/[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 200 by ph10, Wed Aug 1 09:10:40 2007 UTC revision 545 by ph10, Wed Jun 16 10:51:15 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 65  used by pcretest. DEBUG is not defined w Line 66  used by pcretest. DEBUG is not defined w
66    
67  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))  #define SETBIT(a,b) a[b/8] |= (1 << (b%8))
68    
69    /* Maximum length value to check against when making sure that the integer that
70    holds the compiled pattern length does not overflow. We make it a bit less than
71    INT_MAX to allow for adding in group terminating bytes, so that we don't have
72    to check them every time. */
73    
74    #define OFLOW_MAX (INT_MAX - 20)
75    
76    
77  /*************************************************  /*************************************************
78  *      Code parameters and static tables         *  *      Code parameters and static tables         *
# Line 84  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 123  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 133  static const short int escapes[] = { Line 181  static const short int escapes[] = {
181  #endif  #endif
182    
183    
184  /* Tables of names of POSIX character classes and their lengths. The list is  /* Table of special "verbs" like (*PRUNE). This is a short table, so it is
185  terminated by a zero length entry. The first three must be alpha, lower, upper,  searched linearly. Put all the names into a single string, in order to reduce
186  as this is assumed for handling case independence. */  the number of relocations when a shared library is dynamically linked. The
187    string is built from string macros so that it works in UTF-8 mode on EBCDIC
188  static const char *const posix_names[] = {  platforms. */
189    "alpha", "lower", "upper",  
190    "alnum", "ascii", "blank", "cntrl", "digit", "graph",  typedef struct verbitem {
191    "print", "punct", "space", "word",  "xdigit" };    int   len;                 /* Length of verb name */
192      int   op;                  /* Op when no arg, or -1 if arg mandatory */
193      int   op_arg;              /* Op when arg present, or -1 if not allowed */
194    } verbitem;
195    
196    static const char verbnames[] =
197      "\0"                       /* Empty name is a shorthand for MARK */
198      STRING_MARK0
199      STRING_ACCEPT0
200      STRING_COMMIT0
201      STRING_F0
202      STRING_FAIL0
203      STRING_PRUNE0
204      STRING_SKIP0
205      STRING_THEN;
206    
207    static const verbitem verbs[] = {
208      { 0, -1,        OP_MARK },
209      { 4, -1,        OP_MARK },
210      { 6, OP_ACCEPT, -1 },
211      { 6, OP_COMMIT, -1 },
212      { 1, OP_FAIL,   -1 },
213      { 4, OP_FAIL,   -1 },
214      { 5, OP_PRUNE,  OP_PRUNE_ARG },
215      { 4, OP_SKIP,   OP_SKIP_ARG  },
216      { 4, OP_THEN,   OP_THEN_ARG  }
217    };
218    
219    static const int verbcount = sizeof(verbs)/sizeof(verbitem);
220    
221    
222    /* Tables of names of POSIX character classes and their lengths. The names are
223    now all in a single string, to reduce the number of relocations when a shared
224    library is dynamically loaded. The list of lengths is terminated by a zero
225    length entry. The first three must be alpha, lower, upper, as this is assumed
226    for handling case independence. */
227    
228    static const char posix_names[] =
229      STRING_alpha0 STRING_lower0 STRING_upper0 STRING_alnum0
230      STRING_ascii0 STRING_blank0 STRING_cntrl0 STRING_digit0
231      STRING_graph0 STRING_print0 STRING_punct0 STRING_space0
232      STRING_word0  STRING_xdigit;
233    
234  static const uschar posix_name_lengths[] = {  static const uschar posix_name_lengths[] = {
235    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };    5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 4, 6, 0 };
# Line 172  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 179  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 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",    "repeated subpattern is too long\0"    /** DEAD **/
391    "octal value is greater than \\377 (not in UTF-8 mode)",    "octal value is greater than \\377 (not in UTF-8 mode)\0"
392    "internal error: overran compiling workspace",    "internal error: overran compiling workspace\0"
393    "internal error: previously-checked referenced subpattern not found",    "internal error: previously-checked referenced subpattern not found\0"
394    "DEFINE group contains more than one branch",    "DEFINE group contains more than one branch\0"
395    /* 55 */    /* 55 */
396    "repeating a DEFINE group is not allowed",    "repeating a DEFINE group is not allowed\0"
397    "inconsistent NEWLINE options",    "inconsistent NEWLINE options\0"
398    "\\g is not followed by a braced name or an optionally braced non-zero number",    "\\g is not followed by a braced, angle-bracketed, or quoted name/number or by a plain number\0"
399    "(?+ or (?- or (?(+ or (?(- must be followed by a non-zero number"    "a numbered reference must not be zero\0"
400  };    "an argument is not allowed for (*ACCEPT), (*FAIL), or (*COMMIT)\0"
401      /* 60 */
402      "(*VERB) not recognized\0"
403      "number is too big\0"
404      "subpattern name expected\0"
405      "digit expected after (?+\0"
406      "] is an invalid data character in JavaScript compatibility mode\0"
407      /* 65 */
408      "different names for subpatterns of the same number are not allowed\0"
409      "(*MARK) must have an argument\0"
410      "this version of PCRE is not compiled with PCRE_UCP support\0"
411      ;
412    
413  /* Table to identify digits and hex digits. This is used when compiling  /* Table to identify digits and hex digits. This is used when compiling
414  patterns. Note that the tables in chartables are dependent on the locale, and  patterns. Note that the tables in chartables are dependent on the locale, and
# Line 271  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 307  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 388  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 408  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 426  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 451  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 482  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 491  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 || (braced && *(++ptr) != '}'))      if (c < 0)   /* Integer overflow */
701          {
702          *errorcodeptr = ERR61;
703          break;
704          }
705    
706        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 504  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 524  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 545  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 558  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 569  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 579  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 607  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 623  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 657  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 697  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 708  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 733  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 776  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 817  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 827  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 857  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;    /* Handle specials such as (*SKIP) or (*UTF8) etc. */
1133    
1134      if (ptr[1] == CHAR_ASTERISK) ptr += 2;
1135    
1136      /* Handle a normal, unnamed capturing parenthesis. */
1137    
1138      else if (ptr[1] != CHAR_QUESTION_MARK)
1139        {
1140        *count += 1;
1141        if (name == NULL && *count == lorn) return *count;
1142        ptr++;
1143        }
1144    
1145      /* All cases now have (? at the start. Remember when we are in a group
1146      where the parenthesis numbers are duplicated. */
1147    
1148      else if (ptr[2] == CHAR_VERTICAL_LINE)
1149        {
1150        ptr += 3;
1151        dup_parens = TRUE;
1152        }
1153    
1154      /* Handle comments; all characters are allowed until a ket is reached. */
1155    
1156      else if (ptr[2] == CHAR_NUMBER_SIGN)
1157        {
1158        for (ptr += 3; *ptr != 0; ptr++) if (*ptr == CHAR_RIGHT_PARENTHESIS) break;
1159        goto FAIL_EXIT;
1160        }
1161    
1162      /* Handle a condition. If it is an assertion, just carry on so that it
1163      is processed as normal. If not, skip to the closing parenthesis of the
1164      condition (there can't be any nested parens). */
1165    
1166      else if (ptr[2] == CHAR_LEFT_PARENTHESIS)
1167        {
1168        ptr += 2;
1169        if (ptr[1] != CHAR_QUESTION_MARK)
1170          {
1171          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
1172          if (*ptr != 0) ptr++;
1173          }
1174        }
1175    
1176      /* Start with (? but not a condition. */
1177    
1178      else
1179        {
1180        ptr += 2;
1181        if (*ptr == CHAR_P) ptr++;                      /* Allow optional P */
1182    
1183        /* We have to disambiguate (?<! and (?<= from (?<name> for named groups */
1184    
1185        if ((*ptr == CHAR_LESS_THAN_SIGN && ptr[1] != CHAR_EXCLAMATION_MARK &&
1186            ptr[1] != CHAR_EQUALS_SIGN) || *ptr == CHAR_APOSTROPHE)
1187          {
1188          int term;
1189          const uschar *thisname;
1190          *count += 1;
1191          if (name == NULL && *count == lorn) return *count;
1192          term = *ptr++;
1193          if (term == CHAR_LESS_THAN_SIGN) term = CHAR_GREATER_THAN_SIGN;
1194          thisname = ptr;
1195          while (*ptr != term) ptr++;
1196          if (name != NULL && lorn == ptr - thisname &&
1197              strncmp((const char *)name, (const char *)thisname, lorn) == 0)
1198            return *count;
1199          term++;
1200          }
1201        }
1202      }
1203    
1204    /* Past any initial parenthesis handling, scan for parentheses or vertical
1205    bars. */
1206    
1207    for (; *ptr != 0; ptr++)
1208      {
1209    /* Skip over backslashed characters and also entire \Q...\E */    /* Skip over backslashed characters and also entire \Q...\E */
1210    
1211    if (*ptr == '\\')    if (*ptr == CHAR_BACKSLASH)
1212      {      {
1213      if (*(++ptr) == 0) return -1;      if (*(++ptr) == 0) goto FAIL_EXIT;
1214      if (*ptr == 'Q') for (;;)      if (*ptr == CHAR_Q) for (;;)
1215        {        {
1216        while (*(++ptr) != 0 && *ptr != '\\');        while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1217        if (*ptr == 0) return -1;        if (*ptr == 0) goto FAIL_EXIT;
1218        if (*(++ptr) == 'E') break;        if (*(++ptr) == CHAR_E) break;
1219        }        }
1220      continue;      continue;
1221      }      }
1222    
1223    /* Skip over character classes */    /* Skip over character classes; this logic must be similar to the way they
1224      are handled for real. If the first character is '^', skip it. Also, if the
1225      first few characters (either before or after ^) are \Q\E or \E we skip them
1226      too. This makes for compatibility with Perl. Note the use of STR macros to
1227      encode "Q\\E" so that it works in UTF-8 on EBCDIC platforms. */
1228    
1229      if (*ptr == CHAR_LEFT_SQUARE_BRACKET)
1230        {
1231        BOOL negate_class = FALSE;
1232        for (;;)
1233          {
1234          if (ptr[1] == CHAR_BACKSLASH)
1235            {
1236            if (ptr[2] == CHAR_E)
1237              ptr+= 2;
1238            else if (strncmp((const char *)ptr+2,
1239                     STR_Q STR_BACKSLASH STR_E, 3) == 0)
1240              ptr += 4;
1241            else
1242              break;
1243            }
1244          else if (!negate_class && ptr[1] == CHAR_CIRCUMFLEX_ACCENT)
1245            {
1246            negate_class = TRUE;
1247            ptr++;
1248            }
1249          else break;
1250          }
1251    
1252        /* If the next character is ']', it is a data character that must be
1253        skipped, except in JavaScript compatibility mode. */
1254    
1255    if (*ptr == '[')      if (ptr[1] == CHAR_RIGHT_SQUARE_BRACKET &&
1256      {          (cd->external_options & PCRE_JAVASCRIPT_COMPAT) == 0)
1257      while (*(++ptr) != ']')        ptr++;
1258    
1259        while (*(++ptr) != CHAR_RIGHT_SQUARE_BRACKET)
1260        {        {
1261        if (*ptr == '\\')        if (*ptr == 0) return -1;
1262          if (*ptr == CHAR_BACKSLASH)
1263          {          {
1264          if (*(++ptr) == 0) return -1;          if (*(++ptr) == 0) goto FAIL_EXIT;
1265          if (*ptr == 'Q') for (;;)          if (*ptr == CHAR_Q) for (;;)
1266            {            {
1267            while (*(++ptr) != 0 && *ptr != '\\');            while (*(++ptr) != 0 && *ptr != CHAR_BACKSLASH) {};
1268            if (*ptr == 0) return -1;            if (*ptr == 0) goto FAIL_EXIT;
1269            if (*(++ptr) == 'E') break;            if (*(++ptr) == CHAR_E) break;
1270            }            }
1271          continue;          continue;
1272          }          }
# Line 924  for (; *ptr != 0; ptr++) Line 1276  for (; *ptr != 0; ptr++)
1276    
1277    /* Skip comments in /x mode */    /* Skip comments in /x mode */
1278    
1279    if (xmode && *ptr == '#')    if (xmode && *ptr == CHAR_NUMBER_SIGN)
1280      {      {
1281      while (*(++ptr) != 0 && *ptr != '\n');      while (*(++ptr) != 0 && *ptr != CHAR_NL) {};
1282      if (*ptr == 0) return -1;      if (*ptr == 0) goto FAIL_EXIT;
1283      continue;      continue;
1284      }      }
1285    
1286    /* An opening parens must now be a real metacharacter */    /* Check for the special metacharacters */
1287    
1288    if (*ptr != '(') continue;    if (*ptr == CHAR_LEFT_PARENTHESIS)
   if (ptr[1] != '?')  
1289      {      {
1290      count++;      int rc = find_parens_sub(&ptr, cd, name, lorn, xmode, count);
1291      if (name == NULL && count == lorn) return count;      if (rc > 0) return rc;
1292      continue;      if (*ptr == 0) goto FAIL_EXIT;
1293      }      }
1294    
1295    ptr += 2;    else if (*ptr == CHAR_RIGHT_PARENTHESIS)
1296    if (*ptr == 'P') ptr++;                      /* Allow optional P */      {
1297        if (dup_parens && *count < hwm_count) *count = hwm_count;
1298        goto FAIL_EXIT;
1299        }
1300    
1301    /* We have to disambiguate (?<! and (?<= from (?<name> */    else if (*ptr == CHAR_VERTICAL_LINE && dup_parens)
1302        {
1303        if (*count > hwm_count) hwm_count = *count;
1304        *count = start_count;
1305        }
1306      }
1307    
1308    if ((*ptr != '<' || ptr[1] == '!' || ptr[1] == '=') &&  FAIL_EXIT:
1309         *ptr != '\'')  *ptrptr = ptr;
1310      continue;  return -1;
1311    }
1312    
1313    
1314    
1315    
1316    /*************************************************
1317    *       Find forward referenced subpattern       *
1318    *************************************************/
1319    
1320    /* This function scans along a pattern's text looking for capturing
1321    subpatterns, and counting them. If it finds a named pattern that matches the
1322    name it is given, it returns its number. Alternatively, if the name is NULL, it
1323    returns when it reaches a given numbered subpattern. This is used for forward
1324    references to subpatterns. We used to be able to start this scan from the
1325    current compiling point, using the current count value from cd->bracount, and
1326    do it all in a single loop, but the addition of the possibility of duplicate
1327    subpattern numbers means that we have to scan from the very start, in order to
1328    take account of such duplicates, and to use a recursive function to keep track
1329    of the different types of group.
1330    
1331    Arguments:
1332      cd           compile background data
1333      name         name to seek, or NULL if seeking a numbered subpattern
1334      lorn         name length, or subpattern number if name is NULL
1335      xmode        TRUE if we are in /x mode
1336    
1337    Returns:       the number of the found subpattern, or -1 if not found
1338    */
1339    
1340    count++;  static int
1341    find_parens(compile_data *cd, const uschar *name, int lorn, BOOL xmode)
1342    {
1343    uschar *ptr = (uschar *)cd->start_pattern;
1344    int count = 0;
1345    int rc;
1346    
1347    /* If the pattern does not start with an opening parenthesis, the first call
1348    to find_parens_sub() will scan right to the end (if necessary). However, if it
1349    does start with a parenthesis, find_parens_sub() will return when it hits the
1350    matching closing parens. That is why we have to have a loop. */
1351    
1352    if (name == NULL && count == lorn) return count;  for (;;)
1353    term = *ptr++;    {
1354    if (term == '<') term = '>';    rc = find_parens_sub(&ptr, cd, name, lorn, xmode, &count);
1355    thisname = ptr;    if (rc > 0 || *ptr++ == 0) break;
   while (*ptr != term) ptr++;  
   if (name != NULL && lorn == ptr - thisname &&  
       strncmp((const char *)name, (const char *)thisname, lorn) == 0)  
     return count;  
1356    }    }
1357    
1358  return -1;  return rc;
1359  }  }
1360    
1361    
1362    
1363    
1364  /*************************************************  /*************************************************
1365  *      Find first significant op code            *  *      Find first significant op code            *
1366  *************************************************/  *************************************************/
# Line 1016  for (;;) Line 1410  for (;;)
1410    
1411      case OP_CALLOUT:      case OP_CALLOUT:
1412      case OP_CREF:      case OP_CREF:
1413        case OP_NCREF:
1414      case OP_RREF:      case OP_RREF:
1415        case OP_NRREF:
1416      case OP_DEF:      case OP_DEF:
1417      code += _pcre_OP_lengths[*code];      code += _pcre_OP_lengths[*code];
1418      break;      break;
# Line 1032  for (;;) Line 1428  for (;;)
1428    
1429    
1430  /*************************************************  /*************************************************
1431  *        Find the fixed length of a pattern      *  *        Find the fixed length of a branch       *
1432  *************************************************/  *************************************************/
1433    
1434  /* 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,
1435  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.
1436  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
1437    temporarily terminated with OP_END when this function is called.
1438    
1439    This function is called when a backward assertion is encountered, so that if it
1440    fails, the error message can point to the correct place in the pattern.
1441    However, we cannot do this when the assertion contains subroutine calls,
1442    because they can be forward references. We solve this by remembering this case
1443    and doing the check at the end; a flag specifies which mode we are running in.
1444    
1445  Arguments:  Arguments:
1446    code     points to the start of the pattern (the bracket)    code     points to the start of the pattern (the bracket)
1447    options  the compiling options    options  the compiling options
1448      atend    TRUE if called when the pattern is complete
1449      cd       the "compile data" structure
1450    
1451  Returns:   the fixed length, or -1 if there is no fixed length,  Returns:   the fixed length,
1452                 or -1 if there is no fixed length,
1453               or -2 if \C was encountered               or -2 if \C was encountered
1454                 or -3 if an OP_RECURSE item was encountered and atend is FALSE
1455  */  */
1456    
1457  static int  static int
1458  find_fixedlength(uschar *code, int options)  find_fixedlength(uschar *code, int options, BOOL atend, compile_data *cd)
1459  {  {
1460  int length = -1;  int length = -1;
1461    
# Line 1061  branch, check the length against that of Line 1468  branch, check the length against that of
1468  for (;;)  for (;;)
1469    {    {
1470    int d;    int d;
1471      uschar *ce, *cs;
1472    register int op = *cc;    register int op = *cc;
   
1473    switch (op)    switch (op)
1474      {      {
1475      case OP_CBRA:      case OP_CBRA:
1476      case OP_BRA:      case OP_BRA:
1477      case OP_ONCE:      case OP_ONCE:
1478      case OP_COND:      case OP_COND:
1479      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options);      d = find_fixedlength(cc + ((op == OP_CBRA)? 2:0), options, atend, cd);
1480      if (d < 0) return d;      if (d < 0) return d;
1481      branchlength += d;      branchlength += d;
1482      do cc += GET(cc, 1); while (*cc == OP_ALT);      do cc += GET(cc, 1); while (*cc == OP_ALT);
# Line 1092  for (;;) Line 1499  for (;;)
1499      branchlength = 0;      branchlength = 0;
1500      break;      break;
1501    
1502        /* A true recursion implies not fixed length, but a subroutine call may
1503        be OK. If the subroutine is a forward reference, we can't deal with
1504        it until the end of the pattern, so return -3. */
1505    
1506        case OP_RECURSE:
1507        if (!atend) return -3;
1508        cs = ce = (uschar *)cd->start_code + GET(cc, 1);  /* Start subpattern */
1509        do ce += GET(ce, 1); while (*ce == OP_ALT);       /* End subpattern */
1510        if (cc > cs && cc < ce) return -1;                /* Recursion */
1511        d = find_fixedlength(cs + 2, options, atend, cd);
1512        if (d < 0) return d;
1513        branchlength += d;
1514        cc += 1 + LINK_SIZE;
1515        break;
1516    
1517      /* Skip over assertive subpatterns */      /* Skip over assertive subpatterns */
1518    
1519      case OP_ASSERT:      case OP_ASSERT:
# Line 1105  for (;;) Line 1527  for (;;)
1527    
1528      case OP_REVERSE:      case OP_REVERSE:
1529      case OP_CREF:      case OP_CREF:
1530        case OP_NCREF:
1531      case OP_RREF:      case OP_RREF:
1532        case OP_NRREF:
1533      case OP_DEF:      case OP_DEF:
1534      case OP_OPT:      case OP_OPT:
1535      case OP_CALLOUT:      case OP_CALLOUT:
1536      case OP_SOD:      case OP_SOD:
1537      case OP_SOM:      case OP_SOM:
1538        case OP_SET_SOM:
1539      case OP_EOD:      case OP_EOD:
1540      case OP_EODN:      case OP_EODN:
1541      case OP_CIRC:      case OP_CIRC:
# Line 1128  for (;;) Line 1553  for (;;)
1553      branchlength++;      branchlength++;
1554      cc += 2;      cc += 2;
1555  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1556      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1557        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while ((*cc & 0xc0) == 0x80) cc++;  
       }  
1558  #endif  #endif
1559      break;      break;
1560    
# Line 1142  for (;;) Line 1565  for (;;)
1565      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1566      cc += 4;      cc += 4;
1567  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1568      if ((options & PCRE_UTF8) != 0)      if ((options & PCRE_UTF8) != 0 && cc[-1] >= 0xc0)
1569        {        cc += _pcre_utf8_table4[cc[-1] & 0x3f];
       while((*cc & 0x80) == 0x80) cc++;  
       }  
1570  #endif  #endif
1571      break;      break;
1572    
1573      case OP_TYPEEXACT:      case OP_TYPEEXACT:
1574      branchlength += GET2(cc,1);      branchlength += GET2(cc,1);
1575        if (cc[3] == OP_PROP || cc[3] == OP_NOTPROP) cc += 2;
1576      cc += 4;      cc += 4;
1577      break;      break;
1578    
# Line 1168  for (;;) Line 1590  for (;;)
1590      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
1591      case OP_WORDCHAR:      case OP_WORDCHAR:
1592      case OP_ANY:      case OP_ANY:
1593        case OP_ALLANY:
1594      branchlength++;      branchlength++;
1595      cc++;      cc++;
1596      break;      break;
# Line 1222  for (;;) Line 1645  for (;;)
1645    
1646    
1647  /*************************************************  /*************************************************
1648  *    Scan compiled regex for numbered bracket    *  *    Scan compiled regex for specific bracket    *
1649  *************************************************/  *************************************************/
1650    
1651  /* This little function scans through a compiled pattern until it finds a  /* This little function scans through a compiled pattern until it finds a
1652  capturing bracket with the given number.  capturing bracket with the given number, or, if the number is negative, an
1653    instance of OP_REVERSE for a lookbehind. The function is global in the C sense
1654    so that it can be called from pcre_study() when finding the minimum matching
1655    length.
1656    
1657  Arguments:  Arguments:
1658    code        points to start of expression    code        points to start of expression
1659    utf8        TRUE in UTF-8 mode    utf8        TRUE in UTF-8 mode
1660    number      the required bracket number    number      the required bracket number or negative to find a lookbehind
1661    
1662  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
1663  */  */
1664    
1665  static const uschar *  const uschar *
1666  find_bracket(const uschar *code, BOOL utf8, int number)  _pcre_find_bracket(const uschar *code, BOOL utf8, int number)
1667  {  {
1668  for (;;)  for (;;)
1669    {    {
# Line 1250  for (;;) Line 1676  for (;;)
1676    
1677    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1678    
1679      /* Handle recursion */
1680    
1681      else if (c == OP_REVERSE)
1682        {
1683        if (number < 0) return (uschar *)code;
1684        code += _pcre_OP_lengths[c];
1685        }
1686    
1687    /* Handle capturing bracket */    /* Handle capturing bracket */
1688    
1689    else if (c == OP_CBRA)    else if (c == OP_CBRA)
# Line 1259  for (;;) Line 1693  for (;;)
1693      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1694      }      }
1695    
1696    /* 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
1697    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
1698    arrange to skip the extra bytes. */    two bytes of parameters, and for MARK/PRUNE/SKIP/THEN with an argument, we
1699      must add in its length. */
1700    
1701    else    else
1702      {      {
1703        switch(c)
1704          {
1705          case OP_TYPESTAR:
1706          case OP_TYPEMINSTAR:
1707          case OP_TYPEPLUS:
1708          case OP_TYPEMINPLUS:
1709          case OP_TYPEQUERY:
1710          case OP_TYPEMINQUERY:
1711          case OP_TYPEPOSSTAR:
1712          case OP_TYPEPOSPLUS:
1713          case OP_TYPEPOSQUERY:
1714          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1715          break;
1716    
1717          case OP_TYPEUPTO:
1718          case OP_TYPEMINUPTO:
1719          case OP_TYPEEXACT:
1720          case OP_TYPEPOSUPTO:
1721          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1722          break;
1723    
1724          case OP_MARK:
1725          case OP_PRUNE_ARG:
1726          case OP_SKIP_ARG:
1727          case OP_THEN_ARG:
1728          code += code[1];
1729          break;
1730          }
1731    
1732        /* Add in the fixed length from the table */
1733    
1734      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1735    
1736      /* In UTF-8 mode, opcodes that are followed by a character may be followed by
1737      a multi-byte character. The length in the table is a minimum, so we have to
1738      arrange to skip the extra bytes. */
1739    
1740  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1741      if (utf8) switch(c)      if (utf8) switch(c)
1742        {        {
# Line 1287  for (;;) Line 1758  for (;;)
1758        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1759        break;        break;
1760        }        }
1761    #else
1762        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1763  #endif  #endif
1764      }      }
1765    }    }
# Line 1323  for (;;) Line 1796  for (;;)
1796    
1797    if (c == OP_XCLASS) code += GET(code, 1);    if (c == OP_XCLASS) code += GET(code, 1);
1798    
1799    /* 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
1800    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
1801    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
1802    bytes. */    must add in its length. */
1803    
1804    else    else
1805      {      {
1806        switch(c)
1807          {
1808          case OP_TYPESTAR:
1809          case OP_TYPEMINSTAR:
1810          case OP_TYPEPLUS:
1811          case OP_TYPEMINPLUS:
1812          case OP_TYPEQUERY:
1813          case OP_TYPEMINQUERY:
1814          case OP_TYPEPOSSTAR:
1815          case OP_TYPEPOSPLUS:
1816          case OP_TYPEPOSQUERY:
1817          if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
1818          break;
1819    
1820          case OP_TYPEPOSUPTO:
1821          case OP_TYPEUPTO:
1822          case OP_TYPEMINUPTO:
1823          case OP_TYPEEXACT:
1824          if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
1825          break;
1826    
1827          case OP_MARK:
1828          case OP_PRUNE_ARG:
1829          case OP_SKIP_ARG:
1830          case OP_THEN_ARG:
1831          code += code[1];
1832          break;
1833          }
1834    
1835        /* Add in the fixed length from the table */
1836    
1837      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1838    
1839        /* In UTF-8 mode, opcodes that are followed by a character may be followed
1840        by a multi-byte character. The length in the table is a minimum, so we have
1841        to arrange to skip the extra bytes. */
1842    
1843  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1844      if (utf8) switch(c)      if (utf8) switch(c)
1845        {        {
# Line 1352  for (;;) Line 1861  for (;;)
1861        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];        if (code[-1] >= 0xc0) code += _pcre_utf8_table4[code[-1] & 0x3f];
1862        break;        break;
1863        }        }
1864    #else
1865        (void)(utf8);  /* Keep compiler happy by referencing function argument */
1866  #endif  #endif
1867      }      }
1868    }    }
# Line 1367  for (;;) Line 1878  for (;;)
1878  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()
1879  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
1880  group that can match nothing. Note that first_significant_code() skips over  group that can match nothing. Note that first_significant_code() skips over
1881  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
1882  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
1883    bracket whose current branch will already have been scanned.
1884    
1885  Arguments:  Arguments:
1886    code        points to start of search    code        points to start of search
1887    endcode     points to where to stop    endcode     points to where to stop
1888    utf8        TRUE if in UTF8 mode    utf8        TRUE if in UTF8 mode
1889      cd          contains pointers to tables etc.
1890    
1891  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
1892  */  */
1893    
1894  static BOOL  static BOOL
1895  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8)  could_be_empty_branch(const uschar *code, const uschar *endcode, BOOL utf8,
1896      compile_data *cd)
1897  {  {
1898  register int c;  register int c;
1899  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 1390  for (code = first_significant_code(code Line 1904  for (code = first_significant_code(code
1904    
1905    c = *code;    c = *code;
1906    
1907      /* Skip over forward assertions; the other assertions are skipped by
1908      first_significant_code() with a TRUE final argument. */
1909    
1910      if (c == OP_ASSERT)
1911        {
1912        do code += GET(code, 1); while (*code == OP_ALT);
1913        c = *code;
1914        continue;
1915        }
1916    
1917    /* Groups with zero repeats can of course be empty; skip them. */    /* Groups with zero repeats can of course be empty; skip them. */
1918    
1919    if (c == OP_BRAZERO || c == OP_BRAMINZERO)    if (c == OP_BRAZERO || c == OP_BRAMINZERO || c == OP_SKIPZERO)
1920      {      {
1921      code += _pcre_OP_lengths[c];      code += _pcre_OP_lengths[c];
1922      do code += GET(code, 1); while (*code == OP_ALT);      do code += GET(code, 1); while (*code == OP_ALT);
# Line 1400  for (code = first_significant_code(code Line 1924  for (code = first_significant_code(code
1924      continue;      continue;
1925      }      }
1926    
1927      /* For a recursion/subroutine call, if its end has been reached, which
1928      implies a subroutine call, we can scan it. */
1929    
1930      if (c == OP_RECURSE)
1931        {
1932        BOOL empty_branch = FALSE;
1933        const uschar *scode = cd->start_code + GET(code, 1);
1934        if (GET(scode, 1) == 0) return TRUE;    /* Unclosed */
1935        do
1936          {
1937          if (could_be_empty_branch(scode, endcode, utf8, cd))
1938            {
1939            empty_branch = TRUE;
1940            break;
1941            }
1942          scode += GET(scode, 1);
1943          }
1944        while (*scode == OP_ALT);
1945        if (!empty_branch) return FALSE;  /* All branches are non-empty */
1946        continue;
1947        }
1948    
1949    /* For other groups, scan the branches. */    /* For other groups, scan the branches. */
1950    
1951    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE)    if (c == OP_BRA || c == OP_CBRA || c == OP_ONCE || c == OP_COND)
1952      {      {
1953      BOOL empty_branch;      BOOL empty_branch;
1954      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */      if (GET(code, 1) == 0) return TRUE;    /* Hit unclosed bracket */
1955    
1956      /* Scan a closed bracket */      /* If a conditional group has only one branch, there is a second, implied,
1957        empty branch, so just skip over the conditional, because it could be empty.
1958        Otherwise, scan the individual branches of the group. */
1959    
1960      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;  
1961        code += GET(code, 1);        code += GET(code, 1);
1962        else
1963          {
1964          empty_branch = FALSE;
1965          do
1966            {
1967            if (!empty_branch && could_be_empty_branch(code, endcode, utf8, cd))
1968              empty_branch = TRUE;
1969            code += GET(code, 1);
1970            }
1971          while (*code == OP_ALT);
1972          if (!empty_branch) return FALSE;   /* All branches are non-empty */
1973        }        }
1974      while (*code == OP_ALT);  
     if (!empty_branch) return FALSE;   /* All branches are non-empty */  
1975      c = *code;      c = *code;
1976      continue;      continue;
1977      }      }
# Line 1426  for (code = first_significant_code(code Line 1980  for (code = first_significant_code(code
1980    
1981    switch (c)    switch (c)
1982      {      {
1983      /* Check for quantifiers after a class */      /* Check for quantifiers after a class. XCLASS is used for classes that
1984        cannot be represented just by a bit map. This includes negated single
1985        high-valued characters. The length in _pcre_OP_lengths[] is zero; the
1986        actual length is stored in the compiled code, so we must update "code"
1987        here. */
1988    
1989  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
1990      case OP_XCLASS:      case OP_XCLASS:
1991      ccode = code + GET(code, 1);      ccode = code += GET(code, 1);
1992      goto CHECK_CLASS_REPEAT;      goto CHECK_CLASS_REPEAT;
1993  #endif  #endif
1994    
# Line 1474  for (code = first_significant_code(code Line 2032  for (code = first_significant_code(code
2032      case OP_NOT_WORDCHAR:      case OP_NOT_WORDCHAR:
2033      case OP_WORDCHAR:      case OP_WORDCHAR:
2034      case OP_ANY:      case OP_ANY:
2035        case OP_ALLANY:
2036      case OP_ANYBYTE:      case OP_ANYBYTE:
2037      case OP_CHAR:      case OP_CHAR:
2038      case OP_CHARNC:      case OP_CHARNC:
# Line 1492  for (code = first_significant_code(code Line 2051  for (code = first_significant_code(code
2051      case OP_TYPEEXACT:      case OP_TYPEEXACT:
2052      return FALSE;      return FALSE;
2053    
2054        /* These are going to continue, as they may be empty, but we have to
2055        fudge the length for the \p and \P cases. */
2056    
2057        case OP_TYPESTAR:
2058        case OP_TYPEMINSTAR:
2059        case OP_TYPEPOSSTAR:
2060        case OP_TYPEQUERY:
2061        case OP_TYPEMINQUERY:
2062        case OP_TYPEPOSQUERY:
2063        if (code[1] == OP_PROP || code[1] == OP_NOTPROP) code += 2;
2064        break;
2065    
2066        /* Same for these */
2067    
2068        case OP_TYPEUPTO:
2069        case OP_TYPEMINUPTO:
2070        case OP_TYPEPOSUPTO:
2071        if (code[3] == OP_PROP || code[3] == OP_NOTPROP) code += 2;
2072        break;
2073    
2074      /* End of branch */      /* End of branch */
2075    
2076      case OP_KET:      case OP_KET:
# Line 1510  for (code = first_significant_code(code Line 2089  for (code = first_significant_code(code
2089      case OP_QUERY:      case OP_QUERY:
2090      case OP_MINQUERY:      case OP_MINQUERY:
2091      case OP_POSQUERY:      case OP_POSQUERY:
2092        if (utf8 && code[1] >= 0xc0) code += _pcre_utf8_table4[code[1] & 0x3f];
2093        break;
2094    
2095      case OP_UPTO:      case OP_UPTO:
2096      case OP_MINUPTO:      case OP_MINUPTO:
2097      case OP_POSUPTO:      case OP_POSUPTO:
2098      if (utf8) while ((code[2] & 0xc0) == 0x80) code++;      if (utf8 && code[3] >= 0xc0) code += _pcre_utf8_table4[code[3] & 0x3f];
2099      break;      break;
2100  #endif  #endif
2101    
2102        /* MARK, and PRUNE/SKIP/THEN with an argument must skip over the argument
2103        string. */
2104    
2105        case OP_MARK:
2106        case OP_PRUNE_ARG:
2107        case OP_SKIP_ARG:
2108        case OP_THEN_ARG:
2109        code += code[1];
2110        break;
2111    
2112        /* None of the remaining opcodes are required to match a character. */
2113    
2114        default:
2115        break;
2116      }      }
2117    }    }
2118    
# Line 1538  Arguments: Line 2135  Arguments:
2135    endcode     points to where to stop (current RECURSE item)    endcode     points to where to stop (current RECURSE item)
2136    bcptr       points to the chain of current (unclosed) branch starts    bcptr       points to the chain of current (unclosed) branch starts
2137    utf8        TRUE if in UTF-8 mode    utf8        TRUE if in UTF-8 mode
2138      cd          pointers to tables etc
2139    
2140  Returns:      TRUE if what is matched could be empty  Returns:      TRUE if what is matched could be empty
2141  */  */
2142    
2143  static BOOL  static BOOL
2144  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,  could_be_empty(const uschar *code, const uschar *endcode, branch_chain *bcptr,
2145    BOOL utf8)    BOOL utf8, compile_data *cd)
2146  {  {
2147  while (bcptr != NULL && bcptr->current >= code)  while (bcptr != NULL && bcptr->current_branch >= code)
2148    {    {
2149    if (!could_be_empty_branch(bcptr->current, endcode, utf8)) return FALSE;    if (!could_be_empty_branch(bcptr->current_branch, endcode, utf8, cd))
2150        return FALSE;
2151    bcptr = bcptr->outer;    bcptr = bcptr->outer;
2152    }    }
2153  return TRUE;  return TRUE;
# Line 1561  return TRUE; Line 2160  return TRUE;
2160  *************************************************/  *************************************************/
2161    
2162  /* This function is called when the sequence "[:" or "[." or "[=" is  /* This function is called when the sequence "[:" or "[." or "[=" is
2163  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
2164  optional ^ and then a sequence of letters, terminated by a matching ":]" or  sequence of characters terminated by a matching ":]" or ".]" or "=]". If we
2165  ".]" or "=]".  reach an unescaped ']' without the special preceding character, return FALSE.
2166    
2167    Originally, this function only recognized a sequence of letters between the
2168    terminators, but it seems that Perl recognizes any sequence of characters,
2169    though of course unknown POSIX names are subsequently rejected. Perl gives an
2170    "Unknown POSIX class" error for [:f\oo:] for example, where previously PCRE
2171    didn't consider this to be a POSIX class. Likewise for [:1234:].
2172    
2173    The problem in trying to be exactly like Perl is in the handling of escapes. We
2174    have to be sure that [abc[:x\]pqr] is *not* treated as containing a POSIX
2175    class, but [abc[:x\]pqr:]] is (so that an error can be generated). The code
2176    below handles the special case of \], but does not try to do any other escape
2177    processing. This makes it different from Perl for cases such as [:l\ower:]
2178    where Perl recognizes it as the POSIX class "lower" but PCRE does not recognize
2179    "l\ower". This is a lesser evil that not diagnosing bad classes when Perl does,
2180    I think.
2181    
2182  Argument:  Arguments:
2183    ptr      pointer to the initial [    ptr      pointer to the initial [
2184    endptr   where to return the end pointer    endptr   where to return the end pointer
   cd       pointer to compile data  
2185    
2186  Returns:   TRUE or FALSE  Returns:   TRUE or FALSE
2187  */  */
2188    
2189  static BOOL  static BOOL
2190  check_posix_syntax(const uschar *ptr, const uschar **endptr, compile_data *cd)  check_posix_syntax(const uschar *ptr, const uschar **endptr)
2191  {  {
2192  int terminator;          /* Don't combine these lines; the Solaris cc */  int terminator;          /* Don't combine these lines; the Solaris cc */
2193  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */  terminator = *(++ptr);   /* compiler warns about "non-constant" initializer. */
2194  if (*(++ptr) == '^') ptr++;  for (++ptr; *ptr != 0; ptr++)
 while ((cd->ctypes[*ptr] & ctype_letter) != 0) ptr++;  
 if (*ptr == terminator && ptr[1] == ']')  
2195    {    {
2196    *endptr = ptr;    if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET) ptr++; else
2197    return TRUE;      {
2198        if (*ptr == CHAR_RIGHT_SQUARE_BRACKET) return FALSE;
2199        if (*ptr == terminator && ptr[1] == CHAR_RIGHT_SQUARE_BRACKET)
2200          {
2201          *endptr = ptr;
2202          return TRUE;
2203          }
2204        }
2205    }    }
2206  return FALSE;  return FALSE;
2207  }  }
# Line 1608  Returns:     a value representing the na Line 2226  Returns:     a value representing the na
2226  static int  static int
2227  check_posix_name(const uschar *ptr, int len)  check_posix_name(const uschar *ptr, int len)
2228  {  {
2229    const char *pn = posix_names;
2230  register int yield = 0;  register int yield = 0;
2231  while (posix_name_lengths[yield] != 0)  while (posix_name_lengths[yield] != 0)
2232    {    {
2233    if (len == posix_name_lengths[yield] &&    if (len == posix_name_lengths[yield] &&
2234      strncmp((const char *)ptr, posix_names[yield], len) == 0) return yield;      strncmp((const char *)ptr, pn, len) == 0) return yield;
2235      pn += posix_name_lengths[yield] + 1;
2236    yield++;    yield++;
2237    }    }
2238  return -1;  return -1;
# Line 1627  return -1; Line 2247  return -1;
2247  that is referenced. This means that groups can be replicated for fixed  that is referenced. This means that groups can be replicated for fixed
2248  repetition simply by copying (because the recursion is allowed to refer to  repetition simply by copying (because the recursion is allowed to refer to
2249  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
2250  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
2251  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
2252  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
2253  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
2254  the partially compiled regex must be temporarily terminated with OP_END.  is called, the partially compiled regex must be temporarily terminated with
2255    OP_END.
2256    
2257  This function has been extended with the possibility of forward references for  This function has been extended with the possibility of forward references for
2258  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 1654  adjust_recurse(uschar *group, int adjust Line 2275  adjust_recurse(uschar *group, int adjust
2275    uschar *save_hwm)    uschar *save_hwm)
2276  {  {
2277  uschar *ptr = group;  uschar *ptr = group;
2278    
2279  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)  while ((ptr = (uschar *)find_recurse(ptr, utf8)) != NULL)
2280    {    {
2281    int offset;    int offset;
# Line 1707  auto_callout(uschar *code, const uschar Line 2329  auto_callout(uschar *code, const uschar
2329  {  {
2330  *code++ = OP_CALLOUT;  *code++ = OP_CALLOUT;
2331  *code++ = 255;  *code++ = 255;
2332  PUT(code, 0, ptr - cd->start_pattern);  /* Pattern offset */  PUT(code, 0, (int)(ptr - cd->start_pattern));  /* Pattern offset */
2333  PUT(code, LINK_SIZE, 0);                /* Default length */  PUT(code, LINK_SIZE, 0);                       /* Default length */
2334  return code + 2*LINK_SIZE;  return code + 2*LINK_SIZE;
2335  }  }
2336    
# Line 1733  Returns:             nothing Line 2355  Returns:             nothing
2355  static void  static void
2356  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)  complete_callout(uschar *previous_callout, const uschar *ptr, compile_data *cd)
2357  {  {
2358  int length = ptr - cd->start_pattern - GET(previous_callout, 2);  int length = (int)(ptr - cd->start_pattern - GET(previous_callout, 2));
2359  PUT(previous_callout, 2 + LINK_SIZE, length);  PUT(previous_callout, 2 + LINK_SIZE, length);
2360  }  }
2361    
# Line 1765  get_othercase_range(unsigned int *cptr, Line 2387  get_othercase_range(unsigned int *cptr,
2387  unsigned int c, othercase, next;  unsigned int c, othercase, next;
2388    
2389  for (c = *cptr; c <= d; c++)  for (c = *cptr; c <= d; c++)
2390    { if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR) break; }    { if ((othercase = UCD_OTHERCASE(c)) != c) break; }
2391    
2392  if (c > d) return FALSE;  if (c > d) return FALSE;
2393    
# Line 1774  next = othercase + 1; Line 2396  next = othercase + 1;
2396    
2397  for (++c; c <= d; c++)  for (++c; c <= d; c++)
2398    {    {
2399    if (_pcre_ucp_othercase(c) != next) break;    if (UCD_OTHERCASE(c) != next) break;
2400    next++;    next++;
2401    }    }
2402    
# Line 1783  for (++c; c <= d; c++) Line 2405  for (++c; c <= d; c++)
2405    
2406  return TRUE;  return TRUE;
2407  }  }
2408    
2409    
2410    
2411    /*************************************************
2412    *        Check a character and a property        *
2413    *************************************************/
2414    
2415    /* This function is called by check_auto_possessive() when a property item
2416    is adjacent to a fixed character.
2417    
2418    Arguments:
2419      c            the character
2420      ptype        the property type
2421      pdata        the data for the type
2422      negated      TRUE if it's a negated property (\P or \p{^)
2423    
2424    Returns:       TRUE if auto-possessifying is OK
2425    */
2426    
2427    static BOOL
2428    check_char_prop(int c, int ptype, int pdata, BOOL negated)
2429    {
2430    const ucd_record *prop = GET_UCD(c);
2431    switch(ptype)
2432      {
2433      case PT_LAMP:
2434      return (prop->chartype == ucp_Lu ||
2435              prop->chartype == ucp_Ll ||
2436              prop->chartype == ucp_Lt) == negated;
2437    
2438      case PT_GC:
2439      return (pdata == _pcre_ucp_gentype[prop->chartype]) == negated;
2440    
2441      case PT_PC:
2442      return (pdata == prop->chartype) == negated;
2443    
2444      case PT_SC:
2445      return (pdata == prop->script) == negated;
2446    
2447      /* These are specials */
2448    
2449      case PT_ALNUM:
2450      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2451              _pcre_ucp_gentype[prop->chartype] == ucp_N) == negated;
2452    
2453      case PT_SPACE:    /* Perl space */
2454      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2455              c == CHAR_HT || c == CHAR_NL || c == CHAR_FF || c == CHAR_CR)
2456              == negated;
2457    
2458      case PT_PXSPACE:  /* POSIX space */
2459      return (_pcre_ucp_gentype[prop->chartype] == ucp_Z ||
2460              c == CHAR_HT || c == CHAR_NL || c == CHAR_VT ||
2461              c == CHAR_FF || c == CHAR_CR)
2462              == negated;
2463    
2464      case PT_WORD:
2465      return (_pcre_ucp_gentype[prop->chartype] == ucp_L ||
2466              _pcre_ucp_gentype[prop->chartype] == ucp_N ||
2467              c == CHAR_UNDERSCORE) == negated;
2468      }
2469    return FALSE;
2470    }
2471  #endif  /* SUPPORT_UCP */  #endif  /* SUPPORT_UCP */
2472    
2473    
# Line 1796  whether the next thing could possibly ma Line 2481  whether the next thing could possibly ma
2481  sense to automatically possessify the repeated item.  sense to automatically possessify the repeated item.
2482    
2483  Arguments:  Arguments:
2484    op_code       the repeated op code    previous      pointer to the repeated opcode
   this          data for this item, depends on the opcode  
2485    utf8          TRUE in UTF-8 mode    utf8          TRUE in UTF-8 mode
   utf8_char     used for utf8 character bytes, NULL if not relevant  
2486    ptr           next character in pattern    ptr           next character in pattern
2487    options       options bits    options       options bits
2488    cd            contains pointers to tables etc.    cd            contains pointers to tables etc.
# Line 1808  Returns:        TRUE if possessifying is Line 2491  Returns:        TRUE if possessifying is
2491  */  */
2492    
2493  static BOOL  static BOOL
2494  check_auto_possessive(int op_code, int item, BOOL utf8, uschar *utf8_char,  check_auto_possessive(const uschar *previous, BOOL utf8, const uschar *ptr,
2495    const uschar *ptr, int options, compile_data *cd)    int options, compile_data *cd)
2496  {  {
2497  int next;  int c, next;
2498    int op_code = *previous++;
2499    
2500  /* Skip whitespace and comments in extended mode */  /* Skip whitespace and comments in extended mode */
2501    
# Line 1820  if ((options & PCRE_EXTENDED) != 0) Line 2504  if ((options & PCRE_EXTENDED) != 0)
2504    for (;;)    for (;;)
2505      {      {
2506      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2507      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2508        {        {
2509        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2510          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1832  if ((options & PCRE_EXTENDED) != 0) Line 2516  if ((options & PCRE_EXTENDED) != 0)
2516  /* 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
2517  value is a character, a negative value is an escape value. */  value is a character, a negative value is an escape value. */
2518    
2519  if (*ptr == '\\')  if (*ptr == CHAR_BACKSLASH)
2520    {    {
2521    int temperrorcode = 0;    int temperrorcode = 0;
2522    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);    next = check_escape(&ptr, &temperrorcode, cd->bracount, options, FALSE);
# Line 1857  if ((options & PCRE_EXTENDED) != 0) Line 2541  if ((options & PCRE_EXTENDED) != 0)
2541    for (;;)    for (;;)
2542      {      {
2543      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;      while ((cd->ctypes[*ptr] & ctype_space) != 0) ptr++;
2544      if (*ptr == '#')      if (*ptr == CHAR_NUMBER_SIGN)
2545        {        {
2546        while (*(++ptr) != 0)        while (*(++ptr) != 0)
2547          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }          if (IS_NEWLINE(ptr)) { ptr += cd->nllen; break; }
# Line 1868  if ((options & PCRE_EXTENDED) != 0) Line 2552  if ((options & PCRE_EXTENDED) != 0)
2552    
2553  /* If the next thing is itself optional, we have to give up. */  /* If the next thing is itself optional, we have to give up. */
2554    
2555  if (*ptr == '*' || *ptr == '?' || strncmp((char *)ptr, "{0,", 3) == 0)  if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2556    return FALSE;    strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2557        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. */  
   
2558    
2559  /* Handle cases when the next item is a character. */  /* Now compare the next item with the previous opcode. First, handle cases when
2560    the next item is a character. */
2561    
2562  if (next >= 0) switch(op_code)  if (next >= 0) switch(op_code)
2563    {    {
2564    case OP_CHAR:    case OP_CHAR:
2565  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2566    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2567    #else
2568      c = *previous;
2569  #endif  #endif
2570    return item != next;    return c != next;
2571    
2572    /* 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
2573    Unicode property support, we can use it to test the other case of    Unicode property support, we can use it to test the other case of
# Line 1893  if (next >= 0) switch(op_code) Line 2575  if (next >= 0) switch(op_code)
2575    
2576    case OP_CHARNC:    case OP_CHARNC:
2577  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2578    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2579    #else
2580      c = *previous;
2581  #endif  #endif
2582    if (item == next) return FALSE;    if (c == next) return FALSE;
2583  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2584    if (utf8)    if (utf8)
2585      {      {
2586      unsigned int othercase;      unsigned int othercase;
2587      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2588  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2589      othercase = _pcre_ucp_othercase((unsigned int)next);      othercase = UCD_OTHERCASE((unsigned int)next);
2590  #else  #else
2591      othercase = NOTACHAR;      othercase = NOTACHAR;
2592  #endif  #endif
2593      return (unsigned int)item != othercase;      return (unsigned int)c != othercase;
2594      }      }
2595    else    else
2596  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2597    return (item != cd->fcc[next]);  /* Non-UTF-8 mode */    return (c != cd->fcc[next]);  /* Non-UTF-8 mode */
2598    
2599    /* For OP_NOT, "item" must be a single-byte character. */    /* For OP_NOT, its data is always a single-byte character. */
2600    
2601    case OP_NOT:    case OP_NOT:
2602    if (next < 0) return FALSE;  /* Not a character */    if ((c = *previous) == next) return TRUE;
   if (item == next) return TRUE;  
2603    if ((options & PCRE_CASELESS) == 0) return FALSE;    if ((options & PCRE_CASELESS) == 0) return FALSE;
2604  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2605    if (utf8)    if (utf8)
# Line 1924  if (next >= 0) switch(op_code) Line 2607  if (next >= 0) switch(op_code)
2607      unsigned int othercase;      unsigned int othercase;
2608      if (next < 128) othercase = cd->fcc[next]; else      if (next < 128) othercase = cd->fcc[next]; else
2609  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
2610      othercase = _pcre_ucp_othercase(next);      othercase = UCD_OTHERCASE(next);
2611  #else  #else
2612      othercase = NOTACHAR;      othercase = NOTACHAR;
2613  #endif  #endif
2614      return (unsigned int)item == othercase;      return (unsigned int)c == othercase;
2615      }      }
2616    else    else
2617  #endif  /* SUPPORT_UTF8 */  #endif  /* SUPPORT_UTF8 */
2618    return (item == cd->fcc[next]);  /* Non-UTF-8 mode */    return (c == cd->fcc[next]);  /* Non-UTF-8 mode */
2619    
2620      /* Note that OP_DIGIT etc. are generated only when PCRE_UCP is *not* set.
2621      When it is set, \d etc. are converted into OP_(NOT_)PROP codes. */
2622    
2623    case OP_DIGIT:    case OP_DIGIT:
2624    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;    return next > 127 || (cd->ctypes[next] & ctype_digit) == 0;
# Line 1975  if (next >= 0) switch(op_code) Line 2661  if (next >= 0) switch(op_code)
2661      case 0x202f:      case 0x202f:
2662      case 0x205f:      case 0x205f:
2663      case 0x3000:      case 0x3000:
2664      return op_code != OP_HSPACE;      return op_code == OP_NOT_HSPACE;
2665      default:      default:
2666      return op_code == OP_HSPACE;      return op_code != OP_NOT_HSPACE;
2667      }      }
2668    
2669      case OP_ANYNL:
2670    case OP_VSPACE:    case OP_VSPACE:
2671    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2672    switch(next)    switch(next)
# Line 1991  if (next >= 0) switch(op_code) Line 2678  if (next >= 0) switch(op_code)
2678      case 0x85:      case 0x85:
2679      case 0x2028:      case 0x2028:
2680      case 0x2029:      case 0x2029:
2681      return op_code != OP_VSPACE;      return op_code == OP_NOT_VSPACE;
2682      default:      default:
2683      return op_code == OP_VSPACE;      return op_code != OP_NOT_VSPACE;
2684      }      }
2685    
2686    #ifdef SUPPORT_UCP
2687      case OP_PROP:
2688      return check_char_prop(next, previous[0], previous[1], FALSE);
2689    
2690      case OP_NOTPROP:
2691      return check_char_prop(next, previous[0], previous[1], TRUE);
2692    #endif
2693    
2694    default:    default:
2695    return FALSE;    return FALSE;
2696    }    }
2697    
2698    
2699  /* 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
2700    is set, \d turns into ESC_du rather than ESC_d, etc., so ESC_d etc. are
2701    generated only when PCRE_UCP is *not* set, that is, when only ASCII
2702    characteristics are recognized. Similarly, the opcodes OP_DIGIT etc. are
2703    replaced by OP_PROP codes when PCRE_UCP is set. */
2704    
2705  switch(op_code)  switch(op_code)
2706    {    {
2707    case OP_CHAR:    case OP_CHAR:
2708    case OP_CHARNC:    case OP_CHARNC:
2709  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
2710    if (utf8 && item > 127) { GETCHAR(item, utf8_char); }    GETCHARTEST(c, previous);
2711    #else
2712      c = *previous;
2713  #endif  #endif
2714    switch(-next)    switch(-next)
2715      {      {
2716      case ESC_d:      case ESC_d:
2717      return item > 127 || (cd->ctypes[item] & ctype_digit) == 0;      return c > 127 || (cd->ctypes[c] & ctype_digit) == 0;
2718    
2719      case ESC_D:      case ESC_D:
2720      return item <= 127 && (cd->ctypes[item] & ctype_digit) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_digit) != 0;
2721    
2722      case ESC_s:      case ESC_s:
2723      return item > 127 || (cd->ctypes[item] & ctype_space) == 0;      return c > 127 || (cd->ctypes[c] & ctype_space) == 0;
2724    
2725      case ESC_S:      case ESC_S:
2726      return item <= 127 && (cd->ctypes[item] & ctype_space) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_space) != 0;
2727    
2728      case ESC_w:      case ESC_w:
2729      return item > 127 || (cd->ctypes[item] & ctype_word) == 0;      return c > 127 || (cd->ctypes[c] & ctype_word) == 0;
2730    
2731      case ESC_W:      case ESC_W:
2732      return item <= 127 && (cd->ctypes[item] & ctype_word) != 0;      return c <= 127 && (cd->ctypes[c] & ctype_word) != 0;
2733    
2734      case ESC_h:      case ESC_h:
2735      case ESC_H:      case ESC_H:
2736      switch(item)      switch(c)
2737        {        {
2738        case 0x09:        case 0x09:
2739        case 0x20:        case 0x20:
# Line 2060  switch(op_code) Line 2761  switch(op_code)
2761    
2762      case ESC_v:      case ESC_v:
2763      case ESC_V:      case ESC_V:
2764      switch(item)      switch(c)
2765        {        {
2766        case 0x0a:        case 0x0a:
2767        case 0x0b:        case 0x0b:
# Line 2074  switch(op_code) Line 2775  switch(op_code)
2775        return -next == ESC_v;        return -next == ESC_v;
2776        }        }
2777    
2778        /* When PCRE_UCP is set, these values get generated for \d etc. Find
2779        their substitutions and process them. The result will always be either
2780        -ESC_p or -ESC_P. Then fall through to process those values. */
2781    
2782    #ifdef SUPPORT_UCP
2783        case ESC_du:
2784        case ESC_DU:
2785        case ESC_wu:
2786        case ESC_WU:
2787        case ESC_su:
2788        case ESC_SU:
2789          {
2790          int temperrorcode = 0;
2791          ptr = substitutes[-next - ESC_DU];
2792          next = check_escape(&ptr, &temperrorcode, 0, options, FALSE);
2793          if (temperrorcode != 0) return FALSE;
2794          ptr++;    /* For compatibility */
2795          }
2796        /* Fall through */
2797    
2798        case ESC_p:
2799        case ESC_P:
2800          {
2801          int ptype, pdata, errorcodeptr;
2802          BOOL negated;
2803    
2804          ptr--;      /* Make ptr point at the p or P */
2805          ptype = get_ucp(&ptr, &negated, &pdata, &errorcodeptr);
2806          if (ptype < 0) return FALSE;
2807          ptr++;      /* Point past the final curly ket */
2808    
2809          /* If the property item is optional, we have to give up. (When generated
2810          from \d etc by PCRE_UCP, this test will have been applied much earlier,
2811          to the original \d etc. At this point, ptr will point to a zero byte. */
2812    
2813          if (*ptr == CHAR_ASTERISK || *ptr == CHAR_QUESTION_MARK ||
2814            strncmp((char *)ptr, STR_LEFT_CURLY_BRACKET STR_0 STR_COMMA, 3) == 0)
2815              return FALSE;
2816    
2817          /* Do the property check. */
2818    
2819          return check_char_prop(c, ptype, pdata, (next == -ESC_P) != negated);
2820          }
2821    #endif
2822    
2823      default:      default:
2824      return FALSE;      return FALSE;
2825      }      }
2826    
2827      /* In principle, support for Unicode properties should be integrated here as
2828      well. It means re-organizing the above code so as to get hold of the property
2829      values before switching on the op-code. However, I wonder how many patterns
2830      combine ASCII \d etc with Unicode properties? (Note that if PCRE_UCP is set,
2831      these op-codes are never generated.) */
2832    
2833    case OP_DIGIT:    case OP_DIGIT:
2834    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||    return next == -ESC_D || next == -ESC_s || next == -ESC_W ||
2835           next == -ESC_h || next == -ESC_v;           next == -ESC_h || next == -ESC_v || next == -ESC_R;
2836    
2837    case OP_NOT_DIGIT:    case OP_NOT_DIGIT:
2838    return next == -ESC_d;    return next == -ESC_d;
2839    
2840    case OP_WHITESPACE:    case OP_WHITESPACE:
2841    return next == -ESC_S || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_d || next == -ESC_w || next == -ESC_R;
2842    
2843    case OP_NOT_WHITESPACE:    case OP_NOT_WHITESPACE:
2844    return next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_s || next == -ESC_h || next == -ESC_v;
2845    
2846    case OP_HSPACE:    case OP_HSPACE:
2847    return next == -ESC_S || next == -ESC_H || next == -ESC_d || next == -ESC_w;    return next == -ESC_S || next == -ESC_H || next == -ESC_d ||
2848             next == -ESC_w || next == -ESC_v || next == -ESC_R;
2849    
2850    case OP_NOT_HSPACE:    case OP_NOT_HSPACE:
2851    return next == -ESC_h;    return next == -ESC_h;
2852    
2853    /* Can't have \S in here because VT matches \S (Perl anomaly) */    /* Can't have \S in here because VT matches \S (Perl anomaly) */
2854      case OP_ANYNL:
2855    case OP_VSPACE:    case OP_VSPACE:
2856    return next == -ESC_V || next == -ESC_d || next == -ESC_w;    return next == -ESC_V || next == -ESC_d || next == -ESC_w;
2857    
2858    case OP_NOT_VSPACE:    case OP_NOT_VSPACE:
2859    return next == -ESC_v;    return next == -ESC_v || next == -ESC_R;
2860    
2861    case OP_WORDCHAR:    case OP_WORDCHAR:
2862    return next == -ESC_W || next == -ESC_s || next == -ESC_h || next == -ESC_v;    return next == -ESC_W || next == -ESC_s || next == -ESC_h ||
2863             next == -ESC_v || next == -ESC_R;
2864    
2865    case OP_NOT_WORDCHAR:    case OP_NOT_WORDCHAR:
2866    return next == -ESC_w || next == -ESC_d;    return next == -ESC_w || next == -ESC_d;
# Line 2169  BOOL inescq = FALSE; Line 2924  BOOL inescq = FALSE;
2924  BOOL groupsetfirstbyte = FALSE;  BOOL groupsetfirstbyte = FALSE;
2925  const uschar *ptr = *ptrptr;  const uschar *ptr = *ptrptr;
2926  const uschar *tempptr;  const uschar *tempptr;
2927    const uschar *nestptr = NULL;
2928  uschar *previous = NULL;  uschar *previous = NULL;
2929  uschar *previous_callout = NULL;  uschar *previous_callout = NULL;
2930  uschar *save_hwm = NULL;  uschar *save_hwm = NULL;
# Line 2178  uschar classbits[32]; Line 2934  uschar classbits[32];
2934  BOOL class_utf8;  BOOL class_utf8;
2935  BOOL utf8 = (options & PCRE_UTF8) != 0;  BOOL utf8 = (options & PCRE_UTF8) != 0;
2936  uschar *class_utf8data;  uschar *class_utf8data;
2937    uschar *class_utf8data_base;
2938  uschar utf8_char[6];  uschar utf8_char[6];
2939  #else  #else
2940  BOOL utf8 = FALSE;  BOOL utf8 = FALSE;
2941  uschar *utf8_char = NULL;  uschar *utf8_char = NULL;
2942  #endif  #endif
2943    
2944  #ifdef DEBUG  #ifdef PCRE_DEBUG
2945  if (lengthptr != NULL) DPRINTF((">> start branch\n"));  if (lengthptr != NULL) DPRINTF((">> start branch\n"));
2946  #endif  #endif
2947    
# Line 2217  req_caseopt = ((options & PCRE_CASELESS) Line 2974  req_caseopt = ((options & PCRE_CASELESS)
2974  for (;; ptr++)  for (;; ptr++)
2975    {    {
2976    BOOL negate_class;    BOOL negate_class;
2977      BOOL should_flip_negation;
2978    BOOL possessive_quantifier;    BOOL possessive_quantifier;
2979    BOOL is_quantifier;    BOOL is_quantifier;
2980    BOOL is_recurse;    BOOL is_recurse;
# Line 2237  for (;; ptr++) Line 2995  for (;; ptr++)
2995    
2996    c = *ptr;    c = *ptr;
2997    
2998      /* If we are at the end of a nested substitution, revert to the outer level
2999      string. Nesting only happens one level deep. */
3000    
3001      if (c == 0 && nestptr != NULL)
3002        {
3003        ptr = nestptr;
3004        nestptr = NULL;
3005        c = *ptr;
3006        }
3007    
3008    /* 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
3009    previous cycle of this loop. */    previous cycle of this loop. */
3010    
3011    if (lengthptr != NULL)    if (lengthptr != NULL)
3012      {      {
3013  #ifdef DEBUG  #ifdef PCRE_DEBUG
3014      if (code > cd->hwm) cd->hwm = code;                 /* High water info */      if (code > cd->hwm) cd->hwm = code;                 /* High water info */
3015  #endif  #endif
3016      if (code > cd->start_workspace + COMPILE_WORK_SIZE) /* Check for overrun */      if (code > cd->start_workspace + WORK_SIZE_CHECK)   /* Check for overrun */
3017        {        {
3018        *errorcodeptr = ERR52;        *errorcodeptr = ERR52;
3019        goto FAILED;        goto FAILED;
# Line 2258  for (;; ptr++) Line 3026  for (;; ptr++)
3026      */      */
3027    
3028      if (code < last_code) code = last_code;      if (code < last_code) code = last_code;
3029      *lengthptr += code - last_code;  
3030        /* Paranoid check for integer overflow */
3031    
3032        if (OFLOW_MAX - *lengthptr < code - last_code)
3033          {
3034          *errorcodeptr = ERR20;
3035          goto FAILED;
3036          }
3037    
3038        *lengthptr += (int)(code - last_code);
3039      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));
3040    
3041      /* 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 2285  for (;; ptr++) Line 3062  for (;; ptr++)
3062    /* 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
3063    reference list. */    reference list. */
3064    
3065    else if (cd->hwm > cd->start_workspace + COMPILE_WORK_SIZE)    else if (cd->hwm > cd->start_workspace + WORK_SIZE_CHECK)
3066      {      {
3067      *errorcodeptr = ERR52;      *errorcodeptr = ERR52;
3068      goto FAILED;      goto FAILED;
# Line 2295  for (;; ptr++) Line 3072  for (;; ptr++)
3072    
3073    if (inescq && c != 0)    if (inescq && c != 0)
3074      {      {
3075      if (c == '\\' && ptr[1] == 'E')      if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3076        {        {
3077        inescq = FALSE;        inescq = FALSE;
3078        ptr++;        ptr++;
# Line 2321  for (;; ptr++) Line 3098  for (;; ptr++)
3098    /* 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
3099    a quantifier. */    a quantifier. */
3100    
3101    is_quantifier = c == '*' || c == '+' || c == '?' ||    is_quantifier =
3102      (c == '{' && is_counted_repeat(ptr+1));      c == CHAR_ASTERISK || c == CHAR_PLUS || c == CHAR_QUESTION_MARK ||
3103        (c == CHAR_LEFT_CURLY_BRACKET && is_counted_repeat(ptr+1));
3104    
3105    if (!is_quantifier && previous_callout != NULL &&    if (!is_quantifier && previous_callout != NULL &&
3106         after_manual_callout-- <= 0)         after_manual_callout-- <= 0)
# Line 2337  for (;; ptr++) Line 3115  for (;; ptr++)
3115    if ((options & PCRE_EXTENDED) != 0)    if ((options & PCRE_EXTENDED) != 0)
3116      {      {
3117      if ((cd->ctypes[c] & ctype_space) != 0) continue;      if ((cd->ctypes[c] & ctype_space) != 0) continue;
3118      if (c == '#')      if (c == CHAR_NUMBER_SIGN)
3119        {        {
3120        while (*(++ptr) != 0)        while (*(++ptr) != 0)
3121          {          {
# Line 2362  for (;; ptr++) Line 3140  for (;; ptr++)
3140      {      {
3141      /* ===================================================================*/      /* ===================================================================*/
3142      case 0:                        /* The branch terminates at string end */      case 0:                        /* The branch terminates at string end */
3143      case '|':                      /* or | or ) */      case CHAR_VERTICAL_LINE:       /* or | or ) */
3144      case ')':      case CHAR_RIGHT_PARENTHESIS:
3145      *firstbyteptr = firstbyte;      *firstbyteptr = firstbyte;
3146      *reqbyteptr = reqbyte;      *reqbyteptr = reqbyte;
3147      *codeptr = code;      *codeptr = code;
3148      *ptrptr = ptr;      *ptrptr = ptr;
3149      if (lengthptr != NULL)      if (lengthptr != NULL)
3150        {        {
3151        *lengthptr += code - last_code;   /* To include callout length */        if (OFLOW_MAX - *lengthptr < code - last_code)
3152            {
3153            *errorcodeptr = ERR20;
3154            goto FAILED;
3155            }
3156          *lengthptr += (int)(code - last_code);   /* To include callout length */
3157        DPRINTF((">> end branch\n"));        DPRINTF((">> end branch\n"));
3158        }        }
3159      return TRUE;      return TRUE;
# Line 2380  for (;; ptr++) Line 3163  for (;; ptr++)
3163      /* Handle single-character metacharacters. In multiline mode, ^ disables      /* Handle single-character metacharacters. In multiline mode, ^ disables
3164      the setting of any following char as a first character. */      the setting of any following char as a first character. */
3165    
3166      case '^':      case CHAR_CIRCUMFLEX_ACCENT:
3167      if ((options & PCRE_MULTILINE) != 0)      if ((options & PCRE_MULTILINE) != 0)
3168        {        {
3169        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;        if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
# Line 2389  for (;; ptr++) Line 3172  for (;; ptr++)
3172      *code++ = OP_CIRC;      *code++ = OP_CIRC;
3173      break;      break;
3174    
3175      case '$':      case CHAR_DOLLAR_SIGN:
3176      previous = NULL;      previous = NULL;
3177      *code++ = OP_DOLL;      *code++ = OP_DOLL;
3178      break;      break;
# Line 2397  for (;; ptr++) Line 3180  for (;; ptr++)
3180      /* 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
3181      repeats. The value of reqbyte doesn't change either. */      repeats. The value of reqbyte doesn't change either. */
3182    
3183      case '.':      case CHAR_DOT:
3184      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;      if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3185      zerofirstbyte = firstbyte;      zerofirstbyte = firstbyte;
3186      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3187      previous = code;      previous = code;
3188      *code++ = OP_ANY;      *code++ = ((options & PCRE_DOTALL) != 0)? OP_ALLANY: OP_ANY;
3189      break;      break;
3190    
3191    
# Line 2417  for (;; ptr++) Line 3200  for (;; ptr++)
3200      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,
3201      but those above are are explicitly listed afterwards. A flag byte tells      but those above are are explicitly listed afterwards. A flag byte tells
3202      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.
     */  
3203    
3204      case '[':      In JavaScript compatibility mode, an isolated ']' causes an error. In
3205        default (Perl) mode, it is treated as a data character. */
3206    
3207        case CHAR_RIGHT_SQUARE_BRACKET:
3208        if ((cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3209          {
3210          *errorcodeptr = ERR64;
3211          goto FAILED;
3212          }
3213        goto NORMAL_CHAR;
3214    
3215        case CHAR_LEFT_SQUARE_BRACKET:
3216      previous = code;      previous = code;
3217    
3218      /* 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
3219      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. */
3220    
3221      if ((ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&      if ((ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3222          check_posix_syntax(ptr, &tempptr, cd))           ptr[1] == CHAR_EQUALS_SIGN) &&
3223            check_posix_syntax(ptr, &tempptr))
3224        {        {
3225        *errorcodeptr = (ptr[1] == ':')? ERR13 : ERR31;        *errorcodeptr = (ptr[1] == CHAR_COLON)? ERR13 : ERR31;
3226        goto FAILED;        goto FAILED;
3227        }        }
3228    
3229      /* If the first character is '^', set the negation flag and skip it. */      /* If the first character is '^', set the negation flag and skip it. Also,
3230        if the first few characters (either before or after ^) are \Q\E or \E we
3231        skip them too. This makes for compatibility with Perl. */
3232    
3233      if ((c = *(++ptr)) == '^')      negate_class = FALSE;
3234        for (;;)
3235        {        {
       negate_class = TRUE;  
3236        c = *(++ptr);        c = *(++ptr);
3237          if (c == CHAR_BACKSLASH)
3238            {
3239            if (ptr[1] == CHAR_E)
3240              ptr++;
3241            else if (strncmp((const char *)ptr+1,
3242                              STR_Q STR_BACKSLASH STR_E, 3) == 0)
3243              ptr += 3;
3244            else
3245              break;
3246            }
3247          else if (!negate_class && c == CHAR_CIRCUMFLEX_ACCENT)
3248            negate_class = TRUE;
3249          else break;
3250        }        }
3251      else  
3252        /* Empty classes are allowed in JavaScript compatibility mode. Otherwise,
3253        an initial ']' is taken as a data character -- the code below handles
3254        that. In JS mode, [] must always fail, so generate OP_FAIL, whereas
3255        [^] must match any character, so generate OP_ALLANY. */
3256    
3257        if (c == CHAR_RIGHT_SQUARE_BRACKET &&
3258            (cd->external_options & PCRE_JAVASCRIPT_COMPAT) != 0)
3259        {        {
3260        negate_class = FALSE;        *code++ = negate_class? OP_ALLANY : OP_FAIL;
3261          if (firstbyte == REQ_UNSET) firstbyte = REQ_NONE;
3262          zerofirstbyte = firstbyte;
3263          break;
3264        }        }
3265    
3266        /* If a class contains a negative special such as \S, we need to flip the
3267        negation flag at the end, so that support for characters > 255 works
3268        correctly (they are all included in the class). */
3269    
3270        should_flip_negation = FALSE;
3271    
3272      /* 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
3273      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
3274      valued UTF-8 characters, we don't yet do any optimization. */      valued UTF-8 characters, we don't yet do any optimization. */
# Line 2461  for (;; ptr++) Line 3286  for (;; ptr++)
3286  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3287      class_utf8 = FALSE;                       /* No chars >= 256 */      class_utf8 = FALSE;                       /* No chars >= 256 */
3288      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */      class_utf8data = code + LINK_SIZE + 2;    /* For UTF-8 items */
3289        class_utf8data_base = class_utf8data;     /* For resetting in pass 1 */
3290  #endif  #endif
3291    
3292      /* 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 2476  for (;; ptr++) Line 3302  for (;; ptr++)
3302          {                           /* Braces are required because the */          {                           /* Braces are required because the */
3303          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */          GETCHARLEN(c, ptr, ptr);    /* macro generates multiple statements */
3304          }          }
3305    
3306          /* In the pre-compile phase, accumulate the length of any UTF-8 extra
3307          data and reset the pointer. This is so that very large classes that
3308          contain a zillion UTF-8 characters no longer overwrite the work space
3309          (which is on the stack). */
3310    
3311          if (lengthptr != NULL)
3312            {
3313            *lengthptr += class_utf8data - class_utf8data_base;
3314            class_utf8data = class_utf8data_base;
3315            }
3316    
3317  #endif  #endif
3318    
3319        /* Inside \Q...\E everything is literal except \E */        /* Inside \Q...\E everything is literal except \E */
3320    
3321        if (inescq)        if (inescq)
3322          {          {
3323          if (c == '\\' && ptr[1] == 'E')     /* If we are at \E */          if (c == CHAR_BACKSLASH && ptr[1] == CHAR_E)  /* If we are at \E */
3324            {            {
3325            inescq = FALSE;                   /* Reset literal state */            inescq = FALSE;                   /* Reset literal state */
3326            ptr++;                            /* Skip the 'E' */            ptr++;                            /* Skip the 'E' */
# Line 2497  for (;; ptr++) Line 3335  for (;; ptr++)
3335        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl        [.ch.] and [=ch=] ("collating elements") and fault them, as Perl
3336        5.6 and 5.8 do. */        5.6 and 5.8 do. */
3337    
3338        if (c == '[' &&        if (c == CHAR_LEFT_SQUARE_BRACKET &&
3339            (ptr[1] == ':' || ptr[1] == '.' || ptr[1] == '=') &&            (ptr[1] == CHAR_COLON || ptr[1] == CHAR_DOT ||
3340            check_posix_syntax(ptr, &tempptr, cd))             ptr[1] == CHAR_EQUALS_SIGN) && check_posix_syntax(ptr, &tempptr))
3341          {          {
3342          BOOL local_negate = FALSE;          BOOL local_negate = FALSE;
3343          int posix_class, taboffset, tabopt;          int posix_class, taboffset, tabopt;
3344          register const uschar *cbits = cd->cbits;          register const uschar *cbits = cd->cbits;
3345          uschar pbits[32];          uschar pbits[32];
3346    
3347          if (ptr[1] != ':')          if (ptr[1] != CHAR_COLON)
3348            {            {
3349            *errorcodeptr = ERR31;            *errorcodeptr = ERR31;
3350            goto FAILED;            goto FAILED;
3351            }            }
3352    
3353          ptr += 2;          ptr += 2;
3354          if (*ptr == '^')          if (*ptr == CHAR_CIRCUMFLEX_ACCENT)
3355            {            {
3356            local_negate = TRUE;            local_negate = TRUE;
3357              should_flip_negation = TRUE;  /* Note negative special */
3358            ptr++;            ptr++;
3359            }            }
3360    
3361          posix_class = check_posix_name(ptr, tempptr - ptr);          posix_class = check_posix_name(ptr, (int)(tempptr - ptr));
3362          if (posix_class < 0)          if (posix_class < 0)
3363            {            {
3364            *errorcodeptr = ERR30;            *errorcodeptr = ERR30;
# Line 2533  for (;; ptr++) Line 3372  for (;; ptr++)
3372          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)          if ((options & PCRE_CASELESS) != 0 && posix_class <= 2)
3373            posix_class = 0;            posix_class = 0;
3374    
3375          /* 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
3376          because we may be adding and subtracting from it, and we don't want to          different escape sequences that use Unicode properties. */
3377          subtract bits that may be in the main map already. At the end we or the  
3378          result into the bit map that is being built. */  #ifdef SUPPORT_UCP
3379            if ((options & PCRE_UCP) != 0)
3380              {
3381              int pc = posix_class + ((local_negate)? POSIX_SUBSIZE/2 : 0);
3382              if (posix_substitutes[pc] != NULL)
3383                {
3384                nestptr = tempptr + 1;
3385                ptr = posix_substitutes[pc] - 1;
3386                continue;
3387                }
3388              }
3389    #endif
3390            /* In the non-UCP case, we build the bit map for the POSIX class in a
3391            chunk of local store because we may be adding and subtracting from it,
3392            and we don't want to subtract bits that may be in the main map already.
3393            At the end we or the result into the bit map that is being built. */
3394    
3395          posix_class *= 3;          posix_class *= 3;
3396    
# Line 2580  for (;; ptr++) Line 3434  for (;; ptr++)
3434    
3435        /* Backslash may introduce a single character, or it may introduce one        /* Backslash may introduce a single character, or it may introduce one
3436        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
3437        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
3438        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
3439        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
3440        character in them, so set class_charcount bigger than one. */        are either treated as literal characters (by default), or are faulted if
3441          PCRE_EXTRA is set. */
3442    
3443        if (c == '\\')        if (c == CHAR_BACKSLASH)
3444          {          {
3445          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);          c = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3446          if (*errorcodeptr != 0) goto FAILED;          if (*errorcodeptr != 0) goto FAILED;
3447    
3448          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 */  
3449          else if (-c == ESC_Q)            /* Handle start of quoted string */          else if (-c == ESC_Q)            /* Handle start of quoted string */
3450            {            {
3451            if (ptr[1] == '\\' && ptr[2] == 'E')            if (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3452              {              {
3453              ptr += 2; /* avoid empty string */              ptr += 2; /* avoid empty string */
3454              }              }
3455            else inescq = TRUE;            else inescq = TRUE;
3456            continue;            continue;
3457            }            }
3458            else if (-c == ESC_E) continue;  /* Ignore orphan \E */
3459    
3460          if (c < 0)          if (c < 0)
3461            {            {
3462            register const uschar *cbits = cd->cbits;            register const uschar *cbits = cd->cbits;
3463            class_charcount += 2;     /* Greater than 1 is what matters */            class_charcount += 2;     /* Greater than 1 is what matters */
3464    
3465            /* Save time by not doing this in the pre-compile phase. */            switch (-c)
   
           if (lengthptr == NULL) switch (-c)  
3466              {              {
3467    #ifdef SUPPORT_UCP
3468                case ESC_du:     /* These are the values given for \d etc */
3469                case ESC_DU:     /* when PCRE_UCP is set. We replace the */
3470                case ESC_wu:     /* escape sequence with an appropriate \p */
3471                case ESC_WU:     /* or \P to test Unicode properties instead */
3472                case ESC_su:     /* of the default ASCII testing. */
3473                case ESC_SU:
3474                nestptr = ptr;
3475                ptr = substitutes[-c - ESC_DU] - 1;  /* Just before substitute */
3476                class_charcount -= 2;                /* Undo! */
3477                continue;
3478    #endif
3479              case ESC_d:              case ESC_d:
3480              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= cbits[c+cbit_digit];
3481              continue;              continue;
3482    
3483              case ESC_D:              case ESC_D:
3484                should_flip_negation = TRUE;
3485              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_digit];
3486              continue;              continue;
3487    
# Line 2625  for (;; ptr++) Line 3490  for (;; ptr++)
3490              continue;              continue;
3491    
3492              case ESC_W:              case ESC_W:
3493                should_flip_negation = TRUE;
3494              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_word];
3495              continue;              continue;
3496    
# Line 2634  for (;; ptr++) Line 3500  for (;; ptr++)
3500              continue;              continue;
3501    
3502              case ESC_S:              case ESC_S:
3503                should_flip_negation = TRUE;
3504              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];              for (c = 0; c < 32; c++) classbits[c] |= ~cbits[c+cbit_space];
3505              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */              classbits[1] |= 0x08;    /* Perl 5.004 onwards omits VT from \s */
3506              continue;              continue;
3507    
3508              case ESC_E: /* Perl ignores an orphan \E */              case ESC_h:
             continue;  
   
             default:    /* Not recognized; fall through */  
             break;      /* Need "default" setting to stop compiler warning. */  
             }  
   
           /* In the pre-compile phase, just do the recognition. */  
   
           else if (c == -ESC_d || c == -ESC_D || c == -ESC_w ||  
                    c == -ESC_W || c == -ESC_s || c == -ESC_S) continue;  
   
           /* We need to deal with \H, \h, \V, and \v in both phases because  
           they use extra memory. */  
   
           if (-c == ESC_h)  
             {  
3509              SETBIT(classbits, 0x09); /* VT */              SETBIT(classbits, 0x09); /* VT */
3510              SETBIT(classbits, 0x20); /* SPACE */              SETBIT(classbits, 0x20); /* SPACE */
3511              SETBIT(classbits, 0xa0); /* NSBP */              SETBIT(classbits, 0xa0); /* NSBP */
# Line 2678  for (;; ptr++) Line 3529  for (;; ptr++)
3529                }                }
3530  #endif  #endif
3531              continue;              continue;
             }  
3532    
3533            if (-c == ESC_H)              case ESC_H:
             {  
3534              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3535                {                {
3536                int x = 0xff;                int x = 0xff;
# Line 2723  for (;; ptr++) Line 3572  for (;; ptr++)
3572                }                }
3573  #endif  #endif
3574              continue;              continue;
             }  
3575    
3576            if (-c == ESC_v)              case ESC_v:
             {  
3577              SETBIT(classbits, 0x0a); /* LF */              SETBIT(classbits, 0x0a); /* LF */
3578              SETBIT(classbits, 0x0b); /* VT */              SETBIT(classbits, 0x0b); /* VT */
3579              SETBIT(classbits, 0x0c); /* FF */              SETBIT(classbits, 0x0c); /* FF */
# Line 2742  for (;; ptr++) Line 3589  for (;; ptr++)
3589                }                }
3590  #endif  #endif
3591              continue;              continue;
             }  
3592    
3593            if (-c == ESC_V)              case ESC_V:
             {  
3594              for (c = 0; c < 32; c++)              for (c = 0; c < 32; c++)
3595                {                {
3596                int x = 0xff;                int x = 0xff;
# Line 2775  for (;; ptr++) Line 3620  for (;; ptr++)
3620                }                }
3621  #endif  #endif
3622              continue;              continue;
             }  
   
           /* We need to deal with \P and \p in both phases. */  
3623    
3624  #ifdef SUPPORT_UCP  #ifdef SUPPORT_UCP
3625            if (-c == ESC_p || -c == ESC_P)              case ESC_p:
3626              {              case ESC_P:
3627              BOOL negated;                {
3628              int pdata;                BOOL negated;
3629              int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);                int pdata;
3630              if (ptype < 0) goto FAILED;                int ptype = get_ucp(&ptr, &negated, &pdata, errorcodeptr);
3631              class_utf8 = TRUE;                if (ptype < 0) goto FAILED;
3632              *class_utf8data++ = ((-c == ESC_p) != negated)?                class_utf8 = TRUE;
3633                XCL_PROP : XCL_NOTPROP;                *class_utf8data++ = ((-c == ESC_p) != negated)?
3634              *class_utf8data++ = ptype;                  XCL_PROP : XCL_NOTPROP;
3635              *class_utf8data++ = pdata;                *class_utf8data++ = ptype;
3636              class_charcount -= 2;   /* Not a < 256 character */                *class_utf8data++ = pdata;
3637              continue;                class_charcount -= 2;   /* Not a < 256 character */
3638              }                continue;
3639                  }
3640  #endif  #endif
3641            /* Unrecognized escapes are faulted if PCRE is running in its              /* Unrecognized escapes are faulted if PCRE is running in its
3642            strict mode. By default, for compatibility with Perl, they are              strict mode. By default, for compatibility with Perl, they are
3643            treated as literals. */              treated as literals. */
3644    
3645            if ((options & PCRE_EXTRA) != 0)              default:
3646              {              if ((options & PCRE_EXTRA) != 0)
3647              *errorcodeptr = ERR7;                {
3648              goto FAILED;                *errorcodeptr = ERR7;
3649                  goto FAILED;
3650                  }
3651                class_charcount -= 2;  /* Undo the default count from above */
3652                c = *ptr;              /* Get the final character and fall through */
3653                break;
3654              }              }
   
           class_charcount -= 2;  /* Undo the default count from above */  
           c = *ptr;              /* Get the final character and fall through */  
3655            }            }
3656    
3657          /* 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 2820  for (;; ptr++) Line 3665  for (;; ptr++)
3665        entirely. The code for handling \Q and \E is messy. */        entirely. The code for handling \Q and \E is messy. */
3666    
3667        CHECK_RANGE:        CHECK_RANGE:
3668        while (ptr[1] == '\\' && ptr[2] == 'E')        while (ptr[1] == CHAR_BACKSLASH && ptr[2] == CHAR_E)
3669          {          {
3670          inescq = FALSE;          inescq = FALSE;
3671          ptr += 2;          ptr += 2;
# Line 2828  for (;; ptr++) Line 3673  for (;; ptr++)
3673    
3674        oldptr = ptr;        oldptr = ptr;
3675    
3676        if (!inescq && ptr[1] == '-')        /* Remember \r or \n */
3677    
3678          if (c == CHAR_CR || c == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3679    
3680          /* Check for range */
3681    
3682          if (!inescq && ptr[1] == CHAR_MINUS)
3683          {          {
3684          int d;          int d;
3685          ptr += 2;          ptr += 2;
3686          while (*ptr == '\\' && ptr[1] == 'E') ptr += 2;          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E) ptr += 2;
3687    
3688          /* 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
3689          mode. */          mode. */
3690    
3691          while (*ptr == '\\' && ptr[1] == 'Q')          while (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_Q)
3692            {            {
3693            ptr += 2;            ptr += 2;
3694            if (*ptr == '\\' && ptr[1] == 'E') { ptr += 2; continue; }            if (*ptr == CHAR_BACKSLASH && ptr[1] == CHAR_E)
3695                { ptr += 2; continue; }
3696            inescq = TRUE;            inescq = TRUE;
3697            break;            break;
3698            }            }
3699    
3700          if (*ptr == 0 || (!inescq && *ptr == ']'))          if (*ptr == 0 || (!inescq && *ptr == CHAR_RIGHT_SQUARE_BRACKET))
3701            {            {
3702            ptr = oldptr;            ptr = oldptr;
3703            goto LONE_SINGLE_CHARACTER;            goto LONE_SINGLE_CHARACTER;
# Line 2864  for (;; ptr++) Line 3716  for (;; ptr++)
3716          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
3717          in such circumstances. */          in such circumstances. */
3718    
3719          if (!inescq && d == '\\')          if (!inescq && d == CHAR_BACKSLASH)
3720            {            {
3721            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);            d = check_escape(&ptr, errorcodeptr, cd->bracount, options, TRUE);
3722            if (*errorcodeptr != 0) goto FAILED;            if (*errorcodeptr != 0) goto FAILED;
3723    
3724            /* \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 */  
3725    
3726            if (d < 0)            if (d < 0)
3727              {              {
3728              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  
3729                {                {
3730                ptr = oldptr;                ptr = oldptr;
3731                goto LONE_SINGLE_CHARACTER;  /* A few lines below */                goto LONE_SINGLE_CHARACTER;  /* A few lines below */
# Line 2895  for (;; ptr++) Line 3744  for (;; ptr++)
3744    
3745          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */          if (d == c) goto LONE_SINGLE_CHARACTER;  /* A few lines below */
3746    
3747            /* Remember \r or \n */
3748    
3749            if (d == CHAR_CR || d == CHAR_NL) cd->external_flags |= PCRE_HASCRORLF;
3750    
3751          /* 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
3752          matching, we have to use an XCLASS with extra data items. Caseless          matching, we have to use an XCLASS with extra data items. Caseless
3753          matching for characters > 127 is available only if UCP support is          matching for characters > 127 is available only if UCP support is
# Line 3013  for (;; ptr++) Line 3866  for (;; ptr++)
3866          if ((options & PCRE_CASELESS) != 0)          if ((options & PCRE_CASELESS) != 0)
3867            {            {
3868            unsigned int othercase;            unsigned int othercase;
3869            if ((othercase = _pcre_ucp_othercase(c)) != NOTACHAR)            if ((othercase = UCD_OTHERCASE(c)) != c)
3870              {              {
3871              *class_utf8data++ = XCL_SINGLE;              *class_utf8data++ = XCL_SINGLE;
3872              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);              class_utf8data += _pcre_ord2utf8(othercase, class_utf8data);
# Line 3038  for (;; ptr++) Line 3891  for (;; ptr++)
3891          }          }
3892        }        }
3893    
3894      /* 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.
3895        If we are at the end of an internal nested string, revert to the outer
3896        string. */
3897    
3898        while (((c = *(++ptr)) != 0 ||
3899               (nestptr != NULL &&
3900                 (ptr = nestptr, nestptr = NULL, c = *(++ptr)) != 0)) &&
3901               (c != CHAR_RIGHT_SQUARE_BRACKET || inescq));
3902    
3903      while ((c = *(++ptr)) != 0 && (c != ']' || inescq));      /* Check for missing terminating ']' */
3904    
3905      if (c == 0)                          /* Missing terminating ']' */      if (c == 0)
3906        {        {
3907        *errorcodeptr = ERR6;        *errorcodeptr = ERR6;
3908        goto FAILED;        goto FAILED;
3909        }        }
3910    
3911      /* 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
3912      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
3913      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
3914      because OP_NOT and the related opcodes like OP_NOTSTAR operate on      optimize.
3915      single-bytes only. This is an historical hangover. Maybe one day we can  
3916      tidy these opcodes to handle multi-byte characters.      In UTF-8 mode, we can optimize the negative case only if there were no
3917        characters >= 128 because OP_NOT and the related opcodes like OP_NOTSTAR
3918        operate on single-bytes only. This is an historical hangover. Maybe one day
3919        we can tidy these opcodes to handle multi-byte characters.
3920    
3921      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
3922      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 3063  for (;; ptr++) Line 3926  for (;; ptr++)
3926      reqbyte, save the previous value for reinstating. */      reqbyte, save the previous value for reinstating. */
3927    
3928  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3929      if (class_charcount == 1 &&      if (class_charcount == 1 && !class_utf8 &&
3930            (!utf8 ||        (!utf8 || !negate_class || class_lastchar < 128))
           (!class_utf8 && (!negate_class || class_lastchar < 128))))  
   
3931  #else  #else
3932      if (class_charcount == 1)      if (class_charcount == 1)
3933  #endif  #endif
# Line 3109  for (;; ptr++) Line 3970  for (;; ptr++)
3970      zeroreqbyte = reqbyte;      zeroreqbyte = reqbyte;
3971    
3972      /* If there are characters with values > 255, we have to compile an      /* If there are characters with values > 255, we have to compile an
3973      extended class, with its own opcode. If there are no characters < 256,      extended class, with its own opcode, unless there was a negated special
3974      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
3975        characters > 255 are in the class, so any that were explicitly given as
3976        well can be ignored. If (when there are explicit characters > 255 that must
3977        be listed) there are no characters < 256, we can omit the bitmap in the
3978        actual compiled code. */
3979    
3980  #ifdef SUPPORT_UTF8  #ifdef SUPPORT_UTF8
3981      if (class_utf8)      if (class_utf8 && (!should_flip_negation || (options & PCRE_UCP) != 0))
3982        {        {
3983        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */        *class_utf8data++ = XCL_END;    /* Marks the end of extra data */
3984        *code++ = OP_XCLASS;        *code++ = OP_XCLASS;
# Line 3139  for (;; ptr++) Line 4004  for (;; ptr++)
4004        }        }
4005  #endif  #endif
4006    
4007      /* 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
4008      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
4009      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
4010      setting must remain unchanged after any kind of repeat. */      (non-UCP) in the class. Then copy the 32-byte map into the code vector,
4011        negating it if necessary. */
4012    
4013        *code++ = (negate_class == should_flip_negation) ? OP_CLASS : OP_NCLASS;
4014      if (negate_class)      if (negate_class)
4015        {        {
       *code++ = OP_NCLASS;  
4016        if (lengthptr == NULL)    /* Save time in the pre-compile phase */        if (lengthptr == NULL)    /* Save time in the pre-compile phase */
4017          for (c = 0; c < 32; c++) code[c] = ~classbits[c];          for (c = 0; c < 32; c++) code[c] = ~classbits[c];
4018        }        }
4019      else      else
4020        {        {
       *code++ = OP_CLASS;  
4021        memcpy(code, classbits, 32);        memcpy(code, classbits, 32);
4022        }        }
4023      code += 32;      code += 32;
# Line 3163  for (;; ptr++) Line 4028  for (;; ptr++)
4028      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this      /* Various kinds of repeat; '{' is not necessarily a quantifier, but this
4029      has been tested above. */      has been tested above. */
4030    
4031      case '{':      case CHAR_LEFT_CURLY_BRACKET:
4032      if (!is_quantifier) goto NORMAL_CHAR;      if (!is_quantifier) goto NORMAL_CHAR;
4033      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);      ptr = read_repeat_counts(ptr+1, &repeat_min, &repeat_max, errorcodeptr);
4034      if (*errorcodeptr != 0) goto FAILED;      if (*errorcodeptr != 0) goto FAILED;
4035      goto REPEAT;      goto REPEAT;
4036    
4037      case '*':      case CHAR_ASTERISK:
4038      repeat_min = 0;      repeat_min = 0;
4039      repeat_max = -1;      repeat_max = -1;
4040      goto REPEAT;      goto REPEAT;
4041    
4042      case '+':      case CHAR_PLUS:
4043      repeat_min = 1;      repeat_min = 1;
4044      repeat_max = -1;      repeat_max = -1;
4045      goto REPEAT;      goto REPEAT;
4046    
4047      case '?':      case CHAR_QUESTION_MARK:
4048      repeat_min = 0;      repeat_min = 0;
4049      repeat_max = 1;      repeat_max = 1;
4050    
# Line 3214  for (;; ptr++) Line 4079  for (;; ptr++)
4079      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
4080      repeat type to the non-default. */      repeat type to the non-default. */
4081    
4082      if (ptr[1] == '+')      if (ptr[1] == CHAR_PLUS)
4083        {        {
4084        repeat_type = 0;                  /* Force greedy */        repeat_type = 0;                  /* Force greedy */
4085        possessive_quantifier = TRUE;        possessive_quantifier = TRUE;
4086        ptr++;        ptr++;
4087        }        }
4088      else if (ptr[1] == '?')      else if (ptr[1] == CHAR_QUESTION_MARK)
4089        {        {
4090        repeat_type = greedy_non_default;        repeat_type = greedy_non_default;
4091        ptr++;        ptr++;
# Line 3267  for (;; ptr++) Line 4132  for (;; ptr++)
4132    
4133        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4134            repeat_max < 0 &&            repeat_max < 0 &&
4135            check_auto_possessive(*previous, c, utf8, utf8_char, ptr + 1,            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
             options, cd))  
4136          {          {
4137          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4138          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3289  for (;; ptr++) Line 4153  for (;; ptr++)
4153        c = previous[1];        c = previous[1];
4154        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4155            repeat_max < 0 &&            repeat_max < 0 &&
4156            check_auto_possessive(OP_NOT, c, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4157          {          {
4158          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4159          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3313  for (;; ptr++) Line 4177  for (;; ptr++)
4177    
4178        if (!possessive_quantifier &&        if (!possessive_quantifier &&
4179            repeat_max < 0 &&            repeat_max < 0 &&
4180            check_auto_possessive(c, 0, utf8, NULL, ptr + 1, options, cd))            check_auto_possessive(previous, utf8, ptr + 1, options, cd))
4181          {          {
4182          repeat_type = 0;    /* Force greedy */          repeat_type = 0;    /* Force greedy */
4183          possessive_quantifier = TRUE;          possessive_quantifier = TRUE;
# Line 3335  for (;; ptr++) Line 4199  for (;; ptr++)
4199    
4200        if (repeat_max == 0) goto END_REPEAT;        if (repeat_max == 0) goto END_REPEAT;
4201    
4202          /*--------------------------------------------------------------------*/
4203          /* This code is obsolete from release 8.00; the restriction was finally
4204          removed: */
4205    
4206        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4207        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4208    
4209        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4210          /*--------------------------------------------------------------------*/
4211    
4212        /* Combine the op_type with the repeat_type */        /* Combine the op_type with the repeat_type */
4213    
# Line 3485  for (;; ptr++) Line 4354  for (;; ptr++)
4354          goto END_REPEAT;          goto END_REPEAT;
4355          }          }
4356    
4357          /*--------------------------------------------------------------------*/
4358          /* This code is obsolete from release 8.00; the restriction was finally
4359          removed: */
4360    
4361        /* All real repeats make it impossible to handle partial matching (maybe        /* All real repeats make it impossible to handle partial matching (maybe
4362        one day we will be able to remove this restriction). */        one day we will be able to remove this restriction). */
4363    
4364        if (repeat_max != 1) cd->nopartial = TRUE;        /* if (repeat_max != 1) cd->external_flags |= PCRE_NOPARTIAL; */
4365          /*--------------------------------------------------------------------*/
4366    
4367        if (repeat_min == 0 && repeat_max == -1)        if (repeat_min == 0 && repeat_max == -1)
4368          *code++ = OP_CRSTAR + repeat_type;          *code++ = OP_CRSTAR + repeat_type;
# Line 3513  for (;; ptr++) Line 4387  for (;; ptr++)
4387        {        {
4388        register int i;        register int i;
4389        int ketoffset = 0;        int ketoffset = 0;
4390        int len = code - previous;        int len = (int)(code - previous);
4391        uschar *bralink = NULL;        uschar *bralink = NULL;
4392    
4393        /* Repeating a DEFINE group is pointless */        /* Repeating a DEFINE group is pointless */
# Line 3524  for (;; ptr++) Line 4398  for (;; ptr++)
4398          goto FAILED;          goto FAILED;
4399          }          }
4400    
       /* This is a paranoid check to stop integer overflow later on */  
   
       if (len > MAX_DUPLENGTH)  
         {  
         *errorcodeptr = ERR50;  
         goto FAILED;  
         }  
   
4401        /* If the maximum repeat count is unlimited, find the end of the bracket        /* If the maximum repeat count is unlimited, find the end of the bracket
4402        by scanning through from the start, and compute the offset back to it        by scanning through from the start, and compute the offset back to it
4403        from the current code pointer. There may be an OP_OPT setting following        from the current code pointer. There may be an OP_OPT setting following
# Line 3542  for (;; ptr++) Line 4408  for (;; ptr++)
4408          {          {
4409          register uschar *ket = previous;          register uschar *ket = previous;
4410          do ket += GET(ket, 1); while (*ket != OP_KET);          do ket += GET(ket, 1); while (*ket != OP_KET);
4411          ketoffset = code - ket;          ketoffset = (int)(code - ket);
4412          }          }
4413    
4414        /* 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 3554  for (;; ptr++) Line 4420  for (;; ptr++)
4420    
4421        if (repeat_min == 0)        if (repeat_min == 0)
4422          {          {
4423          /* 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
4424          altogether. */          output altogether, like this:
4425    
4426          if (repeat_max == 0)          ** if (repeat_max == 0)
4427            {          **   {
4428            code = previous;          **   code = previous;
4429            goto END_REPEAT;          **   goto END_REPEAT;
4430            }          **   }
4431    
4432            However, that fails when a group is referenced as a subroutine from
4433            elsewhere in the pattern, so now we stick in OP_SKIPZERO in front of it
4434            so that it is skipped on execution. As we don't have a list of which
4435            groups are referenced, we cannot do this selectively.
4436    
4437            If the maximum is 1 or unlimited, we just have to stick in the BRAZERO
4438            and do no more at this point. However, we do need to adjust any
4439            OP_RECURSE calls inside the group that refer to the group itself or any
4440            internal or forward referenced group, because the offset is from the
4441            start of the whole regex. Temporarily terminate the pattern while doing
4442            this. */
4443    
4444          /* If the maximum is 1 or unlimited, we just have to stick in the          if (repeat_max <= 1)    /* Covers 0, 1, and unlimited */
         BRAZERO and do no more at this point. However, we do need to adjust  
         any OP_RECURSE calls inside the group that refer to the group itself or  
         any internal or forward referenced group, because the offset is from  
         the start of the whole regex. Temporarily terminate the pattern while  
         doing this. */  
   
         if (repeat_max <= 1)  
4445            {            {
4446            *code = OP_END;            *code = OP_END;
4447            adjust_recurse(previous, 1, utf8, cd, save_hwm);            adjust_recurse(previous, 1, utf8, cd, save_hwm);
4448            memmove(previous+1, previous, len);            memmove(previous+1, previous, len);
4449            code++;            code++;
4450              if (repeat_max == 0)
4451                {
4452                *previous++ = OP_SKIPZERO;
4453                goto END_REPEAT;
4454                }
4455            *previous++ = OP_BRAZERO + repeat_type;            *previous++ = OP_BRAZERO + repeat_type;
4456            }            }
4457    
# Line 3600  for (;; ptr++) Line 4476  for (;; ptr++)
4476            /* We chain together the bracket offset fields that have to be            /* We chain together the bracket offset fields that have to be
4477            filled in later when the ends of the brackets are reached. */            filled in later when the ends of the brackets are reached. */
4478    
4479            offset = (bralink == NULL)? 0 : previous - bralink;            offset = (bralink == NULL)? 0 : (int)(previous - bralink);
4480            bralink = previous;            bralink = previous;
4481            PUTINC(previous, 0, offset);            PUTINC(previous, 0, offset);
4482            }            }
# Line 3620  for (;; ptr++) Line 4496  for (;; ptr++)
4496          if (repeat_min > 1)          if (repeat_min > 1)
4497            {            {
4498            /* 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
4499            just adjust the length as if we had. */            just adjust the length as if we had. Do some paranoid checks for
4500              potential integer overflow. The INT64_OR_DOUBLE type is a 64-bit
4501              integer type when available, otherwise double. */
4502    
4503            if (lengthptr != NULL)            if (lengthptr != NULL)
4504              *lengthptr += (repeat_min - 1)*length_prevgroup;              {
4505                int delta = (repeat_min - 1)*length_prevgroup;
4506                if ((INT64_OR_DOUBLE)(repeat_min - 1)*
4507                      (INT64_OR_DOUBLE)length_prevgroup >
4508                        (INT64_OR_DOUBLE)INT_MAX ||
4509                    OFLOW_MAX - *lengthptr < delta)
4510                  {
4511                  *errorcodeptr = ERR20;
4512                  goto FAILED;
4513                  }
4514                *lengthptr += delta;
4515                }
4516    
4517            /* This is compiling for real */            /* This is compiling for real */
4518    
# Line 3661  for (;; ptr++) Line 4550  for (;; ptr++)
4550          /* 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
4551          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
4552          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
4553          add 2 + 2*LINKSIZE to allow for the nesting that occurs. */          add 2 + 2*LINKSIZE to allow for the nesting that occurs. Do some
4554            paranoid checks to avoid integer overflow. The INT64_OR_DOUBLE type is
4555            a 64-bit integer type when available, otherwise double. */
4556    
4557          if (lengthptr != NULL && repeat_max > 0)          if (lengthptr != NULL && repeat_max > 0)
4558            *lengthptr += repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -            {
4559              2 - 2*LINK_SIZE;  /* Last one doesn't nest */            int delta = repeat_max * (length_prevgroup + 1 + 2 + 2*LINK_SIZE) -
4560                          2 - 2*LINK_SIZE;   /* Last one doesn't nest */
4561              if ((INT64_OR_DOUBLE)repeat_max *
4562                    (INT64_OR_DOUBLE)(length_prevgroup + 1 + 2 + 2*LINK_SIZE)
4563                      > (INT64_OR_DOUBLE)INT_MAX ||
4564                  OFLOW_MAX - *lengthptr < delta)
4565                {
4566                *errorcodeptr = ERR20;
4567                goto FAILED;
4568                }
4569              *lengthptr += delta;
4570              }
4571    
4572          /* This is compiling for real */          /* This is compiling for real */
4573    
# Line 3683  for (;; ptr++) Line 4585  for (;; ptr++)
4585              {              {
4586              int offset;              int offset;
4587              *code++ = OP_BRA;              *code++ = OP_BRA;
4588              offset = (bralink == NULL)? 0 : code - bralink;              offset = (bralink == NULL)? 0 : (int)(code - bralink);
4589              bralink = code;              bralink = code;
4590              PUTINC(code, 0, offset);              PUTINC(code, 0, offset);
4591              }              }
# Line 3704  for (;; ptr++) Line 4606  for (;; ptr++)
4606          while (bralink != NULL)          while (bralink != NULL)
4607            {            {
4608            int oldlinkoffset;            int oldlinkoffset;
4609            int offset = code - bralink + 1;            int offset = (int)(code - bralink + 1);
4610            uschar *bra = code - offset;            uschar *bra = code - offset;
4611            oldlinkoffset = GET(bra, 1);            oldlinkoffset = GET(bra, 1);
4612            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;            bralink = (oldlinkoffset == 0)? NULL : bralink - oldlinkoffset;
# Line 3735  for (;; ptr++) Line 4637  for (;; ptr++)
4637            uschar *scode = bracode;            uschar *scode = bracode;
4638            do            do
4639              {              {
4640              if (could_be_empty_branch(scode, ketcode, utf8))              if (could_be_empty_branch(scode, ketcode, utf8, cd))
4641                {                {
4642                *bracode += OP_SBRA - OP_BRA;                *bracode += OP_SBRA - OP_BRA;
4643                break;                break;
# Line 3747  for (;; ptr++) Line 4649  for (;; ptr++)
4649          }          }
4650        }        }
4651    
4652        /* If previous is OP_FAIL, it was generated by an empty class [] in
4653        JavaScript mode. The other ways in which OP_FAIL can be generated, that is
4654        by (*FAIL) or (?!) set previous to NULL, which gives a "nothing to repeat"
4655        error above. We can just ignore the repeat in JS case. */
4656    
4657        else if (*previous == OP_FAIL) goto END_REPEAT;
4658    
4659      /* Else there's some kind of shambles */      /* Else there's some kind of shambles */
4660    
4661      else      else
# Line 3771  for (;; ptr++) Line 4680  for (;; ptr++)
4680      if (possessive_quantifier)      if (possessive_quantifier)
4681        {        {
4682        int len;        int len;
4683        if (*tempcode == OP_EXACT || *tempcode == OP_TYPEEXACT ||  
4684            *tempcode == OP_NOTEXACT)        if (*tempcode == OP_TYPEEXACT)
4685            tempcode += _pcre_OP_lengths[*tempcode] +
4686              ((tempcode[3] == OP_PROP || tempcode[3] == OP_NOTPROP)? 2 : 0);
4687    
4688          else if (*tempcode == OP_EXACT || *tempcode == OP_NOTEXACT)
4689            {
4690          tempcode += _pcre_OP_lengths[*tempcode];          tempcode += _pcre_OP_lengths[*tempcode];
4691        len = code - tempcode;  #ifdef SUPPORT_UTF8
4692            if (utf8 && tempcode[-1] >= 0xc0)
4693              tempcode += _pcre_utf8_table4[tempcode[-1] & 0x3f];
4694    #endif
4695            }
4696    
4697          len = (int)(code - tempcode);
4698        if (len > 0) switch (*tempcode)        if (len > 0) switch (*tempcode)
4699          {          {
4700          case OP_STAR:  *tempcode = OP_POSSTAR; break;          case OP_STAR:  *tempcode = OP_POSSTAR; break;
# Line 3792  for (;; ptr++) Line 4712  for (;; ptr++)
4712          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;          case OP_NOTQUERY: *tempcode = OP_NOTPOSQUERY; break;
4713          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;          case OP_NOTUPTO:  *tempcode = OP_NOTPOSUPTO; break;
4714    
4715            /* Because we are moving code along, we must ensure that any
4716            pending recursive references are updated. */
4717    
4718          default:          default:
4719            *code = OP_END;
4720            adjust_recurse(tempcode, 1 + LINK_SIZE, utf8, cd, save_hwm);
4721          memmove(tempcode + 1+LINK_SIZE, tempcode, len);          memmove(tempcode + 1+LINK_SIZE, tempcode, len);
4722          code += 1 + LINK_SIZE;          code += 1 + LINK_SIZE;
4723          len += 1 + LINK_SIZE;          len += 1 + LINK_SIZE;
# Line 3817  for (;; ptr++) Line 4742  for (;; ptr++)
4742      /* ===================================================================*/      /* ===================================================================*/
4743      /* Start of nested parenthesized sub-expression, or comment or lookahead or      /* Start of nested parenthesized sub-expression, or comment or lookahead or
4744      lookbehind or option setting or condition or all the other extended      lookbehind or option setting or condition or all the other extended
4745      parenthesis forms. First deal with the specials; all are introduced by ?,      parenthesis forms.  */
     and the appearance of any of them means that this is not a capturing  
     group. */  
4746    
4747      case '(':      case CHAR_LEFT_PARENTHESIS:
4748      newoptions = options;      newoptions = options;
4749      skipbytes = 0;      skipbytes = 0;
4750      bravalue = OP_CBRA;      bravalue = OP_CBRA;
4751      save_hwm = cd->hwm;      save_hwm = cd->hwm;
4752      reset_bracount = FALSE;      reset_bracount = FALSE;
4753    
4754      if (*(++ptr) == '?')      /* First deal with various "verbs" that can be introduced by '*'. */
4755    
4756        if (*(++ptr) == CHAR_ASTERISK &&
4757             ((cd->ctypes[ptr[1]] & ctype_letter) != 0 || ptr[1] == ':'))
4758          {
4759          int i, namelen;
4760          int arglen = 0;
4761          const char *vn = verbnames;
4762          const uschar *name = ptr + 1;
4763          const uschar *arg = NULL;
4764          previous = NULL;
4765          while ((cd->ctypes[*++ptr] & ctype_letter) != 0) {};
4766          namelen = (int)(ptr - name);
4767    
4768          if (*ptr == CHAR_COLON)
4769            {
4770            arg = ++ptr;
4771            while ((cd->ctypes[*ptr] & (ctype_letter|ctype_digit)) != 0
4772              || *ptr == '_') ptr++;
4773            arglen = (int)(ptr - arg);
4774            }
4775    
4776          if (*ptr != CHAR_RIGHT_PARENTHESIS)
4777            {
4778            *errorcodeptr = ERR60;
4779            goto FAILED;
4780            }
4781    
4782          /* Scan the table of verb names */
4783    
4784          for (i = 0; i < verbcount; i++)
4785            {
4786            if (namelen == verbs[i].len &&
4787                strncmp((char *)name, vn, namelen) == 0)
4788              {
4789              /* Check for open captures before ACCEPT */
4790    
4791              if (verbs[i].op == OP_ACCEPT)
4792                {
4793                open_capitem *oc;
4794                cd->had_accept = TRUE;
4795                for (oc = cd->open_caps; oc != NULL; oc = oc->next)
4796                  {
4797                  *code++ = OP_CLOSE;
4798                  PUT2INC(code, 0, oc->number);
4799                  }
4800                }
4801    
4802              /* Handle the cases with/without an argument */
4803    
4804              if (arglen == 0)
4805                {
4806                if (verbs[i].op < 0)   /* Argument is mandatory */
4807                  {
4808                  *errorcodeptr = ERR66;
4809                  goto FAILED;
4810                  }
4811                *code++ = verbs[i].op;
4812                }
4813    
4814              else
4815                {
4816                if (verbs[i].op_arg < 0)   /* Argument is forbidden */
4817                  {
4818                  *errorcodeptr = ERR59;
4819                  goto FAILED;
4820                  }
4821                *code++ = verbs[i].op_arg;
4822                *code++ = arglen;
4823                memcpy(code, arg, arglen);
4824                code += arglen;
4825                *code++ = 0;
4826                }
4827    
4828              break;  /* Found verb, exit loop */
4829              }
4830    
4831            vn += verbs[i].len + 1;
4832            }
4833    
4834          if (i < verbcount) continue;    /* Successfully handled a verb */
4835          *errorcodeptr = ERR60;          /* Verb not recognized */
4836          goto FAILED;
4837          }
4838    
4839        /* Deal with the extended parentheses; all are introduced by '?', and the
4840        appearance of any of them means that this is not a capturing group. */
4841    
4842        else if (*ptr == CHAR_QUESTION_MARK)
4843        {        {
4844        int i, set, unset, namelen;        int i, set, unset, namelen;
4845        int *optset;        int *optset;
# Line 3837  for (;; ptr++) Line 4848  for (;; ptr++)
4848    
4849        switch (*(++ptr))        switch (*(++ptr))
4850          {          {
4851          case '#':                 /* Comment; skip to ket */          case CHAR_NUMBER_SIGN:                 /* Comment; skip to ket */
4852          ptr++;          ptr++;
4853          while (*ptr != 0 && *ptr != ')') ptr++;          while (*ptr != 0 && *ptr != CHAR_RIGHT_PARENTHESIS) ptr++;
4854          if (*ptr == 0)          if (*ptr == 0)
4855            {            {
4856            *errorcodeptr = ERR18;            *errorcodeptr = ERR18;
# Line 3849  for (;; ptr++) Line 4860  for (;; ptr++)
4860    
4861    
4862          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4863          case '|':                 /* Reset capture count for each branch */          case CHAR_VERTICAL_LINE:  /* Reset capture count for each branch */
4864          reset_bracount = TRUE;          reset_bracount = TRUE;
4865          /* Fall through */          /* Fall through */
4866    
4867          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4868          case ':':                 /* Non-capturing bracket */          case CHAR_COLON:          /* Non-capturing bracket */
4869          bravalue = OP_BRA;          bravalue = OP_BRA;
4870          ptr++;          ptr++;
4871          break;          break;
4872    
4873    
4874          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
4875          case '(':          case CHAR_LEFT_PARENTHESIS:
4876          bravalue = OP_COND;       /* Conditional group */          bravalue = OP_COND;       /* Conditional group */
4877    
4878          /* 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 3881  for (;; ptr++) Line 4892  for (;; ptr++)
4892          the switch. This will take control down to where bracketed groups,          the switch. This will take control down to where bracketed groups,
4893          including assertions, are processed. */          including assertions, are processed. */
4894    
4895          if (ptr[1] == '?' && (ptr[2] == '=' || ptr[2] == '!' || ptr[2] == '<'))          if (ptr[1] == CHAR_QUESTION_MARK && (ptr[2] == CHAR_EQUALS_SIGN ||
4896                ptr[2] == CHAR_EXCLAMATION_MARK || ptr[2] == CHAR_LESS_THAN_SIGN))
4897            break;            break;
4898    
4899          /* 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 3893  for (;; ptr++) Line 4905  for (;; ptr++)
4905    
4906          /* Check for a test for recursion in a named group. */          /* Check for a test for recursion in a named group. */
4907    
4908          if (ptr[1] == 'R' && ptr[2] == '&')          if (ptr[1] == CHAR_R && ptr[2] == CHAR_AMPERSAND)
4909            {            {
4910            terminator = -1;            terminator = -1;
4911            ptr += 2;            ptr += 2;
# Line 3903  for (;; ptr++) Line 4915  for (;; ptr++)
4915          /* 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
4916          syntax (?(<name>) or (?('name') */          syntax (?(<name>) or (?('name') */
4917    
4918          else if (ptr[1] == '<')          else if (ptr[1] == CHAR_LESS_THAN_SIGN)
4919            {            {
4920            terminator = '>';            terminator = CHAR_GREATER_THAN_SIGN;
4921            ptr++;            ptr++;
4922            }            }
4923          else if (ptr[1] == '\'')          else if (ptr[1] == CHAR_APOSTROPHE)
4924            {            {
4925            terminator = '\'';            terminator = CHAR_APOSTROPHE;
4926            ptr++;            ptr++;
4927            }            }
4928          else          else
4929            {            {
4930            terminator = 0;            terminator = 0;
4931            if (ptr[1] == '-' || ptr[1] == '+') refsign = *(++ptr);            if (ptr[1] == CHAR_MINUS || ptr[1] == CHAR_PLUS) refsign = *(++ptr);
4932            }            }
4933    
4934          /* 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 3936  for (;; ptr++) Line 4948  for (;; ptr++)
4948            {            {
4949            if (recno >= 0)            if (recno >= 0)
4950              recno = ((digitab[*ptr] & ctype_digit) != 0)?              recno = ((digitab[*ptr] & ctype_digit) != 0)?
4951                recno * 10 + *ptr - '0' : -1;                recno * 10 + *ptr - CHAR_0 : -1;
4952            ptr++;            ptr++;
4953            }            }
4954          namelen = ptr - name;          namelen = (int)(ptr - name);
4955    
4956          if ((terminator > 0 && *ptr++ != terminator) || *ptr++ != ')')          if ((terminator > 0 && *ptr++ != terminator) ||
4957                *ptr++ != CHAR_RIGHT_PARENTHESIS)
4958            {            {
4959            ptr--;      /* Error offset */            ptr--;      /* Error offset */
4960            *errorcodeptr = ERR26;            *errorcodeptr = ERR26;
# Line 3963  for (;; ptr++) Line 4976  for (;; ptr++)
4976              *errorcodeptr = ERR58;              *errorcodeptr = ERR58;
4977              goto FAILED;              goto FAILED;
4978              }              }
4979            if (refsign == '-')            recno = (refsign == CHAR_MINUS)?
4980                cd->bracount - recno + 1 : recno +cd->bracount;
4981              if (recno <= 0 || recno > cd->final_bracount)
4982              {              {
4983              recno = cd->bracount - recno + 1;              *errorcodeptr = ERR15;
4984              if (recno <= 0)              goto FAILED;
               {  
               *errorcodeptr = ERR15;  
               goto FAILED;  
               }  
4985              }              }
           else recno += cd->bracount;  
4986            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
4987            break;            break;
4988            }            }
4989    
4990          /* Otherwise (did not start with "+" or "-"), start by looking for the          /* Otherwise (did not start with "+" or "-"), start by looking for the
4991          name. */          name. If we find a name, add one to the opcode to change OP_CREF or
4992            OP_RREF into OP_NCREF or OP_NRREF. These behave exactly the same,
4993            except they record that the reference was originally to a name. The
4994            information is used to check duplicate names. */
4995    
4996          slot = cd->name_table;          slot = cd->name_table;
4997          for (i = 0; i < cd->names_found; i++)          for (i = 0; i < cd->names_found; i++)
# Line 3993  for (;; ptr++) Line 5006  for (;; ptr++)
5006            {            {
5007            recno = GET2(slot, 0);            recno = GET2(slot, 0);
5008            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5009              code[1+LINK_SIZE]++;
5010            }            }
5011    
5012          /* Search the pattern for a forward reference */          /* Search the pattern for a forward reference */
5013    
5014          else if ((i = find_parens(ptr, cd->bracount, name, namelen,          else if ((i = find_parens(cd, name, namelen,
5015                          (options & PCRE_EXTENDED) != 0)) > 0)                          (options & PCRE_EXTENDED) != 0)) > 0)
5016            {            {
5017            PUT2(code, 2+LINK_SIZE, i);            PUT2(code, 2+LINK_SIZE, i);
5018              code[1+LINK_SIZE]++;
5019            }            }
5020    
5021          /* If terminator == 0 it means that the name followed directly after          /* If terminator == 0 it means that the name followed directly after
# Line 4018  for (;; ptr++) Line 5033  for (;; ptr++)
5033          /* Check for (?(R) for recursion. Allow digits after R to specify a          /* Check for (?(R) for recursion. Allow digits after R to specify a
5034          specific group number. */          specific group number. */
5035    
5036          else if (*name == 'R')          else if (*name == CHAR_R)
5037            {            {
5038            recno = 0;            recno = 0;
5039            for (i = 1; i < namelen; i++)            for (i = 1; i < namelen; i++)
# Line 4028  for (;; ptr++) Line 5043  for (;; ptr++)
5043                *errorcodeptr = ERR15;                *errorcodeptr = ERR15;
5044                goto FAILED;                goto FAILED;
5045                }                }
5046              recno = recno * 10 + name[i] - '0';              recno = recno * 10 + name[i] - CHAR_0;
5047              }              }
5048            if (recno == 0) recno = RREF_ANY;            if (recno == 0) recno = RREF_ANY;
5049            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */            code[1+LINK_SIZE] = OP_RREF;      /* Change test type */
# Line 4038  for (;; ptr++) Line 5053  for (;; ptr++)
5053          /* Similarly, check for the (?(DEFINE) "condition", which is always          /* Similarly, check for the (?(DEFINE) "condition", which is always
5054          false. */          false. */
5055    
5056          else if (namelen == 6 && strncmp((char *)name, "DEFINE", 6) == 0)          else if (namelen == 6 && strncmp((char *)name, STRING_DEFINE, 6) == 0)
5057            {            {
5058            code[1+LINK_SIZE] = OP_DEF;            code[1+LINK_SIZE] = OP_DEF;
5059            skipbytes = 1;            skipbytes = 1;
5060            }            }
5061    
5062          /* Check for the "name" actually being a subpattern number. */          /* Check for the "name" actually being a subpattern number. We are
5063            in the second pass here, so final_bracount is set. */
5064    
5065          else if (recno > 0)          else if (recno > 0 && recno <= cd->final_bracount)
5066            {            {
5067            PUT2(code, 2+LINK_SIZE, recno);            PUT2(code, 2+LINK_SIZE, recno);
5068            }            }
# Line 4062  for (;; ptr++) Line 5078  for (;; ptr++)
5078    
5079    
5080          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5081          case '=':                 /* Positive lookahead */          case CHAR_EQUALS_SIGN:                 /* Positive lookahead */
5082          bravalue = OP_ASSERT;          bravalue = OP_ASSERT;
5083          ptr++;          ptr++;
5084          break;          break;
5085    
5086    
5087          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5088          case '!':                 /* Negative lookahead */          case CHAR_EXCLAMATION_MARK:            /* Negative lookahead */
         bravalue = OP_ASSERT_NOT;  
5089          ptr++;          ptr++;
5090            if (*ptr == CHAR_RIGHT_PARENTHESIS)    /* Optimize (?!) */
5091              {
5092              *code++ = OP_FAIL;
5093              previous = NULL;
5094              continue;
5095              }
5096            bravalue = OP_ASSERT_NOT;
5097          break;          break;
5098    
5099    
5100          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5101          case '<':                 /* Lookbehind or named define */          case CHAR_LESS_THAN_SIGN:              /* Lookbehind or named define */
5102          switch (ptr[1])          switch (ptr[1])
5103            {            {
5104            case '=':               /* Positive lookbehind */            case CHAR_EQUALS_SIGN:               /* Positive lookbehind */
5105            bravalue = OP_ASSERTBACK;            bravalue = OP_ASSERTBACK;
5106            ptr += 2;            ptr += 2;
5107            break;            break;
5108    
5109            case '!':               /* Negative lookbehind */            case CHAR_EXCLAMATION_MARK:          /* Negative lookbehind */
5110            bravalue = OP_ASSERTBACK_NOT;            bravalue = OP_ASSERTBACK_NOT;
5111            ptr += 2;            ptr += 2;
5112            break;            break;
# Line 4099  for (;; ptr++) Line 5121  for (;; ptr++)
5121    
5122    
5123          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5124          case '>':                 /* One-time brackets */          case CHAR_GREATER_THAN_SIGN:           /* One-time brackets */
5125          bravalue = OP_ONCE;          bravalue = OP_ONCE;
5126          ptr++;          ptr++;
5127          break;          break;
5128    
5129    
5130          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5131          case 'C':                 /* Callout - may be followed by digits; */          case CHAR_C:                 /* Callout - may be followed by digits; */
5132          previous_callout = code;  /* Save for later completion */          previous_callout = code;  /* Save for later completion */
5133          after_manual_callout = 1; /* Skip one item before completing */          after_manual_callout = 1; /* Skip one item before completing */
5134          *code++ = OP_CALLOUT;          *code++ = OP_CALLOUT;
5135            {            {
5136            int n = 0;            int n = 0;
5137            while ((digitab[*(++ptr)] & ctype_digit) != 0)            while ((digitab[*(++ptr)] & ctype_digit) != 0)
5138              n = n * 10 + *ptr - '0';              n = n * 10 + *ptr - CHAR_0;
5139            if (*ptr != ')')            if (*ptr != CHAR_RIGHT_PARENTHESIS)
5140              {              {
5141              *errorcodeptr = ERR39;              *errorcodeptr = ERR39;
5142              goto FAILED;              goto FAILED;
# Line 4125  for (;; ptr++) Line 5147  for (;; ptr++)
5147              goto FAILED;              goto FAILED;
5148              }              }
5149            *code++ = n;            *code++ = n;
5150            PUT(code, 0, ptr - cd->start_pattern + 1);  /* Pattern offset */            PUT(code, 0, (int)(ptr - cd->start_pattern + 1)); /* Pattern offset */
5151            PUT(code, LINK_SIZE, 0);                    /* Default length */            PUT(code, LINK_SIZE, 0);                          /* Default length */
5152            code += 2 * LINK_SIZE;            code += 2 * LINK_SIZE;
5153            }            }
5154          previous = NULL;          previous = NULL;
# Line 4134  for (;; ptr++) Line 5156  for (;; ptr++)
5156    
5157    
5158          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5159          case 'P':                 /* Python-style named subpattern handling */          case CHAR_P:              /* Python-style named subpattern handling */
5160          if (*(++ptr) == '=' || *ptr == '>')  /* Reference or recursion */          if (*(++ptr) == CHAR_EQUALS_SIGN ||
5161                *ptr == CHAR_GREATER_THAN_SIGN)  /* Reference or recursion */
5162            {            {
5163            is_recurse = *ptr == '>';            is_recurse = *ptr == CHAR_GREATER_THAN_SIGN;
5164            terminator = ')';            terminator = CHAR_RIGHT_PARENTHESIS;
5165            goto NAMED_REF_OR_RECURSE;            goto NAMED_REF_OR_RECURSE;
5166            }            }
5167          else if (*ptr != '<')    /* Test for Python-style definition */          else if (*ptr != CHAR_LESS_THAN_SIGN)  /* Test for Python-style defn */
5168            {            {
5169            *errorcodeptr = ERR41;            *errorcodeptr = ERR41;
5170            goto FAILED;            goto FAILED;
# Line 4151  for (;; ptr++) Line 5174  for (;; ptr++)
5174    
5175          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5176          DEFINE_NAME:    /* Come here from (?< handling */          DEFINE_NAME:    /* Come here from (?< handling */
5177          case '\'':          case CHAR_APOSTROPHE:
5178            {            {
5179            terminator = (*ptr == '<')? '>' : '\'';            terminator = (*ptr == CHAR_LESS_THAN_SIGN)?
5180                CHAR_GREATER_THAN_SIGN : CHAR_APOSTROPHE;
5181            name = ++ptr;            name = ++ptr;
5182    
5183            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;            while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5184            namelen = ptr - name;            namelen = (int)(ptr - name);
5185    
5186            /* In the pre-compile phase, just do a syntax check. */            /* In the pre-compile phase, just do a syntax check. */
5187    
# Line 4184  for (;; ptr++) Line 5208  for (;; ptr++)
5208                }                }
5209              }              }
5210    
5211            /* In the real compile, create the entry in the table */            /* In the real compile, create the entry in the table, maintaining
5212              alphabetical order. Duplicate names for different numbers are
5213              permitted only if PCRE_DUPNAMES is set. Duplicate names for the same
5214              number are always OK. (An existing number can be re-used if (?|
5215              appears in the pattern.) In either event, a duplicate name results in
5216              a duplicate entry in the table, even if the number is the same. This
5217              is because the number of names, and hence the table size, is computed
5218              in the pre-compile, and it affects various numbers and pointers which
5219              would all have to be modified, and the compiled code moved down, if
5220              duplicates with the same number were omitted from the table. This
5221              doesn't seem worth the hassle. However, *different* names for the
5222              same number are not permitted. */
5223    
5224            else            else
5225              {              {
5226                BOOL dupname = FALSE;
5227              slot = cd->name_table;              slot = cd->name_table;
5228    
5229              for (i = 0; i < cd->names_found; i++)              for (i = 0; i < cd->names_found; i++)
5230                {                {
5231                int crc = memcmp(name, slot+2, namelen);                int crc = memcmp(name, slot+2, namelen);
# Line 4196  for (;; ptr++) Line 5233  for (;; ptr++)
5233                  {                  {
5234                  if (slot[2+namelen] == 0)                  if (slot[2+namelen] == 0)
5235                    {                    {
5236                    if ((options & PCRE_DUPNAMES) == 0)                    if (GET2(slot, 0) != cd->bracount + 1 &&
5237                          (options & PCRE_DUPNAMES) == 0)
5238                      {                      {
5239                      *errorcodeptr = ERR43;                      *errorcodeptr = ERR43;
5240                      goto FAILED;                      goto FAILED;
5241                      }                      }
5242                      else dupname = TRUE;
5243                    }                    }
5244                  else crc = -1;      /* Current name is substring */                  else crc = -1;      /* Current name is a substring */
5245                  }                  }
5246    
5247                  /* Make space in the table and break the loop for an earlier
5248                  name. For a duplicate or later name, carry on. We do this for
5249                  duplicates so that in the simple case (when ?(| is not used) they
5250                  are in order of their numbers. */
5251    
5252                if (crc < 0)                if (crc < 0)
5253                  {                  {
5254                  memmove(slot + cd->name_entry_size, slot,                  memmove(slot + cd->name_entry_size, slot,
5255                    (cd->names_found - i) * cd->name_entry_size);                    (cd->names_found - i) * cd->name_entry_size);
5256                  break;                  break;
5257                  }                  }
5258    
5259                  /* Continue the loop for a later or duplicate name */
5260    
5261                slot += cd->name_entry_size;                slot += cd->name_entry_size;
5262                }                }
5263    
5264                /* For non-duplicate names, check for a duplicate number before
5265                adding the new name. */
5266    
5267                if (!dupname)
5268                  {
5269                  uschar *cslot = cd->name_table;
5270                  for (i = 0; i < cd->names_found; i++)
5271                    {
5272                    if (cslot != slot)
5273                      {
5274                      if (GET2(cslot, 0) == cd->bracount + 1)
5275                        {
5276                        *errorcodeptr = ERR65;
5277                        goto FAILED;
5278                        }
5279                      }
5280                    else i--;
5281                    cslot += cd->name_entry_size;
5282                    }
5283                  }
5284    
5285              PUT2(slot, 0, cd->bracount + 1);              PUT2(slot, 0, cd->bracount + 1);
5286              memcpy(slot + 2, name, namelen);              memcpy(slot + 2, name, namelen);
5287              slot[2+namelen] = 0;              slot[2+namelen] = 0;
5288              }              }
5289            }            }
5290    
5291          /* In both cases, count the number of names we've encountered. */          /* In both pre-compile and compile, count the number of names we've
5292            encountered. */
5293    
         ptr++;                    /* Move past > or ' */  
5294          cd->names_found++;          cd->names_found++;
5295            ptr++;                    /* Move past > or ' */
5296          goto NUMBERED_GROUP;          goto NUMBERED_GROUP;
5297    
5298    
5299          /* ------------------------------------------------------------ */          /* ------------------------------------------------------------ */
5300          case '&':                 /* Perl recursion/subroutine syntax */          case CHAR_AMPERSAND:            /* Perl recursion/subroutine syntax */
5301          terminator = ')';          terminator = CHAR_RIGHT_PARENTHESIS;
5302          is_recurse = TRUE;          is_recurse = TRUE;
5303          /* Fall through */          /* Fall through */
5304    
5305          /* We come here from the Python syntax above that handles both          /* We come here from the Python syntax above that handles both
5306          references (?P=name) and recursion (?P>name), as well as falling          references (?P=name) and recursion (?P>name), as well as falling
5307          through from the Perl recursion syntax (?&name). */          through from the Perl recursion syntax (?&name). We also come here from
5308            the Perl \k<name> or \k'name' back reference syntax and the \k{name}
5309            .NET syntax, and the Oniguruma \g<...> and \g'...' subroutine syntax. */
5310    
5311          NAMED_REF_OR_RECURSE:          NAMED_REF_OR_RECURSE:
5312          name = ++ptr;          name = ++ptr;
5313          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;          while ((cd->ctypes[*ptr] & ctype_word) != 0) ptr++;
5314          namelen = ptr - name;          namelen = (