2714    
2715      case ESC_w:      case ESC_w:
2716      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2717    
2718      case ESC_W:      case ESC_W:
2719      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2720    
2721      case ESC_h:      case ESC_h:
2722      case ESC_H:      case ESC_H:
2723      switch(item)      switch(c)
2724        {        {
2725        case 0x09:        case 0x09:
2726        case 0x20:        case 0x20:
# Line 2091  switch(op_code) Line 2748  switch(op_code)
2748    
2749      case ESC_v:      case ESC_v:
2750      case ESC_V:      case ESC_V:
2751      switch(item)      switch(c)
2752        {        {
2753        case 0x0a:        case 0x0a:
2754        case 0x0b:        case 0x0b:
# Line 2104  switch(op_code) Line 2761  switch(op_code)
2761        default:        default:
2762        return -next == ESC_v;        return -next == ESC_v;
2763        }        }
2764    
2765        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2766        their substitutions and process them. The result will always be either
2767        -ESC_p or -ESC_P. Then fall through to process those values. */
2768    
2769    #ifdef SUPPORT_UCP
2770        case ESC_du:
2771        case ESC_DU:
2772        case ESC_wu:
2773        case ESC_WU:
2774        case ESC_su:
2775        case ESC_SU:
2776          {
2777          int temperrorcode = 0;
2778          ptr = substitutes[-next - ESC_DU];
2779          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2780          if (temperrorcode != 0) return FALSE;
2781          ptr++;    /* For compatibility */
2782          }
2783        /* Fall through */
2784    
2785        case ESC_p:
2786        case ESC_P:
2787          {
2788          int ptype, pdata, errorcodeptr;
2789          BOOL negated;
2790    
2791          ptr--;      /* Make ptr point at the p or P */
2792          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2793          if (ptype < 0) return FALSE;
2794          ptr++;      /* Point past the final curly ket */
2795    
2796          /* If the property item is optional, we have to give up. (When generated
2797          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2798          to the original \d etc. At this point, ptr will point to a zero byte. */
2799    
2800          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2801            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2802              return FALSE;
2803    
2804          /* Do the property check. */
2805    
2806          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2807          }
2808    #endif
2809    
2810      default:      default:
2811      return FALSE;      return FALSE;
2812      }      }
2813    
2814      /* In principle, support for Unicode properties should be integrated here as
2815      well. It means re-organizing the above code so as to get hold of the property
2816      values before switching on the op-code. However, I wonder how many patterns
2817      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2818      these op-codes are never generated.) */
2819    
2820    case OP_DIGIT:    case OP_DIGIT:
2821    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2822           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2823    
2824    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2825    return next == -ESC_d;    return next == -ESC_d;
2826    
2827    case OP_WHITESPACE:    case OP_WHITESPACE:
2828    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2829    
2830    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2831    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2832    
2833    case OP_HSPACE:    case OP_HSPACE:
2834    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2835             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2836    
2837    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2838    return next == -ESC_h;    return next == -ESC_h;
2839    
2840    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2841      case OP_ANYNL:
2842    case OP_VSPACE:    case OP_VSPACE:
2843    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2844    
2845    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2846    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2847    
2848    case OP_WORDCHAR:    case OP_WORDCHAR:
2849    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2850             next == -ESC_v || next == -ESC_R;
2851    
2852    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2853    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2200  BOOL inescq = FALSE; Line 2911  BOOL inescq = FALSE;
2911  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2912  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2913  const uschar *tempptr;  const uschar *tempptr;
2914    const uschar *nestptr = NULL;
2915  uschar *previous = NULL;  uschar *previous = NULL;
2916  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2917  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2209  uschar classbits[32]; Line 2921  uschar classbits[32];
2921  BOOL class_utf8;  BOOL class_utf8;
2922  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2923  uschar *class_utf8data;  uschar *class_utf8data;
2924    uschar *class_utf8data_base;
2925  uschar utf8_char[6];  uschar utf8_char[6];
2926  #else  #else
2927  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2928  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2929  #endif  #endif
2930    
2931  #ifdef DEBUG  #ifdef PCRE_DEBUG
2932  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2933  #endif  #endif
2934    
# Line 2248  req_caseopt = ((options & PCRE_CASELESS) Line 2961  req_caseopt = ((options & PCRE_CASELESS)
2961  for (;; ptr++)  for (;; ptr++)
2962    {    {
2963    BOOL negate_class;    BOOL negate_class;
2964      BOOL should_flip_negation;
2965    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2966    BOOL is_quantifier;    BOOL is_quantifier;
2967    BOOL is_recurse;    BOOL is_recurse;
# Line 2268  for (;; ptr++) Line 2982  for (;; ptr++)
2982    
2983    c = *ptr;    c = *ptr;
2984    
2985      /* If we are at the end of a nested substitution, revert to the outer level
2986      string. Nesting only happens one level deep. */
2987    
2988      if (c == 0 && nestptr != NULL)
2989        {
2990        ptr = nestptr;
2991        nestptr = NULL;
2992        c = *ptr;
2993        }
2994    
2995    /* 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
2996    previous cycle of this loop. */    previous cycle of this loop. */
2997    
2998    if (lengthptr != NULL)    if (lengthptr != NULL)
2999      {      {
3000  #ifdef DEBUG  #ifdef PCRE_DEBUG
3001      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3002  #endif  #endif
3003      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3004        {        {
3005        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3006        goto FAILED;        goto FAILED;
# Line 2298  for (;; ptr++) Line 3022  for (;; ptr++)
3022        goto FAILED;        goto FAILED;
3023        }        }
3024    
3025      *lengthptr += code - last_code;      *lengthptr += (int)(code - last_code);
3026      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));
3027    
3028      /* 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 2325  for (;; ptr++) Line 3049  for (;; ptr++)
3049    /* 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
3050    reference list. */    reference list. */
3051    
3052    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3053      {      {
3054      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3055      goto FAILED;      goto FAILED;
# Line 2335  for (;; ptr++) Line 3059  for (;; ptr++)
3059    
3060    if (inescq && c != 0)    if (inescq && c != 0)
3061      {      {
3062      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3063        {        {
3064        inescq = FALSE;        inescq = FALSE;
3065        ptr++;        ptr++;
# Line 2361  for (;; ptr++) Line 3085  for (;; ptr++)
3085    /* 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
3086    a quantifier. */    a quantifier. */
3087    
3088    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3089      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3090        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3091    
3092    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3093         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2377  for (;; ptr++) Line 3102  for (;; ptr++)
3102    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3103      {      {
3104      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3105      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3106        {        {
3107        while (*(++ptr) != 0)        while (*(++ptr) != 0)
3108          {          {
# Line 2402  for (;; ptr++) Line 3127  for (;; ptr++)
3127      {      {
3128      /* ===================================================================*/      /* ===================================================================*/
3129      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3130      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3131      case ')':      case CHAR_RIGHT_PARENTHESIS:
3132      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3133      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3134      *codeptr = code;      *codeptr = code;
# Line 2415  for (;; ptr++) Line 3140  for (;; ptr++)
3140          *errorcodeptr = ERR20;          *errorcodeptr = ERR20;
3141          goto FAILED;          goto FAILED;
3142          }          }
3143        *lengthptr += code - last_code;   /* To include callout length */        *lengthptr += (int)(code - last_code);   /* To include callout length */
3144        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3145        }        }
3146      return TRUE;      return TRUE;
# Line 2425  for (;; ptr++) Line 3150  for (;; ptr++)
3150      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3151      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3152    
3153      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3154      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3155        {        {
3156        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2434  for (;; ptr++) Line 3159  for (;; ptr++)
3159      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3160      break;      break;
3161    
3162      case '$':      case CHAR_DOLLAR_SIGN:
3163      previous = NULL;      previous = NULL;
3164      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3165      break;      break;
# Line 2442  for (;; ptr++) Line 3167  for (;; ptr++)
3167      /* 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
3168      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3169    
3170      case '.':      case CHAR_DOT:
3171      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3172      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3173      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3174      previous = code;      previous = code;
3175      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3176      break;      break;
3177    
3178    
# Line 2462  for (;; ptr++) Line 3187  for (;; ptr++)
3187      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,
3188      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3189      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.
     */  
3190    
3191      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3192        default (Perl) mode, it is treated as a data character. */
3193    
3194        case CHAR_RIGHT_SQUARE_BRACKET:
3195        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3196          {
3197          *errorcodeptr = ERR64;
3198          goto FAILED;
3199          }
3200        goto NORMAL_CHAR;
3201    
3202        case CHAR_LEFT_SQUARE_BRACKET:
3203      previous = code;      previous = code;
3204    
3205      /* 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
3206      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. */
3207    
3208      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3209          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3210            check_posix_syntax(ptr, &tempptr))
3211        {        {
3212        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3213        goto FAILED;        goto FAILED;
3214        }        }
3215    
# Line 2485  for (;; ptr++) Line 3221  for (;; ptr++)
3221      for (;;)      for (;;)
3222        {        {
3223        c = *(++ptr);        c = *(++ptr);
3224        if (c == '\\')        if (c == CHAR_BACKSLASH)
3225          {          {
3226          if (ptr[1] == 'E') ptr++;          if (ptr[1] == CHAR_E)
3227            else if (strncmp((const char *)ptr+1, "Q\\E", 3) == 0) ptr += 3;            ptr++;
3228              else break;          else if (strncmp((const char *)ptr+1,
3229                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3230              ptr += 3;
3231            else
3232              break;
3233          }          }
3234        else if (!negate_class && c == '^')        else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3235          negate_class = TRUE;          negate_class = TRUE;
3236        else break;        else break;
3237        }        }
3238    
3239        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3240        an initial ']' is taken as a data character -- the code below handles
3241        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3242        [^] must match any character, so generate OP_ALLANY. */
3243    
3244        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3245            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3246          {
3247          *code++ = negate_class? OP_ALLANY : OP_FAIL;
3248          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3249          zerofirstbyte = firstbyte;
3250          break;
3251          }
3252    
3253        /* If a class contains a negative special such as \S, we need to flip the
3254        negation flag at the end, so that support for characters > 255 works
3255        correctly (they are all included in the class). */
3256    
3257        should_flip_negation = FALSE;
3258    
3259      /* 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
3260      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
3261      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2513  for (;; ptr++) Line 3273  for (;; ptr++)
3273  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3274      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3275      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3276        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3277  #endif  #endif
3278    
3279      /* 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 2528  for (;; ptr++) Line 3289  for (;; ptr++)
3289          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3290          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3291          }          }
3292    
3293          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3294          data and reset the pointer. This is so that very large classes that
3295          contain a zillion UTF-8 characters no longer overwrite the work space
3296          (which is on the stack). */
3297    
3298          if (lengthptr != NULL)
3299            {
3300            *lengthptr += class_utf8data - class_utf8data_base;
3301            class_utf8data = class_utf8data_base;
3302            }
3303    
3304  #endif  #endif
3305    
3306        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3307    
3308        if (inescq)        if (inescq)
3309          {          {
3310          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3311            {            {
3312            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3313            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2549  for (;; ptr++) Line 3322  for (;; ptr++)
3322        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3323        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3324    
3325        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3326            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3327            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3328          {          {
3329          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3330          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3331          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3332          uschar pbits[32];          uschar pbits[32];
3333    
3334          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3335            {            {
3336            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3337            goto FAILED;            goto FAILED;
3338            }            }
3339    
3340          ptr += 2;          ptr += 2;
3341          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3342            {            {
3343            local_negate = TRUE;            local_negate = TRUE;
3344              should_flip_negation = TRUE;  /* Note negative special */
3345            ptr++;            ptr++;
3346            }            }
3347    
3348          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3349          if (posix_class < 0)          if (posix_class < 0)
3350            {            {
3351            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2584  for (;; ptr++) Line 3358  for (;; ptr++)
3358    
3359          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3360            posix_class = 0;            posix_class = 0;
3361    
3362          /* 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
3363          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3364          subtract bits that may be in the main map already. At the end we or the  
3365          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3366            if ((options & PCRE_UCP) != 0)
3367              {
3368              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3369              if (posix_substitutes[pc] != NULL)
3370                {
3371                nestptr = tempptr + 1;
3372                ptr = posix_substitutes[pc] - 1;
3373                continue;
3374                }
3375              }
3376    #endif
3377            /* In the non-UCP case, we build the bit map for the POSIX class in a
3378            chunk of local store because we may be adding and subtracting from it,
3379            and we don't want to subtract bits that may be in the main map already.
3380            At the end we or the result into the bit map that is being built. */
3381    
3382          posix_class *= 3;          posix_class *= 3;
3383    
# Line 2632  for (;; ptr++) Line 3421  for (;; ptr++)
3421    
3422        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3423        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
3424        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
3425        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
3426        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
3427        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3428          PCRE_EXTRA is set. */
3429    
3430        if (c == '\\')        if (c == CHAR_BACKSLASH)
3431          {          {
3432          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3433          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3434    
3435          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 */  
3436          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3437            {            {
3438            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3439              {              {
3440              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3441              }              }
3442            else inescq = TRUE;            else inescq = TRUE;
3443            continue;            continue;
3444            }            }
3445            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3446    
3447          if (c < 0)          if (c < 0)
3448            {            {
3449            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3450            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3451    
3452            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3453              {              {
3454    #ifdef SUPPORT_UCP
3455                case ESC_du:     /* These are the values given for \d etc */
3456                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3457                case ESC_wu:     /* escape sequence with an appropriate \p */
3458                case ESC_WU:     /* or \P to test Unicode properties instead */
3459                case ESC_su:     /* of the default ASCII testing. */
3460                case ESC_SU:
3461                nestptr = ptr;
3462                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3463                class_charcount -= 2;                /* Undo! */
3464                continue;
3465    #endif
3466              case ESC_d:              case ESC_d:
3467              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3468              continue;              continue;
3469    
3470              case ESC_D:              case ESC_D:
3471                should_flip_negation = TRUE;
3472              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3473              continue;              continue;
3474    
# Line 2677  for (;; ptr++) Line 3477  for (;; ptr++)
3477              continue;              continue;
3478    
3479              case ESC_W:              case ESC_W:
3480                should_flip_negation = TRUE;
3481              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3482              continue;              continue;
3483    
# Line 2686  for (;; ptr++) Line 3487  for (;; ptr++)
3487              continue;              continue;
3488    
3489              case ESC_S:              case ESC_S:
3490                should_flip_negation = TRUE;
3491              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3492              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3493              continue;              continue;
   
             case ESC_E: /* Perl ignores an orphan \E */  
             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. */  
3494    
3495            if (-c == ESC_h)              case ESC_h:
             {  
3496              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3497              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3498              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2730  for (;; ptr++) Line 3516  for (;; ptr++)
3516                }                }
3517  #endif  #endif
3518              continue;              continue;
             }  
3519    
3520            if (-c == ESC_H)              case ESC_H:
             {  
3521              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3522                {                {
3523                int x = 0xff;                int x = 0xff;
# Line 2775  for (;; ptr++) Line 3559  for (;; ptr++)
3559                }                }
3560  #endif  #endif
3561              continue;              continue;
             }  
3562    
3563            if (-c == ESC_v)              case ESC_v:
             {  
3564              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3565              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3566              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2794  for (;; ptr++) Line 3576  for (;; ptr++)
3576                }                }
3577  #endif  #endif
3578              continue;              continue;
             }  
3579    
3580            if (-c == ESC_V)              case ESC_V:
             {  
3581              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3582                {                {
3583                int x = 0xff;                int x = 0xff;
# Line 2827  for (;; ptr++) Line 3607  for (;; ptr++)
3607                }                }
3608  #endif  #endif
3609              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3610    
3611  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3612            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3613              {              case ESC_P:
3614              BOOL negated;                {
3615              int pdata;                BOOL negated;
3616              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3617              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3618              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3619              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3620                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3621              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3622              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3623              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3624              continue;                class_charcount -= 2;   /* Not a < 256 character */
3625              }                continue;
3626                  }
3627  #endif  #endif
3628            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3629            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3630            treated as literals. */              treated as literals. */
3631    
3632            if ((options & PCRE_EXTRA) != 0)              default:
3633              {              if ((options & PCRE_EXTRA) != 0)
3634              *errorcodeptr = ERR7;                {
3635              goto FAILED;                *errorcodeptr = ERR7;
3636                  goto FAILED;
3637                  }
3638                class_charcount -= 2;  /* Undo the default count from above */
3639                c = *ptr;              /* Get the final character and fall through */
3640                break;
3641              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3642            }            }
3643    
3644          /* 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 2872  for (;; ptr++) Line 3652  for (;; ptr++)
3652        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3653    
3654        CHECK_RANGE:        CHECK_RANGE:
3655        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3656          {          {
3657          inescq = FALSE;          inescq = FALSE;
3658          ptr += 2;          ptr += 2;
# Line 2880  for (;; ptr++) Line 3660  for (;; ptr++)
3660    
3661        oldptr = ptr;        oldptr = ptr;
3662    
3663        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3664    
3665          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3666    
3667          /* Check for range */
3668    
3669          if (!inescq && ptr[1] == CHAR_MINUS)
3670          {          {
3671          int d;          int d;
3672          ptr += 2;          ptr += 2;
3673          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3674    
3675          /* 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
3676          mode. */          mode. */
3677    
3678          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3679            {            {
3680            ptr += 2;            ptr += 2;
3681            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3682                { ptr += 2; continue; }
3683            inescq = TRUE;            inescq = TRUE;
3684            break;            break;
3685            }            }
3686    
3687          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3688            {            {
3689            ptr = oldptr;            ptr = oldptr;
3690            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2916  for (;; ptr++) Line 3703  for (;; ptr++)
3703          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
3704          in such circumstances. */          in such circumstances. */
3705    
3706          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3707            {            {
3708            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3709            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3710    
3711            /* \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 */  
3712    
3713            if (d < 0)            if (d < 0)
3714              {              {
3715              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  
3716                {                {
3717                ptr = oldptr;                ptr = oldptr;
3718                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2947  for (;; ptr++) Line 3731  for (;; ptr++)
3731    
3732          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3733    
3734            /* Remember \r or \n */
3735    
3736            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3737    
3738          /* 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
3739          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3740          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3065  for (;; ptr++) Line 3853  for (;; ptr++)
3853          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3854            {            {
3855            unsigned int othercase;            unsigned int othercase;
3856            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3857              {              {
3858              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3859              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3090  for (;; ptr++) Line 3878  for (;; ptr++)
3878          }          }
3879        }        }
3880    
3881      /* 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.
3882        If we are at the end of an internal nested string, revert to the outer
3883        string. */
3884    
3885        while (((c = *(++ptr)) != 0 ||
3886               (nestptr != NULL &&
3887                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3888               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3889    
3890      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3891    
3892      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3893        {        {
3894        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3895        goto FAILED;        goto FAILED;
3896        }        }
3897    
3898      /* 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
3899      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
3900      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
3901      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3902      single-bytes only. This is an historical hangover. Maybe one day we can  
3903      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3904        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3905        operate on single-bytes only. This is an historical hangover. Maybe one day
3906        we can tidy these opcodes to handle multi-byte characters.
3907    
3908      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
3909      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 3113  for (;; ptr++) Line 3911  for (;; ptr++)
3911      can cause firstbyte to be set. Otherwise, there can be no first char if      can cause firstbyte to be set. Otherwise, there can be no first char if
3912      this item is first, whatever repeat count may follow. In the case of      this item is first, whatever repeat count may follow. In the case of
3913      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3914    
3915  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3916      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3917            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3918  #else  #else
3919      if (class_charcount == 1)      if (class_charcount == 1)
3920  #endif  #endif
# Line 3161  for (;; ptr++) Line 3957  for (;; ptr++)
3957      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3958    
3959      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3960      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3961      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
3962        characters > 255 are in the class, so any that were explicitly given as
3963        well can be ignored. If (when there are explicit characters > 255 that must
3964        be listed) there are no characters < 256, we can omit the bitmap in the
3965        actual compiled code. */
3966    
3967  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3968      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3969        {        {
3970        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3971        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3191  for (;; ptr++) Line 3991  for (;; ptr++)
3991        }        }
3992  #endif  #endif
3993    
3994      /* 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
3995      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
3996      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
3997      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
3998        negating it if necessary. */
3999    
4000        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4001      if (negate_class)      if (negate_class)
4002        {        {
       *code++ = OP_NCLASS;  
4003        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4004          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4005        }        }
4006      else      else
4007        {        {
       *code++ = OP_CLASS;  
4008        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4009        }        }
4010      code += 32;      code += 32;
# Line 3215  for (;; ptr++) Line 4015  for (;; ptr++)
4015      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4016      has been tested above. */      has been tested above. */
4017    
4018      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4019      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4020      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4021      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4022      goto REPEAT;      goto REPEAT;
4023    
4024      case '*':      case CHAR_ASTERISK:
4025      repeat_min = 0;      repeat_min = 0;
4026      repeat_max = -1;      repeat_max = -1;
4027      goto REPEAT;      goto REPEAT;
4028    
4029      case '+':      case CHAR_PLUS:
4030      repeat_min = 1;      repeat_min = 1;
4031      repeat_max = -1;      repeat_max = -1;
4032      goto REPEAT;      goto REPEAT;
4033    
4034      case '?':      case CHAR_QUESTION_MARK:
4035      repeat_min = 0;      repeat_min = 0;
4036      repeat_max = 1;      repeat_max = 1;
4037    
# Line 3266  for (;; ptr++) Line 4066  for (;; ptr++)
4066      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
4067      repeat type to the non-default. */      repeat type to the non-default. */
4068    
4069      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4070        {        {
4071        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4072        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4073        ptr++;        ptr++;
4074        }        }
4075      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4076        {        {
4077        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4078        ptr++;        ptr++;
# Line 3319  for (;; ptr++) Line 4119  for (;; ptr++)
4119    
4120        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4121            repeat_max < 0 &&            repeat_max < 0 &&
4122            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4123          {          {
4124          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4125          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3341  for (;; ptr++) Line 4140  for (;; ptr++)
4140        c = previous[1];        c = previous[1];
4141        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4142            repeat_max < 0 &&            repeat_max < 0 &&
4143            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4144          {          {
4145          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4146          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3365  for (;; ptr++) Line 4164  for (;; ptr++)
4164    
4165        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4166            repeat_max < 0 &&            repeat_max < 0 &&
4167            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4168          {          {
4169          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4170          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3387  for (;; ptr++) Line 4186  for (;; ptr++)
4186    
4187        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4188    
4189          /*--------------------------------------------------------------------*/
4190          /* This code is obsolete from release 8.00; the restriction was finally
4191          removed: */
4192    
4193        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4194        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4195    
4196        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4197          /*--------------------------------------------------------------------*/
4198    
4199        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4200    
# Line 3537  for (;; ptr++) Line 4341  for (;; ptr++)
4341          goto END_REPEAT;          goto END_REPEAT;
4342          }          }
4343    
4344          /*--------------------------------------------------------------------*/
4345          /* This code is obsolete from release 8.00; the restriction was finally
4346          removed: */
4347    
4348        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4349        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4350    
4351        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4352          /*--------------------------------------------------------------------*/
4353    
4354        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4355          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3565  for (;; ptr++) Line 4374  for (;; ptr++)
4374        {        {
4375        register int i;        register int i;
4376        int ketoffset = 0;        int ketoffset = 0;
4377        int len = code - previous;        int len = (int)(code - previous);
4378        uschar *bralink = NULL;        uschar *bralink = NULL;
4379    
4380        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3586  for (;; ptr++) Line 4395  for (;; ptr++)
4395          {          {
4396          register uschar *ket = previous;          register uschar *ket = previous;
4397          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4398          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4399          }          }
4400    
4401        /* 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 3598  for (;; ptr++) Line 4407  for (;; ptr++)
4407    
4408        if (repeat_min == 0)        if (repeat_min == 0)
4409          {          {
4410          /* 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
4411          altogether. */          output altogether, like this:
   
         if (repeat_max == 0)  
           {  
           code = previous;  
           goto END_REPEAT;  
           }  
4412    
4413          /* If the maximum is 1 or unlimited, we just have to stick in the          ** if (repeat_max == 0)
4414          BRAZERO and do no more at this point. However, we do need to adjust          **   {
4415          any OP_RECURSE calls inside the group that refer to the group itself or          **   code = previous;
4416          any internal or forward referenced group, because the offset is from          **   goto END_REPEAT;
4417          the start of the whole regex. Temporarily terminate the pattern while          **   }
4418          doing this. */  
4419            However, that fails when a group is referenced as a subroutine from
4420            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4421            so that it is skipped on execution. As we don't have a list of which
4422            groups are referenced, we cannot do this selectively.
4423    
4424            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4425            and do no more at this point. However, we do need to adjust any
4426            OP_RECURSE calls inside the group that refer to the group itself or any
4427            internal or forward referenced group, because the offset is from the
4428            start of the whole regex. Temporarily terminate the pattern while doing
4429            this. */
4430    
4431          if (repeat_max <= 1)          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
4432            {            {
4433            *code = OP_END;            *code = OP_END;
4434            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4435            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4436            code++;            code++;
4437              if (repeat_max == 0)
4438                {
4439                *previous++ = OP_SKIPZERO;
4440                goto END_REPEAT;
4441                }
4442            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4443            }            }
4444    
# Line 3644  for (;; ptr++) Line 4463  for (;; ptr++)
4463            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4464            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4465    
4466            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4467            bralink = previous;            bralink = previous;
4468            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4469            }            }
# Line 3665  for (;; ptr++) Line 4484  for (;; ptr++)
4484            {            {
4485            /* 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
4486            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
4487            potential integer overflow. */            potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4488              integer type when available, otherwise double. */
4489    
4490            if (lengthptr != NULL)            if (lengthptr != NULL)
4491              {              {
4492              int delta = (repeat_min - 1)*length_prevgroup;              int delta = (repeat_min - 1)*length_prevgroup;
4493              if ((double)(repeat_min - 1)*(double)length_prevgroup >              if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4494                                                              (double)INT_MAX ||                    (INT64_OR_DOUBLE)length_prevgroup >
4495                        (INT64_OR_DOUBLE)INT_MAX ||
4496                  OFLOW_MAX - *lengthptr < delta)                  OFLOW_MAX - *lengthptr < delta)
4497                {                {
4498                *errorcodeptr = ERR20;                *errorcodeptr = ERR20;
# Line 3717  for (;; ptr++) Line 4538  for (;; ptr++)
4538          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
4539          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
4540          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
4541          paranoid checks to avoid integer overflow. */          paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4542            a 64-bit integer type when available, otherwise double. */
4543    
4544          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4545            {            {
4546            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4547                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */                        2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4548            if ((double)repeat_max *            if ((INT64_OR_DOUBLE)repeat_max *
4549                  (double)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)                  (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4550                    > (double)INT_MAX ||                    > (INT64_OR_DOUBLE)INT_MAX ||
4551                OFLOW_MAX - *lengthptr < delta)                OFLOW_MAX - *lengthptr < delta)
4552              {              {
4553              *errorcodeptr = ERR20;              *errorcodeptr = ERR20;
# Line 3750  for (;; ptr++) Line 4572  for (;; ptr++)
4572              {              {
4573              int offset;              int offset;
4574              *code++ = OP_BRA;              *code++ = OP_BRA;
4575              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4576              bralink = code;              bralink = code;
4577              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4578              }              }
# Line 3771  for (;; ptr++) Line 4593  for (;; ptr++)
4593          while (bralink != NULL)          while (bralink != NULL)
4594            {            {
4595            int oldlinkoffset;            int oldlinkoffset;
4596            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4597            uschar *bra = code - offset;            uschar *bra = code - offset;
4598            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4599            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 3802  for (;; ptr++) Line 4624  for (;; ptr++)
4624            uschar *scode = bracode;            uschar *scode = bracode;
4625            do            do
4626              {              {
4627              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4628                {                {
4629                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4630                break;                break;
# Line 3814  for (;; ptr++) Line 4636  for (;; ptr++)
4636          }          }
4637        }        }
4638    
4639        /* If previous is OP_FAIL, it was generated by an empty class [] in
4640        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4641        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4642        error above. We can just ignore the repeat in JS case. */
4643    
4644        else if (*previous == OP_FAIL) goto END_REPEAT;
4645    
4646      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4647    
4648      else      else
# Line 3838  for (;; ptr++) Line 4667  for (;; ptr++)
4667      if (possessive_quantifier)      if (possessive_quantifier)
4668        {        {
4669        int len;        int len;
4670        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4671            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4672            tempcode += _pcre_OP_lengths[*tempcode] +
4673              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4674    
4675          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4676            {
4677          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4678        len = code - tempcode;  #ifdef SUPPORT_UTF8
4679            if (utf8 && tempcode[-1] >= 0xc0)
4680              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4681    #endif
4682            }
4683    
4684          len = (int)(code - tempcode);
4685        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4686          {          {
4687          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 3859  for (;; ptr++) Line 4699  for (;; ptr++)
4699          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4700          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4701    
4702            /* Because we are moving code along, we must ensure that any
4703            pending recursive references are updated. */
4704    
4705          default:          default:
4706            *code = OP_END;
4707            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4708          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4709          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4710          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3886  for (;; ptr++) Line 4731  for (;; ptr++)
4731      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4732      parenthesis forms.  */      parenthesis forms.  */
4733    
4734      case '(':      case CHAR_LEFT_PARENTHESIS:
4735      newoptions = options;      newoptions = options;
4736      skipbytes = 0;      skipbytes = 0;
4737      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4738      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4739      reset_bracount = FALSE;      reset_bracount = FALSE;
4740    
4741      /* First deal with various "verbs" that can be introduced by '*'. */      /* First deal with various "verbs" that can be introduced by '*'. */
4742    
4743      if (*(++ptr) == '*' && (cd->ctypes[ptr[1]] & ctype_letter) != 0)      if (*(++ptr) == CHAR_ASTERISK &&
4744             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4745        {        {
4746        int i, namelen;        int i, namelen;
4747        const uschar *name = ++ptr;        int arglen = 0;
4748          const char *vn = verbnames;
4749          const uschar *name = ptr + 1;
4750          const uschar *arg = NULL;
4751        previous = NULL;        previous = NULL;
4752        while ((cd->ctypes[*++ptr] & ctype_letter) != 0);        while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4753        if (*ptr == ':')        namelen = (int)(ptr - name);
4754    
4755          if (*ptr == CHAR_COLON)
4756          {          {
4757          *errorcodeptr = ERR59;   /* Not supported */          arg = ++ptr;
4758          goto FAILED;          while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4759          }            || *ptr == '_') ptr++;
4760        if (*ptr != ')')          arglen = (int)(ptr - arg);
4761            }
4762    
4763          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4764          {          {
4765          *errorcodeptr = ERR60;          *errorcodeptr = ERR60;
4766          goto FAILED;          goto FAILED;
4767          }          }
4768        namelen = ptr - name;  
4769          /* Scan the table of verb names */
4770    
4771        for (i = 0; i < verbcount; i++)        for (i = 0; i < verbcount; i++)
4772          {          {
4773          if (namelen == verbs[i].len &&          if (namelen == verbs[i].len &&
4774              strncmp((char *)name, verbs[i].name, namelen) == 0)              strncmp((char *)name, vn, namelen) == 0)
4775            {            {
4776            *code = verbs[i].op;            /* Check for open captures before ACCEPT */
4777            if (*code++ == OP_ACCEPT) cd->had_accept = TRUE;  
4778            break;            if (verbs[i].op == OP_ACCEPT)
4779            }              {
4780          }              open_capitem *oc;
4781        if (i < verbcount) continue;              cd->had_accept = TRUE;
4782        *errorcodeptr = ERR60;              for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4783        goto FAILED;                {
4784        }                *code++ = OP_CLOSE;
4785                  PUT2INC(code, 0, oc->number);
4786                  }
4787                }
4788    
4789              /* Handle the cases with/without an argument */
4790    
4791              if (arglen == 0)
4792                {
4793                if (verbs[i].op < 0)   /* Argument is mandatory */
4794                  {
4795                  *errorcodeptr = ERR66;
4796                  goto FAILED;
4797                  }
4798                *code++ = verbs[i].op;
4799                }
4800    
4801              else
4802                {
4803                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4804                  {
4805                  *errorcodeptr = ERR59;
4806                  goto FAILED;
4807                  }
4808                *code++ = verbs[i].op_arg;
4809                *code++ = arglen;
4810                memcpy(code, arg, arglen);
4811                code += arglen;
4812                *code++ = 0;
4813                }
4814    
4815              break;  /* Found verb, exit loop */
4816              }
4817    
4818            vn += verbs[i].len + 1;
4819            }
4820    
4821          if (i < verbcount) continue;    /* Successfully handled a verb */
4822          *errorcodeptr = ERR60;          /* Verb not recognized */
4823          goto FAILED;
4824          }
4825    
4826      /* Deal with the extended parentheses; all are introduced by '?', and the      /* Deal with the extended parentheses; all are introduced by '?', and the
4827      appearance of any of them means that this is not a capturing group. */      appearance of any of them means that this is not a capturing group. */
4828    
4829      else if (*ptr == '?')      else if (*ptr == CHAR_QUESTION_MARK)
4830        {        {
4831        int i, set, unset, namelen;        int i, set, unset, namelen;
4832        int *optset;        int *optset;
# Line 3939  for (;; ptr++) Line 4835  for (;; ptr++)
4835    
4836        switch (*(++ptr))        switch (*(++ptr))
4837          {          {
4838          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4839          ptr++;          ptr++;
4840          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4841          if (*ptr == 0)          if (*ptr == 0)
4842            {            {
4843            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3951  for (;; ptr++) Line 4847  for (;; ptr++)
4847    
4848    
4849          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4850          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4851          reset_bracount = TRUE;          reset_bracount = TRUE;
4852          /* Fall through */          /* Fall through */
4853    
4854          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4855          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4856          bravalue = OP_BRA;          bravalue = OP_BRA;
4857          ptr++;          ptr++;
4858          break;          break;
4859    
4860    
4861          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4862          case '(':          case CHAR_LEFT_PARENTHESIS:
4863          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4864    
4865          /* 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 3983  for (;; ptr++) Line 4879  for (;; ptr++)
4879          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4880          including assertions, are processed. */          including assertions, are processed. */
4881    
4882          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4883                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4884            break;            break;
4885    
4886          /* 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 3995  for (;; ptr++) Line 4892  for (;; ptr++)
4892    
4893          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4894    
4895          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4896            {            {
4897            terminator = -1;            terminator = -1;
4898            ptr += 2;            ptr += 2;
# Line 4005  for (;; ptr++) Line 4902  for (;; ptr++)
4902          /* 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
4903          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4904    
4905          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4906            {            {
4907            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4908            ptr++;            ptr++;
4909            }            }
4910          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4911            {            {
4912            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4913            ptr++;            ptr++;
4914            }            }
4915          else          else
4916            {            {
4917            terminator = 0;            terminator = 0;
4918            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4919            }            }
4920    
4921          /* 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 4038  for (;; ptr++) Line 4935  for (;; ptr++)
4935            {            {
4936            if (recno >= 0)            if (recno >= 0)
4937              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4938                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4939            ptr++;            ptr++;
4940            }            }
4941          namelen = ptr - name;          namelen = (int)(ptr - name);
4942    
4943          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4944                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4945            {            {
4946            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4947            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 4065  for (;; ptr++) Line 4963  for (;; ptr++)
4963              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4964              goto FAILED;              goto FAILED;
4965              }              }
4966            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4967                cd->bracount - recno + 1 : recno +cd->bracount;
4968              if (recno <= 0 || recno > cd->final_bracount)
4969              {              {
4970              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4971              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4972              }              }
           else recno += cd->bracount;  
4973            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4974            break;            break;
4975            }            }
4976    
4977          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4978          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4979            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4980            except they record that the reference was originally to a name. The
4981            information is used to check duplicate names. */
4982    
4983          slot = cd->name_table;          slot = cd->name_table;
4984          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 4095  for (;; ptr++) Line 4993  for (;; ptr++)
4993            {            {
4994            recno = GET2(slot, 0);            recno = GET2(slot, 0);
4995            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4996              code[1+LINK_SIZE]++;
4997            }            }
4998    
4999          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5000    
5001          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5002                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
5003            {            {
5004            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5005              code[1+LINK_SIZE]++;
5006            }            }
5007    
5008          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4120  for (;; ptr++) Line 5020  for (;; ptr++)
5020          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
5021          specific group number. */          specific group number. */
5022    
5023          else if (*name == 'R')          else if (*name == CHAR_R)
5024            {            {
5025            recno = 0;            recno = 0;
5026            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4130  for (;; ptr++) Line 5030  for (;; ptr++)
5030                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5031                goto FAILED;                goto FAILED;
5032                }                }
5033              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
5034              }              }
5035            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
5036            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4140  for (;; ptr++) Line 5040  for (;; ptr++)
5040          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
5041          false. */          false. */
5042    
5043          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
5044            {            {
5045            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
5046            skipbytes = 1;            skipbytes = 1;
5047            }            }
5048    
5049          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
5050            in the second pass here, so final_bracount is set. */
5051    
5052          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
5053            {            {
5054            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5055            }            }
# Line 4164  for (;; ptr++) Line 5065  for (;; ptr++)
5065    
5066    
5067          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5068          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5069          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5070          ptr++;          ptr++;
5071          break;          break;
5072    
5073    
5074          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5075          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
5076          ptr++;          ptr++;
5077          if (*ptr == ')')          /* Optimize (?!) */          if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
5078            {            {
5079            *code++ = OP_FAIL;            *code++ = OP_FAIL;
5080            previous = NULL;            previous = NULL;
5081            continue;            continue;
5082            }            }
5083          bravalue = OP_ASSERT_NOT;          bravalue = OP_ASSERT_NOT;
5084          break;          break;
5085    
5086    
5087          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5088          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
5089          switch (ptr[1])          switch (ptr[1])
5090            {            {
5091            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5092            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5093            ptr += 2;            ptr += 2;
5094            break;            break;
5095    
5096            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5097            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5098            ptr += 2;            ptr += 2;
5099            break;            break;
# Line 4207  for (;; ptr++) Line 5108  for (;; ptr++)
5108    
5109    
5110          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5111          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
5112          bravalue = OP_ONCE;          bravalue = OP_ONCE;
5113          ptr++;          ptr++;
5114          break;          break;
5115    
5116    
5117          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5118          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5119          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
5120          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
5121          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5122            {            {
5123            int n = 0;            int n = 0;
5124            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
5125              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
5126            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
5127              {              {
5128              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
5129              goto FAILED;              goto FAILED;
# Line 4233  for (;; ptr++) Line 5134  for (;; ptr++)
5134              goto FAILED;              goto FAILED;
5135              }              }
5136            *code++ = n;            *code++ = n;
5137            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5138            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5139            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5140            }            }
5141          previous = NULL;          previous = NULL;
# Line 4242  for (;; ptr++) Line 5143  for (;; ptr++)
5143    
5144    
5145          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5146          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5147          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5148                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5149            {            {
5150            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
5151            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5152            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5153            }            }
5154          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
5155            {            {
5156            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
5157            goto FAILED;            goto FAILED;
# Line 4259  for (;; ptr++) Line 5161  for (;; ptr++)
5161    
5162          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5163          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5164          case '\'':          case CHAR_APOSTROPHE:
5165            {            {
5166            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5167                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5168            name = ++ptr;            name = ++ptr;
5169    
5170            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5171            namelen = ptr - name;            namelen = (int)(ptr - name);
5172    
5173            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5174    
# Line 4292  for (;; ptr++) Line 5195  for (;; ptr++)
5195                }                }
5196              }              }
5197    
5198            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5199              alphabetical order. Duplicate names for different numbers are
5200              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5201              number are always OK. (An existing number can be re-used if (?|
5202              appears in the pattern.) In either event, a duplicate name results in
5203              a duplicate entry in the table, even if the number is the same. This
5204              is because the number of names, and hence the table size, is computed
5205              in the pre-compile, and it affects various numbers and pointers which
5206              would all have to be modified, and the compiled code moved down, if
5207              duplicates with the same number were omitted from the table. This
5208              doesn't seem worth the hassle. However, *different* names for the
5209              same number are not permitted. */
5210    
5211            else            else
5212              {              {
5213                BOOL dupname = FALSE;
5214              slot = cd->name_table;              slot = cd->name_table;
5215    
5216              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5217                {                {
5218                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4304  for (;; ptr++) Line 5220  for (;; ptr++)
5220                  {                  {
5221                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5222                    {                    {
5223                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5224                          (options & PCRE_DUPNAMES) == 0)
5225                      {                      {
5226                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5227                      goto FAILED;                      goto FAILED;
5228                      }                      }
5229                      else dupname = TRUE;
5230                    }                    }
5231                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5232                  }                  }
5233    
5234                  /* Make space in the table and break the loop for an earlier
5235                  name. For a duplicate or later name, carry on. We do this for
5236                  duplicates so that in the simple case (when ?(| is not used) they
5237                  are in order of their numbers. */
5238    
5239                if (crc < 0)                if (crc < 0)
5240                  {                  {
5241                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5242                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5243                  break;                  break;
5244                  }                  }
5245    
5246                  /* Continue the loop for a later or duplicate name */
5247    
5248                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5249                }                }
5250    
5251                /* For non-duplicate names, check for a duplicate number before
5252                adding the new name. */
5253    
5254                if (!dupname)
5255                  {
5256                  uschar *cslot = cd->name_table;
5257                  for (i = 0; i < cd->names_found; i++)
5258                    {
5259                    if (cslot != slot)
5260                      {
5261                      if (GET2(cslot, 0) == cd->bracount + 1)
5262                        {
5263                        *errorcodeptr = ERR65;
5264                        goto FAILED;
5265                        }
5266                      }
5267                    else i--;
5268                    cslot += cd->name_entry_size;
5269                    }
5270                  }
5271    
5272              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5273              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5274              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5275              }              }
5276            }            }
5277    
5278          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5279            encountered. */
5280    
         ptr++;                    /* Move past > or ' */  
5281          cd->names_found++;          cd->names_found++;
5282            ptr++;                    /* Move past > or ' */
5283          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5284    
5285    
5286          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5287          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5288          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5289          is_recurse = TRUE;          is_recurse = TRUE;
5290          /* Fall through */          /* Fall through */
5291    
5292          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
5293          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
5294          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
5295            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
5296            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
5297    
5298          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5299          name = ++ptr;          name = ++ptr;
5300          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5301          namelen = ptr - name;          namelen = (int)(ptr - name);
5302    
5303          /* In the pre-compile phase, do a syntax check and set a dummy          /* In the pre-compile phase, do a syntax check and set a dummy
5304          reference number. */          reference number. */
5305    
5306          if (lengthptr != NULL)          if (lengthptr != NULL)
5307            {            {
5308              if (namelen == 0)
5309                {
5310                *errorcodeptr = ERR62;
5311                goto FAILED;
5312                }
5313            if (*ptr != terminator)            if (*ptr != terminator)
5314              {              {
5315              *errorcodeptr = ERR42;              *errorcodeptr = ERR42;
# Line 4367  for (;; ptr++) Line 5323  for (;; ptr++)
5323            recno = 0;            recno = 0;
5324            }            }
5325    
5326          /* In the real compile, seek the name in the table */          /* In the real compile, seek the name in the table. We check the name
5327            first, and then check that we have reached the end of the name in the
5328            table. That way, if the name that is longer than any in the table,
5329            the comparison will fail without reading beyond the table entry. */
5330    
5331          else          else
5332            {            {
5333            slot = cd->name_table;            slot = cd->name_table;
5334            for (i = 0; i < cd->names_found; i++)            for (i = 0; i < cd->names_found; i++)
5335              {              {
5336              if (strncmp((char *)name, (char *)slot+2, namelen) == 0) break;              if (strncmp((char *)name, (char *)slot+2, namelen) == 0 &&
5337                    slot[2+namelen] == 0)
5338                  break;
5339              slot += cd->name_entry_size;              slot += cd->name_entry_size;
5340              }              }
5341    
# Line 4383  for (;; ptr++) Line 5344  for (;; ptr++)
5344              recno = GET2(slot, 0);              recno = GET2(slot, 0);
5345              }              }
5346            else if ((recno =                /* Forward back reference */            else if ((recno =                /* Forward back reference */
5347                      find_parens(ptr, cd->bracount, name, namelen,                      find_parens(cd, name, namelen,
5348                        (options & PCRE_EXTENDED) != 0)) <= 0)                        (options & PCRE_EXTENDED) != 0)) <= 0)
5349              {              {
5350              *errorcodeptr = ERR15;              *errorcodeptr = ERR15;
# Line 4399  for (;; ptr++) Line 5360  for (;; ptr++)
5360    
5361    
5362          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5363          case 'R':                 /* Recursion */          case CHAR_R:              /* Recursion */
5364          ptr++;                    /* Same as (?0)      */          ptr++;                    /* Same as (?0)      */